1 #define USE_THE_REPOSITORY_VARIABLE
4 #include "environment.h"
17 #include "csum-file.h"
20 #include "run-command.h"
22 #include "object-file.h"
23 #include "object-name.h"
24 #include "object-store-ll.h"
26 #include "commit-reach.h"
30 #define PACK_ID_BITS 16
31 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
33 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
36 * We abuse the setuid bit on directories to mean "do not delta".
38 #define NO_DELTA S_ISUID
41 * The amount of additional space required in order to write an object into the
42 * current pack. This is the hash lengths at the end of the pack, plus the
43 * length of one object ID.
45 #define PACK_SIZE_THRESHOLD (the_hash_algo->rawsz * 3)
48 struct pack_idx_entry idx
;
49 struct hashmap_entry ent
;
50 uint32_t type
: TYPE_BITS
,
51 pack_id
: PACK_ID_BITS
,
55 static int object_entry_hashcmp(const void *map_data UNUSED
,
56 const struct hashmap_entry
*eptr
,
57 const struct hashmap_entry
*entry_or_key
,
60 const struct object_id
*oid
= keydata
;
61 const struct object_entry
*e1
, *e2
;
63 e1
= container_of(eptr
, const struct object_entry
, ent
);
65 return oidcmp(&e1
->idx
.oid
, oid
);
67 e2
= container_of(entry_or_key
, const struct object_entry
, ent
);
68 return oidcmp(&e1
->idx
.oid
, &e2
->idx
.oid
);
71 struct object_entry_pool
{
72 struct object_entry_pool
*next_pool
;
73 struct object_entry
*next_free
;
74 struct object_entry
*end
;
75 struct object_entry entries
[FLEX_ARRAY
]; /* more */
80 struct object_id
*oids
[1024];
81 struct object_entry
*marked
[1024];
82 struct mark_set
*sets
[1024];
95 struct atom_str
*next_atom
;
96 unsigned short str_len
;
97 char str_dat
[FLEX_ARRAY
]; /* more */
102 struct tree_content
*tree
;
103 struct atom_str
*name
;
104 struct tree_entry_ms
{
106 struct object_id oid
;
110 struct tree_content
{
111 unsigned int entry_capacity
; /* must match avail_tree_content */
112 unsigned int entry_count
;
113 unsigned int delta_depth
;
114 struct tree_entry
*entries
[FLEX_ARRAY
]; /* more */
117 struct avail_tree_content
{
118 unsigned int entry_capacity
; /* must match tree_content */
119 struct avail_tree_content
*next_avail
;
123 struct branch
*table_next_branch
;
124 struct branch
*active_next_branch
;
126 struct tree_entry branch_tree
;
127 uintmax_t last_commit
;
131 unsigned pack_id
: PACK_ID_BITS
;
132 struct object_id oid
;
136 struct tag
*next_tag
;
138 unsigned int pack_id
;
139 struct object_id oid
;
143 struct hash_list
*next
;
144 struct object_id oid
;
149 WHENSPEC_RAW_PERMISSIVE
,
154 struct recent_command
{
155 struct recent_command
*prev
;
156 struct recent_command
*next
;
160 typedef void (*mark_set_inserter_t
)(struct mark_set
**s
, struct object_id
*oid
, uintmax_t mark
);
161 typedef void (*each_mark_fn_t
)(uintmax_t mark
, void *obj
, void *cbp
);
163 /* Configured limits on output */
164 static unsigned long max_depth
= 50;
165 static off_t max_packsize
;
166 static int unpack_limit
= 100;
167 static int force_update
;
169 /* Stats and misc. counters */
170 static uintmax_t alloc_count
;
171 static uintmax_t marks_set_count
;
172 static uintmax_t object_count_by_type
[1 << TYPE_BITS
];
173 static uintmax_t duplicate_count_by_type
[1 << TYPE_BITS
];
174 static uintmax_t delta_count_by_type
[1 << TYPE_BITS
];
175 static uintmax_t delta_count_attempts_by_type
[1 << TYPE_BITS
];
176 static unsigned long object_count
;
177 static unsigned long branch_count
;
178 static unsigned long branch_load_count
;
180 static FILE *pack_edges
;
181 static unsigned int show_stats
= 1;
182 static int global_argc
;
183 static const char **global_argv
;
184 static const char *global_prefix
;
187 static struct mem_pool fi_mem_pool
= {
188 .block_alloc
= 2*1024*1024 - sizeof(struct mp_block
),
191 /* Atom management */
192 static unsigned int atom_table_sz
= 4451;
193 static unsigned int atom_cnt
;
194 static struct atom_str
**atom_table
;
196 /* The .pack file being generated */
197 static struct pack_idx_option pack_idx_opts
;
198 static unsigned int pack_id
;
199 static struct hashfile
*pack_file
;
200 static struct packed_git
*pack_data
;
201 static struct packed_git
**all_packs
;
202 static off_t pack_size
;
204 /* Table of objects we've written. */
205 static unsigned int object_entry_alloc
= 5000;
206 static struct object_entry_pool
*blocks
;
207 static struct hashmap object_table
;
208 static struct mark_set
*marks
;
209 static char *export_marks_file
;
210 static char *import_marks_file
;
211 static int import_marks_file_from_stream
;
212 static int import_marks_file_ignore_missing
;
213 static int import_marks_file_done
;
214 static int relative_marks_paths
;
217 static struct last_object last_blob
= {
221 /* Tree management */
222 static unsigned int tree_entry_alloc
= 1000;
223 static void *avail_tree_entry
;
224 static unsigned int avail_tree_table_sz
= 100;
225 static struct avail_tree_content
**avail_tree_table
;
226 static size_t tree_entry_allocd
;
227 static struct strbuf old_tree
= STRBUF_INIT
;
228 static struct strbuf new_tree
= STRBUF_INIT
;
231 static unsigned long max_active_branches
= 5;
232 static unsigned long cur_active_branches
;
233 static unsigned long branch_table_sz
= 1039;
234 static struct branch
**branch_table
;
235 static struct branch
*active_branches
;
238 static struct tag
*first_tag
;
239 static struct tag
*last_tag
;
241 /* Input stream parsing */
242 static whenspec_type whenspec
= WHENSPEC_RAW
;
243 static struct strbuf command_buf
= STRBUF_INIT
;
244 static int unread_command_buf
;
245 static struct recent_command cmd_hist
= {
249 static struct recent_command
*cmd_tail
= &cmd_hist
;
250 static struct recent_command
*rc_free
;
251 static unsigned int cmd_save
= 100;
252 static uintmax_t next_mark
;
253 static struct strbuf new_data
= STRBUF_INIT
;
254 static int seen_data_command
;
255 static int require_explicit_termination
;
256 static int allow_unsafe_features
;
258 /* Signal handling */
259 static volatile sig_atomic_t checkpoint_requested
;
261 /* Submodule marks */
262 static struct string_list sub_marks_from
= STRING_LIST_INIT_DUP
;
263 static struct string_list sub_marks_to
= STRING_LIST_INIT_DUP
;
264 static kh_oid_map_t
*sub_oid_map
;
266 /* Where to write output of cat-blob commands */
267 static int cat_blob_fd
= STDOUT_FILENO
;
269 static void parse_argv(void);
270 static void parse_get_mark(const char *p
);
271 static void parse_cat_blob(const char *p
);
272 static void parse_ls(const char *p
, struct branch
*b
);
274 static void for_each_mark(struct mark_set
*m
, uintmax_t base
, each_mark_fn_t callback
, void *p
)
278 for (k
= 0; k
< 1024; k
++) {
280 for_each_mark(m
->data
.sets
[k
], base
+ (k
<< m
->shift
), callback
, p
);
283 for (k
= 0; k
< 1024; k
++) {
284 if (m
->data
.marked
[k
])
285 callback(base
+ k
, m
->data
.marked
[k
], p
);
290 static void dump_marks_fn(uintmax_t mark
, void *object
, void *cbp
) {
291 struct object_entry
*e
= object
;
294 fprintf(f
, ":%" PRIuMAX
" %s\n", mark
, oid_to_hex(&e
->idx
.oid
));
297 static void write_branch_report(FILE *rpt
, struct branch
*b
)
299 fprintf(rpt
, "%s:\n", b
->name
);
301 fprintf(rpt
, " status :");
303 fputs(" active", rpt
);
304 if (b
->branch_tree
.tree
)
305 fputs(" loaded", rpt
);
306 if (is_null_oid(&b
->branch_tree
.versions
[1].oid
))
307 fputs(" dirty", rpt
);
310 fprintf(rpt
, " tip commit : %s\n", oid_to_hex(&b
->oid
));
311 fprintf(rpt
, " old tree : %s\n",
312 oid_to_hex(&b
->branch_tree
.versions
[0].oid
));
313 fprintf(rpt
, " cur tree : %s\n",
314 oid_to_hex(&b
->branch_tree
.versions
[1].oid
));
315 fprintf(rpt
, " commit clock: %" PRIuMAX
"\n", b
->last_commit
);
317 fputs(" last pack : ", rpt
);
318 if (b
->pack_id
< MAX_PACK_ID
)
319 fprintf(rpt
, "%u", b
->pack_id
);
325 static void write_crash_report(const char *err
)
327 char *loc
= git_pathdup("fast_import_crash_%"PRIuMAX
, (uintmax_t) getpid());
328 FILE *rpt
= fopen(loc
, "w");
331 struct recent_command
*rc
;
334 error_errno("can't write crash report %s", loc
);
339 fprintf(stderr
, "fast-import: dumping crash report to %s\n", loc
);
341 fprintf(rpt
, "fast-import crash report:\n");
342 fprintf(rpt
, " fast-import process: %"PRIuMAX
"\n", (uintmax_t) getpid());
343 fprintf(rpt
, " parent process : %"PRIuMAX
"\n", (uintmax_t) getppid());
344 fprintf(rpt
, " at %s\n", show_date(time(NULL
), 0, DATE_MODE(ISO8601
)));
347 fputs("fatal: ", rpt
);
352 fputs("Most Recent Commands Before Crash\n", rpt
);
353 fputs("---------------------------------\n", rpt
);
354 for (rc
= cmd_hist
.next
; rc
!= &cmd_hist
; rc
= rc
->next
) {
355 if (rc
->next
== &cmd_hist
)
364 fputs("Active Branch LRU\n", rpt
);
365 fputs("-----------------\n", rpt
);
366 fprintf(rpt
, " active_branches = %lu cur, %lu max\n",
368 max_active_branches
);
370 fputs(" pos clock name\n", rpt
);
371 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt
);
372 for (b
= active_branches
, lu
= 0; b
; b
= b
->active_next_branch
)
373 fprintf(rpt
, " %2lu) %6" PRIuMAX
" %s\n",
374 ++lu
, b
->last_commit
, b
->name
);
377 fputs("Inactive Branches\n", rpt
);
378 fputs("-----------------\n", rpt
);
379 for (lu
= 0; lu
< branch_table_sz
; lu
++) {
380 for (b
= branch_table
[lu
]; b
; b
= b
->table_next_branch
)
381 write_branch_report(rpt
, b
);
387 fputs("Annotated Tags\n", rpt
);
388 fputs("--------------\n", rpt
);
389 for (tg
= first_tag
; tg
; tg
= tg
->next_tag
) {
390 fputs(oid_to_hex(&tg
->oid
), rpt
);
392 fputs(tg
->name
, rpt
);
398 fputs("Marks\n", rpt
);
399 fputs("-----\n", rpt
);
400 if (export_marks_file
)
401 fprintf(rpt
, " exported to %s\n", export_marks_file
);
403 for_each_mark(marks
, 0, dump_marks_fn
, rpt
);
406 fputs("-------------------\n", rpt
);
407 fputs("END OF CRASH REPORT\n", rpt
);
412 static void end_packfile(void);
413 static void unkeep_all_packs(void);
414 static void dump_marks(void);
416 static NORETURN
void die_nicely(const char *err
, va_list params
)
420 report_fn die_message_fn
= get_die_message_routine();
423 die_message_fn(err
, params
);
426 char message
[2 * PATH_MAX
];
429 vsnprintf(message
, sizeof(message
), err
, cp
);
430 write_crash_report(message
);
438 #ifndef SIGUSR1 /* Windows, for example */
440 static void set_checkpoint_signal(void)
446 static void checkpoint_signal(int signo UNUSED
)
448 checkpoint_requested
= 1;
451 static void set_checkpoint_signal(void)
455 memset(&sa
, 0, sizeof(sa
));
456 sa
.sa_handler
= checkpoint_signal
;
457 sigemptyset(&sa
.sa_mask
);
458 sa
.sa_flags
= SA_RESTART
;
459 sigaction(SIGUSR1
, &sa
, NULL
);
464 static void alloc_objects(unsigned int cnt
)
466 struct object_entry_pool
*b
;
468 b
= xmalloc(sizeof(struct object_entry_pool
)
469 + cnt
* sizeof(struct object_entry
));
470 b
->next_pool
= blocks
;
471 b
->next_free
= b
->entries
;
472 b
->end
= b
->entries
+ cnt
;
477 static struct object_entry
*new_object(struct object_id
*oid
)
479 struct object_entry
*e
;
481 if (blocks
->next_free
== blocks
->end
)
482 alloc_objects(object_entry_alloc
);
484 e
= blocks
->next_free
++;
485 oidcpy(&e
->idx
.oid
, oid
);
489 static struct object_entry
*find_object(struct object_id
*oid
)
491 return hashmap_get_entry_from_hash(&object_table
, oidhash(oid
), oid
,
492 struct object_entry
, ent
);
495 static struct object_entry
*insert_object(struct object_id
*oid
)
497 struct object_entry
*e
;
498 unsigned int hash
= oidhash(oid
);
500 e
= hashmap_get_entry_from_hash(&object_table
, hash
, oid
,
501 struct object_entry
, ent
);
505 hashmap_entry_init(&e
->ent
, hash
);
506 hashmap_add(&object_table
, &e
->ent
);
512 static void invalidate_pack_id(unsigned int id
)
516 struct hashmap_iter iter
;
517 struct object_entry
*e
;
519 hashmap_for_each_entry(&object_table
, &iter
, e
, ent
) {
520 if (e
->pack_id
== id
)
521 e
->pack_id
= MAX_PACK_ID
;
524 for (lu
= 0; lu
< branch_table_sz
; lu
++) {
527 for (b
= branch_table
[lu
]; b
; b
= b
->table_next_branch
)
528 if (b
->pack_id
== id
)
529 b
->pack_id
= MAX_PACK_ID
;
532 for (t
= first_tag
; t
; t
= t
->next_tag
)
533 if (t
->pack_id
== id
)
534 t
->pack_id
= MAX_PACK_ID
;
537 static unsigned int hc_str(const char *s
, size_t len
)
545 static void insert_mark(struct mark_set
**top
, uintmax_t idnum
, struct object_entry
*oe
)
547 struct mark_set
*s
= *top
;
549 while ((idnum
>> s
->shift
) >= 1024) {
550 s
= mem_pool_calloc(&fi_mem_pool
, 1, sizeof(struct mark_set
));
551 s
->shift
= (*top
)->shift
+ 10;
552 s
->data
.sets
[0] = *top
;
556 uintmax_t i
= idnum
>> s
->shift
;
557 idnum
-= i
<< s
->shift
;
558 if (!s
->data
.sets
[i
]) {
559 s
->data
.sets
[i
] = mem_pool_calloc(&fi_mem_pool
, 1, sizeof(struct mark_set
));
560 s
->data
.sets
[i
]->shift
= s
->shift
- 10;
564 if (!s
->data
.marked
[idnum
])
566 s
->data
.marked
[idnum
] = oe
;
569 static void *find_mark(struct mark_set
*s
, uintmax_t idnum
)
571 uintmax_t orig_idnum
= idnum
;
572 struct object_entry
*oe
= NULL
;
573 if ((idnum
>> s
->shift
) < 1024) {
574 while (s
&& s
->shift
) {
575 uintmax_t i
= idnum
>> s
->shift
;
576 idnum
-= i
<< s
->shift
;
580 oe
= s
->data
.marked
[idnum
];
583 die("mark :%" PRIuMAX
" not declared", orig_idnum
);
587 static struct atom_str
*to_atom(const char *s
, unsigned short len
)
589 unsigned int hc
= hc_str(s
, len
) % atom_table_sz
;
592 for (c
= atom_table
[hc
]; c
; c
= c
->next_atom
)
593 if (c
->str_len
== len
&& !strncmp(s
, c
->str_dat
, len
))
596 c
= mem_pool_alloc(&fi_mem_pool
, sizeof(struct atom_str
) + len
+ 1);
598 memcpy(c
->str_dat
, s
, len
);
600 c
->next_atom
= atom_table
[hc
];
606 static struct branch
*lookup_branch(const char *name
)
608 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
611 for (b
= branch_table
[hc
]; b
; b
= b
->table_next_branch
)
612 if (!strcmp(name
, b
->name
))
617 static struct branch
*new_branch(const char *name
)
619 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
620 struct branch
*b
= lookup_branch(name
);
623 die("Invalid attempt to create duplicate branch: %s", name
);
624 if (check_refname_format(name
, REFNAME_ALLOW_ONELEVEL
))
625 die("Branch name doesn't conform to GIT standards: %s", name
);
627 b
= mem_pool_calloc(&fi_mem_pool
, 1, sizeof(struct branch
));
628 b
->name
= mem_pool_strdup(&fi_mem_pool
, name
);
629 b
->table_next_branch
= branch_table
[hc
];
630 b
->branch_tree
.versions
[0].mode
= S_IFDIR
;
631 b
->branch_tree
.versions
[1].mode
= S_IFDIR
;
634 b
->pack_id
= MAX_PACK_ID
;
635 branch_table
[hc
] = b
;
640 static unsigned int hc_entries(unsigned int cnt
)
642 cnt
= cnt
& 7 ? (cnt
/ 8) + 1 : cnt
/ 8;
643 return cnt
< avail_tree_table_sz
? cnt
: avail_tree_table_sz
- 1;
646 static struct tree_content
*new_tree_content(unsigned int cnt
)
648 struct avail_tree_content
*f
, *l
= NULL
;
649 struct tree_content
*t
;
650 unsigned int hc
= hc_entries(cnt
);
652 for (f
= avail_tree_table
[hc
]; f
; l
= f
, f
= f
->next_avail
)
653 if (f
->entry_capacity
>= cnt
)
658 l
->next_avail
= f
->next_avail
;
660 avail_tree_table
[hc
] = f
->next_avail
;
662 cnt
= cnt
& 7 ? ((cnt
/ 8) + 1) * 8 : cnt
;
663 f
= mem_pool_alloc(&fi_mem_pool
, sizeof(*t
) + sizeof(t
->entries
[0]) * cnt
);
664 f
->entry_capacity
= cnt
;
667 t
= (struct tree_content
*)f
;
673 static void release_tree_entry(struct tree_entry
*e
);
674 static void release_tree_content(struct tree_content
*t
)
676 struct avail_tree_content
*f
= (struct avail_tree_content
*)t
;
677 unsigned int hc
= hc_entries(f
->entry_capacity
);
678 f
->next_avail
= avail_tree_table
[hc
];
679 avail_tree_table
[hc
] = f
;
682 static void release_tree_content_recursive(struct tree_content
*t
)
685 for (i
= 0; i
< t
->entry_count
; i
++)
686 release_tree_entry(t
->entries
[i
]);
687 release_tree_content(t
);
690 static struct tree_content
*grow_tree_content(
691 struct tree_content
*t
,
694 struct tree_content
*r
= new_tree_content(t
->entry_count
+ amt
);
695 r
->entry_count
= t
->entry_count
;
696 r
->delta_depth
= t
->delta_depth
;
697 COPY_ARRAY(r
->entries
, t
->entries
, t
->entry_count
);
698 release_tree_content(t
);
702 static struct tree_entry
*new_tree_entry(void)
704 struct tree_entry
*e
;
706 if (!avail_tree_entry
) {
707 unsigned int n
= tree_entry_alloc
;
708 tree_entry_allocd
+= n
* sizeof(struct tree_entry
);
710 avail_tree_entry
= e
;
712 *((void**)e
) = e
+ 1;
718 e
= avail_tree_entry
;
719 avail_tree_entry
= *((void**)e
);
723 static void release_tree_entry(struct tree_entry
*e
)
726 release_tree_content_recursive(e
->tree
);
727 *((void**)e
) = avail_tree_entry
;
728 avail_tree_entry
= e
;
731 static struct tree_content
*dup_tree_content(struct tree_content
*s
)
733 struct tree_content
*d
;
734 struct tree_entry
*a
, *b
;
739 d
= new_tree_content(s
->entry_count
);
740 for (i
= 0; i
< s
->entry_count
; i
++) {
742 b
= new_tree_entry();
743 memcpy(b
, a
, sizeof(*a
));
744 if (a
->tree
&& is_null_oid(&b
->versions
[1].oid
))
745 b
->tree
= dup_tree_content(a
->tree
);
750 d
->entry_count
= s
->entry_count
;
751 d
->delta_depth
= s
->delta_depth
;
756 static void start_packfile(void)
758 struct strbuf tmp_file
= STRBUF_INIT
;
759 struct packed_git
*p
;
762 pack_fd
= odb_mkstemp(&tmp_file
, "pack/tmp_pack_XXXXXX");
763 FLEX_ALLOC_STR(p
, pack_name
, tmp_file
.buf
);
764 strbuf_release(&tmp_file
);
766 p
->pack_fd
= pack_fd
;
768 pack_file
= hashfd(pack_fd
, p
->pack_name
);
771 pack_size
= write_pack_header(pack_file
, 0);
774 REALLOC_ARRAY(all_packs
, pack_id
+ 1);
775 all_packs
[pack_id
] = p
;
778 static const char *create_index(void)
781 struct pack_idx_entry
**idx
, **c
, **last
;
782 struct object_entry
*e
;
783 struct object_entry_pool
*o
;
785 /* Build the table of object IDs. */
786 ALLOC_ARRAY(idx
, object_count
);
788 for (o
= blocks
; o
; o
= o
->next_pool
)
789 for (e
= o
->next_free
; e
-- != o
->entries
;)
790 if (pack_id
== e
->pack_id
)
792 last
= idx
+ object_count
;
794 die("internal consistency error creating the index");
796 tmpfile
= write_idx_file(NULL
, idx
, object_count
, &pack_idx_opts
,
802 static char *keep_pack(const char *curr_index_name
)
804 static const char *keep_msg
= "fast-import";
805 struct strbuf name
= STRBUF_INIT
;
808 odb_pack_name(&name
, pack_data
->hash
, "keep");
809 keep_fd
= odb_pack_keep(name
.buf
);
811 die_errno("cannot create keep file");
812 write_or_die(keep_fd
, keep_msg
, strlen(keep_msg
));
814 die_errno("failed to write keep file");
816 odb_pack_name(&name
, pack_data
->hash
, "pack");
817 if (finalize_object_file(pack_data
->pack_name
, name
.buf
))
818 die("cannot store pack file");
820 odb_pack_name(&name
, pack_data
->hash
, "idx");
821 if (finalize_object_file(curr_index_name
, name
.buf
))
822 die("cannot store index file");
823 free((void *)curr_index_name
);
824 return strbuf_detach(&name
, NULL
);
827 static void unkeep_all_packs(void)
829 struct strbuf name
= STRBUF_INIT
;
832 for (k
= 0; k
< pack_id
; k
++) {
833 struct packed_git
*p
= all_packs
[k
];
834 odb_pack_name(&name
, p
->hash
, "keep");
835 unlink_or_warn(name
.buf
);
837 strbuf_release(&name
);
840 static int loosen_small_pack(const struct packed_git
*p
)
842 struct child_process unpack
= CHILD_PROCESS_INIT
;
844 if (lseek(p
->pack_fd
, 0, SEEK_SET
) < 0)
845 die_errno("Failed seeking to start of '%s'", p
->pack_name
);
847 unpack
.in
= p
->pack_fd
;
849 unpack
.stdout_to_stderr
= 1;
850 strvec_push(&unpack
.args
, "unpack-objects");
852 strvec_push(&unpack
.args
, "-q");
854 return run_command(&unpack
);
857 static void end_packfile(void)
861 if (running
|| !pack_data
)
865 clear_delta_base_cache();
867 struct packed_git
*new_p
;
868 struct object_id cur_pack_oid
;
874 close_pack_windows(pack_data
);
875 finalize_hashfile(pack_file
, cur_pack_oid
.hash
, FSYNC_COMPONENT_PACK
, 0);
876 fixup_pack_header_footer(pack_data
->pack_fd
, pack_data
->hash
,
877 pack_data
->pack_name
, object_count
,
878 cur_pack_oid
.hash
, pack_size
);
880 if (object_count
<= unpack_limit
) {
881 if (!loosen_small_pack(pack_data
)) {
882 invalidate_pack_id(pack_id
);
887 close(pack_data
->pack_fd
);
888 idx_name
= keep_pack(create_index());
890 /* Register the packfile with core git's machinery. */
891 new_p
= add_packed_git(idx_name
, strlen(idx_name
), 1);
893 die("core git rejected index %s", idx_name
);
894 all_packs
[pack_id
] = new_p
;
895 install_packed_git(the_repository
, new_p
);
898 /* Print the boundary */
900 fprintf(pack_edges
, "%s:", new_p
->pack_name
);
901 for (i
= 0; i
< branch_table_sz
; i
++) {
902 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
) {
903 if (b
->pack_id
== pack_id
)
904 fprintf(pack_edges
, " %s",
905 oid_to_hex(&b
->oid
));
908 for (t
= first_tag
; t
; t
= t
->next_tag
) {
909 if (t
->pack_id
== pack_id
)
910 fprintf(pack_edges
, " %s",
911 oid_to_hex(&t
->oid
));
913 fputc('\n', pack_edges
);
921 close(pack_data
->pack_fd
);
922 unlink_or_warn(pack_data
->pack_name
);
924 FREE_AND_NULL(pack_data
);
927 /* We can't carry a delta across packfiles. */
928 strbuf_release(&last_blob
.data
);
929 last_blob
.offset
= 0;
933 static void cycle_packfile(void)
939 static int store_object(
940 enum object_type type
,
942 struct last_object
*last
,
943 struct object_id
*oidout
,
947 struct object_entry
*e
;
948 unsigned char hdr
[96];
949 struct object_id oid
;
950 unsigned long hdrlen
, deltalen
;
954 hdrlen
= format_object_header((char *)hdr
, sizeof(hdr
), type
,
956 the_hash_algo
->init_fn(&c
);
957 the_hash_algo
->update_fn(&c
, hdr
, hdrlen
);
958 the_hash_algo
->update_fn(&c
, dat
->buf
, dat
->len
);
959 the_hash_algo
->final_oid_fn(&oid
, &c
);
961 oidcpy(oidout
, &oid
);
963 e
= insert_object(&oid
);
965 insert_mark(&marks
, mark
, e
);
967 duplicate_count_by_type
[type
]++;
969 } else if (find_oid_pack(&oid
, get_all_packs(the_repository
))) {
971 e
->pack_id
= MAX_PACK_ID
;
972 e
->idx
.offset
= 1; /* just not zero! */
973 duplicate_count_by_type
[type
]++;
977 if (last
&& last
->data
.len
&& last
->data
.buf
&& last
->depth
< max_depth
978 && dat
->len
> the_hash_algo
->rawsz
) {
980 delta_count_attempts_by_type
[type
]++;
981 delta
= diff_delta(last
->data
.buf
, last
->data
.len
,
983 &deltalen
, dat
->len
- the_hash_algo
->rawsz
);
987 git_deflate_init(&s
, pack_compression_level
);
990 s
.avail_in
= deltalen
;
992 s
.next_in
= (void *)dat
->buf
;
993 s
.avail_in
= dat
->len
;
995 s
.avail_out
= git_deflate_bound(&s
, s
.avail_in
);
996 s
.next_out
= out
= xmalloc(s
.avail_out
);
997 while (git_deflate(&s
, Z_FINISH
) == Z_OK
)
1001 /* Determine if we should auto-checkpoint. */
1003 && (pack_size
+ PACK_SIZE_THRESHOLD
+ s
.total_out
) > max_packsize
)
1004 || (pack_size
+ PACK_SIZE_THRESHOLD
+ s
.total_out
) < pack_size
) {
1006 /* This new object needs to *not* have the current pack_id. */
1007 e
->pack_id
= pack_id
+ 1;
1010 /* We cannot carry a delta into the new pack. */
1012 FREE_AND_NULL(delta
);
1014 git_deflate_init(&s
, pack_compression_level
);
1015 s
.next_in
= (void *)dat
->buf
;
1016 s
.avail_in
= dat
->len
;
1017 s
.avail_out
= git_deflate_bound(&s
, s
.avail_in
);
1018 s
.next_out
= out
= xrealloc(out
, s
.avail_out
);
1019 while (git_deflate(&s
, Z_FINISH
) == Z_OK
)
1021 git_deflate_end(&s
);
1026 e
->pack_id
= pack_id
;
1027 e
->idx
.offset
= pack_size
;
1029 object_count_by_type
[type
]++;
1031 crc32_begin(pack_file
);
1034 off_t ofs
= e
->idx
.offset
- last
->offset
;
1035 unsigned pos
= sizeof(hdr
) - 1;
1037 delta_count_by_type
[type
]++;
1038 e
->depth
= last
->depth
+ 1;
1040 hdrlen
= encode_in_pack_object_header(hdr
, sizeof(hdr
),
1041 OBJ_OFS_DELTA
, deltalen
);
1042 hashwrite(pack_file
, hdr
, hdrlen
);
1043 pack_size
+= hdrlen
;
1045 hdr
[pos
] = ofs
& 127;
1047 hdr
[--pos
] = 128 | (--ofs
& 127);
1048 hashwrite(pack_file
, hdr
+ pos
, sizeof(hdr
) - pos
);
1049 pack_size
+= sizeof(hdr
) - pos
;
1052 hdrlen
= encode_in_pack_object_header(hdr
, sizeof(hdr
),
1054 hashwrite(pack_file
, hdr
, hdrlen
);
1055 pack_size
+= hdrlen
;
1058 hashwrite(pack_file
, out
, s
.total_out
);
1059 pack_size
+= s
.total_out
;
1061 e
->idx
.crc32
= crc32_end(pack_file
);
1066 if (last
->no_swap
) {
1069 strbuf_swap(&last
->data
, dat
);
1071 last
->offset
= e
->idx
.offset
;
1072 last
->depth
= e
->depth
;
1077 static void truncate_pack(struct hashfile_checkpoint
*checkpoint
)
1079 if (hashfile_truncate(pack_file
, checkpoint
))
1080 die_errno("cannot truncate pack to skip duplicate");
1081 pack_size
= checkpoint
->offset
;
1084 static void stream_blob(uintmax_t len
, struct object_id
*oidout
, uintmax_t mark
)
1086 size_t in_sz
= 64 * 1024, out_sz
= 64 * 1024;
1087 unsigned char *in_buf
= xmalloc(in_sz
);
1088 unsigned char *out_buf
= xmalloc(out_sz
);
1089 struct object_entry
*e
;
1090 struct object_id oid
;
1091 unsigned long hdrlen
;
1095 struct hashfile_checkpoint checkpoint
;
1098 /* Determine if we should auto-checkpoint. */
1100 && (pack_size
+ PACK_SIZE_THRESHOLD
+ len
) > max_packsize
)
1101 || (pack_size
+ PACK_SIZE_THRESHOLD
+ len
) < pack_size
)
1104 the_hash_algo
->init_fn(&checkpoint
.ctx
);
1105 hashfile_checkpoint(pack_file
, &checkpoint
);
1106 offset
= checkpoint
.offset
;
1108 hdrlen
= format_object_header((char *)out_buf
, out_sz
, OBJ_BLOB
, len
);
1110 the_hash_algo
->init_fn(&c
);
1111 the_hash_algo
->update_fn(&c
, out_buf
, hdrlen
);
1113 crc32_begin(pack_file
);
1115 git_deflate_init(&s
, pack_compression_level
);
1117 hdrlen
= encode_in_pack_object_header(out_buf
, out_sz
, OBJ_BLOB
, len
);
1119 s
.next_out
= out_buf
+ hdrlen
;
1120 s
.avail_out
= out_sz
- hdrlen
;
1122 while (status
!= Z_STREAM_END
) {
1123 if (0 < len
&& !s
.avail_in
) {
1124 size_t cnt
= in_sz
< len
? in_sz
: (size_t)len
;
1125 size_t n
= fread(in_buf
, 1, cnt
, stdin
);
1126 if (!n
&& feof(stdin
))
1127 die("EOF in data (%" PRIuMAX
" bytes remaining)", len
);
1129 the_hash_algo
->update_fn(&c
, in_buf
, n
);
1135 status
= git_deflate(&s
, len
? 0 : Z_FINISH
);
1137 if (!s
.avail_out
|| status
== Z_STREAM_END
) {
1138 size_t n
= s
.next_out
- out_buf
;
1139 hashwrite(pack_file
, out_buf
, n
);
1141 s
.next_out
= out_buf
;
1142 s
.avail_out
= out_sz
;
1151 die("unexpected deflate failure: %d", status
);
1154 git_deflate_end(&s
);
1155 the_hash_algo
->final_oid_fn(&oid
, &c
);
1158 oidcpy(oidout
, &oid
);
1160 e
= insert_object(&oid
);
1163 insert_mark(&marks
, mark
, e
);
1165 if (e
->idx
.offset
) {
1166 duplicate_count_by_type
[OBJ_BLOB
]++;
1167 truncate_pack(&checkpoint
);
1169 } else if (find_oid_pack(&oid
, get_all_packs(the_repository
))) {
1171 e
->pack_id
= MAX_PACK_ID
;
1172 e
->idx
.offset
= 1; /* just not zero! */
1173 duplicate_count_by_type
[OBJ_BLOB
]++;
1174 truncate_pack(&checkpoint
);
1179 e
->pack_id
= pack_id
;
1180 e
->idx
.offset
= offset
;
1181 e
->idx
.crc32
= crc32_end(pack_file
);
1183 object_count_by_type
[OBJ_BLOB
]++;
1190 /* All calls must be guarded by find_object() or find_mark() to
1191 * ensure the 'struct object_entry' passed was written by this
1192 * process instance. We unpack the entry by the offset, avoiding
1193 * the need for the corresponding .idx file. This unpacking rule
1194 * works because we only use OBJ_REF_DELTA within the packfiles
1195 * created by fast-import.
1197 * oe must not be NULL. Such an oe usually comes from giving
1198 * an unknown SHA-1 to find_object() or an undefined mark to
1199 * find_mark(). Callers must test for this condition and use
1200 * the standard read_sha1_file() when it happens.
1202 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1203 * find_mark(), where the mark was reloaded from an existing marks
1204 * file and is referencing an object that this fast-import process
1205 * instance did not write out to a packfile. Callers must test for
1206 * this condition and use read_sha1_file() instead.
1208 static void *gfi_unpack_entry(
1209 struct object_entry
*oe
,
1210 unsigned long *sizep
)
1212 enum object_type type
;
1213 struct packed_git
*p
= all_packs
[oe
->pack_id
];
1214 if (p
== pack_data
&& p
->pack_size
< (pack_size
+ the_hash_algo
->rawsz
)) {
1215 /* The object is stored in the packfile we are writing to
1216 * and we have modified it since the last time we scanned
1217 * back to read a previously written object. If an old
1218 * window covered [p->pack_size, p->pack_size + rawsz) its
1219 * data is stale and is not valid. Closing all windows
1220 * and updating the packfile length ensures we can read
1221 * the newly written data.
1223 close_pack_windows(p
);
1224 hashflush(pack_file
);
1226 /* We have to offer rawsz bytes additional on the end of
1227 * the packfile as the core unpacker code assumes the
1228 * footer is present at the file end and must promise
1229 * at least rawsz bytes within any window it maps. But
1230 * we don't actually create the footer here.
1232 p
->pack_size
= pack_size
+ the_hash_algo
->rawsz
;
1234 return unpack_entry(the_repository
, p
, oe
->idx
.offset
, &type
, sizep
);
1237 static void load_tree(struct tree_entry
*root
)
1239 struct object_id
*oid
= &root
->versions
[1].oid
;
1240 struct object_entry
*myoe
;
1241 struct tree_content
*t
;
1246 root
->tree
= t
= new_tree_content(8);
1247 if (is_null_oid(oid
))
1250 myoe
= find_object(oid
);
1251 if (myoe
&& myoe
->pack_id
!= MAX_PACK_ID
) {
1252 if (myoe
->type
!= OBJ_TREE
)
1253 die("Not a tree: %s", oid_to_hex(oid
));
1254 t
->delta_depth
= myoe
->depth
;
1255 buf
= gfi_unpack_entry(myoe
, &size
);
1257 die("Can't load tree %s", oid_to_hex(oid
));
1259 enum object_type type
;
1260 buf
= repo_read_object_file(the_repository
, oid
, &type
, &size
);
1261 if (!buf
|| type
!= OBJ_TREE
)
1262 die("Can't load tree %s", oid_to_hex(oid
));
1266 while (c
!= (buf
+ size
)) {
1267 struct tree_entry
*e
= new_tree_entry();
1269 if (t
->entry_count
== t
->entry_capacity
)
1270 root
->tree
= t
= grow_tree_content(t
, t
->entry_count
);
1271 t
->entries
[t
->entry_count
++] = e
;
1274 c
= parse_mode(c
, &e
->versions
[1].mode
);
1276 die("Corrupt mode in %s", oid_to_hex(oid
));
1277 e
->versions
[0].mode
= e
->versions
[1].mode
;
1278 e
->name
= to_atom(c
, strlen(c
));
1279 c
+= e
->name
->str_len
+ 1;
1280 oidread(&e
->versions
[0].oid
, (unsigned char *)c
,
1281 the_repository
->hash_algo
);
1282 oidread(&e
->versions
[1].oid
, (unsigned char *)c
,
1283 the_repository
->hash_algo
);
1284 c
+= the_hash_algo
->rawsz
;
1289 static int tecmp0 (const void *_a
, const void *_b
)
1291 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
1292 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
1293 return base_name_compare(
1294 a
->name
->str_dat
, a
->name
->str_len
, a
->versions
[0].mode
,
1295 b
->name
->str_dat
, b
->name
->str_len
, b
->versions
[0].mode
);
1298 static int tecmp1 (const void *_a
, const void *_b
)
1300 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
1301 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
1302 return base_name_compare(
1303 a
->name
->str_dat
, a
->name
->str_len
, a
->versions
[1].mode
,
1304 b
->name
->str_dat
, b
->name
->str_len
, b
->versions
[1].mode
);
1307 static void mktree(struct tree_content
*t
, int v
, struct strbuf
*b
)
1313 QSORT(t
->entries
, t
->entry_count
, tecmp0
);
1315 QSORT(t
->entries
, t
->entry_count
, tecmp1
);
1317 for (i
= 0; i
< t
->entry_count
; i
++) {
1318 if (t
->entries
[i
]->versions
[v
].mode
)
1319 maxlen
+= t
->entries
[i
]->name
->str_len
+ 34;
1323 strbuf_grow(b
, maxlen
);
1324 for (i
= 0; i
< t
->entry_count
; i
++) {
1325 struct tree_entry
*e
= t
->entries
[i
];
1326 if (!e
->versions
[v
].mode
)
1328 strbuf_addf(b
, "%o %s%c",
1329 (unsigned int)(e
->versions
[v
].mode
& ~NO_DELTA
),
1330 e
->name
->str_dat
, '\0');
1331 strbuf_add(b
, e
->versions
[v
].oid
.hash
, the_hash_algo
->rawsz
);
1335 static void store_tree(struct tree_entry
*root
)
1337 struct tree_content
*t
;
1338 unsigned int i
, j
, del
;
1339 struct last_object lo
= { STRBUF_INIT
, 0, 0, /* no_swap */ 1 };
1340 struct object_entry
*le
= NULL
;
1342 if (!is_null_oid(&root
->versions
[1].oid
))
1349 for (i
= 0; i
< t
->entry_count
; i
++) {
1350 if (t
->entries
[i
]->tree
)
1351 store_tree(t
->entries
[i
]);
1354 if (!(root
->versions
[0].mode
& NO_DELTA
))
1355 le
= find_object(&root
->versions
[0].oid
);
1356 if (S_ISDIR(root
->versions
[0].mode
) && le
&& le
->pack_id
== pack_id
) {
1357 mktree(t
, 0, &old_tree
);
1359 lo
.offset
= le
->idx
.offset
;
1360 lo
.depth
= t
->delta_depth
;
1363 mktree(t
, 1, &new_tree
);
1364 store_object(OBJ_TREE
, &new_tree
, &lo
, &root
->versions
[1].oid
, 0);
1366 t
->delta_depth
= lo
.depth
;
1367 for (i
= 0, j
= 0, del
= 0; i
< t
->entry_count
; i
++) {
1368 struct tree_entry
*e
= t
->entries
[i
];
1369 if (e
->versions
[1].mode
) {
1370 e
->versions
[0].mode
= e
->versions
[1].mode
;
1371 oidcpy(&e
->versions
[0].oid
, &e
->versions
[1].oid
);
1372 t
->entries
[j
++] = e
;
1374 release_tree_entry(e
);
1378 t
->entry_count
-= del
;
1381 static void tree_content_replace(
1382 struct tree_entry
*root
,
1383 const struct object_id
*oid
,
1384 const uint16_t mode
,
1385 struct tree_content
*newtree
)
1388 die("Root cannot be a non-directory");
1389 oidclr(&root
->versions
[0].oid
, the_repository
->hash_algo
);
1390 oidcpy(&root
->versions
[1].oid
, oid
);
1392 release_tree_content_recursive(root
->tree
);
1393 root
->tree
= newtree
;
1396 static int tree_content_set(
1397 struct tree_entry
*root
,
1399 const struct object_id
*oid
,
1400 const uint16_t mode
,
1401 struct tree_content
*subtree
)
1403 struct tree_content
*t
;
1406 struct tree_entry
*e
;
1408 slash1
= strchrnul(p
, '/');
1411 die("Empty path component found in input");
1412 if (!*slash1
&& !S_ISDIR(mode
) && subtree
)
1413 die("Non-directories cannot have subtrees");
1418 for (i
= 0; i
< t
->entry_count
; i
++) {
1420 if (e
->name
->str_len
== n
&& !fspathncmp(p
, e
->name
->str_dat
, n
)) {
1423 && e
->versions
[1].mode
== mode
1424 && oideq(&e
->versions
[1].oid
, oid
))
1426 e
->versions
[1].mode
= mode
;
1427 oidcpy(&e
->versions
[1].oid
, oid
);
1429 release_tree_content_recursive(e
->tree
);
1433 * We need to leave e->versions[0].sha1 alone
1434 * to avoid modifying the preimage tree used
1435 * when writing out the parent directory.
1436 * But after replacing the subdir with a
1437 * completely different one, it's not a good
1438 * delta base any more, and besides, we've
1439 * thrown away the tree entries needed to
1440 * make a delta against it.
1442 * So let's just explicitly disable deltas
1445 if (S_ISDIR(e
->versions
[0].mode
))
1446 e
->versions
[0].mode
|= NO_DELTA
;
1448 oidclr(&root
->versions
[1].oid
, the_repository
->hash_algo
);
1451 if (!S_ISDIR(e
->versions
[1].mode
)) {
1452 e
->tree
= new_tree_content(8);
1453 e
->versions
[1].mode
= S_IFDIR
;
1457 if (tree_content_set(e
, slash1
+ 1, oid
, mode
, subtree
)) {
1458 oidclr(&root
->versions
[1].oid
, the_repository
->hash_algo
);
1465 if (t
->entry_count
== t
->entry_capacity
)
1466 root
->tree
= t
= grow_tree_content(t
, t
->entry_count
);
1467 e
= new_tree_entry();
1468 e
->name
= to_atom(p
, n
);
1469 e
->versions
[0].mode
= 0;
1470 oidclr(&e
->versions
[0].oid
, the_repository
->hash_algo
);
1471 t
->entries
[t
->entry_count
++] = e
;
1473 e
->tree
= new_tree_content(8);
1474 e
->versions
[1].mode
= S_IFDIR
;
1475 tree_content_set(e
, slash1
+ 1, oid
, mode
, subtree
);
1478 e
->versions
[1].mode
= mode
;
1479 oidcpy(&e
->versions
[1].oid
, oid
);
1481 oidclr(&root
->versions
[1].oid
, the_repository
->hash_algo
);
1485 static int tree_content_remove(
1486 struct tree_entry
*root
,
1488 struct tree_entry
*backup_leaf
,
1491 struct tree_content
*t
;
1494 struct tree_entry
*e
;
1496 slash1
= strchrnul(p
, '/');
1502 if (!*p
&& allow_root
) {
1508 for (i
= 0; i
< t
->entry_count
; i
++) {
1510 if (e
->name
->str_len
== n
&& !fspathncmp(p
, e
->name
->str_dat
, n
)) {
1511 if (*slash1
&& !S_ISDIR(e
->versions
[1].mode
))
1513 * If p names a file in some subdirectory, and a
1514 * file or symlink matching the name of the
1515 * parent directory of p exists, then p cannot
1516 * exist and need not be deleted.
1519 if (!*slash1
|| !S_ISDIR(e
->versions
[1].mode
))
1523 if (tree_content_remove(e
, slash1
+ 1, backup_leaf
, 0)) {
1524 for (n
= 0; n
< e
->tree
->entry_count
; n
++) {
1525 if (e
->tree
->entries
[n
]->versions
[1].mode
) {
1526 oidclr(&root
->versions
[1].oid
,
1527 the_repository
->hash_algo
);
1541 memcpy(backup_leaf
, e
, sizeof(*backup_leaf
));
1543 release_tree_content_recursive(e
->tree
);
1545 e
->versions
[1].mode
= 0;
1546 oidclr(&e
->versions
[1].oid
, the_repository
->hash_algo
);
1547 oidclr(&root
->versions
[1].oid
, the_repository
->hash_algo
);
1551 static int tree_content_get(
1552 struct tree_entry
*root
,
1554 struct tree_entry
*leaf
,
1557 struct tree_content
*t
;
1560 struct tree_entry
*e
;
1562 slash1
= strchrnul(p
, '/');
1564 if (!n
&& !allow_root
)
1565 die("Empty path component found in input");
1576 for (i
= 0; i
< t
->entry_count
; i
++) {
1578 if (e
->name
->str_len
== n
&& !fspathncmp(p
, e
->name
->str_dat
, n
)) {
1581 if (!S_ISDIR(e
->versions
[1].mode
))
1585 return tree_content_get(e
, slash1
+ 1, leaf
, 0);
1591 memcpy(leaf
, e
, sizeof(*leaf
));
1592 if (e
->tree
&& is_null_oid(&e
->versions
[1].oid
))
1593 leaf
->tree
= dup_tree_content(e
->tree
);
1599 static int update_branch(struct branch
*b
)
1601 static const char *msg
= "fast-import";
1602 struct ref_transaction
*transaction
;
1603 struct object_id old_oid
;
1604 struct strbuf err
= STRBUF_INIT
;
1606 if (is_null_oid(&b
->oid
)) {
1608 refs_delete_ref(get_main_ref_store(the_repository
),
1609 NULL
, b
->name
, NULL
, 0);
1612 if (refs_read_ref(get_main_ref_store(the_repository
), b
->name
, &old_oid
))
1613 oidclr(&old_oid
, the_repository
->hash_algo
);
1614 if (!force_update
&& !is_null_oid(&old_oid
)) {
1615 struct commit
*old_cmit
, *new_cmit
;
1618 old_cmit
= lookup_commit_reference_gently(the_repository
,
1620 new_cmit
= lookup_commit_reference_gently(the_repository
,
1622 if (!old_cmit
|| !new_cmit
)
1623 return error("Branch %s is missing commits.", b
->name
);
1625 ret
= repo_in_merge_bases(the_repository
, old_cmit
, new_cmit
);
1629 warning("Not updating %s"
1630 " (new tip %s does not contain %s)",
1631 b
->name
, oid_to_hex(&b
->oid
),
1632 oid_to_hex(&old_oid
));
1636 transaction
= ref_store_transaction_begin(get_main_ref_store(the_repository
),
1639 ref_transaction_update(transaction
, b
->name
, &b
->oid
, &old_oid
,
1640 NULL
, NULL
, 0, msg
, &err
) ||
1641 ref_transaction_commit(transaction
, &err
)) {
1642 ref_transaction_free(transaction
);
1643 error("%s", err
.buf
);
1644 strbuf_release(&err
);
1647 ref_transaction_free(transaction
);
1648 strbuf_release(&err
);
1652 static void dump_branches(void)
1657 for (i
= 0; i
< branch_table_sz
; i
++) {
1658 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
)
1659 failure
|= update_branch(b
);
1663 static void dump_tags(void)
1665 static const char *msg
= "fast-import";
1667 struct strbuf ref_name
= STRBUF_INIT
;
1668 struct strbuf err
= STRBUF_INIT
;
1669 struct ref_transaction
*transaction
;
1671 transaction
= ref_store_transaction_begin(get_main_ref_store(the_repository
),
1674 failure
|= error("%s", err
.buf
);
1677 for (t
= first_tag
; t
; t
= t
->next_tag
) {
1678 strbuf_reset(&ref_name
);
1679 strbuf_addf(&ref_name
, "refs/tags/%s", t
->name
);
1681 if (ref_transaction_update(transaction
, ref_name
.buf
,
1682 &t
->oid
, NULL
, NULL
, NULL
,
1684 failure
|= error("%s", err
.buf
);
1688 if (ref_transaction_commit(transaction
, &err
))
1689 failure
|= error("%s", err
.buf
);
1692 ref_transaction_free(transaction
);
1693 strbuf_release(&ref_name
);
1694 strbuf_release(&err
);
1697 static void dump_marks(void)
1699 struct lock_file mark_lock
= LOCK_INIT
;
1702 if (!export_marks_file
|| (import_marks_file
&& !import_marks_file_done
))
1705 if (safe_create_leading_directories_const(export_marks_file
)) {
1706 failure
|= error_errno("unable to create leading directories of %s",
1711 if (hold_lock_file_for_update(&mark_lock
, export_marks_file
, 0) < 0) {
1712 failure
|= error_errno("Unable to write marks file %s",
1717 f
= fdopen_lock_file(&mark_lock
, "w");
1719 int saved_errno
= errno
;
1720 rollback_lock_file(&mark_lock
);
1721 failure
|= error("Unable to write marks file %s: %s",
1722 export_marks_file
, strerror(saved_errno
));
1726 for_each_mark(marks
, 0, dump_marks_fn
, f
);
1727 if (commit_lock_file(&mark_lock
)) {
1728 failure
|= error_errno("Unable to write file %s",
1734 static void insert_object_entry(struct mark_set
**s
, struct object_id
*oid
, uintmax_t mark
)
1736 struct object_entry
*e
;
1737 e
= find_object(oid
);
1739 enum object_type type
= oid_object_info(the_repository
,
1742 die("object not found: %s", oid_to_hex(oid
));
1743 e
= insert_object(oid
);
1745 e
->pack_id
= MAX_PACK_ID
;
1746 e
->idx
.offset
= 1; /* just not zero! */
1748 insert_mark(s
, mark
, e
);
1751 static void insert_oid_entry(struct mark_set
**s
, struct object_id
*oid
, uintmax_t mark
)
1753 insert_mark(s
, mark
, xmemdupz(oid
, sizeof(*oid
)));
1756 static void read_mark_file(struct mark_set
**s
, FILE *f
, mark_set_inserter_t inserter
)
1759 while (fgets(line
, sizeof(line
), f
)) {
1762 struct object_id oid
;
1764 /* Ensure SHA-1 objects are padded with zeros. */
1765 memset(oid
.hash
, 0, sizeof(oid
.hash
));
1767 end
= strchr(line
, '\n');
1768 if (line
[0] != ':' || !end
)
1769 die("corrupt mark line: %s", line
);
1771 mark
= strtoumax(line
+ 1, &end
, 10);
1772 if (!mark
|| end
== line
+ 1
1774 || get_oid_hex_any(end
+ 1, &oid
) == GIT_HASH_UNKNOWN
)
1775 die("corrupt mark line: %s", line
);
1776 inserter(s
, &oid
, mark
);
1780 static void read_marks(void)
1782 FILE *f
= fopen(import_marks_file
, "r");
1785 else if (import_marks_file_ignore_missing
&& errno
== ENOENT
)
1786 goto done
; /* Marks file does not exist */
1788 die_errno("cannot read '%s'", import_marks_file
);
1789 read_mark_file(&marks
, f
, insert_object_entry
);
1792 import_marks_file_done
= 1;
1796 static int read_next_command(void)
1798 static int stdin_eof
= 0;
1801 unread_command_buf
= 0;
1806 if (unread_command_buf
) {
1807 unread_command_buf
= 0;
1809 struct recent_command
*rc
;
1811 stdin_eof
= strbuf_getline_lf(&command_buf
, stdin
);
1815 if (!seen_data_command
1816 && !starts_with(command_buf
.buf
, "feature ")
1817 && !starts_with(command_buf
.buf
, "option ")) {
1826 cmd_hist
.next
= rc
->next
;
1827 cmd_hist
.next
->prev
= &cmd_hist
;
1831 rc
->buf
= xstrdup(command_buf
.buf
);
1832 rc
->prev
= cmd_tail
;
1833 rc
->next
= cmd_hist
.prev
;
1834 rc
->prev
->next
= rc
;
1837 if (command_buf
.buf
[0] == '#')
1843 static void skip_optional_lf(void)
1845 int term_char
= fgetc(stdin
);
1846 if (term_char
!= '\n' && term_char
!= EOF
)
1847 ungetc(term_char
, stdin
);
1850 static void parse_mark(void)
1853 if (skip_prefix(command_buf
.buf
, "mark :", &v
)) {
1854 next_mark
= strtoumax(v
, NULL
, 10);
1855 read_next_command();
1861 static void parse_original_identifier(void)
1864 if (skip_prefix(command_buf
.buf
, "original-oid ", &v
))
1865 read_next_command();
1868 static int parse_data(struct strbuf
*sb
, uintmax_t limit
, uintmax_t *len_res
)
1873 if (!skip_prefix(command_buf
.buf
, "data ", &data
))
1874 die("Expected 'data n' command, found: %s", command_buf
.buf
);
1876 if (skip_prefix(data
, "<<", &data
)) {
1877 char *term
= xstrdup(data
);
1878 size_t term_len
= command_buf
.len
- (data
- command_buf
.buf
);
1881 if (strbuf_getline_lf(&command_buf
, stdin
) == EOF
)
1882 die("EOF in data (terminator '%s' not found)", term
);
1883 if (term_len
== command_buf
.len
1884 && !strcmp(term
, command_buf
.buf
))
1886 strbuf_addbuf(sb
, &command_buf
);
1887 strbuf_addch(sb
, '\n');
1892 uintmax_t len
= strtoumax(data
, NULL
, 10);
1893 size_t n
= 0, length
= (size_t)len
;
1895 if (limit
&& limit
< len
) {
1900 die("data is too large to use in this context");
1902 while (n
< length
) {
1903 size_t s
= strbuf_fread(sb
, length
- n
, stdin
);
1904 if (!s
&& feof(stdin
))
1905 die("EOF in data (%lu bytes remaining)",
1906 (unsigned long)(length
- n
));
1915 static int validate_raw_date(const char *src
, struct strbuf
*result
, int strict
)
1917 const char *orig_src
= src
;
1923 num
= strtoul(src
, &endp
, 10);
1925 * NEEDSWORK: perhaps check for reasonable values? For example, we
1926 * could error on values representing times more than a
1927 * day in the future.
1929 if (errno
|| endp
== src
|| *endp
!= ' ')
1933 if (*src
!= '-' && *src
!= '+')
1936 num
= strtoul(src
+ 1, &endp
, 10);
1938 * NEEDSWORK: check for brokenness other than num > 1400, such as
1939 * (num % 100) >= 60, or ((num % 100) % 15) != 0 ?
1941 if (errno
|| endp
== src
+ 1 || *endp
|| /* did not parse */
1942 (strict
&& (1400 < num
)) /* parsed a broken timezone */
1946 strbuf_addstr(result
, orig_src
);
1950 static char *parse_ident(const char *buf
)
1954 struct strbuf ident
= STRBUF_INIT
;
1956 /* ensure there is a space delimiter even if there is no name */
1960 ltgt
= buf
+ strcspn(buf
, "<>");
1962 die("Missing < in ident string: %s", buf
);
1963 if (ltgt
!= buf
&& ltgt
[-1] != ' ')
1964 die("Missing space before < in ident string: %s", buf
);
1965 ltgt
= ltgt
+ 1 + strcspn(ltgt
+ 1, "<>");
1967 die("Missing > in ident string: %s", buf
);
1970 die("Missing space after > in ident string: %s", buf
);
1972 name_len
= ltgt
- buf
;
1973 strbuf_add(&ident
, buf
, name_len
);
1977 if (validate_raw_date(ltgt
, &ident
, 1) < 0)
1978 die("Invalid raw date \"%s\" in ident: %s", ltgt
, buf
);
1980 case WHENSPEC_RAW_PERMISSIVE
:
1981 if (validate_raw_date(ltgt
, &ident
, 0) < 0)
1982 die("Invalid raw date \"%s\" in ident: %s", ltgt
, buf
);
1984 case WHENSPEC_RFC2822
:
1985 if (parse_date(ltgt
, &ident
) < 0)
1986 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt
, buf
);
1989 if (strcmp("now", ltgt
))
1990 die("Date in ident must be 'now': %s", buf
);
1995 return strbuf_detach(&ident
, NULL
);
1998 static void parse_and_store_blob(
1999 struct last_object
*last
,
2000 struct object_id
*oidout
,
2003 static struct strbuf buf
= STRBUF_INIT
;
2006 if (parse_data(&buf
, big_file_threshold
, &len
))
2007 store_object(OBJ_BLOB
, &buf
, last
, oidout
, mark
);
2010 strbuf_release(&last
->data
);
2014 stream_blob(len
, oidout
, mark
);
2019 static void parse_new_blob(void)
2021 read_next_command();
2023 parse_original_identifier();
2024 parse_and_store_blob(&last_blob
, NULL
, next_mark
);
2027 static void unload_one_branch(void)
2029 while (cur_active_branches
2030 && cur_active_branches
>= max_active_branches
) {
2031 uintmax_t min_commit
= ULONG_MAX
;
2032 struct branch
*e
, *l
= NULL
, *p
= NULL
;
2034 for (e
= active_branches
; e
; e
= e
->active_next_branch
) {
2035 if (e
->last_commit
< min_commit
) {
2037 min_commit
= e
->last_commit
;
2043 e
= p
->active_next_branch
;
2044 p
->active_next_branch
= e
->active_next_branch
;
2046 e
= active_branches
;
2047 active_branches
= e
->active_next_branch
;
2050 e
->active_next_branch
= NULL
;
2051 if (e
->branch_tree
.tree
) {
2052 release_tree_content_recursive(e
->branch_tree
.tree
);
2053 e
->branch_tree
.tree
= NULL
;
2055 cur_active_branches
--;
2059 static void load_branch(struct branch
*b
)
2061 load_tree(&b
->branch_tree
);
2064 b
->active_next_branch
= active_branches
;
2065 active_branches
= b
;
2066 cur_active_branches
++;
2067 branch_load_count
++;
2071 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes
)
2073 unsigned char fanout
= 0;
2074 while ((num_notes
>>= 8))
2079 static void construct_path_with_fanout(const char *hex_sha1
,
2080 unsigned char fanout
, char *path
)
2082 unsigned int i
= 0, j
= 0;
2083 if (fanout
>= the_hash_algo
->rawsz
)
2084 die("Too large fanout (%u)", fanout
);
2086 path
[i
++] = hex_sha1
[j
++];
2087 path
[i
++] = hex_sha1
[j
++];
2091 memcpy(path
+ i
, hex_sha1
+ j
, the_hash_algo
->hexsz
- j
);
2092 path
[i
+ the_hash_algo
->hexsz
- j
] = '\0';
2095 static uintmax_t do_change_note_fanout(
2096 struct tree_entry
*orig_root
, struct tree_entry
*root
,
2097 char *hex_oid
, unsigned int hex_oid_len
,
2098 char *fullpath
, unsigned int fullpath_len
,
2099 unsigned char fanout
)
2101 struct tree_content
*t
;
2102 struct tree_entry
*e
, leaf
;
2103 unsigned int i
, tmp_hex_oid_len
, tmp_fullpath_len
;
2104 uintmax_t num_notes
= 0;
2105 struct object_id oid
;
2106 /* hex oid + '/' between each pair of hex digits + NUL */
2107 char realpath
[GIT_MAX_HEXSZ
+ ((GIT_MAX_HEXSZ
/ 2) - 1) + 1];
2108 const unsigned hexsz
= the_hash_algo
->hexsz
;
2114 for (i
= 0; t
&& i
< t
->entry_count
; i
++) {
2116 tmp_hex_oid_len
= hex_oid_len
+ e
->name
->str_len
;
2117 tmp_fullpath_len
= fullpath_len
;
2120 * We're interested in EITHER existing note entries (entries
2121 * with exactly 40 hex chars in path, not including directory
2122 * separators), OR directory entries that may contain note
2123 * entries (with < 40 hex chars in path).
2124 * Also, each path component in a note entry must be a multiple
2127 if (!e
->versions
[1].mode
||
2128 tmp_hex_oid_len
> hexsz
||
2129 e
->name
->str_len
% 2)
2132 /* This _may_ be a note entry, or a subdir containing notes */
2133 memcpy(hex_oid
+ hex_oid_len
, e
->name
->str_dat
,
2135 if (tmp_fullpath_len
)
2136 fullpath
[tmp_fullpath_len
++] = '/';
2137 memcpy(fullpath
+ tmp_fullpath_len
, e
->name
->str_dat
,
2139 tmp_fullpath_len
+= e
->name
->str_len
;
2140 fullpath
[tmp_fullpath_len
] = '\0';
2142 if (tmp_hex_oid_len
== hexsz
&& !get_oid_hex(hex_oid
, &oid
)) {
2143 /* This is a note entry */
2144 if (fanout
== 0xff) {
2145 /* Counting mode, no rename */
2149 construct_path_with_fanout(hex_oid
, fanout
, realpath
);
2150 if (!strcmp(fullpath
, realpath
)) {
2151 /* Note entry is in correct location */
2156 /* Rename fullpath to realpath */
2157 if (!tree_content_remove(orig_root
, fullpath
, &leaf
, 0))
2158 die("Failed to remove path %s", fullpath
);
2159 tree_content_set(orig_root
, realpath
,
2160 &leaf
.versions
[1].oid
,
2161 leaf
.versions
[1].mode
,
2163 } else if (S_ISDIR(e
->versions
[1].mode
)) {
2164 /* This is a subdir that may contain note entries */
2165 num_notes
+= do_change_note_fanout(orig_root
, e
,
2166 hex_oid
, tmp_hex_oid_len
,
2167 fullpath
, tmp_fullpath_len
, fanout
);
2170 /* The above may have reallocated the current tree_content */
2176 static uintmax_t change_note_fanout(struct tree_entry
*root
,
2177 unsigned char fanout
)
2180 * The size of path is due to one slash between every two hex digits,
2181 * plus the terminating NUL. Note that there is no slash at the end, so
2182 * the number of slashes is one less than half the number of hex
2185 char hex_oid
[GIT_MAX_HEXSZ
], path
[GIT_MAX_HEXSZ
+ (GIT_MAX_HEXSZ
/ 2) - 1 + 1];
2186 return do_change_note_fanout(root
, root
, hex_oid
, 0, path
, 0, fanout
);
2189 static int parse_mapped_oid_hex(const char *hex
, struct object_id
*oid
, const char **end
)
2194 /* Make SHA-1 object IDs have all-zero padding. */
2195 memset(oid
->hash
, 0, sizeof(oid
->hash
));
2197 algo
= parse_oid_hex_any(hex
, oid
, end
);
2198 if (algo
== GIT_HASH_UNKNOWN
)
2201 it
= kh_get_oid_map(sub_oid_map
, *oid
);
2202 /* No such object? */
2203 if (it
== kh_end(sub_oid_map
)) {
2204 /* If we're using the same algorithm, pass it through. */
2205 if (hash_algos
[algo
].format_id
== the_hash_algo
->format_id
)
2209 oidcpy(oid
, kh_value(sub_oid_map
, it
));
2214 * Given a pointer into a string, parse a mark reference:
2216 * idnum ::= ':' bigint;
2218 * Update *endptr to point to the first character after the value.
2220 * Complain if the following character is not what is expected,
2221 * either a space or end of the string.
2223 static uintmax_t parse_mark_ref(const char *p
, char **endptr
)
2229 mark
= strtoumax(p
, endptr
, 10);
2231 die("No value after ':' in mark: %s", command_buf
.buf
);
2236 * Parse the mark reference, and complain if this is not the end of
2239 static uintmax_t parse_mark_ref_eol(const char *p
)
2244 mark
= parse_mark_ref(p
, &end
);
2246 die("Garbage after mark: %s", command_buf
.buf
);
2251 * Parse the mark reference, demanding a trailing space. Update *p to
2252 * point to the first character after the space.
2254 static uintmax_t parse_mark_ref_space(const char **p
)
2259 mark
= parse_mark_ref(*p
, &end
);
2261 die("Missing space after mark: %s", command_buf
.buf
);
2267 * Parse the path string into the strbuf. The path can either be quoted with
2268 * escape sequences or unquoted without escape sequences. Unquoted strings may
2269 * contain spaces only if `is_last_field` is nonzero; otherwise, it stops
2270 * parsing at the first space.
2272 static void parse_path(struct strbuf
*sb
, const char *p
, const char **endp
,
2273 int is_last_field
, const char *field
)
2276 if (unquote_c_style(sb
, p
, endp
))
2277 die("Invalid %s: %s", field
, command_buf
.buf
);
2278 if (strlen(sb
->buf
) != sb
->len
)
2279 die("NUL in %s: %s", field
, command_buf
.buf
);
2282 * Unless we are parsing the last field of a line,
2283 * SP is the end of this field.
2285 *endp
= is_last_field
2287 : strchrnul(p
, ' ');
2288 strbuf_add(sb
, p
, *endp
- p
);
2293 * Parse the path string into the strbuf, and complain if this is not the end of
2294 * the string. Unquoted strings may contain spaces.
2296 static void parse_path_eol(struct strbuf
*sb
, const char *p
, const char *field
)
2300 parse_path(sb
, p
, &end
, 1, field
);
2302 die("Garbage after %s: %s", field
, command_buf
.buf
);
2306 * Parse the path string into the strbuf, and ensure it is followed by a space.
2307 * Unquoted strings may not contain spaces. Update *endp to point to the first
2308 * character after the space.
2310 static void parse_path_space(struct strbuf
*sb
, const char *p
,
2311 const char **endp
, const char *field
)
2313 parse_path(sb
, p
, endp
, 0, field
);
2315 die("Missing space after %s: %s", field
, command_buf
.buf
);
2319 static void file_change_m(const char *p
, struct branch
*b
)
2321 static struct strbuf path
= STRBUF_INIT
;
2322 struct object_entry
*oe
;
2323 struct object_id oid
;
2324 uint16_t mode
, inline_data
= 0;
2326 p
= parse_mode(p
, &mode
);
2328 die("Corrupt mode: %s", command_buf
.buf
);
2333 case S_IFREG
| 0644:
2334 case S_IFREG
| 0755:
2341 die("Corrupt mode: %s", command_buf
.buf
);
2345 oe
= find_mark(marks
, parse_mark_ref_space(&p
));
2346 oidcpy(&oid
, &oe
->idx
.oid
);
2347 } else if (skip_prefix(p
, "inline ", &p
)) {
2349 oe
= NULL
; /* not used with inline_data, but makes gcc happy */
2351 if (parse_mapped_oid_hex(p
, &oid
, &p
))
2352 die("Invalid dataref: %s", command_buf
.buf
);
2353 oe
= find_object(&oid
);
2355 die("Missing space after SHA1: %s", command_buf
.buf
);
2358 strbuf_reset(&path
);
2359 parse_path_eol(&path
, p
, "path");
2361 /* Git does not track empty, non-toplevel directories. */
2362 if (S_ISDIR(mode
) &&
2363 is_empty_tree_oid(&oid
, the_repository
->hash_algo
) &&
2365 tree_content_remove(&b
->branch_tree
, path
.buf
, NULL
, 0);
2369 if (S_ISGITLINK(mode
)) {
2371 die("Git links cannot be specified 'inline': %s",
2374 if (oe
->type
!= OBJ_COMMIT
)
2375 die("Not a commit (actually a %s): %s",
2376 type_name(oe
->type
), command_buf
.buf
);
2379 * Accept the sha1 without checking; it expected to be in
2380 * another repository.
2382 } else if (inline_data
) {
2384 die("Directories cannot be specified 'inline': %s",
2386 while (read_next_command() != EOF
) {
2388 if (skip_prefix(command_buf
.buf
, "cat-blob ", &v
))
2391 parse_and_store_blob(&last_blob
, &oid
, 0);
2396 enum object_type expected
= S_ISDIR(mode
) ?
2398 enum object_type type
= oe
? oe
->type
:
2399 oid_object_info(the_repository
, &oid
,
2402 die("%s not found: %s",
2403 S_ISDIR(mode
) ? "Tree" : "Blob",
2405 if (type
!= expected
)
2406 die("Not a %s (actually a %s): %s",
2407 type_name(expected
), type_name(type
),
2412 tree_content_replace(&b
->branch_tree
, &oid
, mode
, NULL
);
2415 tree_content_set(&b
->branch_tree
, path
.buf
, &oid
, mode
, NULL
);
2418 static void file_change_d(const char *p
, struct branch
*b
)
2420 static struct strbuf path
= STRBUF_INIT
;
2422 strbuf_reset(&path
);
2423 parse_path_eol(&path
, p
, "path");
2424 tree_content_remove(&b
->branch_tree
, path
.buf
, NULL
, 1);
2427 static void file_change_cr(const char *p
, struct branch
*b
, int rename
)
2429 static struct strbuf source
= STRBUF_INIT
;
2430 static struct strbuf dest
= STRBUF_INIT
;
2431 struct tree_entry leaf
;
2433 strbuf_reset(&source
);
2434 parse_path_space(&source
, p
, &p
, "source");
2435 strbuf_reset(&dest
);
2436 parse_path_eol(&dest
, p
, "dest");
2438 memset(&leaf
, 0, sizeof(leaf
));
2440 tree_content_remove(&b
->branch_tree
, source
.buf
, &leaf
, 1);
2442 tree_content_get(&b
->branch_tree
, source
.buf
, &leaf
, 1);
2443 if (!leaf
.versions
[1].mode
)
2444 die("Path %s not in branch", source
.buf
);
2445 if (!*dest
.buf
) { /* C "path/to/subdir" "" */
2446 tree_content_replace(&b
->branch_tree
,
2447 &leaf
.versions
[1].oid
,
2448 leaf
.versions
[1].mode
,
2452 tree_content_set(&b
->branch_tree
, dest
.buf
,
2453 &leaf
.versions
[1].oid
,
2454 leaf
.versions
[1].mode
,
2458 static void note_change_n(const char *p
, struct branch
*b
, unsigned char *old_fanout
)
2460 struct object_entry
*oe
;
2462 struct object_id oid
, commit_oid
;
2463 char path
[GIT_MAX_RAWSZ
* 3];
2464 uint16_t inline_data
= 0;
2465 unsigned char new_fanout
;
2468 * When loading a branch, we don't traverse its tree to count the real
2469 * number of notes (too expensive to do this for all non-note refs).
2470 * This means that recently loaded notes refs might incorrectly have
2471 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2473 * Fix this by traversing the tree and counting the number of notes
2474 * when b->num_notes == 0. If the notes tree is truly empty, the
2475 * calculation should not take long.
2477 if (b
->num_notes
== 0 && *old_fanout
== 0) {
2478 /* Invoke change_note_fanout() in "counting mode". */
2479 b
->num_notes
= change_note_fanout(&b
->branch_tree
, 0xff);
2480 *old_fanout
= convert_num_notes_to_fanout(b
->num_notes
);
2483 /* Now parse the notemodify command. */
2484 /* <dataref> or 'inline' */
2486 oe
= find_mark(marks
, parse_mark_ref_space(&p
));
2487 oidcpy(&oid
, &oe
->idx
.oid
);
2488 } else if (skip_prefix(p
, "inline ", &p
)) {
2490 oe
= NULL
; /* not used with inline_data, but makes gcc happy */
2492 if (parse_mapped_oid_hex(p
, &oid
, &p
))
2493 die("Invalid dataref: %s", command_buf
.buf
);
2494 oe
= find_object(&oid
);
2496 die("Missing space after SHA1: %s", command_buf
.buf
);
2500 s
= lookup_branch(p
);
2502 if (is_null_oid(&s
->oid
))
2503 die("Can't add a note on empty branch.");
2504 oidcpy(&commit_oid
, &s
->oid
);
2505 } else if (*p
== ':') {
2506 uintmax_t commit_mark
= parse_mark_ref_eol(p
);
2507 struct object_entry
*commit_oe
= find_mark(marks
, commit_mark
);
2508 if (commit_oe
->type
!= OBJ_COMMIT
)
2509 die("Mark :%" PRIuMAX
" not a commit", commit_mark
);
2510 oidcpy(&commit_oid
, &commit_oe
->idx
.oid
);
2511 } else if (!repo_get_oid(the_repository
, p
, &commit_oid
)) {
2513 char *buf
= read_object_with_reference(the_repository
,
2517 if (!buf
|| size
< the_hash_algo
->hexsz
+ 6)
2518 die("Not a valid commit: %s", p
);
2521 die("Invalid ref name or SHA1 expression: %s", p
);
2524 read_next_command();
2525 parse_and_store_blob(&last_blob
, &oid
, 0);
2527 if (oe
->type
!= OBJ_BLOB
)
2528 die("Not a blob (actually a %s): %s",
2529 type_name(oe
->type
), command_buf
.buf
);
2530 } else if (!is_null_oid(&oid
)) {
2531 enum object_type type
= oid_object_info(the_repository
, &oid
,
2534 die("Blob not found: %s", command_buf
.buf
);
2535 if (type
!= OBJ_BLOB
)
2536 die("Not a blob (actually a %s): %s",
2537 type_name(type
), command_buf
.buf
);
2540 construct_path_with_fanout(oid_to_hex(&commit_oid
), *old_fanout
, path
);
2541 if (tree_content_remove(&b
->branch_tree
, path
, NULL
, 0))
2544 if (is_null_oid(&oid
))
2545 return; /* nothing to insert */
2548 new_fanout
= convert_num_notes_to_fanout(b
->num_notes
);
2549 construct_path_with_fanout(oid_to_hex(&commit_oid
), new_fanout
, path
);
2550 tree_content_set(&b
->branch_tree
, path
, &oid
, S_IFREG
| 0644, NULL
);
2553 static void file_change_deleteall(struct branch
*b
)
2555 release_tree_content_recursive(b
->branch_tree
.tree
);
2556 oidclr(&b
->branch_tree
.versions
[0].oid
, the_repository
->hash_algo
);
2557 oidclr(&b
->branch_tree
.versions
[1].oid
, the_repository
->hash_algo
);
2558 load_tree(&b
->branch_tree
);
2562 static void parse_from_commit(struct branch
*b
, char *buf
, unsigned long size
)
2564 if (!buf
|| size
< the_hash_algo
->hexsz
+ 6)
2565 die("Not a valid commit: %s", oid_to_hex(&b
->oid
));
2566 if (memcmp("tree ", buf
, 5)
2567 || get_oid_hex(buf
+ 5, &b
->branch_tree
.versions
[1].oid
))
2568 die("The commit %s is corrupt", oid_to_hex(&b
->oid
));
2569 oidcpy(&b
->branch_tree
.versions
[0].oid
,
2570 &b
->branch_tree
.versions
[1].oid
);
2573 static void parse_from_existing(struct branch
*b
)
2575 if (is_null_oid(&b
->oid
)) {
2576 oidclr(&b
->branch_tree
.versions
[0].oid
, the_repository
->hash_algo
);
2577 oidclr(&b
->branch_tree
.versions
[1].oid
, the_repository
->hash_algo
);
2582 buf
= read_object_with_reference(the_repository
,
2583 &b
->oid
, OBJ_COMMIT
, &size
,
2585 parse_from_commit(b
, buf
, size
);
2590 static int parse_objectish(struct branch
*b
, const char *objectish
)
2593 struct object_id oid
;
2595 oidcpy(&oid
, &b
->branch_tree
.versions
[1].oid
);
2597 s
= lookup_branch(objectish
);
2599 die("Can't create a branch from itself: %s", b
->name
);
2601 struct object_id
*t
= &s
->branch_tree
.versions
[1].oid
;
2602 oidcpy(&b
->oid
, &s
->oid
);
2603 oidcpy(&b
->branch_tree
.versions
[0].oid
, t
);
2604 oidcpy(&b
->branch_tree
.versions
[1].oid
, t
);
2605 } else if (*objectish
== ':') {
2606 uintmax_t idnum
= parse_mark_ref_eol(objectish
);
2607 struct object_entry
*oe
= find_mark(marks
, idnum
);
2608 if (oe
->type
!= OBJ_COMMIT
)
2609 die("Mark :%" PRIuMAX
" not a commit", idnum
);
2610 if (!oideq(&b
->oid
, &oe
->idx
.oid
)) {
2611 oidcpy(&b
->oid
, &oe
->idx
.oid
);
2612 if (oe
->pack_id
!= MAX_PACK_ID
) {
2614 char *buf
= gfi_unpack_entry(oe
, &size
);
2615 parse_from_commit(b
, buf
, size
);
2618 parse_from_existing(b
);
2620 } else if (!repo_get_oid(the_repository
, objectish
, &b
->oid
)) {
2621 parse_from_existing(b
);
2622 if (is_null_oid(&b
->oid
))
2626 die("Invalid ref name or SHA1 expression: %s", objectish
);
2628 if (b
->branch_tree
.tree
&& !oideq(&oid
, &b
->branch_tree
.versions
[1].oid
)) {
2629 release_tree_content_recursive(b
->branch_tree
.tree
);
2630 b
->branch_tree
.tree
= NULL
;
2633 read_next_command();
2637 static int parse_from(struct branch
*b
)
2641 if (!skip_prefix(command_buf
.buf
, "from ", &from
))
2644 return parse_objectish(b
, from
);
2647 static int parse_objectish_with_prefix(struct branch
*b
, const char *prefix
)
2651 if (!skip_prefix(command_buf
.buf
, prefix
, &base
))
2654 return parse_objectish(b
, base
);
2657 static struct hash_list
*parse_merge(unsigned int *count
)
2659 struct hash_list
*list
= NULL
, **tail
= &list
, *n
;
2664 while (skip_prefix(command_buf
.buf
, "merge ", &from
)) {
2665 n
= xmalloc(sizeof(*n
));
2666 s
= lookup_branch(from
);
2668 oidcpy(&n
->oid
, &s
->oid
);
2669 else if (*from
== ':') {
2670 uintmax_t idnum
= parse_mark_ref_eol(from
);
2671 struct object_entry
*oe
= find_mark(marks
, idnum
);
2672 if (oe
->type
!= OBJ_COMMIT
)
2673 die("Mark :%" PRIuMAX
" not a commit", idnum
);
2674 oidcpy(&n
->oid
, &oe
->idx
.oid
);
2675 } else if (!repo_get_oid(the_repository
, from
, &n
->oid
)) {
2677 char *buf
= read_object_with_reference(the_repository
,
2681 if (!buf
|| size
< the_hash_algo
->hexsz
+ 6)
2682 die("Not a valid commit: %s", from
);
2685 die("Invalid ref name or SHA1 expression: %s", from
);
2692 read_next_command();
2697 static void parse_new_commit(const char *arg
)
2699 static struct strbuf msg
= STRBUF_INIT
;
2701 char *author
= NULL
;
2702 char *committer
= NULL
;
2703 char *encoding
= NULL
;
2704 struct hash_list
*merge_list
= NULL
;
2705 unsigned int merge_count
;
2706 unsigned char prev_fanout
, new_fanout
;
2709 b
= lookup_branch(arg
);
2711 b
= new_branch(arg
);
2713 read_next_command();
2715 parse_original_identifier();
2716 if (skip_prefix(command_buf
.buf
, "author ", &v
)) {
2717 author
= parse_ident(v
);
2718 read_next_command();
2720 if (skip_prefix(command_buf
.buf
, "committer ", &v
)) {
2721 committer
= parse_ident(v
);
2722 read_next_command();
2725 die("Expected committer but didn't get one");
2726 if (skip_prefix(command_buf
.buf
, "encoding ", &v
)) {
2727 encoding
= xstrdup(v
);
2728 read_next_command();
2730 parse_data(&msg
, 0, NULL
);
2731 read_next_command();
2733 merge_list
= parse_merge(&merge_count
);
2735 /* ensure the branch is active/loaded */
2736 if (!b
->branch_tree
.tree
|| !max_active_branches
) {
2737 unload_one_branch();
2741 prev_fanout
= convert_num_notes_to_fanout(b
->num_notes
);
2744 while (command_buf
.len
> 0) {
2745 if (skip_prefix(command_buf
.buf
, "M ", &v
))
2746 file_change_m(v
, b
);
2747 else if (skip_prefix(command_buf
.buf
, "D ", &v
))
2748 file_change_d(v
, b
);
2749 else if (skip_prefix(command_buf
.buf
, "R ", &v
))
2750 file_change_cr(v
, b
, 1);
2751 else if (skip_prefix(command_buf
.buf
, "C ", &v
))
2752 file_change_cr(v
, b
, 0);
2753 else if (skip_prefix(command_buf
.buf
, "N ", &v
))
2754 note_change_n(v
, b
, &prev_fanout
);
2755 else if (!strcmp("deleteall", command_buf
.buf
))
2756 file_change_deleteall(b
);
2757 else if (skip_prefix(command_buf
.buf
, "ls ", &v
))
2759 else if (skip_prefix(command_buf
.buf
, "cat-blob ", &v
))
2762 unread_command_buf
= 1;
2765 if (read_next_command() == EOF
)
2769 new_fanout
= convert_num_notes_to_fanout(b
->num_notes
);
2770 if (new_fanout
!= prev_fanout
)
2771 b
->num_notes
= change_note_fanout(&b
->branch_tree
, new_fanout
);
2773 /* build the tree and the commit */
2774 store_tree(&b
->branch_tree
);
2775 oidcpy(&b
->branch_tree
.versions
[0].oid
,
2776 &b
->branch_tree
.versions
[1].oid
);
2778 strbuf_reset(&new_data
);
2779 strbuf_addf(&new_data
, "tree %s\n",
2780 oid_to_hex(&b
->branch_tree
.versions
[1].oid
));
2781 if (!is_null_oid(&b
->oid
))
2782 strbuf_addf(&new_data
, "parent %s\n",
2783 oid_to_hex(&b
->oid
));
2784 while (merge_list
) {
2785 struct hash_list
*next
= merge_list
->next
;
2786 strbuf_addf(&new_data
, "parent %s\n",
2787 oid_to_hex(&merge_list
->oid
));
2791 strbuf_addf(&new_data
,
2794 author
? author
: committer
, committer
);
2796 strbuf_addf(&new_data
,
2799 strbuf_addch(&new_data
, '\n');
2800 strbuf_addbuf(&new_data
, &msg
);
2805 if (!store_object(OBJ_COMMIT
, &new_data
, NULL
, &b
->oid
, next_mark
))
2806 b
->pack_id
= pack_id
;
2807 b
->last_commit
= object_count_by_type
[OBJ_COMMIT
];
2810 static void parse_new_tag(const char *arg
)
2812 static struct strbuf msg
= STRBUF_INIT
;
2817 uintmax_t from_mark
= 0;
2818 struct object_id oid
;
2819 enum object_type type
;
2822 t
= mem_pool_calloc(&fi_mem_pool
, 1, sizeof(struct tag
));
2823 t
->name
= mem_pool_strdup(&fi_mem_pool
, arg
);
2825 last_tag
->next_tag
= t
;
2829 read_next_command();
2833 if (!skip_prefix(command_buf
.buf
, "from ", &from
))
2834 die("Expected from command, got %s", command_buf
.buf
);
2835 s
= lookup_branch(from
);
2837 if (is_null_oid(&s
->oid
))
2838 die("Can't tag an empty branch.");
2839 oidcpy(&oid
, &s
->oid
);
2841 } else if (*from
== ':') {
2842 struct object_entry
*oe
;
2843 from_mark
= parse_mark_ref_eol(from
);
2844 oe
= find_mark(marks
, from_mark
);
2846 oidcpy(&oid
, &oe
->idx
.oid
);
2847 } else if (!repo_get_oid(the_repository
, from
, &oid
)) {
2848 struct object_entry
*oe
= find_object(&oid
);
2850 type
= oid_object_info(the_repository
, &oid
, NULL
);
2852 die("Not a valid object: %s", from
);
2856 die("Invalid ref name or SHA1 expression: %s", from
);
2857 read_next_command();
2859 /* original-oid ... */
2860 parse_original_identifier();
2863 if (skip_prefix(command_buf
.buf
, "tagger ", &v
)) {
2864 tagger
= parse_ident(v
);
2865 read_next_command();
2869 /* tag payload/message */
2870 parse_data(&msg
, 0, NULL
);
2872 /* build the tag object */
2873 strbuf_reset(&new_data
);
2875 strbuf_addf(&new_data
,
2879 oid_to_hex(&oid
), type_name(type
), t
->name
);
2881 strbuf_addf(&new_data
,
2882 "tagger %s\n", tagger
);
2883 strbuf_addch(&new_data
, '\n');
2884 strbuf_addbuf(&new_data
, &msg
);
2887 if (store_object(OBJ_TAG
, &new_data
, NULL
, &t
->oid
, next_mark
))
2888 t
->pack_id
= MAX_PACK_ID
;
2890 t
->pack_id
= pack_id
;
2893 static void parse_reset_branch(const char *arg
)
2896 const char *tag_name
;
2898 b
= lookup_branch(arg
);
2900 oidclr(&b
->oid
, the_repository
->hash_algo
);
2901 oidclr(&b
->branch_tree
.versions
[0].oid
, the_repository
->hash_algo
);
2902 oidclr(&b
->branch_tree
.versions
[1].oid
, the_repository
->hash_algo
);
2903 if (b
->branch_tree
.tree
) {
2904 release_tree_content_recursive(b
->branch_tree
.tree
);
2905 b
->branch_tree
.tree
= NULL
;
2909 b
= new_branch(arg
);
2910 read_next_command();
2912 if (b
->delete && skip_prefix(b
->name
, "refs/tags/", &tag_name
)) {
2914 * Elsewhere, we call dump_branches() before dump_tags(),
2915 * and dump_branches() will handle ref deletions first, so
2916 * in order to make sure the deletion actually takes effect,
2917 * we need to remove the tag from our list of tags to update.
2919 * NEEDSWORK: replace list of tags with hashmap for faster
2922 struct tag
*t
, *prev
= NULL
;
2923 for (t
= first_tag
; t
; t
= t
->next_tag
) {
2924 if (!strcmp(t
->name
, tag_name
))
2930 prev
->next_tag
= t
->next_tag
;
2932 first_tag
= t
->next_tag
;
2935 /* There is no mem_pool_free(t) function to call. */
2938 if (command_buf
.len
> 0)
2939 unread_command_buf
= 1;
2942 static void cat_blob_write(const char *buf
, unsigned long size
)
2944 if (write_in_full(cat_blob_fd
, buf
, size
) < 0)
2945 die_errno("Write to frontend failed");
2948 static void cat_blob(struct object_entry
*oe
, struct object_id
*oid
)
2950 struct strbuf line
= STRBUF_INIT
;
2952 enum object_type type
= 0;
2955 if (!oe
|| oe
->pack_id
== MAX_PACK_ID
) {
2956 buf
= repo_read_object_file(the_repository
, oid
, &type
, &size
);
2959 buf
= gfi_unpack_entry(oe
, &size
);
2963 * Output based on batch_one_object() from cat-file.c.
2966 strbuf_reset(&line
);
2967 strbuf_addf(&line
, "%s missing\n", oid_to_hex(oid
));
2968 cat_blob_write(line
.buf
, line
.len
);
2969 strbuf_release(&line
);
2974 die("Can't read object %s", oid_to_hex(oid
));
2975 if (type
!= OBJ_BLOB
)
2976 die("Object %s is a %s but a blob was expected.",
2977 oid_to_hex(oid
), type_name(type
));
2978 strbuf_reset(&line
);
2979 strbuf_addf(&line
, "%s %s %"PRIuMAX
"\n", oid_to_hex(oid
),
2980 type_name(type
), (uintmax_t)size
);
2981 cat_blob_write(line
.buf
, line
.len
);
2982 strbuf_release(&line
);
2983 cat_blob_write(buf
, size
);
2984 cat_blob_write("\n", 1);
2985 if (oe
&& oe
->pack_id
== pack_id
) {
2986 last_blob
.offset
= oe
->idx
.offset
;
2987 strbuf_attach(&last_blob
.data
, buf
, size
, size
);
2988 last_blob
.depth
= oe
->depth
;
2993 static void parse_get_mark(const char *p
)
2995 struct object_entry
*oe
;
2996 char output
[GIT_MAX_HEXSZ
+ 2];
2998 /* get-mark SP <object> LF */
3000 die("Not a mark: %s", p
);
3002 oe
= find_mark(marks
, parse_mark_ref_eol(p
));
3004 die("Unknown mark: %s", command_buf
.buf
);
3006 xsnprintf(output
, sizeof(output
), "%s\n", oid_to_hex(&oe
->idx
.oid
));
3007 cat_blob_write(output
, the_hash_algo
->hexsz
+ 1);
3010 static void parse_cat_blob(const char *p
)
3012 struct object_entry
*oe
;
3013 struct object_id oid
;
3015 /* cat-blob SP <object> LF */
3017 oe
= find_mark(marks
, parse_mark_ref_eol(p
));
3019 die("Unknown mark: %s", command_buf
.buf
);
3020 oidcpy(&oid
, &oe
->idx
.oid
);
3022 if (parse_mapped_oid_hex(p
, &oid
, &p
))
3023 die("Invalid dataref: %s", command_buf
.buf
);
3025 die("Garbage after SHA1: %s", command_buf
.buf
);
3026 oe
= find_object(&oid
);
3032 static struct object_entry
*dereference(struct object_entry
*oe
,
3033 struct object_id
*oid
)
3037 const unsigned hexsz
= the_hash_algo
->hexsz
;
3040 enum object_type type
= oid_object_info(the_repository
, oid
,
3043 die("object not found: %s", oid_to_hex(oid
));
3045 oe
= insert_object(oid
);
3047 oe
->pack_id
= MAX_PACK_ID
;
3051 case OBJ_TREE
: /* easy case. */
3057 die("Not a tree-ish: %s", command_buf
.buf
);
3060 if (oe
->pack_id
!= MAX_PACK_ID
) { /* in a pack being written */
3061 buf
= gfi_unpack_entry(oe
, &size
);
3063 enum object_type unused
;
3064 buf
= repo_read_object_file(the_repository
, oid
, &unused
,
3068 die("Can't load object %s", oid_to_hex(oid
));
3070 /* Peel one layer. */
3073 if (size
< hexsz
+ strlen("object ") ||
3074 get_oid_hex(buf
+ strlen("object "), oid
))
3075 die("Invalid SHA1 in tag: %s", command_buf
.buf
);
3078 if (size
< hexsz
+ strlen("tree ") ||
3079 get_oid_hex(buf
+ strlen("tree "), oid
))
3080 die("Invalid SHA1 in commit: %s", command_buf
.buf
);
3084 return find_object(oid
);
3087 static void insert_mapped_mark(uintmax_t mark
, void *object
, void *cbp
)
3089 struct object_id
*fromoid
= object
;
3090 struct object_id
*tooid
= find_mark(cbp
, mark
);
3094 it
= kh_put_oid_map(sub_oid_map
, *fromoid
, &ret
);
3095 /* We've already seen this object. */
3098 kh_value(sub_oid_map
, it
) = tooid
;
3101 static void build_mark_map_one(struct mark_set
*from
, struct mark_set
*to
)
3103 for_each_mark(from
, 0, insert_mapped_mark
, to
);
3106 static void build_mark_map(struct string_list
*from
, struct string_list
*to
)
3108 struct string_list_item
*fromp
, *top
;
3110 sub_oid_map
= kh_init_oid_map();
3112 for_each_string_list_item(fromp
, from
) {
3113 top
= string_list_lookup(to
, fromp
->string
);
3115 die(_("Missing from marks for submodule '%s'"), fromp
->string
);
3116 } else if (!top
|| !top
->util
) {
3117 die(_("Missing to marks for submodule '%s'"), fromp
->string
);
3119 build_mark_map_one(fromp
->util
, top
->util
);
3123 static struct object_entry
*parse_treeish_dataref(const char **p
)
3125 struct object_id oid
;
3126 struct object_entry
*e
;
3128 if (**p
== ':') { /* <mark> */
3129 e
= find_mark(marks
, parse_mark_ref_space(p
));
3131 die("Unknown mark: %s", command_buf
.buf
);
3132 oidcpy(&oid
, &e
->idx
.oid
);
3133 } else { /* <sha1> */
3134 if (parse_mapped_oid_hex(*p
, &oid
, p
))
3135 die("Invalid dataref: %s", command_buf
.buf
);
3136 e
= find_object(&oid
);
3138 die("Missing space after tree-ish: %s", command_buf
.buf
);
3141 while (!e
|| e
->type
!= OBJ_TREE
)
3142 e
= dereference(e
, &oid
);
3146 static void print_ls(int mode
, const unsigned char *hash
, const char *path
)
3148 static struct strbuf line
= STRBUF_INIT
;
3150 /* See show_tree(). */
3152 S_ISGITLINK(mode
) ? commit_type
:
3153 S_ISDIR(mode
) ? tree_type
:
3157 /* missing SP path LF */
3158 strbuf_reset(&line
);
3159 strbuf_addstr(&line
, "missing ");
3160 quote_c_style(path
, &line
, NULL
, 0);
3161 strbuf_addch(&line
, '\n');
3163 /* mode SP type SP object_name TAB path LF */
3164 strbuf_reset(&line
);
3165 strbuf_addf(&line
, "%06o %s %s\t",
3166 mode
& ~NO_DELTA
, type
, hash_to_hex(hash
));
3167 quote_c_style(path
, &line
, NULL
, 0);
3168 strbuf_addch(&line
, '\n');
3170 cat_blob_write(line
.buf
, line
.len
);
3173 static void parse_ls(const char *p
, struct branch
*b
)
3175 static struct strbuf path
= STRBUF_INIT
;
3176 struct tree_entry
*root
= NULL
;
3177 struct tree_entry leaf
= {NULL
};
3179 /* ls SP (<tree-ish> SP)? <path> */
3182 die("Not in a commit: %s", command_buf
.buf
);
3183 root
= &b
->branch_tree
;
3185 struct object_entry
*e
= parse_treeish_dataref(&p
);
3186 root
= new_tree_entry();
3187 oidcpy(&root
->versions
[1].oid
, &e
->idx
.oid
);
3188 if (!is_null_oid(&root
->versions
[1].oid
))
3189 root
->versions
[1].mode
= S_IFDIR
;
3192 strbuf_reset(&path
);
3193 parse_path_eol(&path
, p
, "path");
3194 tree_content_get(root
, path
.buf
, &leaf
, 1);
3196 * A directory in preparation would have a sha1 of zero
3197 * until it is saved. Save, for simplicity.
3199 if (S_ISDIR(leaf
.versions
[1].mode
))
3202 print_ls(leaf
.versions
[1].mode
, leaf
.versions
[1].oid
.hash
, path
.buf
);
3204 release_tree_content_recursive(leaf
.tree
);
3205 if (!b
|| root
!= &b
->branch_tree
)
3206 release_tree_entry(root
);
3209 static void checkpoint(void)
3211 checkpoint_requested
= 0;
3220 static void parse_checkpoint(void)
3222 checkpoint_requested
= 1;
3226 static void parse_progress(void)
3228 fwrite(command_buf
.buf
, 1, command_buf
.len
, stdout
);
3229 fputc('\n', stdout
);
3234 static void parse_alias(void)
3236 struct object_entry
*e
;
3240 read_next_command();
3245 die(_("Expected 'mark' command, got %s"), command_buf
.buf
);
3248 memset(&b
, 0, sizeof(b
));
3249 if (!parse_objectish_with_prefix(&b
, "to "))
3250 die(_("Expected 'to' command, got %s"), command_buf
.buf
);
3251 e
= find_object(&b
.oid
);
3253 insert_mark(&marks
, next_mark
, e
);
3256 static char* make_fast_import_path(const char *path
)
3258 if (!relative_marks_paths
|| is_absolute_path(path
))
3259 return prefix_filename(global_prefix
, path
);
3260 return git_pathdup("info/fast-import/%s", path
);
3263 static void option_import_marks(const char *marks
,
3264 int from_stream
, int ignore_missing
)
3266 if (import_marks_file
) {
3268 die("Only one import-marks command allowed per stream");
3270 /* read previous mark file */
3271 if(!import_marks_file_from_stream
)
3275 free(import_marks_file
);
3276 import_marks_file
= make_fast_import_path(marks
);
3277 import_marks_file_from_stream
= from_stream
;
3278 import_marks_file_ignore_missing
= ignore_missing
;
3281 static void option_date_format(const char *fmt
)
3283 if (!strcmp(fmt
, "raw"))
3284 whenspec
= WHENSPEC_RAW
;
3285 else if (!strcmp(fmt
, "raw-permissive"))
3286 whenspec
= WHENSPEC_RAW_PERMISSIVE
;
3287 else if (!strcmp(fmt
, "rfc2822"))
3288 whenspec
= WHENSPEC_RFC2822
;
3289 else if (!strcmp(fmt
, "now"))
3290 whenspec
= WHENSPEC_NOW
;
3292 die("unknown --date-format argument %s", fmt
);
3295 static unsigned long ulong_arg(const char *option
, const char *arg
)
3298 unsigned long rv
= strtoul(arg
, &endptr
, 0);
3299 if (strchr(arg
, '-') || endptr
== arg
|| *endptr
)
3300 die("%s: argument must be a non-negative integer", option
);
3304 static void option_depth(const char *depth
)
3306 max_depth
= ulong_arg("--depth", depth
);
3307 if (max_depth
> MAX_DEPTH
)
3308 die("--depth cannot exceed %u", MAX_DEPTH
);
3311 static void option_active_branches(const char *branches
)
3313 max_active_branches
= ulong_arg("--active-branches", branches
);
3316 static void option_export_marks(const char *marks
)
3318 free(export_marks_file
);
3319 export_marks_file
= make_fast_import_path(marks
);
3322 static void option_cat_blob_fd(const char *fd
)
3324 unsigned long n
= ulong_arg("--cat-blob-fd", fd
);
3325 if (n
> (unsigned long) INT_MAX
)
3326 die("--cat-blob-fd cannot exceed %d", INT_MAX
);
3327 cat_blob_fd
= (int) n
;
3330 static void option_export_pack_edges(const char *edges
)
3332 char *fn
= prefix_filename(global_prefix
, edges
);
3335 pack_edges
= xfopen(fn
, "a");
3339 static void option_rewrite_submodules(const char *arg
, struct string_list
*list
)
3341 struct mark_set
*ms
;
3343 char *s
= xstrdup(arg
);
3344 char *f
= strchr(s
, ':');
3346 die(_("Expected format name:filename for submodule rewrite option"));
3349 CALLOC_ARRAY(ms
, 1);
3351 f
= prefix_filename(global_prefix
, f
);
3354 die_errno("cannot read '%s'", f
);
3355 read_mark_file(&ms
, fp
, insert_oid_entry
);
3359 string_list_insert(list
, s
)->util
= ms
;
3364 static int parse_one_option(const char *option
)
3366 if (skip_prefix(option
, "max-pack-size=", &option
)) {
3368 if (!git_parse_ulong(option
, &v
))
3371 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v
);
3373 } else if (v
< 1024 * 1024) {
3374 warning("minimum max-pack-size is 1 MiB");
3378 } else if (skip_prefix(option
, "big-file-threshold=", &option
)) {
3380 if (!git_parse_ulong(option
, &v
))
3382 big_file_threshold
= v
;
3383 } else if (skip_prefix(option
, "depth=", &option
)) {
3384 option_depth(option
);
3385 } else if (skip_prefix(option
, "active-branches=", &option
)) {
3386 option_active_branches(option
);
3387 } else if (skip_prefix(option
, "export-pack-edges=", &option
)) {
3388 option_export_pack_edges(option
);
3389 } else if (!strcmp(option
, "quiet")) {
3391 } else if (!strcmp(option
, "stats")) {
3393 } else if (!strcmp(option
, "allow-unsafe-features")) {
3394 ; /* already handled during early option parsing */
3402 static void check_unsafe_feature(const char *feature
, int from_stream
)
3404 if (from_stream
&& !allow_unsafe_features
)
3405 die(_("feature '%s' forbidden in input without --allow-unsafe-features"),
3409 static int parse_one_feature(const char *feature
, int from_stream
)
3413 if (skip_prefix(feature
, "date-format=", &arg
)) {
3414 option_date_format(arg
);
3415 } else if (skip_prefix(feature
, "import-marks=", &arg
)) {
3416 check_unsafe_feature("import-marks", from_stream
);
3417 option_import_marks(arg
, from_stream
, 0);
3418 } else if (skip_prefix(feature
, "import-marks-if-exists=", &arg
)) {
3419 check_unsafe_feature("import-marks-if-exists", from_stream
);
3420 option_import_marks(arg
, from_stream
, 1);
3421 } else if (skip_prefix(feature
, "export-marks=", &arg
)) {
3422 check_unsafe_feature(feature
, from_stream
);
3423 option_export_marks(arg
);
3424 } else if (!strcmp(feature
, "alias")) {
3425 ; /* Don't die - this feature is supported */
3426 } else if (skip_prefix(feature
, "rewrite-submodules-to=", &arg
)) {
3427 option_rewrite_submodules(arg
, &sub_marks_to
);
3428 } else if (skip_prefix(feature
, "rewrite-submodules-from=", &arg
)) {
3429 option_rewrite_submodules(arg
, &sub_marks_from
);
3430 } else if (!strcmp(feature
, "get-mark")) {
3431 ; /* Don't die - this feature is supported */
3432 } else if (!strcmp(feature
, "cat-blob")) {
3433 ; /* Don't die - this feature is supported */
3434 } else if (!strcmp(feature
, "relative-marks")) {
3435 relative_marks_paths
= 1;
3436 } else if (!strcmp(feature
, "no-relative-marks")) {
3437 relative_marks_paths
= 0;
3438 } else if (!strcmp(feature
, "done")) {
3439 require_explicit_termination
= 1;
3440 } else if (!strcmp(feature
, "force")) {
3442 } else if (!strcmp(feature
, "notes") || !strcmp(feature
, "ls")) {
3443 ; /* do nothing; we have the feature */
3451 static void parse_feature(const char *feature
)
3453 if (seen_data_command
)
3454 die("Got feature command '%s' after data command", feature
);
3456 if (parse_one_feature(feature
, 1))
3459 die("This version of fast-import does not support feature %s.", feature
);
3462 static void parse_option(const char *option
)
3464 if (seen_data_command
)
3465 die("Got option command '%s' after data command", option
);
3467 if (parse_one_option(option
))
3470 die("This version of fast-import does not support option: %s", option
);
3473 static void git_pack_config(void)
3475 int indexversion_value
;
3477 unsigned long packsizelimit_value
;
3479 if (!git_config_get_ulong("pack.depth", &max_depth
)) {
3480 if (max_depth
> MAX_DEPTH
)
3481 max_depth
= MAX_DEPTH
;
3483 if (!git_config_get_int("pack.indexversion", &indexversion_value
)) {
3484 pack_idx_opts
.version
= indexversion_value
;
3485 if (pack_idx_opts
.version
> 2)
3486 git_die_config(the_repository
, "pack.indexversion",
3487 "bad pack.indexVersion=%"PRIu32
, pack_idx_opts
.version
);
3489 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value
))
3490 max_packsize
= packsizelimit_value
;
3492 if (!git_config_get_int("fastimport.unpacklimit", &limit
))
3493 unpack_limit
= limit
;
3494 else if (!git_config_get_int("transfer.unpacklimit", &limit
))
3495 unpack_limit
= limit
;
3497 git_config(git_default_config
, NULL
);
3500 static const char fast_import_usage
[] =
3501 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3503 static void parse_argv(void)
3507 for (i
= 1; i
< global_argc
; i
++) {
3508 const char *a
= global_argv
[i
];
3510 if (*a
!= '-' || !strcmp(a
, "--"))
3513 if (!skip_prefix(a
, "--", &a
))
3514 die("unknown option %s", a
);
3516 if (parse_one_option(a
))
3519 if (parse_one_feature(a
, 0))
3522 if (skip_prefix(a
, "cat-blob-fd=", &a
)) {
3523 option_cat_blob_fd(a
);
3527 die("unknown option --%s", a
);
3529 if (i
!= global_argc
)
3530 usage(fast_import_usage
);
3532 seen_data_command
= 1;
3533 if (import_marks_file
)
3535 build_mark_map(&sub_marks_from
, &sub_marks_to
);
3538 int cmd_fast_import(int argc
,
3541 struct repository
*repo UNUSED
)
3545 if (argc
== 2 && !strcmp(argv
[1], "-h"))
3546 usage(fast_import_usage
);
3548 reset_pack_idx_option(&pack_idx_opts
);
3551 alloc_objects(object_entry_alloc
);
3552 strbuf_init(&command_buf
, 0);
3553 CALLOC_ARRAY(atom_table
, atom_table_sz
);
3554 CALLOC_ARRAY(branch_table
, branch_table_sz
);
3555 CALLOC_ARRAY(avail_tree_table
, avail_tree_table_sz
);
3556 marks
= mem_pool_calloc(&fi_mem_pool
, 1, sizeof(struct mark_set
));
3558 hashmap_init(&object_table
, object_entry_hashcmp
, NULL
, 0);
3561 * We don't parse most options until after we've seen the set of
3562 * "feature" lines at the start of the stream (which allows the command
3563 * line to override stream data). But we must do an early parse of any
3564 * command-line options that impact how we interpret the feature lines.
3566 for (i
= 1; i
< argc
; i
++) {
3567 const char *arg
= argv
[i
];
3568 if (*arg
!= '-' || !strcmp(arg
, "--"))
3570 if (!strcmp(arg
, "--allow-unsafe-features"))
3571 allow_unsafe_features
= 1;
3576 global_prefix
= prefix
;
3578 rc_free
= mem_pool_alloc(&fi_mem_pool
, cmd_save
* sizeof(*rc_free
));
3579 for (i
= 0; i
< (cmd_save
- 1); i
++)
3580 rc_free
[i
].next
= &rc_free
[i
+ 1];
3581 rc_free
[cmd_save
- 1].next
= NULL
;
3584 set_die_routine(die_nicely
);
3585 set_checkpoint_signal();
3586 while (read_next_command() != EOF
) {
3588 if (!strcmp("blob", command_buf
.buf
))
3590 else if (skip_prefix(command_buf
.buf
, "commit ", &v
))
3591 parse_new_commit(v
);
3592 else if (skip_prefix(command_buf
.buf
, "tag ", &v
))
3594 else if (skip_prefix(command_buf
.buf
, "reset ", &v
))
3595 parse_reset_branch(v
);
3596 else if (skip_prefix(command_buf
.buf
, "ls ", &v
))
3598 else if (skip_prefix(command_buf
.buf
, "cat-blob ", &v
))
3600 else if (skip_prefix(command_buf
.buf
, "get-mark ", &v
))
3602 else if (!strcmp("checkpoint", command_buf
.buf
))
3604 else if (!strcmp("done", command_buf
.buf
))
3606 else if (!strcmp("alias", command_buf
.buf
))
3608 else if (starts_with(command_buf
.buf
, "progress "))
3610 else if (skip_prefix(command_buf
.buf
, "feature ", &v
))
3612 else if (skip_prefix(command_buf
.buf
, "option git ", &v
))
3614 else if (starts_with(command_buf
.buf
, "option "))
3615 /* ignore non-git options*/;
3617 die("Unsupported command: %s", command_buf
.buf
);
3619 if (checkpoint_requested
)
3623 /* argv hasn't been parsed yet, do so */
3624 if (!seen_data_command
)
3627 if (require_explicit_termination
&& feof(stdin
))
3628 die("stream ends early");
3641 uintmax_t total_count
= 0, duplicate_count
= 0;
3642 for (i
= 0; i
< ARRAY_SIZE(object_count_by_type
); i
++)
3643 total_count
+= object_count_by_type
[i
];
3644 for (i
= 0; i
< ARRAY_SIZE(duplicate_count_by_type
); i
++)
3645 duplicate_count
+= duplicate_count_by_type
[i
];
3647 fprintf(stderr
, "%s statistics:\n", argv
[0]);
3648 fprintf(stderr
, "---------------------------------------------------------------------\n");
3649 fprintf(stderr
, "Alloc'd objects: %10" PRIuMAX
"\n", alloc_count
);
3650 fprintf(stderr
, "Total objects: %10" PRIuMAX
" (%10" PRIuMAX
" duplicates )\n", total_count
, duplicate_count
);
3651 fprintf(stderr
, " blobs : %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas of %10" PRIuMAX
" attempts)\n", object_count_by_type
[OBJ_BLOB
], duplicate_count_by_type
[OBJ_BLOB
], delta_count_by_type
[OBJ_BLOB
], delta_count_attempts_by_type
[OBJ_BLOB
]);
3652 fprintf(stderr
, " trees : %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas of %10" PRIuMAX
" attempts)\n", object_count_by_type
[OBJ_TREE
], duplicate_count_by_type
[OBJ_TREE
], delta_count_by_type
[OBJ_TREE
], delta_count_attempts_by_type
[OBJ_TREE
]);
3653 fprintf(stderr
, " commits: %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas of %10" PRIuMAX
" attempts)\n", object_count_by_type
[OBJ_COMMIT
], duplicate_count_by_type
[OBJ_COMMIT
], delta_count_by_type
[OBJ_COMMIT
], delta_count_attempts_by_type
[OBJ_COMMIT
]);
3654 fprintf(stderr
, " tags : %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas of %10" PRIuMAX
" attempts)\n", object_count_by_type
[OBJ_TAG
], duplicate_count_by_type
[OBJ_TAG
], delta_count_by_type
[OBJ_TAG
], delta_count_attempts_by_type
[OBJ_TAG
]);
3655 fprintf(stderr
, "Total branches: %10lu (%10lu loads )\n", branch_count
, branch_load_count
);
3656 fprintf(stderr
, " marks: %10" PRIuMAX
" (%10" PRIuMAX
" unique )\n", (((uintmax_t)1) << marks
->shift
) * 1024, marks_set_count
);
3657 fprintf(stderr
, " atoms: %10u\n", atom_cnt
);
3658 fprintf(stderr
, "Memory total: %10" PRIuMAX
" KiB\n", (tree_entry_allocd
+ fi_mem_pool
.pool_alloc
+ alloc_count
*sizeof(struct object_entry
))/1024);
3659 fprintf(stderr
, " pools: %10lu KiB\n", (unsigned long)((tree_entry_allocd
+ fi_mem_pool
.pool_alloc
) /1024));
3660 fprintf(stderr
, " objects: %10" PRIuMAX
" KiB\n", (alloc_count
*sizeof(struct object_entry
))/1024);
3661 fprintf(stderr
, "---------------------------------------------------------------------\n");
3663 fprintf(stderr
, "---------------------------------------------------------------------\n");
3664 fprintf(stderr
, "\n");
3667 return failure
? 1 : 0;