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 object_count;
194 static unsigned long duplicate_count;
195 static unsigned long marks_set_count;
196 static unsigned long object_count_by_type[9];
197 static unsigned long duplicate_count_by_type[9];
200 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
201 static size_t total_allocd;
202 static struct mem_pool *mem_pool;
204 /* Atom management */
205 static unsigned int atom_table_sz = 4451;
206 static unsigned int atom_cnt;
207 static struct atom_str **atom_table;
209 /* The .pack file being generated */
211 static unsigned long pack_offset;
212 static unsigned char pack_sha1[20];
214 /* Table of objects we've written. */
215 static unsigned int object_entry_alloc = 1000;
216 static struct object_entry_pool *blocks;
217 static struct object_entry *object_table[1 << 16];
218 static struct mark_set *marks;
221 static struct last_object last_blob;
223 /* Tree management */
224 static unsigned int tree_entry_alloc = 1000;
225 static void *avail_tree_entry;
226 static unsigned int avail_tree_table_sz = 100;
227 static struct avail_tree_content **avail_tree_table;
230 static unsigned long max_active_branches = 5;
231 static unsigned long cur_active_branches;
232 static unsigned long branch_table_sz = 1039;
233 static struct branch **branch_table;
234 static struct branch *active_branches;
236 /* Input stream parsing */
237 static struct strbuf command_buf;
238 static unsigned long next_mark;
241 static void alloc_objects(int cnt)
243 struct object_entry_pool *b;
245 b = xmalloc(sizeof(struct object_entry_pool)
246 + cnt * sizeof(struct object_entry));
247 b->next_pool = blocks;
248 b->next_free = b->entries;
249 b->end = b->entries + cnt;
254 static struct object_entry* new_object(unsigned char *sha1)
256 struct object_entry *e;
258 if (blocks->next_free == blocks->end)
259 alloc_objects(object_entry_alloc);
261 e = blocks->next_free++;
262 memcpy(e->sha1, sha1, sizeof(e->sha1));
266 static struct object_entry* find_object(unsigned char *sha1)
268 unsigned int h = sha1[0] << 8 | sha1[1];
269 struct object_entry *e;
270 for (e = object_table[h]; e; e = e->next)
271 if (!memcmp(sha1, e->sha1, sizeof(e->sha1)))
276 static struct object_entry* insert_object(unsigned char *sha1)
278 unsigned int h = sha1[0] << 8 | sha1[1];
279 struct object_entry *e = object_table[h];
280 struct object_entry *p = NULL;
283 if (!memcmp(sha1, e->sha1, sizeof(e->sha1)))
289 e = new_object(sha1);
299 static unsigned int hc_str(const char *s, size_t len)
307 static void* pool_alloc(size_t len)
312 for (p = mem_pool; p; p = p->next_pool)
313 if ((p->end - p->next_free >= len))
317 if (len >= (mem_pool_alloc/2)) {
321 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
322 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
323 p->next_pool = mem_pool;
324 p->next_free = p->space;
325 p->end = p->next_free + mem_pool_alloc;
334 static void* pool_calloc(size_t count, size_t size)
336 size_t len = count * size;
337 void *r = pool_alloc(len);
342 static char* pool_strdup(const char *s)
344 char *r = pool_alloc(strlen(s) + 1);
349 static void insert_mark(unsigned long idnum, struct object_entry *oe)
351 struct mark_set *s = marks;
352 while ((idnum >> s->shift) >= 1024) {
353 s = pool_calloc(1, sizeof(struct mark_set));
354 s->shift = marks->shift + 10;
355 s->data.sets[0] = marks;
359 unsigned long i = idnum >> s->shift;
360 idnum -= i << s->shift;
361 if (!s->data.sets[i]) {
362 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
363 s->data.sets[i]->shift = s->shift - 10;
367 if (!s->data.marked[idnum])
369 s->data.marked[idnum] = oe;
372 static struct object_entry* find_mark(unsigned long idnum)
374 unsigned long orig_idnum = idnum;
375 struct mark_set *s = marks;
376 struct object_entry *oe = NULL;
377 if ((idnum >> s->shift) < 1024) {
378 while (s && s->shift) {
379 unsigned long i = idnum >> s->shift;
380 idnum -= i << s->shift;
384 oe = s->data.marked[idnum];
387 die("mark :%lu not declared", orig_idnum);
391 static struct atom_str* to_atom(const char *s, size_t len)
393 unsigned int hc = hc_str(s, len) % atom_table_sz;
396 for (c = atom_table[hc]; c; c = c->next_atom)
397 if (c->str_len == len && !strncmp(s, c->str_dat, len))
400 c = pool_alloc(sizeof(struct atom_str) + len + 1);
402 strncpy(c->str_dat, s, len);
404 c->next_atom = atom_table[hc];
410 static struct branch* lookup_branch(const char *name)
412 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
415 for (b = branch_table[hc]; b; b = b->table_next_branch)
416 if (!strcmp(name, b->name))
421 static struct branch* new_branch(const char *name)
423 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
424 struct branch* b = lookup_branch(name);
427 die("Invalid attempt to create duplicate branch: %s", name);
428 if (check_ref_format(name))
429 die("Branch name doesn't conform to GIT standards: %s", name);
431 b = pool_calloc(1, sizeof(struct branch));
432 b->name = pool_strdup(name);
433 b->table_next_branch = branch_table[hc];
434 branch_table[hc] = b;
439 static unsigned int hc_entries(unsigned int cnt)
441 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
442 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
445 static struct tree_content* new_tree_content(unsigned int cnt)
447 struct avail_tree_content *f, *l = NULL;
448 struct tree_content *t;
449 unsigned int hc = hc_entries(cnt);
451 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
452 if (f->entry_capacity >= cnt)
457 l->next_avail = f->next_avail;
459 avail_tree_table[hc] = f->next_avail;
461 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
462 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
463 f->entry_capacity = cnt;
466 t = (struct tree_content*)f;
471 static void release_tree_entry(struct tree_entry *e);
472 static void release_tree_content(struct tree_content *t)
474 struct avail_tree_content *f = (struct avail_tree_content*)t;
475 unsigned int hc = hc_entries(f->entry_capacity);
476 f->next_avail = avail_tree_table[hc];
477 avail_tree_table[hc] = f;
480 static void release_tree_content_recursive(struct tree_content *t)
483 for (i = 0; i < t->entry_count; i++)
484 release_tree_entry(t->entries[i]);
485 release_tree_content(t);
488 static struct tree_content* grow_tree_content(
489 struct tree_content *t,
492 struct tree_content *r = new_tree_content(t->entry_count + amt);
493 r->entry_count = t->entry_count;
494 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
495 release_tree_content(t);
499 static struct tree_entry* new_tree_entry()
501 struct tree_entry *e;
503 if (!avail_tree_entry) {
504 unsigned int n = tree_entry_alloc;
505 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
507 *((void**)e) = e + 1;
512 e = avail_tree_entry;
513 avail_tree_entry = *((void**)e);
517 static void release_tree_entry(struct tree_entry *e)
520 release_tree_content_recursive(e->tree);
521 *((void**)e) = avail_tree_entry;
522 avail_tree_entry = e;
525 static void yread(int fd, void *buffer, size_t length)
528 while (ret < length) {
529 ssize_t size = xread(fd, (char *) buffer + ret, length - ret);
531 die("Read from descriptor %i: end of stream", fd);
533 die("Read from descriptor %i: %s", fd, strerror(errno));
538 static void ywrite(int fd, void *buffer, size_t length)
541 while (ret < length) {
542 ssize_t size = xwrite(fd, (char *) buffer + ret, length - ret);
544 die("Write to descriptor %i: end of file", fd);
546 die("Write to descriptor %i: %s", fd, strerror(errno));
551 static size_t encode_header(
552 enum object_type type,
559 if (type < OBJ_COMMIT || type > OBJ_DELTA)
560 die("bad type %d", type);
562 c = (type << 4) | (size & 15);
574 static int store_object(
575 enum object_type type,
578 struct last_object *last,
579 unsigned char *sha1out,
583 struct object_entry *e;
584 unsigned char hdr[96];
585 unsigned char sha1[20];
586 unsigned long hdrlen, deltalen;
590 hdrlen = sprintf((char*)hdr,"%s %lu",type_names[type],datlen) + 1;
592 SHA1_Update(&c, hdr, hdrlen);
593 SHA1_Update(&c, dat, datlen);
594 SHA1_Final(sha1, &c);
596 memcpy(sha1out, sha1, sizeof(sha1));
598 e = insert_object(sha1);
600 insert_mark(mark, e);
603 duplicate_count_by_type[type]++;
607 e->offset = pack_offset;
609 object_count_by_type[type]++;
611 if (last && last->data && last->depth < max_depth)
612 delta = diff_delta(last->data, last->len,
618 memset(&s, 0, sizeof(s));
619 deflateInit(&s, zlib_compression_level);
624 s.avail_in = deltalen;
625 hdrlen = encode_header(OBJ_DELTA, deltalen, hdr);
626 ywrite(pack_fd, hdr, hdrlen);
627 ywrite(pack_fd, last->sha1, sizeof(sha1));
628 pack_offset += hdrlen + sizeof(sha1);
634 hdrlen = encode_header(type, datlen, hdr);
635 ywrite(pack_fd, hdr, hdrlen);
636 pack_offset += hdrlen;
639 s.avail_out = deflateBound(&s, s.avail_in);
640 s.next_out = out = xmalloc(s.avail_out);
641 while (deflate(&s, Z_FINISH) == Z_OK)
645 ywrite(pack_fd, out, s.total_out);
646 pack_offset += s.total_out;
656 memcpy(last->sha1, sha1, sizeof(sha1));
661 static const char *get_mode(const char *str, unsigned int *modep)
664 unsigned int mode = 0;
666 while ((c = *str++) != ' ') {
667 if (c < '0' || c > '7')
669 mode = (mode << 3) + (c - '0');
675 static void load_tree(struct tree_entry *root)
677 struct object_entry *myoe;
678 struct tree_content *t;
684 root->tree = t = new_tree_content(8);
685 if (!memcmp(root->sha1, null_sha1, 20))
688 myoe = find_object(root->sha1);
692 buf = read_sha1_file(root->sha1, type, &size);
693 if (!buf || strcmp(type, tree_type))
694 die("Can't load existing tree %s", sha1_to_hex(root->sha1));
698 while (c != (buf + size)) {
699 struct tree_entry *e = new_tree_entry();
701 if (t->entry_count == t->entry_capacity)
702 root->tree = t = grow_tree_content(t, 8);
703 t->entries[t->entry_count++] = e;
706 c = get_mode(c, &e->mode);
708 die("Corrupt mode in %s", sha1_to_hex(root->sha1));
709 e->name = to_atom(c, strlen(c));
710 c += e->name->str_len + 1;
711 memcpy(e->sha1, c, sizeof(e->sha1));
717 static int tecmp (const void *_a, const void *_b)
719 struct tree_entry *a = *((struct tree_entry**)_a);
720 struct tree_entry *b = *((struct tree_entry**)_b);
721 return base_name_compare(
722 a->name->str_dat, a->name->str_len, a->mode,
723 b->name->str_dat, b->name->str_len, b->mode);
726 static void store_tree(struct tree_entry *root)
728 struct tree_content *t = root->tree;
733 if (memcmp(root->sha1, null_sha1, 20))
737 for (i = 0; i < t->entry_count; i++) {
738 maxlen += t->entries[i]->name->str_len + 34;
739 if (t->entries[i]->tree)
740 store_tree(t->entries[i]);
743 qsort(t->entries, t->entry_count, sizeof(t->entries[0]), tecmp);
744 buf = c = xmalloc(maxlen);
745 for (i = 0; i < t->entry_count; i++) {
746 struct tree_entry *e = t->entries[i];
747 c += sprintf(c, "%o", e->mode);
749 strcpy(c, e->name->str_dat);
750 c += e->name->str_len + 1;
751 memcpy(c, e->sha1, 20);
754 store_object(OBJ_TREE, buf, c - buf, NULL, root->sha1, 0);
758 static int tree_content_set(
759 struct tree_entry *root,
761 const unsigned char *sha1,
762 const unsigned int mode)
764 struct tree_content *t = root->tree;
767 struct tree_entry *e;
769 slash1 = strchr(p, '/');
775 for (i = 0; i < t->entry_count; i++) {
777 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
779 if (e->mode == mode && !memcmp(e->sha1, sha1, 20))
782 memcpy(e->sha1, sha1, 20);
784 release_tree_content_recursive(e->tree);
787 memcpy(root->sha1, null_sha1, 20);
790 if (!S_ISDIR(e->mode)) {
791 e->tree = new_tree_content(8);
796 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
797 memcpy(root->sha1, null_sha1, 20);
804 if (t->entry_count == t->entry_capacity)
805 root->tree = t = grow_tree_content(t, 8);
806 e = new_tree_entry();
807 e->name = to_atom(p, n);
808 t->entries[t->entry_count++] = e;
810 e->tree = new_tree_content(8);
812 tree_content_set(e, slash1 + 1, sha1, mode);
816 memcpy(e->sha1, sha1, 20);
818 memcpy(root->sha1, null_sha1, 20);
822 static int tree_content_remove(struct tree_entry *root, const char *p)
824 struct tree_content *t = root->tree;
827 struct tree_entry *e;
829 slash1 = strchr(p, '/');
835 for (i = 0; i < t->entry_count; i++) {
837 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
838 if (!slash1 || !S_ISDIR(e->mode))
842 if (tree_content_remove(e, slash1 + 1)) {
843 if (!e->tree->entry_count)
845 memcpy(root->sha1, null_sha1, 20);
854 for (i++; i < t->entry_count; i++)
855 t->entries[i-1] = t->entries[i];
857 release_tree_entry(e);
858 memcpy(root->sha1, null_sha1, 20);
862 static void init_pack_header()
864 struct pack_header hdr;
866 hdr.hdr_signature = htonl(PACK_SIGNATURE);
867 hdr.hdr_version = htonl(2);
870 ywrite(pack_fd, &hdr, sizeof(hdr));
871 pack_offset = sizeof(hdr);
874 static void fixup_header_footer()
882 if (lseek(pack_fd, 0, SEEK_SET) != 0)
883 die("Failed seeking to start: %s", strerror(errno));
886 yread(pack_fd, hdr, 8);
887 SHA1_Update(&c, hdr, 8);
889 cnt = htonl(object_count);
890 SHA1_Update(&c, &cnt, 4);
891 ywrite(pack_fd, &cnt, 4);
893 buf = xmalloc(128 * 1024);
895 n = xread(pack_fd, buf, 128 * 1024);
898 SHA1_Update(&c, buf, n);
902 SHA1_Final(pack_sha1, &c);
903 ywrite(pack_fd, pack_sha1, sizeof(pack_sha1));
906 static int oecmp (const void *_a, const void *_b)
908 struct object_entry *a = *((struct object_entry**)_a);
909 struct object_entry *b = *((struct object_entry**)_b);
910 return memcmp(a->sha1, b->sha1, sizeof(a->sha1));
913 static void write_index(const char *idx_name)
916 struct object_entry **idx, **c, **last;
917 struct object_entry *e;
918 struct object_entry_pool *o;
919 unsigned int array[256];
922 /* Build the sorted table of object IDs. */
923 idx = xmalloc(object_count * sizeof(struct object_entry*));
925 for (o = blocks; o; o = o->next_pool)
926 for (e = o->entries; e != o->next_free; e++)
928 last = idx + object_count;
929 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
931 /* Generate the fan-out array. */
933 for (i = 0; i < 256; i++) {
934 struct object_entry **next = c;;
935 while (next < last) {
936 if ((*next)->sha1[0] != i)
940 array[i] = htonl(next - idx);
944 f = sha1create("%s", idx_name);
945 sha1write(f, array, 256 * sizeof(int));
946 for (c = idx; c != last; c++) {
947 unsigned int offset = htonl((*c)->offset);
948 sha1write(f, &offset, 4);
949 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
951 sha1write(f, pack_sha1, sizeof(pack_sha1));
952 sha1close(f, NULL, 1);
956 static void dump_branches()
958 static const char *msg = "fast-import";
961 struct ref_lock *lock;
963 for (i = 0; i < branch_table_sz; i++) {
964 for (b = branch_table[i]; b; b = b->table_next_branch) {
965 lock = lock_any_ref_for_update(b->name, NULL, 0);
966 if (!lock || write_ref_sha1(lock, b->sha1, msg) < 0)
967 die("Can't write %s", b->name);
972 static void read_next_command()
974 read_line(&command_buf, stdin, '\n');
977 static void cmd_mark()
979 if (!strncmp("mark :", command_buf.buf, 6)) {
980 next_mark = strtoul(command_buf.buf + 6, NULL, 10);
987 static void* cmd_data (size_t *size)
993 if (strncmp("data ", command_buf.buf, 5))
994 die("Expected 'data n' command, found: %s", command_buf.buf);
996 length = strtoul(command_buf.buf + 5, NULL, 10);
997 buffer = xmalloc(length);
1000 size_t s = fread((char*)buffer + n, 1, length - n, stdin);
1001 if (!s && feof(stdin))
1002 die("EOF in data (%lu bytes remaining)", length - n);
1006 if (fgetc(stdin) != '\n')
1007 die("An lf did not trail the binary data as expected.");
1013 static void cmd_new_blob()
1018 read_next_command();
1022 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1026 static void unload_one_branch()
1028 while (cur_active_branches >= max_active_branches) {
1029 unsigned long min_commit = ULONG_MAX;
1030 struct branch *e, *l = NULL, *p = NULL;
1032 for (e = active_branches; e; e = e->active_next_branch) {
1033 if (e->last_commit < min_commit) {
1035 min_commit = e->last_commit;
1041 e = p->active_next_branch;
1042 p->active_next_branch = e->active_next_branch;
1044 e = active_branches;
1045 active_branches = e->active_next_branch;
1047 e->active_next_branch = NULL;
1048 if (e->branch_tree.tree) {
1049 release_tree_content_recursive(e->branch_tree.tree);
1050 e->branch_tree.tree = NULL;
1052 cur_active_branches--;
1056 static void load_branch(struct branch *b)
1058 load_tree(&b->branch_tree);
1059 b->active_next_branch = active_branches;
1060 active_branches = b;
1061 cur_active_branches++;
1064 static void file_change_m(struct branch *b)
1066 const char *p = command_buf.buf + 2;
1069 struct object_entry *oe;
1070 unsigned char sha1[20];
1074 p = get_mode(p, &mode);
1076 die("Corrupt mode: %s", command_buf.buf);
1078 case S_IFREG | 0644:
1079 case S_IFREG | 0755:
1086 die("Corrupt mode: %s", command_buf.buf);
1091 oe = find_mark(strtoul(p + 1, &x, 10));
1094 if (get_sha1_hex(p, sha1))
1095 die("Invalid SHA1: %s", command_buf.buf);
1096 oe = find_object(sha1);
1100 die("Missing space after SHA1: %s", command_buf.buf);
1102 p_uq = unquote_c_style(p, &endp);
1105 die("Garbage after path in: %s", command_buf.buf);
1110 if (oe->type != OBJ_BLOB)
1111 die("Not a blob (actually a %s): %s",
1112 command_buf.buf, type_names[oe->type]);
1114 if (sha1_object_info(sha1, type, NULL))
1115 die("Blob not found: %s", command_buf.buf);
1116 if (strcmp(blob_type, type))
1117 die("Not a blob (actually a %s): %s",
1118 command_buf.buf, type);
1121 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1127 static void file_change_d(struct branch *b)
1129 const char *p = command_buf.buf + 2;
1133 p_uq = unquote_c_style(p, &endp);
1136 die("Garbage after path in: %s", command_buf.buf);
1139 tree_content_remove(&b->branch_tree, p);
1144 static void cmd_new_commit()
1152 char *author = NULL;
1153 char *committer = NULL;
1156 /* Obtain the branch name from the rest of our command */
1157 sp = strchr(command_buf.buf, ' ') + 1;
1158 str_uq = unquote_c_style(sp, &endp);
1161 die("Garbage after ref in: %s", command_buf.buf);
1164 b = lookup_branch(sp);
1166 die("Branch not declared: %s", sp);
1170 read_next_command();
1172 if (!strncmp("author ", command_buf.buf, 7)) {
1173 author = strdup(command_buf.buf);
1174 read_next_command();
1176 if (!strncmp("committer ", command_buf.buf, 10)) {
1177 committer = strdup(command_buf.buf);
1178 read_next_command();
1181 die("Expected committer but didn't get one");
1182 msg = cmd_data(&msglen);
1184 /* ensure the branch is active/loaded */
1185 if (!b->branch_tree.tree) {
1186 unload_one_branch();
1192 read_next_command();
1193 if (1 == command_buf.len)
1195 else if (!strncmp("M ", command_buf.buf, 2))
1197 else if (!strncmp("D ", command_buf.buf, 2))
1200 die("Unsupported file_change: %s", command_buf.buf);
1203 /* build the tree and the commit */
1204 store_tree(&b->branch_tree);
1205 body = xmalloc(97 + msglen
1207 ? strlen(author) + strlen(committer)
1208 : 2 * strlen(committer)));
1210 sp += sprintf(sp, "tree %s\n", sha1_to_hex(b->branch_tree.sha1));
1211 if (memcmp(b->sha1, null_sha1, 20))
1212 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1214 sp += sprintf(sp, "%s\n", author);
1216 sp += sprintf(sp, "author %s\n", committer + 10);
1217 sp += sprintf(sp, "%s\n\n", committer);
1218 memcpy(sp, msg, msglen);
1225 store_object(OBJ_COMMIT, body, sp - body, NULL, b->sha1, next_mark);
1227 b->last_commit = object_count_by_type[OBJ_COMMIT];
1230 static void cmd_new_branch()
1237 /* Obtain the new branch name from the rest of our command */
1238 sp = strchr(command_buf.buf, ' ') + 1;
1239 str_uq = unquote_c_style(sp, &endp);
1242 die("Garbage after ref in: %s", command_buf.buf);
1248 read_next_command();
1251 if (!strncmp("from ", command_buf.buf, 5)) {
1255 from = strchr(command_buf.buf, ' ') + 1;
1256 str_uq = unquote_c_style(from, &endp);
1259 die("Garbage after string in: %s", command_buf.buf);
1263 s = lookup_branch(from);
1265 die("Can't create a branch from itself: %s", b->name);
1267 memcpy(b->sha1, s->sha1, 20);
1268 memcpy(b->branch_tree.sha1, s->branch_tree.sha1, 20);
1269 } else if (*from == ':') {
1270 unsigned long idnum = strtoul(from + 1, NULL, 10);
1271 struct object_entry *oe = find_mark(idnum);
1272 if (oe->type != OBJ_COMMIT)
1273 die("Mark :%lu not a commit", idnum);
1274 memcpy(b->sha1, oe->sha1, 20);
1275 memcpy(b->branch_tree.sha1, null_sha1, 20);
1276 } else if (!get_sha1(from, b->sha1)) {
1277 if (!memcmp(b->sha1, null_sha1, 20))
1278 memcpy(b->branch_tree.sha1, null_sha1, 20);
1283 buf = read_object_with_reference(b->sha1,
1284 type_names[OBJ_COMMIT], &size, b->sha1);
1285 if (!buf || size < 46)
1286 die("Not a valid commit: %s", from);
1287 if (memcmp("tree ", buf, 5)
1288 || get_sha1_hex(buf + 5, b->branch_tree.sha1))
1289 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1293 die("Invalid ref name or SHA1 expression: %s", from);
1297 read_next_command();
1299 memcpy(b->sha1, null_sha1, 20);
1300 memcpy(b->branch_tree.sha1, null_sha1, 20);
1303 if (command_buf.eof || command_buf.len > 1)
1304 die("An lf did not terminate the branch command as expected.");
1307 static const char fast_import_usage[] =
1308 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] temp.pack";
1310 int main(int argc, const char **argv)
1312 const char *base_name;
1314 unsigned long est_obj_cnt = 1000;
1320 git_config(git_default_config);
1322 for (i = 1; i < argc; i++) {
1323 const char *a = argv[i];
1325 if (*a != '-' || !strcmp(a, "--"))
1327 else if (!strncmp(a, "--objects=", 10))
1328 est_obj_cnt = strtoul(a + 10, NULL, 0);
1329 else if (!strncmp(a, "--depth=", 8))
1330 max_depth = strtoul(a + 8, NULL, 0);
1331 else if (!strncmp(a, "--active-branches=", 18))
1332 max_active_branches = strtoul(a + 18, NULL, 0);
1334 die("unknown option %s", a);
1337 usage(fast_import_usage);
1338 base_name = argv[i];
1340 pack_name = xmalloc(strlen(base_name) + 6);
1341 sprintf(pack_name, "%s.pack", base_name);
1342 idx_name = xmalloc(strlen(base_name) + 5);
1343 sprintf(idx_name, "%s.idx", base_name);
1345 pack_fd = open(pack_name, O_RDWR|O_CREAT|O_EXCL, 0666);
1347 die("Can't create %s: %s", pack_name, strerror(errno));
1350 alloc_objects(est_obj_cnt);
1351 strbuf_init(&command_buf);
1353 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
1354 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
1355 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
1356 marks = pool_calloc(1, sizeof(struct mark_set));
1359 read_next_command();
1360 if (command_buf.eof)
1362 else if (!strcmp("blob", command_buf.buf))
1364 else if (!strncmp("branch ", command_buf.buf, 7))
1366 else if (!strncmp("commit ", command_buf.buf, 7))
1369 die("Unsupported command: %s", command_buf.buf);
1372 fixup_header_footer();
1374 write_index(idx_name);
1377 fprintf(stderr, "%s statistics:\n", argv[0]);
1378 fprintf(stderr, "---------------------------------------------------\n");
1379 fprintf(stderr, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count, alloc_count - est_obj_cnt);
1380 fprintf(stderr, "Total objects: %10lu (%10lu duplicates)\n", object_count, duplicate_count);
1381 fprintf(stderr, " blobs : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB]);
1382 fprintf(stderr, " trees : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE]);
1383 fprintf(stderr, " commits: %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT]);
1384 fprintf(stderr, " tags : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG]);
1385 fprintf(stderr, "Total branches: %10lu\n", branch_count);
1386 fprintf(stderr, " atoms: %10u\n", atom_cnt);
1387 fprintf(stderr, " marks: %10u (%10lu unique )\n", (1 << marks->shift) * 1024, marks_set_count);
1388 fprintf(stderr, "Memory total: %10lu KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
1389 fprintf(stderr, " pools: %10lu KiB\n", total_allocd/1024);
1390 fprintf(stderr, " objects: %10lu KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
1391 fprintf(stderr, "---------------------------------------------------\n");
1393 stat(pack_name, &sb);
1394 fprintf(stderr, "Pack size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1395 stat(idx_name, &sb);
1396 fprintf(stderr, "Index size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1398 fprintf(stderr, "\n");