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
42 static void git_hash_sha1_init(git_hash_ctx
*ctx
)
44 git_SHA1_Init(&ctx
->sha1
);
47 static void git_hash_sha1_update(git_hash_ctx
*ctx
, const void *data
, size_t len
)
49 git_SHA1_Update(&ctx
->sha1
, data
, len
);
52 static void git_hash_sha1_final(unsigned char *hash
, git_hash_ctx
*ctx
)
54 git_SHA1_Final(hash
, &ctx
->sha1
);
57 static void git_hash_unknown_init(git_hash_ctx
*ctx
)
59 die("trying to init unknown hash");
62 static void git_hash_unknown_update(git_hash_ctx
*ctx
, const void *data
, size_t len
)
64 die("trying to update unknown hash");
67 static void git_hash_unknown_final(unsigned char *hash
, git_hash_ctx
*ctx
)
69 die("trying to finalize unknown hash");
72 const struct git_hash_algo hash_algos
[GIT_HASH_NALGOS
] = {
78 git_hash_unknown_init
,
79 git_hash_unknown_update
,
80 git_hash_unknown_final
,
86 /* "sha1", big-endian */
99 * This is meant to hold a *small* number of objects that you would
100 * want read_sha1_file() to be able to return, but yet you do not want
101 * to write them into the object store (e.g. a browse-only
104 static struct cached_object
{
105 unsigned char sha1
[20];
106 enum object_type type
;
110 static int cached_object_nr
, cached_object_alloc
;
112 static struct cached_object empty_tree
= {
113 EMPTY_TREE_SHA1_BIN_LITERAL
,
119 static struct cached_object
*find_cached_object(const unsigned char *sha1
)
122 struct cached_object
*co
= cached_objects
;
124 for (i
= 0; i
< cached_object_nr
; i
++, co
++) {
125 if (!hashcmp(co
->sha1
, sha1
))
128 if (!hashcmp(sha1
, empty_tree
.sha1
))
134 static enum safe_crlf
get_safe_crlf(unsigned flags
)
136 if (flags
& HASH_RENORMALIZE
)
137 return SAFE_CRLF_RENORMALIZE
;
138 else if (flags
& HASH_WRITE_OBJECT
)
141 return SAFE_CRLF_FALSE
;
145 int mkdir_in_gitdir(const char *path
)
147 if (mkdir(path
, 0777)) {
148 int saved_errno
= errno
;
150 struct strbuf sb
= STRBUF_INIT
;
155 * Are we looking at a path in a symlinked worktree
156 * whose original repository does not yet have it?
157 * e.g. .git/rr-cache pointing at its original
158 * repository in which the user hasn't performed any
159 * conflict resolution yet?
161 if (lstat(path
, &st
) || !S_ISLNK(st
.st_mode
) ||
162 strbuf_readlink(&sb
, path
, st
.st_size
) ||
163 !is_absolute_path(sb
.buf
) ||
164 mkdir(sb
.buf
, 0777)) {
171 return adjust_shared_perm(path
);
174 enum scld_error
safe_create_leading_directories(char *path
)
176 char *next_component
= path
+ offset_1st_component(path
);
177 enum scld_error ret
= SCLD_OK
;
179 while (ret
== SCLD_OK
&& next_component
) {
181 char *slash
= next_component
, slash_character
;
183 while (*slash
&& !is_dir_sep(*slash
))
189 next_component
= slash
+ 1;
190 while (is_dir_sep(*next_component
))
192 if (!*next_component
)
195 slash_character
= *slash
;
197 if (!stat(path
, &st
)) {
199 if (!S_ISDIR(st
.st_mode
)) {
203 } else if (mkdir(path
, 0777)) {
204 if (errno
== EEXIST
&&
205 !stat(path
, &st
) && S_ISDIR(st
.st_mode
))
206 ; /* somebody created it since we checked */
207 else if (errno
== ENOENT
)
209 * Either mkdir() failed because
210 * somebody just pruned the containing
211 * directory, or stat() failed because
212 * the file that was in our way was
213 * just removed. Either way, inform
214 * the caller that it might be worth
220 } else if (adjust_shared_perm(path
)) {
223 *slash
= slash_character
;
228 enum scld_error
safe_create_leading_directories_const(const char *path
)
231 /* path points to cache entries, so xstrdup before messing with it */
232 char *buf
= xstrdup(path
);
233 enum scld_error result
= safe_create_leading_directories(buf
);
241 int raceproof_create_file(const char *path
, create_file_fn fn
, void *cb
)
244 * The number of times we will try to remove empty directories
245 * in the way of path. This is only 1 because if another
246 * process is racily creating directories that conflict with
247 * us, we don't want to fight against them.
249 int remove_directories_remaining
= 1;
252 * The number of times that we will try to create the
253 * directories containing path. We are willing to attempt this
254 * more than once, because another process could be trying to
255 * clean up empty directories at the same time as we are
256 * trying to create them.
258 int create_directories_remaining
= 3;
260 /* A scratch copy of path, filled lazily if we need it: */
261 struct strbuf path_copy
= STRBUF_INIT
;
274 if (errno
== EISDIR
&& remove_directories_remaining
-- > 0) {
276 * A directory is in the way. Maybe it is empty; try
280 strbuf_addstr(&path_copy
, path
);
282 if (!remove_dir_recursively(&path_copy
, REMOVE_DIR_EMPTY_ONLY
))
284 } else if (errno
== ENOENT
&& create_directories_remaining
-- > 0) {
286 * Maybe the containing directory didn't exist, or
287 * maybe it was just deleted by a process that is
288 * racing with us to clean up empty directories. Try
291 enum scld_error scld_result
;
294 strbuf_addstr(&path_copy
, path
);
297 scld_result
= safe_create_leading_directories(path_copy
.buf
);
298 if (scld_result
== SCLD_OK
)
300 } while (scld_result
== SCLD_VANISHED
&& create_directories_remaining
-- > 0);
304 strbuf_release(&path_copy
);
309 static void fill_sha1_path(struct strbuf
*buf
, const unsigned char *sha1
)
312 for (i
= 0; i
< 20; i
++) {
313 static char hex
[] = "0123456789abcdef";
314 unsigned int val
= sha1
[i
];
315 strbuf_addch(buf
, hex
[val
>> 4]);
316 strbuf_addch(buf
, hex
[val
& 0xf]);
318 strbuf_addch(buf
, '/');
322 const char *sha1_file_name(const unsigned char *sha1
)
324 static struct strbuf buf
= STRBUF_INIT
;
327 strbuf_addf(&buf
, "%s/", get_object_directory());
329 fill_sha1_path(&buf
, sha1
);
333 struct strbuf
*alt_scratch_buf(struct alternate_object_database
*alt
)
335 strbuf_setlen(&alt
->scratch
, alt
->base_len
);
336 return &alt
->scratch
;
339 static const char *alt_sha1_path(struct alternate_object_database
*alt
,
340 const unsigned char *sha1
)
342 struct strbuf
*buf
= alt_scratch_buf(alt
);
343 fill_sha1_path(buf
, sha1
);
347 struct alternate_object_database
*alt_odb_list
;
348 static struct alternate_object_database
**alt_odb_tail
;
351 * Return non-zero iff the path is usable as an alternate object database.
353 static int alt_odb_usable(struct strbuf
*path
, const char *normalized_objdir
)
355 struct alternate_object_database
*alt
;
357 /* Detect cases where alternate disappeared */
358 if (!is_directory(path
->buf
)) {
359 error("object directory %s does not exist; "
360 "check .git/objects/info/alternates.",
366 * Prevent the common mistake of listing the same
367 * thing twice, or object directory itself.
369 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
370 if (!fspathcmp(path
->buf
, alt
->path
))
373 if (!fspathcmp(path
->buf
, normalized_objdir
))
380 * Prepare alternate object database registry.
382 * The variable alt_odb_list points at the list of struct
383 * alternate_object_database. The elements on this list come from
384 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
385 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
386 * whose contents is similar to that environment variable but can be
387 * LF separated. Its base points at a statically allocated buffer that
388 * contains "/the/directory/corresponding/to/.git/objects/...", while
389 * its name points just after the slash at the end of ".git/objects/"
390 * in the example above, and has enough space to hold 40-byte hex
391 * SHA1, an extra slash for the first level indirection, and the
394 static void read_info_alternates(const char * relative_base
, int depth
);
395 static int link_alt_odb_entry(const char *entry
, const char *relative_base
,
396 int depth
, const char *normalized_objdir
)
398 struct alternate_object_database
*ent
;
399 struct strbuf pathbuf
= STRBUF_INIT
;
401 if (!is_absolute_path(entry
) && relative_base
) {
402 strbuf_realpath(&pathbuf
, relative_base
, 1);
403 strbuf_addch(&pathbuf
, '/');
405 strbuf_addstr(&pathbuf
, entry
);
407 if (strbuf_normalize_path(&pathbuf
) < 0 && relative_base
) {
408 error("unable to normalize alternate object path: %s",
410 strbuf_release(&pathbuf
);
415 * The trailing slash after the directory name is given by
416 * this function at the end. Remove duplicates.
418 while (pathbuf
.len
&& pathbuf
.buf
[pathbuf
.len
- 1] == '/')
419 strbuf_setlen(&pathbuf
, pathbuf
.len
- 1);
421 if (!alt_odb_usable(&pathbuf
, normalized_objdir
)) {
422 strbuf_release(&pathbuf
);
426 ent
= alloc_alt_odb(pathbuf
.buf
);
428 /* add the alternate entry */
430 alt_odb_tail
= &(ent
->next
);
433 /* recursively add alternates */
434 read_info_alternates(pathbuf
.buf
, depth
+ 1);
436 strbuf_release(&pathbuf
);
440 static const char *parse_alt_odb_entry(const char *string
,
448 if (*string
== '#') {
449 /* comment; consume up to next separator */
450 end
= strchrnul(string
, sep
);
451 } else if (*string
== '"' && !unquote_c_style(out
, string
, &end
)) {
453 * quoted path; unquote_c_style has copied the
454 * data for us and set "end". Broken quoting (e.g.,
455 * an entry that doesn't end with a quote) falls
456 * back to the unquoted case below.
459 /* normal, unquoted path */
460 end
= strchrnul(string
, sep
);
461 strbuf_add(out
, string
, end
- string
);
469 static void link_alt_odb_entries(const char *alt
, int sep
,
470 const char *relative_base
, int depth
)
472 struct strbuf objdirbuf
= STRBUF_INIT
;
473 struct strbuf entry
= STRBUF_INIT
;
479 error("%s: ignoring alternate object stores, nesting too deep.",
484 strbuf_add_absolute_path(&objdirbuf
, get_object_directory());
485 if (strbuf_normalize_path(&objdirbuf
) < 0)
486 die("unable to normalize object directory: %s",
490 alt
= parse_alt_odb_entry(alt
, sep
, &entry
);
493 link_alt_odb_entry(entry
.buf
, relative_base
, depth
, objdirbuf
.buf
);
495 strbuf_release(&entry
);
496 strbuf_release(&objdirbuf
);
499 static void read_info_alternates(const char * relative_base
, int depth
)
502 struct strbuf buf
= STRBUF_INIT
;
504 path
= xstrfmt("%s/info/alternates", relative_base
);
505 if (strbuf_read_file(&buf
, path
, 1024) < 0) {
506 warn_on_fopen_errors(path
);
511 link_alt_odb_entries(buf
.buf
, '\n', relative_base
, depth
);
512 strbuf_release(&buf
);
516 struct alternate_object_database
*alloc_alt_odb(const char *dir
)
518 struct alternate_object_database
*ent
;
520 FLEX_ALLOC_STR(ent
, path
, dir
);
521 strbuf_init(&ent
->scratch
, 0);
522 strbuf_addf(&ent
->scratch
, "%s/", dir
);
523 ent
->base_len
= ent
->scratch
.len
;
528 void add_to_alternates_file(const char *reference
)
530 struct lock_file lock
= LOCK_INIT
;
531 char *alts
= git_pathdup("objects/info/alternates");
535 hold_lock_file_for_update(&lock
, alts
, LOCK_DIE_ON_ERROR
);
536 out
= fdopen_lock_file(&lock
, "w");
538 die_errno("unable to fdopen alternates lockfile");
540 in
= fopen(alts
, "r");
542 struct strbuf line
= STRBUF_INIT
;
544 while (strbuf_getline(&line
, in
) != EOF
) {
545 if (!strcmp(reference
, line
.buf
)) {
549 fprintf_or_die(out
, "%s\n", line
.buf
);
552 strbuf_release(&line
);
555 else if (errno
!= ENOENT
)
556 die_errno("unable to read alternates file");
559 rollback_lock_file(&lock
);
561 fprintf_or_die(out
, "%s\n", reference
);
562 if (commit_lock_file(&lock
))
563 die_errno("unable to move new alternates file into place");
565 link_alt_odb_entries(reference
, '\n', NULL
, 0);
570 void add_to_alternates_memory(const char *reference
)
573 * Make sure alternates are initialized, or else our entry may be
574 * overwritten when they are.
578 link_alt_odb_entries(reference
, '\n', NULL
, 0);
582 * Compute the exact path an alternate is at and returns it. In case of
583 * error NULL is returned and the human readable error is added to `err`
584 * `path` may be relative and should point to $GITDIR.
585 * `err` must not be null.
587 char *compute_alternate_path(const char *path
, struct strbuf
*err
)
589 char *ref_git
= NULL
;
590 const char *repo
, *ref_git_s
;
593 ref_git_s
= real_path_if_valid(path
);
596 strbuf_addf(err
, _("path '%s' does not exist"), path
);
600 * Beware: read_gitfile(), real_path() and mkpath()
601 * return static buffer
603 ref_git
= xstrdup(ref_git_s
);
605 repo
= read_gitfile(ref_git
);
607 repo
= read_gitfile(mkpath("%s/.git", ref_git
));
610 ref_git
= xstrdup(repo
);
613 if (!repo
&& is_directory(mkpath("%s/.git/objects", ref_git
))) {
614 char *ref_git_git
= mkpathdup("%s/.git", ref_git
);
616 ref_git
= ref_git_git
;
617 } else if (!is_directory(mkpath("%s/objects", ref_git
))) {
618 struct strbuf sb
= STRBUF_INIT
;
620 if (get_common_dir(&sb
, ref_git
)) {
622 _("reference repository '%s' as a linked "
623 "checkout is not supported yet."),
628 strbuf_addf(err
, _("reference repository '%s' is not a "
629 "local repository."), path
);
633 if (!access(mkpath("%s/shallow", ref_git
), F_OK
)) {
634 strbuf_addf(err
, _("reference repository '%s' is shallow"),
640 if (!access(mkpath("%s/info/grafts", ref_git
), F_OK
)) {
642 _("reference repository '%s' is grafted"),
650 FREE_AND_NULL(ref_git
);
656 int foreach_alt_odb(alt_odb_fn fn
, void *cb
)
658 struct alternate_object_database
*ent
;
662 for (ent
= alt_odb_list
; ent
; ent
= ent
->next
) {
670 void prepare_alt_odb(void)
677 alt
= getenv(ALTERNATE_DB_ENVIRONMENT
);
679 alt_odb_tail
= &alt_odb_list
;
680 link_alt_odb_entries(alt
, PATH_SEP
, NULL
, 0);
682 read_info_alternates(get_object_directory(), 0);
685 /* Returns 1 if we have successfully freshened the file, 0 otherwise. */
686 static int freshen_file(const char *fn
)
689 t
.actime
= t
.modtime
= time(NULL
);
690 return !utime(fn
, &t
);
694 * All of the check_and_freshen functions return 1 if the file exists and was
695 * freshened (if freshening was requested), 0 otherwise. If they return
696 * 0, you should not assume that it is safe to skip a write of the object (it
697 * either does not exist on disk, or has a stale mtime and may be subject to
700 int check_and_freshen_file(const char *fn
, int freshen
)
702 if (access(fn
, F_OK
))
704 if (freshen
&& !freshen_file(fn
))
709 static int check_and_freshen_local(const unsigned char *sha1
, int freshen
)
711 return check_and_freshen_file(sha1_file_name(sha1
), freshen
);
714 static int check_and_freshen_nonlocal(const unsigned char *sha1
, int freshen
)
716 struct alternate_object_database
*alt
;
718 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
719 const char *path
= alt_sha1_path(alt
, sha1
);
720 if (check_and_freshen_file(path
, freshen
))
726 static int check_and_freshen(const unsigned char *sha1
, int freshen
)
728 return check_and_freshen_local(sha1
, freshen
) ||
729 check_and_freshen_nonlocal(sha1
, freshen
);
732 int has_loose_object_nonlocal(const unsigned char *sha1
)
734 return check_and_freshen_nonlocal(sha1
, 0);
737 static int has_loose_object(const unsigned char *sha1
)
739 return check_and_freshen(sha1
, 0);
742 static void mmap_limit_check(size_t length
)
744 static size_t limit
= 0;
746 limit
= git_env_ulong("GIT_MMAP_LIMIT", 0);
751 die("attempting to mmap %"PRIuMAX
" over limit %"PRIuMAX
,
752 (uintmax_t)length
, (uintmax_t)limit
);
755 void *xmmap_gently(void *start
, size_t length
,
756 int prot
, int flags
, int fd
, off_t offset
)
760 mmap_limit_check(length
);
761 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
762 if (ret
== MAP_FAILED
) {
765 release_pack_memory(length
);
766 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
771 void *xmmap(void *start
, size_t length
,
772 int prot
, int flags
, int fd
, off_t offset
)
774 void *ret
= xmmap_gently(start
, length
, prot
, flags
, fd
, offset
);
775 if (ret
== MAP_FAILED
)
776 die_errno("mmap failed");
781 * With an in-core object data in "map", rehash it to make sure the
782 * object name actually matches "sha1" to detect object corruption.
783 * With "map" == NULL, try reading the object named with "sha1" using
784 * the streaming interface and rehash it to do the same.
786 int check_sha1_signature(const unsigned char *sha1
, void *map
,
787 unsigned long size
, const char *type
)
789 struct object_id real_oid
;
790 enum object_type obj_type
;
791 struct git_istream
*st
;
797 hash_sha1_file(map
, size
, type
, real_oid
.hash
);
798 return hashcmp(sha1
, real_oid
.hash
) ? -1 : 0;
801 st
= open_istream(sha1
, &obj_type
, &size
, NULL
);
805 /* Generate the header */
806 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", typename(obj_type
), size
) + 1;
809 the_hash_algo
->init_fn(&c
);
810 the_hash_algo
->update_fn(&c
, hdr
, hdrlen
);
813 ssize_t readlen
= read_istream(st
, buf
, sizeof(buf
));
821 the_hash_algo
->update_fn(&c
, buf
, readlen
);
823 the_hash_algo
->final_fn(real_oid
.hash
, &c
);
825 return hashcmp(sha1
, real_oid
.hash
) ? -1 : 0;
828 int git_open_cloexec(const char *name
, int flags
)
831 static int o_cloexec
= O_CLOEXEC
;
833 fd
= open(name
, flags
| o_cloexec
);
834 if ((o_cloexec
& O_CLOEXEC
) && fd
< 0 && errno
== EINVAL
) {
835 /* Try again w/o O_CLOEXEC: the kernel might not support it */
836 o_cloexec
&= ~O_CLOEXEC
;
837 fd
= open(name
, flags
| o_cloexec
);
840 #if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
842 static int fd_cloexec
= FD_CLOEXEC
;
844 if (!o_cloexec
&& 0 <= fd
&& fd_cloexec
) {
845 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
846 int flags
= fcntl(fd
, F_GETFD
);
847 if (fcntl(fd
, F_SETFD
, flags
| fd_cloexec
))
856 * Find "sha1" as a loose object in the local repository or in an alternate.
857 * Returns 0 on success, negative on failure.
859 * The "path" out-parameter will give the path of the object we found (if any).
860 * Note that it may point to static storage and is only valid until another
861 * call to sha1_file_name(), etc.
863 static int stat_sha1_file(const unsigned char *sha1
, struct stat
*st
,
866 struct alternate_object_database
*alt
;
868 *path
= sha1_file_name(sha1
);
869 if (!lstat(*path
, st
))
874 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
875 *path
= alt_sha1_path(alt
, sha1
);
876 if (!lstat(*path
, st
))
884 * Like stat_sha1_file(), but actually open the object and return the
885 * descriptor. See the caveats on the "path" parameter above.
887 static int open_sha1_file(const unsigned char *sha1
, const char **path
)
890 struct alternate_object_database
*alt
;
891 int most_interesting_errno
;
893 *path
= sha1_file_name(sha1
);
894 fd
= git_open(*path
);
897 most_interesting_errno
= errno
;
900 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
901 *path
= alt_sha1_path(alt
, sha1
);
902 fd
= git_open(*path
);
905 if (most_interesting_errno
== ENOENT
)
906 most_interesting_errno
= errno
;
908 errno
= most_interesting_errno
;
913 * Map the loose object at "path" if it is not NULL, or the path found by
914 * searching for a loose object named "sha1".
916 static void *map_sha1_file_1(const char *path
,
917 const unsigned char *sha1
,
926 fd
= open_sha1_file(sha1
, &path
);
931 if (!fstat(fd
, &st
)) {
932 *size
= xsize_t(st
.st_size
);
934 /* mmap() is forbidden on empty files */
935 error("object file %s is empty", path
);
938 map
= xmmap(NULL
, *size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
945 void *map_sha1_file(const unsigned char *sha1
, unsigned long *size
)
947 return map_sha1_file_1(NULL
, sha1
, size
);
950 static int unpack_sha1_short_header(git_zstream
*stream
,
951 unsigned char *map
, unsigned long mapsize
,
952 void *buffer
, unsigned long bufsiz
)
954 /* Get the data stream */
955 memset(stream
, 0, sizeof(*stream
));
956 stream
->next_in
= map
;
957 stream
->avail_in
= mapsize
;
958 stream
->next_out
= buffer
;
959 stream
->avail_out
= bufsiz
;
961 git_inflate_init(stream
);
962 return git_inflate(stream
, 0);
965 int unpack_sha1_header(git_zstream
*stream
,
966 unsigned char *map
, unsigned long mapsize
,
967 void *buffer
, unsigned long bufsiz
)
969 int status
= unpack_sha1_short_header(stream
, map
, mapsize
,
975 /* Make sure we have the terminating NUL */
976 if (!memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
981 static int unpack_sha1_header_to_strbuf(git_zstream
*stream
, unsigned char *map
,
982 unsigned long mapsize
, void *buffer
,
983 unsigned long bufsiz
, struct strbuf
*header
)
987 status
= unpack_sha1_short_header(stream
, map
, mapsize
, buffer
, bufsiz
);
992 * Check if entire header is unpacked in the first iteration.
994 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
998 * buffer[0..bufsiz] was not large enough. Copy the partial
999 * result out to header, and then append the result of further
1000 * reading the stream.
1002 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
1003 stream
->next_out
= buffer
;
1004 stream
->avail_out
= bufsiz
;
1007 status
= git_inflate(stream
, 0);
1008 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
1009 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
1011 stream
->next_out
= buffer
;
1012 stream
->avail_out
= bufsiz
;
1013 } while (status
!= Z_STREAM_END
);
1017 static void *unpack_sha1_rest(git_zstream
*stream
, void *buffer
, unsigned long size
, const unsigned char *sha1
)
1019 int bytes
= strlen(buffer
) + 1;
1020 unsigned char *buf
= xmallocz(size
);
1024 n
= stream
->total_out
- bytes
;
1027 memcpy(buf
, (char *) buffer
+ bytes
, n
);
1029 if (bytes
<= size
) {
1031 * The above condition must be (bytes <= size), not
1032 * (bytes < size). In other words, even though we
1033 * expect no more output and set avail_out to zero,
1034 * the input zlib stream may have bytes that express
1035 * "this concludes the stream", and we *do* want to
1038 * Otherwise we would not be able to test that we
1039 * consumed all the input to reach the expected size;
1040 * we also want to check that zlib tells us that all
1041 * went well with status == Z_STREAM_END at the end.
1043 stream
->next_out
= buf
+ bytes
;
1044 stream
->avail_out
= size
- bytes
;
1045 while (status
== Z_OK
)
1046 status
= git_inflate(stream
, Z_FINISH
);
1048 if (status
== Z_STREAM_END
&& !stream
->avail_in
) {
1049 git_inflate_end(stream
);
1054 error("corrupt loose object '%s'", sha1_to_hex(sha1
));
1055 else if (stream
->avail_in
)
1056 error("garbage at end of loose object '%s'",
1063 * We used to just use "sscanf()", but that's actually way
1064 * too permissive for what we want to check. So do an anal
1065 * object header parse by hand.
1067 static int parse_sha1_header_extended(const char *hdr
, struct object_info
*oi
,
1070 const char *type_buf
= hdr
;
1072 int type
, type_len
= 0;
1075 * The type can be of any size but is followed by
1087 type
= type_from_string_gently(type_buf
, type_len
, 1);
1089 strbuf_add(oi
->typename
, type_buf
, type_len
);
1091 * Set type to 0 if its an unknown object and
1092 * we're obtaining the type using '--allow-unknown-type'
1095 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
) && (type
< 0))
1098 die("invalid object type");
1103 * The length must follow immediately, and be in canonical
1104 * decimal format (ie "010" is not valid).
1106 size
= *hdr
++ - '0';
1111 unsigned long c
= *hdr
- '0';
1115 size
= size
* 10 + c
;
1123 * The length must be followed by a zero byte
1125 return *hdr
? -1 : type
;
1128 int parse_sha1_header(const char *hdr
, unsigned long *sizep
)
1130 struct object_info oi
= OBJECT_INFO_INIT
;
1133 return parse_sha1_header_extended(hdr
, &oi
, 0);
1136 static int sha1_loose_object_info(const unsigned char *sha1
,
1137 struct object_info
*oi
,
1141 unsigned long mapsize
;
1145 struct strbuf hdrbuf
= STRBUF_INIT
;
1146 unsigned long size_scratch
;
1148 if (oi
->delta_base_sha1
)
1149 hashclr(oi
->delta_base_sha1
);
1152 * If we don't care about type or size, then we don't
1153 * need to look inside the object at all. Note that we
1154 * do not optimize out the stat call, even if the
1155 * caller doesn't care about the disk-size, since our
1156 * return value implicitly indicates whether the
1157 * object even exists.
1159 if (!oi
->typep
&& !oi
->typename
&& !oi
->sizep
&& !oi
->contentp
) {
1162 if (stat_sha1_file(sha1
, &st
, &path
) < 0)
1165 *oi
->disk_sizep
= st
.st_size
;
1169 map
= map_sha1_file(sha1
, &mapsize
);
1174 oi
->sizep
= &size_scratch
;
1177 *oi
->disk_sizep
= mapsize
;
1178 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
)) {
1179 if (unpack_sha1_header_to_strbuf(&stream
, map
, mapsize
, hdr
, sizeof(hdr
), &hdrbuf
) < 0)
1180 status
= error("unable to unpack %s header with --allow-unknown-type",
1182 } else if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1183 status
= error("unable to unpack %s header",
1187 else if (hdrbuf
.len
) {
1188 if ((status
= parse_sha1_header_extended(hdrbuf
.buf
, oi
, flags
)) < 0)
1189 status
= error("unable to parse %s header with --allow-unknown-type",
1191 } else if ((status
= parse_sha1_header_extended(hdr
, oi
, flags
)) < 0)
1192 status
= error("unable to parse %s header", sha1_to_hex(sha1
));
1194 if (status
>= 0 && oi
->contentp
) {
1195 *oi
->contentp
= unpack_sha1_rest(&stream
, hdr
,
1197 if (!*oi
->contentp
) {
1198 git_inflate_end(&stream
);
1202 git_inflate_end(&stream
);
1204 munmap(map
, mapsize
);
1205 if (status
&& oi
->typep
)
1206 *oi
->typep
= status
;
1207 if (oi
->sizep
== &size_scratch
)
1209 strbuf_release(&hdrbuf
);
1210 oi
->whence
= OI_LOOSE
;
1211 return (status
< 0) ? status
: 0;
1214 int sha1_object_info_extended(const unsigned char *sha1
, struct object_info
*oi
, unsigned flags
)
1216 static struct object_info blank_oi
= OBJECT_INFO_INIT
;
1217 struct pack_entry e
;
1219 const unsigned char *real
= (flags
& OBJECT_INFO_LOOKUP_REPLACE
) ?
1220 lookup_replace_object(sha1
) :
1223 if (is_null_sha1(real
))
1229 if (!(flags
& OBJECT_INFO_SKIP_CACHED
)) {
1230 struct cached_object
*co
= find_cached_object(real
);
1233 *(oi
->typep
) = co
->type
;
1235 *(oi
->sizep
) = co
->size
;
1237 *(oi
->disk_sizep
) = 0;
1238 if (oi
->delta_base_sha1
)
1239 hashclr(oi
->delta_base_sha1
);
1241 strbuf_addstr(oi
->typename
, typename(co
->type
));
1243 *oi
->contentp
= xmemdupz(co
->buf
, co
->size
);
1244 oi
->whence
= OI_CACHED
;
1249 if (!find_pack_entry(real
, &e
)) {
1250 /* Most likely it's a loose object. */
1251 if (!sha1_loose_object_info(real
, oi
, flags
))
1254 /* Not a loose object; someone else may have just packed it. */
1255 if (flags
& OBJECT_INFO_QUICK
) {
1258 reprepare_packed_git();
1259 if (!find_pack_entry(real
, &e
))
1264 if (oi
== &blank_oi
)
1266 * We know that the caller doesn't actually need the
1267 * information below, so return early.
1271 rtype
= packed_object_info(e
.p
, e
.offset
, oi
);
1273 mark_bad_packed_object(e
.p
, real
);
1274 return sha1_object_info_extended(real
, oi
, 0);
1275 } else if (oi
->whence
== OI_PACKED
) {
1276 oi
->u
.packed
.offset
= e
.offset
;
1277 oi
->u
.packed
.pack
= e
.p
;
1278 oi
->u
.packed
.is_delta
= (rtype
== OBJ_REF_DELTA
||
1279 rtype
== OBJ_OFS_DELTA
);
1285 /* returns enum object_type or negative */
1286 int sha1_object_info(const unsigned char *sha1
, unsigned long *sizep
)
1288 enum object_type type
;
1289 struct object_info oi
= OBJECT_INFO_INIT
;
1293 if (sha1_object_info_extended(sha1
, &oi
,
1294 OBJECT_INFO_LOOKUP_REPLACE
) < 0)
1299 static void *read_object(const unsigned char *sha1
, enum object_type
*type
,
1300 unsigned long *size
)
1302 struct object_info oi
= OBJECT_INFO_INIT
;
1306 oi
.contentp
= &content
;
1308 if (sha1_object_info_extended(sha1
, &oi
, 0) < 0)
1313 int pretend_sha1_file(void *buf
, unsigned long len
, enum object_type type
,
1314 unsigned char *sha1
)
1316 struct cached_object
*co
;
1318 hash_sha1_file(buf
, len
, typename(type
), sha1
);
1319 if (has_sha1_file(sha1
) || find_cached_object(sha1
))
1321 ALLOC_GROW(cached_objects
, cached_object_nr
+ 1, cached_object_alloc
);
1322 co
= &cached_objects
[cached_object_nr
++];
1325 co
->buf
= xmalloc(len
);
1326 memcpy(co
->buf
, buf
, len
);
1327 hashcpy(co
->sha1
, sha1
);
1332 * This function dies on corrupt objects; the callers who want to
1333 * deal with them should arrange to call read_object() and give error
1334 * messages themselves.
1336 void *read_sha1_file_extended(const unsigned char *sha1
,
1337 enum object_type
*type
,
1338 unsigned long *size
,
1342 const struct packed_git
*p
;
1345 const unsigned char *repl
= lookup_replace
? lookup_replace_object(sha1
)
1349 data
= read_object(repl
, type
, size
);
1353 if (errno
&& errno
!= ENOENT
)
1354 die_errno("failed to read object %s", sha1_to_hex(sha1
));
1356 /* die if we replaced an object with one that does not exist */
1358 die("replacement %s not found for %s",
1359 sha1_to_hex(repl
), sha1_to_hex(sha1
));
1361 if (!stat_sha1_file(repl
, &st
, &path
))
1362 die("loose object %s (stored in %s) is corrupt",
1363 sha1_to_hex(repl
), path
);
1365 if ((p
= has_packed_and_bad(repl
)) != NULL
)
1366 die("packed object %s (stored in %s) is corrupt",
1367 sha1_to_hex(repl
), p
->pack_name
);
1372 void *read_object_with_reference(const unsigned char *sha1
,
1373 const char *required_type_name
,
1374 unsigned long *size
,
1375 unsigned char *actual_sha1_return
)
1377 enum object_type type
, required_type
;
1379 unsigned long isize
;
1380 unsigned char actual_sha1
[20];
1382 required_type
= type_from_string(required_type_name
);
1383 hashcpy(actual_sha1
, sha1
);
1385 int ref_length
= -1;
1386 const char *ref_type
= NULL
;
1388 buffer
= read_sha1_file(actual_sha1
, &type
, &isize
);
1391 if (type
== required_type
) {
1393 if (actual_sha1_return
)
1394 hashcpy(actual_sha1_return
, actual_sha1
);
1397 /* Handle references */
1398 else if (type
== OBJ_COMMIT
)
1400 else if (type
== OBJ_TAG
)
1401 ref_type
= "object ";
1406 ref_length
= strlen(ref_type
);
1408 if (ref_length
+ 40 > isize
||
1409 memcmp(buffer
, ref_type
, ref_length
) ||
1410 get_sha1_hex((char *) buffer
+ ref_length
, actual_sha1
)) {
1415 /* Now we have the ID of the referred-to object in
1416 * actual_sha1. Check again. */
1420 static void write_sha1_file_prepare(const void *buf
, unsigned long len
,
1421 const char *type
, unsigned char *sha1
,
1422 char *hdr
, int *hdrlen
)
1426 /* Generate the header */
1427 *hdrlen
= xsnprintf(hdr
, *hdrlen
, "%s %lu", type
, len
)+1;
1430 the_hash_algo
->init_fn(&c
);
1431 the_hash_algo
->update_fn(&c
, hdr
, *hdrlen
);
1432 the_hash_algo
->update_fn(&c
, buf
, len
);
1433 the_hash_algo
->final_fn(sha1
, &c
);
1437 * Move the just written object into its final resting place.
1439 int finalize_object_file(const char *tmpfile
, const char *filename
)
1443 if (object_creation_mode
== OBJECT_CREATION_USES_RENAMES
)
1445 else if (link(tmpfile
, filename
))
1449 * Coda hack - coda doesn't like cross-directory links,
1450 * so we fall back to a rename, which will mean that it
1451 * won't be able to check collisions, but that's not a
1454 * The same holds for FAT formatted media.
1456 * When this succeeds, we just return. We have nothing
1459 if (ret
&& ret
!= EEXIST
) {
1461 if (!rename(tmpfile
, filename
))
1465 unlink_or_warn(tmpfile
);
1467 if (ret
!= EEXIST
) {
1468 return error_errno("unable to write sha1 filename %s", filename
);
1470 /* FIXME!!! Collision check here ? */
1474 if (adjust_shared_perm(filename
))
1475 return error("unable to set permission to '%s'", filename
);
1479 static int write_buffer(int fd
, const void *buf
, size_t len
)
1481 if (write_in_full(fd
, buf
, len
) < 0)
1482 return error_errno("file write error");
1486 int hash_sha1_file(const void *buf
, unsigned long len
, const char *type
,
1487 unsigned char *sha1
)
1490 int hdrlen
= sizeof(hdr
);
1491 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1495 /* Finalize a file on disk, and close it. */
1496 static void close_sha1_file(int fd
)
1498 if (fsync_object_files
)
1499 fsync_or_die(fd
, "sha1 file");
1501 die_errno("error when closing sha1 file");
1504 /* Size of directory component, including the ending '/' */
1505 static inline int directory_size(const char *filename
)
1507 const char *s
= strrchr(filename
, '/');
1510 return s
- filename
+ 1;
1514 * This creates a temporary file in the same directory as the final
1517 * We want to avoid cross-directory filename renames, because those
1518 * can have problems on various filesystems (FAT, NFS, Coda).
1520 static int create_tmpfile(struct strbuf
*tmp
, const char *filename
)
1522 int fd
, dirlen
= directory_size(filename
);
1525 strbuf_add(tmp
, filename
, dirlen
);
1526 strbuf_addstr(tmp
, "tmp_obj_XXXXXX");
1527 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1528 if (fd
< 0 && dirlen
&& errno
== ENOENT
) {
1530 * Make sure the directory exists; note that the contents
1531 * of the buffer are undefined after mkstemp returns an
1532 * error, so we have to rewrite the whole buffer from
1536 strbuf_add(tmp
, filename
, dirlen
- 1);
1537 if (mkdir(tmp
->buf
, 0777) && errno
!= EEXIST
)
1539 if (adjust_shared_perm(tmp
->buf
))
1543 strbuf_addstr(tmp
, "/tmp_obj_XXXXXX");
1544 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1549 static int write_loose_object(const unsigned char *sha1
, char *hdr
, int hdrlen
,
1550 const void *buf
, unsigned long len
, time_t mtime
)
1553 unsigned char compressed
[4096];
1556 struct object_id parano_oid
;
1557 static struct strbuf tmp_file
= STRBUF_INIT
;
1558 const char *filename
= sha1_file_name(sha1
);
1560 fd
= create_tmpfile(&tmp_file
, filename
);
1562 if (errno
== EACCES
)
1563 return error("insufficient permission for adding an object to repository database %s", get_object_directory());
1565 return error_errno("unable to create temporary file");
1569 git_deflate_init(&stream
, zlib_compression_level
);
1570 stream
.next_out
= compressed
;
1571 stream
.avail_out
= sizeof(compressed
);
1572 the_hash_algo
->init_fn(&c
);
1574 /* First header.. */
1575 stream
.next_in
= (unsigned char *)hdr
;
1576 stream
.avail_in
= hdrlen
;
1577 while (git_deflate(&stream
, 0) == Z_OK
)
1579 the_hash_algo
->update_fn(&c
, hdr
, hdrlen
);
1581 /* Then the data itself.. */
1582 stream
.next_in
= (void *)buf
;
1583 stream
.avail_in
= len
;
1585 unsigned char *in0
= stream
.next_in
;
1586 ret
= git_deflate(&stream
, Z_FINISH
);
1587 the_hash_algo
->update_fn(&c
, in0
, stream
.next_in
- in0
);
1588 if (write_buffer(fd
, compressed
, stream
.next_out
- compressed
) < 0)
1589 die("unable to write sha1 file");
1590 stream
.next_out
= compressed
;
1591 stream
.avail_out
= sizeof(compressed
);
1592 } while (ret
== Z_OK
);
1594 if (ret
!= Z_STREAM_END
)
1595 die("unable to deflate new object %s (%d)", sha1_to_hex(sha1
), ret
);
1596 ret
= git_deflate_end_gently(&stream
);
1598 die("deflateEnd on object %s failed (%d)", sha1_to_hex(sha1
), ret
);
1599 the_hash_algo
->final_fn(parano_oid
.hash
, &c
);
1600 if (hashcmp(sha1
, parano_oid
.hash
) != 0)
1601 die("confused by unstable object source data for %s", sha1_to_hex(sha1
));
1603 close_sha1_file(fd
);
1608 utb
.modtime
= mtime
;
1609 if (utime(tmp_file
.buf
, &utb
) < 0)
1610 warning_errno("failed utime() on %s", tmp_file
.buf
);
1613 return finalize_object_file(tmp_file
.buf
, filename
);
1616 static int freshen_loose_object(const unsigned char *sha1
)
1618 return check_and_freshen(sha1
, 1);
1621 static int freshen_packed_object(const unsigned char *sha1
)
1623 struct pack_entry e
;
1624 if (!find_pack_entry(sha1
, &e
))
1628 if (!freshen_file(e
.p
->pack_name
))
1634 int write_sha1_file(const void *buf
, unsigned long len
, const char *type
, unsigned char *sha1
)
1637 int hdrlen
= sizeof(hdr
);
1639 /* Normally if we have it in the pack then we do not bother writing
1640 * it out into .git/objects/??/?{38} file.
1642 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1643 if (freshen_packed_object(sha1
) || freshen_loose_object(sha1
))
1645 return write_loose_object(sha1
, hdr
, hdrlen
, buf
, len
, 0);
1648 int hash_sha1_file_literally(const void *buf
, unsigned long len
, const char *type
,
1649 struct object_id
*oid
, unsigned flags
)
1652 int hdrlen
, status
= 0;
1654 /* type string, SP, %lu of the length plus NUL must fit this */
1655 hdrlen
= strlen(type
) + 32;
1656 header
= xmalloc(hdrlen
);
1657 write_sha1_file_prepare(buf
, len
, type
, oid
->hash
, header
, &hdrlen
);
1659 if (!(flags
& HASH_WRITE_OBJECT
))
1661 if (freshen_packed_object(oid
->hash
) || freshen_loose_object(oid
->hash
))
1663 status
= write_loose_object(oid
->hash
, header
, hdrlen
, buf
, len
, 0);
1670 int force_object_loose(const unsigned char *sha1
, time_t mtime
)
1674 enum object_type type
;
1679 if (has_loose_object(sha1
))
1681 buf
= read_object(sha1
, &type
, &len
);
1683 return error("cannot read sha1_file for %s", sha1_to_hex(sha1
));
1684 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", typename(type
), len
) + 1;
1685 ret
= write_loose_object(sha1
, hdr
, hdrlen
, buf
, len
, mtime
);
1691 int has_sha1_file_with_flags(const unsigned char *sha1
, int flags
)
1693 if (!startup_info
->have_repository
)
1695 return sha1_object_info_extended(sha1
, NULL
,
1696 flags
| OBJECT_INFO_SKIP_CACHED
) >= 0;
1699 int has_object_file(const struct object_id
*oid
)
1701 return has_sha1_file(oid
->hash
);
1704 int has_object_file_with_flags(const struct object_id
*oid
, int flags
)
1706 return has_sha1_file_with_flags(oid
->hash
, flags
);
1709 static void check_tree(const void *buf
, size_t size
)
1711 struct tree_desc desc
;
1712 struct name_entry entry
;
1714 init_tree_desc(&desc
, buf
, size
);
1715 while (tree_entry(&desc
, &entry
))
1717 * tree_entry() will die() on malformed entries */
1721 static void check_commit(const void *buf
, size_t size
)
1724 memset(&c
, 0, sizeof(c
));
1725 if (parse_commit_buffer(&c
, buf
, size
))
1726 die("corrupt commit");
1729 static void check_tag(const void *buf
, size_t size
)
1732 memset(&t
, 0, sizeof(t
));
1733 if (parse_tag_buffer(&t
, buf
, size
))
1737 static int index_mem(struct object_id
*oid
, void *buf
, size_t size
,
1738 enum object_type type
,
1739 const char *path
, unsigned flags
)
1741 int ret
, re_allocated
= 0;
1742 int write_object
= flags
& HASH_WRITE_OBJECT
;
1748 * Convert blobs to git internal format
1750 if ((type
== OBJ_BLOB
) && path
) {
1751 struct strbuf nbuf
= STRBUF_INIT
;
1752 if (convert_to_git(&the_index
, path
, buf
, size
, &nbuf
,
1753 get_safe_crlf(flags
))) {
1754 buf
= strbuf_detach(&nbuf
, &size
);
1758 if (flags
& HASH_FORMAT_CHECK
) {
1759 if (type
== OBJ_TREE
)
1760 check_tree(buf
, size
);
1761 if (type
== OBJ_COMMIT
)
1762 check_commit(buf
, size
);
1763 if (type
== OBJ_TAG
)
1764 check_tag(buf
, size
);
1768 ret
= write_sha1_file(buf
, size
, typename(type
), oid
->hash
);
1770 ret
= hash_sha1_file(buf
, size
, typename(type
), oid
->hash
);
1776 static int index_stream_convert_blob(struct object_id
*oid
, int fd
,
1777 const char *path
, unsigned flags
)
1780 const int write_object
= flags
& HASH_WRITE_OBJECT
;
1781 struct strbuf sbuf
= STRBUF_INIT
;
1784 assert(would_convert_to_git_filter_fd(path
));
1786 convert_to_git_filter_fd(&the_index
, path
, fd
, &sbuf
,
1787 get_safe_crlf(flags
));
1790 ret
= write_sha1_file(sbuf
.buf
, sbuf
.len
, typename(OBJ_BLOB
),
1793 ret
= hash_sha1_file(sbuf
.buf
, sbuf
.len
, typename(OBJ_BLOB
),
1795 strbuf_release(&sbuf
);
1799 static int index_pipe(struct object_id
*oid
, int fd
, enum object_type type
,
1800 const char *path
, unsigned flags
)
1802 struct strbuf sbuf
= STRBUF_INIT
;
1805 if (strbuf_read(&sbuf
, fd
, 4096) >= 0)
1806 ret
= index_mem(oid
, sbuf
.buf
, sbuf
.len
, type
, path
, flags
);
1809 strbuf_release(&sbuf
);
1813 #define SMALL_FILE_SIZE (32*1024)
1815 static int index_core(struct object_id
*oid
, int fd
, size_t size
,
1816 enum object_type type
, const char *path
,
1822 ret
= index_mem(oid
, "", size
, type
, path
, flags
);
1823 } else if (size
<= SMALL_FILE_SIZE
) {
1824 char *buf
= xmalloc(size
);
1825 ssize_t read_result
= read_in_full(fd
, buf
, size
);
1826 if (read_result
< 0)
1827 ret
= error_errno("read error while indexing %s",
1828 path
? path
: "<unknown>");
1829 else if (read_result
!= size
)
1830 ret
= error("short read while indexing %s",
1831 path
? path
: "<unknown>");
1833 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1836 void *buf
= xmmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1837 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1844 * This creates one packfile per large blob unless bulk-checkin
1845 * machinery is "plugged".
1847 * This also bypasses the usual "convert-to-git" dance, and that is on
1848 * purpose. We could write a streaming version of the converting
1849 * functions and insert that before feeding the data to fast-import
1850 * (or equivalent in-core API described above). However, that is
1851 * somewhat complicated, as we do not know the size of the filter
1852 * result, which we need to know beforehand when writing a git object.
1853 * Since the primary motivation for trying to stream from the working
1854 * tree file and to avoid mmaping it in core is to deal with large
1855 * binary blobs, they generally do not want to get any conversion, and
1856 * callers should avoid this code path when filters are requested.
1858 static int index_stream(struct object_id
*oid
, int fd
, size_t size
,
1859 enum object_type type
, const char *path
,
1862 return index_bulk_checkin(oid
->hash
, fd
, size
, type
, path
, flags
);
1865 int index_fd(struct object_id
*oid
, int fd
, struct stat
*st
,
1866 enum object_type type
, const char *path
, unsigned flags
)
1871 * Call xsize_t() only when needed to avoid potentially unnecessary
1872 * die() for large files.
1874 if (type
== OBJ_BLOB
&& path
&& would_convert_to_git_filter_fd(path
))
1875 ret
= index_stream_convert_blob(oid
, fd
, path
, flags
);
1876 else if (!S_ISREG(st
->st_mode
))
1877 ret
= index_pipe(oid
, fd
, type
, path
, flags
);
1878 else if (st
->st_size
<= big_file_threshold
|| type
!= OBJ_BLOB
||
1879 (path
&& would_convert_to_git(&the_index
, path
)))
1880 ret
= index_core(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1883 ret
= index_stream(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1889 int index_path(struct object_id
*oid
, const char *path
, struct stat
*st
, unsigned flags
)
1892 struct strbuf sb
= STRBUF_INIT
;
1895 switch (st
->st_mode
& S_IFMT
) {
1897 fd
= open(path
, O_RDONLY
);
1899 return error_errno("open(\"%s\")", path
);
1900 if (index_fd(oid
, fd
, st
, OBJ_BLOB
, path
, flags
) < 0)
1901 return error("%s: failed to insert into database",
1905 if (strbuf_readlink(&sb
, path
, st
->st_size
))
1906 return error_errno("readlink(\"%s\")", path
);
1907 if (!(flags
& HASH_WRITE_OBJECT
))
1908 hash_sha1_file(sb
.buf
, sb
.len
, blob_type
, oid
->hash
);
1909 else if (write_sha1_file(sb
.buf
, sb
.len
, blob_type
, oid
->hash
))
1910 rc
= error("%s: failed to insert into database", path
);
1911 strbuf_release(&sb
);
1914 return resolve_gitlink_ref(path
, "HEAD", oid
);
1916 return error("%s: unsupported file type", path
);
1921 int read_pack_header(int fd
, struct pack_header
*header
)
1923 if (read_in_full(fd
, header
, sizeof(*header
)) != sizeof(*header
))
1924 /* "eof before pack header was fully read" */
1925 return PH_ERROR_EOF
;
1927 if (header
->hdr_signature
!= htonl(PACK_SIGNATURE
))
1928 /* "protocol error (pack signature mismatch detected)" */
1929 return PH_ERROR_PACK_SIGNATURE
;
1930 if (!pack_version_ok(header
->hdr_version
))
1931 /* "protocol error (pack version unsupported)" */
1932 return PH_ERROR_PROTOCOL
;
1936 void assert_sha1_type(const unsigned char *sha1
, enum object_type expect
)
1938 enum object_type type
= sha1_object_info(sha1
, NULL
);
1940 die("%s is not a valid object", sha1_to_hex(sha1
));
1942 die("%s is not a valid '%s' object", sha1_to_hex(sha1
),
1946 int for_each_file_in_obj_subdir(unsigned int subdir_nr
,
1947 struct strbuf
*path
,
1948 each_loose_object_fn obj_cb
,
1949 each_loose_cruft_fn cruft_cb
,
1950 each_loose_subdir_fn subdir_cb
,
1953 size_t origlen
, baselen
;
1957 struct object_id oid
;
1959 if (subdir_nr
> 0xff)
1960 BUG("invalid loose object subdirectory: %x", subdir_nr
);
1962 origlen
= path
->len
;
1963 strbuf_complete(path
, '/');
1964 strbuf_addf(path
, "%02x", subdir_nr
);
1966 dir
= opendir(path
->buf
);
1968 if (errno
!= ENOENT
)
1969 r
= error_errno("unable to open %s", path
->buf
);
1970 strbuf_setlen(path
, origlen
);
1974 oid
.hash
[0] = subdir_nr
;
1975 strbuf_addch(path
, '/');
1976 baselen
= path
->len
;
1978 while ((de
= readdir(dir
))) {
1980 if (is_dot_or_dotdot(de
->d_name
))
1983 namelen
= strlen(de
->d_name
);
1984 strbuf_setlen(path
, baselen
);
1985 strbuf_add(path
, de
->d_name
, namelen
);
1986 if (namelen
== GIT_SHA1_HEXSZ
- 2 &&
1987 !hex_to_bytes(oid
.hash
+ 1, de
->d_name
,
1988 GIT_SHA1_RAWSZ
- 1)) {
1990 r
= obj_cb(&oid
, path
->buf
, data
);
1998 r
= cruft_cb(de
->d_name
, path
->buf
, data
);
2005 strbuf_setlen(path
, baselen
- 1);
2006 if (!r
&& subdir_cb
)
2007 r
= subdir_cb(subdir_nr
, path
->buf
, data
);
2009 strbuf_setlen(path
, origlen
);
2014 int for_each_loose_file_in_objdir_buf(struct strbuf
*path
,
2015 each_loose_object_fn obj_cb
,
2016 each_loose_cruft_fn cruft_cb
,
2017 each_loose_subdir_fn subdir_cb
,
2023 for (i
= 0; i
< 256; i
++) {
2024 r
= for_each_file_in_obj_subdir(i
, path
, obj_cb
, cruft_cb
,
2033 int for_each_loose_file_in_objdir(const char *path
,
2034 each_loose_object_fn obj_cb
,
2035 each_loose_cruft_fn cruft_cb
,
2036 each_loose_subdir_fn subdir_cb
,
2039 struct strbuf buf
= STRBUF_INIT
;
2042 strbuf_addstr(&buf
, path
);
2043 r
= for_each_loose_file_in_objdir_buf(&buf
, obj_cb
, cruft_cb
,
2045 strbuf_release(&buf
);
2050 struct loose_alt_odb_data
{
2051 each_loose_object_fn
*cb
;
2055 static int loose_from_alt_odb(struct alternate_object_database
*alt
,
2058 struct loose_alt_odb_data
*data
= vdata
;
2059 struct strbuf buf
= STRBUF_INIT
;
2062 strbuf_addstr(&buf
, alt
->path
);
2063 r
= for_each_loose_file_in_objdir_buf(&buf
,
2064 data
->cb
, NULL
, NULL
,
2066 strbuf_release(&buf
);
2070 int for_each_loose_object(each_loose_object_fn cb
, void *data
, unsigned flags
)
2072 struct loose_alt_odb_data alt
;
2075 r
= for_each_loose_file_in_objdir(get_object_directory(),
2076 cb
, NULL
, NULL
, data
);
2080 if (flags
& FOR_EACH_OBJECT_LOCAL_ONLY
)
2085 return foreach_alt_odb(loose_from_alt_odb
, &alt
);
2088 static int check_stream_sha1(git_zstream
*stream
,
2092 const unsigned char *expected_sha1
)
2095 unsigned char real_sha1
[GIT_MAX_RAWSZ
];
2096 unsigned char buf
[4096];
2097 unsigned long total_read
;
2100 the_hash_algo
->init_fn(&c
);
2101 the_hash_algo
->update_fn(&c
, hdr
, stream
->total_out
);
2104 * We already read some bytes into hdr, but the ones up to the NUL
2105 * do not count against the object's content size.
2107 total_read
= stream
->total_out
- strlen(hdr
) - 1;
2110 * This size comparison must be "<=" to read the final zlib packets;
2111 * see the comment in unpack_sha1_rest for details.
2113 while (total_read
<= size
&&
2114 (status
== Z_OK
|| status
== Z_BUF_ERROR
)) {
2115 stream
->next_out
= buf
;
2116 stream
->avail_out
= sizeof(buf
);
2117 if (size
- total_read
< stream
->avail_out
)
2118 stream
->avail_out
= size
- total_read
;
2119 status
= git_inflate(stream
, Z_FINISH
);
2120 the_hash_algo
->update_fn(&c
, buf
, stream
->next_out
- buf
);
2121 total_read
+= stream
->next_out
- buf
;
2123 git_inflate_end(stream
);
2125 if (status
!= Z_STREAM_END
) {
2126 error("corrupt loose object '%s'", sha1_to_hex(expected_sha1
));
2129 if (stream
->avail_in
) {
2130 error("garbage at end of loose object '%s'",
2131 sha1_to_hex(expected_sha1
));
2135 the_hash_algo
->final_fn(real_sha1
, &c
);
2136 if (hashcmp(expected_sha1
, real_sha1
)) {
2137 error("sha1 mismatch for %s (expected %s)", path
,
2138 sha1_to_hex(expected_sha1
));
2145 int read_loose_object(const char *path
,
2146 const unsigned char *expected_sha1
,
2147 enum object_type
*type
,
2148 unsigned long *size
,
2153 unsigned long mapsize
;
2159 map
= map_sha1_file_1(path
, NULL
, &mapsize
);
2161 error_errno("unable to mmap %s", path
);
2165 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0) {
2166 error("unable to unpack header of %s", path
);
2170 *type
= parse_sha1_header(hdr
, size
);
2172 error("unable to parse header of %s", path
);
2173 git_inflate_end(&stream
);
2177 if (*type
== OBJ_BLOB
) {
2178 if (check_stream_sha1(&stream
, hdr
, *size
, path
, expected_sha1
) < 0)
2181 *contents
= unpack_sha1_rest(&stream
, hdr
, *size
, expected_sha1
);
2183 error("unable to unpack contents of %s", path
);
2184 git_inflate_end(&stream
);
2187 if (check_sha1_signature(expected_sha1
, *contents
,
2188 *size
, typename(*type
))) {
2189 error("sha1 mismatch for %s (expected %s)", path
,
2190 sha1_to_hex(expected_sha1
));
2196 ret
= 0; /* everything checks out */
2200 munmap(map
, mapsize
);