Merge branch 'cc/svn-fe-py-shebang'
[git/git.git] / sha1-file.c
CommitLineData
0fcfd160
LT
1/*
2 * GIT - The information manager from hell
3 *
4 * Copyright (C) Linus Torvalds, 2005
5 *
6 * This handles basic git sha1 object files - packing, unpacking,
7 * creation etc.
8 */
0fcfd160 9#include "cache.h"
b2141fc1 10#include "config.h"
6eac50d8 11#include "string-list.h"
697cc8ef 12#include "lockfile.h"
1f688557 13#include "delta.h"
a733cb60 14#include "pack.h"
8e440259
PE
15#include "blob.h"
16#include "commit.h"
4dd1fbc7 17#include "run-command.h"
8e440259
PE
18#include "tag.h"
19#include "tree.h"
c879daa2 20#include "tree-walk.h"
f35a6d3b 21#include "refs.h"
70f5d5d3 22#include "pack-revindex.h"
628522ec 23#include "sha1-lookup.h"
568508e7 24#include "bulk-checkin.h"
031dc927 25#include "repository.h"
47f351e9 26#include "replace-object.h"
090ea126 27#include "streaming.h"
543c5caa 28#include "dir.h"
12d95ef6 29#include "list.h"
c4c6effa 30#include "mergesort.h"
cf3c6352 31#include "quote.h"
4f39cd82 32#include "packfile.h"
90c62155 33#include "object-store.h"
b14ed5ad 34#include "promisor-remote.h"
e05db0fd 35
1af64f73 36/* The maximum size for an object header. */
37#define MAX_HEADER_LEN 32
38
e1ccd7e2 39
40#define EMPTY_TREE_SHA1_BIN_LITERAL \
41 "\x4b\x82\x5d\xc6\x42\xcb\x6e\xb9\xa0\x60" \
42 "\xe5\x4b\xf8\xd6\x92\x88\xfb\xee\x49\x04"
13eeedb5 43#define EMPTY_TREE_SHA256_BIN_LITERAL \
44 "\x6e\xf1\x9b\x41\x22\x5c\x53\x69\xf1\xc1" \
45 "\x04\xd4\x5d\x8d\x85\xef\xa9\xb0\x57\xb5" \
46 "\x3b\x14\xb4\xb9\xb9\x39\xdd\x74\xde\xcc" \
47 "\x53\x21"
e1ccd7e2 48
49#define EMPTY_BLOB_SHA1_BIN_LITERAL \
50 "\xe6\x9d\xe2\x9b\xb2\xd1\xd6\x43\x4b\x8b" \
51 "\x29\xae\x77\x5a\xd8\xc2\xe4\x8c\x53\x91"
13eeedb5 52#define EMPTY_BLOB_SHA256_BIN_LITERAL \
53 "\x47\x3a\x0f\x4c\x3b\xe8\xa9\x36\x81\xa2" \
54 "\x67\xe3\xb1\xe9\xa7\xdc\xda\x11\x85\x43" \
55 "\x6f\xe1\x41\xf7\x74\x91\x20\xa3\x03\x72" \
56 "\x18\x13"
e1ccd7e2 57
50c5cd58 58const unsigned char null_sha1[GIT_MAX_RAWSZ];
3e56e724 59const struct object_id null_oid;
e1ccd7e2 60static const struct object_id empty_tree_oid = {
8576fde6
JK
61 EMPTY_TREE_SHA1_BIN_LITERAL
62};
e1ccd7e2 63static const struct object_id empty_blob_oid = {
8576fde6
JK
64 EMPTY_BLOB_SHA1_BIN_LITERAL
65};
13eeedb5 66static const struct object_id empty_tree_oid_sha256 = {
67 EMPTY_TREE_SHA256_BIN_LITERAL
68};
69static const struct object_id empty_blob_oid_sha256 = {
70 EMPTY_BLOB_SHA256_BIN_LITERAL
71};
88cd621d 72
ac73cedf 73static void git_hash_sha1_init(git_hash_ctx *ctx)
f50e766b 74{
ac73cedf 75 git_SHA1_Init(&ctx->sha1);
f50e766b 76}
77
ac73cedf 78static void git_hash_sha1_update(git_hash_ctx *ctx, const void *data, size_t len)
f50e766b 79{
ac73cedf 80 git_SHA1_Update(&ctx->sha1, data, len);
f50e766b 81}
82
ac73cedf 83static void git_hash_sha1_final(unsigned char *hash, git_hash_ctx *ctx)
f50e766b 84{
ac73cedf 85 git_SHA1_Final(hash, &ctx->sha1);
f50e766b 86}
87
13eeedb5 88
89static void git_hash_sha256_init(git_hash_ctx *ctx)
90{
91 git_SHA256_Init(&ctx->sha256);
92}
93
94static void git_hash_sha256_update(git_hash_ctx *ctx, const void *data, size_t len)
95{
96 git_SHA256_Update(&ctx->sha256, data, len);
97}
98
99static void git_hash_sha256_final(unsigned char *hash, git_hash_ctx *ctx)
100{
101 git_SHA256_Final(hash, &ctx->sha256);
102}
103
ac73cedf 104static void git_hash_unknown_init(git_hash_ctx *ctx)
f50e766b 105{
1a07e59c 106 BUG("trying to init unknown hash");
f50e766b 107}
108
ac73cedf 109static void git_hash_unknown_update(git_hash_ctx *ctx, const void *data, size_t len)
f50e766b 110{
1a07e59c 111 BUG("trying to update unknown hash");
f50e766b 112}
113
ac73cedf 114static void git_hash_unknown_final(unsigned char *hash, git_hash_ctx *ctx)
f50e766b 115{
1a07e59c 116 BUG("trying to finalize unknown hash");
f50e766b 117}
118
119const struct git_hash_algo hash_algos[GIT_HASH_NALGOS] = {
120 {
121 NULL,
122 0x00000000,
123 0,
124 0,
a2ce0a75 125 0,
f50e766b 126 git_hash_unknown_init,
127 git_hash_unknown_update,
128 git_hash_unknown_final,
129 NULL,
130 NULL,
131 },
132 {
1ccf07cb 133 "sha1",
f50e766b 134 /* "sha1", big-endian */
135 0x73686131,
f50e766b 136 GIT_SHA1_RAWSZ,
137 GIT_SHA1_HEXSZ,
a2ce0a75 138 GIT_SHA1_BLKSZ,
f50e766b 139 git_hash_sha1_init,
140 git_hash_sha1_update,
141 git_hash_sha1_final,
142 &empty_tree_oid,
143 &empty_blob_oid,
144 },
13eeedb5 145 {
146 "sha256",
147 /* "s256", big-endian */
148 0x73323536,
149 GIT_SHA256_RAWSZ,
150 GIT_SHA256_HEXSZ,
151 GIT_SHA256_BLKSZ,
152 git_hash_sha256_init,
153 git_hash_sha256_update,
154 git_hash_sha256_final,
155 &empty_tree_oid_sha256,
156 &empty_blob_oid_sha256,
157 }
f50e766b 158};
159
d8a92ced 160const char *empty_tree_oid_hex(void)
161{
162 static char buf[GIT_MAX_HEXSZ + 1];
163 return oid_to_hex_r(buf, the_hash_algo->empty_tree);
164}
165
166const char *empty_blob_oid_hex(void)
167{
168 static char buf[GIT_MAX_HEXSZ + 1];
169 return oid_to_hex_r(buf, the_hash_algo->empty_blob);
170}
171
2f90b9d9 172int hash_algo_by_name(const char *name)
173{
174 int i;
175 if (!name)
176 return GIT_HASH_UNKNOWN;
177 for (i = 1; i < GIT_HASH_NALGOS; i++)
178 if (!strcmp(name, hash_algos[i].name))
179 return i;
180 return GIT_HASH_UNKNOWN;
181}
182
183int hash_algo_by_id(uint32_t format_id)
184{
185 int i;
186 for (i = 1; i < GIT_HASH_NALGOS; i++)
187 if (format_id == hash_algos[i].format_id)
188 return i;
189 return GIT_HASH_UNKNOWN;
190}
191
95399788 192int hash_algo_by_length(int len)
193{
194 int i;
195 for (i = 1; i < GIT_HASH_NALGOS; i++)
196 if (len == hash_algos[i].rawsz)
197 return i;
198 return GIT_HASH_UNKNOWN;
199}
2f90b9d9 200
c597ba80
NTND
201/*
202 * This is meant to hold a *small* number of objects that you would
cb1c8d1d 203 * want read_object_file() to be able to return, but yet you do not want
c597ba80
NTND
204 * to write them into the object store (e.g. a browse-only
205 * application).
206 */
207static struct cached_object {
62ba93ea 208 struct object_id oid;
c597ba80
NTND
209 enum object_type type;
210 void *buf;
211 unsigned long size;
212} *cached_objects;
213static int cached_object_nr, cached_object_alloc;
214
215static struct cached_object empty_tree = {
62ba93ea 216 { EMPTY_TREE_SHA1_BIN_LITERAL },
c597ba80
NTND
217 OBJ_TREE,
218 "",
219 0
220};
221
62ba93ea 222static struct cached_object *find_cached_object(const struct object_id *oid)
c597ba80
NTND
223{
224 int i;
225 struct cached_object *co = cached_objects;
226
227 for (i = 0; i < cached_object_nr; i++, co++) {
4a7e27e9 228 if (oideq(&co->oid, oid))
c597ba80
NTND
229 return co;
230 }
4a7e27e9 231 if (oideq(oid, the_hash_algo->empty_tree))
c597ba80
NTND
232 return &empty_tree;
233 return NULL;
234}
235
9472935d 236
8462ff43 237static int get_conv_flags(unsigned flags)
9472935d
TB
238{
239 if (flags & HASH_RENORMALIZE)
8462ff43 240 return CONV_EOL_RENORMALIZE;
9472935d 241 else if (flags & HASH_WRITE_OBJECT)
107642fe 242 return global_conv_flags_eol | CONV_WRITE_OBJECT;
9472935d 243 else
8462ff43 244 return 0;
9472935d
TB
245}
246
247
90a6464b
JH
248int mkdir_in_gitdir(const char *path)
249{
250 if (mkdir(path, 0777)) {
251 int saved_errno = errno;
252 struct stat st;
253 struct strbuf sb = STRBUF_INIT;
254
255 if (errno != EEXIST)
256 return -1;
257 /*
258 * Are we looking at a path in a symlinked worktree
259 * whose original repository does not yet have it?
260 * e.g. .git/rr-cache pointing at its original
261 * repository in which the user hasn't performed any
262 * conflict resolution yet?
263 */
264 if (lstat(path, &st) || !S_ISLNK(st.st_mode) ||
265 strbuf_readlink(&sb, path, st.st_size) ||
266 !is_absolute_path(sb.buf) ||
267 mkdir(sb.buf, 0777)) {
268 strbuf_release(&sb);
269 errno = saved_errno;
270 return -1;
271 }
272 strbuf_release(&sb);
273 }
274 return adjust_shared_perm(path);
275}
276
0be0521b 277enum scld_error safe_create_leading_directories(char *path)
b2cb9425 278{
26c8ae2a 279 char *next_component = path + offset_1st_component(path);
0be0521b 280 enum scld_error ret = SCLD_OK;
67d42212 281
0be0521b 282 while (ret == SCLD_OK && next_component) {
f0502332 283 struct stat st;
0f527403 284 char *slash = next_component, slash_character;
f0502332 285
0f527403
MH
286 while (*slash && !is_dir_sep(*slash))
287 slash++;
288
289 if (!*slash)
b2cb9425 290 break;
bf10cf70 291
26c8ae2a 292 next_component = slash + 1;
0f527403 293 while (is_dir_sep(*next_component))
bf10cf70 294 next_component++;
26c8ae2a 295 if (!*next_component)
5f0bdf50 296 break;
831651fd 297
0f527403 298 slash_character = *slash;
831651fd 299 *slash = '\0';
67d42212
JR
300 if (!stat(path, &st)) {
301 /* path exists */
204a047f
MH
302 if (!S_ISDIR(st.st_mode)) {
303 errno = ENOTDIR;
0be0521b 304 ret = SCLD_EXISTS;
204a047f 305 }
53a39721 306 } else if (mkdir(path, 0777)) {
928734d9 307 if (errno == EEXIST &&
9e6f885d 308 !stat(path, &st) && S_ISDIR(st.st_mode))
928734d9 309 ; /* somebody created it since we checked */
18d37e86
MH
310 else if (errno == ENOENT)
311 /*
312 * Either mkdir() failed because
313 * somebody just pruned the containing
314 * directory, or stat() failed because
315 * the file that was in our way was
316 * just removed. Either way, inform
317 * the caller that it might be worth
318 * trying again:
319 */
320 ret = SCLD_VANISHED;
9e6f885d 321 else
0be0521b 322 ret = SCLD_FAILED;
53a39721 323 } else if (adjust_shared_perm(path)) {
0be0521b 324 ret = SCLD_PERMS;
457f06d6 325 }
0f527403 326 *slash = slash_character;
b2cb9425 327 }
9e6f885d 328 return ret;
b2cb9425 329}
723c31fe 330
0be0521b 331enum scld_error safe_create_leading_directories_const(const char *path)
8e21d63b 332{
02944307 333 int save_errno;
8e21d63b
JK
334 /* path points to cache entries, so xstrdup before messing with it */
335 char *buf = xstrdup(path);
0be0521b 336 enum scld_error result = safe_create_leading_directories(buf);
02944307
MH
337
338 save_errno = errno;
8e21d63b 339 free(buf);
02944307 340 errno = save_errno;
8e21d63b
JK
341 return result;
342}
343
177978f5
MH
344int raceproof_create_file(const char *path, create_file_fn fn, void *cb)
345{
346 /*
347 * The number of times we will try to remove empty directories
348 * in the way of path. This is only 1 because if another
349 * process is racily creating directories that conflict with
350 * us, we don't want to fight against them.
351 */
352 int remove_directories_remaining = 1;
353
354 /*
355 * The number of times that we will try to create the
356 * directories containing path. We are willing to attempt this
357 * more than once, because another process could be trying to
358 * clean up empty directories at the same time as we are
359 * trying to create them.
360 */
361 int create_directories_remaining = 3;
362
363 /* A scratch copy of path, filled lazily if we need it: */
364 struct strbuf path_copy = STRBUF_INIT;
365
366 int ret, save_errno;
367
368 /* Sanity check: */
369 assert(*path);
370
371retry_fn:
372 ret = fn(path, cb);
373 save_errno = errno;
374 if (!ret)
375 goto out;
376
377 if (errno == EISDIR && remove_directories_remaining-- > 0) {
378 /*
379 * A directory is in the way. Maybe it is empty; try
380 * to remove it:
381 */
382 if (!path_copy.len)
383 strbuf_addstr(&path_copy, path);
384
385 if (!remove_dir_recursively(&path_copy, REMOVE_DIR_EMPTY_ONLY))
386 goto retry_fn;
387 } else if (errno == ENOENT && create_directories_remaining-- > 0) {
388 /*
389 * Maybe the containing directory didn't exist, or
390 * maybe it was just deleted by a process that is
391 * racing with us to clean up empty directories. Try
392 * to create it:
393 */
394 enum scld_error scld_result;
395
396 if (!path_copy.len)
397 strbuf_addstr(&path_copy, path);
398
399 do {
400 scld_result = safe_create_leading_directories(path_copy.buf);
401 if (scld_result == SCLD_OK)
402 goto retry_fn;
403 } while (scld_result == SCLD_VANISHED && create_directories_remaining-- > 0);
404 }
405
406out:
407 strbuf_release(&path_copy);
408 errno = save_errno;
409 return ret;
410}
411
514c5fdd 412static void fill_loose_path(struct strbuf *buf, const struct object_id *oid)
ace1534d
JH
413{
414 int i;
94b5e093 415 for (i = 0; i < the_hash_algo->rawsz; i++) {
ace1534d 416 static char hex[] = "0123456789abcdef";
514c5fdd 417 unsigned int val = oid->hash[i];
f7b7774f
JK
418 strbuf_addch(buf, hex[val >> 4]);
419 strbuf_addch(buf, hex[val & 0xf]);
afbba2f0 420 if (!i)
f7b7774f 421 strbuf_addch(buf, '/');
ace1534d
JH
422 }
423}
424
f0eaf638
JK
425static const char *odb_loose_path(struct object_directory *odb,
426 struct strbuf *buf,
514c5fdd 427 const struct object_id *oid)
0fcfd160 428{
b69fb867 429 strbuf_reset(buf);
f0eaf638 430 strbuf_addstr(buf, odb->path);
34498471 431 strbuf_addch(buf, '/');
514c5fdd 432 fill_loose_path(buf, oid);
f3f043a1 433 return buf->buf;
0fcfd160
LT
434}
435
f3f043a1 436const char *loose_object_path(struct repository *r, struct strbuf *buf,
514c5fdd 437 const struct object_id *oid)
29ec6af2 438{
514c5fdd 439 return odb_loose_path(r->objects->odb, buf, oid);
0fcfd160
LT
440}
441
4ea82473
JK
442/*
443 * Return non-zero iff the path is usable as an alternate object database.
444 */
13313fc3
SB
445static int alt_odb_usable(struct raw_object_store *o,
446 struct strbuf *path,
447 const char *normalized_objdir)
4ea82473 448{
263db403 449 struct object_directory *odb;
4ea82473
JK
450
451 /* Detect cases where alternate disappeared */
452 if (!is_directory(path->buf)) {
259328b7
NTND
453 error(_("object directory %s does not exist; "
454 "check .git/objects/info/alternates"),
4ea82473
JK
455 path->buf);
456 return 0;
457 }
458
459 /*
460 * Prevent the common mistake of listing the same
461 * thing twice, or object directory itself.
462 */
f0eaf638 463 for (odb = o->odb; odb; odb = odb->next) {
263db403 464 if (!fspathcmp(path->buf, odb->path))
4ea82473
JK
465 return 0;
466 }
467 if (!fspathcmp(path->buf, normalized_objdir))
468 return 0;
469
470 return 1;
471}
472
ddd5d056
JH
473/*
474 * Prepare alternate object database registry.
d5a63b99
JH
475 *
476 * The variable alt_odb_list points at the list of struct
263db403 477 * object_directory. The elements on this list come from
d5a63b99
JH
478 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
479 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
1494e038
JH
480 * whose contents is similar to that environment variable but can be
481 * LF separated. Its base points at a statically allocated buffer that
d5a63b99
JH
482 * contains "/the/directory/corresponding/to/.git/objects/...", while
483 * its name points just after the slash at the end of ".git/objects/"
484 * in the example above, and has enough space to hold 40-byte hex
485 * SHA1, an extra slash for the first level indirection, and the
486 * terminating NUL.
ddd5d056 487 */
77f012e8
SB
488static void read_info_alternates(struct repository *r,
489 const char *relative_base,
490 int depth);
491static int link_alt_odb_entry(struct repository *r, const char *entry,
cfc62fc9 492 const char *relative_base, int depth, const char *normalized_objdir)
ace1534d 493{
263db403 494 struct object_directory *ent;
5bdf0a84 495 struct strbuf pathbuf = STRBUF_INIT;
d5a63b99 496
85dadc38 497 if (!is_absolute_path(entry) && relative_base) {
4ac9006f 498 strbuf_realpath(&pathbuf, relative_base, 1);
5bdf0a84 499 strbuf_addch(&pathbuf, '/');
c2f493a4 500 }
6eac50d8 501 strbuf_addstr(&pathbuf, entry);
c2f493a4 502
37a95862 503 if (strbuf_normalize_path(&pathbuf) < 0 && relative_base) {
259328b7 504 error(_("unable to normalize alternate object path: %s"),
670c359d
JK
505 pathbuf.buf);
506 strbuf_release(&pathbuf);
507 return -1;
508 }
5bdf0a84
HW
509
510 /*
511 * The trailing slash after the directory name is given by
512 * this function at the end. Remove duplicates.
513 */
4ea82473
JK
514 while (pathbuf.len && pathbuf.buf[pathbuf.len - 1] == '/')
515 strbuf_setlen(&pathbuf, pathbuf.len - 1);
5bdf0a84 516
77f012e8 517 if (!alt_odb_usable(r->objects, &pathbuf, normalized_objdir)) {
4ea82473 518 strbuf_release(&pathbuf);
c2f493a4
MW
519 return -1;
520 }
521
f0eaf638
JK
522 ent = xcalloc(1, sizeof(*ent));
523 ent->path = xstrdup(pathbuf.buf);
c2f493a4
MW
524
525 /* add the alternate entry */
f0eaf638
JK
526 *r->objects->odb_tail = ent;
527 r->objects->odb_tail = &(ent->next);
c2f493a4
MW
528 ent->next = NULL;
529
530 /* recursively add alternates */
77f012e8 531 read_info_alternates(r, pathbuf.buf, depth + 1);
c2f493a4 532
4ea82473 533 strbuf_release(&pathbuf);
c2f493a4
MW
534 return 0;
535}
536
cf3c6352
JK
537static const char *parse_alt_odb_entry(const char *string,
538 int sep,
539 struct strbuf *out)
540{
541 const char *end;
542
543 strbuf_reset(out);
544
545 if (*string == '#') {
546 /* comment; consume up to next separator */
547 end = strchrnul(string, sep);
548 } else if (*string == '"' && !unquote_c_style(out, string, &end)) {
549 /*
550 * quoted path; unquote_c_style has copied the
551 * data for us and set "end". Broken quoting (e.g.,
552 * an entry that doesn't end with a quote) falls
553 * back to the unquoted case below.
554 */
555 } else {
556 /* normal, unquoted path */
557 end = strchrnul(string, sep);
558 strbuf_add(out, string, end - string);
559 }
560
561 if (*end)
562 end++;
563 return end;
564}
565
77f012e8
SB
566static void link_alt_odb_entries(struct repository *r, const char *alt,
567 int sep, const char *relative_base, int depth)
c2f493a4 568{
539e7506 569 struct strbuf objdirbuf = STRBUF_INIT;
cf3c6352 570 struct strbuf entry = STRBUF_INIT;
c2f493a4 571
f28e3668
JK
572 if (!alt || !*alt)
573 return;
574
c2f493a4 575 if (depth > 5) {
259328b7 576 error(_("%s: ignoring alternate object stores, nesting too deep"),
c2f493a4
MW
577 relative_base);
578 return;
579 }
580
f0eaf638 581 strbuf_add_absolute_path(&objdirbuf, r->objects->odb->path);
670c359d 582 if (strbuf_normalize_path(&objdirbuf) < 0)
259328b7 583 die(_("unable to normalize object directory: %s"),
670c359d 584 objdirbuf.buf);
539e7506 585
cf3c6352
JK
586 while (*alt) {
587 alt = parse_alt_odb_entry(alt, sep, &entry);
588 if (!entry.len)
9577e7e3 589 continue;
77f012e8 590 link_alt_odb_entry(r, entry.buf,
cfc62fc9 591 relative_base, depth, objdirbuf.buf);
9577e7e3 592 }
cf3c6352 593 strbuf_release(&entry);
539e7506 594 strbuf_release(&objdirbuf);
d5a63b99
JH
595}
596
77f012e8
SB
597static void read_info_alternates(struct repository *r,
598 const char *relative_base,
599 int depth)
d5a63b99 600{
5015f01c 601 char *path;
dc732bd5 602 struct strbuf buf = STRBUF_INIT;
d5a63b99 603
5015f01c 604 path = xstrfmt("%s/info/alternates", relative_base);
dc732bd5 605 if (strbuf_read_file(&buf, path, 1024) < 0) {
f0f7bebe 606 warn_on_fopen_errors(path);
dc732bd5 607 free(path);
9a217f2a 608 return;
ace1534d 609 }
d5a63b99 610
77f012e8 611 link_alt_odb_entries(r, buf.buf, '\n', relative_base, depth);
dc732bd5
JK
612 strbuf_release(&buf);
613 free(path);
ace1534d
JH
614}
615
bef70b22
DB
616void add_to_alternates_file(const char *reference)
617{
f132a127 618 struct lock_file lock = LOCK_INIT;
77b9b1d1
JK
619 char *alts = git_pathdup("objects/info/alternates");
620 FILE *in, *out;
f132a127 621 int found = 0;
77b9b1d1 622
f132a127
623 hold_lock_file_for_update(&lock, alts, LOCK_DIE_ON_ERROR);
624 out = fdopen_lock_file(&lock, "w");
77b9b1d1 625 if (!out)
259328b7 626 die_errno(_("unable to fdopen alternates lockfile"));
77b9b1d1
JK
627
628 in = fopen(alts, "r");
629 if (in) {
630 struct strbuf line = STRBUF_INIT;
77b9b1d1 631
3f163962 632 while (strbuf_getline(&line, in) != EOF) {
77b9b1d1
JK
633 if (!strcmp(reference, line.buf)) {
634 found = 1;
635 break;
636 }
637 fprintf_or_die(out, "%s\n", line.buf);
638 }
639
640 strbuf_release(&line);
641 fclose(in);
77b9b1d1
JK
642 }
643 else if (errno != ENOENT)
259328b7 644 die_errno(_("unable to read alternates file"));
77b9b1d1 645
f132a127
646 if (found) {
647 rollback_lock_file(&lock);
648 } else {
77b9b1d1 649 fprintf_or_die(out, "%s\n", reference);
f132a127 650 if (commit_lock_file(&lock))
259328b7 651 die_errno(_("unable to move new alternates file into place"));
f0eaf638 652 if (the_repository->objects->loaded_alternates)
93d8d1e2
SB
653 link_alt_odb_entries(the_repository, reference,
654 '\n', NULL, 0);
77b9b1d1
JK
655 }
656 free(alts);
bef70b22
DB
657}
658
a5b34d21
JK
659void add_to_alternates_memory(const char *reference)
660{
661 /*
662 * Make sure alternates are initialized, or else our entry may be
663 * overwritten when they are.
664 */
0b209034 665 prepare_alt_odb(the_repository);
a5b34d21 666
93d8d1e2
SB
667 link_alt_odb_entries(the_repository, reference,
668 '\n', NULL, 0);
a5b34d21
JK
669}
670
9eeea7d2
SB
671/*
672 * Compute the exact path an alternate is at and returns it. In case of
673 * error NULL is returned and the human readable error is added to `err`
efde7b72 674 * `path` may be relative and should point to $GIT_DIR.
9eeea7d2
SB
675 * `err` must not be null.
676 */
677char *compute_alternate_path(const char *path, struct strbuf *err)
678{
679 char *ref_git = NULL;
680 const char *repo, *ref_git_s;
681 int seen_error = 0;
682
683 ref_git_s = real_path_if_valid(path);
684 if (!ref_git_s) {
685 seen_error = 1;
686 strbuf_addf(err, _("path '%s' does not exist"), path);
687 goto out;
688 } else
689 /*
690 * Beware: read_gitfile(), real_path() and mkpath()
691 * return static buffer
692 */
693 ref_git = xstrdup(ref_git_s);
694
695 repo = read_gitfile(ref_git);
696 if (!repo)
697 repo = read_gitfile(mkpath("%s/.git", ref_git));
698 if (repo) {
699 free(ref_git);
700 ref_git = xstrdup(repo);
701 }
702
703 if (!repo && is_directory(mkpath("%s/.git/objects", ref_git))) {
704 char *ref_git_git = mkpathdup("%s/.git", ref_git);
705 free(ref_git);
706 ref_git = ref_git_git;
707 } else if (!is_directory(mkpath("%s/objects", ref_git))) {
708 struct strbuf sb = STRBUF_INIT;
709 seen_error = 1;
710 if (get_common_dir(&sb, ref_git)) {
711 strbuf_addf(err,
712 _("reference repository '%s' as a linked "
713 "checkout is not supported yet."),
714 path);
715 goto out;
716 }
717
718 strbuf_addf(err, _("reference repository '%s' is not a "
719 "local repository."), path);
720 goto out;
721 }
722
723 if (!access(mkpath("%s/shallow", ref_git), F_OK)) {
724 strbuf_addf(err, _("reference repository '%s' is shallow"),
725 path);
726 seen_error = 1;
727 goto out;
728 }
729
730 if (!access(mkpath("%s/info/grafts", ref_git), F_OK)) {
731 strbuf_addf(err,
732 _("reference repository '%s' is grafted"),
733 path);
734 seen_error = 1;
735 goto out;
736 }
737
738out:
739 if (seen_error) {
6a83d902 740 FREE_AND_NULL(ref_git);
9eeea7d2
SB
741 }
742
743 return ref_git;
744}
745
709dfa69
JK
746static void fill_alternate_refs_command(struct child_process *cmd,
747 const char *repo_path)
748{
749 const char *value;
750
751 if (!git_config_get_value("core.alternateRefsCommand", &value)) {
752 cmd->use_shell = 1;
753
754 argv_array_push(&cmd->args, value);
755 argv_array_push(&cmd->args, repo_path);
756 } else {
757 cmd->git_cmd = 1;
758
759 argv_array_pushf(&cmd->args, "--git-dir=%s", repo_path);
760 argv_array_push(&cmd->args, "for-each-ref");
761 argv_array_push(&cmd->args, "--format=%(objectname)");
762
763 if (!git_config_get_value("core.alternateRefsPrefixes", &value)) {
764 argv_array_push(&cmd->args, "--");
765 argv_array_split(&cmd->args, value);
766 }
767 }
768
769 cmd->env = local_repo_env;
770 cmd->out = -1;
771}
772
773static void read_alternate_refs(const char *path,
774 alternate_ref_fn *cb,
775 void *data)
776{
777 struct child_process cmd = CHILD_PROCESS_INIT;
778 struct strbuf line = STRBUF_INIT;
779 FILE *fh;
780
781 fill_alternate_refs_command(&cmd, path);
782
783 if (start_command(&cmd))
784 return;
785
786 fh = xfdopen(cmd.out, "r");
787 while (strbuf_getline_lf(&line, fh) != EOF) {
788 struct object_id oid;
789 const char *p;
790
791 if (parse_oid_hex(line.buf, &oid, &p) || *p) {
792 warning(_("invalid line while parsing alternate refs: %s"),
793 line.buf);
794 break;
795 }
796
797 cb(&oid, data);
798 }
799
800 fclose(fh);
801 finish_command(&cmd);
86ad3ea5 802 strbuf_release(&line);
709dfa69
JK
803}
804
805struct alternate_refs_data {
806 alternate_ref_fn *fn;
807 void *data;
808};
809
810static int refs_from_alternate_cb(struct object_directory *e,
811 void *data)
812{
813 struct strbuf path = STRBUF_INIT;
814 size_t base_len;
815 struct alternate_refs_data *cb = data;
816
817 if (!strbuf_realpath(&path, e->path, 0))
818 goto out;
819 if (!strbuf_strip_suffix(&path, "/objects"))
820 goto out;
821 base_len = path.len;
822
823 /* Is this a git repository with refs? */
824 strbuf_addstr(&path, "/refs");
825 if (!is_directory(path.buf))
826 goto out;
827 strbuf_setlen(&path, base_len);
828
829 read_alternate_refs(path.buf, cb->fn, cb->data);
830
831out:
832 strbuf_release(&path);
833 return 0;
834}
835
836void for_each_alternate_ref(alternate_ref_fn fn, void *data)
837{
838 struct alternate_refs_data cb;
839 cb.fn = fn;
840 cb.data = data;
841 foreach_alt_odb(refs_from_alternate_cb, &cb);
842}
843
fe1b2268 844int foreach_alt_odb(alt_odb_fn fn, void *cb)
d79796bc 845{
263db403 846 struct object_directory *ent;
fe1b2268 847 int r = 0;
d79796bc 848
0b209034 849 prepare_alt_odb(the_repository);
f0eaf638 850 for (ent = the_repository->objects->odb->next; ent; ent = ent->next) {
fe1b2268
JK
851 r = fn(ent, cb);
852 if (r)
853 break;
854 }
855 return r;
d79796bc
JH
856}
857
13068bf0 858void prepare_alt_odb(struct repository *r)
c2f493a4 859{
f0eaf638 860 if (r->objects->loaded_alternates)
7dc24aa5
SP
861 return;
862
13068bf0 863 link_alt_odb_entries(r, r->objects->alternate_db, PATH_SEP, NULL, 0);
c2f493a4 864
f0eaf638
JK
865 read_info_alternates(r, r->objects->odb->path, 0);
866 r->objects->loaded_alternates = 1;
c2f493a4
MW
867}
868
3096b2ec 869/* Returns 1 if we have successfully freshened the file, 0 otherwise. */
33d4221c 870static int freshen_file(const char *fn)
ace1534d 871{
33d4221c
JK
872 struct utimbuf t;
873 t.actime = t.modtime = time(NULL);
874 return !utime(fn, &t);
0f4dc14a 875}
ace1534d 876
3096b2ec
JK
877/*
878 * All of the check_and_freshen functions return 1 if the file exists and was
879 * freshened (if freshening was requested), 0 otherwise. If they return
880 * 0, you should not assume that it is safe to skip a write of the object (it
881 * either does not exist on disk, or has a stale mtime and may be subject to
882 * pruning).
883 */
6a5e6f5e 884int check_and_freshen_file(const char *fn, int freshen)
33d4221c
JK
885{
886 if (access(fn, F_OK))
887 return 0;
3096b2ec 888 if (freshen && !freshen_file(fn))
33d4221c
JK
889 return 0;
890 return 1;
891}
892
f0eaf638
JK
893static int check_and_freshen_odb(struct object_directory *odb,
894 const struct object_id *oid,
895 int freshen)
33d4221c 896{
f0eaf638 897 static struct strbuf path = STRBUF_INIT;
514c5fdd 898 odb_loose_path(odb, &path, oid);
f0eaf638
JK
899 return check_and_freshen_file(path.buf, freshen);
900}
ea657730 901
f0eaf638
JK
902static int check_and_freshen_local(const struct object_id *oid, int freshen)
903{
904 return check_and_freshen_odb(the_repository->objects->odb, oid, freshen);
33d4221c
JK
905}
906
6862ebbf 907static int check_and_freshen_nonlocal(const struct object_id *oid, int freshen)
0f4dc14a 908{
263db403 909 struct object_directory *odb;
f3f043a1 910
0b209034 911 prepare_alt_odb(the_repository);
f0eaf638
JK
912 for (odb = the_repository->objects->odb->next; odb; odb = odb->next) {
913 if (check_and_freshen_odb(odb, oid, freshen))
c529d75a 914 return 1;
ace1534d 915 }
c529d75a 916 return 0;
ace1534d
JH
917}
918
6862ebbf 919static int check_and_freshen(const struct object_id *oid, int freshen)
33d4221c 920{
6862ebbf 921 return check_and_freshen_local(oid, freshen) ||
922 check_and_freshen_nonlocal(oid, freshen);
33d4221c
JK
923}
924
6862ebbf 925int has_loose_object_nonlocal(const struct object_id *oid)
33d4221c 926{
6862ebbf 927 return check_and_freshen_nonlocal(oid, 0);
33d4221c
JK
928}
929
6862ebbf 930static int has_loose_object(const struct object_id *oid)
0f4dc14a 931{
6862ebbf 932 return check_and_freshen(oid, 0);
0f4dc14a
BC
933}
934
02710228
SP
935static void mmap_limit_check(size_t length)
936{
937 static size_t limit = 0;
938 if (!limit) {
939 limit = git_env_ulong("GIT_MMAP_LIMIT", 0);
940 if (!limit)
941 limit = SIZE_MAX;
942 }
943 if (length > limit)
259328b7 944 die(_("attempting to mmap %"PRIuMAX" over limit %"PRIuMAX),
02710228
SP
945 (uintmax_t)length, (uintmax_t)limit);
946}
947
1570856b
JK
948void *xmmap_gently(void *start, size_t length,
949 int prot, int flags, int fd, off_t offset)
58ecbd5e 950{
02710228
SP
951 void *ret;
952
953 mmap_limit_check(length);
954 ret = mmap(start, length, prot, flags, fd, offset);
9827d4c1
JK
955 if (ret == MAP_FAILED && !length)
956 ret = NULL;
58ecbd5e
JN
957 return ret;
958}
959
1570856b
JK
960void *xmmap(void *start, size_t length,
961 int prot, int flags, int fd, off_t offset)
962{
963 void *ret = xmmap_gently(start, length, prot, flags, fd, offset);
964 if (ret == MAP_FAILED)
259328b7 965 die_errno(_("mmap failed"));
1570856b
JK
966 return ret;
967}
968
88d0db55 969/*
090ea126 970 * With an in-core object data in "map", rehash it to make sure the
cb1c8d1d
JK
971 * object name actually matches "oid" to detect object corruption.
972 * With "map" == NULL, try reading the object named with "oid" using
090ea126 973 * the streaming interface and rehash it to do the same.
88d0db55 974 */
17e65451 975int check_object_signature(const struct object_id *oid, void *map,
976 unsigned long size, const char *type)
88d0db55 977{
f070facc 978 struct object_id real_oid;
090ea126
NTND
979 enum object_type obj_type;
980 struct git_istream *st;
18e2588e 981 git_hash_ctx c;
1af64f73 982 char hdr[MAX_HEADER_LEN];
090ea126 983 int hdrlen;
88d0db55 984
090ea126 985 if (map) {
f070facc 986 hash_object_file(map, size, type, &real_oid);
9001dc2a 987 return !oideq(oid, &real_oid) ? -1 : 0;
88d0db55
BC
988 }
989
ef7b5195 990 st = open_istream(oid, &obj_type, &size, NULL);
090ea126
NTND
991 if (!st)
992 return -1;
88d0db55 993
090ea126 994 /* Generate the header */
ca473cef 995 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %"PRIuMAX , type_name(obj_type), (uintmax_t)size) + 1;
88d0db55 996
090ea126 997 /* Sha1.. */
18e2588e 998 the_hash_algo->init_fn(&c);
999 the_hash_algo->update_fn(&c, hdr, hdrlen);
090ea126
NTND
1000 for (;;) {
1001 char buf[1024 * 16];
1002 ssize_t readlen = read_istream(st, buf, sizeof(buf));
1f688557 1003
f54fac53
JK
1004 if (readlen < 0) {
1005 close_istream(st);
1006 return -1;
1007 }
090ea126
NTND
1008 if (!readlen)
1009 break;
18e2588e 1010 the_hash_algo->update_fn(&c, buf, readlen);
fa5fc15d 1011 }
18e2588e 1012 the_hash_algo->final_fn(real_oid.hash, &c);
090ea126 1013 close_istream(st);
9001dc2a 1014 return !oideq(oid, &real_oid) ? -1 : 0;
fa5fc15d
SP
1015}
1016
1b8ac5ea 1017int git_open_cloexec(const char *name, int flags)
a0788266 1018{
1e3001a8
JH
1019 int fd;
1020 static int o_cloexec = O_CLOEXEC;
a0788266 1021
1e3001a8
JH
1022 fd = open(name, flags | o_cloexec);
1023 if ((o_cloexec & O_CLOEXEC) && fd < 0 && errno == EINVAL) {
cd66ada0 1024 /* Try again w/o O_CLOEXEC: the kernel might not support it */
1e3001a8
JH
1025 o_cloexec &= ~O_CLOEXEC;
1026 fd = open(name, flags | o_cloexec);
491a8dec 1027 }
491a8dec 1028
9fb9495d 1029#if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
491a8dec 1030 {
1e3001a8 1031 static int fd_cloexec = FD_CLOEXEC;
a0788266 1032
1e3001a8
JH
1033 if (!o_cloexec && 0 <= fd && fd_cloexec) {
1034 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
9fb9495d
EW
1035 int flags = fcntl(fd, F_GETFD);
1036 if (fcntl(fd, F_SETFD, flags | fd_cloexec))
1e3001a8 1037 fd_cloexec = 0;
cd66ada0 1038 }
44d1c19e 1039 }
a0788266 1040#endif
1b8ac5ea 1041 return fd;
a0788266
JS
1042}
1043
3cf8b462 1044/*
514c5fdd 1045 * Find "oid" as a loose object in the local repository or in an alternate.
771e7d57
JK
1046 * Returns 0 on success, negative on failure.
1047 *
1048 * The "path" out-parameter will give the path of the object we found (if any).
1049 * Note that it may point to static storage and is only valid until another
514c5fdd 1050 * call to stat_loose_object().
3cf8b462 1051 */
514c5fdd
JK
1052static int stat_loose_object(struct repository *r, const struct object_id *oid,
1053 struct stat *st, const char **path)
1f688557 1054{
263db403 1055 struct object_directory *odb;
ea657730
CC
1056 static struct strbuf buf = STRBUF_INIT;
1057
d2607fa0 1058 prepare_alt_odb(r);
f0eaf638 1059 for (odb = r->objects->odb; odb; odb = odb->next) {
514c5fdd 1060 *path = odb_loose_path(odb, &buf, oid);
771e7d57 1061 if (!lstat(*path, st))
052fe5ea 1062 return 0;
c7934306
SP
1063 }
1064
3cf8b462
SP
1065 return -1;
1066}
1067
771e7d57 1068/*
514c5fdd 1069 * Like stat_loose_object(), but actually open the object and return the
771e7d57
JK
1070 * descriptor. See the caveats on the "path" parameter above.
1071 */
514c5fdd
JK
1072static int open_loose_object(struct repository *r,
1073 const struct object_id *oid, const char **path)
60bb8b14 1074{
44d1c19e 1075 int fd;
263db403 1076 struct object_directory *odb;
f0eaf638 1077 int most_interesting_errno = ENOENT;
ea657730
CC
1078 static struct strbuf buf = STRBUF_INIT;
1079
ec7283e5 1080 prepare_alt_odb(r);
f0eaf638 1081 for (odb = r->objects->odb; odb; odb = odb->next) {
514c5fdd 1082 *path = odb_loose_path(odb, &buf, oid);
771e7d57 1083 fd = git_open(*path);
44d1c19e
LT
1084 if (fd >= 0)
1085 return fd;
f0eaf638 1086
d6c8a05b
JK
1087 if (most_interesting_errno == ENOENT)
1088 most_interesting_errno = errno;
03e79c88 1089 }
d6c8a05b 1090 errno = most_interesting_errno;
44d1c19e 1091 return -1;
1f688557
JH
1092}
1093
61c7711c 1094static int quick_has_loose(struct repository *r,
d7a24573 1095 const struct object_id *oid)
61c7711c 1096{
61c7711c
JK
1097 struct object_directory *odb;
1098
61c7711c
JK
1099 prepare_alt_odb(r);
1100 for (odb = r->objects->odb; odb; odb = odb->next) {
d7a24573 1101 if (oid_array_lookup(odb_loose_cache(odb, oid), oid) >= 0)
61c7711c
JK
1102 return 1;
1103 }
1104 return 0;
1105}
1106
f6371f92
JK
1107/*
1108 * Map the loose object at "path" if it is not NULL, or the path found by
514c5fdd 1109 * searching for a loose object named "oid".
f6371f92 1110 */
514c5fdd
JK
1111static void *map_loose_object_1(struct repository *r, const char *path,
1112 const struct object_id *oid, unsigned long *size)
27d69a46 1113{
0fcfd160 1114 void *map;
144bde78 1115 int fd;
84dd81c1 1116
f6371f92
JK
1117 if (path)
1118 fd = git_open(path);
8261e1f1 1119 else
514c5fdd 1120 fd = open_loose_object(r, oid, &path);
44d1c19e
LT
1121 map = NULL;
1122 if (fd >= 0) {
1123 struct stat st;
1f688557 1124
44d1c19e
LT
1125 if (!fstat(fd, &st)) {
1126 *size = xsize_t(st.st_size);
33e42de0
MM
1127 if (!*size) {
1128 /* mmap() is forbidden on empty files */
259328b7 1129 error(_("object file %s is empty"), path);
6881925e 1130 close(fd);
33e42de0
MM
1131 return NULL;
1132 }
44d1c19e 1133 map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0);
144bde78 1134 }
44d1c19e 1135 close(fd);
74e34e1f 1136 }
0fcfd160 1137 return map;
74e34e1f
NP
1138}
1139
514c5fdd
JK
1140void *map_loose_object(struct repository *r,
1141 const struct object_id *oid,
1142 unsigned long *size)
068f85e3 1143{
514c5fdd 1144 return map_loose_object_1(r, NULL, oid, size);
068f85e3 1145}
1146
00a7760e
JK
1147static int unpack_loose_short_header(git_zstream *stream,
1148 unsigned char *map, unsigned long mapsize,
1149 void *buffer, unsigned long bufsiz)
47fe3f6e 1150{
c4483576
LT
1151 /* Get the data stream */
1152 memset(stream, 0, sizeof(*stream));
1153 stream->next_in = map;
1154 stream->avail_in = mapsize;
1155 stream->next_out = buffer;
93821bd9 1156 stream->avail_out = bufsiz;
47fe3f6e 1157
39c68542 1158 git_inflate_init(stream);
cc5c54e7 1159 return git_inflate(stream, 0);
9a217f2a
JH
1160}
1161
00a7760e
JK
1162int unpack_loose_header(git_zstream *stream,
1163 unsigned char *map, unsigned long mapsize,
1164 void *buffer, unsigned long bufsiz)
1f688557 1165{
00a7760e
JK
1166 int status = unpack_loose_short_header(stream, map, mapsize,
1167 buffer, bufsiz);
42873078 1168
d21f8426
JH
1169 if (status < Z_OK)
1170 return status;
628522ec 1171
d21f8426
JH
1172 /* Make sure we have the terminating NUL */
1173 if (!memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1174 return -1;
1f688557
JH
1175 return 0;
1176}
1177
00a7760e
JK
1178static int unpack_loose_header_to_strbuf(git_zstream *stream, unsigned char *map,
1179 unsigned long mapsize, void *buffer,
1180 unsigned long bufsiz, struct strbuf *header)
d131b7af 1181{
46f03448
KN
1182 int status;
1183
00a7760e 1184 status = unpack_loose_short_header(stream, map, mapsize, buffer, bufsiz);
d21f8426
JH
1185 if (status < Z_OK)
1186 return -1;
d131b7af 1187
46f03448
KN
1188 /*
1189 * Check if entire header is unpacked in the first iteration.
d131b7af 1190 */
46f03448
KN
1191 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1192 return 0;
d131b7af 1193
46f03448
KN
1194 /*
1195 * buffer[0..bufsiz] was not large enough. Copy the partial
1196 * result out to header, and then append the result of further
1197 * reading the stream.
1198 */
1199 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
1200 stream->next_out = buffer;
1201 stream->avail_out = bufsiz;
d131b7af 1202
46f03448
KN
1203 do {
1204 status = git_inflate(stream, 0);
1205 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
1206 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1207 return 0;
1208 stream->next_out = buffer;
1209 stream->avail_out = bufsiz;
1210 } while (status != Z_STREAM_END);
1211 return -1;
d131b7af
SP
1212}
1213
00a7760e
JK
1214static void *unpack_loose_rest(git_zstream *stream,
1215 void *buffer, unsigned long size,
1216 const struct object_id *oid)
95099731 1217{
5180cacc 1218 int bytes = strlen(buffer) + 1;
3aee68aa 1219 unsigned char *buf = xmallocz(size);
93821bd9 1220 unsigned long n;
7efbff75 1221 int status = Z_OK;
95099731 1222
93821bd9
LT
1223 n = stream->total_out - bytes;
1224 if (n > size)
1225 n = size;
1226 memcpy(buf, (char *) buffer + bytes, n);
1227 bytes = n;
456cdf6e
LT
1228 if (bytes <= size) {
1229 /*
1230 * The above condition must be (bytes <= size), not
1231 * (bytes < size). In other words, even though we
ccf5ace0 1232 * expect no more output and set avail_out to zero,
456cdf6e
LT
1233 * the input zlib stream may have bytes that express
1234 * "this concludes the stream", and we *do* want to
1235 * eat that input.
1236 *
1237 * Otherwise we would not be able to test that we
1238 * consumed all the input to reach the expected size;
1239 * we also want to check that zlib tells us that all
1240 * went well with status == Z_STREAM_END at the end.
1241 */
5180cacc
LT
1242 stream->next_out = buf + bytes;
1243 stream->avail_out = size - bytes;
7efbff75 1244 while (status == Z_OK)
39c68542 1245 status = git_inflate(stream, Z_FINISH);
5180cacc 1246 }
456cdf6e 1247 if (status == Z_STREAM_END && !stream->avail_in) {
39c68542 1248 git_inflate_end(stream);
7efbff75 1249 return buf;
95099731
NTND
1250 }
1251
7efbff75 1252 if (status < 0)
00a7760e 1253 error(_("corrupt loose object '%s'"), oid_to_hex(oid));
7efbff75 1254 else if (stream->avail_in)
259328b7 1255 error(_("garbage at end of loose object '%s'"),
00a7760e 1256 oid_to_hex(oid));
7efbff75
JH
1257 free(buf);
1258 return NULL;
95099731
NTND
1259}
1260
d40d535b 1261/*
5180cacc
LT
1262 * We used to just use "sscanf()", but that's actually way
1263 * too permissive for what we want to check. So do an anal
1264 * object header parse by hand.
d40d535b 1265 */
00a7760e
JK
1266static int parse_loose_header_extended(const char *hdr, struct object_info *oi,
1267 unsigned int flags)
1f688557 1268{
46f03448 1269 const char *type_buf = hdr;
5180cacc 1270 unsigned long size;
46f03448 1271 int type, type_len = 0;
43057304 1272
5180cacc 1273 /*
46f03448 1274 * The type can be of any size but is followed by
21666f1a 1275 * a space.
5180cacc 1276 */
5180cacc
LT
1277 for (;;) {
1278 char c = *hdr++;
d21f8426
JH
1279 if (!c)
1280 return -1;
5180cacc
LT
1281 if (c == ' ')
1282 break;
46f03448 1283 type_len++;
5180cacc 1284 }
1f688557 1285
46f03448 1286 type = type_from_string_gently(type_buf, type_len, 1);
6ca32f47
BW
1287 if (oi->type_name)
1288 strbuf_add(oi->type_name, type_buf, type_len);
46f03448
KN
1289 /*
1290 * Set type to 0 if its an unknown object and
2e3a16b2 1291 * we're obtaining the type using '--allow-unknown-type'
46f03448
KN
1292 * option.
1293 */
19fc5e84 1294 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE) && (type < 0))
46f03448
KN
1295 type = 0;
1296 else if (type < 0)
259328b7 1297 die(_("invalid object type"));
46f03448
KN
1298 if (oi->typep)
1299 *oi->typep = type;
5180cacc
LT
1300
1301 /*
1302 * The length must follow immediately, and be in canonical
1303 * decimal format (ie "010" is not valid).
1304 */
1305 size = *hdr++ - '0';
1306 if (size > 9)
1307 return -1;
1308 if (size) {
1309 for (;;) {
1310 unsigned long c = *hdr - '0';
1311 if (c > 9)
1312 break;
1313 hdr++;
1314 size = size * 10 + c;
1b1005d1 1315 }
c01f51cc 1316 }
46f03448
KN
1317
1318 if (oi->sizep)
1319 *oi->sizep = size;
5180cacc
LT
1320
1321 /*
1322 * The length must be followed by a zero byte
1323 */
46f03448 1324 return *hdr ? -1 : type;
1f688557
JH
1325}
1326
00a7760e 1327int parse_loose_header(const char *hdr, unsigned long *sizep)
bf592c50 1328{
27b5c1a0 1329 struct object_info oi = OBJECT_INFO_INIT;
21666f1a 1330
46f03448 1331 oi.sizep = sizep;
00a7760e 1332 return parse_loose_header_extended(hdr, &oi, 0);
bf592c50
DB
1333}
1334
514c5fdd
JK
1335static int loose_object_info(struct repository *r,
1336 const struct object_id *oid,
1337 struct object_info *oi, int flags)
65c2e0c3 1338{
46f03448
KN
1339 int status = 0;
1340 unsigned long mapsize;
65c2e0c3 1341 void *map;
ef49a7a0 1342 git_zstream stream;
1af64f73 1343 char hdr[MAX_HEADER_LEN];
46f03448 1344 struct strbuf hdrbuf = STRBUF_INIT;
c84a1f3e 1345 unsigned long size_scratch;
65c2e0c3 1346
5d642e75
JK
1347 if (oi->delta_base_sha1)
1348 hashclr(oi->delta_base_sha1);
1349
052fe5ea
JK
1350 /*
1351 * If we don't care about type or size, then we don't
4ef8d1dd
JH
1352 * need to look inside the object at all. Note that we
1353 * do not optimize out the stat call, even if the
1354 * caller doesn't care about the disk-size, since our
1355 * return value implicitly indicates whether the
1356 * object even exists.
052fe5ea 1357 */
6ca32f47 1358 if (!oi->typep && !oi->type_name && !oi->sizep && !oi->contentp) {
771e7d57 1359 const char *path;
4ef8d1dd 1360 struct stat st;
61c7711c 1361 if (!oi->disk_sizep && (flags & OBJECT_INFO_QUICK))
d7a24573 1362 return quick_has_loose(r, oid) ? 0 : -1;
514c5fdd 1363 if (stat_loose_object(r, oid, &st, &path) < 0)
4ef8d1dd
JH
1364 return -1;
1365 if (oi->disk_sizep)
23c339c0 1366 *oi->disk_sizep = st.st_size;
052fe5ea
JK
1367 return 0;
1368 }
1369
514c5fdd 1370 map = map_loose_object(r, oid, &mapsize);
f0df4ed5 1371 if (!map)
dbea72a8 1372 return -1;
c84a1f3e
JT
1373
1374 if (!oi->sizep)
1375 oi->sizep = &size_scratch;
1376
23c339c0
JK
1377 if (oi->disk_sizep)
1378 *oi->disk_sizep = mapsize;
19fc5e84 1379 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE)) {
00a7760e 1380 if (unpack_loose_header_to_strbuf(&stream, map, mapsize, hdr, sizeof(hdr), &hdrbuf) < 0)
259328b7 1381 status = error(_("unable to unpack %s header with --allow-unknown-type"),
514c5fdd 1382 oid_to_hex(oid));
00a7760e 1383 } else if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
259328b7 1384 status = error(_("unable to unpack %s header"),
514c5fdd 1385 oid_to_hex(oid));
46f03448
KN
1386 if (status < 0)
1387 ; /* Do nothing */
1388 else if (hdrbuf.len) {
00a7760e 1389 if ((status = parse_loose_header_extended(hdrbuf.buf, oi, flags)) < 0)
259328b7 1390 status = error(_("unable to parse %s header with --allow-unknown-type"),
514c5fdd 1391 oid_to_hex(oid));
00a7760e 1392 } else if ((status = parse_loose_header_extended(hdr, oi, flags)) < 0)
514c5fdd 1393 status = error(_("unable to parse %s header"), oid_to_hex(oid));
c84a1f3e 1394
b3ea7dd3 1395 if (status >= 0 && oi->contentp) {
00a7760e
JK
1396 *oi->contentp = unpack_loose_rest(&stream, hdr,
1397 *oi->sizep, oid);
b3ea7dd3
JK
1398 if (!*oi->contentp) {
1399 git_inflate_end(&stream);
1400 status = -1;
1401 }
1402 } else
c84a1f3e
JT
1403 git_inflate_end(&stream);
1404
65c2e0c3 1405 munmap(map, mapsize);
46f03448 1406 if (status && oi->typep)
23c339c0 1407 *oi->typep = status;
c84a1f3e
JT
1408 if (oi->sizep == &size_scratch)
1409 oi->sizep = NULL;
46f03448 1410 strbuf_release(&hdrbuf);
3ab0fb06 1411 oi->whence = OI_LOOSE;
93cff9a9 1412 return (status < 0) ? status : 0;
65c2e0c3
JH
1413}
1414
8b4c0103
JT
1415int fetch_if_missing = 1;
1416
9d98354f
SB
1417int oid_object_info_extended(struct repository *r, const struct object_id *oid,
1418 struct object_info *oi, unsigned flags)
f0df4ed5 1419{
cd585e2a 1420 static struct object_info blank_oi = OBJECT_INFO_INIT;
f0df4ed5 1421 struct pack_entry e;
5b086407 1422 int rtype;
b383a13c 1423 const struct object_id *real = oid;
8b4c0103 1424 int already_retried = 0;
f0df4ed5 1425
b383a13c 1426 if (flags & OBJECT_INFO_LOOKUP_REPLACE)
9d98354f 1427 real = lookup_replace_object(r, oid);
f0df4ed5 1428
b383a13c 1429 if (is_null_oid(real))
87b5e236
JK
1430 return -1;
1431
cd585e2a
JT
1432 if (!oi)
1433 oi = &blank_oi;
1434
dfdd4afc 1435 if (!(flags & OBJECT_INFO_SKIP_CACHED)) {
62ba93ea 1436 struct cached_object *co = find_cached_object(real);
dfdd4afc
JT
1437 if (co) {
1438 if (oi->typep)
1439 *(oi->typep) = co->type;
1440 if (oi->sizep)
1441 *(oi->sizep) = co->size;
1442 if (oi->disk_sizep)
1443 *(oi->disk_sizep) = 0;
1444 if (oi->delta_base_sha1)
1445 hashclr(oi->delta_base_sha1);
6ca32f47 1446 if (oi->type_name)
debca9d2 1447 strbuf_addstr(oi->type_name, type_name(co->type));
dfdd4afc
JT
1448 if (oi->contentp)
1449 *oi->contentp = xmemdupz(co->buf, co->size);
1450 oi->whence = OI_CACHED;
1451 return 0;
1452 }
c4d9986f
NTND
1453 }
1454
8b4c0103 1455 while (1) {
42c8ce1c 1456 if (find_pack_entry(r, real, &e))
8b4c0103
JT
1457 break;
1458
024aa469
TI
1459 if (flags & OBJECT_INFO_IGNORE_LOOSE)
1460 return -1;
1461
ddd63e64 1462 /* Most likely it's a loose object. */
514c5fdd 1463 if (!loose_object_info(r, real, oi, flags))
5b086407 1464 return 0;
ddd63e64
SG
1465
1466 /* Not a loose object; someone else may have just packed it. */
2b7750c9 1467 if (!(flags & OBJECT_INFO_QUICK)) {
9d98354f 1468 reprepare_packed_git(r);
42c8ce1c 1469 if (find_pack_entry(r, real, &e))
2b7750c9
JT
1470 break;
1471 }
8b4c0103
JT
1472
1473 /* Check if it is a missing object */
b14ed5ad 1474 if (fetch_if_missing && has_promisor_remote() &&
0f4a4fb1 1475 !already_retried && r == the_repository &&
31f5256c 1476 !(flags & OBJECT_INFO_SKIP_FETCH_OBJECT)) {
8b4c0103 1477 /*
b14ed5ad
CC
1478 * TODO Investigate checking promisor_remote_get_direct()
1479 * TODO return value and stopping on error here.
1480 * TODO Pass a repository struct through
1481 * promisor_remote_get_direct(), such that arbitrary
1482 * repositories work.
8b4c0103 1483 */
b14ed5ad 1484 promisor_remote_get_direct(r, real, 1);
8b4c0103
JT
1485 already_retried = 1;
1486 continue;
dfdd4afc 1487 }
8b4c0103
JT
1488
1489 return -1;
f0df4ed5 1490 }
3d77d877 1491
cd585e2a
JT
1492 if (oi == &blank_oi)
1493 /*
1494 * We know that the caller doesn't actually need the
1495 * information below, so return early.
1496 */
1497 return 0;
9d98354f 1498 rtype = packed_object_info(r, e.p, e.offset, oi);
412916ee 1499 if (rtype < 0) {
b383a13c 1500 mark_bad_packed_object(e.p, real->hash);
9d98354f 1501 return oid_object_info_extended(r, real, oi, 0);
3ab0fb06 1502 } else if (oi->whence == OI_PACKED) {
9a490590
JH
1503 oi->u.packed.offset = e.offset;
1504 oi->u.packed.pack = e.p;
1505 oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA ||
1506 rtype == OBJ_OFS_DELTA);
3d77d877
NP
1507 }
1508
5b086407 1509 return 0;
f0df4ed5
JS
1510}
1511
3fc0dca9 1512/* returns enum object_type or negative */
9d98354f
SB
1513int oid_object_info(struct repository *r,
1514 const struct object_id *oid,
1515 unsigned long *sizep)
9a490590 1516{
5b086407 1517 enum object_type type;
27b5c1a0 1518 struct object_info oi = OBJECT_INFO_INIT;
9a490590 1519
5b086407 1520 oi.typep = &type;
9a490590 1521 oi.sizep = sizep;
9d98354f
SB
1522 if (oid_object_info_extended(r, oid, &oi,
1523 OBJECT_INFO_LOOKUP_REPLACE) < 0)
5b086407
JK
1524 return -1;
1525 return type;
9a490590
JH
1526}
1527
1b9b5c69 1528static void *read_object(struct repository *r,
cba595ab 1529 const struct object_id *oid, enum object_type *type,
f1d8130b
JT
1530 unsigned long *size)
1531{
1532 struct object_info oi = OBJECT_INFO_INIT;
1533 void *content;
1534 oi.typep = type;
1535 oi.sizep = size;
1536 oi.contentp = &content;
1537
cba595ab 1538 if (oid_object_info_extended(r, oid, &oi, 0) < 0)
f1d8130b
JT
1539 return NULL;
1540 return content;
1541}
1542
829e5c3b
PO
1543int pretend_object_file(void *buf, unsigned long len, enum object_type type,
1544 struct object_id *oid)
d66b37bb
JH
1545{
1546 struct cached_object *co;
1547
169c9c01 1548 hash_object_file(buf, len, type_name(type), oid);
98374a07 1549 if (has_object_file(oid) || find_cached_object(oid))
d66b37bb 1550 return 0;
c7353967 1551 ALLOC_GROW(cached_objects, cached_object_nr + 1, cached_object_alloc);
d66b37bb
JH
1552 co = &cached_objects[cached_object_nr++];
1553 co->size = len;
21666f1a 1554 co->type = type;
efa13f7b
JH
1555 co->buf = xmalloc(len);
1556 memcpy(co->buf, buf, len);
62ba93ea 1557 oidcpy(&co->oid, oid);
d66b37bb
JH
1558 return 0;
1559}
1560
b6c4cecc
JH
1561/*
1562 * This function dies on corrupt objects; the callers who want to
1563 * deal with them should arrange to call read_object() and give error
1564 * messages themselves.
1565 */
a3b72c89
SB
1566void *read_object_file_extended(struct repository *r,
1567 const struct object_id *oid,
b4f5aca4 1568 enum object_type *type,
1569 unsigned long *size,
1570 int lookup_replace)
ac939109 1571{
3ba7a065 1572 void *data;
b6c4cecc 1573 const struct packed_git *p;
771e7d57
JK
1574 const char *path;
1575 struct stat st;
1f2e7cea 1576 const struct object_id *repl = lookup_replace ?
a3b72c89 1577 lookup_replace_object(r, oid) : oid;
b6c4cecc 1578
3ba7a065 1579 errno = 0;
cba595ab 1580 data = read_object(r, repl, type, size);
4bbf5a26 1581 if (data)
b6c4cecc 1582 return data;
68095570 1583
25f3af3f 1584 if (errno && errno != ENOENT)
259328b7 1585 die_errno(_("failed to read object %s"), oid_to_hex(oid));
3ba7a065 1586
68095570 1587 /* die if we replaced an object with one that does not exist */
b383a13c 1588 if (repl != oid)
259328b7 1589 die(_("replacement %s not found for %s"),
b383a13c 1590 oid_to_hex(repl), oid_to_hex(oid));
68095570 1591
cba595ab 1592 if (!stat_loose_object(r, repl, &st, &path))
259328b7 1593 die(_("loose object %s (stored in %s) is corrupt"),
b383a13c 1594 oid_to_hex(repl), path);
68095570 1595
a3b72c89 1596 if ((p = has_packed_and_bad(r, repl->hash)) != NULL)
259328b7 1597 die(_("packed object %s (stored in %s) is corrupt"),
b383a13c 1598 oid_to_hex(repl), p->pack_name);
f5552aee 1599
b6c4cecc 1600 return NULL;
ac939109
NP
1601}
1602
d3b4705a
NTND
1603void *read_object_with_reference(struct repository *r,
1604 const struct object_id *oid,
21666f1a 1605 const char *required_type_name,
40469ee9 1606 unsigned long *size,
02f0547e 1607 struct object_id *actual_oid_return)
f4913f91 1608{
21666f1a 1609 enum object_type type, required_type;
f4913f91
JH
1610 void *buffer;
1611 unsigned long isize;
02f0547e 1612 struct object_id actual_oid;
f4913f91 1613
21666f1a 1614 required_type = type_from_string(required_type_name);
02f0547e 1615 oidcpy(&actual_oid, oid);
40469ee9
JH
1616 while (1) {
1617 int ref_length = -1;
1618 const char *ref_type = NULL;
f4913f91 1619
d3b4705a 1620 buffer = repo_read_object_file(r, &actual_oid, &type, &isize);
40469ee9
JH
1621 if (!buffer)
1622 return NULL;
21666f1a 1623 if (type == required_type) {
40469ee9 1624 *size = isize;
02f0547e 1625 if (actual_oid_return)
1626 oidcpy(actual_oid_return, &actual_oid);
40469ee9
JH
1627 return buffer;
1628 }
1629 /* Handle references */
21666f1a 1630 else if (type == OBJ_COMMIT)
40469ee9 1631 ref_type = "tree ";
21666f1a 1632 else if (type == OBJ_TAG)
40469ee9
JH
1633 ref_type = "object ";
1634 else {
1635 free(buffer);
1636 return NULL;
1637 }
1638 ref_length = strlen(ref_type);
f4913f91 1639
94b5e093 1640 if (ref_length + the_hash_algo->hexsz > isize ||
50974ec9 1641 memcmp(buffer, ref_type, ref_length) ||
02f0547e 1642 get_oid_hex((char *) buffer + ref_length, &actual_oid)) {
40469ee9
JH
1643 free(buffer);
1644 return NULL;
1645 }
1cf58e72 1646 free(buffer);
40469ee9 1647 /* Now we have the ID of the referred-to object in
02f0547e 1648 * actual_oid. Check again. */
f4913f91 1649 }
f4913f91
JH
1650}
1651
a09c985e
PO
1652static void write_object_file_prepare(const void *buf, unsigned long len,
1653 const char *type, struct object_id *oid,
1654 char *hdr, int *hdrlen)
d410c0f5 1655{
18e2588e 1656 git_hash_ctx c;
d410c0f5
JH
1657
1658 /* Generate the header */
ca473cef 1659 *hdrlen = xsnprintf(hdr, *hdrlen, "%s %"PRIuMAX , type, (uintmax_t)len)+1;
d410c0f5
JH
1660
1661 /* Sha1.. */
18e2588e 1662 the_hash_algo->init_fn(&c);
1663 the_hash_algo->update_fn(&c, hdr, *hdrlen);
1664 the_hash_algo->update_fn(&c, buf, len);
0fd90dab 1665 the_hash_algo->final_fn(oid->hash, &c);
d410c0f5
JH
1666}
1667
230f1322 1668/*
5a688fe4 1669 * Move the just written object into its final resting place.
230f1322 1670 */
cb5add58 1671int finalize_object_file(const char *tmpfile, const char *filename)
230f1322 1672{
e32c0a9c 1673 int ret = 0;
5a688fe4 1674
348df166 1675 if (object_creation_mode == OBJECT_CREATION_USES_RENAMES)
be66a6c4
JS
1676 goto try_rename;
1677 else if (link(tmpfile, filename))
e32c0a9c 1678 ret = errno;
7ebb6fca
LT
1679
1680 /*
1681 * Coda hack - coda doesn't like cross-directory links,
1682 * so we fall back to a rename, which will mean that it
1683 * won't be able to check collisions, but that's not a
1684 * big deal.
1685 *
1686 * The same holds for FAT formatted media.
1687 *
3be1f18e 1688 * When this succeeds, we just return. We have nothing
7ebb6fca
LT
1689 * left to unlink.
1690 */
1691 if (ret && ret != EEXIST) {
be66a6c4 1692 try_rename:
7ebb6fca 1693 if (!rename(tmpfile, filename))
3be1f18e 1694 goto out;
9e48b389 1695 ret = errno;
230f1322 1696 }
691f1a28 1697 unlink_or_warn(tmpfile);
230f1322
LT
1698 if (ret) {
1699 if (ret != EEXIST) {
2c319886 1700 return error_errno(_("unable to write file %s"), filename);
230f1322
LT
1701 }
1702 /* FIXME!!! Collision check here ? */
1703 }
1704
3be1f18e 1705out:
5256b006 1706 if (adjust_shared_perm(filename))
259328b7 1707 return error(_("unable to set permission to '%s'"), filename);
230f1322
LT
1708 return 0;
1709}
1710
4d548150
LT
1711static int write_buffer(int fd, const void *buf, size_t len)
1712{
d34cf19b 1713 if (write_in_full(fd, buf, len) < 0)
259328b7 1714 return error_errno(_("file write error"));
4d548150
LT
1715 return 0;
1716}
1717
f070facc
PO
1718int hash_object_file(const void *buf, unsigned long len, const char *type,
1719 struct object_id *oid)
abdc3fc8 1720{
1af64f73 1721 char hdr[MAX_HEADER_LEN];
ef1286d3 1722 int hdrlen = sizeof(hdr);
a09c985e 1723 write_object_file_prepare(buf, len, type, oid, hdr, &hdrlen);
abdc3fc8
RS
1724 return 0;
1725}
1726
e9039dd3 1727/* Finalize a file on disk, and close it. */
514c5fdd 1728static void close_loose_object(int fd)
e9039dd3 1729{
aafe9fba 1730 if (fsync_object_files)
76011357 1731 fsync_or_die(fd, "loose object file");
e9039dd3 1732 if (close(fd) != 0)
76011357 1733 die_errno(_("error when closing loose object file"));
e9039dd3
LT
1734}
1735
5723fe7e
LT
1736/* Size of directory component, including the ending '/' */
1737static inline int directory_size(const char *filename)
1738{
1739 const char *s = strrchr(filename, '/');
1740 if (!s)
1741 return 0;
1742 return s - filename + 1;
1743}
1744
1745/*
1746 * This creates a temporary file in the same directory as the final
1747 * 'filename'
1748 *
1749 * We want to avoid cross-directory filename renames, because those
1750 * can have problems on various filesystems (FAT, NFS, Coda).
1751 */
d4b3d11a 1752static int create_tmpfile(struct strbuf *tmp, const char *filename)
5723fe7e
LT
1753{
1754 int fd, dirlen = directory_size(filename);
1755
d4b3d11a
JK
1756 strbuf_reset(tmp);
1757 strbuf_add(tmp, filename, dirlen);
1758 strbuf_addstr(tmp, "tmp_obj_XXXXXX");
1759 fd = git_mkstemp_mode(tmp->buf, 0444);
cbacbf4e 1760 if (fd < 0 && dirlen && errno == ENOENT) {
d4b3d11a
JK
1761 /*
1762 * Make sure the directory exists; note that the contents
1763 * of the buffer are undefined after mkstemp returns an
1764 * error, so we have to rewrite the whole buffer from
1765 * scratch.
1766 */
1767 strbuf_reset(tmp);
1768 strbuf_add(tmp, filename, dirlen - 1);
1769 if (mkdir(tmp->buf, 0777) && errno != EEXIST)
b2476a60 1770 return -1;
d4b3d11a 1771 if (adjust_shared_perm(tmp->buf))
5723fe7e
LT
1772 return -1;
1773
1774 /* Try again */
d4b3d11a
JK
1775 strbuf_addstr(tmp, "/tmp_obj_XXXXXX");
1776 fd = git_mkstemp_mode(tmp->buf, 0444);
5723fe7e
LT
1777 }
1778 return fd;
1779}
1780
3fc7281f
PO
1781static int write_loose_object(const struct object_id *oid, char *hdr,
1782 int hdrlen, const void *buf, unsigned long len,
1783 time_t mtime)
0fcfd160 1784{
915308b1 1785 int fd, ret;
9892beba 1786 unsigned char compressed[4096];
ef49a7a0 1787 git_zstream stream;
18e2588e 1788 git_hash_ctx c;
3fc7281f 1789 struct object_id parano_oid;
d4b3d11a 1790 static struct strbuf tmp_file = STRBUF_INIT;
ea657730
CC
1791 static struct strbuf filename = STRBUF_INIT;
1792
514c5fdd 1793 loose_object_path(the_repository, &filename, oid);
a44c9a5e 1794
ea657730 1795 fd = create_tmpfile(&tmp_file, filename.buf);
aac17941 1796 if (fd < 0) {
35243577 1797 if (errno == EACCES)
259328b7 1798 return error(_("insufficient permission for adding an object to repository database %s"), get_object_directory());
916d081b 1799 else
259328b7 1800 return error_errno(_("unable to create temporary file"));
aac17941
LT
1801 }
1802
0fcfd160 1803 /* Set it up */
55bb5c91 1804 git_deflate_init(&stream, zlib_compression_level);
0fcfd160 1805 stream.next_out = compressed;
9892beba 1806 stream.avail_out = sizeof(compressed);
18e2588e 1807 the_hash_algo->init_fn(&c);
a44c9a5e
LT
1808
1809 /* First header.. */
d65a16f6 1810 stream.next_in = (unsigned char *)hdr;
a44c9a5e 1811 stream.avail_in = hdrlen;
55bb5c91
JH
1812 while (git_deflate(&stream, 0) == Z_OK)
1813 ; /* nothing */
18e2588e 1814 the_hash_algo->update_fn(&c, hdr, hdrlen);
a44c9a5e
LT
1815
1816 /* Then the data itself.. */
c00e657d 1817 stream.next_in = (void *)buf;
a44c9a5e 1818 stream.avail_in = len;
9892beba 1819 do {
748af44c 1820 unsigned char *in0 = stream.next_in;
55bb5c91 1821 ret = git_deflate(&stream, Z_FINISH);
18e2588e 1822 the_hash_algo->update_fn(&c, in0, stream.next_in - in0);
9892beba 1823 if (write_buffer(fd, compressed, stream.next_out - compressed) < 0)
76011357 1824 die(_("unable to write loose object file"));
9892beba
NP
1825 stream.next_out = compressed;
1826 stream.avail_out = sizeof(compressed);
1827 } while (ret == Z_OK);
1828
ac54c277 1829 if (ret != Z_STREAM_END)
259328b7 1830 die(_("unable to deflate new object %s (%d)"), oid_to_hex(oid),
3fc7281f 1831 ret);
55bb5c91 1832 ret = git_deflate_end_gently(&stream);
ac54c277 1833 if (ret != Z_OK)
259328b7 1834 die(_("deflateEnd on object %s failed (%d)"), oid_to_hex(oid),
3fc7281f 1835 ret);
18e2588e 1836 the_hash_algo->final_fn(parano_oid.hash, &c);
9001dc2a 1837 if (!oideq(oid, &parano_oid))
259328b7 1838 die(_("confused by unstable object source data for %s"),
3fc7281f 1839 oid_to_hex(oid));
ac54c277 1840
514c5fdd 1841 close_loose_object(fd);
0fcfd160 1842
bbac7311
NP
1843 if (mtime) {
1844 struct utimbuf utb;
1845 utb.actime = mtime;
1846 utb.modtime = mtime;
d4b3d11a 1847 if (utime(tmp_file.buf, &utb) < 0)
259328b7 1848 warning_errno(_("failed utime() on %s"), tmp_file.buf);
bbac7311
NP
1849 }
1850
ea657730 1851 return finalize_object_file(tmp_file.buf, filename.buf);
0fcfd160 1852}
8237b185 1853
6862ebbf 1854static int freshen_loose_object(const struct object_id *oid)
33d4221c 1855{
6862ebbf 1856 return check_and_freshen(oid, 1);
33d4221c
JK
1857}
1858
6862ebbf 1859static int freshen_packed_object(const struct object_id *oid)
33d4221c
JK
1860{
1861 struct pack_entry e;
544443cb 1862 if (!find_pack_entry(the_repository, oid, &e))
ee1c6c34
JK
1863 return 0;
1864 if (e.p->freshened)
1865 return 1;
1866 if (!freshen_file(e.p->pack_name))
1867 return 0;
1868 e.p->freshened = 1;
1869 return 1;
33d4221c
JK
1870}
1871
a09c985e
PO
1872int write_object_file(const void *buf, unsigned long len, const char *type,
1873 struct object_id *oid)
bbac7311 1874{
1af64f73 1875 char hdr[MAX_HEADER_LEN];
ef1286d3 1876 int hdrlen = sizeof(hdr);
bbac7311
NP
1877
1878 /* Normally if we have it in the pack then we do not bother writing
1879 * it out into .git/objects/??/?{38} file.
1880 */
a09c985e 1881 write_object_file_prepare(buf, len, type, oid, hdr, &hdrlen);
6862ebbf 1882 if (freshen_packed_object(oid) || freshen_loose_object(oid))
bbac7311 1883 return 0;
3fc7281f 1884 return write_loose_object(oid, hdr, hdrlen, buf, len, 0);
bbac7311
NP
1885}
1886
1752cbbc
PO
1887int hash_object_file_literally(const void *buf, unsigned long len,
1888 const char *type, struct object_id *oid,
1889 unsigned flags)
0c3db67c
ES
1890{
1891 char *header;
1892 int hdrlen, status = 0;
1893
1894 /* type string, SP, %lu of the length plus NUL must fit this */
1af64f73 1895 hdrlen = strlen(type) + MAX_HEADER_LEN;
ef1286d3 1896 header = xmalloc(hdrlen);
a09c985e 1897 write_object_file_prepare(buf, len, type, oid, header, &hdrlen);
0c3db67c
ES
1898
1899 if (!(flags & HASH_WRITE_OBJECT))
1900 goto cleanup;
6862ebbf 1901 if (freshen_packed_object(oid) || freshen_loose_object(oid))
0c3db67c 1902 goto cleanup;
3fc7281f 1903 status = write_loose_object(oid, header, hdrlen, buf, len, 0);
0c3db67c
ES
1904
1905cleanup:
1906 free(header);
1907 return status;
1908}
1909
4bdb70a4 1910int force_object_loose(const struct object_id *oid, time_t mtime)
bbac7311 1911{
bbac7311
NP
1912 void *buf;
1913 unsigned long len;
1914 enum object_type type;
1af64f73 1915 char hdr[MAX_HEADER_LEN];
bbac7311 1916 int hdrlen;
1fb23e65 1917 int ret;
bbac7311 1918
6862ebbf 1919 if (has_loose_object(oid))
bbac7311 1920 return 0;
cba595ab 1921 buf = read_object(the_repository, oid, &type, &len);
bbac7311 1922 if (!buf)
2c319886 1923 return error(_("cannot read object for %s"), oid_to_hex(oid));
ca473cef 1924 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %"PRIuMAX , type_name(type), (uintmax_t)len) + 1;
3fc7281f 1925 ret = write_loose_object(oid, hdr, hdrlen, buf, len, mtime);
1fb23e65
BS
1926 free(buf);
1927
1928 return ret;
bbac7311
NP
1929}
1930
cba595ab
JH
1931int repo_has_object_file_with_flags(struct repository *r,
1932 const struct object_id *oid, int flags)
8237b185 1933{
3e8b7d3c
JN
1934 if (!startup_info->have_repository)
1935 return 0;
cba595ab 1936 return oid_object_info_extended(r, oid, NULL,
abef9020 1937 flags | OBJECT_INFO_SKIP_CACHED) >= 0;
b419aa25 1938}
1939
9b45f499
SB
1940int repo_has_object_file(struct repository *r,
1941 const struct object_id *oid)
b419aa25 1942{
cba595ab 1943 return repo_has_object_file_with_flags(r, oid, 0);
5827a035
JK
1944}
1945
c879daa2
NTND
1946static void check_tree(const void *buf, size_t size)
1947{
1948 struct tree_desc desc;
1949 struct name_entry entry;
1950
1951 init_tree_desc(&desc, buf, size);
1952 while (tree_entry(&desc, &entry))
1953 /* do nothing
1954 * tree_entry() will die() on malformed entries */
1955 ;
1956}
1957
1958static void check_commit(const void *buf, size_t size)
1959{
1960 struct commit c;
1961 memset(&c, 0, sizeof(c));
08f4f445 1962 if (parse_commit_buffer(the_repository, &c, buf, size, 0))
259328b7 1963 die(_("corrupt commit"));
c879daa2
NTND
1964}
1965
1966static void check_tag(const void *buf, size_t size)
1967{
1968 struct tag t;
1969 memset(&t, 0, sizeof(t));
0e740fed 1970 if (parse_tag_buffer(the_repository, &t, buf, size))
259328b7 1971 die(_("corrupt tag"));
c879daa2
NTND
1972}
1973
58bf2a4c
NTND
1974static int index_mem(struct index_state *istate,
1975 struct object_id *oid, void *buf, size_t size,
c4ce46fc
JH
1976 enum object_type type,
1977 const char *path, unsigned flags)
e7332f96 1978{
6c510bee 1979 int ret, re_allocated = 0;
c4ce46fc 1980 int write_object = flags & HASH_WRITE_OBJECT;
74400e71 1981
7672db20 1982 if (!type)
edaec3fb 1983 type = OBJ_BLOB;
6c510bee
LT
1984
1985 /*
1986 * Convert blobs to git internal format
1987 */
43df4f86 1988 if ((type == OBJ_BLOB) && path) {
f285a2d7 1989 struct strbuf nbuf = STRBUF_INIT;
58bf2a4c 1990 if (convert_to_git(istate, path, buf, size, &nbuf,
8462ff43 1991 get_conv_flags(flags))) {
b315c5c0 1992 buf = strbuf_detach(&nbuf, &size);
6c510bee
LT
1993 re_allocated = 1;
1994 }
1995 }
c4ce46fc 1996 if (flags & HASH_FORMAT_CHECK) {
c879daa2
NTND
1997 if (type == OBJ_TREE)
1998 check_tree(buf, size);
1999 if (type == OBJ_COMMIT)
2000 check_commit(buf, size);
2001 if (type == OBJ_TAG)
2002 check_tag(buf, size);
2003 }
6c510bee 2004
7672db20 2005 if (write_object)
169c9c01 2006 ret = write_object_file(buf, size, type_name(type), oid);
abdc3fc8 2007 else
169c9c01 2008 ret = hash_object_file(buf, size, type_name(type), oid);
43df4f86 2009 if (re_allocated)
6c510bee 2010 free(buf);
43df4f86
DP
2011 return ret;
2012}
2013
58bf2a4c
NTND
2014static int index_stream_convert_blob(struct index_state *istate,
2015 struct object_id *oid,
2016 int fd,
2017 const char *path,
2018 unsigned flags)
9035d75a
SP
2019{
2020 int ret;
2021 const int write_object = flags & HASH_WRITE_OBJECT;
2022 struct strbuf sbuf = STRBUF_INIT;
2023
2024 assert(path);
58bf2a4c 2025 assert(would_convert_to_git_filter_fd(istate, path));
9035d75a 2026
58bf2a4c 2027 convert_to_git_filter_fd(istate, path, fd, &sbuf,
8462ff43 2028 get_conv_flags(flags));
9035d75a
SP
2029
2030 if (write_object)
169c9c01 2031 ret = write_object_file(sbuf.buf, sbuf.len, type_name(OBJ_BLOB),
a09c985e 2032 oid);
9035d75a 2033 else
169c9c01 2034 ret = hash_object_file(sbuf.buf, sbuf.len, type_name(OBJ_BLOB),
f070facc 2035 oid);
9035d75a
SP
2036 strbuf_release(&sbuf);
2037 return ret;
2038}
2039
58bf2a4c
NTND
2040static int index_pipe(struct index_state *istate, struct object_id *oid,
2041 int fd, enum object_type type,
7b41e1e1
JH
2042 const char *path, unsigned flags)
2043{
2044 struct strbuf sbuf = STRBUF_INIT;
2045 int ret;
2046
2047 if (strbuf_read(&sbuf, fd, 4096) >= 0)
58bf2a4c 2048 ret = index_mem(istate, oid, sbuf.buf, sbuf.len, type, path, flags);
7b41e1e1
JH
2049 else
2050 ret = -1;
2051 strbuf_release(&sbuf);
2052 return ret;
2053}
2054
ea68b0ce
DP
2055#define SMALL_FILE_SIZE (32*1024)
2056
58bf2a4c
NTND
2057static int index_core(struct index_state *istate,
2058 struct object_id *oid, int fd, size_t size,
7b41e1e1
JH
2059 enum object_type type, const char *path,
2060 unsigned flags)
43df4f86
DP
2061{
2062 int ret;
43df4f86 2063
7b41e1e1 2064 if (!size) {
58bf2a4c 2065 ret = index_mem(istate, oid, "", size, type, path, flags);
ea68b0ce
DP
2066 } else if (size <= SMALL_FILE_SIZE) {
2067 char *buf = xmalloc(size);
90dca671
JK
2068 ssize_t read_result = read_in_full(fd, buf, size);
2069 if (read_result < 0)
259328b7 2070 ret = error_errno(_("read error while indexing %s"),
90dca671
JK
2071 path ? path : "<unknown>");
2072 else if (read_result != size)
259328b7 2073 ret = error(_("short read while indexing %s"),
90dca671 2074 path ? path : "<unknown>");
ea68b0ce 2075 else
58bf2a4c 2076 ret = index_mem(istate, oid, buf, size, type, path, flags);
ea68b0ce 2077 free(buf);
08bda208 2078 } else {
43df4f86 2079 void *buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
58bf2a4c 2080 ret = index_mem(istate, oid, buf, size, type, path, flags);
aac17941 2081 munmap(buf, size);
08bda208 2082 }
7b41e1e1
JH
2083 return ret;
2084}
2085
4dd1fbc7 2086/*
568508e7
JH
2087 * This creates one packfile per large blob unless bulk-checkin
2088 * machinery is "plugged".
4dd1fbc7
JH
2089 *
2090 * This also bypasses the usual "convert-to-git" dance, and that is on
2091 * purpose. We could write a streaming version of the converting
2092 * functions and insert that before feeding the data to fast-import
4f22b101
JK
2093 * (or equivalent in-core API described above). However, that is
2094 * somewhat complicated, as we do not know the size of the filter
2095 * result, which we need to know beforehand when writing a git object.
2096 * Since the primary motivation for trying to stream from the working
2097 * tree file and to avoid mmaping it in core is to deal with large
2098 * binary blobs, they generally do not want to get any conversion, and
2099 * callers should avoid this code path when filters are requested.
4dd1fbc7 2100 */
7d5e1dc3 2101static int index_stream(struct object_id *oid, int fd, size_t size,
4dd1fbc7
JH
2102 enum object_type type, const char *path,
2103 unsigned flags)
2104{
68ee6dfc 2105 return index_bulk_checkin(oid, fd, size, type, path, flags);
4dd1fbc7
JH
2106}
2107
58bf2a4c
NTND
2108int index_fd(struct index_state *istate, struct object_id *oid,
2109 int fd, struct stat *st,
7b41e1e1
JH
2110 enum object_type type, const char *path, unsigned flags)
2111{
2112 int ret;
7b41e1e1 2113
9079ab7c
SP
2114 /*
2115 * Call xsize_t() only when needed to avoid potentially unnecessary
2116 * die() for large files.
2117 */
58bf2a4c
NTND
2118 if (type == OBJ_BLOB && path && would_convert_to_git_filter_fd(istate, path))
2119 ret = index_stream_convert_blob(istate, oid, fd, path, flags);
9035d75a 2120 else if (!S_ISREG(st->st_mode))
58bf2a4c 2121 ret = index_pipe(istate, oid, fd, type, path, flags);
9079ab7c 2122 else if (st->st_size <= big_file_threshold || type != OBJ_BLOB ||
58bf2a4c
NTND
2123 (path && would_convert_to_git(istate, path)))
2124 ret = index_core(istate, oid, fd, xsize_t(st->st_size),
2125 type, path, flags);
4dd1fbc7 2126 else
7d5e1dc3 2127 ret = index_stream(oid, fd, xsize_t(st->st_size), type, path,
9079ab7c 2128 flags);
43df4f86 2129 close(fd);
aac17941 2130 return ret;
74400e71 2131}
ec1fcc16 2132
58bf2a4c
NTND
2133int index_path(struct index_state *istate, struct object_id *oid,
2134 const char *path, struct stat *st, unsigned flags)
ec1fcc16
JH
2135{
2136 int fd;
b760d3aa 2137 struct strbuf sb = STRBUF_INIT;
ea8e0297 2138 int rc = 0;
ec1fcc16
JH
2139
2140 switch (st->st_mode & S_IFMT) {
2141 case S_IFREG:
2142 fd = open(path, O_RDONLY);
2143 if (fd < 0)
7616c6ca 2144 return error_errno("open(\"%s\")", path);
58bf2a4c 2145 if (index_fd(istate, oid, fd, st, OBJ_BLOB, path, flags) < 0)
259328b7 2146 return error(_("%s: failed to insert into database"),
ec1fcc16
JH
2147 path);
2148 break;
2149 case S_IFLNK:
7616c6ca
NTND
2150 if (strbuf_readlink(&sb, path, st->st_size))
2151 return error_errno("readlink(\"%s\")", path);
c4ce46fc 2152 if (!(flags & HASH_WRITE_OBJECT))
f070facc 2153 hash_object_file(sb.buf, sb.len, blob_type, oid);
a09c985e 2154 else if (write_object_file(sb.buf, sb.len, blob_type, oid))
259328b7 2155 rc = error(_("%s: failed to insert into database"), path);
b760d3aa 2156 strbuf_release(&sb);
ec1fcc16 2157 break;
f35a6d3b 2158 case S_IFDIR:
a98e6101 2159 return resolve_gitlink_ref(path, "HEAD", oid);
ec1fcc16 2160 default:
259328b7 2161 return error(_("%s: unsupported file type"), path);
ec1fcc16 2162 }
ea8e0297 2163 return rc;
ec1fcc16 2164}
a69e5429
JH
2165
2166int read_pack_header(int fd, struct pack_header *header)
2167{
f48ecd38 2168 if (read_in_full(fd, header, sizeof(*header)) != sizeof(*header))
c697ad14
HO
2169 /* "eof before pack header was fully read" */
2170 return PH_ERROR_EOF;
2171
a69e5429
JH
2172 if (header->hdr_signature != htonl(PACK_SIGNATURE))
2173 /* "protocol error (pack signature mismatch detected)" */
2174 return PH_ERROR_PACK_SIGNATURE;
2175 if (!pack_version_ok(header->hdr_version))
2176 /* "protocol error (pack version unsupported)" */
2177 return PH_ERROR_PROTOCOL;
2178 return 0;
2179}
40d52ff7 2180
e816caa0 2181void assert_oid_type(const struct object_id *oid, enum object_type expect)
40d52ff7 2182{
0df8e965 2183 enum object_type type = oid_object_info(the_repository, oid, NULL);
40d52ff7 2184 if (type < 0)
259328b7 2185 die(_("%s is not a valid object"), oid_to_hex(oid));
40d52ff7 2186 if (type != expect)
259328b7 2187 die(_("%s is not a valid '%s' object"), oid_to_hex(oid),
debca9d2 2188 type_name(expect));
40d52ff7 2189}
27e1e22d 2190
70c49050 2191int for_each_file_in_obj_subdir(unsigned int subdir_nr,
cc817ca3
RS
2192 struct strbuf *path,
2193 each_loose_object_fn obj_cb,
2194 each_loose_cruft_fn cruft_cb,
2195 each_loose_subdir_fn subdir_cb,
2196 void *data)
27e1e22d 2197{
0375f472
RS
2198 size_t origlen, baselen;
2199 DIR *dir;
27e1e22d
JK
2200 struct dirent *de;
2201 int r = 0;
62a24c89 2202 struct object_id oid;
27e1e22d 2203
70c49050
RS
2204 if (subdir_nr > 0xff)
2205 BUG("invalid loose object subdirectory: %x", subdir_nr);
2206
0375f472
RS
2207 origlen = path->len;
2208 strbuf_complete(path, '/');
2209 strbuf_addf(path, "%02x", subdir_nr);
0375f472
RS
2210
2211 dir = opendir(path->buf);
27e1e22d 2212 if (!dir) {
0375f472 2213 if (errno != ENOENT)
259328b7 2214 r = error_errno(_("unable to open %s"), path->buf);
0375f472
RS
2215 strbuf_setlen(path, origlen);
2216 return r;
27e1e22d
JK
2217 }
2218
62a24c89 2219 oid.hash[0] = subdir_nr;
163ee5e6
DS
2220 strbuf_addch(path, '/');
2221 baselen = path->len;
62a24c89 2222
27e1e22d 2223 while ((de = readdir(dir))) {
163ee5e6 2224 size_t namelen;
27e1e22d
JK
2225 if (is_dot_or_dotdot(de->d_name))
2226 continue;
2227
163ee5e6 2228 namelen = strlen(de->d_name);
27e1e22d 2229 strbuf_setlen(path, baselen);
163ee5e6 2230 strbuf_add(path, de->d_name, namelen);
94b5e093 2231 if (namelen == the_hash_algo->hexsz - 2 &&
62a24c89 2232 !hex_to_bytes(oid.hash + 1, de->d_name,
94b5e093 2233 the_hash_algo->rawsz - 1)) {
62a24c89
RS
2234 if (obj_cb) {
2235 r = obj_cb(&oid, path->buf, data);
2236 if (r)
2237 break;
27e1e22d 2238 }
62a24c89 2239 continue;
27e1e22d
JK
2240 }
2241
2242 if (cruft_cb) {
2243 r = cruft_cb(de->d_name, path->buf, data);
2244 if (r)
2245 break;
2246 }
2247 }
094c7e63 2248 closedir(dir);
27e1e22d 2249
163ee5e6 2250 strbuf_setlen(path, baselen - 1);
27e1e22d
JK
2251 if (!r && subdir_cb)
2252 r = subdir_cb(subdir_nr, path->buf, data);
2253
0375f472
RS
2254 strbuf_setlen(path, origlen);
2255
27e1e22d
JK
2256 return r;
2257}
2258
e6f875e0 2259int for_each_loose_file_in_objdir_buf(struct strbuf *path,
27e1e22d
JK
2260 each_loose_object_fn obj_cb,
2261 each_loose_cruft_fn cruft_cb,
2262 each_loose_subdir_fn subdir_cb,
2263 void *data)
2264{
27e1e22d
JK
2265 int r = 0;
2266 int i;
2267
27e1e22d 2268 for (i = 0; i < 256; i++) {
e6f875e0 2269 r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb,
27e1e22d 2270 subdir_cb, data);
27e1e22d
JK
2271 if (r)
2272 break;
2273 }
2274
e6f875e0
JK
2275 return r;
2276}
2277
2278int for_each_loose_file_in_objdir(const char *path,
2279 each_loose_object_fn obj_cb,
2280 each_loose_cruft_fn cruft_cb,
2281 each_loose_subdir_fn subdir_cb,
2282 void *data)
2283{
2284 struct strbuf buf = STRBUF_INIT;
2285 int r;
2286
2287 strbuf_addstr(&buf, path);
2288 r = for_each_loose_file_in_objdir_buf(&buf, obj_cb, cruft_cb,
2289 subdir_cb, data);
27e1e22d 2290 strbuf_release(&buf);
e6f875e0 2291
27e1e22d
JK
2292 return r;
2293}
660c889e 2294
a7ff6f5a
JK
2295int for_each_loose_object(each_loose_object_fn cb, void *data,
2296 enum for_each_object_flags flags)
660c889e 2297{
f0eaf638 2298 struct object_directory *odb;
b0a42642 2299
f0eaf638
JK
2300 prepare_alt_odb(the_repository);
2301 for (odb = the_repository->objects->odb; odb; odb = odb->next) {
2302 int r = for_each_loose_file_in_objdir(odb->path, cb, NULL,
2303 NULL, data);
2304 if (r)
2305 return r;
660c889e 2306
f0eaf638
JK
2307 if (flags & FOR_EACH_OBJECT_LOCAL_ONLY)
2308 break;
2309 }
1385bb7b 2310
f0eaf638 2311 return 0;
660c889e
JK
2312}
2313
3a2e0824
JK
2314static int append_loose_object(const struct object_id *oid, const char *path,
2315 void *data)
660c889e 2316{
3a2e0824
JK
2317 oid_array_append(data, oid);
2318 return 0;
2319}
660c889e 2320
0000d654
RS
2321struct oid_array *odb_loose_cache(struct object_directory *odb,
2322 const struct object_id *oid)
2323{
2324 int subdir_nr = oid->hash[0];
3a2e0824 2325 struct strbuf buf = STRBUF_INIT;
660c889e 2326
3a2e0824
JK
2327 if (subdir_nr < 0 ||
2328 subdir_nr >= ARRAY_SIZE(odb->loose_objects_subdir_seen))
2329 BUG("subdir_nr out of range");
1385bb7b 2330
3a2e0824 2331 if (odb->loose_objects_subdir_seen[subdir_nr])
8be88dbc 2332 return &odb->loose_objects_cache[subdir_nr];
3a2e0824
JK
2333
2334 strbuf_addstr(&buf, odb->path);
2335 for_each_file_in_obj_subdir(subdir_nr, &buf,
2336 append_loose_object,
2337 NULL, NULL,
4cea1ce0 2338 &odb->loose_objects_cache[subdir_nr]);
3a2e0824 2339 odb->loose_objects_subdir_seen[subdir_nr] = 1;
7317aa71 2340 strbuf_release(&buf);
8be88dbc 2341 return &odb->loose_objects_cache[subdir_nr];
660c889e
JK
2342}
2343
d4e19e51
RS
2344void odb_clear_loose_cache(struct object_directory *odb)
2345{
4cea1ce0
RS
2346 int i;
2347
2348 for (i = 0; i < ARRAY_SIZE(odb->loose_objects_cache); i++)
2349 oid_array_clear(&odb->loose_objects_cache[i]);
d4e19e51
RS
2350 memset(&odb->loose_objects_subdir_seen, 0,
2351 sizeof(odb->loose_objects_subdir_seen));
2352}
2353
00a7760e
JK
2354static int check_stream_oid(git_zstream *stream,
2355 const char *hdr,
2356 unsigned long size,
2357 const char *path,
2358 const struct object_id *expected_oid)
f6371f92 2359{
18e2588e 2360 git_hash_ctx c;
00a7760e 2361 struct object_id real_oid;
f6371f92
JK
2362 unsigned char buf[4096];
2363 unsigned long total_read;
2364 int status = Z_OK;
2365
18e2588e 2366 the_hash_algo->init_fn(&c);
2367 the_hash_algo->update_fn(&c, hdr, stream->total_out);
f6371f92
JK
2368
2369 /*
2370 * We already read some bytes into hdr, but the ones up to the NUL
2371 * do not count against the object's content size.
2372 */
2373 total_read = stream->total_out - strlen(hdr) - 1;
2374
2375 /*
2376 * This size comparison must be "<=" to read the final zlib packets;
00a7760e 2377 * see the comment in unpack_loose_rest for details.
f6371f92
JK
2378 */
2379 while (total_read <= size &&
18ad13e5
JH
2380 (status == Z_OK ||
2381 (status == Z_BUF_ERROR && !stream->avail_out))) {
f6371f92
JK
2382 stream->next_out = buf;
2383 stream->avail_out = sizeof(buf);
2384 if (size - total_read < stream->avail_out)
2385 stream->avail_out = size - total_read;
2386 status = git_inflate(stream, Z_FINISH);
18e2588e 2387 the_hash_algo->update_fn(&c, buf, stream->next_out - buf);
f6371f92
JK
2388 total_read += stream->next_out - buf;
2389 }
2390 git_inflate_end(stream);
2391
2392 if (status != Z_STREAM_END) {
00a7760e 2393 error(_("corrupt loose object '%s'"), oid_to_hex(expected_oid));
f6371f92
JK
2394 return -1;
2395 }
cce044df 2396 if (stream->avail_in) {
259328b7 2397 error(_("garbage at end of loose object '%s'"),
00a7760e 2398 oid_to_hex(expected_oid));
cce044df
JK
2399 return -1;
2400 }
f6371f92 2401
00a7760e
JK
2402 the_hash_algo->final_fn(real_oid.hash, &c);
2403 if (!oideq(expected_oid, &real_oid)) {
01f8d594 2404 error(_("hash mismatch for %s (expected %s)"), path,
00a7760e 2405 oid_to_hex(expected_oid));
f6371f92
JK
2406 return -1;
2407 }
2408
2409 return 0;
2410}
2411
2412int read_loose_object(const char *path,
d61d87bd 2413 const struct object_id *expected_oid,
f6371f92
JK
2414 enum object_type *type,
2415 unsigned long *size,
2416 void **contents)
2417{
2418 int ret = -1;
f6371f92
JK
2419 void *map = NULL;
2420 unsigned long mapsize;
2421 git_zstream stream;
1af64f73 2422 char hdr[MAX_HEADER_LEN];
f6371f92
JK
2423
2424 *contents = NULL;
2425
514c5fdd 2426 map = map_loose_object_1(the_repository, path, NULL, &mapsize);
f6371f92 2427 if (!map) {
259328b7 2428 error_errno(_("unable to mmap %s"), path);
f6371f92
JK
2429 goto out;
2430 }
2431
00a7760e 2432 if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0) {
259328b7 2433 error(_("unable to unpack header of %s"), path);
f6371f92
JK
2434 goto out;
2435 }
2436
00a7760e 2437 *type = parse_loose_header(hdr, size);
f6371f92 2438 if (*type < 0) {
259328b7 2439 error(_("unable to parse header of %s"), path);
f6371f92
JK
2440 git_inflate_end(&stream);
2441 goto out;
2442 }
2443
7ac4f3a0 2444 if (*type == OBJ_BLOB && *size > big_file_threshold) {
00a7760e 2445 if (check_stream_oid(&stream, hdr, *size, path, expected_oid) < 0)
f6371f92
JK
2446 goto out;
2447 } else {
00a7760e 2448 *contents = unpack_loose_rest(&stream, hdr, *size, expected_oid);
f6371f92 2449 if (!*contents) {
259328b7 2450 error(_("unable to unpack contents of %s"), path);
f6371f92
JK
2451 git_inflate_end(&stream);
2452 goto out;
2453 }
17e65451 2454 if (check_object_signature(expected_oid, *contents,
debca9d2 2455 *size, type_name(*type))) {
01f8d594 2456 error(_("hash mismatch for %s (expected %s)"), path,
d61d87bd 2457 oid_to_hex(expected_oid));
f6371f92
JK
2458 free(*contents);
2459 goto out;
2460 }
2461 }
2462
2463 ret = 0; /* everything checks out */
2464
2465out:
2466 if (map)
2467 munmap(map, mapsize);
f6371f92
JK
2468 return ret;
2469}