]>
git.ipfire.org Git - thirdparty/git.git/blob - fast-import.c
2 Format of STDIN stream:
11 new_blob ::= 'blob' lf
14 file_content ::= data;
16 new_commit ::= 'commit' sp ref_str lf
18 ('author' sp name '<' email '>' ts tz lf)?
19 'committer' sp name '<' email '>' ts tz lf
21 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
26 file_change ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf
29 mode ::= '644' | '755';
31 new_tag ::= 'tag' sp tag_str lf
32 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
33 'tagger' sp name '<' email '>' ts tz lf
37 # note: the first idnum in a stream should be 1 and subsequent
38 # idnums should not have gaps between values as this will cause
39 # the stream parser to reserve space for the gapped values. An
40 # idnum can be updated in the future to a new object by issuing
41 # a new mark directive with the old idnum.
43 mark ::= 'mark' sp idnum lf;
45 # note: declen indicates the length of binary_data in bytes.
46 # declen does not include the lf preceeding or trailing the
49 data ::= 'data' sp declen lf
53 # note: quoted strings are C-style quoting supporting \c for
54 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
55 # is the signed byte value in octal. Note that the only
56 # characters which must actually be escaped to protect the
57 # stream formatting is: \, " and LF. Otherwise these values
60 ref_str ::= ref | '"' quoted(ref) '"' ;
61 sha1exp_str ::= sha1exp | '"' quoted(sha1exp) '"' ;
62 tag_str ::= tag | '"' quoted(tag) '"' ;
63 path_str ::= path | '"' quoted(path) '"' ;
65 declen ::= # unsigned 32 bit value, ascii base10 notation;
66 binary_data ::= # file content, not interpreted;
68 sp ::= # ASCII space character;
69 lf ::= # ASCII newline (LF) character;
71 # note: a colon (':') must precede the numerical value assigned to
72 # an idnum. This is to distinguish it from a ref or tag name as
73 # GIT does not permit ':' in ref or tag strings.
76 path ::= # GIT style file path, e.g. "a/b/c";
77 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
78 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
79 sha1exp ::= # Any valid GIT SHA1 expression;
80 hexsha1 ::= # SHA1 in hexadecimal format;
82 # note: name and email are UTF8 strings, however name must not
83 # contain '<' or lf and email must not contain any of the
84 # following: '<', '>', lf.
86 name ::= # valid GIT author/committer name;
87 email ::= # valid GIT author/committer email;
88 ts ::= # time since the epoch in seconds, ascii base10 notation;
89 tz ::= # GIT style timezone;
100 #include "csum-file.h"
106 struct object_entry
*next
;
107 enum object_type type
;
108 unsigned long offset
;
109 unsigned char sha1
[20];
112 struct object_entry_pool
114 struct object_entry_pool
*next_pool
;
115 struct object_entry
*next_free
;
116 struct object_entry
*end
;
117 struct object_entry entries
[FLEX_ARRAY
]; /* more */
124 struct object_entry
*marked
[1024];
125 struct mark_set
*sets
[1024];
134 unsigned char sha1
[20];
139 struct mem_pool
*next_pool
;
142 char space
[FLEX_ARRAY
]; /* more */
147 struct atom_str
*next_atom
;
149 char str_dat
[FLEX_ARRAY
]; /* more */
155 struct tree_content
*tree
;
156 struct atom_str
* name
;
158 unsigned char sha1
[20];
163 unsigned int entry_capacity
; /* must match avail_tree_content */
164 unsigned int entry_count
;
165 struct tree_entry
*entries
[FLEX_ARRAY
]; /* more */
168 struct avail_tree_content
170 unsigned int entry_capacity
; /* must match tree_content */
171 struct avail_tree_content
*next_avail
;
176 struct branch
*table_next_branch
;
177 struct branch
*active_next_branch
;
179 unsigned long last_commit
;
180 struct tree_entry branch_tree
;
181 unsigned char sha1
[20];
186 struct tag
*next_tag
;
188 unsigned char sha1
[20];
192 /* Stats and misc. counters */
193 static unsigned long max_depth
= 10;
194 static unsigned long alloc_count
;
195 static unsigned long branch_count
;
196 static unsigned long branch_load_count
;
197 static unsigned long remap_count
;
198 static unsigned long object_count
;
199 static unsigned long duplicate_count
;
200 static unsigned long marks_set_count
;
201 static unsigned long object_count_by_type
[9];
202 static unsigned long duplicate_count_by_type
[9];
205 static size_t mem_pool_alloc
= 2*1024*1024 - sizeof(struct mem_pool
);
206 static size_t total_allocd
;
207 static struct mem_pool
*mem_pool
;
209 /* Atom management */
210 static unsigned int atom_table_sz
= 4451;
211 static unsigned int atom_cnt
;
212 static struct atom_str
**atom_table
;
214 /* The .pack file being generated */
216 static unsigned long pack_size
;
217 static unsigned char pack_sha1
[20];
218 static unsigned char* pack_base
;
219 static unsigned long pack_moff
;
220 static unsigned long pack_mlen
= 128*1024*1024;
221 static unsigned long page_size
;
223 /* Table of objects we've written. */
224 static unsigned int object_entry_alloc
= 1000;
225 static struct object_entry_pool
*blocks
;
226 static struct object_entry
*object_table
[1 << 16];
227 static struct mark_set
*marks
;
228 static const char* mark_file
;
231 static struct last_object last_blob
;
233 /* Tree management */
234 static unsigned int tree_entry_alloc
= 1000;
235 static void *avail_tree_entry
;
236 static unsigned int avail_tree_table_sz
= 100;
237 static struct avail_tree_content
**avail_tree_table
;
240 static unsigned long max_active_branches
= 5;
241 static unsigned long cur_active_branches
;
242 static unsigned long branch_table_sz
= 1039;
243 static struct branch
**branch_table
;
244 static struct branch
*active_branches
;
247 static struct tag
*first_tag
;
248 static struct tag
*last_tag
;
250 /* Input stream parsing */
251 static struct strbuf command_buf
;
252 static unsigned long next_mark
;
253 static FILE* branch_log
;
256 static void alloc_objects(int cnt
)
258 struct object_entry_pool
*b
;
260 b
= xmalloc(sizeof(struct object_entry_pool
)
261 + cnt
* sizeof(struct object_entry
));
262 b
->next_pool
= blocks
;
263 b
->next_free
= b
->entries
;
264 b
->end
= b
->entries
+ cnt
;
269 static struct object_entry
* new_object(unsigned char *sha1
)
271 struct object_entry
*e
;
273 if (blocks
->next_free
== blocks
->end
)
274 alloc_objects(object_entry_alloc
);
276 e
= blocks
->next_free
++;
277 memcpy(e
->sha1
, sha1
, sizeof(e
->sha1
));
281 static struct object_entry
* find_object(unsigned char *sha1
)
283 unsigned int h
= sha1
[0] << 8 | sha1
[1];
284 struct object_entry
*e
;
285 for (e
= object_table
[h
]; e
; e
= e
->next
)
286 if (!memcmp(sha1
, e
->sha1
, sizeof(e
->sha1
)))
291 static struct object_entry
* insert_object(unsigned char *sha1
)
293 unsigned int h
= sha1
[0] << 8 | sha1
[1];
294 struct object_entry
*e
= object_table
[h
];
295 struct object_entry
*p
= NULL
;
298 if (!memcmp(sha1
, e
->sha1
, sizeof(e
->sha1
)))
304 e
= new_object(sha1
);
314 static unsigned int hc_str(const char *s
, size_t len
)
322 static void* pool_alloc(size_t len
)
327 for (p
= mem_pool
; p
; p
= p
->next_pool
)
328 if ((p
->end
- p
->next_free
>= len
))
332 if (len
>= (mem_pool_alloc
/2)) {
336 total_allocd
+= sizeof(struct mem_pool
) + mem_pool_alloc
;
337 p
= xmalloc(sizeof(struct mem_pool
) + mem_pool_alloc
);
338 p
->next_pool
= mem_pool
;
339 p
->next_free
= p
->space
;
340 p
->end
= p
->next_free
+ mem_pool_alloc
;
345 /* round out to a pointer alignment */
346 if (len
& (sizeof(void*) - 1))
347 len
+= sizeof(void*) - (len
& (sizeof(void*) - 1));
352 static void* pool_calloc(size_t count
, size_t size
)
354 size_t len
= count
* size
;
355 void *r
= pool_alloc(len
);
360 static char* pool_strdup(const char *s
)
362 char *r
= pool_alloc(strlen(s
) + 1);
367 static void insert_mark(unsigned long idnum
, struct object_entry
*oe
)
369 struct mark_set
*s
= marks
;
370 while ((idnum
>> s
->shift
) >= 1024) {
371 s
= pool_calloc(1, sizeof(struct mark_set
));
372 s
->shift
= marks
->shift
+ 10;
373 s
->data
.sets
[0] = marks
;
377 unsigned long i
= idnum
>> s
->shift
;
378 idnum
-= i
<< s
->shift
;
379 if (!s
->data
.sets
[i
]) {
380 s
->data
.sets
[i
] = pool_calloc(1, sizeof(struct mark_set
));
381 s
->data
.sets
[i
]->shift
= s
->shift
- 10;
385 if (!s
->data
.marked
[idnum
])
387 s
->data
.marked
[idnum
] = oe
;
390 static struct object_entry
* find_mark(unsigned long idnum
)
392 unsigned long orig_idnum
= idnum
;
393 struct mark_set
*s
= marks
;
394 struct object_entry
*oe
= NULL
;
395 if ((idnum
>> s
->shift
) < 1024) {
396 while (s
&& s
->shift
) {
397 unsigned long i
= idnum
>> s
->shift
;
398 idnum
-= i
<< s
->shift
;
402 oe
= s
->data
.marked
[idnum
];
405 die("mark :%lu not declared", orig_idnum
);
409 static struct atom_str
* to_atom(const char *s
, size_t len
)
411 unsigned int hc
= hc_str(s
, len
) % atom_table_sz
;
414 for (c
= atom_table
[hc
]; c
; c
= c
->next_atom
)
415 if (c
->str_len
== len
&& !strncmp(s
, c
->str_dat
, len
))
418 c
= pool_alloc(sizeof(struct atom_str
) + len
+ 1);
420 strncpy(c
->str_dat
, s
, len
);
422 c
->next_atom
= atom_table
[hc
];
428 static struct branch
* lookup_branch(const char *name
)
430 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
433 for (b
= branch_table
[hc
]; b
; b
= b
->table_next_branch
)
434 if (!strcmp(name
, b
->name
))
439 static struct branch
* new_branch(const char *name
)
441 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
442 struct branch
* b
= lookup_branch(name
);
445 die("Invalid attempt to create duplicate branch: %s", name
);
446 if (check_ref_format(name
))
447 die("Branch name doesn't conform to GIT standards: %s", name
);
449 b
= pool_calloc(1, sizeof(struct branch
));
450 b
->name
= pool_strdup(name
);
451 b
->table_next_branch
= branch_table
[hc
];
452 branch_table
[hc
] = b
;
457 static unsigned int hc_entries(unsigned int cnt
)
459 cnt
= cnt
& 7 ? (cnt
/ 8) + 1 : cnt
/ 8;
460 return cnt
< avail_tree_table_sz
? cnt
: avail_tree_table_sz
- 1;
463 static struct tree_content
* new_tree_content(unsigned int cnt
)
465 struct avail_tree_content
*f
, *l
= NULL
;
466 struct tree_content
*t
;
467 unsigned int hc
= hc_entries(cnt
);
469 for (f
= avail_tree_table
[hc
]; f
; l
= f
, f
= f
->next_avail
)
470 if (f
->entry_capacity
>= cnt
)
475 l
->next_avail
= f
->next_avail
;
477 avail_tree_table
[hc
] = f
->next_avail
;
479 cnt
= cnt
& 7 ? ((cnt
/ 8) + 1) * 8 : cnt
;
480 f
= pool_alloc(sizeof(*t
) + sizeof(t
->entries
[0]) * cnt
);
481 f
->entry_capacity
= cnt
;
484 t
= (struct tree_content
*)f
;
489 static void release_tree_entry(struct tree_entry
*e
);
490 static void release_tree_content(struct tree_content
*t
)
492 struct avail_tree_content
*f
= (struct avail_tree_content
*)t
;
493 unsigned int hc
= hc_entries(f
->entry_capacity
);
494 f
->next_avail
= avail_tree_table
[hc
];
495 avail_tree_table
[hc
] = f
;
498 static void release_tree_content_recursive(struct tree_content
*t
)
501 for (i
= 0; i
< t
->entry_count
; i
++)
502 release_tree_entry(t
->entries
[i
]);
503 release_tree_content(t
);
506 static struct tree_content
* grow_tree_content(
507 struct tree_content
*t
,
510 struct tree_content
*r
= new_tree_content(t
->entry_count
+ amt
);
511 r
->entry_count
= t
->entry_count
;
512 memcpy(r
->entries
,t
->entries
,t
->entry_count
*sizeof(t
->entries
[0]));
513 release_tree_content(t
);
517 static struct tree_entry
* new_tree_entry()
519 struct tree_entry
*e
;
521 if (!avail_tree_entry
) {
522 unsigned int n
= tree_entry_alloc
;
523 total_allocd
+= n
* sizeof(struct tree_entry
);
524 avail_tree_entry
= e
= xmalloc(n
* sizeof(struct tree_entry
));
526 *((void**)e
) = e
+ 1;
532 e
= avail_tree_entry
;
533 avail_tree_entry
= *((void**)e
);
537 static void release_tree_entry(struct tree_entry
*e
)
540 release_tree_content_recursive(e
->tree
);
541 *((void**)e
) = avail_tree_entry
;
542 avail_tree_entry
= e
;
545 static void yread(int fd
, void *buffer
, size_t length
)
548 while (ret
< length
) {
549 ssize_t size
= xread(fd
, (char *) buffer
+ ret
, length
- ret
);
551 die("Read from descriptor %i: end of stream", fd
);
553 die("Read from descriptor %i: %s", fd
, strerror(errno
));
558 static void ywrite(int fd
, void *buffer
, size_t length
)
561 while (ret
< length
) {
562 ssize_t size
= xwrite(fd
, (char *) buffer
+ ret
, length
- ret
);
564 die("Write to descriptor %i: end of file", fd
);
566 die("Write to descriptor %i: %s", fd
, strerror(errno
));
571 static size_t encode_header(
572 enum object_type type
,
579 if (type
< OBJ_COMMIT
|| type
> OBJ_DELTA
)
580 die("bad type %d", type
);
582 c
= (type
<< 4) | (size
& 15);
594 static int store_object(
595 enum object_type type
,
598 struct last_object
*last
,
599 unsigned char *sha1out
,
603 struct object_entry
*e
;
604 unsigned char hdr
[96];
605 unsigned char sha1
[20];
606 unsigned long hdrlen
, deltalen
;
610 hdrlen
= sprintf((char*)hdr
,"%s %lu",type_names
[type
],datlen
) + 1;
612 SHA1_Update(&c
, hdr
, hdrlen
);
613 SHA1_Update(&c
, dat
, datlen
);
614 SHA1_Final(sha1
, &c
);
616 memcpy(sha1out
, sha1
, sizeof(sha1
));
618 e
= insert_object(sha1
);
620 insert_mark(mark
, e
);
623 duplicate_count_by_type
[type
]++;
627 e
->offset
= pack_size
;
629 object_count_by_type
[type
]++;
631 if (last
&& last
->data
&& last
->depth
< max_depth
)
632 delta
= diff_delta(last
->data
, last
->len
,
638 memset(&s
, 0, sizeof(s
));
639 deflateInit(&s
, zlib_compression_level
);
644 s
.avail_in
= deltalen
;
645 hdrlen
= encode_header(OBJ_DELTA
, deltalen
, hdr
);
646 ywrite(pack_fd
, hdr
, hdrlen
);
647 ywrite(pack_fd
, last
->sha1
, sizeof(sha1
));
648 pack_size
+= hdrlen
+ sizeof(sha1
);
654 hdrlen
= encode_header(type
, datlen
, hdr
);
655 ywrite(pack_fd
, hdr
, hdrlen
);
659 s
.avail_out
= deflateBound(&s
, s
.avail_in
);
660 s
.next_out
= out
= xmalloc(s
.avail_out
);
661 while (deflate(&s
, Z_FINISH
) == Z_OK
)
665 ywrite(pack_fd
, out
, s
.total_out
);
666 pack_size
+= s
.total_out
;
676 memcpy(last
->sha1
, sha1
, sizeof(sha1
));
681 static unsigned char* map_pack(unsigned long offset
, unsigned int *left
)
683 if (offset
>= pack_size
)
684 die("object offset outside of pack file");
686 || offset
< pack_moff
687 || (offset
+ 20) >= (pack_moff
+ pack_mlen
)) {
689 munmap(pack_base
, pack_mlen
);
690 pack_moff
= (offset
/ page_size
) * page_size
;
691 pack_base
= mmap(NULL
,pack_mlen
,PROT_READ
,MAP_SHARED
,
693 if (pack_base
== MAP_FAILED
)
694 die("Failed to map generated pack: %s", strerror(errno
));
699 *left
= pack_mlen
- offset
;
700 return pack_base
+ offset
;
703 static unsigned long unpack_object_header(unsigned long offset
,
704 enum object_type
*type
,
705 unsigned long *sizep
)
711 c
= *map_pack(offset
++, NULL
);
712 *type
= (c
>> 4) & 7;
716 c
= *map_pack(offset
++, NULL
);
717 size
+= (c
& 0x7f) << shift
;
724 static void *unpack_non_delta_entry(unsigned long o
, unsigned long sz
)
727 unsigned char *result
;
729 result
= xmalloc(sz
+ 1);
732 memset(&stream
, 0, sizeof(stream
));
733 stream
.next_in
= map_pack(o
, &stream
.avail_in
);
734 stream
.next_out
= result
;
735 stream
.avail_out
= sz
;
737 inflateInit(&stream
);
739 int st
= inflate(&stream
, Z_FINISH
);
740 if (st
== Z_STREAM_END
)
742 if (st
== Z_OK
|| st
== Z_BUF_ERROR
) {
743 o
= stream
.next_in
- pack_base
+ pack_moff
;
744 stream
.next_in
= map_pack(o
, &stream
.avail_in
);
747 die("Error %i from zlib during inflate.", st
);
750 if (stream
.total_out
!= sz
)
751 die("Error after inflate: sizes mismatch");
755 static void *unpack_entry(unsigned long offset
, unsigned long *sizep
);
757 static void *unpack_delta_entry(unsigned long offset
,
758 unsigned long delta_size
,
759 unsigned long *sizep
)
761 struct object_entry
*base_oe
;
762 unsigned char *base_sha1
;
763 void *delta_data
, *base
, *result
;
764 unsigned long base_size
, result_size
;
766 base_sha1
= map_pack(offset
, NULL
);
767 base_oe
= find_object(base_sha1
);
769 die("I'm broken; I can't find a base I know must be here.");
770 base
= unpack_entry(base_oe
->offset
, &base_size
);
771 delta_data
= unpack_non_delta_entry(offset
+ 20, delta_size
);
772 result
= patch_delta(base
, base_size
,
773 delta_data
, delta_size
,
776 die("failed to apply delta");
779 *sizep
= result_size
;
783 static void *unpack_entry(unsigned long offset
, unsigned long *sizep
)
786 enum object_type kind
;
788 offset
= unpack_object_header(offset
, &kind
, &size
);
791 return unpack_delta_entry(offset
, size
, sizep
);
797 return unpack_non_delta_entry(offset
, size
);
799 die("I created an object I can't read!");
803 static const char *get_mode(const char *str
, unsigned int *modep
)
806 unsigned int mode
= 0;
808 while ((c
= *str
++) != ' ') {
809 if (c
< '0' || c
> '7')
811 mode
= (mode
<< 3) + (c
- '0');
817 static void load_tree(struct tree_entry
*root
)
819 struct object_entry
*myoe
;
820 struct tree_content
*t
;
825 root
->tree
= t
= new_tree_content(8);
826 if (!memcmp(root
->sha1
, null_sha1
, 20))
829 myoe
= find_object(root
->sha1
);
831 if (myoe
->type
!= OBJ_TREE
)
832 die("Not a tree: %s", sha1_to_hex(root
->sha1
));
833 buf
= unpack_entry(myoe
->offset
, &size
);
836 buf
= read_sha1_file(root
->sha1
, type
, &size
);
837 if (!buf
|| strcmp(type
, tree_type
))
838 die("Can't load tree %s", sha1_to_hex(root
->sha1
));
842 while (c
!= (buf
+ size
)) {
843 struct tree_entry
*e
= new_tree_entry();
845 if (t
->entry_count
== t
->entry_capacity
)
846 root
->tree
= t
= grow_tree_content(t
, 8);
847 t
->entries
[t
->entry_count
++] = e
;
850 c
= get_mode(c
, &e
->mode
);
852 die("Corrupt mode in %s", sha1_to_hex(root
->sha1
));
853 e
->name
= to_atom(c
, strlen(c
));
854 c
+= e
->name
->str_len
+ 1;
855 memcpy(e
->sha1
, c
, sizeof(e
->sha1
));
861 static int tecmp (const void *_a
, const void *_b
)
863 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
864 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
865 return base_name_compare(
866 a
->name
->str_dat
, a
->name
->str_len
, a
->mode
,
867 b
->name
->str_dat
, b
->name
->str_len
, b
->mode
);
870 static void store_tree(struct tree_entry
*root
)
872 struct tree_content
*t
= root
->tree
;
877 if (memcmp(root
->sha1
, null_sha1
, 20))
881 for (i
= 0; i
< t
->entry_count
; i
++) {
882 maxlen
+= t
->entries
[i
]->name
->str_len
+ 34;
883 if (t
->entries
[i
]->tree
)
884 store_tree(t
->entries
[i
]);
887 qsort(t
->entries
, t
->entry_count
, sizeof(t
->entries
[0]), tecmp
);
888 buf
= c
= xmalloc(maxlen
);
889 for (i
= 0; i
< t
->entry_count
; i
++) {
890 struct tree_entry
*e
= t
->entries
[i
];
891 c
+= sprintf(c
, "%o", e
->mode
);
893 strcpy(c
, e
->name
->str_dat
);
894 c
+= e
->name
->str_len
+ 1;
895 memcpy(c
, e
->sha1
, 20);
898 store_object(OBJ_TREE
, buf
, c
- buf
, NULL
, root
->sha1
, 0);
902 static int tree_content_set(
903 struct tree_entry
*root
,
905 const unsigned char *sha1
,
906 const unsigned int mode
)
908 struct tree_content
*t
= root
->tree
;
911 struct tree_entry
*e
;
913 slash1
= strchr(p
, '/');
919 for (i
= 0; i
< t
->entry_count
; i
++) {
921 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
923 if (e
->mode
== mode
&& !memcmp(e
->sha1
, sha1
, 20))
926 memcpy(e
->sha1
, sha1
, 20);
928 release_tree_content_recursive(e
->tree
);
931 memcpy(root
->sha1
, null_sha1
, 20);
934 if (!S_ISDIR(e
->mode
)) {
935 e
->tree
= new_tree_content(8);
940 if (tree_content_set(e
, slash1
+ 1, sha1
, mode
)) {
941 memcpy(root
->sha1
, null_sha1
, 20);
948 if (t
->entry_count
== t
->entry_capacity
)
949 root
->tree
= t
= grow_tree_content(t
, 8);
950 e
= new_tree_entry();
951 e
->name
= to_atom(p
, n
);
952 t
->entries
[t
->entry_count
++] = e
;
954 e
->tree
= new_tree_content(8);
956 tree_content_set(e
, slash1
+ 1, sha1
, mode
);
960 memcpy(e
->sha1
, sha1
, 20);
962 memcpy(root
->sha1
, null_sha1
, 20);
966 static int tree_content_remove(struct tree_entry
*root
, const char *p
)
968 struct tree_content
*t
= root
->tree
;
971 struct tree_entry
*e
;
973 slash1
= strchr(p
, '/');
979 for (i
= 0; i
< t
->entry_count
; i
++) {
981 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
982 if (!slash1
|| !S_ISDIR(e
->mode
))
986 if (tree_content_remove(e
, slash1
+ 1)) {
987 if (!e
->tree
->entry_count
)
989 memcpy(root
->sha1
, null_sha1
, 20);
998 for (i
++; i
< t
->entry_count
; i
++)
999 t
->entries
[i
-1] = t
->entries
[i
];
1001 release_tree_entry(e
);
1002 memcpy(root
->sha1
, null_sha1
, 20);
1006 static void init_pack_header()
1008 struct pack_header hdr
;
1010 hdr
.hdr_signature
= htonl(PACK_SIGNATURE
);
1011 hdr
.hdr_version
= htonl(2);
1012 hdr
.hdr_entries
= 0;
1014 ywrite(pack_fd
, &hdr
, sizeof(hdr
));
1015 pack_size
= sizeof(hdr
);
1018 static void fixup_header_footer()
1026 if (lseek(pack_fd
, 0, SEEK_SET
) != 0)
1027 die("Failed seeking to start: %s", strerror(errno
));
1030 yread(pack_fd
, hdr
, 8);
1031 SHA1_Update(&c
, hdr
, 8);
1033 cnt
= htonl(object_count
);
1034 SHA1_Update(&c
, &cnt
, 4);
1035 ywrite(pack_fd
, &cnt
, 4);
1037 buf
= xmalloc(128 * 1024);
1039 n
= xread(pack_fd
, buf
, 128 * 1024);
1042 SHA1_Update(&c
, buf
, n
);
1046 SHA1_Final(pack_sha1
, &c
);
1047 ywrite(pack_fd
, pack_sha1
, sizeof(pack_sha1
));
1050 static int oecmp (const void *_a
, const void *_b
)
1052 struct object_entry
*a
= *((struct object_entry
**)_a
);
1053 struct object_entry
*b
= *((struct object_entry
**)_b
);
1054 return memcmp(a
->sha1
, b
->sha1
, sizeof(a
->sha1
));
1057 static void write_index(const char *idx_name
)
1060 struct object_entry
**idx
, **c
, **last
;
1061 struct object_entry
*e
;
1062 struct object_entry_pool
*o
;
1063 unsigned int array
[256];
1066 /* Build the sorted table of object IDs. */
1067 idx
= xmalloc(object_count
* sizeof(struct object_entry
*));
1069 for (o
= blocks
; o
; o
= o
->next_pool
)
1070 for (e
= o
->entries
; e
!= o
->next_free
; e
++)
1072 last
= idx
+ object_count
;
1073 qsort(idx
, object_count
, sizeof(struct object_entry
*), oecmp
);
1075 /* Generate the fan-out array. */
1077 for (i
= 0; i
< 256; i
++) {
1078 struct object_entry
**next
= c
;;
1079 while (next
< last
) {
1080 if ((*next
)->sha1
[0] != i
)
1084 array
[i
] = htonl(next
- idx
);
1088 f
= sha1create("%s", idx_name
);
1089 sha1write(f
, array
, 256 * sizeof(int));
1090 for (c
= idx
; c
!= last
; c
++) {
1091 unsigned int offset
= htonl((*c
)->offset
);
1092 sha1write(f
, &offset
, 4);
1093 sha1write(f
, (*c
)->sha1
, sizeof((*c
)->sha1
));
1095 sha1write(f
, pack_sha1
, sizeof(pack_sha1
));
1096 sha1close(f
, NULL
, 1);
1100 static void dump_branches()
1102 static const char *msg
= "fast-import";
1105 struct ref_lock
*lock
;
1107 for (i
= 0; i
< branch_table_sz
; i
++) {
1108 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
) {
1109 lock
= lock_any_ref_for_update(b
->name
, NULL
, 0);
1110 if (!lock
|| write_ref_sha1(lock
, b
->sha1
, msg
) < 0)
1111 die("Can't write %s", b
->name
);
1116 static void dump_tags()
1118 static const char *msg
= "fast-import";
1120 struct ref_lock
*lock
;
1121 char path
[PATH_MAX
];
1123 for (t
= first_tag
; t
; t
= t
->next_tag
) {
1124 sprintf(path
, "refs/tags/%s", t
->name
);
1125 lock
= lock_any_ref_for_update(path
, NULL
, 0);
1126 if (!lock
|| write_ref_sha1(lock
, t
->sha1
, msg
) < 0)
1127 die("Can't write %s", path
);
1131 static void dump_marks_helper(FILE *f
,
1137 for (k
= 0; k
< 1024; k
++) {
1138 if (m
->data
.sets
[k
])
1139 dump_marks_helper(f
, (base
+ k
) << m
->shift
,
1143 for (k
= 0; k
< 1024; k
++) {
1144 if (m
->data
.marked
[k
])
1145 fprintf(f
, ":%lu %s\n", base
+ k
,
1146 sha1_to_hex(m
->data
.marked
[k
]->sha1
));
1151 static void dump_marks()
1155 FILE *f
= fopen(mark_file
, "w");
1156 dump_marks_helper(f
, 0, marks
);
1161 static void read_next_command()
1163 read_line(&command_buf
, stdin
, '\n');
1166 static void cmd_mark()
1168 if (!strncmp("mark :", command_buf
.buf
, 6)) {
1169 next_mark
= strtoul(command_buf
.buf
+ 6, NULL
, 10);
1170 read_next_command();
1176 static void* cmd_data (size_t *size
)
1182 if (strncmp("data ", command_buf
.buf
, 5))
1183 die("Expected 'data n' command, found: %s", command_buf
.buf
);
1185 length
= strtoul(command_buf
.buf
+ 5, NULL
, 10);
1186 buffer
= xmalloc(length
);
1188 while (n
< length
) {
1189 size_t s
= fread((char*)buffer
+ n
, 1, length
- n
, stdin
);
1190 if (!s
&& feof(stdin
))
1191 die("EOF in data (%lu bytes remaining)", length
- n
);
1195 if (fgetc(stdin
) != '\n')
1196 die("An lf did not trail the binary data as expected.");
1202 static void cmd_new_blob()
1207 read_next_command();
1211 if (store_object(OBJ_BLOB
, d
, l
, &last_blob
, NULL
, next_mark
))
1215 static void unload_one_branch()
1217 while (cur_active_branches
1218 && cur_active_branches
>= max_active_branches
) {
1219 unsigned long min_commit
= ULONG_MAX
;
1220 struct branch
*e
, *l
= NULL
, *p
= NULL
;
1222 for (e
= active_branches
; e
; e
= e
->active_next_branch
) {
1223 if (e
->last_commit
< min_commit
) {
1225 min_commit
= e
->last_commit
;
1231 e
= p
->active_next_branch
;
1232 p
->active_next_branch
= e
->active_next_branch
;
1234 e
= active_branches
;
1235 active_branches
= e
->active_next_branch
;
1237 e
->active_next_branch
= NULL
;
1238 if (e
->branch_tree
.tree
) {
1239 release_tree_content_recursive(e
->branch_tree
.tree
);
1240 e
->branch_tree
.tree
= NULL
;
1242 cur_active_branches
--;
1246 static void load_branch(struct branch
*b
)
1248 load_tree(&b
->branch_tree
);
1249 b
->active_next_branch
= active_branches
;
1250 active_branches
= b
;
1251 cur_active_branches
++;
1252 branch_load_count
++;
1255 static void file_change_m(struct branch
*b
)
1257 const char *p
= command_buf
.buf
+ 2;
1260 struct object_entry
*oe
;
1261 unsigned char sha1
[20];
1265 p
= get_mode(p
, &mode
);
1267 die("Corrupt mode: %s", command_buf
.buf
);
1269 case S_IFREG
| 0644:
1270 case S_IFREG
| 0755:
1277 die("Corrupt mode: %s", command_buf
.buf
);
1282 oe
= find_mark(strtoul(p
+ 1, &x
, 10));
1285 if (get_sha1_hex(p
, sha1
))
1286 die("Invalid SHA1: %s", command_buf
.buf
);
1287 oe
= find_object(sha1
);
1291 die("Missing space after SHA1: %s", command_buf
.buf
);
1293 p_uq
= unquote_c_style(p
, &endp
);
1296 die("Garbage after path in: %s", command_buf
.buf
);
1301 if (oe
->type
!= OBJ_BLOB
)
1302 die("Not a blob (actually a %s): %s",
1303 command_buf
.buf
, type_names
[oe
->type
]);
1305 if (sha1_object_info(sha1
, type
, NULL
))
1306 die("Blob not found: %s", command_buf
.buf
);
1307 if (strcmp(blob_type
, type
))
1308 die("Not a blob (actually a %s): %s",
1309 command_buf
.buf
, type
);
1312 tree_content_set(&b
->branch_tree
, p
, sha1
, S_IFREG
| mode
);
1318 static void file_change_d(struct branch
*b
)
1320 const char *p
= command_buf
.buf
+ 2;
1324 p_uq
= unquote_c_style(p
, &endp
);
1327 die("Garbage after path in: %s", command_buf
.buf
);
1330 tree_content_remove(&b
->branch_tree
, p
);
1335 static void cmd_from(struct branch
*b
)
1337 const char *from
, *endp
;
1341 if (strncmp("from ", command_buf
.buf
, 5))
1345 die("Can't reinitailize branch %s", b
->name
);
1347 from
= strchr(command_buf
.buf
, ' ') + 1;
1348 str_uq
= unquote_c_style(from
, &endp
);
1351 die("Garbage after string in: %s", command_buf
.buf
);
1355 s
= lookup_branch(from
);
1357 die("Can't create a branch from itself: %s", b
->name
);
1359 memcpy(b
->sha1
, s
->sha1
, 20);
1360 memcpy(b
->branch_tree
.sha1
, s
->branch_tree
.sha1
, 20);
1361 } else if (*from
== ':') {
1362 unsigned long idnum
= strtoul(from
+ 1, NULL
, 10);
1363 struct object_entry
*oe
= find_mark(idnum
);
1366 if (oe
->type
!= OBJ_COMMIT
)
1367 die("Mark :%lu not a commit", idnum
);
1368 memcpy(b
->sha1
, oe
->sha1
, 20);
1369 buf
= unpack_entry(oe
->offset
, &size
);
1370 if (!buf
|| size
< 46)
1371 die("Not a valid commit: %s", from
);
1372 if (memcmp("tree ", buf
, 5)
1373 || get_sha1_hex(buf
+ 5, b
->branch_tree
.sha1
))
1374 die("The commit %s is corrupt", sha1_to_hex(b
->sha1
));
1376 } else if (!get_sha1(from
, b
->sha1
)) {
1377 if (!memcmp(b
->sha1
, null_sha1
, 20))
1378 memcpy(b
->branch_tree
.sha1
, null_sha1
, 20);
1383 buf
= read_object_with_reference(b
->sha1
,
1384 type_names
[OBJ_COMMIT
], &size
, b
->sha1
);
1385 if (!buf
|| size
< 46)
1386 die("Not a valid commit: %s", from
);
1387 if (memcmp("tree ", buf
, 5)
1388 || get_sha1_hex(buf
+ 5, b
->branch_tree
.sha1
))
1389 die("The commit %s is corrupt", sha1_to_hex(b
->sha1
));
1393 die("Invalid ref name or SHA1 expression: %s", from
);
1395 read_next_command();
1398 static void cmd_new_commit()
1406 char *author
= NULL
;
1407 char *committer
= NULL
;
1410 /* Obtain the branch name from the rest of our command */
1411 sp
= strchr(command_buf
.buf
, ' ') + 1;
1412 str_uq
= unquote_c_style(sp
, &endp
);
1415 die("Garbage after ref in: %s", command_buf
.buf
);
1418 b
= lookup_branch(sp
);
1424 read_next_command();
1426 if (!strncmp("author ", command_buf
.buf
, 7)) {
1427 author
= strdup(command_buf
.buf
);
1428 read_next_command();
1430 if (!strncmp("committer ", command_buf
.buf
, 10)) {
1431 committer
= strdup(command_buf
.buf
);
1432 read_next_command();
1435 die("Expected committer but didn't get one");
1436 msg
= cmd_data(&msglen
);
1437 read_next_command();
1440 /* ensure the branch is active/loaded */
1441 if (!b
->branch_tree
.tree
|| !max_active_branches
) {
1442 unload_one_branch();
1448 if (1 == command_buf
.len
)
1450 else if (!strncmp("M ", command_buf
.buf
, 2))
1452 else if (!strncmp("D ", command_buf
.buf
, 2))
1455 die("Unsupported file_change: %s", command_buf
.buf
);
1456 read_next_command();
1459 /* build the tree and the commit */
1460 store_tree(&b
->branch_tree
);
1461 body
= xmalloc(97 + msglen
1463 ? strlen(author
) + strlen(committer
)
1464 : 2 * strlen(committer
)));
1466 sp
+= sprintf(sp
, "tree %s\n", sha1_to_hex(b
->branch_tree
.sha1
));
1467 if (memcmp(b
->sha1
, null_sha1
, 20))
1468 sp
+= sprintf(sp
, "parent %s\n", sha1_to_hex(b
->sha1
));
1470 sp
+= sprintf(sp
, "%s\n", author
);
1472 sp
+= sprintf(sp
, "author %s\n", committer
+ 10);
1473 sp
+= sprintf(sp
, "%s\n\n", committer
);
1474 memcpy(sp
, msg
, msglen
);
1481 store_object(OBJ_COMMIT
, body
, sp
- body
, NULL
, b
->sha1
, next_mark
);
1483 b
->last_commit
= object_count_by_type
[OBJ_COMMIT
];
1486 int need_dq
= quote_c_style(b
->name
, NULL
, NULL
, 0);
1487 fprintf(branch_log
, "commit ");
1489 fputc('"', branch_log
);
1490 quote_c_style(b
->name
, NULL
, branch_log
, 0);
1491 fputc('"', branch_log
);
1493 fprintf(branch_log
, "%s", b
->name
);
1494 fprintf(branch_log
," :%lu %s\n",next_mark
,sha1_to_hex(b
->sha1
));
1498 static void cmd_new_tag()
1510 unsigned long from_mark
= 0;
1511 unsigned char sha1
[20];
1513 /* Obtain the new tag name from the rest of our command */
1514 sp
= strchr(command_buf
.buf
, ' ') + 1;
1515 str_uq
= unquote_c_style(sp
, &endp
);
1518 die("Garbage after tag name in: %s", command_buf
.buf
);
1521 t
= pool_alloc(sizeof(struct tag
));
1523 t
->name
= pool_strdup(sp
);
1525 last_tag
->next_tag
= t
;
1531 read_next_command();
1534 if (strncmp("from ", command_buf
.buf
, 5))
1535 die("Expected from command, got %s", command_buf
.buf
);
1537 from
= strchr(command_buf
.buf
, ' ') + 1;
1538 str_uq
= unquote_c_style(from
, &endp
);
1541 die("Garbage after string in: %s", command_buf
.buf
);
1545 s
= lookup_branch(from
);
1547 memcpy(sha1
, s
->sha1
, 20);
1548 } else if (*from
== ':') {
1549 from_mark
= strtoul(from
+ 1, NULL
, 10);
1550 struct object_entry
*oe
= find_mark(from_mark
);
1551 if (oe
->type
!= OBJ_COMMIT
)
1552 die("Mark :%lu not a commit", from_mark
);
1553 memcpy(sha1
, oe
->sha1
, 20);
1554 } else if (!get_sha1(from
, sha1
)) {
1558 buf
= read_object_with_reference(sha1
,
1559 type_names
[OBJ_COMMIT
], &size
, sha1
);
1560 if (!buf
|| size
< 46)
1561 die("Not a valid commit: %s", from
);
1564 die("Invalid ref name or SHA1 expression: %s", from
);
1568 read_next_command();
1571 if (strncmp("tagger ", command_buf
.buf
, 7))
1572 die("Expected tagger command, got %s", command_buf
.buf
);
1573 tagger
= strdup(command_buf
.buf
);
1575 /* tag payload/message */
1576 read_next_command();
1577 msg
= cmd_data(&msglen
);
1579 /* build the tag object */
1580 body
= xmalloc(67 + strlen(t
->name
) + strlen(tagger
) + msglen
);
1582 sp
+= sprintf(sp
, "object %s\n", sha1_to_hex(sha1
));
1583 sp
+= sprintf(sp
, "type %s\n", type_names
[OBJ_COMMIT
]);
1584 sp
+= sprintf(sp
, "tag %s\n", t
->name
);
1585 sp
+= sprintf(sp
, "%s\n\n", tagger
);
1586 memcpy(sp
, msg
, msglen
);
1591 store_object(OBJ_TAG
, body
, sp
- body
, NULL
, t
->sha1
, 0);
1595 int need_dq
= quote_c_style(t
->name
, NULL
, NULL
, 0);
1596 fprintf(branch_log
, "tag ");
1598 fputc('"', branch_log
);
1599 quote_c_style(t
->name
, NULL
, branch_log
, 0);
1600 fputc('"', branch_log
);
1602 fprintf(branch_log
, "%s", t
->name
);
1603 fprintf(branch_log
," :%lu %s\n",from_mark
,sha1_to_hex(t
->sha1
));
1607 static const char fast_import_usage
[] =
1608 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file] [--branch-log=log] temp.pack";
1610 int main(int argc
, const char **argv
)
1612 const char *base_name
;
1614 unsigned long est_obj_cnt
= 1000;
1620 git_config(git_default_config
);
1621 page_size
= getpagesize();
1623 for (i
= 1; i
< argc
; i
++) {
1624 const char *a
= argv
[i
];
1626 if (*a
!= '-' || !strcmp(a
, "--"))
1628 else if (!strncmp(a
, "--objects=", 10))
1629 est_obj_cnt
= strtoul(a
+ 10, NULL
, 0);
1630 else if (!strncmp(a
, "--depth=", 8))
1631 max_depth
= strtoul(a
+ 8, NULL
, 0);
1632 else if (!strncmp(a
, "--active-branches=", 18))
1633 max_active_branches
= strtoul(a
+ 18, NULL
, 0);
1634 else if (!strncmp(a
, "--export-marks=", 15))
1636 else if (!strncmp(a
, "--branch-log=", 13)) {
1637 branch_log
= fopen(a
+ 13, "w");
1639 die("Can't create %s: %s", a
+ 13, strerror(errno
));
1642 die("unknown option %s", a
);
1645 usage(fast_import_usage
);
1646 base_name
= argv
[i
];
1648 pack_name
= xmalloc(strlen(base_name
) + 6);
1649 sprintf(pack_name
, "%s.pack", base_name
);
1650 idx_name
= xmalloc(strlen(base_name
) + 5);
1651 sprintf(idx_name
, "%s.idx", base_name
);
1653 pack_fd
= open(pack_name
, O_RDWR
|O_CREAT
|O_EXCL
, 0666);
1655 die("Can't create %s: %s", pack_name
, strerror(errno
));
1658 alloc_objects(est_obj_cnt
);
1659 strbuf_init(&command_buf
);
1661 atom_table
= xcalloc(atom_table_sz
, sizeof(struct atom_str
*));
1662 branch_table
= xcalloc(branch_table_sz
, sizeof(struct branch
*));
1663 avail_tree_table
= xcalloc(avail_tree_table_sz
, sizeof(struct avail_tree_content
*));
1664 marks
= pool_calloc(1, sizeof(struct mark_set
));
1667 read_next_command();
1668 if (command_buf
.eof
)
1670 else if (!strcmp("blob", command_buf
.buf
))
1672 else if (!strncmp("commit ", command_buf
.buf
, 7))
1674 else if (!strncmp("tag ", command_buf
.buf
, 4))
1677 die("Unsupported command: %s", command_buf
.buf
);
1680 fixup_header_footer();
1682 write_index(idx_name
);
1688 fprintf(stderr
, "%s statistics:\n", argv
[0]);
1689 fprintf(stderr
, "---------------------------------------------------\n");
1690 fprintf(stderr
, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count
, alloc_count
- est_obj_cnt
);
1691 fprintf(stderr
, "Total objects: %10lu (%10lu duplicates)\n", object_count
, duplicate_count
);
1692 fprintf(stderr
, " blobs : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_BLOB
], duplicate_count_by_type
[OBJ_BLOB
]);
1693 fprintf(stderr
, " trees : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_TREE
], duplicate_count_by_type
[OBJ_TREE
]);
1694 fprintf(stderr
, " commits: %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_COMMIT
], duplicate_count_by_type
[OBJ_COMMIT
]);
1695 fprintf(stderr
, " tags : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_TAG
], duplicate_count_by_type
[OBJ_TAG
]);
1696 fprintf(stderr
, "Total branches: %10lu (%10lu loads )\n", branch_count
, branch_load_count
);
1697 fprintf(stderr
, " marks: %10u (%10lu unique )\n", (1 << marks
->shift
) * 1024, marks_set_count
);
1698 fprintf(stderr
, " atoms: %10u\n", atom_cnt
);
1699 fprintf(stderr
, "Memory total: %10lu KiB\n", (total_allocd
+ alloc_count
*sizeof(struct object_entry
))/1024);
1700 fprintf(stderr
, " pools: %10lu KiB\n", total_allocd
/1024);
1701 fprintf(stderr
, " objects: %10lu KiB\n", (alloc_count
*sizeof(struct object_entry
))/1024);
1702 fprintf(stderr
, "Pack remaps: %10lu\n", remap_count
);
1703 fprintf(stderr
, "---------------------------------------------------\n");
1705 stat(pack_name
, &sb
);
1706 fprintf(stderr
, "Pack size: %10lu KiB\n", (unsigned long)(sb
.st_size
/1024));
1707 stat(idx_name
, &sb
);
1708 fprintf(stderr
, "Index size: %10lu KiB\n", (unsigned long)(sb
.st_size
/1024));
1710 fprintf(stderr
, "\n");