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"
33 const unsigned char null_sha1
[GIT_MAX_RAWSZ
];
34 const struct object_id null_oid
;
35 const struct object_id empty_tree_oid
= {
36 EMPTY_TREE_SHA1_BIN_LITERAL
38 const struct object_id empty_blob_oid
= {
39 EMPTY_BLOB_SHA1_BIN_LITERAL
43 * This is meant to hold a *small* number of objects that you would
44 * want read_sha1_file() to be able to return, but yet you do not want
45 * to write them into the object store (e.g. a browse-only
48 static struct cached_object
{
49 unsigned char sha1
[20];
50 enum object_type type
;
54 static int cached_object_nr
, cached_object_alloc
;
56 static struct cached_object empty_tree
= {
57 EMPTY_TREE_SHA1_BIN_LITERAL
,
63 static struct cached_object
*find_cached_object(const unsigned char *sha1
)
66 struct cached_object
*co
= cached_objects
;
68 for (i
= 0; i
< cached_object_nr
; i
++, co
++) {
69 if (!hashcmp(co
->sha1
, sha1
))
72 if (!hashcmp(sha1
, empty_tree
.sha1
))
77 int mkdir_in_gitdir(const char *path
)
79 if (mkdir(path
, 0777)) {
80 int saved_errno
= errno
;
82 struct strbuf sb
= STRBUF_INIT
;
87 * Are we looking at a path in a symlinked worktree
88 * whose original repository does not yet have it?
89 * e.g. .git/rr-cache pointing at its original
90 * repository in which the user hasn't performed any
91 * conflict resolution yet?
93 if (lstat(path
, &st
) || !S_ISLNK(st
.st_mode
) ||
94 strbuf_readlink(&sb
, path
, st
.st_size
) ||
95 !is_absolute_path(sb
.buf
) ||
96 mkdir(sb
.buf
, 0777)) {
103 return adjust_shared_perm(path
);
106 enum scld_error
safe_create_leading_directories(char *path
)
108 char *next_component
= path
+ offset_1st_component(path
);
109 enum scld_error ret
= SCLD_OK
;
111 while (ret
== SCLD_OK
&& next_component
) {
113 char *slash
= next_component
, slash_character
;
115 while (*slash
&& !is_dir_sep(*slash
))
121 next_component
= slash
+ 1;
122 while (is_dir_sep(*next_component
))
124 if (!*next_component
)
127 slash_character
= *slash
;
129 if (!stat(path
, &st
)) {
131 if (!S_ISDIR(st
.st_mode
)) {
135 } else if (mkdir(path
, 0777)) {
136 if (errno
== EEXIST
&&
137 !stat(path
, &st
) && S_ISDIR(st
.st_mode
))
138 ; /* somebody created it since we checked */
139 else if (errno
== ENOENT
)
141 * Either mkdir() failed because
142 * somebody just pruned the containing
143 * directory, or stat() failed because
144 * the file that was in our way was
145 * just removed. Either way, inform
146 * the caller that it might be worth
152 } else if (adjust_shared_perm(path
)) {
155 *slash
= slash_character
;
160 enum scld_error
safe_create_leading_directories_const(const char *path
)
163 /* path points to cache entries, so xstrdup before messing with it */
164 char *buf
= xstrdup(path
);
165 enum scld_error result
= safe_create_leading_directories(buf
);
173 int raceproof_create_file(const char *path
, create_file_fn fn
, void *cb
)
176 * The number of times we will try to remove empty directories
177 * in the way of path. This is only 1 because if another
178 * process is racily creating directories that conflict with
179 * us, we don't want to fight against them.
181 int remove_directories_remaining
= 1;
184 * The number of times that we will try to create the
185 * directories containing path. We are willing to attempt this
186 * more than once, because another process could be trying to
187 * clean up empty directories at the same time as we are
188 * trying to create them.
190 int create_directories_remaining
= 3;
192 /* A scratch copy of path, filled lazily if we need it: */
193 struct strbuf path_copy
= STRBUF_INIT
;
206 if (errno
== EISDIR
&& remove_directories_remaining
-- > 0) {
208 * A directory is in the way. Maybe it is empty; try
212 strbuf_addstr(&path_copy
, path
);
214 if (!remove_dir_recursively(&path_copy
, REMOVE_DIR_EMPTY_ONLY
))
216 } else if (errno
== ENOENT
&& create_directories_remaining
-- > 0) {
218 * Maybe the containing directory didn't exist, or
219 * maybe it was just deleted by a process that is
220 * racing with us to clean up empty directories. Try
223 enum scld_error scld_result
;
226 strbuf_addstr(&path_copy
, path
);
229 scld_result
= safe_create_leading_directories(path_copy
.buf
);
230 if (scld_result
== SCLD_OK
)
232 } while (scld_result
== SCLD_VANISHED
&& create_directories_remaining
-- > 0);
236 strbuf_release(&path_copy
);
241 static void fill_sha1_path(struct strbuf
*buf
, const unsigned char *sha1
)
244 for (i
= 0; i
< 20; i
++) {
245 static char hex
[] = "0123456789abcdef";
246 unsigned int val
= sha1
[i
];
247 strbuf_addch(buf
, hex
[val
>> 4]);
248 strbuf_addch(buf
, hex
[val
& 0xf]);
250 strbuf_addch(buf
, '/');
254 const char *sha1_file_name(const unsigned char *sha1
)
256 static struct strbuf buf
= STRBUF_INIT
;
259 strbuf_addf(&buf
, "%s/", get_object_directory());
261 fill_sha1_path(&buf
, sha1
);
265 struct strbuf
*alt_scratch_buf(struct alternate_object_database
*alt
)
267 strbuf_setlen(&alt
->scratch
, alt
->base_len
);
268 return &alt
->scratch
;
271 static const char *alt_sha1_path(struct alternate_object_database
*alt
,
272 const unsigned char *sha1
)
274 struct strbuf
*buf
= alt_scratch_buf(alt
);
275 fill_sha1_path(buf
, sha1
);
279 struct alternate_object_database
*alt_odb_list
;
280 static struct alternate_object_database
**alt_odb_tail
;
283 * Return non-zero iff the path is usable as an alternate object database.
285 static int alt_odb_usable(struct strbuf
*path
, const char *normalized_objdir
)
287 struct alternate_object_database
*alt
;
289 /* Detect cases where alternate disappeared */
290 if (!is_directory(path
->buf
)) {
291 error("object directory %s does not exist; "
292 "check .git/objects/info/alternates.",
298 * Prevent the common mistake of listing the same
299 * thing twice, or object directory itself.
301 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
302 if (!fspathcmp(path
->buf
, alt
->path
))
305 if (!fspathcmp(path
->buf
, normalized_objdir
))
312 * Prepare alternate object database registry.
314 * The variable alt_odb_list points at the list of struct
315 * alternate_object_database. The elements on this list come from
316 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
317 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
318 * whose contents is similar to that environment variable but can be
319 * LF separated. Its base points at a statically allocated buffer that
320 * contains "/the/directory/corresponding/to/.git/objects/...", while
321 * its name points just after the slash at the end of ".git/objects/"
322 * in the example above, and has enough space to hold 40-byte hex
323 * SHA1, an extra slash for the first level indirection, and the
326 static void read_info_alternates(const char * relative_base
, int depth
);
327 static int link_alt_odb_entry(const char *entry
, const char *relative_base
,
328 int depth
, const char *normalized_objdir
)
330 struct alternate_object_database
*ent
;
331 struct strbuf pathbuf
= STRBUF_INIT
;
333 if (!is_absolute_path(entry
) && relative_base
) {
334 strbuf_realpath(&pathbuf
, relative_base
, 1);
335 strbuf_addch(&pathbuf
, '/');
337 strbuf_addstr(&pathbuf
, entry
);
339 if (strbuf_normalize_path(&pathbuf
) < 0 && relative_base
) {
340 error("unable to normalize alternate object path: %s",
342 strbuf_release(&pathbuf
);
347 * The trailing slash after the directory name is given by
348 * this function at the end. Remove duplicates.
350 while (pathbuf
.len
&& pathbuf
.buf
[pathbuf
.len
- 1] == '/')
351 strbuf_setlen(&pathbuf
, pathbuf
.len
- 1);
353 if (!alt_odb_usable(&pathbuf
, normalized_objdir
)) {
354 strbuf_release(&pathbuf
);
358 ent
= alloc_alt_odb(pathbuf
.buf
);
360 /* add the alternate entry */
362 alt_odb_tail
= &(ent
->next
);
365 /* recursively add alternates */
366 read_info_alternates(pathbuf
.buf
, depth
+ 1);
368 strbuf_release(&pathbuf
);
372 static const char *parse_alt_odb_entry(const char *string
,
380 if (*string
== '#') {
381 /* comment; consume up to next separator */
382 end
= strchrnul(string
, sep
);
383 } else if (*string
== '"' && !unquote_c_style(out
, string
, &end
)) {
385 * quoted path; unquote_c_style has copied the
386 * data for us and set "end". Broken quoting (e.g.,
387 * an entry that doesn't end with a quote) falls
388 * back to the unquoted case below.
391 /* normal, unquoted path */
392 end
= strchrnul(string
, sep
);
393 strbuf_add(out
, string
, end
- string
);
401 static void link_alt_odb_entries(const char *alt
, int sep
,
402 const char *relative_base
, int depth
)
404 struct strbuf objdirbuf
= STRBUF_INIT
;
405 struct strbuf entry
= STRBUF_INIT
;
411 error("%s: ignoring alternate object stores, nesting too deep.",
416 strbuf_add_absolute_path(&objdirbuf
, get_object_directory());
417 if (strbuf_normalize_path(&objdirbuf
) < 0)
418 die("unable to normalize object directory: %s",
422 alt
= parse_alt_odb_entry(alt
, sep
, &entry
);
425 link_alt_odb_entry(entry
.buf
, relative_base
, depth
, objdirbuf
.buf
);
427 strbuf_release(&entry
);
428 strbuf_release(&objdirbuf
);
431 static void read_info_alternates(const char * relative_base
, int depth
)
434 struct strbuf buf
= STRBUF_INIT
;
436 path
= xstrfmt("%s/info/alternates", relative_base
);
437 if (strbuf_read_file(&buf
, path
, 1024) < 0) {
438 warn_on_fopen_errors(path
);
443 link_alt_odb_entries(buf
.buf
, '\n', relative_base
, depth
);
444 strbuf_release(&buf
);
448 struct alternate_object_database
*alloc_alt_odb(const char *dir
)
450 struct alternate_object_database
*ent
;
452 FLEX_ALLOC_STR(ent
, path
, dir
);
453 strbuf_init(&ent
->scratch
, 0);
454 strbuf_addf(&ent
->scratch
, "%s/", dir
);
455 ent
->base_len
= ent
->scratch
.len
;
460 void add_to_alternates_file(const char *reference
)
462 struct lock_file lock
= LOCK_INIT
;
463 char *alts
= git_pathdup("objects/info/alternates");
467 hold_lock_file_for_update(&lock
, alts
, LOCK_DIE_ON_ERROR
);
468 out
= fdopen_lock_file(&lock
, "w");
470 die_errno("unable to fdopen alternates lockfile");
472 in
= fopen(alts
, "r");
474 struct strbuf line
= STRBUF_INIT
;
476 while (strbuf_getline(&line
, in
) != EOF
) {
477 if (!strcmp(reference
, line
.buf
)) {
481 fprintf_or_die(out
, "%s\n", line
.buf
);
484 strbuf_release(&line
);
487 else if (errno
!= ENOENT
)
488 die_errno("unable to read alternates file");
491 rollback_lock_file(&lock
);
493 fprintf_or_die(out
, "%s\n", reference
);
494 if (commit_lock_file(&lock
))
495 die_errno("unable to move new alternates file into place");
497 link_alt_odb_entries(reference
, '\n', NULL
, 0);
502 void add_to_alternates_memory(const char *reference
)
505 * Make sure alternates are initialized, or else our entry may be
506 * overwritten when they are.
510 link_alt_odb_entries(reference
, '\n', NULL
, 0);
514 * Compute the exact path an alternate is at and returns it. In case of
515 * error NULL is returned and the human readable error is added to `err`
516 * `path` may be relative and should point to $GITDIR.
517 * `err` must not be null.
519 char *compute_alternate_path(const char *path
, struct strbuf
*err
)
521 char *ref_git
= NULL
;
522 const char *repo
, *ref_git_s
;
525 ref_git_s
= real_path_if_valid(path
);
528 strbuf_addf(err
, _("path '%s' does not exist"), path
);
532 * Beware: read_gitfile(), real_path() and mkpath()
533 * return static buffer
535 ref_git
= xstrdup(ref_git_s
);
537 repo
= read_gitfile(ref_git
);
539 repo
= read_gitfile(mkpath("%s/.git", ref_git
));
542 ref_git
= xstrdup(repo
);
545 if (!repo
&& is_directory(mkpath("%s/.git/objects", ref_git
))) {
546 char *ref_git_git
= mkpathdup("%s/.git", ref_git
);
548 ref_git
= ref_git_git
;
549 } else if (!is_directory(mkpath("%s/objects", ref_git
))) {
550 struct strbuf sb
= STRBUF_INIT
;
552 if (get_common_dir(&sb
, ref_git
)) {
554 _("reference repository '%s' as a linked "
555 "checkout is not supported yet."),
560 strbuf_addf(err
, _("reference repository '%s' is not a "
561 "local repository."), path
);
565 if (!access(mkpath("%s/shallow", ref_git
), F_OK
)) {
566 strbuf_addf(err
, _("reference repository '%s' is shallow"),
572 if (!access(mkpath("%s/info/grafts", ref_git
), F_OK
)) {
574 _("reference repository '%s' is grafted"),
582 FREE_AND_NULL(ref_git
);
588 int foreach_alt_odb(alt_odb_fn fn
, void *cb
)
590 struct alternate_object_database
*ent
;
594 for (ent
= alt_odb_list
; ent
; ent
= ent
->next
) {
602 void prepare_alt_odb(void)
609 alt
= getenv(ALTERNATE_DB_ENVIRONMENT
);
611 alt_odb_tail
= &alt_odb_list
;
612 link_alt_odb_entries(alt
, PATH_SEP
, NULL
, 0);
614 read_info_alternates(get_object_directory(), 0);
617 /* Returns 1 if we have successfully freshened the file, 0 otherwise. */
618 static int freshen_file(const char *fn
)
621 t
.actime
= t
.modtime
= time(NULL
);
622 return !utime(fn
, &t
);
626 * All of the check_and_freshen functions return 1 if the file exists and was
627 * freshened (if freshening was requested), 0 otherwise. If they return
628 * 0, you should not assume that it is safe to skip a write of the object (it
629 * either does not exist on disk, or has a stale mtime and may be subject to
632 int check_and_freshen_file(const char *fn
, int freshen
)
634 if (access(fn
, F_OK
))
636 if (freshen
&& !freshen_file(fn
))
641 static int check_and_freshen_local(const unsigned char *sha1
, int freshen
)
643 return check_and_freshen_file(sha1_file_name(sha1
), freshen
);
646 static int check_and_freshen_nonlocal(const unsigned char *sha1
, int freshen
)
648 struct alternate_object_database
*alt
;
650 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
651 const char *path
= alt_sha1_path(alt
, sha1
);
652 if (check_and_freshen_file(path
, freshen
))
658 static int check_and_freshen(const unsigned char *sha1
, int freshen
)
660 return check_and_freshen_local(sha1
, freshen
) ||
661 check_and_freshen_nonlocal(sha1
, freshen
);
664 int has_loose_object_nonlocal(const unsigned char *sha1
)
666 return check_and_freshen_nonlocal(sha1
, 0);
669 static int has_loose_object(const unsigned char *sha1
)
671 return check_and_freshen(sha1
, 0);
674 static void mmap_limit_check(size_t length
)
676 static size_t limit
= 0;
678 limit
= git_env_ulong("GIT_MMAP_LIMIT", 0);
683 die("attempting to mmap %"PRIuMAX
" over limit %"PRIuMAX
,
684 (uintmax_t)length
, (uintmax_t)limit
);
687 void *xmmap_gently(void *start
, size_t length
,
688 int prot
, int flags
, int fd
, off_t offset
)
692 mmap_limit_check(length
);
693 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
694 if (ret
== MAP_FAILED
) {
697 release_pack_memory(length
);
698 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
703 void *xmmap(void *start
, size_t length
,
704 int prot
, int flags
, int fd
, off_t offset
)
706 void *ret
= xmmap_gently(start
, length
, prot
, flags
, fd
, offset
);
707 if (ret
== MAP_FAILED
)
708 die_errno("mmap failed");
713 * With an in-core object data in "map", rehash it to make sure the
714 * object name actually matches "sha1" to detect object corruption.
715 * With "map" == NULL, try reading the object named with "sha1" using
716 * the streaming interface and rehash it to do the same.
718 int check_sha1_signature(const unsigned char *sha1
, void *map
,
719 unsigned long size
, const char *type
)
721 unsigned char real_sha1
[20];
722 enum object_type obj_type
;
723 struct git_istream
*st
;
729 hash_sha1_file(map
, size
, type
, real_sha1
);
730 return hashcmp(sha1
, real_sha1
) ? -1 : 0;
733 st
= open_istream(sha1
, &obj_type
, &size
, NULL
);
737 /* Generate the header */
738 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", typename(obj_type
), size
) + 1;
742 git_SHA1_Update(&c
, hdr
, hdrlen
);
745 ssize_t readlen
= read_istream(st
, buf
, sizeof(buf
));
753 git_SHA1_Update(&c
, buf
, readlen
);
755 git_SHA1_Final(real_sha1
, &c
);
757 return hashcmp(sha1
, real_sha1
) ? -1 : 0;
760 int git_open_cloexec(const char *name
, int flags
)
763 static int o_cloexec
= O_CLOEXEC
;
765 fd
= open(name
, flags
| o_cloexec
);
766 if ((o_cloexec
& O_CLOEXEC
) && fd
< 0 && errno
== EINVAL
) {
767 /* Try again w/o O_CLOEXEC: the kernel might not support it */
768 o_cloexec
&= ~O_CLOEXEC
;
769 fd
= open(name
, flags
| o_cloexec
);
772 #if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
774 static int fd_cloexec
= FD_CLOEXEC
;
776 if (!o_cloexec
&& 0 <= fd
&& fd_cloexec
) {
777 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
778 int flags
= fcntl(fd
, F_GETFD
);
779 if (fcntl(fd
, F_SETFD
, flags
| fd_cloexec
))
788 * Find "sha1" as a loose object in the local repository or in an alternate.
789 * Returns 0 on success, negative on failure.
791 * The "path" out-parameter will give the path of the object we found (if any).
792 * Note that it may point to static storage and is only valid until another
793 * call to sha1_file_name(), etc.
795 static int stat_sha1_file(const unsigned char *sha1
, struct stat
*st
,
798 struct alternate_object_database
*alt
;
800 *path
= sha1_file_name(sha1
);
801 if (!lstat(*path
, st
))
806 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
807 *path
= alt_sha1_path(alt
, sha1
);
808 if (!lstat(*path
, st
))
816 * Like stat_sha1_file(), but actually open the object and return the
817 * descriptor. See the caveats on the "path" parameter above.
819 static int open_sha1_file(const unsigned char *sha1
, const char **path
)
822 struct alternate_object_database
*alt
;
823 int most_interesting_errno
;
825 *path
= sha1_file_name(sha1
);
826 fd
= git_open(*path
);
829 most_interesting_errno
= errno
;
832 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
833 *path
= alt_sha1_path(alt
, sha1
);
834 fd
= git_open(*path
);
837 if (most_interesting_errno
== ENOENT
)
838 most_interesting_errno
= errno
;
840 errno
= most_interesting_errno
;
845 * Map the loose object at "path" if it is not NULL, or the path found by
846 * searching for a loose object named "sha1".
848 static void *map_sha1_file_1(const char *path
,
849 const unsigned char *sha1
,
858 fd
= open_sha1_file(sha1
, &path
);
863 if (!fstat(fd
, &st
)) {
864 *size
= xsize_t(st
.st_size
);
866 /* mmap() is forbidden on empty files */
867 error("object file %s is empty", path
);
870 map
= xmmap(NULL
, *size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
877 void *map_sha1_file(const unsigned char *sha1
, unsigned long *size
)
879 return map_sha1_file_1(NULL
, sha1
, size
);
882 static int unpack_sha1_short_header(git_zstream
*stream
,
883 unsigned char *map
, unsigned long mapsize
,
884 void *buffer
, unsigned long bufsiz
)
886 /* Get the data stream */
887 memset(stream
, 0, sizeof(*stream
));
888 stream
->next_in
= map
;
889 stream
->avail_in
= mapsize
;
890 stream
->next_out
= buffer
;
891 stream
->avail_out
= bufsiz
;
893 git_inflate_init(stream
);
894 return git_inflate(stream
, 0);
897 int unpack_sha1_header(git_zstream
*stream
,
898 unsigned char *map
, unsigned long mapsize
,
899 void *buffer
, unsigned long bufsiz
)
901 int status
= unpack_sha1_short_header(stream
, map
, mapsize
,
907 /* Make sure we have the terminating NUL */
908 if (!memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
913 static int unpack_sha1_header_to_strbuf(git_zstream
*stream
, unsigned char *map
,
914 unsigned long mapsize
, void *buffer
,
915 unsigned long bufsiz
, struct strbuf
*header
)
919 status
= unpack_sha1_short_header(stream
, map
, mapsize
, buffer
, bufsiz
);
924 * Check if entire header is unpacked in the first iteration.
926 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
930 * buffer[0..bufsiz] was not large enough. Copy the partial
931 * result out to header, and then append the result of further
932 * reading the stream.
934 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
935 stream
->next_out
= buffer
;
936 stream
->avail_out
= bufsiz
;
939 status
= git_inflate(stream
, 0);
940 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
941 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
943 stream
->next_out
= buffer
;
944 stream
->avail_out
= bufsiz
;
945 } while (status
!= Z_STREAM_END
);
949 static void *unpack_sha1_rest(git_zstream
*stream
, void *buffer
, unsigned long size
, const unsigned char *sha1
)
951 int bytes
= strlen(buffer
) + 1;
952 unsigned char *buf
= xmallocz(size
);
956 n
= stream
->total_out
- bytes
;
959 memcpy(buf
, (char *) buffer
+ bytes
, n
);
963 * The above condition must be (bytes <= size), not
964 * (bytes < size). In other words, even though we
965 * expect no more output and set avail_out to zero,
966 * the input zlib stream may have bytes that express
967 * "this concludes the stream", and we *do* want to
970 * Otherwise we would not be able to test that we
971 * consumed all the input to reach the expected size;
972 * we also want to check that zlib tells us that all
973 * went well with status == Z_STREAM_END at the end.
975 stream
->next_out
= buf
+ bytes
;
976 stream
->avail_out
= size
- bytes
;
977 while (status
== Z_OK
)
978 status
= git_inflate(stream
, Z_FINISH
);
980 if (status
== Z_STREAM_END
&& !stream
->avail_in
) {
981 git_inflate_end(stream
);
986 error("corrupt loose object '%s'", sha1_to_hex(sha1
));
987 else if (stream
->avail_in
)
988 error("garbage at end of loose object '%s'",
995 * We used to just use "sscanf()", but that's actually way
996 * too permissive for what we want to check. So do an anal
997 * object header parse by hand.
999 static int parse_sha1_header_extended(const char *hdr
, struct object_info
*oi
,
1002 const char *type_buf
= hdr
;
1004 int type
, type_len
= 0;
1007 * The type can be of any size but is followed by
1019 type
= type_from_string_gently(type_buf
, type_len
, 1);
1021 strbuf_add(oi
->typename
, type_buf
, type_len
);
1023 * Set type to 0 if its an unknown object and
1024 * we're obtaining the type using '--allow-unknown-type'
1027 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
) && (type
< 0))
1030 die("invalid object type");
1035 * The length must follow immediately, and be in canonical
1036 * decimal format (ie "010" is not valid).
1038 size
= *hdr
++ - '0';
1043 unsigned long c
= *hdr
- '0';
1047 size
= size
* 10 + c
;
1055 * The length must be followed by a zero byte
1057 return *hdr
? -1 : type
;
1060 int parse_sha1_header(const char *hdr
, unsigned long *sizep
)
1062 struct object_info oi
= OBJECT_INFO_INIT
;
1065 return parse_sha1_header_extended(hdr
, &oi
, 0);
1068 static int sha1_loose_object_info(const unsigned char *sha1
,
1069 struct object_info
*oi
,
1073 unsigned long mapsize
;
1077 struct strbuf hdrbuf
= STRBUF_INIT
;
1078 unsigned long size_scratch
;
1080 if (oi
->delta_base_sha1
)
1081 hashclr(oi
->delta_base_sha1
);
1084 * If we don't care about type or size, then we don't
1085 * need to look inside the object at all. Note that we
1086 * do not optimize out the stat call, even if the
1087 * caller doesn't care about the disk-size, since our
1088 * return value implicitly indicates whether the
1089 * object even exists.
1091 if (!oi
->typep
&& !oi
->typename
&& !oi
->sizep
&& !oi
->contentp
) {
1094 if (stat_sha1_file(sha1
, &st
, &path
) < 0)
1097 *oi
->disk_sizep
= st
.st_size
;
1101 map
= map_sha1_file(sha1
, &mapsize
);
1106 oi
->sizep
= &size_scratch
;
1109 *oi
->disk_sizep
= mapsize
;
1110 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
)) {
1111 if (unpack_sha1_header_to_strbuf(&stream
, map
, mapsize
, hdr
, sizeof(hdr
), &hdrbuf
) < 0)
1112 status
= error("unable to unpack %s header with --allow-unknown-type",
1114 } else if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1115 status
= error("unable to unpack %s header",
1119 else if (hdrbuf
.len
) {
1120 if ((status
= parse_sha1_header_extended(hdrbuf
.buf
, oi
, flags
)) < 0)
1121 status
= error("unable to parse %s header with --allow-unknown-type",
1123 } else if ((status
= parse_sha1_header_extended(hdr
, oi
, flags
)) < 0)
1124 status
= error("unable to parse %s header", sha1_to_hex(sha1
));
1126 if (status
>= 0 && oi
->contentp
) {
1127 *oi
->contentp
= unpack_sha1_rest(&stream
, hdr
,
1129 if (!*oi
->contentp
) {
1130 git_inflate_end(&stream
);
1134 git_inflate_end(&stream
);
1136 munmap(map
, mapsize
);
1137 if (status
&& oi
->typep
)
1138 *oi
->typep
= status
;
1139 if (oi
->sizep
== &size_scratch
)
1141 strbuf_release(&hdrbuf
);
1142 oi
->whence
= OI_LOOSE
;
1143 return (status
< 0) ? status
: 0;
1146 int sha1_object_info_extended(const unsigned char *sha1
, struct object_info
*oi
, unsigned flags
)
1148 static struct object_info blank_oi
= OBJECT_INFO_INIT
;
1149 struct pack_entry e
;
1151 const unsigned char *real
= (flags
& OBJECT_INFO_LOOKUP_REPLACE
) ?
1152 lookup_replace_object(sha1
) :
1158 if (!(flags
& OBJECT_INFO_SKIP_CACHED
)) {
1159 struct cached_object
*co
= find_cached_object(real
);
1162 *(oi
->typep
) = co
->type
;
1164 *(oi
->sizep
) = co
->size
;
1166 *(oi
->disk_sizep
) = 0;
1167 if (oi
->delta_base_sha1
)
1168 hashclr(oi
->delta_base_sha1
);
1170 strbuf_addstr(oi
->typename
, typename(co
->type
));
1172 *oi
->contentp
= xmemdupz(co
->buf
, co
->size
);
1173 oi
->whence
= OI_CACHED
;
1178 if (!find_pack_entry(real
, &e
)) {
1179 /* Most likely it's a loose object. */
1180 if (!sha1_loose_object_info(real
, oi
, flags
))
1183 /* Not a loose object; someone else may have just packed it. */
1184 if (flags
& OBJECT_INFO_QUICK
) {
1187 reprepare_packed_git();
1188 if (!find_pack_entry(real
, &e
))
1193 if (oi
== &blank_oi
)
1195 * We know that the caller doesn't actually need the
1196 * information below, so return early.
1200 rtype
= packed_object_info(e
.p
, e
.offset
, oi
);
1202 mark_bad_packed_object(e
.p
, real
);
1203 return sha1_object_info_extended(real
, oi
, 0);
1204 } else if (oi
->whence
== OI_PACKED
) {
1205 oi
->u
.packed
.offset
= e
.offset
;
1206 oi
->u
.packed
.pack
= e
.p
;
1207 oi
->u
.packed
.is_delta
= (rtype
== OBJ_REF_DELTA
||
1208 rtype
== OBJ_OFS_DELTA
);
1214 /* returns enum object_type or negative */
1215 int sha1_object_info(const unsigned char *sha1
, unsigned long *sizep
)
1217 enum object_type type
;
1218 struct object_info oi
= OBJECT_INFO_INIT
;
1222 if (sha1_object_info_extended(sha1
, &oi
,
1223 OBJECT_INFO_LOOKUP_REPLACE
) < 0)
1228 static void *read_object(const unsigned char *sha1
, enum object_type
*type
,
1229 unsigned long *size
)
1231 struct object_info oi
= OBJECT_INFO_INIT
;
1235 oi
.contentp
= &content
;
1237 if (sha1_object_info_extended(sha1
, &oi
, 0) < 0)
1242 int pretend_sha1_file(void *buf
, unsigned long len
, enum object_type type
,
1243 unsigned char *sha1
)
1245 struct cached_object
*co
;
1247 hash_sha1_file(buf
, len
, typename(type
), sha1
);
1248 if (has_sha1_file(sha1
) || find_cached_object(sha1
))
1250 ALLOC_GROW(cached_objects
, cached_object_nr
+ 1, cached_object_alloc
);
1251 co
= &cached_objects
[cached_object_nr
++];
1254 co
->buf
= xmalloc(len
);
1255 memcpy(co
->buf
, buf
, len
);
1256 hashcpy(co
->sha1
, sha1
);
1261 * This function dies on corrupt objects; the callers who want to
1262 * deal with them should arrange to call read_object() and give error
1263 * messages themselves.
1265 void *read_sha1_file_extended(const unsigned char *sha1
,
1266 enum object_type
*type
,
1267 unsigned long *size
,
1271 const struct packed_git
*p
;
1274 const unsigned char *repl
= lookup_replace
? lookup_replace_object(sha1
)
1278 data
= read_object(repl
, type
, size
);
1282 if (errno
&& errno
!= ENOENT
)
1283 die_errno("failed to read object %s", sha1_to_hex(sha1
));
1285 /* die if we replaced an object with one that does not exist */
1287 die("replacement %s not found for %s",
1288 sha1_to_hex(repl
), sha1_to_hex(sha1
));
1290 if (!stat_sha1_file(repl
, &st
, &path
))
1291 die("loose object %s (stored in %s) is corrupt",
1292 sha1_to_hex(repl
), path
);
1294 if ((p
= has_packed_and_bad(repl
)) != NULL
)
1295 die("packed object %s (stored in %s) is corrupt",
1296 sha1_to_hex(repl
), p
->pack_name
);
1301 void *read_object_with_reference(const unsigned char *sha1
,
1302 const char *required_type_name
,
1303 unsigned long *size
,
1304 unsigned char *actual_sha1_return
)
1306 enum object_type type
, required_type
;
1308 unsigned long isize
;
1309 unsigned char actual_sha1
[20];
1311 required_type
= type_from_string(required_type_name
);
1312 hashcpy(actual_sha1
, sha1
);
1314 int ref_length
= -1;
1315 const char *ref_type
= NULL
;
1317 buffer
= read_sha1_file(actual_sha1
, &type
, &isize
);
1320 if (type
== required_type
) {
1322 if (actual_sha1_return
)
1323 hashcpy(actual_sha1_return
, actual_sha1
);
1326 /* Handle references */
1327 else if (type
== OBJ_COMMIT
)
1329 else if (type
== OBJ_TAG
)
1330 ref_type
= "object ";
1335 ref_length
= strlen(ref_type
);
1337 if (ref_length
+ 40 > isize
||
1338 memcmp(buffer
, ref_type
, ref_length
) ||
1339 get_sha1_hex((char *) buffer
+ ref_length
, actual_sha1
)) {
1344 /* Now we have the ID of the referred-to object in
1345 * actual_sha1. Check again. */
1349 static void write_sha1_file_prepare(const void *buf
, unsigned long len
,
1350 const char *type
, unsigned char *sha1
,
1351 char *hdr
, int *hdrlen
)
1355 /* Generate the header */
1356 *hdrlen
= xsnprintf(hdr
, *hdrlen
, "%s %lu", type
, len
)+1;
1360 git_SHA1_Update(&c
, hdr
, *hdrlen
);
1361 git_SHA1_Update(&c
, buf
, len
);
1362 git_SHA1_Final(sha1
, &c
);
1366 * Move the just written object into its final resting place.
1368 int finalize_object_file(const char *tmpfile
, const char *filename
)
1372 if (object_creation_mode
== OBJECT_CREATION_USES_RENAMES
)
1374 else if (link(tmpfile
, filename
))
1378 * Coda hack - coda doesn't like cross-directory links,
1379 * so we fall back to a rename, which will mean that it
1380 * won't be able to check collisions, but that's not a
1383 * The same holds for FAT formatted media.
1385 * When this succeeds, we just return. We have nothing
1388 if (ret
&& ret
!= EEXIST
) {
1390 if (!rename(tmpfile
, filename
))
1394 unlink_or_warn(tmpfile
);
1396 if (ret
!= EEXIST
) {
1397 return error_errno("unable to write sha1 filename %s", filename
);
1399 /* FIXME!!! Collision check here ? */
1403 if (adjust_shared_perm(filename
))
1404 return error("unable to set permission to '%s'", filename
);
1408 static int write_buffer(int fd
, const void *buf
, size_t len
)
1410 if (write_in_full(fd
, buf
, len
) < 0)
1411 return error_errno("file write error");
1415 int hash_sha1_file(const void *buf
, unsigned long len
, const char *type
,
1416 unsigned char *sha1
)
1419 int hdrlen
= sizeof(hdr
);
1420 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1424 /* Finalize a file on disk, and close it. */
1425 static void close_sha1_file(int fd
)
1427 if (fsync_object_files
)
1428 fsync_or_die(fd
, "sha1 file");
1430 die_errno("error when closing sha1 file");
1433 /* Size of directory component, including the ending '/' */
1434 static inline int directory_size(const char *filename
)
1436 const char *s
= strrchr(filename
, '/');
1439 return s
- filename
+ 1;
1443 * This creates a temporary file in the same directory as the final
1446 * We want to avoid cross-directory filename renames, because those
1447 * can have problems on various filesystems (FAT, NFS, Coda).
1449 static int create_tmpfile(struct strbuf
*tmp
, const char *filename
)
1451 int fd
, dirlen
= directory_size(filename
);
1454 strbuf_add(tmp
, filename
, dirlen
);
1455 strbuf_addstr(tmp
, "tmp_obj_XXXXXX");
1456 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1457 if (fd
< 0 && dirlen
&& errno
== ENOENT
) {
1459 * Make sure the directory exists; note that the contents
1460 * of the buffer are undefined after mkstemp returns an
1461 * error, so we have to rewrite the whole buffer from
1465 strbuf_add(tmp
, filename
, dirlen
- 1);
1466 if (mkdir(tmp
->buf
, 0777) && errno
!= EEXIST
)
1468 if (adjust_shared_perm(tmp
->buf
))
1472 strbuf_addstr(tmp
, "/tmp_obj_XXXXXX");
1473 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1478 static int write_loose_object(const unsigned char *sha1
, char *hdr
, int hdrlen
,
1479 const void *buf
, unsigned long len
, time_t mtime
)
1482 unsigned char compressed
[4096];
1485 unsigned char parano_sha1
[20];
1486 static struct strbuf tmp_file
= STRBUF_INIT
;
1487 const char *filename
= sha1_file_name(sha1
);
1489 fd
= create_tmpfile(&tmp_file
, filename
);
1491 if (errno
== EACCES
)
1492 return error("insufficient permission for adding an object to repository database %s", get_object_directory());
1494 return error_errno("unable to create temporary file");
1498 git_deflate_init(&stream
, zlib_compression_level
);
1499 stream
.next_out
= compressed
;
1500 stream
.avail_out
= sizeof(compressed
);
1503 /* First header.. */
1504 stream
.next_in
= (unsigned char *)hdr
;
1505 stream
.avail_in
= hdrlen
;
1506 while (git_deflate(&stream
, 0) == Z_OK
)
1508 git_SHA1_Update(&c
, hdr
, hdrlen
);
1510 /* Then the data itself.. */
1511 stream
.next_in
= (void *)buf
;
1512 stream
.avail_in
= len
;
1514 unsigned char *in0
= stream
.next_in
;
1515 ret
= git_deflate(&stream
, Z_FINISH
);
1516 git_SHA1_Update(&c
, in0
, stream
.next_in
- in0
);
1517 if (write_buffer(fd
, compressed
, stream
.next_out
- compressed
) < 0)
1518 die("unable to write sha1 file");
1519 stream
.next_out
= compressed
;
1520 stream
.avail_out
= sizeof(compressed
);
1521 } while (ret
== Z_OK
);
1523 if (ret
!= Z_STREAM_END
)
1524 die("unable to deflate new object %s (%d)", sha1_to_hex(sha1
), ret
);
1525 ret
= git_deflate_end_gently(&stream
);
1527 die("deflateEnd on object %s failed (%d)", sha1_to_hex(sha1
), ret
);
1528 git_SHA1_Final(parano_sha1
, &c
);
1529 if (hashcmp(sha1
, parano_sha1
) != 0)
1530 die("confused by unstable object source data for %s", sha1_to_hex(sha1
));
1532 close_sha1_file(fd
);
1537 utb
.modtime
= mtime
;
1538 if (utime(tmp_file
.buf
, &utb
) < 0)
1539 warning_errno("failed utime() on %s", tmp_file
.buf
);
1542 return finalize_object_file(tmp_file
.buf
, filename
);
1545 static int freshen_loose_object(const unsigned char *sha1
)
1547 return check_and_freshen(sha1
, 1);
1550 static int freshen_packed_object(const unsigned char *sha1
)
1552 struct pack_entry e
;
1553 if (!find_pack_entry(sha1
, &e
))
1557 if (!freshen_file(e
.p
->pack_name
))
1563 int write_sha1_file(const void *buf
, unsigned long len
, const char *type
, unsigned char *sha1
)
1566 int hdrlen
= sizeof(hdr
);
1568 /* Normally if we have it in the pack then we do not bother writing
1569 * it out into .git/objects/??/?{38} file.
1571 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1572 if (freshen_packed_object(sha1
) || freshen_loose_object(sha1
))
1574 return write_loose_object(sha1
, hdr
, hdrlen
, buf
, len
, 0);
1577 int hash_sha1_file_literally(const void *buf
, unsigned long len
, const char *type
,
1578 struct object_id
*oid
, unsigned flags
)
1581 int hdrlen
, status
= 0;
1583 /* type string, SP, %lu of the length plus NUL must fit this */
1584 hdrlen
= strlen(type
) + 32;
1585 header
= xmalloc(hdrlen
);
1586 write_sha1_file_prepare(buf
, len
, type
, oid
->hash
, header
, &hdrlen
);
1588 if (!(flags
& HASH_WRITE_OBJECT
))
1590 if (freshen_packed_object(oid
->hash
) || freshen_loose_object(oid
->hash
))
1592 status
= write_loose_object(oid
->hash
, header
, hdrlen
, buf
, len
, 0);
1599 int force_object_loose(const unsigned char *sha1
, time_t mtime
)
1603 enum object_type type
;
1608 if (has_loose_object(sha1
))
1610 buf
= read_object(sha1
, &type
, &len
);
1612 return error("cannot read sha1_file for %s", sha1_to_hex(sha1
));
1613 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", typename(type
), len
) + 1;
1614 ret
= write_loose_object(sha1
, hdr
, hdrlen
, buf
, len
, mtime
);
1620 int has_sha1_file_with_flags(const unsigned char *sha1
, int flags
)
1622 if (!startup_info
->have_repository
)
1624 return sha1_object_info_extended(sha1
, NULL
,
1625 flags
| OBJECT_INFO_SKIP_CACHED
) >= 0;
1628 int has_object_file(const struct object_id
*oid
)
1630 return has_sha1_file(oid
->hash
);
1633 int has_object_file_with_flags(const struct object_id
*oid
, int flags
)
1635 return has_sha1_file_with_flags(oid
->hash
, flags
);
1638 static void check_tree(const void *buf
, size_t size
)
1640 struct tree_desc desc
;
1641 struct name_entry entry
;
1643 init_tree_desc(&desc
, buf
, size
);
1644 while (tree_entry(&desc
, &entry
))
1646 * tree_entry() will die() on malformed entries */
1650 static void check_commit(const void *buf
, size_t size
)
1653 memset(&c
, 0, sizeof(c
));
1654 if (parse_commit_buffer(&c
, buf
, size
))
1655 die("corrupt commit");
1658 static void check_tag(const void *buf
, size_t size
)
1661 memset(&t
, 0, sizeof(t
));
1662 if (parse_tag_buffer(&t
, buf
, size
))
1666 static int index_mem(struct object_id
*oid
, void *buf
, size_t size
,
1667 enum object_type type
,
1668 const char *path
, unsigned flags
)
1670 int ret
, re_allocated
= 0;
1671 int write_object
= flags
& HASH_WRITE_OBJECT
;
1677 * Convert blobs to git internal format
1679 if ((type
== OBJ_BLOB
) && path
) {
1680 struct strbuf nbuf
= STRBUF_INIT
;
1681 if (convert_to_git(&the_index
, path
, buf
, size
, &nbuf
,
1682 write_object
? safe_crlf
: SAFE_CRLF_FALSE
)) {
1683 buf
= strbuf_detach(&nbuf
, &size
);
1687 if (flags
& HASH_FORMAT_CHECK
) {
1688 if (type
== OBJ_TREE
)
1689 check_tree(buf
, size
);
1690 if (type
== OBJ_COMMIT
)
1691 check_commit(buf
, size
);
1692 if (type
== OBJ_TAG
)
1693 check_tag(buf
, size
);
1697 ret
= write_sha1_file(buf
, size
, typename(type
), oid
->hash
);
1699 ret
= hash_sha1_file(buf
, size
, typename(type
), oid
->hash
);
1705 static int index_stream_convert_blob(struct object_id
*oid
, int fd
,
1706 const char *path
, unsigned flags
)
1709 const int write_object
= flags
& HASH_WRITE_OBJECT
;
1710 struct strbuf sbuf
= STRBUF_INIT
;
1713 assert(would_convert_to_git_filter_fd(path
));
1715 convert_to_git_filter_fd(&the_index
, path
, fd
, &sbuf
,
1716 write_object
? safe_crlf
: SAFE_CRLF_FALSE
);
1719 ret
= write_sha1_file(sbuf
.buf
, sbuf
.len
, typename(OBJ_BLOB
),
1722 ret
= hash_sha1_file(sbuf
.buf
, sbuf
.len
, typename(OBJ_BLOB
),
1724 strbuf_release(&sbuf
);
1728 static int index_pipe(struct object_id
*oid
, int fd
, enum object_type type
,
1729 const char *path
, unsigned flags
)
1731 struct strbuf sbuf
= STRBUF_INIT
;
1734 if (strbuf_read(&sbuf
, fd
, 4096) >= 0)
1735 ret
= index_mem(oid
, sbuf
.buf
, sbuf
.len
, type
, path
, flags
);
1738 strbuf_release(&sbuf
);
1742 #define SMALL_FILE_SIZE (32*1024)
1744 static int index_core(struct object_id
*oid
, int fd
, size_t size
,
1745 enum object_type type
, const char *path
,
1751 ret
= index_mem(oid
, "", size
, type
, path
, flags
);
1752 } else if (size
<= SMALL_FILE_SIZE
) {
1753 char *buf
= xmalloc(size
);
1754 ssize_t read_result
= read_in_full(fd
, buf
, size
);
1755 if (read_result
< 0)
1756 ret
= error_errno("read error while indexing %s",
1757 path
? path
: "<unknown>");
1758 else if (read_result
!= size
)
1759 ret
= error("short read while indexing %s",
1760 path
? path
: "<unknown>");
1762 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1765 void *buf
= xmmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1766 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1773 * This creates one packfile per large blob unless bulk-checkin
1774 * machinery is "plugged".
1776 * This also bypasses the usual "convert-to-git" dance, and that is on
1777 * purpose. We could write a streaming version of the converting
1778 * functions and insert that before feeding the data to fast-import
1779 * (or equivalent in-core API described above). However, that is
1780 * somewhat complicated, as we do not know the size of the filter
1781 * result, which we need to know beforehand when writing a git object.
1782 * Since the primary motivation for trying to stream from the working
1783 * tree file and to avoid mmaping it in core is to deal with large
1784 * binary blobs, they generally do not want to get any conversion, and
1785 * callers should avoid this code path when filters are requested.
1787 static int index_stream(struct object_id
*oid
, int fd
, size_t size
,
1788 enum object_type type
, const char *path
,
1791 return index_bulk_checkin(oid
->hash
, fd
, size
, type
, path
, flags
);
1794 int index_fd(struct object_id
*oid
, int fd
, struct stat
*st
,
1795 enum object_type type
, const char *path
, unsigned flags
)
1800 * Call xsize_t() only when needed to avoid potentially unnecessary
1801 * die() for large files.
1803 if (type
== OBJ_BLOB
&& path
&& would_convert_to_git_filter_fd(path
))
1804 ret
= index_stream_convert_blob(oid
, fd
, path
, flags
);
1805 else if (!S_ISREG(st
->st_mode
))
1806 ret
= index_pipe(oid
, fd
, type
, path
, flags
);
1807 else if (st
->st_size
<= big_file_threshold
|| type
!= OBJ_BLOB
||
1808 (path
&& would_convert_to_git(&the_index
, path
)))
1809 ret
= index_core(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1812 ret
= index_stream(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1818 int index_path(struct object_id
*oid
, const char *path
, struct stat
*st
, unsigned flags
)
1821 struct strbuf sb
= STRBUF_INIT
;
1824 switch (st
->st_mode
& S_IFMT
) {
1826 fd
= open(path
, O_RDONLY
);
1828 return error_errno("open(\"%s\")", path
);
1829 if (index_fd(oid
, fd
, st
, OBJ_BLOB
, path
, flags
) < 0)
1830 return error("%s: failed to insert into database",
1834 if (strbuf_readlink(&sb
, path
, st
->st_size
))
1835 return error_errno("readlink(\"%s\")", path
);
1836 if (!(flags
& HASH_WRITE_OBJECT
))
1837 hash_sha1_file(sb
.buf
, sb
.len
, blob_type
, oid
->hash
);
1838 else if (write_sha1_file(sb
.buf
, sb
.len
, blob_type
, oid
->hash
))
1839 rc
= error("%s: failed to insert into database", path
);
1840 strbuf_release(&sb
);
1843 return resolve_gitlink_ref(path
, "HEAD", oid
);
1845 return error("%s: unsupported file type", path
);
1850 int read_pack_header(int fd
, struct pack_header
*header
)
1852 if (read_in_full(fd
, header
, sizeof(*header
)) != sizeof(*header
))
1853 /* "eof before pack header was fully read" */
1854 return PH_ERROR_EOF
;
1856 if (header
->hdr_signature
!= htonl(PACK_SIGNATURE
))
1857 /* "protocol error (pack signature mismatch detected)" */
1858 return PH_ERROR_PACK_SIGNATURE
;
1859 if (!pack_version_ok(header
->hdr_version
))
1860 /* "protocol error (pack version unsupported)" */
1861 return PH_ERROR_PROTOCOL
;
1865 void assert_sha1_type(const unsigned char *sha1
, enum object_type expect
)
1867 enum object_type type
= sha1_object_info(sha1
, NULL
);
1869 die("%s is not a valid object", sha1_to_hex(sha1
));
1871 die("%s is not a valid '%s' object", sha1_to_hex(sha1
),
1875 int for_each_file_in_obj_subdir(unsigned int subdir_nr
,
1876 struct strbuf
*path
,
1877 each_loose_object_fn obj_cb
,
1878 each_loose_cruft_fn cruft_cb
,
1879 each_loose_subdir_fn subdir_cb
,
1882 size_t origlen
, baselen
;
1886 struct object_id oid
;
1888 if (subdir_nr
> 0xff)
1889 BUG("invalid loose object subdirectory: %x", subdir_nr
);
1891 origlen
= path
->len
;
1892 strbuf_complete(path
, '/');
1893 strbuf_addf(path
, "%02x", subdir_nr
);
1894 baselen
= path
->len
;
1896 dir
= opendir(path
->buf
);
1898 if (errno
!= ENOENT
)
1899 r
= error_errno("unable to open %s", path
->buf
);
1900 strbuf_setlen(path
, origlen
);
1904 oid
.hash
[0] = subdir_nr
;
1906 while ((de
= readdir(dir
))) {
1907 if (is_dot_or_dotdot(de
->d_name
))
1910 strbuf_setlen(path
, baselen
);
1911 strbuf_addf(path
, "/%s", de
->d_name
);
1913 if (strlen(de
->d_name
) == GIT_SHA1_HEXSZ
- 2 &&
1914 !hex_to_bytes(oid
.hash
+ 1, de
->d_name
,
1915 GIT_SHA1_RAWSZ
- 1)) {
1917 r
= obj_cb(&oid
, path
->buf
, data
);
1925 r
= cruft_cb(de
->d_name
, path
->buf
, data
);
1932 strbuf_setlen(path
, baselen
);
1933 if (!r
&& subdir_cb
)
1934 r
= subdir_cb(subdir_nr
, path
->buf
, data
);
1936 strbuf_setlen(path
, origlen
);
1941 int for_each_loose_file_in_objdir_buf(struct strbuf
*path
,
1942 each_loose_object_fn obj_cb
,
1943 each_loose_cruft_fn cruft_cb
,
1944 each_loose_subdir_fn subdir_cb
,
1950 for (i
= 0; i
< 256; i
++) {
1951 r
= for_each_file_in_obj_subdir(i
, path
, obj_cb
, cruft_cb
,
1960 int for_each_loose_file_in_objdir(const char *path
,
1961 each_loose_object_fn obj_cb
,
1962 each_loose_cruft_fn cruft_cb
,
1963 each_loose_subdir_fn subdir_cb
,
1966 struct strbuf buf
= STRBUF_INIT
;
1969 strbuf_addstr(&buf
, path
);
1970 r
= for_each_loose_file_in_objdir_buf(&buf
, obj_cb
, cruft_cb
,
1972 strbuf_release(&buf
);
1977 struct loose_alt_odb_data
{
1978 each_loose_object_fn
*cb
;
1982 static int loose_from_alt_odb(struct alternate_object_database
*alt
,
1985 struct loose_alt_odb_data
*data
= vdata
;
1986 struct strbuf buf
= STRBUF_INIT
;
1989 strbuf_addstr(&buf
, alt
->path
);
1990 r
= for_each_loose_file_in_objdir_buf(&buf
,
1991 data
->cb
, NULL
, NULL
,
1993 strbuf_release(&buf
);
1997 int for_each_loose_object(each_loose_object_fn cb
, void *data
, unsigned flags
)
1999 struct loose_alt_odb_data alt
;
2002 r
= for_each_loose_file_in_objdir(get_object_directory(),
2003 cb
, NULL
, NULL
, data
);
2007 if (flags
& FOR_EACH_OBJECT_LOCAL_ONLY
)
2012 return foreach_alt_odb(loose_from_alt_odb
, &alt
);
2015 static int check_stream_sha1(git_zstream
*stream
,
2019 const unsigned char *expected_sha1
)
2022 unsigned char real_sha1
[GIT_MAX_RAWSZ
];
2023 unsigned char buf
[4096];
2024 unsigned long total_read
;
2028 git_SHA1_Update(&c
, hdr
, stream
->total_out
);
2031 * We already read some bytes into hdr, but the ones up to the NUL
2032 * do not count against the object's content size.
2034 total_read
= stream
->total_out
- strlen(hdr
) - 1;
2037 * This size comparison must be "<=" to read the final zlib packets;
2038 * see the comment in unpack_sha1_rest for details.
2040 while (total_read
<= size
&&
2041 (status
== Z_OK
|| status
== Z_BUF_ERROR
)) {
2042 stream
->next_out
= buf
;
2043 stream
->avail_out
= sizeof(buf
);
2044 if (size
- total_read
< stream
->avail_out
)
2045 stream
->avail_out
= size
- total_read
;
2046 status
= git_inflate(stream
, Z_FINISH
);
2047 git_SHA1_Update(&c
, buf
, stream
->next_out
- buf
);
2048 total_read
+= stream
->next_out
- buf
;
2050 git_inflate_end(stream
);
2052 if (status
!= Z_STREAM_END
) {
2053 error("corrupt loose object '%s'", sha1_to_hex(expected_sha1
));
2056 if (stream
->avail_in
) {
2057 error("garbage at end of loose object '%s'",
2058 sha1_to_hex(expected_sha1
));
2062 git_SHA1_Final(real_sha1
, &c
);
2063 if (hashcmp(expected_sha1
, real_sha1
)) {
2064 error("sha1 mismatch for %s (expected %s)", path
,
2065 sha1_to_hex(expected_sha1
));
2072 int read_loose_object(const char *path
,
2073 const unsigned char *expected_sha1
,
2074 enum object_type
*type
,
2075 unsigned long *size
,
2080 unsigned long mapsize
;
2086 map
= map_sha1_file_1(path
, NULL
, &mapsize
);
2088 error_errno("unable to mmap %s", path
);
2092 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0) {
2093 error("unable to unpack header of %s", path
);
2097 *type
= parse_sha1_header(hdr
, size
);
2099 error("unable to parse header of %s", path
);
2100 git_inflate_end(&stream
);
2104 if (*type
== OBJ_BLOB
) {
2105 if (check_stream_sha1(&stream
, hdr
, *size
, path
, expected_sha1
) < 0)
2108 *contents
= unpack_sha1_rest(&stream
, hdr
, *size
, expected_sha1
);
2110 error("unable to unpack contents of %s", path
);
2111 git_inflate_end(&stream
);
2114 if (check_sha1_signature(expected_sha1
, *contents
,
2115 *size
, typename(*type
))) {
2116 error("sha1 mismatch for %s (expected %s)", path
,
2117 sha1_to_hex(expected_sha1
));
2123 ret
= 0; /* everything checks out */
2127 munmap(map
, mapsize
);