]>
Commit | Line | Data |
---|---|---|
5e3f94df | 1 | #include "git-compat-util.h" |
1b0c7174 | 2 | #include "tree-walk.h" |
bc96cc87 | 3 | #include "dir.h" |
f394e093 | 4 | #include "gettext.h" |
41771fa4 | 5 | #include "hex.h" |
87bed179 | 6 | #include "object-file.h" |
a034e910 | 7 | #include "object-store-ll.h" |
74ea5c95 | 8 | #include "trace2.h" |
8e440259 | 9 | #include "tree.h" |
64acde94 | 10 | #include "pathspec.h" |
da8be8ce | 11 | #include "json-writer.h" |
f1f63a48 | 12 | #include "environment.h" |
1b0c7174 | 13 | |
4651ece8 LT |
14 | static const char *get_mode(const char *str, unsigned int *modep) |
15 | { | |
16 | unsigned char c; | |
17 | unsigned int mode = 0; | |
18 | ||
64cc1c09 MK |
19 | if (*str == ' ') |
20 | return NULL; | |
21 | ||
4651ece8 LT |
22 | while ((c = *str++) != ' ') { |
23 | if (c < '0' || c > '7') | |
24 | return NULL; | |
25 | mode = (mode << 3) + (c - '0'); | |
26 | } | |
27 | *modep = mode; | |
28 | return str; | |
29 | } | |
30 | ||
8354fa3d | 31 | static int decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size, struct strbuf *err) |
4651ece8 LT |
32 | { |
33 | const char *path; | |
34 | unsigned int mode, len; | |
83e4b757 | 35 | const unsigned hashsz = the_hash_algo->rawsz; |
4651ece8 | 36 | |
83e4b757 | 37 | if (size < hashsz + 3 || buf[size - (hashsz + 1)]) { |
8354fa3d DT |
38 | strbuf_addstr(err, _("too-short tree object")); |
39 | return -1; | |
40 | } | |
64cc1c09 | 41 | |
4651ece8 | 42 | path = get_mode(buf, &mode); |
8354fa3d DT |
43 | if (!path) { |
44 | strbuf_addstr(err, _("malformed mode in tree entry")); | |
45 | return -1; | |
46 | } | |
47 | if (!*path) { | |
48 | strbuf_addstr(err, _("empty filename in tree entry")); | |
49 | return -1; | |
50 | } | |
4651ece8 LT |
51 | len = strlen(path) + 1; |
52 | ||
53 | /* Initialize the descriptor entry */ | |
54 | desc->entry.path = path; | |
ec18b10b | 55 | desc->entry.mode = (desc->flags & TREE_DESC_RAW_MODES) ? mode : canon_mode(mode); |
ea82b2a0 | 56 | desc->entry.pathlen = len - 1; |
92e2cab9 | 57 | oidread(&desc->entry.oid, (const unsigned char *)path + len); |
8354fa3d DT |
58 | |
59 | return 0; | |
4651ece8 LT |
60 | } |
61 | ||
ec18b10b JK |
62 | static int init_tree_desc_internal(struct tree_desc *desc, const void *buffer, |
63 | unsigned long size, struct strbuf *err, | |
64 | enum tree_desc_flags flags) | |
6fda5e51 LT |
65 | { |
66 | desc->buffer = buffer; | |
67 | desc->size = size; | |
ec18b10b | 68 | desc->flags = flags; |
4651ece8 | 69 | if (size) |
8354fa3d DT |
70 | return decode_tree_entry(desc, buffer, size, err); |
71 | return 0; | |
72 | } | |
73 | ||
74 | void init_tree_desc(struct tree_desc *desc, const void *buffer, unsigned long size) | |
75 | { | |
76 | struct strbuf err = STRBUF_INIT; | |
ec18b10b | 77 | if (init_tree_desc_internal(desc, buffer, size, &err, 0)) |
8354fa3d DT |
78 | die("%s", err.buf); |
79 | strbuf_release(&err); | |
80 | } | |
81 | ||
ec18b10b JK |
82 | int init_tree_desc_gently(struct tree_desc *desc, const void *buffer, unsigned long size, |
83 | enum tree_desc_flags flags) | |
8354fa3d DT |
84 | { |
85 | struct strbuf err = STRBUF_INIT; | |
ec18b10b | 86 | int result = init_tree_desc_internal(desc, buffer, size, &err, flags); |
8354fa3d DT |
87 | if (result) |
88 | error("%s", err.buf); | |
89 | strbuf_release(&err); | |
90 | return result; | |
6fda5e51 LT |
91 | } |
92 | ||
5e575807 NTND |
93 | void *fill_tree_descriptor(struct repository *r, |
94 | struct tree_desc *desc, | |
95 | const struct object_id *oid) | |
1b0c7174 JH |
96 | { |
97 | unsigned long size = 0; | |
98 | void *buf = NULL; | |
99 | ||
5c377d3d | 100 | if (oid) { |
6aea6bae | 101 | buf = read_object_with_reference(r, oid, OBJ_TREE, &size, NULL); |
1b0c7174 | 102 | if (!buf) |
5c377d3d | 103 | die("unable to read tree %s", oid_to_hex(oid)); |
1b0c7174 | 104 | } |
6fda5e51 | 105 | init_tree_desc(desc, buf, size); |
1b0c7174 JH |
106 | return buf; |
107 | } | |
108 | ||
1b0c7174 JH |
109 | static void entry_clear(struct name_entry *a) |
110 | { | |
111 | memset(a, 0, sizeof(*a)); | |
112 | } | |
113 | ||
114 | static void entry_extract(struct tree_desc *t, struct name_entry *a) | |
115 | { | |
4651ece8 | 116 | *a = t->entry; |
1b0c7174 JH |
117 | } |
118 | ||
8354fa3d | 119 | static int update_tree_entry_internal(struct tree_desc *desc, struct strbuf *err) |
1b0c7174 | 120 | { |
6fda5e51 | 121 | const void *buf = desc->buffer; |
ea82b2a0 | 122 | const unsigned char *end = (const unsigned char *)desc->entry.path + desc->entry.pathlen + 1 + the_hash_algo->rawsz; |
1b0c7174 | 123 | unsigned long size = desc->size; |
4651ece8 | 124 | unsigned long len = end - (const unsigned char *)buf; |
1b0c7174 JH |
125 | |
126 | if (size < len) | |
2edffef2 | 127 | die(_("too-short tree file")); |
4651ece8 LT |
128 | buf = end; |
129 | size -= len; | |
130 | desc->buffer = buf; | |
131 | desc->size = size; | |
132 | if (size) | |
8354fa3d DT |
133 | return decode_tree_entry(desc, buf, size, err); |
134 | return 0; | |
135 | } | |
136 | ||
137 | void update_tree_entry(struct tree_desc *desc) | |
138 | { | |
139 | struct strbuf err = STRBUF_INIT; | |
140 | if (update_tree_entry_internal(desc, &err)) | |
141 | die("%s", err.buf); | |
142 | strbuf_release(&err); | |
143 | } | |
144 | ||
145 | int update_tree_entry_gently(struct tree_desc *desc) | |
146 | { | |
147 | struct strbuf err = STRBUF_INIT; | |
148 | if (update_tree_entry_internal(desc, &err)) { | |
149 | error("%s", err.buf); | |
150 | strbuf_release(&err); | |
151 | /* Stop processing this tree after error */ | |
152 | desc->size = 0; | |
153 | return -1; | |
154 | } | |
155 | strbuf_release(&err); | |
156 | return 0; | |
1b0c7174 JH |
157 | } |
158 | ||
4c068a98 LT |
159 | int tree_entry(struct tree_desc *desc, struct name_entry *entry) |
160 | { | |
4651ece8 | 161 | if (!desc->size) |
4c068a98 LT |
162 | return 0; |
163 | ||
4651ece8 LT |
164 | *entry = desc->entry; |
165 | update_tree_entry(desc); | |
4c068a98 LT |
166 | return 1; |
167 | } | |
168 | ||
8354fa3d DT |
169 | int tree_entry_gently(struct tree_desc *desc, struct name_entry *entry) |
170 | { | |
171 | if (!desc->size) | |
172 | return 0; | |
173 | ||
174 | *entry = desc->entry; | |
175 | if (update_tree_entry_gently(desc)) | |
176 | return 0; | |
177 | return 1; | |
178 | } | |
179 | ||
da8be8ce DS |
180 | static int traverse_trees_atexit_registered; |
181 | static int traverse_trees_count; | |
182 | static int traverse_trees_cur_depth; | |
183 | static int traverse_trees_max_depth; | |
184 | ||
185 | static void trace2_traverse_trees_statistics_atexit(void) | |
186 | { | |
187 | struct json_writer jw = JSON_WRITER_INIT; | |
188 | ||
189 | jw_object_begin(&jw, 0); | |
190 | jw_object_intmax(&jw, "traverse_trees_count", traverse_trees_count); | |
191 | jw_object_intmax(&jw, "traverse_trees_max_depth", traverse_trees_max_depth); | |
192 | jw_end(&jw); | |
193 | ||
194 | trace2_data_json("traverse_trees", the_repository, "statistics", &jw); | |
195 | ||
196 | jw_release(&jw); | |
197 | } | |
198 | ||
40d934df LT |
199 | void setup_traverse_info(struct traverse_info *info, const char *base) |
200 | { | |
37806080 | 201 | size_t pathlen = strlen(base); |
bcbe5a51 | 202 | static struct traverse_info dummy; |
40d934df LT |
203 | |
204 | memset(info, 0, sizeof(*info)); | |
205 | if (pathlen && base[pathlen-1] == '/') | |
206 | pathlen--; | |
207 | info->pathlen = pathlen ? pathlen + 1 : 0; | |
90553847 JK |
208 | info->name = base; |
209 | info->namelen = pathlen; | |
947208b7 | 210 | if (pathlen) |
bcbe5a51 | 211 | info->prev = &dummy; |
da8be8ce DS |
212 | |
213 | if (trace2_is_enabled() && !traverse_trees_atexit_registered) { | |
214 | atexit(trace2_traverse_trees_statistics_atexit); | |
215 | traverse_trees_atexit_registered = 1; | |
216 | } | |
40d934df LT |
217 | } |
218 | ||
5aa02f98 JK |
219 | char *make_traverse_path(char *path, size_t pathlen, |
220 | const struct traverse_info *info, | |
90553847 | 221 | const char *name, size_t namelen) |
40d934df | 222 | { |
5aa02f98 JK |
223 | /* Always points to the end of the name we're about to add */ |
224 | size_t pos = st_add(info->pathlen, namelen); | |
40d934df | 225 | |
5aa02f98 JK |
226 | if (pos >= pathlen) |
227 | BUG("too small buffer passed to make_traverse_path"); | |
40d934df | 228 | |
5aa02f98 | 229 | path[pos] = 0; |
40d934df | 230 | for (;;) { |
5aa02f98 JK |
231 | if (pos < namelen) |
232 | BUG("traverse_info pathlen does not match strings"); | |
233 | pos -= namelen; | |
234 | memcpy(path + pos, name, namelen); | |
235 | ||
236 | if (!pos) | |
40d934df | 237 | break; |
5aa02f98 JK |
238 | path[--pos] = '/'; |
239 | ||
240 | if (!info) | |
241 | BUG("traverse_info ran out of list items"); | |
90553847 JK |
242 | name = info->name; |
243 | namelen = info->namelen; | |
40d934df | 244 | info = info->prev; |
40d934df LT |
245 | } |
246 | return path; | |
247 | } | |
248 | ||
c43ab062 JK |
249 | void strbuf_make_traverse_path(struct strbuf *out, |
250 | const struct traverse_info *info, | |
251 | const char *name, size_t namelen) | |
252 | { | |
253 | size_t len = traverse_path_len(info, namelen); | |
254 | ||
255 | strbuf_grow(out, len); | |
5aa02f98 JK |
256 | make_traverse_path(out->buf + out->len, out->alloc - out->len, |
257 | info, name, namelen); | |
c43ab062 JK |
258 | strbuf_setlen(out, out->len + len); |
259 | } | |
260 | ||
1ee26571 JH |
261 | struct tree_desc_skip { |
262 | struct tree_desc_skip *prev; | |
263 | const void *ptr; | |
264 | }; | |
265 | ||
266 | struct tree_desc_x { | |
267 | struct tree_desc d; | |
268 | struct tree_desc_skip *skip; | |
269 | }; | |
270 | ||
1ee26571 JH |
271 | static int check_entry_match(const char *a, int a_len, const char *b, int b_len) |
272 | { | |
273 | /* | |
274 | * The caller wants to pick *a* from a tree or nothing. | |
275 | * We are looking at *b* in a tree. | |
276 | * | |
277 | * (0) If a and b are the same name, we are trivially happy. | |
278 | * | |
279 | * There are three possibilities where *a* could be hiding | |
280 | * behind *b*. | |
281 | * | |
282 | * (1) *a* == "t", *b* == "ab" i.e. *b* sorts earlier than *a* no | |
283 | * matter what. | |
284 | * (2) *a* == "t", *b* == "t-2" and "t" is a subtree in the tree; | |
285 | * (3) *a* == "t-2", *b* == "t" and "t-2" is a blob in the tree. | |
286 | * | |
287 | * Otherwise we know *a* won't appear in the tree without | |
288 | * scanning further. | |
289 | */ | |
290 | ||
291 | int cmp = name_compare(a, a_len, b, b_len); | |
292 | ||
293 | /* Most common case first -- reading sync'd trees */ | |
294 | if (!cmp) | |
295 | return cmp; | |
296 | ||
297 | if (0 < cmp) { | |
298 | /* a comes after b; it does not matter if it is case (3) | |
299 | if (b_len < a_len && !memcmp(a, b, b_len) && a[b_len] < '/') | |
300 | return 1; | |
301 | */ | |
302 | return 1; /* keep looking */ | |
303 | } | |
304 | ||
305 | /* b comes after a; are we looking at case (2)? */ | |
306 | if (a_len < b_len && !memcmp(a, b, a_len) && b[a_len] < '/') | |
307 | return 1; /* keep looking */ | |
308 | ||
309 | return -1; /* a cannot appear in the tree */ | |
310 | } | |
311 | ||
312 | /* | |
313 | * From the extended tree_desc, extract the first name entry, while | |
314 | * paying attention to the candidate "first" name. Most importantly, | |
315 | * when looking for an entry, if there are entries that sorts earlier | |
316 | * in the tree object representation than that name, skip them and | |
317 | * process the named entry first. We will remember that we haven't | |
318 | * processed the first entry yet, and in the later call skip the | |
319 | * entry we processed early when update_extended_entry() is called. | |
320 | * | |
321 | * E.g. if the underlying tree object has these entries: | |
322 | * | |
323 | * blob "t-1" | |
324 | * blob "t-2" | |
325 | * tree "t" | |
326 | * blob "t=1" | |
327 | * | |
328 | * and the "first" asks for "t", remember that we still need to | |
329 | * process "t-1" and "t-2" but extract "t". After processing the | |
330 | * entry "t" from this call, the caller will let us know by calling | |
331 | * update_extended_entry() that we can remember "t" has been processed | |
332 | * already. | |
333 | */ | |
334 | ||
335 | static void extended_entry_extract(struct tree_desc_x *t, | |
336 | struct name_entry *a, | |
337 | const char *first, | |
338 | int first_len) | |
339 | { | |
340 | const char *path; | |
341 | int len; | |
342 | struct tree_desc probe; | |
343 | struct tree_desc_skip *skip; | |
344 | ||
345 | /* | |
346 | * Extract the first entry from the tree_desc, but skip the | |
347 | * ones that we already returned in earlier rounds. | |
348 | */ | |
349 | while (1) { | |
350 | if (!t->d.size) { | |
351 | entry_clear(a); | |
352 | break; /* not found */ | |
353 | } | |
354 | entry_extract(&t->d, a); | |
355 | for (skip = t->skip; skip; skip = skip->prev) | |
356 | if (a->path == skip->ptr) | |
357 | break; /* found */ | |
358 | if (!skip) | |
359 | break; | |
360 | /* We have processed this entry already. */ | |
361 | update_tree_entry(&t->d); | |
362 | } | |
363 | ||
364 | if (!first || !a->path) | |
365 | return; | |
366 | ||
367 | /* | |
368 | * The caller wants "first" from this tree, or nothing. | |
369 | */ | |
370 | path = a->path; | |
0de16337 | 371 | len = tree_entry_len(a); |
1ee26571 JH |
372 | switch (check_entry_match(first, first_len, path, len)) { |
373 | case -1: | |
374 | entry_clear(a); | |
375 | case 0: | |
376 | return; | |
377 | default: | |
378 | break; | |
379 | } | |
380 | ||
381 | /* | |
382 | * We need to look-ahead -- we suspect that a subtree whose | |
383 | * name is "first" may be hiding behind the current entry "path". | |
384 | */ | |
385 | probe = t->d; | |
386 | while (probe.size) { | |
387 | entry_extract(&probe, a); | |
388 | path = a->path; | |
0de16337 | 389 | len = tree_entry_len(a); |
1ee26571 JH |
390 | switch (check_entry_match(first, first_len, path, len)) { |
391 | case -1: | |
392 | entry_clear(a); | |
393 | case 0: | |
394 | return; | |
395 | default: | |
396 | update_tree_entry(&probe); | |
397 | break; | |
398 | } | |
399 | /* keep looking */ | |
400 | } | |
401 | entry_clear(a); | |
402 | } | |
403 | ||
404 | static void update_extended_entry(struct tree_desc_x *t, struct name_entry *a) | |
405 | { | |
406 | if (t->d.entry.path == a->path) { | |
407 | update_tree_entry(&t->d); | |
408 | } else { | |
409 | /* we have returned this entry early */ | |
410 | struct tree_desc_skip *skip = xmalloc(sizeof(*skip)); | |
411 | skip->ptr = a->path; | |
412 | skip->prev = t->skip; | |
413 | t->skip = skip; | |
414 | } | |
415 | } | |
416 | ||
417 | static void free_extended_entry(struct tree_desc_x *t) | |
418 | { | |
419 | struct tree_desc_skip *p, *s; | |
420 | ||
421 | for (s = t->skip; s; s = p) { | |
422 | p = s->prev; | |
423 | free(s); | |
424 | } | |
425 | } | |
426 | ||
67022e02 NTND |
427 | static inline int prune_traversal(struct index_state *istate, |
428 | struct name_entry *e, | |
2842c0f9 JH |
429 | struct traverse_info *info, |
430 | struct strbuf *base, | |
431 | int still_interesting) | |
432 | { | |
433 | if (!info->pathspec || still_interesting == 2) | |
434 | return 2; | |
435 | if (still_interesting < 0) | |
436 | return still_interesting; | |
67022e02 | 437 | return tree_entry_interesting(istate, e, base, |
0ad927e9 | 438 | info->pathspec); |
2842c0f9 JH |
439 | } |
440 | ||
67022e02 NTND |
441 | int traverse_trees(struct index_state *istate, |
442 | int n, struct tree_desc *t, | |
443 | struct traverse_info *info) | |
1b0c7174 | 444 | { |
c7cd0e34 | 445 | int ret = 0; |
59c4c7d1 | 446 | struct name_entry *entry; |
1ee26571 | 447 | int i; |
59c4c7d1 | 448 | struct tree_desc_x *tx; |
2842c0f9 JH |
449 | struct strbuf base = STRBUF_INIT; |
450 | int interesting = 1; | |
d9c2bd56 | 451 | char *traverse_path; |
1ee26571 | 452 | |
f1f63a48 JK |
453 | if (traverse_trees_cur_depth > max_allowed_tree_depth) |
454 | return error("exceeded maximum allowed tree depth"); | |
455 | ||
da8be8ce DS |
456 | traverse_trees_count++; |
457 | traverse_trees_cur_depth++; | |
458 | ||
459 | if (traverse_trees_cur_depth > traverse_trees_max_depth) | |
460 | traverse_trees_max_depth = traverse_trees_cur_depth; | |
461 | ||
59c4c7d1 JK |
462 | ALLOC_ARRAY(entry, n); |
463 | ALLOC_ARRAY(tx, n); | |
8dd40c04 JK |
464 | |
465 | for (i = 0; i < n; i++) { | |
1ee26571 | 466 | tx[i].d = t[i]; |
8dd40c04 JK |
467 | tx[i].skip = NULL; |
468 | } | |
1b0c7174 | 469 | |
2842c0f9 | 470 | if (info->prev) { |
c43ab062 JK |
471 | strbuf_make_traverse_path(&base, info->prev, |
472 | info->name, info->namelen); | |
473 | strbuf_addch(&base, '/'); | |
474 | traverse_path = xstrndup(base.buf, base.len); | |
d9c2bd56 | 475 | } else { |
90553847 | 476 | traverse_path = xstrndup(info->name, info->pathlen); |
2842c0f9 | 477 | } |
d9c2bd56 | 478 | info->traverse_path = traverse_path; |
1b0c7174 | 479 | for (;;) { |
a04f8196 | 480 | int trees_used; |
1ee26571 JH |
481 | unsigned long mask, dirmask; |
482 | const char *first = NULL; | |
483 | int first_len = 0; | |
27c0f768 | 484 | struct name_entry *e = NULL; |
1ee26571 JH |
485 | int len; |
486 | ||
487 | for (i = 0; i < n; i++) { | |
488 | e = entry + i; | |
489 | extended_entry_extract(tx + i, e, NULL, 0); | |
490 | } | |
1b0c7174 | 491 | |
1ee26571 JH |
492 | /* |
493 | * A tree may have "t-2" at the current location even | |
494 | * though it may have "t" that is a subtree behind it, | |
495 | * and another tree may return "t". We want to grab | |
496 | * all "t" from all trees to match in such a case. | |
497 | */ | |
1b0c7174 | 498 | for (i = 0; i < n; i++) { |
1ee26571 JH |
499 | e = entry + i; |
500 | if (!e->path) | |
1b0c7174 | 501 | continue; |
0de16337 | 502 | len = tree_entry_len(e); |
1ee26571 JH |
503 | if (!first) { |
504 | first = e->path; | |
505 | first_len = len; | |
506 | continue; | |
507 | } | |
508 | if (name_compare(e->path, len, first, first_len) < 0) { | |
509 | first = e->path; | |
510 | first_len = len; | |
511 | } | |
512 | } | |
513 | ||
514 | if (first) { | |
515 | for (i = 0; i < n; i++) { | |
516 | e = entry + i; | |
517 | extended_entry_extract(tx + i, e, first, first_len); | |
518 | /* Cull the ones that are not the earliest */ | |
519 | if (!e->path) | |
1b0c7174 | 520 | continue; |
0de16337 | 521 | len = tree_entry_len(e); |
1ee26571 JH |
522 | if (name_compare(e->path, len, first, first_len)) |
523 | entry_clear(e); | |
1b0c7174 | 524 | } |
1ee26571 JH |
525 | } |
526 | ||
527 | /* Now we have in entry[i] the earliest name from the trees */ | |
528 | mask = 0; | |
529 | dirmask = 0; | |
530 | for (i = 0; i < n; i++) { | |
531 | if (!entry[i].path) | |
532 | continue; | |
1b0c7174 | 533 | mask |= 1ul << i; |
91e4f036 LT |
534 | if (S_ISDIR(entry[i].mode)) |
535 | dirmask |= 1ul << i; | |
2842c0f9 | 536 | e = &entry[i]; |
1b0c7174 JH |
537 | } |
538 | if (!mask) | |
539 | break; | |
67022e02 | 540 | interesting = prune_traversal(istate, e, info, &base, interesting); |
2842c0f9 JH |
541 | if (interesting < 0) |
542 | break; | |
543 | if (interesting) { | |
a04f8196 SB |
544 | trees_used = info->fn(n, mask, dirmask, entry, info); |
545 | if (trees_used < 0) { | |
c7cd0e34 | 546 | ret = trees_used; |
2842c0f9 JH |
547 | if (!info->show_all_errors) |
548 | break; | |
549 | } | |
a04f8196 | 550 | mask &= trees_used; |
e6c111b4 | 551 | } |
1ee26571 | 552 | for (i = 0; i < n; i++) |
5803c6f8 | 553 | if (mask & (1ul << i)) |
1ee26571 | 554 | update_extended_entry(tx + i, entry + i); |
1b0c7174 | 555 | } |
1ee26571 JH |
556 | for (i = 0; i < n; i++) |
557 | free_extended_entry(tx + i); | |
59c4c7d1 JK |
558 | free(tx); |
559 | free(entry); | |
d9c2bd56 DT |
560 | free(traverse_path); |
561 | info->traverse_path = NULL; | |
2842c0f9 | 562 | strbuf_release(&base); |
da8be8ce DS |
563 | |
564 | traverse_trees_cur_depth--; | |
c7cd0e34 | 565 | return ret; |
1b0c7174 JH |
566 | } |
567 | ||
275721c2 DT |
568 | struct dir_state { |
569 | void *tree; | |
570 | unsigned long size; | |
3b683bcf | 571 | struct object_id oid; |
275721c2 DT |
572 | }; |
573 | ||
50ddb089 NTND |
574 | static int find_tree_entry(struct repository *r, struct tree_desc *t, |
575 | const char *name, struct object_id *result, | |
576 | unsigned short *mode) | |
4dcff634 JH |
577 | { |
578 | int namelen = strlen(name); | |
579 | while (t->size) { | |
580 | const char *entry; | |
0a3faa45 | 581 | struct object_id oid; |
4dcff634 JH |
582 | int entrylen, cmp; |
583 | ||
0a3faa45 | 584 | oidcpy(&oid, tree_entry_extract(t, &entry, mode)); |
0de16337 | 585 | entrylen = tree_entry_len(&t->entry); |
4dcff634 | 586 | update_tree_entry(t); |
4dcff634 JH |
587 | if (entrylen > namelen) |
588 | continue; | |
589 | cmp = memcmp(name, entry, entrylen); | |
590 | if (cmp > 0) | |
591 | continue; | |
592 | if (cmp < 0) | |
593 | break; | |
594 | if (entrylen == namelen) { | |
0a3faa45 | 595 | oidcpy(result, &oid); |
4dcff634 JH |
596 | return 0; |
597 | } | |
598 | if (name[entrylen] != '/') | |
599 | continue; | |
600 | if (!S_ISDIR(*mode)) | |
601 | break; | |
602 | if (++entrylen == namelen) { | |
0a3faa45 | 603 | oidcpy(result, &oid); |
4dcff634 JH |
604 | return 0; |
605 | } | |
50ddb089 | 606 | return get_tree_entry(r, &oid, name + entrylen, result, mode); |
4dcff634 JH |
607 | } |
608 | return -1; | |
609 | } | |
610 | ||
50ddb089 NTND |
611 | int get_tree_entry(struct repository *r, |
612 | const struct object_id *tree_oid, | |
613 | const char *name, | |
614 | struct object_id *oid, | |
615 | unsigned short *mode) | |
4dcff634 JH |
616 | { |
617 | int retval; | |
618 | void *tree; | |
6fda5e51 | 619 | unsigned long size; |
916bc35b | 620 | struct object_id root; |
4dcff634 | 621 | |
6aea6bae | 622 | tree = read_object_with_reference(r, tree_oid, OBJ_TREE, &size, &root); |
4dcff634 JH |
623 | if (!tree) |
624 | return -1; | |
d93b7d1c JK |
625 | |
626 | if (name[0] == '\0') { | |
916bc35b | 627 | oidcpy(oid, &root); |
ef006503 | 628 | free(tree); |
d93b7d1c JK |
629 | return 0; |
630 | } | |
631 | ||
5fb8c05f JH |
632 | if (!size) { |
633 | retval = -1; | |
634 | } else { | |
635 | struct tree_desc t; | |
636 | init_tree_desc(&t, tree, size); | |
50ddb089 | 637 | retval = find_tree_entry(r, &t, name, oid, mode); |
5fb8c05f | 638 | } |
4dcff634 JH |
639 | free(tree); |
640 | return retval; | |
641 | } | |
2c389fc8 | 642 | |
275721c2 DT |
643 | /* |
644 | * This is Linux's built-in max for the number of symlinks to follow. | |
645 | * That limit, of course, does not affect git, but it's a reasonable | |
646 | * choice. | |
647 | */ | |
648 | #define GET_TREE_ENTRY_FOLLOW_SYMLINKS_MAX_LINKS 40 | |
649 | ||
650 | /** | |
651 | * Find a tree entry by following symlinks in tree_sha (which is | |
652 | * assumed to be the root of the repository). In the event that a | |
653 | * symlink points outside the repository (e.g. a link to /foo or a | |
654 | * root-level link to ../foo), the portion of the link which is | |
655 | * outside the repository will be returned in result_path, and *mode | |
656 | * will be set to 0. It is assumed that result_path is uninitialized. | |
657 | * If there are no symlinks, or the end result of the symlink chain | |
658 | * points to an object inside the repository, result will be filled in | |
659 | * with the sha1 of the found object, and *mode will hold the mode of | |
660 | * the object. | |
661 | * | |
d1dd94b3 | 662 | * See the code for enum get_oid_result for a description of |
275721c2 DT |
663 | * the return values. |
664 | */ | |
0dd1f0c3 NTND |
665 | enum get_oid_result get_tree_entry_follow_symlinks(struct repository *r, |
666 | struct object_id *tree_oid, const char *name, | |
667 | struct object_id *result, struct strbuf *result_path, | |
668 | unsigned short *mode) | |
275721c2 DT |
669 | { |
670 | int retval = MISSING_OBJECT; | |
671 | struct dir_state *parents = NULL; | |
672 | size_t parents_alloc = 0; | |
071bcaab | 673 | size_t i, parents_nr = 0; |
0d4a8b5b | 674 | struct object_id current_tree_oid; |
275721c2 DT |
675 | struct strbuf namebuf = STRBUF_INIT; |
676 | struct tree_desc t; | |
677 | int follows_remaining = GET_TREE_ENTRY_FOLLOW_SYMLINKS_MAX_LINKS; | |
275721c2 DT |
678 | |
679 | init_tree_desc(&t, NULL, 0UL); | |
275721c2 | 680 | strbuf_addstr(&namebuf, name); |
3b683bcf | 681 | oidcpy(¤t_tree_oid, tree_oid); |
275721c2 DT |
682 | |
683 | while (1) { | |
684 | int find_result; | |
685 | char *first_slash; | |
686 | char *remainder = NULL; | |
687 | ||
688 | if (!t.buffer) { | |
689 | void *tree; | |
0d4a8b5b | 690 | struct object_id root; |
275721c2 | 691 | unsigned long size; |
0dd1f0c3 | 692 | tree = read_object_with_reference(r, |
d3b4705a | 693 | ¤t_tree_oid, |
6aea6bae | 694 | OBJ_TREE, &size, |
02f0547e | 695 | &root); |
275721c2 DT |
696 | if (!tree) |
697 | goto done; | |
698 | ||
699 | ALLOC_GROW(parents, parents_nr + 1, parents_alloc); | |
700 | parents[parents_nr].tree = tree; | |
701 | parents[parents_nr].size = size; | |
3b683bcf | 702 | oidcpy(&parents[parents_nr].oid, &root); |
275721c2 DT |
703 | parents_nr++; |
704 | ||
705 | if (namebuf.buf[0] == '\0') { | |
3b683bcf | 706 | oidcpy(result, &root); |
275721c2 DT |
707 | retval = FOUND; |
708 | goto done; | |
709 | } | |
710 | ||
711 | if (!size) | |
712 | goto done; | |
713 | ||
714 | /* descend */ | |
715 | init_tree_desc(&t, tree, size); | |
716 | } | |
717 | ||
718 | /* Handle symlinks to e.g. a//b by removing leading slashes */ | |
719 | while (namebuf.buf[0] == '/') { | |
720 | strbuf_remove(&namebuf, 0, 1); | |
721 | } | |
722 | ||
723 | /* Split namebuf into a first component and a remainder */ | |
724 | if ((first_slash = strchr(namebuf.buf, '/'))) { | |
725 | *first_slash = 0; | |
726 | remainder = first_slash + 1; | |
727 | } | |
728 | ||
729 | if (!strcmp(namebuf.buf, "..")) { | |
730 | struct dir_state *parent; | |
731 | /* | |
732 | * We could end up with .. in the namebuf if it | |
733 | * appears in a symlink. | |
734 | */ | |
735 | ||
736 | if (parents_nr == 1) { | |
737 | if (remainder) | |
738 | *first_slash = '/'; | |
739 | strbuf_add(result_path, namebuf.buf, | |
740 | namebuf.len); | |
741 | *mode = 0; | |
742 | retval = FOUND; | |
743 | goto done; | |
744 | } | |
745 | parent = &parents[parents_nr - 1]; | |
746 | free(parent->tree); | |
747 | parents_nr--; | |
748 | parent = &parents[parents_nr - 1]; | |
749 | init_tree_desc(&t, parent->tree, parent->size); | |
750 | strbuf_remove(&namebuf, 0, remainder ? 3 : 2); | |
751 | continue; | |
752 | } | |
753 | ||
754 | /* We could end up here via a symlink to dir/.. */ | |
755 | if (namebuf.buf[0] == '\0') { | |
3b683bcf | 756 | oidcpy(result, &parents[parents_nr - 1].oid); |
275721c2 DT |
757 | retval = FOUND; |
758 | goto done; | |
759 | } | |
760 | ||
761 | /* Look up the first (or only) path component in the tree. */ | |
0dd1f0c3 | 762 | find_result = find_tree_entry(r, &t, namebuf.buf, |
916bc35b | 763 | ¤t_tree_oid, mode); |
275721c2 DT |
764 | if (find_result) { |
765 | goto done; | |
766 | } | |
767 | ||
768 | if (S_ISDIR(*mode)) { | |
769 | if (!remainder) { | |
3b683bcf | 770 | oidcpy(result, ¤t_tree_oid); |
275721c2 DT |
771 | retval = FOUND; |
772 | goto done; | |
773 | } | |
774 | /* Descend the tree */ | |
775 | t.buffer = NULL; | |
776 | strbuf_remove(&namebuf, 0, | |
777 | 1 + first_slash - namebuf.buf); | |
778 | } else if (S_ISREG(*mode)) { | |
779 | if (!remainder) { | |
3b683bcf | 780 | oidcpy(result, ¤t_tree_oid); |
275721c2 DT |
781 | retval = FOUND; |
782 | } else { | |
783 | retval = NOT_DIR; | |
784 | } | |
785 | goto done; | |
786 | } else if (S_ISLNK(*mode)) { | |
787 | /* Follow a symlink */ | |
788 | unsigned long link_len; | |
789 | size_t len; | |
790 | char *contents, *contents_start; | |
791 | struct dir_state *parent; | |
792 | enum object_type type; | |
793 | ||
794 | if (follows_remaining-- == 0) { | |
795 | /* Too many symlinks followed */ | |
796 | retval = SYMLINK_LOOP; | |
797 | goto done; | |
798 | } | |
799 | ||
800 | /* | |
801 | * At this point, we have followed at a least | |
802 | * one symlink, so on error we need to report this. | |
803 | */ | |
804 | retval = DANGLING_SYMLINK; | |
805 | ||
0dd1f0c3 NTND |
806 | contents = repo_read_object_file(r, |
807 | ¤t_tree_oid, &type, | |
b4f5aca4 | 808 | &link_len); |
275721c2 DT |
809 | |
810 | if (!contents) | |
811 | goto done; | |
812 | ||
813 | if (contents[0] == '/') { | |
814 | strbuf_addstr(result_path, contents); | |
815 | free(contents); | |
816 | *mode = 0; | |
817 | retval = FOUND; | |
818 | goto done; | |
819 | } | |
820 | ||
821 | if (remainder) | |
822 | len = first_slash - namebuf.buf; | |
823 | else | |
824 | len = namebuf.len; | |
825 | ||
826 | contents_start = contents; | |
827 | ||
828 | parent = &parents[parents_nr - 1]; | |
829 | init_tree_desc(&t, parent->tree, parent->size); | |
830 | strbuf_splice(&namebuf, 0, len, | |
831 | contents_start, link_len); | |
832 | if (remainder) | |
833 | namebuf.buf[link_len] = '/'; | |
834 | free(contents); | |
835 | } | |
836 | } | |
837 | done: | |
838 | for (i = 0; i < parents_nr; i++) | |
839 | free(parents[i].tree); | |
840 | free(parents); | |
841 | ||
842 | strbuf_release(&namebuf); | |
843 | return retval; | |
844 | } | |
845 | ||
93d93537 NTND |
846 | static int match_entry(const struct pathspec_item *item, |
847 | const struct name_entry *entry, int pathlen, | |
58c4d666 | 848 | const char *match, int matchlen, |
c3a47ca9 | 849 | enum interesting *never_interesting) |
58c4d666 NTND |
850 | { |
851 | int m = -1; /* signals that we haven't called strncmp() */ | |
852 | ||
93d93537 NTND |
853 | if (item->magic & PATHSPEC_ICASE) |
854 | /* | |
855 | * "Never interesting" trick requires exact | |
856 | * matching. We could do something clever with inexact | |
857 | * matching, but it's trickier (and not to forget that | |
858 | * strcasecmp is locale-dependent, at least in | |
859 | * glibc). Just disable it for now. It can't be worse | |
860 | * than the wildcard's codepath of '[Tt][Hi][Is][Ss]' | |
861 | * pattern. | |
862 | */ | |
863 | *never_interesting = entry_not_interesting; | |
864 | else if (*never_interesting != entry_not_interesting) { | |
58c4d666 NTND |
865 | /* |
866 | * We have not seen any match that sorts later | |
867 | * than the current path. | |
868 | */ | |
869 | ||
870 | /* | |
871 | * Does match sort strictly earlier than path | |
872 | * with their common parts? | |
873 | */ | |
874 | m = strncmp(match, entry->path, | |
875 | (matchlen < pathlen) ? matchlen : pathlen); | |
876 | if (m < 0) | |
877 | return 0; | |
878 | ||
879 | /* | |
880 | * If we come here even once, that means there is at | |
881 | * least one pathspec that would sort equal to or | |
882 | * later than the path we are currently looking at. | |
883 | * In other words, if we have never reached this point | |
884 | * after iterating all pathspecs, it means all | |
885 | * pathspecs are either outside of base, or inside the | |
886 | * base but sorts strictly earlier than the current | |
887 | * one. In either case, they will never match the | |
888 | * subsequent entries. In such a case, we initialized | |
889 | * the variable to -1 and that is what will be | |
890 | * returned, allowing the caller to terminate early. | |
891 | */ | |
c3a47ca9 | 892 | *never_interesting = entry_not_interesting; |
58c4d666 NTND |
893 | } |
894 | ||
895 | if (pathlen > matchlen) | |
896 | return 0; | |
897 | ||
898 | if (matchlen > pathlen) { | |
899 | if (match[pathlen] != '/') | |
900 | return 0; | |
35a9f1e9 SG |
901 | /* |
902 | * Reject non-directories as partial pathnames, except | |
903 | * when match is a submodule with a trailing slash and | |
904 | * nothing else (to handle 'submod/' and 'submod' | |
905 | * uniformly). | |
906 | */ | |
907 | if (!S_ISDIR(entry->mode) && | |
908 | (!S_ISGITLINK(entry->mode) || matchlen > pathlen + 1)) | |
58c4d666 NTND |
909 | return 0; |
910 | } | |
911 | ||
912 | if (m == -1) | |
913 | /* | |
914 | * we cheated and did not do strncmp(), so we do | |
915 | * that here. | |
916 | */ | |
93d93537 | 917 | m = ps_strncmp(item, match, entry->path, pathlen); |
58c4d666 NTND |
918 | |
919 | /* | |
920 | * If common part matched earlier then it is a hit, | |
921 | * because we rejected the case where path is not a | |
922 | * leading directory and is shorter than match. | |
923 | */ | |
924 | if (!m) | |
93d93537 NTND |
925 | /* |
926 | * match_entry does not check if the prefix part is | |
927 | * matched case-sensitively. If the entry is a | |
928 | * directory and part of prefix, it'll be rematched | |
929 | * eventually by basecmp with special treatment for | |
930 | * the prefix. | |
931 | */ | |
58c4d666 NTND |
932 | return 1; |
933 | ||
934 | return 0; | |
935 | } | |
936 | ||
93d93537 NTND |
937 | /* :(icase)-aware string compare */ |
938 | static int basecmp(const struct pathspec_item *item, | |
939 | const char *base, const char *match, int len) | |
940 | { | |
941 | if (item->magic & PATHSPEC_ICASE) { | |
942 | int ret, n = len > item->prefix ? item->prefix : len; | |
943 | ret = strncmp(base, match, n); | |
944 | if (ret) | |
945 | return ret; | |
946 | base += n; | |
947 | match += n; | |
948 | len -= n; | |
949 | } | |
950 | return ps_strncmp(item, base, match, len); | |
951 | } | |
952 | ||
953 | static int match_dir_prefix(const struct pathspec_item *item, | |
954 | const char *base, | |
58c4d666 NTND |
955 | const char *match, int matchlen) |
956 | { | |
93d93537 | 957 | if (basecmp(item, base, match, matchlen)) |
58c4d666 NTND |
958 | return 0; |
959 | ||
960 | /* | |
961 | * If the base is a subdirectory of a path which | |
962 | * was specified, all of them are interesting. | |
963 | */ | |
964 | if (!matchlen || | |
965 | base[matchlen] == '/' || | |
966 | match[matchlen - 1] == '/') | |
967 | return 1; | |
968 | ||
969 | /* Just a random prefix match */ | |
970 | return 0; | |
971 | } | |
972 | ||
c904cd89 NTND |
973 | /* |
974 | * Perform matching on the leading non-wildcard part of | |
975 | * pathspec. item->nowildcard_len must be greater than zero. Return | |
976 | * non-zero if base is matched. | |
977 | */ | |
978 | static int match_wildcard_base(const struct pathspec_item *item, | |
979 | const char *base, int baselen, | |
980 | int *matched) | |
981 | { | |
982 | const char *match = item->match; | |
983 | /* the wildcard part is not considered in this function */ | |
984 | int matchlen = item->nowildcard_len; | |
985 | ||
986 | if (baselen) { | |
987 | int dirlen; | |
988 | /* | |
989 | * Return early if base is longer than the | |
990 | * non-wildcard part but it does not match. | |
991 | */ | |
992 | if (baselen >= matchlen) { | |
993 | *matched = matchlen; | |
93d93537 | 994 | return !basecmp(item, base, match, matchlen); |
c904cd89 NTND |
995 | } |
996 | ||
997 | dirlen = matchlen; | |
998 | while (dirlen && match[dirlen - 1] != '/') | |
999 | dirlen--; | |
1000 | ||
1001 | /* | |
1002 | * Return early if base is shorter than the | |
1003 | * non-wildcard part but it does not match. Note that | |
1004 | * base ends with '/' so we are sure it really matches | |
1005 | * directory | |
1006 | */ | |
93d93537 | 1007 | if (basecmp(item, base, match, baselen)) |
c904cd89 NTND |
1008 | return 0; |
1009 | *matched = baselen; | |
1010 | } else | |
1011 | *matched = 0; | |
1012 | /* | |
1013 | * we could have checked entry against the non-wildcard part | |
1014 | * that is not in base and does similar never_interesting | |
1015 | * optimization as in match_entry. For now just be happy with | |
1016 | * base comparison. | |
1017 | */ | |
1018 | return entry_interesting; | |
1019 | } | |
1020 | ||
2c389fc8 NTND |
1021 | /* |
1022 | * Is a tree entry interesting given the pathspec we have? | |
1023 | * | |
30c8c55c | 1024 | * Pre-condition: either baselen == 0 (i.e. empty path) |
1376e507 | 1025 | * or base[baselen-1] == '/' (i.e. with trailing slash). |
2c389fc8 | 1026 | */ |
67022e02 NTND |
1027 | static enum interesting do_match(struct index_state *istate, |
1028 | const struct name_entry *entry, | |
30c8c55c | 1029 | struct strbuf *base, |
ef79b1f8 NTND |
1030 | const struct pathspec *ps, |
1031 | int exclude) | |
2c389fc8 NTND |
1032 | { |
1033 | int i; | |
30c8c55c | 1034 | int pathlen, baselen = base->len; |
c3a47ca9 | 1035 | enum interesting never_interesting = ps->has_wildcard ? |
d688cf07 | 1036 | entry_not_interesting : all_entries_not_interesting; |
2c389fc8 | 1037 | |
5c6933d2 NTND |
1038 | GUARD_PATHSPEC(ps, |
1039 | PATHSPEC_FROMTOP | | |
1040 | PATHSPEC_MAXDEPTH | | |
bd30c2e4 | 1041 | PATHSPEC_LITERAL | |
93d93537 | 1042 | PATHSPEC_GLOB | |
ef79b1f8 | 1043 | PATHSPEC_ICASE | |
5a0b97b3 NTND |
1044 | PATHSPEC_EXCLUDE | |
1045 | PATHSPEC_ATTR); | |
8f4f8f45 | 1046 | |
bc96cc87 | 1047 | if (!ps->nr) { |
6330a171 NTND |
1048 | if (!ps->recursive || |
1049 | !(ps->magic & PATHSPEC_MAXDEPTH) || | |
1050 | ps->max_depth == -1) | |
d688cf07 | 1051 | return all_entries_interesting; |
30c8c55c | 1052 | return within_depth(base->buf, baselen, |
d688cf07 NTND |
1053 | !!S_ISDIR(entry->mode), |
1054 | ps->max_depth) ? | |
1055 | entry_interesting : entry_not_interesting; | |
bc96cc87 | 1056 | } |
2c389fc8 | 1057 | |
0de16337 | 1058 | pathlen = tree_entry_len(entry); |
2c389fc8 | 1059 | |
1376e507 | 1060 | for (i = ps->nr - 1; i >= 0; i--) { |
2c389fc8 NTND |
1061 | const struct pathspec_item *item = ps->items+i; |
1062 | const char *match = item->match; | |
30c8c55c | 1063 | const char *base_str = base->buf; |
c904cd89 | 1064 | int matchlen = item->len, matched = 0; |
2c389fc8 | 1065 | |
ef79b1f8 NTND |
1066 | if ((!exclude && item->magic & PATHSPEC_EXCLUDE) || |
1067 | ( exclude && !(item->magic & PATHSPEC_EXCLUDE))) | |
1068 | continue; | |
1069 | ||
2c389fc8 NTND |
1070 | if (baselen >= matchlen) { |
1071 | /* If it doesn't match, move along... */ | |
93d93537 | 1072 | if (!match_dir_prefix(item, base_str, match, matchlen)) |
d38f2809 | 1073 | goto match_wildcards; |
bc96cc87 | 1074 | |
6330a171 NTND |
1075 | if (!ps->recursive || |
1076 | !(ps->magic & PATHSPEC_MAXDEPTH) || | |
5a0b97b3 NTND |
1077 | ps->max_depth == -1) { |
1078 | if (!item->attr_match_nr) | |
1079 | return all_entries_interesting; | |
1080 | else | |
1081 | goto interesting; | |
1082 | } | |
1083 | ||
1084 | if (within_depth(base_str + matchlen + 1, | |
1085 | baselen - matchlen - 1, | |
1086 | !!S_ISDIR(entry->mode), | |
1087 | ps->max_depth)) | |
1088 | goto interesting; | |
1089 | else | |
1090 | return entry_not_interesting; | |
2c389fc8 NTND |
1091 | } |
1092 | ||
1b740923 | 1093 | /* Either there must be no base, or the base must match. */ |
93d93537 NTND |
1094 | if (baselen == 0 || !basecmp(item, base_str, match, baselen)) { |
1095 | if (match_entry(item, entry, pathlen, | |
58c4d666 NTND |
1096 | match + baselen, matchlen - baselen, |
1097 | &never_interesting)) | |
5a0b97b3 | 1098 | goto interesting; |
f1a2ddbb | 1099 | |
170260ae | 1100 | if (item->nowildcard_len < item->len) { |
bd30c2e4 | 1101 | if (!git_fnmatch(item, match + baselen, entry->path, |
8c6abbcd | 1102 | item->nowildcard_len - baselen)) |
5a0b97b3 | 1103 | goto interesting; |
f1a2ddbb NTND |
1104 | |
1105 | /* | |
1106 | * Match all directories. We'll try to | |
1107 | * match files later on. | |
1108 | */ | |
1109 | if (ps->recursive && S_ISDIR(entry->mode)) | |
d688cf07 | 1110 | return entry_interesting; |
74ed4371 BW |
1111 | |
1112 | /* | |
1113 | * When matching against submodules with | |
1114 | * wildcard characters, ensure that the entry | |
1115 | * at least matches up to the first wild | |
1116 | * character. More accurate matching can then | |
1117 | * be performed in the submodule itself. | |
1118 | */ | |
eef3df5a BW |
1119 | if (ps->recurse_submodules && |
1120 | S_ISGITLINK(entry->mode) && | |
74ed4371 BW |
1121 | !ps_strncmp(item, match + baselen, |
1122 | entry->path, | |
1123 | item->nowildcard_len - baselen)) | |
5a0b97b3 | 1124 | goto interesting; |
f1a2ddbb NTND |
1125 | } |
1126 | ||
1127 | continue; | |
2c389fc8 | 1128 | } |
d38f2809 NTND |
1129 | |
1130 | match_wildcards: | |
170260ae | 1131 | if (item->nowildcard_len == item->len) |
d38f2809 NTND |
1132 | continue; |
1133 | ||
c904cd89 NTND |
1134 | if (item->nowildcard_len && |
1135 | !match_wildcard_base(item, base_str, baselen, &matched)) | |
e4ddb057 | 1136 | continue; |
c904cd89 | 1137 | |
d38f2809 NTND |
1138 | /* |
1139 | * Concatenate base and entry->path into one and do | |
1140 | * fnmatch() on it. | |
c904cd89 NTND |
1141 | * |
1142 | * While we could avoid concatenation in certain cases | |
1143 | * [1], which saves a memcpy and potentially a | |
1144 | * realloc, it turns out not worth it. Measurement on | |
1145 | * linux-2.6 does not show any clear improvements, | |
1146 | * partly because of the nowildcard_len optimization | |
1147 | * in git_fnmatch(). Avoid micro-optimizations here. | |
1148 | * | |
1149 | * [1] if match_wildcard_base() says the base | |
1150 | * directory is already matched, we only need to match | |
1151 | * the rest, which is shorter so _in theory_ faster. | |
d38f2809 NTND |
1152 | */ |
1153 | ||
1154 | strbuf_add(base, entry->path, pathlen); | |
1155 | ||
30c8c55c | 1156 | if (!git_fnmatch(item, match, base->buf, |
8c6abbcd | 1157 | item->nowildcard_len)) { |
30c8c55c | 1158 | strbuf_setlen(base, baselen); |
5a0b97b3 | 1159 | goto interesting; |
d38f2809 | 1160 | } |
74ed4371 BW |
1161 | |
1162 | /* | |
1163 | * When matching against submodules with | |
1164 | * wildcard characters, ensure that the entry | |
1165 | * at least matches up to the first wild | |
1166 | * character. More accurate matching can then | |
1167 | * be performed in the submodule itself. | |
1168 | */ | |
eef3df5a | 1169 | if (ps->recurse_submodules && S_ISGITLINK(entry->mode) && |
30c8c55c | 1170 | !ps_strncmp(item, match, base->buf, |
74ed4371 | 1171 | item->nowildcard_len)) { |
30c8c55c | 1172 | strbuf_setlen(base, baselen); |
5a0b97b3 | 1173 | goto interesting; |
74ed4371 BW |
1174 | } |
1175 | ||
30c8c55c | 1176 | strbuf_setlen(base, baselen); |
d38f2809 NTND |
1177 | |
1178 | /* | |
1179 | * Match all directories. We'll try to match files | |
1180 | * later on. | |
8c69c1f9 NTND |
1181 | * max_depth is ignored but we may consider support it |
1182 | * in future, see | |
3eae30e4 | 1183 | * https://lore.kernel.org/git/7vmxo5l2g4.fsf@alter.siamese.dyndns.org/ |
d38f2809 NTND |
1184 | */ |
1185 | if (ps->recursive && S_ISDIR(entry->mode)) | |
d688cf07 | 1186 | return entry_interesting; |
5a0b97b3 NTND |
1187 | continue; |
1188 | interesting: | |
1189 | if (item->attr_match_nr) { | |
1190 | int ret; | |
1191 | ||
1192 | /* | |
1193 | * Must not return all_entries_not_interesting | |
1194 | * prematurely. We do not know if all entries do not | |
1195 | * match some attributes with current attr API. | |
1196 | */ | |
1197 | never_interesting = entry_not_interesting; | |
1198 | ||
1199 | /* | |
1200 | * Consider all directories interesting (because some | |
1201 | * of those files inside may match some attributes | |
1202 | * even though the parent dir does not) | |
1203 | * | |
1204 | * FIXME: attributes _can_ match directories and we | |
1205 | * can probably return all_entries_interesting or | |
1206 | * all_entries_not_interesting here if matched. | |
1207 | */ | |
1208 | if (S_ISDIR(entry->mode)) | |
1209 | return entry_interesting; | |
1210 | ||
1211 | strbuf_add(base, entry->path, pathlen); | |
30c8c55c JH |
1212 | ret = match_pathspec_attrs(istate, base->buf, |
1213 | base->len, item); | |
1214 | strbuf_setlen(base, baselen); | |
5a0b97b3 NTND |
1215 | if (!ret) |
1216 | continue; | |
1217 | } | |
1218 | return entry_interesting; | |
2c389fc8 NTND |
1219 | } |
1220 | return never_interesting; /* No matches */ | |
1221 | } | |
ef79b1f8 NTND |
1222 | |
1223 | /* | |
1224 | * Is a tree entry interesting given the pathspec we have? | |
1225 | * | |
30c8c55c | 1226 | * Pre-condition: either baselen == 0 (i.e. empty path) |
ef79b1f8 NTND |
1227 | * or base[baselen-1] == '/' (i.e. with trailing slash). |
1228 | */ | |
67022e02 NTND |
1229 | enum interesting tree_entry_interesting(struct index_state *istate, |
1230 | const struct name_entry *entry, | |
0ad927e9 | 1231 | struct strbuf *base, |
ef79b1f8 NTND |
1232 | const struct pathspec *ps) |
1233 | { | |
1234 | enum interesting positive, negative; | |
30c8c55c | 1235 | positive = do_match(istate, entry, base, ps, 0); |
ef79b1f8 NTND |
1236 | |
1237 | /* | |
1238 | * case | entry | positive | negative | result | |
1239 | * -----+-------+----------+----------+------- | |
1240 | * 1 | file | -1 | -1..2 | -1 | |
1241 | * 2 | file | 0 | -1..2 | 0 | |
1242 | * 3 | file | 1 | -1 | 1 | |
1243 | * 4 | file | 1 | 0 | 1 | |
1244 | * 5 | file | 1 | 1 | 0 | |
1245 | * 6 | file | 1 | 2 | 0 | |
1246 | * 7 | file | 2 | -1 | 2 | |
b7845ceb | 1247 | * 8 | file | 2 | 0 | 1 |
ef79b1f8 NTND |
1248 | * 9 | file | 2 | 1 | 0 |
1249 | * 10 | file | 2 | 2 | -1 | |
1250 | * -----+-------+----------+----------+------- | |
1251 | * 11 | dir | -1 | -1..2 | -1 | |
1252 | * 12 | dir | 0 | -1..2 | 0 | |
1253 | * 13 | dir | 1 | -1 | 1 | |
1254 | * 14 | dir | 1 | 0 | 1 | |
1255 | * 15 | dir | 1 | 1 | 1 (*) | |
1256 | * 16 | dir | 1 | 2 | 0 | |
1257 | * 17 | dir | 2 | -1 | 2 | |
b7845ceb | 1258 | * 18 | dir | 2 | 0 | 1 |
ef79b1f8 NTND |
1259 | * 19 | dir | 2 | 1 | 1 (*) |
1260 | * 20 | dir | 2 | 2 | -1 | |
1261 | * | |
1262 | * (*) An exclude pattern interested in a directory does not | |
1263 | * necessarily mean it will exclude all of the directory. In | |
1264 | * wildcard case, it can't decide until looking at individual | |
1265 | * files inside. So don't write such directories off yet. | |
1266 | */ | |
1267 | ||
1268 | if (!(ps->magic & PATHSPEC_EXCLUDE) || | |
1269 | positive <= entry_not_interesting) /* #1, #2, #11, #12 */ | |
1270 | return positive; | |
1271 | ||
30c8c55c | 1272 | negative = do_match(istate, entry, base, ps, 1); |
ef79b1f8 | 1273 | |
b7845ceb NTND |
1274 | /* #8, #18 */ |
1275 | if (positive == all_entries_interesting && | |
1276 | negative == entry_not_interesting) | |
1277 | return entry_interesting; | |
1278 | ||
1279 | /* #3, #4, #7, #13, #14, #17 */ | |
ef79b1f8 NTND |
1280 | if (negative <= entry_not_interesting) |
1281 | return positive; | |
1282 | ||
1283 | /* #15, #19 */ | |
1284 | if (S_ISDIR(entry->mode) && | |
1285 | positive >= entry_interesting && | |
1286 | negative == entry_interesting) | |
1287 | return entry_interesting; | |
1288 | ||
1289 | if ((positive == entry_interesting && | |
1290 | negative >= entry_interesting) || /* #5, #6, #16 */ | |
1291 | (positive == all_entries_interesting && | |
1292 | negative == entry_interesting)) /* #9 */ | |
1293 | return entry_not_interesting; | |
1294 | ||
1295 | return all_entries_not_interesting; /* #10, #20 */ | |
1296 | } |