2 Format of STDIN stream:
14 new_blob ::= 'blob' lf
17 file_content ::= data;
19 new_commit ::= 'commit' sp ref_str lf
21 ('author' sp name '<' email '>' when lf)?
22 'committer' sp name '<' email '>' when lf
24 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
25 ('merge' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)*
30 file_change ::= file_clr
36 file_clr ::= 'deleteall' lf;
37 file_del ::= 'D' sp path_str lf;
38 file_rnm ::= 'R' sp path_str sp path_str lf;
39 file_cpy ::= 'C' sp path_str sp path_str lf;
40 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
41 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
44 new_tag ::= 'tag' sp tag_str lf
45 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
46 'tagger' sp name '<' email '>' when lf
50 reset_branch ::= 'reset' sp ref_str lf
51 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
54 checkpoint ::= 'checkpoint' lf
57 progress ::= 'progress' sp not_lf* lf
60 # note: the first idnum in a stream should be 1 and subsequent
61 # idnums should not have gaps between values as this will cause
62 # the stream parser to reserve space for the gapped values. An
63 # idnum can be updated in the future to a new object by issuing
64 # a new mark directive with the old idnum.
66 mark ::= 'mark' sp idnum lf;
67 data ::= (delimited_data | exact_data)
70 # note: delim may be any string but must not contain lf.
71 # data_line may contain any data but must not be exactly
73 delimited_data ::= 'data' sp '<<' delim lf
77 # note: declen indicates the length of binary_data in bytes.
78 # declen does not include the lf preceeding the binary data.
80 exact_data ::= 'data' sp declen lf
83 # note: quoted strings are C-style quoting supporting \c for
84 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
85 # is the signed byte value in octal. Note that the only
86 # characters which must actually be escaped to protect the
87 # stream formatting is: \, " and LF. Otherwise these values
91 sha1exp_str ::= sha1exp;
93 path_str ::= path | '"' quoted(path) '"' ;
94 mode ::= '100644' | '644'
99 declen ::= # unsigned 32 bit value, ascii base10 notation;
100 bigint ::= # unsigned integer value, ascii base10 notation;
101 binary_data ::= # file content, not interpreted;
103 when ::= raw_when | rfc2822_when;
104 raw_when ::= ts sp tz;
105 rfc2822_when ::= # Valid RFC 2822 date and time;
107 sp ::= # ASCII space character;
108 lf ::= # ASCII newline (LF) character;
110 # note: a colon (':') must precede the numerical value assigned to
111 # an idnum. This is to distinguish it from a ref or tag name as
112 # GIT does not permit ':' in ref or tag strings.
114 idnum ::= ':' bigint;
115 path ::= # GIT style file path, e.g. "a/b/c";
116 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
117 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
118 sha1exp ::= # Any valid GIT SHA1 expression;
119 hexsha1 ::= # SHA1 in hexadecimal format;
121 # note: name and email are UTF8 strings, however name must not
122 # contain '<' or lf and email must not contain any of the
123 # following: '<', '>', lf.
125 name ::= # valid GIT author/committer name;
126 email ::= # valid GIT author/committer email;
127 ts ::= # time since the epoch in seconds, ascii base10 notation;
128 tz ::= # GIT style timezone;
130 # note: comments may appear anywhere in the input, except
131 # within a data command. Any form of the data command
132 # always escapes the related input from comment processing.
134 # In case it is not clear, the '#' that starts the comment
135 # must be the first character on that the line (an lf have
138 comment ::= '#' not_lf* lf;
139 not_lf ::= # Any byte that is not ASCII newline (LF);
151 #include "csum-file.h"
155 #define PACK_ID_BITS 16
156 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
160 struct object_entry
*next
;
162 unsigned type
: TYPE_BITS
;
163 unsigned pack_id
: PACK_ID_BITS
;
164 unsigned char sha1
[20];
167 struct object_entry_pool
169 struct object_entry_pool
*next_pool
;
170 struct object_entry
*next_free
;
171 struct object_entry
*end
;
172 struct object_entry entries
[FLEX_ARRAY
]; /* more */
178 struct object_entry
*marked
[1024];
179 struct mark_set
*sets
[1024];
195 struct mem_pool
*next_pool
;
198 char space
[FLEX_ARRAY
]; /* more */
203 struct atom_str
*next_atom
;
204 unsigned short str_len
;
205 char str_dat
[FLEX_ARRAY
]; /* more */
211 struct tree_content
*tree
;
212 struct atom_str
* name
;
216 unsigned char sha1
[20];
222 unsigned int entry_capacity
; /* must match avail_tree_content */
223 unsigned int entry_count
;
224 unsigned int delta_depth
;
225 struct tree_entry
*entries
[FLEX_ARRAY
]; /* more */
228 struct avail_tree_content
230 unsigned int entry_capacity
; /* must match tree_content */
231 struct avail_tree_content
*next_avail
;
236 struct branch
*table_next_branch
;
237 struct branch
*active_next_branch
;
239 struct tree_entry branch_tree
;
240 uintmax_t last_commit
;
242 unsigned pack_id
: PACK_ID_BITS
;
243 unsigned char sha1
[20];
248 struct tag
*next_tag
;
250 unsigned int pack_id
;
251 unsigned char sha1
[20];
262 struct hash_list
*next
;
263 unsigned char sha1
[20];
272 struct recent_command
274 struct recent_command
*prev
;
275 struct recent_command
*next
;
279 /* Configured limits on output */
280 static unsigned long max_depth
= 10;
281 static off_t max_packsize
= (1LL << 32) - 1;
282 static int force_update
;
284 /* Stats and misc. counters */
285 static uintmax_t alloc_count
;
286 static uintmax_t marks_set_count
;
287 static uintmax_t object_count_by_type
[1 << TYPE_BITS
];
288 static uintmax_t duplicate_count_by_type
[1 << TYPE_BITS
];
289 static uintmax_t delta_count_by_type
[1 << TYPE_BITS
];
290 static unsigned long object_count
;
291 static unsigned long branch_count
;
292 static unsigned long branch_load_count
;
294 static FILE *pack_edges
;
297 static size_t mem_pool_alloc
= 2*1024*1024 - sizeof(struct mem_pool
);
298 static size_t total_allocd
;
299 static struct mem_pool
*mem_pool
;
301 /* Atom management */
302 static unsigned int atom_table_sz
= 4451;
303 static unsigned int atom_cnt
;
304 static struct atom_str
**atom_table
;
306 /* The .pack file being generated */
307 static unsigned int pack_id
;
308 static struct packed_git
*pack_data
;
309 static struct packed_git
**all_packs
;
310 static unsigned long pack_size
;
312 /* Table of objects we've written. */
313 static unsigned int object_entry_alloc
= 5000;
314 static struct object_entry_pool
*blocks
;
315 static struct object_entry
*object_table
[1 << 16];
316 static struct mark_set
*marks
;
317 static const char* mark_file
;
320 static struct last_object last_blob
;
322 /* Tree management */
323 static unsigned int tree_entry_alloc
= 1000;
324 static void *avail_tree_entry
;
325 static unsigned int avail_tree_table_sz
= 100;
326 static struct avail_tree_content
**avail_tree_table
;
327 static struct dbuf old_tree
;
328 static struct dbuf new_tree
;
331 static unsigned long max_active_branches
= 5;
332 static unsigned long cur_active_branches
;
333 static unsigned long branch_table_sz
= 1039;
334 static struct branch
**branch_table
;
335 static struct branch
*active_branches
;
338 static struct tag
*first_tag
;
339 static struct tag
*last_tag
;
341 /* Input stream parsing */
342 static whenspec_type whenspec
= WHENSPEC_RAW
;
343 static struct strbuf command_buf
;
344 static int unread_command_buf
;
345 static struct recent_command cmd_hist
= {&cmd_hist
, &cmd_hist
, NULL
};
346 static struct recent_command
*cmd_tail
= &cmd_hist
;
347 static struct recent_command
*rc_free
;
348 static unsigned int cmd_save
= 100;
349 static uintmax_t next_mark
;
350 static struct dbuf new_data
;
352 static void write_branch_report(FILE *rpt
, struct branch
*b
)
354 fprintf(rpt
, "%s:\n", b
->name
);
356 fprintf(rpt
, " status :");
358 fputs(" active", rpt
);
359 if (b
->branch_tree
.tree
)
360 fputs(" loaded", rpt
);
361 if (is_null_sha1(b
->branch_tree
.versions
[1].sha1
))
362 fputs(" dirty", rpt
);
365 fprintf(rpt
, " tip commit : %s\n", sha1_to_hex(b
->sha1
));
366 fprintf(rpt
, " old tree : %s\n", sha1_to_hex(b
->branch_tree
.versions
[0].sha1
));
367 fprintf(rpt
, " cur tree : %s\n", sha1_to_hex(b
->branch_tree
.versions
[1].sha1
));
368 fprintf(rpt
, " commit clock: %" PRIuMAX
"\n", b
->last_commit
);
370 fputs(" last pack : ", rpt
);
371 if (b
->pack_id
< MAX_PACK_ID
)
372 fprintf(rpt
, "%u", b
->pack_id
);
378 static void write_crash_report(const char *err
)
380 char *loc
= git_path("fast_import_crash_%d", getpid());
381 FILE *rpt
= fopen(loc
, "w");
384 struct recent_command
*rc
;
387 error("can't write crash report %s: %s", loc
, strerror(errno
));
391 fprintf(stderr
, "fast-import: dumping crash report to %s\n", loc
);
393 fprintf(rpt
, "fast-import crash report:\n");
394 fprintf(rpt
, " fast-import process: %d\n", getpid());
396 fprintf(rpt
, " parent process : %d\n", getppid());
398 fprintf(rpt
, " at %s\n", show_date(time(NULL
), 0, DATE_LOCAL
));
401 fputs("fatal: ", rpt
);
406 fputs("Most Recent Commands Before Crash\n", rpt
);
407 fputs("---------------------------------\n", rpt
);
408 for (rc
= cmd_hist
.next
; rc
!= &cmd_hist
; rc
= rc
->next
) {
409 if (rc
->next
== &cmd_hist
)
418 fputs("Active Branch LRU\n", rpt
);
419 fputs("-----------------\n", rpt
);
420 fprintf(rpt
, " active_branches = %lu cur, %lu max\n",
422 max_active_branches
);
424 fputs(" pos clock name\n", rpt
);
425 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt
);
426 for (b
= active_branches
, lu
= 0; b
; b
= b
->active_next_branch
)
427 fprintf(rpt
, " %2lu) %6" PRIuMAX
" %s\n",
428 ++lu
, b
->last_commit
, b
->name
);
431 fputs("Inactive Branches\n", rpt
);
432 fputs("-----------------\n", rpt
);
433 for (lu
= 0; lu
< branch_table_sz
; lu
++) {
434 for (b
= branch_table
[lu
]; b
; b
= b
->table_next_branch
)
435 write_branch_report(rpt
, b
);
439 fputs("-------------------\n", rpt
);
440 fputs("END OF CRASH REPORT\n", rpt
);
444 static NORETURN
void die_nicely(const char *err
, va_list params
)
447 char message
[2 * PATH_MAX
];
449 vsnprintf(message
, sizeof(message
), err
, params
);
450 fputs("fatal: ", stderr
);
451 fputs(message
, stderr
);
456 write_crash_report(message
);
461 static void alloc_objects(unsigned int cnt
)
463 struct object_entry_pool
*b
;
465 b
= xmalloc(sizeof(struct object_entry_pool
)
466 + cnt
* sizeof(struct object_entry
));
467 b
->next_pool
= blocks
;
468 b
->next_free
= b
->entries
;
469 b
->end
= b
->entries
+ cnt
;
474 static struct object_entry
*new_object(unsigned char *sha1
)
476 struct object_entry
*e
;
478 if (blocks
->next_free
== blocks
->end
)
479 alloc_objects(object_entry_alloc
);
481 e
= blocks
->next_free
++;
482 hashcpy(e
->sha1
, sha1
);
486 static struct object_entry
*find_object(unsigned char *sha1
)
488 unsigned int h
= sha1
[0] << 8 | sha1
[1];
489 struct object_entry
*e
;
490 for (e
= object_table
[h
]; e
; e
= e
->next
)
491 if (!hashcmp(sha1
, e
->sha1
))
496 static struct object_entry
*insert_object(unsigned char *sha1
)
498 unsigned int h
= sha1
[0] << 8 | sha1
[1];
499 struct object_entry
*e
= object_table
[h
];
500 struct object_entry
*p
= NULL
;
503 if (!hashcmp(sha1
, e
->sha1
))
509 e
= new_object(sha1
);
519 static unsigned int hc_str(const char *s
, size_t len
)
527 static void *pool_alloc(size_t len
)
532 for (p
= mem_pool
; p
; p
= p
->next_pool
)
533 if ((p
->end
- p
->next_free
>= len
))
537 if (len
>= (mem_pool_alloc
/2)) {
541 total_allocd
+= sizeof(struct mem_pool
) + mem_pool_alloc
;
542 p
= xmalloc(sizeof(struct mem_pool
) + mem_pool_alloc
);
543 p
->next_pool
= mem_pool
;
544 p
->next_free
= p
->space
;
545 p
->end
= p
->next_free
+ mem_pool_alloc
;
550 /* round out to a pointer alignment */
551 if (len
& (sizeof(void*) - 1))
552 len
+= sizeof(void*) - (len
& (sizeof(void*) - 1));
557 static void *pool_calloc(size_t count
, size_t size
)
559 size_t len
= count
* size
;
560 void *r
= pool_alloc(len
);
565 static char *pool_strdup(const char *s
)
567 char *r
= pool_alloc(strlen(s
) + 1);
572 static void size_dbuf(struct dbuf
*b
, size_t maxlen
)
575 if (b
->capacity
>= maxlen
)
579 b
->capacity
= ((maxlen
/ 1024) + 1) * 1024;
580 b
->buffer
= xmalloc(b
->capacity
);
583 static void insert_mark(uintmax_t idnum
, struct object_entry
*oe
)
585 struct mark_set
*s
= marks
;
586 while ((idnum
>> s
->shift
) >= 1024) {
587 s
= pool_calloc(1, sizeof(struct mark_set
));
588 s
->shift
= marks
->shift
+ 10;
589 s
->data
.sets
[0] = marks
;
593 uintmax_t i
= idnum
>> s
->shift
;
594 idnum
-= i
<< s
->shift
;
595 if (!s
->data
.sets
[i
]) {
596 s
->data
.sets
[i
] = pool_calloc(1, sizeof(struct mark_set
));
597 s
->data
.sets
[i
]->shift
= s
->shift
- 10;
601 if (!s
->data
.marked
[idnum
])
603 s
->data
.marked
[idnum
] = oe
;
606 static struct object_entry
*find_mark(uintmax_t idnum
)
608 uintmax_t orig_idnum
= idnum
;
609 struct mark_set
*s
= marks
;
610 struct object_entry
*oe
= NULL
;
611 if ((idnum
>> s
->shift
) < 1024) {
612 while (s
&& s
->shift
) {
613 uintmax_t i
= idnum
>> s
->shift
;
614 idnum
-= i
<< s
->shift
;
618 oe
= s
->data
.marked
[idnum
];
621 die("mark :%" PRIuMAX
" not declared", orig_idnum
);
625 static struct atom_str
*to_atom(const char *s
, unsigned short len
)
627 unsigned int hc
= hc_str(s
, len
) % atom_table_sz
;
630 for (c
= atom_table
[hc
]; c
; c
= c
->next_atom
)
631 if (c
->str_len
== len
&& !strncmp(s
, c
->str_dat
, len
))
634 c
= pool_alloc(sizeof(struct atom_str
) + len
+ 1);
636 strncpy(c
->str_dat
, s
, len
);
638 c
->next_atom
= atom_table
[hc
];
644 static struct branch
*lookup_branch(const char *name
)
646 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
649 for (b
= branch_table
[hc
]; b
; b
= b
->table_next_branch
)
650 if (!strcmp(name
, b
->name
))
655 static struct branch
*new_branch(const char *name
)
657 unsigned int hc
= hc_str(name
, strlen(name
)) % branch_table_sz
;
658 struct branch
* b
= lookup_branch(name
);
661 die("Invalid attempt to create duplicate branch: %s", name
);
662 switch (check_ref_format(name
)) {
663 case 0: break; /* its valid */
664 case -2: break; /* valid, but too few '/', allow anyway */
666 die("Branch name doesn't conform to GIT standards: %s", name
);
669 b
= pool_calloc(1, sizeof(struct branch
));
670 b
->name
= pool_strdup(name
);
671 b
->table_next_branch
= branch_table
[hc
];
672 b
->branch_tree
.versions
[0].mode
= S_IFDIR
;
673 b
->branch_tree
.versions
[1].mode
= S_IFDIR
;
675 b
->pack_id
= MAX_PACK_ID
;
676 branch_table
[hc
] = b
;
681 static unsigned int hc_entries(unsigned int cnt
)
683 cnt
= cnt
& 7 ? (cnt
/ 8) + 1 : cnt
/ 8;
684 return cnt
< avail_tree_table_sz
? cnt
: avail_tree_table_sz
- 1;
687 static struct tree_content
*new_tree_content(unsigned int cnt
)
689 struct avail_tree_content
*f
, *l
= NULL
;
690 struct tree_content
*t
;
691 unsigned int hc
= hc_entries(cnt
);
693 for (f
= avail_tree_table
[hc
]; f
; l
= f
, f
= f
->next_avail
)
694 if (f
->entry_capacity
>= cnt
)
699 l
->next_avail
= f
->next_avail
;
701 avail_tree_table
[hc
] = f
->next_avail
;
703 cnt
= cnt
& 7 ? ((cnt
/ 8) + 1) * 8 : cnt
;
704 f
= pool_alloc(sizeof(*t
) + sizeof(t
->entries
[0]) * cnt
);
705 f
->entry_capacity
= cnt
;
708 t
= (struct tree_content
*)f
;
714 static void release_tree_entry(struct tree_entry
*e
);
715 static void release_tree_content(struct tree_content
*t
)
717 struct avail_tree_content
*f
= (struct avail_tree_content
*)t
;
718 unsigned int hc
= hc_entries(f
->entry_capacity
);
719 f
->next_avail
= avail_tree_table
[hc
];
720 avail_tree_table
[hc
] = f
;
723 static void release_tree_content_recursive(struct tree_content
*t
)
726 for (i
= 0; i
< t
->entry_count
; i
++)
727 release_tree_entry(t
->entries
[i
]);
728 release_tree_content(t
);
731 static struct tree_content
*grow_tree_content(
732 struct tree_content
*t
,
735 struct tree_content
*r
= new_tree_content(t
->entry_count
+ amt
);
736 r
->entry_count
= t
->entry_count
;
737 r
->delta_depth
= t
->delta_depth
;
738 memcpy(r
->entries
,t
->entries
,t
->entry_count
*sizeof(t
->entries
[0]));
739 release_tree_content(t
);
743 static struct tree_entry
*new_tree_entry(void)
745 struct tree_entry
*e
;
747 if (!avail_tree_entry
) {
748 unsigned int n
= tree_entry_alloc
;
749 total_allocd
+= n
* sizeof(struct tree_entry
);
750 avail_tree_entry
= e
= xmalloc(n
* sizeof(struct tree_entry
));
752 *((void**)e
) = e
+ 1;
758 e
= avail_tree_entry
;
759 avail_tree_entry
= *((void**)e
);
763 static void release_tree_entry(struct tree_entry
*e
)
766 release_tree_content_recursive(e
->tree
);
767 *((void**)e
) = avail_tree_entry
;
768 avail_tree_entry
= e
;
771 static struct tree_content
*dup_tree_content(struct tree_content
*s
)
773 struct tree_content
*d
;
774 struct tree_entry
*a
, *b
;
779 d
= new_tree_content(s
->entry_count
);
780 for (i
= 0; i
< s
->entry_count
; i
++) {
782 b
= new_tree_entry();
783 memcpy(b
, a
, sizeof(*a
));
784 if (a
->tree
&& is_null_sha1(b
->versions
[1].sha1
))
785 b
->tree
= dup_tree_content(a
->tree
);
790 d
->entry_count
= s
->entry_count
;
791 d
->delta_depth
= s
->delta_depth
;
796 static void start_packfile(void)
798 static char tmpfile
[PATH_MAX
];
799 struct packed_git
*p
;
800 struct pack_header hdr
;
803 snprintf(tmpfile
, sizeof(tmpfile
),
804 "%s/tmp_pack_XXXXXX", get_object_directory());
805 pack_fd
= xmkstemp(tmpfile
);
806 p
= xcalloc(1, sizeof(*p
) + strlen(tmpfile
) + 2);
807 strcpy(p
->pack_name
, tmpfile
);
808 p
->pack_fd
= pack_fd
;
810 hdr
.hdr_signature
= htonl(PACK_SIGNATURE
);
811 hdr
.hdr_version
= htonl(2);
813 write_or_die(p
->pack_fd
, &hdr
, sizeof(hdr
));
816 pack_size
= sizeof(hdr
);
819 all_packs
= xrealloc(all_packs
, sizeof(*all_packs
) * (pack_id
+ 1));
820 all_packs
[pack_id
] = p
;
823 static int oecmp (const void *a_
, const void *b_
)
825 struct object_entry
*a
= *((struct object_entry
**)a_
);
826 struct object_entry
*b
= *((struct object_entry
**)b_
);
827 return hashcmp(a
->sha1
, b
->sha1
);
830 static char *create_index(void)
832 static char tmpfile
[PATH_MAX
];
835 struct object_entry
**idx
, **c
, **last
, *e
;
836 struct object_entry_pool
*o
;
840 /* Build the sorted table of object IDs. */
841 idx
= xmalloc(object_count
* sizeof(struct object_entry
*));
843 for (o
= blocks
; o
; o
= o
->next_pool
)
844 for (e
= o
->next_free
; e
-- != o
->entries
;)
845 if (pack_id
== e
->pack_id
)
847 last
= idx
+ object_count
;
849 die("internal consistency error creating the index");
850 qsort(idx
, object_count
, sizeof(struct object_entry
*), oecmp
);
852 /* Generate the fan-out array. */
854 for (i
= 0; i
< 256; i
++) {
855 struct object_entry
**next
= c
;;
856 while (next
< last
) {
857 if ((*next
)->sha1
[0] != i
)
861 array
[i
] = htonl(next
- idx
);
865 snprintf(tmpfile
, sizeof(tmpfile
),
866 "%s/tmp_idx_XXXXXX", get_object_directory());
867 idx_fd
= xmkstemp(tmpfile
);
868 f
= sha1fd(idx_fd
, tmpfile
);
869 sha1write(f
, array
, 256 * sizeof(int));
871 for (c
= idx
; c
!= last
; c
++) {
872 uint32_t offset
= htonl((*c
)->offset
);
873 sha1write(f
, &offset
, 4);
874 sha1write(f
, (*c
)->sha1
, sizeof((*c
)->sha1
));
875 SHA1_Update(&ctx
, (*c
)->sha1
, 20);
877 sha1write(f
, pack_data
->sha1
, sizeof(pack_data
->sha1
));
878 sha1close(f
, NULL
, 1);
880 SHA1_Final(pack_data
->sha1
, &ctx
);
884 static char *keep_pack(char *curr_index_name
)
886 static char name
[PATH_MAX
];
887 static const char *keep_msg
= "fast-import";
890 chmod(pack_data
->pack_name
, 0444);
891 chmod(curr_index_name
, 0444);
893 snprintf(name
, sizeof(name
), "%s/pack/pack-%s.keep",
894 get_object_directory(), sha1_to_hex(pack_data
->sha1
));
895 keep_fd
= open(name
, O_RDWR
|O_CREAT
|O_EXCL
, 0600);
897 die("cannot create keep file");
898 write(keep_fd
, keep_msg
, strlen(keep_msg
));
901 snprintf(name
, sizeof(name
), "%s/pack/pack-%s.pack",
902 get_object_directory(), sha1_to_hex(pack_data
->sha1
));
903 if (move_temp_to_file(pack_data
->pack_name
, name
))
904 die("cannot store pack file");
906 snprintf(name
, sizeof(name
), "%s/pack/pack-%s.idx",
907 get_object_directory(), sha1_to_hex(pack_data
->sha1
));
908 if (move_temp_to_file(curr_index_name
, name
))
909 die("cannot store index file");
913 static void unkeep_all_packs(void)
915 static char name
[PATH_MAX
];
918 for (k
= 0; k
< pack_id
; k
++) {
919 struct packed_git
*p
= all_packs
[k
];
920 snprintf(name
, sizeof(name
), "%s/pack/pack-%s.keep",
921 get_object_directory(), sha1_to_hex(p
->sha1
));
926 static void end_packfile(void)
928 struct packed_git
*old_p
= pack_data
, *new_p
;
936 fixup_pack_header_footer(pack_data
->pack_fd
, pack_data
->sha1
,
937 pack_data
->pack_name
, object_count
);
938 close(pack_data
->pack_fd
);
939 idx_name
= keep_pack(create_index());
941 /* Register the packfile with core git's machinary. */
942 new_p
= add_packed_git(idx_name
, strlen(idx_name
), 1);
944 die("core git rejected index %s", idx_name
);
945 new_p
->windows
= old_p
->windows
;
946 all_packs
[pack_id
] = new_p
;
947 install_packed_git(new_p
);
949 /* Print the boundary */
951 fprintf(pack_edges
, "%s:", new_p
->pack_name
);
952 for (i
= 0; i
< branch_table_sz
; i
++) {
953 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
) {
954 if (b
->pack_id
== pack_id
)
955 fprintf(pack_edges
, " %s", sha1_to_hex(b
->sha1
));
958 for (t
= first_tag
; t
; t
= t
->next_tag
) {
959 if (t
->pack_id
== pack_id
)
960 fprintf(pack_edges
, " %s", sha1_to_hex(t
->sha1
));
962 fputc('\n', pack_edges
);
969 unlink(old_p
->pack_name
);
972 /* We can't carry a delta across packfiles. */
973 free(last_blob
.data
);
974 last_blob
.data
= NULL
;
976 last_blob
.offset
= 0;
980 static void cycle_packfile(void)
986 static size_t encode_header(
987 enum object_type type
,
994 if (type
< OBJ_COMMIT
|| type
> OBJ_REF_DELTA
)
995 die("bad type %d", type
);
997 c
= (type
<< 4) | (size
& 15);
1009 static int store_object(
1010 enum object_type type
,
1013 struct last_object
*last
,
1014 unsigned char *sha1out
,
1018 struct object_entry
*e
;
1019 unsigned char hdr
[96];
1020 unsigned char sha1
[20];
1021 unsigned long hdrlen
, deltalen
;
1025 hdrlen
= sprintf((char*)hdr
,"%s %lu", typename(type
),
1026 (unsigned long)datlen
) + 1;
1028 SHA1_Update(&c
, hdr
, hdrlen
);
1029 SHA1_Update(&c
, dat
, datlen
);
1030 SHA1_Final(sha1
, &c
);
1032 hashcpy(sha1out
, sha1
);
1034 e
= insert_object(sha1
);
1036 insert_mark(mark
, e
);
1038 duplicate_count_by_type
[type
]++;
1040 } else if (find_sha1_pack(sha1
, packed_git
)) {
1042 e
->pack_id
= MAX_PACK_ID
;
1043 e
->offset
= 1; /* just not zero! */
1044 duplicate_count_by_type
[type
]++;
1048 if (last
&& last
->data
&& last
->depth
< max_depth
) {
1049 delta
= diff_delta(last
->data
, last
->len
,
1052 if (delta
&& deltalen
>= datlen
) {
1059 memset(&s
, 0, sizeof(s
));
1060 deflateInit(&s
, zlib_compression_level
);
1063 s
.avail_in
= deltalen
;
1066 s
.avail_in
= datlen
;
1068 s
.avail_out
= deflateBound(&s
, s
.avail_in
);
1069 s
.next_out
= out
= xmalloc(s
.avail_out
);
1070 while (deflate(&s
, Z_FINISH
) == Z_OK
)
1074 /* Determine if we should auto-checkpoint. */
1075 if ((pack_size
+ 60 + s
.total_out
) > max_packsize
1076 || (pack_size
+ 60 + s
.total_out
) < pack_size
) {
1078 /* This new object needs to *not* have the current pack_id. */
1079 e
->pack_id
= pack_id
+ 1;
1082 /* We cannot carry a delta into the new pack. */
1087 memset(&s
, 0, sizeof(s
));
1088 deflateInit(&s
, zlib_compression_level
);
1090 s
.avail_in
= datlen
;
1091 s
.avail_out
= deflateBound(&s
, s
.avail_in
);
1092 s
.next_out
= out
= xrealloc(out
, s
.avail_out
);
1093 while (deflate(&s
, Z_FINISH
) == Z_OK
)
1100 e
->pack_id
= pack_id
;
1101 e
->offset
= pack_size
;
1103 object_count_by_type
[type
]++;
1106 unsigned long ofs
= e
->offset
- last
->offset
;
1107 unsigned pos
= sizeof(hdr
) - 1;
1109 delta_count_by_type
[type
]++;
1112 hdrlen
= encode_header(OBJ_OFS_DELTA
, deltalen
, hdr
);
1113 write_or_die(pack_data
->pack_fd
, hdr
, hdrlen
);
1114 pack_size
+= hdrlen
;
1116 hdr
[pos
] = ofs
& 127;
1118 hdr
[--pos
] = 128 | (--ofs
& 127);
1119 write_or_die(pack_data
->pack_fd
, hdr
+ pos
, sizeof(hdr
) - pos
);
1120 pack_size
+= sizeof(hdr
) - pos
;
1124 hdrlen
= encode_header(type
, datlen
, hdr
);
1125 write_or_die(pack_data
->pack_fd
, hdr
, hdrlen
);
1126 pack_size
+= hdrlen
;
1129 write_or_die(pack_data
->pack_fd
, out
, s
.total_out
);
1130 pack_size
+= s
.total_out
;
1138 last
->offset
= e
->offset
;
1144 static void *gfi_unpack_entry(
1145 struct object_entry
*oe
,
1146 unsigned long *sizep
)
1148 enum object_type type
;
1149 struct packed_git
*p
= all_packs
[oe
->pack_id
];
1151 p
->pack_size
= pack_size
+ 20;
1152 return unpack_entry(p
, oe
->offset
, &type
, sizep
);
1155 static const char *get_mode(const char *str
, uint16_t *modep
)
1160 while ((c
= *str
++) != ' ') {
1161 if (c
< '0' || c
> '7')
1163 mode
= (mode
<< 3) + (c
- '0');
1169 static void load_tree(struct tree_entry
*root
)
1171 unsigned char* sha1
= root
->versions
[1].sha1
;
1172 struct object_entry
*myoe
;
1173 struct tree_content
*t
;
1178 root
->tree
= t
= new_tree_content(8);
1179 if (is_null_sha1(sha1
))
1182 myoe
= find_object(sha1
);
1183 if (myoe
&& myoe
->pack_id
!= MAX_PACK_ID
) {
1184 if (myoe
->type
!= OBJ_TREE
)
1185 die("Not a tree: %s", sha1_to_hex(sha1
));
1187 buf
= gfi_unpack_entry(myoe
, &size
);
1189 enum object_type type
;
1190 buf
= read_sha1_file(sha1
, &type
, &size
);
1191 if (!buf
|| type
!= OBJ_TREE
)
1192 die("Can't load tree %s", sha1_to_hex(sha1
));
1196 while (c
!= (buf
+ size
)) {
1197 struct tree_entry
*e
= new_tree_entry();
1199 if (t
->entry_count
== t
->entry_capacity
)
1200 root
->tree
= t
= grow_tree_content(t
, t
->entry_count
);
1201 t
->entries
[t
->entry_count
++] = e
;
1204 c
= get_mode(c
, &e
->versions
[1].mode
);
1206 die("Corrupt mode in %s", sha1_to_hex(sha1
));
1207 e
->versions
[0].mode
= e
->versions
[1].mode
;
1208 e
->name
= to_atom(c
, strlen(c
));
1209 c
+= e
->name
->str_len
+ 1;
1210 hashcpy(e
->versions
[0].sha1
, (unsigned char*)c
);
1211 hashcpy(e
->versions
[1].sha1
, (unsigned char*)c
);
1217 static int tecmp0 (const void *_a
, const void *_b
)
1219 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
1220 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
1221 return base_name_compare(
1222 a
->name
->str_dat
, a
->name
->str_len
, a
->versions
[0].mode
,
1223 b
->name
->str_dat
, b
->name
->str_len
, b
->versions
[0].mode
);
1226 static int tecmp1 (const void *_a
, const void *_b
)
1228 struct tree_entry
*a
= *((struct tree_entry
**)_a
);
1229 struct tree_entry
*b
= *((struct tree_entry
**)_b
);
1230 return base_name_compare(
1231 a
->name
->str_dat
, a
->name
->str_len
, a
->versions
[1].mode
,
1232 b
->name
->str_dat
, b
->name
->str_len
, b
->versions
[1].mode
);
1235 static void mktree(struct tree_content
*t
,
1245 qsort(t
->entries
,t
->entry_count
,sizeof(t
->entries
[0]),tecmp0
);
1247 qsort(t
->entries
,t
->entry_count
,sizeof(t
->entries
[0]),tecmp1
);
1249 for (i
= 0; i
< t
->entry_count
; i
++) {
1250 if (t
->entries
[i
]->versions
[v
].mode
)
1251 maxlen
+= t
->entries
[i
]->name
->str_len
+ 34;
1254 size_dbuf(b
, maxlen
);
1256 for (i
= 0; i
< t
->entry_count
; i
++) {
1257 struct tree_entry
*e
= t
->entries
[i
];
1258 if (!e
->versions
[v
].mode
)
1260 c
+= sprintf(c
, "%o", (unsigned int)e
->versions
[v
].mode
);
1262 strcpy(c
, e
->name
->str_dat
);
1263 c
+= e
->name
->str_len
+ 1;
1264 hashcpy((unsigned char*)c
, e
->versions
[v
].sha1
);
1267 *szp
= c
- (char*)b
->buffer
;
1270 static void store_tree(struct tree_entry
*root
)
1272 struct tree_content
*t
= root
->tree
;
1273 unsigned int i
, j
, del
;
1274 unsigned long new_len
;
1275 struct last_object lo
;
1276 struct object_entry
*le
;
1278 if (!is_null_sha1(root
->versions
[1].sha1
))
1281 for (i
= 0; i
< t
->entry_count
; i
++) {
1282 if (t
->entries
[i
]->tree
)
1283 store_tree(t
->entries
[i
]);
1286 le
= find_object(root
->versions
[0].sha1
);
1287 if (!S_ISDIR(root
->versions
[0].mode
)
1289 || le
->pack_id
!= pack_id
) {
1294 mktree(t
, 0, &lo
.len
, &old_tree
);
1295 lo
.data
= old_tree
.buffer
;
1296 lo
.offset
= le
->offset
;
1297 lo
.depth
= t
->delta_depth
;
1301 mktree(t
, 1, &new_len
, &new_tree
);
1302 store_object(OBJ_TREE
, new_tree
.buffer
, new_len
,
1303 &lo
, root
->versions
[1].sha1
, 0);
1305 t
->delta_depth
= lo
.depth
;
1306 for (i
= 0, j
= 0, del
= 0; i
< t
->entry_count
; i
++) {
1307 struct tree_entry
*e
= t
->entries
[i
];
1308 if (e
->versions
[1].mode
) {
1309 e
->versions
[0].mode
= e
->versions
[1].mode
;
1310 hashcpy(e
->versions
[0].sha1
, e
->versions
[1].sha1
);
1311 t
->entries
[j
++] = e
;
1313 release_tree_entry(e
);
1317 t
->entry_count
-= del
;
1320 static int tree_content_set(
1321 struct tree_entry
*root
,
1323 const unsigned char *sha1
,
1324 const uint16_t mode
,
1325 struct tree_content
*subtree
)
1327 struct tree_content
*t
= root
->tree
;
1330 struct tree_entry
*e
;
1332 slash1
= strchr(p
, '/');
1338 die("Empty path component found in input");
1339 if (!slash1
&& !S_ISDIR(mode
) && subtree
)
1340 die("Non-directories cannot have subtrees");
1342 for (i
= 0; i
< t
->entry_count
; i
++) {
1344 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
1347 && e
->versions
[1].mode
== mode
1348 && !hashcmp(e
->versions
[1].sha1
, sha1
))
1350 e
->versions
[1].mode
= mode
;
1351 hashcpy(e
->versions
[1].sha1
, sha1
);
1353 release_tree_content_recursive(e
->tree
);
1355 hashclr(root
->versions
[1].sha1
);
1358 if (!S_ISDIR(e
->versions
[1].mode
)) {
1359 e
->tree
= new_tree_content(8);
1360 e
->versions
[1].mode
= S_IFDIR
;
1364 if (tree_content_set(e
, slash1
+ 1, sha1
, mode
, subtree
)) {
1365 hashclr(root
->versions
[1].sha1
);
1372 if (t
->entry_count
== t
->entry_capacity
)
1373 root
->tree
= t
= grow_tree_content(t
, t
->entry_count
);
1374 e
= new_tree_entry();
1375 e
->name
= to_atom(p
, n
);
1376 e
->versions
[0].mode
= 0;
1377 hashclr(e
->versions
[0].sha1
);
1378 t
->entries
[t
->entry_count
++] = e
;
1380 e
->tree
= new_tree_content(8);
1381 e
->versions
[1].mode
= S_IFDIR
;
1382 tree_content_set(e
, slash1
+ 1, sha1
, mode
, subtree
);
1385 e
->versions
[1].mode
= mode
;
1386 hashcpy(e
->versions
[1].sha1
, sha1
);
1388 hashclr(root
->versions
[1].sha1
);
1392 static int tree_content_remove(
1393 struct tree_entry
*root
,
1395 struct tree_entry
*backup_leaf
)
1397 struct tree_content
*t
= root
->tree
;
1400 struct tree_entry
*e
;
1402 slash1
= strchr(p
, '/');
1408 for (i
= 0; i
< t
->entry_count
; i
++) {
1410 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
1411 if (!slash1
|| !S_ISDIR(e
->versions
[1].mode
))
1415 if (tree_content_remove(e
, slash1
+ 1, backup_leaf
)) {
1416 for (n
= 0; n
< e
->tree
->entry_count
; n
++) {
1417 if (e
->tree
->entries
[n
]->versions
[1].mode
) {
1418 hashclr(root
->versions
[1].sha1
);
1432 memcpy(backup_leaf
, e
, sizeof(*backup_leaf
));
1434 release_tree_content_recursive(e
->tree
);
1436 e
->versions
[1].mode
= 0;
1437 hashclr(e
->versions
[1].sha1
);
1438 hashclr(root
->versions
[1].sha1
);
1442 static int tree_content_get(
1443 struct tree_entry
*root
,
1445 struct tree_entry
*leaf
)
1447 struct tree_content
*t
= root
->tree
;
1450 struct tree_entry
*e
;
1452 slash1
= strchr(p
, '/');
1458 for (i
= 0; i
< t
->entry_count
; i
++) {
1460 if (e
->name
->str_len
== n
&& !strncmp(p
, e
->name
->str_dat
, n
)) {
1462 memcpy(leaf
, e
, sizeof(*leaf
));
1463 if (e
->tree
&& is_null_sha1(e
->versions
[1].sha1
))
1464 leaf
->tree
= dup_tree_content(e
->tree
);
1469 if (!S_ISDIR(e
->versions
[1].mode
))
1473 return tree_content_get(e
, slash1
+ 1, leaf
);
1479 static int update_branch(struct branch
*b
)
1481 static const char *msg
= "fast-import";
1482 struct ref_lock
*lock
;
1483 unsigned char old_sha1
[20];
1485 if (read_ref(b
->name
, old_sha1
))
1487 lock
= lock_any_ref_for_update(b
->name
, old_sha1
, 0);
1489 return error("Unable to lock %s", b
->name
);
1490 if (!force_update
&& !is_null_sha1(old_sha1
)) {
1491 struct commit
*old_cmit
, *new_cmit
;
1493 old_cmit
= lookup_commit_reference_gently(old_sha1
, 0);
1494 new_cmit
= lookup_commit_reference_gently(b
->sha1
, 0);
1495 if (!old_cmit
|| !new_cmit
) {
1497 return error("Branch %s is missing commits.", b
->name
);
1500 if (!in_merge_bases(old_cmit
, &new_cmit
, 1)) {
1502 warning("Not updating %s"
1503 " (new tip %s does not contain %s)",
1504 b
->name
, sha1_to_hex(b
->sha1
), sha1_to_hex(old_sha1
));
1508 if (write_ref_sha1(lock
, b
->sha1
, msg
) < 0)
1509 return error("Unable to update %s", b
->name
);
1513 static void dump_branches(void)
1518 for (i
= 0; i
< branch_table_sz
; i
++) {
1519 for (b
= branch_table
[i
]; b
; b
= b
->table_next_branch
)
1520 failure
|= update_branch(b
);
1524 static void dump_tags(void)
1526 static const char *msg
= "fast-import";
1528 struct ref_lock
*lock
;
1529 char ref_name
[PATH_MAX
];
1531 for (t
= first_tag
; t
; t
= t
->next_tag
) {
1532 sprintf(ref_name
, "tags/%s", t
->name
);
1533 lock
= lock_ref_sha1(ref_name
, NULL
);
1534 if (!lock
|| write_ref_sha1(lock
, t
->sha1
, msg
) < 0)
1535 failure
|= error("Unable to update %s", ref_name
);
1539 static void dump_marks_helper(FILE *f
,
1545 for (k
= 0; k
< 1024; k
++) {
1546 if (m
->data
.sets
[k
])
1547 dump_marks_helper(f
, (base
+ k
) << m
->shift
,
1551 for (k
= 0; k
< 1024; k
++) {
1552 if (m
->data
.marked
[k
])
1553 fprintf(f
, ":%" PRIuMAX
" %s\n", base
+ k
,
1554 sha1_to_hex(m
->data
.marked
[k
]->sha1
));
1559 static void dump_marks(void)
1561 static struct lock_file mark_lock
;
1568 mark_fd
= hold_lock_file_for_update(&mark_lock
, mark_file
, 0);
1570 failure
|= error("Unable to write marks file %s: %s",
1571 mark_file
, strerror(errno
));
1575 f
= fdopen(mark_fd
, "w");
1577 rollback_lock_file(&mark_lock
);
1578 failure
|= error("Unable to write marks file %s: %s",
1579 mark_file
, strerror(errno
));
1583 dump_marks_helper(f
, 0, marks
);
1585 if (commit_lock_file(&mark_lock
))
1586 failure
|= error("Unable to write marks file %s: %s",
1587 mark_file
, strerror(errno
));
1590 static void read_next_command(void)
1593 if (unread_command_buf
) {
1594 unread_command_buf
= 0;
1595 if (command_buf
.eof
)
1598 struct recent_command
*rc
;
1600 command_buf
.buf
= NULL
;
1601 read_line(&command_buf
, stdin
, '\n');
1602 if (command_buf
.eof
)
1610 cmd_hist
.next
= rc
->next
;
1611 cmd_hist
.next
->prev
= &cmd_hist
;
1615 rc
->buf
= command_buf
.buf
;
1616 rc
->prev
= cmd_tail
;
1617 rc
->next
= cmd_hist
.prev
;
1618 rc
->prev
->next
= rc
;
1621 } while (command_buf
.buf
[0] == '#');
1624 static void skip_optional_lf(void)
1626 int term_char
= fgetc(stdin
);
1627 if (term_char
!= '\n' && term_char
!= EOF
)
1628 ungetc(term_char
, stdin
);
1631 static void cmd_mark(void)
1633 if (!prefixcmp(command_buf
.buf
, "mark :")) {
1634 next_mark
= strtoumax(command_buf
.buf
+ 6, NULL
, 10);
1635 read_next_command();
1641 static void *cmd_data (size_t *size
)
1646 if (prefixcmp(command_buf
.buf
, "data "))
1647 die("Expected 'data n' command, found: %s", command_buf
.buf
);
1649 if (!prefixcmp(command_buf
.buf
+ 5, "<<")) {
1650 char *term
= xstrdup(command_buf
.buf
+ 5 + 2);
1651 size_t sz
= 8192, term_len
= command_buf
.len
- 5 - 2;
1653 buffer
= xmalloc(sz
);
1654 command_buf
.buf
= NULL
;
1656 read_line(&command_buf
, stdin
, '\n');
1657 if (command_buf
.eof
)
1658 die("EOF in data (terminator '%s' not found)", term
);
1659 if (term_len
== command_buf
.len
1660 && !strcmp(term
, command_buf
.buf
))
1662 ALLOC_GROW(buffer
, length
+ command_buf
.len
, sz
);
1663 memcpy(buffer
+ length
,
1665 command_buf
.len
- 1);
1666 length
+= command_buf
.len
- 1;
1667 buffer
[length
++] = '\n';
1673 length
= strtoul(command_buf
.buf
+ 5, NULL
, 10);
1674 buffer
= xmalloc(length
);
1675 while (n
< length
) {
1676 size_t s
= fread(buffer
+ n
, 1, length
- n
, stdin
);
1677 if (!s
&& feof(stdin
))
1678 die("EOF in data (%lu bytes remaining)",
1679 (unsigned long)(length
- n
));
1689 static int validate_raw_date(const char *src
, char *result
, int maxlen
)
1691 const char *orig_src
= src
;
1694 strtoul(src
, &endp
, 10);
1695 if (endp
== src
|| *endp
!= ' ')
1699 if (*src
!= '-' && *src
!= '+')
1703 strtoul(src
+ 1, &endp
, 10);
1704 if (endp
== src
|| *endp
|| (endp
- orig_src
) >= maxlen
)
1707 strcpy(result
, orig_src
);
1711 static char *parse_ident(const char *buf
)
1717 gt
= strrchr(buf
, '>');
1719 die("Missing > in ident string: %s", buf
);
1722 die("Missing space after > in ident string: %s", buf
);
1724 name_len
= gt
- buf
;
1725 ident
= xmalloc(name_len
+ 24);
1726 strncpy(ident
, buf
, name_len
);
1730 if (validate_raw_date(gt
, ident
+ name_len
, 24) < 0)
1731 die("Invalid raw date \"%s\" in ident: %s", gt
, buf
);
1733 case WHENSPEC_RFC2822
:
1734 if (parse_date(gt
, ident
+ name_len
, 24) < 0)
1735 die("Invalid rfc2822 date \"%s\" in ident: %s", gt
, buf
);
1738 if (strcmp("now", gt
))
1739 die("Date in ident must be 'now': %s", buf
);
1740 datestamp(ident
+ name_len
, 24);
1747 static void cmd_new_blob(void)
1752 read_next_command();
1756 if (store_object(OBJ_BLOB
, d
, l
, &last_blob
, NULL
, next_mark
))
1760 static void unload_one_branch(void)
1762 while (cur_active_branches
1763 && cur_active_branches
>= max_active_branches
) {
1764 uintmax_t min_commit
= ULONG_MAX
;
1765 struct branch
*e
, *l
= NULL
, *p
= NULL
;
1767 for (e
= active_branches
; e
; e
= e
->active_next_branch
) {
1768 if (e
->last_commit
< min_commit
) {
1770 min_commit
= e
->last_commit
;
1776 e
= p
->active_next_branch
;
1777 p
->active_next_branch
= e
->active_next_branch
;
1779 e
= active_branches
;
1780 active_branches
= e
->active_next_branch
;
1783 e
->active_next_branch
= NULL
;
1784 if (e
->branch_tree
.tree
) {
1785 release_tree_content_recursive(e
->branch_tree
.tree
);
1786 e
->branch_tree
.tree
= NULL
;
1788 cur_active_branches
--;
1792 static void load_branch(struct branch
*b
)
1794 load_tree(&b
->branch_tree
);
1797 b
->active_next_branch
= active_branches
;
1798 active_branches
= b
;
1799 cur_active_branches
++;
1800 branch_load_count
++;
1804 static void file_change_m(struct branch
*b
)
1806 const char *p
= command_buf
.buf
+ 2;
1809 struct object_entry
*oe
= oe
;
1810 unsigned char sha1
[20];
1811 uint16_t mode
, inline_data
= 0;
1813 p
= get_mode(p
, &mode
);
1815 die("Corrupt mode: %s", command_buf
.buf
);
1817 case S_IFREG
| 0644:
1818 case S_IFREG
| 0755:
1825 die("Corrupt mode: %s", command_buf
.buf
);
1830 oe
= find_mark(strtoumax(p
+ 1, &x
, 10));
1831 hashcpy(sha1
, oe
->sha1
);
1833 } else if (!prefixcmp(p
, "inline")) {
1837 if (get_sha1_hex(p
, sha1
))
1838 die("Invalid SHA1: %s", command_buf
.buf
);
1839 oe
= find_object(sha1
);
1843 die("Missing space after SHA1: %s", command_buf
.buf
);
1845 p_uq
= unquote_c_style(p
, &endp
);
1848 die("Garbage after path in: %s", command_buf
.buf
);
1856 p
= p_uq
= xstrdup(p
);
1857 read_next_command();
1859 if (store_object(OBJ_BLOB
, d
, l
, &last_blob
, sha1
, 0))
1862 if (oe
->type
!= OBJ_BLOB
)
1863 die("Not a blob (actually a %s): %s",
1864 command_buf
.buf
, typename(oe
->type
));
1866 enum object_type type
= sha1_object_info(sha1
, NULL
);
1868 die("Blob not found: %s", command_buf
.buf
);
1869 if (type
!= OBJ_BLOB
)
1870 die("Not a blob (actually a %s): %s",
1871 typename(type
), command_buf
.buf
);
1874 tree_content_set(&b
->branch_tree
, p
, sha1
, S_IFREG
| mode
, NULL
);
1878 static void file_change_d(struct branch
*b
)
1880 const char *p
= command_buf
.buf
+ 2;
1884 p_uq
= unquote_c_style(p
, &endp
);
1887 die("Garbage after path in: %s", command_buf
.buf
);
1890 tree_content_remove(&b
->branch_tree
, p
, NULL
);
1894 static void file_change_cr(struct branch
*b
, int rename
)
1899 struct tree_entry leaf
;
1901 s
= command_buf
.buf
+ 2;
1902 s_uq
= unquote_c_style(s
, &endp
);
1905 die("Missing space after source: %s", command_buf
.buf
);
1908 endp
= strchr(s
, ' ');
1910 die("Missing space after source: %s", command_buf
.buf
);
1911 s_uq
= xmalloc(endp
- s
+ 1);
1912 memcpy(s_uq
, s
, endp
- s
);
1919 die("Missing dest: %s", command_buf
.buf
);
1922 d_uq
= unquote_c_style(d
, &endp
);
1925 die("Garbage after dest in: %s", command_buf
.buf
);
1929 memset(&leaf
, 0, sizeof(leaf
));
1931 tree_content_remove(&b
->branch_tree
, s
, &leaf
);
1933 tree_content_get(&b
->branch_tree
, s
, &leaf
);
1934 if (!leaf
.versions
[1].mode
)
1935 die("Path %s not in branch", s
);
1936 tree_content_set(&b
->branch_tree
, d
,
1937 leaf
.versions
[1].sha1
,
1938 leaf
.versions
[1].mode
,
1945 static void file_change_deleteall(struct branch
*b
)
1947 release_tree_content_recursive(b
->branch_tree
.tree
);
1948 hashclr(b
->branch_tree
.versions
[0].sha1
);
1949 hashclr(b
->branch_tree
.versions
[1].sha1
);
1950 load_tree(&b
->branch_tree
);
1953 static void cmd_from_commit(struct branch
*b
, char *buf
, unsigned long size
)
1955 if (!buf
|| size
< 46)
1956 die("Not a valid commit: %s", sha1_to_hex(b
->sha1
));
1957 if (memcmp("tree ", buf
, 5)
1958 || get_sha1_hex(buf
+ 5, b
->branch_tree
.versions
[1].sha1
))
1959 die("The commit %s is corrupt", sha1_to_hex(b
->sha1
));
1960 hashcpy(b
->branch_tree
.versions
[0].sha1
,
1961 b
->branch_tree
.versions
[1].sha1
);
1964 static void cmd_from_existing(struct branch
*b
)
1966 if (is_null_sha1(b
->sha1
)) {
1967 hashclr(b
->branch_tree
.versions
[0].sha1
);
1968 hashclr(b
->branch_tree
.versions
[1].sha1
);
1973 buf
= read_object_with_reference(b
->sha1
,
1974 commit_type
, &size
, b
->sha1
);
1975 cmd_from_commit(b
, buf
, size
);
1980 static int cmd_from(struct branch
*b
)
1985 if (prefixcmp(command_buf
.buf
, "from "))
1988 if (b
->branch_tree
.tree
) {
1989 release_tree_content_recursive(b
->branch_tree
.tree
);
1990 b
->branch_tree
.tree
= NULL
;
1993 from
= strchr(command_buf
.buf
, ' ') + 1;
1994 s
= lookup_branch(from
);
1996 die("Can't create a branch from itself: %s", b
->name
);
1998 unsigned char *t
= s
->branch_tree
.versions
[1].sha1
;
1999 hashcpy(b
->sha1
, s
->sha1
);
2000 hashcpy(b
->branch_tree
.versions
[0].sha1
, t
);
2001 hashcpy(b
->branch_tree
.versions
[1].sha1
, t
);
2002 } else if (*from
== ':') {
2003 uintmax_t idnum
= strtoumax(from
+ 1, NULL
, 10);
2004 struct object_entry
*oe
= find_mark(idnum
);
2005 if (oe
->type
!= OBJ_COMMIT
)
2006 die("Mark :%" PRIuMAX
" not a commit", idnum
);
2007 hashcpy(b
->sha1
, oe
->sha1
);
2008 if (oe
->pack_id
!= MAX_PACK_ID
) {
2010 char *buf
= gfi_unpack_entry(oe
, &size
);
2011 cmd_from_commit(b
, buf
, size
);
2014 cmd_from_existing(b
);
2015 } else if (!get_sha1(from
, b
->sha1
))
2016 cmd_from_existing(b
);
2018 die("Invalid ref name or SHA1 expression: %s", from
);
2020 read_next_command();
2024 static struct hash_list
*cmd_merge(unsigned int *count
)
2026 struct hash_list
*list
= NULL
, *n
, *e
= e
;
2031 while (!prefixcmp(command_buf
.buf
, "merge ")) {
2032 from
= strchr(command_buf
.buf
, ' ') + 1;
2033 n
= xmalloc(sizeof(*n
));
2034 s
= lookup_branch(from
);
2036 hashcpy(n
->sha1
, s
->sha1
);
2037 else if (*from
== ':') {
2038 uintmax_t idnum
= strtoumax(from
+ 1, NULL
, 10);
2039 struct object_entry
*oe
= find_mark(idnum
);
2040 if (oe
->type
!= OBJ_COMMIT
)
2041 die("Mark :%" PRIuMAX
" not a commit", idnum
);
2042 hashcpy(n
->sha1
, oe
->sha1
);
2043 } else if (!get_sha1(from
, n
->sha1
)) {
2045 char *buf
= read_object_with_reference(n
->sha1
,
2046 commit_type
, &size
, n
->sha1
);
2047 if (!buf
|| size
< 46)
2048 die("Not a valid commit: %s", from
);
2051 die("Invalid ref name or SHA1 expression: %s", from
);
2060 read_next_command();
2065 static void cmd_new_commit(void)
2071 char *author
= NULL
;
2072 char *committer
= NULL
;
2073 struct hash_list
*merge_list
= NULL
;
2074 unsigned int merge_count
;
2076 /* Obtain the branch name from the rest of our command */
2077 sp
= strchr(command_buf
.buf
, ' ') + 1;
2078 b
= lookup_branch(sp
);
2082 read_next_command();
2084 if (!prefixcmp(command_buf
.buf
, "author ")) {
2085 author
= parse_ident(command_buf
.buf
+ 7);
2086 read_next_command();
2088 if (!prefixcmp(command_buf
.buf
, "committer ")) {
2089 committer
= parse_ident(command_buf
.buf
+ 10);
2090 read_next_command();
2093 die("Expected committer but didn't get one");
2094 msg
= cmd_data(&msglen
);
2095 read_next_command();
2097 merge_list
= cmd_merge(&merge_count
);
2099 /* ensure the branch is active/loaded */
2100 if (!b
->branch_tree
.tree
|| !max_active_branches
) {
2101 unload_one_branch();
2106 while (!command_buf
.eof
&& command_buf
.len
> 1) {
2107 if (!prefixcmp(command_buf
.buf
, "M "))
2109 else if (!prefixcmp(command_buf
.buf
, "D "))
2111 else if (!prefixcmp(command_buf
.buf
, "R "))
2112 file_change_cr(b
, 1);
2113 else if (!prefixcmp(command_buf
.buf
, "C "))
2114 file_change_cr(b
, 0);
2115 else if (!strcmp("deleteall", command_buf
.buf
))
2116 file_change_deleteall(b
);
2118 unread_command_buf
= 1;
2121 read_next_command();
2124 /* build the tree and the commit */
2125 store_tree(&b
->branch_tree
);
2126 hashcpy(b
->branch_tree
.versions
[0].sha1
,
2127 b
->branch_tree
.versions
[1].sha1
);
2128 size_dbuf(&new_data
, 114 + msglen
2131 ? strlen(author
) + strlen(committer
)
2132 : 2 * strlen(committer
)));
2133 sp
= new_data
.buffer
;
2134 sp
+= sprintf(sp
, "tree %s\n",
2135 sha1_to_hex(b
->branch_tree
.versions
[1].sha1
));
2136 if (!is_null_sha1(b
->sha1
))
2137 sp
+= sprintf(sp
, "parent %s\n", sha1_to_hex(b
->sha1
));
2138 while (merge_list
) {
2139 struct hash_list
*next
= merge_list
->next
;
2140 sp
+= sprintf(sp
, "parent %s\n", sha1_to_hex(merge_list
->sha1
));
2144 sp
+= sprintf(sp
, "author %s\n", author
? author
: committer
);
2145 sp
+= sprintf(sp
, "committer %s\n", committer
);
2147 memcpy(sp
, msg
, msglen
);
2153 if (!store_object(OBJ_COMMIT
,
2154 new_data
.buffer
, sp
- (char*)new_data
.buffer
,
2155 NULL
, b
->sha1
, next_mark
))
2156 b
->pack_id
= pack_id
;
2157 b
->last_commit
= object_count_by_type
[OBJ_COMMIT
];
2160 static void cmd_new_tag(void)
2169 uintmax_t from_mark
= 0;
2170 unsigned char sha1
[20];
2172 /* Obtain the new tag name from the rest of our command */
2173 sp
= strchr(command_buf
.buf
, ' ') + 1;
2174 t
= pool_alloc(sizeof(struct tag
));
2176 t
->name
= pool_strdup(sp
);
2178 last_tag
->next_tag
= t
;
2182 read_next_command();
2185 if (prefixcmp(command_buf
.buf
, "from "))
2186 die("Expected from command, got %s", command_buf
.buf
);
2187 from
= strchr(command_buf
.buf
, ' ') + 1;
2188 s
= lookup_branch(from
);
2190 hashcpy(sha1
, s
->sha1
);
2191 } else if (*from
== ':') {
2192 struct object_entry
*oe
;
2193 from_mark
= strtoumax(from
+ 1, NULL
, 10);
2194 oe
= find_mark(from_mark
);
2195 if (oe
->type
!= OBJ_COMMIT
)
2196 die("Mark :%" PRIuMAX
" not a commit", from_mark
);
2197 hashcpy(sha1
, oe
->sha1
);
2198 } else if (!get_sha1(from
, sha1
)) {
2202 buf
= read_object_with_reference(sha1
,
2203 commit_type
, &size
, sha1
);
2204 if (!buf
|| size
< 46)
2205 die("Not a valid commit: %s", from
);
2208 die("Invalid ref name or SHA1 expression: %s", from
);
2209 read_next_command();
2212 if (prefixcmp(command_buf
.buf
, "tagger "))
2213 die("Expected tagger command, got %s", command_buf
.buf
);
2214 tagger
= parse_ident(command_buf
.buf
+ 7);
2216 /* tag payload/message */
2217 read_next_command();
2218 msg
= cmd_data(&msglen
);
2220 /* build the tag object */
2221 size_dbuf(&new_data
, 67+strlen(t
->name
)+strlen(tagger
)+msglen
);
2222 sp
= new_data
.buffer
;
2223 sp
+= sprintf(sp
, "object %s\n", sha1_to_hex(sha1
));
2224 sp
+= sprintf(sp
, "type %s\n", commit_type
);
2225 sp
+= sprintf(sp
, "tag %s\n", t
->name
);
2226 sp
+= sprintf(sp
, "tagger %s\n", tagger
);
2228 memcpy(sp
, msg
, msglen
);
2233 if (store_object(OBJ_TAG
, new_data
.buffer
,
2234 sp
- (char*)new_data
.buffer
,
2236 t
->pack_id
= MAX_PACK_ID
;
2238 t
->pack_id
= pack_id
;
2241 static void cmd_reset_branch(void)
2246 /* Obtain the branch name from the rest of our command */
2247 sp
= strchr(command_buf
.buf
, ' ') + 1;
2248 b
= lookup_branch(sp
);
2251 hashclr(b
->branch_tree
.versions
[0].sha1
);
2252 hashclr(b
->branch_tree
.versions
[1].sha1
);
2253 if (b
->branch_tree
.tree
) {
2254 release_tree_content_recursive(b
->branch_tree
.tree
);
2255 b
->branch_tree
.tree
= NULL
;
2260 read_next_command();
2261 if (!cmd_from(b
) && command_buf
.len
> 1)
2262 unread_command_buf
= 1;
2265 static void cmd_checkpoint(void)
2276 static void cmd_progress(void)
2278 fwrite(command_buf
.buf
, 1, command_buf
.len
- 1, stdout
);
2279 fputc('\n', stdout
);
2284 static void import_marks(const char *input_file
)
2287 FILE *f
= fopen(input_file
, "r");
2289 die("cannot read %s: %s", input_file
, strerror(errno
));
2290 while (fgets(line
, sizeof(line
), f
)) {
2293 unsigned char sha1
[20];
2294 struct object_entry
*e
;
2296 end
= strchr(line
, '\n');
2297 if (line
[0] != ':' || !end
)
2298 die("corrupt mark line: %s", line
);
2300 mark
= strtoumax(line
+ 1, &end
, 10);
2301 if (!mark
|| end
== line
+ 1
2302 || *end
!= ' ' || get_sha1(end
+ 1, sha1
))
2303 die("corrupt mark line: %s", line
);
2304 e
= find_object(sha1
);
2306 enum object_type type
= sha1_object_info(sha1
, NULL
);
2308 die("object not found: %s", sha1_to_hex(sha1
));
2309 e
= insert_object(sha1
);
2311 e
->pack_id
= MAX_PACK_ID
;
2312 e
->offset
= 1; /* just not zero! */
2314 insert_mark(mark
, e
);
2319 static const char fast_import_usage
[] =
2320 "git-fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2322 int main(int argc
, const char **argv
)
2324 unsigned int i
, show_stats
= 1;
2326 git_config(git_default_config
);
2327 alloc_objects(object_entry_alloc
);
2328 strbuf_init(&command_buf
);
2329 atom_table
= xcalloc(atom_table_sz
, sizeof(struct atom_str
*));
2330 branch_table
= xcalloc(branch_table_sz
, sizeof(struct branch
*));
2331 avail_tree_table
= xcalloc(avail_tree_table_sz
, sizeof(struct avail_tree_content
*));
2332 marks
= pool_calloc(1, sizeof(struct mark_set
));
2334 for (i
= 1; i
< argc
; i
++) {
2335 const char *a
= argv
[i
];
2337 if (*a
!= '-' || !strcmp(a
, "--"))
2339 else if (!prefixcmp(a
, "--date-format=")) {
2340 const char *fmt
= a
+ 14;
2341 if (!strcmp(fmt
, "raw"))
2342 whenspec
= WHENSPEC_RAW
;
2343 else if (!strcmp(fmt
, "rfc2822"))
2344 whenspec
= WHENSPEC_RFC2822
;
2345 else if (!strcmp(fmt
, "now"))
2346 whenspec
= WHENSPEC_NOW
;
2348 die("unknown --date-format argument %s", fmt
);
2350 else if (!prefixcmp(a
, "--max-pack-size="))
2351 max_packsize
= strtoumax(a
+ 16, NULL
, 0) * 1024 * 1024;
2352 else if (!prefixcmp(a
, "--depth="))
2353 max_depth
= strtoul(a
+ 8, NULL
, 0);
2354 else if (!prefixcmp(a
, "--active-branches="))
2355 max_active_branches
= strtoul(a
+ 18, NULL
, 0);
2356 else if (!prefixcmp(a
, "--import-marks="))
2357 import_marks(a
+ 15);
2358 else if (!prefixcmp(a
, "--export-marks="))
2360 else if (!prefixcmp(a
, "--export-pack-edges=")) {
2363 pack_edges
= fopen(a
+ 20, "a");
2365 die("Cannot open %s: %s", a
+ 20, strerror(errno
));
2366 } else if (!strcmp(a
, "--force"))
2368 else if (!strcmp(a
, "--quiet"))
2370 else if (!strcmp(a
, "--stats"))
2373 die("unknown option %s", a
);
2376 usage(fast_import_usage
);
2378 rc_free
= pool_alloc(cmd_save
* sizeof(*rc_free
));
2379 for (i
= 0; i
< (cmd_save
- 1); i
++)
2380 rc_free
[i
].next
= &rc_free
[i
+ 1];
2381 rc_free
[cmd_save
- 1].next
= NULL
;
2383 prepare_packed_git();
2385 set_die_routine(die_nicely
);
2387 read_next_command();
2388 if (command_buf
.eof
)
2390 else if (!strcmp("blob", command_buf
.buf
))
2392 else if (!prefixcmp(command_buf
.buf
, "commit "))
2394 else if (!prefixcmp(command_buf
.buf
, "tag "))
2396 else if (!prefixcmp(command_buf
.buf
, "reset "))
2398 else if (!strcmp("checkpoint", command_buf
.buf
))
2400 else if (!prefixcmp(command_buf
.buf
, "progress "))
2403 die("Unsupported command: %s", command_buf
.buf
);
2416 uintmax_t total_count
= 0, duplicate_count
= 0;
2417 for (i
= 0; i
< ARRAY_SIZE(object_count_by_type
); i
++)
2418 total_count
+= object_count_by_type
[i
];
2419 for (i
= 0; i
< ARRAY_SIZE(duplicate_count_by_type
); i
++)
2420 duplicate_count
+= duplicate_count_by_type
[i
];
2422 fprintf(stderr
, "%s statistics:\n", argv
[0]);
2423 fprintf(stderr
, "---------------------------------------------------------------------\n");
2424 fprintf(stderr
, "Alloc'd objects: %10" PRIuMAX
"\n", alloc_count
);
2425 fprintf(stderr
, "Total objects: %10" PRIuMAX
" (%10" PRIuMAX
" duplicates )\n", total_count
, duplicate_count
);
2426 fprintf(stderr
, " blobs : %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas)\n", object_count_by_type
[OBJ_BLOB
], duplicate_count_by_type
[OBJ_BLOB
], delta_count_by_type
[OBJ_BLOB
]);
2427 fprintf(stderr
, " trees : %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas)\n", object_count_by_type
[OBJ_TREE
], duplicate_count_by_type
[OBJ_TREE
], delta_count_by_type
[OBJ_TREE
]);
2428 fprintf(stderr
, " commits: %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas)\n", object_count_by_type
[OBJ_COMMIT
], duplicate_count_by_type
[OBJ_COMMIT
], delta_count_by_type
[OBJ_COMMIT
]);
2429 fprintf(stderr
, " tags : %10" PRIuMAX
" (%10" PRIuMAX
" duplicates %10" PRIuMAX
" deltas)\n", object_count_by_type
[OBJ_TAG
], duplicate_count_by_type
[OBJ_TAG
], delta_count_by_type
[OBJ_TAG
]);
2430 fprintf(stderr
, "Total branches: %10lu (%10lu loads )\n", branch_count
, branch_load_count
);
2431 fprintf(stderr
, " marks: %10" PRIuMAX
" (%10" PRIuMAX
" unique )\n", (((uintmax_t)1) << marks
->shift
) * 1024, marks_set_count
);
2432 fprintf(stderr
, " atoms: %10u\n", atom_cnt
);
2433 fprintf(stderr
, "Memory total: %10" PRIuMAX
" KiB\n", (total_allocd
+ alloc_count
*sizeof(struct object_entry
))/1024);
2434 fprintf(stderr
, " pools: %10lu KiB\n", (unsigned long)(total_allocd
/1024));
2435 fprintf(stderr
, " objects: %10" PRIuMAX
" KiB\n", (alloc_count
*sizeof(struct object_entry
))/1024);
2436 fprintf(stderr
, "---------------------------------------------------------------------\n");
2438 fprintf(stderr
, "---------------------------------------------------------------------\n");
2439 fprintf(stderr
, "\n");
2442 return failure
? 1 : 0;