]>
Commit | Line | Data |
---|---|---|
1b0c7174 JH |
1 | #include "cache.h" |
2 | #include "tree-walk.h" | |
e6c111b4 | 3 | #include "unpack-trees.h" |
bc96cc87 | 4 | #include "dir.h" |
8e440259 | 5 | #include "tree.h" |
64acde94 | 6 | #include "pathspec.h" |
1b0c7174 | 7 | |
4651ece8 LT |
8 | static const char *get_mode(const char *str, unsigned int *modep) |
9 | { | |
10 | unsigned char c; | |
11 | unsigned int mode = 0; | |
12 | ||
64cc1c09 MK |
13 | if (*str == ' ') |
14 | return NULL; | |
15 | ||
4651ece8 LT |
16 | while ((c = *str++) != ' ') { |
17 | if (c < '0' || c > '7') | |
18 | return NULL; | |
19 | mode = (mode << 3) + (c - '0'); | |
20 | } | |
21 | *modep = mode; | |
22 | return str; | |
23 | } | |
24 | ||
64cc1c09 | 25 | static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size) |
4651ece8 LT |
26 | { |
27 | const char *path; | |
28 | unsigned int mode, len; | |
29 | ||
64cc1c09 MK |
30 | if (size < 24 || buf[size - 21]) |
31 | die("corrupt tree file"); | |
32 | ||
4651ece8 | 33 | path = get_mode(buf, &mode); |
64cc1c09 | 34 | if (!path || !*path) |
4651ece8 LT |
35 | die("corrupt tree file"); |
36 | len = strlen(path) + 1; | |
37 | ||
38 | /* Initialize the descriptor entry */ | |
39 | desc->entry.path = path; | |
7146e66f | 40 | desc->entry.mode = canon_mode(mode); |
4651ece8 LT |
41 | desc->entry.sha1 = (const unsigned char *)(path + len); |
42 | } | |
43 | ||
6fda5e51 LT |
44 | void init_tree_desc(struct tree_desc *desc, const void *buffer, unsigned long size) |
45 | { | |
46 | desc->buffer = buffer; | |
47 | desc->size = size; | |
4651ece8 LT |
48 | if (size) |
49 | decode_tree_entry(desc, buffer, size); | |
6fda5e51 LT |
50 | } |
51 | ||
1b0c7174 JH |
52 | void *fill_tree_descriptor(struct tree_desc *desc, const unsigned char *sha1) |
53 | { | |
54 | unsigned long size = 0; | |
55 | void *buf = NULL; | |
56 | ||
57 | if (sha1) { | |
8e440259 | 58 | buf = read_object_with_reference(sha1, tree_type, &size, NULL); |
1b0c7174 JH |
59 | if (!buf) |
60 | die("unable to read tree %s", sha1_to_hex(sha1)); | |
61 | } | |
6fda5e51 | 62 | init_tree_desc(desc, buf, size); |
1b0c7174 JH |
63 | return buf; |
64 | } | |
65 | ||
1b0c7174 JH |
66 | static void entry_clear(struct name_entry *a) |
67 | { | |
68 | memset(a, 0, sizeof(*a)); | |
69 | } | |
70 | ||
71 | static void entry_extract(struct tree_desc *t, struct name_entry *a) | |
72 | { | |
4651ece8 | 73 | *a = t->entry; |
1b0c7174 JH |
74 | } |
75 | ||
76 | void update_tree_entry(struct tree_desc *desc) | |
77 | { | |
6fda5e51 | 78 | const void *buf = desc->buffer; |
4651ece8 | 79 | const unsigned char *end = desc->entry.sha1 + 20; |
1b0c7174 | 80 | unsigned long size = desc->size; |
4651ece8 | 81 | unsigned long len = end - (const unsigned char *)buf; |
1b0c7174 JH |
82 | |
83 | if (size < len) | |
84 | die("corrupt tree file"); | |
4651ece8 LT |
85 | buf = end; |
86 | size -= len; | |
87 | desc->buffer = buf; | |
88 | desc->size = size; | |
89 | if (size) | |
90 | decode_tree_entry(desc, buf, size); | |
1b0c7174 JH |
91 | } |
92 | ||
4c068a98 LT |
93 | int tree_entry(struct tree_desc *desc, struct name_entry *entry) |
94 | { | |
4651ece8 | 95 | if (!desc->size) |
4c068a98 LT |
96 | return 0; |
97 | ||
4651ece8 LT |
98 | *entry = desc->entry; |
99 | update_tree_entry(desc); | |
4c068a98 LT |
100 | return 1; |
101 | } | |
102 | ||
40d934df LT |
103 | void setup_traverse_info(struct traverse_info *info, const char *base) |
104 | { | |
105 | int pathlen = strlen(base); | |
bcbe5a51 | 106 | static struct traverse_info dummy; |
40d934df LT |
107 | |
108 | memset(info, 0, sizeof(*info)); | |
109 | if (pathlen && base[pathlen-1] == '/') | |
110 | pathlen--; | |
111 | info->pathlen = pathlen ? pathlen + 1 : 0; | |
112 | info->name.path = base; | |
113 | info->name.sha1 = (void *)(base + pathlen + 1); | |
bcbe5a51 LT |
114 | if (pathlen) |
115 | info->prev = &dummy; | |
40d934df LT |
116 | } |
117 | ||
118 | char *make_traverse_path(char *path, const struct traverse_info *info, const struct name_entry *n) | |
119 | { | |
0de16337 | 120 | int len = tree_entry_len(n); |
40d934df LT |
121 | int pathlen = info->pathlen; |
122 | ||
123 | path[pathlen + len] = 0; | |
124 | for (;;) { | |
125 | memcpy(path + pathlen, n->path, len); | |
126 | if (!pathlen) | |
127 | break; | |
128 | path[--pathlen] = '/'; | |
129 | n = &info->name; | |
0de16337 | 130 | len = tree_entry_len(n); |
40d934df LT |
131 | info = info->prev; |
132 | pathlen -= len; | |
133 | } | |
134 | return path; | |
135 | } | |
136 | ||
1ee26571 JH |
137 | struct tree_desc_skip { |
138 | struct tree_desc_skip *prev; | |
139 | const void *ptr; | |
140 | }; | |
141 | ||
142 | struct tree_desc_x { | |
143 | struct tree_desc d; | |
144 | struct tree_desc_skip *skip; | |
145 | }; | |
146 | ||
1ee26571 JH |
147 | static int check_entry_match(const char *a, int a_len, const char *b, int b_len) |
148 | { | |
149 | /* | |
150 | * The caller wants to pick *a* from a tree or nothing. | |
151 | * We are looking at *b* in a tree. | |
152 | * | |
153 | * (0) If a and b are the same name, we are trivially happy. | |
154 | * | |
155 | * There are three possibilities where *a* could be hiding | |
156 | * behind *b*. | |
157 | * | |
158 | * (1) *a* == "t", *b* == "ab" i.e. *b* sorts earlier than *a* no | |
159 | * matter what. | |
160 | * (2) *a* == "t", *b* == "t-2" and "t" is a subtree in the tree; | |
161 | * (3) *a* == "t-2", *b* == "t" and "t-2" is a blob in the tree. | |
162 | * | |
163 | * Otherwise we know *a* won't appear in the tree without | |
164 | * scanning further. | |
165 | */ | |
166 | ||
167 | int cmp = name_compare(a, a_len, b, b_len); | |
168 | ||
169 | /* Most common case first -- reading sync'd trees */ | |
170 | if (!cmp) | |
171 | return cmp; | |
172 | ||
173 | if (0 < cmp) { | |
174 | /* a comes after b; it does not matter if it is case (3) | |
175 | if (b_len < a_len && !memcmp(a, b, b_len) && a[b_len] < '/') | |
176 | return 1; | |
177 | */ | |
178 | return 1; /* keep looking */ | |
179 | } | |
180 | ||
181 | /* b comes after a; are we looking at case (2)? */ | |
182 | if (a_len < b_len && !memcmp(a, b, a_len) && b[a_len] < '/') | |
183 | return 1; /* keep looking */ | |
184 | ||
185 | return -1; /* a cannot appear in the tree */ | |
186 | } | |
187 | ||
188 | /* | |
189 | * From the extended tree_desc, extract the first name entry, while | |
190 | * paying attention to the candidate "first" name. Most importantly, | |
191 | * when looking for an entry, if there are entries that sorts earlier | |
192 | * in the tree object representation than that name, skip them and | |
193 | * process the named entry first. We will remember that we haven't | |
194 | * processed the first entry yet, and in the later call skip the | |
195 | * entry we processed early when update_extended_entry() is called. | |
196 | * | |
197 | * E.g. if the underlying tree object has these entries: | |
198 | * | |
199 | * blob "t-1" | |
200 | * blob "t-2" | |
201 | * tree "t" | |
202 | * blob "t=1" | |
203 | * | |
204 | * and the "first" asks for "t", remember that we still need to | |
205 | * process "t-1" and "t-2" but extract "t". After processing the | |
206 | * entry "t" from this call, the caller will let us know by calling | |
207 | * update_extended_entry() that we can remember "t" has been processed | |
208 | * already. | |
209 | */ | |
210 | ||
211 | static void extended_entry_extract(struct tree_desc_x *t, | |
212 | struct name_entry *a, | |
213 | const char *first, | |
214 | int first_len) | |
215 | { | |
216 | const char *path; | |
217 | int len; | |
218 | struct tree_desc probe; | |
219 | struct tree_desc_skip *skip; | |
220 | ||
221 | /* | |
222 | * Extract the first entry from the tree_desc, but skip the | |
223 | * ones that we already returned in earlier rounds. | |
224 | */ | |
225 | while (1) { | |
226 | if (!t->d.size) { | |
227 | entry_clear(a); | |
228 | break; /* not found */ | |
229 | } | |
230 | entry_extract(&t->d, a); | |
231 | for (skip = t->skip; skip; skip = skip->prev) | |
232 | if (a->path == skip->ptr) | |
233 | break; /* found */ | |
234 | if (!skip) | |
235 | break; | |
236 | /* We have processed this entry already. */ | |
237 | update_tree_entry(&t->d); | |
238 | } | |
239 | ||
240 | if (!first || !a->path) | |
241 | return; | |
242 | ||
243 | /* | |
244 | * The caller wants "first" from this tree, or nothing. | |
245 | */ | |
246 | path = a->path; | |
0de16337 | 247 | len = tree_entry_len(a); |
1ee26571 JH |
248 | switch (check_entry_match(first, first_len, path, len)) { |
249 | case -1: | |
250 | entry_clear(a); | |
251 | case 0: | |
252 | return; | |
253 | default: | |
254 | break; | |
255 | } | |
256 | ||
257 | /* | |
258 | * We need to look-ahead -- we suspect that a subtree whose | |
259 | * name is "first" may be hiding behind the current entry "path". | |
260 | */ | |
261 | probe = t->d; | |
262 | while (probe.size) { | |
263 | entry_extract(&probe, a); | |
264 | path = a->path; | |
0de16337 | 265 | len = tree_entry_len(a); |
1ee26571 JH |
266 | switch (check_entry_match(first, first_len, path, len)) { |
267 | case -1: | |
268 | entry_clear(a); | |
269 | case 0: | |
270 | return; | |
271 | default: | |
272 | update_tree_entry(&probe); | |
273 | break; | |
274 | } | |
275 | /* keep looking */ | |
276 | } | |
277 | entry_clear(a); | |
278 | } | |
279 | ||
280 | static void update_extended_entry(struct tree_desc_x *t, struct name_entry *a) | |
281 | { | |
282 | if (t->d.entry.path == a->path) { | |
283 | update_tree_entry(&t->d); | |
284 | } else { | |
285 | /* we have returned this entry early */ | |
286 | struct tree_desc_skip *skip = xmalloc(sizeof(*skip)); | |
287 | skip->ptr = a->path; | |
288 | skip->prev = t->skip; | |
289 | t->skip = skip; | |
290 | } | |
291 | } | |
292 | ||
293 | static void free_extended_entry(struct tree_desc_x *t) | |
294 | { | |
295 | struct tree_desc_skip *p, *s; | |
296 | ||
297 | for (s = t->skip; s; s = p) { | |
298 | p = s->prev; | |
299 | free(s); | |
300 | } | |
301 | } | |
302 | ||
2842c0f9 JH |
303 | static inline int prune_traversal(struct name_entry *e, |
304 | struct traverse_info *info, | |
305 | struct strbuf *base, | |
306 | int still_interesting) | |
307 | { | |
308 | if (!info->pathspec || still_interesting == 2) | |
309 | return 2; | |
310 | if (still_interesting < 0) | |
311 | return still_interesting; | |
312 | return tree_entry_interesting(e, base, 0, info->pathspec); | |
313 | } | |
314 | ||
5803c6f8 | 315 | int traverse_trees(int n, struct tree_desc *t, struct traverse_info *info) |
1b0c7174 | 316 | { |
e6c111b4 | 317 | int error = 0; |
1b0c7174 | 318 | struct name_entry *entry = xmalloc(n*sizeof(*entry)); |
1ee26571 JH |
319 | int i; |
320 | struct tree_desc_x *tx = xcalloc(n, sizeof(*tx)); | |
2842c0f9 JH |
321 | struct strbuf base = STRBUF_INIT; |
322 | int interesting = 1; | |
1ee26571 JH |
323 | |
324 | for (i = 0; i < n; i++) | |
325 | tx[i].d = t[i]; | |
1b0c7174 | 326 | |
2842c0f9 JH |
327 | if (info->prev) { |
328 | strbuf_grow(&base, info->pathlen); | |
329 | make_traverse_path(base.buf, info->prev, &info->name); | |
330 | base.buf[info->pathlen-1] = '/'; | |
331 | strbuf_setlen(&base, info->pathlen); | |
332 | } | |
1b0c7174 | 333 | for (;;) { |
a04f8196 | 334 | int trees_used; |
1ee26571 JH |
335 | unsigned long mask, dirmask; |
336 | const char *first = NULL; | |
337 | int first_len = 0; | |
27c0f768 | 338 | struct name_entry *e = NULL; |
1ee26571 JH |
339 | int len; |
340 | ||
341 | for (i = 0; i < n; i++) { | |
342 | e = entry + i; | |
343 | extended_entry_extract(tx + i, e, NULL, 0); | |
344 | } | |
1b0c7174 | 345 | |
1ee26571 JH |
346 | /* |
347 | * A tree may have "t-2" at the current location even | |
348 | * though it may have "t" that is a subtree behind it, | |
349 | * and another tree may return "t". We want to grab | |
350 | * all "t" from all trees to match in such a case. | |
351 | */ | |
1b0c7174 | 352 | for (i = 0; i < n; i++) { |
1ee26571 JH |
353 | e = entry + i; |
354 | if (!e->path) | |
1b0c7174 | 355 | continue; |
0de16337 | 356 | len = tree_entry_len(e); |
1ee26571 JH |
357 | if (!first) { |
358 | first = e->path; | |
359 | first_len = len; | |
360 | continue; | |
361 | } | |
362 | if (name_compare(e->path, len, first, first_len) < 0) { | |
363 | first = e->path; | |
364 | first_len = len; | |
365 | } | |
366 | } | |
367 | ||
368 | if (first) { | |
369 | for (i = 0; i < n; i++) { | |
370 | e = entry + i; | |
371 | extended_entry_extract(tx + i, e, first, first_len); | |
372 | /* Cull the ones that are not the earliest */ | |
373 | if (!e->path) | |
1b0c7174 | 374 | continue; |
0de16337 | 375 | len = tree_entry_len(e); |
1ee26571 JH |
376 | if (name_compare(e->path, len, first, first_len)) |
377 | entry_clear(e); | |
1b0c7174 | 378 | } |
1ee26571 JH |
379 | } |
380 | ||
381 | /* Now we have in entry[i] the earliest name from the trees */ | |
382 | mask = 0; | |
383 | dirmask = 0; | |
384 | for (i = 0; i < n; i++) { | |
385 | if (!entry[i].path) | |
386 | continue; | |
1b0c7174 | 387 | mask |= 1ul << i; |
91e4f036 LT |
388 | if (S_ISDIR(entry[i].mode)) |
389 | dirmask |= 1ul << i; | |
2842c0f9 | 390 | e = &entry[i]; |
1b0c7174 JH |
391 | } |
392 | if (!mask) | |
393 | break; | |
2842c0f9 JH |
394 | interesting = prune_traversal(e, info, &base, interesting); |
395 | if (interesting < 0) | |
396 | break; | |
397 | if (interesting) { | |
a04f8196 SB |
398 | trees_used = info->fn(n, mask, dirmask, entry, info); |
399 | if (trees_used < 0) { | |
400 | error = trees_used; | |
2842c0f9 JH |
401 | if (!info->show_all_errors) |
402 | break; | |
403 | } | |
a04f8196 | 404 | mask &= trees_used; |
e6c111b4 | 405 | } |
1ee26571 | 406 | for (i = 0; i < n; i++) |
5803c6f8 | 407 | if (mask & (1ul << i)) |
1ee26571 | 408 | update_extended_entry(tx + i, entry + i); |
1b0c7174 JH |
409 | } |
410 | free(entry); | |
1ee26571 JH |
411 | for (i = 0; i < n; i++) |
412 | free_extended_entry(tx + i); | |
413 | free(tx); | |
2842c0f9 | 414 | strbuf_release(&base); |
e6c111b4 | 415 | return error; |
1b0c7174 JH |
416 | } |
417 | ||
275721c2 DT |
418 | struct dir_state { |
419 | void *tree; | |
420 | unsigned long size; | |
421 | unsigned char sha1[20]; | |
422 | }; | |
423 | ||
4dcff634 JH |
424 | static int find_tree_entry(struct tree_desc *t, const char *name, unsigned char *result, unsigned *mode) |
425 | { | |
426 | int namelen = strlen(name); | |
427 | while (t->size) { | |
428 | const char *entry; | |
429 | const unsigned char *sha1; | |
430 | int entrylen, cmp; | |
431 | ||
432 | sha1 = tree_entry_extract(t, &entry, mode); | |
0de16337 | 433 | entrylen = tree_entry_len(&t->entry); |
4dcff634 | 434 | update_tree_entry(t); |
4dcff634 JH |
435 | if (entrylen > namelen) |
436 | continue; | |
437 | cmp = memcmp(name, entry, entrylen); | |
438 | if (cmp > 0) | |
439 | continue; | |
440 | if (cmp < 0) | |
441 | break; | |
442 | if (entrylen == namelen) { | |
e702496e | 443 | hashcpy(result, sha1); |
4dcff634 JH |
444 | return 0; |
445 | } | |
446 | if (name[entrylen] != '/') | |
447 | continue; | |
448 | if (!S_ISDIR(*mode)) | |
449 | break; | |
450 | if (++entrylen == namelen) { | |
e702496e | 451 | hashcpy(result, sha1); |
4dcff634 JH |
452 | return 0; |
453 | } | |
454 | return get_tree_entry(sha1, name + entrylen, result, mode); | |
455 | } | |
456 | return -1; | |
457 | } | |
458 | ||
459 | int get_tree_entry(const unsigned char *tree_sha1, const char *name, unsigned char *sha1, unsigned *mode) | |
460 | { | |
461 | int retval; | |
462 | void *tree; | |
6fda5e51 | 463 | unsigned long size; |
d93b7d1c | 464 | unsigned char root[20]; |
4dcff634 | 465 | |
6fda5e51 | 466 | tree = read_object_with_reference(tree_sha1, tree_type, &size, root); |
4dcff634 JH |
467 | if (!tree) |
468 | return -1; | |
d93b7d1c JK |
469 | |
470 | if (name[0] == '\0') { | |
471 | hashcpy(sha1, root); | |
ef006503 | 472 | free(tree); |
d93b7d1c JK |
473 | return 0; |
474 | } | |
475 | ||
5fb8c05f JH |
476 | if (!size) { |
477 | retval = -1; | |
478 | } else { | |
479 | struct tree_desc t; | |
480 | init_tree_desc(&t, tree, size); | |
481 | retval = find_tree_entry(&t, name, sha1, mode); | |
482 | } | |
4dcff634 JH |
483 | free(tree); |
484 | return retval; | |
485 | } | |
2c389fc8 | 486 | |
275721c2 DT |
487 | /* |
488 | * This is Linux's built-in max for the number of symlinks to follow. | |
489 | * That limit, of course, does not affect git, but it's a reasonable | |
490 | * choice. | |
491 | */ | |
492 | #define GET_TREE_ENTRY_FOLLOW_SYMLINKS_MAX_LINKS 40 | |
493 | ||
494 | /** | |
495 | * Find a tree entry by following symlinks in tree_sha (which is | |
496 | * assumed to be the root of the repository). In the event that a | |
497 | * symlink points outside the repository (e.g. a link to /foo or a | |
498 | * root-level link to ../foo), the portion of the link which is | |
499 | * outside the repository will be returned in result_path, and *mode | |
500 | * will be set to 0. It is assumed that result_path is uninitialized. | |
501 | * If there are no symlinks, or the end result of the symlink chain | |
502 | * points to an object inside the repository, result will be filled in | |
503 | * with the sha1 of the found object, and *mode will hold the mode of | |
504 | * the object. | |
505 | * | |
506 | * See the code for enum follow_symlink_result for a description of | |
507 | * the return values. | |
508 | */ | |
509 | enum follow_symlinks_result get_tree_entry_follow_symlinks(unsigned char *tree_sha1, const char *name, unsigned char *result, struct strbuf *result_path, unsigned *mode) | |
510 | { | |
511 | int retval = MISSING_OBJECT; | |
512 | struct dir_state *parents = NULL; | |
513 | size_t parents_alloc = 0; | |
514 | ssize_t parents_nr = 0; | |
515 | unsigned char current_tree_sha1[20]; | |
516 | struct strbuf namebuf = STRBUF_INIT; | |
517 | struct tree_desc t; | |
518 | int follows_remaining = GET_TREE_ENTRY_FOLLOW_SYMLINKS_MAX_LINKS; | |
519 | int i; | |
520 | ||
521 | init_tree_desc(&t, NULL, 0UL); | |
522 | strbuf_init(result_path, 0); | |
523 | strbuf_addstr(&namebuf, name); | |
524 | hashcpy(current_tree_sha1, tree_sha1); | |
525 | ||
526 | while (1) { | |
527 | int find_result; | |
528 | char *first_slash; | |
529 | char *remainder = NULL; | |
530 | ||
531 | if (!t.buffer) { | |
532 | void *tree; | |
533 | unsigned char root[20]; | |
534 | unsigned long size; | |
535 | tree = read_object_with_reference(current_tree_sha1, | |
536 | tree_type, &size, | |
537 | root); | |
538 | if (!tree) | |
539 | goto done; | |
540 | ||
541 | ALLOC_GROW(parents, parents_nr + 1, parents_alloc); | |
542 | parents[parents_nr].tree = tree; | |
543 | parents[parents_nr].size = size; | |
544 | hashcpy(parents[parents_nr].sha1, root); | |
545 | parents_nr++; | |
546 | ||
547 | if (namebuf.buf[0] == '\0') { | |
548 | hashcpy(result, root); | |
549 | retval = FOUND; | |
550 | goto done; | |
551 | } | |
552 | ||
553 | if (!size) | |
554 | goto done; | |
555 | ||
556 | /* descend */ | |
557 | init_tree_desc(&t, tree, size); | |
558 | } | |
559 | ||
560 | /* Handle symlinks to e.g. a//b by removing leading slashes */ | |
561 | while (namebuf.buf[0] == '/') { | |
562 | strbuf_remove(&namebuf, 0, 1); | |
563 | } | |
564 | ||
565 | /* Split namebuf into a first component and a remainder */ | |
566 | if ((first_slash = strchr(namebuf.buf, '/'))) { | |
567 | *first_slash = 0; | |
568 | remainder = first_slash + 1; | |
569 | } | |
570 | ||
571 | if (!strcmp(namebuf.buf, "..")) { | |
572 | struct dir_state *parent; | |
573 | /* | |
574 | * We could end up with .. in the namebuf if it | |
575 | * appears in a symlink. | |
576 | */ | |
577 | ||
578 | if (parents_nr == 1) { | |
579 | if (remainder) | |
580 | *first_slash = '/'; | |
581 | strbuf_add(result_path, namebuf.buf, | |
582 | namebuf.len); | |
583 | *mode = 0; | |
584 | retval = FOUND; | |
585 | goto done; | |
586 | } | |
587 | parent = &parents[parents_nr - 1]; | |
588 | free(parent->tree); | |
589 | parents_nr--; | |
590 | parent = &parents[parents_nr - 1]; | |
591 | init_tree_desc(&t, parent->tree, parent->size); | |
592 | strbuf_remove(&namebuf, 0, remainder ? 3 : 2); | |
593 | continue; | |
594 | } | |
595 | ||
596 | /* We could end up here via a symlink to dir/.. */ | |
597 | if (namebuf.buf[0] == '\0') { | |
598 | hashcpy(result, parents[parents_nr - 1].sha1); | |
599 | retval = FOUND; | |
600 | goto done; | |
601 | } | |
602 | ||
603 | /* Look up the first (or only) path component in the tree. */ | |
604 | find_result = find_tree_entry(&t, namebuf.buf, | |
605 | current_tree_sha1, mode); | |
606 | if (find_result) { | |
607 | goto done; | |
608 | } | |
609 | ||
610 | if (S_ISDIR(*mode)) { | |
611 | if (!remainder) { | |
612 | hashcpy(result, current_tree_sha1); | |
613 | retval = FOUND; | |
614 | goto done; | |
615 | } | |
616 | /* Descend the tree */ | |
617 | t.buffer = NULL; | |
618 | strbuf_remove(&namebuf, 0, | |
619 | 1 + first_slash - namebuf.buf); | |
620 | } else if (S_ISREG(*mode)) { | |
621 | if (!remainder) { | |
622 | hashcpy(result, current_tree_sha1); | |
623 | retval = FOUND; | |
624 | } else { | |
625 | retval = NOT_DIR; | |
626 | } | |
627 | goto done; | |
628 | } else if (S_ISLNK(*mode)) { | |
629 | /* Follow a symlink */ | |
630 | unsigned long link_len; | |
631 | size_t len; | |
632 | char *contents, *contents_start; | |
633 | struct dir_state *parent; | |
634 | enum object_type type; | |
635 | ||
636 | if (follows_remaining-- == 0) { | |
637 | /* Too many symlinks followed */ | |
638 | retval = SYMLINK_LOOP; | |
639 | goto done; | |
640 | } | |
641 | ||
642 | /* | |
643 | * At this point, we have followed at a least | |
644 | * one symlink, so on error we need to report this. | |
645 | */ | |
646 | retval = DANGLING_SYMLINK; | |
647 | ||
648 | contents = read_sha1_file(current_tree_sha1, &type, | |
649 | &link_len); | |
650 | ||
651 | if (!contents) | |
652 | goto done; | |
653 | ||
654 | if (contents[0] == '/') { | |
655 | strbuf_addstr(result_path, contents); | |
656 | free(contents); | |
657 | *mode = 0; | |
658 | retval = FOUND; | |
659 | goto done; | |
660 | } | |
661 | ||
662 | if (remainder) | |
663 | len = first_slash - namebuf.buf; | |
664 | else | |
665 | len = namebuf.len; | |
666 | ||
667 | contents_start = contents; | |
668 | ||
669 | parent = &parents[parents_nr - 1]; | |
670 | init_tree_desc(&t, parent->tree, parent->size); | |
671 | strbuf_splice(&namebuf, 0, len, | |
672 | contents_start, link_len); | |
673 | if (remainder) | |
674 | namebuf.buf[link_len] = '/'; | |
675 | free(contents); | |
676 | } | |
677 | } | |
678 | done: | |
679 | for (i = 0; i < parents_nr; i++) | |
680 | free(parents[i].tree); | |
681 | free(parents); | |
682 | ||
683 | strbuf_release(&namebuf); | |
684 | return retval; | |
685 | } | |
686 | ||
93d93537 NTND |
687 | static int match_entry(const struct pathspec_item *item, |
688 | const struct name_entry *entry, int pathlen, | |
58c4d666 | 689 | const char *match, int matchlen, |
c3a47ca9 | 690 | enum interesting *never_interesting) |
58c4d666 NTND |
691 | { |
692 | int m = -1; /* signals that we haven't called strncmp() */ | |
693 | ||
93d93537 NTND |
694 | if (item->magic & PATHSPEC_ICASE) |
695 | /* | |
696 | * "Never interesting" trick requires exact | |
697 | * matching. We could do something clever with inexact | |
698 | * matching, but it's trickier (and not to forget that | |
699 | * strcasecmp is locale-dependent, at least in | |
700 | * glibc). Just disable it for now. It can't be worse | |
701 | * than the wildcard's codepath of '[Tt][Hi][Is][Ss]' | |
702 | * pattern. | |
703 | */ | |
704 | *never_interesting = entry_not_interesting; | |
705 | else if (*never_interesting != entry_not_interesting) { | |
58c4d666 NTND |
706 | /* |
707 | * We have not seen any match that sorts later | |
708 | * than the current path. | |
709 | */ | |
710 | ||
711 | /* | |
712 | * Does match sort strictly earlier than path | |
713 | * with their common parts? | |
714 | */ | |
715 | m = strncmp(match, entry->path, | |
716 | (matchlen < pathlen) ? matchlen : pathlen); | |
717 | if (m < 0) | |
718 | return 0; | |
719 | ||
720 | /* | |
721 | * If we come here even once, that means there is at | |
722 | * least one pathspec that would sort equal to or | |
723 | * later than the path we are currently looking at. | |
724 | * In other words, if we have never reached this point | |
725 | * after iterating all pathspecs, it means all | |
726 | * pathspecs are either outside of base, or inside the | |
727 | * base but sorts strictly earlier than the current | |
728 | * one. In either case, they will never match the | |
729 | * subsequent entries. In such a case, we initialized | |
730 | * the variable to -1 and that is what will be | |
731 | * returned, allowing the caller to terminate early. | |
732 | */ | |
c3a47ca9 | 733 | *never_interesting = entry_not_interesting; |
58c4d666 NTND |
734 | } |
735 | ||
736 | if (pathlen > matchlen) | |
737 | return 0; | |
738 | ||
739 | if (matchlen > pathlen) { | |
740 | if (match[pathlen] != '/') | |
741 | return 0; | |
74b4f7f2 | 742 | if (!S_ISDIR(entry->mode) && !S_ISGITLINK(entry->mode)) |
58c4d666 NTND |
743 | return 0; |
744 | } | |
745 | ||
746 | if (m == -1) | |
747 | /* | |
748 | * we cheated and did not do strncmp(), so we do | |
749 | * that here. | |
750 | */ | |
93d93537 | 751 | m = ps_strncmp(item, match, entry->path, pathlen); |
58c4d666 NTND |
752 | |
753 | /* | |
754 | * If common part matched earlier then it is a hit, | |
755 | * because we rejected the case where path is not a | |
756 | * leading directory and is shorter than match. | |
757 | */ | |
758 | if (!m) | |
93d93537 NTND |
759 | /* |
760 | * match_entry does not check if the prefix part is | |
761 | * matched case-sensitively. If the entry is a | |
762 | * directory and part of prefix, it'll be rematched | |
763 | * eventually by basecmp with special treatment for | |
764 | * the prefix. | |
765 | */ | |
58c4d666 NTND |
766 | return 1; |
767 | ||
768 | return 0; | |
769 | } | |
770 | ||
93d93537 NTND |
771 | /* :(icase)-aware string compare */ |
772 | static int basecmp(const struct pathspec_item *item, | |
773 | const char *base, const char *match, int len) | |
774 | { | |
775 | if (item->magic & PATHSPEC_ICASE) { | |
776 | int ret, n = len > item->prefix ? item->prefix : len; | |
777 | ret = strncmp(base, match, n); | |
778 | if (ret) | |
779 | return ret; | |
780 | base += n; | |
781 | match += n; | |
782 | len -= n; | |
783 | } | |
784 | return ps_strncmp(item, base, match, len); | |
785 | } | |
786 | ||
787 | static int match_dir_prefix(const struct pathspec_item *item, | |
788 | const char *base, | |
58c4d666 NTND |
789 | const char *match, int matchlen) |
790 | { | |
93d93537 | 791 | if (basecmp(item, base, match, matchlen)) |
58c4d666 NTND |
792 | return 0; |
793 | ||
794 | /* | |
795 | * If the base is a subdirectory of a path which | |
796 | * was specified, all of them are interesting. | |
797 | */ | |
798 | if (!matchlen || | |
799 | base[matchlen] == '/' || | |
800 | match[matchlen - 1] == '/') | |
801 | return 1; | |
802 | ||
803 | /* Just a random prefix match */ | |
804 | return 0; | |
805 | } | |
806 | ||
c904cd89 NTND |
807 | /* |
808 | * Perform matching on the leading non-wildcard part of | |
809 | * pathspec. item->nowildcard_len must be greater than zero. Return | |
810 | * non-zero if base is matched. | |
811 | */ | |
812 | static int match_wildcard_base(const struct pathspec_item *item, | |
813 | const char *base, int baselen, | |
814 | int *matched) | |
815 | { | |
816 | const char *match = item->match; | |
817 | /* the wildcard part is not considered in this function */ | |
818 | int matchlen = item->nowildcard_len; | |
819 | ||
820 | if (baselen) { | |
821 | int dirlen; | |
822 | /* | |
823 | * Return early if base is longer than the | |
824 | * non-wildcard part but it does not match. | |
825 | */ | |
826 | if (baselen >= matchlen) { | |
827 | *matched = matchlen; | |
93d93537 | 828 | return !basecmp(item, base, match, matchlen); |
c904cd89 NTND |
829 | } |
830 | ||
831 | dirlen = matchlen; | |
832 | while (dirlen && match[dirlen - 1] != '/') | |
833 | dirlen--; | |
834 | ||
835 | /* | |
836 | * Return early if base is shorter than the | |
837 | * non-wildcard part but it does not match. Note that | |
838 | * base ends with '/' so we are sure it really matches | |
839 | * directory | |
840 | */ | |
93d93537 | 841 | if (basecmp(item, base, match, baselen)) |
c904cd89 NTND |
842 | return 0; |
843 | *matched = baselen; | |
844 | } else | |
845 | *matched = 0; | |
846 | /* | |
847 | * we could have checked entry against the non-wildcard part | |
848 | * that is not in base and does similar never_interesting | |
849 | * optimization as in match_entry. For now just be happy with | |
850 | * base comparison. | |
851 | */ | |
852 | return entry_interesting; | |
853 | } | |
854 | ||
2c389fc8 NTND |
855 | /* |
856 | * Is a tree entry interesting given the pathspec we have? | |
857 | * | |
1376e507 NTND |
858 | * Pre-condition: either baselen == base_offset (i.e. empty path) |
859 | * or base[baselen-1] == '/' (i.e. with trailing slash). | |
2c389fc8 | 860 | */ |
ef79b1f8 NTND |
861 | static enum interesting do_match(const struct name_entry *entry, |
862 | struct strbuf *base, int base_offset, | |
863 | const struct pathspec *ps, | |
864 | int exclude) | |
2c389fc8 NTND |
865 | { |
866 | int i; | |
1376e507 | 867 | int pathlen, baselen = base->len - base_offset; |
c3a47ca9 | 868 | enum interesting never_interesting = ps->has_wildcard ? |
d688cf07 | 869 | entry_not_interesting : all_entries_not_interesting; |
2c389fc8 | 870 | |
5c6933d2 NTND |
871 | GUARD_PATHSPEC(ps, |
872 | PATHSPEC_FROMTOP | | |
873 | PATHSPEC_MAXDEPTH | | |
bd30c2e4 | 874 | PATHSPEC_LITERAL | |
93d93537 | 875 | PATHSPEC_GLOB | |
ef79b1f8 NTND |
876 | PATHSPEC_ICASE | |
877 | PATHSPEC_EXCLUDE); | |
8f4f8f45 | 878 | |
bc96cc87 | 879 | if (!ps->nr) { |
6330a171 NTND |
880 | if (!ps->recursive || |
881 | !(ps->magic & PATHSPEC_MAXDEPTH) || | |
882 | ps->max_depth == -1) | |
d688cf07 NTND |
883 | return all_entries_interesting; |
884 | return within_depth(base->buf + base_offset, baselen, | |
885 | !!S_ISDIR(entry->mode), | |
886 | ps->max_depth) ? | |
887 | entry_interesting : entry_not_interesting; | |
bc96cc87 | 888 | } |
2c389fc8 | 889 | |
0de16337 | 890 | pathlen = tree_entry_len(entry); |
2c389fc8 | 891 | |
1376e507 | 892 | for (i = ps->nr - 1; i >= 0; i--) { |
2c389fc8 NTND |
893 | const struct pathspec_item *item = ps->items+i; |
894 | const char *match = item->match; | |
1376e507 | 895 | const char *base_str = base->buf + base_offset; |
c904cd89 | 896 | int matchlen = item->len, matched = 0; |
2c389fc8 | 897 | |
ef79b1f8 NTND |
898 | if ((!exclude && item->magic & PATHSPEC_EXCLUDE) || |
899 | ( exclude && !(item->magic & PATHSPEC_EXCLUDE))) | |
900 | continue; | |
901 | ||
2c389fc8 NTND |
902 | if (baselen >= matchlen) { |
903 | /* If it doesn't match, move along... */ | |
93d93537 | 904 | if (!match_dir_prefix(item, base_str, match, matchlen)) |
d38f2809 | 905 | goto match_wildcards; |
bc96cc87 | 906 | |
6330a171 NTND |
907 | if (!ps->recursive || |
908 | !(ps->magic & PATHSPEC_MAXDEPTH) || | |
909 | ps->max_depth == -1) | |
d688cf07 | 910 | return all_entries_interesting; |
bc96cc87 | 911 | |
d688cf07 NTND |
912 | return within_depth(base_str + matchlen + 1, |
913 | baselen - matchlen - 1, | |
914 | !!S_ISDIR(entry->mode), | |
915 | ps->max_depth) ? | |
916 | entry_interesting : entry_not_interesting; | |
2c389fc8 NTND |
917 | } |
918 | ||
1b740923 | 919 | /* Either there must be no base, or the base must match. */ |
93d93537 NTND |
920 | if (baselen == 0 || !basecmp(item, base_str, match, baselen)) { |
921 | if (match_entry(item, entry, pathlen, | |
58c4d666 NTND |
922 | match + baselen, matchlen - baselen, |
923 | &never_interesting)) | |
d688cf07 | 924 | return entry_interesting; |
f1a2ddbb | 925 | |
170260ae | 926 | if (item->nowildcard_len < item->len) { |
bd30c2e4 | 927 | if (!git_fnmatch(item, match + baselen, entry->path, |
8c6abbcd | 928 | item->nowildcard_len - baselen)) |
d688cf07 | 929 | return entry_interesting; |
f1a2ddbb NTND |
930 | |
931 | /* | |
932 | * Match all directories. We'll try to | |
933 | * match files later on. | |
934 | */ | |
935 | if (ps->recursive && S_ISDIR(entry->mode)) | |
d688cf07 | 936 | return entry_interesting; |
f1a2ddbb NTND |
937 | } |
938 | ||
939 | continue; | |
2c389fc8 | 940 | } |
d38f2809 NTND |
941 | |
942 | match_wildcards: | |
170260ae | 943 | if (item->nowildcard_len == item->len) |
d38f2809 NTND |
944 | continue; |
945 | ||
c904cd89 NTND |
946 | if (item->nowildcard_len && |
947 | !match_wildcard_base(item, base_str, baselen, &matched)) | |
e4ddb057 | 948 | continue; |
c904cd89 | 949 | |
d38f2809 NTND |
950 | /* |
951 | * Concatenate base and entry->path into one and do | |
952 | * fnmatch() on it. | |
c904cd89 NTND |
953 | * |
954 | * While we could avoid concatenation in certain cases | |
955 | * [1], which saves a memcpy and potentially a | |
956 | * realloc, it turns out not worth it. Measurement on | |
957 | * linux-2.6 does not show any clear improvements, | |
958 | * partly because of the nowildcard_len optimization | |
959 | * in git_fnmatch(). Avoid micro-optimizations here. | |
960 | * | |
961 | * [1] if match_wildcard_base() says the base | |
962 | * directory is already matched, we only need to match | |
963 | * the rest, which is shorter so _in theory_ faster. | |
d38f2809 NTND |
964 | */ |
965 | ||
966 | strbuf_add(base, entry->path, pathlen); | |
967 | ||
bd30c2e4 | 968 | if (!git_fnmatch(item, match, base->buf + base_offset, |
8c6abbcd | 969 | item->nowildcard_len)) { |
1376e507 | 970 | strbuf_setlen(base, base_offset + baselen); |
d688cf07 | 971 | return entry_interesting; |
d38f2809 | 972 | } |
1376e507 | 973 | strbuf_setlen(base, base_offset + baselen); |
d38f2809 NTND |
974 | |
975 | /* | |
976 | * Match all directories. We'll try to match files | |
977 | * later on. | |
8c69c1f9 NTND |
978 | * max_depth is ignored but we may consider support it |
979 | * in future, see | |
980 | * http://thread.gmane.org/gmane.comp.version-control.git/163757/focus=163840 | |
d38f2809 NTND |
981 | */ |
982 | if (ps->recursive && S_ISDIR(entry->mode)) | |
d688cf07 | 983 | return entry_interesting; |
2c389fc8 NTND |
984 | } |
985 | return never_interesting; /* No matches */ | |
986 | } | |
ef79b1f8 NTND |
987 | |
988 | /* | |
989 | * Is a tree entry interesting given the pathspec we have? | |
990 | * | |
991 | * Pre-condition: either baselen == base_offset (i.e. empty path) | |
992 | * or base[baselen-1] == '/' (i.e. with trailing slash). | |
993 | */ | |
994 | enum interesting tree_entry_interesting(const struct name_entry *entry, | |
995 | struct strbuf *base, int base_offset, | |
996 | const struct pathspec *ps) | |
997 | { | |
998 | enum interesting positive, negative; | |
999 | positive = do_match(entry, base, base_offset, ps, 0); | |
1000 | ||
1001 | /* | |
1002 | * case | entry | positive | negative | result | |
1003 | * -----+-------+----------+----------+------- | |
1004 | * 1 | file | -1 | -1..2 | -1 | |
1005 | * 2 | file | 0 | -1..2 | 0 | |
1006 | * 3 | file | 1 | -1 | 1 | |
1007 | * 4 | file | 1 | 0 | 1 | |
1008 | * 5 | file | 1 | 1 | 0 | |
1009 | * 6 | file | 1 | 2 | 0 | |
1010 | * 7 | file | 2 | -1 | 2 | |
1011 | * 8 | file | 2 | 0 | 2 | |
1012 | * 9 | file | 2 | 1 | 0 | |
1013 | * 10 | file | 2 | 2 | -1 | |
1014 | * -----+-------+----------+----------+------- | |
1015 | * 11 | dir | -1 | -1..2 | -1 | |
1016 | * 12 | dir | 0 | -1..2 | 0 | |
1017 | * 13 | dir | 1 | -1 | 1 | |
1018 | * 14 | dir | 1 | 0 | 1 | |
1019 | * 15 | dir | 1 | 1 | 1 (*) | |
1020 | * 16 | dir | 1 | 2 | 0 | |
1021 | * 17 | dir | 2 | -1 | 2 | |
1022 | * 18 | dir | 2 | 0 | 2 | |
1023 | * 19 | dir | 2 | 1 | 1 (*) | |
1024 | * 20 | dir | 2 | 2 | -1 | |
1025 | * | |
1026 | * (*) An exclude pattern interested in a directory does not | |
1027 | * necessarily mean it will exclude all of the directory. In | |
1028 | * wildcard case, it can't decide until looking at individual | |
1029 | * files inside. So don't write such directories off yet. | |
1030 | */ | |
1031 | ||
1032 | if (!(ps->magic & PATHSPEC_EXCLUDE) || | |
1033 | positive <= entry_not_interesting) /* #1, #2, #11, #12 */ | |
1034 | return positive; | |
1035 | ||
1036 | negative = do_match(entry, base, base_offset, ps, 1); | |
1037 | ||
1038 | /* #3, #4, #7, #8, #13, #14, #17, #18 */ | |
1039 | if (negative <= entry_not_interesting) | |
1040 | return positive; | |
1041 | ||
1042 | /* #15, #19 */ | |
1043 | if (S_ISDIR(entry->mode) && | |
1044 | positive >= entry_interesting && | |
1045 | negative == entry_interesting) | |
1046 | return entry_interesting; | |
1047 | ||
1048 | if ((positive == entry_interesting && | |
1049 | negative >= entry_interesting) || /* #5, #6, #16 */ | |
1050 | (positive == all_entries_interesting && | |
1051 | negative == entry_interesting)) /* #9 */ | |
1052 | return entry_not_interesting; | |
1053 | ||
1054 | return all_entries_not_interesting; /* #10, #20 */ | |
1055 | } |