varint: make it available outside the context of pack
[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"
bad68ec9 15
87b29e5a
JH
16static struct cache_entry *refresh_cache_entry(struct cache_entry *ce, int really);
17
bad68ec9
JH
18/* Index extensions.
19 *
20 * The first letter should be 'A'..'Z' for extensions that are not
21 * necessary for a correct operation (i.e. optimization data).
22 * When new extensions are added that _needs_ to be understood in
23 * order to correctly interpret the index file, pick character that
24 * is outside the range, to cause the reader to abort.
25 */
26
27#define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) )
28#define CACHE_EXT_TREE 0x54524545 /* "TREE" */
b659b49b 29#define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */
e83c5163 30
228e94f9 31struct index_state the_index;
8fd2cb40 32
9cb76b8c
JH
33static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce)
34{
35 istate->cache[nr] = ce;
96872bc2 36 add_name_hash(istate, ce);
9cb76b8c
JH
37}
38
cf558704
LT
39static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce)
40{
41 struct cache_entry *old = istate->cache[nr];
42
96872bc2 43 remove_name_hash(old);
a22c6371 44 set_index_entry(istate, nr, ce);
cf558704
LT
45 istate->cache_changed = 1;
46}
47
81dc2307
PB
48void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name)
49{
50 struct cache_entry *old = istate->cache[nr], *new;
51 int namelen = strlen(new_name);
52
53 new = xmalloc(cache_entry_size(namelen));
54 copy_cache_entry(new, old);
55 new->ce_flags &= ~(CE_STATE_MASK | CE_NAMEMASK);
56 new->ce_flags |= (namelen >= CE_NAMEMASK ? CE_NAMEMASK : namelen);
57 memcpy(new->name, new_name, namelen + 1);
58
59 cache_tree_invalidate_path(istate->cache_tree, old->name);
60 remove_index_entry_at(istate, nr);
61 add_index_entry(istate, new, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
62}
63
415e96c8
JH
64/*
65 * This only updates the "non-critical" parts of the directory
66 * cache, ie the parts that aren't tracked by GIT, and only used
67 * to validate the cache.
68 */
69void fill_stat_cache_info(struct cache_entry *ce, struct stat *st)
70{
fba2f38a
KB
71 ce->ce_ctime.sec = (unsigned int)st->st_ctime;
72 ce->ce_mtime.sec = (unsigned int)st->st_mtime;
c06ff490
KB
73 ce->ce_ctime.nsec = ST_CTIME_NSEC(*st);
74 ce->ce_mtime.nsec = ST_MTIME_NSEC(*st);
7a51ed66
LT
75 ce->ce_dev = st->st_dev;
76 ce->ce_ino = st->st_ino;
77 ce->ce_uid = st->st_uid;
78 ce->ce_gid = st->st_gid;
79 ce->ce_size = st->st_size;
5f73076c
JH
80
81 if (assume_unchanged)
7a51ed66 82 ce->ce_flags |= CE_VALID;
eadb5831
JH
83
84 if (S_ISREG(st->st_mode))
85 ce_mark_uptodate(ce);
415e96c8
JH
86}
87
29e4d363
JH
88static int ce_compare_data(struct cache_entry *ce, struct stat *st)
89{
90 int match = -1;
91 int fd = open(ce->name, O_RDONLY);
92
93 if (fd >= 0) {
94 unsigned char sha1[20];
c4ce46fc 95 if (!index_fd(sha1, fd, st, OBJ_BLOB, ce->name, 0))
a89fccd2 96 match = hashcmp(sha1, ce->sha1);
7f8508e8 97 /* index_fd() closed the file descriptor already */
29e4d363
JH
98 }
99 return match;
100}
101
dc49cd76 102static int ce_compare_link(struct cache_entry *ce, size_t expected_size)
29e4d363
JH
103{
104 int match = -1;
29e4d363
JH
105 void *buffer;
106 unsigned long size;
21666f1a 107 enum object_type type;
a60272b3 108 struct strbuf sb = STRBUF_INIT;
29e4d363 109
a60272b3 110 if (strbuf_readlink(&sb, ce->name, expected_size))
29e4d363 111 return -1;
a60272b3 112
21666f1a 113 buffer = read_sha1_file(ce->sha1, &type, &size);
a60272b3
LT
114 if (buffer) {
115 if (size == sb.len)
116 match = memcmp(buffer, sb.buf, size);
117 free(buffer);
29e4d363 118 }
a60272b3 119 strbuf_release(&sb);
29e4d363
JH
120 return match;
121}
122
f35a6d3b
LT
123static int ce_compare_gitlink(struct cache_entry *ce)
124{
125 unsigned char sha1[20];
126
127 /*
128 * We don't actually require that the .git directory
302b9282 129 * under GITLINK directory be a valid git directory. It
f35a6d3b
LT
130 * might even be missing (in case nobody populated that
131 * sub-project).
132 *
133 * If so, we consider it always to match.
134 */
135 if (resolve_gitlink_ref(ce->name, "HEAD", sha1) < 0)
136 return 0;
137 return hashcmp(sha1, ce->sha1);
138}
139
29e4d363
JH
140static int ce_modified_check_fs(struct cache_entry *ce, struct stat *st)
141{
142 switch (st->st_mode & S_IFMT) {
143 case S_IFREG:
144 if (ce_compare_data(ce, st))
145 return DATA_CHANGED;
146 break;
147 case S_IFLNK:
dc49cd76 148 if (ce_compare_link(ce, xsize_t(st->st_size)))
29e4d363
JH
149 return DATA_CHANGED;
150 break;
a8ee75bc 151 case S_IFDIR:
7a51ed66 152 if (S_ISGITLINK(ce->ce_mode))
c70115b4 153 return ce_compare_gitlink(ce) ? DATA_CHANGED : 0;
29e4d363
JH
154 default:
155 return TYPE_CHANGED;
156 }
157 return 0;
158}
159
87b29e5a 160static int is_empty_blob_sha1(const unsigned char *sha1)
f49c2c22
LT
161{
162 static const unsigned char empty_blob_sha1[20] = {
163 0xe6,0x9d,0xe2,0x9b,0xb2,0xd1,0xd6,0x43,0x4b,0x8b,
164 0x29,0xae,0x77,0x5a,0xd8,0xc2,0xe4,0x8c,0x53,0x91
165 };
166
167 return !hashcmp(sha1, empty_blob_sha1);
168}
169
407c8eb0 170static int ce_match_stat_basic(struct cache_entry *ce, struct stat *st)
734aab75
LT
171{
172 unsigned int changed = 0;
173
7a51ed66
LT
174 if (ce->ce_flags & CE_REMOVE)
175 return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED;
176
177 switch (ce->ce_mode & S_IFMT) {
8ae0a8c5
KS
178 case S_IFREG:
179 changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0;
3e09cdfd
JH
180 /* We consider only the owner x bit to be relevant for
181 * "mode changes"
182 */
183 if (trust_executable_bit &&
7a51ed66 184 (0100 & (ce->ce_mode ^ st->st_mode)))
ffbe1add 185 changed |= MODE_CHANGED;
8ae0a8c5
KS
186 break;
187 case S_IFLNK:
78a8d641
JS
188 if (!S_ISLNK(st->st_mode) &&
189 (has_symlinks || !S_ISREG(st->st_mode)))
190 changed |= TYPE_CHANGED;
8ae0a8c5 191 break;
302b9282 192 case S_IFGITLINK:
c70115b4 193 /* We ignore most of the st_xxx fields for gitlinks */
f35a6d3b
LT
194 if (!S_ISDIR(st->st_mode))
195 changed |= TYPE_CHANGED;
196 else if (ce_compare_gitlink(ce))
197 changed |= DATA_CHANGED;
a8ee75bc 198 return changed;
8ae0a8c5 199 default:
7a51ed66 200 die("internal error: ce_mode is %o", ce->ce_mode);
8ae0a8c5 201 }
fba2f38a 202 if (ce->ce_mtime.sec != (unsigned int)st->st_mtime)
ccc4feb5 203 changed |= MTIME_CHANGED;
fba2f38a 204 if (trust_ctime && ce->ce_ctime.sec != (unsigned int)st->st_ctime)
734aab75 205 changed |= CTIME_CHANGED;
ccc4feb5 206
fba2f38a 207#ifdef USE_NSEC
110c46a9 208 if (ce->ce_mtime.nsec != ST_MTIME_NSEC(*st))
fba2f38a 209 changed |= MTIME_CHANGED;
110c46a9 210 if (trust_ctime && ce->ce_ctime.nsec != ST_CTIME_NSEC(*st))
fba2f38a
KB
211 changed |= CTIME_CHANGED;
212#endif
213
7a51ed66
LT
214 if (ce->ce_uid != (unsigned int) st->st_uid ||
215 ce->ce_gid != (unsigned int) st->st_gid)
734aab75 216 changed |= OWNER_CHANGED;
7a51ed66 217 if (ce->ce_ino != (unsigned int) st->st_ino)
734aab75 218 changed |= INODE_CHANGED;
2cb45e95
LT
219
220#ifdef USE_STDEV
221 /*
222 * st_dev breaks on network filesystems where different
223 * clients will have different views of what "device"
224 * the filesystem is on
225 */
7a51ed66 226 if (ce->ce_dev != (unsigned int) st->st_dev)
2cb45e95
LT
227 changed |= INODE_CHANGED;
228#endif
229
7a51ed66 230 if (ce->ce_size != (unsigned int) st->st_size)
734aab75 231 changed |= DATA_CHANGED;
b0391890 232
f49c2c22
LT
233 /* Racily smudged entry? */
234 if (!ce->ce_size) {
235 if (!is_empty_blob_sha1(ce->sha1))
236 changed |= DATA_CHANGED;
237 }
238
407c8eb0
JH
239 return changed;
240}
241
d1f128b0 242static int is_racy_timestamp(const struct index_state *istate, struct cache_entry *ce)
6d91da6d 243{
050288d5 244 return (!S_ISGITLINK(ce->ce_mode) &&
fba2f38a
KB
245 istate->timestamp.sec &&
246#ifdef USE_NSEC
247 /* nanosecond timestamped files can also be racy! */
248 (istate->timestamp.sec < ce->ce_mtime.sec ||
249 (istate->timestamp.sec == ce->ce_mtime.sec &&
250 istate->timestamp.nsec <= ce->ce_mtime.nsec))
251#else
252 istate->timestamp.sec <= ce->ce_mtime.sec
253#endif
254 );
6d91da6d
JH
255}
256
d1f128b0 257int ie_match_stat(const struct index_state *istate,
4bd5b7da
JH
258 struct cache_entry *ce, struct stat *st,
259 unsigned int options)
407c8eb0 260{
5f73076c 261 unsigned int changed;
4bd5b7da 262 int ignore_valid = options & CE_MATCH_IGNORE_VALID;
56cac48c 263 int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE;
4bd5b7da 264 int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY;
5f73076c
JH
265
266 /*
267 * If it's marked as always valid in the index, it's
268 * valid whatever the checked-out copy says.
56cac48c
NTND
269 *
270 * skip-worktree has the same effect with higher precedence
5f73076c 271 */
56cac48c
NTND
272 if (!ignore_skip_worktree && ce_skip_worktree(ce))
273 return 0;
7a51ed66 274 if (!ignore_valid && (ce->ce_flags & CE_VALID))
5f73076c
JH
275 return 0;
276
331fcb59
JH
277 /*
278 * Intent-to-add entries have not been added, so the index entry
279 * by definition never matches what is in the work tree until it
280 * actually gets added.
281 */
282 if (ce->ce_flags & CE_INTENT_TO_ADD)
283 return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED;
284
5f73076c 285 changed = ce_match_stat_basic(ce, st);
407c8eb0 286
29e4d363
JH
287 /*
288 * Within 1 second of this sequence:
289 * echo xyzzy >file && git-update-index --add file
290 * running this command:
291 * echo frotz >file
292 * would give a falsely clean cache entry. The mtime and
293 * length match the cache, and other stat fields do not change.
294 *
295 * We could detect this at update-index time (the cache entry
296 * being registered/updated records the same time as "now")
297 * and delay the return from git-update-index, but that would
298 * effectively mean we can make at most one commit per second,
299 * which is not acceptable. Instead, we check cache entries
300 * whose mtime are the same as the index file timestamp more
5f73076c 301 * carefully than others.
29e4d363 302 */
6d91da6d 303 if (!changed && is_racy_timestamp(istate, ce)) {
42f77406
JH
304 if (assume_racy_is_modified)
305 changed |= DATA_CHANGED;
306 else
307 changed |= ce_modified_check_fs(ce, st);
308 }
b0391890 309
29e4d363 310 return changed;
b0391890
JH
311}
312
d1f128b0 313int ie_modified(const struct index_state *istate,
4bd5b7da 314 struct cache_entry *ce, struct stat *st, unsigned int options)
b0391890 315{
29e4d363 316 int changed, changed_fs;
4bd5b7da
JH
317
318 changed = ie_match_stat(istate, ce, st, options);
b0391890
JH
319 if (!changed)
320 return 0;
b0391890
JH
321 /*
322 * If the mode or type has changed, there's no point in trying
323 * to refresh the entry - it's not going to match
324 */
325 if (changed & (MODE_CHANGED | TYPE_CHANGED))
326 return changed;
327
c70115b4
JH
328 /*
329 * Immediately after read-tree or update-index --cacheinfo,
330 * the length field is zero, as we have never even read the
331 * lstat(2) information once, and we cannot trust DATA_CHANGED
332 * returned by ie_match_stat() which in turn was returned by
333 * ce_match_stat_basic() to signal that the filesize of the
334 * blob changed. We have to actually go to the filesystem to
335 * see if the contents match, and if so, should answer "unchanged".
336 *
337 * The logic does not apply to gitlinks, as ce_match_stat_basic()
338 * already has checked the actual HEAD from the filesystem in the
339 * subproject. If ie_match_stat() already said it is different,
340 * then we know it is.
b0391890 341 */
c70115b4
JH
342 if ((changed & DATA_CHANGED) &&
343 (S_ISGITLINK(ce->ce_mode) || ce->ce_size != 0))
b0391890
JH
344 return changed;
345
29e4d363
JH
346 changed_fs = ce_modified_check_fs(ce, st);
347 if (changed_fs)
348 return changed | changed_fs;
b0391890
JH
349 return 0;
350}
351
958ba6c9
LT
352int base_name_compare(const char *name1, int len1, int mode1,
353 const char *name2, int len2, int mode2)
354{
355 unsigned char c1, c2;
356 int len = len1 < len2 ? len1 : len2;
357 int cmp;
358
359 cmp = memcmp(name1, name2, len);
360 if (cmp)
361 return cmp;
362 c1 = name1[len];
363 c2 = name2[len];
1833a925 364 if (!c1 && S_ISDIR(mode1))
958ba6c9 365 c1 = '/';
1833a925 366 if (!c2 && S_ISDIR(mode2))
958ba6c9
LT
367 c2 = '/';
368 return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0;
369}
370
0ab9e1e8
LT
371/*
372 * df_name_compare() is identical to base_name_compare(), except it
373 * compares conflicting directory/file entries as equal. Note that
374 * while a directory name compares as equal to a regular file, they
375 * then individually compare _differently_ to a filename that has
376 * a dot after the basename (because '\0' < '.' < '/').
377 *
378 * This is used by routines that want to traverse the git namespace
379 * but then handle conflicting entries together when possible.
380 */
381int df_name_compare(const char *name1, int len1, int mode1,
382 const char *name2, int len2, int mode2)
383{
384 int len = len1 < len2 ? len1 : len2, cmp;
385 unsigned char c1, c2;
386
387 cmp = memcmp(name1, name2, len);
388 if (cmp)
389 return cmp;
390 /* Directories and files compare equal (same length, same name) */
391 if (len1 == len2)
392 return 0;
393 c1 = name1[len];
394 if (!c1 && S_ISDIR(mode1))
395 c1 = '/';
396 c2 = name2[len];
397 if (!c2 && S_ISDIR(mode2))
398 c2 = '/';
399 if (c1 == '/' && !c2)
400 return 0;
401 if (c2 == '/' && !c1)
402 return 0;
403 return c1 - c2;
404}
405
95fd5bf8 406int cache_name_compare(const char *name1, int flags1, const char *name2, int flags2)
eb38c22f 407{
95fd5bf8
LT
408 int len1 = flags1 & CE_NAMEMASK;
409 int len2 = flags2 & CE_NAMEMASK;
eb38c22f
LT
410 int len = len1 < len2 ? len1 : len2;
411 int cmp;
412
413 cmp = memcmp(name1, name2, len);
414 if (cmp)
415 return cmp;
416 if (len1 < len2)
417 return -1;
418 if (len1 > len2)
419 return 1;
5f73076c 420
7b80be15
JH
421 /* Compare stages */
422 flags1 &= CE_STAGEMASK;
423 flags2 &= CE_STAGEMASK;
5f73076c 424
95fd5bf8
LT
425 if (flags1 < flags2)
426 return -1;
427 if (flags1 > flags2)
428 return 1;
eb38c22f
LT
429 return 0;
430}
431
d1f128b0 432int index_name_pos(const struct index_state *istate, const char *name, int namelen)
eb38c22f
LT
433{
434 int first, last;
435
436 first = 0;
4aab5b46 437 last = istate->cache_nr;
eb38c22f
LT
438 while (last > first) {
439 int next = (last + first) >> 1;
4aab5b46 440 struct cache_entry *ce = istate->cache[next];
7a51ed66 441 int cmp = cache_name_compare(name, namelen, ce->name, ce->ce_flags);
eb38c22f 442 if (!cmp)
76e7f4ec 443 return next;
eb38c22f
LT
444 if (cmp < 0) {
445 last = next;
446 continue;
447 }
448 first = next+1;
449 }
76e7f4ec 450 return -first-1;
eb38c22f
LT
451}
452
7b937ca3 453/* Remove entry, return true if there are more entries to go.. */
4aab5b46 454int remove_index_entry_at(struct index_state *istate, int pos)
7b937ca3 455{
cf558704
LT
456 struct cache_entry *ce = istate->cache[pos];
457
cfc5789a 458 record_resolve_undo(istate, ce);
96872bc2 459 remove_name_hash(ce);
4aab5b46
JH
460 istate->cache_changed = 1;
461 istate->cache_nr--;
462 if (pos >= istate->cache_nr)
7b937ca3 463 return 0;
4aab5b46
JH
464 memmove(istate->cache + pos,
465 istate->cache + pos + 1,
466 (istate->cache_nr - pos) * sizeof(struct cache_entry *));
7b937ca3
LT
467 return 1;
468}
469
36419c8e
KB
470/*
471 * Remove all cache ententries marked for removal, that is where
472 * CE_REMOVE is set in ce_flags. This is much more effective than
473 * calling remove_index_entry_at() for each entry to be removed.
474 */
475void remove_marked_cache_entries(struct index_state *istate)
476{
477 struct cache_entry **ce_array = istate->cache;
478 unsigned int i, j;
479
480 for (i = j = 0; i < istate->cache_nr; i++) {
481 if (ce_array[i]->ce_flags & CE_REMOVE)
482 remove_name_hash(ce_array[i]);
483 else
484 ce_array[j++] = ce_array[i];
485 }
486 istate->cache_changed = 1;
487 istate->cache_nr = j;
488}
489
4aab5b46 490int remove_file_from_index(struct index_state *istate, const char *path)
197ee8c9 491{
4aab5b46 492 int pos = index_name_pos(istate, path, strlen(path));
c4e3cca1
JH
493 if (pos < 0)
494 pos = -pos-1;
09d5dc32 495 cache_tree_invalidate_path(istate->cache_tree, path);
4aab5b46
JH
496 while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path))
497 remove_index_entry_at(istate, pos);
197ee8c9
LT
498 return 0;
499}
500
20314271
JS
501static int compare_name(struct cache_entry *ce, const char *path, int namelen)
502{
503 return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen);
504}
505
506static int index_name_pos_also_unmerged(struct index_state *istate,
507 const char *path, int namelen)
508{
509 int pos = index_name_pos(istate, path, namelen);
510 struct cache_entry *ce;
511
512 if (pos >= 0)
513 return pos;
514
515 /* maybe unmerged? */
516 pos = -1 - pos;
517 if (pos >= istate->cache_nr ||
518 compare_name((ce = istate->cache[pos]), path, namelen))
519 return -1;
520
521 /* order of preference: stage 2, 1, 3 */
522 if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr &&
523 ce_stage((ce = istate->cache[pos + 1])) == 2 &&
524 !compare_name(ce, path, namelen))
525 pos++;
526 return pos;
527}
528
1102952b
LT
529static int different_name(struct cache_entry *ce, struct cache_entry *alias)
530{
531 int len = ce_namelen(ce);
532 return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len);
533}
534
535/*
536 * If we add a filename that aliases in the cache, we will use the
537 * name that we already have - but we don't want to update the same
538 * alias twice, because that implies that there were actually two
539 * different files with aliasing names!
540 *
541 * So we use the CE_ADDED flag to verify that the alias was an old
542 * one before we accept it as
543 */
544static struct cache_entry *create_alias_ce(struct cache_entry *ce, struct cache_entry *alias)
545{
546 int len;
547 struct cache_entry *new;
548
549 if (alias->ce_flags & CE_ADDED)
550 die("Will not add file alias '%s' ('%s' already exists in index)", ce->name, alias->name);
551
552 /* Ok, create the new entry using the name of the existing alias */
553 len = ce_namelen(alias);
554 new = xcalloc(1, cache_entry_size(len));
555 memcpy(new->name, alias->name, len);
556 copy_cache_entry(new, ce);
557 free(ce);
558 return new;
559}
560
39425819
JH
561static void record_intent_to_add(struct cache_entry *ce)
562{
563 unsigned char sha1[20];
564 if (write_sha1_file("", 0, blob_type, sha1))
565 die("cannot create an empty blob in the object database");
566 hashcpy(ce->sha1, sha1);
567}
568
38ed1d89 569int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags)
11be42a4 570{
38ed1d89 571 int size, namelen, was_same;
d177cab0 572 mode_t st_mode = st->st_mode;
6835550d 573 struct cache_entry *ce, *alias;
56cac48c 574 unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY;
38ed1d89
JH
575 int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND);
576 int pretend = flags & ADD_CACHE_PRETEND;
39425819
JH
577 int intent_only = flags & ADD_CACHE_INTENT;
578 int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE|
579 (intent_only ? ADD_CACHE_NEW_ONLY : 0));
11be42a4 580
d177cab0 581 if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode))
960b8ad1 582 return error("%s: can only add regular files, symbolic links or git-directories", path);
11be42a4
JS
583
584 namelen = strlen(path);
d177cab0 585 if (S_ISDIR(st_mode)) {
09595258
LT
586 while (namelen && path[namelen-1] == '/')
587 namelen--;
588 }
11be42a4
JS
589 size = cache_entry_size(namelen);
590 ce = xcalloc(1, size);
591 memcpy(ce->name, path, namelen);
7a51ed66 592 ce->ce_flags = namelen;
39425819
JH
593 if (!intent_only)
594 fill_stat_cache_info(ce, st);
388b2acd
JH
595 else
596 ce->ce_flags |= CE_INTENT_TO_ADD;
11be42a4 597
78a8d641 598 if (trust_executable_bit && has_symlinks)
d177cab0 599 ce->ce_mode = create_ce_mode(st_mode);
185c975f 600 else {
78a8d641
JS
601 /* If there is an existing entry, pick the mode bits and type
602 * from it, otherwise assume unexecutable regular file.
11be42a4 603 */
185c975f 604 struct cache_entry *ent;
20314271 605 int pos = index_name_pos_also_unmerged(istate, path, namelen);
185c975f 606
4aab5b46 607 ent = (0 <= pos) ? istate->cache[pos] : NULL;
d177cab0 608 ce->ce_mode = ce_mode_from_stat(ent, st_mode);
11be42a4
JS
609 }
610
dc1ae704
JJ
611 /* When core.ignorecase=true, determine if a directory of the same name but differing
612 * case already exists within the Git repository. If it does, ensure the directory
613 * case of the file being added to the repository matches (is folded into) the existing
614 * entry's directory case.
615 */
616 if (ignore_case) {
617 const char *startPtr = ce->name;
618 const char *ptr = startPtr;
619 while (*ptr) {
620 while (*ptr && *ptr != '/')
621 ++ptr;
622 if (*ptr == '/') {
623 struct cache_entry *foundce;
624 ++ptr;
625 foundce = index_name_exists(&the_index, ce->name, ptr - ce->name, ignore_case);
626 if (foundce) {
627 memcpy((void *)startPtr, foundce->name + (startPtr - ce->name), ptr - startPtr);
628 startPtr = ptr;
629 }
630 }
631 }
632 }
633
6835550d 634 alias = index_name_exists(istate, ce->name, ce_namelen(ce), ignore_case);
d177cab0 635 if (alias && !ce_stage(alias) && !ie_match_stat(istate, alias, st, ce_option)) {
0781b8a9
JH
636 /* Nothing changed, really */
637 free(ce);
125fd984
JH
638 if (!S_ISGITLINK(alias->ce_mode))
639 ce_mark_uptodate(alias);
1102952b 640 alias->ce_flags |= CE_ADDED;
0781b8a9
JH
641 return 0;
642 }
39425819 643 if (!intent_only) {
c4ce46fc 644 if (index_path(ce->sha1, path, st, HASH_WRITE_OBJECT))
39425819
JH
645 return error("unable to index file %s", path);
646 } else
647 record_intent_to_add(ce);
648
1102952b
LT
649 if (ignore_case && alias && different_name(ce, alias))
650 ce = create_alias_ce(ce, alias);
651 ce->ce_flags |= CE_ADDED;
38ed1d89 652
3bf0dd1f 653 /* It was suspected to be racily clean, but it turns out to be Ok */
38ed1d89
JH
654 was_same = (alias &&
655 !ce_stage(alias) &&
656 !hashcmp(alias->sha1, ce->sha1) &&
657 ce->ce_mode == alias->ce_mode);
658
659 if (pretend)
660 ;
39425819 661 else if (add_index_entry(istate, ce, add_option))
960b8ad1 662 return error("unable to add %s to index",path);
38ed1d89 663 if (verbose && !was_same)
11be42a4 664 printf("add '%s'\n", path);
11be42a4
JS
665 return 0;
666}
667
38ed1d89 668int add_file_to_index(struct index_state *istate, const char *path, int flags)
d177cab0
LT
669{
670 struct stat st;
671 if (lstat(path, &st))
d824cbba 672 die_errno("unable to stat '%s'", path);
38ed1d89 673 return add_to_index(istate, path, &st, flags);
d177cab0
LT
674}
675
6640f881
CR
676struct cache_entry *make_cache_entry(unsigned int mode,
677 const unsigned char *sha1, const char *path, int stage,
678 int refresh)
679{
680 int size, len;
681 struct cache_entry *ce;
682
7e7abea9
DP
683 if (!verify_path(path)) {
684 error("Invalid path '%s'", path);
6640f881 685 return NULL;
7e7abea9 686 }
6640f881
CR
687
688 len = strlen(path);
689 size = cache_entry_size(len);
690 ce = xcalloc(1, size);
691
692 hashcpy(ce->sha1, sha1);
693 memcpy(ce->name, path, len);
694 ce->ce_flags = create_ce_flags(len, stage);
695 ce->ce_mode = create_ce_mode(mode);
696
697 if (refresh)
698 return refresh_cache_entry(ce, 0);
699
700 return ce;
701}
702
dbbce55b 703int ce_same_name(struct cache_entry *a, struct cache_entry *b)
7b937ca3
LT
704{
705 int len = ce_namelen(a);
706 return ce_namelen(b) == len && !memcmp(a->name, b->name, len);
707}
708
eb9cb55b 709int ce_path_match(const struct cache_entry *ce, const struct pathspec *pathspec)
c0fd1f51 710{
898bbd9f 711 return match_pathspec_depth(pathspec, ce->name, ce_namelen(ce), 0, NULL);
c0fd1f51
LT
712}
713
8dcf39c4
LT
714/*
715 * We fundamentally don't like some paths: we don't want
716 * dot or dot-dot anywhere, and for obvious reasons don't
717 * want to recurse into ".git" either.
718 *
719 * Also, we don't want double slashes or slashes at the
720 * end that can make pathnames ambiguous.
721 */
722static int verify_dotfile(const char *rest)
723{
724 /*
725 * The first character was '.', but that
726 * has already been discarded, we now test
727 * the rest.
728 */
e0f530ff 729
8dcf39c4 730 /* "." is not allowed */
e0f530ff 731 if (*rest == '\0' || is_dir_sep(*rest))
8dcf39c4
LT
732 return 0;
733
e0f530ff 734 switch (*rest) {
8dcf39c4
LT
735 /*
736 * ".git" followed by NUL or slash is bad. This
737 * shares the path end test with the ".." case.
738 */
739 case 'g':
740 if (rest[1] != 'i')
741 break;
742 if (rest[2] != 't')
743 break;
744 rest += 2;
745 /* fallthrough */
746 case '.':
e0f530ff 747 if (rest[1] == '\0' || is_dir_sep(rest[1]))
8dcf39c4
LT
748 return 0;
749 }
750 return 1;
751}
752
753int verify_path(const char *path)
754{
755 char c;
756
56948cb6
EFL
757 if (has_dos_drive_prefix(path))
758 return 0;
759
8dcf39c4
LT
760 goto inside;
761 for (;;) {
762 if (!c)
763 return 1;
56948cb6 764 if (is_dir_sep(c)) {
8dcf39c4
LT
765inside:
766 c = *path++;
3bdf09c7
JH
767 if ((c == '.' && !verify_dotfile(path)) ||
768 is_dir_sep(c) || c == '\0')
769 return 0;
8dcf39c4
LT
770 }
771 c = *path++;
772 }
773}
774
12676608
LT
775/*
776 * Do we have another file that has the beginning components being a
777 * proper superset of the name we're trying to add?
0f1e4f04 778 */
4aab5b46
JH
779static int has_file_name(struct index_state *istate,
780 const struct cache_entry *ce, int pos, int ok_to_replace)
0f1e4f04 781{
12676608
LT
782 int retval = 0;
783 int len = ce_namelen(ce);
b155725d 784 int stage = ce_stage(ce);
12676608 785 const char *name = ce->name;
0f1e4f04 786
4aab5b46
JH
787 while (pos < istate->cache_nr) {
788 struct cache_entry *p = istate->cache[pos++];
0f1e4f04 789
12676608 790 if (len >= ce_namelen(p))
0f1e4f04 791 break;
12676608
LT
792 if (memcmp(name, p->name, len))
793 break;
b155725d
JH
794 if (ce_stage(p) != stage)
795 continue;
12676608
LT
796 if (p->name[len] != '/')
797 continue;
7a51ed66 798 if (p->ce_flags & CE_REMOVE)
21cd8d00 799 continue;
12676608
LT
800 retval = -1;
801 if (!ok_to_replace)
802 break;
4aab5b46 803 remove_index_entry_at(istate, --pos);
0f1e4f04 804 }
12676608
LT
805 return retval;
806}
0f1e4f04 807
12676608
LT
808/*
809 * Do we have another file with a pathname that is a proper
810 * subset of the name we're trying to add?
811 */
4aab5b46
JH
812static int has_dir_name(struct index_state *istate,
813 const struct cache_entry *ce, int pos, int ok_to_replace)
12676608
LT
814{
815 int retval = 0;
b155725d 816 int stage = ce_stage(ce);
12676608
LT
817 const char *name = ce->name;
818 const char *slash = name + ce_namelen(ce);
0f1e4f04 819
12676608
LT
820 for (;;) {
821 int len;
0f1e4f04 822
12676608
LT
823 for (;;) {
824 if (*--slash == '/')
825 break;
826 if (slash <= ce->name)
827 return retval;
828 }
829 len = slash - name;
0f1e4f04 830
7a51ed66 831 pos = index_name_pos(istate, name, create_ce_flags(len, stage));
12676608 832 if (pos >= 0) {
21cd8d00
JH
833 /*
834 * Found one, but not so fast. This could
835 * be a marker that says "I was here, but
836 * I am being removed". Such an entry is
837 * not a part of the resulting tree, and
838 * it is Ok to have a directory at the same
839 * path.
840 */
077c48df 841 if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) {
21cd8d00
JH
842 retval = -1;
843 if (!ok_to_replace)
844 break;
4aab5b46 845 remove_index_entry_at(istate, pos);
21cd8d00
JH
846 continue;
847 }
12676608 848 }
21cd8d00
JH
849 else
850 pos = -pos-1;
12676608
LT
851
852 /*
853 * Trivial optimization: if we find an entry that
854 * already matches the sub-directory, then we know
b155725d 855 * we're ok, and we can exit.
12676608 856 */
4aab5b46
JH
857 while (pos < istate->cache_nr) {
858 struct cache_entry *p = istate->cache[pos];
b155725d
JH
859 if ((ce_namelen(p) <= len) ||
860 (p->name[len] != '/') ||
861 memcmp(p->name, name, len))
862 break; /* not our subdirectory */
077c48df
JH
863 if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE))
864 /*
865 * p is at the same stage as our entry, and
b155725d
JH
866 * is a subdirectory of what we are looking
867 * at, so we cannot have conflicts at our
868 * level or anything shorter.
869 */
870 return retval;
871 pos++;
192268c1 872 }
0f1e4f04 873 }
12676608
LT
874 return retval;
875}
876
877/* We may be in a situation where we already have path/file and path
878 * is being added, or we already have path and path/file is being
879 * added. Either one would result in a nonsense tree that has path
880 * twice when git-write-tree tries to write it out. Prevent it.
a6080a0a 881 *
12676608
LT
882 * If ok-to-replace is specified, we remove the conflicting entries
883 * from the cache so the caller should recompute the insert position.
884 * When this happens, we return non-zero.
885 */
4aab5b46
JH
886static int check_file_directory_conflict(struct index_state *istate,
887 const struct cache_entry *ce,
888 int pos, int ok_to_replace)
12676608 889{
21cd8d00
JH
890 int retval;
891
892 /*
893 * When ce is an "I am going away" entry, we allow it to be added
894 */
7a51ed66 895 if (ce->ce_flags & CE_REMOVE)
21cd8d00
JH
896 return 0;
897
12676608
LT
898 /*
899 * We check if the path is a sub-path of a subsequent pathname
900 * first, since removing those will not change the position
21cd8d00 901 * in the array.
12676608 902 */
4aab5b46 903 retval = has_file_name(istate, ce, pos, ok_to_replace);
21cd8d00 904
12676608
LT
905 /*
906 * Then check if the path might have a clashing sub-directory
907 * before it.
908 */
4aab5b46 909 return retval + has_dir_name(istate, ce, pos, ok_to_replace);
0f1e4f04
JH
910}
911
af3785dc 912static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option)
197ee8c9
LT
913{
914 int pos;
192268c1
JH
915 int ok_to_add = option & ADD_CACHE_OK_TO_ADD;
916 int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE;
b155725d 917 int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK;
39425819 918 int new_only = option & ADD_CACHE_NEW_ONLY;
5f73076c 919
09d5dc32 920 cache_tree_invalidate_path(istate->cache_tree, ce->name);
7a51ed66 921 pos = index_name_pos(istate, ce->name, ce->ce_flags);
197ee8c9 922
3e09cdfd 923 /* existing match? Just replace it. */
76e7f4ec 924 if (pos >= 0) {
39425819
JH
925 if (!new_only)
926 replace_index_entry(istate, pos, ce);
197ee8c9
LT
927 return 0;
928 }
76e7f4ec 929 pos = -pos-1;
197ee8c9 930
7b937ca3
LT
931 /*
932 * Inserting a merged entry ("stage 0") into the index
933 * will always replace all non-merged entries..
934 */
4aab5b46
JH
935 if (pos < istate->cache_nr && ce_stage(ce) == 0) {
936 while (ce_same_name(istate->cache[pos], ce)) {
7b937ca3 937 ok_to_add = 1;
4aab5b46 938 if (!remove_index_entry_at(istate, pos))
7b937ca3
LT
939 break;
940 }
941 }
942
121481ab
LT
943 if (!ok_to_add)
944 return -1;
8dcf39c4 945 if (!verify_path(ce->name))
7e7abea9 946 return error("Invalid path '%s'", ce->name);
121481ab 947
3e09cdfd 948 if (!skip_df_check &&
4aab5b46 949 check_file_directory_conflict(istate, ce, pos, ok_to_replace)) {
192268c1 950 if (!ok_to_replace)
4aab5b46
JH
951 return error("'%s' appears as both a file and as a directory",
952 ce->name);
7a51ed66 953 pos = index_name_pos(istate, ce->name, ce->ce_flags);
192268c1
JH
954 pos = -pos-1;
955 }
af3785dc
JH
956 return pos + 1;
957}
958
959int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option)
960{
961 int pos;
962
963 if (option & ADD_CACHE_JUST_APPEND)
964 pos = istate->cache_nr;
965 else {
966 int ret;
967 ret = add_index_entry_with_check(istate, ce, option);
968 if (ret <= 0)
969 return ret;
970 pos = ret - 1;
971 }
0f1e4f04 972
197ee8c9 973 /* Make sure the array is big enough .. */
4aab5b46
JH
974 if (istate->cache_nr == istate->cache_alloc) {
975 istate->cache_alloc = alloc_nr(istate->cache_alloc);
976 istate->cache = xrealloc(istate->cache,
977 istate->cache_alloc * sizeof(struct cache_entry *));
197ee8c9
LT
978 }
979
980 /* Add it in.. */
4aab5b46 981 istate->cache_nr++;
af3785dc 982 if (istate->cache_nr > pos + 1)
4aab5b46
JH
983 memmove(istate->cache + pos + 1,
984 istate->cache + pos,
985 (istate->cache_nr - pos - 1) * sizeof(ce));
cf558704 986 set_index_entry(istate, pos, ce);
4aab5b46 987 istate->cache_changed = 1;
197ee8c9
LT
988 return 0;
989}
990
405e5b2f
LT
991/*
992 * "refresh" does not calculate a new sha1 file or bring the
993 * cache up-to-date for mode/content changes. But what it
994 * _does_ do is to "re-match" the stat information of a file
995 * with the cache, so that you can refresh the cache for a
996 * file that hasn't been changed but where the stat entry is
997 * out of date.
998 *
999 * For example, you'd want to do this after doing a "git-read-tree",
1000 * to link up the stat cache details with the proper files.
1001 */
4aab5b46 1002static struct cache_entry *refresh_cache_ent(struct index_state *istate,
4bd5b7da 1003 struct cache_entry *ce,
d05e6970
JK
1004 unsigned int options, int *err,
1005 int *changed_ret)
405e5b2f
LT
1006{
1007 struct stat st;
1008 struct cache_entry *updated;
1009 int changed, size;
4bd5b7da 1010 int ignore_valid = options & CE_MATCH_IGNORE_VALID;
56cac48c 1011 int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE;
405e5b2f 1012
eadb5831
JH
1013 if (ce_uptodate(ce))
1014 return ce;
1015
aa9349d4 1016 /*
56cac48c
NTND
1017 * CE_VALID or CE_SKIP_WORKTREE means the user promised us
1018 * that the change to the work tree does not matter and told
1019 * us not to worry.
aa9349d4 1020 */
56cac48c
NTND
1021 if (!ignore_skip_worktree && ce_skip_worktree(ce)) {
1022 ce_mark_uptodate(ce);
1023 return ce;
1024 }
aa9349d4
MSO
1025 if (!ignore_valid && (ce->ce_flags & CE_VALID)) {
1026 ce_mark_uptodate(ce);
1027 return ce;
1028 }
1029
8fd2cb40 1030 if (lstat(ce->name, &st) < 0) {
ec0cc704
JH
1031 if (err)
1032 *err = errno;
8fd2cb40
JS
1033 return NULL;
1034 }
405e5b2f 1035
4bd5b7da 1036 changed = ie_match_stat(istate, ce, &st, options);
d05e6970
JK
1037 if (changed_ret)
1038 *changed_ret = changed;
405e5b2f 1039 if (!changed) {
4bd5b7da
JH
1040 /*
1041 * The path is unchanged. If we were told to ignore
1042 * valid bit, then we did the actual stat check and
1043 * found that the entry is unmodified. If the entry
1044 * is not marked VALID, this is the place to mark it
1045 * valid again, under "assume unchanged" mode.
1046 */
1047 if (ignore_valid && assume_unchanged &&
7a51ed66 1048 !(ce->ce_flags & CE_VALID))
405e5b2f 1049 ; /* mark this one VALID again */
eadb5831
JH
1050 else {
1051 /*
1052 * We do not mark the index itself "modified"
1053 * because CE_UPTODATE flag is in-core only;
1054 * we are not going to write this change out.
1055 */
125fd984
JH
1056 if (!S_ISGITLINK(ce->ce_mode))
1057 ce_mark_uptodate(ce);
8fd2cb40 1058 return ce;
eadb5831 1059 }
405e5b2f
LT
1060 }
1061
4bd5b7da 1062 if (ie_modified(istate, ce, &st, options)) {
ec0cc704
JH
1063 if (err)
1064 *err = EINVAL;
8fd2cb40
JS
1065 return NULL;
1066 }
405e5b2f
LT
1067
1068 size = ce_size(ce);
1069 updated = xmalloc(size);
1070 memcpy(updated, ce, size);
1071 fill_stat_cache_info(updated, &st);
4bd5b7da
JH
1072 /*
1073 * If ignore_valid is not set, we should leave CE_VALID bit
1074 * alone. Otherwise, paths marked with --no-assume-unchanged
1075 * (i.e. things to be edited) will reacquire CE_VALID bit
1076 * automatically, which is not really what we want.
405e5b2f 1077 */
4bd5b7da 1078 if (!ignore_valid && assume_unchanged &&
7a51ed66
LT
1079 !(ce->ce_flags & CE_VALID))
1080 updated->ce_flags &= ~CE_VALID;
405e5b2f
LT
1081
1082 return updated;
1083}
1084
3deffc52 1085static void show_file(const char * fmt, const char * name, int in_porcelain,
046613c5 1086 int * first, const char *header_msg)
3deffc52
MM
1087{
1088 if (in_porcelain && *first && header_msg) {
1089 printf("%s\n", header_msg);
cd2b8ae9 1090 *first = 0;
3deffc52
MM
1091 }
1092 printf(fmt, name);
1093}
1094
1095int refresh_index(struct index_state *istate, unsigned int flags, const char **pathspec,
046613c5 1096 char *seen, const char *header_msg)
405e5b2f
LT
1097{
1098 int i;
1099 int has_errors = 0;
1100 int really = (flags & REFRESH_REALLY) != 0;
1101 int allow_unmerged = (flags & REFRESH_UNMERGED) != 0;
1102 int quiet = (flags & REFRESH_QUIET) != 0;
1103 int not_new = (flags & REFRESH_IGNORE_MISSING) != 0;
5fdeacb0 1104 int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0;
3deffc52
MM
1105 int first = 1;
1106 int in_porcelain = (flags & REFRESH_IN_PORCELAIN);
4bd5b7da 1107 unsigned int options = really ? CE_MATCH_IGNORE_VALID : 0;
4bd4e730 1108 const char *modified_fmt;
73b7eae6
JK
1109 const char *deleted_fmt;
1110 const char *typechange_fmt;
1111 const char *added_fmt;
4bd4e730 1112 const char *unmerged_fmt;
405e5b2f 1113
4bd4e730 1114 modified_fmt = (in_porcelain ? "M\t%s\n" : "%s: needs update\n");
73b7eae6
JK
1115 deleted_fmt = (in_porcelain ? "D\t%s\n" : "%s: needs update\n");
1116 typechange_fmt = (in_porcelain ? "T\t%s\n" : "%s needs update\n");
1117 added_fmt = (in_porcelain ? "A\t%s\n" : "%s needs update\n");
4bd4e730 1118 unmerged_fmt = (in_porcelain ? "U\t%s\n" : "%s: needs merge\n");
4aab5b46 1119 for (i = 0; i < istate->cache_nr; i++) {
405e5b2f 1120 struct cache_entry *ce, *new;
ec0cc704 1121 int cache_errno = 0;
73b7eae6 1122 int changed = 0;
3d1f148c 1123 int filtered = 0;
ec0cc704 1124
4aab5b46 1125 ce = istate->cache[i];
5fdeacb0
JS
1126 if (ignore_submodules && S_ISGITLINK(ce->ce_mode))
1127 continue;
1128
3d1f148c
JH
1129 if (pathspec &&
1130 !match_pathspec(pathspec, ce->name, strlen(ce->name), 0, seen))
1131 filtered = 1;
1132
405e5b2f 1133 if (ce_stage(ce)) {
4aab5b46
JH
1134 while ((i < istate->cache_nr) &&
1135 ! strcmp(istate->cache[i]->name, ce->name))
405e5b2f
LT
1136 i++;
1137 i--;
1138 if (allow_unmerged)
1139 continue;
3d1f148c
JH
1140 if (!filtered)
1141 show_file(unmerged_fmt, ce->name, in_porcelain,
1142 &first, header_msg);
405e5b2f
LT
1143 has_errors = 1;
1144 continue;
1145 }
1146
3d1f148c 1147 if (filtered)
d616813d
AJ
1148 continue;
1149
73b7eae6 1150 new = refresh_cache_ent(istate, ce, options, &cache_errno, &changed);
8fd2cb40 1151 if (new == ce)
405e5b2f 1152 continue;
8fd2cb40 1153 if (!new) {
73b7eae6
JK
1154 const char *fmt;
1155
8fd2cb40 1156 if (not_new && cache_errno == ENOENT)
405e5b2f 1157 continue;
8fd2cb40 1158 if (really && cache_errno == EINVAL) {
405e5b2f
LT
1159 /* If we are doing --really-refresh that
1160 * means the index is not valid anymore.
1161 */
7a51ed66 1162 ce->ce_flags &= ~CE_VALID;
4aab5b46 1163 istate->cache_changed = 1;
405e5b2f
LT
1164 }
1165 if (quiet)
1166 continue;
73b7eae6
JK
1167
1168 if (cache_errno == ENOENT)
1169 fmt = deleted_fmt;
1170 else if (ce->ce_flags & CE_INTENT_TO_ADD)
1171 fmt = added_fmt; /* must be before other checks */
1172 else if (changed & TYPE_CHANGED)
1173 fmt = typechange_fmt;
1174 else
1175 fmt = modified_fmt;
1176 show_file(fmt,
1177 ce->name, in_porcelain, &first, header_msg);
405e5b2f
LT
1178 has_errors = 1;
1179 continue;
1180 }
cf558704
LT
1181
1182 replace_index_entry(istate, i, new);
405e5b2f
LT
1183 }
1184 return has_errors;
1185}
1186
87b29e5a 1187static struct cache_entry *refresh_cache_entry(struct cache_entry *ce, int really)
ec0cc704 1188{
d05e6970 1189 return refresh_cache_ent(&the_index, ce, really, NULL, NULL);
ec0cc704
JH
1190}
1191
bad68ec9 1192static int verify_hdr(struct cache_header *hdr, unsigned long size)
e83c5163 1193{
9126f009 1194 git_SHA_CTX c;
bad68ec9 1195 unsigned char sha1[20];
e83c5163 1196
ccc4feb5 1197 if (hdr->hdr_signature != htonl(CACHE_SIGNATURE))
e83c5163 1198 return error("bad signature");
06aaaa0b 1199 if (hdr->hdr_version != htonl(2) && hdr->hdr_version != htonl(3))
ca9be054 1200 return error("bad index version");
9126f009
NP
1201 git_SHA1_Init(&c);
1202 git_SHA1_Update(&c, hdr, size - 20);
1203 git_SHA1_Final(sha1, &c);
a89fccd2 1204 if (hashcmp(sha1, (unsigned char *)hdr + size - 20))
ca9be054 1205 return error("bad index file sha1 signature");
e83c5163
LT
1206 return 0;
1207}
1208
4aab5b46
JH
1209static int read_index_extension(struct index_state *istate,
1210 const char *ext, void *data, unsigned long sz)
bad68ec9
JH
1211{
1212 switch (CACHE_EXT(ext)) {
1213 case CACHE_EXT_TREE:
4aab5b46 1214 istate->cache_tree = cache_tree_read(data, sz);
bad68ec9 1215 break;
cfc5789a
JH
1216 case CACHE_EXT_RESOLVE_UNDO:
1217 istate->resolve_undo = resolve_undo_read(data, sz);
1218 break;
bad68ec9
JH
1219 default:
1220 if (*ext < 'A' || 'Z' < *ext)
1221 return error("index uses %.4s extension, which we do not understand",
1222 ext);
1223 fprintf(stderr, "ignoring %.4s extension\n", ext);
1224 break;
1225 }
1226 return 0;
1227}
1228
4aab5b46 1229int read_index(struct index_state *istate)
8fd2cb40 1230{
4aab5b46 1231 return read_index_from(istate, get_index_file());
8fd2cb40
JS
1232}
1233
debed2a6 1234static struct cache_entry *create_from_disk(struct ondisk_cache_entry *ondisk)
7a51ed66 1235{
debed2a6 1236 struct cache_entry *ce;
7fec10b7 1237 size_t len;
06aaaa0b 1238 const char *name;
debed2a6 1239 unsigned int flags;
7fec10b7 1240
7a51ed66 1241 /* On-disk flags are just 16 bits */
debed2a6
RS
1242 flags = ntohs(ondisk->flags);
1243 len = flags & CE_NAMEMASK;
7fec10b7 1244
debed2a6 1245 if (flags & CE_EXTENDED) {
06aaaa0b
NTND
1246 struct ondisk_cache_entry_extended *ondisk2;
1247 int extended_flags;
1248 ondisk2 = (struct ondisk_cache_entry_extended *)ondisk;
1249 extended_flags = ntohs(ondisk2->flags2) << 16;
1250 /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */
1251 if (extended_flags & ~CE_EXTENDED_FLAGS)
1252 die("Unknown index entry format %08x", extended_flags);
debed2a6 1253 flags |= extended_flags;
06aaaa0b
NTND
1254 name = ondisk2->name;
1255 }
1256 else
1257 name = ondisk->name;
1258
7fec10b7 1259 if (len == CE_NAMEMASK)
06aaaa0b 1260 len = strlen(name);
7a51ed66 1261
debed2a6 1262 ce = xmalloc(cache_entry_size(len));
cf558704 1263
debed2a6
RS
1264 ce->ce_ctime.sec = ntohl(ondisk->ctime.sec);
1265 ce->ce_mtime.sec = ntohl(ondisk->mtime.sec);
1266 ce->ce_ctime.nsec = ntohl(ondisk->ctime.nsec);
1267 ce->ce_mtime.nsec = ntohl(ondisk->mtime.nsec);
1268 ce->ce_dev = ntohl(ondisk->dev);
1269 ce->ce_ino = ntohl(ondisk->ino);
1270 ce->ce_mode = ntohl(ondisk->mode);
1271 ce->ce_uid = ntohl(ondisk->uid);
1272 ce->ce_gid = ntohl(ondisk->gid);
1273 ce->ce_size = ntohl(ondisk->size);
1274 ce->ce_flags = flags;
cf558704 1275
debed2a6 1276 hashcpy(ce->sha1, ondisk->sha1);
cf558704 1277
debed2a6
RS
1278 memcpy(ce->name, name, len);
1279 ce->name[len] = '\0';
1280 return ce;
cf558704
LT
1281}
1282
8fd2cb40 1283/* remember to discard_cache() before reading a different cache! */
4aab5b46 1284int read_index_from(struct index_state *istate, const char *path)
e83c5163
LT
1285{
1286 int fd, i;
1287 struct stat st;
debed2a6 1288 unsigned long src_offset;
e83c5163 1289 struct cache_header *hdr;
7a51ed66
LT
1290 void *mmap;
1291 size_t mmap_size;
e83c5163
LT
1292
1293 errno = EBUSY;
913e0e99 1294 if (istate->initialized)
4aab5b46 1295 return istate->cache_nr;
5d1a5c02 1296
e83c5163 1297 errno = ENOENT;
fba2f38a
KB
1298 istate->timestamp.sec = 0;
1299 istate->timestamp.nsec = 0;
8fd2cb40 1300 fd = open(path, O_RDONLY);
5d1a5c02
LT
1301 if (fd < 0) {
1302 if (errno == ENOENT)
1303 return 0;
d824cbba 1304 die_errno("index file open failed");
5d1a5c02 1305 }
e83c5163 1306
3511a377 1307 if (fstat(fd, &st))
d824cbba 1308 die_errno("cannot stat the open index");
3511a377
LFC
1309
1310 errno = EINVAL;
7a51ed66
LT
1311 mmap_size = xsize_t(st.st_size);
1312 if (mmap_size < sizeof(struct cache_header) + 20)
3511a377
LFC
1313 die("index file smaller than expected");
1314
7a51ed66 1315 mmap = xmmap(NULL, mmap_size, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0);
e83c5163 1316 close(fd);
7a51ed66 1317 if (mmap == MAP_FAILED)
0721c314 1318 die_errno("unable to map index file");
e83c5163 1319
7a51ed66
LT
1320 hdr = mmap;
1321 if (verify_hdr(hdr, mmap_size) < 0)
e83c5163
LT
1322 goto unmap;
1323
4aab5b46
JH
1324 istate->cache_nr = ntohl(hdr->hdr_entries);
1325 istate->cache_alloc = alloc_nr(istate->cache_nr);
1326 istate->cache = xcalloc(istate->cache_alloc, sizeof(struct cache_entry *));
913e0e99 1327 istate->initialized = 1;
7a51ed66
LT
1328
1329 src_offset = sizeof(*hdr);
4aab5b46 1330 for (i = 0; i < istate->cache_nr; i++) {
7a51ed66 1331 struct ondisk_cache_entry *disk_ce;
4aab5b46
JH
1332 struct cache_entry *ce;
1333
7a51ed66 1334 disk_ce = (struct ondisk_cache_entry *)((char *)mmap + src_offset);
debed2a6 1335 ce = create_from_disk(disk_ce);
cf558704 1336 set_index_entry(istate, i, ce);
7a51ed66
LT
1337
1338 src_offset += ondisk_ce_size(ce);
e83c5163 1339 }
fba2f38a 1340 istate->timestamp.sec = st.st_mtime;
c06ff490 1341 istate->timestamp.nsec = ST_MTIME_NSEC(st);
fba2f38a 1342
7a51ed66 1343 while (src_offset <= mmap_size - 20 - 8) {
bad68ec9
JH
1344 /* After an array of active_nr index entries,
1345 * there can be arbitrary number of extended
1346 * sections, each of which is prefixed with
1347 * extension name (4-byte) and section length
1348 * in 4-byte network byte order.
1349 */
07cc8eca 1350 uint32_t extsize;
7a51ed66 1351 memcpy(&extsize, (char *)mmap + src_offset + 4, 4);
bad68ec9 1352 extsize = ntohl(extsize);
4aab5b46 1353 if (read_index_extension(istate,
7a51ed66
LT
1354 (const char *) mmap + src_offset,
1355 (char *) mmap + src_offset + 8,
1d7f171c 1356 extsize) < 0)
bad68ec9 1357 goto unmap;
7a51ed66
LT
1358 src_offset += 8;
1359 src_offset += extsize;
bad68ec9 1360 }
7a51ed66 1361 munmap(mmap, mmap_size);
4aab5b46 1362 return istate->cache_nr;
e83c5163
LT
1363
1364unmap:
7a51ed66 1365 munmap(mmap, mmap_size);
e83c5163 1366 errno = EINVAL;
5d1a5c02 1367 die("index file corrupt");
e83c5163
LT
1368}
1369
fa7b3c2f
JH
1370int is_index_unborn(struct index_state *istate)
1371{
debed2a6 1372 return (!istate->cache_nr && !istate->timestamp.sec);
fa7b3c2f
JH
1373}
1374
4aab5b46 1375int discard_index(struct index_state *istate)
6d297f81 1376{
debed2a6
RS
1377 int i;
1378
1379 for (i = 0; i < istate->cache_nr; i++)
1380 free(istate->cache[i]);
cfc5789a 1381 resolve_undo_clear_index(istate);
4aab5b46
JH
1382 istate->cache_nr = 0;
1383 istate->cache_changed = 0;
fba2f38a
KB
1384 istate->timestamp.sec = 0;
1385 istate->timestamp.nsec = 0;
64ca23af 1386 istate->name_hash_initialized = 0;
cf558704 1387 free_hash(&istate->name_hash);
4aab5b46 1388 cache_tree_free(&(istate->cache_tree));
913e0e99 1389 istate->initialized = 0;
6d297f81
JS
1390
1391 /* no need to throw away allocated active_cache */
7a51ed66 1392 return 0;
6d297f81
JS
1393}
1394
d1f128b0 1395int unmerged_index(const struct index_state *istate)
94a5728c
DB
1396{
1397 int i;
1398 for (i = 0; i < istate->cache_nr; i++) {
1399 if (ce_stage(istate->cache[i]))
1400 return 1;
1401 }
1402 return 0;
1403}
1404
4990aadc 1405#define WRITE_BUFFER_SIZE 8192
bf0f910d 1406static unsigned char write_buffer[WRITE_BUFFER_SIZE];
4990aadc
LT
1407static unsigned long write_buffer_len;
1408
9126f009 1409static int ce_write_flush(git_SHA_CTX *context, int fd)
6015c28b
JH
1410{
1411 unsigned int buffered = write_buffer_len;
1412 if (buffered) {
9126f009 1413 git_SHA1_Update(context, write_buffer, buffered);
93822c22 1414 if (write_in_full(fd, write_buffer, buffered) != buffered)
6015c28b
JH
1415 return -1;
1416 write_buffer_len = 0;
1417 }
1418 return 0;
1419}
1420
9126f009 1421static int ce_write(git_SHA_CTX *context, int fd, void *data, unsigned int len)
4990aadc
LT
1422{
1423 while (len) {
1424 unsigned int buffered = write_buffer_len;
1425 unsigned int partial = WRITE_BUFFER_SIZE - buffered;
1426 if (partial > len)
1427 partial = len;
1428 memcpy(write_buffer + buffered, data, partial);
1429 buffered += partial;
1430 if (buffered == WRITE_BUFFER_SIZE) {
6015c28b
JH
1431 write_buffer_len = buffered;
1432 if (ce_write_flush(context, fd))
4990aadc
LT
1433 return -1;
1434 buffered = 0;
1435 }
1436 write_buffer_len = buffered;
1437 len -= partial;
1d7f171c 1438 data = (char *) data + partial;
a6080a0a
JH
1439 }
1440 return 0;
4990aadc
LT
1441}
1442
9126f009 1443static int write_index_ext_header(git_SHA_CTX *context, int fd,
ac58c7b1 1444 unsigned int ext, unsigned int sz)
bad68ec9
JH
1445{
1446 ext = htonl(ext);
1447 sz = htonl(sz);
968a1d65
DR
1448 return ((ce_write(context, fd, &ext, 4) < 0) ||
1449 (ce_write(context, fd, &sz, 4) < 0)) ? -1 : 0;
bad68ec9
JH
1450}
1451
9126f009 1452static int ce_flush(git_SHA_CTX *context, int fd)
4990aadc
LT
1453{
1454 unsigned int left = write_buffer_len;
ca9be054 1455
4990aadc
LT
1456 if (left) {
1457 write_buffer_len = 0;
9126f009 1458 git_SHA1_Update(context, write_buffer, left);
4990aadc 1459 }
ca9be054 1460
2c865d9a
QH
1461 /* Flush first if not enough space for SHA1 signature */
1462 if (left + 20 > WRITE_BUFFER_SIZE) {
93822c22 1463 if (write_in_full(fd, write_buffer, left) != left)
2c865d9a
QH
1464 return -1;
1465 left = 0;
1466 }
1467
ca9be054 1468 /* Append the SHA1 signature at the end */
9126f009 1469 git_SHA1_Final(write_buffer + left, context);
ca9be054 1470 left += 20;
93822c22 1471 return (write_in_full(fd, write_buffer, left) != left) ? -1 : 0;
4990aadc
LT
1472}
1473
407c8eb0
JH
1474static void ce_smudge_racily_clean_entry(struct cache_entry *ce)
1475{
1476 /*
1477 * The only thing we care about in this function is to smudge the
1478 * falsely clean entry due to touch-update-touch race, so we leave
1479 * everything else as they are. We are called for entries whose
1480 * ce_mtime match the index file mtime.
c70115b4
JH
1481 *
1482 * Note that this actually does not do much for gitlinks, for
1483 * which ce_match_stat_basic() always goes to the actual
1484 * contents. The caller checks with is_racy_timestamp() which
1485 * always says "no" for gitlinks, so we are not called for them ;-)
407c8eb0
JH
1486 */
1487 struct stat st;
1488
1489 if (lstat(ce->name, &st) < 0)
1490 return;
1491 if (ce_match_stat_basic(ce, &st))
1492 return;
1493 if (ce_modified_check_fs(ce, &st)) {
4b3511b0
JH
1494 /* This is "racily clean"; smudge it. Note that this
1495 * is a tricky code. At first glance, it may appear
1496 * that it can break with this sequence:
1497 *
1498 * $ echo xyzzy >frotz
1499 * $ git-update-index --add frotz
1500 * $ : >frotz
1501 * $ sleep 3
1502 * $ echo filfre >nitfol
1503 * $ git-update-index --add nitfol
1504 *
b7e58b17 1505 * but it does not. When the second update-index runs,
4b3511b0
JH
1506 * it notices that the entry "frotz" has the same timestamp
1507 * as index, and if we were to smudge it by resetting its
1508 * size to zero here, then the object name recorded
1509 * in index is the 6-byte file but the cached stat information
1510 * becomes zero --- which would then match what we would
a6080a0a 1511 * obtain from the filesystem next time we stat("frotz").
4b3511b0
JH
1512 *
1513 * However, the second update-index, before calling
1514 * this function, notices that the cached size is 6
1515 * bytes and what is on the filesystem is an empty
1516 * file, and never calls us, so the cached size information
1517 * for "frotz" stays 6 which does not match the filesystem.
1518 */
7a51ed66 1519 ce->ce_size = 0;
407c8eb0
JH
1520 }
1521}
1522
9126f009 1523static int ce_write_entry(git_SHA_CTX *c, int fd, struct cache_entry *ce)
7a51ed66
LT
1524{
1525 int size = ondisk_ce_size(ce);
1526 struct ondisk_cache_entry *ondisk = xcalloc(1, size);
06aaaa0b 1527 char *name;
59efba64 1528 int result;
7a51ed66 1529
fba2f38a
KB
1530 ondisk->ctime.sec = htonl(ce->ce_ctime.sec);
1531 ondisk->mtime.sec = htonl(ce->ce_mtime.sec);
fba2f38a
KB
1532 ondisk->ctime.nsec = htonl(ce->ce_ctime.nsec);
1533 ondisk->mtime.nsec = htonl(ce->ce_mtime.nsec);
7a51ed66
LT
1534 ondisk->dev = htonl(ce->ce_dev);
1535 ondisk->ino = htonl(ce->ce_ino);
1536 ondisk->mode = htonl(ce->ce_mode);
1537 ondisk->uid = htonl(ce->ce_uid);
1538 ondisk->gid = htonl(ce->ce_gid);
1539 ondisk->size = htonl(ce->ce_size);
1540 hashcpy(ondisk->sha1, ce->sha1);
1541 ondisk->flags = htons(ce->ce_flags);
06aaaa0b
NTND
1542 if (ce->ce_flags & CE_EXTENDED) {
1543 struct ondisk_cache_entry_extended *ondisk2;
1544 ondisk2 = (struct ondisk_cache_entry_extended *)ondisk;
1545 ondisk2->flags2 = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16);
1546 name = ondisk2->name;
1547 }
1548 else
1549 name = ondisk->name;
1550 memcpy(name, ce->name, ce_namelen(ce));
7a51ed66 1551
59efba64
JN
1552 result = ce_write(c, fd, ondisk, size);
1553 free(ondisk);
1554 return result;
7a51ed66
LT
1555}
1556
483fbe2b
JH
1557static int has_racy_timestamp(struct index_state *istate)
1558{
1559 int entries = istate->cache_nr;
1560 int i;
1561
1562 for (i = 0; i < entries; i++) {
1563 struct cache_entry *ce = istate->cache[i];
1564 if (is_racy_timestamp(istate, ce))
1565 return 1;
1566 }
1567 return 0;
1568}
1569
ccdc4ec3
JH
1570/*
1571 * Opportunisticly update the index but do not complain if we can't
1572 */
1573void update_index_if_able(struct index_state *istate, struct lock_file *lockfile)
1574{
483fbe2b 1575 if ((istate->cache_changed || has_racy_timestamp(istate)) &&
ccdc4ec3
JH
1576 !write_index(istate, lockfile->fd))
1577 commit_locked_index(lockfile);
1578 else
1579 rollback_lock_file(lockfile);
1580}
1581
e1afca4f 1582int write_index(struct index_state *istate, int newfd)
197ee8c9 1583{
9126f009 1584 git_SHA_CTX c;
197ee8c9 1585 struct cache_header hdr;
06aaaa0b 1586 int i, err, removed, extended;
4aab5b46
JH
1587 struct cache_entry **cache = istate->cache;
1588 int entries = istate->cache_nr;
e1afca4f 1589 struct stat st;
025a0709 1590
06aaaa0b 1591 for (i = removed = extended = 0; i < entries; i++) {
7a51ed66 1592 if (cache[i]->ce_flags & CE_REMOVE)
025a0709 1593 removed++;
197ee8c9 1594
06aaaa0b
NTND
1595 /* reduce extended entries if possible */
1596 cache[i]->ce_flags &= ~CE_EXTENDED;
1597 if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) {
1598 extended++;
1599 cache[i]->ce_flags |= CE_EXTENDED;
1600 }
1601 }
1602
ccc4feb5 1603 hdr.hdr_signature = htonl(CACHE_SIGNATURE);
06aaaa0b
NTND
1604 /* for extended format, increase version so older git won't try to read it */
1605 hdr.hdr_version = htonl(extended ? 3 : 2);
025a0709 1606 hdr.hdr_entries = htonl(entries - removed);
197ee8c9 1607
9126f009 1608 git_SHA1_Init(&c);
ca9be054 1609 if (ce_write(&c, newfd, &hdr, sizeof(hdr)) < 0)
197ee8c9
LT
1610 return -1;
1611
1612 for (i = 0; i < entries; i++) {
1613 struct cache_entry *ce = cache[i];
7a51ed66 1614 if (ce->ce_flags & CE_REMOVE)
aa16021e 1615 continue;
e06c43c7 1616 if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce))
407c8eb0 1617 ce_smudge_racily_clean_entry(ce);
7a51ed66 1618 if (ce_write_entry(&c, newfd, ce) < 0)
197ee8c9
LT
1619 return -1;
1620 }
1af1c2b6 1621
bad68ec9 1622 /* Write extension data here */
4aab5b46 1623 if (istate->cache_tree) {
f285a2d7 1624 struct strbuf sb = STRBUF_INIT;
1dffb8fa 1625
1dffb8fa
PH
1626 cache_tree_write(&sb, istate->cache_tree);
1627 err = write_index_ext_header(&c, newfd, CACHE_EXT_TREE, sb.len) < 0
1628 || ce_write(&c, newfd, sb.buf, sb.len) < 0;
1629 strbuf_release(&sb);
1630 if (err)
bad68ec9 1631 return -1;
bad68ec9 1632 }
cfc5789a
JH
1633 if (istate->resolve_undo) {
1634 struct strbuf sb = STRBUF_INIT;
1635
1636 resolve_undo_write(&sb, istate->resolve_undo);
1637 err = write_index_ext_header(&c, newfd, CACHE_EXT_RESOLVE_UNDO,
1638 sb.len) < 0
1639 || ce_write(&c, newfd, sb.buf, sb.len) < 0;
1640 strbuf_release(&sb);
1641 if (err)
1642 return -1;
1643 }
e1afca4f
KB
1644
1645 if (ce_flush(&c, newfd) || fstat(newfd, &st))
1646 return -1;
5bcf109c
KB
1647 istate->timestamp.sec = (unsigned int)st.st_mtime;
1648 istate->timestamp.nsec = ST_MTIME_NSEC(st);
e1afca4f 1649 return 0;
197ee8c9 1650}
e46bbcf6
MV
1651
1652/*
1653 * Read the index file that is potentially unmerged into given
63e8dc5b 1654 * index_state, dropping any unmerged entries. Returns true if
e46bbcf6
MV
1655 * the index is unmerged. Callers who want to refuse to work
1656 * from an unmerged state can call this and check its return value,
1657 * instead of calling read_cache().
1658 */
1659int read_index_unmerged(struct index_state *istate)
1660{
1661 int i;
d1a43f2a 1662 int unmerged = 0;
e46bbcf6
MV
1663
1664 read_index(istate);
e46bbcf6
MV
1665 for (i = 0; i < istate->cache_nr; i++) {
1666 struct cache_entry *ce = istate->cache[i];
d1a43f2a
JH
1667 struct cache_entry *new_ce;
1668 int size, len;
1669
1670 if (!ce_stage(ce))
e46bbcf6 1671 continue;
d1a43f2a
JH
1672 unmerged = 1;
1673 len = strlen(ce->name);
1674 size = cache_entry_size(len);
1675 new_ce = xcalloc(1, size);
d1a43f2a 1676 memcpy(new_ce->name, ce->name, len);
e11d7b59 1677 new_ce->ce_flags = create_ce_flags(len, 0) | CE_CONFLICTED;
d1a43f2a
JH
1678 new_ce->ce_mode = ce->ce_mode;
1679 if (add_index_entry(istate, new_ce, 0))
1680 return error("%s: cannot drop to stage #0",
1681 ce->name);
1682 i = index_name_pos(istate, new_ce->name, len);
e46bbcf6 1683 }
d1a43f2a 1684 return unmerged;
e46bbcf6 1685}
041aee31 1686
98fa4738
JK
1687/*
1688 * Returns 1 if the path is an "other" path with respect to
1689 * the index; that is, the path is not mentioned in the index at all,
1690 * either as a file, a directory with some files in the index,
1691 * or as an unmerged entry.
1692 *
1693 * We helpfully remove a trailing "/" from directories so that
1694 * the output of read_directory can be used as-is.
1695 */
1696int index_name_is_other(const struct index_state *istate, const char *name,
1697 int namelen)
1698{
1699 int pos;
1700 if (namelen && name[namelen - 1] == '/')
1701 namelen--;
1702 pos = index_name_pos(istate, name, namelen);
1703 if (0 <= pos)
1704 return 0; /* exact match */
1705 pos = -pos - 1;
1706 if (pos < istate->cache_nr) {
1707 struct cache_entry *ce = istate->cache[pos];
1708 if (ce_namelen(ce) == namelen &&
1709 !memcmp(ce->name, name, namelen))
1710 return 0; /* Yup, this one exists unmerged */
1711 }
1712 return 1;
1713}