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(void *ctx
)
44 git_SHA1_Init((git_SHA_CTX
*)ctx
);
47 static void git_hash_sha1_update(void *ctx
, const void *data
, size_t len
)
49 git_SHA1_Update((git_SHA_CTX
*)ctx
, data
, len
);
52 static void git_hash_sha1_final(unsigned char *hash
, void *ctx
)
54 git_SHA1_Final(hash
, (git_SHA_CTX
*)ctx
);
57 static void git_hash_unknown_init(void *ctx
)
59 die("trying to init unknown hash");
62 static void git_hash_unknown_update(void *ctx
, const void *data
, size_t len
)
64 die("trying to update unknown hash");
67 static void git_hash_unknown_final(unsigned char *hash
, void *ctx
)
69 die("trying to finalize unknown hash");
72 const struct git_hash_algo hash_algos
[GIT_HASH_NALGOS
] = {
79 git_hash_unknown_init
,
80 git_hash_unknown_update
,
81 git_hash_unknown_final
,
87 /* "sha1", big-endian */
101 * This is meant to hold a *small* number of objects that you would
102 * want read_sha1_file() to be able to return, but yet you do not want
103 * to write them into the object store (e.g. a browse-only
106 static struct cached_object
{
107 unsigned char sha1
[20];
108 enum object_type type
;
112 static int cached_object_nr
, cached_object_alloc
;
114 static struct cached_object empty_tree
= {
115 EMPTY_TREE_SHA1_BIN_LITERAL
,
121 static struct cached_object
*find_cached_object(const unsigned char *sha1
)
124 struct cached_object
*co
= cached_objects
;
126 for (i
= 0; i
< cached_object_nr
; i
++, co
++) {
127 if (!hashcmp(co
->sha1
, sha1
))
130 if (!hashcmp(sha1
, empty_tree
.sha1
))
136 static enum safe_crlf
get_safe_crlf(unsigned flags
)
138 if (flags
& HASH_RENORMALIZE
)
139 return SAFE_CRLF_RENORMALIZE
;
140 else if (flags
& HASH_WRITE_OBJECT
)
143 return SAFE_CRLF_FALSE
;
147 int mkdir_in_gitdir(const char *path
)
149 if (mkdir(path
, 0777)) {
150 int saved_errno
= errno
;
152 struct strbuf sb
= STRBUF_INIT
;
157 * Are we looking at a path in a symlinked worktree
158 * whose original repository does not yet have it?
159 * e.g. .git/rr-cache pointing at its original
160 * repository in which the user hasn't performed any
161 * conflict resolution yet?
163 if (lstat(path
, &st
) || !S_ISLNK(st
.st_mode
) ||
164 strbuf_readlink(&sb
, path
, st
.st_size
) ||
165 !is_absolute_path(sb
.buf
) ||
166 mkdir(sb
.buf
, 0777)) {
173 return adjust_shared_perm(path
);
176 enum scld_error
safe_create_leading_directories(char *path
)
178 char *next_component
= path
+ offset_1st_component(path
);
179 enum scld_error ret
= SCLD_OK
;
181 while (ret
== SCLD_OK
&& next_component
) {
183 char *slash
= next_component
, slash_character
;
185 while (*slash
&& !is_dir_sep(*slash
))
191 next_component
= slash
+ 1;
192 while (is_dir_sep(*next_component
))
194 if (!*next_component
)
197 slash_character
= *slash
;
199 if (!stat(path
, &st
)) {
201 if (!S_ISDIR(st
.st_mode
)) {
205 } else if (mkdir(path
, 0777)) {
206 if (errno
== EEXIST
&&
207 !stat(path
, &st
) && S_ISDIR(st
.st_mode
))
208 ; /* somebody created it since we checked */
209 else if (errno
== ENOENT
)
211 * Either mkdir() failed because
212 * somebody just pruned the containing
213 * directory, or stat() failed because
214 * the file that was in our way was
215 * just removed. Either way, inform
216 * the caller that it might be worth
222 } else if (adjust_shared_perm(path
)) {
225 *slash
= slash_character
;
230 enum scld_error
safe_create_leading_directories_const(const char *path
)
233 /* path points to cache entries, so xstrdup before messing with it */
234 char *buf
= xstrdup(path
);
235 enum scld_error result
= safe_create_leading_directories(buf
);
243 int raceproof_create_file(const char *path
, create_file_fn fn
, void *cb
)
246 * The number of times we will try to remove empty directories
247 * in the way of path. This is only 1 because if another
248 * process is racily creating directories that conflict with
249 * us, we don't want to fight against them.
251 int remove_directories_remaining
= 1;
254 * The number of times that we will try to create the
255 * directories containing path. We are willing to attempt this
256 * more than once, because another process could be trying to
257 * clean up empty directories at the same time as we are
258 * trying to create them.
260 int create_directories_remaining
= 3;
262 /* A scratch copy of path, filled lazily if we need it: */
263 struct strbuf path_copy
= STRBUF_INIT
;
276 if (errno
== EISDIR
&& remove_directories_remaining
-- > 0) {
278 * A directory is in the way. Maybe it is empty; try
282 strbuf_addstr(&path_copy
, path
);
284 if (!remove_dir_recursively(&path_copy
, REMOVE_DIR_EMPTY_ONLY
))
286 } else if (errno
== ENOENT
&& create_directories_remaining
-- > 0) {
288 * Maybe the containing directory didn't exist, or
289 * maybe it was just deleted by a process that is
290 * racing with us to clean up empty directories. Try
293 enum scld_error scld_result
;
296 strbuf_addstr(&path_copy
, path
);
299 scld_result
= safe_create_leading_directories(path_copy
.buf
);
300 if (scld_result
== SCLD_OK
)
302 } while (scld_result
== SCLD_VANISHED
&& create_directories_remaining
-- > 0);
306 strbuf_release(&path_copy
);
311 static void fill_sha1_path(struct strbuf
*buf
, const unsigned char *sha1
)
314 for (i
= 0; i
< 20; i
++) {
315 static char hex
[] = "0123456789abcdef";
316 unsigned int val
= sha1
[i
];
317 strbuf_addch(buf
, hex
[val
>> 4]);
318 strbuf_addch(buf
, hex
[val
& 0xf]);
320 strbuf_addch(buf
, '/');
324 const char *sha1_file_name(const unsigned char *sha1
)
326 static struct strbuf buf
= STRBUF_INIT
;
329 strbuf_addf(&buf
, "%s/", get_object_directory());
331 fill_sha1_path(&buf
, sha1
);
335 struct strbuf
*alt_scratch_buf(struct alternate_object_database
*alt
)
337 strbuf_setlen(&alt
->scratch
, alt
->base_len
);
338 return &alt
->scratch
;
341 static const char *alt_sha1_path(struct alternate_object_database
*alt
,
342 const unsigned char *sha1
)
344 struct strbuf
*buf
= alt_scratch_buf(alt
);
345 fill_sha1_path(buf
, sha1
);
349 struct alternate_object_database
*alt_odb_list
;
350 static struct alternate_object_database
**alt_odb_tail
;
353 * Return non-zero iff the path is usable as an alternate object database.
355 static int alt_odb_usable(struct strbuf
*path
, const char *normalized_objdir
)
357 struct alternate_object_database
*alt
;
359 /* Detect cases where alternate disappeared */
360 if (!is_directory(path
->buf
)) {
361 error("object directory %s does not exist; "
362 "check .git/objects/info/alternates.",
368 * Prevent the common mistake of listing the same
369 * thing twice, or object directory itself.
371 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
372 if (!fspathcmp(path
->buf
, alt
->path
))
375 if (!fspathcmp(path
->buf
, normalized_objdir
))
382 * Prepare alternate object database registry.
384 * The variable alt_odb_list points at the list of struct
385 * alternate_object_database. The elements on this list come from
386 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
387 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
388 * whose contents is similar to that environment variable but can be
389 * LF separated. Its base points at a statically allocated buffer that
390 * contains "/the/directory/corresponding/to/.git/objects/...", while
391 * its name points just after the slash at the end of ".git/objects/"
392 * in the example above, and has enough space to hold 40-byte hex
393 * SHA1, an extra slash for the first level indirection, and the
396 static void read_info_alternates(const char * relative_base
, int depth
);
397 static int link_alt_odb_entry(const char *entry
, const char *relative_base
,
398 int depth
, const char *normalized_objdir
)
400 struct alternate_object_database
*ent
;
401 struct strbuf pathbuf
= STRBUF_INIT
;
403 if (!is_absolute_path(entry
) && relative_base
) {
404 strbuf_realpath(&pathbuf
, relative_base
, 1);
405 strbuf_addch(&pathbuf
, '/');
407 strbuf_addstr(&pathbuf
, entry
);
409 if (strbuf_normalize_path(&pathbuf
) < 0 && relative_base
) {
410 error("unable to normalize alternate object path: %s",
412 strbuf_release(&pathbuf
);
417 * The trailing slash after the directory name is given by
418 * this function at the end. Remove duplicates.
420 while (pathbuf
.len
&& pathbuf
.buf
[pathbuf
.len
- 1] == '/')
421 strbuf_setlen(&pathbuf
, pathbuf
.len
- 1);
423 if (!alt_odb_usable(&pathbuf
, normalized_objdir
)) {
424 strbuf_release(&pathbuf
);
428 ent
= alloc_alt_odb(pathbuf
.buf
);
430 /* add the alternate entry */
432 alt_odb_tail
= &(ent
->next
);
435 /* recursively add alternates */
436 read_info_alternates(pathbuf
.buf
, depth
+ 1);
438 strbuf_release(&pathbuf
);
442 static const char *parse_alt_odb_entry(const char *string
,
450 if (*string
== '#') {
451 /* comment; consume up to next separator */
452 end
= strchrnul(string
, sep
);
453 } else if (*string
== '"' && !unquote_c_style(out
, string
, &end
)) {
455 * quoted path; unquote_c_style has copied the
456 * data for us and set "end". Broken quoting (e.g.,
457 * an entry that doesn't end with a quote) falls
458 * back to the unquoted case below.
461 /* normal, unquoted path */
462 end
= strchrnul(string
, sep
);
463 strbuf_add(out
, string
, end
- string
);
471 static void link_alt_odb_entries(const char *alt
, int sep
,
472 const char *relative_base
, int depth
)
474 struct strbuf objdirbuf
= STRBUF_INIT
;
475 struct strbuf entry
= STRBUF_INIT
;
481 error("%s: ignoring alternate object stores, nesting too deep.",
486 strbuf_add_absolute_path(&objdirbuf
, get_object_directory());
487 if (strbuf_normalize_path(&objdirbuf
) < 0)
488 die("unable to normalize object directory: %s",
492 alt
= parse_alt_odb_entry(alt
, sep
, &entry
);
495 link_alt_odb_entry(entry
.buf
, relative_base
, depth
, objdirbuf
.buf
);
497 strbuf_release(&entry
);
498 strbuf_release(&objdirbuf
);
501 static void read_info_alternates(const char * relative_base
, int depth
)
504 struct strbuf buf
= STRBUF_INIT
;
506 path
= xstrfmt("%s/info/alternates", relative_base
);
507 if (strbuf_read_file(&buf
, path
, 1024) < 0) {
508 warn_on_fopen_errors(path
);
513 link_alt_odb_entries(buf
.buf
, '\n', relative_base
, depth
);
514 strbuf_release(&buf
);
518 struct alternate_object_database
*alloc_alt_odb(const char *dir
)
520 struct alternate_object_database
*ent
;
522 FLEX_ALLOC_STR(ent
, path
, dir
);
523 strbuf_init(&ent
->scratch
, 0);
524 strbuf_addf(&ent
->scratch
, "%s/", dir
);
525 ent
->base_len
= ent
->scratch
.len
;
530 void add_to_alternates_file(const char *reference
)
532 struct lock_file lock
= LOCK_INIT
;
533 char *alts
= git_pathdup("objects/info/alternates");
537 hold_lock_file_for_update(&lock
, alts
, LOCK_DIE_ON_ERROR
);
538 out
= fdopen_lock_file(&lock
, "w");
540 die_errno("unable to fdopen alternates lockfile");
542 in
= fopen(alts
, "r");
544 struct strbuf line
= STRBUF_INIT
;
546 while (strbuf_getline(&line
, in
) != EOF
) {
547 if (!strcmp(reference
, line
.buf
)) {
551 fprintf_or_die(out
, "%s\n", line
.buf
);
554 strbuf_release(&line
);
557 else if (errno
!= ENOENT
)
558 die_errno("unable to read alternates file");
561 rollback_lock_file(&lock
);
563 fprintf_or_die(out
, "%s\n", reference
);
564 if (commit_lock_file(&lock
))
565 die_errno("unable to move new alternates file into place");
567 link_alt_odb_entries(reference
, '\n', NULL
, 0);
572 void add_to_alternates_memory(const char *reference
)
575 * Make sure alternates are initialized, or else our entry may be
576 * overwritten when they are.
580 link_alt_odb_entries(reference
, '\n', NULL
, 0);
584 * Compute the exact path an alternate is at and returns it. In case of
585 * error NULL is returned and the human readable error is added to `err`
586 * `path` may be relative and should point to $GITDIR.
587 * `err` must not be null.
589 char *compute_alternate_path(const char *path
, struct strbuf
*err
)
591 char *ref_git
= NULL
;
592 const char *repo
, *ref_git_s
;
595 ref_git_s
= real_path_if_valid(path
);
598 strbuf_addf(err
, _("path '%s' does not exist"), path
);
602 * Beware: read_gitfile(), real_path() and mkpath()
603 * return static buffer
605 ref_git
= xstrdup(ref_git_s
);
607 repo
= read_gitfile(ref_git
);
609 repo
= read_gitfile(mkpath("%s/.git", ref_git
));
612 ref_git
= xstrdup(repo
);
615 if (!repo
&& is_directory(mkpath("%s/.git/objects", ref_git
))) {
616 char *ref_git_git
= mkpathdup("%s/.git", ref_git
);
618 ref_git
= ref_git_git
;
619 } else if (!is_directory(mkpath("%s/objects", ref_git
))) {
620 struct strbuf sb
= STRBUF_INIT
;
622 if (get_common_dir(&sb
, ref_git
)) {
624 _("reference repository '%s' as a linked "
625 "checkout is not supported yet."),
630 strbuf_addf(err
, _("reference repository '%s' is not a "
631 "local repository."), path
);
635 if (!access(mkpath("%s/shallow", ref_git
), F_OK
)) {
636 strbuf_addf(err
, _("reference repository '%s' is shallow"),
642 if (!access(mkpath("%s/info/grafts", ref_git
), F_OK
)) {
644 _("reference repository '%s' is grafted"),
652 FREE_AND_NULL(ref_git
);
658 int foreach_alt_odb(alt_odb_fn fn
, void *cb
)
660 struct alternate_object_database
*ent
;
664 for (ent
= alt_odb_list
; ent
; ent
= ent
->next
) {
672 void prepare_alt_odb(void)
679 alt
= getenv(ALTERNATE_DB_ENVIRONMENT
);
681 alt_odb_tail
= &alt_odb_list
;
682 link_alt_odb_entries(alt
, PATH_SEP
, NULL
, 0);
684 read_info_alternates(get_object_directory(), 0);
687 /* Returns 1 if we have successfully freshened the file, 0 otherwise. */
688 static int freshen_file(const char *fn
)
691 t
.actime
= t
.modtime
= time(NULL
);
692 return !utime(fn
, &t
);
696 * All of the check_and_freshen functions return 1 if the file exists and was
697 * freshened (if freshening was requested), 0 otherwise. If they return
698 * 0, you should not assume that it is safe to skip a write of the object (it
699 * either does not exist on disk, or has a stale mtime and may be subject to
702 int check_and_freshen_file(const char *fn
, int freshen
)
704 if (access(fn
, F_OK
))
706 if (freshen
&& !freshen_file(fn
))
711 static int check_and_freshen_local(const unsigned char *sha1
, int freshen
)
713 return check_and_freshen_file(sha1_file_name(sha1
), freshen
);
716 static int check_and_freshen_nonlocal(const unsigned char *sha1
, int freshen
)
718 struct alternate_object_database
*alt
;
720 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
721 const char *path
= alt_sha1_path(alt
, sha1
);
722 if (check_and_freshen_file(path
, freshen
))
728 static int check_and_freshen(const unsigned char *sha1
, int freshen
)
730 return check_and_freshen_local(sha1
, freshen
) ||
731 check_and_freshen_nonlocal(sha1
, freshen
);
734 int has_loose_object_nonlocal(const unsigned char *sha1
)
736 return check_and_freshen_nonlocal(sha1
, 0);
739 static int has_loose_object(const unsigned char *sha1
)
741 return check_and_freshen(sha1
, 0);
744 static void mmap_limit_check(size_t length
)
746 static size_t limit
= 0;
748 limit
= git_env_ulong("GIT_MMAP_LIMIT", 0);
753 die("attempting to mmap %"PRIuMAX
" over limit %"PRIuMAX
,
754 (uintmax_t)length
, (uintmax_t)limit
);
757 void *xmmap_gently(void *start
, size_t length
,
758 int prot
, int flags
, int fd
, off_t offset
)
762 mmap_limit_check(length
);
763 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
764 if (ret
== MAP_FAILED
) {
767 release_pack_memory(length
);
768 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
773 void *xmmap(void *start
, size_t length
,
774 int prot
, int flags
, int fd
, off_t offset
)
776 void *ret
= xmmap_gently(start
, length
, prot
, flags
, fd
, offset
);
777 if (ret
== MAP_FAILED
)
778 die_errno("mmap failed");
783 * With an in-core object data in "map", rehash it to make sure the
784 * object name actually matches "sha1" to detect object corruption.
785 * With "map" == NULL, try reading the object named with "sha1" using
786 * the streaming interface and rehash it to do the same.
788 int check_sha1_signature(const unsigned char *sha1
, void *map
,
789 unsigned long size
, const char *type
)
791 unsigned char real_sha1
[20];
792 enum object_type obj_type
;
793 struct git_istream
*st
;
799 hash_sha1_file(map
, size
, type
, real_sha1
);
800 return hashcmp(sha1
, real_sha1
) ? -1 : 0;
803 st
= open_istream(sha1
, &obj_type
, &size
, NULL
);
807 /* Generate the header */
808 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", typename(obj_type
), size
) + 1;
812 git_SHA1_Update(&c
, hdr
, hdrlen
);
815 ssize_t readlen
= read_istream(st
, buf
, sizeof(buf
));
823 git_SHA1_Update(&c
, buf
, readlen
);
825 git_SHA1_Final(real_sha1
, &c
);
827 return hashcmp(sha1
, real_sha1
) ? -1 : 0;
830 int git_open_cloexec(const char *name
, int flags
)
833 static int o_cloexec
= O_CLOEXEC
;
835 fd
= open(name
, flags
| o_cloexec
);
836 if ((o_cloexec
& O_CLOEXEC
) && fd
< 0 && errno
== EINVAL
) {
837 /* Try again w/o O_CLOEXEC: the kernel might not support it */
838 o_cloexec
&= ~O_CLOEXEC
;
839 fd
= open(name
, flags
| o_cloexec
);
842 #if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
844 static int fd_cloexec
= FD_CLOEXEC
;
846 if (!o_cloexec
&& 0 <= fd
&& fd_cloexec
) {
847 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
848 int flags
= fcntl(fd
, F_GETFD
);
849 if (fcntl(fd
, F_SETFD
, flags
| fd_cloexec
))
858 * Find "sha1" as a loose object in the local repository or in an alternate.
859 * Returns 0 on success, negative on failure.
861 * The "path" out-parameter will give the path of the object we found (if any).
862 * Note that it may point to static storage and is only valid until another
863 * call to sha1_file_name(), etc.
865 static int stat_sha1_file(const unsigned char *sha1
, struct stat
*st
,
868 struct alternate_object_database
*alt
;
870 *path
= sha1_file_name(sha1
);
871 if (!lstat(*path
, st
))
876 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
877 *path
= alt_sha1_path(alt
, sha1
);
878 if (!lstat(*path
, st
))
886 * Like stat_sha1_file(), but actually open the object and return the
887 * descriptor. See the caveats on the "path" parameter above.
889 static int open_sha1_file(const unsigned char *sha1
, const char **path
)
892 struct alternate_object_database
*alt
;
893 int most_interesting_errno
;
895 *path
= sha1_file_name(sha1
);
896 fd
= git_open(*path
);
899 most_interesting_errno
= errno
;
902 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
903 *path
= alt_sha1_path(alt
, sha1
);
904 fd
= git_open(*path
);
907 if (most_interesting_errno
== ENOENT
)
908 most_interesting_errno
= errno
;
910 errno
= most_interesting_errno
;
915 * Map the loose object at "path" if it is not NULL, or the path found by
916 * searching for a loose object named "sha1".
918 static void *map_sha1_file_1(const char *path
,
919 const unsigned char *sha1
,
928 fd
= open_sha1_file(sha1
, &path
);
933 if (!fstat(fd
, &st
)) {
934 *size
= xsize_t(st
.st_size
);
936 /* mmap() is forbidden on empty files */
937 error("object file %s is empty", path
);
940 map
= xmmap(NULL
, *size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
947 void *map_sha1_file(const unsigned char *sha1
, unsigned long *size
)
949 return map_sha1_file_1(NULL
, sha1
, size
);
952 static int unpack_sha1_short_header(git_zstream
*stream
,
953 unsigned char *map
, unsigned long mapsize
,
954 void *buffer
, unsigned long bufsiz
)
956 /* Get the data stream */
957 memset(stream
, 0, sizeof(*stream
));
958 stream
->next_in
= map
;
959 stream
->avail_in
= mapsize
;
960 stream
->next_out
= buffer
;
961 stream
->avail_out
= bufsiz
;
963 git_inflate_init(stream
);
964 return git_inflate(stream
, 0);
967 int unpack_sha1_header(git_zstream
*stream
,
968 unsigned char *map
, unsigned long mapsize
,
969 void *buffer
, unsigned long bufsiz
)
971 int status
= unpack_sha1_short_header(stream
, map
, mapsize
,
977 /* Make sure we have the terminating NUL */
978 if (!memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
983 static int unpack_sha1_header_to_strbuf(git_zstream
*stream
, unsigned char *map
,
984 unsigned long mapsize
, void *buffer
,
985 unsigned long bufsiz
, struct strbuf
*header
)
989 status
= unpack_sha1_short_header(stream
, map
, mapsize
, buffer
, bufsiz
);
994 * Check if entire header is unpacked in the first iteration.
996 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
1000 * buffer[0..bufsiz] was not large enough. Copy the partial
1001 * result out to header, and then append the result of further
1002 * reading the stream.
1004 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
1005 stream
->next_out
= buffer
;
1006 stream
->avail_out
= bufsiz
;
1009 status
= git_inflate(stream
, 0);
1010 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
1011 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
1013 stream
->next_out
= buffer
;
1014 stream
->avail_out
= bufsiz
;
1015 } while (status
!= Z_STREAM_END
);
1019 static void *unpack_sha1_rest(git_zstream
*stream
, void *buffer
, unsigned long size
, const unsigned char *sha1
)
1021 int bytes
= strlen(buffer
) + 1;
1022 unsigned char *buf
= xmallocz(size
);
1026 n
= stream
->total_out
- bytes
;
1029 memcpy(buf
, (char *) buffer
+ bytes
, n
);
1031 if (bytes
<= size
) {
1033 * The above condition must be (bytes <= size), not
1034 * (bytes < size). In other words, even though we
1035 * expect no more output and set avail_out to zero,
1036 * the input zlib stream may have bytes that express
1037 * "this concludes the stream", and we *do* want to
1040 * Otherwise we would not be able to test that we
1041 * consumed all the input to reach the expected size;
1042 * we also want to check that zlib tells us that all
1043 * went well with status == Z_STREAM_END at the end.
1045 stream
->next_out
= buf
+ bytes
;
1046 stream
->avail_out
= size
- bytes
;
1047 while (status
== Z_OK
)
1048 status
= git_inflate(stream
, Z_FINISH
);
1050 if (status
== Z_STREAM_END
&& !stream
->avail_in
) {
1051 git_inflate_end(stream
);
1056 error("corrupt loose object '%s'", sha1_to_hex(sha1
));
1057 else if (stream
->avail_in
)
1058 error("garbage at end of loose object '%s'",
1065 * We used to just use "sscanf()", but that's actually way
1066 * too permissive for what we want to check. So do an anal
1067 * object header parse by hand.
1069 static int parse_sha1_header_extended(const char *hdr
, struct object_info
*oi
,
1072 const char *type_buf
= hdr
;
1074 int type
, type_len
= 0;
1077 * The type can be of any size but is followed by
1089 type
= type_from_string_gently(type_buf
, type_len
, 1);
1091 strbuf_add(oi
->typename
, type_buf
, type_len
);
1093 * Set type to 0 if its an unknown object and
1094 * we're obtaining the type using '--allow-unknown-type'
1097 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
) && (type
< 0))
1100 die("invalid object type");
1105 * The length must follow immediately, and be in canonical
1106 * decimal format (ie "010" is not valid).
1108 size
= *hdr
++ - '0';
1113 unsigned long c
= *hdr
- '0';
1117 size
= size
* 10 + c
;
1125 * The length must be followed by a zero byte
1127 return *hdr
? -1 : type
;
1130 int parse_sha1_header(const char *hdr
, unsigned long *sizep
)
1132 struct object_info oi
= OBJECT_INFO_INIT
;
1135 return parse_sha1_header_extended(hdr
, &oi
, 0);
1138 static int sha1_loose_object_info(const unsigned char *sha1
,
1139 struct object_info
*oi
,
1143 unsigned long mapsize
;
1147 struct strbuf hdrbuf
= STRBUF_INIT
;
1148 unsigned long size_scratch
;
1150 if (oi
->delta_base_sha1
)
1151 hashclr(oi
->delta_base_sha1
);
1154 * If we don't care about type or size, then we don't
1155 * need to look inside the object at all. Note that we
1156 * do not optimize out the stat call, even if the
1157 * caller doesn't care about the disk-size, since our
1158 * return value implicitly indicates whether the
1159 * object even exists.
1161 if (!oi
->typep
&& !oi
->typename
&& !oi
->sizep
&& !oi
->contentp
) {
1164 if (stat_sha1_file(sha1
, &st
, &path
) < 0)
1167 *oi
->disk_sizep
= st
.st_size
;
1171 map
= map_sha1_file(sha1
, &mapsize
);
1176 oi
->sizep
= &size_scratch
;
1179 *oi
->disk_sizep
= mapsize
;
1180 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
)) {
1181 if (unpack_sha1_header_to_strbuf(&stream
, map
, mapsize
, hdr
, sizeof(hdr
), &hdrbuf
) < 0)
1182 status
= error("unable to unpack %s header with --allow-unknown-type",
1184 } else if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1185 status
= error("unable to unpack %s header",
1189 else if (hdrbuf
.len
) {
1190 if ((status
= parse_sha1_header_extended(hdrbuf
.buf
, oi
, flags
)) < 0)
1191 status
= error("unable to parse %s header with --allow-unknown-type",
1193 } else if ((status
= parse_sha1_header_extended(hdr
, oi
, flags
)) < 0)
1194 status
= error("unable to parse %s header", sha1_to_hex(sha1
));
1196 if (status
>= 0 && oi
->contentp
) {
1197 *oi
->contentp
= unpack_sha1_rest(&stream
, hdr
,
1199 if (!*oi
->contentp
) {
1200 git_inflate_end(&stream
);
1204 git_inflate_end(&stream
);
1206 munmap(map
, mapsize
);
1207 if (status
&& oi
->typep
)
1208 *oi
->typep
= status
;
1209 if (oi
->sizep
== &size_scratch
)
1211 strbuf_release(&hdrbuf
);
1212 oi
->whence
= OI_LOOSE
;
1213 return (status
< 0) ? status
: 0;
1216 int sha1_object_info_extended(const unsigned char *sha1
, struct object_info
*oi
, unsigned flags
)
1218 static struct object_info blank_oi
= OBJECT_INFO_INIT
;
1219 struct pack_entry e
;
1221 const unsigned char *real
= (flags
& OBJECT_INFO_LOOKUP_REPLACE
) ?
1222 lookup_replace_object(sha1
) :
1225 if (is_null_sha1(real
))
1231 if (!(flags
& OBJECT_INFO_SKIP_CACHED
)) {
1232 struct cached_object
*co
= find_cached_object(real
);
1235 *(oi
->typep
) = co
->type
;
1237 *(oi
->sizep
) = co
->size
;
1239 *(oi
->disk_sizep
) = 0;
1240 if (oi
->delta_base_sha1
)
1241 hashclr(oi
->delta_base_sha1
);
1243 strbuf_addstr(oi
->typename
, typename(co
->type
));
1245 *oi
->contentp
= xmemdupz(co
->buf
, co
->size
);
1246 oi
->whence
= OI_CACHED
;
1251 if (!find_pack_entry(real
, &e
)) {
1252 /* Most likely it's a loose object. */
1253 if (!sha1_loose_object_info(real
, oi
, flags
))
1256 /* Not a loose object; someone else may have just packed it. */
1257 if (flags
& OBJECT_INFO_QUICK
) {
1260 reprepare_packed_git();
1261 if (!find_pack_entry(real
, &e
))
1266 if (oi
== &blank_oi
)
1268 * We know that the caller doesn't actually need the
1269 * information below, so return early.
1273 rtype
= packed_object_info(e
.p
, e
.offset
, oi
);
1275 mark_bad_packed_object(e
.p
, real
);
1276 return sha1_object_info_extended(real
, oi
, 0);
1277 } else if (oi
->whence
== OI_PACKED
) {
1278 oi
->u
.packed
.offset
= e
.offset
;
1279 oi
->u
.packed
.pack
= e
.p
;
1280 oi
->u
.packed
.is_delta
= (rtype
== OBJ_REF_DELTA
||
1281 rtype
== OBJ_OFS_DELTA
);
1287 /* returns enum object_type or negative */
1288 int sha1_object_info(const unsigned char *sha1
, unsigned long *sizep
)
1290 enum object_type type
;
1291 struct object_info oi
= OBJECT_INFO_INIT
;
1295 if (sha1_object_info_extended(sha1
, &oi
,
1296 OBJECT_INFO_LOOKUP_REPLACE
) < 0)
1301 static void *read_object(const unsigned char *sha1
, enum object_type
*type
,
1302 unsigned long *size
)
1304 struct object_info oi
= OBJECT_INFO_INIT
;
1308 oi
.contentp
= &content
;
1310 if (sha1_object_info_extended(sha1
, &oi
, 0) < 0)
1315 int pretend_sha1_file(void *buf
, unsigned long len
, enum object_type type
,
1316 unsigned char *sha1
)
1318 struct cached_object
*co
;
1320 hash_sha1_file(buf
, len
, typename(type
), sha1
);
1321 if (has_sha1_file(sha1
) || find_cached_object(sha1
))
1323 ALLOC_GROW(cached_objects
, cached_object_nr
+ 1, cached_object_alloc
);
1324 co
= &cached_objects
[cached_object_nr
++];
1327 co
->buf
= xmalloc(len
);
1328 memcpy(co
->buf
, buf
, len
);
1329 hashcpy(co
->sha1
, sha1
);
1334 * This function dies on corrupt objects; the callers who want to
1335 * deal with them should arrange to call read_object() and give error
1336 * messages themselves.
1338 void *read_sha1_file_extended(const unsigned char *sha1
,
1339 enum object_type
*type
,
1340 unsigned long *size
,
1344 const struct packed_git
*p
;
1347 const unsigned char *repl
= lookup_replace
? lookup_replace_object(sha1
)
1351 data
= read_object(repl
, type
, size
);
1355 if (errno
&& errno
!= ENOENT
)
1356 die_errno("failed to read object %s", sha1_to_hex(sha1
));
1358 /* die if we replaced an object with one that does not exist */
1360 die("replacement %s not found for %s",
1361 sha1_to_hex(repl
), sha1_to_hex(sha1
));
1363 if (!stat_sha1_file(repl
, &st
, &path
))
1364 die("loose object %s (stored in %s) is corrupt",
1365 sha1_to_hex(repl
), path
);
1367 if ((p
= has_packed_and_bad(repl
)) != NULL
)
1368 die("packed object %s (stored in %s) is corrupt",
1369 sha1_to_hex(repl
), p
->pack_name
);
1374 void *read_object_with_reference(const unsigned char *sha1
,
1375 const char *required_type_name
,
1376 unsigned long *size
,
1377 unsigned char *actual_sha1_return
)
1379 enum object_type type
, required_type
;
1381 unsigned long isize
;
1382 unsigned char actual_sha1
[20];
1384 required_type
= type_from_string(required_type_name
);
1385 hashcpy(actual_sha1
, sha1
);
1387 int ref_length
= -1;
1388 const char *ref_type
= NULL
;
1390 buffer
= read_sha1_file(actual_sha1
, &type
, &isize
);
1393 if (type
== required_type
) {
1395 if (actual_sha1_return
)
1396 hashcpy(actual_sha1_return
, actual_sha1
);
1399 /* Handle references */
1400 else if (type
== OBJ_COMMIT
)
1402 else if (type
== OBJ_TAG
)
1403 ref_type
= "object ";
1408 ref_length
= strlen(ref_type
);
1410 if (ref_length
+ 40 > isize
||
1411 memcmp(buffer
, ref_type
, ref_length
) ||
1412 get_sha1_hex((char *) buffer
+ ref_length
, actual_sha1
)) {
1417 /* Now we have the ID of the referred-to object in
1418 * actual_sha1. Check again. */
1422 static void write_sha1_file_prepare(const void *buf
, unsigned long len
,
1423 const char *type
, unsigned char *sha1
,
1424 char *hdr
, int *hdrlen
)
1428 /* Generate the header */
1429 *hdrlen
= xsnprintf(hdr
, *hdrlen
, "%s %lu", type
, len
)+1;
1433 git_SHA1_Update(&c
, hdr
, *hdrlen
);
1434 git_SHA1_Update(&c
, buf
, len
);
1435 git_SHA1_Final(sha1
, &c
);
1439 * Move the just written object into its final resting place.
1441 int finalize_object_file(const char *tmpfile
, const char *filename
)
1445 if (object_creation_mode
== OBJECT_CREATION_USES_RENAMES
)
1447 else if (link(tmpfile
, filename
))
1451 * Coda hack - coda doesn't like cross-directory links,
1452 * so we fall back to a rename, which will mean that it
1453 * won't be able to check collisions, but that's not a
1456 * The same holds for FAT formatted media.
1458 * When this succeeds, we just return. We have nothing
1461 if (ret
&& ret
!= EEXIST
) {
1463 if (!rename(tmpfile
, filename
))
1467 unlink_or_warn(tmpfile
);
1469 if (ret
!= EEXIST
) {
1470 return error_errno("unable to write sha1 filename %s", filename
);
1472 /* FIXME!!! Collision check here ? */
1476 if (adjust_shared_perm(filename
))
1477 return error("unable to set permission to '%s'", filename
);
1481 static int write_buffer(int fd
, const void *buf
, size_t len
)
1483 if (write_in_full(fd
, buf
, len
) < 0)
1484 return error_errno("file write error");
1488 int hash_sha1_file(const void *buf
, unsigned long len
, const char *type
,
1489 unsigned char *sha1
)
1492 int hdrlen
= sizeof(hdr
);
1493 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1497 /* Finalize a file on disk, and close it. */
1498 static void close_sha1_file(int fd
)
1500 if (fsync_object_files
)
1501 fsync_or_die(fd
, "sha1 file");
1503 die_errno("error when closing sha1 file");
1506 /* Size of directory component, including the ending '/' */
1507 static inline int directory_size(const char *filename
)
1509 const char *s
= strrchr(filename
, '/');
1512 return s
- filename
+ 1;
1516 * This creates a temporary file in the same directory as the final
1519 * We want to avoid cross-directory filename renames, because those
1520 * can have problems on various filesystems (FAT, NFS, Coda).
1522 static int create_tmpfile(struct strbuf
*tmp
, const char *filename
)
1524 int fd
, dirlen
= directory_size(filename
);
1527 strbuf_add(tmp
, filename
, dirlen
);
1528 strbuf_addstr(tmp
, "tmp_obj_XXXXXX");
1529 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1530 if (fd
< 0 && dirlen
&& errno
== ENOENT
) {
1532 * Make sure the directory exists; note that the contents
1533 * of the buffer are undefined after mkstemp returns an
1534 * error, so we have to rewrite the whole buffer from
1538 strbuf_add(tmp
, filename
, dirlen
- 1);
1539 if (mkdir(tmp
->buf
, 0777) && errno
!= EEXIST
)
1541 if (adjust_shared_perm(tmp
->buf
))
1545 strbuf_addstr(tmp
, "/tmp_obj_XXXXXX");
1546 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1551 static int write_loose_object(const unsigned char *sha1
, char *hdr
, int hdrlen
,
1552 const void *buf
, unsigned long len
, time_t mtime
)
1555 unsigned char compressed
[4096];
1558 unsigned char parano_sha1
[20];
1559 static struct strbuf tmp_file
= STRBUF_INIT
;
1560 const char *filename
= sha1_file_name(sha1
);
1562 fd
= create_tmpfile(&tmp_file
, filename
);
1564 if (errno
== EACCES
)
1565 return error("insufficient permission for adding an object to repository database %s", get_object_directory());
1567 return error_errno("unable to create temporary file");
1571 git_deflate_init(&stream
, zlib_compression_level
);
1572 stream
.next_out
= compressed
;
1573 stream
.avail_out
= sizeof(compressed
);
1576 /* First header.. */
1577 stream
.next_in
= (unsigned char *)hdr
;
1578 stream
.avail_in
= hdrlen
;
1579 while (git_deflate(&stream
, 0) == Z_OK
)
1581 git_SHA1_Update(&c
, hdr
, hdrlen
);
1583 /* Then the data itself.. */
1584 stream
.next_in
= (void *)buf
;
1585 stream
.avail_in
= len
;
1587 unsigned char *in0
= stream
.next_in
;
1588 ret
= git_deflate(&stream
, Z_FINISH
);
1589 git_SHA1_Update(&c
, in0
, stream
.next_in
- in0
);
1590 if (write_buffer(fd
, compressed
, stream
.next_out
- compressed
) < 0)
1591 die("unable to write sha1 file");
1592 stream
.next_out
= compressed
;
1593 stream
.avail_out
= sizeof(compressed
);
1594 } while (ret
== Z_OK
);
1596 if (ret
!= Z_STREAM_END
)
1597 die("unable to deflate new object %s (%d)", sha1_to_hex(sha1
), ret
);
1598 ret
= git_deflate_end_gently(&stream
);
1600 die("deflateEnd on object %s failed (%d)", sha1_to_hex(sha1
), ret
);
1601 git_SHA1_Final(parano_sha1
, &c
);
1602 if (hashcmp(sha1
, parano_sha1
) != 0)
1603 die("confused by unstable object source data for %s", sha1_to_hex(sha1
));
1605 close_sha1_file(fd
);
1610 utb
.modtime
= mtime
;
1611 if (utime(tmp_file
.buf
, &utb
) < 0)
1612 warning_errno("failed utime() on %s", tmp_file
.buf
);
1615 return finalize_object_file(tmp_file
.buf
, filename
);
1618 static int freshen_loose_object(const unsigned char *sha1
)
1620 return check_and_freshen(sha1
, 1);
1623 static int freshen_packed_object(const unsigned char *sha1
)
1625 struct pack_entry e
;
1626 if (!find_pack_entry(sha1
, &e
))
1630 if (!freshen_file(e
.p
->pack_name
))
1636 int write_sha1_file(const void *buf
, unsigned long len
, const char *type
, unsigned char *sha1
)
1639 int hdrlen
= sizeof(hdr
);
1641 /* Normally if we have it in the pack then we do not bother writing
1642 * it out into .git/objects/??/?{38} file.
1644 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1645 if (freshen_packed_object(sha1
) || freshen_loose_object(sha1
))
1647 return write_loose_object(sha1
, hdr
, hdrlen
, buf
, len
, 0);
1650 int hash_sha1_file_literally(const void *buf
, unsigned long len
, const char *type
,
1651 struct object_id
*oid
, unsigned flags
)
1654 int hdrlen
, status
= 0;
1656 /* type string, SP, %lu of the length plus NUL must fit this */
1657 hdrlen
= strlen(type
) + 32;
1658 header
= xmalloc(hdrlen
);
1659 write_sha1_file_prepare(buf
, len
, type
, oid
->hash
, header
, &hdrlen
);
1661 if (!(flags
& HASH_WRITE_OBJECT
))
1663 if (freshen_packed_object(oid
->hash
) || freshen_loose_object(oid
->hash
))
1665 status
= write_loose_object(oid
->hash
, header
, hdrlen
, buf
, len
, 0);
1672 int force_object_loose(const unsigned char *sha1
, time_t mtime
)
1676 enum object_type type
;
1681 if (has_loose_object(sha1
))
1683 buf
= read_object(sha1
, &type
, &len
);
1685 return error("cannot read sha1_file for %s", sha1_to_hex(sha1
));
1686 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", typename(type
), len
) + 1;
1687 ret
= write_loose_object(sha1
, hdr
, hdrlen
, buf
, len
, mtime
);
1693 int has_sha1_file_with_flags(const unsigned char *sha1
, int flags
)
1695 if (!startup_info
->have_repository
)
1697 return sha1_object_info_extended(sha1
, NULL
,
1698 flags
| OBJECT_INFO_SKIP_CACHED
) >= 0;
1701 int has_object_file(const struct object_id
*oid
)
1703 return has_sha1_file(oid
->hash
);
1706 int has_object_file_with_flags(const struct object_id
*oid
, int flags
)
1708 return has_sha1_file_with_flags(oid
->hash
, flags
);
1711 static void check_tree(const void *buf
, size_t size
)
1713 struct tree_desc desc
;
1714 struct name_entry entry
;
1716 init_tree_desc(&desc
, buf
, size
);
1717 while (tree_entry(&desc
, &entry
))
1719 * tree_entry() will die() on malformed entries */
1723 static void check_commit(const void *buf
, size_t size
)
1726 memset(&c
, 0, sizeof(c
));
1727 if (parse_commit_buffer(&c
, buf
, size
))
1728 die("corrupt commit");
1731 static void check_tag(const void *buf
, size_t size
)
1734 memset(&t
, 0, sizeof(t
));
1735 if (parse_tag_buffer(&t
, buf
, size
))
1739 static int index_mem(struct object_id
*oid
, void *buf
, size_t size
,
1740 enum object_type type
,
1741 const char *path
, unsigned flags
)
1743 int ret
, re_allocated
= 0;
1744 int write_object
= flags
& HASH_WRITE_OBJECT
;
1750 * Convert blobs to git internal format
1752 if ((type
== OBJ_BLOB
) && path
) {
1753 struct strbuf nbuf
= STRBUF_INIT
;
1754 if (convert_to_git(&the_index
, path
, buf
, size
, &nbuf
,
1755 get_safe_crlf(flags
))) {
1756 buf
= strbuf_detach(&nbuf
, &size
);
1760 if (flags
& HASH_FORMAT_CHECK
) {
1761 if (type
== OBJ_TREE
)
1762 check_tree(buf
, size
);
1763 if (type
== OBJ_COMMIT
)
1764 check_commit(buf
, size
);
1765 if (type
== OBJ_TAG
)
1766 check_tag(buf
, size
);
1770 ret
= write_sha1_file(buf
, size
, typename(type
), oid
->hash
);
1772 ret
= hash_sha1_file(buf
, size
, typename(type
), oid
->hash
);
1778 static int index_stream_convert_blob(struct object_id
*oid
, int fd
,
1779 const char *path
, unsigned flags
)
1782 const int write_object
= flags
& HASH_WRITE_OBJECT
;
1783 struct strbuf sbuf
= STRBUF_INIT
;
1786 assert(would_convert_to_git_filter_fd(path
));
1788 convert_to_git_filter_fd(&the_index
, path
, fd
, &sbuf
,
1789 get_safe_crlf(flags
));
1792 ret
= write_sha1_file(sbuf
.buf
, sbuf
.len
, typename(OBJ_BLOB
),
1795 ret
= hash_sha1_file(sbuf
.buf
, sbuf
.len
, typename(OBJ_BLOB
),
1797 strbuf_release(&sbuf
);
1801 static int index_pipe(struct object_id
*oid
, int fd
, enum object_type type
,
1802 const char *path
, unsigned flags
)
1804 struct strbuf sbuf
= STRBUF_INIT
;
1807 if (strbuf_read(&sbuf
, fd
, 4096) >= 0)
1808 ret
= index_mem(oid
, sbuf
.buf
, sbuf
.len
, type
, path
, flags
);
1811 strbuf_release(&sbuf
);
1815 #define SMALL_FILE_SIZE (32*1024)
1817 static int index_core(struct object_id
*oid
, int fd
, size_t size
,
1818 enum object_type type
, const char *path
,
1824 ret
= index_mem(oid
, "", size
, type
, path
, flags
);
1825 } else if (size
<= SMALL_FILE_SIZE
) {
1826 char *buf
= xmalloc(size
);
1827 ssize_t read_result
= read_in_full(fd
, buf
, size
);
1828 if (read_result
< 0)
1829 ret
= error_errno("read error while indexing %s",
1830 path
? path
: "<unknown>");
1831 else if (read_result
!= size
)
1832 ret
= error("short read while indexing %s",
1833 path
? path
: "<unknown>");
1835 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1838 void *buf
= xmmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1839 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1846 * This creates one packfile per large blob unless bulk-checkin
1847 * machinery is "plugged".
1849 * This also bypasses the usual "convert-to-git" dance, and that is on
1850 * purpose. We could write a streaming version of the converting
1851 * functions and insert that before feeding the data to fast-import
1852 * (or equivalent in-core API described above). However, that is
1853 * somewhat complicated, as we do not know the size of the filter
1854 * result, which we need to know beforehand when writing a git object.
1855 * Since the primary motivation for trying to stream from the working
1856 * tree file and to avoid mmaping it in core is to deal with large
1857 * binary blobs, they generally do not want to get any conversion, and
1858 * callers should avoid this code path when filters are requested.
1860 static int index_stream(struct object_id
*oid
, int fd
, size_t size
,
1861 enum object_type type
, const char *path
,
1864 return index_bulk_checkin(oid
->hash
, fd
, size
, type
, path
, flags
);
1867 int index_fd(struct object_id
*oid
, int fd
, struct stat
*st
,
1868 enum object_type type
, const char *path
, unsigned flags
)
1873 * Call xsize_t() only when needed to avoid potentially unnecessary
1874 * die() for large files.
1876 if (type
== OBJ_BLOB
&& path
&& would_convert_to_git_filter_fd(path
))
1877 ret
= index_stream_convert_blob(oid
, fd
, path
, flags
);
1878 else if (!S_ISREG(st
->st_mode
))
1879 ret
= index_pipe(oid
, fd
, type
, path
, flags
);
1880 else if (st
->st_size
<= big_file_threshold
|| type
!= OBJ_BLOB
||
1881 (path
&& would_convert_to_git(&the_index
, path
)))
1882 ret
= index_core(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1885 ret
= index_stream(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1891 int index_path(struct object_id
*oid
, const char *path
, struct stat
*st
, unsigned flags
)
1894 struct strbuf sb
= STRBUF_INIT
;
1897 switch (st
->st_mode
& S_IFMT
) {
1899 fd
= open(path
, O_RDONLY
);
1901 return error_errno("open(\"%s\")", path
);
1902 if (index_fd(oid
, fd
, st
, OBJ_BLOB
, path
, flags
) < 0)
1903 return error("%s: failed to insert into database",
1907 if (strbuf_readlink(&sb
, path
, st
->st_size
))
1908 return error_errno("readlink(\"%s\")", path
);
1909 if (!(flags
& HASH_WRITE_OBJECT
))
1910 hash_sha1_file(sb
.buf
, sb
.len
, blob_type
, oid
->hash
);
1911 else if (write_sha1_file(sb
.buf
, sb
.len
, blob_type
, oid
->hash
))
1912 rc
= error("%s: failed to insert into database", path
);
1913 strbuf_release(&sb
);
1916 return resolve_gitlink_ref(path
, "HEAD", oid
);
1918 return error("%s: unsupported file type", path
);
1923 int read_pack_header(int fd
, struct pack_header
*header
)
1925 if (read_in_full(fd
, header
, sizeof(*header
)) != sizeof(*header
))
1926 /* "eof before pack header was fully read" */
1927 return PH_ERROR_EOF
;
1929 if (header
->hdr_signature
!= htonl(PACK_SIGNATURE
))
1930 /* "protocol error (pack signature mismatch detected)" */
1931 return PH_ERROR_PACK_SIGNATURE
;
1932 if (!pack_version_ok(header
->hdr_version
))
1933 /* "protocol error (pack version unsupported)" */
1934 return PH_ERROR_PROTOCOL
;
1938 void assert_sha1_type(const unsigned char *sha1
, enum object_type expect
)
1940 enum object_type type
= sha1_object_info(sha1
, NULL
);
1942 die("%s is not a valid object", sha1_to_hex(sha1
));
1944 die("%s is not a valid '%s' object", sha1_to_hex(sha1
),
1948 int for_each_file_in_obj_subdir(unsigned int subdir_nr
,
1949 struct strbuf
*path
,
1950 each_loose_object_fn obj_cb
,
1951 each_loose_cruft_fn cruft_cb
,
1952 each_loose_subdir_fn subdir_cb
,
1955 size_t origlen
, baselen
;
1959 struct object_id oid
;
1961 if (subdir_nr
> 0xff)
1962 BUG("invalid loose object subdirectory: %x", subdir_nr
);
1964 origlen
= path
->len
;
1965 strbuf_complete(path
, '/');
1966 strbuf_addf(path
, "%02x", subdir_nr
);
1968 dir
= opendir(path
->buf
);
1970 if (errno
!= ENOENT
)
1971 r
= error_errno("unable to open %s", path
->buf
);
1972 strbuf_setlen(path
, origlen
);
1976 oid
.hash
[0] = subdir_nr
;
1977 strbuf_addch(path
, '/');
1978 baselen
= path
->len
;
1980 while ((de
= readdir(dir
))) {
1982 if (is_dot_or_dotdot(de
->d_name
))
1985 namelen
= strlen(de
->d_name
);
1986 strbuf_setlen(path
, baselen
);
1987 strbuf_add(path
, de
->d_name
, namelen
);
1988 if (namelen
== GIT_SHA1_HEXSZ
- 2 &&
1989 !hex_to_bytes(oid
.hash
+ 1, de
->d_name
,
1990 GIT_SHA1_RAWSZ
- 1)) {
1992 r
= obj_cb(&oid
, path
->buf
, data
);
2000 r
= cruft_cb(de
->d_name
, path
->buf
, data
);
2007 strbuf_setlen(path
, baselen
- 1);
2008 if (!r
&& subdir_cb
)
2009 r
= subdir_cb(subdir_nr
, path
->buf
, data
);
2011 strbuf_setlen(path
, origlen
);
2016 int for_each_loose_file_in_objdir_buf(struct strbuf
*path
,
2017 each_loose_object_fn obj_cb
,
2018 each_loose_cruft_fn cruft_cb
,
2019 each_loose_subdir_fn subdir_cb
,
2025 for (i
= 0; i
< 256; i
++) {
2026 r
= for_each_file_in_obj_subdir(i
, path
, obj_cb
, cruft_cb
,
2035 int for_each_loose_file_in_objdir(const char *path
,
2036 each_loose_object_fn obj_cb
,
2037 each_loose_cruft_fn cruft_cb
,
2038 each_loose_subdir_fn subdir_cb
,
2041 struct strbuf buf
= STRBUF_INIT
;
2044 strbuf_addstr(&buf
, path
);
2045 r
= for_each_loose_file_in_objdir_buf(&buf
, obj_cb
, cruft_cb
,
2047 strbuf_release(&buf
);
2052 struct loose_alt_odb_data
{
2053 each_loose_object_fn
*cb
;
2057 static int loose_from_alt_odb(struct alternate_object_database
*alt
,
2060 struct loose_alt_odb_data
*data
= vdata
;
2061 struct strbuf buf
= STRBUF_INIT
;
2064 strbuf_addstr(&buf
, alt
->path
);
2065 r
= for_each_loose_file_in_objdir_buf(&buf
,
2066 data
->cb
, NULL
, NULL
,
2068 strbuf_release(&buf
);
2072 int for_each_loose_object(each_loose_object_fn cb
, void *data
, unsigned flags
)
2074 struct loose_alt_odb_data alt
;
2077 r
= for_each_loose_file_in_objdir(get_object_directory(),
2078 cb
, NULL
, NULL
, data
);
2082 if (flags
& FOR_EACH_OBJECT_LOCAL_ONLY
)
2087 return foreach_alt_odb(loose_from_alt_odb
, &alt
);
2090 static int check_stream_sha1(git_zstream
*stream
,
2094 const unsigned char *expected_sha1
)
2097 unsigned char real_sha1
[GIT_MAX_RAWSZ
];
2098 unsigned char buf
[4096];
2099 unsigned long total_read
;
2103 git_SHA1_Update(&c
, hdr
, stream
->total_out
);
2106 * We already read some bytes into hdr, but the ones up to the NUL
2107 * do not count against the object's content size.
2109 total_read
= stream
->total_out
- strlen(hdr
) - 1;
2112 * This size comparison must be "<=" to read the final zlib packets;
2113 * see the comment in unpack_sha1_rest for details.
2115 while (total_read
<= size
&&
2116 (status
== Z_OK
|| status
== Z_BUF_ERROR
)) {
2117 stream
->next_out
= buf
;
2118 stream
->avail_out
= sizeof(buf
);
2119 if (size
- total_read
< stream
->avail_out
)
2120 stream
->avail_out
= size
- total_read
;
2121 status
= git_inflate(stream
, Z_FINISH
);
2122 git_SHA1_Update(&c
, buf
, stream
->next_out
- buf
);
2123 total_read
+= stream
->next_out
- buf
;
2125 git_inflate_end(stream
);
2127 if (status
!= Z_STREAM_END
) {
2128 error("corrupt loose object '%s'", sha1_to_hex(expected_sha1
));
2131 if (stream
->avail_in
) {
2132 error("garbage at end of loose object '%s'",
2133 sha1_to_hex(expected_sha1
));
2137 git_SHA1_Final(real_sha1
, &c
);
2138 if (hashcmp(expected_sha1
, real_sha1
)) {
2139 error("sha1 mismatch for %s (expected %s)", path
,
2140 sha1_to_hex(expected_sha1
));
2147 int read_loose_object(const char *path
,
2148 const unsigned char *expected_sha1
,
2149 enum object_type
*type
,
2150 unsigned long *size
,
2155 unsigned long mapsize
;
2161 map
= map_sha1_file_1(path
, NULL
, &mapsize
);
2163 error_errno("unable to mmap %s", path
);
2167 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0) {
2168 error("unable to unpack header of %s", path
);
2172 *type
= parse_sha1_header(hdr
, size
);
2174 error("unable to parse header of %s", path
);
2175 git_inflate_end(&stream
);
2179 if (*type
== OBJ_BLOB
) {
2180 if (check_stream_sha1(&stream
, hdr
, *size
, path
, expected_sha1
) < 0)
2183 *contents
= unpack_sha1_rest(&stream
, hdr
, *size
, expected_sha1
);
2185 error("unable to unpack contents of %s", path
);
2186 git_inflate_end(&stream
);
2189 if (check_sha1_signature(expected_sha1
, *contents
,
2190 *size
, typename(*type
))) {
2191 error("sha1 mismatch for %s (expected %s)", path
,
2192 sha1_to_hex(expected_sha1
));
2198 ret
= 0; /* everything checks out */
2202 munmap(map
, mapsize
);