2 Format of STDIN stream:
14 new_blob ::= 'blob' lf
17 file_content ::= data;
19 new_commit ::= 'commit' sp ref_str lf
21 ('author' sp name '<' email '>' when lf)?
22 'committer' sp name '<' email '>' when lf
24 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
25 ('merge' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)*
30 file_change ::= file_clr
36 file_clr ::= 'deleteall' lf;
37 file_del ::= 'D' sp path_str lf;
38 file_rnm ::= 'R' sp path_str sp path_str lf;
39 file_cpy ::= 'C' sp path_str sp path_str lf;
40 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
41 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
44 new_tag ::= 'tag' sp tag_str lf
45 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
46 'tagger' sp name '<' email '>' when lf
50 reset_branch ::= 'reset' sp ref_str lf
51 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
54 checkpoint ::= 'checkpoint' lf
57 progress ::= 'progress' sp not_lf* lf
60 # note: the first idnum in a stream should be 1 and subsequent
61 # idnums should not have gaps between values as this will cause
62 # the stream parser to reserve space for the gapped values. An
63 # idnum can be updated in the future to a new object by issuing
64 # a new mark directive with the old idnum.
66 mark ::= 'mark' sp idnum lf;
67 data ::= (delimited_data | exact_data)
70 # note: delim may be any string but must not contain lf.
71 # data_line may contain any data but must not be exactly
73 delimited_data ::= 'data' sp '<<' delim lf
77 # note: declen indicates the length of binary_data in bytes.
78 # declen does not include the lf preceeding the binary data.
80 exact_data ::= 'data' sp declen lf
83 # note: quoted strings are C-style quoting supporting \c for
84 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
85 # is the signed byte value in octal. Note that the only
86 # characters which must actually be escaped to protect the
87 # stream formatting is: \, " and LF. Otherwise these values
91 sha1exp_str ::= sha1exp;
93 path_str ::= path | '"' quoted(path) '"' ;
94 mode ::= '100644' | '644'
99 declen ::= # unsigned 32 bit value, ascii base10 notation;
100 bigint ::= # unsigned integer value, ascii base10 notation;
101 binary_data ::= # file content, not interpreted;
103 when ::= raw_when | rfc2822_when;
104 raw_when ::= ts sp tz;
105 rfc2822_when ::= # Valid RFC 2822 date and time;
107 sp ::= # ASCII space character;
108 lf ::= # ASCII newline (LF) character;
110 # note: a colon (':') must precede the numerical value assigned to
111 # an idnum. This is to distinguish it from a ref or tag name as
112 # GIT does not permit ':' in ref or tag strings.
114 idnum ::= ':' bigint;
115 path ::= # GIT style file path, e.g. "a/b/c";
116 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
117 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
118 sha1exp ::= # Any valid GIT SHA1 expression;
119 hexsha1 ::= # SHA1 in hexadecimal format;
121 # note: name and email are UTF8 strings, however name must not
122 # contain '<' or lf and email must not contain any of the
123 # following: '<', '>', lf.
125 name ::= # valid GIT author/committer name;
126 email ::= # valid GIT author/committer email;
127 ts ::= # time since the epoch in seconds, ascii base10 notation;
128 tz ::= # GIT style timezone;
130 # note: comments may appear anywhere in the input, except
131 # within a data command. Any form of the data command
132 # always escapes the related input from comment processing.
134 # In case it is not clear, the '#' that starts the comment
135 # must be the first character on that the line (an lf have
138 comment ::= '#' not_lf* lf;
139 not_lf ::= # Any byte that is not ASCII newline (LF);
151 #include "csum-file.h"
155 #define PACK_ID_BITS 16
156 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
160 struct object_entry
*next
;
162 unsigned type
: TYPE_BITS
;
163 unsigned pack_id
: PACK_ID_BITS
;
164 unsigned char sha1
[20];
167 struct object_entry_pool
169 struct object_entry_pool
*next_pool
;
170 struct object_entry
*next_free
;
171 struct object_entry
*end
;
172 struct object_entry entries
[FLEX_ARRAY
]; /* more */
178 struct object_entry
*marked
[1024];
179 struct mark_set
*sets
[1024];
195 struct mem_pool
*next_pool
;
198 char space
[FLEX_ARRAY
]; /* more */
203 struct atom_str
*next_atom
;
204 unsigned short str_len
;
205 char str_dat
[FLEX_ARRAY
]; /* more */
211 struct tree_content
*tree
;
212 struct atom_str
* name
;
216 unsigned char sha1
[20];
222 unsigned int entry_capacity
; /* must match avail_tree_content */
223 unsigned int entry_count
;
224 unsigned int delta_depth
;
225 struct tree_entry
*entries
[FLEX_ARRAY
]; /* more */
228 struct avail_tree_content
230 unsigned int entry_capacity
; /* must match tree_content */
231 struct avail_tree_content
*next_avail
;
236 struct branch
*table_next_branch
;
237 struct branch
*active_next_branch
;
239 struct tree_entry branch_tree
;
240 uintmax_t last_commit
;
242 unsigned pack_id
: PACK_ID_BITS
;
243 unsigned char sha1
[20];
248 struct tag
*next_tag
;
250 unsigned int pack_id
;
251 unsigned char sha1
[20];
262 struct hash_list
*next
;
263 unsigned char sha1
[20];
272 /* Configured limits on output */
273 static unsigned long max_depth
= 10;
274 static off_t max_packsize
= (1LL << 32) - 1;
275 static int force_update
;
277 /* Stats and misc. counters */
278 static uintmax_t alloc_count
;
279 static uintmax_t marks_set_count
;
280 static uintmax_t object_count_by_type
[1 << TYPE_BITS
];
281 static uintmax_t duplicate_count_by_type
[1 << TYPE_BITS
];
282 static uintmax_t delta_count_by_type
[1 << TYPE_BITS
];
283 static unsigned long object_count
;
284 static unsigned long branch_count
;
285 static unsigned long branch_load_count
;
287 static FILE *pack_edges
;
290 static size_t mem_pool_alloc
= 2*1024*1024 - sizeof(struct mem_pool
);
291 static size_t total_allocd
;
292 static struct mem_pool
*mem_pool
;
294 /* Atom management */
295 static unsigned int atom_table_sz
= 4451;
296 static unsigned int atom_cnt
;
297 static struct atom_str
**atom_table
;
299 /* The .pack file being generated */
300 static unsigned int pack_id
;
301 static struct packed_git
*pack_data
;
302 static struct packed_git
**all_packs
;
303 static unsigned long pack_size
;
305 /* Table of objects we've written. */
306 static unsigned int object_entry_alloc
= 5000;
307 static struct object_entry_pool
*blocks
;
308 static struct object_entry
*object_table
[1 << 16];
309 static struct mark_set
*marks
;
310 static const char* mark_file
;
313 static struct last_object last_blob
;
315 /* Tree management */
316 static unsigned int tree_entry_alloc
= 1000;
317 static void *avail_tree_entry
;
318 static unsigned int avail_tree_table_sz
= 100;
319 static struct avail_tree_content
**avail_tree_table
;
320 static struct dbuf old_tree
;
321 static struct dbuf new_tree
;
324 static unsigned long max_active_branches
= 5;
325 static unsigned long cur_active_branches
;
326 static unsigned long branch_table_sz
= 1039;
327 static struct branch
**branch_table
;
328 static struct branch
*active_branches
;
331 static struct tag
*first_tag
;
332 static struct tag
*last_tag
;
334 /* Input stream parsing */
335 static whenspec_type whenspec
= WHENSPEC_RAW
;
336 static struct strbuf command_buf
;
337 static int unread_command_buf
;
338 static uintmax_t next_mark
;
339 static struct dbuf new_data
;
341 static void write_branch_report(FILE *rpt
, struct branch
*b
)
343 fprintf(rpt
, "%s:\n", b
->name
);
345 fprintf(rpt
, " status :");
347 fputs(" active", rpt
);
348 if (b
->branch_tree
.tree
)
349 fputs(" loaded", rpt
);
350 if (is_null_sha1(b
->branch_tree
.versions
[1].sha1
))
351 fputs(" dirty", rpt
);
354 fprintf(rpt
, " tip commit : %s\n", sha1_to_hex(b
->sha1
));
355 fprintf(rpt
, " old tree : %s\n", sha1_to_hex(b
->branch_tree
.versions
[0].sha1
));
356 fprintf(rpt
, " cur tree : %s\n", sha1_to_hex(b
->branch_tree
.versions
[1].sha1
));
357 fprintf(rpt
, " commit clock: %" PRIuMAX
"\n", b
->last_commit
);
359 fputs(" last pack : ", rpt
);
360 if (b
->pack_id
< MAX_PACK_ID
)
361 fprintf(rpt
, "%u", b
->pack_id
);
367 static void write_crash_report(const char *err
, va_list params
)
369 char *loc
= git_path("fast_import_crash_%d", getpid());
370 FILE *rpt
= fopen(loc
, "w");
375 error("can't write crash report %s: %s", loc
, strerror(errno
));
379 fprintf(stderr
, "fast-import: dumping crash report to %s\n", loc
);
381 fprintf(rpt
, "fast-import crash report:\n");
382 fprintf(rpt
, " fast-import process: %d\n", getpid());
383 fprintf(rpt
, " parent process : %d\n", getppid());
384 fprintf(rpt
, " at %s\n", show_date(time(NULL
), 0, DATE_LOCAL
));
387 fputs("fatal: ", rpt
);
388 vfprintf(rpt
, err
, params
);
392 fputs("Active Branch LRU\n", rpt
);
393 fputs("-----------------\n", rpt
);
394 fprintf(rpt
, " active_branches = %lu cur, %lu max\n",
396 max_active_branches
);
398 fputs(" pos clock name\n", rpt
);
399 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt
);
400 for (b
= active_branches
, lu
= 0; b
; b
= b
->active_next_branch
)
401 fprintf(rpt
, " %2lu) %6" PRIuMAX
" %s\n",
402 ++lu
, b
->last_commit
, b
->name
);
405 fputs("Inactive Branches\n", rpt
);
406 fputs("-----------------\n", rpt
);
407 for (lu
= 0; lu
< branch_table_sz
; lu
++) {
408 for (b
= branch_table
[lu
]; b
; b
= b
->table_next_branch
)
409 write_branch_report(rpt
, b
);
413 fputs("-------------------\n", rpt
);
414 fputs("END OF CRASH REPORT\n", rpt
);
418 static NORETURN
void die_nicely(const char *err
, va_list params
)
422 fputs("fatal: ", stderr
);
423 vfprintf(stderr
, err
, params
);
428 write_crash_report(err
, params
);
434 static void alloc_objects(unsigned int cnt
)
436 struct object_entry_pool
*b
;
438 b
= xmalloc(sizeof(struct object_entry_pool
)
439 + cnt
* sizeof(struct object_entry
));
440 b
->next_pool
= blocks
;
441 b
->next_free
= b
->entries
;
442 b
->end
= b
->entries
+ cnt
;
447 static struct object_entry
*new_object(unsigned char *sha1
)
449 struct object_entry
*e
;
451 if (blocks
->next_free
== blocks
->end
)
452 alloc_objects(object_entry_alloc
);
454 e
= blocks
->next_free
++;
455 hashcpy(e
->sha1
, sha1
);
459 static struct object_entry
*find_object(unsigned char *sha1
)
461 unsigned int h
= sha1
[0] << 8 | sha1
[1];
462 struct object_entry
*e
;
463 for (e
= object_table
[h
]; e
; e
= e
->next
)
464 if (!hashcmp(sha1
, e
->sha1
))
469 static struct object_entry
*insert_object(unsigned char *sha1
)
471 unsigned int h
= sha1
[0] << 8 | sha1
[1];
472 struct object_entry
*e
= object_table
[h
];
473 struct object_entry
*p
= NULL
;
476 if (!hashcmp(sha1
, e
->sha1
))
482 e
= new_object(sha1
);
492 static unsigned int hc_str(const char *s
, size_t len
)
500 static void *pool_alloc(size_t len
)
505 for (p
= mem_pool
; p
; p
= p
->next_pool
)
506 if ((p
->end
- p
->next_free
>= len
))
510 if (len
>= (mem_pool_alloc
/2)) {
514 total_allocd
+= sizeof(struct mem_pool
) + mem_pool_alloc
;
515 p
= xmalloc(sizeof(struct mem_pool
) + mem_pool_alloc
);
516 p
->next_pool
= mem_pool
;
517 p
->next_free
= p
->space
;
518 p
->end
= p
->next_free
+ mem_pool_alloc
;
523 /* round out to a pointer alignment */
524 if (len
& (sizeof(void*) - 1))
525 len
+= sizeof(void*) - (len
& (sizeof(void*) - 1));
530 static void *pool_calloc(size_t count
, size_t size
)
532 size_t len
= count
* size
;
533 void *r
= pool_alloc(len
);
538 static char *pool_strdup(const char *s
)
540 char *r
= pool_alloc(strlen(s
) + 1);
545 static void size_dbuf(struct dbuf
*b
, size_t maxlen
)
548 if (b
->capacity
>= maxlen
)
552 b
->capacity
= ((maxlen
/ 1024) + 1) * 1024;
553 b
->buffer
= xmalloc(b
->capacity
);
556 static void insert_mark(uintmax_t idnum
, struct object_entry
*oe
)
558 struct mark_set
*s
= marks
;
559 while ((idnum
>> s
->shift
) >= 1024) {
560 s
= pool_calloc(1, sizeof(struct mark_set
));
561 s
->shift
= marks
->shift
+ 10;
562 s
->data
.sets
[0] = marks
;
566 uintmax_t i
= idnum
>> s
->shift
;
567 idnum
-= i
<< s
->shift
;
568 if (!s
->data
.sets
[i
]) {
569 s
->data
.sets
[i
] = pool_calloc(1, sizeof(struct mark_set
));
570 s
->data
.sets
[i
]->shift
= s
->shift
- 10;
574 if (!s
->data
.marked
[idnum
])
576 s
->data
.marked
[idnum
] = oe
;
579 static struct object_entry
*find_mark(uintmax_t idnum
)
581 uintmax_t orig_idnum
= idnum
;
582 struct mark_set
*s
= marks
;
583 struct object_entry
*oe
= NULL
;
584 if ((idnum
>> s
->shift
) < 1024) {
585 while (s
&& s
->shift
) {
586 uintmax_t i
= idnum
>> s
->shift
;
587 idnum
-= i
<< s
->shift
;
591 oe
= s
->data
.marked
[idnum
];
594 die("mark :%" PRIuMAX
" not declared", orig_idnum
);
598 static struct atom_str
*to_atom(const char *s
, unsigned short len
)
600 unsigned int hc
= hc_str(s
, len
) % atom_table_sz
;
603 for (c
= atom_table
[hc
]; c
; c
= c
->next_atom
)
604 if (c
->str_len
== len
&& !strncmp(s
, c
->str_dat
, len
))
607 c
= pool_alloc(sizeof(struct atom_str
) + len
+ 1);
609 strncpy(c
->str_dat
, s
, len
);
611 c
->next_atom
= atom_table
[hc
];
617 static struct branch
*lookup_branch(const char *name
)
619 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
622 for (b
= branch_table
[hc
]; b
; b
= b
->table_next_branch
)
623 if (!strcmp(name
, b
->name
))
628 static struct branch
*new_branch(const char *name
)
630 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
631 struct branch
* b
= lookup_branch(name
);
634 die("Invalid attempt to create duplicate branch: %s", name
);
635 switch (check_ref_format(name
)) {
636 case 0: break; /* its valid */
637 case -2: break; /* valid, but too few '/', allow anyway */
639 die("Branch name doesn't conform to GIT standards: %s", name
);
642 b
= pool_calloc(1, sizeof(struct branch
));
643 b
->name
= pool_strdup(name
);
644 b
->table_next_branch
= branch_table
[hc
];
645 b
->branch_tree
.versions
[0].mode
= S_IFDIR
;
646 b
->branch_tree
.versions
[1].mode
= S_IFDIR
;
648 b
->pack_id
= MAX_PACK_ID
;
649 branch_table
[hc
] = b
;
654 static unsigned int hc_entries(unsigned int cnt
)
656 cnt
= cnt
& 7 ? (cnt
/ 8) + 1 : cnt
/ 8;
657 return cnt
< avail_tree_table_sz
? cnt
: avail_tree_table_sz
- 1;
660 static struct tree_content
*new_tree_content(unsigned int cnt
)
662 struct avail_tree_content
*f
, *l
= NULL
;
663 struct tree_content
*t
;
664 unsigned int hc
= hc_entries(cnt
);
666 for (f
= avail_tree_table
[hc
]; f
; l
= f
, f
= f
->next_avail
)
667 if (f
->entry_capacity
>= cnt
)
672 l
->next_avail
= f
->next_avail
;
674 avail_tree_table
[hc
] = f
->next_avail
;
676 cnt
= cnt
& 7 ? ((cnt
/ 8) + 1) * 8 : cnt
;
677 f
= pool_alloc(sizeof(*t
) + sizeof(t
->entries
[0]) * cnt
);
678 f
->entry_capacity
= cnt
;
681 t
= (struct tree_content
*)f
;
687 static void release_tree_entry(struct tree_entry
*e
);
688 static void release_tree_content(struct tree_content
*t
)
690 struct avail_tree_content
*f
= (struct avail_tree_content
*)t
;
691 unsigned int hc
= hc_entries(f
->entry_capacity
);
692 f
->next_avail
= avail_tree_table
[hc
];
693 avail_tree_table
[hc
] = f
;
696 static void release_tree_content_recursive(struct tree_content
*t
)
699 for (i
= 0; i
< t
->entry_count
; i
++)
700 release_tree_entry(t
->entries
[i
]);
701 release_tree_content(t
);
704 static struct tree_content
*grow_tree_content(
705 struct tree_content
*t
,
708 struct tree_content
*r
= new_tree_content(t
->entry_count
+ amt
);
709 r
->entry_count
= t
->entry_count
;
710 r
->delta_depth
= t
->delta_depth
;
711 memcpy(r
->entries
,t
->entries
,t
->entry_count
*sizeof(t
->entries
[0]));
712 release_tree_content(t
);
716 static struct tree_entry
*new_tree_entry(void)
718 struct tree_entry
*e
;
720 if (!avail_tree_entry
) {
721 unsigned int n
= tree_entry_alloc
;
722 total_allocd
+= n
* sizeof(struct tree_entry
);
723 avail_tree_entry
= e
= xmalloc(n
* sizeof(struct tree_entry
));
725 *((void**)e
) = e
+ 1;
731 e
= avail_tree_entry
;
732 avail_tree_entry
= *((void**)e
);
736 static void release_tree_entry(struct tree_entry
*e
)
739 release_tree_content_recursive(e
->tree
);
740 *((void**)e
) = avail_tree_entry
;
741 avail_tree_entry
= e
;
744 static struct tree_content
*dup_tree_content(struct tree_content
*s
)
746 struct tree_content
*d
;
747 struct tree_entry
*a
, *b
;
752 d
= new_tree_content(s
->entry_count
);
753 for (i
= 0; i
< s
->entry_count
; i
++) {
755 b
= new_tree_entry();
756 memcpy(b
, a
, sizeof(*a
));
757 if (a
->tree
&& is_null_sha1(b
->versions
[1].sha1
))
758 b
->tree
= dup_tree_content(a
->tree
);
763 d
->entry_count
= s
->entry_count
;
764 d
->delta_depth
= s
->delta_depth
;
769 static void start_packfile(void)
771 static char tmpfile
[PATH_MAX
];
772 struct packed_git
*p
;
773 struct pack_header hdr
;
776 snprintf(tmpfile
, sizeof(tmpfile
),
777 "%s/tmp_pack_XXXXXX", get_object_directory());
778 pack_fd
= xmkstemp(tmpfile
);
779 p
= xcalloc(1, sizeof(*p
) + strlen(tmpfile
) + 2);
780 strcpy(p
->pack_name
, tmpfile
);
781 p
->pack_fd
= pack_fd
;
783 hdr
.hdr_signature
= htonl(PACK_SIGNATURE
);
784 hdr
.hdr_version
= htonl(2);
786 write_or_die(p
->pack_fd
, &hdr
, sizeof(hdr
));
789 pack_size
= sizeof(hdr
);
792 all_packs
= xrealloc(all_packs
, sizeof(*all_packs
) * (pack_id
+ 1));
793 all_packs
[pack_id
] = p
;
796 static int oecmp (const void *a_
, const void *b_
)
798 struct object_entry
*a
= *((struct object_entry
**)a_
);
799 struct object_entry
*b
= *((struct object_entry
**)b_
);
800 return hashcmp(a
->sha1
, b
->sha1
);
803 static char *create_index(void)
805 static char tmpfile
[PATH_MAX
];
808 struct object_entry
**idx
, **c
, **last
, *e
;
809 struct object_entry_pool
*o
;
813 /* Build the sorted table of object IDs. */
814 idx
= xmalloc(object_count
* sizeof(struct object_entry
*));
816 for (o
= blocks
; o
; o
= o
->next_pool
)
817 for (e
= o
->next_free
; e
-- != o
->entries
;)
818 if (pack_id
== e
->pack_id
)
820 last
= idx
+ object_count
;
822 die("internal consistency error creating the index");
823 qsort(idx
, object_count
, sizeof(struct object_entry
*), oecmp
);
825 /* Generate the fan-out array. */
827 for (i
= 0; i
< 256; i
++) {
828 struct object_entry
**next
= c
;;
829 while (next
< last
) {
830 if ((*next
)->sha1
[0] != i
)
834 array
[i
] = htonl(next
- idx
);
838 snprintf(tmpfile
, sizeof(tmpfile
),
839 "%s/tmp_idx_XXXXXX", get_object_directory());
840 idx_fd
= xmkstemp(tmpfile
);
841 f
= sha1fd(idx_fd
, tmpfile
);
842 sha1write(f
, array
, 256 * sizeof(int));
844 for (c
= idx
; c
!= last
; c
++) {
845 uint32_t offset
= htonl((*c
)->offset
);
846 sha1write(f
, &offset
, 4);
847 sha1write(f
, (*c
)->sha1
, sizeof((*c
)->sha1
));
848 SHA1_Update(&ctx
, (*c
)->sha1
, 20);
850 sha1write(f
, pack_data
->sha1
, sizeof(pack_data
->sha1
));
851 sha1close(f
, NULL
, 1);
853 SHA1_Final(pack_data
->sha1
, &ctx
);
857 static char *keep_pack(char *curr_index_name
)
859 static char name
[PATH_MAX
];
860 static const char *keep_msg
= "fast-import";
863 chmod(pack_data
->pack_name
, 0444);
864 chmod(curr_index_name
, 0444);
866 snprintf(name
, sizeof(name
), "%s/pack/pack-%s.keep",
867 get_object_directory(), sha1_to_hex(pack_data
->sha1
));
868 keep_fd
= open(name
, O_RDWR
|O_CREAT
|O_EXCL
, 0600);
870 die("cannot create keep file");
871 write(keep_fd
, keep_msg
, strlen(keep_msg
));
874 snprintf(name
, sizeof(name
), "%s/pack/pack-%s.pack",
875 get_object_directory(), sha1_to_hex(pack_data
->sha1
));
876 if (move_temp_to_file(pack_data
->pack_name
, name
))
877 die("cannot store pack file");
879 snprintf(name
, sizeof(name
), "%s/pack/pack-%s.idx",
880 get_object_directory(), sha1_to_hex(pack_data
->sha1
));
881 if (move_temp_to_file(curr_index_name
, name
))
882 die("cannot store index file");
886 static void unkeep_all_packs(void)
888 static char name
[PATH_MAX
];
891 for (k
= 0; k
< pack_id
; k
++) {
892 struct packed_git
*p
= all_packs
[k
];
893 snprintf(name
, sizeof(name
), "%s/pack/pack-%s.keep",
894 get_object_directory(), sha1_to_hex(p
->sha1
));
899 static void end_packfile(void)
901 struct packed_git
*old_p
= pack_data
, *new_p
;
909 fixup_pack_header_footer(pack_data
->pack_fd
, pack_data
->sha1
,
910 pack_data
->pack_name
, object_count
);
911 close(pack_data
->pack_fd
);
912 idx_name
= keep_pack(create_index());
914 /* Register the packfile with core git's machinary. */
915 new_p
= add_packed_git(idx_name
, strlen(idx_name
), 1);
917 die("core git rejected index %s", idx_name
);
918 new_p
->windows
= old_p
->windows
;
919 all_packs
[pack_id
] = new_p
;
920 install_packed_git(new_p
);
922 /* Print the boundary */
924 fprintf(pack_edges
, "%s:", new_p
->pack_name
);
925 for (i
= 0; i
< branch_table_sz
; i
++) {
926 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
) {
927 if (b
->pack_id
== pack_id
)
928 fprintf(pack_edges
, " %s", sha1_to_hex(b
->sha1
));
931 for (t
= first_tag
; t
; t
= t
->next_tag
) {
932 if (t
->pack_id
== pack_id
)
933 fprintf(pack_edges
, " %s", sha1_to_hex(t
->sha1
));
935 fputc('\n', pack_edges
);
942 unlink(old_p
->pack_name
);
945 /* We can't carry a delta across packfiles. */
946 free(last_blob
.data
);
947 last_blob
.data
= NULL
;
949 last_blob
.offset
= 0;
953 static void cycle_packfile(void)
959 static size_t encode_header(
960 enum object_type type
,
967 if (type
< OBJ_COMMIT
|| type
> OBJ_REF_DELTA
)
968 die("bad type %d", type
);
970 c
= (type
<< 4) | (size
& 15);
982 static int store_object(
983 enum object_type type
,
986 struct last_object
*last
,
987 unsigned char *sha1out
,
991 struct object_entry
*e
;
992 unsigned char hdr
[96];
993 unsigned char sha1
[20];
994 unsigned long hdrlen
, deltalen
;
998 hdrlen
= sprintf((char*)hdr
,"%s %lu", typename(type
),
999 (unsigned long)datlen
) + 1;
1001 SHA1_Update(&c
, hdr
, hdrlen
);
1002 SHA1_Update(&c
, dat
, datlen
);
1003 SHA1_Final(sha1
, &c
);
1005 hashcpy(sha1out
, sha1
);
1007 e
= insert_object(sha1
);
1009 insert_mark(mark
, e
);
1011 duplicate_count_by_type
[type
]++;
1013 } else if (find_sha1_pack(sha1
, packed_git
)) {
1015 e
->pack_id
= MAX_PACK_ID
;
1016 e
->offset
= 1; /* just not zero! */
1017 duplicate_count_by_type
[type
]++;
1021 if (last
&& last
->data
&& last
->depth
< max_depth
) {
1022 delta
= diff_delta(last
->data
, last
->len
,
1025 if (delta
&& deltalen
>= datlen
) {
1032 memset(&s
, 0, sizeof(s
));
1033 deflateInit(&s
, zlib_compression_level
);
1036 s
.avail_in
= deltalen
;
1039 s
.avail_in
= datlen
;
1041 s
.avail_out
= deflateBound(&s
, s
.avail_in
);
1042 s
.next_out
= out
= xmalloc(s
.avail_out
);
1043 while (deflate(&s
, Z_FINISH
) == Z_OK
)
1047 /* Determine if we should auto-checkpoint. */
1048 if ((pack_size
+ 60 + s
.total_out
) > max_packsize
1049 || (pack_size
+ 60 + s
.total_out
) < pack_size
) {
1051 /* This new object needs to *not* have the current pack_id. */
1052 e
->pack_id
= pack_id
+ 1;
1055 /* We cannot carry a delta into the new pack. */
1060 memset(&s
, 0, sizeof(s
));
1061 deflateInit(&s
, zlib_compression_level
);
1063 s
.avail_in
= datlen
;
1064 s
.avail_out
= deflateBound(&s
, s
.avail_in
);
1065 s
.next_out
= out
= xrealloc(out
, s
.avail_out
);
1066 while (deflate(&s
, Z_FINISH
) == Z_OK
)
1073 e
->pack_id
= pack_id
;
1074 e
->offset
= pack_size
;
1076 object_count_by_type
[type
]++;
1079 unsigned long ofs
= e
->offset
- last
->offset
;
1080 unsigned pos
= sizeof(hdr
) - 1;
1082 delta_count_by_type
[type
]++;
1085 hdrlen
= encode_header(OBJ_OFS_DELTA
, deltalen
, hdr
);
1086 write_or_die(pack_data
->pack_fd
, hdr
, hdrlen
);
1087 pack_size
+= hdrlen
;
1089 hdr
[pos
] = ofs
& 127;
1091 hdr
[--pos
] = 128 | (--ofs
& 127);
1092 write_or_die(pack_data
->pack_fd
, hdr
+ pos
, sizeof(hdr
) - pos
);
1093 pack_size
+= sizeof(hdr
) - pos
;
1097 hdrlen
= encode_header(type
, datlen
, hdr
);
1098 write_or_die(pack_data
->pack_fd
, hdr
, hdrlen
);
1099 pack_size
+= hdrlen
;
1102 write_or_die(pack_data
->pack_fd
, out
, s
.total_out
);
1103 pack_size
+= s
.total_out
;
1111 last
->offset
= e
->offset
;
1117 static void *gfi_unpack_entry(
1118 struct object_entry
*oe
,
1119 unsigned long *sizep
)
1121 enum object_type type
;
1122 struct packed_git
*p
= all_packs
[oe
->pack_id
];
1124 p
->pack_size
= pack_size
+ 20;
1125 return unpack_entry(p
, oe
->offset
, &type
, sizep
);
1128 static const char *get_mode(const char *str
, uint16_t *modep
)
1133 while ((c
= *str
++) != ' ') {
1134 if (c
< '0' || c
> '7')
1136 mode
= (mode
<< 3) + (c
- '0');
1142 static void load_tree(struct tree_entry
*root
)
1144 unsigned char* sha1
= root
->versions
[1].sha1
;
1145 struct object_entry
*myoe
;
1146 struct tree_content
*t
;
1151 root
->tree
= t
= new_tree_content(8);
1152 if (is_null_sha1(sha1
))
1155 myoe
= find_object(sha1
);
1156 if (myoe
&& myoe
->pack_id
!= MAX_PACK_ID
) {
1157 if (myoe
->type
!= OBJ_TREE
)
1158 die("Not a tree: %s", sha1_to_hex(sha1
));
1160 buf
= gfi_unpack_entry(myoe
, &size
);
1162 enum object_type type
;
1163 buf
= read_sha1_file(sha1
, &type
, &size
);
1164 if (!buf
|| type
!= OBJ_TREE
)
1165 die("Can't load tree %s", sha1_to_hex(sha1
));
1169 while (c
!= (buf
+ size
)) {
1170 struct tree_entry
*e
= new_tree_entry();
1172 if (t
->entry_count
== t
->entry_capacity
)
1173 root
->tree
= t
= grow_tree_content(t
, t
->entry_count
);
1174 t
->entries
[t
->entry_count
++] = e
;
1177 c
= get_mode(c
, &e
->versions
[1].mode
);
1179 die("Corrupt mode in %s", sha1_to_hex(sha1
));
1180 e
->versions
[0].mode
= e
->versions
[1].mode
;
1181 e
->name
= to_atom(c
, strlen(c
));
1182 c
+= e
->name
->str_len
+ 1;
1183 hashcpy(e
->versions
[0].sha1
, (unsigned char*)c
);
1184 hashcpy(e
->versions
[1].sha1
, (unsigned char*)c
);
1190 static int tecmp0 (const void *_a
, const void *_b
)
1192 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
1193 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
1194 return base_name_compare(
1195 a
->name
->str_dat
, a
->name
->str_len
, a
->versions
[0].mode
,
1196 b
->name
->str_dat
, b
->name
->str_len
, b
->versions
[0].mode
);
1199 static int tecmp1 (const void *_a
, const void *_b
)
1201 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
1202 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
1203 return base_name_compare(
1204 a
->name
->str_dat
, a
->name
->str_len
, a
->versions
[1].mode
,
1205 b
->name
->str_dat
, b
->name
->str_len
, b
->versions
[1].mode
);
1208 static void mktree(struct tree_content
*t
,
1218 qsort(t
->entries
,t
->entry_count
,sizeof(t
->entries
[0]),tecmp0
);
1220 qsort(t
->entries
,t
->entry_count
,sizeof(t
->entries
[0]),tecmp1
);
1222 for (i
= 0; i
< t
->entry_count
; i
++) {
1223 if (t
->entries
[i
]->versions
[v
].mode
)
1224 maxlen
+= t
->entries
[i
]->name
->str_len
+ 34;
1227 size_dbuf(b
, maxlen
);
1229 for (i
= 0; i
< t
->entry_count
; i
++) {
1230 struct tree_entry
*e
= t
->entries
[i
];
1231 if (!e
->versions
[v
].mode
)
1233 c
+= sprintf(c
, "%o", (unsigned int)e
->versions
[v
].mode
);
1235 strcpy(c
, e
->name
->str_dat
);
1236 c
+= e
->name
->str_len
+ 1;
1237 hashcpy((unsigned char*)c
, e
->versions
[v
].sha1
);
1240 *szp
= c
- (char*)b
->buffer
;
1243 static void store_tree(struct tree_entry
*root
)
1245 struct tree_content
*t
= root
->tree
;
1246 unsigned int i
, j
, del
;
1247 unsigned long new_len
;
1248 struct last_object lo
;
1249 struct object_entry
*le
;
1251 if (!is_null_sha1(root
->versions
[1].sha1
))
1254 for (i
= 0; i
< t
->entry_count
; i
++) {
1255 if (t
->entries
[i
]->tree
)
1256 store_tree(t
->entries
[i
]);
1259 le
= find_object(root
->versions
[0].sha1
);
1260 if (!S_ISDIR(root
->versions
[0].mode
)
1262 || le
->pack_id
!= pack_id
) {
1267 mktree(t
, 0, &lo
.len
, &old_tree
);
1268 lo
.data
= old_tree
.buffer
;
1269 lo
.offset
= le
->offset
;
1270 lo
.depth
= t
->delta_depth
;
1274 mktree(t
, 1, &new_len
, &new_tree
);
1275 store_object(OBJ_TREE
, new_tree
.buffer
, new_len
,
1276 &lo
, root
->versions
[1].sha1
, 0);
1278 t
->delta_depth
= lo
.depth
;
1279 for (i
= 0, j
= 0, del
= 0; i
< t
->entry_count
; i
++) {
1280 struct tree_entry
*e
= t
->entries
[i
];
1281 if (e
->versions
[1].mode
) {
1282 e
->versions
[0].mode
= e
->versions
[1].mode
;
1283 hashcpy(e
->versions
[0].sha1
, e
->versions
[1].sha1
);
1284 t
->entries
[j
++] = e
;
1286 release_tree_entry(e
);
1290 t
->entry_count
-= del
;
1293 static int tree_content_set(
1294 struct tree_entry
*root
,
1296 const unsigned char *sha1
,
1297 const uint16_t mode
,
1298 struct tree_content
*subtree
)
1300 struct tree_content
*t
= root
->tree
;
1303 struct tree_entry
*e
;
1305 slash1
= strchr(p
, '/');
1311 die("Empty path component found in input");
1312 if (!slash1
&& !S_ISDIR(mode
) && subtree
)
1313 die("Non-directories cannot have subtrees");
1315 for (i
= 0; i
< t
->entry_count
; i
++) {
1317 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
1320 && e
->versions
[1].mode
== mode
1321 && !hashcmp(e
->versions
[1].sha1
, sha1
))
1323 e
->versions
[1].mode
= mode
;
1324 hashcpy(e
->versions
[1].sha1
, sha1
);
1326 release_tree_content_recursive(e
->tree
);
1328 hashclr(root
->versions
[1].sha1
);
1331 if (!S_ISDIR(e
->versions
[1].mode
)) {
1332 e
->tree
= new_tree_content(8);
1333 e
->versions
[1].mode
= S_IFDIR
;
1337 if (tree_content_set(e
, slash1
+ 1, sha1
, mode
, subtree
)) {
1338 hashclr(root
->versions
[1].sha1
);
1345 if (t
->entry_count
== t
->entry_capacity
)
1346 root
->tree
= t
= grow_tree_content(t
, t
->entry_count
);
1347 e
= new_tree_entry();
1348 e
->name
= to_atom(p
, n
);
1349 e
->versions
[0].mode
= 0;
1350 hashclr(e
->versions
[0].sha1
);
1351 t
->entries
[t
->entry_count
++] = e
;
1353 e
->tree
= new_tree_content(8);
1354 e
->versions
[1].mode
= S_IFDIR
;
1355 tree_content_set(e
, slash1
+ 1, sha1
, mode
, subtree
);
1358 e
->versions
[1].mode
= mode
;
1359 hashcpy(e
->versions
[1].sha1
, sha1
);
1361 hashclr(root
->versions
[1].sha1
);
1365 static int tree_content_remove(
1366 struct tree_entry
*root
,
1368 struct tree_entry
*backup_leaf
)
1370 struct tree_content
*t
= root
->tree
;
1373 struct tree_entry
*e
;
1375 slash1
= strchr(p
, '/');
1381 for (i
= 0; i
< t
->entry_count
; i
++) {
1383 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
1384 if (!slash1
|| !S_ISDIR(e
->versions
[1].mode
))
1388 if (tree_content_remove(e
, slash1
+ 1, backup_leaf
)) {
1389 for (n
= 0; n
< e
->tree
->entry_count
; n
++) {
1390 if (e
->tree
->entries
[n
]->versions
[1].mode
) {
1391 hashclr(root
->versions
[1].sha1
);
1405 memcpy(backup_leaf
, e
, sizeof(*backup_leaf
));
1407 release_tree_content_recursive(e
->tree
);
1409 e
->versions
[1].mode
= 0;
1410 hashclr(e
->versions
[1].sha1
);
1411 hashclr(root
->versions
[1].sha1
);
1415 static int tree_content_get(
1416 struct tree_entry
*root
,
1418 struct tree_entry
*leaf
)
1420 struct tree_content
*t
= root
->tree
;
1423 struct tree_entry
*e
;
1425 slash1
= strchr(p
, '/');
1431 for (i
= 0; i
< t
->entry_count
; i
++) {
1433 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
1435 memcpy(leaf
, e
, sizeof(*leaf
));
1436 if (e
->tree
&& is_null_sha1(e
->versions
[1].sha1
))
1437 leaf
->tree
= dup_tree_content(e
->tree
);
1442 if (!S_ISDIR(e
->versions
[1].mode
))
1446 return tree_content_get(e
, slash1
+ 1, leaf
);
1452 static int update_branch(struct branch
*b
)
1454 static const char *msg
= "fast-import";
1455 struct ref_lock
*lock
;
1456 unsigned char old_sha1
[20];
1458 if (read_ref(b
->name
, old_sha1
))
1460 lock
= lock_any_ref_for_update(b
->name
, old_sha1
, 0);
1462 return error("Unable to lock %s", b
->name
);
1463 if (!force_update
&& !is_null_sha1(old_sha1
)) {
1464 struct commit
*old_cmit
, *new_cmit
;
1466 old_cmit
= lookup_commit_reference_gently(old_sha1
, 0);
1467 new_cmit
= lookup_commit_reference_gently(b
->sha1
, 0);
1468 if (!old_cmit
|| !new_cmit
) {
1470 return error("Branch %s is missing commits.", b
->name
);
1473 if (!in_merge_bases(old_cmit
, &new_cmit
, 1)) {
1475 warning("Not updating %s"
1476 " (new tip %s does not contain %s)",
1477 b
->name
, sha1_to_hex(b
->sha1
), sha1_to_hex(old_sha1
));
1481 if (write_ref_sha1(lock
, b
->sha1
, msg
) < 0)
1482 return error("Unable to update %s", b
->name
);
1486 static void dump_branches(void)
1491 for (i
= 0; i
< branch_table_sz
; i
++) {
1492 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
)
1493 failure
|= update_branch(b
);
1497 static void dump_tags(void)
1499 static const char *msg
= "fast-import";
1501 struct ref_lock
*lock
;
1502 char ref_name
[PATH_MAX
];
1504 for (t
= first_tag
; t
; t
= t
->next_tag
) {
1505 sprintf(ref_name
, "tags/%s", t
->name
);
1506 lock
= lock_ref_sha1(ref_name
, NULL
);
1507 if (!lock
|| write_ref_sha1(lock
, t
->sha1
, msg
) < 0)
1508 failure
|= error("Unable to update %s", ref_name
);
1512 static void dump_marks_helper(FILE *f
,
1518 for (k
= 0; k
< 1024; k
++) {
1519 if (m
->data
.sets
[k
])
1520 dump_marks_helper(f
, (base
+ k
) << m
->shift
,
1524 for (k
= 0; k
< 1024; k
++) {
1525 if (m
->data
.marked
[k
])
1526 fprintf(f
, ":%" PRIuMAX
" %s\n", base
+ k
,
1527 sha1_to_hex(m
->data
.marked
[k
]->sha1
));
1532 static void dump_marks(void)
1534 static struct lock_file mark_lock
;
1541 mark_fd
= hold_lock_file_for_update(&mark_lock
, mark_file
, 0);
1543 failure
|= error("Unable to write marks file %s: %s",
1544 mark_file
, strerror(errno
));
1548 f
= fdopen(mark_fd
, "w");
1550 rollback_lock_file(&mark_lock
);
1551 failure
|= error("Unable to write marks file %s: %s",
1552 mark_file
, strerror(errno
));
1556 dump_marks_helper(f
, 0, marks
);
1558 if (commit_lock_file(&mark_lock
))
1559 failure
|= error("Unable to write marks file %s: %s",
1560 mark_file
, strerror(errno
));
1563 static void read_next_command(void)
1566 if (unread_command_buf
)
1567 unread_command_buf
= 0;
1569 read_line(&command_buf
, stdin
, '\n');
1570 } while (!command_buf
.eof
&& command_buf
.buf
[0] == '#');
1573 static void skip_optional_lf()
1575 int term_char
= fgetc(stdin
);
1576 if (term_char
!= '\n' && term_char
!= EOF
)
1577 ungetc(term_char
, stdin
);
1580 static void cmd_mark(void)
1582 if (!prefixcmp(command_buf
.buf
, "mark :")) {
1583 next_mark
= strtoumax(command_buf
.buf
+ 6, NULL
, 10);
1584 read_next_command();
1590 static void *cmd_data (size_t *size
)
1595 if (prefixcmp(command_buf
.buf
, "data "))
1596 die("Expected 'data n' command, found: %s", command_buf
.buf
);
1598 if (!prefixcmp(command_buf
.buf
+ 5, "<<")) {
1599 char *term
= xstrdup(command_buf
.buf
+ 5 + 2);
1600 size_t sz
= 8192, term_len
= command_buf
.len
- 5 - 2;
1602 buffer
= xmalloc(sz
);
1604 read_line(&command_buf
, stdin
, '\n');
1605 if (command_buf
.eof
)
1606 die("EOF in data (terminator '%s' not found)", term
);
1607 if (term_len
== command_buf
.len
1608 && !strcmp(term
, command_buf
.buf
))
1610 ALLOC_GROW(buffer
, length
+ command_buf
.len
, sz
);
1611 memcpy(buffer
+ length
,
1613 command_buf
.len
- 1);
1614 length
+= command_buf
.len
- 1;
1615 buffer
[length
++] = '\n';
1621 length
= strtoul(command_buf
.buf
+ 5, NULL
, 10);
1622 buffer
= xmalloc(length
);
1623 while (n
< length
) {
1624 size_t s
= fread(buffer
+ n
, 1, length
- n
, stdin
);
1625 if (!s
&& feof(stdin
))
1626 die("EOF in data (%lu bytes remaining)",
1627 (unsigned long)(length
- n
));
1637 static int validate_raw_date(const char *src
, char *result
, int maxlen
)
1639 const char *orig_src
= src
;
1642 strtoul(src
, &endp
, 10);
1643 if (endp
== src
|| *endp
!= ' ')
1647 if (*src
!= '-' && *src
!= '+')
1651 strtoul(src
+ 1, &endp
, 10);
1652 if (endp
== src
|| *endp
|| (endp
- orig_src
) >= maxlen
)
1655 strcpy(result
, orig_src
);
1659 static char *parse_ident(const char *buf
)
1665 gt
= strrchr(buf
, '>');
1667 die("Missing > in ident string: %s", buf
);
1670 die("Missing space after > in ident string: %s", buf
);
1672 name_len
= gt
- buf
;
1673 ident
= xmalloc(name_len
+ 24);
1674 strncpy(ident
, buf
, name_len
);
1678 if (validate_raw_date(gt
, ident
+ name_len
, 24) < 0)
1679 die("Invalid raw date \"%s\" in ident: %s", gt
, buf
);
1681 case WHENSPEC_RFC2822
:
1682 if (parse_date(gt
, ident
+ name_len
, 24) < 0)
1683 die("Invalid rfc2822 date \"%s\" in ident: %s", gt
, buf
);
1686 if (strcmp("now", gt
))
1687 die("Date in ident must be 'now': %s", buf
);
1688 datestamp(ident
+ name_len
, 24);
1695 static void cmd_new_blob(void)
1700 read_next_command();
1704 if (store_object(OBJ_BLOB
, d
, l
, &last_blob
, NULL
, next_mark
))
1708 static void unload_one_branch(void)
1710 while (cur_active_branches
1711 && cur_active_branches
>= max_active_branches
) {
1712 uintmax_t min_commit
= ULONG_MAX
;
1713 struct branch
*e
, *l
= NULL
, *p
= NULL
;
1715 for (e
= active_branches
; e
; e
= e
->active_next_branch
) {
1716 if (e
->last_commit
< min_commit
) {
1718 min_commit
= e
->last_commit
;
1724 e
= p
->active_next_branch
;
1725 p
->active_next_branch
= e
->active_next_branch
;
1727 e
= active_branches
;
1728 active_branches
= e
->active_next_branch
;
1731 e
->active_next_branch
= NULL
;
1732 if (e
->branch_tree
.tree
) {
1733 release_tree_content_recursive(e
->branch_tree
.tree
);
1734 e
->branch_tree
.tree
= NULL
;
1736 cur_active_branches
--;
1740 static void load_branch(struct branch
*b
)
1742 load_tree(&b
->branch_tree
);
1745 b
->active_next_branch
= active_branches
;
1746 active_branches
= b
;
1747 cur_active_branches
++;
1748 branch_load_count
++;
1752 static void file_change_m(struct branch
*b
)
1754 const char *p
= command_buf
.buf
+ 2;
1757 struct object_entry
*oe
= oe
;
1758 unsigned char sha1
[20];
1759 uint16_t mode
, inline_data
= 0;
1761 p
= get_mode(p
, &mode
);
1763 die("Corrupt mode: %s", command_buf
.buf
);
1765 case S_IFREG
| 0644:
1766 case S_IFREG
| 0755:
1773 die("Corrupt mode: %s", command_buf
.buf
);
1778 oe
= find_mark(strtoumax(p
+ 1, &x
, 10));
1779 hashcpy(sha1
, oe
->sha1
);
1781 } else if (!prefixcmp(p
, "inline")) {
1785 if (get_sha1_hex(p
, sha1
))
1786 die("Invalid SHA1: %s", command_buf
.buf
);
1787 oe
= find_object(sha1
);
1791 die("Missing space after SHA1: %s", command_buf
.buf
);
1793 p_uq
= unquote_c_style(p
, &endp
);
1796 die("Garbage after path in: %s", command_buf
.buf
);
1804 p
= p_uq
= xstrdup(p
);
1805 read_next_command();
1807 if (store_object(OBJ_BLOB
, d
, l
, &last_blob
, sha1
, 0))
1810 if (oe
->type
!= OBJ_BLOB
)
1811 die("Not a blob (actually a %s): %s",
1812 command_buf
.buf
, typename(oe
->type
));
1814 enum object_type type
= sha1_object_info(sha1
, NULL
);
1816 die("Blob not found: %s", command_buf
.buf
);
1817 if (type
!= OBJ_BLOB
)
1818 die("Not a blob (actually a %s): %s",
1819 typename(type
), command_buf
.buf
);
1822 tree_content_set(&b
->branch_tree
, p
, sha1
, S_IFREG
| mode
, NULL
);
1826 static void file_change_d(struct branch
*b
)
1828 const char *p
= command_buf
.buf
+ 2;
1832 p_uq
= unquote_c_style(p
, &endp
);
1835 die("Garbage after path in: %s", command_buf
.buf
);
1838 tree_content_remove(&b
->branch_tree
, p
, NULL
);
1842 static void file_change_cr(struct branch
*b
, int rename
)
1847 struct tree_entry leaf
;
1849 s
= command_buf
.buf
+ 2;
1850 s_uq
= unquote_c_style(s
, &endp
);
1853 die("Missing space after source: %s", command_buf
.buf
);
1856 endp
= strchr(s
, ' ');
1858 die("Missing space after source: %s", command_buf
.buf
);
1859 s_uq
= xmalloc(endp
- s
+ 1);
1860 memcpy(s_uq
, s
, endp
- s
);
1867 die("Missing dest: %s", command_buf
.buf
);
1870 d_uq
= unquote_c_style(d
, &endp
);
1873 die("Garbage after dest in: %s", command_buf
.buf
);
1877 memset(&leaf
, 0, sizeof(leaf
));
1879 tree_content_remove(&b
->branch_tree
, s
, &leaf
);
1881 tree_content_get(&b
->branch_tree
, s
, &leaf
);
1882 if (!leaf
.versions
[1].mode
)
1883 die("Path %s not in branch", s
);
1884 tree_content_set(&b
->branch_tree
, d
,
1885 leaf
.versions
[1].sha1
,
1886 leaf
.versions
[1].mode
,
1893 static void file_change_deleteall(struct branch
*b
)
1895 release_tree_content_recursive(b
->branch_tree
.tree
);
1896 hashclr(b
->branch_tree
.versions
[0].sha1
);
1897 hashclr(b
->branch_tree
.versions
[1].sha1
);
1898 load_tree(&b
->branch_tree
);
1901 static void cmd_from_commit(struct branch
*b
, char *buf
, unsigned long size
)
1903 if (!buf
|| size
< 46)
1904 die("Not a valid commit: %s", sha1_to_hex(b
->sha1
));
1905 if (memcmp("tree ", buf
, 5)
1906 || get_sha1_hex(buf
+ 5, b
->branch_tree
.versions
[1].sha1
))
1907 die("The commit %s is corrupt", sha1_to_hex(b
->sha1
));
1908 hashcpy(b
->branch_tree
.versions
[0].sha1
,
1909 b
->branch_tree
.versions
[1].sha1
);
1912 static void cmd_from_existing(struct branch
*b
)
1914 if (is_null_sha1(b
->sha1
)) {
1915 hashclr(b
->branch_tree
.versions
[0].sha1
);
1916 hashclr(b
->branch_tree
.versions
[1].sha1
);
1921 buf
= read_object_with_reference(b
->sha1
,
1922 commit_type
, &size
, b
->sha1
);
1923 cmd_from_commit(b
, buf
, size
);
1928 static int cmd_from(struct branch
*b
)
1933 if (prefixcmp(command_buf
.buf
, "from "))
1936 if (b
->branch_tree
.tree
) {
1937 release_tree_content_recursive(b
->branch_tree
.tree
);
1938 b
->branch_tree
.tree
= NULL
;
1941 from
= strchr(command_buf
.buf
, ' ') + 1;
1942 s
= lookup_branch(from
);
1944 die("Can't create a branch from itself: %s", b
->name
);
1946 unsigned char *t
= s
->branch_tree
.versions
[1].sha1
;
1947 hashcpy(b
->sha1
, s
->sha1
);
1948 hashcpy(b
->branch_tree
.versions
[0].sha1
, t
);
1949 hashcpy(b
->branch_tree
.versions
[1].sha1
, t
);
1950 } else if (*from
== ':') {
1951 uintmax_t idnum
= strtoumax(from
+ 1, NULL
, 10);
1952 struct object_entry
*oe
= find_mark(idnum
);
1953 if (oe
->type
!= OBJ_COMMIT
)
1954 die("Mark :%" PRIuMAX
" not a commit", idnum
);
1955 hashcpy(b
->sha1
, oe
->sha1
);
1956 if (oe
->pack_id
!= MAX_PACK_ID
) {
1958 char *buf
= gfi_unpack_entry(oe
, &size
);
1959 cmd_from_commit(b
, buf
, size
);
1962 cmd_from_existing(b
);
1963 } else if (!get_sha1(from
, b
->sha1
))
1964 cmd_from_existing(b
);
1966 die("Invalid ref name or SHA1 expression: %s", from
);
1968 read_next_command();
1972 static struct hash_list
*cmd_merge(unsigned int *count
)
1974 struct hash_list
*list
= NULL
, *n
, *e
= e
;
1979 while (!prefixcmp(command_buf
.buf
, "merge ")) {
1980 from
= strchr(command_buf
.buf
, ' ') + 1;
1981 n
= xmalloc(sizeof(*n
));
1982 s
= lookup_branch(from
);
1984 hashcpy(n
->sha1
, s
->sha1
);
1985 else if (*from
== ':') {
1986 uintmax_t idnum
= strtoumax(from
+ 1, NULL
, 10);
1987 struct object_entry
*oe
= find_mark(idnum
);
1988 if (oe
->type
!= OBJ_COMMIT
)
1989 die("Mark :%" PRIuMAX
" not a commit", idnum
);
1990 hashcpy(n
->sha1
, oe
->sha1
);
1991 } else if (!get_sha1(from
, n
->sha1
)) {
1993 char *buf
= read_object_with_reference(n
->sha1
,
1994 commit_type
, &size
, n
->sha1
);
1995 if (!buf
|| size
< 46)
1996 die("Not a valid commit: %s", from
);
1999 die("Invalid ref name or SHA1 expression: %s", from
);
2008 read_next_command();
2013 static void cmd_new_commit(void)
2019 char *author
= NULL
;
2020 char *committer
= NULL
;
2021 struct hash_list
*merge_list
= NULL
;
2022 unsigned int merge_count
;
2024 /* Obtain the branch name from the rest of our command */
2025 sp
= strchr(command_buf
.buf
, ' ') + 1;
2026 b
= lookup_branch(sp
);
2030 read_next_command();
2032 if (!prefixcmp(command_buf
.buf
, "author ")) {
2033 author
= parse_ident(command_buf
.buf
+ 7);
2034 read_next_command();
2036 if (!prefixcmp(command_buf
.buf
, "committer ")) {
2037 committer
= parse_ident(command_buf
.buf
+ 10);
2038 read_next_command();
2041 die("Expected committer but didn't get one");
2042 msg
= cmd_data(&msglen
);
2043 read_next_command();
2045 merge_list
= cmd_merge(&merge_count
);
2047 /* ensure the branch is active/loaded */
2048 if (!b
->branch_tree
.tree
|| !max_active_branches
) {
2049 unload_one_branch();
2054 while (!command_buf
.eof
&& command_buf
.len
> 1) {
2055 if (!prefixcmp(command_buf
.buf
, "M "))
2057 else if (!prefixcmp(command_buf
.buf
, "D "))
2059 else if (!prefixcmp(command_buf
.buf
, "R "))
2060 file_change_cr(b
, 1);
2061 else if (!prefixcmp(command_buf
.buf
, "C "))
2062 file_change_cr(b
, 0);
2063 else if (!strcmp("deleteall", command_buf
.buf
))
2064 file_change_deleteall(b
);
2066 unread_command_buf
= 1;
2069 read_next_command();
2072 /* build the tree and the commit */
2073 store_tree(&b
->branch_tree
);
2074 hashcpy(b
->branch_tree
.versions
[0].sha1
,
2075 b
->branch_tree
.versions
[1].sha1
);
2076 size_dbuf(&new_data
, 114 + msglen
2079 ? strlen(author
) + strlen(committer
)
2080 : 2 * strlen(committer
)));
2081 sp
= new_data
.buffer
;
2082 sp
+= sprintf(sp
, "tree %s\n",
2083 sha1_to_hex(b
->branch_tree
.versions
[1].sha1
));
2084 if (!is_null_sha1(b
->sha1
))
2085 sp
+= sprintf(sp
, "parent %s\n", sha1_to_hex(b
->sha1
));
2086 while (merge_list
) {
2087 struct hash_list
*next
= merge_list
->next
;
2088 sp
+= sprintf(sp
, "parent %s\n", sha1_to_hex(merge_list
->sha1
));
2092 sp
+= sprintf(sp
, "author %s\n", author
? author
: committer
);
2093 sp
+= sprintf(sp
, "committer %s\n", committer
);
2095 memcpy(sp
, msg
, msglen
);
2101 if (!store_object(OBJ_COMMIT
,
2102 new_data
.buffer
, sp
- (char*)new_data
.buffer
,
2103 NULL
, b
->sha1
, next_mark
))
2104 b
->pack_id
= pack_id
;
2105 b
->last_commit
= object_count_by_type
[OBJ_COMMIT
];
2108 static void cmd_new_tag(void)
2117 uintmax_t from_mark
= 0;
2118 unsigned char sha1
[20];
2120 /* Obtain the new tag name from the rest of our command */
2121 sp
= strchr(command_buf
.buf
, ' ') + 1;
2122 t
= pool_alloc(sizeof(struct tag
));
2124 t
->name
= pool_strdup(sp
);
2126 last_tag
->next_tag
= t
;
2130 read_next_command();
2133 if (prefixcmp(command_buf
.buf
, "from "))
2134 die("Expected from command, got %s", command_buf
.buf
);
2135 from
= strchr(command_buf
.buf
, ' ') + 1;
2136 s
= lookup_branch(from
);
2138 hashcpy(sha1
, s
->sha1
);
2139 } else if (*from
== ':') {
2140 struct object_entry
*oe
;
2141 from_mark
= strtoumax(from
+ 1, NULL
, 10);
2142 oe
= find_mark(from_mark
);
2143 if (oe
->type
!= OBJ_COMMIT
)
2144 die("Mark :%" PRIuMAX
" not a commit", from_mark
);
2145 hashcpy(sha1
, oe
->sha1
);
2146 } else if (!get_sha1(from
, sha1
)) {
2150 buf
= read_object_with_reference(sha1
,
2151 commit_type
, &size
, sha1
);
2152 if (!buf
|| size
< 46)
2153 die("Not a valid commit: %s", from
);
2156 die("Invalid ref name or SHA1 expression: %s", from
);
2157 read_next_command();
2160 if (prefixcmp(command_buf
.buf
, "tagger "))
2161 die("Expected tagger command, got %s", command_buf
.buf
);
2162 tagger
= parse_ident(command_buf
.buf
+ 7);
2164 /* tag payload/message */
2165 read_next_command();
2166 msg
= cmd_data(&msglen
);
2168 /* build the tag object */
2169 size_dbuf(&new_data
, 67+strlen(t
->name
)+strlen(tagger
)+msglen
);
2170 sp
= new_data
.buffer
;
2171 sp
+= sprintf(sp
, "object %s\n", sha1_to_hex(sha1
));
2172 sp
+= sprintf(sp
, "type %s\n", commit_type
);
2173 sp
+= sprintf(sp
, "tag %s\n", t
->name
);
2174 sp
+= sprintf(sp
, "tagger %s\n", tagger
);
2176 memcpy(sp
, msg
, msglen
);
2181 if (store_object(OBJ_TAG
, new_data
.buffer
,
2182 sp
- (char*)new_data
.buffer
,
2184 t
->pack_id
= MAX_PACK_ID
;
2186 t
->pack_id
= pack_id
;
2189 static void cmd_reset_branch(void)
2194 /* Obtain the branch name from the rest of our command */
2195 sp
= strchr(command_buf
.buf
, ' ') + 1;
2196 b
= lookup_branch(sp
);
2199 hashclr(b
->branch_tree
.versions
[0].sha1
);
2200 hashclr(b
->branch_tree
.versions
[1].sha1
);
2201 if (b
->branch_tree
.tree
) {
2202 release_tree_content_recursive(b
->branch_tree
.tree
);
2203 b
->branch_tree
.tree
= NULL
;
2208 read_next_command();
2209 if (!cmd_from(b
) && command_buf
.len
> 1)
2210 unread_command_buf
= 1;
2213 static void cmd_checkpoint(void)
2224 static void cmd_progress(void)
2226 fwrite(command_buf
.buf
, 1, command_buf
.len
- 1, stdout
);
2227 fputc('\n', stdout
);
2232 static void import_marks(const char *input_file
)
2235 FILE *f
= fopen(input_file
, "r");
2237 die("cannot read %s: %s", input_file
, strerror(errno
));
2238 while (fgets(line
, sizeof(line
), f
)) {
2241 unsigned char sha1
[20];
2242 struct object_entry
*e
;
2244 end
= strchr(line
, '\n');
2245 if (line
[0] != ':' || !end
)
2246 die("corrupt mark line: %s", line
);
2248 mark
= strtoumax(line
+ 1, &end
, 10);
2249 if (!mark
|| end
== line
+ 1
2250 || *end
!= ' ' || get_sha1(end
+ 1, sha1
))
2251 die("corrupt mark line: %s", line
);
2252 e
= find_object(sha1
);
2254 enum object_type type
= sha1_object_info(sha1
, NULL
);
2256 die("object not found: %s", sha1_to_hex(sha1
));
2257 e
= insert_object(sha1
);
2259 e
->pack_id
= MAX_PACK_ID
;
2260 e
->offset
= 1; /* just not zero! */
2262 insert_mark(mark
, e
);
2267 static const char fast_import_usage
[] =
2268 "git-fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2270 int main(int argc
, const char **argv
)
2272 int i
, show_stats
= 1;
2274 git_config(git_default_config
);
2275 alloc_objects(object_entry_alloc
);
2276 strbuf_init(&command_buf
);
2277 atom_table
= xcalloc(atom_table_sz
, sizeof(struct atom_str
*));
2278 branch_table
= xcalloc(branch_table_sz
, sizeof(struct branch
*));
2279 avail_tree_table
= xcalloc(avail_tree_table_sz
, sizeof(struct avail_tree_content
*));
2280 marks
= pool_calloc(1, sizeof(struct mark_set
));
2282 for (i
= 1; i
< argc
; i
++) {
2283 const char *a
= argv
[i
];
2285 if (*a
!= '-' || !strcmp(a
, "--"))
2287 else if (!prefixcmp(a
, "--date-format=")) {
2288 const char *fmt
= a
+ 14;
2289 if (!strcmp(fmt
, "raw"))
2290 whenspec
= WHENSPEC_RAW
;
2291 else if (!strcmp(fmt
, "rfc2822"))
2292 whenspec
= WHENSPEC_RFC2822
;
2293 else if (!strcmp(fmt
, "now"))
2294 whenspec
= WHENSPEC_NOW
;
2296 die("unknown --date-format argument %s", fmt
);
2298 else if (!prefixcmp(a
, "--max-pack-size="))
2299 max_packsize
= strtoumax(a
+ 16, NULL
, 0) * 1024 * 1024;
2300 else if (!prefixcmp(a
, "--depth="))
2301 max_depth
= strtoul(a
+ 8, NULL
, 0);
2302 else if (!prefixcmp(a
, "--active-branches="))
2303 max_active_branches
= strtoul(a
+ 18, NULL
, 0);
2304 else if (!prefixcmp(a
, "--import-marks="))
2305 import_marks(a
+ 15);
2306 else if (!prefixcmp(a
, "--export-marks="))
2308 else if (!prefixcmp(a
, "--export-pack-edges=")) {
2311 pack_edges
= fopen(a
+ 20, "a");
2313 die("Cannot open %s: %s", a
+ 20, strerror(errno
));
2314 } else if (!strcmp(a
, "--force"))
2316 else if (!strcmp(a
, "--quiet"))
2318 else if (!strcmp(a
, "--stats"))
2321 die("unknown option %s", a
);
2324 usage(fast_import_usage
);
2326 prepare_packed_git();
2328 set_die_routine(die_nicely
);
2330 read_next_command();
2331 if (command_buf
.eof
)
2333 else if (!strcmp("blob", command_buf
.buf
))
2335 else if (!prefixcmp(command_buf
.buf
, "commit "))
2337 else if (!prefixcmp(command_buf
.buf
, "tag "))
2339 else if (!prefixcmp(command_buf
.buf
, "reset "))
2341 else if (!strcmp("checkpoint", command_buf
.buf
))
2343 else if (!prefixcmp(command_buf
.buf
, "progress "))
2346 die("Unsupported command: %s", command_buf
.buf
);
2359 uintmax_t total_count
= 0, duplicate_count
= 0;
2360 for (i
= 0; i
< ARRAY_SIZE(object_count_by_type
); i
++)
2361 total_count
+= object_count_by_type
[i
];
2362 for (i
= 0; i
< ARRAY_SIZE(duplicate_count_by_type
); i
++)
2363 duplicate_count
+= duplicate_count_by_type
[i
];
2365 fprintf(stderr
, "%s statistics:\n", argv
[0]);
2366 fprintf(stderr
, "---------------------------------------------------------------------\n");
2367 fprintf(stderr
, "Alloc'd objects: %10" PRIuMAX
"\n", alloc_count
);
2368 fprintf(stderr
, "Total objects: %10" PRIuMAX
" (%10" PRIuMAX
" duplicates )\n", total_count
, duplicate_count
);
2369 fprintf(stderr
, " blobs : %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas)\n", object_count_by_type
[OBJ_BLOB
], duplicate_count_by_type
[OBJ_BLOB
], delta_count_by_type
[OBJ_BLOB
]);
2370 fprintf(stderr
, " trees : %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas)\n", object_count_by_type
[OBJ_TREE
], duplicate_count_by_type
[OBJ_TREE
], delta_count_by_type
[OBJ_TREE
]);
2371 fprintf(stderr
, " commits: %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas)\n", object_count_by_type
[OBJ_COMMIT
], duplicate_count_by_type
[OBJ_COMMIT
], delta_count_by_type
[OBJ_COMMIT
]);
2372 fprintf(stderr
, " tags : %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas)\n", object_count_by_type
[OBJ_TAG
], duplicate_count_by_type
[OBJ_TAG
], delta_count_by_type
[OBJ_TAG
]);
2373 fprintf(stderr
, "Total branches: %10lu (%10lu loads )\n", branch_count
, branch_load_count
);
2374 fprintf(stderr
, " marks: %10" PRIuMAX
" (%10" PRIuMAX
" unique )\n", (((uintmax_t)1) << marks
->shift
) * 1024, marks_set_count
);
2375 fprintf(stderr
, " atoms: %10u\n", atom_cnt
);
2376 fprintf(stderr
, "Memory total: %10" PRIuMAX
" KiB\n", (total_allocd
+ alloc_count
*sizeof(struct object_entry
))/1024);
2377 fprintf(stderr
, " pools: %10lu KiB\n", (unsigned long)(total_allocd
/1024));
2378 fprintf(stderr
, " objects: %10" PRIuMAX
" KiB\n", (alloc_count
*sizeof(struct object_entry
))/1024);
2379 fprintf(stderr
, "---------------------------------------------------------------------\n");
2381 fprintf(stderr
, "---------------------------------------------------------------------\n");
2382 fprintf(stderr
, "\n");
2385 return failure
? 1 : 0;