lockfile.h: extract new header file for the functions in lockfile.c
[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"
697cc8ef 8#include "lockfile.h"
bad68ec9 9#include "cache-tree.h"
f35a6d3b 10#include "refs.h"
d616813d 11#include "dir.h"
041aee31
JH
12#include "tree.h"
13#include "commit.h"
39425819 14#include "blob.h"
cfc5789a 15#include "resolve-undo.h"
6c9cd161
JH
16#include "strbuf.h"
17#include "varint.h"
5fc2fc8f 18#include "split-index.h"
c18b80a0 19#include "sigchain.h"
bad68ec9 20
25762726
BK
21static struct cache_entry *refresh_cache_entry(struct cache_entry *ce,
22 unsigned int options);
87b29e5a 23
b60e188c
TG
24/* Mask for the name length in ce_flags in the on-disk index */
25
26#define CE_NAMEMASK (0x0fff)
27
bad68ec9
JH
28/* Index extensions.
29 *
30 * The first letter should be 'A'..'Z' for extensions that are not
31 * necessary for a correct operation (i.e. optimization data).
32 * When new extensions are added that _needs_ to be understood in
33 * order to correctly interpret the index file, pick character that
34 * is outside the range, to cause the reader to abort.
35 */
36
37#define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) )
38#define CACHE_EXT_TREE 0x54524545 /* "TREE" */
b659b49b 39#define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */
5fc2fc8f
NTND
40#define CACHE_EXT_LINK 0x6c696e6b /* "link" */
41
42/* changes that can be kept in $GIT_DIR/index (basically all extensions) */
e0cf0d7d 43#define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \
c18b80a0
NTND
44 CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \
45 SPLIT_INDEX_ORDERED)
e83c5163 46
228e94f9 47struct index_state the_index;
626f35c8 48static const char *alternate_index_output;
8fd2cb40 49
9cb76b8c
JH
50static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce)
51{
52 istate->cache[nr] = ce;
96872bc2 53 add_name_hash(istate, ce);
9cb76b8c
JH
54}
55
cf558704
LT
56static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce)
57{
58 struct cache_entry *old = istate->cache[nr];
59
078a58e8 60 replace_index_entry_in_base(istate, old, ce);
2092678c 61 remove_name_hash(istate, old);
5699d17e 62 free(old);
a22c6371 63 set_index_entry(istate, nr, ce);
078a58e8 64 ce->ce_flags |= CE_UPDATE_IN_BASE;
e636a7b4 65 istate->cache_changed |= CE_ENTRY_CHANGED;
cf558704
LT
66}
67
81dc2307
PB
68void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name)
69{
70 struct cache_entry *old = istate->cache[nr], *new;
71 int namelen = strlen(new_name);
72
73 new = xmalloc(cache_entry_size(namelen));
74 copy_cache_entry(new, old);
419a597f 75 new->ce_flags &= ~CE_HASHED;
b60e188c 76 new->ce_namelen = namelen;
5fc2fc8f 77 new->index = 0;
81dc2307
PB
78 memcpy(new->name, new_name, namelen + 1);
79
a5400efe 80 cache_tree_invalidate_path(istate, old->name);
81dc2307
PB
81 remove_index_entry_at(istate, nr);
82 add_index_entry(istate, new, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
83}
84
c21d39d7
MH
85void fill_stat_data(struct stat_data *sd, struct stat *st)
86{
87 sd->sd_ctime.sec = (unsigned int)st->st_ctime;
88 sd->sd_mtime.sec = (unsigned int)st->st_mtime;
89 sd->sd_ctime.nsec = ST_CTIME_NSEC(*st);
90 sd->sd_mtime.nsec = ST_MTIME_NSEC(*st);
91 sd->sd_dev = st->st_dev;
92 sd->sd_ino = st->st_ino;
93 sd->sd_uid = st->st_uid;
94 sd->sd_gid = st->st_gid;
95 sd->sd_size = st->st_size;
96}
97
98int match_stat_data(const struct stat_data *sd, struct stat *st)
99{
100 int changed = 0;
101
102 if (sd->sd_mtime.sec != (unsigned int)st->st_mtime)
103 changed |= MTIME_CHANGED;
104 if (trust_ctime && check_stat &&
105 sd->sd_ctime.sec != (unsigned int)st->st_ctime)
106 changed |= CTIME_CHANGED;
107
108#ifdef USE_NSEC
109 if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st))
110 changed |= MTIME_CHANGED;
111 if (trust_ctime && check_stat &&
112 sd->sd_ctime.nsec != ST_CTIME_NSEC(*st))
113 changed |= CTIME_CHANGED;
114#endif
115
116 if (check_stat) {
117 if (sd->sd_uid != (unsigned int) st->st_uid ||
118 sd->sd_gid != (unsigned int) st->st_gid)
119 changed |= OWNER_CHANGED;
120 if (sd->sd_ino != (unsigned int) st->st_ino)
121 changed |= INODE_CHANGED;
122 }
123
124#ifdef USE_STDEV
125 /*
126 * st_dev breaks on network filesystems where different
127 * clients will have different views of what "device"
128 * the filesystem is on
129 */
130 if (check_stat && sd->sd_dev != (unsigned int) st->st_dev)
131 changed |= INODE_CHANGED;
132#endif
133
134 if (sd->sd_size != (unsigned int) st->st_size)
135 changed |= DATA_CHANGED;
136
137 return changed;
138}
139
415e96c8
JH
140/*
141 * This only updates the "non-critical" parts of the directory
142 * cache, ie the parts that aren't tracked by GIT, and only used
143 * to validate the cache.
144 */
145void fill_stat_cache_info(struct cache_entry *ce, struct stat *st)
146{
c21d39d7 147 fill_stat_data(&ce->ce_stat_data, st);
5f73076c
JH
148
149 if (assume_unchanged)
7a51ed66 150 ce->ce_flags |= CE_VALID;
eadb5831
JH
151
152 if (S_ISREG(st->st_mode))
153 ce_mark_uptodate(ce);
415e96c8
JH
154}
155
21a6b9fa 156static int ce_compare_data(const struct cache_entry *ce, struct stat *st)
29e4d363
JH
157{
158 int match = -1;
159 int fd = open(ce->name, O_RDONLY);
160
161 if (fd >= 0) {
162 unsigned char sha1[20];
c4ce46fc 163 if (!index_fd(sha1, fd, st, OBJ_BLOB, ce->name, 0))
a89fccd2 164 match = hashcmp(sha1, ce->sha1);
7f8508e8 165 /* index_fd() closed the file descriptor already */
29e4d363
JH
166 }
167 return match;
168}
169
21a6b9fa 170static int ce_compare_link(const struct cache_entry *ce, size_t expected_size)
29e4d363
JH
171{
172 int match = -1;
29e4d363
JH
173 void *buffer;
174 unsigned long size;
21666f1a 175 enum object_type type;
a60272b3 176 struct strbuf sb = STRBUF_INIT;
29e4d363 177
a60272b3 178 if (strbuf_readlink(&sb, ce->name, expected_size))
29e4d363 179 return -1;
a60272b3 180
21666f1a 181 buffer = read_sha1_file(ce->sha1, &type, &size);
a60272b3
LT
182 if (buffer) {
183 if (size == sb.len)
184 match = memcmp(buffer, sb.buf, size);
185 free(buffer);
29e4d363 186 }
a60272b3 187 strbuf_release(&sb);
29e4d363
JH
188 return match;
189}
190
21a6b9fa 191static int ce_compare_gitlink(const struct cache_entry *ce)
f35a6d3b
LT
192{
193 unsigned char sha1[20];
194
195 /*
196 * We don't actually require that the .git directory
302b9282 197 * under GITLINK directory be a valid git directory. It
f35a6d3b
LT
198 * might even be missing (in case nobody populated that
199 * sub-project).
200 *
201 * If so, we consider it always to match.
202 */
203 if (resolve_gitlink_ref(ce->name, "HEAD", sha1) < 0)
204 return 0;
205 return hashcmp(sha1, ce->sha1);
206}
207
21a6b9fa 208static int ce_modified_check_fs(const struct cache_entry *ce, struct stat *st)
29e4d363
JH
209{
210 switch (st->st_mode & S_IFMT) {
211 case S_IFREG:
212 if (ce_compare_data(ce, st))
213 return DATA_CHANGED;
214 break;
215 case S_IFLNK:
dc49cd76 216 if (ce_compare_link(ce, xsize_t(st->st_size)))
29e4d363
JH
217 return DATA_CHANGED;
218 break;
a8ee75bc 219 case S_IFDIR:
7a51ed66 220 if (S_ISGITLINK(ce->ce_mode))
c70115b4 221 return ce_compare_gitlink(ce) ? DATA_CHANGED : 0;
29e4d363
JH
222 default:
223 return TYPE_CHANGED;
224 }
225 return 0;
226}
227
21a6b9fa 228static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st)
734aab75
LT
229{
230 unsigned int changed = 0;
231
7a51ed66
LT
232 if (ce->ce_flags & CE_REMOVE)
233 return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED;
234
235 switch (ce->ce_mode & S_IFMT) {
8ae0a8c5
KS
236 case S_IFREG:
237 changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0;
3e09cdfd
JH
238 /* We consider only the owner x bit to be relevant for
239 * "mode changes"
240 */
241 if (trust_executable_bit &&
7a51ed66 242 (0100 & (ce->ce_mode ^ st->st_mode)))
ffbe1add 243 changed |= MODE_CHANGED;
8ae0a8c5
KS
244 break;
245 case S_IFLNK:
78a8d641
JS
246 if (!S_ISLNK(st->st_mode) &&
247 (has_symlinks || !S_ISREG(st->st_mode)))
248 changed |= TYPE_CHANGED;
8ae0a8c5 249 break;
302b9282 250 case S_IFGITLINK:
c70115b4 251 /* We ignore most of the st_xxx fields for gitlinks */
f35a6d3b
LT
252 if (!S_ISDIR(st->st_mode))
253 changed |= TYPE_CHANGED;
254 else if (ce_compare_gitlink(ce))
255 changed |= DATA_CHANGED;
a8ee75bc 256 return changed;
8ae0a8c5 257 default:
7a51ed66 258 die("internal error: ce_mode is %o", ce->ce_mode);
8ae0a8c5 259 }
2cb45e95 260
c21d39d7 261 changed |= match_stat_data(&ce->ce_stat_data, st);
b0391890 262
f49c2c22 263 /* Racily smudged entry? */
c21d39d7 264 if (!ce->ce_stat_data.sd_size) {
f49c2c22
LT
265 if (!is_empty_blob_sha1(ce->sha1))
266 changed |= DATA_CHANGED;
267 }
268
407c8eb0
JH
269 return changed;
270}
271
21a6b9fa
RS
272static int is_racy_timestamp(const struct index_state *istate,
273 const struct cache_entry *ce)
6d91da6d 274{
050288d5 275 return (!S_ISGITLINK(ce->ce_mode) &&
fba2f38a
KB
276 istate->timestamp.sec &&
277#ifdef USE_NSEC
278 /* nanosecond timestamped files can also be racy! */
c21d39d7
MH
279 (istate->timestamp.sec < ce->ce_stat_data.sd_mtime.sec ||
280 (istate->timestamp.sec == ce->ce_stat_data.sd_mtime.sec &&
281 istate->timestamp.nsec <= ce->ce_stat_data.sd_mtime.nsec))
fba2f38a 282#else
c21d39d7 283 istate->timestamp.sec <= ce->ce_stat_data.sd_mtime.sec
fba2f38a
KB
284#endif
285 );
6d91da6d
JH
286}
287
d1f128b0 288int ie_match_stat(const struct index_state *istate,
21a6b9fa 289 const struct cache_entry *ce, struct stat *st,
4bd5b7da 290 unsigned int options)
407c8eb0 291{
5f73076c 292 unsigned int changed;
4bd5b7da 293 int ignore_valid = options & CE_MATCH_IGNORE_VALID;
56cac48c 294 int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE;
4bd5b7da 295 int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY;
5f73076c
JH
296
297 /*
298 * If it's marked as always valid in the index, it's
299 * valid whatever the checked-out copy says.
56cac48c
NTND
300 *
301 * skip-worktree has the same effect with higher precedence
5f73076c 302 */
56cac48c
NTND
303 if (!ignore_skip_worktree && ce_skip_worktree(ce))
304 return 0;
7a51ed66 305 if (!ignore_valid && (ce->ce_flags & CE_VALID))
5f73076c
JH
306 return 0;
307
331fcb59
JH
308 /*
309 * Intent-to-add entries have not been added, so the index entry
310 * by definition never matches what is in the work tree until it
311 * actually gets added.
312 */
313 if (ce->ce_flags & CE_INTENT_TO_ADD)
314 return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED;
315
5f73076c 316 changed = ce_match_stat_basic(ce, st);
407c8eb0 317
29e4d363
JH
318 /*
319 * Within 1 second of this sequence:
320 * echo xyzzy >file && git-update-index --add file
321 * running this command:
322 * echo frotz >file
323 * would give a falsely clean cache entry. The mtime and
324 * length match the cache, and other stat fields do not change.
325 *
326 * We could detect this at update-index time (the cache entry
327 * being registered/updated records the same time as "now")
328 * and delay the return from git-update-index, but that would
329 * effectively mean we can make at most one commit per second,
330 * which is not acceptable. Instead, we check cache entries
331 * whose mtime are the same as the index file timestamp more
5f73076c 332 * carefully than others.
29e4d363 333 */
6d91da6d 334 if (!changed && is_racy_timestamp(istate, ce)) {
42f77406
JH
335 if (assume_racy_is_modified)
336 changed |= DATA_CHANGED;
337 else
338 changed |= ce_modified_check_fs(ce, st);
339 }
b0391890 340
29e4d363 341 return changed;
b0391890
JH
342}
343
d1f128b0 344int ie_modified(const struct index_state *istate,
21a6b9fa
RS
345 const struct cache_entry *ce,
346 struct stat *st, unsigned int options)
b0391890 347{
29e4d363 348 int changed, changed_fs;
4bd5b7da
JH
349
350 changed = ie_match_stat(istate, ce, st, options);
b0391890
JH
351 if (!changed)
352 return 0;
b0391890
JH
353 /*
354 * If the mode or type has changed, there's no point in trying
355 * to refresh the entry - it's not going to match
356 */
357 if (changed & (MODE_CHANGED | TYPE_CHANGED))
358 return changed;
359
c70115b4
JH
360 /*
361 * Immediately after read-tree or update-index --cacheinfo,
362 * the length field is zero, as we have never even read the
363 * lstat(2) information once, and we cannot trust DATA_CHANGED
364 * returned by ie_match_stat() which in turn was returned by
365 * ce_match_stat_basic() to signal that the filesize of the
366 * blob changed. We have to actually go to the filesystem to
367 * see if the contents match, and if so, should answer "unchanged".
368 *
369 * The logic does not apply to gitlinks, as ce_match_stat_basic()
370 * already has checked the actual HEAD from the filesystem in the
371 * subproject. If ie_match_stat() already said it is different,
372 * then we know it is.
b0391890 373 */
c70115b4 374 if ((changed & DATA_CHANGED) &&
c21d39d7 375 (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0))
b0391890
JH
376 return changed;
377
29e4d363
JH
378 changed_fs = ce_modified_check_fs(ce, st);
379 if (changed_fs)
380 return changed | changed_fs;
b0391890
JH
381 return 0;
382}
383
958ba6c9
LT
384int base_name_compare(const char *name1, int len1, int mode1,
385 const char *name2, int len2, int mode2)
386{
387 unsigned char c1, c2;
388 int len = len1 < len2 ? len1 : len2;
389 int cmp;
390
391 cmp = memcmp(name1, name2, len);
392 if (cmp)
393 return cmp;
394 c1 = name1[len];
395 c2 = name2[len];
1833a925 396 if (!c1 && S_ISDIR(mode1))
958ba6c9 397 c1 = '/';
1833a925 398 if (!c2 && S_ISDIR(mode2))
958ba6c9
LT
399 c2 = '/';
400 return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0;
401}
402
0ab9e1e8
LT
403/*
404 * df_name_compare() is identical to base_name_compare(), except it
405 * compares conflicting directory/file entries as equal. Note that
406 * while a directory name compares as equal to a regular file, they
407 * then individually compare _differently_ to a filename that has
408 * a dot after the basename (because '\0' < '.' < '/').
409 *
410 * This is used by routines that want to traverse the git namespace
411 * but then handle conflicting entries together when possible.
412 */
413int df_name_compare(const char *name1, int len1, int mode1,
414 const char *name2, int len2, int mode2)
415{
416 int len = len1 < len2 ? len1 : len2, cmp;
417 unsigned char c1, c2;
418
419 cmp = memcmp(name1, name2, len);
420 if (cmp)
421 return cmp;
422 /* Directories and files compare equal (same length, same name) */
423 if (len1 == len2)
424 return 0;
425 c1 = name1[len];
426 if (!c1 && S_ISDIR(mode1))
427 c1 = '/';
428 c2 = name2[len];
429 if (!c2 && S_ISDIR(mode2))
430 c2 = '/';
431 if (c1 == '/' && !c2)
432 return 0;
433 if (c2 == '/' && !c1)
434 return 0;
435 return c1 - c2;
436}
437
ccdd4a0f 438int name_compare(const char *name1, size_t len1, const char *name2, size_t len2)
eb38c22f 439{
ccdd4a0f
JM
440 size_t min_len = (len1 < len2) ? len1 : len2;
441 int cmp = memcmp(name1, name2, min_len);
eb38c22f
LT
442 if (cmp)
443 return cmp;
444 if (len1 < len2)
445 return -1;
446 if (len1 > len2)
447 return 1;
ccdd4a0f
JM
448 return 0;
449}
450
451int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2)
452{
453 int cmp;
454
455 cmp = name_compare(name1, len1, name2, len2);
456 if (cmp)
457 return cmp;
5f73076c 458
b60e188c 459 if (stage1 < stage2)
95fd5bf8 460 return -1;
b60e188c 461 if (stage1 > stage2)
95fd5bf8 462 return 1;
eb38c22f
LT
463 return 0;
464}
465
357e9c69 466static int index_name_stage_pos(const struct index_state *istate, const char *name, int namelen, int stage)
eb38c22f
LT
467{
468 int first, last;
469
470 first = 0;
4aab5b46 471 last = istate->cache_nr;
eb38c22f
LT
472 while (last > first) {
473 int next = (last + first) >> 1;
4aab5b46 474 struct cache_entry *ce = istate->cache[next];
b60e188c 475 int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce));
eb38c22f 476 if (!cmp)
76e7f4ec 477 return next;
eb38c22f
LT
478 if (cmp < 0) {
479 last = next;
480 continue;
481 }
482 first = next+1;
483 }
76e7f4ec 484 return -first-1;
eb38c22f
LT
485}
486
b60e188c
TG
487int index_name_pos(const struct index_state *istate, const char *name, int namelen)
488{
489 return index_name_stage_pos(istate, name, namelen, 0);
490}
491
7b937ca3 492/* Remove entry, return true if there are more entries to go.. */
4aab5b46 493int remove_index_entry_at(struct index_state *istate, int pos)
7b937ca3 494{
cf558704
LT
495 struct cache_entry *ce = istate->cache[pos];
496
cfc5789a 497 record_resolve_undo(istate, ce);
2092678c 498 remove_name_hash(istate, ce);
045113a5 499 save_or_free_index_entry(istate, ce);
e636a7b4 500 istate->cache_changed |= CE_ENTRY_REMOVED;
4aab5b46
JH
501 istate->cache_nr--;
502 if (pos >= istate->cache_nr)
7b937ca3 503 return 0;
4aab5b46
JH
504 memmove(istate->cache + pos,
505 istate->cache + pos + 1,
506 (istate->cache_nr - pos) * sizeof(struct cache_entry *));
7b937ca3
LT
507 return 1;
508}
509
36419c8e 510/*
98e023de 511 * Remove all cache entries marked for removal, that is where
36419c8e
KB
512 * CE_REMOVE is set in ce_flags. This is much more effective than
513 * calling remove_index_entry_at() for each entry to be removed.
514 */
515void remove_marked_cache_entries(struct index_state *istate)
516{
517 struct cache_entry **ce_array = istate->cache;
518 unsigned int i, j;
519
520 for (i = j = 0; i < istate->cache_nr; i++) {
5699d17e 521 if (ce_array[i]->ce_flags & CE_REMOVE) {
2092678c 522 remove_name_hash(istate, ce_array[i]);
045113a5 523 save_or_free_index_entry(istate, ce_array[i]);
5699d17e 524 }
36419c8e
KB
525 else
526 ce_array[j++] = ce_array[i];
527 }
ad837d9e
NTND
528 if (j == istate->cache_nr)
529 return;
e636a7b4 530 istate->cache_changed |= CE_ENTRY_REMOVED;
36419c8e
KB
531 istate->cache_nr = j;
532}
533
4aab5b46 534int remove_file_from_index(struct index_state *istate, const char *path)
197ee8c9 535{
4aab5b46 536 int pos = index_name_pos(istate, path, strlen(path));
c4e3cca1
JH
537 if (pos < 0)
538 pos = -pos-1;
a5400efe 539 cache_tree_invalidate_path(istate, path);
4aab5b46
JH
540 while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path))
541 remove_index_entry_at(istate, pos);
197ee8c9
LT
542 return 0;
543}
544
20314271
JS
545static int compare_name(struct cache_entry *ce, const char *path, int namelen)
546{
547 return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen);
548}
549
550static int index_name_pos_also_unmerged(struct index_state *istate,
551 const char *path, int namelen)
552{
553 int pos = index_name_pos(istate, path, namelen);
554 struct cache_entry *ce;
555
556 if (pos >= 0)
557 return pos;
558
559 /* maybe unmerged? */
560 pos = -1 - pos;
561 if (pos >= istate->cache_nr ||
562 compare_name((ce = istate->cache[pos]), path, namelen))
563 return -1;
564
565 /* order of preference: stage 2, 1, 3 */
566 if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr &&
567 ce_stage((ce = istate->cache[pos + 1])) == 2 &&
568 !compare_name(ce, path, namelen))
569 pos++;
570 return pos;
571}
572
1102952b
LT
573static int different_name(struct cache_entry *ce, struct cache_entry *alias)
574{
575 int len = ce_namelen(ce);
576 return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len);
577}
578
579/*
580 * If we add a filename that aliases in the cache, we will use the
581 * name that we already have - but we don't want to update the same
582 * alias twice, because that implies that there were actually two
583 * different files with aliasing names!
584 *
585 * So we use the CE_ADDED flag to verify that the alias was an old
586 * one before we accept it as
587 */
045113a5
NTND
588static struct cache_entry *create_alias_ce(struct index_state *istate,
589 struct cache_entry *ce,
590 struct cache_entry *alias)
1102952b
LT
591{
592 int len;
593 struct cache_entry *new;
594
595 if (alias->ce_flags & CE_ADDED)
596 die("Will not add file alias '%s' ('%s' already exists in index)", ce->name, alias->name);
597
598 /* Ok, create the new entry using the name of the existing alias */
599 len = ce_namelen(alias);
600 new = xcalloc(1, cache_entry_size(len));
601 memcpy(new->name, alias->name, len);
602 copy_cache_entry(new, ce);
045113a5 603 save_or_free_index_entry(istate, ce);
1102952b
LT
604 return new;
605}
606
b4b313f9 607void set_object_name_for_intent_to_add_entry(struct cache_entry *ce)
39425819
JH
608{
609 unsigned char sha1[20];
610 if (write_sha1_file("", 0, blob_type, sha1))
611 die("cannot create an empty blob in the object database");
612 hashcpy(ce->sha1, sha1);
613}
614
38ed1d89 615int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags)
11be42a4 616{
38ed1d89 617 int size, namelen, was_same;
d177cab0 618 mode_t st_mode = st->st_mode;
6835550d 619 struct cache_entry *ce, *alias;
56cac48c 620 unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY;
38ed1d89
JH
621 int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND);
622 int pretend = flags & ADD_CACHE_PRETEND;
39425819
JH
623 int intent_only = flags & ADD_CACHE_INTENT;
624 int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE|
625 (intent_only ? ADD_CACHE_NEW_ONLY : 0));
11be42a4 626
d177cab0 627 if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode))
960b8ad1 628 return error("%s: can only add regular files, symbolic links or git-directories", path);
11be42a4
JS
629
630 namelen = strlen(path);
d177cab0 631 if (S_ISDIR(st_mode)) {
09595258
LT
632 while (namelen && path[namelen-1] == '/')
633 namelen--;
634 }
11be42a4
JS
635 size = cache_entry_size(namelen);
636 ce = xcalloc(1, size);
637 memcpy(ce->name, path, namelen);
b60e188c 638 ce->ce_namelen = namelen;
39425819
JH
639 if (!intent_only)
640 fill_stat_cache_info(ce, st);
388b2acd
JH
641 else
642 ce->ce_flags |= CE_INTENT_TO_ADD;
11be42a4 643
78a8d641 644 if (trust_executable_bit && has_symlinks)
d177cab0 645 ce->ce_mode = create_ce_mode(st_mode);
185c975f 646 else {
78a8d641
JS
647 /* If there is an existing entry, pick the mode bits and type
648 * from it, otherwise assume unexecutable regular file.
11be42a4 649 */
185c975f 650 struct cache_entry *ent;
20314271 651 int pos = index_name_pos_also_unmerged(istate, path, namelen);
185c975f 652
4aab5b46 653 ent = (0 <= pos) ? istate->cache[pos] : NULL;
d177cab0 654 ce->ce_mode = ce_mode_from_stat(ent, st_mode);
11be42a4
JS
655 }
656
dc1ae704
JJ
657 /* When core.ignorecase=true, determine if a directory of the same name but differing
658 * case already exists within the Git repository. If it does, ensure the directory
659 * case of the file being added to the repository matches (is folded into) the existing
660 * entry's directory case.
661 */
662 if (ignore_case) {
663 const char *startPtr = ce->name;
664 const char *ptr = startPtr;
665 while (*ptr) {
666 while (*ptr && *ptr != '/')
667 ++ptr;
668 if (*ptr == '/') {
669 struct cache_entry *foundce;
670 ++ptr;
d28eec26 671 foundce = index_dir_exists(istate, ce->name, ptr - ce->name - 1);
dc1ae704
JJ
672 if (foundce) {
673 memcpy((void *)startPtr, foundce->name + (startPtr - ce->name), ptr - startPtr);
674 startPtr = ptr;
675 }
676 }
677 }
678 }
679
ebbd7439 680 alias = index_file_exists(istate, ce->name, ce_namelen(ce), ignore_case);
d177cab0 681 if (alias && !ce_stage(alias) && !ie_match_stat(istate, alias, st, ce_option)) {
0781b8a9
JH
682 /* Nothing changed, really */
683 free(ce);
125fd984
JH
684 if (!S_ISGITLINK(alias->ce_mode))
685 ce_mark_uptodate(alias);
1102952b 686 alias->ce_flags |= CE_ADDED;
0781b8a9
JH
687 return 0;
688 }
39425819 689 if (!intent_only) {
c4ce46fc 690 if (index_path(ce->sha1, path, st, HASH_WRITE_OBJECT))
39425819
JH
691 return error("unable to index file %s", path);
692 } else
b4b313f9 693 set_object_name_for_intent_to_add_entry(ce);
39425819 694
1102952b 695 if (ignore_case && alias && different_name(ce, alias))
045113a5 696 ce = create_alias_ce(istate, ce, alias);
1102952b 697 ce->ce_flags |= CE_ADDED;
38ed1d89 698
3bf0dd1f 699 /* It was suspected to be racily clean, but it turns out to be Ok */
38ed1d89
JH
700 was_same = (alias &&
701 !ce_stage(alias) &&
702 !hashcmp(alias->sha1, ce->sha1) &&
703 ce->ce_mode == alias->ce_mode);
704
705 if (pretend)
706 ;
39425819 707 else if (add_index_entry(istate, ce, add_option))
960b8ad1 708 return error("unable to add %s to index",path);
38ed1d89 709 if (verbose && !was_same)
11be42a4 710 printf("add '%s'\n", path);
11be42a4
JS
711 return 0;
712}
713
38ed1d89 714int add_file_to_index(struct index_state *istate, const char *path, int flags)
d177cab0
LT
715{
716 struct stat st;
717 if (lstat(path, &st))
d824cbba 718 die_errno("unable to stat '%s'", path);
38ed1d89 719 return add_to_index(istate, path, &st, flags);
d177cab0
LT
720}
721
6640f881
CR
722struct cache_entry *make_cache_entry(unsigned int mode,
723 const unsigned char *sha1, const char *path, int stage,
25762726 724 unsigned int refresh_options)
6640f881
CR
725{
726 int size, len;
727 struct cache_entry *ce;
728
7e7abea9
DP
729 if (!verify_path(path)) {
730 error("Invalid path '%s'", path);
6640f881 731 return NULL;
7e7abea9 732 }
6640f881
CR
733
734 len = strlen(path);
735 size = cache_entry_size(len);
736 ce = xcalloc(1, size);
737
738 hashcpy(ce->sha1, sha1);
739 memcpy(ce->name, path, len);
b60e188c
TG
740 ce->ce_flags = create_ce_flags(stage);
741 ce->ce_namelen = len;
6640f881
CR
742 ce->ce_mode = create_ce_mode(mode);
743
25762726 744 return refresh_cache_entry(ce, refresh_options);
6640f881
CR
745}
746
9c5e6c80 747int ce_same_name(const struct cache_entry *a, const struct cache_entry *b)
7b937ca3
LT
748{
749 int len = ce_namelen(a);
750 return ce_namelen(b) == len && !memcmp(a->name, b->name, len);
751}
752
8dcf39c4
LT
753/*
754 * We fundamentally don't like some paths: we don't want
755 * dot or dot-dot anywhere, and for obvious reasons don't
756 * want to recurse into ".git" either.
757 *
758 * Also, we don't want double slashes or slashes at the
759 * end that can make pathnames ambiguous.
760 */
761static int verify_dotfile(const char *rest)
762{
763 /*
764 * The first character was '.', but that
765 * has already been discarded, we now test
766 * the rest.
767 */
e0f530ff 768
8dcf39c4 769 /* "." is not allowed */
e0f530ff 770 if (*rest == '\0' || is_dir_sep(*rest))
8dcf39c4
LT
771 return 0;
772
e0f530ff 773 switch (*rest) {
8dcf39c4
LT
774 /*
775 * ".git" followed by NUL or slash is bad. This
776 * shares the path end test with the ".." case.
777 */
778 case 'g':
779 if (rest[1] != 'i')
780 break;
781 if (rest[2] != 't')
782 break;
783 rest += 2;
784 /* fallthrough */
785 case '.':
e0f530ff 786 if (rest[1] == '\0' || is_dir_sep(rest[1]))
8dcf39c4
LT
787 return 0;
788 }
789 return 1;
790}
791
792int verify_path(const char *path)
793{
794 char c;
795
56948cb6
EFL
796 if (has_dos_drive_prefix(path))
797 return 0;
798
8dcf39c4
LT
799 goto inside;
800 for (;;) {
801 if (!c)
802 return 1;
56948cb6 803 if (is_dir_sep(c)) {
8dcf39c4
LT
804inside:
805 c = *path++;
3bdf09c7
JH
806 if ((c == '.' && !verify_dotfile(path)) ||
807 is_dir_sep(c) || c == '\0')
808 return 0;
8dcf39c4
LT
809 }
810 c = *path++;
811 }
812}
813
12676608
LT
814/*
815 * Do we have another file that has the beginning components being a
816 * proper superset of the name we're trying to add?
0f1e4f04 817 */
4aab5b46
JH
818static int has_file_name(struct index_state *istate,
819 const struct cache_entry *ce, int pos, int ok_to_replace)
0f1e4f04 820{
12676608
LT
821 int retval = 0;
822 int len = ce_namelen(ce);
b155725d 823 int stage = ce_stage(ce);
12676608 824 const char *name = ce->name;
0f1e4f04 825
4aab5b46
JH
826 while (pos < istate->cache_nr) {
827 struct cache_entry *p = istate->cache[pos++];
0f1e4f04 828
12676608 829 if (len >= ce_namelen(p))
0f1e4f04 830 break;
12676608
LT
831 if (memcmp(name, p->name, len))
832 break;
b155725d
JH
833 if (ce_stage(p) != stage)
834 continue;
12676608
LT
835 if (p->name[len] != '/')
836 continue;
7a51ed66 837 if (p->ce_flags & CE_REMOVE)
21cd8d00 838 continue;
12676608
LT
839 retval = -1;
840 if (!ok_to_replace)
841 break;
4aab5b46 842 remove_index_entry_at(istate, --pos);
0f1e4f04 843 }
12676608
LT
844 return retval;
845}
0f1e4f04 846
12676608
LT
847/*
848 * Do we have another file with a pathname that is a proper
849 * subset of the name we're trying to add?
850 */
4aab5b46
JH
851static int has_dir_name(struct index_state *istate,
852 const struct cache_entry *ce, int pos, int ok_to_replace)
12676608
LT
853{
854 int retval = 0;
b155725d 855 int stage = ce_stage(ce);
12676608
LT
856 const char *name = ce->name;
857 const char *slash = name + ce_namelen(ce);
0f1e4f04 858
12676608
LT
859 for (;;) {
860 int len;
0f1e4f04 861
12676608
LT
862 for (;;) {
863 if (*--slash == '/')
864 break;
865 if (slash <= ce->name)
866 return retval;
867 }
868 len = slash - name;
0f1e4f04 869
b60e188c 870 pos = index_name_stage_pos(istate, name, len, stage);
12676608 871 if (pos >= 0) {
21cd8d00
JH
872 /*
873 * Found one, but not so fast. This could
874 * be a marker that says "I was here, but
875 * I am being removed". Such an entry is
876 * not a part of the resulting tree, and
877 * it is Ok to have a directory at the same
878 * path.
879 */
077c48df 880 if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) {
21cd8d00
JH
881 retval = -1;
882 if (!ok_to_replace)
883 break;
4aab5b46 884 remove_index_entry_at(istate, pos);
21cd8d00
JH
885 continue;
886 }
12676608 887 }
21cd8d00
JH
888 else
889 pos = -pos-1;
12676608
LT
890
891 /*
892 * Trivial optimization: if we find an entry that
893 * already matches the sub-directory, then we know
b155725d 894 * we're ok, and we can exit.
12676608 895 */
4aab5b46
JH
896 while (pos < istate->cache_nr) {
897 struct cache_entry *p = istate->cache[pos];
b155725d
JH
898 if ((ce_namelen(p) <= len) ||
899 (p->name[len] != '/') ||
900 memcmp(p->name, name, len))
901 break; /* not our subdirectory */
077c48df
JH
902 if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE))
903 /*
904 * p is at the same stage as our entry, and
b155725d
JH
905 * is a subdirectory of what we are looking
906 * at, so we cannot have conflicts at our
907 * level or anything shorter.
908 */
909 return retval;
910 pos++;
192268c1 911 }
0f1e4f04 912 }
12676608
LT
913 return retval;
914}
915
916/* We may be in a situation where we already have path/file and path
917 * is being added, or we already have path and path/file is being
918 * added. Either one would result in a nonsense tree that has path
919 * twice when git-write-tree tries to write it out. Prevent it.
a6080a0a 920 *
12676608
LT
921 * If ok-to-replace is specified, we remove the conflicting entries
922 * from the cache so the caller should recompute the insert position.
923 * When this happens, we return non-zero.
924 */
4aab5b46
JH
925static int check_file_directory_conflict(struct index_state *istate,
926 const struct cache_entry *ce,
927 int pos, int ok_to_replace)
12676608 928{
21cd8d00
JH
929 int retval;
930
931 /*
932 * When ce is an "I am going away" entry, we allow it to be added
933 */
7a51ed66 934 if (ce->ce_flags & CE_REMOVE)
21cd8d00
JH
935 return 0;
936
12676608
LT
937 /*
938 * We check if the path is a sub-path of a subsequent pathname
939 * first, since removing those will not change the position
21cd8d00 940 * in the array.
12676608 941 */
4aab5b46 942 retval = has_file_name(istate, ce, pos, ok_to_replace);
21cd8d00 943
12676608
LT
944 /*
945 * Then check if the path might have a clashing sub-directory
946 * before it.
947 */
4aab5b46 948 return retval + has_dir_name(istate, ce, pos, ok_to_replace);
0f1e4f04
JH
949}
950
af3785dc 951static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option)
197ee8c9
LT
952{
953 int pos;
192268c1
JH
954 int ok_to_add = option & ADD_CACHE_OK_TO_ADD;
955 int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE;
b155725d 956 int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK;
39425819 957 int new_only = option & ADD_CACHE_NEW_ONLY;
5f73076c 958
ce7c614b
NTND
959 if (!(option & ADD_CACHE_KEEP_CACHE_TREE))
960 cache_tree_invalidate_path(istate, ce->name);
b60e188c 961 pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce));
197ee8c9 962
3e09cdfd 963 /* existing match? Just replace it. */
76e7f4ec 964 if (pos >= 0) {
39425819
JH
965 if (!new_only)
966 replace_index_entry(istate, pos, ce);
197ee8c9
LT
967 return 0;
968 }
76e7f4ec 969 pos = -pos-1;
197ee8c9 970
7b937ca3
LT
971 /*
972 * Inserting a merged entry ("stage 0") into the index
973 * will always replace all non-merged entries..
974 */
4aab5b46
JH
975 if (pos < istate->cache_nr && ce_stage(ce) == 0) {
976 while (ce_same_name(istate->cache[pos], ce)) {
7b937ca3 977 ok_to_add = 1;
4aab5b46 978 if (!remove_index_entry_at(istate, pos))
7b937ca3
LT
979 break;
980 }
981 }
982
121481ab
LT
983 if (!ok_to_add)
984 return -1;
8dcf39c4 985 if (!verify_path(ce->name))
7e7abea9 986 return error("Invalid path '%s'", ce->name);
121481ab 987
3e09cdfd 988 if (!skip_df_check &&
4aab5b46 989 check_file_directory_conflict(istate, ce, pos, ok_to_replace)) {
192268c1 990 if (!ok_to_replace)
4aab5b46
JH
991 return error("'%s' appears as both a file and as a directory",
992 ce->name);
b60e188c 993 pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce));
192268c1
JH
994 pos = -pos-1;
995 }
af3785dc
JH
996 return pos + 1;
997}
998
999int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option)
1000{
1001 int pos;
1002
1003 if (option & ADD_CACHE_JUST_APPEND)
1004 pos = istate->cache_nr;
1005 else {
1006 int ret;
1007 ret = add_index_entry_with_check(istate, ce, option);
1008 if (ret <= 0)
1009 return ret;
1010 pos = ret - 1;
1011 }
0f1e4f04 1012
197ee8c9 1013 /* Make sure the array is big enough .. */
999f5660 1014 ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc);
197ee8c9
LT
1015
1016 /* Add it in.. */
4aab5b46 1017 istate->cache_nr++;
af3785dc 1018 if (istate->cache_nr > pos + 1)
4aab5b46
JH
1019 memmove(istate->cache + pos + 1,
1020 istate->cache + pos,
1021 (istate->cache_nr - pos - 1) * sizeof(ce));
cf558704 1022 set_index_entry(istate, pos, ce);
e636a7b4 1023 istate->cache_changed |= CE_ENTRY_ADDED;
197ee8c9
LT
1024 return 0;
1025}
1026
405e5b2f
LT
1027/*
1028 * "refresh" does not calculate a new sha1 file or bring the
1029 * cache up-to-date for mode/content changes. But what it
1030 * _does_ do is to "re-match" the stat information of a file
1031 * with the cache, so that you can refresh the cache for a
1032 * file that hasn't been changed but where the stat entry is
1033 * out of date.
1034 *
1035 * For example, you'd want to do this after doing a "git-read-tree",
1036 * to link up the stat cache details with the proper files.
1037 */
4aab5b46 1038static struct cache_entry *refresh_cache_ent(struct index_state *istate,
4bd5b7da 1039 struct cache_entry *ce,
d05e6970
JK
1040 unsigned int options, int *err,
1041 int *changed_ret)
405e5b2f
LT
1042{
1043 struct stat st;
1044 struct cache_entry *updated;
1045 int changed, size;
25762726 1046 int refresh = options & CE_MATCH_REFRESH;
4bd5b7da 1047 int ignore_valid = options & CE_MATCH_IGNORE_VALID;
56cac48c 1048 int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE;
2e2e7ec1 1049 int ignore_missing = options & CE_MATCH_IGNORE_MISSING;
405e5b2f 1050
25762726 1051 if (!refresh || ce_uptodate(ce))
eadb5831
JH
1052 return ce;
1053
aa9349d4 1054 /*
56cac48c
NTND
1055 * CE_VALID or CE_SKIP_WORKTREE means the user promised us
1056 * that the change to the work tree does not matter and told
1057 * us not to worry.
aa9349d4 1058 */
56cac48c
NTND
1059 if (!ignore_skip_worktree && ce_skip_worktree(ce)) {
1060 ce_mark_uptodate(ce);
1061 return ce;
1062 }
aa9349d4
MSO
1063 if (!ignore_valid && (ce->ce_flags & CE_VALID)) {
1064 ce_mark_uptodate(ce);
1065 return ce;
1066 }
1067
ccad42d4
RS
1068 if (has_symlink_leading_path(ce->name, ce_namelen(ce))) {
1069 if (ignore_missing)
1070 return ce;
1071 if (err)
1072 *err = ENOENT;
1073 return NULL;
1074 }
1075
8fd2cb40 1076 if (lstat(ce->name, &st) < 0) {
2e2e7ec1
BK
1077 if (ignore_missing && errno == ENOENT)
1078 return ce;
ec0cc704
JH
1079 if (err)
1080 *err = errno;
8fd2cb40
JS
1081 return NULL;
1082 }
405e5b2f 1083
4bd5b7da 1084 changed = ie_match_stat(istate, ce, &st, options);
d05e6970
JK
1085 if (changed_ret)
1086 *changed_ret = changed;
405e5b2f 1087 if (!changed) {
4bd5b7da
JH
1088 /*
1089 * The path is unchanged. If we were told to ignore
1090 * valid bit, then we did the actual stat check and
1091 * found that the entry is unmodified. If the entry
1092 * is not marked VALID, this is the place to mark it
1093 * valid again, under "assume unchanged" mode.
1094 */
1095 if (ignore_valid && assume_unchanged &&
7a51ed66 1096 !(ce->ce_flags & CE_VALID))
405e5b2f 1097 ; /* mark this one VALID again */
eadb5831
JH
1098 else {
1099 /*
1100 * We do not mark the index itself "modified"
1101 * because CE_UPTODATE flag is in-core only;
1102 * we are not going to write this change out.
1103 */
125fd984
JH
1104 if (!S_ISGITLINK(ce->ce_mode))
1105 ce_mark_uptodate(ce);
8fd2cb40 1106 return ce;
eadb5831 1107 }
405e5b2f
LT
1108 }
1109
4bd5b7da 1110 if (ie_modified(istate, ce, &st, options)) {
ec0cc704
JH
1111 if (err)
1112 *err = EINVAL;
8fd2cb40
JS
1113 return NULL;
1114 }
405e5b2f
LT
1115
1116 size = ce_size(ce);
1117 updated = xmalloc(size);
1118 memcpy(updated, ce, size);
1119 fill_stat_cache_info(updated, &st);
4bd5b7da
JH
1120 /*
1121 * If ignore_valid is not set, we should leave CE_VALID bit
1122 * alone. Otherwise, paths marked with --no-assume-unchanged
1123 * (i.e. things to be edited) will reacquire CE_VALID bit
1124 * automatically, which is not really what we want.
405e5b2f 1125 */
4bd5b7da 1126 if (!ignore_valid && assume_unchanged &&
7a51ed66
LT
1127 !(ce->ce_flags & CE_VALID))
1128 updated->ce_flags &= ~CE_VALID;
405e5b2f 1129
e636a7b4 1130 /* istate->cache_changed is updated in the caller */
405e5b2f
LT
1131 return updated;
1132}
1133
3deffc52 1134static void show_file(const char * fmt, const char * name, int in_porcelain,
046613c5 1135 int * first, const char *header_msg)
3deffc52
MM
1136{
1137 if (in_porcelain && *first && header_msg) {
1138 printf("%s\n", header_msg);
cd2b8ae9 1139 *first = 0;
3deffc52
MM
1140 }
1141 printf(fmt, name);
1142}
1143
9b2d6149
NTND
1144int refresh_index(struct index_state *istate, unsigned int flags,
1145 const struct pathspec *pathspec,
046613c5 1146 char *seen, const char *header_msg)
405e5b2f
LT
1147{
1148 int i;
1149 int has_errors = 0;
1150 int really = (flags & REFRESH_REALLY) != 0;
1151 int allow_unmerged = (flags & REFRESH_UNMERGED) != 0;
1152 int quiet = (flags & REFRESH_QUIET) != 0;
1153 int not_new = (flags & REFRESH_IGNORE_MISSING) != 0;
5fdeacb0 1154 int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0;
3deffc52
MM
1155 int first = 1;
1156 int in_porcelain = (flags & REFRESH_IN_PORCELAIN);
25762726
BK
1157 unsigned int options = (CE_MATCH_REFRESH |
1158 (really ? CE_MATCH_IGNORE_VALID : 0) |
2e2e7ec1 1159 (not_new ? CE_MATCH_IGNORE_MISSING : 0));
4bd4e730 1160 const char *modified_fmt;
73b7eae6
JK
1161 const char *deleted_fmt;
1162 const char *typechange_fmt;
1163 const char *added_fmt;
4bd4e730 1164 const char *unmerged_fmt;
405e5b2f 1165
4bd4e730 1166 modified_fmt = (in_porcelain ? "M\t%s\n" : "%s: needs update\n");
73b7eae6
JK
1167 deleted_fmt = (in_porcelain ? "D\t%s\n" : "%s: needs update\n");
1168 typechange_fmt = (in_porcelain ? "T\t%s\n" : "%s needs update\n");
1169 added_fmt = (in_porcelain ? "A\t%s\n" : "%s needs update\n");
4bd4e730 1170 unmerged_fmt = (in_porcelain ? "U\t%s\n" : "%s: needs merge\n");
4aab5b46 1171 for (i = 0; i < istate->cache_nr; i++) {
405e5b2f 1172 struct cache_entry *ce, *new;
ec0cc704 1173 int cache_errno = 0;
73b7eae6 1174 int changed = 0;
3d1f148c 1175 int filtered = 0;
ec0cc704 1176
4aab5b46 1177 ce = istate->cache[i];
5fdeacb0
JS
1178 if (ignore_submodules && S_ISGITLINK(ce->ce_mode))
1179 continue;
1180
429bb40a 1181 if (pathspec && !ce_path_match(ce, pathspec, seen))
3d1f148c
JH
1182 filtered = 1;
1183
405e5b2f 1184 if (ce_stage(ce)) {
4aab5b46
JH
1185 while ((i < istate->cache_nr) &&
1186 ! strcmp(istate->cache[i]->name, ce->name))
405e5b2f
LT
1187 i++;
1188 i--;
1189 if (allow_unmerged)
1190 continue;
3d1f148c
JH
1191 if (!filtered)
1192 show_file(unmerged_fmt, ce->name, in_porcelain,
1193 &first, header_msg);
405e5b2f
LT
1194 has_errors = 1;
1195 continue;
1196 }
1197
3d1f148c 1198 if (filtered)
d616813d
AJ
1199 continue;
1200
73b7eae6 1201 new = refresh_cache_ent(istate, ce, options, &cache_errno, &changed);
8fd2cb40 1202 if (new == ce)
405e5b2f 1203 continue;
8fd2cb40 1204 if (!new) {
73b7eae6
JK
1205 const char *fmt;
1206
8fd2cb40 1207 if (really && cache_errno == EINVAL) {
405e5b2f
LT
1208 /* If we are doing --really-refresh that
1209 * means the index is not valid anymore.
1210 */
7a51ed66 1211 ce->ce_flags &= ~CE_VALID;
078a58e8 1212 ce->ce_flags |= CE_UPDATE_IN_BASE;
e636a7b4 1213 istate->cache_changed |= CE_ENTRY_CHANGED;
405e5b2f
LT
1214 }
1215 if (quiet)
1216 continue;
73b7eae6
JK
1217
1218 if (cache_errno == ENOENT)
1219 fmt = deleted_fmt;
1220 else if (ce->ce_flags & CE_INTENT_TO_ADD)
1221 fmt = added_fmt; /* must be before other checks */
1222 else if (changed & TYPE_CHANGED)
1223 fmt = typechange_fmt;
1224 else
1225 fmt = modified_fmt;
1226 show_file(fmt,
1227 ce->name, in_porcelain, &first, header_msg);
405e5b2f
LT
1228 has_errors = 1;
1229 continue;
1230 }
cf558704
LT
1231
1232 replace_index_entry(istate, i, new);
405e5b2f
LT
1233 }
1234 return has_errors;
1235}
1236
25762726
BK
1237static struct cache_entry *refresh_cache_entry(struct cache_entry *ce,
1238 unsigned int options)
ec0cc704 1239{
25762726 1240 return refresh_cache_ent(&the_index, ce, options, NULL, NULL);
ec0cc704
JH
1241}
1242
db3b313c
JH
1243
1244/*****************************************************************
1245 * Index File I/O
1246 *****************************************************************/
1247
9d227781
JH
1248#define INDEX_FORMAT_DEFAULT 3
1249
136347d7
TG
1250static unsigned int get_index_format_default(void)
1251{
1252 char *envversion = getenv("GIT_INDEX_VERSION");
3c09d684 1253 char *endp;
b27a5720 1254 int value;
3c09d684
TG
1255 unsigned int version = INDEX_FORMAT_DEFAULT;
1256
136347d7 1257 if (!envversion) {
b27a5720
TA
1258 if (!git_config_get_int("index.version", &value))
1259 version = value;
3c09d684
TG
1260 if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) {
1261 warning(_("index.version set, but the value is invalid.\n"
136347d7 1262 "Using version %i"), INDEX_FORMAT_DEFAULT);
3c09d684 1263 return INDEX_FORMAT_DEFAULT;
136347d7
TG
1264 }
1265 return version;
1266 }
3c09d684
TG
1267
1268 version = strtoul(envversion, &endp, 10);
1269 if (*endp ||
1270 version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) {
1271 warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n"
1272 "Using version %i"), INDEX_FORMAT_DEFAULT);
1273 version = INDEX_FORMAT_DEFAULT;
1274 }
1275 return version;
136347d7
TG
1276}
1277
db3b313c
JH
1278/*
1279 * dev/ino/uid/gid/size are also just tracked to the low 32 bits
1280 * Again - this is just a (very strong in practice) heuristic that
1281 * the inode hasn't changed.
1282 *
1283 * We save the fields in big-endian order to allow using the
1284 * index file over NFS transparently.
1285 */
1286struct ondisk_cache_entry {
1287 struct cache_time ctime;
1288 struct cache_time mtime;
7800c1eb
TG
1289 uint32_t dev;
1290 uint32_t ino;
1291 uint32_t mode;
1292 uint32_t uid;
1293 uint32_t gid;
1294 uint32_t size;
db3b313c 1295 unsigned char sha1[20];
7800c1eb 1296 uint16_t flags;
db3b313c
JH
1297 char name[FLEX_ARRAY]; /* more */
1298};
1299
1300/*
1301 * This struct is used when CE_EXTENDED bit is 1
1302 * The struct must match ondisk_cache_entry exactly from
1303 * ctime till flags
1304 */
1305struct ondisk_cache_entry_extended {
1306 struct cache_time ctime;
1307 struct cache_time mtime;
7800c1eb
TG
1308 uint32_t dev;
1309 uint32_t ino;
1310 uint32_t mode;
1311 uint32_t uid;
1312 uint32_t gid;
1313 uint32_t size;
db3b313c 1314 unsigned char sha1[20];
7800c1eb
TG
1315 uint16_t flags;
1316 uint16_t flags2;
db3b313c
JH
1317 char name[FLEX_ARRAY]; /* more */
1318};
1319
6c9cd161 1320/* These are only used for v3 or lower */
db3b313c
JH
1321#define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,name) + (len) + 8) & ~7)
1322#define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len)
1323#define ondisk_cache_entry_extended_size(len) align_flex_name(ondisk_cache_entry_extended,len)
1324#define ondisk_ce_size(ce) (((ce)->ce_flags & CE_EXTENDED) ? \
1325 ondisk_cache_entry_extended_size(ce_namelen(ce)) : \
1326 ondisk_cache_entry_size(ce_namelen(ce)))
1327
bad68ec9 1328static int verify_hdr(struct cache_header *hdr, unsigned long size)
e83c5163 1329{
9126f009 1330 git_SHA_CTX c;
bad68ec9 1331 unsigned char sha1[20];
0136bac9 1332 int hdr_version;
e83c5163 1333
ccc4feb5 1334 if (hdr->hdr_signature != htonl(CACHE_SIGNATURE))
e83c5163 1335 return error("bad signature");
0136bac9 1336 hdr_version = ntohl(hdr->hdr_version);
b82a7b5b 1337 if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version)
0136bac9 1338 return error("bad index version %d", hdr_version);
9126f009
NP
1339 git_SHA1_Init(&c);
1340 git_SHA1_Update(&c, hdr, size - 20);
1341 git_SHA1_Final(sha1, &c);
a89fccd2 1342 if (hashcmp(sha1, (unsigned char *)hdr + size - 20))
ca9be054 1343 return error("bad index file sha1 signature");
e83c5163
LT
1344 return 0;
1345}
1346
4aab5b46
JH
1347static int read_index_extension(struct index_state *istate,
1348 const char *ext, void *data, unsigned long sz)
bad68ec9
JH
1349{
1350 switch (CACHE_EXT(ext)) {
1351 case CACHE_EXT_TREE:
4aab5b46 1352 istate->cache_tree = cache_tree_read(data, sz);
bad68ec9 1353 break;
cfc5789a
JH
1354 case CACHE_EXT_RESOLVE_UNDO:
1355 istate->resolve_undo = resolve_undo_read(data, sz);
1356 break;
5fc2fc8f
NTND
1357 case CACHE_EXT_LINK:
1358 if (read_link_extension(istate, data, sz))
1359 return -1;
1360 break;
bad68ec9
JH
1361 default:
1362 if (*ext < 'A' || 'Z' < *ext)
1363 return error("index uses %.4s extension, which we do not understand",
1364 ext);
1365 fprintf(stderr, "ignoring %.4s extension\n", ext);
1366 break;
1367 }
1368 return 0;
1369}
1370
216aab1e
MH
1371int hold_locked_index(struct lock_file *lk, int die_on_error)
1372{
1373 return hold_lock_file_for_update(lk, get_index_file(),
1374 die_on_error
1375 ? LOCK_DIE_ON_ERROR
1376 : 0);
1377}
1378
4aab5b46 1379int read_index(struct index_state *istate)
8fd2cb40 1380{
4aab5b46 1381 return read_index_from(istate, get_index_file());
8fd2cb40
JS
1382}
1383
3fc22b53
JH
1384static struct cache_entry *cache_entry_from_ondisk(struct ondisk_cache_entry *ondisk,
1385 unsigned int flags,
1386 const char *name,
1387 size_t len)
1388{
1389 struct cache_entry *ce = xmalloc(cache_entry_size(len));
1390
c3d8da57
JK
1391 ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec);
1392 ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec);
1393 ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec);
1394 ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec);
1395 ce->ce_stat_data.sd_dev = get_be32(&ondisk->dev);
1396 ce->ce_stat_data.sd_ino = get_be32(&ondisk->ino);
1397 ce->ce_mode = get_be32(&ondisk->mode);
1398 ce->ce_stat_data.sd_uid = get_be32(&ondisk->uid);
1399 ce->ce_stat_data.sd_gid = get_be32(&ondisk->gid);
1400 ce->ce_stat_data.sd_size = get_be32(&ondisk->size);
b60e188c
TG
1401 ce->ce_flags = flags & ~CE_NAMEMASK;
1402 ce->ce_namelen = len;
5fc2fc8f 1403 ce->index = 0;
3fc22b53
JH
1404 hashcpy(ce->sha1, ondisk->sha1);
1405 memcpy(ce->name, name, len);
1406 ce->name[len] = '\0';
1407 return ce;
1408}
1409
6c9cd161
JH
1410/*
1411 * Adjacent cache entries tend to share the leading paths, so it makes
1412 * sense to only store the differences in later entries. In the v4
1413 * on-disk format of the index, each on-disk cache entry stores the
1414 * number of bytes to be stripped from the end of the previous name,
1415 * and the bytes to append to the result, to come up with its name.
1416 */
1417static unsigned long expand_name_field(struct strbuf *name, const char *cp_)
1418{
1419 const unsigned char *ep, *cp = (const unsigned char *)cp_;
1420 size_t len = decode_varint(&cp);
1421
1422 if (name->len < len)
1423 die("malformed name field in the index");
1424 strbuf_remove(name, name->len - len, len);
1425 for (ep = cp; *ep; ep++)
1426 ; /* find the end */
1427 strbuf_add(name, cp, ep - cp);
1428 return (const char *)ep + 1 - cp_;
1429}
1430
936f53d0 1431static struct cache_entry *create_from_disk(struct ondisk_cache_entry *ondisk,
6c9cd161
JH
1432 unsigned long *ent_size,
1433 struct strbuf *previous_name)
7a51ed66 1434{
debed2a6 1435 struct cache_entry *ce;
7fec10b7 1436 size_t len;
06aaaa0b 1437 const char *name;
debed2a6 1438 unsigned int flags;
7fec10b7 1439
7a51ed66 1440 /* On-disk flags are just 16 bits */
c3d8da57 1441 flags = get_be16(&ondisk->flags);
debed2a6 1442 len = flags & CE_NAMEMASK;
7fec10b7 1443
debed2a6 1444 if (flags & CE_EXTENDED) {
06aaaa0b
NTND
1445 struct ondisk_cache_entry_extended *ondisk2;
1446 int extended_flags;
1447 ondisk2 = (struct ondisk_cache_entry_extended *)ondisk;
c3d8da57 1448 extended_flags = get_be16(&ondisk2->flags2) << 16;
06aaaa0b
NTND
1449 /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */
1450 if (extended_flags & ~CE_EXTENDED_FLAGS)
1451 die("Unknown index entry format %08x", extended_flags);
debed2a6 1452 flags |= extended_flags;
06aaaa0b
NTND
1453 name = ondisk2->name;
1454 }
1455 else
1456 name = ondisk->name;
1457
6c9cd161
JH
1458 if (!previous_name) {
1459 /* v3 and earlier */
1460 if (len == CE_NAMEMASK)
1461 len = strlen(name);
1462 ce = cache_entry_from_ondisk(ondisk, flags, name, len);
1463
1464 *ent_size = ondisk_ce_size(ce);
1465 } else {
1466 unsigned long consumed;
1467 consumed = expand_name_field(previous_name, name);
1468 ce = cache_entry_from_ondisk(ondisk, flags,
1469 previous_name->buf,
1470 previous_name->len);
1471
1472 *ent_size = (name - ((char *)ondisk)) + consumed;
1473 }
debed2a6 1474 return ce;
cf558704
LT
1475}
1476
15999d0b
JSP
1477static void check_ce_order(struct cache_entry *ce, struct cache_entry *next_ce)
1478{
1479 int name_compare = strcmp(ce->name, next_ce->name);
1480 if (0 < name_compare)
1481 die("unordered stage entries in index");
1482 if (!name_compare) {
1483 if (!ce_stage(ce))
1484 die("multiple stage entries for merged file '%s'",
1485 ce->name);
1486 if (ce_stage(ce) > ce_stage(next_ce))
1487 die("unordered stage entries for '%s'",
1488 ce->name);
1489 }
1490}
1491
8fd2cb40 1492/* remember to discard_cache() before reading a different cache! */
3e52f70b 1493int do_read_index(struct index_state *istate, const char *path, int must_exist)
e83c5163
LT
1494{
1495 int fd, i;
1496 struct stat st;
debed2a6 1497 unsigned long src_offset;
e83c5163 1498 struct cache_header *hdr;
7a51ed66
LT
1499 void *mmap;
1500 size_t mmap_size;
6c9cd161 1501 struct strbuf previous_name_buf = STRBUF_INIT, *previous_name;
e83c5163 1502
913e0e99 1503 if (istate->initialized)
4aab5b46 1504 return istate->cache_nr;
5d1a5c02 1505
fba2f38a
KB
1506 istate->timestamp.sec = 0;
1507 istate->timestamp.nsec = 0;
8fd2cb40 1508 fd = open(path, O_RDONLY);
5d1a5c02 1509 if (fd < 0) {
5fc2fc8f 1510 if (!must_exist && errno == ENOENT)
5d1a5c02 1511 return 0;
5fc2fc8f 1512 die_errno("%s: index file open failed", path);
5d1a5c02 1513 }
e83c5163 1514
3511a377 1515 if (fstat(fd, &st))
d824cbba 1516 die_errno("cannot stat the open index");
3511a377 1517
7a51ed66
LT
1518 mmap_size = xsize_t(st.st_size);
1519 if (mmap_size < sizeof(struct cache_header) + 20)
3511a377
LFC
1520 die("index file smaller than expected");
1521
7a51ed66 1522 mmap = xmmap(NULL, mmap_size, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0);
7a51ed66 1523 if (mmap == MAP_FAILED)
0721c314 1524 die_errno("unable to map index file");
57d84f8d 1525 close(fd);
e83c5163 1526
7a51ed66
LT
1527 hdr = mmap;
1528 if (verify_hdr(hdr, mmap_size) < 0)
e83c5163
LT
1529 goto unmap;
1530
e93021b2 1531 hashcpy(istate->sha1, (const unsigned char *)hdr + mmap_size - 20);
9d227781 1532 istate->version = ntohl(hdr->hdr_version);
4aab5b46
JH
1533 istate->cache_nr = ntohl(hdr->hdr_entries);
1534 istate->cache_alloc = alloc_nr(istate->cache_nr);
c4aa3167 1535 istate->cache = xcalloc(istate->cache_alloc, sizeof(*istate->cache));
913e0e99 1536 istate->initialized = 1;
7a51ed66 1537
9d227781 1538 if (istate->version == 4)
6c9cd161
JH
1539 previous_name = &previous_name_buf;
1540 else
1541 previous_name = NULL;
1542
7a51ed66 1543 src_offset = sizeof(*hdr);
4aab5b46 1544 for (i = 0; i < istate->cache_nr; i++) {
7a51ed66 1545 struct ondisk_cache_entry *disk_ce;
4aab5b46 1546 struct cache_entry *ce;
936f53d0 1547 unsigned long consumed;
4aab5b46 1548
7a51ed66 1549 disk_ce = (struct ondisk_cache_entry *)((char *)mmap + src_offset);
6c9cd161 1550 ce = create_from_disk(disk_ce, &consumed, previous_name);
cf558704 1551 set_index_entry(istate, i, ce);
7a51ed66 1552
15999d0b
JSP
1553 if (i > 0)
1554 check_ce_order(istate->cache[i - 1], ce);
1555
936f53d0 1556 src_offset += consumed;
e83c5163 1557 }
6c9cd161 1558 strbuf_release(&previous_name_buf);
fba2f38a 1559 istate->timestamp.sec = st.st_mtime;
c06ff490 1560 istate->timestamp.nsec = ST_MTIME_NSEC(st);
fba2f38a 1561
7a51ed66 1562 while (src_offset <= mmap_size - 20 - 8) {
bad68ec9
JH
1563 /* After an array of active_nr index entries,
1564 * there can be arbitrary number of extended
1565 * sections, each of which is prefixed with
1566 * extension name (4-byte) and section length
1567 * in 4-byte network byte order.
1568 */
07cc8eca 1569 uint32_t extsize;
7a51ed66 1570 memcpy(&extsize, (char *)mmap + src_offset + 4, 4);
bad68ec9 1571 extsize = ntohl(extsize);
4aab5b46 1572 if (read_index_extension(istate,
7a51ed66
LT
1573 (const char *) mmap + src_offset,
1574 (char *) mmap + src_offset + 8,
1d7f171c 1575 extsize) < 0)
bad68ec9 1576 goto unmap;
7a51ed66
LT
1577 src_offset += 8;
1578 src_offset += extsize;
bad68ec9 1579 }
7a51ed66 1580 munmap(mmap, mmap_size);
4aab5b46 1581 return istate->cache_nr;
e83c5163
LT
1582
1583unmap:
7a51ed66 1584 munmap(mmap, mmap_size);
5d1a5c02 1585 die("index file corrupt");
e83c5163
LT
1586}
1587
5fc2fc8f
NTND
1588int read_index_from(struct index_state *istate, const char *path)
1589{
1590 struct split_index *split_index;
1591 int ret;
1592
1593 /* istate->initialized covers both .git/index and .git/sharedindex.xxx */
1594 if (istate->initialized)
1595 return istate->cache_nr;
1596
1597 ret = do_read_index(istate, path, 0);
1598 split_index = istate->split_index;
1599 if (!split_index)
1600 return ret;
1601
1602 if (is_null_sha1(split_index->base_sha1))
1603 return ret;
5fc2fc8f
NTND
1604
1605 if (split_index->base)
1606 discard_index(split_index->base);
1607 else
1608 split_index->base = xcalloc(1, sizeof(*split_index->base));
1609 ret = do_read_index(split_index->base,
1610 git_path("sharedindex.%s",
1611 sha1_to_hex(split_index->base_sha1)), 1);
1612 if (hashcmp(split_index->base_sha1, split_index->base->sha1))
1613 die("broken index, expect %s in %s, got %s",
1614 sha1_to_hex(split_index->base_sha1),
1615 git_path("sharedindex.%s",
1616 sha1_to_hex(split_index->base_sha1)),
1617 sha1_to_hex(split_index->base->sha1));
1618 merge_base_index(istate);
1619 return ret;
1620}
1621
fa7b3c2f
JH
1622int is_index_unborn(struct index_state *istate)
1623{
debed2a6 1624 return (!istate->cache_nr && !istate->timestamp.sec);
fa7b3c2f
JH
1625}
1626
4aab5b46 1627int discard_index(struct index_state *istate)
6d297f81 1628{
debed2a6
RS
1629 int i;
1630
5fc2fc8f
NTND
1631 for (i = 0; i < istate->cache_nr; i++) {
1632 if (istate->cache[i]->index &&
1633 istate->split_index &&
1634 istate->split_index->base &&
1635 istate->cache[i]->index <= istate->split_index->base->cache_nr &&
1636 istate->cache[i] == istate->split_index->base->cache[istate->cache[i]->index - 1])
1637 continue;
debed2a6 1638 free(istate->cache[i]);
5fc2fc8f 1639 }
cfc5789a 1640 resolve_undo_clear_index(istate);
4aab5b46
JH
1641 istate->cache_nr = 0;
1642 istate->cache_changed = 0;
fba2f38a
KB
1643 istate->timestamp.sec = 0;
1644 istate->timestamp.nsec = 0;
2092678c 1645 free_name_hash(istate);
4aab5b46 1646 cache_tree_free(&(istate->cache_tree));
913e0e99 1647 istate->initialized = 0;
a0fc4db0
RS
1648 free(istate->cache);
1649 istate->cache = NULL;
1650 istate->cache_alloc = 0;
5fc2fc8f 1651 discard_split_index(istate);
7a51ed66 1652 return 0;
6d297f81
JS
1653}
1654
d1f128b0 1655int unmerged_index(const struct index_state *istate)
94a5728c
DB
1656{
1657 int i;
1658 for (i = 0; i < istate->cache_nr; i++) {
1659 if (ce_stage(istate->cache[i]))
1660 return 1;
1661 }
1662 return 0;
1663}
1664
4990aadc 1665#define WRITE_BUFFER_SIZE 8192
bf0f910d 1666static unsigned char write_buffer[WRITE_BUFFER_SIZE];
4990aadc
LT
1667static unsigned long write_buffer_len;
1668
9126f009 1669static int ce_write_flush(git_SHA_CTX *context, int fd)
6015c28b
JH
1670{
1671 unsigned int buffered = write_buffer_len;
1672 if (buffered) {
9126f009 1673 git_SHA1_Update(context, write_buffer, buffered);
93822c22 1674 if (write_in_full(fd, write_buffer, buffered) != buffered)
6015c28b
JH
1675 return -1;
1676 write_buffer_len = 0;
1677 }
1678 return 0;
1679}
1680
9126f009 1681static int ce_write(git_SHA_CTX *context, int fd, void *data, unsigned int len)
4990aadc
LT
1682{
1683 while (len) {
1684 unsigned int buffered = write_buffer_len;
1685 unsigned int partial = WRITE_BUFFER_SIZE - buffered;
1686 if (partial > len)
1687 partial = len;
1688 memcpy(write_buffer + buffered, data, partial);
1689 buffered += partial;
1690 if (buffered == WRITE_BUFFER_SIZE) {
6015c28b
JH
1691 write_buffer_len = buffered;
1692 if (ce_write_flush(context, fd))
4990aadc
LT
1693 return -1;
1694 buffered = 0;
1695 }
1696 write_buffer_len = buffered;
1697 len -= partial;
1d7f171c 1698 data = (char *) data + partial;
a6080a0a
JH
1699 }
1700 return 0;
4990aadc
LT
1701}
1702
9126f009 1703static int write_index_ext_header(git_SHA_CTX *context, int fd,
ac58c7b1 1704 unsigned int ext, unsigned int sz)
bad68ec9
JH
1705{
1706 ext = htonl(ext);
1707 sz = htonl(sz);
968a1d65
DR
1708 return ((ce_write(context, fd, &ext, 4) < 0) ||
1709 (ce_write(context, fd, &sz, 4) < 0)) ? -1 : 0;
bad68ec9
JH
1710}
1711
e93021b2 1712static int ce_flush(git_SHA_CTX *context, int fd, unsigned char *sha1)
4990aadc
LT
1713{
1714 unsigned int left = write_buffer_len;
ca9be054 1715
4990aadc
LT
1716 if (left) {
1717 write_buffer_len = 0;
9126f009 1718 git_SHA1_Update(context, write_buffer, left);
4990aadc 1719 }
ca9be054 1720
2c865d9a
QH
1721 /* Flush first if not enough space for SHA1 signature */
1722 if (left + 20 > WRITE_BUFFER_SIZE) {
93822c22 1723 if (write_in_full(fd, write_buffer, left) != left)
2c865d9a
QH
1724 return -1;
1725 left = 0;
1726 }
1727
ca9be054 1728 /* Append the SHA1 signature at the end */
9126f009 1729 git_SHA1_Final(write_buffer + left, context);
e93021b2 1730 hashcpy(sha1, write_buffer + left);
ca9be054 1731 left += 20;
93822c22 1732 return (write_in_full(fd, write_buffer, left) != left) ? -1 : 0;
4990aadc
LT
1733}
1734
407c8eb0
JH
1735static void ce_smudge_racily_clean_entry(struct cache_entry *ce)
1736{
1737 /*
1738 * The only thing we care about in this function is to smudge the
1739 * falsely clean entry due to touch-update-touch race, so we leave
1740 * everything else as they are. We are called for entries whose
c21d39d7 1741 * ce_stat_data.sd_mtime match the index file mtime.
c70115b4
JH
1742 *
1743 * Note that this actually does not do much for gitlinks, for
1744 * which ce_match_stat_basic() always goes to the actual
1745 * contents. The caller checks with is_racy_timestamp() which
1746 * always says "no" for gitlinks, so we are not called for them ;-)
407c8eb0
JH
1747 */
1748 struct stat st;
1749
1750 if (lstat(ce->name, &st) < 0)
1751 return;
1752 if (ce_match_stat_basic(ce, &st))
1753 return;
1754 if (ce_modified_check_fs(ce, &st)) {
4b3511b0
JH
1755 /* This is "racily clean"; smudge it. Note that this
1756 * is a tricky code. At first glance, it may appear
1757 * that it can break with this sequence:
1758 *
1759 * $ echo xyzzy >frotz
1760 * $ git-update-index --add frotz
1761 * $ : >frotz
1762 * $ sleep 3
1763 * $ echo filfre >nitfol
1764 * $ git-update-index --add nitfol
1765 *
b7e58b17 1766 * but it does not. When the second update-index runs,
4b3511b0
JH
1767 * it notices that the entry "frotz" has the same timestamp
1768 * as index, and if we were to smudge it by resetting its
1769 * size to zero here, then the object name recorded
1770 * in index is the 6-byte file but the cached stat information
1771 * becomes zero --- which would then match what we would
a6080a0a 1772 * obtain from the filesystem next time we stat("frotz").
4b3511b0
JH
1773 *
1774 * However, the second update-index, before calling
1775 * this function, notices that the cached size is 6
1776 * bytes and what is on the filesystem is an empty
1777 * file, and never calls us, so the cached size information
1778 * for "frotz" stays 6 which does not match the filesystem.
1779 */
c21d39d7 1780 ce->ce_stat_data.sd_size = 0;
407c8eb0
JH
1781 }
1782}
1783
f136f7bf
JH
1784/* Copy miscellaneous fields but not the name */
1785static char *copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk,
1786 struct cache_entry *ce)
7a51ed66 1787{
b60e188c
TG
1788 short flags;
1789
c21d39d7
MH
1790 ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec);
1791 ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec);
1792 ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec);
1793 ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec);
1794 ondisk->dev = htonl(ce->ce_stat_data.sd_dev);
1795 ondisk->ino = htonl(ce->ce_stat_data.sd_ino);
7a51ed66 1796 ondisk->mode = htonl(ce->ce_mode);
c21d39d7
MH
1797 ondisk->uid = htonl(ce->ce_stat_data.sd_uid);
1798 ondisk->gid = htonl(ce->ce_stat_data.sd_gid);
1799 ondisk->size = htonl(ce->ce_stat_data.sd_size);
7a51ed66 1800 hashcpy(ondisk->sha1, ce->sha1);
b60e188c 1801
ce51bf09 1802 flags = ce->ce_flags & ~CE_NAMEMASK;
b60e188c
TG
1803 flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce));
1804 ondisk->flags = htons(flags);
06aaaa0b
NTND
1805 if (ce->ce_flags & CE_EXTENDED) {
1806 struct ondisk_cache_entry_extended *ondisk2;
1807 ondisk2 = (struct ondisk_cache_entry_extended *)ondisk;
1808 ondisk2->flags2 = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16);
f136f7bf 1809 return ondisk2->name;
06aaaa0b 1810 }
f136f7bf
JH
1811 else {
1812 return ondisk->name;
1813 }
1814}
1815
9d227781
JH
1816static int ce_write_entry(git_SHA_CTX *c, int fd, struct cache_entry *ce,
1817 struct strbuf *previous_name)
f136f7bf 1818{
9d227781
JH
1819 int size;
1820 struct ondisk_cache_entry *ondisk;
b3c96fb1 1821 int saved_namelen = saved_namelen; /* compiler workaround */
f136f7bf
JH
1822 char *name;
1823 int result;
1824
b3c96fb1
NTND
1825 if (ce->ce_flags & CE_STRIP_NAME) {
1826 saved_namelen = ce_namelen(ce);
1827 ce->ce_namelen = 0;
1828 }
1829
9d227781
JH
1830 if (!previous_name) {
1831 size = ondisk_ce_size(ce);
1832 ondisk = xcalloc(1, size);
1833 name = copy_cache_entry_to_ondisk(ondisk, ce);
1834 memcpy(name, ce->name, ce_namelen(ce));
1835 } else {
1836 int common, to_remove, prefix_size;
1837 unsigned char to_remove_vi[16];
1838 for (common = 0;
1839 (ce->name[common] &&
1840 common < previous_name->len &&
1841 ce->name[common] == previous_name->buf[common]);
1842 common++)
1843 ; /* still matching */
1844 to_remove = previous_name->len - common;
1845 prefix_size = encode_varint(to_remove, to_remove_vi);
1846
1847 if (ce->ce_flags & CE_EXTENDED)
1848 size = offsetof(struct ondisk_cache_entry_extended, name);
1849 else
1850 size = offsetof(struct ondisk_cache_entry, name);
1851 size += prefix_size + (ce_namelen(ce) - common + 1);
1852
1853 ondisk = xcalloc(1, size);
1854 name = copy_cache_entry_to_ondisk(ondisk, ce);
1855 memcpy(name, to_remove_vi, prefix_size);
1856 memcpy(name + prefix_size, ce->name + common, ce_namelen(ce) - common);
1857
1858 strbuf_splice(previous_name, common, to_remove,
1859 ce->name + common, ce_namelen(ce) - common);
06aaaa0b 1860 }
b3c96fb1
NTND
1861 if (ce->ce_flags & CE_STRIP_NAME) {
1862 ce->ce_namelen = saved_namelen;
1863 ce->ce_flags &= ~CE_STRIP_NAME;
1864 }
7a51ed66 1865
59efba64
JN
1866 result = ce_write(c, fd, ondisk, size);
1867 free(ondisk);
1868 return result;
7a51ed66
LT
1869}
1870
426ddeea
YM
1871/*
1872 * This function verifies if index_state has the correct sha1 of the
1873 * index file. Don't die if we have any other failure, just return 0.
1874 */
1875static int verify_index_from(const struct index_state *istate, const char *path)
1876{
1877 int fd;
1878 ssize_t n;
1879 struct stat st;
1880 unsigned char sha1[20];
1881
1882 if (!istate->initialized)
1883 return 0;
1884
1885 fd = open(path, O_RDONLY);
1886 if (fd < 0)
1887 return 0;
1888
1889 if (fstat(fd, &st))
1890 goto out;
1891
1892 if (st.st_size < sizeof(struct cache_header) + 20)
1893 goto out;
1894
1895 n = pread_in_full(fd, sha1, 20, st.st_size - 20);
1896 if (n != 20)
1897 goto out;
1898
1899 if (hashcmp(istate->sha1, sha1))
1900 goto out;
1901
1902 close(fd);
1903 return 1;
1904
1905out:
1906 close(fd);
1907 return 0;
1908}
1909
1910static int verify_index(const struct index_state *istate)
1911{
1912 return verify_index_from(istate, get_index_file());
1913}
1914
483fbe2b
JH
1915static int has_racy_timestamp(struct index_state *istate)
1916{
1917 int entries = istate->cache_nr;
1918 int i;
1919
1920 for (i = 0; i < entries; i++) {
1921 struct cache_entry *ce = istate->cache[i];
1922 if (is_racy_timestamp(istate, ce))
1923 return 1;
1924 }
1925 return 0;
1926}
1927
ccdc4ec3 1928/*
98e023de 1929 * Opportunistically update the index but do not complain if we can't
ccdc4ec3
JH
1930 */
1931void update_index_if_able(struct index_state *istate, struct lock_file *lockfile)
1932{
483fbe2b 1933 if ((istate->cache_changed || has_racy_timestamp(istate)) &&
788cef81 1934 verify_index(istate) &&
03b86647 1935 write_locked_index(istate, lockfile, COMMIT_LOCK))
ccdc4ec3
JH
1936 rollback_lock_file(lockfile);
1937}
1938
c18b80a0
NTND
1939static int do_write_index(struct index_state *istate, int newfd,
1940 int strip_extensions)
197ee8c9 1941{
9126f009 1942 git_SHA_CTX c;
197ee8c9 1943 struct cache_header hdr;
9d227781 1944 int i, err, removed, extended, hdr_version;
4aab5b46
JH
1945 struct cache_entry **cache = istate->cache;
1946 int entries = istate->cache_nr;
e1afca4f 1947 struct stat st;
9d227781 1948 struct strbuf previous_name_buf = STRBUF_INIT, *previous_name;
025a0709 1949
06aaaa0b 1950 for (i = removed = extended = 0; i < entries; i++) {
7a51ed66 1951 if (cache[i]->ce_flags & CE_REMOVE)
025a0709 1952 removed++;
197ee8c9 1953
06aaaa0b
NTND
1954 /* reduce extended entries if possible */
1955 cache[i]->ce_flags &= ~CE_EXTENDED;
1956 if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) {
1957 extended++;
1958 cache[i]->ce_flags |= CE_EXTENDED;
1959 }
1960 }
1961
d6e3c181 1962 if (!istate->version) {
136347d7 1963 istate->version = get_index_format_default();
d6e3c181
NTND
1964 if (getenv("GIT_TEST_SPLIT_INDEX"))
1965 init_split_index(istate);
1966 }
9d227781
JH
1967
1968 /* demote version 3 to version 2 when the latter suffices */
1969 if (istate->version == 3 || istate->version == 2)
1970 istate->version = extended ? 3 : 2;
1971
1972 hdr_version = istate->version;
1973
ccc4feb5 1974 hdr.hdr_signature = htonl(CACHE_SIGNATURE);
9d227781 1975 hdr.hdr_version = htonl(hdr_version);
025a0709 1976 hdr.hdr_entries = htonl(entries - removed);
197ee8c9 1977
9126f009 1978 git_SHA1_Init(&c);
ca9be054 1979 if (ce_write(&c, newfd, &hdr, sizeof(hdr)) < 0)
197ee8c9
LT
1980 return -1;
1981
9d227781 1982 previous_name = (hdr_version == 4) ? &previous_name_buf : NULL;
197ee8c9
LT
1983 for (i = 0; i < entries; i++) {
1984 struct cache_entry *ce = cache[i];
7a51ed66 1985 if (ce->ce_flags & CE_REMOVE)
aa16021e 1986 continue;
e06c43c7 1987 if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce))
407c8eb0 1988 ce_smudge_racily_clean_entry(ce);
83bd7437
JK
1989 if (is_null_sha1(ce->sha1)) {
1990 static const char msg[] = "cache entry has null sha1: %s";
1991 static int allow = -1;
1992
1993 if (allow < 0)
1994 allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0);
1995 if (allow)
1996 warning(msg, ce->name);
1997 else
1998 return error(msg, ce->name);
1999 }
9d227781 2000 if (ce_write_entry(&c, newfd, ce, previous_name) < 0)
197ee8c9
LT
2001 return -1;
2002 }
9d227781 2003 strbuf_release(&previous_name_buf);
1af1c2b6 2004
bad68ec9 2005 /* Write extension data here */
c18b80a0 2006 if (!strip_extensions && istate->split_index) {
5fc2fc8f
NTND
2007 struct strbuf sb = STRBUF_INIT;
2008
2009 err = write_link_extension(&sb, istate) < 0 ||
2010 write_index_ext_header(&c, newfd, CACHE_EXT_LINK,
2011 sb.len) < 0 ||
2012 ce_write(&c, newfd, sb.buf, sb.len) < 0;
2013 strbuf_release(&sb);
2014 if (err)
2015 return -1;
2016 }
c18b80a0 2017 if (!strip_extensions && istate->cache_tree) {
f285a2d7 2018 struct strbuf sb = STRBUF_INIT;
1dffb8fa 2019
1dffb8fa
PH
2020 cache_tree_write(&sb, istate->cache_tree);
2021 err = write_index_ext_header(&c, newfd, CACHE_EXT_TREE, sb.len) < 0
2022 || ce_write(&c, newfd, sb.buf, sb.len) < 0;
2023 strbuf_release(&sb);
2024 if (err)
bad68ec9 2025 return -1;
bad68ec9 2026 }
c18b80a0 2027 if (!strip_extensions && istate->resolve_undo) {
cfc5789a
JH
2028 struct strbuf sb = STRBUF_INIT;
2029
2030 resolve_undo_write(&sb, istate->resolve_undo);
2031 err = write_index_ext_header(&c, newfd, CACHE_EXT_RESOLVE_UNDO,
2032 sb.len) < 0
2033 || ce_write(&c, newfd, sb.buf, sb.len) < 0;
2034 strbuf_release(&sb);
2035 if (err)
2036 return -1;
2037 }
e1afca4f 2038
e93021b2 2039 if (ce_flush(&c, newfd, istate->sha1) || fstat(newfd, &st))
e1afca4f 2040 return -1;
5bcf109c
KB
2041 istate->timestamp.sec = (unsigned int)st.st_mtime;
2042 istate->timestamp.nsec = ST_MTIME_NSEC(st);
e1afca4f 2043 return 0;
197ee8c9 2044}
e46bbcf6 2045
626f35c8
NTND
2046void set_alternate_index_output(const char *name)
2047{
2048 alternate_index_output = name;
2049}
2050
2051static int commit_locked_index(struct lock_file *lk)
2052{
751baced
MH
2053 if (alternate_index_output)
2054 return commit_lock_file_to(lk, alternate_index_output);
2055 else
626f35c8 2056 return commit_lock_file(lk);
626f35c8
NTND
2057}
2058
03b86647
NTND
2059static int do_write_locked_index(struct index_state *istate, struct lock_file *lock,
2060 unsigned flags)
2061{
c18b80a0 2062 int ret = do_write_index(istate, lock->fd, 0);
03b86647
NTND
2063 if (ret)
2064 return ret;
2065 assert((flags & (COMMIT_LOCK | CLOSE_LOCK)) !=
2066 (COMMIT_LOCK | CLOSE_LOCK));
2067 if (flags & COMMIT_LOCK)
2068 return commit_locked_index(lock);
2069 else if (flags & CLOSE_LOCK)
2070 return close_lock_file(lock);
2071 else
2072 return ret;
2073}
2074
5fc2fc8f
NTND
2075static int write_split_index(struct index_state *istate,
2076 struct lock_file *lock,
2077 unsigned flags)
2078{
2079 int ret;
2080 prepare_to_write_split_index(istate);
2081 ret = do_write_locked_index(istate, lock, flags);
2082 finish_writing_split_index(istate);
2083 return ret;
2084}
2085
c18b80a0
NTND
2086static char *temporary_sharedindex;
2087
2088static void remove_temporary_sharedindex(void)
2089{
2090 if (temporary_sharedindex) {
2091 unlink_or_warn(temporary_sharedindex);
2092 free(temporary_sharedindex);
2093 temporary_sharedindex = NULL;
2094 }
2095}
2096
2097static void remove_temporary_sharedindex_on_signal(int signo)
2098{
2099 remove_temporary_sharedindex();
2100 sigchain_pop(signo);
2101 raise(signo);
2102}
2103
a0a96756
NTND
2104static int write_shared_index(struct index_state *istate,
2105 struct lock_file *lock, unsigned flags)
c18b80a0
NTND
2106{
2107 struct split_index *si = istate->split_index;
2108 static int installed_handler;
2109 int fd, ret;
2110
2111 temporary_sharedindex = git_pathdup("sharedindex_XXXXXX");
a0a96756
NTND
2112 fd = mkstemp(temporary_sharedindex);
2113 if (fd < 0) {
2114 free(temporary_sharedindex);
2115 temporary_sharedindex = NULL;
2116 hashclr(si->base_sha1);
2117 return do_write_locked_index(istate, lock, flags);
2118 }
c18b80a0
NTND
2119 if (!installed_handler) {
2120 atexit(remove_temporary_sharedindex);
2121 sigchain_push_common(remove_temporary_sharedindex_on_signal);
2122 }
2123 move_cache_to_base_index(istate);
2124 ret = do_write_index(si->base, fd, 1);
2125 close(fd);
2126 if (ret) {
2127 remove_temporary_sharedindex();
2128 return ret;
2129 }
2130 ret = rename(temporary_sharedindex,
2131 git_path("sharedindex.%s", sha1_to_hex(si->base->sha1)));
2132 free(temporary_sharedindex);
2133 temporary_sharedindex = NULL;
2134 if (!ret)
2135 hashcpy(si->base_sha1, si->base->sha1);
2136 return ret;
2137}
2138
03b86647
NTND
2139int write_locked_index(struct index_state *istate, struct lock_file *lock,
2140 unsigned flags)
2141{
5fc2fc8f
NTND
2142 struct split_index *si = istate->split_index;
2143
5165dd59
NTND
2144 if (!si || alternate_index_output ||
2145 (istate->cache_changed & ~EXTMASK)) {
5fc2fc8f
NTND
2146 if (si)
2147 hashclr(si->base_sha1);
2148 return do_write_locked_index(istate, lock, flags);
2149 }
2150
d6e3c181
NTND
2151 if (getenv("GIT_TEST_SPLIT_INDEX")) {
2152 int v = si->base_sha1[0];
2153 if ((v & 15) < 6)
2154 istate->cache_changed |= SPLIT_INDEX_ORDERED;
2155 }
c18b80a0 2156 if (istate->cache_changed & SPLIT_INDEX_ORDERED) {
a0a96756 2157 int ret = write_shared_index(istate, lock, flags);
c18b80a0
NTND
2158 if (ret)
2159 return ret;
2160 }
2161
5fc2fc8f 2162 return write_split_index(istate, lock, flags);
03b86647
NTND
2163}
2164
e46bbcf6
MV
2165/*
2166 * Read the index file that is potentially unmerged into given
63e8dc5b 2167 * index_state, dropping any unmerged entries. Returns true if
e46bbcf6
MV
2168 * the index is unmerged. Callers who want to refuse to work
2169 * from an unmerged state can call this and check its return value,
2170 * instead of calling read_cache().
2171 */
2172int read_index_unmerged(struct index_state *istate)
2173{
2174 int i;
d1a43f2a 2175 int unmerged = 0;
e46bbcf6
MV
2176
2177 read_index(istate);
e46bbcf6
MV
2178 for (i = 0; i < istate->cache_nr; i++) {
2179 struct cache_entry *ce = istate->cache[i];
d1a43f2a
JH
2180 struct cache_entry *new_ce;
2181 int size, len;
2182
2183 if (!ce_stage(ce))
e46bbcf6 2184 continue;
d1a43f2a 2185 unmerged = 1;
68c4f6a5 2186 len = ce_namelen(ce);
d1a43f2a
JH
2187 size = cache_entry_size(len);
2188 new_ce = xcalloc(1, size);
d1a43f2a 2189 memcpy(new_ce->name, ce->name, len);
b60e188c
TG
2190 new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED;
2191 new_ce->ce_namelen = len;
d1a43f2a
JH
2192 new_ce->ce_mode = ce->ce_mode;
2193 if (add_index_entry(istate, new_ce, 0))
2194 return error("%s: cannot drop to stage #0",
5699d17e 2195 new_ce->name);
e46bbcf6 2196 }
d1a43f2a 2197 return unmerged;
e46bbcf6 2198}
041aee31 2199
98fa4738
JK
2200/*
2201 * Returns 1 if the path is an "other" path with respect to
2202 * the index; that is, the path is not mentioned in the index at all,
2203 * either as a file, a directory with some files in the index,
2204 * or as an unmerged entry.
2205 *
2206 * We helpfully remove a trailing "/" from directories so that
2207 * the output of read_directory can be used as-is.
2208 */
2209int index_name_is_other(const struct index_state *istate, const char *name,
2210 int namelen)
2211{
2212 int pos;
2213 if (namelen && name[namelen - 1] == '/')
2214 namelen--;
2215 pos = index_name_pos(istate, name, namelen);
2216 if (0 <= pos)
2217 return 0; /* exact match */
2218 pos = -pos - 1;
2219 if (pos < istate->cache_nr) {
2220 struct cache_entry *ce = istate->cache[pos];
2221 if (ce_namelen(ce) == namelen &&
2222 !memcmp(ce->name, name, namelen))
2223 return 0; /* Yup, this one exists unmerged */
2224 }
2225 return 1;
2226}
29fb37b2 2227
ff366825 2228void *read_blob_data_from_index(struct index_state *istate, const char *path, unsigned long *size)
29fb37b2
LF
2229{
2230 int pos, len;
2231 unsigned long sz;
2232 enum object_type type;
2233 void *data;
2234
2235 len = strlen(path);
2236 pos = index_name_pos(istate, path, len);
2237 if (pos < 0) {
2238 /*
2239 * We might be in the middle of a merge, in which
2240 * case we would read stage #2 (ours).
2241 */
2242 int i;
2243 for (i = -pos - 1;
2244 (pos < 0 && i < istate->cache_nr &&
2245 !strcmp(istate->cache[i]->name, path));
2246 i++)
2247 if (ce_stage(istate->cache[i]) == 2)
2248 pos = i;
2249 }
2250 if (pos < 0)
2251 return NULL;
2252 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
2253 if (!data || type != OBJ_BLOB) {
2254 free(data);
2255 return NULL;
2256 }
ff366825
LF
2257 if (size)
2258 *size = sz;
29fb37b2
LF
2259 return data;
2260}
38612532
MH
2261
2262void stat_validity_clear(struct stat_validity *sv)
2263{
2264 free(sv->sd);
2265 sv->sd = NULL;
2266}
2267
2268int stat_validity_check(struct stat_validity *sv, const char *path)
2269{
2270 struct stat st;
2271
2272 if (stat(path, &st) < 0)
2273 return sv->sd == NULL;
2274 if (!sv->sd)
2275 return 0;
2276 return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st);
2277}
2278
2279void stat_validity_update(struct stat_validity *sv, int fd)
2280{
2281 struct stat st;
2282
2283 if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode))
2284 stat_validity_clear(sv);
2285 else {
2286 if (!sv->sd)
2287 sv->sd = xcalloc(1, sizeof(struct stat_data));
2288 fill_stat_data(sv->sd, &st);
2289 }
2290}