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;
530 e = avail_tree_entry;
531 avail_tree_entry = *((void**)e);
535 static void release_tree_entry(struct tree_entry *e)
538 release_tree_content_recursive(e->tree);
539 *((void**)e) = avail_tree_entry;
540 avail_tree_entry = e;
543 static void yread(int fd, void *buffer, size_t length)
546 while (ret < length) {
547 ssize_t size = xread(fd, (char *) buffer + ret, length - ret);
549 die("Read from descriptor %i: end of stream", fd);
551 die("Read from descriptor %i: %s", fd, strerror(errno));
556 static void ywrite(int fd, void *buffer, size_t length)
559 while (ret < length) {
560 ssize_t size = xwrite(fd, (char *) buffer + ret, length - ret);
562 die("Write to descriptor %i: end of file", fd);
564 die("Write to descriptor %i: %s", fd, strerror(errno));
569 static size_t encode_header(
570 enum object_type type,
577 if (type < OBJ_COMMIT || type > OBJ_DELTA)
578 die("bad type %d", type);
580 c = (type << 4) | (size & 15);
592 static int store_object(
593 enum object_type type,
596 struct last_object *last,
597 unsigned char *sha1out,
601 struct object_entry *e;
602 unsigned char hdr[96];
603 unsigned char sha1[20];
604 unsigned long hdrlen, deltalen;
608 hdrlen = sprintf((char*)hdr,"%s %lu",type_names[type],datlen) + 1;
610 SHA1_Update(&c, hdr, hdrlen);
611 SHA1_Update(&c, dat, datlen);
612 SHA1_Final(sha1, &c);
614 memcpy(sha1out, sha1, sizeof(sha1));
616 e = insert_object(sha1);
618 insert_mark(mark, e);
621 duplicate_count_by_type[type]++;
625 e->offset = pack_size;
627 object_count_by_type[type]++;
629 if (last && last->data && last->depth < max_depth)
630 delta = diff_delta(last->data, last->len,
636 memset(&s, 0, sizeof(s));
637 deflateInit(&s, zlib_compression_level);
642 s.avail_in = deltalen;
643 hdrlen = encode_header(OBJ_DELTA, deltalen, hdr);
644 ywrite(pack_fd, hdr, hdrlen);
645 ywrite(pack_fd, last->sha1, sizeof(sha1));
646 pack_size += hdrlen + sizeof(sha1);
652 hdrlen = encode_header(type, datlen, hdr);
653 ywrite(pack_fd, hdr, hdrlen);
657 s.avail_out = deflateBound(&s, s.avail_in);
658 s.next_out = out = xmalloc(s.avail_out);
659 while (deflate(&s, Z_FINISH) == Z_OK)
663 ywrite(pack_fd, out, s.total_out);
664 pack_size += s.total_out;
674 memcpy(last->sha1, sha1, sizeof(sha1));
679 static void* map_pack(unsigned long offset)
681 if (offset >= pack_size)
682 die("object offset outside of pack file");
683 if (offset >= pack_mlen) {
685 munmap(pack_base, pack_mlen);
686 /* round out how much we map to 16 MB units */
687 pack_mlen = pack_size;
688 if (pack_mlen & ((1 << 24) - 1))
689 pack_mlen = ((pack_mlen >> 24) + 1) << 24;
690 pack_base = mmap(NULL,pack_mlen,PROT_READ,MAP_SHARED,pack_fd,0);
691 if (pack_base == MAP_FAILED)
692 die("Failed to map generated pack: %s", strerror(errno));
695 return (char*)pack_base + offset;
698 static unsigned long unpack_object_header(unsigned long offset,
699 enum object_type *type,
700 unsigned long *sizep)
706 c = *(unsigned char*)map_pack(offset++);
707 *type = (c >> 4) & 7;
711 c = *(unsigned char*)map_pack(offset++);
712 size += (c & 0x7f) << shift;
719 static void *unpack_non_delta_entry(unsigned long o, unsigned long sz)
722 unsigned char *result;
724 result = xmalloc(sz + 1);
727 memset(&stream, 0, sizeof(stream));
728 stream.next_in = map_pack(o);
729 stream.avail_in = pack_mlen - o;
730 stream.next_out = result;
731 stream.avail_out = sz;
733 inflateInit(&stream);
735 int st = inflate(&stream, Z_FINISH);
736 if (st == Z_STREAM_END)
739 o = stream.next_in - (unsigned char*)pack_base;
740 stream.next_in = map_pack(o);
741 stream.avail_in = pack_mlen - o;
744 die("Error from zlib during inflate.");
747 if (stream.total_out != sz)
748 die("Error after inflate: sizes mismatch");
752 static void *unpack_entry(unsigned long offset, unsigned long *sizep);
754 static void *unpack_delta_entry(unsigned long offset,
755 unsigned long delta_size,
756 unsigned long *sizep)
758 struct object_entry *base_oe;
759 unsigned char *base_sha1;
760 void *delta_data, *base, *result;
761 unsigned long base_size, result_size;
763 base_sha1 = (unsigned char*)map_pack(offset + 20) - 20;
764 base_oe = find_object(base_sha1);
766 die("I'm broken; I can't find a base I know must be here.");
767 base = unpack_entry(base_oe->offset, &base_size);
768 delta_data = unpack_non_delta_entry(offset + 20, delta_size);
769 result = patch_delta(base, base_size,
770 delta_data, delta_size,
773 die("failed to apply delta");
776 *sizep = result_size;
780 static void *unpack_entry(unsigned long offset, unsigned long *sizep)
783 enum object_type kind;
785 offset = unpack_object_header(offset, &kind, &size);
788 return unpack_delta_entry(offset, size, sizep);
794 return unpack_non_delta_entry(offset, size);
796 die("I created an object I can't read!");
800 static const char *get_mode(const char *str, unsigned int *modep)
803 unsigned int mode = 0;
805 while ((c = *str++) != ' ') {
806 if (c < '0' || c > '7')
808 mode = (mode << 3) + (c - '0');
814 static void load_tree(struct tree_entry *root)
816 struct object_entry *myoe;
817 struct tree_content *t;
822 root->tree = t = new_tree_content(8);
823 if (!memcmp(root->sha1, null_sha1, 20))
826 myoe = find_object(root->sha1);
828 if (myoe->type != OBJ_TREE)
829 die("Not a tree: %s", sha1_to_hex(root->sha1));
830 buf = unpack_entry(myoe->offset, &size);
833 buf = read_sha1_file(root->sha1, type, &size);
834 if (!buf || strcmp(type, tree_type))
835 die("Can't load tree %s", sha1_to_hex(root->sha1));
839 while (c != (buf + size)) {
840 struct tree_entry *e = new_tree_entry();
842 if (t->entry_count == t->entry_capacity)
843 root->tree = t = grow_tree_content(t, 8);
844 t->entries[t->entry_count++] = e;
847 c = get_mode(c, &e->mode);
849 die("Corrupt mode in %s", sha1_to_hex(root->sha1));
850 e->name = to_atom(c, strlen(c));
851 c += e->name->str_len + 1;
852 memcpy(e->sha1, c, sizeof(e->sha1));
858 static int tecmp (const void *_a, const void *_b)
860 struct tree_entry *a = *((struct tree_entry**)_a);
861 struct tree_entry *b = *((struct tree_entry**)_b);
862 return base_name_compare(
863 a->name->str_dat, a->name->str_len, a->mode,
864 b->name->str_dat, b->name->str_len, b->mode);
867 static void store_tree(struct tree_entry *root)
869 struct tree_content *t = root->tree;
874 if (memcmp(root->sha1, null_sha1, 20))
878 for (i = 0; i < t->entry_count; i++) {
879 maxlen += t->entries[i]->name->str_len + 34;
880 if (t->entries[i]->tree)
881 store_tree(t->entries[i]);
884 qsort(t->entries, t->entry_count, sizeof(t->entries[0]), tecmp);
885 buf = c = xmalloc(maxlen);
886 for (i = 0; i < t->entry_count; i++) {
887 struct tree_entry *e = t->entries[i];
888 c += sprintf(c, "%o", e->mode);
890 strcpy(c, e->name->str_dat);
891 c += e->name->str_len + 1;
892 memcpy(c, e->sha1, 20);
895 store_object(OBJ_TREE, buf, c - buf, NULL, root->sha1, 0);
899 static int tree_content_set(
900 struct tree_entry *root,
902 const unsigned char *sha1,
903 const unsigned int mode)
905 struct tree_content *t = root->tree;
908 struct tree_entry *e;
910 slash1 = strchr(p, '/');
916 for (i = 0; i < t->entry_count; i++) {
918 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
920 if (e->mode == mode && !memcmp(e->sha1, sha1, 20))
923 memcpy(e->sha1, sha1, 20);
925 release_tree_content_recursive(e->tree);
928 memcpy(root->sha1, null_sha1, 20);
931 if (!S_ISDIR(e->mode)) {
932 e->tree = new_tree_content(8);
937 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
938 memcpy(root->sha1, null_sha1, 20);
945 if (t->entry_count == t->entry_capacity)
946 root->tree = t = grow_tree_content(t, 8);
947 e = new_tree_entry();
948 e->name = to_atom(p, n);
949 t->entries[t->entry_count++] = e;
951 e->tree = new_tree_content(8);
953 tree_content_set(e, slash1 + 1, sha1, mode);
957 memcpy(e->sha1, sha1, 20);
959 memcpy(root->sha1, null_sha1, 20);
963 static int tree_content_remove(struct tree_entry *root, const char *p)
965 struct tree_content *t = root->tree;
968 struct tree_entry *e;
970 slash1 = strchr(p, '/');
976 for (i = 0; i < t->entry_count; i++) {
978 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
979 if (!slash1 || !S_ISDIR(e->mode))
983 if (tree_content_remove(e, slash1 + 1)) {
984 if (!e->tree->entry_count)
986 memcpy(root->sha1, null_sha1, 20);
995 for (i++; i < t->entry_count; i++)
996 t->entries[i-1] = t->entries[i];
998 release_tree_entry(e);
999 memcpy(root->sha1, null_sha1, 20);
1003 static void init_pack_header()
1005 struct pack_header hdr;
1007 hdr.hdr_signature = htonl(PACK_SIGNATURE);
1008 hdr.hdr_version = htonl(2);
1009 hdr.hdr_entries = 0;
1011 ywrite(pack_fd, &hdr, sizeof(hdr));
1012 pack_size = sizeof(hdr);
1015 static void fixup_header_footer()
1023 if (lseek(pack_fd, 0, SEEK_SET) != 0)
1024 die("Failed seeking to start: %s", strerror(errno));
1027 yread(pack_fd, hdr, 8);
1028 SHA1_Update(&c, hdr, 8);
1030 cnt = htonl(object_count);
1031 SHA1_Update(&c, &cnt, 4);
1032 ywrite(pack_fd, &cnt, 4);
1034 buf = xmalloc(128 * 1024);
1036 n = xread(pack_fd, buf, 128 * 1024);
1039 SHA1_Update(&c, buf, n);
1043 SHA1_Final(pack_sha1, &c);
1044 ywrite(pack_fd, pack_sha1, sizeof(pack_sha1));
1047 static int oecmp (const void *_a, const void *_b)
1049 struct object_entry *a = *((struct object_entry**)_a);
1050 struct object_entry *b = *((struct object_entry**)_b);
1051 return memcmp(a->sha1, b->sha1, sizeof(a->sha1));
1054 static void write_index(const char *idx_name)
1057 struct object_entry **idx, **c, **last;
1058 struct object_entry *e;
1059 struct object_entry_pool *o;
1060 unsigned int array[256];
1063 /* Build the sorted table of object IDs. */
1064 idx = xmalloc(object_count * sizeof(struct object_entry*));
1066 for (o = blocks; o; o = o->next_pool)
1067 for (e = o->entries; e != o->next_free; e++)
1069 last = idx + object_count;
1070 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
1072 /* Generate the fan-out array. */
1074 for (i = 0; i < 256; i++) {
1075 struct object_entry **next = c;;
1076 while (next < last) {
1077 if ((*next)->sha1[0] != i)
1081 array[i] = htonl(next - idx);
1085 f = sha1create("%s", idx_name);
1086 sha1write(f, array, 256 * sizeof(int));
1087 for (c = idx; c != last; c++) {
1088 unsigned int offset = htonl((*c)->offset);
1089 sha1write(f, &offset, 4);
1090 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
1092 sha1write(f, pack_sha1, sizeof(pack_sha1));
1093 sha1close(f, NULL, 1);
1097 static void dump_branches()
1099 static const char *msg = "fast-import";
1102 struct ref_lock *lock;
1104 for (i = 0; i < branch_table_sz; i++) {
1105 for (b = branch_table[i]; b; b = b->table_next_branch) {
1106 lock = lock_any_ref_for_update(b->name, NULL, 0);
1107 if (!lock || write_ref_sha1(lock, b->sha1, msg) < 0)
1108 die("Can't write %s", b->name);
1113 static void dump_tags()
1115 static const char *msg = "fast-import";
1117 struct ref_lock *lock;
1118 char path[PATH_MAX];
1120 for (t = first_tag; t; t = t->next_tag) {
1121 sprintf(path, "refs/tags/%s", t->name);
1122 lock = lock_any_ref_for_update(path, NULL, 0);
1123 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1124 die("Can't write %s", path);
1128 static void dump_marks_helper(FILE *f,
1134 for (k = 0; k < 1024; k++) {
1135 if (m->data.sets[k])
1136 dump_marks_helper(f, (base + k) << m->shift,
1140 for (k = 0; k < 1024; k++) {
1141 if (m->data.marked[k])
1142 fprintf(f, ":%lu %s\n", base + k,
1143 sha1_to_hex(m->data.marked[k]->sha1));
1148 static void dump_marks()
1152 FILE *f = fopen(mark_file, "w");
1153 dump_marks_helper(f, 0, marks);
1158 static void read_next_command()
1160 read_line(&command_buf, stdin, '\n');
1163 static void cmd_mark()
1165 if (!strncmp("mark :", command_buf.buf, 6)) {
1166 next_mark = strtoul(command_buf.buf + 6, NULL, 10);
1167 read_next_command();
1173 static void* cmd_data (size_t *size)
1179 if (strncmp("data ", command_buf.buf, 5))
1180 die("Expected 'data n' command, found: %s", command_buf.buf);
1182 length = strtoul(command_buf.buf + 5, NULL, 10);
1183 buffer = xmalloc(length);
1185 while (n < length) {
1186 size_t s = fread((char*)buffer + n, 1, length - n, stdin);
1187 if (!s && feof(stdin))
1188 die("EOF in data (%lu bytes remaining)", length - n);
1192 if (fgetc(stdin) != '\n')
1193 die("An lf did not trail the binary data as expected.");
1199 static void cmd_new_blob()
1204 read_next_command();
1208 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1212 static void unload_one_branch()
1214 while (cur_active_branches
1215 && cur_active_branches >= max_active_branches) {
1216 unsigned long min_commit = ULONG_MAX;
1217 struct branch *e, *l = NULL, *p = NULL;
1219 for (e = active_branches; e; e = e->active_next_branch) {
1220 if (e->last_commit < min_commit) {
1222 min_commit = e->last_commit;
1228 e = p->active_next_branch;
1229 p->active_next_branch = e->active_next_branch;
1231 e = active_branches;
1232 active_branches = e->active_next_branch;
1234 e->active_next_branch = NULL;
1235 if (e->branch_tree.tree) {
1236 release_tree_content_recursive(e->branch_tree.tree);
1237 e->branch_tree.tree = NULL;
1239 cur_active_branches--;
1243 static void load_branch(struct branch *b)
1245 load_tree(&b->branch_tree);
1246 b->active_next_branch = active_branches;
1247 active_branches = b;
1248 cur_active_branches++;
1249 branch_load_count++;
1252 static void file_change_m(struct branch *b)
1254 const char *p = command_buf.buf + 2;
1257 struct object_entry *oe;
1258 unsigned char sha1[20];
1262 p = get_mode(p, &mode);
1264 die("Corrupt mode: %s", command_buf.buf);
1266 case S_IFREG | 0644:
1267 case S_IFREG | 0755:
1274 die("Corrupt mode: %s", command_buf.buf);
1279 oe = find_mark(strtoul(p + 1, &x, 10));
1282 if (get_sha1_hex(p, sha1))
1283 die("Invalid SHA1: %s", command_buf.buf);
1284 oe = find_object(sha1);
1288 die("Missing space after SHA1: %s", command_buf.buf);
1290 p_uq = unquote_c_style(p, &endp);
1293 die("Garbage after path in: %s", command_buf.buf);
1298 if (oe->type != OBJ_BLOB)
1299 die("Not a blob (actually a %s): %s",
1300 command_buf.buf, type_names[oe->type]);
1302 if (sha1_object_info(sha1, type, NULL))
1303 die("Blob not found: %s", command_buf.buf);
1304 if (strcmp(blob_type, type))
1305 die("Not a blob (actually a %s): %s",
1306 command_buf.buf, type);
1309 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1315 static void file_change_d(struct branch *b)
1317 const char *p = command_buf.buf + 2;
1321 p_uq = unquote_c_style(p, &endp);
1324 die("Garbage after path in: %s", command_buf.buf);
1327 tree_content_remove(&b->branch_tree, p);
1332 static void cmd_from(struct branch *b)
1334 const char *from, *endp;
1338 if (strncmp("from ", command_buf.buf, 5))
1342 die("Can't reinitailize branch %s", b->name);
1344 from = strchr(command_buf.buf, ' ') + 1;
1345 str_uq = unquote_c_style(from, &endp);
1348 die("Garbage after string in: %s", command_buf.buf);
1352 s = lookup_branch(from);
1354 die("Can't create a branch from itself: %s", b->name);
1356 memcpy(b->sha1, s->sha1, 20);
1357 memcpy(b->branch_tree.sha1, s->branch_tree.sha1, 20);
1358 } else if (*from == ':') {
1359 unsigned long idnum = strtoul(from + 1, NULL, 10);
1360 struct object_entry *oe = find_mark(idnum);
1363 if (oe->type != OBJ_COMMIT)
1364 die("Mark :%lu not a commit", idnum);
1365 memcpy(b->sha1, oe->sha1, 20);
1366 buf = unpack_entry(oe->offset, &size);
1367 if (!buf || size < 46)
1368 die("Not a valid commit: %s", from);
1369 if (memcmp("tree ", buf, 5)
1370 || get_sha1_hex(buf + 5, b->branch_tree.sha1))
1371 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1373 } else if (!get_sha1(from, b->sha1)) {
1374 if (!memcmp(b->sha1, null_sha1, 20))
1375 memcpy(b->branch_tree.sha1, null_sha1, 20);
1380 buf = read_object_with_reference(b->sha1,
1381 type_names[OBJ_COMMIT], &size, b->sha1);
1382 if (!buf || size < 46)
1383 die("Not a valid commit: %s", from);
1384 if (memcmp("tree ", buf, 5)
1385 || get_sha1_hex(buf + 5, b->branch_tree.sha1))
1386 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1390 die("Invalid ref name or SHA1 expression: %s", from);
1392 read_next_command();
1395 static void cmd_new_commit()
1403 char *author = NULL;
1404 char *committer = NULL;
1407 /* Obtain the branch name from the rest of our command */
1408 sp = strchr(command_buf.buf, ' ') + 1;
1409 str_uq = unquote_c_style(sp, &endp);
1412 die("Garbage after ref in: %s", command_buf.buf);
1415 b = lookup_branch(sp);
1421 read_next_command();
1423 if (!strncmp("author ", command_buf.buf, 7)) {
1424 author = strdup(command_buf.buf);
1425 read_next_command();
1427 if (!strncmp("committer ", command_buf.buf, 10)) {
1428 committer = strdup(command_buf.buf);
1429 read_next_command();
1432 die("Expected committer but didn't get one");
1433 msg = cmd_data(&msglen);
1434 read_next_command();
1437 /* ensure the branch is active/loaded */
1438 if (!b->branch_tree.tree || !max_active_branches) {
1439 unload_one_branch();
1445 if (1 == command_buf.len)
1447 else if (!strncmp("M ", command_buf.buf, 2))
1449 else if (!strncmp("D ", command_buf.buf, 2))
1452 die("Unsupported file_change: %s", command_buf.buf);
1453 read_next_command();
1456 /* build the tree and the commit */
1457 store_tree(&b->branch_tree);
1458 body = xmalloc(97 + msglen
1460 ? strlen(author) + strlen(committer)
1461 : 2 * strlen(committer)));
1463 sp += sprintf(sp, "tree %s\n", sha1_to_hex(b->branch_tree.sha1));
1464 if (memcmp(b->sha1, null_sha1, 20))
1465 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1467 sp += sprintf(sp, "%s\n", author);
1469 sp += sprintf(sp, "author %s\n", committer + 10);
1470 sp += sprintf(sp, "%s\n\n", committer);
1471 memcpy(sp, msg, msglen);
1478 store_object(OBJ_COMMIT, body, sp - body, NULL, b->sha1, next_mark);
1480 b->last_commit = object_count_by_type[OBJ_COMMIT];
1483 int need_dq = quote_c_style(b->name, NULL, NULL, 0);
1484 fprintf(branch_log, "commit ");
1486 fputc('"', branch_log);
1487 quote_c_style(b->name, NULL, branch_log, 0);
1488 fputc('"', branch_log);
1490 fprintf(branch_log, "%s", b->name);
1491 fprintf(branch_log," :%lu %s\n",next_mark,sha1_to_hex(b->sha1));
1495 static void cmd_new_tag()
1507 unsigned long from_mark = 0;
1508 unsigned char sha1[20];
1510 /* Obtain the new tag name from the rest of our command */
1511 sp = strchr(command_buf.buf, ' ') + 1;
1512 str_uq = unquote_c_style(sp, &endp);
1515 die("Garbage after tag name in: %s", command_buf.buf);
1518 t = pool_alloc(sizeof(struct tag));
1520 t->name = pool_strdup(sp);
1522 last_tag->next_tag = t;
1528 read_next_command();
1531 if (strncmp("from ", command_buf.buf, 5))
1532 die("Expected from command, got %s", command_buf.buf);
1534 from = strchr(command_buf.buf, ' ') + 1;
1535 str_uq = unquote_c_style(from, &endp);
1538 die("Garbage after string in: %s", command_buf.buf);
1542 s = lookup_branch(from);
1544 memcpy(sha1, s->sha1, 20);
1545 } else if (*from == ':') {
1546 from_mark = strtoul(from + 1, NULL, 10);
1547 struct object_entry *oe = find_mark(from_mark);
1548 if (oe->type != OBJ_COMMIT)
1549 die("Mark :%lu not a commit", from_mark);
1550 memcpy(sha1, oe->sha1, 20);
1551 } else if (!get_sha1(from, sha1)) {
1555 buf = read_object_with_reference(sha1,
1556 type_names[OBJ_COMMIT], &size, sha1);
1557 if (!buf || size < 46)
1558 die("Not a valid commit: %s", from);
1561 die("Invalid ref name or SHA1 expression: %s", from);
1565 read_next_command();
1568 if (strncmp("tagger ", command_buf.buf, 7))
1569 die("Expected tagger command, got %s", command_buf.buf);
1570 tagger = strdup(command_buf.buf);
1572 /* tag payload/message */
1573 read_next_command();
1574 msg = cmd_data(&msglen);
1576 /* build the tag object */
1577 body = xmalloc(67 + strlen(t->name) + strlen(tagger) + msglen);
1579 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
1580 sp += sprintf(sp, "type %s\n", type_names[OBJ_COMMIT]);
1581 sp += sprintf(sp, "tag %s\n", t->name);
1582 sp += sprintf(sp, "%s\n\n", tagger);
1583 memcpy(sp, msg, msglen);
1588 store_object(OBJ_TAG, body, sp - body, NULL, t->sha1, 0);
1592 int need_dq = quote_c_style(t->name, NULL, NULL, 0);
1593 fprintf(branch_log, "tag ");
1595 fputc('"', branch_log);
1596 quote_c_style(t->name, NULL, branch_log, 0);
1597 fputc('"', branch_log);
1599 fprintf(branch_log, "%s", t->name);
1600 fprintf(branch_log," :%lu %s\n",from_mark,sha1_to_hex(t->sha1));
1604 static const char fast_import_usage[] =
1605 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file] [--branch-log=log] temp.pack";
1607 int main(int argc, const char **argv)
1609 const char *base_name;
1611 unsigned long est_obj_cnt = 1000;
1617 git_config(git_default_config);
1619 for (i = 1; i < argc; i++) {
1620 const char *a = argv[i];
1622 if (*a != '-' || !strcmp(a, "--"))
1624 else if (!strncmp(a, "--objects=", 10))
1625 est_obj_cnt = strtoul(a + 10, NULL, 0);
1626 else if (!strncmp(a, "--depth=", 8))
1627 max_depth = strtoul(a + 8, NULL, 0);
1628 else if (!strncmp(a, "--active-branches=", 18))
1629 max_active_branches = strtoul(a + 18, NULL, 0);
1630 else if (!strncmp(a, "--export-marks=", 15))
1632 else if (!strncmp(a, "--branch-log=", 13)) {
1633 branch_log = fopen(a + 13, "w");
1635 die("Can't create %s: %s", a + 13, strerror(errno));
1638 die("unknown option %s", a);
1641 usage(fast_import_usage);
1642 base_name = argv[i];
1644 pack_name = xmalloc(strlen(base_name) + 6);
1645 sprintf(pack_name, "%s.pack", base_name);
1646 idx_name = xmalloc(strlen(base_name) + 5);
1647 sprintf(idx_name, "%s.idx", base_name);
1649 pack_fd = open(pack_name, O_RDWR|O_CREAT|O_EXCL, 0666);
1651 die("Can't create %s: %s", pack_name, strerror(errno));
1654 alloc_objects(est_obj_cnt);
1655 strbuf_init(&command_buf);
1657 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
1658 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
1659 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
1660 marks = pool_calloc(1, sizeof(struct mark_set));
1663 read_next_command();
1664 if (command_buf.eof)
1666 else if (!strcmp("blob", command_buf.buf))
1668 else if (!strncmp("commit ", command_buf.buf, 7))
1670 else if (!strncmp("tag ", command_buf.buf, 4))
1673 die("Unsupported command: %s", command_buf.buf);
1676 fixup_header_footer();
1678 write_index(idx_name);
1684 fprintf(stderr, "%s statistics:\n", argv[0]);
1685 fprintf(stderr, "---------------------------------------------------\n");
1686 fprintf(stderr, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count, alloc_count - est_obj_cnt);
1687 fprintf(stderr, "Total objects: %10lu (%10lu duplicates)\n", object_count, duplicate_count);
1688 fprintf(stderr, " blobs : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB]);
1689 fprintf(stderr, " trees : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE]);
1690 fprintf(stderr, " commits: %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT]);
1691 fprintf(stderr, " tags : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG]);
1692 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
1693 fprintf(stderr, " marks: %10u (%10lu unique )\n", (1 << marks->shift) * 1024, marks_set_count);
1694 fprintf(stderr, " atoms: %10u\n", atom_cnt);
1695 fprintf(stderr, "Memory total: %10lu KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
1696 fprintf(stderr, " pools: %10lu KiB\n", total_allocd/1024);
1697 fprintf(stderr, " objects: %10lu KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
1698 fprintf(stderr, "Pack remaps: %10lu\n", remap_count);
1699 fprintf(stderr, "---------------------------------------------------\n");
1701 stat(pack_name, &sb);
1702 fprintf(stderr, "Pack size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1703 stat(idx_name, &sb);
1704 fprintf(stderr, "Index size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1706 fprintf(stderr, "\n");