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 "repository.h"
26 #include "replace-object.h"
27 #include "streaming.h"
30 #include "mergesort.h"
33 #include "fetch-object.h"
34 #include "object-store.h"
36 /* The maximum size for an object header. */
37 #define MAX_HEADER_LEN 32
40 #define EMPTY_TREE_SHA1_BIN_LITERAL \
41 "\x4b\x82\x5d\xc6\x42\xcb\x6e\xb9\xa0\x60" \
42 "\xe5\x4b\xf8\xd6\x92\x88\xfb\xee\x49\x04"
44 #define EMPTY_BLOB_SHA1_BIN_LITERAL \
45 "\xe6\x9d\xe2\x9b\xb2\xd1\xd6\x43\x4b\x8b" \
46 "\x29\xae\x77\x5a\xd8\xc2\xe4\x8c\x53\x91"
48 const unsigned char null_sha1
[GIT_MAX_RAWSZ
];
49 const struct object_id null_oid
;
50 static const struct object_id empty_tree_oid
= {
51 EMPTY_TREE_SHA1_BIN_LITERAL
53 static const struct object_id empty_blob_oid
= {
54 EMPTY_BLOB_SHA1_BIN_LITERAL
57 static void git_hash_sha1_init(git_hash_ctx
*ctx
)
59 git_SHA1_Init(&ctx
->sha1
);
62 static void git_hash_sha1_update(git_hash_ctx
*ctx
, const void *data
, size_t len
)
64 git_SHA1_Update(&ctx
->sha1
, data
, len
);
67 static void git_hash_sha1_final(unsigned char *hash
, git_hash_ctx
*ctx
)
69 git_SHA1_Final(hash
, &ctx
->sha1
);
72 static void git_hash_unknown_init(git_hash_ctx
*ctx
)
74 BUG("trying to init unknown hash");
77 static void git_hash_unknown_update(git_hash_ctx
*ctx
, const void *data
, size_t len
)
79 BUG("trying to update unknown hash");
82 static void git_hash_unknown_final(unsigned char *hash
, git_hash_ctx
*ctx
)
84 BUG("trying to finalize unknown hash");
87 const struct git_hash_algo hash_algos
[GIT_HASH_NALGOS
] = {
93 git_hash_unknown_init
,
94 git_hash_unknown_update
,
95 git_hash_unknown_final
,
101 /* "sha1", big-endian */
106 git_hash_sha1_update
,
113 const char *empty_tree_oid_hex(void)
115 static char buf
[GIT_MAX_HEXSZ
+ 1];
116 return oid_to_hex_r(buf
, the_hash_algo
->empty_tree
);
119 const char *empty_blob_oid_hex(void)
121 static char buf
[GIT_MAX_HEXSZ
+ 1];
122 return oid_to_hex_r(buf
, the_hash_algo
->empty_blob
);
125 int hash_algo_by_name(const char *name
)
129 return GIT_HASH_UNKNOWN
;
130 for (i
= 1; i
< GIT_HASH_NALGOS
; i
++)
131 if (!strcmp(name
, hash_algos
[i
].name
))
133 return GIT_HASH_UNKNOWN
;
136 int hash_algo_by_id(uint32_t format_id
)
139 for (i
= 1; i
< GIT_HASH_NALGOS
; i
++)
140 if (format_id
== hash_algos
[i
].format_id
)
142 return GIT_HASH_UNKNOWN
;
147 * This is meant to hold a *small* number of objects that you would
148 * want read_sha1_file() to be able to return, but yet you do not want
149 * to write them into the object store (e.g. a browse-only
152 static struct cached_object
{
153 struct object_id oid
;
154 enum object_type type
;
158 static int cached_object_nr
, cached_object_alloc
;
160 static struct cached_object empty_tree
= {
161 { EMPTY_TREE_SHA1_BIN_LITERAL
},
167 static struct cached_object
*find_cached_object(const struct object_id
*oid
)
170 struct cached_object
*co
= cached_objects
;
172 for (i
= 0; i
< cached_object_nr
; i
++, co
++) {
173 if (oideq(&co
->oid
, oid
))
176 if (oideq(oid
, the_hash_algo
->empty_tree
))
182 static int get_conv_flags(unsigned flags
)
184 if (flags
& HASH_RENORMALIZE
)
185 return CONV_EOL_RENORMALIZE
;
186 else if (flags
& HASH_WRITE_OBJECT
)
187 return global_conv_flags_eol
| CONV_WRITE_OBJECT
;
193 int mkdir_in_gitdir(const char *path
)
195 if (mkdir(path
, 0777)) {
196 int saved_errno
= errno
;
198 struct strbuf sb
= STRBUF_INIT
;
203 * Are we looking at a path in a symlinked worktree
204 * whose original repository does not yet have it?
205 * e.g. .git/rr-cache pointing at its original
206 * repository in which the user hasn't performed any
207 * conflict resolution yet?
209 if (lstat(path
, &st
) || !S_ISLNK(st
.st_mode
) ||
210 strbuf_readlink(&sb
, path
, st
.st_size
) ||
211 !is_absolute_path(sb
.buf
) ||
212 mkdir(sb
.buf
, 0777)) {
219 return adjust_shared_perm(path
);
222 enum scld_error
safe_create_leading_directories(char *path
)
224 char *next_component
= path
+ offset_1st_component(path
);
225 enum scld_error ret
= SCLD_OK
;
227 while (ret
== SCLD_OK
&& next_component
) {
229 char *slash
= next_component
, slash_character
;
231 while (*slash
&& !is_dir_sep(*slash
))
237 next_component
= slash
+ 1;
238 while (is_dir_sep(*next_component
))
240 if (!*next_component
)
243 slash_character
= *slash
;
245 if (!stat(path
, &st
)) {
247 if (!S_ISDIR(st
.st_mode
)) {
251 } else if (mkdir(path
, 0777)) {
252 if (errno
== EEXIST
&&
253 !stat(path
, &st
) && S_ISDIR(st
.st_mode
))
254 ; /* somebody created it since we checked */
255 else if (errno
== ENOENT
)
257 * Either mkdir() failed because
258 * somebody just pruned the containing
259 * directory, or stat() failed because
260 * the file that was in our way was
261 * just removed. Either way, inform
262 * the caller that it might be worth
268 } else if (adjust_shared_perm(path
)) {
271 *slash
= slash_character
;
276 enum scld_error
safe_create_leading_directories_const(const char *path
)
279 /* path points to cache entries, so xstrdup before messing with it */
280 char *buf
= xstrdup(path
);
281 enum scld_error result
= safe_create_leading_directories(buf
);
289 int raceproof_create_file(const char *path
, create_file_fn fn
, void *cb
)
292 * The number of times we will try to remove empty directories
293 * in the way of path. This is only 1 because if another
294 * process is racily creating directories that conflict with
295 * us, we don't want to fight against them.
297 int remove_directories_remaining
= 1;
300 * The number of times that we will try to create the
301 * directories containing path. We are willing to attempt this
302 * more than once, because another process could be trying to
303 * clean up empty directories at the same time as we are
304 * trying to create them.
306 int create_directories_remaining
= 3;
308 /* A scratch copy of path, filled lazily if we need it: */
309 struct strbuf path_copy
= STRBUF_INIT
;
322 if (errno
== EISDIR
&& remove_directories_remaining
-- > 0) {
324 * A directory is in the way. Maybe it is empty; try
328 strbuf_addstr(&path_copy
, path
);
330 if (!remove_dir_recursively(&path_copy
, REMOVE_DIR_EMPTY_ONLY
))
332 } else if (errno
== ENOENT
&& create_directories_remaining
-- > 0) {
334 * Maybe the containing directory didn't exist, or
335 * maybe it was just deleted by a process that is
336 * racing with us to clean up empty directories. Try
339 enum scld_error scld_result
;
342 strbuf_addstr(&path_copy
, path
);
345 scld_result
= safe_create_leading_directories(path_copy
.buf
);
346 if (scld_result
== SCLD_OK
)
348 } while (scld_result
== SCLD_VANISHED
&& create_directories_remaining
-- > 0);
352 strbuf_release(&path_copy
);
357 static void fill_sha1_path(struct strbuf
*buf
, const unsigned char *sha1
)
360 for (i
= 0; i
< the_hash_algo
->rawsz
; i
++) {
361 static char hex
[] = "0123456789abcdef";
362 unsigned int val
= sha1
[i
];
363 strbuf_addch(buf
, hex
[val
>> 4]);
364 strbuf_addch(buf
, hex
[val
& 0xf]);
366 strbuf_addch(buf
, '/');
370 void sha1_file_name(struct repository
*r
, struct strbuf
*buf
, const unsigned char *sha1
)
372 strbuf_addstr(buf
, r
->objects
->objectdir
);
373 strbuf_addch(buf
, '/');
374 fill_sha1_path(buf
, sha1
);
377 struct strbuf
*alt_scratch_buf(struct alternate_object_database
*alt
)
379 strbuf_setlen(&alt
->scratch
, alt
->base_len
);
380 return &alt
->scratch
;
383 static const char *alt_sha1_path(struct alternate_object_database
*alt
,
384 const unsigned char *sha1
)
386 struct strbuf
*buf
= alt_scratch_buf(alt
);
387 fill_sha1_path(buf
, sha1
);
392 * Return non-zero iff the path is usable as an alternate object database.
394 static int alt_odb_usable(struct raw_object_store
*o
,
396 const char *normalized_objdir
)
398 struct alternate_object_database
*alt
;
400 /* Detect cases where alternate disappeared */
401 if (!is_directory(path
->buf
)) {
402 error(_("object directory %s does not exist; "
403 "check .git/objects/info/alternates"),
409 * Prevent the common mistake of listing the same
410 * thing twice, or object directory itself.
412 for (alt
= o
->alt_odb_list
; alt
; alt
= alt
->next
) {
413 if (!fspathcmp(path
->buf
, alt
->path
))
416 if (!fspathcmp(path
->buf
, normalized_objdir
))
423 * Prepare alternate object database registry.
425 * The variable alt_odb_list points at the list of struct
426 * alternate_object_database. The elements on this list come from
427 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
428 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
429 * whose contents is similar to that environment variable but can be
430 * LF separated. Its base points at a statically allocated buffer that
431 * contains "/the/directory/corresponding/to/.git/objects/...", while
432 * its name points just after the slash at the end of ".git/objects/"
433 * in the example above, and has enough space to hold 40-byte hex
434 * SHA1, an extra slash for the first level indirection, and the
437 static void read_info_alternates(struct repository
*r
,
438 const char *relative_base
,
440 static int link_alt_odb_entry(struct repository
*r
, const char *entry
,
441 const char *relative_base
, int depth
, const char *normalized_objdir
)
443 struct alternate_object_database
*ent
;
444 struct strbuf pathbuf
= STRBUF_INIT
;
446 if (!is_absolute_path(entry
) && relative_base
) {
447 strbuf_realpath(&pathbuf
, relative_base
, 1);
448 strbuf_addch(&pathbuf
, '/');
450 strbuf_addstr(&pathbuf
, entry
);
452 if (strbuf_normalize_path(&pathbuf
) < 0 && relative_base
) {
453 error(_("unable to normalize alternate object path: %s"),
455 strbuf_release(&pathbuf
);
460 * The trailing slash after the directory name is given by
461 * this function at the end. Remove duplicates.
463 while (pathbuf
.len
&& pathbuf
.buf
[pathbuf
.len
- 1] == '/')
464 strbuf_setlen(&pathbuf
, pathbuf
.len
- 1);
466 if (!alt_odb_usable(r
->objects
, &pathbuf
, normalized_objdir
)) {
467 strbuf_release(&pathbuf
);
471 ent
= alloc_alt_odb(pathbuf
.buf
);
473 /* add the alternate entry */
474 *r
->objects
->alt_odb_tail
= ent
;
475 r
->objects
->alt_odb_tail
= &(ent
->next
);
478 /* recursively add alternates */
479 read_info_alternates(r
, pathbuf
.buf
, depth
+ 1);
481 strbuf_release(&pathbuf
);
485 static const char *parse_alt_odb_entry(const char *string
,
493 if (*string
== '#') {
494 /* comment; consume up to next separator */
495 end
= strchrnul(string
, sep
);
496 } else if (*string
== '"' && !unquote_c_style(out
, string
, &end
)) {
498 * quoted path; unquote_c_style has copied the
499 * data for us and set "end". Broken quoting (e.g.,
500 * an entry that doesn't end with a quote) falls
501 * back to the unquoted case below.
504 /* normal, unquoted path */
505 end
= strchrnul(string
, sep
);
506 strbuf_add(out
, string
, end
- string
);
514 static void link_alt_odb_entries(struct repository
*r
, const char *alt
,
515 int sep
, const char *relative_base
, int depth
)
517 struct strbuf objdirbuf
= STRBUF_INIT
;
518 struct strbuf entry
= STRBUF_INIT
;
524 error(_("%s: ignoring alternate object stores, nesting too deep"),
529 strbuf_add_absolute_path(&objdirbuf
, r
->objects
->objectdir
);
530 if (strbuf_normalize_path(&objdirbuf
) < 0)
531 die(_("unable to normalize object directory: %s"),
535 alt
= parse_alt_odb_entry(alt
, sep
, &entry
);
538 link_alt_odb_entry(r
, entry
.buf
,
539 relative_base
, depth
, objdirbuf
.buf
);
541 strbuf_release(&entry
);
542 strbuf_release(&objdirbuf
);
545 static void read_info_alternates(struct repository
*r
,
546 const char *relative_base
,
550 struct strbuf buf
= STRBUF_INIT
;
552 path
= xstrfmt("%s/info/alternates", relative_base
);
553 if (strbuf_read_file(&buf
, path
, 1024) < 0) {
554 warn_on_fopen_errors(path
);
559 link_alt_odb_entries(r
, buf
.buf
, '\n', relative_base
, depth
);
560 strbuf_release(&buf
);
564 struct alternate_object_database
*alloc_alt_odb(const char *dir
)
566 struct alternate_object_database
*ent
;
568 FLEX_ALLOC_STR(ent
, path
, dir
);
569 strbuf_init(&ent
->scratch
, 0);
570 strbuf_addf(&ent
->scratch
, "%s/", dir
);
571 ent
->base_len
= ent
->scratch
.len
;
576 void add_to_alternates_file(const char *reference
)
578 struct lock_file lock
= LOCK_INIT
;
579 char *alts
= git_pathdup("objects/info/alternates");
583 hold_lock_file_for_update(&lock
, alts
, LOCK_DIE_ON_ERROR
);
584 out
= fdopen_lock_file(&lock
, "w");
586 die_errno(_("unable to fdopen alternates lockfile"));
588 in
= fopen(alts
, "r");
590 struct strbuf line
= STRBUF_INIT
;
592 while (strbuf_getline(&line
, in
) != EOF
) {
593 if (!strcmp(reference
, line
.buf
)) {
597 fprintf_or_die(out
, "%s\n", line
.buf
);
600 strbuf_release(&line
);
603 else if (errno
!= ENOENT
)
604 die_errno(_("unable to read alternates file"));
607 rollback_lock_file(&lock
);
609 fprintf_or_die(out
, "%s\n", reference
);
610 if (commit_lock_file(&lock
))
611 die_errno(_("unable to move new alternates file into place"));
612 if (the_repository
->objects
->alt_odb_tail
)
613 link_alt_odb_entries(the_repository
, reference
,
619 void add_to_alternates_memory(const char *reference
)
622 * Make sure alternates are initialized, or else our entry may be
623 * overwritten when they are.
625 prepare_alt_odb(the_repository
);
627 link_alt_odb_entries(the_repository
, reference
,
632 * Compute the exact path an alternate is at and returns it. In case of
633 * error NULL is returned and the human readable error is added to `err`
634 * `path` may be relative and should point to $GIT_DIR.
635 * `err` must not be null.
637 char *compute_alternate_path(const char *path
, struct strbuf
*err
)
639 char *ref_git
= NULL
;
640 const char *repo
, *ref_git_s
;
643 ref_git_s
= real_path_if_valid(path
);
646 strbuf_addf(err
, _("path '%s' does not exist"), path
);
650 * Beware: read_gitfile(), real_path() and mkpath()
651 * return static buffer
653 ref_git
= xstrdup(ref_git_s
);
655 repo
= read_gitfile(ref_git
);
657 repo
= read_gitfile(mkpath("%s/.git", ref_git
));
660 ref_git
= xstrdup(repo
);
663 if (!repo
&& is_directory(mkpath("%s/.git/objects", ref_git
))) {
664 char *ref_git_git
= mkpathdup("%s/.git", ref_git
);
666 ref_git
= ref_git_git
;
667 } else if (!is_directory(mkpath("%s/objects", ref_git
))) {
668 struct strbuf sb
= STRBUF_INIT
;
670 if (get_common_dir(&sb
, ref_git
)) {
672 _("reference repository '%s' as a linked "
673 "checkout is not supported yet."),
678 strbuf_addf(err
, _("reference repository '%s' is not a "
679 "local repository."), path
);
683 if (!access(mkpath("%s/shallow", ref_git
), F_OK
)) {
684 strbuf_addf(err
, _("reference repository '%s' is shallow"),
690 if (!access(mkpath("%s/info/grafts", ref_git
), F_OK
)) {
692 _("reference repository '%s' is grafted"),
700 FREE_AND_NULL(ref_git
);
706 int foreach_alt_odb(alt_odb_fn fn
, void *cb
)
708 struct alternate_object_database
*ent
;
711 prepare_alt_odb(the_repository
);
712 for (ent
= the_repository
->objects
->alt_odb_list
; ent
; ent
= ent
->next
) {
720 void prepare_alt_odb(struct repository
*r
)
722 if (r
->objects
->alt_odb_tail
)
725 r
->objects
->alt_odb_tail
= &r
->objects
->alt_odb_list
;
726 link_alt_odb_entries(r
, r
->objects
->alternate_db
, PATH_SEP
, NULL
, 0);
728 read_info_alternates(r
, r
->objects
->objectdir
, 0);
731 /* Returns 1 if we have successfully freshened the file, 0 otherwise. */
732 static int freshen_file(const char *fn
)
735 t
.actime
= t
.modtime
= time(NULL
);
736 return !utime(fn
, &t
);
740 * All of the check_and_freshen functions return 1 if the file exists and was
741 * freshened (if freshening was requested), 0 otherwise. If they return
742 * 0, you should not assume that it is safe to skip a write of the object (it
743 * either does not exist on disk, or has a stale mtime and may be subject to
746 int check_and_freshen_file(const char *fn
, int freshen
)
748 if (access(fn
, F_OK
))
750 if (freshen
&& !freshen_file(fn
))
755 static int check_and_freshen_local(const struct object_id
*oid
, int freshen
)
757 static struct strbuf buf
= STRBUF_INIT
;
760 sha1_file_name(the_repository
, &buf
, oid
->hash
);
762 return check_and_freshen_file(buf
.buf
, freshen
);
765 static int check_and_freshen_nonlocal(const struct object_id
*oid
, int freshen
)
767 struct alternate_object_database
*alt
;
768 prepare_alt_odb(the_repository
);
769 for (alt
= the_repository
->objects
->alt_odb_list
; alt
; alt
= alt
->next
) {
770 const char *path
= alt_sha1_path(alt
, oid
->hash
);
771 if (check_and_freshen_file(path
, freshen
))
777 static int check_and_freshen(const struct object_id
*oid
, int freshen
)
779 return check_and_freshen_local(oid
, freshen
) ||
780 check_and_freshen_nonlocal(oid
, freshen
);
783 int has_loose_object_nonlocal(const struct object_id
*oid
)
785 return check_and_freshen_nonlocal(oid
, 0);
788 static int has_loose_object(const struct object_id
*oid
)
790 return check_and_freshen(oid
, 0);
793 static void mmap_limit_check(size_t length
)
795 static size_t limit
= 0;
797 limit
= git_env_ulong("GIT_MMAP_LIMIT", 0);
802 die(_("attempting to mmap %"PRIuMAX
" over limit %"PRIuMAX
),
803 (uintmax_t)length
, (uintmax_t)limit
);
806 void *xmmap_gently(void *start
, size_t length
,
807 int prot
, int flags
, int fd
, off_t offset
)
811 mmap_limit_check(length
);
812 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
813 if (ret
== MAP_FAILED
) {
816 release_pack_memory(length
);
817 ret
= mmap(start
, length
, prot
, flags
, fd
, offset
);
822 void *xmmap(void *start
, size_t length
,
823 int prot
, int flags
, int fd
, off_t offset
)
825 void *ret
= xmmap_gently(start
, length
, prot
, flags
, fd
, offset
);
826 if (ret
== MAP_FAILED
)
827 die_errno(_("mmap failed"));
832 * With an in-core object data in "map", rehash it to make sure the
833 * object name actually matches "sha1" to detect object corruption.
834 * With "map" == NULL, try reading the object named with "sha1" using
835 * the streaming interface and rehash it to do the same.
837 int check_object_signature(const struct object_id
*oid
, void *map
,
838 unsigned long size
, const char *type
)
840 struct object_id real_oid
;
841 enum object_type obj_type
;
842 struct git_istream
*st
;
844 char hdr
[MAX_HEADER_LEN
];
848 hash_object_file(map
, size
, type
, &real_oid
);
849 return !oideq(oid
, &real_oid
) ? -1 : 0;
852 st
= open_istream(oid
, &obj_type
, &size
, NULL
);
856 /* Generate the header */
857 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", type_name(obj_type
), size
) + 1;
860 the_hash_algo
->init_fn(&c
);
861 the_hash_algo
->update_fn(&c
, hdr
, hdrlen
);
864 ssize_t readlen
= read_istream(st
, buf
, sizeof(buf
));
872 the_hash_algo
->update_fn(&c
, buf
, readlen
);
874 the_hash_algo
->final_fn(real_oid
.hash
, &c
);
876 return !oideq(oid
, &real_oid
) ? -1 : 0;
879 int git_open_cloexec(const char *name
, int flags
)
882 static int o_cloexec
= O_CLOEXEC
;
884 fd
= open(name
, flags
| o_cloexec
);
885 if ((o_cloexec
& O_CLOEXEC
) && fd
< 0 && errno
== EINVAL
) {
886 /* Try again w/o O_CLOEXEC: the kernel might not support it */
887 o_cloexec
&= ~O_CLOEXEC
;
888 fd
= open(name
, flags
| o_cloexec
);
891 #if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
893 static int fd_cloexec
= FD_CLOEXEC
;
895 if (!o_cloexec
&& 0 <= fd
&& fd_cloexec
) {
896 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
897 int flags
= fcntl(fd
, F_GETFD
);
898 if (fcntl(fd
, F_SETFD
, flags
| fd_cloexec
))
907 * Find "sha1" as a loose object in the local repository or in an alternate.
908 * Returns 0 on success, negative on failure.
910 * The "path" out-parameter will give the path of the object we found (if any).
911 * Note that it may point to static storage and is only valid until another
912 * call to sha1_file_name(), etc.
914 static int stat_sha1_file(struct repository
*r
, const unsigned char *sha1
,
915 struct stat
*st
, const char **path
)
917 struct alternate_object_database
*alt
;
918 static struct strbuf buf
= STRBUF_INIT
;
921 sha1_file_name(r
, &buf
, sha1
);
924 if (!lstat(*path
, st
))
929 for (alt
= r
->objects
->alt_odb_list
; alt
; alt
= alt
->next
) {
930 *path
= alt_sha1_path(alt
, sha1
);
931 if (!lstat(*path
, st
))
939 * Like stat_sha1_file(), but actually open the object and return the
940 * descriptor. See the caveats on the "path" parameter above.
942 static int open_sha1_file(struct repository
*r
,
943 const unsigned char *sha1
, const char **path
)
946 struct alternate_object_database
*alt
;
947 int most_interesting_errno
;
948 static struct strbuf buf
= STRBUF_INIT
;
951 sha1_file_name(r
, &buf
, sha1
);
954 fd
= git_open(*path
);
957 most_interesting_errno
= errno
;
960 for (alt
= r
->objects
->alt_odb_list
; alt
; alt
= alt
->next
) {
961 *path
= alt_sha1_path(alt
, sha1
);
962 fd
= git_open(*path
);
965 if (most_interesting_errno
== ENOENT
)
966 most_interesting_errno
= errno
;
968 errno
= most_interesting_errno
;
973 * Map the loose object at "path" if it is not NULL, or the path found by
974 * searching for a loose object named "sha1".
976 static void *map_sha1_file_1(struct repository
*r
, const char *path
,
977 const unsigned char *sha1
, unsigned long *size
)
985 fd
= open_sha1_file(r
, sha1
, &path
);
990 if (!fstat(fd
, &st
)) {
991 *size
= xsize_t(st
.st_size
);
993 /* mmap() is forbidden on empty files */
994 error(_("object file %s is empty"), path
);
997 map
= xmmap(NULL
, *size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1004 void *map_sha1_file(struct repository
*r
,
1005 const unsigned char *sha1
, unsigned long *size
)
1007 return map_sha1_file_1(r
, NULL
, sha1
, size
);
1010 static int unpack_sha1_short_header(git_zstream
*stream
,
1011 unsigned char *map
, unsigned long mapsize
,
1012 void *buffer
, unsigned long bufsiz
)
1014 /* Get the data stream */
1015 memset(stream
, 0, sizeof(*stream
));
1016 stream
->next_in
= map
;
1017 stream
->avail_in
= mapsize
;
1018 stream
->next_out
= buffer
;
1019 stream
->avail_out
= bufsiz
;
1021 git_inflate_init(stream
);
1022 return git_inflate(stream
, 0);
1025 int unpack_sha1_header(git_zstream
*stream
,
1026 unsigned char *map
, unsigned long mapsize
,
1027 void *buffer
, unsigned long bufsiz
)
1029 int status
= unpack_sha1_short_header(stream
, map
, mapsize
,
1035 /* Make sure we have the terminating NUL */
1036 if (!memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
1041 static int unpack_sha1_header_to_strbuf(git_zstream
*stream
, unsigned char *map
,
1042 unsigned long mapsize
, void *buffer
,
1043 unsigned long bufsiz
, struct strbuf
*header
)
1047 status
= unpack_sha1_short_header(stream
, map
, mapsize
, buffer
, bufsiz
);
1052 * Check if entire header is unpacked in the first iteration.
1054 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
1058 * buffer[0..bufsiz] was not large enough. Copy the partial
1059 * result out to header, and then append the result of further
1060 * reading the stream.
1062 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
1063 stream
->next_out
= buffer
;
1064 stream
->avail_out
= bufsiz
;
1067 status
= git_inflate(stream
, 0);
1068 strbuf_add(header
, buffer
, stream
->next_out
- (unsigned char *)buffer
);
1069 if (memchr(buffer
, '\0', stream
->next_out
- (unsigned char *)buffer
))
1071 stream
->next_out
= buffer
;
1072 stream
->avail_out
= bufsiz
;
1073 } while (status
!= Z_STREAM_END
);
1077 static void *unpack_sha1_rest(git_zstream
*stream
, void *buffer
, unsigned long size
, const unsigned char *sha1
)
1079 int bytes
= strlen(buffer
) + 1;
1080 unsigned char *buf
= xmallocz(size
);
1084 n
= stream
->total_out
- bytes
;
1087 memcpy(buf
, (char *) buffer
+ bytes
, n
);
1089 if (bytes
<= size
) {
1091 * The above condition must be (bytes <= size), not
1092 * (bytes < size). In other words, even though we
1093 * expect no more output and set avail_out to zero,
1094 * the input zlib stream may have bytes that express
1095 * "this concludes the stream", and we *do* want to
1098 * Otherwise we would not be able to test that we
1099 * consumed all the input to reach the expected size;
1100 * we also want to check that zlib tells us that all
1101 * went well with status == Z_STREAM_END at the end.
1103 stream
->next_out
= buf
+ bytes
;
1104 stream
->avail_out
= size
- bytes
;
1105 while (status
== Z_OK
)
1106 status
= git_inflate(stream
, Z_FINISH
);
1108 if (status
== Z_STREAM_END
&& !stream
->avail_in
) {
1109 git_inflate_end(stream
);
1114 error(_("corrupt loose object '%s'"), sha1_to_hex(sha1
));
1115 else if (stream
->avail_in
)
1116 error(_("garbage at end of loose object '%s'"),
1123 * We used to just use "sscanf()", but that's actually way
1124 * too permissive for what we want to check. So do an anal
1125 * object header parse by hand.
1127 static int parse_sha1_header_extended(const char *hdr
, struct object_info
*oi
,
1130 const char *type_buf
= hdr
;
1132 int type
, type_len
= 0;
1135 * The type can be of any size but is followed by
1147 type
= type_from_string_gently(type_buf
, type_len
, 1);
1149 strbuf_add(oi
->type_name
, type_buf
, type_len
);
1151 * Set type to 0 if its an unknown object and
1152 * we're obtaining the type using '--allow-unknown-type'
1155 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
) && (type
< 0))
1158 die(_("invalid object type"));
1163 * The length must follow immediately, and be in canonical
1164 * decimal format (ie "010" is not valid).
1166 size
= *hdr
++ - '0';
1171 unsigned long c
= *hdr
- '0';
1175 size
= size
* 10 + c
;
1183 * The length must be followed by a zero byte
1185 return *hdr
? -1 : type
;
1188 int parse_sha1_header(const char *hdr
, unsigned long *sizep
)
1190 struct object_info oi
= OBJECT_INFO_INIT
;
1193 return parse_sha1_header_extended(hdr
, &oi
, 0);
1196 static int sha1_loose_object_info(struct repository
*r
,
1197 const unsigned char *sha1
,
1198 struct object_info
*oi
, int flags
)
1201 unsigned long mapsize
;
1204 char hdr
[MAX_HEADER_LEN
];
1205 struct strbuf hdrbuf
= STRBUF_INIT
;
1206 unsigned long size_scratch
;
1208 if (oi
->delta_base_sha1
)
1209 hashclr(oi
->delta_base_sha1
);
1212 * If we don't care about type or size, then we don't
1213 * need to look inside the object at all. Note that we
1214 * do not optimize out the stat call, even if the
1215 * caller doesn't care about the disk-size, since our
1216 * return value implicitly indicates whether the
1217 * object even exists.
1219 if (!oi
->typep
&& !oi
->type_name
&& !oi
->sizep
&& !oi
->contentp
) {
1222 if (stat_sha1_file(r
, sha1
, &st
, &path
) < 0)
1225 *oi
->disk_sizep
= st
.st_size
;
1229 map
= map_sha1_file(r
, sha1
, &mapsize
);
1234 oi
->sizep
= &size_scratch
;
1237 *oi
->disk_sizep
= mapsize
;
1238 if ((flags
& OBJECT_INFO_ALLOW_UNKNOWN_TYPE
)) {
1239 if (unpack_sha1_header_to_strbuf(&stream
, map
, mapsize
, hdr
, sizeof(hdr
), &hdrbuf
) < 0)
1240 status
= error(_("unable to unpack %s header with --allow-unknown-type"),
1242 } else if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1243 status
= error(_("unable to unpack %s header"),
1247 else if (hdrbuf
.len
) {
1248 if ((status
= parse_sha1_header_extended(hdrbuf
.buf
, oi
, flags
)) < 0)
1249 status
= error(_("unable to parse %s header with --allow-unknown-type"),
1251 } else if ((status
= parse_sha1_header_extended(hdr
, oi
, flags
)) < 0)
1252 status
= error(_("unable to parse %s header"), sha1_to_hex(sha1
));
1254 if (status
>= 0 && oi
->contentp
) {
1255 *oi
->contentp
= unpack_sha1_rest(&stream
, hdr
,
1257 if (!*oi
->contentp
) {
1258 git_inflate_end(&stream
);
1262 git_inflate_end(&stream
);
1264 munmap(map
, mapsize
);
1265 if (status
&& oi
->typep
)
1266 *oi
->typep
= status
;
1267 if (oi
->sizep
== &size_scratch
)
1269 strbuf_release(&hdrbuf
);
1270 oi
->whence
= OI_LOOSE
;
1271 return (status
< 0) ? status
: 0;
1274 int fetch_if_missing
= 1;
1276 int oid_object_info_extended(struct repository
*r
, const struct object_id
*oid
,
1277 struct object_info
*oi
, unsigned flags
)
1279 static struct object_info blank_oi
= OBJECT_INFO_INIT
;
1280 struct pack_entry e
;
1282 const struct object_id
*real
= oid
;
1283 int already_retried
= 0;
1285 if (flags
& OBJECT_INFO_LOOKUP_REPLACE
)
1286 real
= lookup_replace_object(r
, oid
);
1288 if (is_null_oid(real
))
1294 if (!(flags
& OBJECT_INFO_SKIP_CACHED
)) {
1295 struct cached_object
*co
= find_cached_object(real
);
1298 *(oi
->typep
) = co
->type
;
1300 *(oi
->sizep
) = co
->size
;
1302 *(oi
->disk_sizep
) = 0;
1303 if (oi
->delta_base_sha1
)
1304 hashclr(oi
->delta_base_sha1
);
1306 strbuf_addstr(oi
->type_name
, type_name(co
->type
));
1308 *oi
->contentp
= xmemdupz(co
->buf
, co
->size
);
1309 oi
->whence
= OI_CACHED
;
1315 if (find_pack_entry(r
, real
, &e
))
1318 if (flags
& OBJECT_INFO_IGNORE_LOOSE
)
1321 /* Most likely it's a loose object. */
1322 if (!sha1_loose_object_info(r
, real
->hash
, oi
, flags
))
1325 /* Not a loose object; someone else may have just packed it. */
1326 if (!(flags
& OBJECT_INFO_QUICK
)) {
1327 reprepare_packed_git(r
);
1328 if (find_pack_entry(r
, real
, &e
))
1332 /* Check if it is a missing object */
1333 if (fetch_if_missing
&& repository_format_partial_clone
&&
1334 !already_retried
&& r
== the_repository
) {
1336 * TODO Investigate having fetch_object() return
1337 * TODO error/success and stopping the music here.
1338 * TODO Pass a repository struct through fetch_object,
1339 * such that arbitrary repositories work.
1341 fetch_objects(repository_format_partial_clone
, real
, 1);
1342 already_retried
= 1;
1349 if (oi
== &blank_oi
)
1351 * We know that the caller doesn't actually need the
1352 * information below, so return early.
1355 rtype
= packed_object_info(r
, e
.p
, e
.offset
, oi
);
1357 mark_bad_packed_object(e
.p
, real
->hash
);
1358 return oid_object_info_extended(r
, real
, oi
, 0);
1359 } else if (oi
->whence
== OI_PACKED
) {
1360 oi
->u
.packed
.offset
= e
.offset
;
1361 oi
->u
.packed
.pack
= e
.p
;
1362 oi
->u
.packed
.is_delta
= (rtype
== OBJ_REF_DELTA
||
1363 rtype
== OBJ_OFS_DELTA
);
1369 /* returns enum object_type or negative */
1370 int oid_object_info(struct repository
*r
,
1371 const struct object_id
*oid
,
1372 unsigned long *sizep
)
1374 enum object_type type
;
1375 struct object_info oi
= OBJECT_INFO_INIT
;
1379 if (oid_object_info_extended(r
, oid
, &oi
,
1380 OBJECT_INFO_LOOKUP_REPLACE
) < 0)
1385 static void *read_object(const unsigned char *sha1
, enum object_type
*type
,
1386 unsigned long *size
)
1388 struct object_id oid
;
1389 struct object_info oi
= OBJECT_INFO_INIT
;
1393 oi
.contentp
= &content
;
1395 hashcpy(oid
.hash
, sha1
);
1397 if (oid_object_info_extended(the_repository
, &oid
, &oi
, 0) < 0)
1402 int pretend_object_file(void *buf
, unsigned long len
, enum object_type type
,
1403 struct object_id
*oid
)
1405 struct cached_object
*co
;
1407 hash_object_file(buf
, len
, type_name(type
), oid
);
1408 if (has_sha1_file(oid
->hash
) || find_cached_object(oid
))
1410 ALLOC_GROW(cached_objects
, cached_object_nr
+ 1, cached_object_alloc
);
1411 co
= &cached_objects
[cached_object_nr
++];
1414 co
->buf
= xmalloc(len
);
1415 memcpy(co
->buf
, buf
, len
);
1416 oidcpy(&co
->oid
, oid
);
1421 * This function dies on corrupt objects; the callers who want to
1422 * deal with them should arrange to call read_object() and give error
1423 * messages themselves.
1425 void *read_object_file_extended(const struct object_id
*oid
,
1426 enum object_type
*type
,
1427 unsigned long *size
,
1431 const struct packed_git
*p
;
1434 const struct object_id
*repl
= lookup_replace
?
1435 lookup_replace_object(the_repository
, oid
) : oid
;
1438 data
= read_object(repl
->hash
, type
, size
);
1442 if (errno
&& errno
!= ENOENT
)
1443 die_errno(_("failed to read object %s"), oid_to_hex(oid
));
1445 /* die if we replaced an object with one that does not exist */
1447 die(_("replacement %s not found for %s"),
1448 oid_to_hex(repl
), oid_to_hex(oid
));
1450 if (!stat_sha1_file(the_repository
, repl
->hash
, &st
, &path
))
1451 die(_("loose object %s (stored in %s) is corrupt"),
1452 oid_to_hex(repl
), path
);
1454 if ((p
= has_packed_and_bad(repl
->hash
)) != NULL
)
1455 die(_("packed object %s (stored in %s) is corrupt"),
1456 oid_to_hex(repl
), p
->pack_name
);
1461 void *read_object_with_reference(const struct object_id
*oid
,
1462 const char *required_type_name
,
1463 unsigned long *size
,
1464 struct object_id
*actual_oid_return
)
1466 enum object_type type
, required_type
;
1468 unsigned long isize
;
1469 struct object_id actual_oid
;
1471 required_type
= type_from_string(required_type_name
);
1472 oidcpy(&actual_oid
, oid
);
1474 int ref_length
= -1;
1475 const char *ref_type
= NULL
;
1477 buffer
= read_object_file(&actual_oid
, &type
, &isize
);
1480 if (type
== required_type
) {
1482 if (actual_oid_return
)
1483 oidcpy(actual_oid_return
, &actual_oid
);
1486 /* Handle references */
1487 else if (type
== OBJ_COMMIT
)
1489 else if (type
== OBJ_TAG
)
1490 ref_type
= "object ";
1495 ref_length
= strlen(ref_type
);
1497 if (ref_length
+ the_hash_algo
->hexsz
> isize
||
1498 memcmp(buffer
, ref_type
, ref_length
) ||
1499 get_oid_hex((char *) buffer
+ ref_length
, &actual_oid
)) {
1504 /* Now we have the ID of the referred-to object in
1505 * actual_oid. Check again. */
1509 static void write_object_file_prepare(const void *buf
, unsigned long len
,
1510 const char *type
, struct object_id
*oid
,
1511 char *hdr
, int *hdrlen
)
1515 /* Generate the header */
1516 *hdrlen
= xsnprintf(hdr
, *hdrlen
, "%s %lu", type
, len
)+1;
1519 the_hash_algo
->init_fn(&c
);
1520 the_hash_algo
->update_fn(&c
, hdr
, *hdrlen
);
1521 the_hash_algo
->update_fn(&c
, buf
, len
);
1522 the_hash_algo
->final_fn(oid
->hash
, &c
);
1526 * Move the just written object into its final resting place.
1528 int finalize_object_file(const char *tmpfile
, const char *filename
)
1532 if (object_creation_mode
== OBJECT_CREATION_USES_RENAMES
)
1534 else if (link(tmpfile
, filename
))
1538 * Coda hack - coda doesn't like cross-directory links,
1539 * so we fall back to a rename, which will mean that it
1540 * won't be able to check collisions, but that's not a
1543 * The same holds for FAT formatted media.
1545 * When this succeeds, we just return. We have nothing
1548 if (ret
&& ret
!= EEXIST
) {
1550 if (!rename(tmpfile
, filename
))
1554 unlink_or_warn(tmpfile
);
1556 if (ret
!= EEXIST
) {
1557 return error_errno(_("unable to write sha1 filename %s"), filename
);
1559 /* FIXME!!! Collision check here ? */
1563 if (adjust_shared_perm(filename
))
1564 return error(_("unable to set permission to '%s'"), filename
);
1568 static int write_buffer(int fd
, const void *buf
, size_t len
)
1570 if (write_in_full(fd
, buf
, len
) < 0)
1571 return error_errno(_("file write error"));
1575 int hash_object_file(const void *buf
, unsigned long len
, const char *type
,
1576 struct object_id
*oid
)
1578 char hdr
[MAX_HEADER_LEN
];
1579 int hdrlen
= sizeof(hdr
);
1580 write_object_file_prepare(buf
, len
, type
, oid
, hdr
, &hdrlen
);
1584 /* Finalize a file on disk, and close it. */
1585 static void close_sha1_file(int fd
)
1587 if (fsync_object_files
)
1588 fsync_or_die(fd
, "sha1 file");
1590 die_errno(_("error when closing sha1 file"));
1593 /* Size of directory component, including the ending '/' */
1594 static inline int directory_size(const char *filename
)
1596 const char *s
= strrchr(filename
, '/');
1599 return s
- filename
+ 1;
1603 * This creates a temporary file in the same directory as the final
1606 * We want to avoid cross-directory filename renames, because those
1607 * can have problems on various filesystems (FAT, NFS, Coda).
1609 static int create_tmpfile(struct strbuf
*tmp
, const char *filename
)
1611 int fd
, dirlen
= directory_size(filename
);
1614 strbuf_add(tmp
, filename
, dirlen
);
1615 strbuf_addstr(tmp
, "tmp_obj_XXXXXX");
1616 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1617 if (fd
< 0 && dirlen
&& errno
== ENOENT
) {
1619 * Make sure the directory exists; note that the contents
1620 * of the buffer are undefined after mkstemp returns an
1621 * error, so we have to rewrite the whole buffer from
1625 strbuf_add(tmp
, filename
, dirlen
- 1);
1626 if (mkdir(tmp
->buf
, 0777) && errno
!= EEXIST
)
1628 if (adjust_shared_perm(tmp
->buf
))
1632 strbuf_addstr(tmp
, "/tmp_obj_XXXXXX");
1633 fd
= git_mkstemp_mode(tmp
->buf
, 0444);
1638 static int write_loose_object(const struct object_id
*oid
, char *hdr
,
1639 int hdrlen
, const void *buf
, unsigned long len
,
1643 unsigned char compressed
[4096];
1646 struct object_id parano_oid
;
1647 static struct strbuf tmp_file
= STRBUF_INIT
;
1648 static struct strbuf filename
= STRBUF_INIT
;
1650 strbuf_reset(&filename
);
1651 sha1_file_name(the_repository
, &filename
, oid
->hash
);
1653 fd
= create_tmpfile(&tmp_file
, filename
.buf
);
1655 if (errno
== EACCES
)
1656 return error(_("insufficient permission for adding an object to repository database %s"), get_object_directory());
1658 return error_errno(_("unable to create temporary file"));
1662 git_deflate_init(&stream
, zlib_compression_level
);
1663 stream
.next_out
= compressed
;
1664 stream
.avail_out
= sizeof(compressed
);
1665 the_hash_algo
->init_fn(&c
);
1667 /* First header.. */
1668 stream
.next_in
= (unsigned char *)hdr
;
1669 stream
.avail_in
= hdrlen
;
1670 while (git_deflate(&stream
, 0) == Z_OK
)
1672 the_hash_algo
->update_fn(&c
, hdr
, hdrlen
);
1674 /* Then the data itself.. */
1675 stream
.next_in
= (void *)buf
;
1676 stream
.avail_in
= len
;
1678 unsigned char *in0
= stream
.next_in
;
1679 ret
= git_deflate(&stream
, Z_FINISH
);
1680 the_hash_algo
->update_fn(&c
, in0
, stream
.next_in
- in0
);
1681 if (write_buffer(fd
, compressed
, stream
.next_out
- compressed
) < 0)
1682 die(_("unable to write sha1 file"));
1683 stream
.next_out
= compressed
;
1684 stream
.avail_out
= sizeof(compressed
);
1685 } while (ret
== Z_OK
);
1687 if (ret
!= Z_STREAM_END
)
1688 die(_("unable to deflate new object %s (%d)"), oid_to_hex(oid
),
1690 ret
= git_deflate_end_gently(&stream
);
1692 die(_("deflateEnd on object %s failed (%d)"), oid_to_hex(oid
),
1694 the_hash_algo
->final_fn(parano_oid
.hash
, &c
);
1695 if (!oideq(oid
, ¶no_oid
))
1696 die(_("confused by unstable object source data for %s"),
1699 close_sha1_file(fd
);
1704 utb
.modtime
= mtime
;
1705 if (utime(tmp_file
.buf
, &utb
) < 0)
1706 warning_errno(_("failed utime() on %s"), tmp_file
.buf
);
1709 return finalize_object_file(tmp_file
.buf
, filename
.buf
);
1712 static int freshen_loose_object(const struct object_id
*oid
)
1714 return check_and_freshen(oid
, 1);
1717 static int freshen_packed_object(const struct object_id
*oid
)
1719 struct pack_entry e
;
1720 if (!find_pack_entry(the_repository
, oid
, &e
))
1724 if (!freshen_file(e
.p
->pack_name
))
1730 int write_object_file(const void *buf
, unsigned long len
, const char *type
,
1731 struct object_id
*oid
)
1733 char hdr
[MAX_HEADER_LEN
];
1734 int hdrlen
= sizeof(hdr
);
1736 /* Normally if we have it in the pack then we do not bother writing
1737 * it out into .git/objects/??/?{38} file.
1739 write_object_file_prepare(buf
, len
, type
, oid
, hdr
, &hdrlen
);
1740 if (freshen_packed_object(oid
) || freshen_loose_object(oid
))
1742 return write_loose_object(oid
, hdr
, hdrlen
, buf
, len
, 0);
1745 int hash_object_file_literally(const void *buf
, unsigned long len
,
1746 const char *type
, struct object_id
*oid
,
1750 int hdrlen
, status
= 0;
1752 /* type string, SP, %lu of the length plus NUL must fit this */
1753 hdrlen
= strlen(type
) + MAX_HEADER_LEN
;
1754 header
= xmalloc(hdrlen
);
1755 write_object_file_prepare(buf
, len
, type
, oid
, header
, &hdrlen
);
1757 if (!(flags
& HASH_WRITE_OBJECT
))
1759 if (freshen_packed_object(oid
) || freshen_loose_object(oid
))
1761 status
= write_loose_object(oid
, header
, hdrlen
, buf
, len
, 0);
1768 int force_object_loose(const struct object_id
*oid
, time_t mtime
)
1772 enum object_type type
;
1773 char hdr
[MAX_HEADER_LEN
];
1777 if (has_loose_object(oid
))
1779 buf
= read_object(oid
->hash
, &type
, &len
);
1781 return error(_("cannot read sha1_file for %s"), oid_to_hex(oid
));
1782 hdrlen
= xsnprintf(hdr
, sizeof(hdr
), "%s %lu", type_name(type
), len
) + 1;
1783 ret
= write_loose_object(oid
, hdr
, hdrlen
, buf
, len
, mtime
);
1789 int has_sha1_file_with_flags(const unsigned char *sha1
, int flags
)
1791 struct object_id oid
;
1792 if (!startup_info
->have_repository
)
1794 hashcpy(oid
.hash
, sha1
);
1795 return oid_object_info_extended(the_repository
, &oid
, NULL
,
1796 flags
| OBJECT_INFO_SKIP_CACHED
) >= 0;
1799 int has_object_file(const struct object_id
*oid
)
1801 return has_sha1_file(oid
->hash
);
1804 int has_object_file_with_flags(const struct object_id
*oid
, int flags
)
1806 return has_sha1_file_with_flags(oid
->hash
, flags
);
1809 static void check_tree(const void *buf
, size_t size
)
1811 struct tree_desc desc
;
1812 struct name_entry entry
;
1814 init_tree_desc(&desc
, buf
, size
);
1815 while (tree_entry(&desc
, &entry
))
1817 * tree_entry() will die() on malformed entries */
1821 static void check_commit(const void *buf
, size_t size
)
1824 memset(&c
, 0, sizeof(c
));
1825 if (parse_commit_buffer(the_repository
, &c
, buf
, size
, 0))
1826 die(_("corrupt commit"));
1829 static void check_tag(const void *buf
, size_t size
)
1832 memset(&t
, 0, sizeof(t
));
1833 if (parse_tag_buffer(the_repository
, &t
, buf
, size
))
1834 die(_("corrupt tag"));
1837 static int index_mem(struct object_id
*oid
, void *buf
, size_t size
,
1838 enum object_type type
,
1839 const char *path
, unsigned flags
)
1841 int ret
, re_allocated
= 0;
1842 int write_object
= flags
& HASH_WRITE_OBJECT
;
1848 * Convert blobs to git internal format
1850 if ((type
== OBJ_BLOB
) && path
) {
1851 struct strbuf nbuf
= STRBUF_INIT
;
1852 if (convert_to_git(&the_index
, path
, buf
, size
, &nbuf
,
1853 get_conv_flags(flags
))) {
1854 buf
= strbuf_detach(&nbuf
, &size
);
1858 if (flags
& HASH_FORMAT_CHECK
) {
1859 if (type
== OBJ_TREE
)
1860 check_tree(buf
, size
);
1861 if (type
== OBJ_COMMIT
)
1862 check_commit(buf
, size
);
1863 if (type
== OBJ_TAG
)
1864 check_tag(buf
, size
);
1868 ret
= write_object_file(buf
, size
, type_name(type
), oid
);
1870 ret
= hash_object_file(buf
, size
, type_name(type
), oid
);
1876 static int index_stream_convert_blob(struct object_id
*oid
, int fd
,
1877 const char *path
, unsigned flags
)
1880 const int write_object
= flags
& HASH_WRITE_OBJECT
;
1881 struct strbuf sbuf
= STRBUF_INIT
;
1884 assert(would_convert_to_git_filter_fd(&the_index
, path
));
1886 convert_to_git_filter_fd(&the_index
, path
, fd
, &sbuf
,
1887 get_conv_flags(flags
));
1890 ret
= write_object_file(sbuf
.buf
, sbuf
.len
, type_name(OBJ_BLOB
),
1893 ret
= hash_object_file(sbuf
.buf
, sbuf
.len
, type_name(OBJ_BLOB
),
1895 strbuf_release(&sbuf
);
1899 static int index_pipe(struct object_id
*oid
, int fd
, enum object_type type
,
1900 const char *path
, unsigned flags
)
1902 struct strbuf sbuf
= STRBUF_INIT
;
1905 if (strbuf_read(&sbuf
, fd
, 4096) >= 0)
1906 ret
= index_mem(oid
, sbuf
.buf
, sbuf
.len
, type
, path
, flags
);
1909 strbuf_release(&sbuf
);
1913 #define SMALL_FILE_SIZE (32*1024)
1915 static int index_core(struct object_id
*oid
, int fd
, size_t size
,
1916 enum object_type type
, const char *path
,
1922 ret
= index_mem(oid
, "", size
, type
, path
, flags
);
1923 } else if (size
<= SMALL_FILE_SIZE
) {
1924 char *buf
= xmalloc(size
);
1925 ssize_t read_result
= read_in_full(fd
, buf
, size
);
1926 if (read_result
< 0)
1927 ret
= error_errno(_("read error while indexing %s"),
1928 path
? path
: "<unknown>");
1929 else if (read_result
!= size
)
1930 ret
= error(_("short read while indexing %s"),
1931 path
? path
: "<unknown>");
1933 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1936 void *buf
= xmmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1937 ret
= index_mem(oid
, buf
, size
, type
, path
, flags
);
1944 * This creates one packfile per large blob unless bulk-checkin
1945 * machinery is "plugged".
1947 * This also bypasses the usual "convert-to-git" dance, and that is on
1948 * purpose. We could write a streaming version of the converting
1949 * functions and insert that before feeding the data to fast-import
1950 * (or equivalent in-core API described above). However, that is
1951 * somewhat complicated, as we do not know the size of the filter
1952 * result, which we need to know beforehand when writing a git object.
1953 * Since the primary motivation for trying to stream from the working
1954 * tree file and to avoid mmaping it in core is to deal with large
1955 * binary blobs, they generally do not want to get any conversion, and
1956 * callers should avoid this code path when filters are requested.
1958 static int index_stream(struct object_id
*oid
, int fd
, size_t size
,
1959 enum object_type type
, const char *path
,
1962 return index_bulk_checkin(oid
, fd
, size
, type
, path
, flags
);
1965 int index_fd(struct object_id
*oid
, int fd
, struct stat
*st
,
1966 enum object_type type
, const char *path
, unsigned flags
)
1971 * Call xsize_t() only when needed to avoid potentially unnecessary
1972 * die() for large files.
1974 if (type
== OBJ_BLOB
&& path
&& would_convert_to_git_filter_fd(&the_index
, path
))
1975 ret
= index_stream_convert_blob(oid
, fd
, path
, flags
);
1976 else if (!S_ISREG(st
->st_mode
))
1977 ret
= index_pipe(oid
, fd
, type
, path
, flags
);
1978 else if (st
->st_size
<= big_file_threshold
|| type
!= OBJ_BLOB
||
1979 (path
&& would_convert_to_git(&the_index
, path
)))
1980 ret
= index_core(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1983 ret
= index_stream(oid
, fd
, xsize_t(st
->st_size
), type
, path
,
1989 int index_path(struct object_id
*oid
, const char *path
, struct stat
*st
, unsigned flags
)
1992 struct strbuf sb
= STRBUF_INIT
;
1995 switch (st
->st_mode
& S_IFMT
) {
1997 fd
= open(path
, O_RDONLY
);
1999 return error_errno("open(\"%s\")", path
);
2000 if (index_fd(oid
, fd
, st
, OBJ_BLOB
, path
, flags
) < 0)
2001 return error(_("%s: failed to insert into database"),
2005 if (strbuf_readlink(&sb
, path
, st
->st_size
))
2006 return error_errno("readlink(\"%s\")", path
);
2007 if (!(flags
& HASH_WRITE_OBJECT
))
2008 hash_object_file(sb
.buf
, sb
.len
, blob_type
, oid
);
2009 else if (write_object_file(sb
.buf
, sb
.len
, blob_type
, oid
))
2010 rc
= error(_("%s: failed to insert into database"), path
);
2011 strbuf_release(&sb
);
2014 return resolve_gitlink_ref(path
, "HEAD", oid
);
2016 return error(_("%s: unsupported file type"), path
);
2021 int read_pack_header(int fd
, struct pack_header
*header
)
2023 if (read_in_full(fd
, header
, sizeof(*header
)) != sizeof(*header
))
2024 /* "eof before pack header was fully read" */
2025 return PH_ERROR_EOF
;
2027 if (header
->hdr_signature
!= htonl(PACK_SIGNATURE
))
2028 /* "protocol error (pack signature mismatch detected)" */
2029 return PH_ERROR_PACK_SIGNATURE
;
2030 if (!pack_version_ok(header
->hdr_version
))
2031 /* "protocol error (pack version unsupported)" */
2032 return PH_ERROR_PROTOCOL
;
2036 void assert_oid_type(const struct object_id
*oid
, enum object_type expect
)
2038 enum object_type type
= oid_object_info(the_repository
, oid
, NULL
);
2040 die(_("%s is not a valid object"), oid_to_hex(oid
));
2042 die(_("%s is not a valid '%s' object"), oid_to_hex(oid
),
2046 int for_each_file_in_obj_subdir(unsigned int subdir_nr
,
2047 struct strbuf
*path
,
2048 each_loose_object_fn obj_cb
,
2049 each_loose_cruft_fn cruft_cb
,
2050 each_loose_subdir_fn subdir_cb
,
2053 size_t origlen
, baselen
;
2057 struct object_id oid
;
2059 if (subdir_nr
> 0xff)
2060 BUG("invalid loose object subdirectory: %x", subdir_nr
);
2062 origlen
= path
->len
;
2063 strbuf_complete(path
, '/');
2064 strbuf_addf(path
, "%02x", subdir_nr
);
2066 dir
= opendir(path
->buf
);
2068 if (errno
!= ENOENT
)
2069 r
= error_errno(_("unable to open %s"), path
->buf
);
2070 strbuf_setlen(path
, origlen
);
2074 oid
.hash
[0] = subdir_nr
;
2075 strbuf_addch(path
, '/');
2076 baselen
= path
->len
;
2078 while ((de
= readdir(dir
))) {
2080 if (is_dot_or_dotdot(de
->d_name
))
2083 namelen
= strlen(de
->d_name
);
2084 strbuf_setlen(path
, baselen
);
2085 strbuf_add(path
, de
->d_name
, namelen
);
2086 if (namelen
== the_hash_algo
->hexsz
- 2 &&
2087 !hex_to_bytes(oid
.hash
+ 1, de
->d_name
,
2088 the_hash_algo
->rawsz
- 1)) {
2090 r
= obj_cb(&oid
, path
->buf
, data
);
2098 r
= cruft_cb(de
->d_name
, path
->buf
, data
);
2105 strbuf_setlen(path
, baselen
- 1);
2106 if (!r
&& subdir_cb
)
2107 r
= subdir_cb(subdir_nr
, path
->buf
, data
);
2109 strbuf_setlen(path
, origlen
);
2114 int for_each_loose_file_in_objdir_buf(struct strbuf
*path
,
2115 each_loose_object_fn obj_cb
,
2116 each_loose_cruft_fn cruft_cb
,
2117 each_loose_subdir_fn subdir_cb
,
2123 for (i
= 0; i
< 256; i
++) {
2124 r
= for_each_file_in_obj_subdir(i
, path
, obj_cb
, cruft_cb
,
2133 int for_each_loose_file_in_objdir(const char *path
,
2134 each_loose_object_fn obj_cb
,
2135 each_loose_cruft_fn cruft_cb
,
2136 each_loose_subdir_fn subdir_cb
,
2139 struct strbuf buf
= STRBUF_INIT
;
2142 strbuf_addstr(&buf
, path
);
2143 r
= for_each_loose_file_in_objdir_buf(&buf
, obj_cb
, cruft_cb
,
2145 strbuf_release(&buf
);
2150 struct loose_alt_odb_data
{
2151 each_loose_object_fn
*cb
;
2155 static int loose_from_alt_odb(struct alternate_object_database
*alt
,
2158 struct loose_alt_odb_data
*data
= vdata
;
2159 struct strbuf buf
= STRBUF_INIT
;
2162 strbuf_addstr(&buf
, alt
->path
);
2163 r
= for_each_loose_file_in_objdir_buf(&buf
,
2164 data
->cb
, NULL
, NULL
,
2166 strbuf_release(&buf
);
2170 int for_each_loose_object(each_loose_object_fn cb
, void *data
,
2171 enum for_each_object_flags flags
)
2173 struct loose_alt_odb_data alt
;
2176 r
= for_each_loose_file_in_objdir(get_object_directory(),
2177 cb
, NULL
, NULL
, data
);
2181 if (flags
& FOR_EACH_OBJECT_LOCAL_ONLY
)
2186 return foreach_alt_odb(loose_from_alt_odb
, &alt
);
2189 static int check_stream_sha1(git_zstream
*stream
,
2193 const unsigned char *expected_sha1
)
2196 unsigned char real_sha1
[GIT_MAX_RAWSZ
];
2197 unsigned char buf
[4096];
2198 unsigned long total_read
;
2201 the_hash_algo
->init_fn(&c
);
2202 the_hash_algo
->update_fn(&c
, hdr
, stream
->total_out
);
2205 * We already read some bytes into hdr, but the ones up to the NUL
2206 * do not count against the object's content size.
2208 total_read
= stream
->total_out
- strlen(hdr
) - 1;
2211 * This size comparison must be "<=" to read the final zlib packets;
2212 * see the comment in unpack_sha1_rest for details.
2214 while (total_read
<= size
&&
2215 (status
== Z_OK
|| status
== Z_BUF_ERROR
)) {
2216 stream
->next_out
= buf
;
2217 stream
->avail_out
= sizeof(buf
);
2218 if (size
- total_read
< stream
->avail_out
)
2219 stream
->avail_out
= size
- total_read
;
2220 status
= git_inflate(stream
, Z_FINISH
);
2221 the_hash_algo
->update_fn(&c
, buf
, stream
->next_out
- buf
);
2222 total_read
+= stream
->next_out
- buf
;
2224 git_inflate_end(stream
);
2226 if (status
!= Z_STREAM_END
) {
2227 error(_("corrupt loose object '%s'"), sha1_to_hex(expected_sha1
));
2230 if (stream
->avail_in
) {
2231 error(_("garbage at end of loose object '%s'"),
2232 sha1_to_hex(expected_sha1
));
2236 the_hash_algo
->final_fn(real_sha1
, &c
);
2237 if (!hasheq(expected_sha1
, real_sha1
)) {
2238 error(_("sha1 mismatch for %s (expected %s)"), path
,
2239 sha1_to_hex(expected_sha1
));
2246 int read_loose_object(const char *path
,
2247 const struct object_id
*expected_oid
,
2248 enum object_type
*type
,
2249 unsigned long *size
,
2254 unsigned long mapsize
;
2256 char hdr
[MAX_HEADER_LEN
];
2260 map
= map_sha1_file_1(the_repository
, path
, NULL
, &mapsize
);
2262 error_errno(_("unable to mmap %s"), path
);
2266 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0) {
2267 error(_("unable to unpack header of %s"), path
);
2271 *type
= parse_sha1_header(hdr
, size
);
2273 error(_("unable to parse header of %s"), path
);
2274 git_inflate_end(&stream
);
2278 if (*type
== OBJ_BLOB
&& *size
> big_file_threshold
) {
2279 if (check_stream_sha1(&stream
, hdr
, *size
, path
, expected_oid
->hash
) < 0)
2282 *contents
= unpack_sha1_rest(&stream
, hdr
, *size
, expected_oid
->hash
);
2284 error(_("unable to unpack contents of %s"), path
);
2285 git_inflate_end(&stream
);
2288 if (check_object_signature(expected_oid
, *contents
,
2289 *size
, type_name(*type
))) {
2290 error(_("sha1 mismatch for %s (expected %s)"), path
,
2291 oid_to_hex(expected_oid
));
2297 ret
= 0; /* everything checks out */
2301 munmap(map
, mapsize
);