1 /*
2  * Copyright (C) the libgit2 contributors. All rights reserved.
3  *
4  * This file is part of libgit2, distributed under the GNU GPL v2 with
5  * a Linking Exception. For full terms see the included COPYING file.
6  */
7 
8 #include "index.h"
9 
10 #include <stddef.h>
11 
12 #include "repository.h"
13 #include "tree.h"
14 #include "tree-cache.h"
15 #include "hash.h"
16 #include "iterator.h"
17 #include "pathspec.h"
18 #include "ignore.h"
19 #include "blob.h"
20 #include "idxmap.h"
21 #include "diff.h"
22 #include "varint.h"
23 
24 #include "git2/odb.h"
25 #include "git2/oid.h"
26 #include "git2/blob.h"
27 #include "git2/config.h"
28 #include "git2/sys/index.h"
29 
30 static int index_apply_to_wd_diff(git_index *index, int action, const git_strarray *paths,
31 				  unsigned int flags,
32 				  git_index_matched_path_cb cb, void *payload);
33 
34 #define minimal_entry_size (offsetof(struct entry_short, path))
35 
36 static const size_t INDEX_FOOTER_SIZE = GIT_OID_RAWSZ;
37 static const size_t INDEX_HEADER_SIZE = 12;
38 
39 static const unsigned int INDEX_VERSION_NUMBER_DEFAULT = 2;
40 static const unsigned int INDEX_VERSION_NUMBER_LB = 2;
41 static const unsigned int INDEX_VERSION_NUMBER_EXT = 3;
42 static const unsigned int INDEX_VERSION_NUMBER_COMP = 4;
43 static const unsigned int INDEX_VERSION_NUMBER_UB = 4;
44 
45 static const unsigned int INDEX_HEADER_SIG = 0x44495243;
46 static const char INDEX_EXT_TREECACHE_SIG[] = {'T', 'R', 'E', 'E'};
47 static const char INDEX_EXT_UNMERGED_SIG[] = {'R', 'E', 'U', 'C'};
48 static const char INDEX_EXT_CONFLICT_NAME_SIG[] = {'N', 'A', 'M', 'E'};
49 
50 #define INDEX_OWNER(idx) ((git_repository *)(GIT_REFCOUNT_OWNER(idx)))
51 
52 struct index_header {
53 	uint32_t signature;
54 	uint32_t version;
55 	uint32_t entry_count;
56 };
57 
58 struct index_extension {
59 	char signature[4];
60 	uint32_t extension_size;
61 };
62 
63 struct entry_time {
64 	uint32_t seconds;
65 	uint32_t nanoseconds;
66 };
67 
68 struct entry_short {
69 	struct entry_time ctime;
70 	struct entry_time mtime;
71 	uint32_t dev;
72 	uint32_t ino;
73 	uint32_t mode;
74 	uint32_t uid;
75 	uint32_t gid;
76 	uint32_t file_size;
77 	git_oid oid;
78 	uint16_t flags;
79 	char path[1]; /* arbitrary length */
80 };
81 
82 struct entry_long {
83 	struct entry_time ctime;
84 	struct entry_time mtime;
85 	uint32_t dev;
86 	uint32_t ino;
87 	uint32_t mode;
88 	uint32_t uid;
89 	uint32_t gid;
90 	uint32_t file_size;
91 	git_oid oid;
92 	uint16_t flags;
93 	uint16_t flags_extended;
94 	char path[1]; /* arbitrary length */
95 };
96 
97 struct entry_srch_key {
98 	const char *path;
99 	size_t pathlen;
100 	int stage;
101 };
102 
103 struct entry_internal {
104 	git_index_entry entry;
105 	size_t pathlen;
106 	char path[GIT_FLEX_ARRAY];
107 };
108 
109 struct reuc_entry_internal {
110 	git_index_reuc_entry entry;
111 	size_t pathlen;
112 	char path[GIT_FLEX_ARRAY];
113 };
114 
115 bool git_index__enforce_unsaved_safety = false;
116 
117 /* local declarations */
118 static int read_extension(size_t *read_len, git_index *index, const char *buffer, size_t buffer_size);
119 static int read_header(struct index_header *dest, const void *buffer);
120 
121 static int parse_index(git_index *index, const char *buffer, size_t buffer_size);
122 static bool is_index_extended(git_index *index);
123 static int write_index(git_oid *checksum, git_index *index, git_filebuf *file);
124 
125 static void index_entry_free(git_index_entry *entry);
126 static void index_entry_reuc_free(git_index_reuc_entry *reuc);
127 
index_map_set(git_idxmap * map,git_index_entry * e,bool ignore_case)128 GIT_INLINE(int) index_map_set(git_idxmap *map, git_index_entry *e, bool ignore_case)
129 {
130 	if (ignore_case)
131 		return git_idxmap_icase_set((git_idxmap_icase *) map, e, e);
132 	else
133 		return git_idxmap_set(map, e, e);
134 }
135 
index_map_delete(git_idxmap * map,git_index_entry * e,bool ignore_case)136 GIT_INLINE(int) index_map_delete(git_idxmap *map, git_index_entry *e, bool ignore_case)
137 {
138 	if (ignore_case)
139 		return git_idxmap_icase_delete((git_idxmap_icase *) map, e);
140 	else
141 		return git_idxmap_delete(map, e);
142 }
143 
index_map_resize(git_idxmap * map,size_t count,bool ignore_case)144 GIT_INLINE(int) index_map_resize(git_idxmap *map, size_t count, bool ignore_case)
145 {
146 	if (ignore_case)
147 		return git_idxmap_icase_resize((git_idxmap_icase *) map, count);
148 	else
149 		return git_idxmap_resize(map, count);
150 }
151 
git_index_entry_srch(const void * key,const void * array_member)152 int git_index_entry_srch(const void *key, const void *array_member)
153 {
154 	const struct entry_srch_key *srch_key = key;
155 	const struct entry_internal *entry = array_member;
156 	int cmp;
157 	size_t len1, len2, len;
158 
159 	len1 = srch_key->pathlen;
160 	len2 = entry->pathlen;
161 	len = len1 < len2 ? len1 : len2;
162 
163 	cmp = memcmp(srch_key->path, entry->path, len);
164 	if (cmp)
165 		return cmp;
166 	if (len1 < len2)
167 		return -1;
168 	if (len1 > len2)
169 		return 1;
170 
171 	if (srch_key->stage != GIT_INDEX_STAGE_ANY)
172 		return srch_key->stage - GIT_INDEX_ENTRY_STAGE(&entry->entry);
173 
174 	return 0;
175 }
176 
git_index_entry_isrch(const void * key,const void * array_member)177 int git_index_entry_isrch(const void *key, const void *array_member)
178 {
179 	const struct entry_srch_key *srch_key = key;
180 	const struct entry_internal *entry = array_member;
181 	int cmp;
182 	size_t len1, len2, len;
183 
184 	len1 = srch_key->pathlen;
185 	len2 = entry->pathlen;
186 	len = len1 < len2 ? len1 : len2;
187 
188 	cmp = strncasecmp(srch_key->path, entry->path, len);
189 
190 	if (cmp)
191 		return cmp;
192 	if (len1 < len2)
193 		return -1;
194 	if (len1 > len2)
195 		return 1;
196 
197 	if (srch_key->stage != GIT_INDEX_STAGE_ANY)
198 		return srch_key->stage - GIT_INDEX_ENTRY_STAGE(&entry->entry);
199 
200 	return 0;
201 }
202 
index_entry_srch_path(const void * path,const void * array_member)203 static int index_entry_srch_path(const void *path, const void *array_member)
204 {
205 	const git_index_entry *entry = array_member;
206 
207 	return strcmp((const char *)path, entry->path);
208 }
209 
index_entry_isrch_path(const void * path,const void * array_member)210 static int index_entry_isrch_path(const void *path, const void *array_member)
211 {
212 	const git_index_entry *entry = array_member;
213 
214 	return strcasecmp((const char *)path, entry->path);
215 }
216 
git_index_entry_cmp(const void * a,const void * b)217 int git_index_entry_cmp(const void *a, const void *b)
218 {
219 	int diff;
220 	const git_index_entry *entry_a = a;
221 	const git_index_entry *entry_b = b;
222 
223 	diff = strcmp(entry_a->path, entry_b->path);
224 
225 	if (diff == 0)
226 		diff = (GIT_INDEX_ENTRY_STAGE(entry_a) - GIT_INDEX_ENTRY_STAGE(entry_b));
227 
228 	return diff;
229 }
230 
git_index_entry_icmp(const void * a,const void * b)231 int git_index_entry_icmp(const void *a, const void *b)
232 {
233 	int diff;
234 	const git_index_entry *entry_a = a;
235 	const git_index_entry *entry_b = b;
236 
237 	diff = strcasecmp(entry_a->path, entry_b->path);
238 
239 	if (diff == 0)
240 		diff = (GIT_INDEX_ENTRY_STAGE(entry_a) - GIT_INDEX_ENTRY_STAGE(entry_b));
241 
242 	return diff;
243 }
244 
conflict_name_cmp(const void * a,const void * b)245 static int conflict_name_cmp(const void *a, const void *b)
246 {
247 	const git_index_name_entry *name_a = a;
248 	const git_index_name_entry *name_b = b;
249 
250 	if (name_a->ancestor && !name_b->ancestor)
251 		return 1;
252 
253 	if (!name_a->ancestor && name_b->ancestor)
254 		return -1;
255 
256 	if (name_a->ancestor)
257 		return strcmp(name_a->ancestor, name_b->ancestor);
258 
259 	if (!name_a->ours || !name_b->ours)
260 		return 0;
261 
262 	return strcmp(name_a->ours, name_b->ours);
263 }
264 
265 /**
266  * TODO: enable this when resolving case insensitive conflicts
267  */
268 #if 0
269 static int conflict_name_icmp(const void *a, const void *b)
270 {
271 	const git_index_name_entry *name_a = a;
272 	const git_index_name_entry *name_b = b;
273 
274 	if (name_a->ancestor && !name_b->ancestor)
275 		return 1;
276 
277 	if (!name_a->ancestor && name_b->ancestor)
278 		return -1;
279 
280 	if (name_a->ancestor)
281 		return strcasecmp(name_a->ancestor, name_b->ancestor);
282 
283 	if (!name_a->ours || !name_b->ours)
284 		return 0;
285 
286 	return strcasecmp(name_a->ours, name_b->ours);
287 }
288 #endif
289 
reuc_srch(const void * key,const void * array_member)290 static int reuc_srch(const void *key, const void *array_member)
291 {
292 	const git_index_reuc_entry *reuc = array_member;
293 
294 	return strcmp(key, reuc->path);
295 }
296 
reuc_isrch(const void * key,const void * array_member)297 static int reuc_isrch(const void *key, const void *array_member)
298 {
299 	const git_index_reuc_entry *reuc = array_member;
300 
301 	return strcasecmp(key, reuc->path);
302 }
303 
reuc_cmp(const void * a,const void * b)304 static int reuc_cmp(const void *a, const void *b)
305 {
306 	const git_index_reuc_entry *info_a = a;
307 	const git_index_reuc_entry *info_b = b;
308 
309 	return strcmp(info_a->path, info_b->path);
310 }
311 
reuc_icmp(const void * a,const void * b)312 static int reuc_icmp(const void *a, const void *b)
313 {
314 	const git_index_reuc_entry *info_a = a;
315 	const git_index_reuc_entry *info_b = b;
316 
317 	return strcasecmp(info_a->path, info_b->path);
318 }
319 
index_entry_reuc_free(git_index_reuc_entry * reuc)320 static void index_entry_reuc_free(git_index_reuc_entry *reuc)
321 {
322 	git__free(reuc);
323 }
324 
index_entry_free(git_index_entry * entry)325 static void index_entry_free(git_index_entry *entry)
326 {
327 	if (!entry)
328 		return;
329 
330 	memset(&entry->id, 0, sizeof(entry->id));
331 	git__free(entry);
332 }
333 
git_index__create_mode(unsigned int mode)334 unsigned int git_index__create_mode(unsigned int mode)
335 {
336 	if (S_ISLNK(mode))
337 		return S_IFLNK;
338 
339 	if (S_ISDIR(mode) || (mode & S_IFMT) == (S_IFLNK | S_IFDIR))
340 		return (S_IFLNK | S_IFDIR);
341 
342 	return S_IFREG | GIT_PERMS_CANONICAL(mode);
343 }
344 
index_merge_mode(git_index * index,git_index_entry * existing,unsigned int mode)345 static unsigned int index_merge_mode(
346 	git_index *index, git_index_entry *existing, unsigned int mode)
347 {
348 	if (index->no_symlinks && S_ISREG(mode) &&
349 		existing && S_ISLNK(existing->mode))
350 		return existing->mode;
351 
352 	if (index->distrust_filemode && S_ISREG(mode))
353 		return (existing && S_ISREG(existing->mode)) ?
354 			existing->mode : git_index__create_mode(0666);
355 
356 	return git_index__create_mode(mode);
357 }
358 
index_find_in_entries(size_t * out,git_vector * entries,git_vector_cmp entry_srch,const char * path,size_t path_len,int stage)359 GIT_INLINE(int) index_find_in_entries(
360 	size_t *out, git_vector *entries, git_vector_cmp entry_srch,
361 	const char *path, size_t path_len, int stage)
362 {
363 	struct entry_srch_key srch_key;
364 	srch_key.path = path;
365 	srch_key.pathlen = !path_len ? strlen(path) : path_len;
366 	srch_key.stage = stage;
367 	return git_vector_bsearch2(out, entries, entry_srch, &srch_key);
368 }
369 
index_find(size_t * out,git_index * index,const char * path,size_t path_len,int stage)370 GIT_INLINE(int) index_find(
371 	size_t *out, git_index *index,
372 	const char *path, size_t path_len, int stage)
373 {
374 	git_vector_sort(&index->entries);
375 
376 	return index_find_in_entries(
377 		out, &index->entries, index->entries_search, path, path_len, stage);
378 }
379 
git_index__set_ignore_case(git_index * index,bool ignore_case)380 void git_index__set_ignore_case(git_index *index, bool ignore_case)
381 {
382 	index->ignore_case = ignore_case;
383 
384 	if (ignore_case) {
385 		index->entries_cmp_path    = git__strcasecmp_cb;
386 		index->entries_search      = git_index_entry_isrch;
387 		index->entries_search_path = index_entry_isrch_path;
388 		index->reuc_search         = reuc_isrch;
389 	} else {
390 		index->entries_cmp_path    = git__strcmp_cb;
391 		index->entries_search      = git_index_entry_srch;
392 		index->entries_search_path = index_entry_srch_path;
393 		index->reuc_search         = reuc_srch;
394 	}
395 
396 	git_vector_set_cmp(&index->entries,
397 		ignore_case ? git_index_entry_icmp : git_index_entry_cmp);
398 	git_vector_sort(&index->entries);
399 
400 	git_vector_set_cmp(&index->reuc, ignore_case ? reuc_icmp : reuc_cmp);
401 	git_vector_sort(&index->reuc);
402 }
403 
git_index_open(git_index ** index_out,const char * index_path)404 int git_index_open(git_index **index_out, const char *index_path)
405 {
406 	git_index *index;
407 	int error = -1;
408 
409 	assert(index_out);
410 
411 	index = git__calloc(1, sizeof(git_index));
412 	GIT_ERROR_CHECK_ALLOC(index);
413 
414 	if (git_pool_init(&index->tree_pool, 1) < 0)
415 		goto fail;
416 
417 	if (index_path != NULL) {
418 		index->index_file_path = git__strdup(index_path);
419 		if (!index->index_file_path)
420 			goto fail;
421 
422 		/* Check if index file is stored on disk already */
423 		if (git_path_exists(index->index_file_path) == true)
424 			index->on_disk = 1;
425 	}
426 
427 	if (git_vector_init(&index->entries, 32, git_index_entry_cmp) < 0 ||
428 	    git_idxmap_new(&index->entries_map) < 0 ||
429 	    git_vector_init(&index->names, 8, conflict_name_cmp) < 0 ||
430 	    git_vector_init(&index->reuc, 8, reuc_cmp) < 0 ||
431 	    git_vector_init(&index->deleted, 8, git_index_entry_cmp) < 0)
432 		goto fail;
433 
434 	index->entries_cmp_path = git__strcmp_cb;
435 	index->entries_search = git_index_entry_srch;
436 	index->entries_search_path = index_entry_srch_path;
437 	index->reuc_search = reuc_srch;
438 	index->version = INDEX_VERSION_NUMBER_DEFAULT;
439 
440 	if (index_path != NULL && (error = git_index_read(index, true)) < 0)
441 		goto fail;
442 
443 	*index_out = index;
444 	GIT_REFCOUNT_INC(index);
445 
446 	return 0;
447 
448 fail:
449 	git_pool_clear(&index->tree_pool);
450 	git_index_free(index);
451 	return error;
452 }
453 
git_index_new(git_index ** out)454 int git_index_new(git_index **out)
455 {
456 	return git_index_open(out, NULL);
457 }
458 
index_free(git_index * index)459 static void index_free(git_index *index)
460 {
461 	/* index iterators increment the refcount of the index, so if we
462 	 * get here then there should be no outstanding iterators.
463 	 */
464 	assert(!git_atomic_get(&index->readers));
465 
466 	git_index_clear(index);
467 	git_idxmap_free(index->entries_map);
468 	git_vector_free(&index->entries);
469 	git_vector_free(&index->names);
470 	git_vector_free(&index->reuc);
471 	git_vector_free(&index->deleted);
472 
473 	git__free(index->index_file_path);
474 
475 	git__memzero(index, sizeof(*index));
476 	git__free(index);
477 }
478 
git_index_free(git_index * index)479 void git_index_free(git_index *index)
480 {
481 	if (index == NULL)
482 		return;
483 
484 	GIT_REFCOUNT_DEC(index, index_free);
485 }
486 
487 /* call with locked index */
index_free_deleted(git_index * index)488 static void index_free_deleted(git_index *index)
489 {
490 	int readers = (int)git_atomic_get(&index->readers);
491 	size_t i;
492 
493 	if (readers > 0 || !index->deleted.length)
494 		return;
495 
496 	for (i = 0; i < index->deleted.length; ++i) {
497 		git_index_entry *ie = git__swap(index->deleted.contents[i], NULL);
498 		index_entry_free(ie);
499 	}
500 
501 	git_vector_clear(&index->deleted);
502 }
503 
504 /* call with locked index */
index_remove_entry(git_index * index,size_t pos)505 static int index_remove_entry(git_index *index, size_t pos)
506 {
507 	int error = 0;
508 	git_index_entry *entry = git_vector_get(&index->entries, pos);
509 
510 	if (entry != NULL) {
511 		git_tree_cache_invalidate_path(index->tree, entry->path);
512 		index_map_delete(index->entries_map, entry, index->ignore_case);
513 	}
514 
515 	error = git_vector_remove(&index->entries, pos);
516 
517 	if (!error) {
518 		if (git_atomic_get(&index->readers) > 0) {
519 			error = git_vector_insert(&index->deleted, entry);
520 		} else {
521 			index_entry_free(entry);
522 		}
523 
524 		index->dirty = 1;
525 	}
526 
527 	return error;
528 }
529 
git_index_clear(git_index * index)530 int git_index_clear(git_index *index)
531 {
532 	int error = 0;
533 
534 	assert(index);
535 
536 	index->dirty = 1;
537 	index->tree = NULL;
538 	git_pool_clear(&index->tree_pool);
539 
540 	git_idxmap_clear(index->entries_map);
541 	while (!error && index->entries.length > 0)
542 		error = index_remove_entry(index, index->entries.length - 1);
543 
544 	if (error)
545 		goto done;
546 
547 	index_free_deleted(index);
548 
549 	if ((error = git_index_name_clear(index)) < 0 ||
550 		(error = git_index_reuc_clear(index)) < 0)
551 	    goto done;
552 
553 	git_futils_filestamp_set(&index->stamp, NULL);
554 
555 done:
556 	return error;
557 }
558 
create_index_error(int error,const char * msg)559 static int create_index_error(int error, const char *msg)
560 {
561 	git_error_set_str(GIT_ERROR_INDEX, msg);
562 	return error;
563 }
564 
git_index_set_caps(git_index * index,int caps)565 int git_index_set_caps(git_index *index, int caps)
566 {
567 	unsigned int old_ignore_case;
568 
569 	assert(index);
570 
571 	old_ignore_case = index->ignore_case;
572 
573 	if (caps == GIT_INDEX_CAPABILITY_FROM_OWNER) {
574 		git_repository *repo = INDEX_OWNER(index);
575 		int val;
576 
577 		if (!repo)
578 			return create_index_error(
579 				-1, "cannot access repository to set index caps");
580 
581 		if (!git_repository__configmap_lookup(&val, repo, GIT_CONFIGMAP_IGNORECASE))
582 			index->ignore_case = (val != 0);
583 		if (!git_repository__configmap_lookup(&val, repo, GIT_CONFIGMAP_FILEMODE))
584 			index->distrust_filemode = (val == 0);
585 		if (!git_repository__configmap_lookup(&val, repo, GIT_CONFIGMAP_SYMLINKS))
586 			index->no_symlinks = (val == 0);
587 	}
588 	else {
589 		index->ignore_case = ((caps & GIT_INDEX_CAPABILITY_IGNORE_CASE) != 0);
590 		index->distrust_filemode = ((caps & GIT_INDEX_CAPABILITY_NO_FILEMODE) != 0);
591 		index->no_symlinks = ((caps & GIT_INDEX_CAPABILITY_NO_SYMLINKS) != 0);
592 	}
593 
594 	if (old_ignore_case != index->ignore_case) {
595 		git_index__set_ignore_case(index, (bool)index->ignore_case);
596 	}
597 
598 	return 0;
599 }
600 
git_index_caps(const git_index * index)601 int git_index_caps(const git_index *index)
602 {
603 	return ((index->ignore_case ? GIT_INDEX_CAPABILITY_IGNORE_CASE : 0) |
604 			(index->distrust_filemode ? GIT_INDEX_CAPABILITY_NO_FILEMODE : 0) |
605 			(index->no_symlinks ? GIT_INDEX_CAPABILITY_NO_SYMLINKS : 0));
606 }
607 
git_index_checksum(git_index * index)608 const git_oid *git_index_checksum(git_index *index)
609 {
610 	return &index->checksum;
611 }
612 
613 /**
614  * Returns 1 for changed, 0 for not changed and <0 for errors
615  */
compare_checksum(git_index * index)616 static int compare_checksum(git_index *index)
617 {
618 	int fd;
619 	ssize_t bytes_read;
620 	git_oid checksum = {{ 0 }};
621 
622 	if ((fd = p_open(index->index_file_path, O_RDONLY)) < 0)
623 		return fd;
624 
625 	if (p_lseek(fd, -20, SEEK_END) < 0) {
626 		p_close(fd);
627 		git_error_set(GIT_ERROR_OS, "failed to seek to end of file");
628 		return -1;
629 	}
630 
631 	bytes_read = p_read(fd, &checksum, GIT_OID_RAWSZ);
632 	p_close(fd);
633 
634 	if (bytes_read < 0)
635 		return -1;
636 
637 	return !!git_oid_cmp(&checksum, &index->checksum);
638 }
639 
git_index_read(git_index * index,int force)640 int git_index_read(git_index *index, int force)
641 {
642 	int error = 0, updated;
643 	git_buf buffer = GIT_BUF_INIT;
644 	git_futils_filestamp stamp = index->stamp;
645 
646 	if (!index->index_file_path)
647 		return create_index_error(-1,
648 			"failed to read index: The index is in-memory only");
649 
650 	index->on_disk = git_path_exists(index->index_file_path);
651 
652 	if (!index->on_disk) {
653 		if (force && (error = git_index_clear(index)) < 0)
654 			return error;
655 
656 		index->dirty = 0;
657 		return 0;
658 	}
659 
660 	if ((updated = git_futils_filestamp_check(&stamp, index->index_file_path) < 0) ||
661 	    ((updated = compare_checksum(index)) < 0)) {
662 		git_error_set(
663 			GIT_ERROR_INDEX,
664 			"failed to read index: '%s' no longer exists",
665 			index->index_file_path);
666 		return updated;
667 	}
668 
669 	if (!updated && !force)
670 		return 0;
671 
672 	error = git_futils_readbuffer(&buffer, index->index_file_path);
673 	if (error < 0)
674 		return error;
675 
676 	index->tree = NULL;
677 	git_pool_clear(&index->tree_pool);
678 
679 	error = git_index_clear(index);
680 
681 	if (!error)
682 		error = parse_index(index, buffer.ptr, buffer.size);
683 
684 	if (!error) {
685 		git_futils_filestamp_set(&index->stamp, &stamp);
686 		index->dirty = 0;
687 	}
688 
689 	git_buf_dispose(&buffer);
690 	return error;
691 }
692 
git_index_read_safely(git_index * index)693 int git_index_read_safely(git_index *index)
694 {
695 	if (git_index__enforce_unsaved_safety && index->dirty) {
696 		git_error_set(GIT_ERROR_INDEX,
697 			"the index has unsaved changes that would be overwritten by this operation");
698 		return GIT_EINDEXDIRTY;
699 	}
700 
701 	return git_index_read(index, false);
702 }
703 
git_index__changed_relative_to(git_index * index,const git_oid * checksum)704 int git_index__changed_relative_to(
705 	git_index *index, const git_oid *checksum)
706 {
707 	/* attempt to update index (ignoring errors) */
708 	if (git_index_read(index, false) < 0)
709 		git_error_clear();
710 
711 	return !!git_oid_cmp(&index->checksum, checksum);
712 }
713 
is_racy_entry(git_index * index,const git_index_entry * entry)714 static bool is_racy_entry(git_index *index, const git_index_entry *entry)
715 {
716 	/* Git special-cases submodules in the check */
717 	if (S_ISGITLINK(entry->mode))
718 		return false;
719 
720 	return git_index_entry_newer_than_index(entry, index);
721 }
722 
723 /*
724  * Force the next diff to take a look at those entries which have the
725  * same timestamp as the current index.
726  */
truncate_racily_clean(git_index * index)727 static int truncate_racily_clean(git_index *index)
728 {
729 	size_t i;
730 	int error;
731 	git_index_entry *entry;
732 	git_diff_options diff_opts = GIT_DIFF_OPTIONS_INIT;
733 	git_diff *diff = NULL;
734 	git_vector paths = GIT_VECTOR_INIT;
735 	git_diff_delta *delta;
736 
737 	/* Nothing to do if there's no repo to talk about */
738 	if (!INDEX_OWNER(index))
739 		return 0;
740 
741 	/* If there's no workdir, we can't know where to even check */
742 	if (!git_repository_workdir(INDEX_OWNER(index)))
743 		return 0;
744 
745 	diff_opts.flags |= GIT_DIFF_INCLUDE_TYPECHANGE | GIT_DIFF_IGNORE_SUBMODULES | GIT_DIFF_DISABLE_PATHSPEC_MATCH;
746 	git_vector_foreach(&index->entries, i, entry) {
747 		if ((entry->flags_extended & GIT_INDEX_ENTRY_UPTODATE) == 0 &&
748 			is_racy_entry(index, entry))
749 			git_vector_insert(&paths, (char *)entry->path);
750 	}
751 
752 	if (paths.length == 0)
753 		goto done;
754 
755 	diff_opts.pathspec.count = paths.length;
756 	diff_opts.pathspec.strings = (char **)paths.contents;
757 
758 	if ((error = git_diff_index_to_workdir(&diff, INDEX_OWNER(index), index, &diff_opts)) < 0)
759 		return error;
760 
761 	git_vector_foreach(&diff->deltas, i, delta) {
762 		entry = (git_index_entry *)git_index_get_bypath(index, delta->old_file.path, 0);
763 
764 		/* Ensure that we have a stage 0 for this file (ie, it's not a
765 		 * conflict), otherwise smudging it is quite pointless.
766 		 */
767 		if (entry) {
768 			entry->file_size = 0;
769 			index->dirty = 1;
770 		}
771 	}
772 
773 done:
774 	git_diff_free(diff);
775 	git_vector_free(&paths);
776 	return 0;
777 }
778 
git_index_version(git_index * index)779 unsigned git_index_version(git_index *index)
780 {
781 	assert(index);
782 
783 	return index->version;
784 }
785 
git_index_set_version(git_index * index,unsigned int version)786 int git_index_set_version(git_index *index, unsigned int version)
787 {
788 	assert(index);
789 
790 	if (version < INDEX_VERSION_NUMBER_LB ||
791 	    version > INDEX_VERSION_NUMBER_UB) {
792 		git_error_set(GIT_ERROR_INDEX, "invalid version number");
793 		return -1;
794 	}
795 
796 	index->version = version;
797 
798 	return 0;
799 }
800 
git_index_write(git_index * index)801 int git_index_write(git_index *index)
802 {
803 	git_indexwriter writer = GIT_INDEXWRITER_INIT;
804 	int error;
805 
806 	truncate_racily_clean(index);
807 
808 	if ((error = git_indexwriter_init(&writer, index)) == 0 &&
809 		(error = git_indexwriter_commit(&writer)) == 0)
810 		index->dirty = 0;
811 
812 	git_indexwriter_cleanup(&writer);
813 
814 	return error;
815 }
816 
git_index_path(const git_index * index)817 const char * git_index_path(const git_index *index)
818 {
819 	assert(index);
820 	return index->index_file_path;
821 }
822 
git_index_write_tree(git_oid * oid,git_index * index)823 int git_index_write_tree(git_oid *oid, git_index *index)
824 {
825 	git_repository *repo;
826 
827 	assert(oid && index);
828 
829 	repo = INDEX_OWNER(index);
830 
831 	if (repo == NULL)
832 		return create_index_error(-1, "Failed to write tree. "
833 		  "the index file is not backed up by an existing repository");
834 
835 	return git_tree__write_index(oid, index, repo);
836 }
837 
git_index_write_tree_to(git_oid * oid,git_index * index,git_repository * repo)838 int git_index_write_tree_to(
839 	git_oid *oid, git_index *index, git_repository *repo)
840 {
841 	assert(oid && index && repo);
842 	return git_tree__write_index(oid, index, repo);
843 }
844 
git_index_entrycount(const git_index * index)845 size_t git_index_entrycount(const git_index *index)
846 {
847 	assert(index);
848 	return index->entries.length;
849 }
850 
git_index_get_byindex(git_index * index,size_t n)851 const git_index_entry *git_index_get_byindex(
852 	git_index *index, size_t n)
853 {
854 	assert(index);
855 	git_vector_sort(&index->entries);
856 	return git_vector_get(&index->entries, n);
857 }
858 
git_index_get_bypath(git_index * index,const char * path,int stage)859 const git_index_entry *git_index_get_bypath(
860 	git_index *index, const char *path, int stage)
861 {
862 	git_index_entry key = {{ 0 }};
863 	git_index_entry *value;
864 
865 	assert(index);
866 
867 	key.path = path;
868 	GIT_INDEX_ENTRY_STAGE_SET(&key, stage);
869 
870 	if (index->ignore_case)
871 		value = git_idxmap_icase_get((git_idxmap_icase *) index->entries_map, &key);
872 	else
873 		value = git_idxmap_get(index->entries_map, &key);
874 
875 	if (!value) {
876 	    git_error_set(GIT_ERROR_INDEX, "index does not contain '%s'", path);
877 	    return NULL;
878 	}
879 
880 	return value;
881 }
882 
git_index_entry__init_from_stat(git_index_entry * entry,struct stat * st,bool trust_mode)883 void git_index_entry__init_from_stat(
884 	git_index_entry *entry, struct stat *st, bool trust_mode)
885 {
886 	entry->ctime.seconds = (int32_t)st->st_ctime;
887 	entry->mtime.seconds = (int32_t)st->st_mtime;
888 #if defined(GIT_USE_NSEC)
889 	entry->mtime.nanoseconds = st->st_mtime_nsec;
890 	entry->ctime.nanoseconds = st->st_ctime_nsec;
891 #endif
892 	entry->dev  = st->st_rdev;
893 	entry->ino  = st->st_ino;
894 	entry->mode = (!trust_mode && S_ISREG(st->st_mode)) ?
895 		git_index__create_mode(0666) : git_index__create_mode(st->st_mode);
896 	entry->uid  = st->st_uid;
897 	entry->gid  = st->st_gid;
898 	entry->file_size = (uint32_t)st->st_size;
899 }
900 
index_entry_adjust_namemask(git_index_entry * entry,size_t path_length)901 static void index_entry_adjust_namemask(
902 		git_index_entry *entry,
903 		size_t path_length)
904 {
905 	entry->flags &= ~GIT_INDEX_ENTRY_NAMEMASK;
906 
907 	if (path_length < GIT_INDEX_ENTRY_NAMEMASK)
908 		entry->flags |= path_length & GIT_INDEX_ENTRY_NAMEMASK;
909 	else
910 		entry->flags |= GIT_INDEX_ENTRY_NAMEMASK;
911 }
912 
913 /* When `from_workdir` is true, we will validate the paths to avoid placing
914  * paths that are invalid for the working directory on the current filesystem
915  * (eg, on Windows, we will disallow `GIT~1`, `AUX`, `COM1`, etc).  This
916  * function will *always* prevent `.git` and directory traversal `../` from
917  * being added to the index.
918  */
index_entry_create(git_index_entry ** out,git_repository * repo,const char * path,struct stat * st,bool from_workdir)919 static int index_entry_create(
920 	git_index_entry **out,
921 	git_repository *repo,
922 	const char *path,
923 	struct stat *st,
924 	bool from_workdir)
925 {
926 	size_t pathlen = strlen(path), alloclen;
927 	struct entry_internal *entry;
928 	unsigned int path_valid_flags = GIT_PATH_REJECT_INDEX_DEFAULTS;
929 	uint16_t mode = 0;
930 
931 	/* always reject placing `.git` in the index and directory traversal.
932 	 * when requested, disallow platform-specific filenames and upgrade to
933 	 * the platform-specific `.git` tests (eg, `git~1`, etc).
934 	 */
935 	if (from_workdir)
936 		path_valid_flags |= GIT_PATH_REJECT_WORKDIR_DEFAULTS;
937 	if (st)
938 		mode = st->st_mode;
939 
940 	if (!git_path_isvalid(repo, path, mode, path_valid_flags)) {
941 		git_error_set(GIT_ERROR_INDEX, "invalid path: '%s'", path);
942 		return -1;
943 	}
944 
945 	GIT_ERROR_CHECK_ALLOC_ADD(&alloclen, sizeof(struct entry_internal), pathlen);
946 	GIT_ERROR_CHECK_ALLOC_ADD(&alloclen, alloclen, 1);
947 	entry = git__calloc(1, alloclen);
948 	GIT_ERROR_CHECK_ALLOC(entry);
949 
950 	entry->pathlen = pathlen;
951 	memcpy(entry->path, path, pathlen);
952 	entry->entry.path = entry->path;
953 
954 	*out = (git_index_entry *)entry;
955 	return 0;
956 }
957 
index_entry_init(git_index_entry ** entry_out,git_index * index,const char * rel_path)958 static int index_entry_init(
959 	git_index_entry **entry_out,
960 	git_index *index,
961 	const char *rel_path)
962 {
963 	int error = 0;
964 	git_index_entry *entry = NULL;
965 	git_buf path = GIT_BUF_INIT;
966 	struct stat st;
967 	git_oid oid;
968 	git_repository *repo;
969 
970 	if (INDEX_OWNER(index) == NULL)
971 		return create_index_error(-1,
972 			"could not initialize index entry. "
973 			"Index is not backed up by an existing repository.");
974 
975 	/*
976 	 * FIXME: this is duplicated with the work in
977 	 * git_blob__create_from_paths. It should accept an optional stat
978 	 * structure so we can pass in the one we have to do here.
979 	 */
980 	repo = INDEX_OWNER(index);
981 	if (git_repository__ensure_not_bare(repo, "create blob from file") < 0)
982 		return GIT_EBAREREPO;
983 
984 	if (git_buf_joinpath(&path, git_repository_workdir(repo), rel_path) < 0)
985 		return -1;
986 
987 	error = git_path_lstat(path.ptr, &st);
988 	git_buf_dispose(&path);
989 
990 	if (error < 0)
991 		return error;
992 
993 	if (index_entry_create(&entry, INDEX_OWNER(index), rel_path, &st, true) < 0)
994 		return -1;
995 
996 	/* write the blob to disk and get the oid and stat info */
997 	error = git_blob__create_from_paths(
998 		&oid, &st, INDEX_OWNER(index), NULL, rel_path, 0, true);
999 
1000 	if (error < 0) {
1001 		index_entry_free(entry);
1002 		return error;
1003 	}
1004 
1005 	entry->id = oid;
1006 	git_index_entry__init_from_stat(entry, &st, !index->distrust_filemode);
1007 
1008 	*entry_out = (git_index_entry *)entry;
1009 	return 0;
1010 }
1011 
reuc_entry_alloc(const char * path)1012 static git_index_reuc_entry *reuc_entry_alloc(const char *path)
1013 {
1014 	size_t pathlen = strlen(path),
1015 		structlen = sizeof(struct reuc_entry_internal),
1016 		alloclen;
1017 	struct reuc_entry_internal *entry;
1018 
1019 	if (GIT_ADD_SIZET_OVERFLOW(&alloclen, structlen, pathlen) ||
1020 		GIT_ADD_SIZET_OVERFLOW(&alloclen, alloclen, 1))
1021 		return NULL;
1022 
1023 	entry = git__calloc(1, alloclen);
1024 	if (!entry)
1025 		return NULL;
1026 
1027 	entry->pathlen = pathlen;
1028 	memcpy(entry->path, path, pathlen);
1029 	entry->entry.path = entry->path;
1030 
1031 	return (git_index_reuc_entry *)entry;
1032 }
1033 
index_entry_reuc_init(git_index_reuc_entry ** reuc_out,const char * path,int ancestor_mode,const git_oid * ancestor_oid,int our_mode,const git_oid * our_oid,int their_mode,const git_oid * their_oid)1034 static int index_entry_reuc_init(git_index_reuc_entry **reuc_out,
1035 	const char *path,
1036 	int ancestor_mode, const git_oid *ancestor_oid,
1037 	int our_mode, const git_oid *our_oid,
1038 	int their_mode, const git_oid *their_oid)
1039 {
1040 	git_index_reuc_entry *reuc = NULL;
1041 
1042 	assert(reuc_out && path);
1043 
1044 	*reuc_out = reuc = reuc_entry_alloc(path);
1045 	GIT_ERROR_CHECK_ALLOC(reuc);
1046 
1047 	if ((reuc->mode[0] = ancestor_mode) > 0) {
1048 		assert(ancestor_oid);
1049 		git_oid_cpy(&reuc->oid[0], ancestor_oid);
1050 	}
1051 
1052 	if ((reuc->mode[1] = our_mode) > 0) {
1053 		assert(our_oid);
1054 		git_oid_cpy(&reuc->oid[1], our_oid);
1055 	}
1056 
1057 	if ((reuc->mode[2] = their_mode) > 0) {
1058 		assert(their_oid);
1059 		git_oid_cpy(&reuc->oid[2], their_oid);
1060 	}
1061 
1062 	return 0;
1063 }
1064 
index_entry_cpy(git_index_entry * tgt,const git_index_entry * src)1065 static void index_entry_cpy(
1066 	git_index_entry *tgt,
1067 	const git_index_entry *src)
1068 {
1069 	const char *tgt_path = tgt->path;
1070 	memcpy(tgt, src, sizeof(*tgt));
1071 	tgt->path = tgt_path;
1072 }
1073 
index_entry_dup(git_index_entry ** out,git_index * index,const git_index_entry * src)1074 static int index_entry_dup(
1075 	git_index_entry **out,
1076 	git_index *index,
1077 	const git_index_entry *src)
1078 {
1079 	if (index_entry_create(out, INDEX_OWNER(index), src->path, NULL, false) < 0)
1080 		return -1;
1081 
1082 	index_entry_cpy(*out, src);
1083 	return 0;
1084 }
1085 
index_entry_cpy_nocache(git_index_entry * tgt,const git_index_entry * src)1086 static void index_entry_cpy_nocache(
1087 	git_index_entry *tgt,
1088 	const git_index_entry *src)
1089 {
1090 	git_oid_cpy(&tgt->id, &src->id);
1091 	tgt->mode = src->mode;
1092 	tgt->flags = src->flags;
1093 	tgt->flags_extended = (src->flags_extended & GIT_INDEX_ENTRY_EXTENDED_FLAGS);
1094 }
1095 
index_entry_dup_nocache(git_index_entry ** out,git_index * index,const git_index_entry * src)1096 static int index_entry_dup_nocache(
1097 	git_index_entry **out,
1098 	git_index *index,
1099 	const git_index_entry *src)
1100 {
1101 	if (index_entry_create(out, INDEX_OWNER(index), src->path, NULL, false) < 0)
1102 		return -1;
1103 
1104 	index_entry_cpy_nocache(*out, src);
1105 	return 0;
1106 }
1107 
has_file_name(git_index * index,const git_index_entry * entry,size_t pos,int ok_to_replace)1108 static int has_file_name(git_index *index,
1109 	 const git_index_entry *entry, size_t pos, int ok_to_replace)
1110 {
1111 	size_t len = strlen(entry->path);
1112 	int stage = GIT_INDEX_ENTRY_STAGE(entry);
1113 	const char *name = entry->path;
1114 
1115 	while (pos < index->entries.length) {
1116 		struct entry_internal *p = index->entries.contents[pos++];
1117 
1118 		if (len >= p->pathlen)
1119 			break;
1120 		if (memcmp(name, p->path, len))
1121 			break;
1122 		if (GIT_INDEX_ENTRY_STAGE(&p->entry) != stage)
1123 			continue;
1124 		if (p->path[len] != '/')
1125 			continue;
1126 		if (!ok_to_replace)
1127 			return -1;
1128 
1129 		if (index_remove_entry(index, --pos) < 0)
1130 			break;
1131 	}
1132 	return 0;
1133 }
1134 
1135 /*
1136  * Do we have another file with a pathname that is a proper
1137  * subset of the name we're trying to add?
1138  */
has_dir_name(git_index * index,const git_index_entry * entry,int ok_to_replace)1139 static int has_dir_name(git_index *index,
1140 		const git_index_entry *entry, int ok_to_replace)
1141 {
1142 	int stage = GIT_INDEX_ENTRY_STAGE(entry);
1143 	const char *name = entry->path;
1144 	const char *slash = name + strlen(name);
1145 
1146 	for (;;) {
1147 		size_t len, pos;
1148 
1149 		for (;;) {
1150 			if (*--slash == '/')
1151 				break;
1152 			if (slash <= entry->path)
1153 				return 0;
1154 		}
1155 		len = slash - name;
1156 
1157 		if (!index_find(&pos, index, name, len, stage)) {
1158 			if (!ok_to_replace)
1159 				return -1;
1160 
1161 			if (index_remove_entry(index, pos) < 0)
1162 				break;
1163 			continue;
1164 		}
1165 
1166 		/*
1167 		 * Trivial optimization: if we find an entry that
1168 		 * already matches the sub-directory, then we know
1169 		 * we're ok, and we can exit.
1170 		 */
1171 		for (; pos < index->entries.length; ++pos) {
1172 			struct entry_internal *p = index->entries.contents[pos];
1173 
1174 			if (p->pathlen <= len ||
1175 			    p->path[len] != '/' ||
1176 			    memcmp(p->path, name, len))
1177 				break; /* not our subdirectory */
1178 
1179 			if (GIT_INDEX_ENTRY_STAGE(&p->entry) == stage)
1180 				return 0;
1181 		}
1182 	}
1183 
1184 	return 0;
1185 }
1186 
check_file_directory_collision(git_index * index,git_index_entry * entry,size_t pos,int ok_to_replace)1187 static int check_file_directory_collision(git_index *index,
1188 		git_index_entry *entry, size_t pos, int ok_to_replace)
1189 {
1190 	if (has_file_name(index, entry, pos, ok_to_replace) < 0 ||
1191 	    has_dir_name(index, entry, ok_to_replace) < 0) {
1192 		git_error_set(GIT_ERROR_INDEX,
1193 			"'%s' appears as both a file and a directory", entry->path);
1194 		return -1;
1195 	}
1196 
1197 	return 0;
1198 }
1199 
canonicalize_directory_path(git_index * index,git_index_entry * entry,git_index_entry * existing)1200 static int canonicalize_directory_path(
1201 	git_index *index,
1202 	git_index_entry *entry,
1203 	git_index_entry *existing)
1204 {
1205 	const git_index_entry *match, *best = NULL;
1206 	char *search, *sep;
1207 	size_t pos, search_len, best_len;
1208 
1209 	if (!index->ignore_case)
1210 		return 0;
1211 
1212 	/* item already exists in the index, simply re-use the existing case */
1213 	if (existing) {
1214 		memcpy((char *)entry->path, existing->path, strlen(existing->path));
1215 		return 0;
1216 	}
1217 
1218 	/* nothing to do */
1219 	if (strchr(entry->path, '/') == NULL)
1220 		return 0;
1221 
1222 	if ((search = git__strdup(entry->path)) == NULL)
1223 		return -1;
1224 
1225 	/* starting at the parent directory and descending to the root, find the
1226 	 * common parent directory.
1227 	 */
1228 	while (!best && (sep = strrchr(search, '/'))) {
1229 		sep[1] = '\0';
1230 
1231 		search_len = strlen(search);
1232 
1233 		git_vector_bsearch2(
1234 			&pos, &index->entries, index->entries_search_path, search);
1235 
1236 		while ((match = git_vector_get(&index->entries, pos))) {
1237 			if (GIT_INDEX_ENTRY_STAGE(match) != 0) {
1238 				/* conflicts do not contribute to canonical paths */
1239 			} else if (strncmp(search, match->path, search_len) == 0) {
1240 				/* prefer an exact match to the input filename */
1241 				best = match;
1242 				best_len = search_len;
1243 				break;
1244 			} else if (strncasecmp(search, match->path, search_len) == 0) {
1245 				/* continue walking, there may be a path with an exact
1246 				 * (case sensitive) match later in the index, but use this
1247 				 * as the best match until that happens.
1248 				 */
1249 				if (!best) {
1250 					best = match;
1251 					best_len = search_len;
1252 				}
1253 			} else {
1254 				break;
1255 			}
1256 
1257 			pos++;
1258 		}
1259 
1260 		sep[0] = '\0';
1261 	}
1262 
1263 	if (best)
1264 		memcpy((char *)entry->path, best->path, best_len);
1265 
1266 	git__free(search);
1267 	return 0;
1268 }
1269 
index_no_dups(void ** old,void * new)1270 static int index_no_dups(void **old, void *new)
1271 {
1272 	const git_index_entry *entry = new;
1273 	GIT_UNUSED(old);
1274 	git_error_set(GIT_ERROR_INDEX, "'%s' appears multiple times at stage %d",
1275 		entry->path, GIT_INDEX_ENTRY_STAGE(entry));
1276 	return GIT_EEXISTS;
1277 }
1278 
index_existing_and_best(git_index_entry ** existing,size_t * existing_position,git_index_entry ** best,git_index * index,const git_index_entry * entry)1279 static void index_existing_and_best(
1280 	git_index_entry **existing,
1281 	size_t *existing_position,
1282 	git_index_entry **best,
1283 	git_index *index,
1284 	const git_index_entry *entry)
1285 {
1286 	git_index_entry *e;
1287 	size_t pos;
1288 	int error;
1289 
1290 	error = index_find(&pos,
1291 		index, entry->path, 0, GIT_INDEX_ENTRY_STAGE(entry));
1292 
1293 	if (error == 0) {
1294 		*existing = index->entries.contents[pos];
1295 		*existing_position = pos;
1296 		*best = index->entries.contents[pos];
1297 		return;
1298 	}
1299 
1300 	*existing = NULL;
1301 	*existing_position = 0;
1302 	*best = NULL;
1303 
1304 	if (GIT_INDEX_ENTRY_STAGE(entry) == 0) {
1305 		for (; pos < index->entries.length; pos++) {
1306 			int (*strcomp)(const char *a, const char *b) =
1307 				index->ignore_case ? git__strcasecmp : git__strcmp;
1308 
1309 			e = index->entries.contents[pos];
1310 
1311 			if (strcomp(entry->path, e->path) != 0)
1312 				break;
1313 
1314 			if (GIT_INDEX_ENTRY_STAGE(e) == GIT_INDEX_STAGE_ANCESTOR) {
1315 				*best = e;
1316 				continue;
1317 			} else {
1318 				*best = e;
1319 				break;
1320 			}
1321 		}
1322 	}
1323 }
1324 
1325 /* index_insert takes ownership of the new entry - if it can't insert
1326  * it, then it will return an error **and also free the entry**.  When
1327  * it replaces an existing entry, it will update the entry_ptr with the
1328  * actual entry in the index (and free the passed in one).
1329  *
1330  * trust_path is whether we use the given path, or whether (on case
1331  * insensitive systems only) we try to canonicalize the given path to
1332  * be within an existing directory.
1333  *
1334  * trust_mode is whether we trust the mode in entry_ptr.
1335  *
1336  * trust_id is whether we trust the id or it should be validated.
1337  */
index_insert(git_index * index,git_index_entry ** entry_ptr,int replace,bool trust_path,bool trust_mode,bool trust_id)1338 static int index_insert(
1339 	git_index *index,
1340 	git_index_entry **entry_ptr,
1341 	int replace,
1342 	bool trust_path,
1343 	bool trust_mode,
1344 	bool trust_id)
1345 {
1346 	git_index_entry *existing, *best, *entry;
1347 	size_t path_length, position;
1348 	int error;
1349 
1350 	assert(index && entry_ptr);
1351 
1352 	entry = *entry_ptr;
1353 
1354 	/* Make sure that the path length flag is correct */
1355 	path_length = ((struct entry_internal *)entry)->pathlen;
1356 	index_entry_adjust_namemask(entry, path_length);
1357 
1358 	/* This entry is now up-to-date and should not be checked for raciness */
1359 	entry->flags_extended |= GIT_INDEX_ENTRY_UPTODATE;
1360 
1361 	git_vector_sort(&index->entries);
1362 
1363 	/*
1364 	 * Look if an entry with this path already exists, either staged, or (if
1365 	 * this entry is a regular staged item) as the "ours" side of a conflict.
1366 	 */
1367 	index_existing_and_best(&existing, &position, &best, index, entry);
1368 
1369 	/* Update the file mode */
1370 	entry->mode = trust_mode ?
1371 		git_index__create_mode(entry->mode) :
1372 		index_merge_mode(index, best, entry->mode);
1373 
1374 	/* Canonicalize the directory name */
1375 	if (!trust_path && (error = canonicalize_directory_path(index, entry, best)) < 0)
1376 		goto out;
1377 
1378 	/* Ensure that the given id exists (unless it's a submodule) */
1379 	if (!trust_id && INDEX_OWNER(index) &&
1380 	    (entry->mode & GIT_FILEMODE_COMMIT) != GIT_FILEMODE_COMMIT) {
1381 
1382 		if (!git_object__is_valid(INDEX_OWNER(index), &entry->id,
1383 					  git_object__type_from_filemode(entry->mode))) {
1384 			error = -1;
1385 			goto out;
1386 		}
1387 	}
1388 
1389 	/* Look for tree / blob name collisions, removing conflicts if requested */
1390 	if ((error = check_file_directory_collision(index, entry, position, replace)) < 0)
1391 		goto out;
1392 
1393 	/*
1394 	 * If we are replacing an existing item, overwrite the existing entry
1395 	 * and return it in place of the passed in one.
1396 	 */
1397 	if (existing) {
1398 		if (replace) {
1399 			index_entry_cpy(existing, entry);
1400 
1401 			if (trust_path)
1402 				memcpy((char *)existing->path, entry->path, strlen(entry->path));
1403 		}
1404 
1405 		index_entry_free(entry);
1406 		*entry_ptr = existing;
1407 	} else {
1408 		/*
1409 		 * If replace is not requested or no existing entry exists, insert
1410 		 * at the sorted position.  (Since we re-sort after each insert to
1411 		 * check for dups, this is actually cheaper in the long run.)
1412 		 */
1413 		if ((error = git_vector_insert_sorted(&index->entries, entry, index_no_dups)) < 0 ||
1414 		    (error = index_map_set(index->entries_map, entry, index->ignore_case)) < 0)
1415 			goto out;
1416 	}
1417 
1418 	index->dirty = 1;
1419 
1420 out:
1421 	if (error < 0) {
1422 		index_entry_free(*entry_ptr);
1423 		*entry_ptr = NULL;
1424 	}
1425 
1426 	return error;
1427 }
1428 
index_conflict_to_reuc(git_index * index,const char * path)1429 static int index_conflict_to_reuc(git_index *index, const char *path)
1430 {
1431 	const git_index_entry *conflict_entries[3];
1432 	int ancestor_mode, our_mode, their_mode;
1433 	git_oid const *ancestor_oid, *our_oid, *their_oid;
1434 	int ret;
1435 
1436 	if ((ret = git_index_conflict_get(&conflict_entries[0],
1437 		&conflict_entries[1], &conflict_entries[2], index, path)) < 0)
1438 		return ret;
1439 
1440 	ancestor_mode = conflict_entries[0] == NULL ? 0 : conflict_entries[0]->mode;
1441 	our_mode = conflict_entries[1] == NULL ? 0 : conflict_entries[1]->mode;
1442 	their_mode = conflict_entries[2] == NULL ? 0 : conflict_entries[2]->mode;
1443 
1444 	ancestor_oid = conflict_entries[0] == NULL ? NULL : &conflict_entries[0]->id;
1445 	our_oid = conflict_entries[1] == NULL ? NULL : &conflict_entries[1]->id;
1446 	their_oid = conflict_entries[2] == NULL ? NULL : &conflict_entries[2]->id;
1447 
1448 	if ((ret = git_index_reuc_add(index, path, ancestor_mode, ancestor_oid,
1449 		our_mode, our_oid, their_mode, their_oid)) >= 0)
1450 		ret = git_index_conflict_remove(index, path);
1451 
1452 	return ret;
1453 }
1454 
is_file_or_link(const int filemode)1455 GIT_INLINE(bool) is_file_or_link(const int filemode)
1456 {
1457 	return (filemode == GIT_FILEMODE_BLOB ||
1458 		filemode == GIT_FILEMODE_BLOB_EXECUTABLE ||
1459 		filemode == GIT_FILEMODE_LINK);
1460 }
1461 
valid_filemode(const int filemode)1462 GIT_INLINE(bool) valid_filemode(const int filemode)
1463 {
1464 	return (is_file_or_link(filemode) || filemode == GIT_FILEMODE_COMMIT);
1465 }
1466 
git_index_add_from_buffer(git_index * index,const git_index_entry * source_entry,const void * buffer,size_t len)1467 int git_index_add_from_buffer(
1468     git_index *index, const git_index_entry *source_entry,
1469     const void *buffer, size_t len)
1470 {
1471 	git_index_entry *entry = NULL;
1472 	int error = 0;
1473 	git_oid id;
1474 
1475 	assert(index && source_entry->path);
1476 
1477 	if (INDEX_OWNER(index) == NULL)
1478 		return create_index_error(-1,
1479 			"could not initialize index entry. "
1480 			"Index is not backed up by an existing repository.");
1481 
1482 	if (!is_file_or_link(source_entry->mode)) {
1483 		git_error_set(GIT_ERROR_INDEX, "invalid filemode");
1484 		return -1;
1485 	}
1486 
1487 	if (len > UINT32_MAX) {
1488 		git_error_set(GIT_ERROR_INDEX, "buffer is too large");
1489 		return -1;
1490 	}
1491 
1492 	if (index_entry_dup(&entry, index, source_entry) < 0)
1493 		return -1;
1494 
1495 	error = git_blob_create_from_buffer(&id, INDEX_OWNER(index), buffer, len);
1496 	if (error < 0) {
1497 		index_entry_free(entry);
1498 		return error;
1499 	}
1500 
1501 	git_oid_cpy(&entry->id, &id);
1502 	entry->file_size = (uint32_t)len;
1503 
1504 	if ((error = index_insert(index, &entry, 1, true, true, true)) < 0)
1505 		return error;
1506 
1507 	/* Adding implies conflict was resolved, move conflict entries to REUC */
1508 	if ((error = index_conflict_to_reuc(index, entry->path)) < 0 && error != GIT_ENOTFOUND)
1509 		return error;
1510 
1511 	git_tree_cache_invalidate_path(index->tree, entry->path);
1512 	return 0;
1513 }
1514 
add_repo_as_submodule(git_index_entry ** out,git_index * index,const char * path)1515 static int add_repo_as_submodule(git_index_entry **out, git_index *index, const char *path)
1516 {
1517 	git_repository *sub;
1518 	git_buf abspath = GIT_BUF_INIT;
1519 	git_repository *repo = INDEX_OWNER(index);
1520 	git_reference *head;
1521 	git_index_entry *entry;
1522 	struct stat st;
1523 	int error;
1524 
1525 	if ((error = git_buf_joinpath(&abspath, git_repository_workdir(repo), path)) < 0)
1526 		return error;
1527 
1528 	if ((error = p_stat(abspath.ptr, &st)) < 0) {
1529 		git_error_set(GIT_ERROR_OS, "failed to stat repository dir");
1530 		return -1;
1531 	}
1532 
1533 	if (index_entry_create(&entry, INDEX_OWNER(index), path, &st, true) < 0)
1534 		return -1;
1535 
1536 	git_index_entry__init_from_stat(entry, &st, !index->distrust_filemode);
1537 
1538 	if ((error = git_repository_open(&sub, abspath.ptr)) < 0)
1539 		return error;
1540 
1541 	if ((error = git_repository_head(&head, sub)) < 0)
1542 		return error;
1543 
1544 	git_oid_cpy(&entry->id, git_reference_target(head));
1545 	entry->mode = GIT_FILEMODE_COMMIT;
1546 
1547 	git_reference_free(head);
1548 	git_repository_free(sub);
1549 	git_buf_dispose(&abspath);
1550 
1551 	*out = entry;
1552 	return 0;
1553 }
1554 
git_index_add_bypath(git_index * index,const char * path)1555 int git_index_add_bypath(git_index *index, const char *path)
1556 {
1557 	git_index_entry *entry = NULL;
1558 	int ret;
1559 
1560 	assert(index && path);
1561 
1562 	if ((ret = index_entry_init(&entry, index, path)) == 0)
1563 		ret = index_insert(index, &entry, 1, false, false, true);
1564 
1565 	/* If we were given a directory, let's see if it's a submodule */
1566 	if (ret < 0 && ret != GIT_EDIRECTORY)
1567 		return ret;
1568 
1569 	if (ret == GIT_EDIRECTORY) {
1570 		git_submodule *sm;
1571 		git_error_state err;
1572 
1573 		git_error_state_capture(&err, ret);
1574 
1575 		ret = git_submodule_lookup(&sm, INDEX_OWNER(index), path);
1576 		if (ret == GIT_ENOTFOUND)
1577 			return git_error_state_restore(&err);
1578 
1579 		git_error_state_free(&err);
1580 
1581 		/*
1582 		 * EEXISTS means that there is a repository at that path, but it's not known
1583 		 * as a submodule. We add its HEAD as an entry and don't register it.
1584 		 */
1585 		if (ret == GIT_EEXISTS) {
1586 			if ((ret = add_repo_as_submodule(&entry, index, path)) < 0)
1587 				return ret;
1588 
1589 			if ((ret = index_insert(index, &entry, 1, false, false, true)) < 0)
1590 				return ret;
1591 		} else if (ret < 0) {
1592 			return ret;
1593 		} else {
1594 			ret = git_submodule_add_to_index(sm, false);
1595 			git_submodule_free(sm);
1596 			return ret;
1597 		}
1598 	}
1599 
1600 	/* Adding implies conflict was resolved, move conflict entries to REUC */
1601 	if ((ret = index_conflict_to_reuc(index, path)) < 0 && ret != GIT_ENOTFOUND)
1602 		return ret;
1603 
1604 	git_tree_cache_invalidate_path(index->tree, entry->path);
1605 	return 0;
1606 }
1607 
git_index_remove_bypath(git_index * index,const char * path)1608 int git_index_remove_bypath(git_index *index, const char *path)
1609 {
1610 	int ret;
1611 
1612 	assert(index && path);
1613 
1614 	if (((ret = git_index_remove(index, path, 0)) < 0 &&
1615 		ret != GIT_ENOTFOUND) ||
1616 		((ret = index_conflict_to_reuc(index, path)) < 0 &&
1617 		ret != GIT_ENOTFOUND))
1618 		return ret;
1619 
1620 	if (ret == GIT_ENOTFOUND)
1621 		git_error_clear();
1622 
1623 	return 0;
1624 }
1625 
git_index__fill(git_index * index,const git_vector * source_entries)1626 int git_index__fill(git_index *index, const git_vector *source_entries)
1627 {
1628 	const git_index_entry *source_entry = NULL;
1629 	int error = 0;
1630 	size_t i;
1631 
1632 	assert(index);
1633 
1634 	if (!source_entries->length)
1635 		return 0;
1636 
1637 	if (git_vector_size_hint(&index->entries, source_entries->length) < 0 ||
1638 	    index_map_resize(index->entries_map, (size_t)(source_entries->length * 1.3),
1639 			     index->ignore_case) < 0)
1640 		return -1;
1641 
1642 	git_vector_foreach(source_entries, i, source_entry) {
1643 		git_index_entry *entry = NULL;
1644 
1645 		if ((error = index_entry_dup(&entry, index, source_entry)) < 0)
1646 			break;
1647 
1648 		index_entry_adjust_namemask(entry, ((struct entry_internal *)entry)->pathlen);
1649 		entry->flags_extended |= GIT_INDEX_ENTRY_UPTODATE;
1650 		entry->mode = git_index__create_mode(entry->mode);
1651 
1652 		if ((error = git_vector_insert(&index->entries, entry)) < 0)
1653 			break;
1654 
1655 		if ((error = index_map_set(index->entries_map, entry, index->ignore_case)) < 0)
1656 			break;
1657 
1658 		index->dirty = 1;
1659 	}
1660 
1661 	if (!error)
1662 		git_vector_sort(&index->entries);
1663 
1664 	return error;
1665 }
1666 
1667 
git_index_add(git_index * index,const git_index_entry * source_entry)1668 int git_index_add(git_index *index, const git_index_entry *source_entry)
1669 {
1670 	git_index_entry *entry = NULL;
1671 	int ret;
1672 
1673 	assert(index && source_entry && source_entry->path);
1674 
1675 	if (!valid_filemode(source_entry->mode)) {
1676 		git_error_set(GIT_ERROR_INDEX, "invalid entry mode");
1677 		return -1;
1678 	}
1679 
1680 	if ((ret = index_entry_dup(&entry, index, source_entry)) < 0 ||
1681 		(ret = index_insert(index, &entry, 1, true, true, false)) < 0)
1682 		return ret;
1683 
1684 	git_tree_cache_invalidate_path(index->tree, entry->path);
1685 	return 0;
1686 }
1687 
git_index_remove(git_index * index,const char * path,int stage)1688 int git_index_remove(git_index *index, const char *path, int stage)
1689 {
1690 	int error;
1691 	size_t position;
1692 	git_index_entry remove_key = {{ 0 }};
1693 
1694 	remove_key.path = path;
1695 	GIT_INDEX_ENTRY_STAGE_SET(&remove_key, stage);
1696 
1697 	index_map_delete(index->entries_map, &remove_key, index->ignore_case);
1698 
1699 	if (index_find(&position, index, path, 0, stage) < 0) {
1700 		git_error_set(
1701 			GIT_ERROR_INDEX, "index does not contain %s at stage %d", path, stage);
1702 		error = GIT_ENOTFOUND;
1703 	} else {
1704 		error = index_remove_entry(index, position);
1705 	}
1706 
1707 	return error;
1708 }
1709 
git_index_remove_directory(git_index * index,const char * dir,int stage)1710 int git_index_remove_directory(git_index *index, const char *dir, int stage)
1711 {
1712 	git_buf pfx = GIT_BUF_INIT;
1713 	int error = 0;
1714 	size_t pos;
1715 	git_index_entry *entry;
1716 
1717 	if (!(error = git_buf_sets(&pfx, dir)) &&
1718 		!(error = git_path_to_dir(&pfx)))
1719 		index_find(&pos, index, pfx.ptr, pfx.size, GIT_INDEX_STAGE_ANY);
1720 
1721 	while (!error) {
1722 		entry = git_vector_get(&index->entries, pos);
1723 		if (!entry || git__prefixcmp(entry->path, pfx.ptr) != 0)
1724 			break;
1725 
1726 		if (GIT_INDEX_ENTRY_STAGE(entry) != stage) {
1727 			++pos;
1728 			continue;
1729 		}
1730 
1731 		error = index_remove_entry(index, pos);
1732 
1733 		/* removed entry at 'pos' so we don't need to increment */
1734 	}
1735 
1736 	git_buf_dispose(&pfx);
1737 
1738 	return error;
1739 }
1740 
git_index_find_prefix(size_t * at_pos,git_index * index,const char * prefix)1741 int git_index_find_prefix(size_t *at_pos, git_index *index, const char *prefix)
1742 {
1743 	int error = 0;
1744 	size_t pos;
1745 	const git_index_entry *entry;
1746 
1747 	index_find(&pos, index, prefix, strlen(prefix), GIT_INDEX_STAGE_ANY);
1748 	entry = git_vector_get(&index->entries, pos);
1749 	if (!entry || git__prefixcmp(entry->path, prefix) != 0)
1750 		error = GIT_ENOTFOUND;
1751 
1752 	if (!error && at_pos)
1753 		*at_pos = pos;
1754 
1755 	return error;
1756 }
1757 
git_index__find_pos(size_t * out,git_index * index,const char * path,size_t path_len,int stage)1758 int git_index__find_pos(
1759 	size_t *out, git_index *index, const char *path, size_t path_len, int stage)
1760 {
1761 	assert(index && path);
1762 	return index_find(out, index, path, path_len, stage);
1763 }
1764 
git_index_find(size_t * at_pos,git_index * index,const char * path)1765 int git_index_find(size_t *at_pos, git_index *index, const char *path)
1766 {
1767 	size_t pos;
1768 
1769 	assert(index && path);
1770 
1771 	if (git_vector_bsearch2(
1772 			&pos, &index->entries, index->entries_search_path, path) < 0) {
1773 		git_error_set(GIT_ERROR_INDEX, "index does not contain %s", path);
1774 		return GIT_ENOTFOUND;
1775 	}
1776 
1777 	/* Since our binary search only looked at path, we may be in the
1778 	 * middle of a list of stages.
1779 	 */
1780 	for (; pos > 0; --pos) {
1781 		const git_index_entry *prev = git_vector_get(&index->entries, pos - 1);
1782 
1783 		if (index->entries_cmp_path(prev->path, path) != 0)
1784 			break;
1785 	}
1786 
1787 	if (at_pos)
1788 		*at_pos = pos;
1789 
1790 	return 0;
1791 }
1792 
git_index_conflict_add(git_index * index,const git_index_entry * ancestor_entry,const git_index_entry * our_entry,const git_index_entry * their_entry)1793 int git_index_conflict_add(git_index *index,
1794 	const git_index_entry *ancestor_entry,
1795 	const git_index_entry *our_entry,
1796 	const git_index_entry *their_entry)
1797 {
1798 	git_index_entry *entries[3] = { 0 };
1799 	unsigned short i;
1800 	int ret = 0;
1801 
1802 	assert (index);
1803 
1804 	if ((ancestor_entry &&
1805 			(ret = index_entry_dup(&entries[0], index, ancestor_entry)) < 0) ||
1806 		(our_entry &&
1807 			(ret = index_entry_dup(&entries[1], index, our_entry)) < 0) ||
1808 		(their_entry &&
1809 			(ret = index_entry_dup(&entries[2], index, their_entry)) < 0))
1810 		goto on_error;
1811 
1812 	/* Validate entries */
1813 	for (i = 0; i < 3; i++) {
1814 		if (entries[i] && !valid_filemode(entries[i]->mode)) {
1815 			git_error_set(GIT_ERROR_INDEX, "invalid filemode for stage %d entry",
1816 				i + 1);
1817 			ret = -1;
1818 			goto on_error;
1819 		}
1820 	}
1821 
1822 	/* Remove existing index entries for each path */
1823 	for (i = 0; i < 3; i++) {
1824 		if (entries[i] == NULL)
1825 			continue;
1826 
1827 		if ((ret = git_index_remove(index, entries[i]->path, 0)) != 0) {
1828 			if (ret != GIT_ENOTFOUND)
1829 				goto on_error;
1830 
1831 			git_error_clear();
1832 			ret = 0;
1833 		}
1834 	}
1835 
1836 	/* Add the conflict entries */
1837 	for (i = 0; i < 3; i++) {
1838 		if (entries[i] == NULL)
1839 			continue;
1840 
1841 		/* Make sure stage is correct */
1842 		GIT_INDEX_ENTRY_STAGE_SET(entries[i], i + 1);
1843 
1844 		if ((ret = index_insert(index, &entries[i], 1, true, true, false)) < 0)
1845 			goto on_error;
1846 
1847 		entries[i] = NULL; /* don't free if later entry fails */
1848 	}
1849 
1850 	return 0;
1851 
1852 on_error:
1853 	for (i = 0; i < 3; i++) {
1854 		if (entries[i] != NULL)
1855 			index_entry_free(entries[i]);
1856 	}
1857 
1858 	return ret;
1859 }
1860 
index_conflict__get_byindex(const git_index_entry ** ancestor_out,const git_index_entry ** our_out,const git_index_entry ** their_out,git_index * index,size_t n)1861 static int index_conflict__get_byindex(
1862 	const git_index_entry **ancestor_out,
1863 	const git_index_entry **our_out,
1864 	const git_index_entry **their_out,
1865 	git_index *index,
1866 	size_t n)
1867 {
1868 	const git_index_entry *conflict_entry;
1869 	const char *path = NULL;
1870 	size_t count;
1871 	int stage, len = 0;
1872 
1873 	assert(ancestor_out && our_out && their_out && index);
1874 
1875 	*ancestor_out = NULL;
1876 	*our_out = NULL;
1877 	*their_out = NULL;
1878 
1879 	for (count = git_index_entrycount(index); n < count; ++n) {
1880 		conflict_entry = git_vector_get(&index->entries, n);
1881 
1882 		if (path && index->entries_cmp_path(conflict_entry->path, path) != 0)
1883 			break;
1884 
1885 		stage = GIT_INDEX_ENTRY_STAGE(conflict_entry);
1886 		path = conflict_entry->path;
1887 
1888 		switch (stage) {
1889 		case 3:
1890 			*their_out = conflict_entry;
1891 			len++;
1892 			break;
1893 		case 2:
1894 			*our_out = conflict_entry;
1895 			len++;
1896 			break;
1897 		case 1:
1898 			*ancestor_out = conflict_entry;
1899 			len++;
1900 			break;
1901 		default:
1902 			break;
1903 		};
1904 	}
1905 
1906 	return len;
1907 }
1908 
git_index_conflict_get(const git_index_entry ** ancestor_out,const git_index_entry ** our_out,const git_index_entry ** their_out,git_index * index,const char * path)1909 int git_index_conflict_get(
1910 	const git_index_entry **ancestor_out,
1911 	const git_index_entry **our_out,
1912 	const git_index_entry **their_out,
1913 	git_index *index,
1914 	const char *path)
1915 {
1916 	size_t pos;
1917 	int len = 0;
1918 
1919 	assert(ancestor_out && our_out && their_out && index && path);
1920 
1921 	*ancestor_out = NULL;
1922 	*our_out = NULL;
1923 	*their_out = NULL;
1924 
1925 	if (git_index_find(&pos, index, path) < 0)
1926 		return GIT_ENOTFOUND;
1927 
1928 	if ((len = index_conflict__get_byindex(
1929 		ancestor_out, our_out, their_out, index, pos)) < 0)
1930 		return len;
1931 	else if (len == 0)
1932 		return GIT_ENOTFOUND;
1933 
1934 	return 0;
1935 }
1936 
index_conflict_remove(git_index * index,const char * path)1937 static int index_conflict_remove(git_index *index, const char *path)
1938 {
1939 	size_t pos = 0;
1940 	git_index_entry *conflict_entry;
1941 	int error = 0;
1942 
1943 	if (path != NULL && git_index_find(&pos, index, path) < 0)
1944 		return GIT_ENOTFOUND;
1945 
1946 	while ((conflict_entry = git_vector_get(&index->entries, pos)) != NULL) {
1947 
1948 		if (path != NULL &&
1949 			index->entries_cmp_path(conflict_entry->path, path) != 0)
1950 			break;
1951 
1952 		if (GIT_INDEX_ENTRY_STAGE(conflict_entry) == 0) {
1953 			pos++;
1954 			continue;
1955 		}
1956 
1957 		if ((error = index_remove_entry(index, pos)) < 0)
1958 			break;
1959 	}
1960 
1961 	return error;
1962 }
1963 
git_index_conflict_remove(git_index * index,const char * path)1964 int git_index_conflict_remove(git_index *index, const char *path)
1965 {
1966 	assert(index && path);
1967 	return index_conflict_remove(index, path);
1968 }
1969 
git_index_conflict_cleanup(git_index * index)1970 int git_index_conflict_cleanup(git_index *index)
1971 {
1972 	assert(index);
1973 	return index_conflict_remove(index, NULL);
1974 }
1975 
git_index_has_conflicts(const git_index * index)1976 int git_index_has_conflicts(const git_index *index)
1977 {
1978 	size_t i;
1979 	git_index_entry *entry;
1980 
1981 	assert(index);
1982 
1983 	git_vector_foreach(&index->entries, i, entry) {
1984 		if (GIT_INDEX_ENTRY_STAGE(entry) > 0)
1985 			return 1;
1986 	}
1987 
1988 	return 0;
1989 }
1990 
git_index_iterator_new(git_index_iterator ** iterator_out,git_index * index)1991 int git_index_iterator_new(
1992 	git_index_iterator **iterator_out,
1993 	git_index *index)
1994 {
1995 	git_index_iterator *it;
1996 	int error;
1997 
1998 	assert(iterator_out && index);
1999 
2000 	it = git__calloc(1, sizeof(git_index_iterator));
2001 	GIT_ERROR_CHECK_ALLOC(it);
2002 
2003 	if ((error = git_index_snapshot_new(&it->snap, index)) < 0) {
2004 		git__free(it);
2005 		return error;
2006 	}
2007 
2008 	it->index = index;
2009 
2010 	*iterator_out = it;
2011 	return 0;
2012 }
2013 
git_index_iterator_next(const git_index_entry ** out,git_index_iterator * it)2014 int git_index_iterator_next(
2015 	const git_index_entry **out,
2016 	git_index_iterator *it)
2017 {
2018 	assert(out && it);
2019 
2020 	if (it->cur >= git_vector_length(&it->snap))
2021 		return GIT_ITEROVER;
2022 
2023 	*out = (git_index_entry *)git_vector_get(&it->snap, it->cur++);
2024 	return 0;
2025 }
2026 
git_index_iterator_free(git_index_iterator * it)2027 void git_index_iterator_free(git_index_iterator *it)
2028 {
2029 	if (it == NULL)
2030 		return;
2031 
2032 	git_index_snapshot_release(&it->snap, it->index);
2033 	git__free(it);
2034 }
2035 
git_index_conflict_iterator_new(git_index_conflict_iterator ** iterator_out,git_index * index)2036 int git_index_conflict_iterator_new(
2037 	git_index_conflict_iterator **iterator_out,
2038 	git_index *index)
2039 {
2040 	git_index_conflict_iterator *it = NULL;
2041 
2042 	assert(iterator_out && index);
2043 
2044 	it = git__calloc(1, sizeof(git_index_conflict_iterator));
2045 	GIT_ERROR_CHECK_ALLOC(it);
2046 
2047 	it->index = index;
2048 
2049 	*iterator_out = it;
2050 	return 0;
2051 }
2052 
git_index_conflict_next(const git_index_entry ** ancestor_out,const git_index_entry ** our_out,const git_index_entry ** their_out,git_index_conflict_iterator * iterator)2053 int git_index_conflict_next(
2054 	const git_index_entry **ancestor_out,
2055 	const git_index_entry **our_out,
2056 	const git_index_entry **their_out,
2057 	git_index_conflict_iterator *iterator)
2058 {
2059 	const git_index_entry *entry;
2060 	int len;
2061 
2062 	assert(ancestor_out && our_out && their_out && iterator);
2063 
2064 	*ancestor_out = NULL;
2065 	*our_out = NULL;
2066 	*their_out = NULL;
2067 
2068 	while (iterator->cur < iterator->index->entries.length) {
2069 		entry = git_index_get_byindex(iterator->index, iterator->cur);
2070 
2071 		if (git_index_entry_is_conflict(entry)) {
2072 			if ((len = index_conflict__get_byindex(
2073 				ancestor_out,
2074 				our_out,
2075 				their_out,
2076 				iterator->index,
2077 				iterator->cur)) < 0)
2078 				return len;
2079 
2080 			iterator->cur += len;
2081 			return 0;
2082 		}
2083 
2084 		iterator->cur++;
2085 	}
2086 
2087 	return GIT_ITEROVER;
2088 }
2089 
git_index_conflict_iterator_free(git_index_conflict_iterator * iterator)2090 void git_index_conflict_iterator_free(git_index_conflict_iterator *iterator)
2091 {
2092 	if (iterator == NULL)
2093 		return;
2094 
2095 	git__free(iterator);
2096 }
2097 
git_index_name_entrycount(git_index * index)2098 size_t git_index_name_entrycount(git_index *index)
2099 {
2100 	assert(index);
2101 	return index->names.length;
2102 }
2103 
git_index_name_get_byindex(git_index * index,size_t n)2104 const git_index_name_entry *git_index_name_get_byindex(
2105 	git_index *index, size_t n)
2106 {
2107 	assert(index);
2108 
2109 	git_vector_sort(&index->names);
2110 	return git_vector_get(&index->names, n);
2111 }
2112 
index_name_entry_free(git_index_name_entry * ne)2113 static void index_name_entry_free(git_index_name_entry *ne)
2114 {
2115 	if (!ne)
2116 		return;
2117 	git__free(ne->ancestor);
2118 	git__free(ne->ours);
2119 	git__free(ne->theirs);
2120 	git__free(ne);
2121 }
2122 
git_index_name_add(git_index * index,const char * ancestor,const char * ours,const char * theirs)2123 int git_index_name_add(git_index *index,
2124 	const char *ancestor, const char *ours, const char *theirs)
2125 {
2126 	git_index_name_entry *conflict_name;
2127 
2128 	assert((ancestor && ours) || (ancestor && theirs) || (ours && theirs));
2129 
2130 	conflict_name = git__calloc(1, sizeof(git_index_name_entry));
2131 	GIT_ERROR_CHECK_ALLOC(conflict_name);
2132 
2133 	if ((ancestor && !(conflict_name->ancestor = git__strdup(ancestor))) ||
2134 		(ours     && !(conflict_name->ours     = git__strdup(ours))) ||
2135 		(theirs   && !(conflict_name->theirs   = git__strdup(theirs))) ||
2136 		git_vector_insert(&index->names, conflict_name) < 0)
2137 	{
2138 		index_name_entry_free(conflict_name);
2139 		return -1;
2140 	}
2141 
2142 	index->dirty = 1;
2143 	return 0;
2144 }
2145 
git_index_name_clear(git_index * index)2146 int git_index_name_clear(git_index *index)
2147 {
2148 	size_t i;
2149 	git_index_name_entry *conflict_name;
2150 
2151 	assert(index);
2152 
2153 	git_vector_foreach(&index->names, i, conflict_name)
2154 		index_name_entry_free(conflict_name);
2155 
2156 	git_vector_clear(&index->names);
2157 
2158 	index->dirty = 1;
2159 
2160 	return 0;
2161 }
2162 
git_index_reuc_entrycount(git_index * index)2163 size_t git_index_reuc_entrycount(git_index *index)
2164 {
2165 	assert(index);
2166 	return index->reuc.length;
2167 }
2168 
index_reuc_on_dup(void ** old,void * new)2169 static int index_reuc_on_dup(void **old, void *new)
2170 {
2171 	index_entry_reuc_free(*old);
2172 	*old = new;
2173 	return GIT_EEXISTS;
2174 }
2175 
index_reuc_insert(git_index * index,git_index_reuc_entry * reuc)2176 static int index_reuc_insert(
2177 	git_index *index,
2178 	git_index_reuc_entry *reuc)
2179 {
2180 	int res;
2181 
2182 	assert(index && reuc && reuc->path != NULL);
2183 	assert(git_vector_is_sorted(&index->reuc));
2184 
2185 	res = git_vector_insert_sorted(&index->reuc, reuc, &index_reuc_on_dup);
2186 	index->dirty = 1;
2187 
2188 	return res == GIT_EEXISTS ? 0 : res;
2189 }
2190 
git_index_reuc_add(git_index * index,const char * path,int ancestor_mode,const git_oid * ancestor_oid,int our_mode,const git_oid * our_oid,int their_mode,const git_oid * their_oid)2191 int git_index_reuc_add(git_index *index, const char *path,
2192 	int ancestor_mode, const git_oid *ancestor_oid,
2193 	int our_mode, const git_oid *our_oid,
2194 	int their_mode, const git_oid *their_oid)
2195 {
2196 	git_index_reuc_entry *reuc = NULL;
2197 	int error = 0;
2198 
2199 	assert(index && path);
2200 
2201 	if ((error = index_entry_reuc_init(&reuc, path, ancestor_mode,
2202 			ancestor_oid, our_mode, our_oid, their_mode, their_oid)) < 0 ||
2203 		(error = index_reuc_insert(index, reuc)) < 0)
2204 		index_entry_reuc_free(reuc);
2205 
2206 	return error;
2207 }
2208 
git_index_reuc_find(size_t * at_pos,git_index * index,const char * path)2209 int git_index_reuc_find(size_t *at_pos, git_index *index, const char *path)
2210 {
2211 	return git_vector_bsearch2(at_pos, &index->reuc, index->reuc_search, path);
2212 }
2213 
git_index_reuc_get_bypath(git_index * index,const char * path)2214 const git_index_reuc_entry *git_index_reuc_get_bypath(
2215 	git_index *index, const char *path)
2216 {
2217 	size_t pos;
2218 	assert(index && path);
2219 
2220 	if (!index->reuc.length)
2221 		return NULL;
2222 
2223 	assert(git_vector_is_sorted(&index->reuc));
2224 
2225 	if (git_index_reuc_find(&pos, index, path) < 0)
2226 		return NULL;
2227 
2228 	return git_vector_get(&index->reuc, pos);
2229 }
2230 
git_index_reuc_get_byindex(git_index * index,size_t n)2231 const git_index_reuc_entry *git_index_reuc_get_byindex(
2232 	git_index *index, size_t n)
2233 {
2234 	assert(index);
2235 	assert(git_vector_is_sorted(&index->reuc));
2236 
2237 	return git_vector_get(&index->reuc, n);
2238 }
2239 
git_index_reuc_remove(git_index * index,size_t position)2240 int git_index_reuc_remove(git_index *index, size_t position)
2241 {
2242 	int error;
2243 	git_index_reuc_entry *reuc;
2244 
2245 	assert(git_vector_is_sorted(&index->reuc));
2246 
2247 	reuc = git_vector_get(&index->reuc, position);
2248 	error = git_vector_remove(&index->reuc, position);
2249 
2250 	if (!error)
2251 		index_entry_reuc_free(reuc);
2252 
2253 	index->dirty = 1;
2254 	return error;
2255 }
2256 
git_index_reuc_clear(git_index * index)2257 int git_index_reuc_clear(git_index *index)
2258 {
2259 	size_t i;
2260 
2261 	assert(index);
2262 
2263 	for (i = 0; i < index->reuc.length; ++i)
2264 		index_entry_reuc_free(git__swap(index->reuc.contents[i], NULL));
2265 
2266 	git_vector_clear(&index->reuc);
2267 
2268 	index->dirty = 1;
2269 
2270 	return 0;
2271 }
2272 
index_error_invalid(const char * message)2273 static int index_error_invalid(const char *message)
2274 {
2275 	git_error_set(GIT_ERROR_INDEX, "invalid data in index - %s", message);
2276 	return -1;
2277 }
2278 
read_reuc(git_index * index,const char * buffer,size_t size)2279 static int read_reuc(git_index *index, const char *buffer, size_t size)
2280 {
2281 	const char *endptr;
2282 	size_t len;
2283 	int i;
2284 
2285 	/* If called multiple times, the vector might already be initialized */
2286 	if (index->reuc._alloc_size == 0 &&
2287 		git_vector_init(&index->reuc, 16, reuc_cmp) < 0)
2288 		return -1;
2289 
2290 	while (size) {
2291 		git_index_reuc_entry *lost;
2292 
2293 		len = p_strnlen(buffer, size) + 1;
2294 		if (size <= len)
2295 			return index_error_invalid("reading reuc entries");
2296 
2297 		lost = reuc_entry_alloc(buffer);
2298 		GIT_ERROR_CHECK_ALLOC(lost);
2299 
2300 		size -= len;
2301 		buffer += len;
2302 
2303 		/* read 3 ASCII octal numbers for stage entries */
2304 		for (i = 0; i < 3; i++) {
2305 			int64_t tmp;
2306 
2307 			if (git__strntol64(&tmp, buffer, size, &endptr, 8) < 0 ||
2308 				!endptr || endptr == buffer || *endptr ||
2309 				tmp < 0 || tmp > UINT32_MAX) {
2310 				index_entry_reuc_free(lost);
2311 				return index_error_invalid("reading reuc entry stage");
2312 			}
2313 
2314 			lost->mode[i] = (uint32_t)tmp;
2315 
2316 			len = (endptr + 1) - buffer;
2317 			if (size <= len) {
2318 				index_entry_reuc_free(lost);
2319 				return index_error_invalid("reading reuc entry stage");
2320 			}
2321 
2322 			size -= len;
2323 			buffer += len;
2324 		}
2325 
2326 		/* read up to 3 OIDs for stage entries */
2327 		for (i = 0; i < 3; i++) {
2328 			if (!lost->mode[i])
2329 				continue;
2330 			if (size < 20) {
2331 				index_entry_reuc_free(lost);
2332 				return index_error_invalid("reading reuc entry oid");
2333 			}
2334 
2335 			git_oid_fromraw(&lost->oid[i], (const unsigned char *) buffer);
2336 			size -= 20;
2337 			buffer += 20;
2338 		}
2339 
2340 		/* entry was read successfully - insert into reuc vector */
2341 		if (git_vector_insert(&index->reuc, lost) < 0)
2342 			return -1;
2343 	}
2344 
2345 	/* entries are guaranteed to be sorted on-disk */
2346 	git_vector_set_sorted(&index->reuc, true);
2347 
2348 	return 0;
2349 }
2350 
2351 
read_conflict_names(git_index * index,const char * buffer,size_t size)2352 static int read_conflict_names(git_index *index, const char *buffer, size_t size)
2353 {
2354 	size_t len;
2355 
2356 	/* This gets called multiple times, the vector might already be initialized */
2357 	if (index->names._alloc_size == 0 &&
2358 		git_vector_init(&index->names, 16, conflict_name_cmp) < 0)
2359 		return -1;
2360 
2361 #define read_conflict_name(ptr) \
2362 	len = p_strnlen(buffer, size) + 1; \
2363 	if (size < len) { \
2364 		index_error_invalid("reading conflict name entries"); \
2365 		goto out_err; \
2366 	} \
2367 	if (len == 1) \
2368 		ptr = NULL; \
2369 	else { \
2370 		ptr = git__malloc(len); \
2371 		GIT_ERROR_CHECK_ALLOC(ptr); \
2372 		memcpy(ptr, buffer, len); \
2373 	} \
2374 	\
2375 	buffer += len; \
2376 	size -= len;
2377 
2378 	while (size) {
2379 		git_index_name_entry *conflict_name = git__calloc(1, sizeof(git_index_name_entry));
2380 		GIT_ERROR_CHECK_ALLOC(conflict_name);
2381 
2382 		read_conflict_name(conflict_name->ancestor);
2383 		read_conflict_name(conflict_name->ours);
2384 		read_conflict_name(conflict_name->theirs);
2385 
2386 		if (git_vector_insert(&index->names, conflict_name) < 0)
2387 			goto out_err;
2388 
2389 		continue;
2390 
2391 out_err:
2392 		git__free(conflict_name->ancestor);
2393 		git__free(conflict_name->ours);
2394 		git__free(conflict_name->theirs);
2395 		git__free(conflict_name);
2396 		return -1;
2397 	}
2398 
2399 #undef read_conflict_name
2400 
2401 	/* entries are guaranteed to be sorted on-disk */
2402 	git_vector_set_sorted(&index->names, true);
2403 
2404 	return 0;
2405 }
2406 
index_entry_size(size_t path_len,size_t varint_len,uint32_t flags)2407 static size_t index_entry_size(size_t path_len, size_t varint_len, uint32_t flags)
2408 {
2409 	if (varint_len) {
2410 		if (flags & GIT_INDEX_ENTRY_EXTENDED)
2411 			return offsetof(struct entry_long, path) + path_len + 1 + varint_len;
2412 		else
2413 			return offsetof(struct entry_short, path) + path_len + 1 + varint_len;
2414 	} else {
2415 #define entry_size(type,len) ((offsetof(type, path) + (len) + 8) & ~7)
2416 		if (flags & GIT_INDEX_ENTRY_EXTENDED)
2417 			return entry_size(struct entry_long, path_len);
2418 		else
2419 			return entry_size(struct entry_short, path_len);
2420 #undef entry_size
2421 	}
2422 }
2423 
read_entry(git_index_entry ** out,size_t * out_size,git_index * index,const void * buffer,size_t buffer_size,const char * last)2424 static int read_entry(
2425 	git_index_entry **out,
2426 	size_t *out_size,
2427 	git_index *index,
2428 	const void *buffer,
2429 	size_t buffer_size,
2430 	const char *last)
2431 {
2432 	size_t path_length, entry_size;
2433 	const char *path_ptr;
2434 	struct entry_short source;
2435 	git_index_entry entry = {{0}};
2436 	bool compressed = index->version >= INDEX_VERSION_NUMBER_COMP;
2437 	char *tmp_path = NULL;
2438 
2439 	if (INDEX_FOOTER_SIZE + minimal_entry_size > buffer_size)
2440 		return -1;
2441 
2442 	/* buffer is not guaranteed to be aligned */
2443 	memcpy(&source, buffer, sizeof(struct entry_short));
2444 
2445 	entry.ctime.seconds = (git_time_t)ntohl(source.ctime.seconds);
2446 	entry.ctime.nanoseconds = ntohl(source.ctime.nanoseconds);
2447 	entry.mtime.seconds = (git_time_t)ntohl(source.mtime.seconds);
2448 	entry.mtime.nanoseconds = ntohl(source.mtime.nanoseconds);
2449 	entry.dev = ntohl(source.dev);
2450 	entry.ino = ntohl(source.ino);
2451 	entry.mode = ntohl(source.mode);
2452 	entry.uid = ntohl(source.uid);
2453 	entry.gid = ntohl(source.gid);
2454 	entry.file_size = ntohl(source.file_size);
2455 	git_oid_cpy(&entry.id, &source.oid);
2456 	entry.flags = ntohs(source.flags);
2457 
2458 	if (entry.flags & GIT_INDEX_ENTRY_EXTENDED) {
2459 		uint16_t flags_raw;
2460 		size_t flags_offset;
2461 
2462 		flags_offset = offsetof(struct entry_long, flags_extended);
2463 		memcpy(&flags_raw, (const char *) buffer + flags_offset,
2464 			sizeof(flags_raw));
2465 		flags_raw = ntohs(flags_raw);
2466 
2467 		memcpy(&entry.flags_extended, &flags_raw, sizeof(flags_raw));
2468 		path_ptr = (const char *) buffer + offsetof(struct entry_long, path);
2469 	} else
2470 		path_ptr = (const char *) buffer + offsetof(struct entry_short, path);
2471 
2472 	if (!compressed) {
2473 		path_length = entry.flags & GIT_INDEX_ENTRY_NAMEMASK;
2474 
2475 		/* if this is a very long string, we must find its
2476 		 * real length without overflowing */
2477 		if (path_length == 0xFFF) {
2478 			const char *path_end;
2479 
2480 			path_end = memchr(path_ptr, '\0', buffer_size);
2481 			if (path_end == NULL)
2482 				return -1;
2483 
2484 			path_length = path_end - path_ptr;
2485 		}
2486 
2487 		entry_size = index_entry_size(path_length, 0, entry.flags);
2488 		entry.path = (char *)path_ptr;
2489 	} else {
2490 		size_t varint_len, last_len, prefix_len, suffix_len, path_len;
2491 		uintmax_t strip_len;
2492 
2493 		strip_len = git_decode_varint((const unsigned char *)path_ptr, &varint_len);
2494 		last_len = strlen(last);
2495 
2496 		if (varint_len == 0 || last_len < strip_len)
2497 			return index_error_invalid("incorrect prefix length");
2498 
2499 		prefix_len = last_len - (size_t)strip_len;
2500 		suffix_len = strlen(path_ptr + varint_len);
2501 
2502 		GIT_ERROR_CHECK_ALLOC_ADD(&path_len, prefix_len, suffix_len);
2503 		GIT_ERROR_CHECK_ALLOC_ADD(&path_len, path_len, 1);
2504 
2505 		if (path_len > GIT_PATH_MAX)
2506 			return index_error_invalid("unreasonable path length");
2507 
2508 		tmp_path = git__malloc(path_len);
2509 		GIT_ERROR_CHECK_ALLOC(tmp_path);
2510 
2511 		memcpy(tmp_path, last, prefix_len);
2512 		memcpy(tmp_path + prefix_len, path_ptr + varint_len, suffix_len + 1);
2513 		entry_size = index_entry_size(suffix_len, varint_len, entry.flags);
2514 		entry.path = tmp_path;
2515 	}
2516 
2517 	if (entry_size == 0)
2518 		return -1;
2519 
2520 	if (INDEX_FOOTER_SIZE + entry_size > buffer_size)
2521 		return -1;
2522 
2523 	if (index_entry_dup(out, index, &entry) < 0) {
2524 		git__free(tmp_path);
2525 		return -1;
2526 	}
2527 
2528 	git__free(tmp_path);
2529 	*out_size = entry_size;
2530 	return 0;
2531 }
2532 
read_header(struct index_header * dest,const void * buffer)2533 static int read_header(struct index_header *dest, const void *buffer)
2534 {
2535 	const struct index_header *source = buffer;
2536 
2537 	dest->signature = ntohl(source->signature);
2538 	if (dest->signature != INDEX_HEADER_SIG)
2539 		return index_error_invalid("incorrect header signature");
2540 
2541 	dest->version = ntohl(source->version);
2542 	if (dest->version < INDEX_VERSION_NUMBER_LB ||
2543 		dest->version > INDEX_VERSION_NUMBER_UB)
2544 		return index_error_invalid("incorrect header version");
2545 
2546 	dest->entry_count = ntohl(source->entry_count);
2547 	return 0;
2548 }
2549 
read_extension(size_t * read_len,git_index * index,const char * buffer,size_t buffer_size)2550 static int read_extension(size_t *read_len, git_index *index, const char *buffer, size_t buffer_size)
2551 {
2552 	struct index_extension dest;
2553 	size_t total_size;
2554 
2555 	/* buffer is not guaranteed to be aligned */
2556 	memcpy(&dest, buffer, sizeof(struct index_extension));
2557 	dest.extension_size = ntohl(dest.extension_size);
2558 
2559 	total_size = dest.extension_size + sizeof(struct index_extension);
2560 
2561 	if (dest.extension_size > total_size ||
2562 		buffer_size < total_size ||
2563 		buffer_size - total_size < INDEX_FOOTER_SIZE) {
2564 		index_error_invalid("extension is truncated");
2565 		return -1;
2566 	}
2567 
2568 	/* optional extension */
2569 	if (dest.signature[0] >= 'A' && dest.signature[0] <= 'Z') {
2570 		/* tree cache */
2571 		if (memcmp(dest.signature, INDEX_EXT_TREECACHE_SIG, 4) == 0) {
2572 			if (git_tree_cache_read(&index->tree, buffer + 8, dest.extension_size, &index->tree_pool) < 0)
2573 				return -1;
2574 		} else if (memcmp(dest.signature, INDEX_EXT_UNMERGED_SIG, 4) == 0) {
2575 			if (read_reuc(index, buffer + 8, dest.extension_size) < 0)
2576 				return -1;
2577 		} else if (memcmp(dest.signature, INDEX_EXT_CONFLICT_NAME_SIG, 4) == 0) {
2578 			if (read_conflict_names(index, buffer + 8, dest.extension_size) < 0)
2579 				return -1;
2580 		}
2581 		/* else, unsupported extension. We cannot parse this, but we can skip
2582 		 * it by returning `total_size */
2583 	} else {
2584 		/* we cannot handle non-ignorable extensions;
2585 		 * in fact they aren't even defined in the standard */
2586 		git_error_set(GIT_ERROR_INDEX, "unsupported mandatory extension: '%.4s'", dest.signature);
2587 		return -1;
2588 	}
2589 
2590 	*read_len = total_size;
2591 
2592 	return 0;
2593 }
2594 
parse_index(git_index * index,const char * buffer,size_t buffer_size)2595 static int parse_index(git_index *index, const char *buffer, size_t buffer_size)
2596 {
2597 	int error = 0;
2598 	unsigned int i;
2599 	struct index_header header = { 0 };
2600 	git_oid checksum_calculated, checksum_expected;
2601 	const char *last = NULL;
2602 	const char *empty = "";
2603 
2604 #define seek_forward(_increase) { \
2605 	if (_increase >= buffer_size) { \
2606 		error = index_error_invalid("ran out of data while parsing"); \
2607 		goto done; } \
2608 	buffer += _increase; \
2609 	buffer_size -= _increase;\
2610 }
2611 
2612 	if (buffer_size < INDEX_HEADER_SIZE + INDEX_FOOTER_SIZE)
2613 		return index_error_invalid("insufficient buffer space");
2614 
2615 	/* Precalculate the SHA1 of the files's contents -- we'll match it to
2616 	 * the provided SHA1 in the footer */
2617 	git_hash_buf(&checksum_calculated, buffer, buffer_size - INDEX_FOOTER_SIZE);
2618 
2619 	/* Parse header */
2620 	if ((error = read_header(&header, buffer)) < 0)
2621 		return error;
2622 
2623 	index->version = header.version;
2624 	if (index->version >= INDEX_VERSION_NUMBER_COMP)
2625 		last = empty;
2626 
2627 	seek_forward(INDEX_HEADER_SIZE);
2628 
2629 	assert(!index->entries.length);
2630 
2631 	if ((error = index_map_resize(index->entries_map, header.entry_count, index->ignore_case)) < 0)
2632 		return error;
2633 
2634 	/* Parse all the entries */
2635 	for (i = 0; i < header.entry_count && buffer_size > INDEX_FOOTER_SIZE; ++i) {
2636 		git_index_entry *entry = NULL;
2637 		size_t entry_size;
2638 
2639 		if ((error = read_entry(&entry, &entry_size, index, buffer, buffer_size, last)) < 0) {
2640 			error = index_error_invalid("invalid entry");
2641 			goto done;
2642 		}
2643 
2644 		if ((error = git_vector_insert(&index->entries, entry)) < 0) {
2645 			index_entry_free(entry);
2646 			goto done;
2647 		}
2648 
2649 		if ((error = index_map_set(index->entries_map, entry, index->ignore_case)) < 0) {
2650 			index_entry_free(entry);
2651 			goto done;
2652 		}
2653 		error = 0;
2654 
2655 		if (index->version >= INDEX_VERSION_NUMBER_COMP)
2656 			last = entry->path;
2657 
2658 		seek_forward(entry_size);
2659 	}
2660 
2661 	if (i != header.entry_count) {
2662 		error = index_error_invalid("header entries changed while parsing");
2663 		goto done;
2664 	}
2665 
2666 	/* There's still space for some extensions! */
2667 	while (buffer_size > INDEX_FOOTER_SIZE) {
2668 		size_t extension_size;
2669 
2670 		if ((error = read_extension(&extension_size, index, buffer, buffer_size)) < 0) {
2671 			goto done;
2672 		}
2673 
2674 		seek_forward(extension_size);
2675 	}
2676 
2677 	if (buffer_size != INDEX_FOOTER_SIZE) {
2678 		error = index_error_invalid(
2679 			"buffer size does not match index footer size");
2680 		goto done;
2681 	}
2682 
2683 	/* 160-bit SHA-1 over the content of the index file before this checksum. */
2684 	git_oid_fromraw(&checksum_expected, (const unsigned char *)buffer);
2685 
2686 	if (git_oid__cmp(&checksum_calculated, &checksum_expected) != 0) {
2687 		error = index_error_invalid(
2688 			"calculated checksum does not match expected");
2689 		goto done;
2690 	}
2691 
2692 	git_oid_cpy(&index->checksum, &checksum_calculated);
2693 
2694 #undef seek_forward
2695 
2696 	/* Entries are stored case-sensitively on disk, so re-sort now if
2697 	 * in-memory index is supposed to be case-insensitive
2698 	 */
2699 	git_vector_set_sorted(&index->entries, !index->ignore_case);
2700 	git_vector_sort(&index->entries);
2701 
2702 	index->dirty = 0;
2703 done:
2704 	return error;
2705 }
2706 
is_index_extended(git_index * index)2707 static bool is_index_extended(git_index *index)
2708 {
2709 	size_t i, extended;
2710 	git_index_entry *entry;
2711 
2712 	extended = 0;
2713 
2714 	git_vector_foreach(&index->entries, i, entry) {
2715 		entry->flags &= ~GIT_INDEX_ENTRY_EXTENDED;
2716 		if (entry->flags_extended & GIT_INDEX_ENTRY_EXTENDED_FLAGS) {
2717 			extended++;
2718 			entry->flags |= GIT_INDEX_ENTRY_EXTENDED;
2719 		}
2720 	}
2721 
2722 	return (extended > 0);
2723 }
2724 
write_disk_entry(git_filebuf * file,git_index_entry * entry,const char * last)2725 static int write_disk_entry(git_filebuf *file, git_index_entry *entry, const char *last)
2726 {
2727 	void *mem = NULL;
2728 	struct entry_short ondisk;
2729 	size_t path_len, disk_size;
2730 	int varint_len = 0;
2731 	char *path;
2732 	const char *path_start = entry->path;
2733 	size_t same_len = 0;
2734 
2735 	path_len = ((struct entry_internal *)entry)->pathlen;
2736 
2737 	if (last) {
2738 		const char *last_c = last;
2739 
2740 		while (*path_start == *last_c) {
2741 			if (!*path_start || !*last_c)
2742 				break;
2743 			++path_start;
2744 			++last_c;
2745 			++same_len;
2746 		}
2747 		path_len -= same_len;
2748 		varint_len = git_encode_varint(NULL, 0, strlen(last) - same_len);
2749 	}
2750 
2751 	disk_size = index_entry_size(path_len, varint_len, entry->flags);
2752 
2753 	if (git_filebuf_reserve(file, &mem, disk_size) < 0)
2754 		return -1;
2755 
2756 	memset(mem, 0x0, disk_size);
2757 
2758 	/**
2759 	 * Yes, we have to truncate.
2760 	 *
2761 	 * The on-disk format for Index entries clearly defines
2762 	 * the time and size fields to be 4 bytes each -- so even if
2763 	 * we store these values with 8 bytes on-memory, they must
2764 	 * be truncated to 4 bytes before writing to disk.
2765 	 *
2766 	 * In 2038 I will be either too dead or too rich to care about this
2767 	 */
2768 	ondisk.ctime.seconds = htonl((uint32_t)entry->ctime.seconds);
2769 	ondisk.mtime.seconds = htonl((uint32_t)entry->mtime.seconds);
2770 	ondisk.ctime.nanoseconds = htonl(entry->ctime.nanoseconds);
2771 	ondisk.mtime.nanoseconds = htonl(entry->mtime.nanoseconds);
2772 	ondisk.dev = htonl(entry->dev);
2773 	ondisk.ino = htonl(entry->ino);
2774 	ondisk.mode = htonl(entry->mode);
2775 	ondisk.uid = htonl(entry->uid);
2776 	ondisk.gid = htonl(entry->gid);
2777 	ondisk.file_size = htonl((uint32_t)entry->file_size);
2778 
2779 	git_oid_cpy(&ondisk.oid, &entry->id);
2780 
2781 	ondisk.flags = htons(entry->flags);
2782 
2783 	if (entry->flags & GIT_INDEX_ENTRY_EXTENDED) {
2784 		const size_t path_offset = offsetof(struct entry_long, path);
2785 		struct entry_long ondisk_ext;
2786 		memcpy(&ondisk_ext, &ondisk, sizeof(struct entry_short));
2787 		ondisk_ext.flags_extended = htons(entry->flags_extended &
2788 			GIT_INDEX_ENTRY_EXTENDED_FLAGS);
2789 		memcpy(mem, &ondisk_ext, path_offset);
2790 		path = (char *)mem + path_offset;
2791 		disk_size -= path_offset;
2792 	} else {
2793 		const size_t path_offset = offsetof(struct entry_short, path);
2794 		memcpy(mem, &ondisk, path_offset);
2795 		path = (char *)mem + path_offset;
2796 		disk_size -= path_offset;
2797 	}
2798 
2799 	if (last) {
2800 		varint_len = git_encode_varint((unsigned char *) path,
2801 					  disk_size, strlen(last) - same_len);
2802 		assert(varint_len > 0);
2803 		path += varint_len;
2804 		disk_size -= varint_len;
2805 
2806 		/*
2807 		 * If using path compression, we are not allowed
2808 		 * to have additional trailing NULs.
2809 		 */
2810 		assert(disk_size == path_len + 1);
2811 	} else {
2812 		/*
2813 		 * If no path compression is used, we do have
2814 		 * NULs as padding. As such, simply assert that
2815 		 * we have enough space left to write the path.
2816 		 */
2817 		assert(disk_size > path_len);
2818 	}
2819 
2820 	memcpy(path, path_start, path_len + 1);
2821 
2822 	return 0;
2823 }
2824 
write_entries(git_index * index,git_filebuf * file)2825 static int write_entries(git_index *index, git_filebuf *file)
2826 {
2827 	int error = 0;
2828 	size_t i;
2829 	git_vector case_sorted, *entries;
2830 	git_index_entry *entry;
2831 	const char *last = NULL;
2832 
2833 	/* If index->entries is sorted case-insensitively, then we need
2834 	 * to re-sort it case-sensitively before writing */
2835 	if (index->ignore_case) {
2836 		git_vector_dup(&case_sorted, &index->entries, git_index_entry_cmp);
2837 		git_vector_sort(&case_sorted);
2838 		entries = &case_sorted;
2839 	} else {
2840 		entries = &index->entries;
2841 	}
2842 
2843 	if (index->version >= INDEX_VERSION_NUMBER_COMP)
2844 		last = "";
2845 
2846 	git_vector_foreach(entries, i, entry) {
2847 		if ((error = write_disk_entry(file, entry, last)) < 0)
2848 			break;
2849 		if (index->version >= INDEX_VERSION_NUMBER_COMP)
2850 			last = entry->path;
2851 	}
2852 
2853 	if (index->ignore_case)
2854 		git_vector_free(&case_sorted);
2855 
2856 	return error;
2857 }
2858 
write_extension(git_filebuf * file,struct index_extension * header,git_buf * data)2859 static int write_extension(git_filebuf *file, struct index_extension *header, git_buf *data)
2860 {
2861 	struct index_extension ondisk;
2862 
2863 	memset(&ondisk, 0x0, sizeof(struct index_extension));
2864 	memcpy(&ondisk, header, 4);
2865 	ondisk.extension_size = htonl(header->extension_size);
2866 
2867 	git_filebuf_write(file, &ondisk, sizeof(struct index_extension));
2868 	return git_filebuf_write(file, data->ptr, data->size);
2869 }
2870 
create_name_extension_data(git_buf * name_buf,git_index_name_entry * conflict_name)2871 static int create_name_extension_data(git_buf *name_buf, git_index_name_entry *conflict_name)
2872 {
2873 	int error = 0;
2874 
2875 	if (conflict_name->ancestor == NULL)
2876 		error = git_buf_put(name_buf, "\0", 1);
2877 	else
2878 		error = git_buf_put(name_buf, conflict_name->ancestor, strlen(conflict_name->ancestor) + 1);
2879 
2880 	if (error != 0)
2881 		goto on_error;
2882 
2883 	if (conflict_name->ours == NULL)
2884 		error = git_buf_put(name_buf, "\0", 1);
2885 	else
2886 		error = git_buf_put(name_buf, conflict_name->ours, strlen(conflict_name->ours) + 1);
2887 
2888 	if (error != 0)
2889 		goto on_error;
2890 
2891 	if (conflict_name->theirs == NULL)
2892 		error = git_buf_put(name_buf, "\0", 1);
2893 	else
2894 		error = git_buf_put(name_buf, conflict_name->theirs, strlen(conflict_name->theirs) + 1);
2895 
2896 on_error:
2897 	return error;
2898 }
2899 
write_name_extension(git_index * index,git_filebuf * file)2900 static int write_name_extension(git_index *index, git_filebuf *file)
2901 {
2902 	git_buf name_buf = GIT_BUF_INIT;
2903 	git_vector *out = &index->names;
2904 	git_index_name_entry *conflict_name;
2905 	struct index_extension extension;
2906 	size_t i;
2907 	int error = 0;
2908 
2909 	git_vector_foreach(out, i, conflict_name) {
2910 		if ((error = create_name_extension_data(&name_buf, conflict_name)) < 0)
2911 			goto done;
2912 	}
2913 
2914 	memset(&extension, 0x0, sizeof(struct index_extension));
2915 	memcpy(&extension.signature, INDEX_EXT_CONFLICT_NAME_SIG, 4);
2916 	extension.extension_size = (uint32_t)name_buf.size;
2917 
2918 	error = write_extension(file, &extension, &name_buf);
2919 
2920 	git_buf_dispose(&name_buf);
2921 
2922 done:
2923 	return error;
2924 }
2925 
create_reuc_extension_data(git_buf * reuc_buf,git_index_reuc_entry * reuc)2926 static int create_reuc_extension_data(git_buf *reuc_buf, git_index_reuc_entry *reuc)
2927 {
2928 	int i;
2929 	int error = 0;
2930 
2931 	if ((error = git_buf_put(reuc_buf, reuc->path, strlen(reuc->path) + 1)) < 0)
2932 		return error;
2933 
2934 	for (i = 0; i < 3; i++) {
2935 		if ((error = git_buf_printf(reuc_buf, "%o", reuc->mode[i])) < 0 ||
2936 			(error = git_buf_put(reuc_buf, "\0", 1)) < 0)
2937 			return error;
2938 	}
2939 
2940 	for (i = 0; i < 3; i++) {
2941 		if (reuc->mode[i] && (error = git_buf_put(reuc_buf, (char *)&reuc->oid[i].id, GIT_OID_RAWSZ)) < 0)
2942 			return error;
2943 	}
2944 
2945 	return 0;
2946 }
2947 
write_reuc_extension(git_index * index,git_filebuf * file)2948 static int write_reuc_extension(git_index *index, git_filebuf *file)
2949 {
2950 	git_buf reuc_buf = GIT_BUF_INIT;
2951 	git_vector *out = &index->reuc;
2952 	git_index_reuc_entry *reuc;
2953 	struct index_extension extension;
2954 	size_t i;
2955 	int error = 0;
2956 
2957 	git_vector_foreach(out, i, reuc) {
2958 		if ((error = create_reuc_extension_data(&reuc_buf, reuc)) < 0)
2959 			goto done;
2960 	}
2961 
2962 	memset(&extension, 0x0, sizeof(struct index_extension));
2963 	memcpy(&extension.signature, INDEX_EXT_UNMERGED_SIG, 4);
2964 	extension.extension_size = (uint32_t)reuc_buf.size;
2965 
2966 	error = write_extension(file, &extension, &reuc_buf);
2967 
2968 	git_buf_dispose(&reuc_buf);
2969 
2970 done:
2971 	return error;
2972 }
2973 
write_tree_extension(git_index * index,git_filebuf * file)2974 static int write_tree_extension(git_index *index, git_filebuf *file)
2975 {
2976 	struct index_extension extension;
2977 	git_buf buf = GIT_BUF_INIT;
2978 	int error;
2979 
2980 	if (index->tree == NULL)
2981 		return 0;
2982 
2983 	if ((error = git_tree_cache_write(&buf, index->tree)) < 0)
2984 		return error;
2985 
2986 	memset(&extension, 0x0, sizeof(struct index_extension));
2987 	memcpy(&extension.signature, INDEX_EXT_TREECACHE_SIG, 4);
2988 	extension.extension_size = (uint32_t)buf.size;
2989 
2990 	error = write_extension(file, &extension, &buf);
2991 
2992 	git_buf_dispose(&buf);
2993 
2994 	return error;
2995 }
2996 
clear_uptodate(git_index * index)2997 static void clear_uptodate(git_index *index)
2998 {
2999 	git_index_entry *entry;
3000 	size_t i;
3001 
3002 	git_vector_foreach(&index->entries, i, entry)
3003 		entry->flags_extended &= ~GIT_INDEX_ENTRY_UPTODATE;
3004 }
3005 
write_index(git_oid * checksum,git_index * index,git_filebuf * file)3006 static int write_index(git_oid *checksum, git_index *index, git_filebuf *file)
3007 {
3008 	git_oid hash_final;
3009 	struct index_header header;
3010 	bool is_extended;
3011 	uint32_t index_version_number;
3012 
3013 	assert(index && file);
3014 
3015 	if (index->version <= INDEX_VERSION_NUMBER_EXT)  {
3016 		is_extended = is_index_extended(index);
3017 		index_version_number = is_extended ? INDEX_VERSION_NUMBER_EXT : INDEX_VERSION_NUMBER_LB;
3018 	} else {
3019 		index_version_number = index->version;
3020 	}
3021 
3022 	header.signature = htonl(INDEX_HEADER_SIG);
3023 	header.version = htonl(index_version_number);
3024 	header.entry_count = htonl((uint32_t)index->entries.length);
3025 
3026 	if (git_filebuf_write(file, &header, sizeof(struct index_header)) < 0)
3027 		return -1;
3028 
3029 	if (write_entries(index, file) < 0)
3030 		return -1;
3031 
3032 	/* write the tree cache extension */
3033 	if (index->tree != NULL && write_tree_extension(index, file) < 0)
3034 		return -1;
3035 
3036 	/* write the rename conflict extension */
3037 	if (index->names.length > 0 && write_name_extension(index, file) < 0)
3038 		return -1;
3039 
3040 	/* write the reuc extension */
3041 	if (index->reuc.length > 0 && write_reuc_extension(index, file) < 0)
3042 		return -1;
3043 
3044 	/* get out the hash for all the contents we've appended to the file */
3045 	git_filebuf_hash(&hash_final, file);
3046 	git_oid_cpy(checksum, &hash_final);
3047 
3048 	/* write it at the end of the file */
3049 	if (git_filebuf_write(file, hash_final.id, GIT_OID_RAWSZ) < 0)
3050 		return -1;
3051 
3052 	/* file entries are no longer up to date */
3053 	clear_uptodate(index);
3054 
3055 	return 0;
3056 }
3057 
git_index_entry_stage(const git_index_entry * entry)3058 int git_index_entry_stage(const git_index_entry *entry)
3059 {
3060 	return GIT_INDEX_ENTRY_STAGE(entry);
3061 }
3062 
git_index_entry_is_conflict(const git_index_entry * entry)3063 int git_index_entry_is_conflict(const git_index_entry *entry)
3064 {
3065 	return (GIT_INDEX_ENTRY_STAGE(entry) > 0);
3066 }
3067 
3068 typedef struct read_tree_data {
3069 	git_index *index;
3070 	git_vector *old_entries;
3071 	git_vector *new_entries;
3072 	git_vector_cmp entry_cmp;
3073 	git_tree_cache *tree;
3074 } read_tree_data;
3075 
read_tree_cb(const char * root,const git_tree_entry * tentry,void * payload)3076 static int read_tree_cb(
3077 	const char *root, const git_tree_entry *tentry, void *payload)
3078 {
3079 	read_tree_data *data = payload;
3080 	git_index_entry *entry = NULL, *old_entry;
3081 	git_buf path = GIT_BUF_INIT;
3082 	size_t pos;
3083 
3084 	if (git_tree_entry__is_tree(tentry))
3085 		return 0;
3086 
3087 	if (git_buf_joinpath(&path, root, tentry->filename) < 0)
3088 		return -1;
3089 
3090 	if (index_entry_create(&entry, INDEX_OWNER(data->index), path.ptr, NULL, false) < 0)
3091 		return -1;
3092 
3093 	entry->mode = tentry->attr;
3094 	git_oid_cpy(&entry->id, git_tree_entry_id(tentry));
3095 
3096 	/* look for corresponding old entry and copy data to new entry */
3097 	if (data->old_entries != NULL &&
3098 		!index_find_in_entries(
3099 			&pos, data->old_entries, data->entry_cmp, path.ptr, 0, 0) &&
3100 		(old_entry = git_vector_get(data->old_entries, pos)) != NULL &&
3101 		entry->mode == old_entry->mode &&
3102 		git_oid_equal(&entry->id, &old_entry->id))
3103 	{
3104 		index_entry_cpy(entry, old_entry);
3105 		entry->flags_extended = 0;
3106 	}
3107 
3108 	index_entry_adjust_namemask(entry, path.size);
3109 	git_buf_dispose(&path);
3110 
3111 	if (git_vector_insert(data->new_entries, entry) < 0) {
3112 		index_entry_free(entry);
3113 		return -1;
3114 	}
3115 
3116 	return 0;
3117 }
3118 
git_index_read_tree(git_index * index,const git_tree * tree)3119 int git_index_read_tree(git_index *index, const git_tree *tree)
3120 {
3121 	int error = 0;
3122 	git_vector entries = GIT_VECTOR_INIT;
3123 	git_idxmap *entries_map;
3124 	read_tree_data data;
3125 	size_t i;
3126 	git_index_entry *e;
3127 
3128 	if (git_idxmap_new(&entries_map) < 0)
3129 		return -1;
3130 
3131 	git_vector_set_cmp(&entries, index->entries._cmp); /* match sort */
3132 
3133 	data.index = index;
3134 	data.old_entries = &index->entries;
3135 	data.new_entries = &entries;
3136 	data.entry_cmp   = index->entries_search;
3137 
3138 	index->tree = NULL;
3139 	git_pool_clear(&index->tree_pool);
3140 
3141 	git_vector_sort(&index->entries);
3142 
3143 	if ((error = git_tree_walk(tree, GIT_TREEWALK_POST, read_tree_cb, &data)) < 0)
3144 		goto cleanup;
3145 
3146 	if ((error = index_map_resize(entries_map, entries.length, index->ignore_case)) < 0)
3147 		goto cleanup;
3148 
3149 	git_vector_foreach(&entries, i, e) {
3150 		if ((error = index_map_set(entries_map, e, index->ignore_case)) < 0) {
3151 			git_error_set(GIT_ERROR_INDEX, "failed to insert entry into map");
3152 			return error;
3153 		}
3154 	}
3155 
3156 	error = 0;
3157 
3158 	git_vector_sort(&entries);
3159 
3160 	if ((error = git_index_clear(index)) < 0) {
3161 		/* well, this isn't good */;
3162 	} else {
3163 		git_vector_swap(&entries, &index->entries);
3164 		entries_map = git__swap(index->entries_map, entries_map);
3165 	}
3166 
3167 	index->dirty = 1;
3168 
3169 cleanup:
3170 	git_vector_free(&entries);
3171 	git_idxmap_free(entries_map);
3172 	if (error < 0)
3173 		return error;
3174 
3175 	error = git_tree_cache_read_tree(&index->tree, tree, &index->tree_pool);
3176 
3177 	return error;
3178 }
3179 
git_index_read_iterator(git_index * index,git_iterator * new_iterator,size_t new_length_hint)3180 static int git_index_read_iterator(
3181 	git_index *index,
3182 	git_iterator *new_iterator,
3183 	size_t new_length_hint)
3184 {
3185 	git_vector new_entries = GIT_VECTOR_INIT,
3186 		remove_entries = GIT_VECTOR_INIT;
3187 	git_idxmap *new_entries_map = NULL;
3188 	git_iterator *index_iterator = NULL;
3189 	git_iterator_options opts = GIT_ITERATOR_OPTIONS_INIT;
3190 	const git_index_entry *old_entry, *new_entry;
3191 	git_index_entry *entry;
3192 	size_t i;
3193 	int error;
3194 
3195 	assert((new_iterator->flags & GIT_ITERATOR_DONT_IGNORE_CASE));
3196 
3197 	if ((error = git_vector_init(&new_entries, new_length_hint, index->entries._cmp)) < 0 ||
3198 	    (error = git_vector_init(&remove_entries, index->entries.length, NULL)) < 0 ||
3199 	    (error = git_idxmap_new(&new_entries_map)) < 0)
3200 		goto done;
3201 
3202 	if (new_length_hint && (error = index_map_resize(new_entries_map, new_length_hint,
3203 							 index->ignore_case)) < 0)
3204 		goto done;
3205 
3206 	opts.flags = GIT_ITERATOR_DONT_IGNORE_CASE |
3207 		GIT_ITERATOR_INCLUDE_CONFLICTS;
3208 
3209 	if ((error = git_iterator_for_index(&index_iterator,
3210 			git_index_owner(index), index, &opts)) < 0 ||
3211 		((error = git_iterator_current(&old_entry, index_iterator)) < 0 &&
3212 			error != GIT_ITEROVER) ||
3213 		((error = git_iterator_current(&new_entry, new_iterator)) < 0 &&
3214 			error != GIT_ITEROVER))
3215 		goto done;
3216 
3217 	while (true) {
3218 		git_index_entry
3219 			*dup_entry = NULL,
3220 			*add_entry = NULL,
3221 			*remove_entry = NULL;
3222 		int diff;
3223 
3224 		error = 0;
3225 
3226 		if (old_entry && new_entry)
3227 			diff = git_index_entry_cmp(old_entry, new_entry);
3228 		else if (!old_entry && new_entry)
3229 			diff = 1;
3230 		else if (old_entry && !new_entry)
3231 			diff = -1;
3232 		else
3233 			break;
3234 
3235 		if (diff < 0) {
3236 			remove_entry = (git_index_entry *)old_entry;
3237 		} else if (diff > 0) {
3238 			dup_entry = (git_index_entry *)new_entry;
3239 		} else {
3240 			/* Path and stage are equal, if the OID is equal, keep it to
3241 			 * keep the stat cache data.
3242 			 */
3243 			if (git_oid_equal(&old_entry->id, &new_entry->id) &&
3244 				old_entry->mode == new_entry->mode) {
3245 				add_entry = (git_index_entry *)old_entry;
3246 			} else {
3247 				dup_entry = (git_index_entry *)new_entry;
3248 				remove_entry = (git_index_entry *)old_entry;
3249 			}
3250 		}
3251 
3252 		if (dup_entry) {
3253 			if ((error = index_entry_dup_nocache(&add_entry, index, dup_entry)) < 0)
3254 				goto done;
3255 
3256 			index_entry_adjust_namemask(add_entry,
3257 				((struct entry_internal *)add_entry)->pathlen);
3258 		}
3259 
3260 		/* invalidate this path in the tree cache if this is new (to
3261 		 * invalidate the parent trees)
3262 		 */
3263 		if (dup_entry && !remove_entry && index->tree)
3264 			git_tree_cache_invalidate_path(index->tree, dup_entry->path);
3265 
3266 		if (add_entry) {
3267 			if ((error = git_vector_insert(&new_entries, add_entry)) == 0)
3268 				error = index_map_set(new_entries_map, add_entry,
3269 						      index->ignore_case);
3270 		}
3271 
3272 		if (remove_entry && error >= 0)
3273 			error = git_vector_insert(&remove_entries, remove_entry);
3274 
3275 		if (error < 0) {
3276 			git_error_set(GIT_ERROR_INDEX, "failed to insert entry");
3277 			goto done;
3278 		}
3279 
3280 		if (diff <= 0) {
3281 			if ((error = git_iterator_advance(&old_entry, index_iterator)) < 0 &&
3282 				error != GIT_ITEROVER)
3283 				goto done;
3284 		}
3285 
3286 		if (diff >= 0) {
3287 			if ((error = git_iterator_advance(&new_entry, new_iterator)) < 0 &&
3288 				error != GIT_ITEROVER)
3289 				goto done;
3290 		}
3291 	}
3292 
3293 	if ((error = git_index_name_clear(index)) < 0 ||
3294 		(error = git_index_reuc_clear(index)) < 0)
3295 	    goto done;
3296 
3297 	git_vector_swap(&new_entries, &index->entries);
3298 	new_entries_map = git__swap(index->entries_map, new_entries_map);
3299 
3300 	git_vector_foreach(&remove_entries, i, entry) {
3301 		if (index->tree)
3302 			git_tree_cache_invalidate_path(index->tree, entry->path);
3303 
3304 		index_entry_free(entry);
3305 	}
3306 
3307 	clear_uptodate(index);
3308 
3309 	index->dirty = 1;
3310 	error = 0;
3311 
3312 done:
3313 	git_idxmap_free(new_entries_map);
3314 	git_vector_free(&new_entries);
3315 	git_vector_free(&remove_entries);
3316 	git_iterator_free(index_iterator);
3317 	return error;
3318 }
3319 
git_index_read_index(git_index * index,const git_index * new_index)3320 int git_index_read_index(
3321 	git_index *index,
3322 	const git_index *new_index)
3323 {
3324 	git_iterator *new_iterator = NULL;
3325 	git_iterator_options opts = GIT_ITERATOR_OPTIONS_INIT;
3326 	int error;
3327 
3328 	opts.flags = GIT_ITERATOR_DONT_IGNORE_CASE |
3329 		GIT_ITERATOR_INCLUDE_CONFLICTS;
3330 
3331 	if ((error = git_iterator_for_index(&new_iterator,
3332 		git_index_owner(new_index), (git_index *)new_index, &opts)) < 0 ||
3333 		(error = git_index_read_iterator(index, new_iterator,
3334 		new_index->entries.length)) < 0)
3335 		goto done;
3336 
3337 done:
3338 	git_iterator_free(new_iterator);
3339 	return error;
3340 }
3341 
git_index_owner(const git_index * index)3342 git_repository *git_index_owner(const git_index *index)
3343 {
3344 	return INDEX_OWNER(index);
3345 }
3346 
3347 enum {
3348 	INDEX_ACTION_NONE = 0,
3349 	INDEX_ACTION_UPDATE = 1,
3350 	INDEX_ACTION_REMOVE = 2,
3351 	INDEX_ACTION_ADDALL = 3,
3352 };
3353 
git_index_add_all(git_index * index,const git_strarray * paths,unsigned int flags,git_index_matched_path_cb cb,void * payload)3354 int git_index_add_all(
3355 	git_index *index,
3356 	const git_strarray *paths,
3357 	unsigned int flags,
3358 	git_index_matched_path_cb cb,
3359 	void *payload)
3360 {
3361 	int error;
3362 	git_repository *repo;
3363 	git_iterator *wditer = NULL;
3364 	git_pathspec ps;
3365 	bool no_fnmatch = (flags & GIT_INDEX_ADD_DISABLE_PATHSPEC_MATCH) != 0;
3366 
3367 	assert(index);
3368 
3369 	repo = INDEX_OWNER(index);
3370 	if ((error = git_repository__ensure_not_bare(repo, "index add all")) < 0)
3371 		return error;
3372 
3373 	if ((error = git_pathspec__init(&ps, paths)) < 0)
3374 		return error;
3375 
3376 	/* optionally check that pathspec doesn't mention any ignored files */
3377 	if ((flags & GIT_INDEX_ADD_CHECK_PATHSPEC) != 0 &&
3378 		(flags & GIT_INDEX_ADD_FORCE) == 0 &&
3379 		(error = git_ignore__check_pathspec_for_exact_ignores(
3380 			repo, &ps.pathspec, no_fnmatch)) < 0)
3381 		goto cleanup;
3382 
3383 	error = index_apply_to_wd_diff(index, INDEX_ACTION_ADDALL, paths, flags, cb, payload);
3384 
3385 	if (error)
3386 		git_error_set_after_callback(error);
3387 
3388 cleanup:
3389 	git_iterator_free(wditer);
3390 	git_pathspec__clear(&ps);
3391 
3392 	return error;
3393 }
3394 
3395 struct foreach_diff_data {
3396 	git_index *index;
3397 	const git_pathspec *pathspec;
3398 	unsigned int flags;
3399 	git_index_matched_path_cb cb;
3400 	void *payload;
3401 };
3402 
apply_each_file(const git_diff_delta * delta,float progress,void * payload)3403 static int apply_each_file(const git_diff_delta *delta, float progress, void *payload)
3404 {
3405 	struct foreach_diff_data *data = payload;
3406 	const char *match, *path;
3407 	int error = 0;
3408 
3409 	GIT_UNUSED(progress);
3410 
3411 	path = delta->old_file.path;
3412 
3413 	/* We only want those which match the pathspecs */
3414 	if (!git_pathspec__match(
3415 		    &data->pathspec->pathspec, path, false, (bool)data->index->ignore_case,
3416 		    &match, NULL))
3417 		return 0;
3418 
3419 	if (data->cb)
3420 		error = data->cb(path, match, data->payload);
3421 
3422 	if (error > 0) /* skip this entry */
3423 		return 0;
3424 	if (error < 0) /* actual error */
3425 		return error;
3426 
3427 	/* If the workdir item does not exist, remove it from the index. */
3428 	if ((delta->new_file.flags & GIT_DIFF_FLAG_EXISTS) == 0)
3429 		error = git_index_remove_bypath(data->index, path);
3430 	else
3431 		error = git_index_add_bypath(data->index, delta->new_file.path);
3432 
3433 	return error;
3434 }
3435 
index_apply_to_wd_diff(git_index * index,int action,const git_strarray * paths,unsigned int flags,git_index_matched_path_cb cb,void * payload)3436 static int index_apply_to_wd_diff(git_index *index, int action, const git_strarray *paths,
3437 				  unsigned int flags,
3438 				  git_index_matched_path_cb cb, void *payload)
3439 {
3440 	int error;
3441 	git_diff *diff;
3442 	git_pathspec ps;
3443 	git_repository *repo;
3444 	git_diff_options opts = GIT_DIFF_OPTIONS_INIT;
3445 	struct foreach_diff_data data = {
3446 		index,
3447 		NULL,
3448 		flags,
3449 		cb,
3450 		payload,
3451 	};
3452 
3453 	assert(index);
3454 	assert(action == INDEX_ACTION_UPDATE || action == INDEX_ACTION_ADDALL);
3455 
3456 	repo = INDEX_OWNER(index);
3457 
3458 	if (!repo) {
3459 		return create_index_error(-1,
3460 			"cannot run update; the index is not backed up by a repository.");
3461 	}
3462 
3463 	/*
3464 	 * We do the matching ourselves intead of passing the list to
3465 	 * diff because we want to tell the callback which one
3466 	 * matched, which we do not know if we ask diff to filter for us.
3467 	 */
3468 	if ((error = git_pathspec__init(&ps, paths)) < 0)
3469 		return error;
3470 
3471 	opts.flags = GIT_DIFF_INCLUDE_TYPECHANGE;
3472 	if (action == INDEX_ACTION_ADDALL) {
3473 		opts.flags |= GIT_DIFF_INCLUDE_UNTRACKED |
3474 			GIT_DIFF_RECURSE_UNTRACKED_DIRS;
3475 
3476 		if (flags == GIT_INDEX_ADD_FORCE)
3477 			opts.flags |= GIT_DIFF_INCLUDE_IGNORED;
3478 	}
3479 
3480 	if ((error = git_diff_index_to_workdir(&diff, repo, index, &opts)) < 0)
3481 		goto cleanup;
3482 
3483 	data.pathspec = &ps;
3484 	error = git_diff_foreach(diff, apply_each_file, NULL, NULL, NULL, &data);
3485 	git_diff_free(diff);
3486 
3487 	if (error) /* make sure error is set if callback stopped iteration */
3488 		git_error_set_after_callback(error);
3489 
3490 cleanup:
3491 	git_pathspec__clear(&ps);
3492 	return error;
3493 }
3494 
index_apply_to_all(git_index * index,int action,const git_strarray * paths,git_index_matched_path_cb cb,void * payload)3495 static int index_apply_to_all(
3496 	git_index *index,
3497 	int action,
3498 	const git_strarray *paths,
3499 	git_index_matched_path_cb cb,
3500 	void *payload)
3501 {
3502 	int error = 0;
3503 	size_t i;
3504 	git_pathspec ps;
3505 	const char *match;
3506 	git_buf path = GIT_BUF_INIT;
3507 
3508 	assert(index);
3509 
3510 	if ((error = git_pathspec__init(&ps, paths)) < 0)
3511 		return error;
3512 
3513 	git_vector_sort(&index->entries);
3514 
3515 	for (i = 0; !error && i < index->entries.length; ++i) {
3516 		git_index_entry *entry = git_vector_get(&index->entries, i);
3517 
3518 		/* check if path actually matches */
3519 		if (!git_pathspec__match(
3520 				&ps.pathspec, entry->path, false, (bool)index->ignore_case,
3521 				&match, NULL))
3522 			continue;
3523 
3524 		/* issue notification callback if requested */
3525 		if (cb && (error = cb(entry->path, match, payload)) != 0) {
3526 			if (error > 0) { /* return > 0 means skip this one */
3527 				error = 0;
3528 				continue;
3529 			}
3530 			if (error < 0)   /* return < 0 means abort */
3531 				break;
3532 		}
3533 
3534 		/* index manipulation may alter entry, so don't depend on it */
3535 		if ((error = git_buf_sets(&path, entry->path)) < 0)
3536 			break;
3537 
3538 		switch (action) {
3539 		case INDEX_ACTION_NONE:
3540 			break;
3541 		case INDEX_ACTION_UPDATE:
3542 			error = git_index_add_bypath(index, path.ptr);
3543 
3544 			if (error == GIT_ENOTFOUND) {
3545 				git_error_clear();
3546 
3547 				error = git_index_remove_bypath(index, path.ptr);
3548 
3549 				if (!error) /* back up foreach if we removed this */
3550 					i--;
3551 			}
3552 			break;
3553 		case INDEX_ACTION_REMOVE:
3554 			if (!(error = git_index_remove_bypath(index, path.ptr)))
3555 				i--; /* back up foreach if we removed this */
3556 			break;
3557 		default:
3558 			git_error_set(GIT_ERROR_INVALID, "unknown index action %d", action);
3559 			error = -1;
3560 			break;
3561 		}
3562 	}
3563 
3564 	git_buf_dispose(&path);
3565 	git_pathspec__clear(&ps);
3566 
3567 	return error;
3568 }
3569 
git_index_remove_all(git_index * index,const git_strarray * pathspec,git_index_matched_path_cb cb,void * payload)3570 int git_index_remove_all(
3571 	git_index *index,
3572 	const git_strarray *pathspec,
3573 	git_index_matched_path_cb cb,
3574 	void *payload)
3575 {
3576 	int error = index_apply_to_all(
3577 		index, INDEX_ACTION_REMOVE, pathspec, cb, payload);
3578 
3579 	if (error) /* make sure error is set if callback stopped iteration */
3580 		git_error_set_after_callback(error);
3581 
3582 	return error;
3583 }
3584 
git_index_update_all(git_index * index,const git_strarray * pathspec,git_index_matched_path_cb cb,void * payload)3585 int git_index_update_all(
3586 	git_index *index,
3587 	const git_strarray *pathspec,
3588 	git_index_matched_path_cb cb,
3589 	void *payload)
3590 {
3591 	int error = index_apply_to_wd_diff(index, INDEX_ACTION_UPDATE, pathspec, 0, cb, payload);
3592 	if (error) /* make sure error is set if callback stopped iteration */
3593 		git_error_set_after_callback(error);
3594 
3595 	return error;
3596 }
3597 
git_index_snapshot_new(git_vector * snap,git_index * index)3598 int git_index_snapshot_new(git_vector *snap, git_index *index)
3599 {
3600 	int error;
3601 
3602 	GIT_REFCOUNT_INC(index);
3603 
3604 	git_atomic_inc(&index->readers);
3605 	git_vector_sort(&index->entries);
3606 
3607 	error = git_vector_dup(snap, &index->entries, index->entries._cmp);
3608 
3609 	if (error < 0)
3610 		git_index_snapshot_release(snap, index);
3611 
3612 	return error;
3613 }
3614 
git_index_snapshot_release(git_vector * snap,git_index * index)3615 void git_index_snapshot_release(git_vector *snap, git_index *index)
3616 {
3617 	git_vector_free(snap);
3618 
3619 	git_atomic_dec(&index->readers);
3620 
3621 	git_index_free(index);
3622 }
3623 
git_index_snapshot_find(size_t * out,git_vector * entries,git_vector_cmp entry_srch,const char * path,size_t path_len,int stage)3624 int git_index_snapshot_find(
3625 	size_t *out, git_vector *entries, git_vector_cmp entry_srch,
3626 	const char *path, size_t path_len, int stage)
3627 {
3628 	return index_find_in_entries(out, entries, entry_srch, path, path_len, stage);
3629 }
3630 
git_indexwriter_init(git_indexwriter * writer,git_index * index)3631 int git_indexwriter_init(
3632 	git_indexwriter *writer,
3633 	git_index *index)
3634 {
3635 	int error;
3636 
3637 	GIT_REFCOUNT_INC(index);
3638 
3639 	writer->index = index;
3640 
3641 	if (!index->index_file_path)
3642 		return create_index_error(-1,
3643 			"failed to write index: The index is in-memory only");
3644 
3645 	if ((error = git_filebuf_open(
3646 		&writer->file, index->index_file_path, GIT_FILEBUF_HASH_CONTENTS, GIT_INDEX_FILE_MODE)) < 0) {
3647 
3648 		if (error == GIT_ELOCKED)
3649 			git_error_set(GIT_ERROR_INDEX, "the index is locked; this might be due to a concurrent or crashed process");
3650 
3651 		return error;
3652 	}
3653 
3654 	writer->should_write = 1;
3655 
3656 	return 0;
3657 }
3658 
git_indexwriter_init_for_operation(git_indexwriter * writer,git_repository * repo,unsigned int * checkout_strategy)3659 int git_indexwriter_init_for_operation(
3660 	git_indexwriter *writer,
3661 	git_repository *repo,
3662 	unsigned int *checkout_strategy)
3663 {
3664 	git_index *index;
3665 	int error;
3666 
3667 	if ((error = git_repository_index__weakptr(&index, repo)) < 0 ||
3668 		(error = git_indexwriter_init(writer, index)) < 0)
3669 		return error;
3670 
3671 	writer->should_write = (*checkout_strategy & GIT_CHECKOUT_DONT_WRITE_INDEX) == 0;
3672 	*checkout_strategy |= GIT_CHECKOUT_DONT_WRITE_INDEX;
3673 
3674 	return 0;
3675 }
3676 
git_indexwriter_commit(git_indexwriter * writer)3677 int git_indexwriter_commit(git_indexwriter *writer)
3678 {
3679 	int error;
3680 	git_oid checksum = {{ 0 }};
3681 
3682 	if (!writer->should_write)
3683 		return 0;
3684 
3685 	git_vector_sort(&writer->index->entries);
3686 	git_vector_sort(&writer->index->reuc);
3687 
3688 	if ((error = write_index(&checksum, writer->index, &writer->file)) < 0) {
3689 		git_indexwriter_cleanup(writer);
3690 		return error;
3691 	}
3692 
3693 	if ((error = git_filebuf_commit(&writer->file)) < 0)
3694 		return error;
3695 
3696 	if ((error = git_futils_filestamp_check(
3697 		&writer->index->stamp, writer->index->index_file_path)) < 0) {
3698 		git_error_set(GIT_ERROR_OS, "could not read index timestamp");
3699 		return -1;
3700 	}
3701 
3702 	writer->index->dirty = 0;
3703 	writer->index->on_disk = 1;
3704 	git_oid_cpy(&writer->index->checksum, &checksum);
3705 
3706 	git_index_free(writer->index);
3707 	writer->index = NULL;
3708 
3709 	return 0;
3710 }
3711 
git_indexwriter_cleanup(git_indexwriter * writer)3712 void git_indexwriter_cleanup(git_indexwriter *writer)
3713 {
3714 	git_filebuf_cleanup(&writer->file);
3715 
3716 	git_index_free(writer->index);
3717 	writer->index = NULL;
3718 }
3719 
3720 /* Deprecated functions */
3721 
3722 #ifndef GIT_DEPRECATE_HARD
git_index_add_frombuffer(git_index * index,const git_index_entry * source_entry,const void * buffer,size_t len)3723 int git_index_add_frombuffer(
3724     git_index *index, const git_index_entry *source_entry,
3725     const void *buffer, size_t len)
3726 {
3727 	return git_index_add_from_buffer(index, source_entry, buffer, len);
3728 }
3729 #endif
3730