Git 1.9.0
[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"
bad68ec9 8#include "cache-tree.h"
f35a6d3b 9#include "refs.h"
d616813d 10#include "dir.h"
041aee31
JH
11#include "tree.h"
12#include "commit.h"
39425819 13#include "blob.h"
cfc5789a 14#include "resolve-undo.h"
6c9cd161
JH
15#include "strbuf.h"
16#include "varint.h"
bad68ec9 17
87b29e5a
JH
18static struct cache_entry *refresh_cache_entry(struct cache_entry *ce, int really);
19
b60e188c
TG
20/* Mask for the name length in ce_flags in the on-disk index */
21
22#define CE_NAMEMASK (0x0fff)
23
bad68ec9
JH
24/* Index extensions.
25 *
26 * The first letter should be 'A'..'Z' for extensions that are not
27 * necessary for a correct operation (i.e. optimization data).
28 * When new extensions are added that _needs_ to be understood in
29 * order to correctly interpret the index file, pick character that
30 * is outside the range, to cause the reader to abort.
31 */
32
33#define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) )
34#define CACHE_EXT_TREE 0x54524545 /* "TREE" */
b659b49b 35#define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */
e83c5163 36
228e94f9 37struct index_state the_index;
8fd2cb40 38
9cb76b8c
JH
39static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce)
40{
41 istate->cache[nr] = ce;
96872bc2 42 add_name_hash(istate, ce);
9cb76b8c
JH
43}
44
cf558704
LT
45static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce)
46{
47 struct cache_entry *old = istate->cache[nr];
48
2092678c 49 remove_name_hash(istate, old);
a22c6371 50 set_index_entry(istate, nr, ce);
cf558704
LT
51 istate->cache_changed = 1;
52}
53
81dc2307
PB
54void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name)
55{
56 struct cache_entry *old = istate->cache[nr], *new;
57 int namelen = strlen(new_name);
58
59 new = xmalloc(cache_entry_size(namelen));
60 copy_cache_entry(new, old);
b60e188c
TG
61 new->ce_flags &= ~CE_STATE_MASK;
62 new->ce_namelen = namelen;
81dc2307
PB
63 memcpy(new->name, new_name, namelen + 1);
64
65 cache_tree_invalidate_path(istate->cache_tree, old->name);
66 remove_index_entry_at(istate, nr);
67 add_index_entry(istate, new, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
68}
69
c21d39d7
MH
70void fill_stat_data(struct stat_data *sd, struct stat *st)
71{
72 sd->sd_ctime.sec = (unsigned int)st->st_ctime;
73 sd->sd_mtime.sec = (unsigned int)st->st_mtime;
74 sd->sd_ctime.nsec = ST_CTIME_NSEC(*st);
75 sd->sd_mtime.nsec = ST_MTIME_NSEC(*st);
76 sd->sd_dev = st->st_dev;
77 sd->sd_ino = st->st_ino;
78 sd->sd_uid = st->st_uid;
79 sd->sd_gid = st->st_gid;
80 sd->sd_size = st->st_size;
81}
82
83int match_stat_data(const struct stat_data *sd, struct stat *st)
84{
85 int changed = 0;
86
87 if (sd->sd_mtime.sec != (unsigned int)st->st_mtime)
88 changed |= MTIME_CHANGED;
89 if (trust_ctime && check_stat &&
90 sd->sd_ctime.sec != (unsigned int)st->st_ctime)
91 changed |= CTIME_CHANGED;
92
93#ifdef USE_NSEC
94 if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st))
95 changed |= MTIME_CHANGED;
96 if (trust_ctime && check_stat &&
97 sd->sd_ctime.nsec != ST_CTIME_NSEC(*st))
98 changed |= CTIME_CHANGED;
99#endif
100
101 if (check_stat) {
102 if (sd->sd_uid != (unsigned int) st->st_uid ||
103 sd->sd_gid != (unsigned int) st->st_gid)
104 changed |= OWNER_CHANGED;
105 if (sd->sd_ino != (unsigned int) st->st_ino)
106 changed |= INODE_CHANGED;
107 }
108
109#ifdef USE_STDEV
110 /*
111 * st_dev breaks on network filesystems where different
112 * clients will have different views of what "device"
113 * the filesystem is on
114 */
115 if (check_stat && sd->sd_dev != (unsigned int) st->st_dev)
116 changed |= INODE_CHANGED;
117#endif
118
119 if (sd->sd_size != (unsigned int) st->st_size)
120 changed |= DATA_CHANGED;
121
122 return changed;
123}
124
415e96c8
JH
125/*
126 * This only updates the "non-critical" parts of the directory
127 * cache, ie the parts that aren't tracked by GIT, and only used
128 * to validate the cache.
129 */
130void fill_stat_cache_info(struct cache_entry *ce, struct stat *st)
131{
c21d39d7 132 fill_stat_data(&ce->ce_stat_data, st);
5f73076c
JH
133
134 if (assume_unchanged)
7a51ed66 135 ce->ce_flags |= CE_VALID;
eadb5831
JH
136
137 if (S_ISREG(st->st_mode))
138 ce_mark_uptodate(ce);
415e96c8
JH
139}
140
21a6b9fa 141static int ce_compare_data(const struct cache_entry *ce, struct stat *st)
29e4d363
JH
142{
143 int match = -1;
144 int fd = open(ce->name, O_RDONLY);
145
146 if (fd >= 0) {
147 unsigned char sha1[20];
c4ce46fc 148 if (!index_fd(sha1, fd, st, OBJ_BLOB, ce->name, 0))
a89fccd2 149 match = hashcmp(sha1, ce->sha1);
7f8508e8 150 /* index_fd() closed the file descriptor already */
29e4d363
JH
151 }
152 return match;
153}
154
21a6b9fa 155static int ce_compare_link(const struct cache_entry *ce, size_t expected_size)
29e4d363
JH
156{
157 int match = -1;
29e4d363
JH
158 void *buffer;
159 unsigned long size;
21666f1a 160 enum object_type type;
a60272b3 161 struct strbuf sb = STRBUF_INIT;
29e4d363 162
a60272b3 163 if (strbuf_readlink(&sb, ce->name, expected_size))
29e4d363 164 return -1;
a60272b3 165
21666f1a 166 buffer = read_sha1_file(ce->sha1, &type, &size);
a60272b3
LT
167 if (buffer) {
168 if (size == sb.len)
169 match = memcmp(buffer, sb.buf, size);
170 free(buffer);
29e4d363 171 }
a60272b3 172 strbuf_release(&sb);
29e4d363
JH
173 return match;
174}
175
21a6b9fa 176static int ce_compare_gitlink(const struct cache_entry *ce)
f35a6d3b
LT
177{
178 unsigned char sha1[20];
179
180 /*
181 * We don't actually require that the .git directory
302b9282 182 * under GITLINK directory be a valid git directory. It
f35a6d3b
LT
183 * might even be missing (in case nobody populated that
184 * sub-project).
185 *
186 * If so, we consider it always to match.
187 */
188 if (resolve_gitlink_ref(ce->name, "HEAD", sha1) < 0)
189 return 0;
190 return hashcmp(sha1, ce->sha1);
191}
192
21a6b9fa 193static int ce_modified_check_fs(const struct cache_entry *ce, struct stat *st)
29e4d363
JH
194{
195 switch (st->st_mode & S_IFMT) {
196 case S_IFREG:
197 if (ce_compare_data(ce, st))
198 return DATA_CHANGED;
199 break;
200 case S_IFLNK:
dc49cd76 201 if (ce_compare_link(ce, xsize_t(st->st_size)))
29e4d363
JH
202 return DATA_CHANGED;
203 break;
a8ee75bc 204 case S_IFDIR:
7a51ed66 205 if (S_ISGITLINK(ce->ce_mode))
c70115b4 206 return ce_compare_gitlink(ce) ? DATA_CHANGED : 0;
29e4d363
JH
207 default:
208 return TYPE_CHANGED;
209 }
210 return 0;
211}
212
21a6b9fa 213static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st)
734aab75
LT
214{
215 unsigned int changed = 0;
216
7a51ed66
LT
217 if (ce->ce_flags & CE_REMOVE)
218 return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED;
219
220 switch (ce->ce_mode & S_IFMT) {
8ae0a8c5
KS
221 case S_IFREG:
222 changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0;
3e09cdfd
JH
223 /* We consider only the owner x bit to be relevant for
224 * "mode changes"
225 */
226 if (trust_executable_bit &&
7a51ed66 227 (0100 & (ce->ce_mode ^ st->st_mode)))
ffbe1add 228 changed |= MODE_CHANGED;
8ae0a8c5
KS
229 break;
230 case S_IFLNK:
78a8d641
JS
231 if (!S_ISLNK(st->st_mode) &&
232 (has_symlinks || !S_ISREG(st->st_mode)))
233 changed |= TYPE_CHANGED;
8ae0a8c5 234 break;
302b9282 235 case S_IFGITLINK:
c70115b4 236 /* We ignore most of the st_xxx fields for gitlinks */
f35a6d3b
LT
237 if (!S_ISDIR(st->st_mode))
238 changed |= TYPE_CHANGED;
239 else if (ce_compare_gitlink(ce))
240 changed |= DATA_CHANGED;
a8ee75bc 241 return changed;
8ae0a8c5 242 default:
7a51ed66 243 die("internal error: ce_mode is %o", ce->ce_mode);
8ae0a8c5 244 }
2cb45e95 245
c21d39d7 246 changed |= match_stat_data(&ce->ce_stat_data, st);
b0391890 247
f49c2c22 248 /* Racily smudged entry? */
c21d39d7 249 if (!ce->ce_stat_data.sd_size) {
f49c2c22
LT
250 if (!is_empty_blob_sha1(ce->sha1))
251 changed |= DATA_CHANGED;
252 }
253
407c8eb0
JH
254 return changed;
255}
256
21a6b9fa
RS
257static int is_racy_timestamp(const struct index_state *istate,
258 const struct cache_entry *ce)
6d91da6d 259{
050288d5 260 return (!S_ISGITLINK(ce->ce_mode) &&
fba2f38a
KB
261 istate->timestamp.sec &&
262#ifdef USE_NSEC
263 /* nanosecond timestamped files can also be racy! */
c21d39d7
MH
264 (istate->timestamp.sec < ce->ce_stat_data.sd_mtime.sec ||
265 (istate->timestamp.sec == ce->ce_stat_data.sd_mtime.sec &&
266 istate->timestamp.nsec <= ce->ce_stat_data.sd_mtime.nsec))
fba2f38a 267#else
c21d39d7 268 istate->timestamp.sec <= ce->ce_stat_data.sd_mtime.sec
fba2f38a
KB
269#endif
270 );
6d91da6d
JH
271}
272
d1f128b0 273int ie_match_stat(const struct index_state *istate,
21a6b9fa 274 const struct cache_entry *ce, struct stat *st,
4bd5b7da 275 unsigned int options)
407c8eb0 276{
5f73076c 277 unsigned int changed;
4bd5b7da 278 int ignore_valid = options & CE_MATCH_IGNORE_VALID;
56cac48c 279 int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE;
4bd5b7da 280 int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY;
5f73076c
JH
281
282 /*
283 * If it's marked as always valid in the index, it's
284 * valid whatever the checked-out copy says.
56cac48c
NTND
285 *
286 * skip-worktree has the same effect with higher precedence
5f73076c 287 */
56cac48c
NTND
288 if (!ignore_skip_worktree && ce_skip_worktree(ce))
289 return 0;
7a51ed66 290 if (!ignore_valid && (ce->ce_flags & CE_VALID))
5f73076c
JH
291 return 0;
292
331fcb59
JH
293 /*
294 * Intent-to-add entries have not been added, so the index entry
295 * by definition never matches what is in the work tree until it
296 * actually gets added.
297 */
298 if (ce->ce_flags & CE_INTENT_TO_ADD)
299 return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED;
300
5f73076c 301 changed = ce_match_stat_basic(ce, st);
407c8eb0 302
29e4d363
JH
303 /*
304 * Within 1 second of this sequence:
305 * echo xyzzy >file && git-update-index --add file
306 * running this command:
307 * echo frotz >file
308 * would give a falsely clean cache entry. The mtime and
309 * length match the cache, and other stat fields do not change.
310 *
311 * We could detect this at update-index time (the cache entry
312 * being registered/updated records the same time as "now")
313 * and delay the return from git-update-index, but that would
314 * effectively mean we can make at most one commit per second,
315 * which is not acceptable. Instead, we check cache entries
316 * whose mtime are the same as the index file timestamp more
5f73076c 317 * carefully than others.
29e4d363 318 */
6d91da6d 319 if (!changed && is_racy_timestamp(istate, ce)) {
42f77406
JH
320 if (assume_racy_is_modified)
321 changed |= DATA_CHANGED;
322 else
323 changed |= ce_modified_check_fs(ce, st);
324 }
b0391890 325
29e4d363 326 return changed;
b0391890
JH
327}
328
d1f128b0 329int ie_modified(const struct index_state *istate,
21a6b9fa
RS
330 const struct cache_entry *ce,
331 struct stat *st, unsigned int options)
b0391890 332{
29e4d363 333 int changed, changed_fs;
4bd5b7da
JH
334
335 changed = ie_match_stat(istate, ce, st, options);
b0391890
JH
336 if (!changed)
337 return 0;
b0391890
JH
338 /*
339 * If the mode or type has changed, there's no point in trying
340 * to refresh the entry - it's not going to match
341 */
342 if (changed & (MODE_CHANGED | TYPE_CHANGED))
343 return changed;
344
c70115b4
JH
345 /*
346 * Immediately after read-tree or update-index --cacheinfo,
347 * the length field is zero, as we have never even read the
348 * lstat(2) information once, and we cannot trust DATA_CHANGED
349 * returned by ie_match_stat() which in turn was returned by
350 * ce_match_stat_basic() to signal that the filesize of the
351 * blob changed. We have to actually go to the filesystem to
352 * see if the contents match, and if so, should answer "unchanged".
353 *
354 * The logic does not apply to gitlinks, as ce_match_stat_basic()
355 * already has checked the actual HEAD from the filesystem in the
356 * subproject. If ie_match_stat() already said it is different,
357 * then we know it is.
b0391890 358 */
c70115b4 359 if ((changed & DATA_CHANGED) &&
c21d39d7 360 (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0))
b0391890
JH
361 return changed;
362
29e4d363
JH
363 changed_fs = ce_modified_check_fs(ce, st);
364 if (changed_fs)
365 return changed | changed_fs;
b0391890
JH
366 return 0;
367}
368
958ba6c9
LT
369int base_name_compare(const char *name1, int len1, int mode1,
370 const char *name2, int len2, int mode2)
371{
372 unsigned char c1, c2;
373 int len = len1 < len2 ? len1 : len2;
374 int cmp;
375
376 cmp = memcmp(name1, name2, len);
377 if (cmp)
378 return cmp;
379 c1 = name1[len];
380 c2 = name2[len];
1833a925 381 if (!c1 && S_ISDIR(mode1))
958ba6c9 382 c1 = '/';
1833a925 383 if (!c2 && S_ISDIR(mode2))
958ba6c9
LT
384 c2 = '/';
385 return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0;
386}
387
0ab9e1e8
LT
388/*
389 * df_name_compare() is identical to base_name_compare(), except it
390 * compares conflicting directory/file entries as equal. Note that
391 * while a directory name compares as equal to a regular file, they
392 * then individually compare _differently_ to a filename that has
393 * a dot after the basename (because '\0' < '.' < '/').
394 *
395 * This is used by routines that want to traverse the git namespace
396 * but then handle conflicting entries together when possible.
397 */
398int df_name_compare(const char *name1, int len1, int mode1,
399 const char *name2, int len2, int mode2)
400{
401 int len = len1 < len2 ? len1 : len2, cmp;
402 unsigned char c1, c2;
403
404 cmp = memcmp(name1, name2, len);
405 if (cmp)
406 return cmp;
407 /* Directories and files compare equal (same length, same name) */
408 if (len1 == len2)
409 return 0;
410 c1 = name1[len];
411 if (!c1 && S_ISDIR(mode1))
412 c1 = '/';
413 c2 = name2[len];
414 if (!c2 && S_ISDIR(mode2))
415 c2 = '/';
416 if (c1 == '/' && !c2)
417 return 0;
418 if (c2 == '/' && !c1)
419 return 0;
420 return c1 - c2;
421}
422
b60e188c 423int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2)
eb38c22f 424{
b60e188c
TG
425 int len = len1 < len2 ? len1 : len2;
426 int cmp;
eb38c22f
LT
427
428 cmp = memcmp(name1, name2, len);
429 if (cmp)
430 return cmp;
431 if (len1 < len2)
432 return -1;
433 if (len1 > len2)
434 return 1;
5f73076c 435
b60e188c 436 if (stage1 < stage2)
95fd5bf8 437 return -1;
b60e188c 438 if (stage1 > stage2)
95fd5bf8 439 return 1;
eb38c22f
LT
440 return 0;
441}
442
b60e188c
TG
443int cache_name_compare(const char *name1, int len1, const char *name2, int len2)
444{
445 return cache_name_stage_compare(name1, len1, 0, name2, len2, 0);
446}
447
357e9c69 448static int index_name_stage_pos(const struct index_state *istate, const char *name, int namelen, int stage)
eb38c22f
LT
449{
450 int first, last;
451
452 first = 0;
4aab5b46 453 last = istate->cache_nr;
eb38c22f
LT
454 while (last > first) {
455 int next = (last + first) >> 1;
4aab5b46 456 struct cache_entry *ce = istate->cache[next];
b60e188c 457 int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce));
eb38c22f 458 if (!cmp)
76e7f4ec 459 return next;
eb38c22f
LT
460 if (cmp < 0) {
461 last = next;
462 continue;
463 }
464 first = next+1;
465 }
76e7f4ec 466 return -first-1;
eb38c22f
LT
467}
468
b60e188c
TG
469int index_name_pos(const struct index_state *istate, const char *name, int namelen)
470{
471 return index_name_stage_pos(istate, name, namelen, 0);
472}
473
7b937ca3 474/* Remove entry, return true if there are more entries to go.. */
4aab5b46 475int remove_index_entry_at(struct index_state *istate, int pos)
7b937ca3 476{
cf558704
LT
477 struct cache_entry *ce = istate->cache[pos];
478
cfc5789a 479 record_resolve_undo(istate, ce);
2092678c 480 remove_name_hash(istate, ce);
4aab5b46
JH
481 istate->cache_changed = 1;
482 istate->cache_nr--;
483 if (pos >= istate->cache_nr)
7b937ca3 484 return 0;
4aab5b46
JH
485 memmove(istate->cache + pos,
486 istate->cache + pos + 1,
487 (istate->cache_nr - pos) * sizeof(struct cache_entry *));
7b937ca3
LT
488 return 1;
489}
490
36419c8e 491/*
98e023de 492 * Remove all cache entries marked for removal, that is where
36419c8e
KB
493 * CE_REMOVE is set in ce_flags. This is much more effective than
494 * calling remove_index_entry_at() for each entry to be removed.
495 */
496void remove_marked_cache_entries(struct index_state *istate)
497{
498 struct cache_entry **ce_array = istate->cache;
499 unsigned int i, j;
500
501 for (i = j = 0; i < istate->cache_nr; i++) {
502 if (ce_array[i]->ce_flags & CE_REMOVE)
2092678c 503 remove_name_hash(istate, ce_array[i]);
36419c8e
KB
504 else
505 ce_array[j++] = ce_array[i];
506 }
507 istate->cache_changed = 1;
508 istate->cache_nr = j;
509}
510
4aab5b46 511int remove_file_from_index(struct index_state *istate, const char *path)
197ee8c9 512{
4aab5b46 513 int pos = index_name_pos(istate, path, strlen(path));
c4e3cca1
JH
514 if (pos < 0)
515 pos = -pos-1;
09d5dc32 516 cache_tree_invalidate_path(istate->cache_tree, path);
4aab5b46
JH
517 while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path))
518 remove_index_entry_at(istate, pos);
197ee8c9
LT
519 return 0;
520}
521
20314271
JS
522static int compare_name(struct cache_entry *ce, const char *path, int namelen)
523{
524 return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen);
525}
526
527static int index_name_pos_also_unmerged(struct index_state *istate,
528 const char *path, int namelen)
529{
530 int pos = index_name_pos(istate, path, namelen);
531 struct cache_entry *ce;
532
533 if (pos >= 0)
534 return pos;
535
536 /* maybe unmerged? */
537 pos = -1 - pos;
538 if (pos >= istate->cache_nr ||
539 compare_name((ce = istate->cache[pos]), path, namelen))
540 return -1;
541
542 /* order of preference: stage 2, 1, 3 */
543 if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr &&
544 ce_stage((ce = istate->cache[pos + 1])) == 2 &&
545 !compare_name(ce, path, namelen))
546 pos++;
547 return pos;
548}
549
1102952b
LT
550static int different_name(struct cache_entry *ce, struct cache_entry *alias)
551{
552 int len = ce_namelen(ce);
553 return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len);
554}
555
556/*
557 * If we add a filename that aliases in the cache, we will use the
558 * name that we already have - but we don't want to update the same
559 * alias twice, because that implies that there were actually two
560 * different files with aliasing names!
561 *
562 * So we use the CE_ADDED flag to verify that the alias was an old
563 * one before we accept it as
564 */
565static struct cache_entry *create_alias_ce(struct cache_entry *ce, struct cache_entry *alias)
566{
567 int len;
568 struct cache_entry *new;
569
570 if (alias->ce_flags & CE_ADDED)
571 die("Will not add file alias '%s' ('%s' already exists in index)", ce->name, alias->name);
572
573 /* Ok, create the new entry using the name of the existing alias */
574 len = ce_namelen(alias);
575 new = xcalloc(1, cache_entry_size(len));
576 memcpy(new->name, alias->name, len);
577 copy_cache_entry(new, ce);
578 free(ce);
579 return new;
580}
581
39425819
JH
582static void record_intent_to_add(struct cache_entry *ce)
583{
584 unsigned char sha1[20];
585 if (write_sha1_file("", 0, blob_type, sha1))
586 die("cannot create an empty blob in the object database");
587 hashcpy(ce->sha1, sha1);
588}
589
38ed1d89 590int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags)
11be42a4 591{
38ed1d89 592 int size, namelen, was_same;
d177cab0 593 mode_t st_mode = st->st_mode;
6835550d 594 struct cache_entry *ce, *alias;
56cac48c 595 unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY;
38ed1d89
JH
596 int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND);
597 int pretend = flags & ADD_CACHE_PRETEND;
39425819
JH
598 int intent_only = flags & ADD_CACHE_INTENT;
599 int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE|
600 (intent_only ? ADD_CACHE_NEW_ONLY : 0));
11be42a4 601
d177cab0 602 if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode))
960b8ad1 603 return error("%s: can only add regular files, symbolic links or git-directories", path);
11be42a4
JS
604
605 namelen = strlen(path);
d177cab0 606 if (S_ISDIR(st_mode)) {
09595258
LT
607 while (namelen && path[namelen-1] == '/')
608 namelen--;
609 }
11be42a4
JS
610 size = cache_entry_size(namelen);
611 ce = xcalloc(1, size);
612 memcpy(ce->name, path, namelen);
b60e188c 613 ce->ce_namelen = namelen;
39425819
JH
614 if (!intent_only)
615 fill_stat_cache_info(ce, st);
388b2acd
JH
616 else
617 ce->ce_flags |= CE_INTENT_TO_ADD;
11be42a4 618
78a8d641 619 if (trust_executable_bit && has_symlinks)
d177cab0 620 ce->ce_mode = create_ce_mode(st_mode);
185c975f 621 else {
78a8d641
JS
622 /* If there is an existing entry, pick the mode bits and type
623 * from it, otherwise assume unexecutable regular file.
11be42a4 624 */
185c975f 625 struct cache_entry *ent;
20314271 626 int pos = index_name_pos_also_unmerged(istate, path, namelen);
185c975f 627
4aab5b46 628 ent = (0 <= pos) ? istate->cache[pos] : NULL;
d177cab0 629 ce->ce_mode = ce_mode_from_stat(ent, st_mode);
11be42a4
JS
630 }
631
dc1ae704
JJ
632 /* When core.ignorecase=true, determine if a directory of the same name but differing
633 * case already exists within the Git repository. If it does, ensure the directory
634 * case of the file being added to the repository matches (is folded into) the existing
635 * entry's directory case.
636 */
637 if (ignore_case) {
638 const char *startPtr = ce->name;
639 const char *ptr = startPtr;
640 while (*ptr) {
641 while (*ptr && *ptr != '/')
642 ++ptr;
643 if (*ptr == '/') {
644 struct cache_entry *foundce;
645 ++ptr;
d28eec26 646 foundce = index_dir_exists(istate, ce->name, ptr - ce->name - 1);
dc1ae704
JJ
647 if (foundce) {
648 memcpy((void *)startPtr, foundce->name + (startPtr - ce->name), ptr - startPtr);
649 startPtr = ptr;
650 }
651 }
652 }
653 }
654
ebbd7439 655 alias = index_file_exists(istate, ce->name, ce_namelen(ce), ignore_case);
d177cab0 656 if (alias && !ce_stage(alias) && !ie_match_stat(istate, alias, st, ce_option)) {
0781b8a9
JH
657 /* Nothing changed, really */
658 free(ce);
125fd984
JH
659 if (!S_ISGITLINK(alias->ce_mode))
660 ce_mark_uptodate(alias);
1102952b 661 alias->ce_flags |= CE_ADDED;
0781b8a9
JH
662 return 0;
663 }
39425819 664 if (!intent_only) {
c4ce46fc 665 if (index_path(ce->sha1, path, st, HASH_WRITE_OBJECT))
39425819
JH
666 return error("unable to index file %s", path);
667 } else
668 record_intent_to_add(ce);
669
1102952b
LT
670 if (ignore_case && alias && different_name(ce, alias))
671 ce = create_alias_ce(ce, alias);
672 ce->ce_flags |= CE_ADDED;
38ed1d89 673
3bf0dd1f 674 /* It was suspected to be racily clean, but it turns out to be Ok */
38ed1d89
JH
675 was_same = (alias &&
676 !ce_stage(alias) &&
677 !hashcmp(alias->sha1, ce->sha1) &&
678 ce->ce_mode == alias->ce_mode);
679
680 if (pretend)
681 ;
39425819 682 else if (add_index_entry(istate, ce, add_option))
960b8ad1 683 return error("unable to add %s to index",path);
38ed1d89 684 if (verbose && !was_same)
11be42a4 685 printf("add '%s'\n", path);
11be42a4
JS
686 return 0;
687}
688
38ed1d89 689int add_file_to_index(struct index_state *istate, const char *path, int flags)
d177cab0
LT
690{
691 struct stat st;
692 if (lstat(path, &st))
d824cbba 693 die_errno("unable to stat '%s'", path);
38ed1d89 694 return add_to_index(istate, path, &st, flags);
d177cab0
LT
695}
696
6640f881
CR
697struct cache_entry *make_cache_entry(unsigned int mode,
698 const unsigned char *sha1, const char *path, int stage,
699 int refresh)
700{
701 int size, len;
702 struct cache_entry *ce;
703
7e7abea9
DP
704 if (!verify_path(path)) {
705 error("Invalid path '%s'", path);
6640f881 706 return NULL;
7e7abea9 707 }
6640f881
CR
708
709 len = strlen(path);
710 size = cache_entry_size(len);
711 ce = xcalloc(1, size);
712
713 hashcpy(ce->sha1, sha1);
714 memcpy(ce->name, path, len);
b60e188c
TG
715 ce->ce_flags = create_ce_flags(stage);
716 ce->ce_namelen = len;
6640f881
CR
717 ce->ce_mode = create_ce_mode(mode);
718
719 if (refresh)
720 return refresh_cache_entry(ce, 0);
721
722 return ce;
723}
724
9c5e6c80 725int ce_same_name(const struct cache_entry *a, const struct cache_entry *b)
7b937ca3
LT
726{
727 int len = ce_namelen(a);
728 return ce_namelen(b) == len && !memcmp(a->name, b->name, len);
729}
730
eb9cb55b 731int ce_path_match(const struct cache_entry *ce, const struct pathspec *pathspec)
c0fd1f51 732{
898bbd9f 733 return match_pathspec_depth(pathspec, ce->name, ce_namelen(ce), 0, NULL);
c0fd1f51
LT
734}
735
8dcf39c4
LT
736/*
737 * We fundamentally don't like some paths: we don't want
738 * dot or dot-dot anywhere, and for obvious reasons don't
739 * want to recurse into ".git" either.
740 *
741 * Also, we don't want double slashes or slashes at the
742 * end that can make pathnames ambiguous.
743 */
744static int verify_dotfile(const char *rest)
745{
746 /*
747 * The first character was '.', but that
748 * has already been discarded, we now test
749 * the rest.
750 */
e0f530ff 751
8dcf39c4 752 /* "." is not allowed */
e0f530ff 753 if (*rest == '\0' || is_dir_sep(*rest))
8dcf39c4
LT
754 return 0;
755
e0f530ff 756 switch (*rest) {
8dcf39c4
LT
757 /*
758 * ".git" followed by NUL or slash is bad. This
759 * shares the path end test with the ".." case.
760 */
761 case 'g':
762 if (rest[1] != 'i')
763 break;
764 if (rest[2] != 't')
765 break;
766 rest += 2;
767 /* fallthrough */
768 case '.':
e0f530ff 769 if (rest[1] == '\0' || is_dir_sep(rest[1]))
8dcf39c4
LT
770 return 0;
771 }
772 return 1;
773}
774
775int verify_path(const char *path)
776{
777 char c;
778
56948cb6
EFL
779 if (has_dos_drive_prefix(path))
780 return 0;
781
8dcf39c4
LT
782 goto inside;
783 for (;;) {
784 if (!c)
785 return 1;
56948cb6 786 if (is_dir_sep(c)) {
8dcf39c4
LT
787inside:
788 c = *path++;
3bdf09c7
JH
789 if ((c == '.' && !verify_dotfile(path)) ||
790 is_dir_sep(c) || c == '\0')
791 return 0;
8dcf39c4
LT
792 }
793 c = *path++;
794 }
795}
796
12676608
LT
797/*
798 * Do we have another file that has the beginning components being a
799 * proper superset of the name we're trying to add?
0f1e4f04 800 */
4aab5b46
JH
801static int has_file_name(struct index_state *istate,
802 const struct cache_entry *ce, int pos, int ok_to_replace)
0f1e4f04 803{
12676608
LT
804 int retval = 0;
805 int len = ce_namelen(ce);
b155725d 806 int stage = ce_stage(ce);
12676608 807 const char *name = ce->name;
0f1e4f04 808
4aab5b46
JH
809 while (pos < istate->cache_nr) {
810 struct cache_entry *p = istate->cache[pos++];
0f1e4f04 811
12676608 812 if (len >= ce_namelen(p))
0f1e4f04 813 break;
12676608
LT
814 if (memcmp(name, p->name, len))
815 break;
b155725d
JH
816 if (ce_stage(p) != stage)
817 continue;
12676608
LT
818 if (p->name[len] != '/')
819 continue;
7a51ed66 820 if (p->ce_flags & CE_REMOVE)
21cd8d00 821 continue;
12676608
LT
822 retval = -1;
823 if (!ok_to_replace)
824 break;
4aab5b46 825 remove_index_entry_at(istate, --pos);
0f1e4f04 826 }
12676608
LT
827 return retval;
828}
0f1e4f04 829
12676608
LT
830/*
831 * Do we have another file with a pathname that is a proper
832 * subset of the name we're trying to add?
833 */
4aab5b46
JH
834static int has_dir_name(struct index_state *istate,
835 const struct cache_entry *ce, int pos, int ok_to_replace)
12676608
LT
836{
837 int retval = 0;
b155725d 838 int stage = ce_stage(ce);
12676608
LT
839 const char *name = ce->name;
840 const char *slash = name + ce_namelen(ce);
0f1e4f04 841
12676608
LT
842 for (;;) {
843 int len;
0f1e4f04 844
12676608
LT
845 for (;;) {
846 if (*--slash == '/')
847 break;
848 if (slash <= ce->name)
849 return retval;
850 }
851 len = slash - name;
0f1e4f04 852
b60e188c 853 pos = index_name_stage_pos(istate, name, len, stage);
12676608 854 if (pos >= 0) {
21cd8d00
JH
855 /*
856 * Found one, but not so fast. This could
857 * be a marker that says "I was here, but
858 * I am being removed". Such an entry is
859 * not a part of the resulting tree, and
860 * it is Ok to have a directory at the same
861 * path.
862 */
077c48df 863 if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) {
21cd8d00
JH
864 retval = -1;
865 if (!ok_to_replace)
866 break;
4aab5b46 867 remove_index_entry_at(istate, pos);
21cd8d00
JH
868 continue;
869 }
12676608 870 }
21cd8d00
JH
871 else
872 pos = -pos-1;
12676608
LT
873
874 /*
875 * Trivial optimization: if we find an entry that
876 * already matches the sub-directory, then we know
b155725d 877 * we're ok, and we can exit.
12676608 878 */
4aab5b46
JH
879 while (pos < istate->cache_nr) {
880 struct cache_entry *p = istate->cache[pos];
b155725d
JH
881 if ((ce_namelen(p) <= len) ||
882 (p->name[len] != '/') ||
883 memcmp(p->name, name, len))
884 break; /* not our subdirectory */
077c48df
JH
885 if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE))
886 /*
887 * p is at the same stage as our entry, and
b155725d
JH
888 * is a subdirectory of what we are looking
889 * at, so we cannot have conflicts at our
890 * level or anything shorter.
891 */
892 return retval;
893 pos++;
192268c1 894 }
0f1e4f04 895 }
12676608
LT
896 return retval;
897}
898
899/* We may be in a situation where we already have path/file and path
900 * is being added, or we already have path and path/file is being
901 * added. Either one would result in a nonsense tree that has path
902 * twice when git-write-tree tries to write it out. Prevent it.
a6080a0a 903 *
12676608
LT
904 * If ok-to-replace is specified, we remove the conflicting entries
905 * from the cache so the caller should recompute the insert position.
906 * When this happens, we return non-zero.
907 */
4aab5b46
JH
908static int check_file_directory_conflict(struct index_state *istate,
909 const struct cache_entry *ce,
910 int pos, int ok_to_replace)
12676608 911{
21cd8d00
JH
912 int retval;
913
914 /*
915 * When ce is an "I am going away" entry, we allow it to be added
916 */
7a51ed66 917 if (ce->ce_flags & CE_REMOVE)
21cd8d00
JH
918 return 0;
919
12676608
LT
920 /*
921 * We check if the path is a sub-path of a subsequent pathname
922 * first, since removing those will not change the position
21cd8d00 923 * in the array.
12676608 924 */
4aab5b46 925 retval = has_file_name(istate, ce, pos, ok_to_replace);
21cd8d00 926
12676608
LT
927 /*
928 * Then check if the path might have a clashing sub-directory
929 * before it.
930 */
4aab5b46 931 return retval + has_dir_name(istate, ce, pos, ok_to_replace);
0f1e4f04
JH
932}
933
af3785dc 934static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option)
197ee8c9
LT
935{
936 int pos;
192268c1
JH
937 int ok_to_add = option & ADD_CACHE_OK_TO_ADD;
938 int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE;
b155725d 939 int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK;
39425819 940 int new_only = option & ADD_CACHE_NEW_ONLY;
5f73076c 941
09d5dc32 942 cache_tree_invalidate_path(istate->cache_tree, ce->name);
b60e188c 943 pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce));
197ee8c9 944
3e09cdfd 945 /* existing match? Just replace it. */
76e7f4ec 946 if (pos >= 0) {
39425819
JH
947 if (!new_only)
948 replace_index_entry(istate, pos, ce);
197ee8c9
LT
949 return 0;
950 }
76e7f4ec 951 pos = -pos-1;
197ee8c9 952
7b937ca3
LT
953 /*
954 * Inserting a merged entry ("stage 0") into the index
955 * will always replace all non-merged entries..
956 */
4aab5b46
JH
957 if (pos < istate->cache_nr && ce_stage(ce) == 0) {
958 while (ce_same_name(istate->cache[pos], ce)) {
7b937ca3 959 ok_to_add = 1;
4aab5b46 960 if (!remove_index_entry_at(istate, pos))
7b937ca3
LT
961 break;
962 }
963 }
964
121481ab
LT
965 if (!ok_to_add)
966 return -1;
8dcf39c4 967 if (!verify_path(ce->name))
7e7abea9 968 return error("Invalid path '%s'", ce->name);
121481ab 969
3e09cdfd 970 if (!skip_df_check &&
4aab5b46 971 check_file_directory_conflict(istate, ce, pos, ok_to_replace)) {
192268c1 972 if (!ok_to_replace)
4aab5b46
JH
973 return error("'%s' appears as both a file and as a directory",
974 ce->name);
b60e188c 975 pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce));
192268c1
JH
976 pos = -pos-1;
977 }
af3785dc
JH
978 return pos + 1;
979}
980
981int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option)
982{
983 int pos;
984
985 if (option & ADD_CACHE_JUST_APPEND)
986 pos = istate->cache_nr;
987 else {
988 int ret;
989 ret = add_index_entry_with_check(istate, ce, option);
990 if (ret <= 0)
991 return ret;
992 pos = ret - 1;
993 }
0f1e4f04 994
197ee8c9 995 /* Make sure the array is big enough .. */
4aab5b46
JH
996 if (istate->cache_nr == istate->cache_alloc) {
997 istate->cache_alloc = alloc_nr(istate->cache_alloc);
998 istate->cache = xrealloc(istate->cache,
c4aa3167 999 istate->cache_alloc * sizeof(*istate->cache));
197ee8c9
LT
1000 }
1001
1002 /* Add it in.. */
4aab5b46 1003 istate->cache_nr++;
af3785dc 1004 if (istate->cache_nr > pos + 1)
4aab5b46
JH
1005 memmove(istate->cache + pos + 1,
1006 istate->cache + pos,
1007 (istate->cache_nr - pos - 1) * sizeof(ce));
cf558704 1008 set_index_entry(istate, pos, ce);
4aab5b46 1009 istate->cache_changed = 1;
197ee8c9
LT
1010 return 0;
1011}
1012
405e5b2f
LT
1013/*
1014 * "refresh" does not calculate a new sha1 file or bring the
1015 * cache up-to-date for mode/content changes. But what it
1016 * _does_ do is to "re-match" the stat information of a file
1017 * with the cache, so that you can refresh the cache for a
1018 * file that hasn't been changed but where the stat entry is
1019 * out of date.
1020 *
1021 * For example, you'd want to do this after doing a "git-read-tree",
1022 * to link up the stat cache details with the proper files.
1023 */
4aab5b46 1024static struct cache_entry *refresh_cache_ent(struct index_state *istate,
4bd5b7da 1025 struct cache_entry *ce,
d05e6970
JK
1026 unsigned int options, int *err,
1027 int *changed_ret)
405e5b2f
LT
1028{
1029 struct stat st;
1030 struct cache_entry *updated;
1031 int changed, size;
4bd5b7da 1032 int ignore_valid = options & CE_MATCH_IGNORE_VALID;
56cac48c 1033 int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE;
405e5b2f 1034
eadb5831
JH
1035 if (ce_uptodate(ce))
1036 return ce;
1037
aa9349d4 1038 /*
56cac48c
NTND
1039 * CE_VALID or CE_SKIP_WORKTREE means the user promised us
1040 * that the change to the work tree does not matter and told
1041 * us not to worry.
aa9349d4 1042 */
56cac48c
NTND
1043 if (!ignore_skip_worktree && ce_skip_worktree(ce)) {
1044 ce_mark_uptodate(ce);
1045 return ce;
1046 }
aa9349d4
MSO
1047 if (!ignore_valid && (ce->ce_flags & CE_VALID)) {
1048 ce_mark_uptodate(ce);
1049 return ce;
1050 }
1051
8fd2cb40 1052 if (lstat(ce->name, &st) < 0) {
ec0cc704
JH
1053 if (err)
1054 *err = errno;
8fd2cb40
JS
1055 return NULL;
1056 }
405e5b2f 1057
4bd5b7da 1058 changed = ie_match_stat(istate, ce, &st, options);
d05e6970
JK
1059 if (changed_ret)
1060 *changed_ret = changed;
405e5b2f 1061 if (!changed) {
4bd5b7da
JH
1062 /*
1063 * The path is unchanged. If we were told to ignore
1064 * valid bit, then we did the actual stat check and
1065 * found that the entry is unmodified. If the entry
1066 * is not marked VALID, this is the place to mark it
1067 * valid again, under "assume unchanged" mode.
1068 */
1069 if (ignore_valid && assume_unchanged &&
7a51ed66 1070 !(ce->ce_flags & CE_VALID))
405e5b2f 1071 ; /* mark this one VALID again */
eadb5831
JH
1072 else {
1073 /*
1074 * We do not mark the index itself "modified"
1075 * because CE_UPTODATE flag is in-core only;
1076 * we are not going to write this change out.
1077 */
125fd984
JH
1078 if (!S_ISGITLINK(ce->ce_mode))
1079 ce_mark_uptodate(ce);
8fd2cb40 1080 return ce;
eadb5831 1081 }
405e5b2f
LT
1082 }
1083
4bd5b7da 1084 if (ie_modified(istate, ce, &st, options)) {
ec0cc704
JH
1085 if (err)
1086 *err = EINVAL;
8fd2cb40
JS
1087 return NULL;
1088 }
405e5b2f
LT
1089
1090 size = ce_size(ce);
1091 updated = xmalloc(size);
1092 memcpy(updated, ce, size);
1093 fill_stat_cache_info(updated, &st);
4bd5b7da
JH
1094 /*
1095 * If ignore_valid is not set, we should leave CE_VALID bit
1096 * alone. Otherwise, paths marked with --no-assume-unchanged
1097 * (i.e. things to be edited) will reacquire CE_VALID bit
1098 * automatically, which is not really what we want.
405e5b2f 1099 */
4bd5b7da 1100 if (!ignore_valid && assume_unchanged &&
7a51ed66
LT
1101 !(ce->ce_flags & CE_VALID))
1102 updated->ce_flags &= ~CE_VALID;
405e5b2f
LT
1103
1104 return updated;
1105}
1106
3deffc52 1107static void show_file(const char * fmt, const char * name, int in_porcelain,
046613c5 1108 int * first, const char *header_msg)
3deffc52
MM
1109{
1110 if (in_porcelain && *first && header_msg) {
1111 printf("%s\n", header_msg);
cd2b8ae9 1112 *first = 0;
3deffc52
MM
1113 }
1114 printf(fmt, name);
1115}
1116
9b2d6149
NTND
1117int refresh_index(struct index_state *istate, unsigned int flags,
1118 const struct pathspec *pathspec,
046613c5 1119 char *seen, const char *header_msg)
405e5b2f
LT
1120{
1121 int i;
1122 int has_errors = 0;
1123 int really = (flags & REFRESH_REALLY) != 0;
1124 int allow_unmerged = (flags & REFRESH_UNMERGED) != 0;
1125 int quiet = (flags & REFRESH_QUIET) != 0;
1126 int not_new = (flags & REFRESH_IGNORE_MISSING) != 0;
5fdeacb0 1127 int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0;
3deffc52
MM
1128 int first = 1;
1129 int in_porcelain = (flags & REFRESH_IN_PORCELAIN);
4bd5b7da 1130 unsigned int options = really ? CE_MATCH_IGNORE_VALID : 0;
4bd4e730 1131 const char *modified_fmt;
73b7eae6
JK
1132 const char *deleted_fmt;
1133 const char *typechange_fmt;
1134 const char *added_fmt;
4bd4e730 1135 const char *unmerged_fmt;
405e5b2f 1136
4bd4e730 1137 modified_fmt = (in_porcelain ? "M\t%s\n" : "%s: needs update\n");
73b7eae6
JK
1138 deleted_fmt = (in_porcelain ? "D\t%s\n" : "%s: needs update\n");
1139 typechange_fmt = (in_porcelain ? "T\t%s\n" : "%s needs update\n");
1140 added_fmt = (in_porcelain ? "A\t%s\n" : "%s needs update\n");
4bd4e730 1141 unmerged_fmt = (in_porcelain ? "U\t%s\n" : "%s: needs merge\n");
4aab5b46 1142 for (i = 0; i < istate->cache_nr; i++) {
405e5b2f 1143 struct cache_entry *ce, *new;
ec0cc704 1144 int cache_errno = 0;
73b7eae6 1145 int changed = 0;
3d1f148c 1146 int filtered = 0;
ec0cc704 1147
4aab5b46 1148 ce = istate->cache[i];
5fdeacb0
JS
1149 if (ignore_submodules && S_ISGITLINK(ce->ce_mode))
1150 continue;
1151
3d1f148c 1152 if (pathspec &&
9b2d6149 1153 !match_pathspec_depth(pathspec, ce->name, ce_namelen(ce), 0, seen))
3d1f148c
JH
1154 filtered = 1;
1155
405e5b2f 1156 if (ce_stage(ce)) {
4aab5b46
JH
1157 while ((i < istate->cache_nr) &&
1158 ! strcmp(istate->cache[i]->name, ce->name))
405e5b2f
LT
1159 i++;
1160 i--;
1161 if (allow_unmerged)
1162 continue;
3d1f148c
JH
1163 if (!filtered)
1164 show_file(unmerged_fmt, ce->name, in_porcelain,
1165 &first, header_msg);
405e5b2f
LT
1166 has_errors = 1;
1167 continue;
1168 }
1169
3d1f148c 1170 if (filtered)
d616813d
AJ
1171 continue;
1172
73b7eae6 1173 new = refresh_cache_ent(istate, ce, options, &cache_errno, &changed);
8fd2cb40 1174 if (new == ce)
405e5b2f 1175 continue;
8fd2cb40 1176 if (!new) {
73b7eae6
JK
1177 const char *fmt;
1178
8fd2cb40 1179 if (not_new && cache_errno == ENOENT)
405e5b2f 1180 continue;
8fd2cb40 1181 if (really && cache_errno == EINVAL) {
405e5b2f
LT
1182 /* If we are doing --really-refresh that
1183 * means the index is not valid anymore.
1184 */
7a51ed66 1185 ce->ce_flags &= ~CE_VALID;
4aab5b46 1186 istate->cache_changed = 1;
405e5b2f
LT
1187 }
1188 if (quiet)
1189 continue;
73b7eae6
JK
1190
1191 if (cache_errno == ENOENT)
1192 fmt = deleted_fmt;
1193 else if (ce->ce_flags & CE_INTENT_TO_ADD)
1194 fmt = added_fmt; /* must be before other checks */
1195 else if (changed & TYPE_CHANGED)
1196 fmt = typechange_fmt;
1197 else
1198 fmt = modified_fmt;
1199 show_file(fmt,
1200 ce->name, in_porcelain, &first, header_msg);
405e5b2f
LT
1201 has_errors = 1;
1202 continue;
1203 }
cf558704
LT
1204
1205 replace_index_entry(istate, i, new);
405e5b2f
LT
1206 }
1207 return has_errors;
1208}
1209
87b29e5a 1210static struct cache_entry *refresh_cache_entry(struct cache_entry *ce, int really)
ec0cc704 1211{
d05e6970 1212 return refresh_cache_ent(&the_index, ce, really, NULL, NULL);
ec0cc704
JH
1213}
1214
db3b313c
JH
1215
1216/*****************************************************************
1217 * Index File I/O
1218 *****************************************************************/
1219
9d227781
JH
1220#define INDEX_FORMAT_DEFAULT 3
1221
db3b313c
JH
1222/*
1223 * dev/ino/uid/gid/size are also just tracked to the low 32 bits
1224 * Again - this is just a (very strong in practice) heuristic that
1225 * the inode hasn't changed.
1226 *
1227 * We save the fields in big-endian order to allow using the
1228 * index file over NFS transparently.
1229 */
1230struct ondisk_cache_entry {
1231 struct cache_time ctime;
1232 struct cache_time mtime;
7800c1eb
TG
1233 uint32_t dev;
1234 uint32_t ino;
1235 uint32_t mode;
1236 uint32_t uid;
1237 uint32_t gid;
1238 uint32_t size;
db3b313c 1239 unsigned char sha1[20];
7800c1eb 1240 uint16_t flags;
db3b313c
JH
1241 char name[FLEX_ARRAY]; /* more */
1242};
1243
1244/*
1245 * This struct is used when CE_EXTENDED bit is 1
1246 * The struct must match ondisk_cache_entry exactly from
1247 * ctime till flags
1248 */
1249struct ondisk_cache_entry_extended {
1250 struct cache_time ctime;
1251 struct cache_time mtime;
7800c1eb
TG
1252 uint32_t dev;
1253 uint32_t ino;
1254 uint32_t mode;
1255 uint32_t uid;
1256 uint32_t gid;
1257 uint32_t size;
db3b313c 1258 unsigned char sha1[20];
7800c1eb
TG
1259 uint16_t flags;
1260 uint16_t flags2;
db3b313c
JH
1261 char name[FLEX_ARRAY]; /* more */
1262};
1263
6c9cd161 1264/* These are only used for v3 or lower */
db3b313c
JH
1265#define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,name) + (len) + 8) & ~7)
1266#define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len)
1267#define ondisk_cache_entry_extended_size(len) align_flex_name(ondisk_cache_entry_extended,len)
1268#define ondisk_ce_size(ce) (((ce)->ce_flags & CE_EXTENDED) ? \
1269 ondisk_cache_entry_extended_size(ce_namelen(ce)) : \
1270 ondisk_cache_entry_size(ce_namelen(ce)))
1271
bad68ec9 1272static int verify_hdr(struct cache_header *hdr, unsigned long size)
e83c5163 1273{
9126f009 1274 git_SHA_CTX c;
bad68ec9 1275 unsigned char sha1[20];
0136bac9 1276 int hdr_version;
e83c5163 1277
ccc4feb5 1278 if (hdr->hdr_signature != htonl(CACHE_SIGNATURE))
e83c5163 1279 return error("bad signature");
0136bac9 1280 hdr_version = ntohl(hdr->hdr_version);
b82a7b5b 1281 if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version)
0136bac9 1282 return error("bad index version %d", hdr_version);
9126f009
NP
1283 git_SHA1_Init(&c);
1284 git_SHA1_Update(&c, hdr, size - 20);
1285 git_SHA1_Final(sha1, &c);
a89fccd2 1286 if (hashcmp(sha1, (unsigned char *)hdr + size - 20))
ca9be054 1287 return error("bad index file sha1 signature");
e83c5163
LT
1288 return 0;
1289}
1290
4aab5b46
JH
1291static int read_index_extension(struct index_state *istate,
1292 const char *ext, void *data, unsigned long sz)
bad68ec9
JH
1293{
1294 switch (CACHE_EXT(ext)) {
1295 case CACHE_EXT_TREE:
4aab5b46 1296 istate->cache_tree = cache_tree_read(data, sz);
bad68ec9 1297 break;
cfc5789a
JH
1298 case CACHE_EXT_RESOLVE_UNDO:
1299 istate->resolve_undo = resolve_undo_read(data, sz);
1300 break;
bad68ec9
JH
1301 default:
1302 if (*ext < 'A' || 'Z' < *ext)
1303 return error("index uses %.4s extension, which we do not understand",
1304 ext);
1305 fprintf(stderr, "ignoring %.4s extension\n", ext);
1306 break;
1307 }
1308 return 0;
1309}
1310
4aab5b46 1311int read_index(struct index_state *istate)
8fd2cb40 1312{
4aab5b46 1313 return read_index_from(istate, get_index_file());
8fd2cb40
JS
1314}
1315
d60c49c2
JH
1316#ifndef NEEDS_ALIGNED_ACCESS
1317#define ntoh_s(var) ntohs(var)
1318#define ntoh_l(var) ntohl(var)
1319#else
1320static inline uint16_t ntoh_s_force_align(void *p)
1321{
1322 uint16_t x;
1323 memcpy(&x, p, sizeof(x));
1324 return ntohs(x);
1325}
1326static inline uint32_t ntoh_l_force_align(void *p)
1327{
1328 uint32_t x;
1329 memcpy(&x, p, sizeof(x));
1330 return ntohl(x);
1331}
1332#define ntoh_s(var) ntoh_s_force_align(&(var))
1333#define ntoh_l(var) ntoh_l_force_align(&(var))
1334#endif
1335
3fc22b53
JH
1336static struct cache_entry *cache_entry_from_ondisk(struct ondisk_cache_entry *ondisk,
1337 unsigned int flags,
1338 const char *name,
1339 size_t len)
1340{
1341 struct cache_entry *ce = xmalloc(cache_entry_size(len));
1342
c21d39d7
MH
1343 ce->ce_stat_data.sd_ctime.sec = ntoh_l(ondisk->ctime.sec);
1344 ce->ce_stat_data.sd_mtime.sec = ntoh_l(ondisk->mtime.sec);
1345 ce->ce_stat_data.sd_ctime.nsec = ntoh_l(ondisk->ctime.nsec);
1346 ce->ce_stat_data.sd_mtime.nsec = ntoh_l(ondisk->mtime.nsec);
1347 ce->ce_stat_data.sd_dev = ntoh_l(ondisk->dev);
1348 ce->ce_stat_data.sd_ino = ntoh_l(ondisk->ino);
3fc22b53 1349 ce->ce_mode = ntoh_l(ondisk->mode);
c21d39d7
MH
1350 ce->ce_stat_data.sd_uid = ntoh_l(ondisk->uid);
1351 ce->ce_stat_data.sd_gid = ntoh_l(ondisk->gid);
1352 ce->ce_stat_data.sd_size = ntoh_l(ondisk->size);
b60e188c
TG
1353 ce->ce_flags = flags & ~CE_NAMEMASK;
1354 ce->ce_namelen = len;
3fc22b53
JH
1355 hashcpy(ce->sha1, ondisk->sha1);
1356 memcpy(ce->name, name, len);
1357 ce->name[len] = '\0';
1358 return ce;
1359}
1360
6c9cd161
JH
1361/*
1362 * Adjacent cache entries tend to share the leading paths, so it makes
1363 * sense to only store the differences in later entries. In the v4
1364 * on-disk format of the index, each on-disk cache entry stores the
1365 * number of bytes to be stripped from the end of the previous name,
1366 * and the bytes to append to the result, to come up with its name.
1367 */
1368static unsigned long expand_name_field(struct strbuf *name, const char *cp_)
1369{
1370 const unsigned char *ep, *cp = (const unsigned char *)cp_;
1371 size_t len = decode_varint(&cp);
1372
1373 if (name->len < len)
1374 die("malformed name field in the index");
1375 strbuf_remove(name, name->len - len, len);
1376 for (ep = cp; *ep; ep++)
1377 ; /* find the end */
1378 strbuf_add(name, cp, ep - cp);
1379 return (const char *)ep + 1 - cp_;
1380}
1381
936f53d0 1382static struct cache_entry *create_from_disk(struct ondisk_cache_entry *ondisk,
6c9cd161
JH
1383 unsigned long *ent_size,
1384 struct strbuf *previous_name)
7a51ed66 1385{
debed2a6 1386 struct cache_entry *ce;
7fec10b7 1387 size_t len;
06aaaa0b 1388 const char *name;
debed2a6 1389 unsigned int flags;
7fec10b7 1390
7a51ed66 1391 /* On-disk flags are just 16 bits */
d60c49c2 1392 flags = ntoh_s(ondisk->flags);
debed2a6 1393 len = flags & CE_NAMEMASK;
7fec10b7 1394
debed2a6 1395 if (flags & CE_EXTENDED) {
06aaaa0b
NTND
1396 struct ondisk_cache_entry_extended *ondisk2;
1397 int extended_flags;
1398 ondisk2 = (struct ondisk_cache_entry_extended *)ondisk;
d60c49c2 1399 extended_flags = ntoh_s(ondisk2->flags2) << 16;
06aaaa0b
NTND
1400 /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */
1401 if (extended_flags & ~CE_EXTENDED_FLAGS)
1402 die("Unknown index entry format %08x", extended_flags);
debed2a6 1403 flags |= extended_flags;
06aaaa0b
NTND
1404 name = ondisk2->name;
1405 }
1406 else
1407 name = ondisk->name;
1408
6c9cd161
JH
1409 if (!previous_name) {
1410 /* v3 and earlier */
1411 if (len == CE_NAMEMASK)
1412 len = strlen(name);
1413 ce = cache_entry_from_ondisk(ondisk, flags, name, len);
1414
1415 *ent_size = ondisk_ce_size(ce);
1416 } else {
1417 unsigned long consumed;
1418 consumed = expand_name_field(previous_name, name);
1419 ce = cache_entry_from_ondisk(ondisk, flags,
1420 previous_name->buf,
1421 previous_name->len);
1422
1423 *ent_size = (name - ((char *)ondisk)) + consumed;
1424 }
debed2a6 1425 return ce;
cf558704
LT
1426}
1427
8fd2cb40 1428/* remember to discard_cache() before reading a different cache! */
4aab5b46 1429int read_index_from(struct index_state *istate, const char *path)
e83c5163
LT
1430{
1431 int fd, i;
1432 struct stat st;
debed2a6 1433 unsigned long src_offset;
e83c5163 1434 struct cache_header *hdr;
7a51ed66
LT
1435 void *mmap;
1436 size_t mmap_size;
6c9cd161 1437 struct strbuf previous_name_buf = STRBUF_INIT, *previous_name;
e83c5163 1438
913e0e99 1439 if (istate->initialized)
4aab5b46 1440 return istate->cache_nr;
5d1a5c02 1441
fba2f38a
KB
1442 istate->timestamp.sec = 0;
1443 istate->timestamp.nsec = 0;
8fd2cb40 1444 fd = open(path, O_RDONLY);
5d1a5c02
LT
1445 if (fd < 0) {
1446 if (errno == ENOENT)
1447 return 0;
d824cbba 1448 die_errno("index file open failed");
5d1a5c02 1449 }
e83c5163 1450
3511a377 1451 if (fstat(fd, &st))
d824cbba 1452 die_errno("cannot stat the open index");
3511a377 1453
7a51ed66
LT
1454 mmap_size = xsize_t(st.st_size);
1455 if (mmap_size < sizeof(struct cache_header) + 20)
3511a377
LFC
1456 die("index file smaller than expected");
1457
7a51ed66 1458 mmap = xmmap(NULL, mmap_size, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0);
7a51ed66 1459 if (mmap == MAP_FAILED)
0721c314 1460 die_errno("unable to map index file");
57d84f8d 1461 close(fd);
e83c5163 1462
7a51ed66
LT
1463 hdr = mmap;
1464 if (verify_hdr(hdr, mmap_size) < 0)
e83c5163
LT
1465 goto unmap;
1466
9d227781 1467 istate->version = ntohl(hdr->hdr_version);
4aab5b46
JH
1468 istate->cache_nr = ntohl(hdr->hdr_entries);
1469 istate->cache_alloc = alloc_nr(istate->cache_nr);
c4aa3167 1470 istate->cache = xcalloc(istate->cache_alloc, sizeof(*istate->cache));
913e0e99 1471 istate->initialized = 1;
7a51ed66 1472
9d227781 1473 if (istate->version == 4)
6c9cd161
JH
1474 previous_name = &previous_name_buf;
1475 else
1476 previous_name = NULL;
1477
7a51ed66 1478 src_offset = sizeof(*hdr);
4aab5b46 1479 for (i = 0; i < istate->cache_nr; i++) {
7a51ed66 1480 struct ondisk_cache_entry *disk_ce;
4aab5b46 1481 struct cache_entry *ce;
936f53d0 1482 unsigned long consumed;
4aab5b46 1483
7a51ed66 1484 disk_ce = (struct ondisk_cache_entry *)((char *)mmap + src_offset);
6c9cd161 1485 ce = create_from_disk(disk_ce, &consumed, previous_name);
cf558704 1486 set_index_entry(istate, i, ce);
7a51ed66 1487
936f53d0 1488 src_offset += consumed;
e83c5163 1489 }
6c9cd161 1490 strbuf_release(&previous_name_buf);
fba2f38a 1491 istate->timestamp.sec = st.st_mtime;
c06ff490 1492 istate->timestamp.nsec = ST_MTIME_NSEC(st);
fba2f38a 1493
7a51ed66 1494 while (src_offset <= mmap_size - 20 - 8) {
bad68ec9
JH
1495 /* After an array of active_nr index entries,
1496 * there can be arbitrary number of extended
1497 * sections, each of which is prefixed with
1498 * extension name (4-byte) and section length
1499 * in 4-byte network byte order.
1500 */
07cc8eca 1501 uint32_t extsize;
7a51ed66 1502 memcpy(&extsize, (char *)mmap + src_offset + 4, 4);
bad68ec9 1503 extsize = ntohl(extsize);
4aab5b46 1504 if (read_index_extension(istate,
7a51ed66
LT
1505 (const char *) mmap + src_offset,
1506 (char *) mmap + src_offset + 8,
1d7f171c 1507 extsize) < 0)
bad68ec9 1508 goto unmap;
7a51ed66
LT
1509 src_offset += 8;
1510 src_offset += extsize;
bad68ec9 1511 }
7a51ed66 1512 munmap(mmap, mmap_size);
4aab5b46 1513 return istate->cache_nr;
e83c5163
LT
1514
1515unmap:
7a51ed66 1516 munmap(mmap, mmap_size);
5d1a5c02 1517 die("index file corrupt");
e83c5163
LT
1518}
1519
fa7b3c2f
JH
1520int is_index_unborn(struct index_state *istate)
1521{
debed2a6 1522 return (!istate->cache_nr && !istate->timestamp.sec);
fa7b3c2f
JH
1523}
1524
4aab5b46 1525int discard_index(struct index_state *istate)
6d297f81 1526{
debed2a6
RS
1527 int i;
1528
1529 for (i = 0; i < istate->cache_nr; i++)
1530 free(istate->cache[i]);
cfc5789a 1531 resolve_undo_clear_index(istate);
4aab5b46
JH
1532 istate->cache_nr = 0;
1533 istate->cache_changed = 0;
fba2f38a
KB
1534 istate->timestamp.sec = 0;
1535 istate->timestamp.nsec = 0;
2092678c 1536 free_name_hash(istate);
4aab5b46 1537 cache_tree_free(&(istate->cache_tree));
913e0e99 1538 istate->initialized = 0;
a0fc4db0
RS
1539 free(istate->cache);
1540 istate->cache = NULL;
1541 istate->cache_alloc = 0;
7a51ed66 1542 return 0;
6d297f81
JS
1543}
1544
d1f128b0 1545int unmerged_index(const struct index_state *istate)
94a5728c
DB
1546{
1547 int i;
1548 for (i = 0; i < istate->cache_nr; i++) {
1549 if (ce_stage(istate->cache[i]))
1550 return 1;
1551 }
1552 return 0;
1553}
1554
4990aadc 1555#define WRITE_BUFFER_SIZE 8192
bf0f910d 1556static unsigned char write_buffer[WRITE_BUFFER_SIZE];
4990aadc
LT
1557static unsigned long write_buffer_len;
1558
9126f009 1559static int ce_write_flush(git_SHA_CTX *context, int fd)
6015c28b
JH
1560{
1561 unsigned int buffered = write_buffer_len;
1562 if (buffered) {
9126f009 1563 git_SHA1_Update(context, write_buffer, buffered);
93822c22 1564 if (write_in_full(fd, write_buffer, buffered) != buffered)
6015c28b
JH
1565 return -1;
1566 write_buffer_len = 0;
1567 }
1568 return 0;
1569}
1570
9126f009 1571static int ce_write(git_SHA_CTX *context, int fd, void *data, unsigned int len)
4990aadc
LT
1572{
1573 while (len) {
1574 unsigned int buffered = write_buffer_len;
1575 unsigned int partial = WRITE_BUFFER_SIZE - buffered;
1576 if (partial > len)
1577 partial = len;
1578 memcpy(write_buffer + buffered, data, partial);
1579 buffered += partial;
1580 if (buffered == WRITE_BUFFER_SIZE) {
6015c28b
JH
1581 write_buffer_len = buffered;
1582 if (ce_write_flush(context, fd))
4990aadc
LT
1583 return -1;
1584 buffered = 0;
1585 }
1586 write_buffer_len = buffered;
1587 len -= partial;
1d7f171c 1588 data = (char *) data + partial;
a6080a0a
JH
1589 }
1590 return 0;
4990aadc
LT
1591}
1592
9126f009 1593static int write_index_ext_header(git_SHA_CTX *context, int fd,
ac58c7b1 1594 unsigned int ext, unsigned int sz)
bad68ec9
JH
1595{
1596 ext = htonl(ext);
1597 sz = htonl(sz);
968a1d65
DR
1598 return ((ce_write(context, fd, &ext, 4) < 0) ||
1599 (ce_write(context, fd, &sz, 4) < 0)) ? -1 : 0;
bad68ec9
JH
1600}
1601
9126f009 1602static int ce_flush(git_SHA_CTX *context, int fd)
4990aadc
LT
1603{
1604 unsigned int left = write_buffer_len;
ca9be054 1605
4990aadc
LT
1606 if (left) {
1607 write_buffer_len = 0;
9126f009 1608 git_SHA1_Update(context, write_buffer, left);
4990aadc 1609 }
ca9be054 1610
2c865d9a
QH
1611 /* Flush first if not enough space for SHA1 signature */
1612 if (left + 20 > WRITE_BUFFER_SIZE) {
93822c22 1613 if (write_in_full(fd, write_buffer, left) != left)
2c865d9a
QH
1614 return -1;
1615 left = 0;
1616 }
1617
ca9be054 1618 /* Append the SHA1 signature at the end */
9126f009 1619 git_SHA1_Final(write_buffer + left, context);
ca9be054 1620 left += 20;
93822c22 1621 return (write_in_full(fd, write_buffer, left) != left) ? -1 : 0;
4990aadc
LT
1622}
1623
407c8eb0
JH
1624static void ce_smudge_racily_clean_entry(struct cache_entry *ce)
1625{
1626 /*
1627 * The only thing we care about in this function is to smudge the
1628 * falsely clean entry due to touch-update-touch race, so we leave
1629 * everything else as they are. We are called for entries whose
c21d39d7 1630 * ce_stat_data.sd_mtime match the index file mtime.
c70115b4
JH
1631 *
1632 * Note that this actually does not do much for gitlinks, for
1633 * which ce_match_stat_basic() always goes to the actual
1634 * contents. The caller checks with is_racy_timestamp() which
1635 * always says "no" for gitlinks, so we are not called for them ;-)
407c8eb0
JH
1636 */
1637 struct stat st;
1638
1639 if (lstat(ce->name, &st) < 0)
1640 return;
1641 if (ce_match_stat_basic(ce, &st))
1642 return;
1643 if (ce_modified_check_fs(ce, &st)) {
4b3511b0
JH
1644 /* This is "racily clean"; smudge it. Note that this
1645 * is a tricky code. At first glance, it may appear
1646 * that it can break with this sequence:
1647 *
1648 * $ echo xyzzy >frotz
1649 * $ git-update-index --add frotz
1650 * $ : >frotz
1651 * $ sleep 3
1652 * $ echo filfre >nitfol
1653 * $ git-update-index --add nitfol
1654 *
b7e58b17 1655 * but it does not. When the second update-index runs,
4b3511b0
JH
1656 * it notices that the entry "frotz" has the same timestamp
1657 * as index, and if we were to smudge it by resetting its
1658 * size to zero here, then the object name recorded
1659 * in index is the 6-byte file but the cached stat information
1660 * becomes zero --- which would then match what we would
a6080a0a 1661 * obtain from the filesystem next time we stat("frotz").
4b3511b0
JH
1662 *
1663 * However, the second update-index, before calling
1664 * this function, notices that the cached size is 6
1665 * bytes and what is on the filesystem is an empty
1666 * file, and never calls us, so the cached size information
1667 * for "frotz" stays 6 which does not match the filesystem.
1668 */
c21d39d7 1669 ce->ce_stat_data.sd_size = 0;
407c8eb0
JH
1670 }
1671}
1672
f136f7bf
JH
1673/* Copy miscellaneous fields but not the name */
1674static char *copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk,
1675 struct cache_entry *ce)
7a51ed66 1676{
b60e188c
TG
1677 short flags;
1678
c21d39d7
MH
1679 ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec);
1680 ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec);
1681 ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec);
1682 ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec);
1683 ondisk->dev = htonl(ce->ce_stat_data.sd_dev);
1684 ondisk->ino = htonl(ce->ce_stat_data.sd_ino);
7a51ed66 1685 ondisk->mode = htonl(ce->ce_mode);
c21d39d7
MH
1686 ondisk->uid = htonl(ce->ce_stat_data.sd_uid);
1687 ondisk->gid = htonl(ce->ce_stat_data.sd_gid);
1688 ondisk->size = htonl(ce->ce_stat_data.sd_size);
7a51ed66 1689 hashcpy(ondisk->sha1, ce->sha1);
b60e188c
TG
1690
1691 flags = ce->ce_flags;
1692 flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce));
1693 ondisk->flags = htons(flags);
06aaaa0b
NTND
1694 if (ce->ce_flags & CE_EXTENDED) {
1695 struct ondisk_cache_entry_extended *ondisk2;
1696 ondisk2 = (struct ondisk_cache_entry_extended *)ondisk;
1697 ondisk2->flags2 = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16);
f136f7bf 1698 return ondisk2->name;
06aaaa0b 1699 }
f136f7bf
JH
1700 else {
1701 return ondisk->name;
1702 }
1703}
1704
9d227781
JH
1705static int ce_write_entry(git_SHA_CTX *c, int fd, struct cache_entry *ce,
1706 struct strbuf *previous_name)
f136f7bf 1707{
9d227781
JH
1708 int size;
1709 struct ondisk_cache_entry *ondisk;
f136f7bf
JH
1710 char *name;
1711 int result;
1712
9d227781
JH
1713 if (!previous_name) {
1714 size = ondisk_ce_size(ce);
1715 ondisk = xcalloc(1, size);
1716 name = copy_cache_entry_to_ondisk(ondisk, ce);
1717 memcpy(name, ce->name, ce_namelen(ce));
1718 } else {
1719 int common, to_remove, prefix_size;
1720 unsigned char to_remove_vi[16];
1721 for (common = 0;
1722 (ce->name[common] &&
1723 common < previous_name->len &&
1724 ce->name[common] == previous_name->buf[common]);
1725 common++)
1726 ; /* still matching */
1727 to_remove = previous_name->len - common;
1728 prefix_size = encode_varint(to_remove, to_remove_vi);
1729
1730 if (ce->ce_flags & CE_EXTENDED)
1731 size = offsetof(struct ondisk_cache_entry_extended, name);
1732 else
1733 size = offsetof(struct ondisk_cache_entry, name);
1734 size += prefix_size + (ce_namelen(ce) - common + 1);
1735
1736 ondisk = xcalloc(1, size);
1737 name = copy_cache_entry_to_ondisk(ondisk, ce);
1738 memcpy(name, to_remove_vi, prefix_size);
1739 memcpy(name + prefix_size, ce->name + common, ce_namelen(ce) - common);
1740
1741 strbuf_splice(previous_name, common, to_remove,
1742 ce->name + common, ce_namelen(ce) - common);
06aaaa0b 1743 }
7a51ed66 1744
59efba64
JN
1745 result = ce_write(c, fd, ondisk, size);
1746 free(ondisk);
1747 return result;
7a51ed66
LT
1748}
1749
483fbe2b
JH
1750static int has_racy_timestamp(struct index_state *istate)
1751{
1752 int entries = istate->cache_nr;
1753 int i;
1754
1755 for (i = 0; i < entries; i++) {
1756 struct cache_entry *ce = istate->cache[i];
1757 if (is_racy_timestamp(istate, ce))
1758 return 1;
1759 }
1760 return 0;
1761}
1762
ccdc4ec3 1763/*
98e023de 1764 * Opportunistically update the index but do not complain if we can't
ccdc4ec3
JH
1765 */
1766void update_index_if_able(struct index_state *istate, struct lock_file *lockfile)
1767{
483fbe2b 1768 if ((istate->cache_changed || has_racy_timestamp(istate)) &&
ccdc4ec3
JH
1769 !write_index(istate, lockfile->fd))
1770 commit_locked_index(lockfile);
1771 else
1772 rollback_lock_file(lockfile);
1773}
1774
e1afca4f 1775int write_index(struct index_state *istate, int newfd)
197ee8c9 1776{
9126f009 1777 git_SHA_CTX c;
197ee8c9 1778 struct cache_header hdr;
9d227781 1779 int i, err, removed, extended, hdr_version;
4aab5b46
JH
1780 struct cache_entry **cache = istate->cache;
1781 int entries = istate->cache_nr;
e1afca4f 1782 struct stat st;
9d227781 1783 struct strbuf previous_name_buf = STRBUF_INIT, *previous_name;
025a0709 1784
06aaaa0b 1785 for (i = removed = extended = 0; i < entries; i++) {
7a51ed66 1786 if (cache[i]->ce_flags & CE_REMOVE)
025a0709 1787 removed++;
197ee8c9 1788
06aaaa0b
NTND
1789 /* reduce extended entries if possible */
1790 cache[i]->ce_flags &= ~CE_EXTENDED;
1791 if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) {
1792 extended++;
1793 cache[i]->ce_flags |= CE_EXTENDED;
1794 }
1795 }
1796
9d227781
JH
1797 if (!istate->version)
1798 istate->version = INDEX_FORMAT_DEFAULT;
1799
1800 /* demote version 3 to version 2 when the latter suffices */
1801 if (istate->version == 3 || istate->version == 2)
1802 istate->version = extended ? 3 : 2;
1803
1804 hdr_version = istate->version;
1805
ccc4feb5 1806 hdr.hdr_signature = htonl(CACHE_SIGNATURE);
9d227781 1807 hdr.hdr_version = htonl(hdr_version);
025a0709 1808 hdr.hdr_entries = htonl(entries - removed);
197ee8c9 1809
9126f009 1810 git_SHA1_Init(&c);
ca9be054 1811 if (ce_write(&c, newfd, &hdr, sizeof(hdr)) < 0)
197ee8c9
LT
1812 return -1;
1813
9d227781 1814 previous_name = (hdr_version == 4) ? &previous_name_buf : NULL;
197ee8c9
LT
1815 for (i = 0; i < entries; i++) {
1816 struct cache_entry *ce = cache[i];
7a51ed66 1817 if (ce->ce_flags & CE_REMOVE)
aa16021e 1818 continue;
e06c43c7 1819 if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce))
407c8eb0 1820 ce_smudge_racily_clean_entry(ce);
83bd7437
JK
1821 if (is_null_sha1(ce->sha1)) {
1822 static const char msg[] = "cache entry has null sha1: %s";
1823 static int allow = -1;
1824
1825 if (allow < 0)
1826 allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0);
1827 if (allow)
1828 warning(msg, ce->name);
1829 else
1830 return error(msg, ce->name);
1831 }
9d227781 1832 if (ce_write_entry(&c, newfd, ce, previous_name) < 0)
197ee8c9
LT
1833 return -1;
1834 }
9d227781 1835 strbuf_release(&previous_name_buf);
1af1c2b6 1836
bad68ec9 1837 /* Write extension data here */
4aab5b46 1838 if (istate->cache_tree) {
f285a2d7 1839 struct strbuf sb = STRBUF_INIT;
1dffb8fa 1840
1dffb8fa
PH
1841 cache_tree_write(&sb, istate->cache_tree);
1842 err = write_index_ext_header(&c, newfd, CACHE_EXT_TREE, sb.len) < 0
1843 || ce_write(&c, newfd, sb.buf, sb.len) < 0;
1844 strbuf_release(&sb);
1845 if (err)
bad68ec9 1846 return -1;
bad68ec9 1847 }
cfc5789a
JH
1848 if (istate->resolve_undo) {
1849 struct strbuf sb = STRBUF_INIT;
1850
1851 resolve_undo_write(&sb, istate->resolve_undo);
1852 err = write_index_ext_header(&c, newfd, CACHE_EXT_RESOLVE_UNDO,
1853 sb.len) < 0
1854 || ce_write(&c, newfd, sb.buf, sb.len) < 0;
1855 strbuf_release(&sb);
1856 if (err)
1857 return -1;
1858 }
e1afca4f
KB
1859
1860 if (ce_flush(&c, newfd) || fstat(newfd, &st))
1861 return -1;
5bcf109c
KB
1862 istate->timestamp.sec = (unsigned int)st.st_mtime;
1863 istate->timestamp.nsec = ST_MTIME_NSEC(st);
e1afca4f 1864 return 0;
197ee8c9 1865}
e46bbcf6
MV
1866
1867/*
1868 * Read the index file that is potentially unmerged into given
63e8dc5b 1869 * index_state, dropping any unmerged entries. Returns true if
e46bbcf6
MV
1870 * the index is unmerged. Callers who want to refuse to work
1871 * from an unmerged state can call this and check its return value,
1872 * instead of calling read_cache().
1873 */
1874int read_index_unmerged(struct index_state *istate)
1875{
1876 int i;
d1a43f2a 1877 int unmerged = 0;
e46bbcf6
MV
1878
1879 read_index(istate);
e46bbcf6
MV
1880 for (i = 0; i < istate->cache_nr; i++) {
1881 struct cache_entry *ce = istate->cache[i];
d1a43f2a
JH
1882 struct cache_entry *new_ce;
1883 int size, len;
1884
1885 if (!ce_stage(ce))
e46bbcf6 1886 continue;
d1a43f2a 1887 unmerged = 1;
68c4f6a5 1888 len = ce_namelen(ce);
d1a43f2a
JH
1889 size = cache_entry_size(len);
1890 new_ce = xcalloc(1, size);
d1a43f2a 1891 memcpy(new_ce->name, ce->name, len);
b60e188c
TG
1892 new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED;
1893 new_ce->ce_namelen = len;
d1a43f2a
JH
1894 new_ce->ce_mode = ce->ce_mode;
1895 if (add_index_entry(istate, new_ce, 0))
1896 return error("%s: cannot drop to stage #0",
1897 ce->name);
1898 i = index_name_pos(istate, new_ce->name, len);
e46bbcf6 1899 }
d1a43f2a 1900 return unmerged;
e46bbcf6 1901}
041aee31 1902
98fa4738
JK
1903/*
1904 * Returns 1 if the path is an "other" path with respect to
1905 * the index; that is, the path is not mentioned in the index at all,
1906 * either as a file, a directory with some files in the index,
1907 * or as an unmerged entry.
1908 *
1909 * We helpfully remove a trailing "/" from directories so that
1910 * the output of read_directory can be used as-is.
1911 */
1912int index_name_is_other(const struct index_state *istate, const char *name,
1913 int namelen)
1914{
1915 int pos;
1916 if (namelen && name[namelen - 1] == '/')
1917 namelen--;
1918 pos = index_name_pos(istate, name, namelen);
1919 if (0 <= pos)
1920 return 0; /* exact match */
1921 pos = -pos - 1;
1922 if (pos < istate->cache_nr) {
1923 struct cache_entry *ce = istate->cache[pos];
1924 if (ce_namelen(ce) == namelen &&
1925 !memcmp(ce->name, name, namelen))
1926 return 0; /* Yup, this one exists unmerged */
1927 }
1928 return 1;
1929}
29fb37b2 1930
ff366825 1931void *read_blob_data_from_index(struct index_state *istate, const char *path, unsigned long *size)
29fb37b2
LF
1932{
1933 int pos, len;
1934 unsigned long sz;
1935 enum object_type type;
1936 void *data;
1937
1938 len = strlen(path);
1939 pos = index_name_pos(istate, path, len);
1940 if (pos < 0) {
1941 /*
1942 * We might be in the middle of a merge, in which
1943 * case we would read stage #2 (ours).
1944 */
1945 int i;
1946 for (i = -pos - 1;
1947 (pos < 0 && i < istate->cache_nr &&
1948 !strcmp(istate->cache[i]->name, path));
1949 i++)
1950 if (ce_stage(istate->cache[i]) == 2)
1951 pos = i;
1952 }
1953 if (pos < 0)
1954 return NULL;
1955 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
1956 if (!data || type != OBJ_BLOB) {
1957 free(data);
1958 return NULL;
1959 }
ff366825
LF
1960 if (size)
1961 *size = sz;
29fb37b2
LF
1962 return data;
1963}
38612532
MH
1964
1965void stat_validity_clear(struct stat_validity *sv)
1966{
1967 free(sv->sd);
1968 sv->sd = NULL;
1969}
1970
1971int stat_validity_check(struct stat_validity *sv, const char *path)
1972{
1973 struct stat st;
1974
1975 if (stat(path, &st) < 0)
1976 return sv->sd == NULL;
1977 if (!sv->sd)
1978 return 0;
1979 return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st);
1980}
1981
1982void stat_validity_update(struct stat_validity *sv, int fd)
1983{
1984 struct stat st;
1985
1986 if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode))
1987 stat_validity_clear(sv);
1988 else {
1989 if (!sv->sd)
1990 sv->sd = xcalloc(1, sizeof(struct stat_data));
1991 fill_stat_data(sv->sd, &st);
1992 }
1993}