2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
6 * This handles basic git sha1 object files - packing, unpacking,
11 #include "string-list.h"
17 #include "run-command.h"
20 #include "tree-walk.h"
22 #include "pack-revindex.h"
23 #include "sha1-lookup.h"
24 #include "bulk-checkin.h"
25 #include "streaming.h"
29 #include "mergesort.h"
32 #include "fetch-object.h"
34 const unsigned char null_sha1
[GIT_MAX_RAWSZ
];
35 const struct object_id null_oid
;
36 const struct object_id empty_tree_oid
= {
37 EMPTY_TREE_SHA1_BIN_LITERAL
39 const struct object_id empty_blob_oid
= {
40 EMPTY_BLOB_SHA1_BIN_LITERAL
43 static void git_hash_sha1_init(void *ctx
)
45 git_SHA1_Init((git_SHA_CTX
*)ctx
);
48 static void git_hash_sha1_update(void *ctx
, const void *data
, size_t len
)
50 git_SHA1_Update((git_SHA_CTX
*)ctx
, data
, len
);
53 static void git_hash_sha1_final(unsigned char *hash
, void *ctx
)
55 git_SHA1_Final(hash
, (git_SHA_CTX
*)ctx
);
58 static void git_hash_unknown_init(void *ctx
)
60 die("trying to init unknown hash");
63 static void git_hash_unknown_update(void *ctx
, const void *data
, size_t len
)
65 die("trying to update unknown hash");
68 static void git_hash_unknown_final(unsigned char *hash
, void *ctx
)
70 die("trying to finalize unknown hash");
73 const struct git_hash_algo hash_algos
[GIT_HASH_NALGOS
] = {
80 git_hash_unknown_init
,
81 git_hash_unknown_update
,
82 git_hash_unknown_final
,
88 /* "sha1", big-endian */
102 * This is meant to hold a *small* number of objects that you would
103 * want read_sha1_file() to be able to return, but yet you do not want
104 * to write them into the object store (e.g. a browse-only
107 static struct cached_object
{
108 unsigned char sha1
[20];
109 enum object_type type
;
113 static int cached_object_nr
, cached_object_alloc
;
115 static struct cached_object empty_tree
= {
116 EMPTY_TREE_SHA1_BIN_LITERAL
,
122 static struct cached_object
*find_cached_object(const unsigned char *sha1
)
125 struct cached_object
*co
= cached_objects
;
127 for (i
= 0; i
< cached_object_nr
; i
++, co
++) {
128 if (!hashcmp(co
->sha1
, sha1
))
131 if (!hashcmp(sha1
, empty_tree
.sha1
))
137 static enum safe_crlf
get_safe_crlf(unsigned flags
)
139 if (flags
& HASH_RENORMALIZE
)
140 return SAFE_CRLF_RENORMALIZE
;
141 else if (flags
& HASH_WRITE_OBJECT
)
144 return SAFE_CRLF_FALSE
;
148 int mkdir_in_gitdir(const char *path
)
150 if (mkdir(path
, 0777)) {
151 int saved_errno
= errno
;
153 struct strbuf sb
= STRBUF_INIT
;
158 * Are we looking at a path in a symlinked worktree
159 * whose original repository does not yet have it?
160 * e.g. .git/rr-cache pointing at its original
161 * repository in which the user hasn't performed any
162 * conflict resolution yet?
164 if (lstat(path
, &st
) || !S_ISLNK(st
.st_mode
) ||
165 strbuf_readlink(&sb
, path
, st
.st_size
) ||
166 !is_absolute_path(sb
.buf
) ||
167 mkdir(sb
.buf
, 0777)) {
174 return adjust_shared_perm(path
);
177 enum scld_error
safe_create_leading_directories(char *path
)
179 char *next_component
= path
+ offset_1st_component(path
);
180 enum scld_error ret
= SCLD_OK
;
182 while (ret
== SCLD_OK
&& next_component
) {
184 char *slash
= next_component
, slash_character
;
186 while (*slash
&& !is_dir_sep(*slash
))
192 next_component
= slash
+ 1;
193 while (is_dir_sep(*next_component
))
195 if (!*next_component
)
198 slash_character
= *slash
;
200 if (!stat(path
, &st
)) {
202 if (!S_ISDIR(st
.st_mode
)) {
206 } else if (mkdir(path
, 0777)) {
207 if (errno
== EEXIST
&&
208 !stat(path
, &st
) && S_ISDIR(st
.st_mode
))
209 ; /* somebody created it since we checked */
210 else if (errno
== ENOENT
)
212 * Either mkdir() failed because
213 * somebody just pruned the containing
214 * directory, or stat() failed because
215 * the file that was in our way was
216 * just removed. Either way, inform
217 * the caller that it might be worth
223 } else if (adjust_shared_perm(path
)) {
226 *slash
= slash_character
;
231 enum scld_error
safe_create_leading_directories_const(const char *path
)
234 /* path points to cache entries, so xstrdup before messing with it */
235 char *buf
= xstrdup(path
);
236 enum scld_error result
= safe_create_leading_directories(buf
);
244 int raceproof_create_file(const char *path
, create_file_fn fn
, void *cb
)
247 * The number of times we will try to remove empty directories
248 * in the way of path. This is only 1 because if another
249 * process is racily creating directories that conflict with
250 * us, we don't want to fight against them.
252 int remove_directories_remaining
= 1;
255 * The number of times that we will try to create the
256 * directories containing path. We are willing to attempt this
257 * more than once, because another process could be trying to
258 * clean up empty directories at the same time as we are
259 * trying to create them.
261 int create_directories_remaining
= 3;
263 /* A scratch copy of path, filled lazily if we need it: */
264 struct strbuf path_copy
= STRBUF_INIT
;
277 if (errno
== EISDIR
&& remove_directories_remaining
-- > 0) {
279 * A directory is in the way. Maybe it is empty; try
283 strbuf_addstr(&path_copy
, path
);
285 if (!remove_dir_recursively(&path_copy
, REMOVE_DIR_EMPTY_ONLY
))
287 } else if (errno
== ENOENT
&& create_directories_remaining
-- > 0) {
289 * Maybe the containing directory didn't exist, or
290 * maybe it was just deleted by a process that is
291 * racing with us to clean up empty directories. Try
294 enum scld_error scld_result
;
297 strbuf_addstr(&path_copy
, path
);
300 scld_result
= safe_create_leading_directories(path_copy
.buf
);
301 if (scld_result
== SCLD_OK
)
303 } while (scld_result
== SCLD_VANISHED
&& create_directories_remaining
-- > 0);
307 strbuf_release(&path_copy
);
312 static void fill_sha1_path(struct strbuf
*buf
, const unsigned char *sha1
)
315 for (i
= 0; i
< 20; i
++) {
316 static char hex
[] = "0123456789abcdef";
317 unsigned int val
= sha1
[i
];
318 strbuf_addch(buf
, hex
[val
>> 4]);
319 strbuf_addch(buf
, hex
[val
& 0xf]);
321 strbuf_addch(buf
, '/');
325 const char *sha1_file_name(const unsigned char *sha1
)
327 static struct strbuf buf
= STRBUF_INIT
;
330 strbuf_addf(&buf
, "%s/", get_object_directory());
332 fill_sha1_path(&buf
, sha1
);
336 struct strbuf
*alt_scratch_buf(struct alternate_object_database
*alt
)
338 strbuf_setlen(&alt
->scratch
, alt
->base_len
);
339 return &alt
->scratch
;
342 static const char *alt_sha1_path(struct alternate_object_database
*alt
,
343 const unsigned char *sha1
)
345 struct strbuf
*buf
= alt_scratch_buf(alt
);
346 fill_sha1_path(buf
, sha1
);
350 struct alternate_object_database
*alt_odb_list
;
351 static struct alternate_object_database
**alt_odb_tail
;
354 * Return non-zero iff the path is usable as an alternate object database.
356 static int alt_odb_usable(struct strbuf
*path
, const char *normalized_objdir
)
358 struct alternate_object_database
*alt
;
360 /* Detect cases where alternate disappeared */
361 if (!is_directory(path
->buf
)) {
362 error("object directory %s does not exist; "
363 "check .git/objects/info/alternates.",
369 * Prevent the common mistake of listing the same
370 * thing twice, or object directory itself.
372 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
373 if (!fspathcmp(path
->buf
, alt
->path
))
376 if (!fspathcmp(path
->buf
, normalized_objdir
))
383 * Prepare alternate object database registry.
385 * The variable alt_odb_list points at the list of struct
386 * alternate_object_database. The elements on this list come from
387 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
388 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
389 * whose contents is similar to that environment variable but can be
390 * LF separated. Its base points at a statically allocated buffer that
391 * contains "/the/directory/corresponding/to/.git/objects/...", while
392 * its name points just after the slash at the end of ".git/objects/"
393 * in the example above, and has enough space to hold 40-byte hex
394 * SHA1, an extra slash for the first level indirection, and the
397 static void read_info_alternates(const char * relative_base
, int depth
);
398 static int link_alt_odb_entry(const char *entry
, const char *relative_base
,
399 int depth
, const char *normalized_objdir
)
401 struct alternate_object_database
*ent
;
402 struct strbuf pathbuf
= STRBUF_INIT
;
404 if (!is_absolute_path(entry
) && relative_base
) {
405 strbuf_realpath(&pathbuf
, relative_base
, 1);
406 strbuf_addch(&pathbuf
, '/');
408 strbuf_addstr(&pathbuf
, entry
);
410 if (strbuf_normalize_path(&pathbuf
) < 0 && relative_base
) {
411 error("unable to normalize alternate object path: %s",
413 strbuf_release(&pathbuf
);
418 * The trailing slash after the directory name is given by
419 * this function at the end. Remove duplicates.
421 while (pathbuf
.len
&& pathbuf
.buf
[pathbuf
.len
- 1] == '/')
422 strbuf_setlen(&pathbuf
, pathbuf
.len
- 1);
424 if (!alt_odb_usable(&pathbuf
, normalized_objdir
)) {
425 strbuf_release(&pathbuf
);
429 ent
= alloc_alt_odb(pathbuf
.buf
);
431 /* add the alternate entry */
433 alt_odb_tail
= &(ent
->next
);
436 /* recursively add alternates */
437 read_info_alternates(pathbuf
.buf
, depth
+ 1);
439 strbuf_release(&pathbuf
);
443 static const char *parse_alt_odb_entry(const char *string
,
451 if (*string
== '#') {
452 /* comment; consume up to next separator */
453 end
= strchrnul(string
, sep
);
454 } else if (*string
== '"' && !unquote_c_style(out
, string
, &end
)) {
456 * quoted path; unquote_c_style has copied the
457 * data for us and set "end". Broken quoting (e.g.,
458 * an entry that doesn't end with a quote) falls
459 * back to the unquoted case below.
462 /* normal, unquoted path */
463 end
= strchrnul(string
, sep
);
464 strbuf_add(out
, string
, end
- string
);
472 static void link_alt_odb_entries(const char *alt
, int sep
,
473 const char *relative_base
, int depth
)
475 struct strbuf objdirbuf
= STRBUF_INIT
;
476 struct strbuf entry
= STRBUF_INIT
;
482 error("%s: ignoring alternate object stores, nesting too deep.",
487 strbuf_add_absolute_path(&objdirbuf
, get_object_directory());
488 if (strbuf_normalize_path(&objdirbuf
) < 0)
489 die("unable to normalize object directory: %s",
493 alt
= parse_alt_odb_entry(alt
, sep
, &entry
);
496 link_alt_odb_entry(entry
.buf
, relative_base
, depth
, objdirbuf
.buf
);
498 strbuf_release(&entry
);
499 strbuf_release(&objdirbuf
);
502 static void read_info_alternates(const char * relative_base
, int depth
)
505 struct strbuf buf
= STRBUF_INIT
;
507 path
= xstrfmt("%s/info/alternates", relative_base
);
508 if (strbuf_read_file(&buf
, path
, 1024) < 0) {
509 warn_on_fopen_errors(path
);
514 link_alt_odb_entries(buf
.buf
, '\n', relative_base
, depth
);
515 strbuf_release(&buf
);
519 struct alternate_object_database
*alloc_alt_odb(const char *dir
)
521 struct alternate_object_database
*ent
;
523 FLEX_ALLOC_STR(ent
, path
, dir
);
524 strbuf_init(&ent
->scratch
, 0);
525 strbuf_addf(&ent
->scratch
, "%s/", dir
);
526 ent
->base_len
= ent
->scratch
.len
;
531 void add_to_alternates_file(const char *reference
)
533 struct lock_file lock
= LOCK_INIT
;
534 char *alts
= git_pathdup("objects/info/alternates");
538 hold_lock_file_for_update(&lock
, alts
, LOCK_DIE_ON_ERROR
);
539 out
= fdopen_lock_file(&lock
, "w");
541 die_errno("unable to fdopen alternates lockfile");
543 in
= fopen(alts
, "r");
545 struct strbuf line
= STRBUF_INIT
;
547 while (strbuf_getline(&line
, in
) != EOF
) {
548 if (!strcmp(reference
, line
.buf
)) {
552 fprintf_or_die(out
, "%s\n", line
.buf
);
555 strbuf_release(&line
);
558 else if (errno
!= ENOENT
)
559 die_errno("unable to read alternates file");
562 rollback_lock_file(&lock
);
564 fprintf_or_die(out
, "%s\n", reference
);
565 if (commit_lock_file(&lock
))
566 die_errno("unable to move new alternates file into place");
568 link_alt_odb_entries(reference
, '\n', NULL
, 0);
573 void add_to_alternates_memory(const char *reference
)
576 * Make sure alternates are initialized, or else our entry may be
577 * overwritten when they are.
581 link_alt_odb_entries(reference
, '\n', NULL
, 0);
585 * Compute the exact path an alternate is at and returns it. In case of
586 * error NULL is returned and the human readable error is added to `err`
587 * `path` may be relative and should point to $GITDIR.
588 * `err` must not be null.
590 char *compute_alternate_path(const char *path
, struct strbuf
*err
)
592 char *ref_git
= NULL
;
593 const char *repo
, *ref_git_s
;
596 ref_git_s
= real_path_if_valid(path
);
599 strbuf_addf(err
, _("path '%s' does not exist"), path
);
603 * Beware: read_gitfile(), real_path() and mkpath()
604 * return static buffer
606 ref_git
= xstrdup(ref_git_s
);
608 repo
= read_gitfile(ref_git
);
610 repo
= read_gitfile(mkpath("%s/.git", ref_git
));
613 ref_git
= xstrdup(repo
);
616 if (!repo
&& is_directory(mkpath("%s/.git/objects", ref_git
))) {
617 char *ref_git_git
= mkpathdup("%s/.git", ref_git
);
619 ref_git
= ref_git_git
;
620 } else if (!is_directory(mkpath("%s/objects", ref_git
))) {
621 struct strbuf sb
= STRBUF_INIT
;
623 if (get_common_dir(&sb
, ref_git
)) {
625 _("reference repository '%s' as a linked "
626 "checkout is not supported yet."),
631 strbuf_addf(err
, _("reference repository '%s' is not a "
632 "local repository."), path
);
636 if (!access(mkpath("%s/shallow", ref_git
), F_OK
)) {
637 strbuf_addf(err
, _("reference repository '%s' is shallow"),
643 if (!access(mkpath("%s/info/grafts", ref_git
), F_OK
)) {
645 _("reference repository '%s' is grafted"),
653 FREE_AND_NULL(ref_git
);
659 int foreach_alt_odb(alt_odb_fn fn
, void *cb
)
661 struct alternate_object_database
*ent
;
665 for (ent
= alt_odb_list
; ent
; ent
= ent
->next
) {
673 void prepare_alt_odb(void)
680 alt
= getenv(ALTERNATE_DB_ENVIRONMENT
);
682 alt_odb_tail
= &alt_odb_list
;
683 link_alt_odb_entries(alt
, PATH_SEP
, NULL
, 0);
685 read_info_alternates(get_object_directory(), 0);
688 /* Returns 1 if we have successfully freshened the file, 0 otherwise. */
689 static int freshen_file(const char *fn
)
692 t
.actime
= t
.modtime
= time(NULL
);
693 return !utime(fn
, &t
);
697 * All of the check_and_freshen functions return 1 if the file exists and was
698 * freshened (if freshening was requested), 0 otherwise. If they return
699 * 0, you should not assume that it is safe to skip a write of the object (it
700 * either does not exist on disk, or has a stale mtime and may be subject to
703 int check_and_freshen_file(const char *fn
, int freshen
)
705 if (access(fn
, F_OK
))
707 if (freshen
&& !freshen_file(fn
))
712 static int check_and_freshen_local(const unsigned char *sha1
, int freshen
)
714 return check_and_freshen_file(sha1_file_name(sha1
), freshen
);
717 static int check_and_freshen_nonlocal(const unsigned char *sha1
, int freshen
)
719 struct alternate_object_database
*alt
;
721 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
722 const char *path
= alt_sha1_path(alt
, sha1
);
723 if (check_and_freshen_file(path
, freshen
))
729 static int check_and_freshen(const unsigned char *sha1
, int freshen
)
731 return check_and_freshen_local(sha1
, freshen
) ||
732 check_and_freshen_nonlocal(sha1
, freshen
);
735 int has_loose_object_nonlocal(const unsigned char *sha1
)
737 return check_and_freshen_nonlocal(sha1
, 0);
740 static int has_loose_object(const unsigned char *sha1
)
742 return check_and_freshen(sha1
, 0);
745 static void mmap_limit_check(size_t length
)
747 static size_t limit
= 0;
749 limit
= git_env_ulong("GIT_MMAP_LIMIT", 0);
754 die("attempting to mmap %"PRIuMAX
" over limit %"PRIuMAX
,
755 (uintmax_t)length
, (uintmax_t)limit
);
758 void *xmmap_gently(void *start
, size_t length
,
759 int prot
, int flags
, int fd
, off_t offset
)
763 mmap_limit_check(length
);
764 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
765 if (ret
== MAP_FAILED
) {
768 release_pack_memory(length
);
769 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
774 void *xmmap(void *start
, size_t length
,
775 int prot
, int flags
, int fd
, off_t offset
)
777 void *ret
= xmmap_gently(start
, length
, prot
, flags
, fd
, offset
);
778 if (ret
== MAP_FAILED
)
779 die_errno("mmap failed");
784 * With an in-core object data in "map", rehash it to make sure the
785 * object name actually matches "sha1" to detect object corruption.
786 * With "map" == NULL, try reading the object named with "sha1" using
787 * the streaming interface and rehash it to do the same.
789 int check_sha1_signature(const unsigned char *sha1
, void *map
,
790 unsigned long size
, const char *type
)
792 unsigned char real_sha1
[20];
793 enum object_type obj_type
;
794 struct git_istream
*st
;
800 hash_sha1_file(map
, size
, type
, real_sha1
);
801 return hashcmp(sha1
, real_sha1
) ? -1 : 0;
804 st
= open_istream(sha1
, &obj_type
, &size
, NULL
);
808 /* Generate the header */
809 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", typename(obj_type
), size
) + 1;
813 git_SHA1_Update(&c
, hdr
, hdrlen
);
816 ssize_t readlen
= read_istream(st
, buf
, sizeof(buf
));
824 git_SHA1_Update(&c
, buf
, readlen
);
826 git_SHA1_Final(real_sha1
, &c
);
828 return hashcmp(sha1
, real_sha1
) ? -1 : 0;
831 int git_open_cloexec(const char *name
, int flags
)
834 static int o_cloexec
= O_CLOEXEC
;
836 fd
= open(name
, flags
| o_cloexec
);
837 if ((o_cloexec
& O_CLOEXEC
) && fd
< 0 && errno
== EINVAL
) {
838 /* Try again w/o O_CLOEXEC: the kernel might not support it */
839 o_cloexec
&= ~O_CLOEXEC
;
840 fd
= open(name
, flags
| o_cloexec
);
843 #if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
845 static int fd_cloexec
= FD_CLOEXEC
;
847 if (!o_cloexec
&& 0 <= fd
&& fd_cloexec
) {
848 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
849 int flags
= fcntl(fd
, F_GETFD
);
850 if (fcntl(fd
, F_SETFD
, flags
| fd_cloexec
))
859 * Find "sha1" as a loose object in the local repository or in an alternate.
860 * Returns 0 on success, negative on failure.
862 * The "path" out-parameter will give the path of the object we found (if any).
863 * Note that it may point to static storage and is only valid until another
864 * call to sha1_file_name(), etc.
866 static int stat_sha1_file(const unsigned char *sha1
, struct stat
*st
,
869 struct alternate_object_database
*alt
;
871 *path
= sha1_file_name(sha1
);
872 if (!lstat(*path
, st
))
877 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
878 *path
= alt_sha1_path(alt
, sha1
);
879 if (!lstat(*path
, st
))
887 * Like stat_sha1_file(), but actually open the object and return the
888 * descriptor. See the caveats on the "path" parameter above.
890 static int open_sha1_file(const unsigned char *sha1
, const char **path
)
893 struct alternate_object_database
*alt
;
894 int most_interesting_errno
;
896 *path
= sha1_file_name(sha1
);
897 fd
= git_open(*path
);
900 most_interesting_errno
= errno
;
903 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
904 *path
= alt_sha1_path(alt
, sha1
);
905 fd
= git_open(*path
);
908 if (most_interesting_errno
== ENOENT
)
909 most_interesting_errno
= errno
;
911 errno
= most_interesting_errno
;
916 * Map the loose object at "path" if it is not NULL, or the path found by
917 * searching for a loose object named "sha1".
919 static void *map_sha1_file_1(const char *path
,
920 const unsigned char *sha1
,
929 fd
= open_sha1_file(sha1
, &path
);
934 if (!fstat(fd
, &st
)) {
935 *size
= xsize_t(st
.st_size
);
937 /* mmap() is forbidden on empty files */
938 error("object file %s is empty", path
);
941 map
= xmmap(NULL
, *size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
948 void *map_sha1_file(const unsigned char *sha1
, unsigned long *size
)
950 return map_sha1_file_1(NULL
, sha1
, size
);
953 static int unpack_sha1_short_header(git_zstream
*stream
,
954 unsigned char *map
, unsigned long mapsize
,
955 void *buffer
, unsigned long bufsiz
)
957 /* Get the data stream */
958 memset(stream
, 0, sizeof(*stream
));
959 stream
->next_in
= map
;
960 stream
->avail_in
= mapsize
;
961 stream
->next_out
= buffer
;
962 stream
->avail_out
= bufsiz
;
964 git_inflate_init(stream
);
965 return git_inflate(stream
, 0);
968 int unpack_sha1_header(git_zstream
*stream
,
969 unsigned char *map
, unsigned long mapsize
,
970 void *buffer
, unsigned long bufsiz
)
972 int status
= unpack_sha1_short_header(stream
, map
, mapsize
,
978 /* Make sure we have the terminating NUL */
979 if (!memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
984 static int unpack_sha1_header_to_strbuf(git_zstream
*stream
, unsigned char *map
,
985 unsigned long mapsize
, void *buffer
,
986 unsigned long bufsiz
, struct strbuf
*header
)
990 status
= unpack_sha1_short_header(stream
, map
, mapsize
, buffer
, bufsiz
);
995 * Check if entire header is unpacked in the first iteration.
997 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
1001 * buffer[0..bufsiz] was not large enough. Copy the partial
1002 * result out to header, and then append the result of further
1003 * reading the stream.
1005 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
1006 stream
->next_out
= buffer
;
1007 stream
->avail_out
= bufsiz
;
1010 status
= git_inflate(stream
, 0);
1011 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
1012 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
1014 stream
->next_out
= buffer
;
1015 stream
->avail_out
= bufsiz
;
1016 } while (status
!= Z_STREAM_END
);
1020 static void *unpack_sha1_rest(git_zstream
*stream
, void *buffer
, unsigned long size
, const unsigned char *sha1
)
1022 int bytes
= strlen(buffer
) + 1;
1023 unsigned char *buf
= xmallocz(size
);
1027 n
= stream
->total_out
- bytes
;
1030 memcpy(buf
, (char *) buffer
+ bytes
, n
);
1032 if (bytes
<= size
) {
1034 * The above condition must be (bytes <= size), not
1035 * (bytes < size). In other words, even though we
1036 * expect no more output and set avail_out to zero,
1037 * the input zlib stream may have bytes that express
1038 * "this concludes the stream", and we *do* want to
1041 * Otherwise we would not be able to test that we
1042 * consumed all the input to reach the expected size;
1043 * we also want to check that zlib tells us that all
1044 * went well with status == Z_STREAM_END at the end.
1046 stream
->next_out
= buf
+ bytes
;
1047 stream
->avail_out
= size
- bytes
;
1048 while (status
== Z_OK
)
1049 status
= git_inflate(stream
, Z_FINISH
);
1051 if (status
== Z_STREAM_END
&& !stream
->avail_in
) {
1052 git_inflate_end(stream
);
1057 error("corrupt loose object '%s'", sha1_to_hex(sha1
));
1058 else if (stream
->avail_in
)
1059 error("garbage at end of loose object '%s'",
1066 * We used to just use "sscanf()", but that's actually way
1067 * too permissive for what we want to check. So do an anal
1068 * object header parse by hand.
1070 static int parse_sha1_header_extended(const char *hdr
, struct object_info
*oi
,
1073 const char *type_buf
= hdr
;
1075 int type
, type_len
= 0;
1078 * The type can be of any size but is followed by
1090 type
= type_from_string_gently(type_buf
, type_len
, 1);
1092 strbuf_add(oi
->typename
, type_buf
, type_len
);
1094 * Set type to 0 if its an unknown object and
1095 * we're obtaining the type using '--allow-unknown-type'
1098 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
) && (type
< 0))
1101 die("invalid object type");
1106 * The length must follow immediately, and be in canonical
1107 * decimal format (ie "010" is not valid).
1109 size
= *hdr
++ - '0';
1114 unsigned long c
= *hdr
- '0';
1118 size
= size
* 10 + c
;
1126 * The length must be followed by a zero byte
1128 return *hdr
? -1 : type
;
1131 int parse_sha1_header(const char *hdr
, unsigned long *sizep
)
1133 struct object_info oi
= OBJECT_INFO_INIT
;
1136 return parse_sha1_header_extended(hdr
, &oi
, 0);
1139 static int sha1_loose_object_info(const unsigned char *sha1
,
1140 struct object_info
*oi
,
1144 unsigned long mapsize
;
1148 struct strbuf hdrbuf
= STRBUF_INIT
;
1149 unsigned long size_scratch
;
1151 if (oi
->delta_base_sha1
)
1152 hashclr(oi
->delta_base_sha1
);
1155 * If we don't care about type or size, then we don't
1156 * need to look inside the object at all. Note that we
1157 * do not optimize out the stat call, even if the
1158 * caller doesn't care about the disk-size, since our
1159 * return value implicitly indicates whether the
1160 * object even exists.
1162 if (!oi
->typep
&& !oi
->typename
&& !oi
->sizep
&& !oi
->contentp
) {
1165 if (stat_sha1_file(sha1
, &st
, &path
) < 0)
1168 *oi
->disk_sizep
= st
.st_size
;
1172 map
= map_sha1_file(sha1
, &mapsize
);
1177 oi
->sizep
= &size_scratch
;
1180 *oi
->disk_sizep
= mapsize
;
1181 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
)) {
1182 if (unpack_sha1_header_to_strbuf(&stream
, map
, mapsize
, hdr
, sizeof(hdr
), &hdrbuf
) < 0)
1183 status
= error("unable to unpack %s header with --allow-unknown-type",
1185 } else if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1186 status
= error("unable to unpack %s header",
1190 else if (hdrbuf
.len
) {
1191 if ((status
= parse_sha1_header_extended(hdrbuf
.buf
, oi
, flags
)) < 0)
1192 status
= error("unable to parse %s header with --allow-unknown-type",
1194 } else if ((status
= parse_sha1_header_extended(hdr
, oi
, flags
)) < 0)
1195 status
= error("unable to parse %s header", sha1_to_hex(sha1
));
1197 if (status
>= 0 && oi
->contentp
) {
1198 *oi
->contentp
= unpack_sha1_rest(&stream
, hdr
,
1200 if (!*oi
->contentp
) {
1201 git_inflate_end(&stream
);
1205 git_inflate_end(&stream
);
1207 munmap(map
, mapsize
);
1208 if (status
&& oi
->typep
)
1209 *oi
->typep
= status
;
1210 if (oi
->sizep
== &size_scratch
)
1212 strbuf_release(&hdrbuf
);
1213 oi
->whence
= OI_LOOSE
;
1214 return (status
< 0) ? status
: 0;
1217 int fetch_if_missing
= 1;
1219 int sha1_object_info_extended(const unsigned char *sha1
, struct object_info
*oi
, unsigned flags
)
1221 static struct object_info blank_oi
= OBJECT_INFO_INIT
;
1222 struct pack_entry e
;
1224 const unsigned char *real
= (flags
& OBJECT_INFO_LOOKUP_REPLACE
) ?
1225 lookup_replace_object(sha1
) :
1227 int already_retried
= 0;
1229 if (is_null_sha1(real
))
1235 if (!(flags
& OBJECT_INFO_SKIP_CACHED
)) {
1236 struct cached_object
*co
= find_cached_object(real
);
1239 *(oi
->typep
) = co
->type
;
1241 *(oi
->sizep
) = co
->size
;
1243 *(oi
->disk_sizep
) = 0;
1244 if (oi
->delta_base_sha1
)
1245 hashclr(oi
->delta_base_sha1
);
1247 strbuf_addstr(oi
->typename
, typename(co
->type
));
1249 *oi
->contentp
= xmemdupz(co
->buf
, co
->size
);
1250 oi
->whence
= OI_CACHED
;
1256 if (find_pack_entry(real
, &e
))
1259 /* Most likely it's a loose object. */
1260 if (!sha1_loose_object_info(real
, oi
, flags
))
1263 /* Not a loose object; someone else may have just packed it. */
1264 reprepare_packed_git();
1265 if (find_pack_entry(real
, &e
))
1268 /* Check if it is a missing object */
1269 if (fetch_if_missing
&& repository_format_partial_clone
&&
1272 * TODO Investigate haveing fetch_object() return
1273 * TODO error/success and stopping the music here.
1275 fetch_object(repository_format_partial_clone
, real
);
1276 already_retried
= 1;
1283 if (oi
== &blank_oi
)
1285 * We know that the caller doesn't actually need the
1286 * information below, so return early.
1289 rtype
= packed_object_info(e
.p
, e
.offset
, oi
);
1291 mark_bad_packed_object(e
.p
, real
);
1292 return sha1_object_info_extended(real
, oi
, 0);
1293 } else if (oi
->whence
== OI_PACKED
) {
1294 oi
->u
.packed
.offset
= e
.offset
;
1295 oi
->u
.packed
.pack
= e
.p
;
1296 oi
->u
.packed
.is_delta
= (rtype
== OBJ_REF_DELTA
||
1297 rtype
== OBJ_OFS_DELTA
);
1303 /* returns enum object_type or negative */
1304 int sha1_object_info(const unsigned char *sha1
, unsigned long *sizep
)
1306 enum object_type type
;
1307 struct object_info oi
= OBJECT_INFO_INIT
;
1311 if (sha1_object_info_extended(sha1
, &oi
,
1312 OBJECT_INFO_LOOKUP_REPLACE
) < 0)
1317 static void *read_object(const unsigned char *sha1
, enum object_type
*type
,
1318 unsigned long *size
)
1320 struct object_info oi
= OBJECT_INFO_INIT
;
1324 oi
.contentp
= &content
;
1326 if (sha1_object_info_extended(sha1
, &oi
, 0) < 0)
1331 int pretend_sha1_file(void *buf
, unsigned long len
, enum object_type type
,
1332 unsigned char *sha1
)
1334 struct cached_object
*co
;
1336 hash_sha1_file(buf
, len
, typename(type
), sha1
);
1337 if (has_sha1_file(sha1
) || find_cached_object(sha1
))
1339 ALLOC_GROW(cached_objects
, cached_object_nr
+ 1, cached_object_alloc
);
1340 co
= &cached_objects
[cached_object_nr
++];
1343 co
->buf
= xmalloc(len
);
1344 memcpy(co
->buf
, buf
, len
);
1345 hashcpy(co
->sha1
, sha1
);
1350 * This function dies on corrupt objects; the callers who want to
1351 * deal with them should arrange to call read_object() and give error
1352 * messages themselves.
1354 void *read_sha1_file_extended(const unsigned char *sha1
,
1355 enum object_type
*type
,
1356 unsigned long *size
,
1360 const struct packed_git
*p
;
1363 const unsigned char *repl
= lookup_replace
? lookup_replace_object(sha1
)
1367 data
= read_object(repl
, type
, size
);
1371 if (errno
&& errno
!= ENOENT
)
1372 die_errno("failed to read object %s", sha1_to_hex(sha1
));
1374 /* die if we replaced an object with one that does not exist */
1376 die("replacement %s not found for %s",
1377 sha1_to_hex(repl
), sha1_to_hex(sha1
));
1379 if (!stat_sha1_file(repl
, &st
, &path
))
1380 die("loose object %s (stored in %s) is corrupt",
1381 sha1_to_hex(repl
), path
);
1383 if ((p
= has_packed_and_bad(repl
)) != NULL
)
1384 die("packed object %s (stored in %s) is corrupt",
1385 sha1_to_hex(repl
), p
->pack_name
);
1390 void *read_object_with_reference(const unsigned char *sha1
,
1391 const char *required_type_name
,
1392 unsigned long *size
,
1393 unsigned char *actual_sha1_return
)
1395 enum object_type type
, required_type
;
1397 unsigned long isize
;
1398 unsigned char actual_sha1
[20];
1400 required_type
= type_from_string(required_type_name
);
1401 hashcpy(actual_sha1
, sha1
);
1403 int ref_length
= -1;
1404 const char *ref_type
= NULL
;
1406 buffer
= read_sha1_file(actual_sha1
, &type
, &isize
);
1409 if (type
== required_type
) {
1411 if (actual_sha1_return
)
1412 hashcpy(actual_sha1_return
, actual_sha1
);
1415 /* Handle references */
1416 else if (type
== OBJ_COMMIT
)
1418 else if (type
== OBJ_TAG
)
1419 ref_type
= "object ";
1424 ref_length
= strlen(ref_type
);
1426 if (ref_length
+ 40 > isize
||
1427 memcmp(buffer
, ref_type
, ref_length
) ||
1428 get_sha1_hex((char *) buffer
+ ref_length
, actual_sha1
)) {
1433 /* Now we have the ID of the referred-to object in
1434 * actual_sha1. Check again. */
1438 static void write_sha1_file_prepare(const void *buf
, unsigned long len
,
1439 const char *type
, unsigned char *sha1
,
1440 char *hdr
, int *hdrlen
)
1444 /* Generate the header */
1445 *hdrlen
= xsnprintf(hdr
, *hdrlen
, "%s %lu", type
, len
)+1;
1449 git_SHA1_Update(&c
, hdr
, *hdrlen
);
1450 git_SHA1_Update(&c
, buf
, len
);
1451 git_SHA1_Final(sha1
, &c
);
1455 * Move the just written object into its final resting place.
1457 int finalize_object_file(const char *tmpfile
, const char *filename
)
1461 if (object_creation_mode
== OBJECT_CREATION_USES_RENAMES
)
1463 else if (link(tmpfile
, filename
))
1467 * Coda hack - coda doesn't like cross-directory links,
1468 * so we fall back to a rename, which will mean that it
1469 * won't be able to check collisions, but that's not a
1472 * The same holds for FAT formatted media.
1474 * When this succeeds, we just return. We have nothing
1477 if (ret
&& ret
!= EEXIST
) {
1479 if (!rename(tmpfile
, filename
))
1483 unlink_or_warn(tmpfile
);
1485 if (ret
!= EEXIST
) {
1486 return error_errno("unable to write sha1 filename %s", filename
);
1488 /* FIXME!!! Collision check here ? */
1492 if (adjust_shared_perm(filename
))
1493 return error("unable to set permission to '%s'", filename
);
1497 static int write_buffer(int fd
, const void *buf
, size_t len
)
1499 if (write_in_full(fd
, buf
, len
) < 0)
1500 return error_errno("file write error");
1504 int hash_sha1_file(const void *buf
, unsigned long len
, const char *type
,
1505 unsigned char *sha1
)
1508 int hdrlen
= sizeof(hdr
);
1509 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1513 /* Finalize a file on disk, and close it. */
1514 static void close_sha1_file(int fd
)
1516 if (fsync_object_files
)
1517 fsync_or_die(fd
, "sha1 file");
1519 die_errno("error when closing sha1 file");
1522 /* Size of directory component, including the ending '/' */
1523 static inline int directory_size(const char *filename
)
1525 const char *s
= strrchr(filename
, '/');
1528 return s
- filename
+ 1;
1532 * This creates a temporary file in the same directory as the final
1535 * We want to avoid cross-directory filename renames, because those
1536 * can have problems on various filesystems (FAT, NFS, Coda).
1538 static int create_tmpfile(struct strbuf
*tmp
, const char *filename
)
1540 int fd
, dirlen
= directory_size(filename
);
1543 strbuf_add(tmp
, filename
, dirlen
);
1544 strbuf_addstr(tmp
, "tmp_obj_XXXXXX");
1545 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1546 if (fd
< 0 && dirlen
&& errno
== ENOENT
) {
1548 * Make sure the directory exists; note that the contents
1549 * of the buffer are undefined after mkstemp returns an
1550 * error, so we have to rewrite the whole buffer from
1554 strbuf_add(tmp
, filename
, dirlen
- 1);
1555 if (mkdir(tmp
->buf
, 0777) && errno
!= EEXIST
)
1557 if (adjust_shared_perm(tmp
->buf
))
1561 strbuf_addstr(tmp
, "/tmp_obj_XXXXXX");
1562 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1567 static int write_loose_object(const unsigned char *sha1
, char *hdr
, int hdrlen
,
1568 const void *buf
, unsigned long len
, time_t mtime
)
1571 unsigned char compressed
[4096];
1574 unsigned char parano_sha1
[20];
1575 static struct strbuf tmp_file
= STRBUF_INIT
;
1576 const char *filename
= sha1_file_name(sha1
);
1578 fd
= create_tmpfile(&tmp_file
, filename
);
1580 if (errno
== EACCES
)
1581 return error("insufficient permission for adding an object to repository database %s", get_object_directory());
1583 return error_errno("unable to create temporary file");
1587 git_deflate_init(&stream
, zlib_compression_level
);
1588 stream
.next_out
= compressed
;
1589 stream
.avail_out
= sizeof(compressed
);
1592 /* First header.. */
1593 stream
.next_in
= (unsigned char *)hdr
;
1594 stream
.avail_in
= hdrlen
;
1595 while (git_deflate(&stream
, 0) == Z_OK
)
1597 git_SHA1_Update(&c
, hdr
, hdrlen
);
1599 /* Then the data itself.. */
1600 stream
.next_in
= (void *)buf
;
1601 stream
.avail_in
= len
;
1603 unsigned char *in0
= stream
.next_in
;
1604 ret
= git_deflate(&stream
, Z_FINISH
);
1605 git_SHA1_Update(&c
, in0
, stream
.next_in
- in0
);
1606 if (write_buffer(fd
, compressed
, stream
.next_out
- compressed
) < 0)
1607 die("unable to write sha1 file");
1608 stream
.next_out
= compressed
;
1609 stream
.avail_out
= sizeof(compressed
);
1610 } while (ret
== Z_OK
);
1612 if (ret
!= Z_STREAM_END
)
1613 die("unable to deflate new object %s (%d)", sha1_to_hex(sha1
), ret
);
1614 ret
= git_deflate_end_gently(&stream
);
1616 die("deflateEnd on object %s failed (%d)", sha1_to_hex(sha1
), ret
);
1617 git_SHA1_Final(parano_sha1
, &c
);
1618 if (hashcmp(sha1
, parano_sha1
) != 0)
1619 die("confused by unstable object source data for %s", sha1_to_hex(sha1
));
1621 close_sha1_file(fd
);
1626 utb
.modtime
= mtime
;
1627 if (utime(tmp_file
.buf
, &utb
) < 0)
1628 warning_errno("failed utime() on %s", tmp_file
.buf
);
1631 return finalize_object_file(tmp_file
.buf
, filename
);
1634 static int freshen_loose_object(const unsigned char *sha1
)
1636 return check_and_freshen(sha1
, 1);
1639 static int freshen_packed_object(const unsigned char *sha1
)
1641 struct pack_entry e
;
1642 if (!find_pack_entry(sha1
, &e
))
1646 if (!freshen_file(e
.p
->pack_name
))
1652 int write_sha1_file(const void *buf
, unsigned long len
, const char *type
, unsigned char *sha1
)
1655 int hdrlen
= sizeof(hdr
);
1657 /* Normally if we have it in the pack then we do not bother writing
1658 * it out into .git/objects/??/?{38} file.
1660 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1661 if (freshen_packed_object(sha1
) || freshen_loose_object(sha1
))
1663 return write_loose_object(sha1
, hdr
, hdrlen
, buf
, len
, 0);
1666 int hash_sha1_file_literally(const void *buf
, unsigned long len
, const char *type
,
1667 struct object_id
*oid
, unsigned flags
)
1670 int hdrlen
, status
= 0;
1672 /* type string, SP, %lu of the length plus NUL must fit this */
1673 hdrlen
= strlen(type
) + 32;
1674 header
= xmalloc(hdrlen
);
1675 write_sha1_file_prepare(buf
, len
, type
, oid
->hash
, header
, &hdrlen
);
1677 if (!(flags
& HASH_WRITE_OBJECT
))
1679 if (freshen_packed_object(oid
->hash
) || freshen_loose_object(oid
->hash
))
1681 status
= write_loose_object(oid
->hash
, header
, hdrlen
, buf
, len
, 0);
1688 int force_object_loose(const unsigned char *sha1
, time_t mtime
)
1692 enum object_type type
;
1697 if (has_loose_object(sha1
))
1699 buf
= read_object(sha1
, &type
, &len
);
1701 return error("cannot read sha1_file for %s", sha1_to_hex(sha1
));
1702 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", typename(type
), len
) + 1;
1703 ret
= write_loose_object(sha1
, hdr
, hdrlen
, buf
, len
, mtime
);
1709 int has_sha1_file_with_flags(const unsigned char *sha1
, int flags
)
1711 if (!startup_info
->have_repository
)
1713 return sha1_object_info_extended(sha1
, NULL
,
1714 flags
| OBJECT_INFO_SKIP_CACHED
) >= 0;
1717 int has_object_file(const struct object_id
*oid
)
1719 return has_sha1_file(oid
->hash
);
1722 int has_object_file_with_flags(const struct object_id
*oid
, int flags
)
1724 return has_sha1_file_with_flags(oid
->hash
, flags
);
1727 static void check_tree(const void *buf
, size_t size
)
1729 struct tree_desc desc
;
1730 struct name_entry entry
;
1732 init_tree_desc(&desc
, buf
, size
);
1733 while (tree_entry(&desc
, &entry
))
1735 * tree_entry() will die() on malformed entries */
1739 static void check_commit(const void *buf
, size_t size
)
1742 memset(&c
, 0, sizeof(c
));
1743 if (parse_commit_buffer(&c
, buf
, size
))
1744 die("corrupt commit");
1747 static void check_tag(const void *buf
, size_t size
)
1750 memset(&t
, 0, sizeof(t
));
1751 if (parse_tag_buffer(&t
, buf
, size
))
1755 static int index_mem(struct object_id
*oid
, void *buf
, size_t size
,
1756 enum object_type type
,
1757 const char *path
, unsigned flags
)
1759 int ret
, re_allocated
= 0;
1760 int write_object
= flags
& HASH_WRITE_OBJECT
;
1766 * Convert blobs to git internal format
1768 if ((type
== OBJ_BLOB
) && path
) {
1769 struct strbuf nbuf
= STRBUF_INIT
;
1770 if (convert_to_git(&the_index
, path
, buf
, size
, &nbuf
,
1771 get_safe_crlf(flags
))) {
1772 buf
= strbuf_detach(&nbuf
, &size
);
1776 if (flags
& HASH_FORMAT_CHECK
) {
1777 if (type
== OBJ_TREE
)
1778 check_tree(buf
, size
);
1779 if (type
== OBJ_COMMIT
)
1780 check_commit(buf
, size
);
1781 if (type
== OBJ_TAG
)
1782 check_tag(buf
, size
);
1786 ret
= write_sha1_file(buf
, size
, typename(type
), oid
->hash
);
1788 ret
= hash_sha1_file(buf
, size
, typename(type
), oid
->hash
);
1794 static int index_stream_convert_blob(struct object_id
*oid
, int fd
,
1795 const char *path
, unsigned flags
)
1798 const int write_object
= flags
& HASH_WRITE_OBJECT
;
1799 struct strbuf sbuf
= STRBUF_INIT
;
1802 assert(would_convert_to_git_filter_fd(path
));
1804 convert_to_git_filter_fd(&the_index
, path
, fd
, &sbuf
,
1805 get_safe_crlf(flags
));
1808 ret
= write_sha1_file(sbuf
.buf
, sbuf
.len
, typename(OBJ_BLOB
),
1811 ret
= hash_sha1_file(sbuf
.buf
, sbuf
.len
, typename(OBJ_BLOB
),
1813 strbuf_release(&sbuf
);
1817 static int index_pipe(struct object_id
*oid
, int fd
, enum object_type type
,
1818 const char *path
, unsigned flags
)
1820 struct strbuf sbuf
= STRBUF_INIT
;
1823 if (strbuf_read(&sbuf
, fd
, 4096) >= 0)
1824 ret
= index_mem(oid
, sbuf
.buf
, sbuf
.len
, type
, path
, flags
);
1827 strbuf_release(&sbuf
);
1831 #define SMALL_FILE_SIZE (32*1024)
1833 static int index_core(struct object_id
*oid
, int fd
, size_t size
,
1834 enum object_type type
, const char *path
,
1840 ret
= index_mem(oid
, "", size
, type
, path
, flags
);
1841 } else if (size
<= SMALL_FILE_SIZE
) {
1842 char *buf
= xmalloc(size
);
1843 ssize_t read_result
= read_in_full(fd
, buf
, size
);
1844 if (read_result
< 0)
1845 ret
= error_errno("read error while indexing %s",
1846 path
? path
: "<unknown>");
1847 else if (read_result
!= size
)
1848 ret
= error("short read while indexing %s",
1849 path
? path
: "<unknown>");
1851 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1854 void *buf
= xmmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1855 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1862 * This creates one packfile per large blob unless bulk-checkin
1863 * machinery is "plugged".
1865 * This also bypasses the usual "convert-to-git" dance, and that is on
1866 * purpose. We could write a streaming version of the converting
1867 * functions and insert that before feeding the data to fast-import
1868 * (or equivalent in-core API described above). However, that is
1869 * somewhat complicated, as we do not know the size of the filter
1870 * result, which we need to know beforehand when writing a git object.
1871 * Since the primary motivation for trying to stream from the working
1872 * tree file and to avoid mmaping it in core is to deal with large
1873 * binary blobs, they generally do not want to get any conversion, and
1874 * callers should avoid this code path when filters are requested.
1876 static int index_stream(struct object_id
*oid
, int fd
, size_t size
,
1877 enum object_type type
, const char *path
,
1880 return index_bulk_checkin(oid
->hash
, fd
, size
, type
, path
, flags
);
1883 int index_fd(struct object_id
*oid
, int fd
, struct stat
*st
,
1884 enum object_type type
, const char *path
, unsigned flags
)
1889 * Call xsize_t() only when needed to avoid potentially unnecessary
1890 * die() for large files.
1892 if (type
== OBJ_BLOB
&& path
&& would_convert_to_git_filter_fd(path
))
1893 ret
= index_stream_convert_blob(oid
, fd
, path
, flags
);
1894 else if (!S_ISREG(st
->st_mode
))
1895 ret
= index_pipe(oid
, fd
, type
, path
, flags
);
1896 else if (st
->st_size
<= big_file_threshold
|| type
!= OBJ_BLOB
||
1897 (path
&& would_convert_to_git(&the_index
, path
)))
1898 ret
= index_core(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1901 ret
= index_stream(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1907 int index_path(struct object_id
*oid
, const char *path
, struct stat
*st
, unsigned flags
)
1910 struct strbuf sb
= STRBUF_INIT
;
1913 switch (st
->st_mode
& S_IFMT
) {
1915 fd
= open(path
, O_RDONLY
);
1917 return error_errno("open(\"%s\")", path
);
1918 if (index_fd(oid
, fd
, st
, OBJ_BLOB
, path
, flags
) < 0)
1919 return error("%s: failed to insert into database",
1923 if (strbuf_readlink(&sb
, path
, st
->st_size
))
1924 return error_errno("readlink(\"%s\")", path
);
1925 if (!(flags
& HASH_WRITE_OBJECT
))
1926 hash_sha1_file(sb
.buf
, sb
.len
, blob_type
, oid
->hash
);
1927 else if (write_sha1_file(sb
.buf
, sb
.len
, blob_type
, oid
->hash
))
1928 rc
= error("%s: failed to insert into database", path
);
1929 strbuf_release(&sb
);
1932 return resolve_gitlink_ref(path
, "HEAD", oid
);
1934 return error("%s: unsupported file type", path
);
1939 int read_pack_header(int fd
, struct pack_header
*header
)
1941 if (read_in_full(fd
, header
, sizeof(*header
)) != sizeof(*header
))
1942 /* "eof before pack header was fully read" */
1943 return PH_ERROR_EOF
;
1945 if (header
->hdr_signature
!= htonl(PACK_SIGNATURE
))
1946 /* "protocol error (pack signature mismatch detected)" */
1947 return PH_ERROR_PACK_SIGNATURE
;
1948 if (!pack_version_ok(header
->hdr_version
))
1949 /* "protocol error (pack version unsupported)" */
1950 return PH_ERROR_PROTOCOL
;
1954 void assert_sha1_type(const unsigned char *sha1
, enum object_type expect
)
1956 enum object_type type
= sha1_object_info(sha1
, NULL
);
1958 die("%s is not a valid object", sha1_to_hex(sha1
));
1960 die("%s is not a valid '%s' object", sha1_to_hex(sha1
),
1964 int for_each_file_in_obj_subdir(unsigned int subdir_nr
,
1965 struct strbuf
*path
,
1966 each_loose_object_fn obj_cb
,
1967 each_loose_cruft_fn cruft_cb
,
1968 each_loose_subdir_fn subdir_cb
,
1971 size_t origlen
, baselen
;
1975 struct object_id oid
;
1977 if (subdir_nr
> 0xff)
1978 BUG("invalid loose object subdirectory: %x", subdir_nr
);
1980 origlen
= path
->len
;
1981 strbuf_complete(path
, '/');
1982 strbuf_addf(path
, "%02x", subdir_nr
);
1984 dir
= opendir(path
->buf
);
1986 if (errno
!= ENOENT
)
1987 r
= error_errno("unable to open %s", path
->buf
);
1988 strbuf_setlen(path
, origlen
);
1992 oid
.hash
[0] = subdir_nr
;
1993 strbuf_addch(path
, '/');
1994 baselen
= path
->len
;
1996 while ((de
= readdir(dir
))) {
1998 if (is_dot_or_dotdot(de
->d_name
))
2001 namelen
= strlen(de
->d_name
);
2002 strbuf_setlen(path
, baselen
);
2003 strbuf_add(path
, de
->d_name
, namelen
);
2004 if (namelen
== GIT_SHA1_HEXSZ
- 2 &&
2005 !hex_to_bytes(oid
.hash
+ 1, de
->d_name
,
2006 GIT_SHA1_RAWSZ
- 1)) {
2008 r
= obj_cb(&oid
, path
->buf
, data
);
2016 r
= cruft_cb(de
->d_name
, path
->buf
, data
);
2023 strbuf_setlen(path
, baselen
- 1);
2024 if (!r
&& subdir_cb
)
2025 r
= subdir_cb(subdir_nr
, path
->buf
, data
);
2027 strbuf_setlen(path
, origlen
);
2032 int for_each_loose_file_in_objdir_buf(struct strbuf
*path
,
2033 each_loose_object_fn obj_cb
,
2034 each_loose_cruft_fn cruft_cb
,
2035 each_loose_subdir_fn subdir_cb
,
2041 for (i
= 0; i
< 256; i
++) {
2042 r
= for_each_file_in_obj_subdir(i
, path
, obj_cb
, cruft_cb
,
2051 int for_each_loose_file_in_objdir(const char *path
,
2052 each_loose_object_fn obj_cb
,
2053 each_loose_cruft_fn cruft_cb
,
2054 each_loose_subdir_fn subdir_cb
,
2057 struct strbuf buf
= STRBUF_INIT
;
2060 strbuf_addstr(&buf
, path
);
2061 r
= for_each_loose_file_in_objdir_buf(&buf
, obj_cb
, cruft_cb
,
2063 strbuf_release(&buf
);
2068 struct loose_alt_odb_data
{
2069 each_loose_object_fn
*cb
;
2073 static int loose_from_alt_odb(struct alternate_object_database
*alt
,
2076 struct loose_alt_odb_data
*data
= vdata
;
2077 struct strbuf buf
= STRBUF_INIT
;
2080 strbuf_addstr(&buf
, alt
->path
);
2081 r
= for_each_loose_file_in_objdir_buf(&buf
,
2082 data
->cb
, NULL
, NULL
,
2084 strbuf_release(&buf
);
2088 int for_each_loose_object(each_loose_object_fn cb
, void *data
, unsigned flags
)
2090 struct loose_alt_odb_data alt
;
2093 r
= for_each_loose_file_in_objdir(get_object_directory(),
2094 cb
, NULL
, NULL
, data
);
2098 if (flags
& FOR_EACH_OBJECT_LOCAL_ONLY
)
2103 return foreach_alt_odb(loose_from_alt_odb
, &alt
);
2106 static int check_stream_sha1(git_zstream
*stream
,
2110 const unsigned char *expected_sha1
)
2113 unsigned char real_sha1
[GIT_MAX_RAWSZ
];
2114 unsigned char buf
[4096];
2115 unsigned long total_read
;
2119 git_SHA1_Update(&c
, hdr
, stream
->total_out
);
2122 * We already read some bytes into hdr, but the ones up to the NUL
2123 * do not count against the object's content size.
2125 total_read
= stream
->total_out
- strlen(hdr
) - 1;
2128 * This size comparison must be "<=" to read the final zlib packets;
2129 * see the comment in unpack_sha1_rest for details.
2131 while (total_read
<= size
&&
2132 (status
== Z_OK
|| status
== Z_BUF_ERROR
)) {
2133 stream
->next_out
= buf
;
2134 stream
->avail_out
= sizeof(buf
);
2135 if (size
- total_read
< stream
->avail_out
)
2136 stream
->avail_out
= size
- total_read
;
2137 status
= git_inflate(stream
, Z_FINISH
);
2138 git_SHA1_Update(&c
, buf
, stream
->next_out
- buf
);
2139 total_read
+= stream
->next_out
- buf
;
2141 git_inflate_end(stream
);
2143 if (status
!= Z_STREAM_END
) {
2144 error("corrupt loose object '%s'", sha1_to_hex(expected_sha1
));
2147 if (stream
->avail_in
) {
2148 error("garbage at end of loose object '%s'",
2149 sha1_to_hex(expected_sha1
));
2153 git_SHA1_Final(real_sha1
, &c
);
2154 if (hashcmp(expected_sha1
, real_sha1
)) {
2155 error("sha1 mismatch for %s (expected %s)", path
,
2156 sha1_to_hex(expected_sha1
));
2163 int read_loose_object(const char *path
,
2164 const unsigned char *expected_sha1
,
2165 enum object_type
*type
,
2166 unsigned long *size
,
2171 unsigned long mapsize
;
2177 map
= map_sha1_file_1(path
, NULL
, &mapsize
);
2179 error_errno("unable to mmap %s", path
);
2183 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0) {
2184 error("unable to unpack header of %s", path
);
2188 *type
= parse_sha1_header(hdr
, size
);
2190 error("unable to parse header of %s", path
);
2191 git_inflate_end(&stream
);
2195 if (*type
== OBJ_BLOB
) {
2196 if (check_stream_sha1(&stream
, hdr
, *size
, path
, expected_sha1
) < 0)
2199 *contents
= unpack_sha1_rest(&stream
, hdr
, *size
, expected_sha1
);
2201 error("unable to unpack contents of %s", path
);
2202 git_inflate_end(&stream
);
2205 if (check_sha1_signature(expected_sha1
, *contents
,
2206 *size
, typename(*type
))) {
2207 error("sha1 mismatch for %s (expected %s)", path
,
2208 sha1_to_hex(expected_sha1
));
2214 ret
= 0; /* everything checks out */
2218 munmap(map
, mapsize
);