tree-walk: use size_t consistently
[git/git.git] / tree-walk.c
CommitLineData
1b0c7174
JH
1#include "cache.h"
2#include "tree-walk.h"
e6c111b4 3#include "unpack-trees.h"
bc96cc87 4#include "dir.h"
cbd53a21 5#include "object-store.h"
8e440259 6#include "tree.h"
64acde94 7#include "pathspec.h"
1b0c7174 8
4651ece8
LT
9static const char *get_mode(const char *str, unsigned int *modep)
10{
11 unsigned char c;
12 unsigned int mode = 0;
13
64cc1c09
MK
14 if (*str == ' ')
15 return NULL;
16
4651ece8
LT
17 while ((c = *str++) != ' ') {
18 if (c < '0' || c > '7')
19 return NULL;
20 mode = (mode << 3) + (c - '0');
21 }
22 *modep = mode;
23 return str;
24}
25
8354fa3d 26static int decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size, struct strbuf *err)
4651ece8
LT
27{
28 const char *path;
29 unsigned int mode, len;
83e4b757 30 const unsigned hashsz = the_hash_algo->rawsz;
4651ece8 31
83e4b757 32 if (size < hashsz + 3 || buf[size - (hashsz + 1)]) {
8354fa3d
DT
33 strbuf_addstr(err, _("too-short tree object"));
34 return -1;
35 }
64cc1c09 36
4651ece8 37 path = get_mode(buf, &mode);
8354fa3d
DT
38 if (!path) {
39 strbuf_addstr(err, _("malformed mode in tree entry"));
40 return -1;
41 }
42 if (!*path) {
43 strbuf_addstr(err, _("empty filename in tree entry"));
44 return -1;
45 }
4651ece8
LT
46 len = strlen(path) + 1;
47
48 /* Initialize the descriptor entry */
49 desc->entry.path = path;
7146e66f 50 desc->entry.mode = canon_mode(mode);
ea82b2a0 51 desc->entry.pathlen = len - 1;
52 hashcpy(desc->entry.oid.hash, (const unsigned char *)path + len);
8354fa3d
DT
53
54 return 0;
4651ece8
LT
55}
56
8354fa3d 57static int init_tree_desc_internal(struct tree_desc *desc, const void *buffer, unsigned long size, struct strbuf *err)
6fda5e51
LT
58{
59 desc->buffer = buffer;
60 desc->size = size;
4651ece8 61 if (size)
8354fa3d
DT
62 return decode_tree_entry(desc, buffer, size, err);
63 return 0;
64}
65
66void init_tree_desc(struct tree_desc *desc, const void *buffer, unsigned long size)
67{
68 struct strbuf err = STRBUF_INIT;
69 if (init_tree_desc_internal(desc, buffer, size, &err))
70 die("%s", err.buf);
71 strbuf_release(&err);
72}
73
74int init_tree_desc_gently(struct tree_desc *desc, const void *buffer, unsigned long size)
75{
76 struct strbuf err = STRBUF_INIT;
77 int result = init_tree_desc_internal(desc, buffer, size, &err);
78 if (result)
79 error("%s", err.buf);
80 strbuf_release(&err);
81 return result;
6fda5e51
LT
82}
83
5c377d3d 84void *fill_tree_descriptor(struct tree_desc *desc, const struct object_id *oid)
1b0c7174
JH
85{
86 unsigned long size = 0;
87 void *buf = NULL;
88
5c377d3d 89 if (oid) {
02f0547e 90 buf = read_object_with_reference(oid, tree_type, &size, NULL);
1b0c7174 91 if (!buf)
5c377d3d 92 die("unable to read tree %s", oid_to_hex(oid));
1b0c7174 93 }
6fda5e51 94 init_tree_desc(desc, buf, size);
1b0c7174
JH
95 return buf;
96}
97
1b0c7174
JH
98static void entry_clear(struct name_entry *a)
99{
100 memset(a, 0, sizeof(*a));
101}
102
103static void entry_extract(struct tree_desc *t, struct name_entry *a)
104{
4651ece8 105 *a = t->entry;
1b0c7174
JH
106}
107
8354fa3d 108static int update_tree_entry_internal(struct tree_desc *desc, struct strbuf *err)
1b0c7174 109{
6fda5e51 110 const void *buf = desc->buffer;
ea82b2a0 111 const unsigned char *end = (const unsigned char *)desc->entry.path + desc->entry.pathlen + 1 + the_hash_algo->rawsz;
1b0c7174 112 unsigned long size = desc->size;
4651ece8 113 unsigned long len = end - (const unsigned char *)buf;
1b0c7174
JH
114
115 if (size < len)
2edffef2 116 die(_("too-short tree file"));
4651ece8
LT
117 buf = end;
118 size -= len;
119 desc->buffer = buf;
120 desc->size = size;
121 if (size)
8354fa3d
DT
122 return decode_tree_entry(desc, buf, size, err);
123 return 0;
124}
125
126void update_tree_entry(struct tree_desc *desc)
127{
128 struct strbuf err = STRBUF_INIT;
129 if (update_tree_entry_internal(desc, &err))
130 die("%s", err.buf);
131 strbuf_release(&err);
132}
133
134int update_tree_entry_gently(struct tree_desc *desc)
135{
136 struct strbuf err = STRBUF_INIT;
137 if (update_tree_entry_internal(desc, &err)) {
138 error("%s", err.buf);
139 strbuf_release(&err);
140 /* Stop processing this tree after error */
141 desc->size = 0;
142 return -1;
143 }
144 strbuf_release(&err);
145 return 0;
1b0c7174
JH
146}
147
4c068a98
LT
148int tree_entry(struct tree_desc *desc, struct name_entry *entry)
149{
4651ece8 150 if (!desc->size)
4c068a98
LT
151 return 0;
152
4651ece8
LT
153 *entry = desc->entry;
154 update_tree_entry(desc);
4c068a98
LT
155 return 1;
156}
157
8354fa3d
DT
158int tree_entry_gently(struct tree_desc *desc, struct name_entry *entry)
159{
160 if (!desc->size)
161 return 0;
162
163 *entry = desc->entry;
164 if (update_tree_entry_gently(desc))
165 return 0;
166 return 1;
167}
168
40d934df
LT
169void setup_traverse_info(struct traverse_info *info, const char *base)
170{
37806080 171 size_t pathlen = strlen(base);
bcbe5a51 172 static struct traverse_info dummy;
40d934df
LT
173
174 memset(info, 0, sizeof(*info));
175 if (pathlen && base[pathlen-1] == '/')
176 pathlen--;
177 info->pathlen = pathlen ? pathlen + 1 : 0;
90553847
JK
178 info->name = base;
179 info->namelen = pathlen;
947208b7 180 if (pathlen)
bcbe5a51 181 info->prev = &dummy;
40d934df
LT
182}
183
90553847
JK
184char *make_traverse_path(char *path, const struct traverse_info *info,
185 const char *name, size_t namelen)
40d934df 186{
37806080 187 size_t pathlen = info->pathlen;
40d934df 188
90553847 189 path[pathlen + namelen] = 0;
40d934df 190 for (;;) {
90553847 191 memcpy(path + pathlen, name, namelen);
40d934df
LT
192 if (!pathlen)
193 break;
194 path[--pathlen] = '/';
90553847
JK
195 name = info->name;
196 namelen = info->namelen;
40d934df 197 info = info->prev;
90553847 198 pathlen -= namelen;
40d934df
LT
199 }
200 return path;
201}
202
1ee26571
JH
203struct tree_desc_skip {
204 struct tree_desc_skip *prev;
205 const void *ptr;
206};
207
208struct tree_desc_x {
209 struct tree_desc d;
210 struct tree_desc_skip *skip;
211};
212
1ee26571
JH
213static int check_entry_match(const char *a, int a_len, const char *b, int b_len)
214{
215 /*
216 * The caller wants to pick *a* from a tree or nothing.
217 * We are looking at *b* in a tree.
218 *
219 * (0) If a and b are the same name, we are trivially happy.
220 *
221 * There are three possibilities where *a* could be hiding
222 * behind *b*.
223 *
224 * (1) *a* == "t", *b* == "ab" i.e. *b* sorts earlier than *a* no
225 * matter what.
226 * (2) *a* == "t", *b* == "t-2" and "t" is a subtree in the tree;
227 * (3) *a* == "t-2", *b* == "t" and "t-2" is a blob in the tree.
228 *
229 * Otherwise we know *a* won't appear in the tree without
230 * scanning further.
231 */
232
233 int cmp = name_compare(a, a_len, b, b_len);
234
235 /* Most common case first -- reading sync'd trees */
236 if (!cmp)
237 return cmp;
238
239 if (0 < cmp) {
240 /* a comes after b; it does not matter if it is case (3)
241 if (b_len < a_len && !memcmp(a, b, b_len) && a[b_len] < '/')
242 return 1;
243 */
244 return 1; /* keep looking */
245 }
246
247 /* b comes after a; are we looking at case (2)? */
248 if (a_len < b_len && !memcmp(a, b, a_len) && b[a_len] < '/')
249 return 1; /* keep looking */
250
251 return -1; /* a cannot appear in the tree */
252}
253
254/*
255 * From the extended tree_desc, extract the first name entry, while
256 * paying attention to the candidate "first" name. Most importantly,
257 * when looking for an entry, if there are entries that sorts earlier
258 * in the tree object representation than that name, skip them and
259 * process the named entry first. We will remember that we haven't
260 * processed the first entry yet, and in the later call skip the
261 * entry we processed early when update_extended_entry() is called.
262 *
263 * E.g. if the underlying tree object has these entries:
264 *
265 * blob "t-1"
266 * blob "t-2"
267 * tree "t"
268 * blob "t=1"
269 *
270 * and the "first" asks for "t", remember that we still need to
271 * process "t-1" and "t-2" but extract "t". After processing the
272 * entry "t" from this call, the caller will let us know by calling
273 * update_extended_entry() that we can remember "t" has been processed
274 * already.
275 */
276
277static void extended_entry_extract(struct tree_desc_x *t,
278 struct name_entry *a,
279 const char *first,
280 int first_len)
281{
282 const char *path;
283 int len;
284 struct tree_desc probe;
285 struct tree_desc_skip *skip;
286
287 /*
288 * Extract the first entry from the tree_desc, but skip the
289 * ones that we already returned in earlier rounds.
290 */
291 while (1) {
292 if (!t->d.size) {
293 entry_clear(a);
294 break; /* not found */
295 }
296 entry_extract(&t->d, a);
297 for (skip = t->skip; skip; skip = skip->prev)
298 if (a->path == skip->ptr)
299 break; /* found */
300 if (!skip)
301 break;
302 /* We have processed this entry already. */
303 update_tree_entry(&t->d);
304 }
305
306 if (!first || !a->path)
307 return;
308
309 /*
310 * The caller wants "first" from this tree, or nothing.
311 */
312 path = a->path;
0de16337 313 len = tree_entry_len(a);
1ee26571
JH
314 switch (check_entry_match(first, first_len, path, len)) {
315 case -1:
316 entry_clear(a);
317 case 0:
318 return;
319 default:
320 break;
321 }
322
323 /*
324 * We need to look-ahead -- we suspect that a subtree whose
325 * name is "first" may be hiding behind the current entry "path".
326 */
327 probe = t->d;
328 while (probe.size) {
329 entry_extract(&probe, a);
330 path = a->path;
0de16337 331 len = tree_entry_len(a);
1ee26571
JH
332 switch (check_entry_match(first, first_len, path, len)) {
333 case -1:
334 entry_clear(a);
335 case 0:
336 return;
337 default:
338 update_tree_entry(&probe);
339 break;
340 }
341 /* keep looking */
342 }
343 entry_clear(a);
344}
345
346static void update_extended_entry(struct tree_desc_x *t, struct name_entry *a)
347{
348 if (t->d.entry.path == a->path) {
349 update_tree_entry(&t->d);
350 } else {
351 /* we have returned this entry early */
352 struct tree_desc_skip *skip = xmalloc(sizeof(*skip));
353 skip->ptr = a->path;
354 skip->prev = t->skip;
355 t->skip = skip;
356 }
357}
358
359static void free_extended_entry(struct tree_desc_x *t)
360{
361 struct tree_desc_skip *p, *s;
362
363 for (s = t->skip; s; s = p) {
364 p = s->prev;
365 free(s);
366 }
367}
368
67022e02
NTND
369static inline int prune_traversal(struct index_state *istate,
370 struct name_entry *e,
2842c0f9
JH
371 struct traverse_info *info,
372 struct strbuf *base,
373 int still_interesting)
374{
375 if (!info->pathspec || still_interesting == 2)
376 return 2;
377 if (still_interesting < 0)
378 return still_interesting;
67022e02
NTND
379 return tree_entry_interesting(istate, e, base,
380 0, info->pathspec);
2842c0f9
JH
381}
382
67022e02
NTND
383int traverse_trees(struct index_state *istate,
384 int n, struct tree_desc *t,
385 struct traverse_info *info)
1b0c7174 386{
e6c111b4 387 int error = 0;
1b0c7174 388 struct name_entry *entry = xmalloc(n*sizeof(*entry));
1ee26571
JH
389 int i;
390 struct tree_desc_x *tx = xcalloc(n, sizeof(*tx));
2842c0f9
JH
391 struct strbuf base = STRBUF_INIT;
392 int interesting = 1;
d9c2bd56 393 char *traverse_path;
1ee26571
JH
394
395 for (i = 0; i < n; i++)
396 tx[i].d = t[i];
1b0c7174 397
2842c0f9
JH
398 if (info->prev) {
399 strbuf_grow(&base, info->pathlen);
90553847
JK
400 make_traverse_path(base.buf, info->prev, info->name,
401 info->namelen);
2842c0f9
JH
402 base.buf[info->pathlen-1] = '/';
403 strbuf_setlen(&base, info->pathlen);
d9c2bd56
DT
404 traverse_path = xstrndup(base.buf, info->pathlen);
405 } else {
90553847 406 traverse_path = xstrndup(info->name, info->pathlen);
2842c0f9 407 }
d9c2bd56 408 info->traverse_path = traverse_path;
1b0c7174 409 for (;;) {
a04f8196 410 int trees_used;
1ee26571
JH
411 unsigned long mask, dirmask;
412 const char *first = NULL;
413 int first_len = 0;
27c0f768 414 struct name_entry *e = NULL;
1ee26571
JH
415 int len;
416
417 for (i = 0; i < n; i++) {
418 e = entry + i;
419 extended_entry_extract(tx + i, e, NULL, 0);
420 }
1b0c7174 421
1ee26571
JH
422 /*
423 * A tree may have "t-2" at the current location even
424 * though it may have "t" that is a subtree behind it,
425 * and another tree may return "t". We want to grab
426 * all "t" from all trees to match in such a case.
427 */
1b0c7174 428 for (i = 0; i < n; i++) {
1ee26571
JH
429 e = entry + i;
430 if (!e->path)
1b0c7174 431 continue;
0de16337 432 len = tree_entry_len(e);
1ee26571
JH
433 if (!first) {
434 first = e->path;
435 first_len = len;
436 continue;
437 }
438 if (name_compare(e->path, len, first, first_len) < 0) {
439 first = e->path;
440 first_len = len;
441 }
442 }
443
444 if (first) {
445 for (i = 0; i < n; i++) {
446 e = entry + i;
447 extended_entry_extract(tx + i, e, first, first_len);
448 /* Cull the ones that are not the earliest */
449 if (!e->path)
1b0c7174 450 continue;
0de16337 451 len = tree_entry_len(e);
1ee26571
JH
452 if (name_compare(e->path, len, first, first_len))
453 entry_clear(e);
1b0c7174 454 }
1ee26571
JH
455 }
456
457 /* Now we have in entry[i] the earliest name from the trees */
458 mask = 0;
459 dirmask = 0;
460 for (i = 0; i < n; i++) {
461 if (!entry[i].path)
462 continue;
1b0c7174 463 mask |= 1ul << i;
91e4f036
LT
464 if (S_ISDIR(entry[i].mode))
465 dirmask |= 1ul << i;
2842c0f9 466 e = &entry[i];
1b0c7174
JH
467 }
468 if (!mask)
469 break;
67022e02 470 interesting = prune_traversal(istate, e, info, &base, interesting);
2842c0f9
JH
471 if (interesting < 0)
472 break;
473 if (interesting) {
a04f8196
SB
474 trees_used = info->fn(n, mask, dirmask, entry, info);
475 if (trees_used < 0) {
476 error = trees_used;
2842c0f9
JH
477 if (!info->show_all_errors)
478 break;
479 }
a04f8196 480 mask &= trees_used;
e6c111b4 481 }
1ee26571 482 for (i = 0; i < n; i++)
5803c6f8 483 if (mask & (1ul << i))
1ee26571 484 update_extended_entry(tx + i, entry + i);
1b0c7174
JH
485 }
486 free(entry);
1ee26571
JH
487 for (i = 0; i < n; i++)
488 free_extended_entry(tx + i);
489 free(tx);
d9c2bd56
DT
490 free(traverse_path);
491 info->traverse_path = NULL;
2842c0f9 492 strbuf_release(&base);
e6c111b4 493 return error;
1b0c7174
JH
494}
495
275721c2
DT
496struct dir_state {
497 void *tree;
498 unsigned long size;
3b683bcf 499 struct object_id oid;
275721c2
DT
500};
501
5ec1e728 502static int find_tree_entry(struct tree_desc *t, const char *name, struct object_id *result, unsigned short *mode)
4dcff634
JH
503{
504 int namelen = strlen(name);
505 while (t->size) {
506 const char *entry;
0a3faa45 507 struct object_id oid;
4dcff634
JH
508 int entrylen, cmp;
509
0a3faa45 510 oidcpy(&oid, tree_entry_extract(t, &entry, mode));
0de16337 511 entrylen = tree_entry_len(&t->entry);
4dcff634 512 update_tree_entry(t);
4dcff634
JH
513 if (entrylen > namelen)
514 continue;
515 cmp = memcmp(name, entry, entrylen);
516 if (cmp > 0)
517 continue;
518 if (cmp < 0)
519 break;
520 if (entrylen == namelen) {
0a3faa45 521 oidcpy(result, &oid);
4dcff634
JH
522 return 0;
523 }
524 if (name[entrylen] != '/')
525 continue;
526 if (!S_ISDIR(*mode))
527 break;
528 if (++entrylen == namelen) {
0a3faa45 529 oidcpy(result, &oid);
4dcff634
JH
530 return 0;
531 }
0a3faa45 532 return get_tree_entry(&oid, name + entrylen, result, mode);
4dcff634
JH
533 }
534 return -1;
535}
536
5ec1e728 537int get_tree_entry(const struct object_id *tree_oid, const char *name, struct object_id *oid, unsigned short *mode)
4dcff634
JH
538{
539 int retval;
540 void *tree;
6fda5e51 541 unsigned long size;
916bc35b 542 struct object_id root;
4dcff634 543
02f0547e 544 tree = read_object_with_reference(tree_oid, tree_type, &size, &root);
4dcff634
JH
545 if (!tree)
546 return -1;
d93b7d1c
JK
547
548 if (name[0] == '\0') {
916bc35b 549 oidcpy(oid, &root);
ef006503 550 free(tree);
d93b7d1c
JK
551 return 0;
552 }
553
5fb8c05f
JH
554 if (!size) {
555 retval = -1;
556 } else {
557 struct tree_desc t;
558 init_tree_desc(&t, tree, size);
916bc35b 559 retval = find_tree_entry(&t, name, oid, mode);
5fb8c05f 560 }
4dcff634
JH
561 free(tree);
562 return retval;
563}
2c389fc8 564
275721c2
DT
565/*
566 * This is Linux's built-in max for the number of symlinks to follow.
567 * That limit, of course, does not affect git, but it's a reasonable
568 * choice.
569 */
570#define GET_TREE_ENTRY_FOLLOW_SYMLINKS_MAX_LINKS 40
571
572/**
573 * Find a tree entry by following symlinks in tree_sha (which is
574 * assumed to be the root of the repository). In the event that a
575 * symlink points outside the repository (e.g. a link to /foo or a
576 * root-level link to ../foo), the portion of the link which is
577 * outside the repository will be returned in result_path, and *mode
578 * will be set to 0. It is assumed that result_path is uninitialized.
579 * If there are no symlinks, or the end result of the symlink chain
580 * points to an object inside the repository, result will be filled in
581 * with the sha1 of the found object, and *mode will hold the mode of
582 * the object.
583 *
d1dd94b3 584 * See the code for enum get_oid_result for a description of
275721c2
DT
585 * the return values.
586 */
5ec1e728 587enum get_oid_result get_tree_entry_follow_symlinks(struct object_id *tree_oid, const char *name, struct object_id *result, struct strbuf *result_path, unsigned short *mode)
275721c2
DT
588{
589 int retval = MISSING_OBJECT;
590 struct dir_state *parents = NULL;
591 size_t parents_alloc = 0;
071bcaab 592 size_t i, parents_nr = 0;
0d4a8b5b 593 struct object_id current_tree_oid;
275721c2
DT
594 struct strbuf namebuf = STRBUF_INIT;
595 struct tree_desc t;
596 int follows_remaining = GET_TREE_ENTRY_FOLLOW_SYMLINKS_MAX_LINKS;
275721c2
DT
597
598 init_tree_desc(&t, NULL, 0UL);
275721c2 599 strbuf_addstr(&namebuf, name);
3b683bcf 600 oidcpy(&current_tree_oid, tree_oid);
275721c2
DT
601
602 while (1) {
603 int find_result;
604 char *first_slash;
605 char *remainder = NULL;
606
607 if (!t.buffer) {
608 void *tree;
0d4a8b5b 609 struct object_id root;
275721c2 610 unsigned long size;
02f0547e 611 tree = read_object_with_reference(&current_tree_oid,
275721c2 612 tree_type, &size,
02f0547e 613 &root);
275721c2
DT
614 if (!tree)
615 goto done;
616
617 ALLOC_GROW(parents, parents_nr + 1, parents_alloc);
618 parents[parents_nr].tree = tree;
619 parents[parents_nr].size = size;
3b683bcf 620 oidcpy(&parents[parents_nr].oid, &root);
275721c2
DT
621 parents_nr++;
622
623 if (namebuf.buf[0] == '\0') {
3b683bcf 624 oidcpy(result, &root);
275721c2
DT
625 retval = FOUND;
626 goto done;
627 }
628
629 if (!size)
630 goto done;
631
632 /* descend */
633 init_tree_desc(&t, tree, size);
634 }
635
636 /* Handle symlinks to e.g. a//b by removing leading slashes */
637 while (namebuf.buf[0] == '/') {
638 strbuf_remove(&namebuf, 0, 1);
639 }
640
641 /* Split namebuf into a first component and a remainder */
642 if ((first_slash = strchr(namebuf.buf, '/'))) {
643 *first_slash = 0;
644 remainder = first_slash + 1;
645 }
646
647 if (!strcmp(namebuf.buf, "..")) {
648 struct dir_state *parent;
649 /*
650 * We could end up with .. in the namebuf if it
651 * appears in a symlink.
652 */
653
654 if (parents_nr == 1) {
655 if (remainder)
656 *first_slash = '/';
657 strbuf_add(result_path, namebuf.buf,
658 namebuf.len);
659 *mode = 0;
660 retval = FOUND;
661 goto done;
662 }
663 parent = &parents[parents_nr - 1];
664 free(parent->tree);
665 parents_nr--;
666 parent = &parents[parents_nr - 1];
667 init_tree_desc(&t, parent->tree, parent->size);
668 strbuf_remove(&namebuf, 0, remainder ? 3 : 2);
669 continue;
670 }
671
672 /* We could end up here via a symlink to dir/.. */
673 if (namebuf.buf[0] == '\0') {
3b683bcf 674 oidcpy(result, &parents[parents_nr - 1].oid);
275721c2
DT
675 retval = FOUND;
676 goto done;
677 }
678
679 /* Look up the first (or only) path component in the tree. */
680 find_result = find_tree_entry(&t, namebuf.buf,
916bc35b 681 &current_tree_oid, mode);
275721c2
DT
682 if (find_result) {
683 goto done;
684 }
685
686 if (S_ISDIR(*mode)) {
687 if (!remainder) {
3b683bcf 688 oidcpy(result, &current_tree_oid);
275721c2
DT
689 retval = FOUND;
690 goto done;
691 }
692 /* Descend the tree */
693 t.buffer = NULL;
694 strbuf_remove(&namebuf, 0,
695 1 + first_slash - namebuf.buf);
696 } else if (S_ISREG(*mode)) {
697 if (!remainder) {
3b683bcf 698 oidcpy(result, &current_tree_oid);
275721c2
DT
699 retval = FOUND;
700 } else {
701 retval = NOT_DIR;
702 }
703 goto done;
704 } else if (S_ISLNK(*mode)) {
705 /* Follow a symlink */
706 unsigned long link_len;
707 size_t len;
708 char *contents, *contents_start;
709 struct dir_state *parent;
710 enum object_type type;
711
712 if (follows_remaining-- == 0) {
713 /* Too many symlinks followed */
714 retval = SYMLINK_LOOP;
715 goto done;
716 }
717
718 /*
719 * At this point, we have followed at a least
720 * one symlink, so on error we need to report this.
721 */
722 retval = DANGLING_SYMLINK;
723
b4f5aca4 724 contents = read_object_file(&current_tree_oid, &type,
725 &link_len);
275721c2
DT
726
727 if (!contents)
728 goto done;
729
730 if (contents[0] == '/') {
731 strbuf_addstr(result_path, contents);
732 free(contents);
733 *mode = 0;
734 retval = FOUND;
735 goto done;
736 }
737
738 if (remainder)
739 len = first_slash - namebuf.buf;
740 else
741 len = namebuf.len;
742
743 contents_start = contents;
744
745 parent = &parents[parents_nr - 1];
746 init_tree_desc(&t, parent->tree, parent->size);
747 strbuf_splice(&namebuf, 0, len,
748 contents_start, link_len);
749 if (remainder)
750 namebuf.buf[link_len] = '/';
751 free(contents);
752 }
753 }
754done:
755 for (i = 0; i < parents_nr; i++)
756 free(parents[i].tree);
757 free(parents);
758
759 strbuf_release(&namebuf);
760 return retval;
761}
762
93d93537
NTND
763static int match_entry(const struct pathspec_item *item,
764 const struct name_entry *entry, int pathlen,
58c4d666 765 const char *match, int matchlen,
c3a47ca9 766 enum interesting *never_interesting)
58c4d666
NTND
767{
768 int m = -1; /* signals that we haven't called strncmp() */
769
93d93537
NTND
770 if (item->magic & PATHSPEC_ICASE)
771 /*
772 * "Never interesting" trick requires exact
773 * matching. We could do something clever with inexact
774 * matching, but it's trickier (and not to forget that
775 * strcasecmp is locale-dependent, at least in
776 * glibc). Just disable it for now. It can't be worse
777 * than the wildcard's codepath of '[Tt][Hi][Is][Ss]'
778 * pattern.
779 */
780 *never_interesting = entry_not_interesting;
781 else if (*never_interesting != entry_not_interesting) {
58c4d666
NTND
782 /*
783 * We have not seen any match that sorts later
784 * than the current path.
785 */
786
787 /*
788 * Does match sort strictly earlier than path
789 * with their common parts?
790 */
791 m = strncmp(match, entry->path,
792 (matchlen < pathlen) ? matchlen : pathlen);
793 if (m < 0)
794 return 0;
795
796 /*
797 * If we come here even once, that means there is at
798 * least one pathspec that would sort equal to or
799 * later than the path we are currently looking at.
800 * In other words, if we have never reached this point
801 * after iterating all pathspecs, it means all
802 * pathspecs are either outside of base, or inside the
803 * base but sorts strictly earlier than the current
804 * one. In either case, they will never match the
805 * subsequent entries. In such a case, we initialized
806 * the variable to -1 and that is what will be
807 * returned, allowing the caller to terminate early.
808 */
c3a47ca9 809 *never_interesting = entry_not_interesting;
58c4d666
NTND
810 }
811
812 if (pathlen > matchlen)
813 return 0;
814
815 if (matchlen > pathlen) {
816 if (match[pathlen] != '/')
817 return 0;
74b4f7f2 818 if (!S_ISDIR(entry->mode) && !S_ISGITLINK(entry->mode))
58c4d666
NTND
819 return 0;
820 }
821
822 if (m == -1)
823 /*
824 * we cheated and did not do strncmp(), so we do
825 * that here.
826 */
93d93537 827 m = ps_strncmp(item, match, entry->path, pathlen);
58c4d666
NTND
828
829 /*
830 * If common part matched earlier then it is a hit,
831 * because we rejected the case where path is not a
832 * leading directory and is shorter than match.
833 */
834 if (!m)
93d93537
NTND
835 /*
836 * match_entry does not check if the prefix part is
837 * matched case-sensitively. If the entry is a
838 * directory and part of prefix, it'll be rematched
839 * eventually by basecmp with special treatment for
840 * the prefix.
841 */
58c4d666
NTND
842 return 1;
843
844 return 0;
845}
846
93d93537
NTND
847/* :(icase)-aware string compare */
848static int basecmp(const struct pathspec_item *item,
849 const char *base, const char *match, int len)
850{
851 if (item->magic & PATHSPEC_ICASE) {
852 int ret, n = len > item->prefix ? item->prefix : len;
853 ret = strncmp(base, match, n);
854 if (ret)
855 return ret;
856 base += n;
857 match += n;
858 len -= n;
859 }
860 return ps_strncmp(item, base, match, len);
861}
862
863static int match_dir_prefix(const struct pathspec_item *item,
864 const char *base,
58c4d666
NTND
865 const char *match, int matchlen)
866{
93d93537 867 if (basecmp(item, base, match, matchlen))
58c4d666
NTND
868 return 0;
869
870 /*
871 * If the base is a subdirectory of a path which
872 * was specified, all of them are interesting.
873 */
874 if (!matchlen ||
875 base[matchlen] == '/' ||
876 match[matchlen - 1] == '/')
877 return 1;
878
879 /* Just a random prefix match */
880 return 0;
881}
882
c904cd89
NTND
883/*
884 * Perform matching on the leading non-wildcard part of
885 * pathspec. item->nowildcard_len must be greater than zero. Return
886 * non-zero if base is matched.
887 */
888static int match_wildcard_base(const struct pathspec_item *item,
889 const char *base, int baselen,
890 int *matched)
891{
892 const char *match = item->match;
893 /* the wildcard part is not considered in this function */
894 int matchlen = item->nowildcard_len;
895
896 if (baselen) {
897 int dirlen;
898 /*
899 * Return early if base is longer than the
900 * non-wildcard part but it does not match.
901 */
902 if (baselen >= matchlen) {
903 *matched = matchlen;
93d93537 904 return !basecmp(item, base, match, matchlen);
c904cd89
NTND
905 }
906
907 dirlen = matchlen;
908 while (dirlen && match[dirlen - 1] != '/')
909 dirlen--;
910
911 /*
912 * Return early if base is shorter than the
913 * non-wildcard part but it does not match. Note that
914 * base ends with '/' so we are sure it really matches
915 * directory
916 */
93d93537 917 if (basecmp(item, base, match, baselen))
c904cd89
NTND
918 return 0;
919 *matched = baselen;
920 } else
921 *matched = 0;
922 /*
923 * we could have checked entry against the non-wildcard part
924 * that is not in base and does similar never_interesting
925 * optimization as in match_entry. For now just be happy with
926 * base comparison.
927 */
928 return entry_interesting;
929}
930
2c389fc8
NTND
931/*
932 * Is a tree entry interesting given the pathspec we have?
933 *
1376e507
NTND
934 * Pre-condition: either baselen == base_offset (i.e. empty path)
935 * or base[baselen-1] == '/' (i.e. with trailing slash).
2c389fc8 936 */
67022e02
NTND
937static enum interesting do_match(struct index_state *istate,
938 const struct name_entry *entry,
ef79b1f8
NTND
939 struct strbuf *base, int base_offset,
940 const struct pathspec *ps,
941 int exclude)
2c389fc8
NTND
942{
943 int i;
1376e507 944 int pathlen, baselen = base->len - base_offset;
c3a47ca9 945 enum interesting never_interesting = ps->has_wildcard ?
d688cf07 946 entry_not_interesting : all_entries_not_interesting;
2c389fc8 947
5c6933d2
NTND
948 GUARD_PATHSPEC(ps,
949 PATHSPEC_FROMTOP |
950 PATHSPEC_MAXDEPTH |
bd30c2e4 951 PATHSPEC_LITERAL |
93d93537 952 PATHSPEC_GLOB |
ef79b1f8 953 PATHSPEC_ICASE |
5a0b97b3
NTND
954 PATHSPEC_EXCLUDE |
955 PATHSPEC_ATTR);
8f4f8f45 956
bc96cc87 957 if (!ps->nr) {
6330a171
NTND
958 if (!ps->recursive ||
959 !(ps->magic & PATHSPEC_MAXDEPTH) ||
960 ps->max_depth == -1)
d688cf07
NTND
961 return all_entries_interesting;
962 return within_depth(base->buf + base_offset, baselen,
963 !!S_ISDIR(entry->mode),
964 ps->max_depth) ?
965 entry_interesting : entry_not_interesting;
bc96cc87 966 }
2c389fc8 967
0de16337 968 pathlen = tree_entry_len(entry);
2c389fc8 969
1376e507 970 for (i = ps->nr - 1; i >= 0; i--) {
2c389fc8
NTND
971 const struct pathspec_item *item = ps->items+i;
972 const char *match = item->match;
1376e507 973 const char *base_str = base->buf + base_offset;
c904cd89 974 int matchlen = item->len, matched = 0;
2c389fc8 975
ef79b1f8
NTND
976 if ((!exclude && item->magic & PATHSPEC_EXCLUDE) ||
977 ( exclude && !(item->magic & PATHSPEC_EXCLUDE)))
978 continue;
979
2c389fc8
NTND
980 if (baselen >= matchlen) {
981 /* If it doesn't match, move along... */
93d93537 982 if (!match_dir_prefix(item, base_str, match, matchlen))
d38f2809 983 goto match_wildcards;
bc96cc87 984
6330a171
NTND
985 if (!ps->recursive ||
986 !(ps->magic & PATHSPEC_MAXDEPTH) ||
5a0b97b3
NTND
987 ps->max_depth == -1) {
988 if (!item->attr_match_nr)
989 return all_entries_interesting;
990 else
991 goto interesting;
992 }
993
994 if (within_depth(base_str + matchlen + 1,
995 baselen - matchlen - 1,
996 !!S_ISDIR(entry->mode),
997 ps->max_depth))
998 goto interesting;
999 else
1000 return entry_not_interesting;
2c389fc8
NTND
1001 }
1002
1b740923 1003 /* Either there must be no base, or the base must match. */
93d93537
NTND
1004 if (baselen == 0 || !basecmp(item, base_str, match, baselen)) {
1005 if (match_entry(item, entry, pathlen,
58c4d666
NTND
1006 match + baselen, matchlen - baselen,
1007 &never_interesting))
5a0b97b3 1008 goto interesting;
f1a2ddbb 1009
170260ae 1010 if (item->nowildcard_len < item->len) {
bd30c2e4 1011 if (!git_fnmatch(item, match + baselen, entry->path,
8c6abbcd 1012 item->nowildcard_len - baselen))
5a0b97b3 1013 goto interesting;
f1a2ddbb
NTND
1014
1015 /*
1016 * Match all directories. We'll try to
1017 * match files later on.
1018 */
1019 if (ps->recursive && S_ISDIR(entry->mode))
d688cf07 1020 return entry_interesting;
74ed4371
BW
1021
1022 /*
1023 * When matching against submodules with
1024 * wildcard characters, ensure that the entry
1025 * at least matches up to the first wild
1026 * character. More accurate matching can then
1027 * be performed in the submodule itself.
1028 */
eef3df5a
BW
1029 if (ps->recurse_submodules &&
1030 S_ISGITLINK(entry->mode) &&
74ed4371
BW
1031 !ps_strncmp(item, match + baselen,
1032 entry->path,
1033 item->nowildcard_len - baselen))
5a0b97b3 1034 goto interesting;
f1a2ddbb
NTND
1035 }
1036
1037 continue;
2c389fc8 1038 }
d38f2809
NTND
1039
1040match_wildcards:
170260ae 1041 if (item->nowildcard_len == item->len)
d38f2809
NTND
1042 continue;
1043
c904cd89
NTND
1044 if (item->nowildcard_len &&
1045 !match_wildcard_base(item, base_str, baselen, &matched))
e4ddb057 1046 continue;
c904cd89 1047
d38f2809
NTND
1048 /*
1049 * Concatenate base and entry->path into one and do
1050 * fnmatch() on it.
c904cd89
NTND
1051 *
1052 * While we could avoid concatenation in certain cases
1053 * [1], which saves a memcpy and potentially a
1054 * realloc, it turns out not worth it. Measurement on
1055 * linux-2.6 does not show any clear improvements,
1056 * partly because of the nowildcard_len optimization
1057 * in git_fnmatch(). Avoid micro-optimizations here.
1058 *
1059 * [1] if match_wildcard_base() says the base
1060 * directory is already matched, we only need to match
1061 * the rest, which is shorter so _in theory_ faster.
d38f2809
NTND
1062 */
1063
1064 strbuf_add(base, entry->path, pathlen);
1065
bd30c2e4 1066 if (!git_fnmatch(item, match, base->buf + base_offset,
8c6abbcd 1067 item->nowildcard_len)) {
1376e507 1068 strbuf_setlen(base, base_offset + baselen);
5a0b97b3 1069 goto interesting;
d38f2809 1070 }
74ed4371
BW
1071
1072 /*
1073 * When matching against submodules with
1074 * wildcard characters, ensure that the entry
1075 * at least matches up to the first wild
1076 * character. More accurate matching can then
1077 * be performed in the submodule itself.
1078 */
eef3df5a 1079 if (ps->recurse_submodules && S_ISGITLINK(entry->mode) &&
74ed4371
BW
1080 !ps_strncmp(item, match, base->buf + base_offset,
1081 item->nowildcard_len)) {
1082 strbuf_setlen(base, base_offset + baselen);
5a0b97b3 1083 goto interesting;
74ed4371
BW
1084 }
1085
1376e507 1086 strbuf_setlen(base, base_offset + baselen);
d38f2809
NTND
1087
1088 /*
1089 * Match all directories. We'll try to match files
1090 * later on.
8c69c1f9
NTND
1091 * max_depth is ignored but we may consider support it
1092 * in future, see
5840eb9d 1093 * https://public-inbox.org/git/7vmxo5l2g4.fsf@alter.siamese.dyndns.org/
d38f2809
NTND
1094 */
1095 if (ps->recursive && S_ISDIR(entry->mode))
d688cf07 1096 return entry_interesting;
5a0b97b3
NTND
1097 continue;
1098interesting:
1099 if (item->attr_match_nr) {
1100 int ret;
1101
1102 /*
1103 * Must not return all_entries_not_interesting
1104 * prematurely. We do not know if all entries do not
1105 * match some attributes with current attr API.
1106 */
1107 never_interesting = entry_not_interesting;
1108
1109 /*
1110 * Consider all directories interesting (because some
1111 * of those files inside may match some attributes
1112 * even though the parent dir does not)
1113 *
1114 * FIXME: attributes _can_ match directories and we
1115 * can probably return all_entries_interesting or
1116 * all_entries_not_interesting here if matched.
1117 */
1118 if (S_ISDIR(entry->mode))
1119 return entry_interesting;
1120
1121 strbuf_add(base, entry->path, pathlen);
1122 ret = match_pathspec_attrs(istate, base->buf + base_offset,
1123 base->len - base_offset, item);
1124 strbuf_setlen(base, base_offset + baselen);
1125 if (!ret)
1126 continue;
1127 }
1128 return entry_interesting;
2c389fc8
NTND
1129 }
1130 return never_interesting; /* No matches */
1131}
ef79b1f8
NTND
1132
1133/*
1134 * Is a tree entry interesting given the pathspec we have?
1135 *
1136 * Pre-condition: either baselen == base_offset (i.e. empty path)
1137 * or base[baselen-1] == '/' (i.e. with trailing slash).
1138 */
67022e02
NTND
1139enum interesting tree_entry_interesting(struct index_state *istate,
1140 const struct name_entry *entry,
ef79b1f8
NTND
1141 struct strbuf *base, int base_offset,
1142 const struct pathspec *ps)
1143{
1144 enum interesting positive, negative;
67022e02 1145 positive = do_match(istate, entry, base, base_offset, ps, 0);
ef79b1f8
NTND
1146
1147 /*
1148 * case | entry | positive | negative | result
1149 * -----+-------+----------+----------+-------
1150 * 1 | file | -1 | -1..2 | -1
1151 * 2 | file | 0 | -1..2 | 0
1152 * 3 | file | 1 | -1 | 1
1153 * 4 | file | 1 | 0 | 1
1154 * 5 | file | 1 | 1 | 0
1155 * 6 | file | 1 | 2 | 0
1156 * 7 | file | 2 | -1 | 2
b7845ceb 1157 * 8 | file | 2 | 0 | 1
ef79b1f8
NTND
1158 * 9 | file | 2 | 1 | 0
1159 * 10 | file | 2 | 2 | -1
1160 * -----+-------+----------+----------+-------
1161 * 11 | dir | -1 | -1..2 | -1
1162 * 12 | dir | 0 | -1..2 | 0
1163 * 13 | dir | 1 | -1 | 1
1164 * 14 | dir | 1 | 0 | 1
1165 * 15 | dir | 1 | 1 | 1 (*)
1166 * 16 | dir | 1 | 2 | 0
1167 * 17 | dir | 2 | -1 | 2
b7845ceb 1168 * 18 | dir | 2 | 0 | 1
ef79b1f8
NTND
1169 * 19 | dir | 2 | 1 | 1 (*)
1170 * 20 | dir | 2 | 2 | -1
1171 *
1172 * (*) An exclude pattern interested in a directory does not
1173 * necessarily mean it will exclude all of the directory. In
1174 * wildcard case, it can't decide until looking at individual
1175 * files inside. So don't write such directories off yet.
1176 */
1177
1178 if (!(ps->magic & PATHSPEC_EXCLUDE) ||
1179 positive <= entry_not_interesting) /* #1, #2, #11, #12 */
1180 return positive;
1181
67022e02 1182 negative = do_match(istate, entry, base, base_offset, ps, 1);
ef79b1f8 1183
b7845ceb
NTND
1184 /* #8, #18 */
1185 if (positive == all_entries_interesting &&
1186 negative == entry_not_interesting)
1187 return entry_interesting;
1188
1189 /* #3, #4, #7, #13, #14, #17 */
ef79b1f8
NTND
1190 if (negative <= entry_not_interesting)
1191 return positive;
1192
1193 /* #15, #19 */
1194 if (S_ISDIR(entry->mode) &&
1195 positive >= entry_interesting &&
1196 negative == entry_interesting)
1197 return entry_interesting;
1198
1199 if ((positive == entry_interesting &&
1200 negative >= entry_interesting) || /* #5, #6, #16 */
1201 (positive == all_entries_interesting &&
1202 negative == entry_interesting)) /* #9 */
1203 return entry_not_interesting;
1204
1205 return all_entries_not_interesting; /* #10, #20 */
1206}