2 Format of STDIN stream:
12 new_blob ::= 'blob' lf
15 file_content ::= data;
17 new_branch ::= 'branch' sp ref_str lf
18 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
21 new_commit ::= 'commit' sp ref_str lf
23 ('author' sp name '<' email '>' ts tz lf)?
24 'committer' sp name '<' email '>' ts tz lf
30 file_change ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf
33 mode ::= '644' | '755';
35 new_tag ::= 'tag' sp tag_str lf
36 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
37 'tagger' sp name '<' email '>' ts tz lf
41 # note: the first idnum in a stream should be 1 and subsequent
42 # idnums should not have gaps between values as this will cause
43 # the stream parser to reserve space for the gapped values. An
44 # idnum can be updated in the future to a new object by issuing
45 # a new mark directive with the old idnum.
47 mark ::= 'mark' sp idnum lf;
49 # note: declen indicates the length of binary_data in bytes.
50 # declen does not include the lf preceeding or trailing the
53 data ::= 'data' sp declen lf
57 # note: quoted strings are C-style quoting supporting \c for
58 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
59 # is the signed byte value in octal. Note that the only
60 # characters which must actually be escaped to protect the
61 # stream formatting is: \, " and LF. Otherwise these values
64 ref_str ::= ref | '"' quoted(ref) '"' ;
65 sha1exp_str ::= sha1exp | '"' quoted(sha1exp) '"' ;
66 tag_str ::= tag | '"' quoted(tag) '"' ;
67 path_str ::= path | '"' quoted(path) '"' ;
69 declen ::= # unsigned 32 bit value, ascii base10 notation;
70 binary_data ::= # file content, not interpreted;
72 sp ::= # ASCII space character;
73 lf ::= # ASCII newline (LF) character;
75 # note: a colon (':') must precede the numerical value assigned to
76 # an idnum. This is to distinguish it from a ref or tag name as
77 # GIT does not permit ':' in ref or tag strings.
80 path ::= # GIT style file path, e.g. "a/b/c";
81 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
82 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
83 sha1exp ::= # Any valid GIT SHA1 expression;
84 hexsha1 ::= # SHA1 in hexadecimal format;
86 # note: name and email are UTF8 strings, however name must not
87 # contain '<' or lf and email must not contain any of the
88 # following: '<', '>', lf.
90 name ::= # valid GIT author/committer name;
91 email ::= # valid GIT author/committer email;
92 ts ::= # time since the epoch in seconds, ascii base10 notation;
93 tz ::= # GIT style timezone;
104 #include "csum-file.h"
110 struct object_entry
*next
;
111 enum object_type type
;
112 unsigned long offset
;
113 unsigned char sha1
[20];
116 struct object_entry_pool
118 struct object_entry_pool
*next_pool
;
119 struct object_entry
*next_free
;
120 struct object_entry
*end
;
121 struct object_entry entries
[FLEX_ARRAY
]; /* more */
128 struct object_entry
*marked
[1024];
129 struct mark_set
*sets
[1024];
138 unsigned char sha1
[20];
143 struct mem_pool
*next_pool
;
146 char space
[FLEX_ARRAY
]; /* more */
151 struct atom_str
*next_atom
;
153 char str_dat
[FLEX_ARRAY
]; /* more */
159 struct tree_content
*tree
;
160 struct atom_str
* name
;
162 unsigned char sha1
[20];
167 unsigned int entry_capacity
; /* must match avail_tree_content */
168 unsigned int entry_count
;
169 struct tree_entry
*entries
[FLEX_ARRAY
]; /* more */
172 struct avail_tree_content
174 unsigned int entry_capacity
; /* must match tree_content */
175 struct avail_tree_content
*next_avail
;
180 struct branch
*table_next_branch
;
181 struct branch
*active_next_branch
;
183 unsigned long last_commit
;
184 struct tree_entry branch_tree
;
185 unsigned char sha1
[20];
190 struct tag
*next_tag
;
192 unsigned char sha1
[20];
196 /* Stats and misc. counters */
197 static unsigned long max_depth
= 10;
198 static unsigned long alloc_count
;
199 static unsigned long branch_count
;
200 static unsigned long branch_load_count
;
201 static unsigned long object_count
;
202 static unsigned long duplicate_count
;
203 static unsigned long marks_set_count
;
204 static unsigned long object_count_by_type
[9];
205 static unsigned long duplicate_count_by_type
[9];
208 static size_t mem_pool_alloc
= 2*1024*1024 - sizeof(struct mem_pool
);
209 static size_t total_allocd
;
210 static struct mem_pool
*mem_pool
;
212 /* Atom management */
213 static unsigned int atom_table_sz
= 4451;
214 static unsigned int atom_cnt
;
215 static struct atom_str
**atom_table
;
217 /* The .pack file being generated */
219 static unsigned long pack_offset
;
220 static unsigned char pack_sha1
[20];
222 /* Table of objects we've written. */
223 static unsigned int object_entry_alloc
= 1000;
224 static struct object_entry_pool
*blocks
;
225 static struct object_entry
*object_table
[1 << 16];
226 static struct mark_set
*marks
;
229 static struct last_object last_blob
;
231 /* Tree management */
232 static unsigned int tree_entry_alloc
= 1000;
233 static void *avail_tree_entry
;
234 static unsigned int avail_tree_table_sz
= 100;
235 static struct avail_tree_content
**avail_tree_table
;
238 static unsigned long max_active_branches
= 5;
239 static unsigned long cur_active_branches
;
240 static unsigned long branch_table_sz
= 1039;
241 static struct branch
**branch_table
;
242 static struct branch
*active_branches
;
245 static struct tag
*first_tag
;
246 static struct tag
*last_tag
;
248 /* Input stream parsing */
249 static struct strbuf command_buf
;
250 static unsigned long next_mark
;
253 static void alloc_objects(int cnt
)
255 struct object_entry_pool
*b
;
257 b
= xmalloc(sizeof(struct object_entry_pool
)
258 + cnt
* sizeof(struct object_entry
));
259 b
->next_pool
= blocks
;
260 b
->next_free
= b
->entries
;
261 b
->end
= b
->entries
+ cnt
;
266 static struct object_entry
* new_object(unsigned char *sha1
)
268 struct object_entry
*e
;
270 if (blocks
->next_free
== blocks
->end
)
271 alloc_objects(object_entry_alloc
);
273 e
= blocks
->next_free
++;
274 memcpy(e
->sha1
, sha1
, sizeof(e
->sha1
));
278 static struct object_entry
* find_object(unsigned char *sha1
)
280 unsigned int h
= sha1
[0] << 8 | sha1
[1];
281 struct object_entry
*e
;
282 for (e
= object_table
[h
]; e
; e
= e
->next
)
283 if (!memcmp(sha1
, e
->sha1
, sizeof(e
->sha1
)))
288 static struct object_entry
* insert_object(unsigned char *sha1
)
290 unsigned int h
= sha1
[0] << 8 | sha1
[1];
291 struct object_entry
*e
= object_table
[h
];
292 struct object_entry
*p
= NULL
;
295 if (!memcmp(sha1
, e
->sha1
, sizeof(e
->sha1
)))
301 e
= new_object(sha1
);
311 static unsigned int hc_str(const char *s
, size_t len
)
319 static void* pool_alloc(size_t len
)
324 for (p
= mem_pool
; p
; p
= p
->next_pool
)
325 if ((p
->end
- p
->next_free
>= len
))
329 if (len
>= (mem_pool_alloc
/2)) {
333 total_allocd
+= sizeof(struct mem_pool
) + mem_pool_alloc
;
334 p
= xmalloc(sizeof(struct mem_pool
) + mem_pool_alloc
);
335 p
->next_pool
= mem_pool
;
336 p
->next_free
= p
->space
;
337 p
->end
= p
->next_free
+ mem_pool_alloc
;
346 static void* pool_calloc(size_t count
, size_t size
)
348 size_t len
= count
* size
;
349 void *r
= pool_alloc(len
);
354 static char* pool_strdup(const char *s
)
356 char *r
= pool_alloc(strlen(s
) + 1);
361 static void insert_mark(unsigned long idnum
, struct object_entry
*oe
)
363 struct mark_set
*s
= marks
;
364 while ((idnum
>> s
->shift
) >= 1024) {
365 s
= pool_calloc(1, sizeof(struct mark_set
));
366 s
->shift
= marks
->shift
+ 10;
367 s
->data
.sets
[0] = marks
;
371 unsigned long i
= idnum
>> s
->shift
;
372 idnum
-= i
<< s
->shift
;
373 if (!s
->data
.sets
[i
]) {
374 s
->data
.sets
[i
] = pool_calloc(1, sizeof(struct mark_set
));
375 s
->data
.sets
[i
]->shift
= s
->shift
- 10;
379 if (!s
->data
.marked
[idnum
])
381 s
->data
.marked
[idnum
] = oe
;
384 static struct object_entry
* find_mark(unsigned long idnum
)
386 unsigned long orig_idnum
= idnum
;
387 struct mark_set
*s
= marks
;
388 struct object_entry
*oe
= NULL
;
389 if ((idnum
>> s
->shift
) < 1024) {
390 while (s
&& s
->shift
) {
391 unsigned long i
= idnum
>> s
->shift
;
392 idnum
-= i
<< s
->shift
;
396 oe
= s
->data
.marked
[idnum
];
399 die("mark :%lu not declared", orig_idnum
);
403 static struct atom_str
* to_atom(const char *s
, size_t len
)
405 unsigned int hc
= hc_str(s
, len
) % atom_table_sz
;
408 for (c
= atom_table
[hc
]; c
; c
= c
->next_atom
)
409 if (c
->str_len
== len
&& !strncmp(s
, c
->str_dat
, len
))
412 c
= pool_alloc(sizeof(struct atom_str
) + len
+ 1);
414 strncpy(c
->str_dat
, s
, len
);
416 c
->next_atom
= atom_table
[hc
];
422 static struct branch
* lookup_branch(const char *name
)
424 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
427 for (b
= branch_table
[hc
]; b
; b
= b
->table_next_branch
)
428 if (!strcmp(name
, b
->name
))
433 static struct branch
* new_branch(const char *name
)
435 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
436 struct branch
* b
= lookup_branch(name
);
439 die("Invalid attempt to create duplicate branch: %s", name
);
440 if (check_ref_format(name
))
441 die("Branch name doesn't conform to GIT standards: %s", name
);
443 b
= pool_calloc(1, sizeof(struct branch
));
444 b
->name
= pool_strdup(name
);
445 b
->table_next_branch
= branch_table
[hc
];
446 branch_table
[hc
] = b
;
451 static unsigned int hc_entries(unsigned int cnt
)
453 cnt
= cnt
& 7 ? (cnt
/ 8) + 1 : cnt
/ 8;
454 return cnt
< avail_tree_table_sz
? cnt
: avail_tree_table_sz
- 1;
457 static struct tree_content
* new_tree_content(unsigned int cnt
)
459 struct avail_tree_content
*f
, *l
= NULL
;
460 struct tree_content
*t
;
461 unsigned int hc
= hc_entries(cnt
);
463 for (f
= avail_tree_table
[hc
]; f
; l
= f
, f
= f
->next_avail
)
464 if (f
->entry_capacity
>= cnt
)
469 l
->next_avail
= f
->next_avail
;
471 avail_tree_table
[hc
] = f
->next_avail
;
473 cnt
= cnt
& 7 ? ((cnt
/ 8) + 1) * 8 : cnt
;
474 f
= pool_alloc(sizeof(*t
) + sizeof(t
->entries
[0]) * cnt
);
475 f
->entry_capacity
= cnt
;
478 t
= (struct tree_content
*)f
;
483 static void release_tree_entry(struct tree_entry
*e
);
484 static void release_tree_content(struct tree_content
*t
)
486 struct avail_tree_content
*f
= (struct avail_tree_content
*)t
;
487 unsigned int hc
= hc_entries(f
->entry_capacity
);
488 f
->next_avail
= avail_tree_table
[hc
];
489 avail_tree_table
[hc
] = f
;
492 static void release_tree_content_recursive(struct tree_content
*t
)
495 for (i
= 0; i
< t
->entry_count
; i
++)
496 release_tree_entry(t
->entries
[i
]);
497 release_tree_content(t
);
500 static struct tree_content
* grow_tree_content(
501 struct tree_content
*t
,
504 struct tree_content
*r
= new_tree_content(t
->entry_count
+ amt
);
505 r
->entry_count
= t
->entry_count
;
506 memcpy(r
->entries
,t
->entries
,t
->entry_count
*sizeof(t
->entries
[0]));
507 release_tree_content(t
);
511 static struct tree_entry
* new_tree_entry()
513 struct tree_entry
*e
;
515 if (!avail_tree_entry
) {
516 unsigned int n
= tree_entry_alloc
;
517 avail_tree_entry
= e
= xmalloc(n
* sizeof(struct tree_entry
));
519 *((void**)e
) = e
+ 1;
524 e
= avail_tree_entry
;
525 avail_tree_entry
= *((void**)e
);
529 static void release_tree_entry(struct tree_entry
*e
)
532 release_tree_content_recursive(e
->tree
);
533 *((void**)e
) = avail_tree_entry
;
534 avail_tree_entry
= e
;
537 static void yread(int fd
, void *buffer
, size_t length
)
540 while (ret
< length
) {
541 ssize_t size
= xread(fd
, (char *) buffer
+ ret
, length
- ret
);
543 die("Read from descriptor %i: end of stream", fd
);
545 die("Read from descriptor %i: %s", fd
, strerror(errno
));
550 static void ywrite(int fd
, void *buffer
, size_t length
)
553 while (ret
< length
) {
554 ssize_t size
= xwrite(fd
, (char *) buffer
+ ret
, length
- ret
);
556 die("Write to descriptor %i: end of file", fd
);
558 die("Write to descriptor %i: %s", fd
, strerror(errno
));
563 static size_t encode_header(
564 enum object_type type
,
571 if (type
< OBJ_COMMIT
|| type
> OBJ_DELTA
)
572 die("bad type %d", type
);
574 c
= (type
<< 4) | (size
& 15);
586 static int store_object(
587 enum object_type type
,
590 struct last_object
*last
,
591 unsigned char *sha1out
,
595 struct object_entry
*e
;
596 unsigned char hdr
[96];
597 unsigned char sha1
[20];
598 unsigned long hdrlen
, deltalen
;
602 hdrlen
= sprintf((char*)hdr
,"%s %lu",type_names
[type
],datlen
) + 1;
604 SHA1_Update(&c
, hdr
, hdrlen
);
605 SHA1_Update(&c
, dat
, datlen
);
606 SHA1_Final(sha1
, &c
);
608 memcpy(sha1out
, sha1
, sizeof(sha1
));
610 e
= insert_object(sha1
);
612 insert_mark(mark
, e
);
615 duplicate_count_by_type
[type
]++;
619 e
->offset
= pack_offset
;
621 object_count_by_type
[type
]++;
623 if (last
&& last
->data
&& last
->depth
< max_depth
)
624 delta
= diff_delta(last
->data
, last
->len
,
630 memset(&s
, 0, sizeof(s
));
631 deflateInit(&s
, zlib_compression_level
);
636 s
.avail_in
= deltalen
;
637 hdrlen
= encode_header(OBJ_DELTA
, deltalen
, hdr
);
638 ywrite(pack_fd
, hdr
, hdrlen
);
639 ywrite(pack_fd
, last
->sha1
, sizeof(sha1
));
640 pack_offset
+= hdrlen
+ sizeof(sha1
);
646 hdrlen
= encode_header(type
, datlen
, hdr
);
647 ywrite(pack_fd
, hdr
, hdrlen
);
648 pack_offset
+= hdrlen
;
651 s
.avail_out
= deflateBound(&s
, s
.avail_in
);
652 s
.next_out
= out
= xmalloc(s
.avail_out
);
653 while (deflate(&s
, Z_FINISH
) == Z_OK
)
657 ywrite(pack_fd
, out
, s
.total_out
);
658 pack_offset
+= s
.total_out
;
668 memcpy(last
->sha1
, sha1
, sizeof(sha1
));
673 static const char *get_mode(const char *str
, unsigned int *modep
)
676 unsigned int mode
= 0;
678 while ((c
= *str
++) != ' ') {
679 if (c
< '0' || c
> '7')
681 mode
= (mode
<< 3) + (c
- '0');
687 static void load_tree(struct tree_entry
*root
)
689 struct object_entry
*myoe
;
690 struct tree_content
*t
;
696 root
->tree
= t
= new_tree_content(8);
697 if (!memcmp(root
->sha1
, null_sha1
, 20))
700 myoe
= find_object(root
->sha1
);
704 buf
= read_sha1_file(root
->sha1
, type
, &size
);
705 if (!buf
|| strcmp(type
, tree_type
))
706 die("Can't load existing tree %s", sha1_to_hex(root
->sha1
));
710 while (c
!= (buf
+ size
)) {
711 struct tree_entry
*e
= new_tree_entry();
713 if (t
->entry_count
== t
->entry_capacity
)
714 root
->tree
= t
= grow_tree_content(t
, 8);
715 t
->entries
[t
->entry_count
++] = e
;
718 c
= get_mode(c
, &e
->mode
);
720 die("Corrupt mode in %s", sha1_to_hex(root
->sha1
));
721 e
->name
= to_atom(c
, strlen(c
));
722 c
+= e
->name
->str_len
+ 1;
723 memcpy(e
->sha1
, c
, sizeof(e
->sha1
));
729 static int tecmp (const void *_a
, const void *_b
)
731 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
732 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
733 return base_name_compare(
734 a
->name
->str_dat
, a
->name
->str_len
, a
->mode
,
735 b
->name
->str_dat
, b
->name
->str_len
, b
->mode
);
738 static void store_tree(struct tree_entry
*root
)
740 struct tree_content
*t
= root
->tree
;
745 if (memcmp(root
->sha1
, null_sha1
, 20))
749 for (i
= 0; i
< t
->entry_count
; i
++) {
750 maxlen
+= t
->entries
[i
]->name
->str_len
+ 34;
751 if (t
->entries
[i
]->tree
)
752 store_tree(t
->entries
[i
]);
755 qsort(t
->entries
, t
->entry_count
, sizeof(t
->entries
[0]), tecmp
);
756 buf
= c
= xmalloc(maxlen
);
757 for (i
= 0; i
< t
->entry_count
; i
++) {
758 struct tree_entry
*e
= t
->entries
[i
];
759 c
+= sprintf(c
, "%o", e
->mode
);
761 strcpy(c
, e
->name
->str_dat
);
762 c
+= e
->name
->str_len
+ 1;
763 memcpy(c
, e
->sha1
, 20);
766 store_object(OBJ_TREE
, buf
, c
- buf
, NULL
, root
->sha1
, 0);
770 static int tree_content_set(
771 struct tree_entry
*root
,
773 const unsigned char *sha1
,
774 const unsigned int mode
)
776 struct tree_content
*t
= root
->tree
;
779 struct tree_entry
*e
;
781 slash1
= strchr(p
, '/');
787 for (i
= 0; i
< t
->entry_count
; i
++) {
789 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
791 if (e
->mode
== mode
&& !memcmp(e
->sha1
, sha1
, 20))
794 memcpy(e
->sha1
, sha1
, 20);
796 release_tree_content_recursive(e
->tree
);
799 memcpy(root
->sha1
, null_sha1
, 20);
802 if (!S_ISDIR(e
->mode
)) {
803 e
->tree
= new_tree_content(8);
808 if (tree_content_set(e
, slash1
+ 1, sha1
, mode
)) {
809 memcpy(root
->sha1
, null_sha1
, 20);
816 if (t
->entry_count
== t
->entry_capacity
)
817 root
->tree
= t
= grow_tree_content(t
, 8);
818 e
= new_tree_entry();
819 e
->name
= to_atom(p
, n
);
820 t
->entries
[t
->entry_count
++] = e
;
822 e
->tree
= new_tree_content(8);
824 tree_content_set(e
, slash1
+ 1, sha1
, mode
);
828 memcpy(e
->sha1
, sha1
, 20);
830 memcpy(root
->sha1
, null_sha1
, 20);
834 static int tree_content_remove(struct tree_entry
*root
, const char *p
)
836 struct tree_content
*t
= root
->tree
;
839 struct tree_entry
*e
;
841 slash1
= strchr(p
, '/');
847 for (i
= 0; i
< t
->entry_count
; i
++) {
849 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
850 if (!slash1
|| !S_ISDIR(e
->mode
))
854 if (tree_content_remove(e
, slash1
+ 1)) {
855 if (!e
->tree
->entry_count
)
857 memcpy(root
->sha1
, null_sha1
, 20);
866 for (i
++; i
< t
->entry_count
; i
++)
867 t
->entries
[i
-1] = t
->entries
[i
];
869 release_tree_entry(e
);
870 memcpy(root
->sha1
, null_sha1
, 20);
874 static void init_pack_header()
876 struct pack_header hdr
;
878 hdr
.hdr_signature
= htonl(PACK_SIGNATURE
);
879 hdr
.hdr_version
= htonl(2);
882 ywrite(pack_fd
, &hdr
, sizeof(hdr
));
883 pack_offset
= sizeof(hdr
);
886 static void fixup_header_footer()
894 if (lseek(pack_fd
, 0, SEEK_SET
) != 0)
895 die("Failed seeking to start: %s", strerror(errno
));
898 yread(pack_fd
, hdr
, 8);
899 SHA1_Update(&c
, hdr
, 8);
901 cnt
= htonl(object_count
);
902 SHA1_Update(&c
, &cnt
, 4);
903 ywrite(pack_fd
, &cnt
, 4);
905 buf
= xmalloc(128 * 1024);
907 n
= xread(pack_fd
, buf
, 128 * 1024);
910 SHA1_Update(&c
, buf
, n
);
914 SHA1_Final(pack_sha1
, &c
);
915 ywrite(pack_fd
, pack_sha1
, sizeof(pack_sha1
));
918 static int oecmp (const void *_a
, const void *_b
)
920 struct object_entry
*a
= *((struct object_entry
**)_a
);
921 struct object_entry
*b
= *((struct object_entry
**)_b
);
922 return memcmp(a
->sha1
, b
->sha1
, sizeof(a
->sha1
));
925 static void write_index(const char *idx_name
)
928 struct object_entry
**idx
, **c
, **last
;
929 struct object_entry
*e
;
930 struct object_entry_pool
*o
;
931 unsigned int array
[256];
934 /* Build the sorted table of object IDs. */
935 idx
= xmalloc(object_count
* sizeof(struct object_entry
*));
937 for (o
= blocks
; o
; o
= o
->next_pool
)
938 for (e
= o
->entries
; e
!= o
->next_free
; e
++)
940 last
= idx
+ object_count
;
941 qsort(idx
, object_count
, sizeof(struct object_entry
*), oecmp
);
943 /* Generate the fan-out array. */
945 for (i
= 0; i
< 256; i
++) {
946 struct object_entry
**next
= c
;;
947 while (next
< last
) {
948 if ((*next
)->sha1
[0] != i
)
952 array
[i
] = htonl(next
- idx
);
956 f
= sha1create("%s", idx_name
);
957 sha1write(f
, array
, 256 * sizeof(int));
958 for (c
= idx
; c
!= last
; c
++) {
959 unsigned int offset
= htonl((*c
)->offset
);
960 sha1write(f
, &offset
, 4);
961 sha1write(f
, (*c
)->sha1
, sizeof((*c
)->sha1
));
963 sha1write(f
, pack_sha1
, sizeof(pack_sha1
));
964 sha1close(f
, NULL
, 1);
968 static void dump_branches()
970 static const char *msg
= "fast-import";
973 struct ref_lock
*lock
;
975 for (i
= 0; i
< branch_table_sz
; i
++) {
976 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
) {
977 lock
= lock_any_ref_for_update(b
->name
, NULL
, 0);
978 if (!lock
|| write_ref_sha1(lock
, b
->sha1
, msg
) < 0)
979 die("Can't write %s", b
->name
);
984 static void dump_tags()
986 static const char *msg
= "fast-import";
988 struct ref_lock
*lock
;
991 for (t
= first_tag
; t
; t
= t
->next_tag
) {
992 sprintf(path
, "refs/tags/%s", t
->name
);
993 lock
= lock_any_ref_for_update(path
, NULL
, 0);
994 if (!lock
|| write_ref_sha1(lock
, t
->sha1
, msg
) < 0)
995 die("Can't write %s", path
);
999 static void read_next_command()
1001 read_line(&command_buf
, stdin
, '\n');
1004 static void cmd_mark()
1006 if (!strncmp("mark :", command_buf
.buf
, 6)) {
1007 next_mark
= strtoul(command_buf
.buf
+ 6, NULL
, 10);
1008 read_next_command();
1014 static void* cmd_data (size_t *size
)
1020 if (strncmp("data ", command_buf
.buf
, 5))
1021 die("Expected 'data n' command, found: %s", command_buf
.buf
);
1023 length
= strtoul(command_buf
.buf
+ 5, NULL
, 10);
1024 buffer
= xmalloc(length
);
1026 while (n
< length
) {
1027 size_t s
= fread((char*)buffer
+ n
, 1, length
- n
, stdin
);
1028 if (!s
&& feof(stdin
))
1029 die("EOF in data (%lu bytes remaining)", length
- n
);
1033 if (fgetc(stdin
) != '\n')
1034 die("An lf did not trail the binary data as expected.");
1040 static void cmd_new_blob()
1045 read_next_command();
1049 if (store_object(OBJ_BLOB
, d
, l
, &last_blob
, NULL
, next_mark
))
1053 static void unload_one_branch()
1055 while (cur_active_branches
>= max_active_branches
) {
1056 unsigned long min_commit
= ULONG_MAX
;
1057 struct branch
*e
, *l
= NULL
, *p
= NULL
;
1059 for (e
= active_branches
; e
; e
= e
->active_next_branch
) {
1060 if (e
->last_commit
< min_commit
) {
1062 min_commit
= e
->last_commit
;
1068 e
= p
->active_next_branch
;
1069 p
->active_next_branch
= e
->active_next_branch
;
1071 e
= active_branches
;
1072 active_branches
= e
->active_next_branch
;
1074 e
->active_next_branch
= NULL
;
1075 if (e
->branch_tree
.tree
) {
1076 release_tree_content_recursive(e
->branch_tree
.tree
);
1077 e
->branch_tree
.tree
= NULL
;
1079 cur_active_branches
--;
1083 static void load_branch(struct branch
*b
)
1085 load_tree(&b
->branch_tree
);
1086 b
->active_next_branch
= active_branches
;
1087 active_branches
= b
;
1088 cur_active_branches
++;
1089 branch_load_count
++;
1092 static void file_change_m(struct branch
*b
)
1094 const char *p
= command_buf
.buf
+ 2;
1097 struct object_entry
*oe
;
1098 unsigned char sha1
[20];
1102 p
= get_mode(p
, &mode
);
1104 die("Corrupt mode: %s", command_buf
.buf
);
1106 case S_IFREG
| 0644:
1107 case S_IFREG
| 0755:
1114 die("Corrupt mode: %s", command_buf
.buf
);
1119 oe
= find_mark(strtoul(p
+ 1, &x
, 10));
1122 if (get_sha1_hex(p
, sha1
))
1123 die("Invalid SHA1: %s", command_buf
.buf
);
1124 oe
= find_object(sha1
);
1128 die("Missing space after SHA1: %s", command_buf
.buf
);
1130 p_uq
= unquote_c_style(p
, &endp
);
1133 die("Garbage after path in: %s", command_buf
.buf
);
1138 if (oe
->type
!= OBJ_BLOB
)
1139 die("Not a blob (actually a %s): %s",
1140 command_buf
.buf
, type_names
[oe
->type
]);
1142 if (sha1_object_info(sha1
, type
, NULL
))
1143 die("Blob not found: %s", command_buf
.buf
);
1144 if (strcmp(blob_type
, type
))
1145 die("Not a blob (actually a %s): %s",
1146 command_buf
.buf
, type
);
1149 tree_content_set(&b
->branch_tree
, p
, sha1
, S_IFREG
| mode
);
1155 static void file_change_d(struct branch
*b
)
1157 const char *p
= command_buf
.buf
+ 2;
1161 p_uq
= unquote_c_style(p
, &endp
);
1164 die("Garbage after path in: %s", command_buf
.buf
);
1167 tree_content_remove(&b
->branch_tree
, p
);
1172 static void cmd_new_commit()
1180 char *author
= NULL
;
1181 char *committer
= NULL
;
1184 /* Obtain the branch name from the rest of our command */
1185 sp
= strchr(command_buf
.buf
, ' ') + 1;
1186 str_uq
= unquote_c_style(sp
, &endp
);
1189 die("Garbage after ref in: %s", command_buf
.buf
);
1192 b
= lookup_branch(sp
);
1194 die("Branch not declared: %s", sp
);
1198 read_next_command();
1200 if (!strncmp("author ", command_buf
.buf
, 7)) {
1201 author
= strdup(command_buf
.buf
);
1202 read_next_command();
1204 if (!strncmp("committer ", command_buf
.buf
, 10)) {
1205 committer
= strdup(command_buf
.buf
);
1206 read_next_command();
1209 die("Expected committer but didn't get one");
1210 msg
= cmd_data(&msglen
);
1212 /* ensure the branch is active/loaded */
1213 if (!b
->branch_tree
.tree
) {
1214 unload_one_branch();
1220 read_next_command();
1221 if (1 == command_buf
.len
)
1223 else if (!strncmp("M ", command_buf
.buf
, 2))
1225 else if (!strncmp("D ", command_buf
.buf
, 2))
1228 die("Unsupported file_change: %s", command_buf
.buf
);
1231 /* build the tree and the commit */
1232 store_tree(&b
->branch_tree
);
1233 body
= xmalloc(97 + msglen
1235 ? strlen(author
) + strlen(committer
)
1236 : 2 * strlen(committer
)));
1238 sp
+= sprintf(sp
, "tree %s\n", sha1_to_hex(b
->branch_tree
.sha1
));
1239 if (memcmp(b
->sha1
, null_sha1
, 20))
1240 sp
+= sprintf(sp
, "parent %s\n", sha1_to_hex(b
->sha1
));
1242 sp
+= sprintf(sp
, "%s\n", author
);
1244 sp
+= sprintf(sp
, "author %s\n", committer
+ 10);
1245 sp
+= sprintf(sp
, "%s\n\n", committer
);
1246 memcpy(sp
, msg
, msglen
);
1253 store_object(OBJ_COMMIT
, body
, sp
- body
, NULL
, b
->sha1
, next_mark
);
1255 b
->last_commit
= object_count_by_type
[OBJ_COMMIT
];
1258 static void cmd_new_branch()
1265 /* Obtain the new branch name from the rest of our command */
1266 sp
= strchr(command_buf
.buf
, ' ') + 1;
1267 str_uq
= unquote_c_style(sp
, &endp
);
1270 die("Garbage after ref in: %s", command_buf
.buf
);
1276 read_next_command();
1279 if (!strncmp("from ", command_buf
.buf
, 5)) {
1283 from
= strchr(command_buf
.buf
, ' ') + 1;
1284 str_uq
= unquote_c_style(from
, &endp
);
1287 die("Garbage after string in: %s", command_buf
.buf
);
1291 s
= lookup_branch(from
);
1293 die("Can't create a branch from itself: %s", b
->name
);
1295 memcpy(b
->sha1
, s
->sha1
, 20);
1296 memcpy(b
->branch_tree
.sha1
, s
->branch_tree
.sha1
, 20);
1297 } else if (*from
== ':') {
1298 unsigned long idnum
= strtoul(from
+ 1, NULL
, 10);
1299 struct object_entry
*oe
= find_mark(idnum
);
1300 if (oe
->type
!= OBJ_COMMIT
)
1301 die("Mark :%lu not a commit", idnum
);
1302 memcpy(b
->sha1
, oe
->sha1
, 20);
1303 memcpy(b
->branch_tree
.sha1
, null_sha1
, 20);
1304 } else if (!get_sha1(from
, b
->sha1
)) {
1305 if (!memcmp(b
->sha1
, null_sha1
, 20))
1306 memcpy(b
->branch_tree
.sha1
, null_sha1
, 20);
1311 buf
= read_object_with_reference(b
->sha1
,
1312 type_names
[OBJ_COMMIT
], &size
, b
->sha1
);
1313 if (!buf
|| size
< 46)
1314 die("Not a valid commit: %s", from
);
1315 if (memcmp("tree ", buf
, 5)
1316 || get_sha1_hex(buf
+ 5, b
->branch_tree
.sha1
))
1317 die("The commit %s is corrupt", sha1_to_hex(b
->sha1
));
1321 die("Invalid ref name or SHA1 expression: %s", from
);
1325 read_next_command();
1327 memcpy(b
->sha1
, null_sha1
, 20);
1328 memcpy(b
->branch_tree
.sha1
, null_sha1
, 20);
1331 if (command_buf
.eof
|| command_buf
.len
> 1)
1332 die("An lf did not terminate the branch command as expected.");
1335 static void cmd_new_tag()
1347 unsigned char sha1
[20];
1349 /* Obtain the new tag name from the rest of our command */
1350 sp
= strchr(command_buf
.buf
, ' ') + 1;
1351 str_uq
= unquote_c_style(sp
, &endp
);
1354 die("Garbage after tag name in: %s", command_buf
.buf
);
1357 t
= pool_alloc(sizeof(struct tag
));
1359 t
->name
= pool_strdup(sp
);
1361 last_tag
->next_tag
= t
;
1367 read_next_command();
1370 if (strncmp("from ", command_buf
.buf
, 5))
1371 die("Expected from command, got %s", command_buf
.buf
);
1373 from
= strchr(command_buf
.buf
, ' ') + 1;
1374 str_uq
= unquote_c_style(from
, &endp
);
1377 die("Garbage after string in: %s", command_buf
.buf
);
1381 s
= lookup_branch(from
);
1383 memcpy(sha1
, s
->sha1
, 20);
1384 } else if (*from
== ':') {
1385 unsigned long idnum
= strtoul(from
+ 1, NULL
, 10);
1386 struct object_entry
*oe
= find_mark(idnum
);
1387 if (oe
->type
!= OBJ_COMMIT
)
1388 die("Mark :%lu not a commit", idnum
);
1389 memcpy(sha1
, oe
->sha1
, 20);
1390 } else if (!get_sha1(from
, sha1
)) {
1394 buf
= read_object_with_reference(sha1
,
1395 type_names
[OBJ_COMMIT
], &size
, sha1
);
1396 if (!buf
|| size
< 46)
1397 die("Not a valid commit: %s", from
);
1400 die("Invalid ref name or SHA1 expression: %s", from
);
1404 read_next_command();
1407 if (strncmp("tagger ", command_buf
.buf
, 7))
1408 die("Expected tagger command, got %s", command_buf
.buf
);
1409 tagger
= strdup(command_buf
.buf
);
1411 /* tag payload/message */
1412 read_next_command();
1413 msg
= cmd_data(&msglen
);
1415 /* build the tag object */
1416 body
= xmalloc(67 + strlen(t
->name
) + strlen(tagger
) + msglen
);
1418 sp
+= sprintf(sp
, "object %s\n", sha1_to_hex(sha1
));
1419 sp
+= sprintf(sp
, "type %s\n", type_names
[OBJ_COMMIT
]);
1420 sp
+= sprintf(sp
, "tag %s\n", t
->name
);
1421 sp
+= sprintf(sp
, "%s\n\n", tagger
);
1422 memcpy(sp
, msg
, msglen
);
1427 store_object(OBJ_TAG
, body
, sp
- body
, NULL
, t
->sha1
, 0);
1431 static const char fast_import_usage
[] =
1432 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] temp.pack";
1434 int main(int argc
, const char **argv
)
1436 const char *base_name
;
1438 unsigned long est_obj_cnt
= 1000;
1444 git_config(git_default_config
);
1446 for (i
= 1; i
< argc
; i
++) {
1447 const char *a
= argv
[i
];
1449 if (*a
!= '-' || !strcmp(a
, "--"))
1451 else if (!strncmp(a
, "--objects=", 10))
1452 est_obj_cnt
= strtoul(a
+ 10, NULL
, 0);
1453 else if (!strncmp(a
, "--depth=", 8))
1454 max_depth
= strtoul(a
+ 8, NULL
, 0);
1455 else if (!strncmp(a
, "--active-branches=", 18))
1456 max_active_branches
= strtoul(a
+ 18, NULL
, 0);
1458 die("unknown option %s", a
);
1461 usage(fast_import_usage
);
1462 base_name
= argv
[i
];
1464 pack_name
= xmalloc(strlen(base_name
) + 6);
1465 sprintf(pack_name
, "%s.pack", base_name
);
1466 idx_name
= xmalloc(strlen(base_name
) + 5);
1467 sprintf(idx_name
, "%s.idx", base_name
);
1469 pack_fd
= open(pack_name
, O_RDWR
|O_CREAT
|O_EXCL
, 0666);
1471 die("Can't create %s: %s", pack_name
, strerror(errno
));
1474 alloc_objects(est_obj_cnt
);
1475 strbuf_init(&command_buf
);
1477 atom_table
= xcalloc(atom_table_sz
, sizeof(struct atom_str
*));
1478 branch_table
= xcalloc(branch_table_sz
, sizeof(struct branch
*));
1479 avail_tree_table
= xcalloc(avail_tree_table_sz
, sizeof(struct avail_tree_content
*));
1480 marks
= pool_calloc(1, sizeof(struct mark_set
));
1483 read_next_command();
1484 if (command_buf
.eof
)
1486 else if (!strcmp("blob", command_buf
.buf
))
1488 else if (!strncmp("branch ", command_buf
.buf
, 7))
1490 else if (!strncmp("commit ", command_buf
.buf
, 7))
1492 else if (!strncmp("tag ", command_buf
.buf
, 4))
1495 die("Unsupported command: %s", command_buf
.buf
);
1498 fixup_header_footer();
1500 write_index(idx_name
);
1504 fprintf(stderr
, "%s statistics:\n", argv
[0]);
1505 fprintf(stderr
, "---------------------------------------------------\n");
1506 fprintf(stderr
, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count
, alloc_count
- est_obj_cnt
);
1507 fprintf(stderr
, "Total objects: %10lu (%10lu duplicates)\n", object_count
, duplicate_count
);
1508 fprintf(stderr
, " blobs : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_BLOB
], duplicate_count_by_type
[OBJ_BLOB
]);
1509 fprintf(stderr
, " trees : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_TREE
], duplicate_count_by_type
[OBJ_TREE
]);
1510 fprintf(stderr
, " commits: %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_COMMIT
], duplicate_count_by_type
[OBJ_COMMIT
]);
1511 fprintf(stderr
, " tags : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_TAG
], duplicate_count_by_type
[OBJ_TAG
]);
1512 fprintf(stderr
, "Total branches: %10lu (%10lu loads )\n", branch_count
, branch_load_count
);
1513 fprintf(stderr
, " marks: %10u (%10lu unique )\n", (1 << marks
->shift
) * 1024, marks_set_count
);
1514 fprintf(stderr
, " atoms: %10u\n", atom_cnt
);
1515 fprintf(stderr
, "Memory total: %10lu KiB\n", (total_allocd
+ alloc_count
*sizeof(struct object_entry
))/1024);
1516 fprintf(stderr
, " pools: %10lu KiB\n", total_allocd
/1024);
1517 fprintf(stderr
, " objects: %10lu KiB\n", (alloc_count
*sizeof(struct object_entry
))/1024);
1518 fprintf(stderr
, "---------------------------------------------------\n");
1520 stat(pack_name
, &sb
);
1521 fprintf(stderr
, "Pack size: %10lu KiB\n", (unsigned long)(sb
.st_size
/1024));
1522 stat(idx_name
, &sb
);
1523 fprintf(stderr
, "Index size: %10lu KiB\n", (unsigned long)(sb
.st_size
/1024));
1525 fprintf(stderr
, "\n");