]>
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
17 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
19 ('author' sp name '<' email '>' ts tz lf)?
20 'committer' sp name '<' email '>' ts tz 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 void* pack_base
;
219 static size_t pack_mlen
;
221 /* Table of objects we've written. */
222 static unsigned int object_entry_alloc
= 1000;
223 static struct object_entry_pool
*blocks
;
224 static struct object_entry
*object_table
[1 << 16];
225 static struct mark_set
*marks
;
228 static struct last_object last_blob
;
230 /* Tree management */
231 static unsigned int tree_entry_alloc
= 1000;
232 static void *avail_tree_entry
;
233 static unsigned int avail_tree_table_sz
= 100;
234 static struct avail_tree_content
**avail_tree_table
;
237 static unsigned long max_active_branches
= 5;
238 static unsigned long cur_active_branches
;
239 static unsigned long branch_table_sz
= 1039;
240 static struct branch
**branch_table
;
241 static struct branch
*active_branches
;
244 static struct tag
*first_tag
;
245 static struct tag
*last_tag
;
247 /* Input stream parsing */
248 static struct strbuf command_buf
;
249 static unsigned long next_mark
;
252 static void alloc_objects(int cnt
)
254 struct object_entry_pool
*b
;
256 b
= xmalloc(sizeof(struct object_entry_pool
)
257 + cnt
* sizeof(struct object_entry
));
258 b
->next_pool
= blocks
;
259 b
->next_free
= b
->entries
;
260 b
->end
= b
->entries
+ cnt
;
265 static struct object_entry
* new_object(unsigned char *sha1
)
267 struct object_entry
*e
;
269 if (blocks
->next_free
== blocks
->end
)
270 alloc_objects(object_entry_alloc
);
272 e
= blocks
->next_free
++;
273 memcpy(e
->sha1
, sha1
, sizeof(e
->sha1
));
277 static struct object_entry
* find_object(unsigned char *sha1
)
279 unsigned int h
= sha1
[0] << 8 | sha1
[1];
280 struct object_entry
*e
;
281 for (e
= object_table
[h
]; e
; e
= e
->next
)
282 if (!memcmp(sha1
, e
->sha1
, sizeof(e
->sha1
)))
287 static struct object_entry
* insert_object(unsigned char *sha1
)
289 unsigned int h
= sha1
[0] << 8 | sha1
[1];
290 struct object_entry
*e
= object_table
[h
];
291 struct object_entry
*p
= NULL
;
294 if (!memcmp(sha1
, e
->sha1
, sizeof(e
->sha1
)))
300 e
= new_object(sha1
);
310 static unsigned int hc_str(const char *s
, size_t len
)
318 static void* pool_alloc(size_t len
)
323 for (p
= mem_pool
; p
; p
= p
->next_pool
)
324 if ((p
->end
- p
->next_free
>= len
))
328 if (len
>= (mem_pool_alloc
/2)) {
332 total_allocd
+= sizeof(struct mem_pool
) + mem_pool_alloc
;
333 p
= xmalloc(sizeof(struct mem_pool
) + mem_pool_alloc
);
334 p
->next_pool
= mem_pool
;
335 p
->next_free
= p
->space
;
336 p
->end
= p
->next_free
+ mem_pool_alloc
;
341 /* round out to a pointer alignment */
342 if (len
& (sizeof(void*) - 1))
343 len
+= sizeof(void*) - (len
& (sizeof(void*) - 1));
348 static void* pool_calloc(size_t count
, size_t size
)
350 size_t len
= count
* size
;
351 void *r
= pool_alloc(len
);
356 static char* pool_strdup(const char *s
)
358 char *r
= pool_alloc(strlen(s
) + 1);
363 static void insert_mark(unsigned long idnum
, struct object_entry
*oe
)
365 struct mark_set
*s
= marks
;
366 while ((idnum
>> s
->shift
) >= 1024) {
367 s
= pool_calloc(1, sizeof(struct mark_set
));
368 s
->shift
= marks
->shift
+ 10;
369 s
->data
.sets
[0] = marks
;
373 unsigned long i
= idnum
>> s
->shift
;
374 idnum
-= i
<< s
->shift
;
375 if (!s
->data
.sets
[i
]) {
376 s
->data
.sets
[i
] = pool_calloc(1, sizeof(struct mark_set
));
377 s
->data
.sets
[i
]->shift
= s
->shift
- 10;
381 if (!s
->data
.marked
[idnum
])
383 s
->data
.marked
[idnum
] = oe
;
386 static struct object_entry
* find_mark(unsigned long idnum
)
388 unsigned long orig_idnum
= idnum
;
389 struct mark_set
*s
= marks
;
390 struct object_entry
*oe
= NULL
;
391 if ((idnum
>> s
->shift
) < 1024) {
392 while (s
&& s
->shift
) {
393 unsigned long i
= idnum
>> s
->shift
;
394 idnum
-= i
<< s
->shift
;
398 oe
= s
->data
.marked
[idnum
];
401 die("mark :%lu not declared", orig_idnum
);
405 static struct atom_str
* to_atom(const char *s
, size_t len
)
407 unsigned int hc
= hc_str(s
, len
) % atom_table_sz
;
410 for (c
= atom_table
[hc
]; c
; c
= c
->next_atom
)
411 if (c
->str_len
== len
&& !strncmp(s
, c
->str_dat
, len
))
414 c
= pool_alloc(sizeof(struct atom_str
) + len
+ 1);
416 strncpy(c
->str_dat
, s
, len
);
418 c
->next_atom
= atom_table
[hc
];
424 static struct branch
* lookup_branch(const char *name
)
426 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
429 for (b
= branch_table
[hc
]; b
; b
= b
->table_next_branch
)
430 if (!strcmp(name
, b
->name
))
435 static struct branch
* new_branch(const char *name
)
437 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
438 struct branch
* b
= lookup_branch(name
);
441 die("Invalid attempt to create duplicate branch: %s", name
);
442 if (check_ref_format(name
))
443 die("Branch name doesn't conform to GIT standards: %s", name
);
445 b
= pool_calloc(1, sizeof(struct branch
));
446 b
->name
= pool_strdup(name
);
447 b
->table_next_branch
= branch_table
[hc
];
448 branch_table
[hc
] = b
;
453 static unsigned int hc_entries(unsigned int cnt
)
455 cnt
= cnt
& 7 ? (cnt
/ 8) + 1 : cnt
/ 8;
456 return cnt
< avail_tree_table_sz
? cnt
: avail_tree_table_sz
- 1;
459 static struct tree_content
* new_tree_content(unsigned int cnt
)
461 struct avail_tree_content
*f
, *l
= NULL
;
462 struct tree_content
*t
;
463 unsigned int hc
= hc_entries(cnt
);
465 for (f
= avail_tree_table
[hc
]; f
; l
= f
, f
= f
->next_avail
)
466 if (f
->entry_capacity
>= cnt
)
471 l
->next_avail
= f
->next_avail
;
473 avail_tree_table
[hc
] = f
->next_avail
;
475 cnt
= cnt
& 7 ? ((cnt
/ 8) + 1) * 8 : cnt
;
476 f
= pool_alloc(sizeof(*t
) + sizeof(t
->entries
[0]) * cnt
);
477 f
->entry_capacity
= cnt
;
480 t
= (struct tree_content
*)f
;
485 static void release_tree_entry(struct tree_entry
*e
);
486 static void release_tree_content(struct tree_content
*t
)
488 struct avail_tree_content
*f
= (struct avail_tree_content
*)t
;
489 unsigned int hc
= hc_entries(f
->entry_capacity
);
490 f
->next_avail
= avail_tree_table
[hc
];
491 avail_tree_table
[hc
] = f
;
494 static void release_tree_content_recursive(struct tree_content
*t
)
497 for (i
= 0; i
< t
->entry_count
; i
++)
498 release_tree_entry(t
->entries
[i
]);
499 release_tree_content(t
);
502 static struct tree_content
* grow_tree_content(
503 struct tree_content
*t
,
506 struct tree_content
*r
= new_tree_content(t
->entry_count
+ amt
);
507 r
->entry_count
= t
->entry_count
;
508 memcpy(r
->entries
,t
->entries
,t
->entry_count
*sizeof(t
->entries
[0]));
509 release_tree_content(t
);
513 static struct tree_entry
* new_tree_entry()
515 struct tree_entry
*e
;
517 if (!avail_tree_entry
) {
518 unsigned int n
= tree_entry_alloc
;
519 avail_tree_entry
= e
= xmalloc(n
* sizeof(struct tree_entry
));
521 *((void**)e
) = e
+ 1;
526 e
= avail_tree_entry
;
527 avail_tree_entry
= *((void**)e
);
531 static void release_tree_entry(struct tree_entry
*e
)
534 release_tree_content_recursive(e
->tree
);
535 *((void**)e
) = avail_tree_entry
;
536 avail_tree_entry
= e
;
539 static void yread(int fd
, void *buffer
, size_t length
)
542 while (ret
< length
) {
543 ssize_t size
= xread(fd
, (char *) buffer
+ ret
, length
- ret
);
545 die("Read from descriptor %i: end of stream", fd
);
547 die("Read from descriptor %i: %s", fd
, strerror(errno
));
552 static void ywrite(int fd
, void *buffer
, size_t length
)
555 while (ret
< length
) {
556 ssize_t size
= xwrite(fd
, (char *) buffer
+ ret
, length
- ret
);
558 die("Write to descriptor %i: end of file", fd
);
560 die("Write to descriptor %i: %s", fd
, strerror(errno
));
565 static size_t encode_header(
566 enum object_type type
,
573 if (type
< OBJ_COMMIT
|| type
> OBJ_DELTA
)
574 die("bad type %d", type
);
576 c
= (type
<< 4) | (size
& 15);
588 static int store_object(
589 enum object_type type
,
592 struct last_object
*last
,
593 unsigned char *sha1out
,
597 struct object_entry
*e
;
598 unsigned char hdr
[96];
599 unsigned char sha1
[20];
600 unsigned long hdrlen
, deltalen
;
604 hdrlen
= sprintf((char*)hdr
,"%s %lu",type_names
[type
],datlen
) + 1;
606 SHA1_Update(&c
, hdr
, hdrlen
);
607 SHA1_Update(&c
, dat
, datlen
);
608 SHA1_Final(sha1
, &c
);
610 memcpy(sha1out
, sha1
, sizeof(sha1
));
612 e
= insert_object(sha1
);
614 insert_mark(mark
, e
);
617 duplicate_count_by_type
[type
]++;
621 e
->offset
= pack_size
;
623 object_count_by_type
[type
]++;
625 if (last
&& last
->data
&& last
->depth
< max_depth
)
626 delta
= diff_delta(last
->data
, last
->len
,
632 memset(&s
, 0, sizeof(s
));
633 deflateInit(&s
, zlib_compression_level
);
638 s
.avail_in
= deltalen
;
639 hdrlen
= encode_header(OBJ_DELTA
, deltalen
, hdr
);
640 ywrite(pack_fd
, hdr
, hdrlen
);
641 ywrite(pack_fd
, last
->sha1
, sizeof(sha1
));
642 pack_size
+= hdrlen
+ sizeof(sha1
);
648 hdrlen
= encode_header(type
, datlen
, hdr
);
649 ywrite(pack_fd
, hdr
, hdrlen
);
653 s
.avail_out
= deflateBound(&s
, s
.avail_in
);
654 s
.next_out
= out
= xmalloc(s
.avail_out
);
655 while (deflate(&s
, Z_FINISH
) == Z_OK
)
659 ywrite(pack_fd
, out
, s
.total_out
);
660 pack_size
+= s
.total_out
;
670 memcpy(last
->sha1
, sha1
, sizeof(sha1
));
675 static void* map_pack(unsigned long offset
)
677 if (offset
>= pack_size
)
678 die("object offset outside of pack file");
679 if (offset
>= pack_mlen
) {
681 munmap(pack_base
, pack_mlen
);
682 /* round out how much we map to 16 MB units */
683 pack_mlen
= pack_size
;
684 if (pack_mlen
& ((1 << 24) - 1))
685 pack_mlen
= ((pack_mlen
>> 24) + 1) << 24;
686 pack_base
= mmap(NULL
,pack_mlen
,PROT_READ
,MAP_SHARED
,pack_fd
,0);
687 if (pack_base
== MAP_FAILED
)
688 die("Failed to map generated pack: %s", strerror(errno
));
691 return (char*)pack_base
+ offset
;
694 static unsigned long unpack_object_header(unsigned long offset
,
695 enum object_type
*type
,
696 unsigned long *sizep
)
702 c
= *(unsigned char*)map_pack(offset
++);
703 *type
= (c
>> 4) & 7;
707 c
= *(unsigned char*)map_pack(offset
++);
708 size
+= (c
& 0x7f) << shift
;
715 static void *unpack_non_delta_entry(unsigned long o
, unsigned long sz
)
718 unsigned char *result
;
720 result
= xmalloc(sz
+ 1);
723 memset(&stream
, 0, sizeof(stream
));
724 stream
.next_in
= map_pack(o
);
725 stream
.avail_in
= pack_mlen
- o
;
726 stream
.next_out
= result
;
727 stream
.avail_out
= sz
;
729 inflateInit(&stream
);
731 int st
= inflate(&stream
, Z_FINISH
);
732 if (st
== Z_STREAM_END
)
735 o
= stream
.next_in
- (unsigned char*)pack_base
;
736 stream
.next_in
= map_pack(o
);
737 stream
.avail_in
= pack_mlen
- o
;
740 die("Error from zlib during inflate.");
743 if (stream
.total_out
!= sz
)
744 die("Error after inflate: sizes mismatch");
748 static void *unpack_entry(unsigned long offset
, unsigned long *sizep
);
750 static void *unpack_delta_entry(unsigned long offset
,
751 unsigned long delta_size
,
752 unsigned long *sizep
)
754 struct object_entry
*base_oe
;
755 unsigned char *base_sha1
;
756 void *delta_data
, *base
, *result
;
757 unsigned long base_size
, result_size
;
759 base_sha1
= (unsigned char*)map_pack(offset
+ 20) - 20;
760 base_oe
= find_object(base_sha1
);
762 die("I'm broken; I can't find a base I know must be here.");
763 base
= unpack_entry(base_oe
->offset
, &base_size
);
764 delta_data
= unpack_non_delta_entry(offset
+ 20, delta_size
);
765 result
= patch_delta(base
, base_size
,
766 delta_data
, delta_size
,
769 die("failed to apply delta");
772 *sizep
= result_size
;
776 static void *unpack_entry(unsigned long offset
, unsigned long *sizep
)
779 enum object_type kind
;
781 offset
= unpack_object_header(offset
, &kind
, &size
);
784 return unpack_delta_entry(offset
, size
, sizep
);
790 return unpack_non_delta_entry(offset
, size
);
792 die("I created an object I can't read!");
796 static const char *get_mode(const char *str
, unsigned int *modep
)
799 unsigned int mode
= 0;
801 while ((c
= *str
++) != ' ') {
802 if (c
< '0' || c
> '7')
804 mode
= (mode
<< 3) + (c
- '0');
810 static void load_tree(struct tree_entry
*root
)
812 struct object_entry
*myoe
;
813 struct tree_content
*t
;
818 root
->tree
= t
= new_tree_content(8);
819 if (!memcmp(root
->sha1
, null_sha1
, 20))
822 myoe
= find_object(root
->sha1
);
824 if (myoe
->type
!= OBJ_TREE
)
825 die("Not a tree: %s", sha1_to_hex(root
->sha1
));
826 buf
= unpack_entry(myoe
->offset
, &size
);
829 buf
= read_sha1_file(root
->sha1
, type
, &size
);
830 if (!buf
|| strcmp(type
, tree_type
))
831 die("Can't load tree %s", sha1_to_hex(root
->sha1
));
835 while (c
!= (buf
+ size
)) {
836 struct tree_entry
*e
= new_tree_entry();
838 if (t
->entry_count
== t
->entry_capacity
)
839 root
->tree
= t
= grow_tree_content(t
, 8);
840 t
->entries
[t
->entry_count
++] = e
;
843 c
= get_mode(c
, &e
->mode
);
845 die("Corrupt mode in %s", sha1_to_hex(root
->sha1
));
846 e
->name
= to_atom(c
, strlen(c
));
847 c
+= e
->name
->str_len
+ 1;
848 memcpy(e
->sha1
, c
, sizeof(e
->sha1
));
854 static int tecmp (const void *_a
, const void *_b
)
856 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
857 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
858 return base_name_compare(
859 a
->name
->str_dat
, a
->name
->str_len
, a
->mode
,
860 b
->name
->str_dat
, b
->name
->str_len
, b
->mode
);
863 static void store_tree(struct tree_entry
*root
)
865 struct tree_content
*t
= root
->tree
;
870 if (memcmp(root
->sha1
, null_sha1
, 20))
874 for (i
= 0; i
< t
->entry_count
; i
++) {
875 maxlen
+= t
->entries
[i
]->name
->str_len
+ 34;
876 if (t
->entries
[i
]->tree
)
877 store_tree(t
->entries
[i
]);
880 qsort(t
->entries
, t
->entry_count
, sizeof(t
->entries
[0]), tecmp
);
881 buf
= c
= xmalloc(maxlen
);
882 for (i
= 0; i
< t
->entry_count
; i
++) {
883 struct tree_entry
*e
= t
->entries
[i
];
884 c
+= sprintf(c
, "%o", e
->mode
);
886 strcpy(c
, e
->name
->str_dat
);
887 c
+= e
->name
->str_len
+ 1;
888 memcpy(c
, e
->sha1
, 20);
891 store_object(OBJ_TREE
, buf
, c
- buf
, NULL
, root
->sha1
, 0);
895 static int tree_content_set(
896 struct tree_entry
*root
,
898 const unsigned char *sha1
,
899 const unsigned int mode
)
901 struct tree_content
*t
= root
->tree
;
904 struct tree_entry
*e
;
906 slash1
= strchr(p
, '/');
912 for (i
= 0; i
< t
->entry_count
; i
++) {
914 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
916 if (e
->mode
== mode
&& !memcmp(e
->sha1
, sha1
, 20))
919 memcpy(e
->sha1
, sha1
, 20);
921 release_tree_content_recursive(e
->tree
);
924 memcpy(root
->sha1
, null_sha1
, 20);
927 if (!S_ISDIR(e
->mode
)) {
928 e
->tree
= new_tree_content(8);
933 if (tree_content_set(e
, slash1
+ 1, sha1
, mode
)) {
934 memcpy(root
->sha1
, null_sha1
, 20);
941 if (t
->entry_count
== t
->entry_capacity
)
942 root
->tree
= t
= grow_tree_content(t
, 8);
943 e
= new_tree_entry();
944 e
->name
= to_atom(p
, n
);
945 t
->entries
[t
->entry_count
++] = e
;
947 e
->tree
= new_tree_content(8);
949 tree_content_set(e
, slash1
+ 1, sha1
, mode
);
953 memcpy(e
->sha1
, sha1
, 20);
955 memcpy(root
->sha1
, null_sha1
, 20);
959 static int tree_content_remove(struct tree_entry
*root
, const char *p
)
961 struct tree_content
*t
= root
->tree
;
964 struct tree_entry
*e
;
966 slash1
= strchr(p
, '/');
972 for (i
= 0; i
< t
->entry_count
; i
++) {
974 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
975 if (!slash1
|| !S_ISDIR(e
->mode
))
979 if (tree_content_remove(e
, slash1
+ 1)) {
980 if (!e
->tree
->entry_count
)
982 memcpy(root
->sha1
, null_sha1
, 20);
991 for (i
++; i
< t
->entry_count
; i
++)
992 t
->entries
[i
-1] = t
->entries
[i
];
994 release_tree_entry(e
);
995 memcpy(root
->sha1
, null_sha1
, 20);
999 static void init_pack_header()
1001 struct pack_header hdr
;
1003 hdr
.hdr_signature
= htonl(PACK_SIGNATURE
);
1004 hdr
.hdr_version
= htonl(2);
1005 hdr
.hdr_entries
= 0;
1007 ywrite(pack_fd
, &hdr
, sizeof(hdr
));
1008 pack_size
= sizeof(hdr
);
1011 static void fixup_header_footer()
1019 if (lseek(pack_fd
, 0, SEEK_SET
) != 0)
1020 die("Failed seeking to start: %s", strerror(errno
));
1023 yread(pack_fd
, hdr
, 8);
1024 SHA1_Update(&c
, hdr
, 8);
1026 cnt
= htonl(object_count
);
1027 SHA1_Update(&c
, &cnt
, 4);
1028 ywrite(pack_fd
, &cnt
, 4);
1030 buf
= xmalloc(128 * 1024);
1032 n
= xread(pack_fd
, buf
, 128 * 1024);
1035 SHA1_Update(&c
, buf
, n
);
1039 SHA1_Final(pack_sha1
, &c
);
1040 ywrite(pack_fd
, pack_sha1
, sizeof(pack_sha1
));
1043 static int oecmp (const void *_a
, const void *_b
)
1045 struct object_entry
*a
= *((struct object_entry
**)_a
);
1046 struct object_entry
*b
= *((struct object_entry
**)_b
);
1047 return memcmp(a
->sha1
, b
->sha1
, sizeof(a
->sha1
));
1050 static void write_index(const char *idx_name
)
1053 struct object_entry
**idx
, **c
, **last
;
1054 struct object_entry
*e
;
1055 struct object_entry_pool
*o
;
1056 unsigned int array
[256];
1059 /* Build the sorted table of object IDs. */
1060 idx
= xmalloc(object_count
* sizeof(struct object_entry
*));
1062 for (o
= blocks
; o
; o
= o
->next_pool
)
1063 for (e
= o
->entries
; e
!= o
->next_free
; e
++)
1065 last
= idx
+ object_count
;
1066 qsort(idx
, object_count
, sizeof(struct object_entry
*), oecmp
);
1068 /* Generate the fan-out array. */
1070 for (i
= 0; i
< 256; i
++) {
1071 struct object_entry
**next
= c
;;
1072 while (next
< last
) {
1073 if ((*next
)->sha1
[0] != i
)
1077 array
[i
] = htonl(next
- idx
);
1081 f
= sha1create("%s", idx_name
);
1082 sha1write(f
, array
, 256 * sizeof(int));
1083 for (c
= idx
; c
!= last
; c
++) {
1084 unsigned int offset
= htonl((*c
)->offset
);
1085 sha1write(f
, &offset
, 4);
1086 sha1write(f
, (*c
)->sha1
, sizeof((*c
)->sha1
));
1088 sha1write(f
, pack_sha1
, sizeof(pack_sha1
));
1089 sha1close(f
, NULL
, 1);
1093 static void dump_branches()
1095 static const char *msg
= "fast-import";
1098 struct ref_lock
*lock
;
1100 for (i
= 0; i
< branch_table_sz
; i
++) {
1101 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
) {
1102 lock
= lock_any_ref_for_update(b
->name
, NULL
, 0);
1103 if (!lock
|| write_ref_sha1(lock
, b
->sha1
, msg
) < 0)
1104 die("Can't write %s", b
->name
);
1109 static void dump_tags()
1111 static const char *msg
= "fast-import";
1113 struct ref_lock
*lock
;
1114 char path
[PATH_MAX
];
1116 for (t
= first_tag
; t
; t
= t
->next_tag
) {
1117 sprintf(path
, "refs/tags/%s", t
->name
);
1118 lock
= lock_any_ref_for_update(path
, NULL
, 0);
1119 if (!lock
|| write_ref_sha1(lock
, t
->sha1
, msg
) < 0)
1120 die("Can't write %s", path
);
1124 static void read_next_command()
1126 read_line(&command_buf
, stdin
, '\n');
1129 static void cmd_mark()
1131 if (!strncmp("mark :", command_buf
.buf
, 6)) {
1132 next_mark
= strtoul(command_buf
.buf
+ 6, NULL
, 10);
1133 read_next_command();
1139 static void* cmd_data (size_t *size
)
1145 if (strncmp("data ", command_buf
.buf
, 5))
1146 die("Expected 'data n' command, found: %s", command_buf
.buf
);
1148 length
= strtoul(command_buf
.buf
+ 5, NULL
, 10);
1149 buffer
= xmalloc(length
);
1151 while (n
< length
) {
1152 size_t s
= fread((char*)buffer
+ n
, 1, length
- n
, stdin
);
1153 if (!s
&& feof(stdin
))
1154 die("EOF in data (%lu bytes remaining)", length
- n
);
1158 if (fgetc(stdin
) != '\n')
1159 die("An lf did not trail the binary data as expected.");
1165 static void cmd_new_blob()
1170 read_next_command();
1174 if (store_object(OBJ_BLOB
, d
, l
, &last_blob
, NULL
, next_mark
))
1178 static void unload_one_branch()
1180 while (cur_active_branches
1181 && cur_active_branches
>= max_active_branches
) {
1182 unsigned long min_commit
= ULONG_MAX
;
1183 struct branch
*e
, *l
= NULL
, *p
= NULL
;
1185 for (e
= active_branches
; e
; e
= e
->active_next_branch
) {
1186 if (e
->last_commit
< min_commit
) {
1188 min_commit
= e
->last_commit
;
1194 e
= p
->active_next_branch
;
1195 p
->active_next_branch
= e
->active_next_branch
;
1197 e
= active_branches
;
1198 active_branches
= e
->active_next_branch
;
1200 e
->active_next_branch
= NULL
;
1201 if (e
->branch_tree
.tree
) {
1202 release_tree_content_recursive(e
->branch_tree
.tree
);
1203 e
->branch_tree
.tree
= NULL
;
1205 cur_active_branches
--;
1209 static void load_branch(struct branch
*b
)
1211 load_tree(&b
->branch_tree
);
1212 b
->active_next_branch
= active_branches
;
1213 active_branches
= b
;
1214 cur_active_branches
++;
1215 branch_load_count
++;
1218 static void file_change_m(struct branch
*b
)
1220 const char *p
= command_buf
.buf
+ 2;
1223 struct object_entry
*oe
;
1224 unsigned char sha1
[20];
1228 p
= get_mode(p
, &mode
);
1230 die("Corrupt mode: %s", command_buf
.buf
);
1232 case S_IFREG
| 0644:
1233 case S_IFREG
| 0755:
1240 die("Corrupt mode: %s", command_buf
.buf
);
1245 oe
= find_mark(strtoul(p
+ 1, &x
, 10));
1248 if (get_sha1_hex(p
, sha1
))
1249 die("Invalid SHA1: %s", command_buf
.buf
);
1250 oe
= find_object(sha1
);
1254 die("Missing space after SHA1: %s", command_buf
.buf
);
1256 p_uq
= unquote_c_style(p
, &endp
);
1259 die("Garbage after path in: %s", command_buf
.buf
);
1264 if (oe
->type
!= OBJ_BLOB
)
1265 die("Not a blob (actually a %s): %s",
1266 command_buf
.buf
, type_names
[oe
->type
]);
1268 if (sha1_object_info(sha1
, type
, NULL
))
1269 die("Blob not found: %s", command_buf
.buf
);
1270 if (strcmp(blob_type
, type
))
1271 die("Not a blob (actually a %s): %s",
1272 command_buf
.buf
, type
);
1275 tree_content_set(&b
->branch_tree
, p
, sha1
, S_IFREG
| mode
);
1281 static void file_change_d(struct branch
*b
)
1283 const char *p
= command_buf
.buf
+ 2;
1287 p_uq
= unquote_c_style(p
, &endp
);
1290 die("Garbage after path in: %s", command_buf
.buf
);
1293 tree_content_remove(&b
->branch_tree
, p
);
1298 static void cmd_from(struct branch
*b
)
1300 const char *from
, *endp
;
1304 if (strncmp("from ", command_buf
.buf
, 5))
1308 die("Can't reinitailize branch %s", b
->name
);
1310 from
= strchr(command_buf
.buf
, ' ') + 1;
1311 str_uq
= unquote_c_style(from
, &endp
);
1314 die("Garbage after string in: %s", command_buf
.buf
);
1318 s
= lookup_branch(from
);
1320 die("Can't create a branch from itself: %s", b
->name
);
1322 memcpy(b
->sha1
, s
->sha1
, 20);
1323 memcpy(b
->branch_tree
.sha1
, s
->branch_tree
.sha1
, 20);
1324 } else if (*from
== ':') {
1325 unsigned long idnum
= strtoul(from
+ 1, NULL
, 10);
1326 struct object_entry
*oe
= find_mark(idnum
);
1329 if (oe
->type
!= OBJ_COMMIT
)
1330 die("Mark :%lu not a commit", idnum
);
1331 memcpy(b
->sha1
, oe
->sha1
, 20);
1332 buf
= unpack_entry(oe
->offset
, &size
);
1333 if (!buf
|| size
< 46)
1334 die("Not a valid commit: %s", from
);
1335 if (memcmp("tree ", buf
, 5)
1336 || get_sha1_hex(buf
+ 5, b
->branch_tree
.sha1
))
1337 die("The commit %s is corrupt", sha1_to_hex(b
->sha1
));
1339 } else if (!get_sha1(from
, b
->sha1
)) {
1340 if (!memcmp(b
->sha1
, null_sha1
, 20))
1341 memcpy(b
->branch_tree
.sha1
, null_sha1
, 20);
1346 buf
= read_object_with_reference(b
->sha1
,
1347 type_names
[OBJ_COMMIT
], &size
, b
->sha1
);
1348 if (!buf
|| size
< 46)
1349 die("Not a valid commit: %s", from
);
1350 if (memcmp("tree ", buf
, 5)
1351 || get_sha1_hex(buf
+ 5, b
->branch_tree
.sha1
))
1352 die("The commit %s is corrupt", sha1_to_hex(b
->sha1
));
1356 die("Invalid ref name or SHA1 expression: %s", from
);
1358 read_next_command();
1361 static void cmd_new_commit()
1369 char *author
= NULL
;
1370 char *committer
= NULL
;
1373 /* Obtain the branch name from the rest of our command */
1374 sp
= strchr(command_buf
.buf
, ' ') + 1;
1375 str_uq
= unquote_c_style(sp
, &endp
);
1378 die("Garbage after ref in: %s", command_buf
.buf
);
1381 b
= lookup_branch(sp
);
1387 read_next_command();
1390 if (!strncmp("author ", command_buf
.buf
, 7)) {
1391 author
= strdup(command_buf
.buf
);
1392 read_next_command();
1394 if (!strncmp("committer ", command_buf
.buf
, 10)) {
1395 committer
= strdup(command_buf
.buf
);
1396 read_next_command();
1399 die("Expected committer but didn't get one");
1400 msg
= cmd_data(&msglen
);
1402 /* ensure the branch is active/loaded */
1403 if (!b
->branch_tree
.tree
|| !max_active_branches
) {
1404 unload_one_branch();
1410 read_next_command();
1411 if (1 == command_buf
.len
)
1413 else if (!strncmp("M ", command_buf
.buf
, 2))
1415 else if (!strncmp("D ", command_buf
.buf
, 2))
1418 die("Unsupported file_change: %s", command_buf
.buf
);
1421 /* build the tree and the commit */
1422 store_tree(&b
->branch_tree
);
1423 body
= xmalloc(97 + msglen
1425 ? strlen(author
) + strlen(committer
)
1426 : 2 * strlen(committer
)));
1428 sp
+= sprintf(sp
, "tree %s\n", sha1_to_hex(b
->branch_tree
.sha1
));
1429 if (memcmp(b
->sha1
, null_sha1
, 20))
1430 sp
+= sprintf(sp
, "parent %s\n", sha1_to_hex(b
->sha1
));
1432 sp
+= sprintf(sp
, "%s\n", author
);
1434 sp
+= sprintf(sp
, "author %s\n", committer
+ 10);
1435 sp
+= sprintf(sp
, "%s\n\n", committer
);
1436 memcpy(sp
, msg
, msglen
);
1443 store_object(OBJ_COMMIT
, body
, sp
- body
, NULL
, b
->sha1
, next_mark
);
1445 b
->last_commit
= object_count_by_type
[OBJ_COMMIT
];
1448 static void cmd_new_tag()
1460 unsigned char sha1
[20];
1462 /* Obtain the new tag name from the rest of our command */
1463 sp
= strchr(command_buf
.buf
, ' ') + 1;
1464 str_uq
= unquote_c_style(sp
, &endp
);
1467 die("Garbage after tag name in: %s", command_buf
.buf
);
1470 t
= pool_alloc(sizeof(struct tag
));
1472 t
->name
= pool_strdup(sp
);
1474 last_tag
->next_tag
= t
;
1480 read_next_command();
1483 if (strncmp("from ", command_buf
.buf
, 5))
1484 die("Expected from command, got %s", command_buf
.buf
);
1486 from
= strchr(command_buf
.buf
, ' ') + 1;
1487 str_uq
= unquote_c_style(from
, &endp
);
1490 die("Garbage after string in: %s", command_buf
.buf
);
1494 s
= lookup_branch(from
);
1496 memcpy(sha1
, s
->sha1
, 20);
1497 } else if (*from
== ':') {
1498 unsigned long idnum
= strtoul(from
+ 1, NULL
, 10);
1499 struct object_entry
*oe
= find_mark(idnum
);
1500 if (oe
->type
!= OBJ_COMMIT
)
1501 die("Mark :%lu not a commit", idnum
);
1502 memcpy(sha1
, oe
->sha1
, 20);
1503 } else if (!get_sha1(from
, sha1
)) {
1507 buf
= read_object_with_reference(sha1
,
1508 type_names
[OBJ_COMMIT
], &size
, sha1
);
1509 if (!buf
|| size
< 46)
1510 die("Not a valid commit: %s", from
);
1513 die("Invalid ref name or SHA1 expression: %s", from
);
1517 read_next_command();
1520 if (strncmp("tagger ", command_buf
.buf
, 7))
1521 die("Expected tagger command, got %s", command_buf
.buf
);
1522 tagger
= strdup(command_buf
.buf
);
1524 /* tag payload/message */
1525 read_next_command();
1526 msg
= cmd_data(&msglen
);
1528 /* build the tag object */
1529 body
= xmalloc(67 + strlen(t
->name
) + strlen(tagger
) + msglen
);
1531 sp
+= sprintf(sp
, "object %s\n", sha1_to_hex(sha1
));
1532 sp
+= sprintf(sp
, "type %s\n", type_names
[OBJ_COMMIT
]);
1533 sp
+= sprintf(sp
, "tag %s\n", t
->name
);
1534 sp
+= sprintf(sp
, "%s\n\n", tagger
);
1535 memcpy(sp
, msg
, msglen
);
1540 store_object(OBJ_TAG
, body
, sp
- body
, NULL
, t
->sha1
, 0);
1544 static const char fast_import_usage
[] =
1545 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] temp.pack";
1547 int main(int argc
, const char **argv
)
1549 const char *base_name
;
1551 unsigned long est_obj_cnt
= 1000;
1557 git_config(git_default_config
);
1559 for (i
= 1; i
< argc
; i
++) {
1560 const char *a
= argv
[i
];
1562 if (*a
!= '-' || !strcmp(a
, "--"))
1564 else if (!strncmp(a
, "--objects=", 10))
1565 est_obj_cnt
= strtoul(a
+ 10, NULL
, 0);
1566 else if (!strncmp(a
, "--depth=", 8))
1567 max_depth
= strtoul(a
+ 8, NULL
, 0);
1568 else if (!strncmp(a
, "--active-branches=", 18))
1569 max_active_branches
= strtoul(a
+ 18, NULL
, 0);
1571 die("unknown option %s", a
);
1574 usage(fast_import_usage
);
1575 base_name
= argv
[i
];
1577 pack_name
= xmalloc(strlen(base_name
) + 6);
1578 sprintf(pack_name
, "%s.pack", base_name
);
1579 idx_name
= xmalloc(strlen(base_name
) + 5);
1580 sprintf(idx_name
, "%s.idx", base_name
);
1582 pack_fd
= open(pack_name
, O_RDWR
|O_CREAT
|O_EXCL
, 0666);
1584 die("Can't create %s: %s", pack_name
, strerror(errno
));
1587 alloc_objects(est_obj_cnt
);
1588 strbuf_init(&command_buf
);
1590 atom_table
= xcalloc(atom_table_sz
, sizeof(struct atom_str
*));
1591 branch_table
= xcalloc(branch_table_sz
, sizeof(struct branch
*));
1592 avail_tree_table
= xcalloc(avail_tree_table_sz
, sizeof(struct avail_tree_content
*));
1593 marks
= pool_calloc(1, sizeof(struct mark_set
));
1596 read_next_command();
1597 if (command_buf
.eof
)
1599 else if (!strcmp("blob", command_buf
.buf
))
1601 else if (!strncmp("commit ", command_buf
.buf
, 7))
1603 else if (!strncmp("tag ", command_buf
.buf
, 4))
1606 die("Unsupported command: %s", command_buf
.buf
);
1609 fixup_header_footer();
1611 write_index(idx_name
);
1615 fprintf(stderr
, "%s statistics:\n", argv
[0]);
1616 fprintf(stderr
, "---------------------------------------------------\n");
1617 fprintf(stderr
, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count
, alloc_count
- est_obj_cnt
);
1618 fprintf(stderr
, "Total objects: %10lu (%10lu duplicates)\n", object_count
, duplicate_count
);
1619 fprintf(stderr
, " blobs : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_BLOB
], duplicate_count_by_type
[OBJ_BLOB
]);
1620 fprintf(stderr
, " trees : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_TREE
], duplicate_count_by_type
[OBJ_TREE
]);
1621 fprintf(stderr
, " commits: %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_COMMIT
], duplicate_count_by_type
[OBJ_COMMIT
]);
1622 fprintf(stderr
, " tags : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_TAG
], duplicate_count_by_type
[OBJ_TAG
]);
1623 fprintf(stderr
, "Total branches: %10lu (%10lu loads )\n", branch_count
, branch_load_count
);
1624 fprintf(stderr
, " marks: %10u (%10lu unique )\n", (1 << marks
->shift
) * 1024, marks_set_count
);
1625 fprintf(stderr
, " atoms: %10u\n", atom_cnt
);
1626 fprintf(stderr
, "Memory total: %10lu KiB\n", (total_allocd
+ alloc_count
*sizeof(struct object_entry
))/1024);
1627 fprintf(stderr
, " pools: %10lu KiB\n", total_allocd
/1024);
1628 fprintf(stderr
, " objects: %10lu KiB\n", (alloc_count
*sizeof(struct object_entry
))/1024);
1629 fprintf(stderr
, "Pack remaps: %10lu\n", remap_count
);
1630 fprintf(stderr
, "---------------------------------------------------\n");
1632 stat(pack_name
, &sb
);
1633 fprintf(stderr
, "Pack size: %10lu KiB\n", (unsigned long)(sb
.st_size
/1024));
1634 stat(idx_name
, &sb
);
1635 fprintf(stderr
, "Index size: %10lu KiB\n", (unsigned long)(sb
.st_size
/1024));
1637 fprintf(stderr
, "\n");