2 Format of STDIN stream:
12 new_blob ::= 'blob' lf
15 file_content ::= data;
17 new_branch ::= 'branch' sp ref_str lf
18 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
21 new_commit ::= 'commit' sp ref_str lf
23 ('author' sp name '<' email '>' ts tz lf)?
24 'committer' sp name '<' email '>' ts tz lf
30 file_change ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf
33 mode ::= '644' | '755';
35 new_tag ::= 'tag' sp tag_str lf
36 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
37 'tagger' sp name '<' email '>' ts tz lf
41 # note: the first idnum in a stream should be 1 and subsequent
42 # idnums should not have gaps between values as this will cause
43 # the stream parser to reserve space for the gapped values. An
44 # idnum can be updated in the future to a new object by issuing
45 # a new mark directive with the old idnum.
47 mark ::= 'mark' sp idnum lf;
49 # note: declen indicates the length of binary_data in bytes.
50 # declen does not include the lf preceeding or trailing the
53 data ::= 'data' sp declen lf
57 # note: quoted strings are C-style quoting supporting \c for
58 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
59 # is the signed byte value in octal. Note that the only
60 # characters which must actually be escaped to protect the
61 # stream formatting is: \, " and LF. Otherwise these values
64 ref_str ::= ref | '"' quoted(ref) '"' ;
65 sha1exp_str ::= sha1exp | '"' quoted(sha1exp) '"' ;
66 tag_str ::= tag | '"' quoted(tag) '"' ;
67 path_str ::= path | '"' quoted(path) '"' ;
69 declen ::= # unsigned 32 bit value, ascii base10 notation;
70 binary_data ::= # file content, not interpreted;
72 sp ::= # ASCII space character;
73 lf ::= # ASCII newline (LF) character;
75 # note: a colon (':') must precede the numerical value assigned to
76 # an idnum. This is to distinguish it from a ref or tag name as
77 # GIT does not permit ':' in ref or tag strings.
80 path ::= # GIT style file path, e.g. "a/b/c";
81 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
82 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
83 sha1exp ::= # Any valid GIT SHA1 expression;
84 hexsha1 ::= # SHA1 in hexadecimal format;
86 # note: name and email are UTF8 strings, however name must not
87 # contain '<' or lf and email must not contain any of the
88 # following: '<', '>', lf.
90 name ::= # valid GIT author/committer name;
91 email ::= # valid GIT author/committer email;
92 ts ::= # time since the epoch in seconds, ascii base10 notation;
93 tz ::= # GIT style timezone;
104 #include "csum-file.h"
110 struct object_entry *next;
111 enum object_type type;
112 unsigned long offset;
113 unsigned char sha1[20];
116 struct object_entry_pool
118 struct object_entry_pool *next_pool;
119 struct object_entry *next_free;
120 struct object_entry *end;
121 struct object_entry entries[FLEX_ARRAY]; /* more */
128 struct object_entry *marked[1024];
129 struct mark_set *sets[1024];
138 unsigned char sha1[20];
143 struct mem_pool *next_pool;
146 char space[FLEX_ARRAY]; /* more */
151 struct atom_str *next_atom;
153 char str_dat[FLEX_ARRAY]; /* more */
159 struct tree_content *tree;
160 struct atom_str* name;
162 unsigned char sha1[20];
167 unsigned int entry_capacity; /* must match avail_tree_content */
168 unsigned int entry_count;
169 struct tree_entry *entries[FLEX_ARRAY]; /* more */
172 struct avail_tree_content
174 unsigned int entry_capacity; /* must match tree_content */
175 struct avail_tree_content *next_avail;
180 struct branch *table_next_branch;
181 struct branch *active_next_branch;
183 unsigned long last_commit;
184 struct tree_entry branch_tree;
185 unsigned char sha1[20];
189 /* Stats and misc. counters */
190 static unsigned long max_depth = 10;
191 static unsigned long alloc_count;
192 static unsigned long branch_count;
193 static unsigned long branch_load_count;
194 static unsigned long object_count;
195 static unsigned long duplicate_count;
196 static unsigned long marks_set_count;
197 static unsigned long object_count_by_type[9];
198 static unsigned long duplicate_count_by_type[9];
201 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
202 static size_t total_allocd;
203 static struct mem_pool *mem_pool;
205 /* Atom management */
206 static unsigned int atom_table_sz = 4451;
207 static unsigned int atom_cnt;
208 static struct atom_str **atom_table;
210 /* The .pack file being generated */
212 static unsigned long pack_offset;
213 static unsigned char pack_sha1[20];
215 /* Table of objects we've written. */
216 static unsigned int object_entry_alloc = 1000;
217 static struct object_entry_pool *blocks;
218 static struct object_entry *object_table[1 << 16];
219 static struct mark_set *marks;
222 static struct last_object last_blob;
224 /* Tree management */
225 static unsigned int tree_entry_alloc = 1000;
226 static void *avail_tree_entry;
227 static unsigned int avail_tree_table_sz = 100;
228 static struct avail_tree_content **avail_tree_table;
231 static unsigned long max_active_branches = 5;
232 static unsigned long cur_active_branches;
233 static unsigned long branch_table_sz = 1039;
234 static struct branch **branch_table;
235 static struct branch *active_branches;
237 /* Input stream parsing */
238 static struct strbuf command_buf;
239 static unsigned long next_mark;
242 static void alloc_objects(int cnt)
244 struct object_entry_pool *b;
246 b = xmalloc(sizeof(struct object_entry_pool)
247 + cnt * sizeof(struct object_entry));
248 b->next_pool = blocks;
249 b->next_free = b->entries;
250 b->end = b->entries + cnt;
255 static struct object_entry* new_object(unsigned char *sha1)
257 struct object_entry *e;
259 if (blocks->next_free == blocks->end)
260 alloc_objects(object_entry_alloc);
262 e = blocks->next_free++;
263 memcpy(e->sha1, sha1, sizeof(e->sha1));
267 static struct object_entry* find_object(unsigned char *sha1)
269 unsigned int h = sha1[0] << 8 | sha1[1];
270 struct object_entry *e;
271 for (e = object_table[h]; e; e = e->next)
272 if (!memcmp(sha1, e->sha1, sizeof(e->sha1)))
277 static struct object_entry* insert_object(unsigned char *sha1)
279 unsigned int h = sha1[0] << 8 | sha1[1];
280 struct object_entry *e = object_table[h];
281 struct object_entry *p = NULL;
284 if (!memcmp(sha1, e->sha1, sizeof(e->sha1)))
290 e = new_object(sha1);
300 static unsigned int hc_str(const char *s, size_t len)
308 static void* pool_alloc(size_t len)
313 for (p = mem_pool; p; p = p->next_pool)
314 if ((p->end - p->next_free >= len))
318 if (len >= (mem_pool_alloc/2)) {
322 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
323 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
324 p->next_pool = mem_pool;
325 p->next_free = p->space;
326 p->end = p->next_free + mem_pool_alloc;
335 static void* pool_calloc(size_t count, size_t size)
337 size_t len = count * size;
338 void *r = pool_alloc(len);
343 static char* pool_strdup(const char *s)
345 char *r = pool_alloc(strlen(s) + 1);
350 static void insert_mark(unsigned long idnum, struct object_entry *oe)
352 struct mark_set *s = marks;
353 while ((idnum >> s->shift) >= 1024) {
354 s = pool_calloc(1, sizeof(struct mark_set));
355 s->shift = marks->shift + 10;
356 s->data.sets[0] = marks;
360 unsigned long i = idnum >> s->shift;
361 idnum -= i << s->shift;
362 if (!s->data.sets[i]) {
363 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
364 s->data.sets[i]->shift = s->shift - 10;
368 if (!s->data.marked[idnum])
370 s->data.marked[idnum] = oe;
373 static struct object_entry* find_mark(unsigned long idnum)
375 unsigned long orig_idnum = idnum;
376 struct mark_set *s = marks;
377 struct object_entry *oe = NULL;
378 if ((idnum >> s->shift) < 1024) {
379 while (s && s->shift) {
380 unsigned long i = idnum >> s->shift;
381 idnum -= i << s->shift;
385 oe = s->data.marked[idnum];
388 die("mark :%lu not declared", orig_idnum);
392 static struct atom_str* to_atom(const char *s, size_t len)
394 unsigned int hc = hc_str(s, len) % atom_table_sz;
397 for (c = atom_table[hc]; c; c = c->next_atom)
398 if (c->str_len == len && !strncmp(s, c->str_dat, len))
401 c = pool_alloc(sizeof(struct atom_str) + len + 1);
403 strncpy(c->str_dat, s, len);
405 c->next_atom = atom_table[hc];
411 static struct branch* lookup_branch(const char *name)
413 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
416 for (b = branch_table[hc]; b; b = b->table_next_branch)
417 if (!strcmp(name, b->name))
422 static struct branch* new_branch(const char *name)
424 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
425 struct branch* b = lookup_branch(name);
428 die("Invalid attempt to create duplicate branch: %s", name);
429 if (check_ref_format(name))
430 die("Branch name doesn't conform to GIT standards: %s", name);
432 b = pool_calloc(1, sizeof(struct branch));
433 b->name = pool_strdup(name);
434 b->table_next_branch = branch_table[hc];
435 branch_table[hc] = b;
440 static unsigned int hc_entries(unsigned int cnt)
442 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
443 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
446 static struct tree_content* new_tree_content(unsigned int cnt)
448 struct avail_tree_content *f, *l = NULL;
449 struct tree_content *t;
450 unsigned int hc = hc_entries(cnt);
452 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
453 if (f->entry_capacity >= cnt)
458 l->next_avail = f->next_avail;
460 avail_tree_table[hc] = f->next_avail;
462 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
463 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
464 f->entry_capacity = cnt;
467 t = (struct tree_content*)f;
472 static void release_tree_entry(struct tree_entry *e);
473 static void release_tree_content(struct tree_content *t)
475 struct avail_tree_content *f = (struct avail_tree_content*)t;
476 unsigned int hc = hc_entries(f->entry_capacity);
477 f->next_avail = avail_tree_table[hc];
478 avail_tree_table[hc] = f;
481 static void release_tree_content_recursive(struct tree_content *t)
484 for (i = 0; i < t->entry_count; i++)
485 release_tree_entry(t->entries[i]);
486 release_tree_content(t);
489 static struct tree_content* grow_tree_content(
490 struct tree_content *t,
493 struct tree_content *r = new_tree_content(t->entry_count + amt);
494 r->entry_count = t->entry_count;
495 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
496 release_tree_content(t);
500 static struct tree_entry* new_tree_entry()
502 struct tree_entry *e;
504 if (!avail_tree_entry) {
505 unsigned int n = tree_entry_alloc;
506 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
508 *((void**)e) = e + 1;
513 e = avail_tree_entry;
514 avail_tree_entry = *((void**)e);
518 static void release_tree_entry(struct tree_entry *e)
521 release_tree_content_recursive(e->tree);
522 *((void**)e) = avail_tree_entry;
523 avail_tree_entry = e;
526 static void yread(int fd, void *buffer, size_t length)
529 while (ret < length) {
530 ssize_t size = xread(fd, (char *) buffer + ret, length - ret);
532 die("Read from descriptor %i: end of stream", fd);
534 die("Read from descriptor %i: %s", fd, strerror(errno));
539 static void ywrite(int fd, void *buffer, size_t length)
542 while (ret < length) {
543 ssize_t size = xwrite(fd, (char *) buffer + ret, length - ret);
545 die("Write to descriptor %i: end of file", fd);
547 die("Write to descriptor %i: %s", fd, strerror(errno));
552 static size_t encode_header(
553 enum object_type type,
560 if (type < OBJ_COMMIT || type > OBJ_DELTA)
561 die("bad type %d", type);
563 c = (type << 4) | (size & 15);
575 static int store_object(
576 enum object_type type,
579 struct last_object *last,
580 unsigned char *sha1out,
584 struct object_entry *e;
585 unsigned char hdr[96];
586 unsigned char sha1[20];
587 unsigned long hdrlen, deltalen;
591 hdrlen = sprintf((char*)hdr,"%s %lu",type_names[type],datlen) + 1;
593 SHA1_Update(&c, hdr, hdrlen);
594 SHA1_Update(&c, dat, datlen);
595 SHA1_Final(sha1, &c);
597 memcpy(sha1out, sha1, sizeof(sha1));
599 e = insert_object(sha1);
601 insert_mark(mark, e);
604 duplicate_count_by_type[type]++;
608 e->offset = pack_offset;
610 object_count_by_type[type]++;
612 if (last && last->data && last->depth < max_depth)
613 delta = diff_delta(last->data, last->len,
619 memset(&s, 0, sizeof(s));
620 deflateInit(&s, zlib_compression_level);
625 s.avail_in = deltalen;
626 hdrlen = encode_header(OBJ_DELTA, deltalen, hdr);
627 ywrite(pack_fd, hdr, hdrlen);
628 ywrite(pack_fd, last->sha1, sizeof(sha1));
629 pack_offset += hdrlen + sizeof(sha1);
635 hdrlen = encode_header(type, datlen, hdr);
636 ywrite(pack_fd, hdr, hdrlen);
637 pack_offset += hdrlen;
640 s.avail_out = deflateBound(&s, s.avail_in);
641 s.next_out = out = xmalloc(s.avail_out);
642 while (deflate(&s, Z_FINISH) == Z_OK)
646 ywrite(pack_fd, out, s.total_out);
647 pack_offset += s.total_out;
657 memcpy(last->sha1, sha1, sizeof(sha1));
662 static const char *get_mode(const char *str, unsigned int *modep)
665 unsigned int mode = 0;
667 while ((c = *str++) != ' ') {
668 if (c < '0' || c > '7')
670 mode = (mode << 3) + (c - '0');
676 static void load_tree(struct tree_entry *root)
678 struct object_entry *myoe;
679 struct tree_content *t;
685 root->tree = t = new_tree_content(8);
686 if (!memcmp(root->sha1, null_sha1, 20))
689 myoe = find_object(root->sha1);
693 buf = read_sha1_file(root->sha1, type, &size);
694 if (!buf || strcmp(type, tree_type))
695 die("Can't load existing tree %s", sha1_to_hex(root->sha1));
699 while (c != (buf + size)) {
700 struct tree_entry *e = new_tree_entry();
702 if (t->entry_count == t->entry_capacity)
703 root->tree = t = grow_tree_content(t, 8);
704 t->entries[t->entry_count++] = e;
707 c = get_mode(c, &e->mode);
709 die("Corrupt mode in %s", sha1_to_hex(root->sha1));
710 e->name = to_atom(c, strlen(c));
711 c += e->name->str_len + 1;
712 memcpy(e->sha1, c, sizeof(e->sha1));
718 static int tecmp (const void *_a, const void *_b)
720 struct tree_entry *a = *((struct tree_entry**)_a);
721 struct tree_entry *b = *((struct tree_entry**)_b);
722 return base_name_compare(
723 a->name->str_dat, a->name->str_len, a->mode,
724 b->name->str_dat, b->name->str_len, b->mode);
727 static void store_tree(struct tree_entry *root)
729 struct tree_content *t = root->tree;
734 if (memcmp(root->sha1, null_sha1, 20))
738 for (i = 0; i < t->entry_count; i++) {
739 maxlen += t->entries[i]->name->str_len + 34;
740 if (t->entries[i]->tree)
741 store_tree(t->entries[i]);
744 qsort(t->entries, t->entry_count, sizeof(t->entries[0]), tecmp);
745 buf = c = xmalloc(maxlen);
746 for (i = 0; i < t->entry_count; i++) {
747 struct tree_entry *e = t->entries[i];
748 c += sprintf(c, "%o", e->mode);
750 strcpy(c, e->name->str_dat);
751 c += e->name->str_len + 1;
752 memcpy(c, e->sha1, 20);
755 store_object(OBJ_TREE, buf, c - buf, NULL, root->sha1, 0);
759 static int tree_content_set(
760 struct tree_entry *root,
762 const unsigned char *sha1,
763 const unsigned int mode)
765 struct tree_content *t = root->tree;
768 struct tree_entry *e;
770 slash1 = strchr(p, '/');
776 for (i = 0; i < t->entry_count; i++) {
778 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
780 if (e->mode == mode && !memcmp(e->sha1, sha1, 20))
783 memcpy(e->sha1, sha1, 20);
785 release_tree_content_recursive(e->tree);
788 memcpy(root->sha1, null_sha1, 20);
791 if (!S_ISDIR(e->mode)) {
792 e->tree = new_tree_content(8);
797 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
798 memcpy(root->sha1, null_sha1, 20);
805 if (t->entry_count == t->entry_capacity)
806 root->tree = t = grow_tree_content(t, 8);
807 e = new_tree_entry();
808 e->name = to_atom(p, n);
809 t->entries[t->entry_count++] = e;
811 e->tree = new_tree_content(8);
813 tree_content_set(e, slash1 + 1, sha1, mode);
817 memcpy(e->sha1, sha1, 20);
819 memcpy(root->sha1, null_sha1, 20);
823 static int tree_content_remove(struct tree_entry *root, const char *p)
825 struct tree_content *t = root->tree;
828 struct tree_entry *e;
830 slash1 = strchr(p, '/');
836 for (i = 0; i < t->entry_count; i++) {
838 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
839 if (!slash1 || !S_ISDIR(e->mode))
843 if (tree_content_remove(e, slash1 + 1)) {
844 if (!e->tree->entry_count)
846 memcpy(root->sha1, null_sha1, 20);
855 for (i++; i < t->entry_count; i++)
856 t->entries[i-1] = t->entries[i];
858 release_tree_entry(e);
859 memcpy(root->sha1, null_sha1, 20);
863 static void init_pack_header()
865 struct pack_header hdr;
867 hdr.hdr_signature = htonl(PACK_SIGNATURE);
868 hdr.hdr_version = htonl(2);
871 ywrite(pack_fd, &hdr, sizeof(hdr));
872 pack_offset = sizeof(hdr);
875 static void fixup_header_footer()
883 if (lseek(pack_fd, 0, SEEK_SET) != 0)
884 die("Failed seeking to start: %s", strerror(errno));
887 yread(pack_fd, hdr, 8);
888 SHA1_Update(&c, hdr, 8);
890 cnt = htonl(object_count);
891 SHA1_Update(&c, &cnt, 4);
892 ywrite(pack_fd, &cnt, 4);
894 buf = xmalloc(128 * 1024);
896 n = xread(pack_fd, buf, 128 * 1024);
899 SHA1_Update(&c, buf, n);
903 SHA1_Final(pack_sha1, &c);
904 ywrite(pack_fd, pack_sha1, sizeof(pack_sha1));
907 static int oecmp (const void *_a, const void *_b)
909 struct object_entry *a = *((struct object_entry**)_a);
910 struct object_entry *b = *((struct object_entry**)_b);
911 return memcmp(a->sha1, b->sha1, sizeof(a->sha1));
914 static void write_index(const char *idx_name)
917 struct object_entry **idx, **c, **last;
918 struct object_entry *e;
919 struct object_entry_pool *o;
920 unsigned int array[256];
923 /* Build the sorted table of object IDs. */
924 idx = xmalloc(object_count * sizeof(struct object_entry*));
926 for (o = blocks; o; o = o->next_pool)
927 for (e = o->entries; e != o->next_free; e++)
929 last = idx + object_count;
930 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
932 /* Generate the fan-out array. */
934 for (i = 0; i < 256; i++) {
935 struct object_entry **next = c;;
936 while (next < last) {
937 if ((*next)->sha1[0] != i)
941 array[i] = htonl(next - idx);
945 f = sha1create("%s", idx_name);
946 sha1write(f, array, 256 * sizeof(int));
947 for (c = idx; c != last; c++) {
948 unsigned int offset = htonl((*c)->offset);
949 sha1write(f, &offset, 4);
950 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
952 sha1write(f, pack_sha1, sizeof(pack_sha1));
953 sha1close(f, NULL, 1);
957 static void dump_branches()
959 static const char *msg = "fast-import";
962 struct ref_lock *lock;
964 for (i = 0; i < branch_table_sz; i++) {
965 for (b = branch_table[i]; b; b = b->table_next_branch) {
966 lock = lock_any_ref_for_update(b->name, NULL, 0);
967 if (!lock || write_ref_sha1(lock, b->sha1, msg) < 0)
968 die("Can't write %s", b->name);
973 static void read_next_command()
975 read_line(&command_buf, stdin, '\n');
978 static void cmd_mark()
980 if (!strncmp("mark :", command_buf.buf, 6)) {
981 next_mark = strtoul(command_buf.buf + 6, NULL, 10);
988 static void* cmd_data (size_t *size)
994 if (strncmp("data ", command_buf.buf, 5))
995 die("Expected 'data n' command, found: %s", command_buf.buf);
997 length = strtoul(command_buf.buf + 5, NULL, 10);
998 buffer = xmalloc(length);
1000 while (n < length) {
1001 size_t s = fread((char*)buffer + n, 1, length - n, stdin);
1002 if (!s && feof(stdin))
1003 die("EOF in data (%lu bytes remaining)", length - n);
1007 if (fgetc(stdin) != '\n')
1008 die("An lf did not trail the binary data as expected.");
1014 static void cmd_new_blob()
1019 read_next_command();
1023 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1027 static void unload_one_branch()
1029 while (cur_active_branches >= max_active_branches) {
1030 unsigned long min_commit = ULONG_MAX;
1031 struct branch *e, *l = NULL, *p = NULL;
1033 for (e = active_branches; e; e = e->active_next_branch) {
1034 if (e->last_commit < min_commit) {
1036 min_commit = e->last_commit;
1042 e = p->active_next_branch;
1043 p->active_next_branch = e->active_next_branch;
1045 e = active_branches;
1046 active_branches = e->active_next_branch;
1048 e->active_next_branch = NULL;
1049 if (e->branch_tree.tree) {
1050 release_tree_content_recursive(e->branch_tree.tree);
1051 e->branch_tree.tree = NULL;
1053 cur_active_branches--;
1057 static void load_branch(struct branch *b)
1059 load_tree(&b->branch_tree);
1060 b->active_next_branch = active_branches;
1061 active_branches = b;
1062 cur_active_branches++;
1063 branch_load_count++;
1066 static void file_change_m(struct branch *b)
1068 const char *p = command_buf.buf + 2;
1071 struct object_entry *oe;
1072 unsigned char sha1[20];
1076 p = get_mode(p, &mode);
1078 die("Corrupt mode: %s", command_buf.buf);
1080 case S_IFREG | 0644:
1081 case S_IFREG | 0755:
1088 die("Corrupt mode: %s", command_buf.buf);
1093 oe = find_mark(strtoul(p + 1, &x, 10));
1096 if (get_sha1_hex(p, sha1))
1097 die("Invalid SHA1: %s", command_buf.buf);
1098 oe = find_object(sha1);
1102 die("Missing space after SHA1: %s", command_buf.buf);
1104 p_uq = unquote_c_style(p, &endp);
1107 die("Garbage after path in: %s", command_buf.buf);
1112 if (oe->type != OBJ_BLOB)
1113 die("Not a blob (actually a %s): %s",
1114 command_buf.buf, type_names[oe->type]);
1116 if (sha1_object_info(sha1, type, NULL))
1117 die("Blob not found: %s", command_buf.buf);
1118 if (strcmp(blob_type, type))
1119 die("Not a blob (actually a %s): %s",
1120 command_buf.buf, type);
1123 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1129 static void file_change_d(struct branch *b)
1131 const char *p = command_buf.buf + 2;
1135 p_uq = unquote_c_style(p, &endp);
1138 die("Garbage after path in: %s", command_buf.buf);
1141 tree_content_remove(&b->branch_tree, p);
1146 static void cmd_new_commit()
1154 char *author = NULL;
1155 char *committer = NULL;
1158 /* Obtain the branch name from the rest of our command */
1159 sp = strchr(command_buf.buf, ' ') + 1;
1160 str_uq = unquote_c_style(sp, &endp);
1163 die("Garbage after ref in: %s", command_buf.buf);
1166 b = lookup_branch(sp);
1168 die("Branch not declared: %s", sp);
1172 read_next_command();
1174 if (!strncmp("author ", command_buf.buf, 7)) {
1175 author = strdup(command_buf.buf);
1176 read_next_command();
1178 if (!strncmp("committer ", command_buf.buf, 10)) {
1179 committer = strdup(command_buf.buf);
1180 read_next_command();
1183 die("Expected committer but didn't get one");
1184 msg = cmd_data(&msglen);
1186 /* ensure the branch is active/loaded */
1187 if (!b->branch_tree.tree) {
1188 unload_one_branch();
1194 read_next_command();
1195 if (1 == command_buf.len)
1197 else if (!strncmp("M ", command_buf.buf, 2))
1199 else if (!strncmp("D ", command_buf.buf, 2))
1202 die("Unsupported file_change: %s", command_buf.buf);
1205 /* build the tree and the commit */
1206 store_tree(&b->branch_tree);
1207 body = xmalloc(97 + msglen
1209 ? strlen(author) + strlen(committer)
1210 : 2 * strlen(committer)));
1212 sp += sprintf(sp, "tree %s\n", sha1_to_hex(b->branch_tree.sha1));
1213 if (memcmp(b->sha1, null_sha1, 20))
1214 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1216 sp += sprintf(sp, "%s\n", author);
1218 sp += sprintf(sp, "author %s\n", committer + 10);
1219 sp += sprintf(sp, "%s\n\n", committer);
1220 memcpy(sp, msg, msglen);
1227 store_object(OBJ_COMMIT, body, sp - body, NULL, b->sha1, next_mark);
1229 b->last_commit = object_count_by_type[OBJ_COMMIT];
1232 static void cmd_new_branch()
1239 /* Obtain the new branch name from the rest of our command */
1240 sp = strchr(command_buf.buf, ' ') + 1;
1241 str_uq = unquote_c_style(sp, &endp);
1244 die("Garbage after ref in: %s", command_buf.buf);
1250 read_next_command();
1253 if (!strncmp("from ", command_buf.buf, 5)) {
1257 from = strchr(command_buf.buf, ' ') + 1;
1258 str_uq = unquote_c_style(from, &endp);
1261 die("Garbage after string in: %s", command_buf.buf);
1265 s = lookup_branch(from);
1267 die("Can't create a branch from itself: %s", b->name);
1269 memcpy(b->sha1, s->sha1, 20);
1270 memcpy(b->branch_tree.sha1, s->branch_tree.sha1, 20);
1271 } else if (*from == ':') {
1272 unsigned long idnum = strtoul(from + 1, NULL, 10);
1273 struct object_entry *oe = find_mark(idnum);
1274 if (oe->type != OBJ_COMMIT)
1275 die("Mark :%lu not a commit", idnum);
1276 memcpy(b->sha1, oe->sha1, 20);
1277 memcpy(b->branch_tree.sha1, null_sha1, 20);
1278 } else if (!get_sha1(from, b->sha1)) {
1279 if (!memcmp(b->sha1, null_sha1, 20))
1280 memcpy(b->branch_tree.sha1, null_sha1, 20);
1285 buf = read_object_with_reference(b->sha1,
1286 type_names[OBJ_COMMIT], &size, b->sha1);
1287 if (!buf || size < 46)
1288 die("Not a valid commit: %s", from);
1289 if (memcmp("tree ", buf, 5)
1290 || get_sha1_hex(buf + 5, b->branch_tree.sha1))
1291 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1295 die("Invalid ref name or SHA1 expression: %s", from);
1299 read_next_command();
1301 memcpy(b->sha1, null_sha1, 20);
1302 memcpy(b->branch_tree.sha1, null_sha1, 20);
1305 if (command_buf.eof || command_buf.len > 1)
1306 die("An lf did not terminate the branch command as expected.");
1309 static const char fast_import_usage[] =
1310 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] temp.pack";
1312 int main(int argc, const char **argv)
1314 const char *base_name;
1316 unsigned long est_obj_cnt = 1000;
1322 git_config(git_default_config);
1324 for (i = 1; i < argc; i++) {
1325 const char *a = argv[i];
1327 if (*a != '-' || !strcmp(a, "--"))
1329 else if (!strncmp(a, "--objects=", 10))
1330 est_obj_cnt = strtoul(a + 10, NULL, 0);
1331 else if (!strncmp(a, "--depth=", 8))
1332 max_depth = strtoul(a + 8, NULL, 0);
1333 else if (!strncmp(a, "--active-branches=", 18))
1334 max_active_branches = strtoul(a + 18, NULL, 0);
1336 die("unknown option %s", a);
1339 usage(fast_import_usage);
1340 base_name = argv[i];
1342 pack_name = xmalloc(strlen(base_name) + 6);
1343 sprintf(pack_name, "%s.pack", base_name);
1344 idx_name = xmalloc(strlen(base_name) + 5);
1345 sprintf(idx_name, "%s.idx", base_name);
1347 pack_fd = open(pack_name, O_RDWR|O_CREAT|O_EXCL, 0666);
1349 die("Can't create %s: %s", pack_name, strerror(errno));
1352 alloc_objects(est_obj_cnt);
1353 strbuf_init(&command_buf);
1355 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
1356 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
1357 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
1358 marks = pool_calloc(1, sizeof(struct mark_set));
1361 read_next_command();
1362 if (command_buf.eof)
1364 else if (!strcmp("blob", command_buf.buf))
1366 else if (!strncmp("branch ", command_buf.buf, 7))
1368 else if (!strncmp("commit ", command_buf.buf, 7))
1371 die("Unsupported command: %s", command_buf.buf);
1374 fixup_header_footer();
1376 write_index(idx_name);
1379 fprintf(stderr, "%s statistics:\n", argv[0]);
1380 fprintf(stderr, "---------------------------------------------------\n");
1381 fprintf(stderr, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count, alloc_count - est_obj_cnt);
1382 fprintf(stderr, "Total objects: %10lu (%10lu duplicates)\n", object_count, duplicate_count);
1383 fprintf(stderr, " blobs : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB]);
1384 fprintf(stderr, " trees : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE]);
1385 fprintf(stderr, " commits: %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT]);
1386 fprintf(stderr, " tags : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG]);
1387 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
1388 fprintf(stderr, " marks: %10u (%10lu unique )\n", (1 << marks->shift) * 1024, marks_set_count);
1389 fprintf(stderr, " atoms: %10u\n", atom_cnt);
1390 fprintf(stderr, "Memory total: %10lu KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
1391 fprintf(stderr, " pools: %10lu KiB\n", total_allocd/1024);
1392 fprintf(stderr, " objects: %10lu KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
1393 fprintf(stderr, "---------------------------------------------------\n");
1395 stat(pack_name, &sb);
1396 fprintf(stderr, "Pack size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1397 stat(idx_name, &sb);
1398 fprintf(stderr, "Index size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1400 fprintf(stderr, "\n");