Second batch post 2.19
[git/git.git] / read-cache.c
CommitLineData
8bc9a0c7
LT
1/*
2 * GIT - The information manager from hell
3 *
4 * Copyright (C) Linus Torvalds, 2005
5 */
4aab5b46 6#define NO_THE_INDEX_COMPATIBILITY_MACROS
e83c5163 7#include "cache.h"
b2141fc1 8#include "config.h"
cffbfad5
EN
9#include "diff.h"
10#include "diffcore.h"
f6ecc62d 11#include "tempfile.h"
697cc8ef 12#include "lockfile.h"
bad68ec9 13#include "cache-tree.h"
f35a6d3b 14#include "refs.h"
d616813d 15#include "dir.h"
cbd53a21 16#include "object-store.h"
041aee31
JH
17#include "tree.h"
18#include "commit.h"
39425819 19#include "blob.h"
cfc5789a 20#include "resolve-undo.h"
6c9cd161
JH
21#include "strbuf.h"
22#include "varint.h"
5fc2fc8f 23#include "split-index.h"
a42643aa 24#include "utf8.h"
883e248b 25#include "fsmonitor.h"
bad68ec9 26
b60e188c
TG
27/* Mask for the name length in ce_flags in the on-disk index */
28
29#define CE_NAMEMASK (0x0fff)
30
bad68ec9
JH
31/* Index extensions.
32 *
33 * The first letter should be 'A'..'Z' for extensions that are not
34 * necessary for a correct operation (i.e. optimization data).
35 * When new extensions are added that _needs_ to be understood in
36 * order to correctly interpret the index file, pick character that
37 * is outside the range, to cause the reader to abort.
38 */
39
40#define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) )
41#define CACHE_EXT_TREE 0x54524545 /* "TREE" */
b659b49b 42#define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */
5fc2fc8f 43#define CACHE_EXT_LINK 0x6c696e6b /* "link" */
83c094ad 44#define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */
883e248b 45#define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */
5fc2fc8f
NTND
46
47/* changes that can be kept in $GIT_DIR/index (basically all extensions) */
e0cf0d7d 48#define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \
c18b80a0 49 CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \
883e248b 50 SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED)
e83c5163 51
8e72d675
JM
52
53/*
54 * This is an estimate of the pathname length in the index. We use
55 * this for V4 index files to guess the un-deltafied size of the index
56 * in memory because of pathname deltafication. This is not required
57 * for V2/V3 index formats because their pathnames are not compressed.
58 * If the initial amount of memory set aside is not sufficient, the
59 * mem pool will allocate extra memory.
60 */
61#define CACHE_ENTRY_PATH_LENGTH 80
62
63static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len)
64{
65 struct cache_entry *ce;
66 ce = mem_pool_alloc(mem_pool, cache_entry_size(len));
67 ce->mem_pool_allocated = 1;
68 return ce;
69}
70
71static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len)
72{
73 struct cache_entry * ce;
74 ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len));
75 ce->mem_pool_allocated = 1;
76 return ce;
77}
78
79static struct mem_pool *find_mem_pool(struct index_state *istate)
80{
81 struct mem_pool **pool_ptr;
82
83 if (istate->split_index && istate->split_index->base)
84 pool_ptr = &istate->split_index->base->ce_mem_pool;
85 else
86 pool_ptr = &istate->ce_mem_pool;
87
88 if (!*pool_ptr)
89 mem_pool_init(pool_ptr, 0);
90
91 return *pool_ptr;
92}
93
228e94f9 94struct index_state the_index;
626f35c8 95static const char *alternate_index_output;
8fd2cb40 96
9cb76b8c
JH
97static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce)
98{
99 istate->cache[nr] = ce;
96872bc2 100 add_name_hash(istate, ce);
9cb76b8c
JH
101}
102
cf558704
LT
103static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce)
104{
105 struct cache_entry *old = istate->cache[nr];
106
078a58e8 107 replace_index_entry_in_base(istate, old, ce);
2092678c 108 remove_name_hash(istate, old);
a849735b 109 discard_cache_entry(old);
0e267b7a 110 ce->ce_flags &= ~CE_HASHED;
a22c6371 111 set_index_entry(istate, nr, ce);
078a58e8 112 ce->ce_flags |= CE_UPDATE_IN_BASE;
883e248b 113 mark_fsmonitor_invalid(istate, ce);
e636a7b4 114 istate->cache_changed |= CE_ENTRY_CHANGED;
cf558704
LT
115}
116
81dc2307
PB
117void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name)
118{
285c2e25 119 struct cache_entry *old_entry = istate->cache[nr], *new_entry;
81dc2307
PB
120 int namelen = strlen(new_name);
121
a849735b 122 new_entry = make_empty_cache_entry(istate, namelen);
285c2e25
BW
123 copy_cache_entry(new_entry, old_entry);
124 new_entry->ce_flags &= ~CE_HASHED;
125 new_entry->ce_namelen = namelen;
126 new_entry->index = 0;
127 memcpy(new_entry->name, new_name, namelen + 1);
81dc2307 128
285c2e25
BW
129 cache_tree_invalidate_path(istate, old_entry->name);
130 untracked_cache_remove_from_index(istate, old_entry->name);
81dc2307 131 remove_index_entry_at(istate, nr);
285c2e25 132 add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
81dc2307
PB
133}
134
c21d39d7
MH
135void fill_stat_data(struct stat_data *sd, struct stat *st)
136{
137 sd->sd_ctime.sec = (unsigned int)st->st_ctime;
138 sd->sd_mtime.sec = (unsigned int)st->st_mtime;
139 sd->sd_ctime.nsec = ST_CTIME_NSEC(*st);
140 sd->sd_mtime.nsec = ST_MTIME_NSEC(*st);
141 sd->sd_dev = st->st_dev;
142 sd->sd_ino = st->st_ino;
143 sd->sd_uid = st->st_uid;
144 sd->sd_gid = st->st_gid;
145 sd->sd_size = st->st_size;
146}
147
148int match_stat_data(const struct stat_data *sd, struct stat *st)
149{
150 int changed = 0;
151
152 if (sd->sd_mtime.sec != (unsigned int)st->st_mtime)
153 changed |= MTIME_CHANGED;
154 if (trust_ctime && check_stat &&
155 sd->sd_ctime.sec != (unsigned int)st->st_ctime)
156 changed |= CTIME_CHANGED;
157
158#ifdef USE_NSEC
159 if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st))
160 changed |= MTIME_CHANGED;
161 if (trust_ctime && check_stat &&
162 sd->sd_ctime.nsec != ST_CTIME_NSEC(*st))
163 changed |= CTIME_CHANGED;
164#endif
165
166 if (check_stat) {
167 if (sd->sd_uid != (unsigned int) st->st_uid ||
168 sd->sd_gid != (unsigned int) st->st_gid)
169 changed |= OWNER_CHANGED;
170 if (sd->sd_ino != (unsigned int) st->st_ino)
171 changed |= INODE_CHANGED;
172 }
173
174#ifdef USE_STDEV
175 /*
176 * st_dev breaks on network filesystems where different
177 * clients will have different views of what "device"
178 * the filesystem is on
179 */
180 if (check_stat && sd->sd_dev != (unsigned int) st->st_dev)
181 changed |= INODE_CHANGED;
182#endif
183
184 if (sd->sd_size != (unsigned int) st->st_size)
185 changed |= DATA_CHANGED;
186
187 return changed;
188}
189
415e96c8
JH
190/*
191 * This only updates the "non-critical" parts of the directory
192 * cache, ie the parts that aren't tracked by GIT, and only used
193 * to validate the cache.
194 */
195void fill_stat_cache_info(struct cache_entry *ce, struct stat *st)
196{
c21d39d7 197 fill_stat_data(&ce->ce_stat_data, st);
5f73076c
JH
198
199 if (assume_unchanged)
7a51ed66 200 ce->ce_flags |= CE_VALID;
eadb5831 201
883e248b 202 if (S_ISREG(st->st_mode)) {
eadb5831 203 ce_mark_uptodate(ce);
883e248b
BP
204 mark_fsmonitor_valid(ce);
205 }
415e96c8
JH
206}
207
21a6b9fa 208static int ce_compare_data(const struct cache_entry *ce, struct stat *st)
29e4d363
JH
209{
210 int match = -1;
1b8ac5ea 211 int fd = git_open_cloexec(ce->name, O_RDONLY);
29e4d363
JH
212
213 if (fd >= 0) {
bebfecb9 214 struct object_id oid;
e3506559 215 if (!index_fd(&oid, fd, st, OBJ_BLOB, ce->name, 0))
6a29d7b7 216 match = !oideq(&oid, &ce->oid);
7f8508e8 217 /* index_fd() closed the file descriptor already */
29e4d363
JH
218 }
219 return match;
220}
221
21a6b9fa 222static int ce_compare_link(const struct cache_entry *ce, size_t expected_size)
29e4d363
JH
223{
224 int match = -1;
29e4d363
JH
225 void *buffer;
226 unsigned long size;
21666f1a 227 enum object_type type;
a60272b3 228 struct strbuf sb = STRBUF_INIT;
29e4d363 229
a60272b3 230 if (strbuf_readlink(&sb, ce->name, expected_size))
29e4d363 231 return -1;
a60272b3 232
b4f5aca4 233 buffer = read_object_file(&ce->oid, &type, &size);
a60272b3
LT
234 if (buffer) {
235 if (size == sb.len)
236 match = memcmp(buffer, sb.buf, size);
237 free(buffer);
29e4d363 238 }
a60272b3 239 strbuf_release(&sb);
29e4d363
JH
240 return match;
241}
242
21a6b9fa 243static int ce_compare_gitlink(const struct cache_entry *ce)
f35a6d3b 244{
1053fe82 245 struct object_id oid;
f35a6d3b
LT
246
247 /*
248 * We don't actually require that the .git directory
302b9282 249 * under GITLINK directory be a valid git directory. It
f35a6d3b
LT
250 * might even be missing (in case nobody populated that
251 * sub-project).
252 *
253 * If so, we consider it always to match.
254 */
a98e6101 255 if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0)
f35a6d3b 256 return 0;
6a29d7b7 257 return !oideq(&oid, &ce->oid);
f35a6d3b
LT
258}
259
21a6b9fa 260static int ce_modified_check_fs(const struct cache_entry *ce, struct stat *st)
29e4d363
JH
261{
262 switch (st->st_mode & S_IFMT) {
263 case S_IFREG:
264 if (ce_compare_data(ce, st))
265 return DATA_CHANGED;
266 break;
267 case S_IFLNK:
dc49cd76 268 if (ce_compare_link(ce, xsize_t(st->st_size)))
29e4d363
JH
269 return DATA_CHANGED;
270 break;
a8ee75bc 271 case S_IFDIR:
7a51ed66 272 if (S_ISGITLINK(ce->ce_mode))
c70115b4 273 return ce_compare_gitlink(ce) ? DATA_CHANGED : 0;
1cf01a34 274 /* else fallthrough */
29e4d363
JH
275 default:
276 return TYPE_CHANGED;
277 }
278 return 0;
279}
280
21a6b9fa 281static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st)
734aab75
LT
282{
283 unsigned int changed = 0;
284
7a51ed66
LT
285 if (ce->ce_flags & CE_REMOVE)
286 return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED;
287
288 switch (ce->ce_mode & S_IFMT) {
8ae0a8c5
KS
289 case S_IFREG:
290 changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0;
3e09cdfd
JH
291 /* We consider only the owner x bit to be relevant for
292 * "mode changes"
293 */
294 if (trust_executable_bit &&
7a51ed66 295 (0100 & (ce->ce_mode ^ st->st_mode)))
ffbe1add 296 changed |= MODE_CHANGED;
8ae0a8c5
KS
297 break;
298 case S_IFLNK:
78a8d641
JS
299 if (!S_ISLNK(st->st_mode) &&
300 (has_symlinks || !S_ISREG(st->st_mode)))
301 changed |= TYPE_CHANGED;
8ae0a8c5 302 break;
302b9282 303 case S_IFGITLINK:
c70115b4 304 /* We ignore most of the st_xxx fields for gitlinks */
f35a6d3b
LT
305 if (!S_ISDIR(st->st_mode))
306 changed |= TYPE_CHANGED;
307 else if (ce_compare_gitlink(ce))
308 changed |= DATA_CHANGED;
a8ee75bc 309 return changed;
8ae0a8c5 310 default:
7a51ed66 311 die("internal error: ce_mode is %o", ce->ce_mode);
8ae0a8c5 312 }
2cb45e95 313
c21d39d7 314 changed |= match_stat_data(&ce->ce_stat_data, st);
b0391890 315
f49c2c22 316 /* Racily smudged entry? */
c21d39d7 317 if (!ce->ce_stat_data.sd_size) {
99d1a986 318 if (!is_empty_blob_sha1(ce->oid.hash))
f49c2c22
LT
319 changed |= DATA_CHANGED;
320 }
321
407c8eb0
JH
322 return changed;
323}
324
2bb4cda1
NTND
325static int is_racy_stat(const struct index_state *istate,
326 const struct stat_data *sd)
6d91da6d 327{
2bb4cda1 328 return (istate->timestamp.sec &&
fba2f38a
KB
329#ifdef USE_NSEC
330 /* nanosecond timestamped files can also be racy! */
2bb4cda1
NTND
331 (istate->timestamp.sec < sd->sd_mtime.sec ||
332 (istate->timestamp.sec == sd->sd_mtime.sec &&
333 istate->timestamp.nsec <= sd->sd_mtime.nsec))
fba2f38a 334#else
2bb4cda1 335 istate->timestamp.sec <= sd->sd_mtime.sec
fba2f38a 336#endif
2bb4cda1
NTND
337 );
338}
339
340static int is_racy_timestamp(const struct index_state *istate,
341 const struct cache_entry *ce)
342{
343 return (!S_ISGITLINK(ce->ce_mode) &&
344 is_racy_stat(istate, &ce->ce_stat_data));
6d91da6d
JH
345}
346
ed4efab1
NTND
347int match_stat_data_racy(const struct index_state *istate,
348 const struct stat_data *sd, struct stat *st)
349{
350 if (is_racy_stat(istate, sd))
351 return MTIME_CHANGED;
352 return match_stat_data(sd, st);
6d91da6d
JH
353}
354
883e248b 355int ie_match_stat(struct index_state *istate,
21a6b9fa 356 const struct cache_entry *ce, struct stat *st,
4bd5b7da 357 unsigned int options)
407c8eb0 358{
5f73076c 359 unsigned int changed;
4bd5b7da 360 int ignore_valid = options & CE_MATCH_IGNORE_VALID;
56cac48c 361 int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE;
4bd5b7da 362 int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY;
883e248b 363 int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR;
5f73076c 364
883e248b
BP
365 if (!ignore_fsmonitor)
366 refresh_fsmonitor(istate);
5f73076c
JH
367 /*
368 * If it's marked as always valid in the index, it's
369 * valid whatever the checked-out copy says.
56cac48c
NTND
370 *
371 * skip-worktree has the same effect with higher precedence
5f73076c 372 */
56cac48c
NTND
373 if (!ignore_skip_worktree && ce_skip_worktree(ce))
374 return 0;
7a51ed66 375 if (!ignore_valid && (ce->ce_flags & CE_VALID))
5f73076c 376 return 0;
883e248b
BP
377 if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID))
378 return 0;
5f73076c 379
331fcb59
JH
380 /*
381 * Intent-to-add entries have not been added, so the index entry
382 * by definition never matches what is in the work tree until it
383 * actually gets added.
384 */
895ff3b2 385 if (ce_intent_to_add(ce))
331fcb59
JH
386 return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED;
387
5f73076c 388 changed = ce_match_stat_basic(ce, st);
407c8eb0 389
29e4d363
JH
390 /*
391 * Within 1 second of this sequence:
392 * echo xyzzy >file && git-update-index --add file
393 * running this command:
394 * echo frotz >file
395 * would give a falsely clean cache entry. The mtime and
396 * length match the cache, and other stat fields do not change.
397 *
398 * We could detect this at update-index time (the cache entry
399 * being registered/updated records the same time as "now")
400 * and delay the return from git-update-index, but that would
401 * effectively mean we can make at most one commit per second,
402 * which is not acceptable. Instead, we check cache entries
403 * whose mtime are the same as the index file timestamp more
5f73076c 404 * carefully than others.
29e4d363 405 */
6d91da6d 406 if (!changed && is_racy_timestamp(istate, ce)) {
42f77406
JH
407 if (assume_racy_is_modified)
408 changed |= DATA_CHANGED;
409 else
410 changed |= ce_modified_check_fs(ce, st);
411 }
b0391890 412
29e4d363 413 return changed;
b0391890
JH
414}
415
883e248b 416int ie_modified(struct index_state *istate,
21a6b9fa
RS
417 const struct cache_entry *ce,
418 struct stat *st, unsigned int options)
b0391890 419{
29e4d363 420 int changed, changed_fs;
4bd5b7da
JH
421
422 changed = ie_match_stat(istate, ce, st, options);
b0391890
JH
423 if (!changed)
424 return 0;
b0391890
JH
425 /*
426 * If the mode or type has changed, there's no point in trying
427 * to refresh the entry - it's not going to match
428 */
429 if (changed & (MODE_CHANGED | TYPE_CHANGED))
430 return changed;
431
c70115b4
JH
432 /*
433 * Immediately after read-tree or update-index --cacheinfo,
434 * the length field is zero, as we have never even read the
435 * lstat(2) information once, and we cannot trust DATA_CHANGED
436 * returned by ie_match_stat() which in turn was returned by
437 * ce_match_stat_basic() to signal that the filesize of the
438 * blob changed. We have to actually go to the filesystem to
439 * see if the contents match, and if so, should answer "unchanged".
440 *
441 * The logic does not apply to gitlinks, as ce_match_stat_basic()
442 * already has checked the actual HEAD from the filesystem in the
443 * subproject. If ie_match_stat() already said it is different,
444 * then we know it is.
b0391890 445 */
c70115b4 446 if ((changed & DATA_CHANGED) &&
c21d39d7 447 (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0))
b0391890
JH
448 return changed;
449
29e4d363
JH
450 changed_fs = ce_modified_check_fs(ce, st);
451 if (changed_fs)
452 return changed | changed_fs;
b0391890
JH
453 return 0;
454}
455
958ba6c9
LT
456int base_name_compare(const char *name1, int len1, int mode1,
457 const char *name2, int len2, int mode2)
458{
459 unsigned char c1, c2;
460 int len = len1 < len2 ? len1 : len2;
461 int cmp;
462
463 cmp = memcmp(name1, name2, len);
464 if (cmp)
465 return cmp;
466 c1 = name1[len];
467 c2 = name2[len];
1833a925 468 if (!c1 && S_ISDIR(mode1))
958ba6c9 469 c1 = '/';
1833a925 470 if (!c2 && S_ISDIR(mode2))
958ba6c9
LT
471 c2 = '/';
472 return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0;
473}
474
0ab9e1e8
LT
475/*
476 * df_name_compare() is identical to base_name_compare(), except it
477 * compares conflicting directory/file entries as equal. Note that
478 * while a directory name compares as equal to a regular file, they
479 * then individually compare _differently_ to a filename that has
480 * a dot after the basename (because '\0' < '.' < '/').
481 *
482 * This is used by routines that want to traverse the git namespace
483 * but then handle conflicting entries together when possible.
484 */
485int df_name_compare(const char *name1, int len1, int mode1,
486 const char *name2, int len2, int mode2)
487{
488 int len = len1 < len2 ? len1 : len2, cmp;
489 unsigned char c1, c2;
490
491 cmp = memcmp(name1, name2, len);
492 if (cmp)
493 return cmp;
494 /* Directories and files compare equal (same length, same name) */
495 if (len1 == len2)
496 return 0;
497 c1 = name1[len];
498 if (!c1 && S_ISDIR(mode1))
499 c1 = '/';
500 c2 = name2[len];
501 if (!c2 && S_ISDIR(mode2))
502 c2 = '/';
503 if (c1 == '/' && !c2)
504 return 0;
505 if (c2 == '/' && !c1)
506 return 0;
507 return c1 - c2;
508}
509
ccdd4a0f 510int name_compare(const char *name1, size_t len1, const char *name2, size_t len2)
eb38c22f 511{
ccdd4a0f
JM
512 size_t min_len = (len1 < len2) ? len1 : len2;
513 int cmp = memcmp(name1, name2, min_len);
eb38c22f
LT
514 if (cmp)
515 return cmp;
516 if (len1 < len2)
517 return -1;
518 if (len1 > len2)
519 return 1;
ccdd4a0f
JM
520 return 0;
521}
522
523int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2)
524{
525 int cmp;
526
527 cmp = name_compare(name1, len1, name2, len2);
528 if (cmp)
529 return cmp;
5f73076c 530
b60e188c 531 if (stage1 < stage2)
95fd5bf8 532 return -1;
b60e188c 533 if (stage1 > stage2)
95fd5bf8 534 return 1;
eb38c22f
LT
535 return 0;
536}
537
357e9c69 538static int index_name_stage_pos(const struct index_state *istate, const char *name, int namelen, int stage)
eb38c22f
LT
539{
540 int first, last;
541
542 first = 0;
4aab5b46 543 last = istate->cache_nr;
eb38c22f
LT
544 while (last > first) {
545 int next = (last + first) >> 1;
4aab5b46 546 struct cache_entry *ce = istate->cache[next];
b60e188c 547 int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce));
eb38c22f 548 if (!cmp)
76e7f4ec 549 return next;
eb38c22f
LT
550 if (cmp < 0) {
551 last = next;
552 continue;
553 }
554 first = next+1;
555 }
76e7f4ec 556 return -first-1;
eb38c22f
LT
557}
558
b60e188c
TG
559int index_name_pos(const struct index_state *istate, const char *name, int namelen)
560{
561 return index_name_stage_pos(istate, name, namelen, 0);
562}
563
4aab5b46 564int remove_index_entry_at(struct index_state *istate, int pos)
7b937ca3 565{
cf558704
LT
566 struct cache_entry *ce = istate->cache[pos];
567
cfc5789a 568 record_resolve_undo(istate, ce);
2092678c 569 remove_name_hash(istate, ce);
045113a5 570 save_or_free_index_entry(istate, ce);
e636a7b4 571 istate->cache_changed |= CE_ENTRY_REMOVED;
4aab5b46
JH
572 istate->cache_nr--;
573 if (pos >= istate->cache_nr)
7b937ca3 574 return 0;
f331ab9d
RS
575 MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1,
576 istate->cache_nr - pos);
7b937ca3
LT
577 return 1;
578}
579
36419c8e 580/*
98e023de 581 * Remove all cache entries marked for removal, that is where
36419c8e
KB
582 * CE_REMOVE is set in ce_flags. This is much more effective than
583 * calling remove_index_entry_at() for each entry to be removed.
584 */
585void remove_marked_cache_entries(struct index_state *istate)
586{
587 struct cache_entry **ce_array = istate->cache;
588 unsigned int i, j;
589
590 for (i = j = 0; i < istate->cache_nr; i++) {
5699d17e 591 if (ce_array[i]->ce_flags & CE_REMOVE) {
2092678c 592 remove_name_hash(istate, ce_array[i]);
045113a5 593 save_or_free_index_entry(istate, ce_array[i]);
5699d17e 594 }
36419c8e
KB
595 else
596 ce_array[j++] = ce_array[i];
597 }
ad837d9e
NTND
598 if (j == istate->cache_nr)
599 return;
e636a7b4 600 istate->cache_changed |= CE_ENTRY_REMOVED;
36419c8e
KB
601 istate->cache_nr = j;
602}
603
4aab5b46 604int remove_file_from_index(struct index_state *istate, const char *path)
197ee8c9 605{
4aab5b46 606 int pos = index_name_pos(istate, path, strlen(path));
c4e3cca1
JH
607 if (pos < 0)
608 pos = -pos-1;
a5400efe 609 cache_tree_invalidate_path(istate, path);
e931371a 610 untracked_cache_remove_from_index(istate, path);
4aab5b46
JH
611 while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path))
612 remove_index_entry_at(istate, pos);
197ee8c9
LT
613 return 0;
614}
615
20314271
JS
616static int compare_name(struct cache_entry *ce, const char *path, int namelen)
617{
618 return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen);
619}
620
621static int index_name_pos_also_unmerged(struct index_state *istate,
622 const char *path, int namelen)
623{
624 int pos = index_name_pos(istate, path, namelen);
625 struct cache_entry *ce;
626
627 if (pos >= 0)
628 return pos;
629
630 /* maybe unmerged? */
631 pos = -1 - pos;
632 if (pos >= istate->cache_nr ||
633 compare_name((ce = istate->cache[pos]), path, namelen))
634 return -1;
635
636 /* order of preference: stage 2, 1, 3 */
637 if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr &&
638 ce_stage((ce = istate->cache[pos + 1])) == 2 &&
639 !compare_name(ce, path, namelen))
640 pos++;
641 return pos;
642}
643
1102952b
LT
644static int different_name(struct cache_entry *ce, struct cache_entry *alias)
645{
646 int len = ce_namelen(ce);
647 return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len);
648}
649
650/*
651 * If we add a filename that aliases in the cache, we will use the
652 * name that we already have - but we don't want to update the same
653 * alias twice, because that implies that there were actually two
654 * different files with aliasing names!
655 *
656 * So we use the CE_ADDED flag to verify that the alias was an old
657 * one before we accept it as
658 */
045113a5
NTND
659static struct cache_entry *create_alias_ce(struct index_state *istate,
660 struct cache_entry *ce,
661 struct cache_entry *alias)
1102952b
LT
662{
663 int len;
285c2e25 664 struct cache_entry *new_entry;
1102952b
LT
665
666 if (alias->ce_flags & CE_ADDED)
667 die("Will not add file alias '%s' ('%s' already exists in index)", ce->name, alias->name);
668
669 /* Ok, create the new entry using the name of the existing alias */
670 len = ce_namelen(alias);
a849735b 671 new_entry = make_empty_cache_entry(istate, len);
285c2e25
BW
672 memcpy(new_entry->name, alias->name, len);
673 copy_cache_entry(new_entry, ce);
045113a5 674 save_or_free_index_entry(istate, ce);
285c2e25 675 return new_entry;
1102952b
LT
676}
677
b4b313f9 678void set_object_name_for_intent_to_add_entry(struct cache_entry *ce)
39425819 679{
a09c985e
PO
680 struct object_id oid;
681 if (write_object_file("", 0, blob_type, &oid))
39425819 682 die("cannot create an empty blob in the object database");
a09c985e 683 oidcpy(&ce->oid, &oid);
39425819
JH
684}
685
610d55af 686int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags)
11be42a4 687{
a849735b 688 int namelen, was_same;
d177cab0 689 mode_t st_mode = st->st_mode;
9472935d 690 struct cache_entry *ce, *alias = NULL;
56cac48c 691 unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY;
38ed1d89
JH
692 int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND);
693 int pretend = flags & ADD_CACHE_PRETEND;
39425819
JH
694 int intent_only = flags & ADD_CACHE_INTENT;
695 int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE|
696 (intent_only ? ADD_CACHE_NEW_ONLY : 0));
9472935d
TB
697 int newflags = HASH_WRITE_OBJECT;
698
699 if (flags & HASH_RENORMALIZE)
700 newflags |= HASH_RENORMALIZE;
11be42a4 701
d177cab0 702 if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode))
960b8ad1 703 return error("%s: can only add regular files, symbolic links or git-directories", path);
11be42a4
JS
704
705 namelen = strlen(path);
d177cab0 706 if (S_ISDIR(st_mode)) {
09595258
LT
707 while (namelen && path[namelen-1] == '/')
708 namelen--;
709 }
a849735b 710 ce = make_empty_cache_entry(istate, namelen);
11be42a4 711 memcpy(ce->name, path, namelen);
b60e188c 712 ce->ce_namelen = namelen;
39425819
JH
713 if (!intent_only)
714 fill_stat_cache_info(ce, st);
388b2acd
JH
715 else
716 ce->ce_flags |= CE_INTENT_TO_ADD;
11be42a4 717
610d55af
TG
718
719 if (trust_executable_bit && has_symlinks) {
d177cab0 720 ce->ce_mode = create_ce_mode(st_mode);
610d55af 721 } else {
78a8d641
JS
722 /* If there is an existing entry, pick the mode bits and type
723 * from it, otherwise assume unexecutable regular file.
11be42a4 724 */
185c975f 725 struct cache_entry *ent;
20314271 726 int pos = index_name_pos_also_unmerged(istate, path, namelen);
185c975f 727
4aab5b46 728 ent = (0 <= pos) ? istate->cache[pos] : NULL;
d177cab0 729 ce->ce_mode = ce_mode_from_stat(ent, st_mode);
11be42a4
JS
730 }
731
dc1ae704
JJ
732 /* When core.ignorecase=true, determine if a directory of the same name but differing
733 * case already exists within the Git repository. If it does, ensure the directory
734 * case of the file being added to the repository matches (is folded into) the existing
735 * entry's directory case.
736 */
737 if (ignore_case) {
41284eb0 738 adjust_dirname_case(istate, ce->name);
dc1ae704 739 }
9472935d
TB
740 if (!(flags & HASH_RENORMALIZE)) {
741 alias = index_file_exists(istate, ce->name,
742 ce_namelen(ce), ignore_case);
743 if (alias &&
744 !ce_stage(alias) &&
745 !ie_match_stat(istate, alias, st, ce_option)) {
746 /* Nothing changed, really */
747 if (!S_ISGITLINK(alias->ce_mode))
748 ce_mark_uptodate(alias);
749 alias->ce_flags |= CE_ADDED;
dc1ae704 750
a849735b 751 discard_cache_entry(ce);
9472935d
TB
752 return 0;
753 }
0781b8a9 754 }
39425819 755 if (!intent_only) {
9472935d 756 if (index_path(&ce->oid, path, st, newflags)) {
a849735b 757 discard_cache_entry(ce);
39425819 758 return error("unable to index file %s", path);
2d9426b0 759 }
39425819 760 } else
b4b313f9 761 set_object_name_for_intent_to_add_entry(ce);
39425819 762
1102952b 763 if (ignore_case && alias && different_name(ce, alias))
045113a5 764 ce = create_alias_ce(istate, ce, alias);
1102952b 765 ce->ce_flags |= CE_ADDED;
38ed1d89 766
3bf0dd1f 767 /* It was suspected to be racily clean, but it turns out to be Ok */
38ed1d89
JH
768 was_same = (alias &&
769 !ce_stage(alias) &&
4a7e27e9 770 oideq(&alias->oid, &ce->oid) &&
38ed1d89
JH
771 ce->ce_mode == alias->ce_mode);
772
773 if (pretend)
a849735b 774 discard_cache_entry(ce);
067178ed 775 else if (add_index_entry(istate, ce, add_option)) {
a849735b 776 discard_cache_entry(ce);
067178ed
JH
777 return error("unable to add %s to index", path);
778 }
38ed1d89 779 if (verbose && !was_same)
11be42a4 780 printf("add '%s'\n", path);
11be42a4
JS
781 return 0;
782}
783
610d55af 784int add_file_to_index(struct index_state *istate, const char *path, int flags)
d177cab0
LT
785{
786 struct stat st;
787 if (lstat(path, &st))
d824cbba 788 die_errno("unable to stat '%s'", path);
610d55af 789 return add_to_index(istate, path, &st, flags);
d177cab0
LT
790}
791
a849735b
JM
792struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len)
793{
8e72d675 794 return mem_pool__ce_calloc(find_mem_pool(istate), len);
a849735b
JM
795}
796
797struct cache_entry *make_empty_transient_cache_entry(size_t len)
798{
799 return xcalloc(1, cache_entry_size(len));
800}
801
802struct cache_entry *make_cache_entry(struct index_state *istate,
803 unsigned int mode,
825ed4d9
JM
804 const struct object_id *oid,
805 const char *path,
806 int stage,
807 unsigned int refresh_options)
6640f881 808{
bc1c2caa 809 struct cache_entry *ce, *ret;
a849735b 810 int len;
6640f881 811
10ecfa76 812 if (!verify_path(path, mode)) {
7e7abea9 813 error("Invalid path '%s'", path);
6640f881 814 return NULL;
7e7abea9 815 }
6640f881
CR
816
817 len = strlen(path);
a849735b 818 ce = make_empty_cache_entry(istate, len);
6640f881 819
825ed4d9 820 oidcpy(&ce->oid, oid);
6640f881 821 memcpy(ce->name, path, len);
b60e188c
TG
822 ce->ce_flags = create_ce_flags(stage);
823 ce->ce_namelen = len;
6640f881
CR
824 ce->ce_mode = create_ce_mode(mode);
825
768d7965 826 ret = refresh_cache_entry(&the_index, ce, refresh_options);
915e44c6 827 if (ret != ce)
a849735b 828 discard_cache_entry(ce);
915e44c6 829 return ret;
6640f881
CR
830}
831
a849735b
JM
832struct cache_entry *make_transient_cache_entry(unsigned int mode, const struct object_id *oid,
833 const char *path, int stage)
834{
835 struct cache_entry *ce;
836 int len;
837
838 if (!verify_path(path, mode)) {
839 error("Invalid path '%s'", path);
840 return NULL;
841 }
842
843 len = strlen(path);
844 ce = make_empty_transient_cache_entry(len);
845
846 oidcpy(&ce->oid, oid);
847 memcpy(ce->name, path, len);
848 ce->ce_flags = create_ce_flags(stage);
849 ce->ce_namelen = len;
850 ce->ce_mode = create_ce_mode(mode);
851
852 return ce;
853}
854
d9d70966
TG
855/*
856 * Chmod an index entry with either +x or -x.
857 *
858 * Returns -1 if the chmod for the particular cache entry failed (if it's
859 * not a regular file), -2 if an invalid flip argument is passed in, 0
860 * otherwise.
861 */
862int chmod_index_entry(struct index_state *istate, struct cache_entry *ce,
863 char flip)
864{
865 if (!S_ISREG(ce->ce_mode))
866 return -1;
867 switch (flip) {
868 case '+':
869 ce->ce_mode |= 0111;
870 break;
871 case '-':
872 ce->ce_mode &= ~0111;
873 break;
874 default:
875 return -2;
876 }
877 cache_tree_invalidate_path(istate, ce->name);
878 ce->ce_flags |= CE_UPDATE_IN_BASE;
883e248b 879 mark_fsmonitor_invalid(istate, ce);
d9d70966
TG
880 istate->cache_changed |= CE_ENTRY_CHANGED;
881
882 return 0;
883}
884
9c5e6c80 885int ce_same_name(const struct cache_entry *a, const struct cache_entry *b)
7b937ca3
LT
886{
887 int len = ce_namelen(a);
888 return ce_namelen(b) == len && !memcmp(a->name, b->name, len);
889}
890
8dcf39c4
LT
891/*
892 * We fundamentally don't like some paths: we don't want
893 * dot or dot-dot anywhere, and for obvious reasons don't
894 * want to recurse into ".git" either.
895 *
896 * Also, we don't want double slashes or slashes at the
897 * end that can make pathnames ambiguous.
898 */
10ecfa76 899static int verify_dotfile(const char *rest, unsigned mode)
8dcf39c4
LT
900{
901 /*
902 * The first character was '.', but that
903 * has already been discarded, we now test
904 * the rest.
905 */
e0f530ff 906
8dcf39c4 907 /* "." is not allowed */
e0f530ff 908 if (*rest == '\0' || is_dir_sep(*rest))
8dcf39c4
LT
909 return 0;
910
e0f530ff 911 switch (*rest) {
8dcf39c4 912 /*
641084b6
JK
913 * ".git" followed by NUL or slash is bad. Note that we match
914 * case-insensitively here, even if ignore_case is not set.
915 * This outlaws ".GIT" everywhere out of an abundance of caution,
916 * since there's really no good reason to allow it.
10ecfa76
JK
917 *
918 * Once we've seen ".git", we can also find ".gitmodules", etc (also
919 * case-insensitively).
8dcf39c4
LT
920 */
921 case 'g':
cc2fc7c2
JK
922 case 'G':
923 if (rest[1] != 'i' && rest[1] != 'I')
8dcf39c4 924 break;
cc2fc7c2 925 if (rest[2] != 't' && rest[2] != 'T')
8dcf39c4 926 break;
e19e5e66
JK
927 if (rest[3] == '\0' || is_dir_sep(rest[3]))
928 return 0;
10ecfa76
JK
929 if (S_ISLNK(mode)) {
930 rest += 3;
931 if (skip_iprefix(rest, "modules", &rest) &&
932 (*rest == '\0' || is_dir_sep(*rest)))
933 return 0;
934 }
e19e5e66 935 break;
8dcf39c4 936 case '.':
e0f530ff 937 if (rest[1] == '\0' || is_dir_sep(rest[1]))
8dcf39c4
LT
938 return 0;
939 }
940 return 1;
941}
942
10ecfa76 943int verify_path(const char *path, unsigned mode)
8dcf39c4
LT
944{
945 char c;
946
56948cb6
EFL
947 if (has_dos_drive_prefix(path))
948 return 0;
949
8dcf39c4
LT
950 goto inside;
951 for (;;) {
952 if (!c)
953 return 1;
56948cb6 954 if (is_dir_sep(c)) {
8dcf39c4 955inside:
10ecfa76
JK
956 if (protect_hfs) {
957 if (is_hfs_dotgit(path))
958 return 0;
959 if (S_ISLNK(mode)) {
960 if (is_hfs_dotgitmodules(path))
961 return 0;
962 }
963 }
964 if (protect_ntfs) {
965 if (is_ntfs_dotgit(path))
966 return 0;
967 if (S_ISLNK(mode)) {
968 if (is_ntfs_dotgitmodules(path))
969 return 0;
970 }
971 }
972
8dcf39c4 973 c = *path++;
10ecfa76 974 if ((c == '.' && !verify_dotfile(path, mode)) ||
3bdf09c7
JH
975 is_dir_sep(c) || c == '\0')
976 return 0;
8dcf39c4
LT
977 }
978 c = *path++;
979 }
980}
981
12676608
LT
982/*
983 * Do we have another file that has the beginning components being a
984 * proper superset of the name we're trying to add?
0f1e4f04 985 */
4aab5b46
JH
986static int has_file_name(struct index_state *istate,
987 const struct cache_entry *ce, int pos, int ok_to_replace)
0f1e4f04 988{
12676608
LT
989 int retval = 0;
990 int len = ce_namelen(ce);
b155725d 991 int stage = ce_stage(ce);
12676608 992 const char *name = ce->name;
0f1e4f04 993
4aab5b46
JH
994 while (pos < istate->cache_nr) {
995 struct cache_entry *p = istate->cache[pos++];
0f1e4f04 996
12676608 997 if (len >= ce_namelen(p))
0f1e4f04 998 break;
12676608
LT
999 if (memcmp(name, p->name, len))
1000 break;
b155725d
JH
1001 if (ce_stage(p) != stage)
1002 continue;
12676608
LT
1003 if (p->name[len] != '/')
1004 continue;
7a51ed66 1005 if (p->ce_flags & CE_REMOVE)
21cd8d00 1006 continue;
12676608
LT
1007 retval = -1;
1008 if (!ok_to_replace)
1009 break;
4aab5b46 1010 remove_index_entry_at(istate, --pos);
0f1e4f04 1011 }
12676608
LT
1012 return retval;
1013}
0f1e4f04 1014
a6db3fbb
JH
1015
1016/*
1017 * Like strcmp(), but also return the offset of the first change.
1018 * If strings are equal, return the length.
1019 */
1020int strcmp_offset(const char *s1, const char *s2, size_t *first_change)
1021{
1022 size_t k;
1023
1024 if (!first_change)
1025 return strcmp(s1, s2);
1026
1027 for (k = 0; s1[k] == s2[k]; k++)
1028 if (s1[k] == '\0')
1029 break;
1030
1031 *first_change = k;
1032 return (unsigned char)s1[k] - (unsigned char)s2[k];
1033}
1034
12676608
LT
1035/*
1036 * Do we have another file with a pathname that is a proper
1037 * subset of the name we're trying to add?
06b6d81b
JH
1038 *
1039 * That is, is there another file in the index with a path
1040 * that matches a sub-directory in the given entry?
12676608 1041 */
4aab5b46
JH
1042static int has_dir_name(struct index_state *istate,
1043 const struct cache_entry *ce, int pos, int ok_to_replace)
12676608
LT
1044{
1045 int retval = 0;
b155725d 1046 int stage = ce_stage(ce);
12676608
LT
1047 const char *name = ce->name;
1048 const char *slash = name + ce_namelen(ce);
06b6d81b
JH
1049 size_t len_eq_last;
1050 int cmp_last = 0;
1051
1052 /*
1053 * We are frequently called during an iteration on a sorted
1054 * list of pathnames and while building a new index. Therefore,
1055 * there is a high probability that this entry will eventually
1056 * be appended to the index, rather than inserted in the middle.
1057 * If we can confirm that, we can avoid binary searches on the
1058 * components of the pathname.
1059 *
1060 * Compare the entry's full path with the last path in the index.
1061 */
1062 if (istate->cache_nr > 0) {
1063 cmp_last = strcmp_offset(name,
1064 istate->cache[istate->cache_nr - 1]->name,
1065 &len_eq_last);
1066 if (cmp_last > 0) {
1067 if (len_eq_last == 0) {
1068 /*
1069 * The entry sorts AFTER the last one in the
1070 * index and their paths have no common prefix,
1071 * so there cannot be a F/D conflict.
1072 */
1073 return retval;
1074 } else {
1075 /*
1076 * The entry sorts AFTER the last one in the
1077 * index, but has a common prefix. Fall through
1078 * to the loop below to disect the entry's path
1079 * and see where the difference is.
1080 */
1081 }
1082 } else if (cmp_last == 0) {
1083 /*
1084 * The entry exactly matches the last one in the
1085 * index, but because of multiple stage and CE_REMOVE
1086 * items, we fall through and let the regular search
1087 * code handle it.
1088 */
1089 }
1090 }
0f1e4f04 1091
12676608 1092 for (;;) {
b986df5c 1093 size_t len;
0f1e4f04 1094
12676608
LT
1095 for (;;) {
1096 if (*--slash == '/')
1097 break;
1098 if (slash <= ce->name)
1099 return retval;
1100 }
1101 len = slash - name;
0f1e4f04 1102
b986df5c
JH
1103 if (cmp_last > 0) {
1104 /*
1105 * (len + 1) is a directory boundary (including
1106 * the trailing slash). And since the loop is
1107 * decrementing "slash", the first iteration is
1108 * the longest directory prefix; subsequent
1109 * iterations consider parent directories.
1110 */
1111
1112 if (len + 1 <= len_eq_last) {
1113 /*
1114 * The directory prefix (including the trailing
1115 * slash) also appears as a prefix in the last
1116 * entry, so the remainder cannot collide (because
1117 * strcmp said the whole path was greater).
1118 *
1119 * EQ: last: xxx/A
1120 * this: xxx/B
1121 *
1122 * LT: last: xxx/file_A
1123 * this: xxx/file_B
1124 */
1125 return retval;
1126 }
1127
1128 if (len > len_eq_last) {
1129 /*
1130 * This part of the directory prefix (excluding
1131 * the trailing slash) is longer than the known
1132 * equal portions, so this sub-directory cannot
1133 * collide with a file.
1134 *
1135 * GT: last: xxxA
1136 * this: xxxB/file
1137 */
1138 return retval;
1139 }
1140
1141 if (istate->cache_nr > 0 &&
1142 ce_namelen(istate->cache[istate->cache_nr - 1]) > len) {
1143 /*
1144 * The directory prefix lines up with part of
1145 * a longer file or directory name, but sorts
1146 * after it, so this sub-directory cannot
1147 * collide with a file.
1148 *
1149 * last: xxx/yy-file (because '-' sorts before '/')
1150 * this: xxx/yy/abc
1151 */
1152 return retval;
1153 }
1154
1155 /*
1156 * This is a possible collision. Fall through and
1157 * let the regular search code handle it.
1158 *
1159 * last: xxx
1160 * this: xxx/file
1161 */
1162 }
1163
b60e188c 1164 pos = index_name_stage_pos(istate, name, len, stage);
12676608 1165 if (pos >= 0) {
21cd8d00
JH
1166 /*
1167 * Found one, but not so fast. This could
1168 * be a marker that says "I was here, but
1169 * I am being removed". Such an entry is
1170 * not a part of the resulting tree, and
1171 * it is Ok to have a directory at the same
1172 * path.
1173 */
077c48df 1174 if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) {
21cd8d00
JH
1175 retval = -1;
1176 if (!ok_to_replace)
1177 break;
4aab5b46 1178 remove_index_entry_at(istate, pos);
21cd8d00
JH
1179 continue;
1180 }
12676608 1181 }
21cd8d00
JH
1182 else
1183 pos = -pos-1;
12676608
LT
1184
1185 /*
1186 * Trivial optimization: if we find an entry that
1187 * already matches the sub-directory, then we know
b155725d 1188 * we're ok, and we can exit.
12676608 1189 */
4aab5b46
JH
1190 while (pos < istate->cache_nr) {
1191 struct cache_entry *p = istate->cache[pos];
b155725d
JH
1192 if ((ce_namelen(p) <= len) ||
1193 (p->name[len] != '/') ||
1194 memcmp(p->name, name, len))
1195 break; /* not our subdirectory */
077c48df
JH
1196 if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE))
1197 /*
1198 * p is at the same stage as our entry, and
b155725d
JH
1199 * is a subdirectory of what we are looking
1200 * at, so we cannot have conflicts at our
1201 * level or anything shorter.
1202 */
1203 return retval;
1204 pos++;
192268c1 1205 }
0f1e4f04 1206 }
12676608
LT
1207 return retval;
1208}
1209
1210/* We may be in a situation where we already have path/file and path
1211 * is being added, or we already have path and path/file is being
1212 * added. Either one would result in a nonsense tree that has path
1213 * twice when git-write-tree tries to write it out. Prevent it.
a6080a0a 1214 *
12676608
LT
1215 * If ok-to-replace is specified, we remove the conflicting entries
1216 * from the cache so the caller should recompute the insert position.
1217 * When this happens, we return non-zero.
1218 */
4aab5b46
JH
1219static int check_file_directory_conflict(struct index_state *istate,
1220 const struct cache_entry *ce,
1221 int pos, int ok_to_replace)
12676608 1222{
21cd8d00
JH
1223 int retval;
1224
1225 /*
1226 * When ce is an "I am going away" entry, we allow it to be added
1227 */
7a51ed66 1228 if (ce->ce_flags & CE_REMOVE)
21cd8d00
JH
1229 return 0;
1230
12676608
LT
1231 /*
1232 * We check if the path is a sub-path of a subsequent pathname
1233 * first, since removing those will not change the position
21cd8d00 1234 * in the array.
12676608 1235 */
4aab5b46 1236 retval = has_file_name(istate, ce, pos, ok_to_replace);
21cd8d00 1237
12676608
LT
1238 /*
1239 * Then check if the path might have a clashing sub-directory
1240 * before it.
1241 */
4aab5b46 1242 return retval + has_dir_name(istate, ce, pos, ok_to_replace);
0f1e4f04
JH
1243}
1244
af3785dc 1245static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option)
197ee8c9
LT
1246{
1247 int pos;
192268c1
JH
1248 int ok_to_add = option & ADD_CACHE_OK_TO_ADD;
1249 int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE;
b155725d 1250 int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK;
39425819 1251 int new_only = option & ADD_CACHE_NEW_ONLY;
5f73076c 1252
ce7c614b
NTND
1253 if (!(option & ADD_CACHE_KEEP_CACHE_TREE))
1254 cache_tree_invalidate_path(istate, ce->name);
e5494631
JH
1255
1256 /*
1257 * If this entry's path sorts after the last entry in the index,
1258 * we can avoid searching for it.
1259 */
1260 if (istate->cache_nr > 0 &&
1261 strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0)
1262 pos = -istate->cache_nr - 1;
1263 else
1264 pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce));
197ee8c9 1265
3e09cdfd 1266 /* existing match? Just replace it. */
76e7f4ec 1267 if (pos >= 0) {
39425819
JH
1268 if (!new_only)
1269 replace_index_entry(istate, pos, ce);
197ee8c9
LT
1270 return 0;
1271 }
76e7f4ec 1272 pos = -pos-1;
197ee8c9 1273
ffcc9ba7
NTND
1274 if (!(option & ADD_CACHE_KEEP_CACHE_TREE))
1275 untracked_cache_add_to_index(istate, ce->name);
e931371a 1276
7b937ca3
LT
1277 /*
1278 * Inserting a merged entry ("stage 0") into the index
1279 * will always replace all non-merged entries..
1280 */
4aab5b46
JH
1281 if (pos < istate->cache_nr && ce_stage(ce) == 0) {
1282 while (ce_same_name(istate->cache[pos], ce)) {
7b937ca3 1283 ok_to_add = 1;
4aab5b46 1284 if (!remove_index_entry_at(istate, pos))
7b937ca3
LT
1285 break;
1286 }
1287 }
1288
121481ab
LT
1289 if (!ok_to_add)
1290 return -1;
10ecfa76 1291 if (!verify_path(ce->name, ce->ce_mode))
7e7abea9 1292 return error("Invalid path '%s'", ce->name);
121481ab 1293
3e09cdfd 1294 if (!skip_df_check &&
4aab5b46 1295 check_file_directory_conflict(istate, ce, pos, ok_to_replace)) {
192268c1 1296 if (!ok_to_replace)
4aab5b46
JH
1297 return error("'%s' appears as both a file and as a directory",
1298 ce->name);
b60e188c 1299 pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce));
192268c1
JH
1300 pos = -pos-1;
1301 }
af3785dc
JH
1302 return pos + 1;
1303}
1304
1305int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option)
1306{
1307 int pos;
1308
1309 if (option & ADD_CACHE_JUST_APPEND)
1310 pos = istate->cache_nr;
1311 else {
1312 int ret;
1313 ret = add_index_entry_with_check(istate, ce, option);
1314 if (ret <= 0)
1315 return ret;
1316 pos = ret - 1;
1317 }
0f1e4f04 1318
197ee8c9 1319 /* Make sure the array is big enough .. */
999f5660 1320 ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc);
197ee8c9
LT
1321
1322 /* Add it in.. */
4aab5b46 1323 istate->cache_nr++;
af3785dc 1324 if (istate->cache_nr > pos + 1)
f919ffeb
SG
1325 MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos,
1326 istate->cache_nr - pos - 1);
cf558704 1327 set_index_entry(istate, pos, ce);
e636a7b4 1328 istate->cache_changed |= CE_ENTRY_ADDED;
197ee8c9
LT
1329 return 0;
1330}
1331
405e5b2f
LT
1332/*
1333 * "refresh" does not calculate a new sha1 file or bring the
1334 * cache up-to-date for mode/content changes. But what it
1335 * _does_ do is to "re-match" the stat information of a file
1336 * with the cache, so that you can refresh the cache for a
1337 * file that hasn't been changed but where the stat entry is
1338 * out of date.
1339 *
1340 * For example, you'd want to do this after doing a "git-read-tree",
1341 * to link up the stat cache details with the proper files.
1342 */
4aab5b46 1343static struct cache_entry *refresh_cache_ent(struct index_state *istate,
4bd5b7da 1344 struct cache_entry *ce,
d05e6970
JK
1345 unsigned int options, int *err,
1346 int *changed_ret)
405e5b2f
LT
1347{
1348 struct stat st;
1349 struct cache_entry *updated;
a849735b 1350 int changed;
25762726 1351 int refresh = options & CE_MATCH_REFRESH;
4bd5b7da 1352 int ignore_valid = options & CE_MATCH_IGNORE_VALID;
56cac48c 1353 int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE;
2e2e7ec1 1354 int ignore_missing = options & CE_MATCH_IGNORE_MISSING;
883e248b 1355 int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR;
405e5b2f 1356
25762726 1357 if (!refresh || ce_uptodate(ce))
eadb5831
JH
1358 return ce;
1359
883e248b
BP
1360 if (!ignore_fsmonitor)
1361 refresh_fsmonitor(istate);
aa9349d4 1362 /*
56cac48c
NTND
1363 * CE_VALID or CE_SKIP_WORKTREE means the user promised us
1364 * that the change to the work tree does not matter and told
1365 * us not to worry.
aa9349d4 1366 */
56cac48c
NTND
1367 if (!ignore_skip_worktree && ce_skip_worktree(ce)) {
1368 ce_mark_uptodate(ce);
1369 return ce;
1370 }
aa9349d4
MSO
1371 if (!ignore_valid && (ce->ce_flags & CE_VALID)) {
1372 ce_mark_uptodate(ce);
1373 return ce;
1374 }
883e248b
BP
1375 if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) {
1376 ce_mark_uptodate(ce);
1377 return ce;
1378 }
aa9349d4 1379
ccad42d4
RS
1380 if (has_symlink_leading_path(ce->name, ce_namelen(ce))) {
1381 if (ignore_missing)
1382 return ce;
1383 if (err)
1384 *err = ENOENT;
1385 return NULL;
1386 }
1387
8fd2cb40 1388 if (lstat(ce->name, &st) < 0) {
2e2e7ec1
BK
1389 if (ignore_missing && errno == ENOENT)
1390 return ce;
ec0cc704
JH
1391 if (err)
1392 *err = errno;
8fd2cb40
JS
1393 return NULL;
1394 }
405e5b2f 1395
4bd5b7da 1396 changed = ie_match_stat(istate, ce, &st, options);
d05e6970
JK
1397 if (changed_ret)
1398 *changed_ret = changed;
405e5b2f 1399 if (!changed) {
4bd5b7da
JH
1400 /*
1401 * The path is unchanged. If we were told to ignore
1402 * valid bit, then we did the actual stat check and
1403 * found that the entry is unmodified. If the entry
1404 * is not marked VALID, this is the place to mark it
1405 * valid again, under "assume unchanged" mode.
1406 */
1407 if (ignore_valid && assume_unchanged &&
7a51ed66 1408 !(ce->ce_flags & CE_VALID))
405e5b2f 1409 ; /* mark this one VALID again */
eadb5831
JH
1410 else {
1411 /*
1412 * We do not mark the index itself "modified"
1413 * because CE_UPTODATE flag is in-core only;
1414 * we are not going to write this change out.
1415 */
883e248b 1416 if (!S_ISGITLINK(ce->ce_mode)) {
125fd984 1417 ce_mark_uptodate(ce);
883e248b
BP
1418 mark_fsmonitor_valid(ce);
1419 }
8fd2cb40 1420 return ce;
eadb5831 1421 }
405e5b2f
LT
1422 }
1423
4bd5b7da 1424 if (ie_modified(istate, ce, &st, options)) {
ec0cc704
JH
1425 if (err)
1426 *err = EINVAL;
8fd2cb40
JS
1427 return NULL;
1428 }
405e5b2f 1429
a849735b 1430 updated = make_empty_cache_entry(istate, ce_namelen(ce));
0e267b7a
BP
1431 copy_cache_entry(updated, ce);
1432 memcpy(updated->name, ce->name, ce->ce_namelen + 1);
405e5b2f 1433 fill_stat_cache_info(updated, &st);
4bd5b7da
JH
1434 /*
1435 * If ignore_valid is not set, we should leave CE_VALID bit
1436 * alone. Otherwise, paths marked with --no-assume-unchanged
1437 * (i.e. things to be edited) will reacquire CE_VALID bit
1438 * automatically, which is not really what we want.
405e5b2f 1439 */
4bd5b7da 1440 if (!ignore_valid && assume_unchanged &&
7a51ed66
LT
1441 !(ce->ce_flags & CE_VALID))
1442 updated->ce_flags &= ~CE_VALID;
405e5b2f 1443
e636a7b4 1444 /* istate->cache_changed is updated in the caller */
405e5b2f
LT
1445 return updated;
1446}
1447
3deffc52 1448static void show_file(const char * fmt, const char * name, int in_porcelain,
046613c5 1449 int * first, const char *header_msg)
3deffc52
MM
1450{
1451 if (in_porcelain && *first && header_msg) {
1452 printf("%s\n", header_msg);
cd2b8ae9 1453 *first = 0;
3deffc52
MM
1454 }
1455 printf(fmt, name);
1456}
1457
9b2d6149
NTND
1458int refresh_index(struct index_state *istate, unsigned int flags,
1459 const struct pathspec *pathspec,
046613c5 1460 char *seen, const char *header_msg)
405e5b2f
LT
1461{
1462 int i;
1463 int has_errors = 0;
1464 int really = (flags & REFRESH_REALLY) != 0;
1465 int allow_unmerged = (flags & REFRESH_UNMERGED) != 0;
1466 int quiet = (flags & REFRESH_QUIET) != 0;
1467 int not_new = (flags & REFRESH_IGNORE_MISSING) != 0;
5fdeacb0 1468 int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0;
3deffc52
MM
1469 int first = 1;
1470 int in_porcelain = (flags & REFRESH_IN_PORCELAIN);
25762726
BK
1471 unsigned int options = (CE_MATCH_REFRESH |
1472 (really ? CE_MATCH_IGNORE_VALID : 0) |
2e2e7ec1 1473 (not_new ? CE_MATCH_IGNORE_MISSING : 0));
4bd4e730 1474 const char *modified_fmt;
73b7eae6
JK
1475 const char *deleted_fmt;
1476 const char *typechange_fmt;
1477 const char *added_fmt;
4bd4e730 1478 const char *unmerged_fmt;
405e5b2f 1479
c46c406a 1480 trace_performance_enter();
4bd4e730 1481 modified_fmt = (in_porcelain ? "M\t%s\n" : "%s: needs update\n");
73b7eae6
JK
1482 deleted_fmt = (in_porcelain ? "D\t%s\n" : "%s: needs update\n");
1483 typechange_fmt = (in_porcelain ? "T\t%s\n" : "%s needs update\n");
1484 added_fmt = (in_porcelain ? "A\t%s\n" : "%s needs update\n");
4bd4e730 1485 unmerged_fmt = (in_porcelain ? "U\t%s\n" : "%s: needs merge\n");
4aab5b46 1486 for (i = 0; i < istate->cache_nr; i++) {
285c2e25 1487 struct cache_entry *ce, *new_entry;
ec0cc704 1488 int cache_errno = 0;
73b7eae6 1489 int changed = 0;
3d1f148c 1490 int filtered = 0;
ec0cc704 1491
4aab5b46 1492 ce = istate->cache[i];
5fdeacb0
JS
1493 if (ignore_submodules && S_ISGITLINK(ce->ce_mode))
1494 continue;
1495
6d2df284 1496 if (pathspec && !ce_path_match(&the_index, ce, pathspec, seen))
3d1f148c
JH
1497 filtered = 1;
1498
405e5b2f 1499 if (ce_stage(ce)) {
4aab5b46
JH
1500 while ((i < istate->cache_nr) &&
1501 ! strcmp(istate->cache[i]->name, ce->name))
405e5b2f
LT
1502 i++;
1503 i--;
1504 if (allow_unmerged)
1505 continue;
3d1f148c
JH
1506 if (!filtered)
1507 show_file(unmerged_fmt, ce->name, in_porcelain,
1508 &first, header_msg);
405e5b2f
LT
1509 has_errors = 1;
1510 continue;
1511 }
1512
3d1f148c 1513 if (filtered)
d616813d
AJ
1514 continue;
1515
285c2e25
BW
1516 new_entry = refresh_cache_ent(istate, ce, options, &cache_errno, &changed);
1517 if (new_entry == ce)
405e5b2f 1518 continue;
285c2e25 1519 if (!new_entry) {
73b7eae6
JK
1520 const char *fmt;
1521
8fd2cb40 1522 if (really && cache_errno == EINVAL) {
405e5b2f
LT
1523 /* If we are doing --really-refresh that
1524 * means the index is not valid anymore.
1525 */
7a51ed66 1526 ce->ce_flags &= ~CE_VALID;
078a58e8 1527 ce->ce_flags |= CE_UPDATE_IN_BASE;
883e248b 1528 mark_fsmonitor_invalid(istate, ce);
e636a7b4 1529 istate->cache_changed |= CE_ENTRY_CHANGED;
405e5b2f
LT
1530 }
1531 if (quiet)
1532 continue;
73b7eae6
JK
1533
1534 if (cache_errno == ENOENT)
1535 fmt = deleted_fmt;
895ff3b2 1536 else if (ce_intent_to_add(ce))
73b7eae6
JK
1537 fmt = added_fmt; /* must be before other checks */
1538 else if (changed & TYPE_CHANGED)
1539 fmt = typechange_fmt;
1540 else
1541 fmt = modified_fmt;
1542 show_file(fmt,
1543 ce->name, in_porcelain, &first, header_msg);
405e5b2f
LT
1544 has_errors = 1;
1545 continue;
1546 }
cf558704 1547
285c2e25 1548 replace_index_entry(istate, i, new_entry);
405e5b2f 1549 }
c46c406a 1550 trace_performance_leave("refresh index");
405e5b2f
LT
1551 return has_errors;
1552}
1553
768d7965
JM
1554struct cache_entry *refresh_cache_entry(struct index_state *istate,
1555 struct cache_entry *ce,
1556 unsigned int options)
ec0cc704 1557{
768d7965 1558 return refresh_cache_ent(istate, ce, options, NULL, NULL);
ec0cc704
JH
1559}
1560
db3b313c
JH
1561
1562/*****************************************************************
1563 * Index File I/O
1564 *****************************************************************/
1565
9d227781
JH
1566#define INDEX_FORMAT_DEFAULT 3
1567
136347d7
TG
1568static unsigned int get_index_format_default(void)
1569{
1570 char *envversion = getenv("GIT_INDEX_VERSION");
3c09d684 1571 char *endp;
b27a5720 1572 int value;
3c09d684
TG
1573 unsigned int version = INDEX_FORMAT_DEFAULT;
1574
136347d7 1575 if (!envversion) {
b27a5720
TA
1576 if (!git_config_get_int("index.version", &value))
1577 version = value;
3c09d684
TG
1578 if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) {
1579 warning(_("index.version set, but the value is invalid.\n"
136347d7 1580 "Using version %i"), INDEX_FORMAT_DEFAULT);
3c09d684 1581 return INDEX_FORMAT_DEFAULT;
136347d7
TG
1582 }
1583 return version;
1584 }
3c09d684
TG
1585
1586 version = strtoul(envversion, &endp, 10);
1587 if (*endp ||
1588 version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) {
1589 warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n"
1590 "Using version %i"), INDEX_FORMAT_DEFAULT);
1591 version = INDEX_FORMAT_DEFAULT;
1592 }
1593 return version;
136347d7
TG
1594}
1595
db3b313c
JH
1596/*
1597 * dev/ino/uid/gid/size are also just tracked to the low 32 bits
1598 * Again - this is just a (very strong in practice) heuristic that
1599 * the inode hasn't changed.
1600 *
1601 * We save the fields in big-endian order to allow using the
1602 * index file over NFS transparently.
1603 */
1604struct ondisk_cache_entry {
1605 struct cache_time ctime;
1606 struct cache_time mtime;
7800c1eb
TG
1607 uint32_t dev;
1608 uint32_t ino;
1609 uint32_t mode;
1610 uint32_t uid;
1611 uint32_t gid;
1612 uint32_t size;
db3b313c 1613 unsigned char sha1[20];
7800c1eb 1614 uint16_t flags;
db3b313c
JH
1615 char name[FLEX_ARRAY]; /* more */
1616};
1617
1618/*
1619 * This struct is used when CE_EXTENDED bit is 1
1620 * The struct must match ondisk_cache_entry exactly from
1621 * ctime till flags
1622 */
1623struct ondisk_cache_entry_extended {
1624 struct cache_time ctime;
1625 struct cache_time mtime;
7800c1eb
TG
1626 uint32_t dev;
1627 uint32_t ino;
1628 uint32_t mode;
1629 uint32_t uid;
1630 uint32_t gid;
1631 uint32_t size;
db3b313c 1632 unsigned char sha1[20];
7800c1eb
TG
1633 uint16_t flags;
1634 uint16_t flags2;
db3b313c
JH
1635 char name[FLEX_ARRAY]; /* more */
1636};
1637
6c9cd161 1638/* These are only used for v3 or lower */
ce012deb 1639#define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len)
db3b313c
JH
1640#define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,name) + (len) + 8) & ~7)
1641#define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len)
1642#define ondisk_cache_entry_extended_size(len) align_flex_name(ondisk_cache_entry_extended,len)
1643#define ondisk_ce_size(ce) (((ce)->ce_flags & CE_EXTENDED) ? \
1644 ondisk_cache_entry_extended_size(ce_namelen(ce)) : \
1645 ondisk_cache_entry_size(ce_namelen(ce)))
1646
a33fc72f
JH
1647/* Allow fsck to force verification of the index checksum. */
1648int verify_index_checksum;
1649
00ec50e5
BP
1650/* Allow fsck to force verification of the cache entry order. */
1651int verify_ce_order;
1652
bad68ec9 1653static int verify_hdr(struct cache_header *hdr, unsigned long size)
e83c5163 1654{
aab61359 1655 git_hash_ctx c;
1656 unsigned char hash[GIT_MAX_RAWSZ];
0136bac9 1657 int hdr_version;
e83c5163 1658
ccc4feb5 1659 if (hdr->hdr_signature != htonl(CACHE_SIGNATURE))
e83c5163 1660 return error("bad signature");
0136bac9 1661 hdr_version = ntohl(hdr->hdr_version);
b82a7b5b 1662 if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version)
0136bac9 1663 return error("bad index version %d", hdr_version);
a33fc72f
JH
1664
1665 if (!verify_index_checksum)
1666 return 0;
1667
aab61359 1668 the_hash_algo->init_fn(&c);
1669 the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz);
1670 the_hash_algo->final_fn(hash, &c);
67947c34 1671 if (!hasheq(hash, (unsigned char *)hdr + size - the_hash_algo->rawsz))
ca9be054 1672 return error("bad index file sha1 signature");
e83c5163
LT
1673 return 0;
1674}
1675
4aab5b46
JH
1676static int read_index_extension(struct index_state *istate,
1677 const char *ext, void *data, unsigned long sz)
bad68ec9
JH
1678{
1679 switch (CACHE_EXT(ext)) {
1680 case CACHE_EXT_TREE:
4aab5b46 1681 istate->cache_tree = cache_tree_read(data, sz);
bad68ec9 1682 break;
cfc5789a
JH
1683 case CACHE_EXT_RESOLVE_UNDO:
1684 istate->resolve_undo = resolve_undo_read(data, sz);
1685 break;
5fc2fc8f
NTND
1686 case CACHE_EXT_LINK:
1687 if (read_link_extension(istate, data, sz))
1688 return -1;
1689 break;
f9e6c649
NTND
1690 case CACHE_EXT_UNTRACKED:
1691 istate->untracked = read_untracked_extension(data, sz);
1692 break;
883e248b
BP
1693 case CACHE_EXT_FSMONITOR:
1694 read_fsmonitor_extension(istate, data, sz);
1695 break;
bad68ec9
JH
1696 default:
1697 if (*ext < 'A' || 'Z' < *ext)
1698 return error("index uses %.4s extension, which we do not understand",
1699 ext);
1700 fprintf(stderr, "ignoring %.4s extension\n", ext);
1701 break;
1702 }
1703 return 0;
1704}
1705
b3e83cc7 1706int hold_locked_index(struct lock_file *lk, int lock_flags)
216aab1e 1707{
b3e83cc7 1708 return hold_lock_file_for_update(lk, get_index_file(), lock_flags);
216aab1e
MH
1709}
1710
4aab5b46 1711int read_index(struct index_state *istate)
8fd2cb40 1712{
a125a223 1713 return read_index_from(istate, get_index_file(), get_git_dir());
8fd2cb40
JS
1714}
1715
8e72d675 1716static struct cache_entry *cache_entry_from_ondisk(struct mem_pool *mem_pool,
a849735b 1717 struct ondisk_cache_entry *ondisk,
3fc22b53
JH
1718 unsigned int flags,
1719 const char *name,
1720 size_t len)
1721{
8e72d675 1722 struct cache_entry *ce = mem_pool__ce_alloc(mem_pool, len);
3fc22b53 1723
c3d8da57
JK
1724 ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec);
1725 ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec);
1726 ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec);
1727 ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec);
1728 ce->ce_stat_data.sd_dev = get_be32(&ondisk->dev);
1729 ce->ce_stat_data.sd_ino = get_be32(&ondisk->ino);
1730 ce->ce_mode = get_be32(&ondisk->mode);
1731 ce->ce_stat_data.sd_uid = get_be32(&ondisk->uid);
1732 ce->ce_stat_data.sd_gid = get_be32(&ondisk->gid);
1733 ce->ce_stat_data.sd_size = get_be32(&ondisk->size);
b60e188c
TG
1734 ce->ce_flags = flags & ~CE_NAMEMASK;
1735 ce->ce_namelen = len;
5fc2fc8f 1736 ce->index = 0;
99d1a986 1737 hashcpy(ce->oid.hash, ondisk->sha1);
3fc22b53
JH
1738 memcpy(ce->name, name, len);
1739 ce->name[len] = '\0';
1740 return ce;
1741}
1742
6c9cd161
JH
1743/*
1744 * Adjacent cache entries tend to share the leading paths, so it makes
1745 * sense to only store the differences in later entries. In the v4
1746 * on-disk format of the index, each on-disk cache entry stores the
1747 * number of bytes to be stripped from the end of the previous name,
1748 * and the bytes to append to the result, to come up with its name.
1749 */
1750static unsigned long expand_name_field(struct strbuf *name, const char *cp_)
1751{
1752 const unsigned char *ep, *cp = (const unsigned char *)cp_;
1753 size_t len = decode_varint(&cp);
1754
1755 if (name->len < len)
1756 die("malformed name field in the index");
1757 strbuf_remove(name, name->len - len, len);
1758 for (ep = cp; *ep; ep++)
1759 ; /* find the end */
1760 strbuf_add(name, cp, ep - cp);
1761 return (const char *)ep + 1 - cp_;
1762}
1763
8e72d675 1764static struct cache_entry *create_from_disk(struct mem_pool *mem_pool,
a849735b 1765 struct ondisk_cache_entry *ondisk,
6c9cd161
JH
1766 unsigned long *ent_size,
1767 struct strbuf *previous_name)
7a51ed66 1768{
debed2a6 1769 struct cache_entry *ce;
7fec10b7 1770 size_t len;
06aaaa0b 1771 const char *name;
debed2a6 1772 unsigned int flags;
7fec10b7 1773
7a51ed66 1774 /* On-disk flags are just 16 bits */
c3d8da57 1775 flags = get_be16(&ondisk->flags);
debed2a6 1776 len = flags & CE_NAMEMASK;
7fec10b7 1777
debed2a6 1778 if (flags & CE_EXTENDED) {
06aaaa0b
NTND
1779 struct ondisk_cache_entry_extended *ondisk2;
1780 int extended_flags;
1781 ondisk2 = (struct ondisk_cache_entry_extended *)ondisk;
c3d8da57 1782 extended_flags = get_be16(&ondisk2->flags2) << 16;
06aaaa0b
NTND
1783 /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */
1784 if (extended_flags & ~CE_EXTENDED_FLAGS)
1785 die("Unknown index entry format %08x", extended_flags);
debed2a6 1786 flags |= extended_flags;
06aaaa0b
NTND
1787 name = ondisk2->name;
1788 }
1789 else
1790 name = ondisk->name;
1791
6c9cd161
JH
1792 if (!previous_name) {
1793 /* v3 and earlier */
1794 if (len == CE_NAMEMASK)
1795 len = strlen(name);
8e72d675 1796 ce = cache_entry_from_ondisk(mem_pool, ondisk, flags, name, len);
6c9cd161
JH
1797
1798 *ent_size = ondisk_ce_size(ce);
1799 } else {
1800 unsigned long consumed;
1801 consumed = expand_name_field(previous_name, name);
8e72d675 1802 ce = cache_entry_from_ondisk(mem_pool, ondisk, flags,
6c9cd161
JH
1803 previous_name->buf,
1804 previous_name->len);
1805
1806 *ent_size = (name - ((char *)ondisk)) + consumed;
1807 }
debed2a6 1808 return ce;
cf558704
LT
1809}
1810
03f15a79 1811static void check_ce_order(struct index_state *istate)
15999d0b 1812{
03f15a79
TG
1813 unsigned int i;
1814
00ec50e5
BP
1815 if (!verify_ce_order)
1816 return;
1817
03f15a79
TG
1818 for (i = 1; i < istate->cache_nr; i++) {
1819 struct cache_entry *ce = istate->cache[i - 1];
1820 struct cache_entry *next_ce = istate->cache[i];
1821 int name_compare = strcmp(ce->name, next_ce->name);
1822
1823 if (0 < name_compare)
1824 die("unordered stage entries in index");
1825 if (!name_compare) {
1826 if (!ce_stage(ce))
1827 die("multiple stage entries for merged file '%s'",
1828 ce->name);
1829 if (ce_stage(ce) > ce_stage(next_ce))
1830 die("unordered stage entries for '%s'",
1831 ce->name);
1832 }
15999d0b
JSP
1833 }
1834}
1835
435ec090
CC
1836static void tweak_untracked_cache(struct index_state *istate)
1837{
1838 switch (git_config_get_untracked_cache()) {
1839 case -1: /* keep: do nothing */
1840 break;
1841 case 0: /* false */
1842 remove_untracked_cache(istate);
1843 break;
1844 case 1: /* true */
1845 add_untracked_cache(istate);
1846 break;
1847 default: /* unknown value: do nothing */
1848 break;
1849 }
1850}
1851
43925312
CC
1852static void tweak_split_index(struct index_state *istate)
1853{
1854 switch (git_config_get_split_index()) {
1855 case -1: /* unset: do nothing */
1856 break;
1857 case 0: /* false */
1858 remove_split_index(istate);
1859 break;
1860 case 1: /* true */
1861 add_split_index(istate);
1862 break;
1863 default: /* unknown value: do nothing */
1864 break;
1865 }
1866}
1867
435ec090
CC
1868static void post_read_index_from(struct index_state *istate)
1869{
1870 check_ce_order(istate);
1871 tweak_untracked_cache(istate);
43925312 1872 tweak_split_index(istate);
883e248b 1873 tweak_fsmonitor(istate);
435ec090
CC
1874}
1875
8e72d675
JM
1876static size_t estimate_cache_size_from_compressed(unsigned int entries)
1877{
1878 return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH);
1879}
1880
1881static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries)
1882{
1883 long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry);
1884
1885 /*
1886 * Account for potential alignment differences.
1887 */
1888 per_entry += align_padding_size(sizeof(struct cache_entry), -sizeof(struct ondisk_cache_entry));
1889 return ondisk_size + entries * per_entry;
1890}
1891
8fd2cb40 1892/* remember to discard_cache() before reading a different cache! */
3e52f70b 1893int do_read_index(struct index_state *istate, const char *path, int must_exist)
e83c5163
LT
1894{
1895 int fd, i;
1896 struct stat st;
debed2a6 1897 unsigned long src_offset;
e83c5163 1898 struct cache_header *hdr;
7a51ed66
LT
1899 void *mmap;
1900 size_t mmap_size;
6c9cd161 1901 struct strbuf previous_name_buf = STRBUF_INIT, *previous_name;
e83c5163 1902
913e0e99 1903 if (istate->initialized)
4aab5b46 1904 return istate->cache_nr;
5d1a5c02 1905
fba2f38a
KB
1906 istate->timestamp.sec = 0;
1907 istate->timestamp.nsec = 0;
8fd2cb40 1908 fd = open(path, O_RDONLY);
5d1a5c02 1909 if (fd < 0) {
5fc2fc8f 1910 if (!must_exist && errno == ENOENT)
5d1a5c02 1911 return 0;
5fc2fc8f 1912 die_errno("%s: index file open failed", path);
5d1a5c02 1913 }
e83c5163 1914
3511a377 1915 if (fstat(fd, &st))
d824cbba 1916 die_errno("cannot stat the open index");
3511a377 1917
7a51ed66 1918 mmap_size = xsize_t(st.st_size);
aab61359 1919 if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz)
3511a377
LFC
1920 die("index file smaller than expected");
1921
a1293ef7 1922 mmap = xmmap(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0);
7a51ed66 1923 if (mmap == MAP_FAILED)
0721c314 1924 die_errno("unable to map index file");
57d84f8d 1925 close(fd);
e83c5163 1926
7a51ed66
LT
1927 hdr = mmap;
1928 if (verify_hdr(hdr, mmap_size) < 0)
e83c5163
LT
1929 goto unmap;
1930
75691ea3 1931 hashcpy(istate->oid.hash, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz);
9d227781 1932 istate->version = ntohl(hdr->hdr_version);
4aab5b46
JH
1933 istate->cache_nr = ntohl(hdr->hdr_entries);
1934 istate->cache_alloc = alloc_nr(istate->cache_nr);
c4aa3167 1935 istate->cache = xcalloc(istate->cache_alloc, sizeof(*istate->cache));
913e0e99 1936 istate->initialized = 1;
7a51ed66 1937
8e72d675 1938 if (istate->version == 4) {
6c9cd161 1939 previous_name = &previous_name_buf;
8e72d675
JM
1940 mem_pool_init(&istate->ce_mem_pool,
1941 estimate_cache_size_from_compressed(istate->cache_nr));
1942 } else {
6c9cd161 1943 previous_name = NULL;
8e72d675
JM
1944 mem_pool_init(&istate->ce_mem_pool,
1945 estimate_cache_size(mmap_size, istate->cache_nr));
1946 }
6c9cd161 1947
7a51ed66 1948 src_offset = sizeof(*hdr);
4aab5b46 1949 for (i = 0; i < istate->cache_nr; i++) {
7a51ed66 1950 struct ondisk_cache_entry *disk_ce;
4aab5b46 1951 struct cache_entry *ce;
936f53d0 1952 unsigned long consumed;
4aab5b46 1953
7a51ed66 1954 disk_ce = (struct ondisk_cache_entry *)((char *)mmap + src_offset);
8e72d675 1955 ce = create_from_disk(istate->ce_mem_pool, disk_ce, &consumed, previous_name);
cf558704 1956 set_index_entry(istate, i, ce);
7a51ed66 1957
936f53d0 1958 src_offset += consumed;
e83c5163 1959 }
6c9cd161 1960 strbuf_release(&previous_name_buf);
fba2f38a 1961 istate->timestamp.sec = st.st_mtime;
c06ff490 1962 istate->timestamp.nsec = ST_MTIME_NSEC(st);
fba2f38a 1963
aab61359 1964 while (src_offset <= mmap_size - the_hash_algo->rawsz - 8) {
bad68ec9
JH
1965 /* After an array of active_nr index entries,
1966 * there can be arbitrary number of extended
1967 * sections, each of which is prefixed with
1968 * extension name (4-byte) and section length
1969 * in 4-byte network byte order.
1970 */
07cc8eca 1971 uint32_t extsize;
7a51ed66 1972 memcpy(&extsize, (char *)mmap + src_offset + 4, 4);
bad68ec9 1973 extsize = ntohl(extsize);
4aab5b46 1974 if (read_index_extension(istate,
7a51ed66
LT
1975 (const char *) mmap + src_offset,
1976 (char *) mmap + src_offset + 8,
1d7f171c 1977 extsize) < 0)
bad68ec9 1978 goto unmap;
7a51ed66
LT
1979 src_offset += 8;
1980 src_offset += extsize;
bad68ec9 1981 }
7a51ed66 1982 munmap(mmap, mmap_size);
4aab5b46 1983 return istate->cache_nr;
e83c5163
LT
1984
1985unmap:
7a51ed66 1986 munmap(mmap, mmap_size);
5d1a5c02 1987 die("index file corrupt");
e83c5163
LT
1988}
1989
0d59ffb4
CC
1990/*
1991 * Signal that the shared index is used by updating its mtime.
1992 *
1993 * This way, shared index can be removed if they have not been used
1994 * for some time.
1995 */
a125a223 1996static void freshen_shared_index(const char *shared_index, int warn)
0d59ffb4 1997{
0d59ffb4
CC
1998 if (!check_and_freshen_file(shared_index, 1) && warn)
1999 warning("could not freshen shared index '%s'", shared_index);
2000}
2001
a125a223
TG
2002int read_index_from(struct index_state *istate, const char *path,
2003 const char *gitdir)
5fc2fc8f
NTND
2004{
2005 struct split_index *split_index;
2006 int ret;
2182abd9 2007 char *base_oid_hex;
a125a223 2008 char *base_path;
5fc2fc8f
NTND
2009
2010 /* istate->initialized covers both .git/index and .git/sharedindex.xxx */
2011 if (istate->initialized)
2012 return istate->cache_nr;
2013
c46c406a 2014 trace_performance_enter();
5fc2fc8f 2015 ret = do_read_index(istate, path, 0);
c46c406a 2016 trace_performance_leave("read cache %s", path);
435ec090 2017
5fc2fc8f 2018 split_index = istate->split_index;
2182abd9 2019 if (!split_index || is_null_oid(&split_index->base_oid)) {
435ec090 2020 post_read_index_from(istate);
5fc2fc8f 2021 return ret;
03f15a79 2022 }
5fc2fc8f 2023
c46c406a 2024 trace_performance_enter();
5fc2fc8f
NTND
2025 if (split_index->base)
2026 discard_index(split_index->base);
2027 else
2028 split_index->base = xcalloc(1, sizeof(*split_index->base));
de6ae5f9 2029
2182abd9 2030 base_oid_hex = oid_to_hex(&split_index->base_oid);
2031 base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex);
de6ae5f9 2032 ret = do_read_index(split_index->base, base_path, 1);
9001dc2a 2033 if (!oideq(&split_index->base_oid, &split_index->base->oid))
5fc2fc8f 2034 die("broken index, expect %s in %s, got %s",
2182abd9 2035 base_oid_hex, base_path,
75691ea3 2036 oid_to_hex(&split_index->base->oid));
de6ae5f9 2037
a125a223 2038 freshen_shared_index(base_path, 0);
5fc2fc8f 2039 merge_base_index(istate);
435ec090 2040 post_read_index_from(istate);
a125a223 2041 free(base_path);
c46c406a 2042 trace_performance_leave("read cache %s", base_path);
5fc2fc8f
NTND
2043 return ret;
2044}
2045
fa7b3c2f
JH
2046int is_index_unborn(struct index_state *istate)
2047{
debed2a6 2048 return (!istate->cache_nr && !istate->timestamp.sec);
fa7b3c2f
JH
2049}
2050
4aab5b46 2051int discard_index(struct index_state *istate)
6d297f81 2052{
8e72d675
JM
2053 /*
2054 * Cache entries in istate->cache[] should have been allocated
2055 * from the memory pool associated with this index, or from an
2056 * associated split_index. There is no need to free individual
8616a2d0
JM
2057 * cache entries. validate_cache_entries can detect when this
2058 * assertion does not hold.
8e72d675 2059 */
8616a2d0 2060 validate_cache_entries(istate);
debed2a6 2061
cfc5789a 2062 resolve_undo_clear_index(istate);
4aab5b46
JH
2063 istate->cache_nr = 0;
2064 istate->cache_changed = 0;
fba2f38a
KB
2065 istate->timestamp.sec = 0;
2066 istate->timestamp.nsec = 0;
2092678c 2067 free_name_hash(istate);
4aab5b46 2068 cache_tree_free(&(istate->cache_tree));
913e0e99 2069 istate->initialized = 0;
6a83d902 2070 FREE_AND_NULL(istate->cache);
a0fc4db0 2071 istate->cache_alloc = 0;
5fc2fc8f 2072 discard_split_index(istate);
f9e6c649
NTND
2073 free_untracked_cache(istate->untracked);
2074 istate->untracked = NULL;
8e72d675
JM
2075
2076 if (istate->ce_mem_pool) {
8616a2d0 2077 mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries());
8e72d675
JM
2078 istate->ce_mem_pool = NULL;
2079 }
2080
7a51ed66 2081 return 0;
6d297f81
JS
2082}
2083
8616a2d0
JM
2084/*
2085 * Validate the cache entries of this index.
2086 * All cache entries associated with this index
2087 * should have been allocated by the memory pool
2088 * associated with this index, or by a referenced
2089 * split index.
2090 */
2091void validate_cache_entries(const struct index_state *istate)
2092{
2093 int i;
2094
2095 if (!should_validate_cache_entries() ||!istate || !istate->initialized)
2096 return;
2097
2098 for (i = 0; i < istate->cache_nr; i++) {
2099 if (!istate) {
2100 die("internal error: cache entry is not allocated from expected memory pool");
2101 } else if (!istate->ce_mem_pool ||
2102 !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) {
2103 if (!istate->split_index ||
2104 !istate->split_index->base ||
2105 !istate->split_index->base->ce_mem_pool ||
2106 !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) {
2107 die("internal error: cache entry is not allocated from expected memory pool");
2108 }
2109 }
2110 }
2111
2112 if (istate->split_index)
2113 validate_cache_entries(istate->split_index->base);
2114}
2115
d1f128b0 2116int unmerged_index(const struct index_state *istate)
94a5728c
DB
2117{
2118 int i;
2119 for (i = 0; i < istate->cache_nr; i++) {
2120 if (ce_stage(istate->cache[i]))
2121 return 1;
2122 }
2123 return 0;
2124}
2125
e1f8694f
EN
2126int index_has_changes(const struct index_state *istate,
2127 struct tree *tree,
2128 struct strbuf *sb)
cffbfad5 2129{
e1f8694f 2130 struct object_id cmp;
cffbfad5
EN
2131 int i;
2132
1b9fbefb
EN
2133 if (istate != &the_index) {
2134 BUG("index_has_changes cannot yet accept istate != &the_index; do_diff_cache needs updating first.");
2135 }
e1f8694f
EN
2136 if (tree)
2137 cmp = tree->object.oid;
2138 if (tree || !get_oid_tree("HEAD", &cmp)) {
cffbfad5
EN
2139 struct diff_options opt;
2140
2141 diff_setup(&opt);
2142 opt.flags.exit_with_status = 1;
2143 if (!sb)
2144 opt.flags.quick = 1;
e1f8694f 2145 do_diff_cache(&cmp, &opt);
cffbfad5
EN
2146 diffcore_std(&opt);
2147 for (i = 0; sb && i < diff_queued_diff.nr; i++) {
2148 if (i)
2149 strbuf_addch(sb, ' ');
2150 strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path);
2151 }
2152 diff_flush(&opt);
2153 return opt.flags.has_changes != 0;
2154 } else {
1b9fbefb 2155 for (i = 0; sb && i < istate->cache_nr; i++) {
cffbfad5
EN
2156 if (i)
2157 strbuf_addch(sb, ' ');
1b9fbefb 2158 strbuf_addstr(sb, istate->cache[i]->name);
cffbfad5 2159 }
1b9fbefb 2160 return !!istate->cache_nr;
cffbfad5
EN
2161 }
2162}
2163
4990aadc 2164#define WRITE_BUFFER_SIZE 8192
bf0f910d 2165static unsigned char write_buffer[WRITE_BUFFER_SIZE];
4990aadc
LT
2166static unsigned long write_buffer_len;
2167
aab61359 2168static int ce_write_flush(git_hash_ctx *context, int fd)
6015c28b
JH
2169{
2170 unsigned int buffered = write_buffer_len;
2171 if (buffered) {
aab61359 2172 the_hash_algo->update_fn(context, write_buffer, buffered);
06f46f23 2173 if (write_in_full(fd, write_buffer, buffered) < 0)
6015c28b
JH
2174 return -1;
2175 write_buffer_len = 0;
2176 }
2177 return 0;
2178}
2179
aab61359 2180static int ce_write(git_hash_ctx *context, int fd, void *data, unsigned int len)
4990aadc
LT
2181{
2182 while (len) {
2183 unsigned int buffered = write_buffer_len;
2184 unsigned int partial = WRITE_BUFFER_SIZE - buffered;
2185 if (partial > len)
2186 partial = len;
2187 memcpy(write_buffer + buffered, data, partial);
2188 buffered += partial;
2189 if (buffered == WRITE_BUFFER_SIZE) {
6015c28b
JH
2190 write_buffer_len = buffered;
2191 if (ce_write_flush(context, fd))
4990aadc
LT
2192 return -1;
2193 buffered = 0;
2194 }
2195 write_buffer_len = buffered;
2196 len -= partial;
1d7f171c 2197 data = (char *) data + partial;
a6080a0a
JH
2198 }
2199 return 0;
4990aadc
LT
2200}
2201
aab61359 2202static int write_index_ext_header(git_hash_ctx *context, int fd,
ac58c7b1 2203 unsigned int ext, unsigned int sz)
bad68ec9
JH
2204{
2205 ext = htonl(ext);
2206 sz = htonl(sz);
968a1d65
DR
2207 return ((ce_write(context, fd, &ext, 4) < 0) ||
2208 (ce_write(context, fd, &sz, 4) < 0)) ? -1 : 0;
bad68ec9
JH
2209}
2210
aab61359 2211static int ce_flush(git_hash_ctx *context, int fd, unsigned char *hash)
4990aadc
LT
2212{
2213 unsigned int left = write_buffer_len;
ca9be054 2214
4990aadc
LT
2215 if (left) {
2216 write_buffer_len = 0;
aab61359 2217 the_hash_algo->update_fn(context, write_buffer, left);
4990aadc 2218 }
ca9be054 2219
aab61359 2220 /* Flush first if not enough space for hash signature */
2221 if (left + the_hash_algo->rawsz > WRITE_BUFFER_SIZE) {
06f46f23 2222 if (write_in_full(fd, write_buffer, left) < 0)
2c865d9a
QH
2223 return -1;
2224 left = 0;
2225 }
2226
aab61359 2227 /* Append the hash signature at the end */
2228 the_hash_algo->final_fn(write_buffer + left, context);
2229 hashcpy(hash, write_buffer + left);
2230 left += the_hash_algo->rawsz;
06f46f23 2231 return (write_in_full(fd, write_buffer, left) < 0) ? -1 : 0;
4990aadc
LT
2232}
2233
407c8eb0
JH
2234static void ce_smudge_racily_clean_entry(struct cache_entry *ce)
2235{
2236 /*
2237 * The only thing we care about in this function is to smudge the
2238 * falsely clean entry due to touch-update-touch race, so we leave
2239 * everything else as they are. We are called for entries whose
c21d39d7 2240 * ce_stat_data.sd_mtime match the index file mtime.
c70115b4
JH
2241 *
2242 * Note that this actually does not do much for gitlinks, for
2243 * which ce_match_stat_basic() always goes to the actual
2244 * contents. The caller checks with is_racy_timestamp() which
2245 * always says "no" for gitlinks, so we are not called for them ;-)
407c8eb0
JH
2246 */
2247 struct stat st;
2248
2249 if (lstat(ce->name, &st) < 0)
2250 return;
2251 if (ce_match_stat_basic(ce, &st))
2252 return;
2253 if (ce_modified_check_fs(ce, &st)) {
4b3511b0
JH
2254 /* This is "racily clean"; smudge it. Note that this
2255 * is a tricky code. At first glance, it may appear
2256 * that it can break with this sequence:
2257 *
2258 * $ echo xyzzy >frotz
2259 * $ git-update-index --add frotz
2260 * $ : >frotz
2261 * $ sleep 3
2262 * $ echo filfre >nitfol
2263 * $ git-update-index --add nitfol
2264 *
b7e58b17 2265 * but it does not. When the second update-index runs,
4b3511b0
JH
2266 * it notices that the entry "frotz" has the same timestamp
2267 * as index, and if we were to smudge it by resetting its
2268 * size to zero here, then the object name recorded
2269 * in index is the 6-byte file but the cached stat information
2270 * becomes zero --- which would then match what we would
a6080a0a 2271 * obtain from the filesystem next time we stat("frotz").
4b3511b0
JH
2272 *
2273 * However, the second update-index, before calling
2274 * this function, notices that the cached size is 6
2275 * bytes and what is on the filesystem is an empty
2276 * file, and never calls us, so the cached size information
2277 * for "frotz" stays 6 which does not match the filesystem.
2278 */
c21d39d7 2279 ce->ce_stat_data.sd_size = 0;
407c8eb0
JH
2280 }
2281}
2282
f136f7bf 2283/* Copy miscellaneous fields but not the name */
ce012deb 2284static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk,
f136f7bf 2285 struct cache_entry *ce)
7a51ed66 2286{
b60e188c
TG
2287 short flags;
2288
c21d39d7
MH
2289 ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec);
2290 ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec);
2291 ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec);
2292 ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec);
2293 ondisk->dev = htonl(ce->ce_stat_data.sd_dev);
2294 ondisk->ino = htonl(ce->ce_stat_data.sd_ino);
7a51ed66 2295 ondisk->mode = htonl(ce->ce_mode);
c21d39d7
MH
2296 ondisk->uid = htonl(ce->ce_stat_data.sd_uid);
2297 ondisk->gid = htonl(ce->ce_stat_data.sd_gid);
2298 ondisk->size = htonl(ce->ce_stat_data.sd_size);
99d1a986 2299 hashcpy(ondisk->sha1, ce->oid.hash);
b60e188c 2300
ce51bf09 2301 flags = ce->ce_flags & ~CE_NAMEMASK;
b60e188c
TG
2302 flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce));
2303 ondisk->flags = htons(flags);
06aaaa0b
NTND
2304 if (ce->ce_flags & CE_EXTENDED) {
2305 struct ondisk_cache_entry_extended *ondisk2;
2306 ondisk2 = (struct ondisk_cache_entry_extended *)ondisk;
2307 ondisk2->flags2 = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16);
f136f7bf
JH
2308 }
2309}
2310
aab61359 2311static int ce_write_entry(git_hash_ctx *c, int fd, struct cache_entry *ce,
ce012deb 2312 struct strbuf *previous_name, struct ondisk_cache_entry *ondisk)
f136f7bf 2313{
9d227781 2314 int size;
f136f7bf 2315 int result;
00a4b035
RJ
2316 unsigned int saved_namelen;
2317 int stripped_name = 0;
ce012deb 2318 static unsigned char padding[8] = { 0x00 };
f136f7bf 2319
b3c96fb1
NTND
2320 if (ce->ce_flags & CE_STRIP_NAME) {
2321 saved_namelen = ce_namelen(ce);
2322 ce->ce_namelen = 0;
00a4b035 2323 stripped_name = 1;
b3c96fb1
NTND
2324 }
2325
ce012deb
KW
2326 if (ce->ce_flags & CE_EXTENDED)
2327 size = offsetof(struct ondisk_cache_entry_extended, name);
2328 else
2329 size = offsetof(struct ondisk_cache_entry, name);
2330
9d227781 2331 if (!previous_name) {
ce012deb
KW
2332 int len = ce_namelen(ce);
2333 copy_cache_entry_to_ondisk(ondisk, ce);
2334 result = ce_write(c, fd, ondisk, size);
2335 if (!result)
2336 result = ce_write(c, fd, ce->name, len);
2337 if (!result)
2338 result = ce_write(c, fd, padding, align_padding_size(size, len));
9d227781
JH
2339 } else {
2340 int common, to_remove, prefix_size;
2341 unsigned char to_remove_vi[16];
2342 for (common = 0;
2343 (ce->name[common] &&
2344 common < previous_name->len &&
2345 ce->name[common] == previous_name->buf[common]);
2346 common++)
2347 ; /* still matching */
2348 to_remove = previous_name->len - common;
2349 prefix_size = encode_varint(to_remove, to_remove_vi);
2350
ce012deb
KW
2351 copy_cache_entry_to_ondisk(ondisk, ce);
2352 result = ce_write(c, fd, ondisk, size);
2353 if (!result)
2354 result = ce_write(c, fd, to_remove_vi, prefix_size);
2355 if (!result)
0b90b881
TG
2356 result = ce_write(c, fd, ce->name + common, ce_namelen(ce) - common);
2357 if (!result)
2358 result = ce_write(c, fd, padding, 1);
9d227781
JH
2359
2360 strbuf_splice(previous_name, common, to_remove,
2361 ce->name + common, ce_namelen(ce) - common);
06aaaa0b 2362 }
00a4b035 2363 if (stripped_name) {
b3c96fb1
NTND
2364 ce->ce_namelen = saved_namelen;
2365 ce->ce_flags &= ~CE_STRIP_NAME;
2366 }
7a51ed66 2367
59efba64 2368 return result;
7a51ed66
LT
2369}
2370
426ddeea
YM
2371/*
2372 * This function verifies if index_state has the correct sha1 of the
2373 * index file. Don't die if we have any other failure, just return 0.
2374 */
2375static int verify_index_from(const struct index_state *istate, const char *path)
2376{
2377 int fd;
2378 ssize_t n;
2379 struct stat st;
aab61359 2380 unsigned char hash[GIT_MAX_RAWSZ];
426ddeea
YM
2381
2382 if (!istate->initialized)
2383 return 0;
2384
2385 fd = open(path, O_RDONLY);
2386 if (fd < 0)
2387 return 0;
2388
2389 if (fstat(fd, &st))
2390 goto out;
2391
aab61359 2392 if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz)
426ddeea
YM
2393 goto out;
2394
aab61359 2395 n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz);
2396 if (n != the_hash_algo->rawsz)
426ddeea
YM
2397 goto out;
2398
67947c34 2399 if (!hasheq(istate->oid.hash, hash))
426ddeea
YM
2400 goto out;
2401
2402 close(fd);
2403 return 1;
2404
2405out:
2406 close(fd);
2407 return 0;
2408}
2409
2410static int verify_index(const struct index_state *istate)
2411{
2412 return verify_index_from(istate, get_index_file());
2413}
2414
483fbe2b
JH
2415static int has_racy_timestamp(struct index_state *istate)
2416{
2417 int entries = istate->cache_nr;
2418 int i;
2419
2420 for (i = 0; i < entries; i++) {
2421 struct cache_entry *ce = istate->cache[i];
2422 if (is_racy_timestamp(istate, ce))
2423 return 1;
2424 }
2425 return 0;
2426}
2427
ccdc4ec3
JH
2428void update_index_if_able(struct index_state *istate, struct lock_file *lockfile)
2429{
483fbe2b 2430 if ((istate->cache_changed || has_racy_timestamp(istate)) &&
df60cf57
2431 verify_index(istate))
2432 write_locked_index(istate, lockfile, COMMIT_LOCK);
b74c90fb
2433 else
2434 rollback_lock_file(lockfile);
ccdc4ec3
JH
2435}
2436
812d6b00
2437/*
2438 * On success, `tempfile` is closed. If it is the temporary file
2439 * of a `struct lock_file`, we will therefore effectively perform
2440 * a 'close_lock_file_gently()`. Since that is an implementation
2441 * detail of lockfiles, callers of `do_write_index()` should not
2442 * rely on it.
2443 */
9f41c7a6 2444static int do_write_index(struct index_state *istate, struct tempfile *tempfile,
c18b80a0 2445 int strip_extensions)
197ee8c9 2446{
ca54d9ba 2447 uint64_t start = getnanotime();
9f41c7a6 2448 int newfd = tempfile->fd;
aab61359 2449 git_hash_ctx c;
197ee8c9 2450 struct cache_header hdr;
b50386c7 2451 int i, err = 0, removed, extended, hdr_version;
4aab5b46
JH
2452 struct cache_entry **cache = istate->cache;
2453 int entries = istate->cache_nr;
e1afca4f 2454 struct stat st;
ce012deb 2455 struct ondisk_cache_entry_extended ondisk;
9d227781 2456 struct strbuf previous_name_buf = STRBUF_INIT, *previous_name;
4bddd983 2457 int drop_cache_tree = istate->drop_cache_tree;
025a0709 2458
06aaaa0b 2459 for (i = removed = extended = 0; i < entries; i++) {
7a51ed66 2460 if (cache[i]->ce_flags & CE_REMOVE)
025a0709 2461 removed++;
197ee8c9 2462
06aaaa0b
NTND
2463 /* reduce extended entries if possible */
2464 cache[i]->ce_flags &= ~CE_EXTENDED;
2465 if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) {
2466 extended++;
2467 cache[i]->ce_flags |= CE_EXTENDED;
2468 }
2469 }
2470
d6e3c181 2471 if (!istate->version) {
136347d7 2472 istate->version = get_index_format_default();
4c2db938 2473 if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0))
d6e3c181
NTND
2474 init_split_index(istate);
2475 }
9d227781
JH
2476
2477 /* demote version 3 to version 2 when the latter suffices */
2478 if (istate->version == 3 || istate->version == 2)
2479 istate->version = extended ? 3 : 2;
2480
2481 hdr_version = istate->version;
2482
ccc4feb5 2483 hdr.hdr_signature = htonl(CACHE_SIGNATURE);
9d227781 2484 hdr.hdr_version = htonl(hdr_version);
025a0709 2485 hdr.hdr_entries = htonl(entries - removed);
197ee8c9 2486
aab61359 2487 the_hash_algo->init_fn(&c);
ca9be054 2488 if (ce_write(&c, newfd, &hdr, sizeof(hdr)) < 0)
197ee8c9
LT
2489 return -1;
2490
9d227781 2491 previous_name = (hdr_version == 4) ? &previous_name_buf : NULL;
ce012deb 2492
197ee8c9
LT
2493 for (i = 0; i < entries; i++) {
2494 struct cache_entry *ce = cache[i];
7a51ed66 2495 if (ce->ce_flags & CE_REMOVE)
aa16021e 2496 continue;
e06c43c7 2497 if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce))
407c8eb0 2498 ce_smudge_racily_clean_entry(ce);
99d1a986 2499 if (is_null_oid(&ce->oid)) {
83bd7437
JK
2500 static const char msg[] = "cache entry has null sha1: %s";
2501 static int allow = -1;
2502
2503 if (allow < 0)
2504 allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0);
2505 if (allow)
2506 warning(msg, ce->name);
2507 else
b50386c7 2508 err = error(msg, ce->name);
a96d3cc3
JK
2509
2510 drop_cache_tree = 1;
83bd7437 2511 }
ce012deb 2512 if (ce_write_entry(&c, newfd, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0)
b50386c7
KW
2513 err = -1;
2514
2515 if (err)
2516 break;
197ee8c9 2517 }
9d227781 2518 strbuf_release(&previous_name_buf);
1af1c2b6 2519
b50386c7
KW
2520 if (err)
2521 return err;
2522
bad68ec9 2523 /* Write extension data here */
c18b80a0 2524 if (!strip_extensions && istate->split_index) {
5fc2fc8f
NTND
2525 struct strbuf sb = STRBUF_INIT;
2526
2527 err = write_link_extension(&sb, istate) < 0 ||
2528 write_index_ext_header(&c, newfd, CACHE_EXT_LINK,
2529 sb.len) < 0 ||
2530 ce_write(&c, newfd, sb.buf, sb.len) < 0;
2531 strbuf_release(&sb);
2532 if (err)
2533 return -1;
2534 }
a96d3cc3 2535 if (!strip_extensions && !drop_cache_tree && istate->cache_tree) {
f285a2d7 2536 struct strbuf sb = STRBUF_INIT;
1dffb8fa 2537
1dffb8fa
PH
2538 cache_tree_write(&sb, istate->cache_tree);
2539 err = write_index_ext_header(&c, newfd, CACHE_EXT_TREE, sb.len) < 0
2540 || ce_write(&c, newfd, sb.buf, sb.len) < 0;
2541 strbuf_release(&sb);
2542 if (err)
bad68ec9 2543 return -1;
bad68ec9 2544 }
c18b80a0 2545 if (!strip_extensions && istate->resolve_undo) {
cfc5789a
JH
2546 struct strbuf sb = STRBUF_INIT;
2547
2548 resolve_undo_write(&sb, istate->resolve_undo);
2549 err = write_index_ext_header(&c, newfd, CACHE_EXT_RESOLVE_UNDO,
2550 sb.len) < 0
2551 || ce_write(&c, newfd, sb.buf, sb.len) < 0;
2552 strbuf_release(&sb);
2553 if (err)
2554 return -1;
2555 }
83c094ad
NTND
2556 if (!strip_extensions && istate->untracked) {
2557 struct strbuf sb = STRBUF_INIT;
2558
2559 write_untracked_extension(&sb, istate->untracked);
2560 err = write_index_ext_header(&c, newfd, CACHE_EXT_UNTRACKED,
2561 sb.len) < 0 ||
2562 ce_write(&c, newfd, sb.buf, sb.len) < 0;
2563 strbuf_release(&sb);
2564 if (err)
2565 return -1;
2566 }
883e248b
BP
2567 if (!strip_extensions && istate->fsmonitor_last_update) {
2568 struct strbuf sb = STRBUF_INIT;
2569
2570 write_fsmonitor_extension(&sb, istate);
2571 err = write_index_ext_header(&c, newfd, CACHE_EXT_FSMONITOR, sb.len) < 0
2572 || ce_write(&c, newfd, sb.buf, sb.len) < 0;
2573 strbuf_release(&sb);
2574 if (err)
2575 return -1;
2576 }
e1afca4f 2577
75691ea3 2578 if (ce_flush(&c, newfd, istate->oid.hash))
9f41c7a6 2579 return -1;
49bd0fc2
JK
2580 if (close_tempfile_gently(tempfile)) {
2581 error(_("could not close '%s'"), tempfile->filename.buf);
49bd0fc2
JK
2582 return -1;
2583 }
9f41c7a6 2584 if (stat(tempfile->filename.buf, &st))
e1afca4f 2585 return -1;
5bcf109c
KB
2586 istate->timestamp.sec = (unsigned int)st.st_mtime;
2587 istate->timestamp.nsec = ST_MTIME_NSEC(st);
ca54d9ba 2588 trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed);
e1afca4f 2589 return 0;
197ee8c9 2590}
e46bbcf6 2591
626f35c8
NTND
2592void set_alternate_index_output(const char *name)
2593{
2594 alternate_index_output = name;
2595}
2596
2597static int commit_locked_index(struct lock_file *lk)
2598{
751baced
MH
2599 if (alternate_index_output)
2600 return commit_lock_file_to(lk, alternate_index_output);
2601 else
626f35c8 2602 return commit_lock_file(lk);
626f35c8
NTND
2603}
2604
03b86647
NTND
2605static int do_write_locked_index(struct index_state *istate, struct lock_file *lock,
2606 unsigned flags)
2607{
076aa2cb 2608 int ret = do_write_index(istate, lock->tempfile, 0);
03b86647
NTND
2609 if (ret)
2610 return ret;
03b86647
NTND
2611 if (flags & COMMIT_LOCK)
2612 return commit_locked_index(lock);
812d6b00 2613 return close_lock_file_gently(lock);
03b86647
NTND
2614}
2615
5fc2fc8f
NTND
2616static int write_split_index(struct index_state *istate,
2617 struct lock_file *lock,
2618 unsigned flags)
2619{
2620 int ret;
2621 prepare_to_write_split_index(istate);
2622 ret = do_write_locked_index(istate, lock, flags);
2623 finish_writing_split_index(istate);
2624 return ret;
2625}
2626
b9683722
CC
2627static const char *shared_index_expire = "2.weeks.ago";
2628
2629static unsigned long get_shared_index_expire_date(void)
2630{
2631 static unsigned long shared_index_expire_date;
2632 static int shared_index_expire_date_prepared;
2633
2634 if (!shared_index_expire_date_prepared) {
2635 git_config_get_expiry("splitindex.sharedindexexpire",
2636 &shared_index_expire);
2637 shared_index_expire_date = approxidate(shared_index_expire);
2638 shared_index_expire_date_prepared = 1;
2639 }
2640
2641 return shared_index_expire_date;
2642}
2643
2644static int should_delete_shared_index(const char *shared_index_path)
2645{
2646 struct stat st;
2647 unsigned long expiration;
2648
2649 /* Check timestamp */
2650 expiration = get_shared_index_expire_date();
2651 if (!expiration)
2652 return 0;
2653 if (stat(shared_index_path, &st))
78bde923 2654 return error_errno(_("could not stat '%s'"), shared_index_path);
b9683722
CC
2655 if (st.st_mtime > expiration)
2656 return 0;
2657
2658 return 1;
2659}
2660
2661static int clean_shared_index_files(const char *current_hex)
2662{
2663 struct dirent *de;
2664 DIR *dir = opendir(get_git_dir());
2665
2666 if (!dir)
2667 return error_errno(_("unable to open git dir: %s"), get_git_dir());
2668
2669 while ((de = readdir(dir)) != NULL) {
2670 const char *sha1_hex;
2671 const char *shared_index_path;
2672 if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex))
2673 continue;
2674 if (!strcmp(sha1_hex, current_hex))
2675 continue;
2676 shared_index_path = git_path("%s", de->d_name);
2677 if (should_delete_shared_index(shared_index_path) > 0 &&
2678 unlink(shared_index_path))
2679 warning_errno(_("unable to unlink: %s"), shared_index_path);
2680 }
2681 closedir(dir);
2682
2683 return 0;
2684}
2685
a0a96756 2686static int write_shared_index(struct index_state *istate,
59f9d2dd 2687 struct tempfile **temp)
c18b80a0
NTND
2688{
2689 struct split_index *si = istate->split_index;
076aa2cb 2690 int ret;
c18b80a0 2691
c18b80a0 2692 move_cache_to_base_index(istate);
7db2d08c 2693 ret = do_write_index(si->base, *temp, 1);
59f9d2dd 2694 if (ret)
c18b80a0 2695 return ret;
7db2d08c 2696 ret = adjust_shared_perm(get_tempfile_path(*temp));
df801f3f 2697 if (ret) {
7db2d08c 2698 error("cannot fix permission bits on %s", get_tempfile_path(*temp));
df801f3f
CC
2699 return ret;
2700 }
7db2d08c 2701 ret = rename_tempfile(temp,
75691ea3 2702 git_path("sharedindex.%s", oid_to_hex(&si->base->oid)));
b9683722 2703 if (!ret) {
75691ea3 2704 oidcpy(&si->base_oid, &si->base->oid);
2705 clean_shared_index_files(oid_to_hex(&si->base->oid));
b9683722
CC
2706 }
2707
c18b80a0
NTND
2708 return ret;
2709}
2710
e6a1dd77
CC
2711static const int default_max_percent_split_change = 20;
2712
2713static int too_many_not_shared_entries(struct index_state *istate)
2714{
2715 int i, not_shared = 0;
2716 int max_split = git_config_get_max_percent_split_change();
2717
2718 switch (max_split) {
2719 case -1:
2720 /* not or badly configured: use the default value */
2721 max_split = default_max_percent_split_change;
2722 break;
2723 case 0:
2724 return 1; /* 0% means always write a new shared index */
2725 case 100:
2726 return 0; /* 100% means never write a new shared index */
2727 default:
2728 break; /* just use the configured value */
2729 }
2730
2731 /* Count not shared entries */
2732 for (i = 0; i < istate->cache_nr; i++) {
2733 struct cache_entry *ce = istate->cache[i];
2734 if (!ce->index)
2735 not_shared++;
2736 }
2737
2738 return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100;
2739}
2740
03b86647
NTND
2741int write_locked_index(struct index_state *istate, struct lock_file *lock,
2742 unsigned flags)
2743{
0d59ffb4 2744 int new_shared_index, ret;
5fc2fc8f
NTND
2745 struct split_index *si = istate->split_index;
2746
4592e608
NTND
2747 if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0))
2748 cache_tree_verify(istate);
2749
61000814
2750 if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) {
2751 if (flags & COMMIT_LOCK)
2752 rollback_lock_file(lock);
2753 return 0;
2754 }
2755
3bd28eb2
AV
2756 if (istate->fsmonitor_last_update)
2757 fill_fsmonitor_bitmap(istate);
2758
5165dd59
NTND
2759 if (!si || alternate_index_output ||
2760 (istate->cache_changed & ~EXTMASK)) {
5fc2fc8f 2761 if (si)
2182abd9 2762 oidclr(&si->base_oid);
df60cf57
2763 ret = do_write_locked_index(istate, lock, flags);
2764 goto out;
5fc2fc8f
NTND
2765 }
2766
4c2db938 2767 if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0)) {
2182abd9 2768 int v = si->base_oid.hash[0];
d6e3c181
NTND
2769 if ((v & 15) < 6)
2770 istate->cache_changed |= SPLIT_INDEX_ORDERED;
2771 }
e6a1dd77
CC
2772 if (too_many_not_shared_entries(istate))
2773 istate->cache_changed |= SPLIT_INDEX_ORDERED;
0d59ffb4
CC
2774
2775 new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED;
2776
2777 if (new_shared_index) {
59f9d2dd
NTND
2778 struct tempfile *temp;
2779 int saved_errno;
2780
2781 temp = mks_tempfile(git_path("sharedindex_XXXXXX"));
2782 if (!temp) {
2182abd9 2783 oidclr(&si->base_oid);
59f9d2dd 2784 ret = do_write_locked_index(istate, lock, flags);
ef5b3a6c
NTND
2785 goto out;
2786 }
2787 ret = write_shared_index(istate, &temp);
59f9d2dd
NTND
2788
2789 saved_errno = errno;
2790 if (is_tempfile_active(temp))
2791 delete_tempfile(&temp);
2792 errno = saved_errno;
2793
c18b80a0 2794 if (ret)
df60cf57 2795 goto out;
c18b80a0
NTND
2796 }
2797
0d59ffb4
CC
2798 ret = write_split_index(istate, lock, flags);
2799
2800 /* Freshen the shared index only if the split-index was written */
a125a223
TG
2801 if (!ret && !new_shared_index) {
2802 const char *shared_index = git_path("sharedindex.%s",
2182abd9 2803 oid_to_hex(&si->base_oid));
a125a223
TG
2804 freshen_shared_index(shared_index, 1);
2805 }
0d59ffb4 2806
df60cf57
2807out:
2808 if (flags & COMMIT_LOCK)
2809 rollback_lock_file(lock);
0d59ffb4 2810 return ret;
03b86647
NTND
2811}
2812
e46bbcf6
MV
2813/*
2814 * Read the index file that is potentially unmerged into given
ad376204
EN
2815 * index_state, dropping any unmerged entries to stage #0 (potentially
2816 * resulting in a path appearing as both a file and a directory in the
2817 * index; the caller is responsible to clear out the extra entries
2818 * before writing the index to a tree). Returns true if the index is
2819 * unmerged. Callers who want to refuse to work from an unmerged
2820 * state can call this and check its return value, instead of calling
2821 * read_cache().
e46bbcf6
MV
2822 */
2823int read_index_unmerged(struct index_state *istate)
2824{
2825 int i;
d1a43f2a 2826 int unmerged = 0;
e46bbcf6
MV
2827
2828 read_index(istate);
e46bbcf6
MV
2829 for (i = 0; i < istate->cache_nr; i++) {
2830 struct cache_entry *ce = istate->cache[i];
d1a43f2a 2831 struct cache_entry *new_ce;
a849735b 2832 int len;
d1a43f2a
JH
2833
2834 if (!ce_stage(ce))
e46bbcf6 2835 continue;
d1a43f2a 2836 unmerged = 1;
68c4f6a5 2837 len = ce_namelen(ce);
a849735b 2838 new_ce = make_empty_cache_entry(istate, len);
d1a43f2a 2839 memcpy(new_ce->name, ce->name, len);
b60e188c
TG
2840 new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED;
2841 new_ce->ce_namelen = len;
d1a43f2a 2842 new_ce->ce_mode = ce->ce_mode;
ad376204 2843 if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK))
d1a43f2a 2844 return error("%s: cannot drop to stage #0",
5699d17e 2845 new_ce->name);
e46bbcf6 2846 }
d1a43f2a 2847 return unmerged;
e46bbcf6 2848}
041aee31 2849
98fa4738
JK
2850/*
2851 * Returns 1 if the path is an "other" path with respect to
2852 * the index; that is, the path is not mentioned in the index at all,
2853 * either as a file, a directory with some files in the index,
2854 * or as an unmerged entry.
2855 *
2856 * We helpfully remove a trailing "/" from directories so that
2857 * the output of read_directory can be used as-is.
2858 */
2859int index_name_is_other(const struct index_state *istate, const char *name,
2860 int namelen)
2861{
2862 int pos;
2863 if (namelen && name[namelen - 1] == '/')
2864 namelen--;
2865 pos = index_name_pos(istate, name, namelen);
2866 if (0 <= pos)
2867 return 0; /* exact match */
2868 pos = -pos - 1;
2869 if (pos < istate->cache_nr) {
2870 struct cache_entry *ce = istate->cache[pos];
2871 if (ce_namelen(ce) == namelen &&
2872 !memcmp(ce->name, name, namelen))
2873 return 0; /* Yup, this one exists unmerged */
2874 }
2875 return 1;
2876}
29fb37b2 2877
87542508
BW
2878void *read_blob_data_from_index(const struct index_state *istate,
2879 const char *path, unsigned long *size)
29fb37b2
LF
2880{
2881 int pos, len;
2882 unsigned long sz;
2883 enum object_type type;
2884 void *data;
2885
2886 len = strlen(path);
2887 pos = index_name_pos(istate, path, len);
2888 if (pos < 0) {
2889 /*
2890 * We might be in the middle of a merge, in which
2891 * case we would read stage #2 (ours).
2892 */
2893 int i;
2894 for (i = -pos - 1;
2895 (pos < 0 && i < istate->cache_nr &&
2896 !strcmp(istate->cache[i]->name, path));
2897 i++)
2898 if (ce_stage(istate->cache[i]) == 2)
2899 pos = i;
2900 }
2901 if (pos < 0)
2902 return NULL;
b4f5aca4 2903 data = read_object_file(&istate->cache[pos]->oid, &type, &sz);
29fb37b2
LF
2904 if (!data || type != OBJ_BLOB) {
2905 free(data);
2906 return NULL;
2907 }
ff366825
LF
2908 if (size)
2909 *size = sz;
29fb37b2
LF
2910 return data;
2911}
38612532
MH
2912
2913void stat_validity_clear(struct stat_validity *sv)
2914{
6a83d902 2915 FREE_AND_NULL(sv->sd);
38612532
MH
2916}
2917
2918int stat_validity_check(struct stat_validity *sv, const char *path)
2919{
2920 struct stat st;
2921
2922 if (stat(path, &st) < 0)
2923 return sv->sd == NULL;
2924 if (!sv->sd)
2925 return 0;
2926 return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st);
2927}
2928
2929void stat_validity_update(struct stat_validity *sv, int fd)
2930{
2931 struct stat st;
2932
2933 if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode))
2934 stat_validity_clear(sv);
2935 else {
2936 if (!sv->sd)
2937 sv->sd = xcalloc(1, sizeof(struct stat_data));
2938 fill_stat_data(sv->sd, &st);
2939 }
2940}
edf3b905
DT
2941
2942void move_index_extensions(struct index_state *dst, struct index_state *src)
2943{
2944 dst->untracked = src->untracked;
2945 src->untracked = NULL;
836ef2b6
NTND
2946 dst->cache_tree = src->cache_tree;
2947 src->cache_tree = NULL;
edf3b905 2948}
a849735b 2949
8e72d675
JM
2950struct cache_entry *dup_cache_entry(const struct cache_entry *ce,
2951 struct index_state *istate)
2952{
2953 unsigned int size = ce_size(ce);
2954 int mem_pool_allocated;
2955 struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce));
2956 mem_pool_allocated = new_entry->mem_pool_allocated;
2957
2958 memcpy(new_entry, ce, size);
2959 new_entry->mem_pool_allocated = mem_pool_allocated;
2960 return new_entry;
2961}
2962
a849735b
JM
2963void discard_cache_entry(struct cache_entry *ce)
2964{
8616a2d0
JM
2965 if (ce && should_validate_cache_entries())
2966 memset(ce, 0xCD, cache_entry_size(ce->ce_namelen));
2967
8e72d675
JM
2968 if (ce && ce->mem_pool_allocated)
2969 return;
2970
a849735b
JM
2971 free(ce);
2972}
8616a2d0
JM
2973
2974int should_validate_cache_entries(void)
2975{
2976 static int validate_index_cache_entries = -1;
2977
2978 if (validate_index_cache_entries < 0) {
2979 if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES"))
2980 validate_index_cache_entries = 1;
2981 else
2982 validate_index_cache_entries = 0;
2983 }
2984
2985 return validate_index_cache_entries;
2986}