]>
git.ipfire.org Git - thirdparty/git.git/blob - fast-import.c
8328e004bb2fad1c1a7a831fedf0ef2aab5c21f7
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 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
;
226 static const char* mark_file
;
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
;
251 static FILE* branch_log
;
254 static void alloc_objects(int cnt
)
256 struct object_entry_pool
*b
;
258 b
= xmalloc(sizeof(struct object_entry_pool
)
259 + cnt
* sizeof(struct object_entry
));
260 b
->next_pool
= blocks
;
261 b
->next_free
= b
->entries
;
262 b
->end
= b
->entries
+ cnt
;
267 static struct object_entry
* new_object(unsigned char *sha1
)
269 struct object_entry
*e
;
271 if (blocks
->next_free
== blocks
->end
)
272 alloc_objects(object_entry_alloc
);
274 e
= blocks
->next_free
++;
275 memcpy(e
->sha1
, sha1
, sizeof(e
->sha1
));
279 static struct object_entry
* find_object(unsigned char *sha1
)
281 unsigned int h
= sha1
[0] << 8 | sha1
[1];
282 struct object_entry
*e
;
283 for (e
= object_table
[h
]; e
; e
= e
->next
)
284 if (!memcmp(sha1
, e
->sha1
, sizeof(e
->sha1
)))
289 static struct object_entry
* insert_object(unsigned char *sha1
)
291 unsigned int h
= sha1
[0] << 8 | sha1
[1];
292 struct object_entry
*e
= object_table
[h
];
293 struct object_entry
*p
= NULL
;
296 if (!memcmp(sha1
, e
->sha1
, sizeof(e
->sha1
)))
302 e
= new_object(sha1
);
312 static unsigned int hc_str(const char *s
, size_t len
)
320 static void* pool_alloc(size_t len
)
325 for (p
= mem_pool
; p
; p
= p
->next_pool
)
326 if ((p
->end
- p
->next_free
>= len
))
330 if (len
>= (mem_pool_alloc
/2)) {
334 total_allocd
+= sizeof(struct mem_pool
) + mem_pool_alloc
;
335 p
= xmalloc(sizeof(struct mem_pool
) + mem_pool_alloc
);
336 p
->next_pool
= mem_pool
;
337 p
->next_free
= p
->space
;
338 p
->end
= p
->next_free
+ mem_pool_alloc
;
343 /* round out to a pointer alignment */
344 if (len
& (sizeof(void*) - 1))
345 len
+= sizeof(void*) - (len
& (sizeof(void*) - 1));
350 static void* pool_calloc(size_t count
, size_t size
)
352 size_t len
= count
* size
;
353 void *r
= pool_alloc(len
);
358 static char* pool_strdup(const char *s
)
360 char *r
= pool_alloc(strlen(s
) + 1);
365 static void insert_mark(unsigned long idnum
, struct object_entry
*oe
)
367 struct mark_set
*s
= marks
;
368 while ((idnum
>> s
->shift
) >= 1024) {
369 s
= pool_calloc(1, sizeof(struct mark_set
));
370 s
->shift
= marks
->shift
+ 10;
371 s
->data
.sets
[0] = marks
;
375 unsigned long i
= idnum
>> s
->shift
;
376 idnum
-= i
<< s
->shift
;
377 if (!s
->data
.sets
[i
]) {
378 s
->data
.sets
[i
] = pool_calloc(1, sizeof(struct mark_set
));
379 s
->data
.sets
[i
]->shift
= s
->shift
- 10;
383 if (!s
->data
.marked
[idnum
])
385 s
->data
.marked
[idnum
] = oe
;
388 static struct object_entry
* find_mark(unsigned long idnum
)
390 unsigned long orig_idnum
= idnum
;
391 struct mark_set
*s
= marks
;
392 struct object_entry
*oe
= NULL
;
393 if ((idnum
>> s
->shift
) < 1024) {
394 while (s
&& s
->shift
) {
395 unsigned long i
= idnum
>> s
->shift
;
396 idnum
-= i
<< s
->shift
;
400 oe
= s
->data
.marked
[idnum
];
403 die("mark :%lu not declared", orig_idnum
);
407 static struct atom_str
* to_atom(const char *s
, size_t len
)
409 unsigned int hc
= hc_str(s
, len
) % atom_table_sz
;
412 for (c
= atom_table
[hc
]; c
; c
= c
->next_atom
)
413 if (c
->str_len
== len
&& !strncmp(s
, c
->str_dat
, len
))
416 c
= pool_alloc(sizeof(struct atom_str
) + len
+ 1);
418 strncpy(c
->str_dat
, s
, len
);
420 c
->next_atom
= atom_table
[hc
];
426 static struct branch
* lookup_branch(const char *name
)
428 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
431 for (b
= branch_table
[hc
]; b
; b
= b
->table_next_branch
)
432 if (!strcmp(name
, b
->name
))
437 static struct branch
* new_branch(const char *name
)
439 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
440 struct branch
* b
= lookup_branch(name
);
443 die("Invalid attempt to create duplicate branch: %s", name
);
444 if (check_ref_format(name
))
445 die("Branch name doesn't conform to GIT standards: %s", name
);
447 b
= pool_calloc(1, sizeof(struct branch
));
448 b
->name
= pool_strdup(name
);
449 b
->table_next_branch
= branch_table
[hc
];
450 branch_table
[hc
] = b
;
455 static unsigned int hc_entries(unsigned int cnt
)
457 cnt
= cnt
& 7 ? (cnt
/ 8) + 1 : cnt
/ 8;
458 return cnt
< avail_tree_table_sz
? cnt
: avail_tree_table_sz
- 1;
461 static struct tree_content
* new_tree_content(unsigned int cnt
)
463 struct avail_tree_content
*f
, *l
= NULL
;
464 struct tree_content
*t
;
465 unsigned int hc
= hc_entries(cnt
);
467 for (f
= avail_tree_table
[hc
]; f
; l
= f
, f
= f
->next_avail
)
468 if (f
->entry_capacity
>= cnt
)
473 l
->next_avail
= f
->next_avail
;
475 avail_tree_table
[hc
] = f
->next_avail
;
477 cnt
= cnt
& 7 ? ((cnt
/ 8) + 1) * 8 : cnt
;
478 f
= pool_alloc(sizeof(*t
) + sizeof(t
->entries
[0]) * cnt
);
479 f
->entry_capacity
= cnt
;
482 t
= (struct tree_content
*)f
;
487 static void release_tree_entry(struct tree_entry
*e
);
488 static void release_tree_content(struct tree_content
*t
)
490 struct avail_tree_content
*f
= (struct avail_tree_content
*)t
;
491 unsigned int hc
= hc_entries(f
->entry_capacity
);
492 f
->next_avail
= avail_tree_table
[hc
];
493 avail_tree_table
[hc
] = f
;
496 static void release_tree_content_recursive(struct tree_content
*t
)
499 for (i
= 0; i
< t
->entry_count
; i
++)
500 release_tree_entry(t
->entries
[i
]);
501 release_tree_content(t
);
504 static struct tree_content
* grow_tree_content(
505 struct tree_content
*t
,
508 struct tree_content
*r
= new_tree_content(t
->entry_count
+ amt
);
509 r
->entry_count
= t
->entry_count
;
510 memcpy(r
->entries
,t
->entries
,t
->entry_count
*sizeof(t
->entries
[0]));
511 release_tree_content(t
);
515 static struct tree_entry
* new_tree_entry()
517 struct tree_entry
*e
;
519 if (!avail_tree_entry
) {
520 unsigned int n
= tree_entry_alloc
;
521 total_allocd
+= n
* sizeof(struct tree_entry
);
522 avail_tree_entry
= e
= xmalloc(n
* sizeof(struct tree_entry
));
524 *((void**)e
) = e
+ 1;
529 e
= avail_tree_entry
;
530 avail_tree_entry
= *((void**)e
);
534 static void release_tree_entry(struct tree_entry
*e
)
537 release_tree_content_recursive(e
->tree
);
538 *((void**)e
) = avail_tree_entry
;
539 avail_tree_entry
= e
;
542 static void yread(int fd
, void *buffer
, size_t length
)
545 while (ret
< length
) {
546 ssize_t size
= xread(fd
, (char *) buffer
+ ret
, length
- ret
);
548 die("Read from descriptor %i: end of stream", fd
);
550 die("Read from descriptor %i: %s", fd
, strerror(errno
));
555 static void ywrite(int fd
, void *buffer
, size_t length
)
558 while (ret
< length
) {
559 ssize_t size
= xwrite(fd
, (char *) buffer
+ ret
, length
- ret
);
561 die("Write to descriptor %i: end of file", fd
);
563 die("Write to descriptor %i: %s", fd
, strerror(errno
));
568 static size_t encode_header(
569 enum object_type type
,
576 if (type
< OBJ_COMMIT
|| type
> OBJ_DELTA
)
577 die("bad type %d", type
);
579 c
= (type
<< 4) | (size
& 15);
591 static int store_object(
592 enum object_type type
,
595 struct last_object
*last
,
596 unsigned char *sha1out
,
600 struct object_entry
*e
;
601 unsigned char hdr
[96];
602 unsigned char sha1
[20];
603 unsigned long hdrlen
, deltalen
;
607 hdrlen
= sprintf((char*)hdr
,"%s %lu",type_names
[type
],datlen
) + 1;
609 SHA1_Update(&c
, hdr
, hdrlen
);
610 SHA1_Update(&c
, dat
, datlen
);
611 SHA1_Final(sha1
, &c
);
613 memcpy(sha1out
, sha1
, sizeof(sha1
));
615 e
= insert_object(sha1
);
617 insert_mark(mark
, e
);
620 duplicate_count_by_type
[type
]++;
624 e
->offset
= pack_size
;
626 object_count_by_type
[type
]++;
628 if (last
&& last
->data
&& last
->depth
< max_depth
)
629 delta
= diff_delta(last
->data
, last
->len
,
635 memset(&s
, 0, sizeof(s
));
636 deflateInit(&s
, zlib_compression_level
);
641 s
.avail_in
= deltalen
;
642 hdrlen
= encode_header(OBJ_DELTA
, deltalen
, hdr
);
643 ywrite(pack_fd
, hdr
, hdrlen
);
644 ywrite(pack_fd
, last
->sha1
, sizeof(sha1
));
645 pack_size
+= hdrlen
+ sizeof(sha1
);
651 hdrlen
= encode_header(type
, datlen
, hdr
);
652 ywrite(pack_fd
, hdr
, hdrlen
);
656 s
.avail_out
= deflateBound(&s
, s
.avail_in
);
657 s
.next_out
= out
= xmalloc(s
.avail_out
);
658 while (deflate(&s
, Z_FINISH
) == Z_OK
)
662 ywrite(pack_fd
, out
, s
.total_out
);
663 pack_size
+= s
.total_out
;
673 memcpy(last
->sha1
, sha1
, sizeof(sha1
));
678 static void* map_pack(unsigned long offset
)
680 if (offset
>= pack_size
)
681 die("object offset outside of pack file");
682 if (offset
>= pack_mlen
) {
684 munmap(pack_base
, pack_mlen
);
685 /* round out how much we map to 16 MB units */
686 pack_mlen
= pack_size
;
687 if (pack_mlen
& ((1 << 24) - 1))
688 pack_mlen
= ((pack_mlen
>> 24) + 1) << 24;
689 pack_base
= mmap(NULL
,pack_mlen
,PROT_READ
,MAP_SHARED
,pack_fd
,0);
690 if (pack_base
== MAP_FAILED
)
691 die("Failed to map generated pack: %s", strerror(errno
));
694 return (char*)pack_base
+ offset
;
697 static unsigned long unpack_object_header(unsigned long offset
,
698 enum object_type
*type
,
699 unsigned long *sizep
)
705 c
= *(unsigned char*)map_pack(offset
++);
706 *type
= (c
>> 4) & 7;
710 c
= *(unsigned char*)map_pack(offset
++);
711 size
+= (c
& 0x7f) << shift
;
718 static void *unpack_non_delta_entry(unsigned long o
, unsigned long sz
)
721 unsigned char *result
;
723 result
= xmalloc(sz
+ 1);
726 memset(&stream
, 0, sizeof(stream
));
727 stream
.next_in
= map_pack(o
);
728 stream
.avail_in
= pack_mlen
- o
;
729 stream
.next_out
= result
;
730 stream
.avail_out
= sz
;
732 inflateInit(&stream
);
734 int st
= inflate(&stream
, Z_FINISH
);
735 if (st
== Z_STREAM_END
)
738 o
= stream
.next_in
- (unsigned char*)pack_base
;
739 stream
.next_in
= map_pack(o
);
740 stream
.avail_in
= pack_mlen
- o
;
743 die("Error from zlib during inflate.");
746 if (stream
.total_out
!= sz
)
747 die("Error after inflate: sizes mismatch");
751 static void *unpack_entry(unsigned long offset
, unsigned long *sizep
);
753 static void *unpack_delta_entry(unsigned long offset
,
754 unsigned long delta_size
,
755 unsigned long *sizep
)
757 struct object_entry
*base_oe
;
758 unsigned char *base_sha1
;
759 void *delta_data
, *base
, *result
;
760 unsigned long base_size
, result_size
;
762 base_sha1
= (unsigned char*)map_pack(offset
+ 20) - 20;
763 base_oe
= find_object(base_sha1
);
765 die("I'm broken; I can't find a base I know must be here.");
766 base
= unpack_entry(base_oe
->offset
, &base_size
);
767 delta_data
= unpack_non_delta_entry(offset
+ 20, delta_size
);
768 result
= patch_delta(base
, base_size
,
769 delta_data
, delta_size
,
772 die("failed to apply delta");
775 *sizep
= result_size
;
779 static void *unpack_entry(unsigned long offset
, unsigned long *sizep
)
782 enum object_type kind
;
784 offset
= unpack_object_header(offset
, &kind
, &size
);
787 return unpack_delta_entry(offset
, size
, sizep
);
793 return unpack_non_delta_entry(offset
, size
);
795 die("I created an object I can't read!");
799 static const char *get_mode(const char *str
, unsigned int *modep
)
802 unsigned int mode
= 0;
804 while ((c
= *str
++) != ' ') {
805 if (c
< '0' || c
> '7')
807 mode
= (mode
<< 3) + (c
- '0');
813 static void load_tree(struct tree_entry
*root
)
815 struct object_entry
*myoe
;
816 struct tree_content
*t
;
821 root
->tree
= t
= new_tree_content(8);
822 if (!memcmp(root
->sha1
, null_sha1
, 20))
825 myoe
= find_object(root
->sha1
);
827 if (myoe
->type
!= OBJ_TREE
)
828 die("Not a tree: %s", sha1_to_hex(root
->sha1
));
829 buf
= unpack_entry(myoe
->offset
, &size
);
832 buf
= read_sha1_file(root
->sha1
, type
, &size
);
833 if (!buf
|| strcmp(type
, tree_type
))
834 die("Can't load tree %s", sha1_to_hex(root
->sha1
));
838 while (c
!= (buf
+ size
)) {
839 struct tree_entry
*e
= new_tree_entry();
841 if (t
->entry_count
== t
->entry_capacity
)
842 root
->tree
= t
= grow_tree_content(t
, 8);
843 t
->entries
[t
->entry_count
++] = e
;
846 c
= get_mode(c
, &e
->mode
);
848 die("Corrupt mode in %s", sha1_to_hex(root
->sha1
));
849 e
->name
= to_atom(c
, strlen(c
));
850 c
+= e
->name
->str_len
+ 1;
851 memcpy(e
->sha1
, c
, sizeof(e
->sha1
));
857 static int tecmp (const void *_a
, const void *_b
)
859 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
860 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
861 return base_name_compare(
862 a
->name
->str_dat
, a
->name
->str_len
, a
->mode
,
863 b
->name
->str_dat
, b
->name
->str_len
, b
->mode
);
866 static void store_tree(struct tree_entry
*root
)
868 struct tree_content
*t
= root
->tree
;
873 if (memcmp(root
->sha1
, null_sha1
, 20))
877 for (i
= 0; i
< t
->entry_count
; i
++) {
878 maxlen
+= t
->entries
[i
]->name
->str_len
+ 34;
879 if (t
->entries
[i
]->tree
)
880 store_tree(t
->entries
[i
]);
883 qsort(t
->entries
, t
->entry_count
, sizeof(t
->entries
[0]), tecmp
);
884 buf
= c
= xmalloc(maxlen
);
885 for (i
= 0; i
< t
->entry_count
; i
++) {
886 struct tree_entry
*e
= t
->entries
[i
];
887 c
+= sprintf(c
, "%o", e
->mode
);
889 strcpy(c
, e
->name
->str_dat
);
890 c
+= e
->name
->str_len
+ 1;
891 memcpy(c
, e
->sha1
, 20);
894 store_object(OBJ_TREE
, buf
, c
- buf
, NULL
, root
->sha1
, 0);
898 static int tree_content_set(
899 struct tree_entry
*root
,
901 const unsigned char *sha1
,
902 const unsigned int mode
)
904 struct tree_content
*t
= root
->tree
;
907 struct tree_entry
*e
;
909 slash1
= strchr(p
, '/');
915 for (i
= 0; i
< t
->entry_count
; i
++) {
917 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
919 if (e
->mode
== mode
&& !memcmp(e
->sha1
, sha1
, 20))
922 memcpy(e
->sha1
, sha1
, 20);
924 release_tree_content_recursive(e
->tree
);
927 memcpy(root
->sha1
, null_sha1
, 20);
930 if (!S_ISDIR(e
->mode
)) {
931 e
->tree
= new_tree_content(8);
936 if (tree_content_set(e
, slash1
+ 1, sha1
, mode
)) {
937 memcpy(root
->sha1
, null_sha1
, 20);
944 if (t
->entry_count
== t
->entry_capacity
)
945 root
->tree
= t
= grow_tree_content(t
, 8);
946 e
= new_tree_entry();
947 e
->name
= to_atom(p
, n
);
948 t
->entries
[t
->entry_count
++] = e
;
950 e
->tree
= new_tree_content(8);
952 tree_content_set(e
, slash1
+ 1, sha1
, mode
);
956 memcpy(e
->sha1
, sha1
, 20);
958 memcpy(root
->sha1
, null_sha1
, 20);
962 static int tree_content_remove(struct tree_entry
*root
, const char *p
)
964 struct tree_content
*t
= root
->tree
;
967 struct tree_entry
*e
;
969 slash1
= strchr(p
, '/');
975 for (i
= 0; i
< t
->entry_count
; i
++) {
977 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
978 if (!slash1
|| !S_ISDIR(e
->mode
))
982 if (tree_content_remove(e
, slash1
+ 1)) {
983 if (!e
->tree
->entry_count
)
985 memcpy(root
->sha1
, null_sha1
, 20);
994 for (i
++; i
< t
->entry_count
; i
++)
995 t
->entries
[i
-1] = t
->entries
[i
];
997 release_tree_entry(e
);
998 memcpy(root
->sha1
, null_sha1
, 20);
1002 static void init_pack_header()
1004 struct pack_header hdr
;
1006 hdr
.hdr_signature
= htonl(PACK_SIGNATURE
);
1007 hdr
.hdr_version
= htonl(2);
1008 hdr
.hdr_entries
= 0;
1010 ywrite(pack_fd
, &hdr
, sizeof(hdr
));
1011 pack_size
= sizeof(hdr
);
1014 static void fixup_header_footer()
1022 if (lseek(pack_fd
, 0, SEEK_SET
) != 0)
1023 die("Failed seeking to start: %s", strerror(errno
));
1026 yread(pack_fd
, hdr
, 8);
1027 SHA1_Update(&c
, hdr
, 8);
1029 cnt
= htonl(object_count
);
1030 SHA1_Update(&c
, &cnt
, 4);
1031 ywrite(pack_fd
, &cnt
, 4);
1033 buf
= xmalloc(128 * 1024);
1035 n
= xread(pack_fd
, buf
, 128 * 1024);
1038 SHA1_Update(&c
, buf
, n
);
1042 SHA1_Final(pack_sha1
, &c
);
1043 ywrite(pack_fd
, pack_sha1
, sizeof(pack_sha1
));
1046 static int oecmp (const void *_a
, const void *_b
)
1048 struct object_entry
*a
= *((struct object_entry
**)_a
);
1049 struct object_entry
*b
= *((struct object_entry
**)_b
);
1050 return memcmp(a
->sha1
, b
->sha1
, sizeof(a
->sha1
));
1053 static void write_index(const char *idx_name
)
1056 struct object_entry
**idx
, **c
, **last
;
1057 struct object_entry
*e
;
1058 struct object_entry_pool
*o
;
1059 unsigned int array
[256];
1062 /* Build the sorted table of object IDs. */
1063 idx
= xmalloc(object_count
* sizeof(struct object_entry
*));
1065 for (o
= blocks
; o
; o
= o
->next_pool
)
1066 for (e
= o
->entries
; e
!= o
->next_free
; e
++)
1068 last
= idx
+ object_count
;
1069 qsort(idx
, object_count
, sizeof(struct object_entry
*), oecmp
);
1071 /* Generate the fan-out array. */
1073 for (i
= 0; i
< 256; i
++) {
1074 struct object_entry
**next
= c
;;
1075 while (next
< last
) {
1076 if ((*next
)->sha1
[0] != i
)
1080 array
[i
] = htonl(next
- idx
);
1084 f
= sha1create("%s", idx_name
);
1085 sha1write(f
, array
, 256 * sizeof(int));
1086 for (c
= idx
; c
!= last
; c
++) {
1087 unsigned int offset
= htonl((*c
)->offset
);
1088 sha1write(f
, &offset
, 4);
1089 sha1write(f
, (*c
)->sha1
, sizeof((*c
)->sha1
));
1091 sha1write(f
, pack_sha1
, sizeof(pack_sha1
));
1092 sha1close(f
, NULL
, 1);
1096 static void dump_branches()
1098 static const char *msg
= "fast-import";
1101 struct ref_lock
*lock
;
1103 for (i
= 0; i
< branch_table_sz
; i
++) {
1104 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
) {
1105 lock
= lock_any_ref_for_update(b
->name
, NULL
, 0);
1106 if (!lock
|| write_ref_sha1(lock
, b
->sha1
, msg
) < 0)
1107 die("Can't write %s", b
->name
);
1112 static void dump_tags()
1114 static const char *msg
= "fast-import";
1116 struct ref_lock
*lock
;
1117 char path
[PATH_MAX
];
1119 for (t
= first_tag
; t
; t
= t
->next_tag
) {
1120 sprintf(path
, "refs/tags/%s", t
->name
);
1121 lock
= lock_any_ref_for_update(path
, NULL
, 0);
1122 if (!lock
|| write_ref_sha1(lock
, t
->sha1
, msg
) < 0)
1123 die("Can't write %s", path
);
1127 static void dump_marks_helper(FILE *f
,
1133 for (k
= 0; k
< 1024; k
++) {
1134 if (m
->data
.sets
[k
])
1135 dump_marks_helper(f
, (base
+ k
) << m
->shift
,
1139 for (k
= 0; k
< 1024; k
++) {
1140 if (m
->data
.marked
[k
])
1141 fprintf(f
, ":%lu %s\n", base
+ k
,
1142 sha1_to_hex(m
->data
.marked
[k
]->sha1
));
1147 static void dump_marks()
1151 FILE *f
= fopen(mark_file
, "w");
1152 dump_marks_helper(f
, 0, marks
);
1157 static void read_next_command()
1159 read_line(&command_buf
, stdin
, '\n');
1162 static void cmd_mark()
1164 if (!strncmp("mark :", command_buf
.buf
, 6)) {
1165 next_mark
= strtoul(command_buf
.buf
+ 6, NULL
, 10);
1166 read_next_command();
1172 static void* cmd_data (size_t *size
)
1178 if (strncmp("data ", command_buf
.buf
, 5))
1179 die("Expected 'data n' command, found: %s", command_buf
.buf
);
1181 length
= strtoul(command_buf
.buf
+ 5, NULL
, 10);
1182 buffer
= xmalloc(length
);
1184 while (n
< length
) {
1185 size_t s
= fread((char*)buffer
+ n
, 1, length
- n
, stdin
);
1186 if (!s
&& feof(stdin
))
1187 die("EOF in data (%lu bytes remaining)", length
- n
);
1191 if (fgetc(stdin
) != '\n')
1192 die("An lf did not trail the binary data as expected.");
1198 static void cmd_new_blob()
1203 read_next_command();
1207 if (store_object(OBJ_BLOB
, d
, l
, &last_blob
, NULL
, next_mark
))
1211 static void unload_one_branch()
1213 while (cur_active_branches
1214 && cur_active_branches
>= max_active_branches
) {
1215 unsigned long min_commit
= ULONG_MAX
;
1216 struct branch
*e
, *l
= NULL
, *p
= NULL
;
1218 for (e
= active_branches
; e
; e
= e
->active_next_branch
) {
1219 if (e
->last_commit
< min_commit
) {
1221 min_commit
= e
->last_commit
;
1227 e
= p
->active_next_branch
;
1228 p
->active_next_branch
= e
->active_next_branch
;
1230 e
= active_branches
;
1231 active_branches
= e
->active_next_branch
;
1233 e
->active_next_branch
= NULL
;
1234 if (e
->branch_tree
.tree
) {
1235 release_tree_content_recursive(e
->branch_tree
.tree
);
1236 e
->branch_tree
.tree
= NULL
;
1238 cur_active_branches
--;
1242 static void load_branch(struct branch
*b
)
1244 load_tree(&b
->branch_tree
);
1245 b
->active_next_branch
= active_branches
;
1246 active_branches
= b
;
1247 cur_active_branches
++;
1248 branch_load_count
++;
1251 static void file_change_m(struct branch
*b
)
1253 const char *p
= command_buf
.buf
+ 2;
1256 struct object_entry
*oe
;
1257 unsigned char sha1
[20];
1261 p
= get_mode(p
, &mode
);
1263 die("Corrupt mode: %s", command_buf
.buf
);
1265 case S_IFREG
| 0644:
1266 case S_IFREG
| 0755:
1273 die("Corrupt mode: %s", command_buf
.buf
);
1278 oe
= find_mark(strtoul(p
+ 1, &x
, 10));
1281 if (get_sha1_hex(p
, sha1
))
1282 die("Invalid SHA1: %s", command_buf
.buf
);
1283 oe
= find_object(sha1
);
1287 die("Missing space after SHA1: %s", command_buf
.buf
);
1289 p_uq
= unquote_c_style(p
, &endp
);
1292 die("Garbage after path in: %s", command_buf
.buf
);
1297 if (oe
->type
!= OBJ_BLOB
)
1298 die("Not a blob (actually a %s): %s",
1299 command_buf
.buf
, type_names
[oe
->type
]);
1301 if (sha1_object_info(sha1
, type
, NULL
))
1302 die("Blob not found: %s", command_buf
.buf
);
1303 if (strcmp(blob_type
, type
))
1304 die("Not a blob (actually a %s): %s",
1305 command_buf
.buf
, type
);
1308 tree_content_set(&b
->branch_tree
, p
, sha1
, S_IFREG
| mode
);
1314 static void file_change_d(struct branch
*b
)
1316 const char *p
= command_buf
.buf
+ 2;
1320 p_uq
= unquote_c_style(p
, &endp
);
1323 die("Garbage after path in: %s", command_buf
.buf
);
1326 tree_content_remove(&b
->branch_tree
, p
);
1331 static void cmd_from(struct branch
*b
)
1333 const char *from
, *endp
;
1337 if (strncmp("from ", command_buf
.buf
, 5))
1341 die("Can't reinitailize branch %s", b
->name
);
1343 from
= strchr(command_buf
.buf
, ' ') + 1;
1344 str_uq
= unquote_c_style(from
, &endp
);
1347 die("Garbage after string in: %s", command_buf
.buf
);
1351 s
= lookup_branch(from
);
1353 die("Can't create a branch from itself: %s", b
->name
);
1355 memcpy(b
->sha1
, s
->sha1
, 20);
1356 memcpy(b
->branch_tree
.sha1
, s
->branch_tree
.sha1
, 20);
1357 } else if (*from
== ':') {
1358 unsigned long idnum
= strtoul(from
+ 1, NULL
, 10);
1359 struct object_entry
*oe
= find_mark(idnum
);
1362 if (oe
->type
!= OBJ_COMMIT
)
1363 die("Mark :%lu not a commit", idnum
);
1364 memcpy(b
->sha1
, oe
->sha1
, 20);
1365 buf
= unpack_entry(oe
->offset
, &size
);
1366 if (!buf
|| size
< 46)
1367 die("Not a valid commit: %s", from
);
1368 if (memcmp("tree ", buf
, 5)
1369 || get_sha1_hex(buf
+ 5, b
->branch_tree
.sha1
))
1370 die("The commit %s is corrupt", sha1_to_hex(b
->sha1
));
1372 } else if (!get_sha1(from
, b
->sha1
)) {
1373 if (!memcmp(b
->sha1
, null_sha1
, 20))
1374 memcpy(b
->branch_tree
.sha1
, null_sha1
, 20);
1379 buf
= read_object_with_reference(b
->sha1
,
1380 type_names
[OBJ_COMMIT
], &size
, b
->sha1
);
1381 if (!buf
|| size
< 46)
1382 die("Not a valid commit: %s", from
);
1383 if (memcmp("tree ", buf
, 5)
1384 || get_sha1_hex(buf
+ 5, b
->branch_tree
.sha1
))
1385 die("The commit %s is corrupt", sha1_to_hex(b
->sha1
));
1389 die("Invalid ref name or SHA1 expression: %s", from
);
1391 read_next_command();
1394 static void cmd_new_commit()
1402 char *author
= NULL
;
1403 char *committer
= NULL
;
1406 /* Obtain the branch name from the rest of our command */
1407 sp
= strchr(command_buf
.buf
, ' ') + 1;
1408 str_uq
= unquote_c_style(sp
, &endp
);
1411 die("Garbage after ref in: %s", command_buf
.buf
);
1414 b
= lookup_branch(sp
);
1420 read_next_command();
1422 if (!strncmp("author ", command_buf
.buf
, 7)) {
1423 author
= strdup(command_buf
.buf
);
1424 read_next_command();
1426 if (!strncmp("committer ", command_buf
.buf
, 10)) {
1427 committer
= strdup(command_buf
.buf
);
1428 read_next_command();
1431 die("Expected committer but didn't get one");
1432 msg
= cmd_data(&msglen
);
1433 read_next_command();
1436 /* ensure the branch is active/loaded */
1437 if (!b
->branch_tree
.tree
|| !max_active_branches
) {
1438 unload_one_branch();
1444 if (1 == command_buf
.len
)
1446 else if (!strncmp("M ", command_buf
.buf
, 2))
1448 else if (!strncmp("D ", command_buf
.buf
, 2))
1451 die("Unsupported file_change: %s", command_buf
.buf
);
1452 read_next_command();
1455 /* build the tree and the commit */
1456 store_tree(&b
->branch_tree
);
1457 body
= xmalloc(97 + msglen
1459 ? strlen(author
) + strlen(committer
)
1460 : 2 * strlen(committer
)));
1462 sp
+= sprintf(sp
, "tree %s\n", sha1_to_hex(b
->branch_tree
.sha1
));
1463 if (memcmp(b
->sha1
, null_sha1
, 20))
1464 sp
+= sprintf(sp
, "parent %s\n", sha1_to_hex(b
->sha1
));
1466 sp
+= sprintf(sp
, "%s\n", author
);
1468 sp
+= sprintf(sp
, "author %s\n", committer
+ 10);
1469 sp
+= sprintf(sp
, "%s\n\n", committer
);
1470 memcpy(sp
, msg
, msglen
);
1477 store_object(OBJ_COMMIT
, body
, sp
- body
, NULL
, b
->sha1
, next_mark
);
1479 b
->last_commit
= object_count_by_type
[OBJ_COMMIT
];
1482 int need_dq
= quote_c_style(b
->name
, NULL
, NULL
, 0);
1483 fprintf(branch_log
, "commit ");
1485 fputc('"', branch_log
);
1486 quote_c_style(b
->name
, NULL
, branch_log
, 0);
1487 fputc('"', branch_log
);
1489 fprintf(branch_log
, "%s", b
->name
);
1490 fprintf(branch_log
," :%lu %s\n",next_mark
,sha1_to_hex(b
->sha1
));
1494 static void cmd_new_tag()
1506 unsigned long from_mark
= 0;
1507 unsigned char sha1
[20];
1509 /* Obtain the new tag name from the rest of our command */
1510 sp
= strchr(command_buf
.buf
, ' ') + 1;
1511 str_uq
= unquote_c_style(sp
, &endp
);
1514 die("Garbage after tag name in: %s", command_buf
.buf
);
1517 t
= pool_alloc(sizeof(struct tag
));
1519 t
->name
= pool_strdup(sp
);
1521 last_tag
->next_tag
= t
;
1527 read_next_command();
1530 if (strncmp("from ", command_buf
.buf
, 5))
1531 die("Expected from command, got %s", command_buf
.buf
);
1533 from
= strchr(command_buf
.buf
, ' ') + 1;
1534 str_uq
= unquote_c_style(from
, &endp
);
1537 die("Garbage after string in: %s", command_buf
.buf
);
1541 s
= lookup_branch(from
);
1543 memcpy(sha1
, s
->sha1
, 20);
1544 } else if (*from
== ':') {
1545 from_mark
= strtoul(from
+ 1, NULL
, 10);
1546 struct object_entry
*oe
= find_mark(from_mark
);
1547 if (oe
->type
!= OBJ_COMMIT
)
1548 die("Mark :%lu not a commit", from_mark
);
1549 memcpy(sha1
, oe
->sha1
, 20);
1550 } else if (!get_sha1(from
, sha1
)) {
1554 buf
= read_object_with_reference(sha1
,
1555 type_names
[OBJ_COMMIT
], &size
, sha1
);
1556 if (!buf
|| size
< 46)
1557 die("Not a valid commit: %s", from
);
1560 die("Invalid ref name or SHA1 expression: %s", from
);
1564 read_next_command();
1567 if (strncmp("tagger ", command_buf
.buf
, 7))
1568 die("Expected tagger command, got %s", command_buf
.buf
);
1569 tagger
= strdup(command_buf
.buf
);
1571 /* tag payload/message */
1572 read_next_command();
1573 msg
= cmd_data(&msglen
);
1575 /* build the tag object */
1576 body
= xmalloc(67 + strlen(t
->name
) + strlen(tagger
) + msglen
);
1578 sp
+= sprintf(sp
, "object %s\n", sha1_to_hex(sha1
));
1579 sp
+= sprintf(sp
, "type %s\n", type_names
[OBJ_COMMIT
]);
1580 sp
+= sprintf(sp
, "tag %s\n", t
->name
);
1581 sp
+= sprintf(sp
, "%s\n\n", tagger
);
1582 memcpy(sp
, msg
, msglen
);
1587 store_object(OBJ_TAG
, body
, sp
- body
, NULL
, t
->sha1
, 0);
1591 int need_dq
= quote_c_style(t
->name
, NULL
, NULL
, 0);
1592 fprintf(branch_log
, "tag ");
1594 fputc('"', branch_log
);
1595 quote_c_style(t
->name
, NULL
, branch_log
, 0);
1596 fputc('"', branch_log
);
1598 fprintf(branch_log
, "%s", t
->name
);
1599 fprintf(branch_log
," :%lu %s\n",from_mark
,sha1_to_hex(t
->sha1
));
1603 static const char fast_import_usage
[] =
1604 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file] [--branch-log=log] temp.pack";
1606 int main(int argc
, const char **argv
)
1608 const char *base_name
;
1610 unsigned long est_obj_cnt
= 1000;
1616 git_config(git_default_config
);
1618 for (i
= 1; i
< argc
; i
++) {
1619 const char *a
= argv
[i
];
1621 if (*a
!= '-' || !strcmp(a
, "--"))
1623 else if (!strncmp(a
, "--objects=", 10))
1624 est_obj_cnt
= strtoul(a
+ 10, NULL
, 0);
1625 else if (!strncmp(a
, "--depth=", 8))
1626 max_depth
= strtoul(a
+ 8, NULL
, 0);
1627 else if (!strncmp(a
, "--active-branches=", 18))
1628 max_active_branches
= strtoul(a
+ 18, NULL
, 0);
1629 else if (!strncmp(a
, "--export-marks=", 15))
1631 else if (!strncmp(a
, "--branch-log=", 13)) {
1632 branch_log
= fopen(a
+ 13, "w");
1634 die("Can't create %s: %s", a
+ 13, strerror(errno
));
1637 die("unknown option %s", a
);
1640 usage(fast_import_usage
);
1641 base_name
= argv
[i
];
1643 pack_name
= xmalloc(strlen(base_name
) + 6);
1644 sprintf(pack_name
, "%s.pack", base_name
);
1645 idx_name
= xmalloc(strlen(base_name
) + 5);
1646 sprintf(idx_name
, "%s.idx", base_name
);
1648 pack_fd
= open(pack_name
, O_RDWR
|O_CREAT
|O_EXCL
, 0666);
1650 die("Can't create %s: %s", pack_name
, strerror(errno
));
1653 alloc_objects(est_obj_cnt
);
1654 strbuf_init(&command_buf
);
1656 atom_table
= xcalloc(atom_table_sz
, sizeof(struct atom_str
*));
1657 branch_table
= xcalloc(branch_table_sz
, sizeof(struct branch
*));
1658 avail_tree_table
= xcalloc(avail_tree_table_sz
, sizeof(struct avail_tree_content
*));
1659 marks
= pool_calloc(1, sizeof(struct mark_set
));
1662 read_next_command();
1663 if (command_buf
.eof
)
1665 else if (!strcmp("blob", command_buf
.buf
))
1667 else if (!strncmp("commit ", command_buf
.buf
, 7))
1669 else if (!strncmp("tag ", command_buf
.buf
, 4))
1672 die("Unsupported command: %s", command_buf
.buf
);
1675 fixup_header_footer();
1677 write_index(idx_name
);
1683 fprintf(stderr
, "%s statistics:\n", argv
[0]);
1684 fprintf(stderr
, "---------------------------------------------------\n");
1685 fprintf(stderr
, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count
, alloc_count
- est_obj_cnt
);
1686 fprintf(stderr
, "Total objects: %10lu (%10lu duplicates)\n", object_count
, duplicate_count
);
1687 fprintf(stderr
, " blobs : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_BLOB
], duplicate_count_by_type
[OBJ_BLOB
]);
1688 fprintf(stderr
, " trees : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_TREE
], duplicate_count_by_type
[OBJ_TREE
]);
1689 fprintf(stderr
, " commits: %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_COMMIT
], duplicate_count_by_type
[OBJ_COMMIT
]);
1690 fprintf(stderr
, " tags : %10lu (%10lu duplicates)\n", object_count_by_type
[OBJ_TAG
], duplicate_count_by_type
[OBJ_TAG
]);
1691 fprintf(stderr
, "Total branches: %10lu (%10lu loads )\n", branch_count
, branch_load_count
);
1692 fprintf(stderr
, " marks: %10u (%10lu unique )\n", (1 << marks
->shift
) * 1024, marks_set_count
);
1693 fprintf(stderr
, " atoms: %10u\n", atom_cnt
);
1694 fprintf(stderr
, "Memory total: %10lu KiB\n", (total_allocd
+ alloc_count
*sizeof(struct object_entry
))/1024);
1695 fprintf(stderr
, " pools: %10lu KiB\n", total_allocd
/1024);
1696 fprintf(stderr
, " objects: %10lu KiB\n", (alloc_count
*sizeof(struct object_entry
))/1024);
1697 fprintf(stderr
, "Pack remaps: %10lu\n", remap_count
);
1698 fprintf(stderr
, "---------------------------------------------------\n");
1700 stat(pack_name
, &sb
);
1701 fprintf(stderr
, "Pack size: %10lu KiB\n", (unsigned long)(sb
.st_size
/1024));
1702 stat(idx_name
, &sb
);
1703 fprintf(stderr
, "Index size: %10lu KiB\n", (unsigned long)(sb
.st_size
/1024));
1705 fprintf(stderr
, "\n");