16 #include "streaming.h"
17 #include "thread-utils.h"
19 #include "object-store.h"
20 #include "replace-object.h"
21 #include "promisor-remote.h"
23 static const char index_pack_usage
[] =
24 "git index-pack [-v] [-o <index-file>] [--keep | --keep=<msg>] [--[no-]rev-index] [--verify] [--strict] (<pack-file> | --stdin [--fix-thin] [<pack-file>])";
27 struct pack_idx_entry idx
;
29 unsigned char hdr_size
;
31 signed char real_type
;
40 /* Initialized by make_base(). */
41 struct base_data
*base
;
42 struct object_entry
*obj
;
43 int ref_first
, ref_last
;
44 int ofs_first
, ofs_last
;
46 * Threads should increment retain_data if they are about to call
47 * patch_delta() using this struct's data as a base, and decrement this
48 * when they are done. While retain_data is nonzero, this struct's data
49 * will not be freed even if the delta base cache limit is exceeded.
53 * The number of direct children that have not been fully processed
54 * (entered work_head, entered done_head, left done_head). When this
55 * number reaches zero, this struct base_data can be freed.
57 int children_remaining
;
59 /* Not initialized by make_base(). */
60 struct list_head list
;
66 * Stack of struct base_data that have unprocessed children.
67 * threaded_second_pass() uses this as a source of work (the other being the
70 * Guarded by work_mutex.
72 static LIST_HEAD(work_head
);
75 * Stack of struct base_data that have children, all of whom have been
76 * processed or are being processed, and at least one child is being processed.
77 * These struct base_data must be kept around until the last child is
80 * Guarded by work_mutex.
82 static LIST_HEAD(done_head
);
85 * All threads share one delta base cache.
87 * base_cache_used is guarded by work_mutex, and base_cache_limit is read-only
90 static size_t base_cache_used
;
91 static size_t base_cache_limit
;
98 /* Remember to update object flag allocation in object.h */
99 #define FLAG_LINK (1u<<20)
100 #define FLAG_CHECKED (1u<<21)
102 struct ofs_delta_entry
{
107 struct ref_delta_entry
{
108 struct object_id oid
;
112 static struct object_entry
*objects
;
113 static struct object_stat
*obj_stat
;
114 static struct ofs_delta_entry
*ofs_deltas
;
115 static struct ref_delta_entry
*ref_deltas
;
116 static struct thread_local nothread_data
;
117 static int nr_objects
;
118 static int nr_ofs_deltas
;
119 static int nr_ref_deltas
;
120 static int ref_deltas_alloc
;
121 static int nr_resolved_deltas
;
122 static int nr_threads
;
124 static int from_stdin
;
126 static int do_fsck_object
;
127 static struct fsck_options fsck_options
= FSCK_OPTIONS_MISSING_GITMODULES
;
129 static const char *progress_title
;
130 static int show_resolving_progress
;
131 static int show_stat
;
132 static int check_self_contained_and_connected
;
134 static struct progress
*progress
;
136 /* We always read in 4kB chunks. */
137 static unsigned char input_buffer
[4096];
138 static unsigned int input_offset
, input_len
;
139 static off_t consumed_bytes
;
140 static off_t max_input_size
;
141 static unsigned deepest_delta
;
142 static git_hash_ctx input_ctx
;
143 static uint32_t input_crc32
;
144 static int input_fd
, output_fd
;
145 static const char *curr_pack
;
147 static struct thread_local
*thread_data
;
148 static int nr_dispatched
;
149 static int threads_active
;
151 static pthread_mutex_t read_mutex
;
152 #define read_lock() lock_mutex(&read_mutex)
153 #define read_unlock() unlock_mutex(&read_mutex)
155 static pthread_mutex_t counter_mutex
;
156 #define counter_lock() lock_mutex(&counter_mutex)
157 #define counter_unlock() unlock_mutex(&counter_mutex)
159 static pthread_mutex_t work_mutex
;
160 #define work_lock() lock_mutex(&work_mutex)
161 #define work_unlock() unlock_mutex(&work_mutex)
163 static pthread_mutex_t deepest_delta_mutex
;
164 #define deepest_delta_lock() lock_mutex(&deepest_delta_mutex)
165 #define deepest_delta_unlock() unlock_mutex(&deepest_delta_mutex)
167 static pthread_key_t key
;
169 static inline void lock_mutex(pthread_mutex_t
*mutex
)
172 pthread_mutex_lock(mutex
);
175 static inline void unlock_mutex(pthread_mutex_t
*mutex
)
178 pthread_mutex_unlock(mutex
);
182 * Mutex and conditional variable can't be statically-initialized on Windows.
184 static void init_thread(void)
187 init_recursive_mutex(&read_mutex
);
188 pthread_mutex_init(&counter_mutex
, NULL
);
189 pthread_mutex_init(&work_mutex
, NULL
);
191 pthread_mutex_init(&deepest_delta_mutex
, NULL
);
192 pthread_key_create(&key
, NULL
);
193 CALLOC_ARRAY(thread_data
, nr_threads
);
194 for (i
= 0; i
< nr_threads
; i
++) {
195 thread_data
[i
].pack_fd
= xopen(curr_pack
, O_RDONLY
);
201 static void cleanup_thread(void)
207 pthread_mutex_destroy(&read_mutex
);
208 pthread_mutex_destroy(&counter_mutex
);
209 pthread_mutex_destroy(&work_mutex
);
211 pthread_mutex_destroy(&deepest_delta_mutex
);
212 for (i
= 0; i
< nr_threads
; i
++)
213 close(thread_data
[i
].pack_fd
);
214 pthread_key_delete(key
);
218 static int mark_link(struct object
*obj
, enum object_type type
,
219 void *data
, struct fsck_options
*options
)
224 if (type
!= OBJ_ANY
&& obj
->type
!= type
)
225 die(_("object type mismatch at %s"), oid_to_hex(&obj
->oid
));
227 obj
->flags
|= FLAG_LINK
;
231 /* The content of each linked object must have been checked
232 or it must be already present in the object database */
233 static unsigned check_object(struct object
*obj
)
238 if (!(obj
->flags
& FLAG_LINK
))
241 if (!(obj
->flags
& FLAG_CHECKED
)) {
243 int type
= oid_object_info(the_repository
, &obj
->oid
, &size
);
245 die(_("did not receive expected object %s"),
246 oid_to_hex(&obj
->oid
));
247 if (type
!= obj
->type
)
248 die(_("object %s: expected type %s, found %s"),
249 oid_to_hex(&obj
->oid
),
250 type_name(obj
->type
), type_name(type
));
251 obj
->flags
|= FLAG_CHECKED
;
258 static unsigned check_objects(void)
260 unsigned i
, max
, foreign_nr
= 0;
262 max
= get_max_object_index();
265 progress
= start_delayed_progress(_("Checking objects"), max
);
267 for (i
= 0; i
< max
; i
++) {
268 foreign_nr
+= check_object(get_indexed_object(i
));
269 display_progress(progress
, i
+ 1);
272 stop_progress(&progress
);
277 /* Discard current buffer used content. */
278 static void flush(void)
282 write_or_die(output_fd
, input_buffer
, input_offset
);
283 the_hash_algo
->update_fn(&input_ctx
, input_buffer
, input_offset
);
284 memmove(input_buffer
, input_buffer
+ input_offset
, input_len
);
290 * Make sure at least "min" bytes are available in the buffer, and
291 * return the pointer to the buffer.
293 static void *fill(int min
)
295 if (min
<= input_len
)
296 return input_buffer
+ input_offset
;
297 if (min
> sizeof(input_buffer
))
298 die(Q_("cannot fill %d byte",
299 "cannot fill %d bytes",
304 ssize_t ret
= xread(input_fd
, input_buffer
+ input_len
,
305 sizeof(input_buffer
) - input_len
);
309 die_errno(_("read error on input"));
313 display_throughput(progress
, consumed_bytes
+ input_len
);
314 } while (input_len
< min
);
318 static void use(int bytes
)
320 if (bytes
> input_len
)
321 die(_("used more bytes than were available"));
322 input_crc32
= crc32(input_crc32
, input_buffer
+ input_offset
, bytes
);
324 input_offset
+= bytes
;
326 /* make sure off_t is sufficiently large not to wrap */
327 if (signed_add_overflows(consumed_bytes
, bytes
))
328 die(_("pack too large for current definition of off_t"));
329 consumed_bytes
+= bytes
;
330 if (max_input_size
&& consumed_bytes
> max_input_size
) {
331 struct strbuf size_limit
= STRBUF_INIT
;
332 strbuf_humanise_bytes(&size_limit
, max_input_size
);
333 die(_("pack exceeds maximum allowed size (%s)"),
338 static const char *open_pack_file(const char *pack_name
)
343 struct strbuf tmp_file
= STRBUF_INIT
;
344 output_fd
= odb_mkstemp(&tmp_file
,
345 "pack/tmp_pack_XXXXXX");
346 pack_name
= strbuf_detach(&tmp_file
, NULL
);
348 output_fd
= xopen(pack_name
, O_CREAT
|O_EXCL
|O_RDWR
, 0600);
350 nothread_data
.pack_fd
= output_fd
;
352 input_fd
= xopen(pack_name
, O_RDONLY
);
354 nothread_data
.pack_fd
= input_fd
;
356 the_hash_algo
->init_fn(&input_ctx
);
360 static void parse_pack_header(void)
362 struct pack_header
*hdr
= fill(sizeof(struct pack_header
));
364 /* Header consistency check */
365 if (hdr
->hdr_signature
!= htonl(PACK_SIGNATURE
))
366 die(_("pack signature mismatch"));
367 if (!pack_version_ok(hdr
->hdr_version
))
368 die(_("pack version %"PRIu32
" unsupported"),
369 ntohl(hdr
->hdr_version
));
371 nr_objects
= ntohl(hdr
->hdr_entries
);
372 use(sizeof(struct pack_header
));
375 __attribute__((format (printf
, 2, 3)))
376 static NORETURN
void bad_object(off_t offset
, const char *format
, ...)
381 va_start(params
, format
);
382 vsnprintf(buf
, sizeof(buf
), format
, params
);
384 die(_("pack has bad object at offset %"PRIuMAX
": %s"),
385 (uintmax_t)offset
, buf
);
388 static inline struct thread_local
*get_thread_data(void)
392 return pthread_getspecific(key
);
393 assert(!threads_active
&&
394 "This should only be reached when all threads are gone");
396 return ¬hread_data
;
399 static void set_thread_data(struct thread_local
*data
)
402 pthread_setspecific(key
, data
);
405 static void free_base_data(struct base_data
*c
)
408 FREE_AND_NULL(c
->data
);
409 base_cache_used
-= c
->size
;
413 static void prune_base_data(struct base_data
*retain
)
415 struct list_head
*pos
;
417 if (base_cache_used
<= base_cache_limit
)
420 list_for_each_prev(pos
, &done_head
) {
421 struct base_data
*b
= list_entry(pos
, struct base_data
, list
);
422 if (b
->retain_data
|| b
== retain
)
426 if (base_cache_used
<= base_cache_limit
)
431 list_for_each_prev(pos
, &work_head
) {
432 struct base_data
*b
= list_entry(pos
, struct base_data
, list
);
433 if (b
->retain_data
|| b
== retain
)
437 if (base_cache_used
<= base_cache_limit
)
443 static int is_delta_type(enum object_type type
)
445 return (type
== OBJ_REF_DELTA
|| type
== OBJ_OFS_DELTA
);
448 static void *unpack_entry_data(off_t offset
, unsigned long size
,
449 enum object_type type
, struct object_id
*oid
)
451 static char fixed_buf
[8192];
459 if (!is_delta_type(type
)) {
460 hdrlen
= format_object_header(hdr
, sizeof(hdr
), type
, size
);
461 the_hash_algo
->init_fn(&c
);
462 the_hash_algo
->update_fn(&c
, hdr
, hdrlen
);
465 if (type
== OBJ_BLOB
&& size
> big_file_threshold
)
468 buf
= xmallocz(size
);
470 memset(&stream
, 0, sizeof(stream
));
471 git_inflate_init(&stream
);
472 stream
.next_out
= buf
;
473 stream
.avail_out
= buf
== fixed_buf
? sizeof(fixed_buf
) : size
;
476 unsigned char *last_out
= stream
.next_out
;
477 stream
.next_in
= fill(1);
478 stream
.avail_in
= input_len
;
479 status
= git_inflate(&stream
, 0);
480 use(input_len
- stream
.avail_in
);
482 the_hash_algo
->update_fn(&c
, last_out
, stream
.next_out
- last_out
);
483 if (buf
== fixed_buf
) {
484 stream
.next_out
= buf
;
485 stream
.avail_out
= sizeof(fixed_buf
);
487 } while (status
== Z_OK
);
488 if (stream
.total_out
!= size
|| status
!= Z_STREAM_END
)
489 bad_object(offset
, _("inflate returned %d"), status
);
490 git_inflate_end(&stream
);
492 the_hash_algo
->final_oid_fn(oid
, &c
);
493 return buf
== fixed_buf
? NULL
: buf
;
496 static void *unpack_raw_entry(struct object_entry
*obj
,
498 struct object_id
*ref_oid
,
499 struct object_id
*oid
)
502 unsigned long size
, c
;
507 obj
->idx
.offset
= consumed_bytes
;
508 input_crc32
= crc32(0, NULL
, 0);
513 obj
->type
= (c
>> 4) & 7;
520 size
+= (c
& 0x7f) << shift
;
527 oidread(ref_oid
, fill(the_hash_algo
->rawsz
));
528 use(the_hash_algo
->rawsz
);
534 base_offset
= c
& 127;
537 if (!base_offset
|| MSB(base_offset
, 7))
538 bad_object(obj
->idx
.offset
, _("offset value overflow for delta base object"));
542 base_offset
= (base_offset
<< 7) + (c
& 127);
544 *ofs_offset
= obj
->idx
.offset
- base_offset
;
545 if (*ofs_offset
<= 0 || *ofs_offset
>= obj
->idx
.offset
)
546 bad_object(obj
->idx
.offset
, _("delta base offset is out of bound"));
554 bad_object(obj
->idx
.offset
, _("unknown object type %d"), obj
->type
);
556 obj
->hdr_size
= consumed_bytes
- obj
->idx
.offset
;
558 data
= unpack_entry_data(obj
->idx
.offset
, obj
->size
, obj
->type
, oid
);
559 obj
->idx
.crc32
= input_crc32
;
563 static void *unpack_data(struct object_entry
*obj
,
564 int (*consume
)(const unsigned char *, unsigned long, void *),
567 off_t from
= obj
[0].idx
.offset
+ obj
[0].hdr_size
;
568 off_t len
= obj
[1].idx
.offset
- from
;
569 unsigned char *data
, *inbuf
;
573 data
= xmallocz(consume
? 64*1024 : obj
->size
);
574 inbuf
= xmalloc((len
< 64*1024) ? (int)len
: 64*1024);
576 memset(&stream
, 0, sizeof(stream
));
577 git_inflate_init(&stream
);
578 stream
.next_out
= data
;
579 stream
.avail_out
= consume
? 64*1024 : obj
->size
;
582 ssize_t n
= (len
< 64*1024) ? (ssize_t
)len
: 64*1024;
583 n
= xpread(get_thread_data()->pack_fd
, inbuf
, n
, from
);
585 die_errno(_("cannot pread pack file"));
587 die(Q_("premature end of pack file, %"PRIuMAX
" byte missing",
588 "premature end of pack file, %"PRIuMAX
" bytes missing",
593 stream
.next_in
= inbuf
;
596 status
= git_inflate(&stream
, 0);
599 status
= git_inflate(&stream
, 0);
600 if (consume(data
, stream
.next_out
- data
, cb_data
)) {
605 stream
.next_out
= data
;
606 stream
.avail_out
= 64*1024;
607 } while (status
== Z_OK
&& stream
.avail_in
);
609 } while (len
&& status
== Z_OK
&& !stream
.avail_in
);
611 /* This has been inflated OK when first encountered, so... */
612 if (status
!= Z_STREAM_END
|| stream
.total_out
!= obj
->size
)
613 die(_("serious inflate inconsistency"));
615 git_inflate_end(&stream
);
623 static void *get_data_from_pack(struct object_entry
*obj
)
625 return unpack_data(obj
, NULL
, NULL
);
628 static int compare_ofs_delta_bases(off_t offset1
, off_t offset2
,
629 enum object_type type1
,
630 enum object_type type2
)
632 int cmp
= type1
- type2
;
635 return offset1
< offset2
? -1 :
636 offset1
> offset2
? 1 :
640 static int find_ofs_delta(const off_t offset
)
642 int first
= 0, last
= nr_ofs_deltas
;
644 while (first
< last
) {
645 int next
= first
+ (last
- first
) / 2;
646 struct ofs_delta_entry
*delta
= &ofs_deltas
[next
];
649 cmp
= compare_ofs_delta_bases(offset
, delta
->offset
,
651 objects
[delta
->obj_no
].type
);
663 static void find_ofs_delta_children(off_t offset
,
664 int *first_index
, int *last_index
)
666 int first
= find_ofs_delta(offset
);
668 int end
= nr_ofs_deltas
- 1;
675 while (first
> 0 && ofs_deltas
[first
- 1].offset
== offset
)
677 while (last
< end
&& ofs_deltas
[last
+ 1].offset
== offset
)
679 *first_index
= first
;
683 static int compare_ref_delta_bases(const struct object_id
*oid1
,
684 const struct object_id
*oid2
,
685 enum object_type type1
,
686 enum object_type type2
)
688 int cmp
= type1
- type2
;
691 return oidcmp(oid1
, oid2
);
694 static int find_ref_delta(const struct object_id
*oid
)
696 int first
= 0, last
= nr_ref_deltas
;
698 while (first
< last
) {
699 int next
= first
+ (last
- first
) / 2;
700 struct ref_delta_entry
*delta
= &ref_deltas
[next
];
703 cmp
= compare_ref_delta_bases(oid
, &delta
->oid
,
705 objects
[delta
->obj_no
].type
);
717 static void find_ref_delta_children(const struct object_id
*oid
,
718 int *first_index
, int *last_index
)
720 int first
= find_ref_delta(oid
);
722 int end
= nr_ref_deltas
- 1;
729 while (first
> 0 && oideq(&ref_deltas
[first
- 1].oid
, oid
))
731 while (last
< end
&& oideq(&ref_deltas
[last
+ 1].oid
, oid
))
733 *first_index
= first
;
737 struct compare_data
{
738 struct object_entry
*entry
;
739 struct git_istream
*st
;
741 unsigned long buf_size
;
744 static int compare_objects(const unsigned char *buf
, unsigned long size
,
747 struct compare_data
*data
= cb_data
;
749 if (data
->buf_size
< size
) {
751 data
->buf
= xmalloc(size
);
752 data
->buf_size
= size
;
756 ssize_t len
= read_istream(data
->st
, data
->buf
, size
);
758 die(_("SHA1 COLLISION FOUND WITH %s !"),
759 oid_to_hex(&data
->entry
->idx
.oid
));
761 die(_("unable to read %s"),
762 oid_to_hex(&data
->entry
->idx
.oid
));
763 if (memcmp(buf
, data
->buf
, len
))
764 die(_("SHA1 COLLISION FOUND WITH %s !"),
765 oid_to_hex(&data
->entry
->idx
.oid
));
772 static int check_collison(struct object_entry
*entry
)
774 struct compare_data data
;
775 enum object_type type
;
778 if (entry
->size
<= big_file_threshold
|| entry
->type
!= OBJ_BLOB
)
781 memset(&data
, 0, sizeof(data
));
783 data
.st
= open_istream(the_repository
, &entry
->idx
.oid
, &type
, &size
,
787 if (size
!= entry
->size
|| type
!= entry
->type
)
788 die(_("SHA1 COLLISION FOUND WITH %s !"),
789 oid_to_hex(&entry
->idx
.oid
));
790 unpack_data(entry
, compare_objects
, &data
);
791 close_istream(data
.st
);
796 static void sha1_object(const void *data
, struct object_entry
*obj_entry
,
797 unsigned long size
, enum object_type type
,
798 const struct object_id
*oid
)
800 void *new_data
= NULL
;
801 int collision_test_needed
= 0;
803 assert(data
|| obj_entry
);
805 if (startup_info
->have_repository
) {
807 collision_test_needed
=
808 has_object_file_with_flags(oid
, OBJECT_INFO_QUICK
);
812 if (collision_test_needed
&& !data
) {
814 if (!check_collison(obj_entry
))
815 collision_test_needed
= 0;
818 if (collision_test_needed
) {
820 enum object_type has_type
;
821 unsigned long has_size
;
823 has_type
= oid_object_info(the_repository
, oid
, &has_size
);
825 die(_("cannot read existing object info %s"), oid_to_hex(oid
));
826 if (has_type
!= type
|| has_size
!= size
)
827 die(_("SHA1 COLLISION FOUND WITH %s !"), oid_to_hex(oid
));
828 has_data
= read_object_file(oid
, &has_type
, &has_size
);
831 data
= new_data
= get_data_from_pack(obj_entry
);
833 die(_("cannot read existing object %s"), oid_to_hex(oid
));
834 if (size
!= has_size
|| type
!= has_type
||
835 memcmp(data
, has_data
, size
) != 0)
836 die(_("SHA1 COLLISION FOUND WITH %s !"), oid_to_hex(oid
));
840 if (strict
|| do_fsck_object
) {
842 if (type
== OBJ_BLOB
) {
843 struct blob
*blob
= lookup_blob(the_repository
, oid
);
845 blob
->object
.flags
|= FLAG_CHECKED
;
847 die(_("invalid blob object %s"), oid_to_hex(oid
));
848 if (do_fsck_object
&&
849 fsck_object(&blob
->object
, (void *)data
, size
, &fsck_options
))
850 die(_("fsck error in packed object"));
854 void *buf
= (void *) data
;
856 assert(data
&& "data can only be NULL for large _blobs_");
859 * we do not need to free the memory here, as the
860 * buf is deleted by the caller.
862 obj
= parse_object_buffer(the_repository
, oid
, type
,
866 die(_("invalid %s"), type_name(type
));
867 if (do_fsck_object
&&
868 fsck_object(obj
, buf
, size
, &fsck_options
))
869 die(_("fsck error in packed object"));
870 if (strict
&& fsck_walk(obj
, NULL
, &fsck_options
))
871 die(_("Not all child objects of %s are reachable"), oid_to_hex(&obj
->oid
));
873 if (obj
->type
== OBJ_TREE
) {
874 struct tree
*item
= (struct tree
*) obj
;
878 if (obj
->type
== OBJ_COMMIT
) {
879 struct commit
*commit
= (struct commit
*) obj
;
880 if (detach_commit_buffer(commit
, NULL
) != data
)
881 BUG("parse_object_buffer transmogrified our buffer");
883 obj
->flags
|= FLAG_CHECKED
;
892 * Ensure that this node has been reconstructed and return its contents.
894 * In the typical and best case, this node would already be reconstructed
895 * (through the invocation to resolve_delta() in threaded_second_pass()) and it
896 * would not be pruned. However, if pruning of this node was necessary due to
897 * reaching delta_base_cache_limit, this function will find the closest
898 * ancestor with reconstructed data that has not been pruned (or if there is
899 * none, the ultimate base object), and reconstruct each node in the delta
900 * chain in order to generate the reconstructed data for this node.
902 static void *get_base_data(struct base_data
*c
)
905 struct object_entry
*obj
= c
->obj
;
906 struct base_data
**delta
= NULL
;
907 int delta_nr
= 0, delta_alloc
= 0;
909 while (is_delta_type(c
->obj
->type
) && !c
->data
) {
910 ALLOC_GROW(delta
, delta_nr
+ 1, delta_alloc
);
911 delta
[delta_nr
++] = c
;
915 c
->data
= get_data_from_pack(obj
);
917 base_cache_used
+= c
->size
;
920 for (; delta_nr
> 0; delta_nr
--) {
922 c
= delta
[delta_nr
- 1];
924 base
= get_base_data(c
->base
);
925 raw
= get_data_from_pack(obj
);
926 c
->data
= patch_delta(
932 bad_object(obj
->idx
.offset
, _("failed to apply delta"));
933 base_cache_used
+= c
->size
;
941 static struct base_data
*make_base(struct object_entry
*obj
,
942 struct base_data
*parent
)
944 struct base_data
*base
= xcalloc(1, sizeof(struct base_data
));
947 find_ref_delta_children(&obj
->idx
.oid
,
948 &base
->ref_first
, &base
->ref_last
);
949 find_ofs_delta_children(obj
->idx
.offset
,
950 &base
->ofs_first
, &base
->ofs_last
);
951 base
->children_remaining
= base
->ref_last
- base
->ref_first
+
952 base
->ofs_last
- base
->ofs_first
+ 2;
956 static struct base_data
*resolve_delta(struct object_entry
*delta_obj
,
957 struct base_data
*base
)
959 void *delta_data
, *result_data
;
960 struct base_data
*result
;
961 unsigned long result_size
;
964 int i
= delta_obj
- objects
;
965 int j
= base
->obj
- objects
;
966 obj_stat
[i
].delta_depth
= obj_stat
[j
].delta_depth
+ 1;
967 deepest_delta_lock();
968 if (deepest_delta
< obj_stat
[i
].delta_depth
)
969 deepest_delta
= obj_stat
[i
].delta_depth
;
970 deepest_delta_unlock();
971 obj_stat
[i
].base_object_no
= j
;
973 delta_data
= get_data_from_pack(delta_obj
);
975 result_data
= patch_delta(base
->data
, base
->size
,
976 delta_data
, delta_obj
->size
, &result_size
);
979 bad_object(delta_obj
->idx
.offset
, _("failed to apply delta"));
980 hash_object_file(the_hash_algo
, result_data
, result_size
,
981 delta_obj
->real_type
, &delta_obj
->idx
.oid
);
982 sha1_object(result_data
, NULL
, result_size
, delta_obj
->real_type
,
983 &delta_obj
->idx
.oid
);
985 result
= make_base(delta_obj
, base
);
986 result
->data
= result_data
;
987 result
->size
= result_size
;
990 nr_resolved_deltas
++;
996 static int compare_ofs_delta_entry(const void *a
, const void *b
)
998 const struct ofs_delta_entry
*delta_a
= a
;
999 const struct ofs_delta_entry
*delta_b
= b
;
1001 return delta_a
->offset
< delta_b
->offset
? -1 :
1002 delta_a
->offset
> delta_b
->offset
? 1 :
1006 static int compare_ref_delta_entry(const void *a
, const void *b
)
1008 const struct ref_delta_entry
*delta_a
= a
;
1009 const struct ref_delta_entry
*delta_b
= b
;
1011 return oidcmp(&delta_a
->oid
, &delta_b
->oid
);
1014 static void *threaded_second_pass(void *data
)
1017 set_thread_data(data
);
1019 struct base_data
*parent
= NULL
;
1020 struct object_entry
*child_obj
;
1021 struct base_data
*child
;
1024 display_progress(progress
, nr_resolved_deltas
);
1028 if (list_empty(&work_head
)) {
1030 * Take an object from the object array.
1032 while (nr_dispatched
< nr_objects
&&
1033 is_delta_type(objects
[nr_dispatched
].type
))
1035 if (nr_dispatched
>= nr_objects
) {
1039 child_obj
= &objects
[nr_dispatched
++];
1042 * Peek at the top of the stack, and take a child from
1045 parent
= list_first_entry(&work_head
, struct base_data
,
1048 if (parent
->ref_first
<= parent
->ref_last
) {
1049 int offset
= ref_deltas
[parent
->ref_first
++].obj_no
;
1050 child_obj
= objects
+ offset
;
1051 if (child_obj
->real_type
!= OBJ_REF_DELTA
)
1052 die("REF_DELTA at offset %"PRIuMAX
" already resolved (duplicate base %s?)",
1053 (uintmax_t) child_obj
->idx
.offset
,
1054 oid_to_hex(&parent
->obj
->idx
.oid
));
1055 child_obj
->real_type
= parent
->obj
->real_type
;
1057 child_obj
= objects
+
1058 ofs_deltas
[parent
->ofs_first
++].obj_no
;
1059 assert(child_obj
->real_type
== OBJ_OFS_DELTA
);
1060 child_obj
->real_type
= parent
->obj
->real_type
;
1063 if (parent
->ref_first
> parent
->ref_last
&&
1064 parent
->ofs_first
> parent
->ofs_last
) {
1066 * This parent has run out of children, so move
1069 list_del(&parent
->list
);
1070 list_add(&parent
->list
, &done_head
);
1074 * Ensure that the parent has data, since we will need
1077 * NEEDSWORK: If parent data needs to be reloaded, this
1078 * prolongs the time that the current thread spends in
1079 * the mutex. A mitigating factor is that parent data
1080 * needs to be reloaded only if the delta base cache
1081 * limit is exceeded, so in the typical case, this does
1084 get_base_data(parent
);
1085 parent
->retain_data
++;
1090 child
= resolve_delta(child_obj
, parent
);
1091 if (!child
->children_remaining
)
1092 FREE_AND_NULL(child
->data
);
1094 child
= make_base(child_obj
, NULL
);
1095 if (child
->children_remaining
) {
1097 * Since this child has its own delta children,
1098 * we will need this data in the future.
1099 * Inflate now so that future iterations will
1100 * have access to this object's data while
1101 * outside the work mutex.
1103 child
->data
= get_data_from_pack(child_obj
);
1104 child
->size
= child_obj
->size
;
1110 parent
->retain_data
--;
1113 * This child has its own children, so add it to
1116 list_add(&child
->list
, &work_head
);
1117 base_cache_used
+= child
->size
;
1118 prune_base_data(NULL
);
1119 free_base_data(child
);
1122 * This child does not have its own children. It may be
1123 * the last descendant of its ancestors; free those
1126 struct base_data
*p
= parent
;
1129 struct base_data
*next_p
;
1131 p
->children_remaining
--;
1132 if (p
->children_remaining
)
1142 FREE_AND_NULL(child
);
1151 * - find locations of all objects;
1152 * - calculate SHA1 of all non-delta objects;
1153 * - remember base (SHA1 or offset) for all deltas.
1155 static void parse_pack_objects(unsigned char *hash
)
1157 int i
, nr_delays
= 0;
1158 struct ofs_delta_entry
*ofs_delta
= ofs_deltas
;
1159 struct object_id ref_delta_oid
;
1163 progress
= start_progress(
1164 progress_title
? progress_title
:
1165 from_stdin
? _("Receiving objects") : _("Indexing objects"),
1167 for (i
= 0; i
< nr_objects
; i
++) {
1168 struct object_entry
*obj
= &objects
[i
];
1169 void *data
= unpack_raw_entry(obj
, &ofs_delta
->offset
,
1172 obj
->real_type
= obj
->type
;
1173 if (obj
->type
== OBJ_OFS_DELTA
) {
1175 ofs_delta
->obj_no
= i
;
1177 } else if (obj
->type
== OBJ_REF_DELTA
) {
1178 ALLOC_GROW(ref_deltas
, nr_ref_deltas
+ 1, ref_deltas_alloc
);
1179 oidcpy(&ref_deltas
[nr_ref_deltas
].oid
, &ref_delta_oid
);
1180 ref_deltas
[nr_ref_deltas
].obj_no
= i
;
1183 /* large blobs, check later */
1184 obj
->real_type
= OBJ_BAD
;
1187 sha1_object(data
, NULL
, obj
->size
, obj
->type
,
1190 display_progress(progress
, i
+1);
1192 objects
[i
].idx
.offset
= consumed_bytes
;
1193 stop_progress(&progress
);
1195 /* Check pack integrity */
1197 the_hash_algo
->final_fn(hash
, &input_ctx
);
1198 if (!hasheq(fill(the_hash_algo
->rawsz
), hash
))
1199 die(_("pack is corrupted (SHA1 mismatch)"));
1200 use(the_hash_algo
->rawsz
);
1202 /* If input_fd is a file, we should have reached its end now. */
1203 if (fstat(input_fd
, &st
))
1204 die_errno(_("cannot fstat packfile"));
1205 if (S_ISREG(st
.st_mode
) &&
1206 lseek(input_fd
, 0, SEEK_CUR
) - input_len
!= st
.st_size
)
1207 die(_("pack has junk at the end"));
1209 for (i
= 0; i
< nr_objects
; i
++) {
1210 struct object_entry
*obj
= &objects
[i
];
1211 if (obj
->real_type
!= OBJ_BAD
)
1213 obj
->real_type
= obj
->type
;
1214 sha1_object(NULL
, obj
, obj
->size
, obj
->type
,
1219 die(_("confusion beyond insanity in parse_pack_objects()"));
1224 * - for all non-delta objects, look if it is used as a base for
1226 * - if used as a base, uncompress the object and apply all deltas,
1227 * recursively checking if the resulting object is used as a base
1228 * for some more deltas.
1230 static void resolve_deltas(void)
1234 if (!nr_ofs_deltas
&& !nr_ref_deltas
)
1237 /* Sort deltas by base SHA1/offset for fast searching */
1238 QSORT(ofs_deltas
, nr_ofs_deltas
, compare_ofs_delta_entry
);
1239 QSORT(ref_deltas
, nr_ref_deltas
, compare_ref_delta_entry
);
1241 if (verbose
|| show_resolving_progress
)
1242 progress
= start_progress(_("Resolving deltas"),
1243 nr_ref_deltas
+ nr_ofs_deltas
);
1246 base_cache_limit
= delta_base_cache_limit
* nr_threads
;
1247 if (nr_threads
> 1 || getenv("GIT_FORCE_THREADS")) {
1249 for (i
= 0; i
< nr_threads
; i
++) {
1250 int ret
= pthread_create(&thread_data
[i
].thread
, NULL
,
1251 threaded_second_pass
, thread_data
+ i
);
1253 die(_("unable to create thread: %s"),
1256 for (i
= 0; i
< nr_threads
; i
++)
1257 pthread_join(thread_data
[i
].thread
, NULL
);
1261 threaded_second_pass(¬hread_data
);
1266 * - append objects to convert thin pack to full pack if required
1267 * - write the final pack hash
1269 static void fix_unresolved_deltas(struct hashfile
*f
);
1270 static void conclude_pack(int fix_thin_pack
, const char *curr_pack
, unsigned char *pack_hash
)
1272 if (nr_ref_deltas
+ nr_ofs_deltas
== nr_resolved_deltas
) {
1273 stop_progress(&progress
);
1274 /* Flush remaining pack final hash. */
1279 if (fix_thin_pack
) {
1281 unsigned char read_hash
[GIT_MAX_RAWSZ
], tail_hash
[GIT_MAX_RAWSZ
];
1282 struct strbuf msg
= STRBUF_INIT
;
1283 int nr_unresolved
= nr_ofs_deltas
+ nr_ref_deltas
- nr_resolved_deltas
;
1284 int nr_objects_initial
= nr_objects
;
1285 if (nr_unresolved
<= 0)
1286 die(_("confusion beyond insanity"));
1287 REALLOC_ARRAY(objects
, nr_objects
+ nr_unresolved
+ 1);
1288 memset(objects
+ nr_objects
+ 1, 0,
1289 nr_unresolved
* sizeof(*objects
));
1290 f
= hashfd(output_fd
, curr_pack
);
1291 fix_unresolved_deltas(f
);
1292 strbuf_addf(&msg
, Q_("completed with %d local object",
1293 "completed with %d local objects",
1294 nr_objects
- nr_objects_initial
),
1295 nr_objects
- nr_objects_initial
);
1296 stop_progress_msg(&progress
, msg
.buf
);
1297 strbuf_release(&msg
);
1298 finalize_hashfile(f
, tail_hash
, FSYNC_COMPONENT_PACK
, 0);
1299 hashcpy(read_hash
, pack_hash
);
1300 fixup_pack_header_footer(output_fd
, pack_hash
,
1301 curr_pack
, nr_objects
,
1302 read_hash
, consumed_bytes
-the_hash_algo
->rawsz
);
1303 if (!hasheq(read_hash
, tail_hash
))
1304 die(_("Unexpected tail checksum for %s "
1305 "(disk corruption?)"), curr_pack
);
1307 if (nr_ofs_deltas
+ nr_ref_deltas
!= nr_resolved_deltas
)
1308 die(Q_("pack has %d unresolved delta",
1309 "pack has %d unresolved deltas",
1310 nr_ofs_deltas
+ nr_ref_deltas
- nr_resolved_deltas
),
1311 nr_ofs_deltas
+ nr_ref_deltas
- nr_resolved_deltas
);
1314 static int write_compressed(struct hashfile
*f
, void *in
, unsigned int size
)
1318 unsigned char outbuf
[4096];
1320 git_deflate_init(&stream
, zlib_compression_level
);
1321 stream
.next_in
= in
;
1322 stream
.avail_in
= size
;
1325 stream
.next_out
= outbuf
;
1326 stream
.avail_out
= sizeof(outbuf
);
1327 status
= git_deflate(&stream
, Z_FINISH
);
1328 hashwrite(f
, outbuf
, sizeof(outbuf
) - stream
.avail_out
);
1329 } while (status
== Z_OK
);
1331 if (status
!= Z_STREAM_END
)
1332 die(_("unable to deflate appended object (%d)"), status
);
1333 size
= stream
.total_out
;
1334 git_deflate_end(&stream
);
1338 static struct object_entry
*append_obj_to_pack(struct hashfile
*f
,
1339 const unsigned char *sha1
, void *buf
,
1340 unsigned long size
, enum object_type type
)
1342 struct object_entry
*obj
= &objects
[nr_objects
++];
1343 unsigned char header
[10];
1344 unsigned long s
= size
;
1346 unsigned char c
= (type
<< 4) | (s
& 15);
1349 header
[n
++] = c
| 0x80;
1355 hashwrite(f
, header
, n
);
1357 obj
[0].hdr_size
= n
;
1359 obj
[0].real_type
= type
;
1360 obj
[1].idx
.offset
= obj
[0].idx
.offset
+ n
;
1361 obj
[1].idx
.offset
+= write_compressed(f
, buf
, size
);
1362 obj
[0].idx
.crc32
= crc32_end(f
);
1364 oidread(&obj
->idx
.oid
, sha1
);
1368 static int delta_pos_compare(const void *_a
, const void *_b
)
1370 struct ref_delta_entry
*a
= *(struct ref_delta_entry
**)_a
;
1371 struct ref_delta_entry
*b
= *(struct ref_delta_entry
**)_b
;
1372 return a
->obj_no
- b
->obj_no
;
1375 static void fix_unresolved_deltas(struct hashfile
*f
)
1377 struct ref_delta_entry
**sorted_by_pos
;
1381 * Since many unresolved deltas may well be themselves base objects
1382 * for more unresolved deltas, we really want to include the
1383 * smallest number of base objects that would cover as much delta
1384 * as possible by picking the
1385 * trunc deltas first, allowing for other deltas to resolve without
1386 * additional base objects. Since most base objects are to be found
1387 * before deltas depending on them, a good heuristic is to start
1388 * resolving deltas in the same order as their position in the pack.
1390 ALLOC_ARRAY(sorted_by_pos
, nr_ref_deltas
);
1391 for (i
= 0; i
< nr_ref_deltas
; i
++)
1392 sorted_by_pos
[i
] = &ref_deltas
[i
];
1393 QSORT(sorted_by_pos
, nr_ref_deltas
, delta_pos_compare
);
1395 if (has_promisor_remote()) {
1397 * Prefetch the delta bases.
1399 struct oid_array to_fetch
= OID_ARRAY_INIT
;
1400 for (i
= 0; i
< nr_ref_deltas
; i
++) {
1401 struct ref_delta_entry
*d
= sorted_by_pos
[i
];
1402 if (!oid_object_info_extended(the_repository
, &d
->oid
,
1404 OBJECT_INFO_FOR_PREFETCH
))
1406 oid_array_append(&to_fetch
, &d
->oid
);
1408 promisor_remote_get_direct(the_repository
,
1409 to_fetch
.oid
, to_fetch
.nr
);
1410 oid_array_clear(&to_fetch
);
1413 for (i
= 0; i
< nr_ref_deltas
; i
++) {
1414 struct ref_delta_entry
*d
= sorted_by_pos
[i
];
1415 enum object_type type
;
1419 if (objects
[d
->obj_no
].real_type
!= OBJ_REF_DELTA
)
1421 data
= read_object_file(&d
->oid
, &type
, &size
);
1425 if (check_object_signature(the_repository
, &d
->oid
, data
, size
,
1427 die(_("local object %s is corrupt"), oid_to_hex(&d
->oid
));
1430 * Add this as an object to the objects array and call
1431 * threaded_second_pass() (which will pick up the added
1434 append_obj_to_pack(f
, d
->oid
.hash
, data
, size
, type
);
1436 threaded_second_pass(NULL
);
1438 display_progress(progress
, nr_resolved_deltas
);
1440 free(sorted_by_pos
);
1443 static const char *derive_filename(const char *pack_name
, const char *strip
,
1444 const char *suffix
, struct strbuf
*buf
)
1447 if (!strip_suffix(pack_name
, strip
, &len
) || !len
||
1448 pack_name
[len
- 1] != '.')
1449 die(_("packfile name '%s' does not end with '.%s'"),
1451 strbuf_add(buf
, pack_name
, len
);
1452 strbuf_addstr(buf
, suffix
);
1456 static void write_special_file(const char *suffix
, const char *msg
,
1457 const char *pack_name
, const unsigned char *hash
,
1458 const char **report
)
1460 struct strbuf name_buf
= STRBUF_INIT
;
1461 const char *filename
;
1463 int msg_len
= strlen(msg
);
1466 filename
= derive_filename(pack_name
, "pack", suffix
, &name_buf
);
1468 filename
= odb_pack_name(&name_buf
, hash
, suffix
);
1470 fd
= odb_pack_keep(filename
);
1472 if (errno
!= EEXIST
)
1473 die_errno(_("cannot write %s file '%s'"),
1477 write_or_die(fd
, msg
, msg_len
);
1478 write_or_die(fd
, "\n", 1);
1481 die_errno(_("cannot close written %s file '%s'"),
1486 strbuf_release(&name_buf
);
1489 static void rename_tmp_packfile(const char **final_name
,
1490 const char *curr_name
,
1491 struct strbuf
*name
, unsigned char *hash
,
1492 const char *ext
, int make_read_only_if_same
)
1494 if (*final_name
!= curr_name
) {
1496 *final_name
= odb_pack_name(name
, hash
, ext
);
1497 if (finalize_object_file(curr_name
, *final_name
))
1498 die(_("unable to rename temporary '*.%s' file to '%s'"),
1500 } else if (make_read_only_if_same
) {
1501 chmod(*final_name
, 0444);
1505 static void final(const char *final_pack_name
, const char *curr_pack_name
,
1506 const char *final_index_name
, const char *curr_index_name
,
1507 const char *final_rev_index_name
, const char *curr_rev_index_name
,
1508 const char *keep_msg
, const char *promisor_msg
,
1509 unsigned char *hash
)
1511 const char *report
= "pack";
1512 struct strbuf pack_name
= STRBUF_INIT
;
1513 struct strbuf index_name
= STRBUF_INIT
;
1514 struct strbuf rev_index_name
= STRBUF_INIT
;
1520 fsync_component_or_die(FSYNC_COMPONENT_PACK
, output_fd
, curr_pack_name
);
1521 err
= close(output_fd
);
1523 die_errno(_("error while closing pack file"));
1527 write_special_file("keep", keep_msg
, final_pack_name
, hash
,
1530 write_special_file("promisor", promisor_msg
, final_pack_name
,
1533 rename_tmp_packfile(&final_pack_name
, curr_pack_name
, &pack_name
,
1534 hash
, "pack", from_stdin
);
1535 if (curr_rev_index_name
)
1536 rename_tmp_packfile(&final_rev_index_name
, curr_rev_index_name
,
1537 &rev_index_name
, hash
, "rev", 1);
1538 rename_tmp_packfile(&final_index_name
, curr_index_name
, &index_name
,
1541 if (do_fsck_object
) {
1542 struct packed_git
*p
;
1543 p
= add_packed_git(final_index_name
, strlen(final_index_name
), 0);
1545 install_packed_git(the_repository
, p
);
1549 printf("%s\n", hash_to_hex(hash
));
1551 struct strbuf buf
= STRBUF_INIT
;
1553 strbuf_addf(&buf
, "%s\t%s\n", report
, hash_to_hex(hash
));
1554 write_or_die(1, buf
.buf
, buf
.len
);
1555 strbuf_release(&buf
);
1558 * Let's just mimic git-unpack-objects here and write
1559 * the last part of the input buffer to stdout.
1562 err
= xwrite(1, input_buffer
+ input_offset
, input_len
);
1566 input_offset
+= err
;
1570 strbuf_release(&rev_index_name
);
1571 strbuf_release(&index_name
);
1572 strbuf_release(&pack_name
);
1575 static int git_index_pack_config(const char *k
, const char *v
, void *cb
)
1577 struct pack_idx_option
*opts
= cb
;
1579 if (!strcmp(k
, "pack.indexversion")) {
1580 opts
->version
= git_config_int(k
, v
);
1581 if (opts
->version
> 2)
1582 die(_("bad pack.indexVersion=%"PRIu32
), opts
->version
);
1585 if (!strcmp(k
, "pack.threads")) {
1586 nr_threads
= git_config_int(k
, v
);
1588 die(_("invalid number of threads specified (%d)"),
1590 if (!HAVE_THREADS
&& nr_threads
!= 1) {
1591 warning(_("no threads support, ignoring %s"), k
);
1596 if (!strcmp(k
, "pack.writereverseindex")) {
1597 if (git_config_bool(k
, v
))
1598 opts
->flags
|= WRITE_REV
;
1600 opts
->flags
&= ~WRITE_REV
;
1602 return git_default_config(k
, v
, cb
);
1605 static int cmp_uint32(const void *a_
, const void *b_
)
1607 uint32_t a
= *((uint32_t *)a_
);
1608 uint32_t b
= *((uint32_t *)b_
);
1610 return (a
< b
) ? -1 : (a
!= b
);
1613 static void read_v2_anomalous_offsets(struct packed_git
*p
,
1614 struct pack_idx_option
*opts
)
1616 const uint32_t *idx1
, *idx2
;
1619 /* The address of the 4-byte offset table */
1620 idx1
= (((const uint32_t *)((const uint8_t *)p
->index_data
+ p
->crc_offset
))
1621 + (size_t)p
->num_objects
/* CRC32 table */
1624 /* The address of the 8-byte offset table */
1625 idx2
= idx1
+ p
->num_objects
;
1627 for (i
= 0; i
< p
->num_objects
; i
++) {
1628 uint32_t off
= ntohl(idx1
[i
]);
1629 if (!(off
& 0x80000000))
1631 off
= off
& 0x7fffffff;
1632 check_pack_index_ptr(p
, &idx2
[off
* 2]);
1636 * The real offset is ntohl(idx2[off * 2]) in high 4
1637 * octets, and ntohl(idx2[off * 2 + 1]) in low 4
1638 * octets. But idx2[off * 2] is Zero!!!
1640 ALLOC_GROW(opts
->anomaly
, opts
->anomaly_nr
+ 1, opts
->anomaly_alloc
);
1641 opts
->anomaly
[opts
->anomaly_nr
++] = ntohl(idx2
[off
* 2 + 1]);
1644 QSORT(opts
->anomaly
, opts
->anomaly_nr
, cmp_uint32
);
1647 static void read_idx_option(struct pack_idx_option
*opts
, const char *pack_name
)
1649 struct packed_git
*p
= add_packed_git(pack_name
, strlen(pack_name
), 1);
1652 die(_("Cannot open existing pack file '%s'"), pack_name
);
1653 if (open_pack_index(p
))
1654 die(_("Cannot open existing pack idx file for '%s'"), pack_name
);
1656 /* Read the attributes from the existing idx file */
1657 opts
->version
= p
->index_version
;
1659 if (opts
->version
== 2)
1660 read_v2_anomalous_offsets(p
, opts
);
1663 * Get rid of the idx file as we do not need it anymore.
1664 * NEEDSWORK: extract this bit from free_pack_by_name() in
1665 * object-file.c, perhaps? It shouldn't matter very much as we
1666 * know we haven't installed this pack (hence we never have
1667 * read anything from it).
1669 close_pack_index(p
);
1673 static void show_pack_info(int stat_only
)
1675 int i
, baseobjects
= nr_objects
- nr_ref_deltas
- nr_ofs_deltas
;
1676 unsigned long *chain_histogram
= NULL
;
1679 CALLOC_ARRAY(chain_histogram
, deepest_delta
);
1681 for (i
= 0; i
< nr_objects
; i
++) {
1682 struct object_entry
*obj
= &objects
[i
];
1684 if (is_delta_type(obj
->type
))
1685 chain_histogram
[obj_stat
[i
].delta_depth
- 1]++;
1688 printf("%s %-6s %"PRIuMAX
" %"PRIuMAX
" %"PRIuMAX
,
1689 oid_to_hex(&obj
->idx
.oid
),
1690 type_name(obj
->real_type
), (uintmax_t)obj
->size
,
1691 (uintmax_t)(obj
[1].idx
.offset
- obj
->idx
.offset
),
1692 (uintmax_t)obj
->idx
.offset
);
1693 if (is_delta_type(obj
->type
)) {
1694 struct object_entry
*bobj
= &objects
[obj_stat
[i
].base_object_no
];
1695 printf(" %u %s", obj_stat
[i
].delta_depth
,
1696 oid_to_hex(&bobj
->idx
.oid
));
1702 printf_ln(Q_("non delta: %d object",
1703 "non delta: %d objects",
1706 for (i
= 0; i
< deepest_delta
; i
++) {
1707 if (!chain_histogram
[i
])
1709 printf_ln(Q_("chain length = %d: %lu object",
1710 "chain length = %d: %lu objects",
1711 chain_histogram
[i
]),
1713 chain_histogram
[i
]);
1715 free(chain_histogram
);
1718 int cmd_index_pack(int argc
, const char **argv
, const char *prefix
)
1720 int i
, fix_thin_pack
= 0, verify
= 0, stat_only
= 0, rev_index
;
1721 const char *curr_index
;
1722 const char *curr_rev_index
= NULL
;
1723 const char *index_name
= NULL
, *pack_name
= NULL
, *rev_index_name
= NULL
;
1724 const char *keep_msg
= NULL
;
1725 const char *promisor_msg
= NULL
;
1726 struct strbuf index_name_buf
= STRBUF_INIT
;
1727 struct strbuf rev_index_name_buf
= STRBUF_INIT
;
1728 struct pack_idx_entry
**idx_objects
;
1729 struct pack_idx_option opts
;
1730 unsigned char pack_hash
[GIT_MAX_RAWSZ
];
1731 unsigned foreign_nr
= 1; /* zero is a "good" value, assume bad */
1732 int report_end_of_input
= 0;
1736 * index-pack never needs to fetch missing objects except when
1737 * REF_DELTA bases are missing (which are explicitly handled). It only
1738 * accesses the repo to do hash collision checks and to check which
1739 * REF_DELTA bases need to be fetched.
1741 fetch_if_missing
= 0;
1743 if (argc
== 2 && !strcmp(argv
[1], "-h"))
1744 usage(index_pack_usage
);
1746 read_replace_refs
= 0;
1747 fsck_options
.walk
= mark_link
;
1749 reset_pack_idx_option(&opts
);
1750 git_config(git_index_pack_config
, &opts
);
1751 if (prefix
&& chdir(prefix
))
1752 die(_("Cannot come back to cwd"));
1754 if (git_env_bool(GIT_TEST_WRITE_REV_INDEX
, 0))
1757 rev_index
= !!(opts
.flags
& (WRITE_REV_VERIFY
| WRITE_REV
));
1759 for (i
= 1; i
< argc
; i
++) {
1760 const char *arg
= argv
[i
];
1763 if (!strcmp(arg
, "--stdin")) {
1765 } else if (!strcmp(arg
, "--fix-thin")) {
1767 } else if (skip_to_optional_arg(arg
, "--strict", &arg
)) {
1770 fsck_set_msg_types(&fsck_options
, arg
);
1771 } else if (!strcmp(arg
, "--check-self-contained-and-connected")) {
1773 check_self_contained_and_connected
= 1;
1774 } else if (!strcmp(arg
, "--fsck-objects")) {
1776 } else if (!strcmp(arg
, "--verify")) {
1778 } else if (!strcmp(arg
, "--verify-stat")) {
1781 } else if (!strcmp(arg
, "--verify-stat-only")) {
1785 } else if (skip_to_optional_arg(arg
, "--keep", &keep_msg
)) {
1786 ; /* nothing to do */
1787 } else if (skip_to_optional_arg(arg
, "--promisor", &promisor_msg
)) {
1788 ; /* already parsed */
1789 } else if (starts_with(arg
, "--threads=")) {
1791 nr_threads
= strtoul(arg
+10, &end
, 0);
1792 if (!arg
[10] || *end
|| nr_threads
< 0)
1793 usage(index_pack_usage
);
1794 if (!HAVE_THREADS
&& nr_threads
!= 1) {
1795 warning(_("no threads support, ignoring %s"), arg
);
1798 } else if (starts_with(arg
, "--pack_header=")) {
1799 struct pack_header
*hdr
;
1802 hdr
= (struct pack_header
*)input_buffer
;
1803 hdr
->hdr_signature
= htonl(PACK_SIGNATURE
);
1804 hdr
->hdr_version
= htonl(strtoul(arg
+ 14, &c
, 10));
1806 die(_("bad %s"), arg
);
1807 hdr
->hdr_entries
= htonl(strtoul(c
+ 1, &c
, 10));
1809 die(_("bad %s"), arg
);
1810 input_len
= sizeof(*hdr
);
1811 } else if (!strcmp(arg
, "-v")) {
1813 } else if (!strcmp(arg
, "--progress-title")) {
1814 if (progress_title
|| (i
+1) >= argc
)
1815 usage(index_pack_usage
);
1816 progress_title
= argv
[++i
];
1817 } else if (!strcmp(arg
, "--show-resolving-progress")) {
1818 show_resolving_progress
= 1;
1819 } else if (!strcmp(arg
, "--report-end-of-input")) {
1820 report_end_of_input
= 1;
1821 } else if (!strcmp(arg
, "-o")) {
1822 if (index_name
|| (i
+1) >= argc
)
1823 usage(index_pack_usage
);
1824 index_name
= argv
[++i
];
1825 } else if (starts_with(arg
, "--index-version=")) {
1827 opts
.version
= strtoul(arg
+ 16, &c
, 10);
1828 if (opts
.version
> 2)
1829 die(_("bad %s"), arg
);
1831 opts
.off32_limit
= strtoul(c
+1, &c
, 0);
1832 if (*c
|| opts
.off32_limit
& 0x80000000)
1833 die(_("bad %s"), arg
);
1834 } else if (skip_prefix(arg
, "--max-input-size=", &arg
)) {
1835 max_input_size
= strtoumax(arg
, NULL
, 10);
1836 } else if (skip_prefix(arg
, "--object-format=", &arg
)) {
1837 hash_algo
= hash_algo_by_name(arg
);
1838 if (hash_algo
== GIT_HASH_UNKNOWN
)
1839 die(_("unknown hash algorithm '%s'"), arg
);
1840 repo_set_hash_algo(the_repository
, hash_algo
);
1841 } else if (!strcmp(arg
, "--rev-index")) {
1843 } else if (!strcmp(arg
, "--no-rev-index")) {
1846 usage(index_pack_usage
);
1851 usage(index_pack_usage
);
1855 if (!pack_name
&& !from_stdin
)
1856 usage(index_pack_usage
);
1857 if (fix_thin_pack
&& !from_stdin
)
1858 die(_("the option '%s' requires '%s'"), "--fix-thin", "--stdin");
1859 if (from_stdin
&& !startup_info
->have_repository
)
1860 die(_("--stdin requires a git repository"));
1861 if (from_stdin
&& hash_algo
)
1862 die(_("options '%s' and '%s' cannot be used together"), "--object-format", "--stdin");
1863 if (!index_name
&& pack_name
)
1864 index_name
= derive_filename(pack_name
, "pack", "idx", &index_name_buf
);
1866 opts
.flags
&= ~(WRITE_REV
| WRITE_REV_VERIFY
);
1868 opts
.flags
|= verify
? WRITE_REV_VERIFY
: WRITE_REV
;
1870 rev_index_name
= derive_filename(index_name
,
1872 &rev_index_name_buf
);
1877 die(_("--verify with no packfile name given"));
1878 read_idx_option(&opts
, index_name
);
1879 opts
.flags
|= WRITE_IDX_VERIFY
| WRITE_IDX_STRICT
;
1882 opts
.flags
|= WRITE_IDX_STRICT
;
1884 if (HAVE_THREADS
&& !nr_threads
) {
1885 nr_threads
= online_cpus();
1887 * Experiments show that going above 20 threads doesn't help,
1888 * no matter how many cores you have. Below that, we tend to
1889 * max at half the number of online_cpus(), presumably because
1890 * half of those are hyperthreads rather than full cores. We'll
1891 * never reduce the level below "3", though, to match a
1892 * historical value that nobody complained about.
1895 ; /* too few cores to consider capping */
1896 else if (nr_threads
< 6)
1897 nr_threads
= 3; /* historic cap */
1898 else if (nr_threads
< 40)
1901 nr_threads
= 20; /* hard cap */
1904 curr_pack
= open_pack_file(pack_name
);
1905 parse_pack_header();
1906 CALLOC_ARRAY(objects
, st_add(nr_objects
, 1));
1908 CALLOC_ARRAY(obj_stat
, st_add(nr_objects
, 1));
1909 CALLOC_ARRAY(ofs_deltas
, nr_objects
);
1910 parse_pack_objects(pack_hash
);
1911 if (report_end_of_input
)
1912 write_in_full(2, "\0", 1);
1914 conclude_pack(fix_thin_pack
, curr_pack
, pack_hash
);
1918 foreign_nr
= check_objects();
1921 show_pack_info(stat_only
);
1923 ALLOC_ARRAY(idx_objects
, nr_objects
);
1924 for (i
= 0; i
< nr_objects
; i
++)
1925 idx_objects
[i
] = &objects
[i
].idx
;
1926 curr_index
= write_idx_file(index_name
, idx_objects
, nr_objects
, &opts
, pack_hash
);
1928 curr_rev_index
= write_rev_file(rev_index_name
, idx_objects
,
1929 nr_objects
, pack_hash
,
1934 final(pack_name
, curr_pack
,
1935 index_name
, curr_index
,
1936 rev_index_name
, curr_rev_index
,
1937 keep_msg
, promisor_msg
,
1942 if (do_fsck_object
&& fsck_finish(&fsck_options
))
1943 die(_("fsck error in pack objects"));
1947 strbuf_release(&index_name_buf
);
1948 strbuf_release(&rev_index_name_buf
);
1950 free((void *) curr_pack
);
1952 free((void *) curr_index
);
1953 if (!rev_index_name
)
1954 free((void *) curr_rev_index
);
1957 * Let the caller know this pack is not self contained
1959 if (check_self_contained_and_connected
&& foreign_nr
)