2 Format of STDIN stream:
12 new_blob ::= 'blob' lf
15 file_content ::= data;
17 new_commit ::= 'commit' sp ref_str lf
19 ('author' sp name '<' email '>' ts tz lf)?
20 'committer' sp name '<' email '>' ts tz lf
22 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
23 ('merge' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)*
28 file_change ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf
31 mode ::= '644' | '755';
33 new_tag ::= 'tag' sp tag_str lf
34 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
35 'tagger' sp name '<' email '>' ts tz lf
39 reset_branch ::= 'reset' sp ref_str lf
40 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
43 checkpoint ::= 'checkpoint' lf
46 # note: the first idnum in a stream should be 1 and subsequent
47 # idnums should not have gaps between values as this will cause
48 # the stream parser to reserve space for the gapped values. An
49 # idnum can be updated in the future to a new object by issuing
50 # a new mark directive with the old idnum.
52 mark ::= 'mark' sp idnum lf;
54 # note: declen indicates the length of binary_data in bytes.
55 # declen does not include the lf preceeding or trailing the
58 data ::= 'data' sp declen lf
62 # note: quoted strings are C-style quoting supporting \c for
63 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
64 # is the signed byte value in octal. Note that the only
65 # characters which must actually be escaped to protect the
66 # stream formatting is: \, " and LF. Otherwise these values
69 ref_str ::= ref | '"' quoted(ref) '"' ;
70 sha1exp_str ::= sha1exp | '"' quoted(sha1exp) '"' ;
71 tag_str ::= tag | '"' quoted(tag) '"' ;
72 path_str ::= path | '"' quoted(path) '"' ;
74 declen ::= # unsigned 32 bit value, ascii base10 notation;
75 binary_data ::= # file content, not interpreted;
77 sp ::= # ASCII space character;
78 lf ::= # ASCII newline (LF) character;
80 # note: a colon (':') must precede the numerical value assigned to
81 # an idnum. This is to distinguish it from a ref or tag name as
82 # GIT does not permit ':' in ref or tag strings.
85 path ::= # GIT style file path, e.g. "a/b/c";
86 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
87 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
88 sha1exp ::= # Any valid GIT SHA1 expression;
89 hexsha1 ::= # SHA1 in hexadecimal format;
91 # note: name and email are UTF8 strings, however name must not
92 # contain '<' or lf and email must not contain any of the
93 # following: '<', '>', lf.
95 name ::= # valid GIT author/committer name;
96 email ::= # valid GIT author/committer email;
97 ts ::= # time since the epoch in seconds, ascii base10 notation;
98 tz ::= # GIT style timezone;
109 #include "csum-file.h"
115 struct object_entry *next;
116 unsigned long offset;
117 unsigned type : TYPE_BITS;
118 unsigned pack_id : 16;
119 unsigned char sha1[20];
122 struct object_entry_pool
124 struct object_entry_pool *next_pool;
125 struct object_entry *next_free;
126 struct object_entry *end;
127 struct object_entry entries[FLEX_ARRAY]; /* more */
134 struct object_entry *marked[1024];
135 struct mark_set *sets[1024];
143 unsigned long offset;
150 struct mem_pool *next_pool;
153 char space[FLEX_ARRAY]; /* more */
158 struct atom_str *next_atom;
160 char str_dat[FLEX_ARRAY]; /* more */
166 struct tree_content *tree;
167 struct atom_str* name;
171 unsigned char sha1[20];
177 unsigned int entry_capacity; /* must match avail_tree_content */
178 unsigned int entry_count;
179 unsigned int delta_depth;
180 struct tree_entry *entries[FLEX_ARRAY]; /* more */
183 struct avail_tree_content
185 unsigned int entry_capacity; /* must match tree_content */
186 struct avail_tree_content *next_avail;
191 struct branch *table_next_branch;
192 struct branch *active_next_branch;
194 unsigned long last_commit;
195 struct tree_entry branch_tree;
196 unsigned char sha1[20];
201 struct tag *next_tag;
203 unsigned char sha1[20];
214 struct hash_list *next;
215 unsigned char sha1[20];
218 /* Stats and misc. counters */
219 static unsigned long max_depth = 10;
220 static unsigned long alloc_count;
221 static unsigned long branch_count;
222 static unsigned long branch_load_count;
223 static unsigned long object_count;
224 static unsigned long marks_set_count;
225 static unsigned long object_count_by_type[1 << TYPE_BITS];
226 static unsigned long duplicate_count_by_type[1 << TYPE_BITS];
227 static unsigned long delta_count_by_type[1 << TYPE_BITS];
230 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
231 static size_t total_allocd;
232 static struct mem_pool *mem_pool;
234 /* Atom management */
235 static unsigned int atom_table_sz = 4451;
236 static unsigned int atom_cnt;
237 static struct atom_str **atom_table;
239 /* The .pack file being generated */
240 static const char *base_name;
241 static unsigned int pack_id;
242 static char *idx_name;
243 static struct packed_git *pack_data;
244 static struct packed_git **all_packs;
246 static unsigned long pack_size;
247 static unsigned char pack_sha1[20];
249 /* Table of objects we've written. */
250 static unsigned int object_entry_alloc = 5000;
251 static struct object_entry_pool *blocks;
252 static struct object_entry *object_table[1 << 16];
253 static struct mark_set *marks;
254 static const char* mark_file;
257 static struct last_object last_blob;
259 /* Tree management */
260 static unsigned int tree_entry_alloc = 1000;
261 static void *avail_tree_entry;
262 static unsigned int avail_tree_table_sz = 100;
263 static struct avail_tree_content **avail_tree_table;
264 static struct dbuf old_tree;
265 static struct dbuf new_tree;
268 static unsigned long max_active_branches = 5;
269 static unsigned long cur_active_branches;
270 static unsigned long branch_table_sz = 1039;
271 static struct branch **branch_table;
272 static struct branch *active_branches;
275 static struct tag *first_tag;
276 static struct tag *last_tag;
278 /* Input stream parsing */
279 static struct strbuf command_buf;
280 static unsigned long next_mark;
281 static struct dbuf new_data;
282 static FILE* branch_log;
285 static void alloc_objects(unsigned int cnt)
287 struct object_entry_pool *b;
289 b = xmalloc(sizeof(struct object_entry_pool)
290 + cnt * sizeof(struct object_entry));
291 b->next_pool = blocks;
292 b->next_free = b->entries;
293 b->end = b->entries + cnt;
298 static struct object_entry* new_object(unsigned char *sha1)
300 struct object_entry *e;
302 if (blocks->next_free == blocks->end)
303 alloc_objects(object_entry_alloc);
305 e = blocks->next_free++;
306 e->pack_id = pack_id;
307 hashcpy(e->sha1, sha1);
311 static struct object_entry* find_object(unsigned char *sha1)
313 unsigned int h = sha1[0] << 8 | sha1[1];
314 struct object_entry *e;
315 for (e = object_table[h]; e; e = e->next)
316 if (!hashcmp(sha1, e->sha1))
321 static struct object_entry* insert_object(unsigned char *sha1)
323 unsigned int h = sha1[0] << 8 | sha1[1];
324 struct object_entry *e = object_table[h];
325 struct object_entry *p = NULL;
328 if (!hashcmp(sha1, e->sha1))
334 e = new_object(sha1);
344 static unsigned int hc_str(const char *s, size_t len)
352 static void* pool_alloc(size_t len)
357 for (p = mem_pool; p; p = p->next_pool)
358 if ((p->end - p->next_free >= len))
362 if (len >= (mem_pool_alloc/2)) {
366 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
367 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
368 p->next_pool = mem_pool;
369 p->next_free = p->space;
370 p->end = p->next_free + mem_pool_alloc;
375 /* round out to a pointer alignment */
376 if (len & (sizeof(void*) - 1))
377 len += sizeof(void*) - (len & (sizeof(void*) - 1));
382 static void* pool_calloc(size_t count, size_t size)
384 size_t len = count * size;
385 void *r = pool_alloc(len);
390 static char* pool_strdup(const char *s)
392 char *r = pool_alloc(strlen(s) + 1);
397 static void size_dbuf(struct dbuf *b, size_t maxlen)
400 if (b->capacity >= maxlen)
404 b->capacity = ((maxlen / 1024) + 1) * 1024;
405 b->buffer = xmalloc(b->capacity);
408 static void insert_mark(unsigned long idnum, struct object_entry *oe)
410 struct mark_set *s = marks;
411 while ((idnum >> s->shift) >= 1024) {
412 s = pool_calloc(1, sizeof(struct mark_set));
413 s->shift = marks->shift + 10;
414 s->data.sets[0] = marks;
418 unsigned long i = idnum >> s->shift;
419 idnum -= i << s->shift;
420 if (!s->data.sets[i]) {
421 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
422 s->data.sets[i]->shift = s->shift - 10;
426 if (!s->data.marked[idnum])
428 s->data.marked[idnum] = oe;
431 static struct object_entry* find_mark(unsigned long idnum)
433 unsigned long orig_idnum = idnum;
434 struct mark_set *s = marks;
435 struct object_entry *oe = NULL;
436 if ((idnum >> s->shift) < 1024) {
437 while (s && s->shift) {
438 unsigned long i = idnum >> s->shift;
439 idnum -= i << s->shift;
443 oe = s->data.marked[idnum];
446 die("mark :%lu not declared", orig_idnum);
450 static struct atom_str* to_atom(const char *s, size_t len)
452 unsigned int hc = hc_str(s, len) % atom_table_sz;
455 for (c = atom_table[hc]; c; c = c->next_atom)
456 if (c->str_len == len && !strncmp(s, c->str_dat, len))
459 c = pool_alloc(sizeof(struct atom_str) + len + 1);
461 strncpy(c->str_dat, s, len);
463 c->next_atom = atom_table[hc];
469 static struct branch* lookup_branch(const char *name)
471 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
474 for (b = branch_table[hc]; b; b = b->table_next_branch)
475 if (!strcmp(name, b->name))
480 static struct branch* new_branch(const char *name)
482 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
483 struct branch* b = lookup_branch(name);
486 die("Invalid attempt to create duplicate branch: %s", name);
487 if (check_ref_format(name))
488 die("Branch name doesn't conform to GIT standards: %s", name);
490 b = pool_calloc(1, sizeof(struct branch));
491 b->name = pool_strdup(name);
492 b->table_next_branch = branch_table[hc];
493 b->branch_tree.versions[0].mode = S_IFDIR;
494 b->branch_tree.versions[1].mode = S_IFDIR;
495 branch_table[hc] = b;
500 static unsigned int hc_entries(unsigned int cnt)
502 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
503 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
506 static struct tree_content* new_tree_content(unsigned int cnt)
508 struct avail_tree_content *f, *l = NULL;
509 struct tree_content *t;
510 unsigned int hc = hc_entries(cnt);
512 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
513 if (f->entry_capacity >= cnt)
518 l->next_avail = f->next_avail;
520 avail_tree_table[hc] = f->next_avail;
522 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
523 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
524 f->entry_capacity = cnt;
527 t = (struct tree_content*)f;
533 static void release_tree_entry(struct tree_entry *e);
534 static void release_tree_content(struct tree_content *t)
536 struct avail_tree_content *f = (struct avail_tree_content*)t;
537 unsigned int hc = hc_entries(f->entry_capacity);
538 f->next_avail = avail_tree_table[hc];
539 avail_tree_table[hc] = f;
542 static void release_tree_content_recursive(struct tree_content *t)
545 for (i = 0; i < t->entry_count; i++)
546 release_tree_entry(t->entries[i]);
547 release_tree_content(t);
550 static struct tree_content* grow_tree_content(
551 struct tree_content *t,
554 struct tree_content *r = new_tree_content(t->entry_count + amt);
555 r->entry_count = t->entry_count;
556 r->delta_depth = t->delta_depth;
557 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
558 release_tree_content(t);
562 static struct tree_entry* new_tree_entry()
564 struct tree_entry *e;
566 if (!avail_tree_entry) {
567 unsigned int n = tree_entry_alloc;
568 total_allocd += n * sizeof(struct tree_entry);
569 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
571 *((void**)e) = e + 1;
577 e = avail_tree_entry;
578 avail_tree_entry = *((void**)e);
582 static void release_tree_entry(struct tree_entry *e)
585 release_tree_content_recursive(e->tree);
586 *((void**)e) = avail_tree_entry;
587 avail_tree_entry = e;
590 static void yread(int fd, void *buffer, size_t length)
593 while (ret < length) {
594 ssize_t size = xread(fd, (char *) buffer + ret, length - ret);
596 die("Read from descriptor %i: end of stream", fd);
598 die("Read from descriptor %i: %s", fd, strerror(errno));
603 static void start_packfile()
605 struct packed_git *p;
606 struct pack_header hdr;
608 idx_name = xmalloc(strlen(base_name) + 11);
609 p = xcalloc(1, sizeof(*p) + strlen(base_name) + 13);
610 sprintf(p->pack_name, "%s%5.5i.pack", base_name, pack_id + 1);
611 sprintf(idx_name, "%s%5.5i.idx", base_name, pack_id + 1);
613 pack_fd = open(p->pack_name, O_RDWR|O_CREAT|O_EXCL, 0666);
615 die("Can't create %s: %s", p->pack_name, strerror(errno));
616 p->pack_fd = pack_fd;
618 hdr.hdr_signature = htonl(PACK_SIGNATURE);
619 hdr.hdr_version = htonl(2);
621 write_or_die(pack_fd, &hdr, sizeof(hdr));
624 pack_size = sizeof(hdr);
627 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
628 all_packs[pack_id] = p;
631 static void fixup_header_footer()
638 if (lseek(pack_fd, 0, SEEK_SET) != 0)
639 die("Failed seeking to start: %s", strerror(errno));
642 yread(pack_fd, hdr, 8);
643 SHA1_Update(&c, hdr, 8);
645 cnt = htonl(object_count);
646 SHA1_Update(&c, &cnt, 4);
647 write_or_die(pack_fd, &cnt, 4);
649 buf = xmalloc(128 * 1024);
651 size_t n = xread(pack_fd, buf, 128 * 1024);
654 SHA1_Update(&c, buf, n);
658 SHA1_Final(pack_sha1, &c);
659 write_or_die(pack_fd, pack_sha1, sizeof(pack_sha1));
662 static int oecmp (const void *a_, const void *b_)
664 struct object_entry *a = *((struct object_entry**)a_);
665 struct object_entry *b = *((struct object_entry**)b_);
666 return hashcmp(a->sha1, b->sha1);
669 static void write_index(const char *idx_name)
672 struct object_entry **idx, **c, **last, *e;
673 struct object_entry_pool *o;
674 unsigned int array[256];
677 /* Build the sorted table of object IDs. */
678 idx = xmalloc(object_count * sizeof(struct object_entry*));
680 for (o = blocks; o; o = o->next_pool)
681 for (e = o->entries; e != o->next_free; e++)
682 if (pack_id == e->pack_id)
684 last = idx + object_count;
685 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
687 /* Generate the fan-out array. */
689 for (i = 0; i < 256; i++) {
690 struct object_entry **next = c;;
691 while (next < last) {
692 if ((*next)->sha1[0] != i)
696 array[i] = htonl(next - idx);
700 f = sha1create("%s", idx_name);
701 sha1write(f, array, 256 * sizeof(int));
702 for (c = idx; c != last; c++) {
703 unsigned int offset = htonl((*c)->offset);
704 sha1write(f, &offset, 4);
705 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
707 sha1write(f, pack_sha1, sizeof(pack_sha1));
708 sha1close(f, NULL, 1);
712 static void end_packfile()
714 struct packed_git *old_p = pack_data, *new_p;
716 fixup_header_footer();
717 write_index(idx_name);
719 /* Register the packfile with core git's machinary. */
720 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
722 die("core git rejected index %s", idx_name);
723 new_p->windows = old_p->windows;
724 new_p->pack_fd = old_p->pack_fd;
725 all_packs[pack_id++] = new_p;
726 install_packed_git(new_p);
730 /* We can't carry a delta across packfiles. */
731 free(last_blob.data);
732 last_blob.data = NULL;
734 last_blob.offset = 0;
738 static size_t encode_header(
739 enum object_type type,
746 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
747 die("bad type %d", type);
749 c = (type << 4) | (size & 15);
761 static int store_object(
762 enum object_type type,
765 struct last_object *last,
766 unsigned char *sha1out,
770 struct object_entry *e;
771 unsigned char hdr[96];
772 unsigned char sha1[20];
773 unsigned long hdrlen, deltalen;
777 hdrlen = sprintf((char*)hdr,"%s %lu",type_names[type],datlen) + 1;
779 SHA1_Update(&c, hdr, hdrlen);
780 SHA1_Update(&c, dat, datlen);
781 SHA1_Final(sha1, &c);
783 hashcpy(sha1out, sha1);
785 e = insert_object(sha1);
787 insert_mark(mark, e);
789 duplicate_count_by_type[type]++;
793 e->offset = pack_size;
795 object_count_by_type[type]++;
797 if (last && last->data && last->depth < max_depth)
798 delta = diff_delta(last->data, last->len,
804 memset(&s, 0, sizeof(s));
805 deflateInit(&s, zlib_compression_level);
808 unsigned long ofs = e->offset - last->offset;
809 unsigned pos = sizeof(hdr) - 1;
811 delta_count_by_type[type]++;
814 s.avail_in = deltalen;
816 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
817 write_or_die(pack_fd, hdr, hdrlen);
820 hdr[pos] = ofs & 127;
822 hdr[--pos] = 128 | (--ofs & 127);
823 write_or_die(pack_fd, hdr + pos, sizeof(hdr) - pos);
824 pack_size += sizeof(hdr) - pos;
830 hdrlen = encode_header(type, datlen, hdr);
831 write_or_die(pack_fd, hdr, hdrlen);
835 s.avail_out = deflateBound(&s, s.avail_in);
836 s.next_out = out = xmalloc(s.avail_out);
837 while (deflate(&s, Z_FINISH) == Z_OK)
841 write_or_die(pack_fd, out, s.total_out);
842 pack_size += s.total_out;
848 if (last->data && !last->no_free)
851 last->offset = e->offset;
857 static void *gfi_unpack_entry(
858 struct object_entry *oe,
859 unsigned long *sizep)
861 static char type[20];
862 struct packed_git *p = all_packs[oe->pack_id];
864 p->pack_size = pack_size + 20;
865 return unpack_entry(p, oe->offset, type, sizep);
868 static const char *get_mode(const char *str, unsigned int *modep)
871 unsigned int mode = 0;
873 while ((c = *str++) != ' ') {
874 if (c < '0' || c > '7')
876 mode = (mode << 3) + (c - '0');
882 static void load_tree(struct tree_entry *root)
884 unsigned char* sha1 = root->versions[1].sha1;
885 struct object_entry *myoe;
886 struct tree_content *t;
891 root->tree = t = new_tree_content(8);
892 if (is_null_sha1(sha1))
895 myoe = find_object(sha1);
897 if (myoe->type != OBJ_TREE)
898 die("Not a tree: %s", sha1_to_hex(sha1));
900 buf = gfi_unpack_entry(myoe, &size);
903 buf = read_sha1_file(sha1, type, &size);
904 if (!buf || strcmp(type, tree_type))
905 die("Can't load tree %s", sha1_to_hex(sha1));
909 while (c != (buf + size)) {
910 struct tree_entry *e = new_tree_entry();
912 if (t->entry_count == t->entry_capacity)
913 root->tree = t = grow_tree_content(t, 8);
914 t->entries[t->entry_count++] = e;
917 c = get_mode(c, &e->versions[1].mode);
919 die("Corrupt mode in %s", sha1_to_hex(sha1));
920 e->versions[0].mode = e->versions[1].mode;
921 e->name = to_atom(c, strlen(c));
922 c += e->name->str_len + 1;
923 hashcpy(e->versions[0].sha1, (unsigned char*)c);
924 hashcpy(e->versions[1].sha1, (unsigned char*)c);
930 static int tecmp0 (const void *_a, const void *_b)
932 struct tree_entry *a = *((struct tree_entry**)_a);
933 struct tree_entry *b = *((struct tree_entry**)_b);
934 return base_name_compare(
935 a->name->str_dat, a->name->str_len, a->versions[0].mode,
936 b->name->str_dat, b->name->str_len, b->versions[0].mode);
939 static int tecmp1 (const void *_a, const void *_b)
941 struct tree_entry *a = *((struct tree_entry**)_a);
942 struct tree_entry *b = *((struct tree_entry**)_b);
943 return base_name_compare(
944 a->name->str_dat, a->name->str_len, a->versions[1].mode,
945 b->name->str_dat, b->name->str_len, b->versions[1].mode);
948 static void mktree(struct tree_content *t,
958 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
960 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
962 for (i = 0; i < t->entry_count; i++) {
963 if (t->entries[i]->versions[v].mode)
964 maxlen += t->entries[i]->name->str_len + 34;
967 size_dbuf(b, maxlen);
969 for (i = 0; i < t->entry_count; i++) {
970 struct tree_entry *e = t->entries[i];
971 if (!e->versions[v].mode)
973 c += sprintf(c, "%o", e->versions[v].mode);
975 strcpy(c, e->name->str_dat);
976 c += e->name->str_len + 1;
977 hashcpy((unsigned char*)c, e->versions[v].sha1);
980 *szp = c - (char*)b->buffer;
983 static void store_tree(struct tree_entry *root)
985 struct tree_content *t = root->tree;
986 unsigned int i, j, del;
987 unsigned long new_len;
988 struct last_object lo;
989 struct object_entry *le;
991 if (!is_null_sha1(root->versions[1].sha1))
994 for (i = 0; i < t->entry_count; i++) {
995 if (t->entries[i]->tree)
996 store_tree(t->entries[i]);
999 le = find_object(root->versions[0].sha1);
1000 if (!S_ISDIR(root->versions[0].mode)
1002 || le->pack_id != pack_id) {
1006 mktree(t, 0, &lo.len, &old_tree);
1007 lo.data = old_tree.buffer;
1008 lo.offset = le->offset;
1009 lo.depth = t->delta_depth;
1013 mktree(t, 1, &new_len, &new_tree);
1014 store_object(OBJ_TREE, new_tree.buffer, new_len,
1015 &lo, root->versions[1].sha1, 0);
1017 t->delta_depth = lo.depth;
1018 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1019 struct tree_entry *e = t->entries[i];
1020 if (e->versions[1].mode) {
1021 e->versions[0].mode = e->versions[1].mode;
1022 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1023 t->entries[j++] = e;
1025 release_tree_entry(e);
1029 t->entry_count -= del;
1032 static int tree_content_set(
1033 struct tree_entry *root,
1035 const unsigned char *sha1,
1036 const unsigned int mode)
1038 struct tree_content *t = root->tree;
1041 struct tree_entry *e;
1043 slash1 = strchr(p, '/');
1049 for (i = 0; i < t->entry_count; i++) {
1051 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1053 if (e->versions[1].mode == mode
1054 && !hashcmp(e->versions[1].sha1, sha1))
1056 e->versions[1].mode = mode;
1057 hashcpy(e->versions[1].sha1, sha1);
1059 release_tree_content_recursive(e->tree);
1062 hashclr(root->versions[1].sha1);
1065 if (!S_ISDIR(e->versions[1].mode)) {
1066 e->tree = new_tree_content(8);
1067 e->versions[1].mode = S_IFDIR;
1071 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
1072 hashclr(root->versions[1].sha1);
1079 if (t->entry_count == t->entry_capacity)
1080 root->tree = t = grow_tree_content(t, 8);
1081 e = new_tree_entry();
1082 e->name = to_atom(p, n);
1083 e->versions[0].mode = 0;
1084 hashclr(e->versions[0].sha1);
1085 t->entries[t->entry_count++] = e;
1087 e->tree = new_tree_content(8);
1088 e->versions[1].mode = S_IFDIR;
1089 tree_content_set(e, slash1 + 1, sha1, mode);
1092 e->versions[1].mode = mode;
1093 hashcpy(e->versions[1].sha1, sha1);
1095 hashclr(root->versions[1].sha1);
1099 static int tree_content_remove(struct tree_entry *root, const char *p)
1101 struct tree_content *t = root->tree;
1104 struct tree_entry *e;
1106 slash1 = strchr(p, '/');
1112 for (i = 0; i < t->entry_count; i++) {
1114 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1115 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1119 if (tree_content_remove(e, slash1 + 1)) {
1120 for (n = 0; n < e->tree->entry_count; n++) {
1121 if (e->tree->entries[n]->versions[1].mode) {
1122 hashclr(root->versions[1].sha1);
1135 release_tree_content_recursive(e->tree);
1138 e->versions[1].mode = 0;
1139 hashclr(e->versions[1].sha1);
1140 hashclr(root->versions[1].sha1);
1144 static void dump_branches()
1146 static const char *msg = "fast-import";
1149 struct ref_lock *lock;
1151 for (i = 0; i < branch_table_sz; i++) {
1152 for (b = branch_table[i]; b; b = b->table_next_branch) {
1153 lock = lock_any_ref_for_update(b->name, NULL);
1154 if (!lock || write_ref_sha1(lock, b->sha1, msg) < 0)
1155 die("Can't write %s", b->name);
1160 static void dump_tags()
1162 static const char *msg = "fast-import";
1164 struct ref_lock *lock;
1165 char path[PATH_MAX];
1167 for (t = first_tag; t; t = t->next_tag) {
1168 sprintf(path, "refs/tags/%s", t->name);
1169 lock = lock_any_ref_for_update(path, NULL);
1170 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1171 die("Can't write %s", path);
1175 static void dump_marks_helper(FILE *f,
1181 for (k = 0; k < 1024; k++) {
1182 if (m->data.sets[k])
1183 dump_marks_helper(f, (base + k) << m->shift,
1187 for (k = 0; k < 1024; k++) {
1188 if (m->data.marked[k])
1189 fprintf(f, ":%lu %s\n", base + k,
1190 sha1_to_hex(m->data.marked[k]->sha1));
1195 static void dump_marks()
1199 FILE *f = fopen(mark_file, "w");
1200 dump_marks_helper(f, 0, marks);
1205 static void read_next_command()
1207 read_line(&command_buf, stdin, '\n');
1210 static void cmd_mark()
1212 if (!strncmp("mark :", command_buf.buf, 6)) {
1213 next_mark = strtoul(command_buf.buf + 6, NULL, 10);
1214 read_next_command();
1220 static void* cmd_data (size_t *size)
1226 if (strncmp("data ", command_buf.buf, 5))
1227 die("Expected 'data n' command, found: %s", command_buf.buf);
1229 length = strtoul(command_buf.buf + 5, NULL, 10);
1230 buffer = xmalloc(length);
1232 while (n < length) {
1233 size_t s = fread((char*)buffer + n, 1, length - n, stdin);
1234 if (!s && feof(stdin))
1235 die("EOF in data (%lu bytes remaining)", length - n);
1239 if (fgetc(stdin) != '\n')
1240 die("An lf did not trail the binary data as expected.");
1246 static void cmd_new_blob()
1251 read_next_command();
1255 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1259 static void unload_one_branch()
1261 while (cur_active_branches
1262 && cur_active_branches >= max_active_branches) {
1263 unsigned long min_commit = ULONG_MAX;
1264 struct branch *e, *l = NULL, *p = NULL;
1266 for (e = active_branches; e; e = e->active_next_branch) {
1267 if (e->last_commit < min_commit) {
1269 min_commit = e->last_commit;
1275 e = p->active_next_branch;
1276 p->active_next_branch = e->active_next_branch;
1278 e = active_branches;
1279 active_branches = e->active_next_branch;
1281 e->active_next_branch = NULL;
1282 if (e->branch_tree.tree) {
1283 release_tree_content_recursive(e->branch_tree.tree);
1284 e->branch_tree.tree = NULL;
1286 cur_active_branches--;
1290 static void load_branch(struct branch *b)
1292 load_tree(&b->branch_tree);
1293 b->active_next_branch = active_branches;
1294 active_branches = b;
1295 cur_active_branches++;
1296 branch_load_count++;
1299 static void file_change_m(struct branch *b)
1301 const char *p = command_buf.buf + 2;
1304 struct object_entry *oe;
1305 unsigned char sha1[20];
1309 p = get_mode(p, &mode);
1311 die("Corrupt mode: %s", command_buf.buf);
1313 case S_IFREG | 0644:
1314 case S_IFREG | 0755:
1321 die("Corrupt mode: %s", command_buf.buf);
1326 oe = find_mark(strtoul(p + 1, &x, 10));
1327 hashcpy(sha1, oe->sha1);
1330 if (get_sha1_hex(p, sha1))
1331 die("Invalid SHA1: %s", command_buf.buf);
1332 oe = find_object(sha1);
1336 die("Missing space after SHA1: %s", command_buf.buf);
1338 p_uq = unquote_c_style(p, &endp);
1341 die("Garbage after path in: %s", command_buf.buf);
1346 if (oe->type != OBJ_BLOB)
1347 die("Not a blob (actually a %s): %s",
1348 command_buf.buf, type_names[oe->type]);
1350 if (sha1_object_info(sha1, type, NULL))
1351 die("Blob not found: %s", command_buf.buf);
1352 if (strcmp(blob_type, type))
1353 die("Not a blob (actually a %s): %s",
1354 command_buf.buf, type);
1357 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1363 static void file_change_d(struct branch *b)
1365 const char *p = command_buf.buf + 2;
1369 p_uq = unquote_c_style(p, &endp);
1372 die("Garbage after path in: %s", command_buf.buf);
1375 tree_content_remove(&b->branch_tree, p);
1380 static void cmd_from(struct branch *b)
1382 const char *from, *endp;
1386 if (strncmp("from ", command_buf.buf, 5))
1390 die("Can't reinitailize branch %s", b->name);
1392 from = strchr(command_buf.buf, ' ') + 1;
1393 str_uq = unquote_c_style(from, &endp);
1396 die("Garbage after string in: %s", command_buf.buf);
1400 s = lookup_branch(from);
1402 die("Can't create a branch from itself: %s", b->name);
1404 unsigned char *t = s->branch_tree.versions[1].sha1;
1405 hashcpy(b->sha1, s->sha1);
1406 hashcpy(b->branch_tree.versions[0].sha1, t);
1407 hashcpy(b->branch_tree.versions[1].sha1, t);
1408 } else if (*from == ':') {
1409 unsigned long idnum = strtoul(from + 1, NULL, 10);
1410 struct object_entry *oe = find_mark(idnum);
1413 if (oe->type != OBJ_COMMIT)
1414 die("Mark :%lu not a commit", idnum);
1415 hashcpy(b->sha1, oe->sha1);
1416 buf = gfi_unpack_entry(oe, &size);
1417 if (!buf || size < 46)
1418 die("Not a valid commit: %s", from);
1419 if (memcmp("tree ", buf, 5)
1420 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1421 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1423 hashcpy(b->branch_tree.versions[0].sha1,
1424 b->branch_tree.versions[1].sha1);
1425 } else if (!get_sha1(from, b->sha1)) {
1426 if (is_null_sha1(b->sha1)) {
1427 hashclr(b->branch_tree.versions[0].sha1);
1428 hashclr(b->branch_tree.versions[1].sha1);
1433 buf = read_object_with_reference(b->sha1,
1434 type_names[OBJ_COMMIT], &size, b->sha1);
1435 if (!buf || size < 46)
1436 die("Not a valid commit: %s", from);
1437 if (memcmp("tree ", buf, 5)
1438 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1439 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1441 hashcpy(b->branch_tree.versions[0].sha1,
1442 b->branch_tree.versions[1].sha1);
1445 die("Invalid ref name or SHA1 expression: %s", from);
1447 read_next_command();
1450 static struct hash_list* cmd_merge(unsigned int *count)
1452 struct hash_list *list = NULL, *n, *e;
1453 const char *from, *endp;
1458 while (!strncmp("merge ", command_buf.buf, 6)) {
1459 from = strchr(command_buf.buf, ' ') + 1;
1460 str_uq = unquote_c_style(from, &endp);
1463 die("Garbage after string in: %s", command_buf.buf);
1467 n = xmalloc(sizeof(*n));
1468 s = lookup_branch(from);
1470 hashcpy(n->sha1, s->sha1);
1471 else if (*from == ':') {
1472 unsigned long idnum = strtoul(from + 1, NULL, 10);
1473 struct object_entry *oe = find_mark(idnum);
1474 if (oe->type != OBJ_COMMIT)
1475 die("Mark :%lu not a commit", idnum);
1476 hashcpy(n->sha1, oe->sha1);
1477 } else if (get_sha1(from, n->sha1))
1478 die("Invalid ref name or SHA1 expression: %s", from);
1487 read_next_command();
1492 static void cmd_new_commit()
1500 char *author = NULL;
1501 char *committer = NULL;
1502 struct hash_list *merge_list = NULL;
1503 unsigned int merge_count;
1505 /* Obtain the branch name from the rest of our command */
1506 sp = strchr(command_buf.buf, ' ') + 1;
1507 str_uq = unquote_c_style(sp, &endp);
1510 die("Garbage after ref in: %s", command_buf.buf);
1513 b = lookup_branch(sp);
1519 read_next_command();
1521 if (!strncmp("author ", command_buf.buf, 7)) {
1522 author = strdup(command_buf.buf);
1523 read_next_command();
1525 if (!strncmp("committer ", command_buf.buf, 10)) {
1526 committer = strdup(command_buf.buf);
1527 read_next_command();
1530 die("Expected committer but didn't get one");
1531 msg = cmd_data(&msglen);
1532 read_next_command();
1534 merge_list = cmd_merge(&merge_count);
1536 /* ensure the branch is active/loaded */
1537 if (!b->branch_tree.tree || !max_active_branches) {
1538 unload_one_branch();
1544 if (1 == command_buf.len)
1546 else if (!strncmp("M ", command_buf.buf, 2))
1548 else if (!strncmp("D ", command_buf.buf, 2))
1551 die("Unsupported file_change: %s", command_buf.buf);
1552 read_next_command();
1555 /* build the tree and the commit */
1556 store_tree(&b->branch_tree);
1557 hashcpy(b->branch_tree.versions[0].sha1,
1558 b->branch_tree.versions[1].sha1);
1559 size_dbuf(&new_data, 97 + msglen
1562 ? strlen(author) + strlen(committer)
1563 : 2 * strlen(committer)));
1564 sp = new_data.buffer;
1565 sp += sprintf(sp, "tree %s\n",
1566 sha1_to_hex(b->branch_tree.versions[1].sha1));
1567 if (!is_null_sha1(b->sha1))
1568 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1569 while (merge_list) {
1570 struct hash_list *next = merge_list->next;
1571 sp += sprintf(sp, "parent %s\n", sha1_to_hex(merge_list->sha1));
1576 sp += sprintf(sp, "%s\n", author);
1578 sp += sprintf(sp, "author %s\n", committer + 10);
1579 sp += sprintf(sp, "%s\n\n", committer);
1580 memcpy(sp, msg, msglen);
1587 store_object(OBJ_COMMIT,
1588 new_data.buffer, sp - (char*)new_data.buffer,
1589 NULL, b->sha1, next_mark);
1590 b->last_commit = object_count_by_type[OBJ_COMMIT];
1593 int need_dq = quote_c_style(b->name, NULL, NULL, 0);
1594 fprintf(branch_log, "commit ");
1596 fputc('"', branch_log);
1597 quote_c_style(b->name, NULL, branch_log, 0);
1598 fputc('"', branch_log);
1600 fprintf(branch_log, "%s", b->name);
1601 fprintf(branch_log," :%lu %s\n",next_mark,sha1_to_hex(b->sha1));
1605 static void cmd_new_tag()
1616 unsigned long from_mark = 0;
1617 unsigned char sha1[20];
1619 /* Obtain the new tag name from the rest of our command */
1620 sp = strchr(command_buf.buf, ' ') + 1;
1621 str_uq = unquote_c_style(sp, &endp);
1624 die("Garbage after tag name in: %s", command_buf.buf);
1627 t = pool_alloc(sizeof(struct tag));
1629 t->name = pool_strdup(sp);
1631 last_tag->next_tag = t;
1637 read_next_command();
1640 if (strncmp("from ", command_buf.buf, 5))
1641 die("Expected from command, got %s", command_buf.buf);
1643 from = strchr(command_buf.buf, ' ') + 1;
1644 str_uq = unquote_c_style(from, &endp);
1647 die("Garbage after string in: %s", command_buf.buf);
1651 s = lookup_branch(from);
1653 hashcpy(sha1, s->sha1);
1654 } else if (*from == ':') {
1655 from_mark = strtoul(from + 1, NULL, 10);
1656 struct object_entry *oe = find_mark(from_mark);
1657 if (oe->type != OBJ_COMMIT)
1658 die("Mark :%lu not a commit", from_mark);
1659 hashcpy(sha1, oe->sha1);
1660 } else if (!get_sha1(from, sha1)) {
1664 buf = read_object_with_reference(sha1,
1665 type_names[OBJ_COMMIT], &size, sha1);
1666 if (!buf || size < 46)
1667 die("Not a valid commit: %s", from);
1670 die("Invalid ref name or SHA1 expression: %s", from);
1674 read_next_command();
1677 if (strncmp("tagger ", command_buf.buf, 7))
1678 die("Expected tagger command, got %s", command_buf.buf);
1679 tagger = strdup(command_buf.buf);
1681 /* tag payload/message */
1682 read_next_command();
1683 msg = cmd_data(&msglen);
1685 /* build the tag object */
1686 size_dbuf(&new_data, 67+strlen(t->name)+strlen(tagger)+msglen);
1687 sp = new_data.buffer;
1688 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
1689 sp += sprintf(sp, "type %s\n", type_names[OBJ_COMMIT]);
1690 sp += sprintf(sp, "tag %s\n", t->name);
1691 sp += sprintf(sp, "%s\n\n", tagger);
1692 memcpy(sp, msg, msglen);
1697 store_object(OBJ_TAG, new_data.buffer, sp - (char*)new_data.buffer,
1701 int need_dq = quote_c_style(t->name, NULL, NULL, 0);
1702 fprintf(branch_log, "tag ");
1704 fputc('"', branch_log);
1705 quote_c_style(t->name, NULL, branch_log, 0);
1706 fputc('"', branch_log);
1708 fprintf(branch_log, "%s", t->name);
1709 fprintf(branch_log," :%lu %s\n",from_mark,sha1_to_hex(t->sha1));
1713 static void cmd_reset_branch()
1720 /* Obtain the branch name from the rest of our command */
1721 sp = strchr(command_buf.buf, ' ') + 1;
1722 str_uq = unquote_c_style(sp, &endp);
1725 die("Garbage after ref in: %s", command_buf.buf);
1728 b = lookup_branch(sp);
1731 if (b->branch_tree.tree) {
1732 release_tree_content_recursive(b->branch_tree.tree);
1733 b->branch_tree.tree = NULL;
1740 read_next_command();
1744 static void cmd_checkpoint()
1750 read_next_command();
1753 static const char fast_import_usage[] =
1754 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file] [--branch-log=log] temp.pack";
1756 int main(int argc, const char **argv)
1759 unsigned long est_obj_cnt = object_entry_alloc;
1760 unsigned long duplicate_count;
1763 git_config(git_default_config);
1765 for (i = 1; i < argc; i++) {
1766 const char *a = argv[i];
1768 if (*a != '-' || !strcmp(a, "--"))
1770 else if (!strncmp(a, "--objects=", 10))
1771 est_obj_cnt = strtoul(a + 10, NULL, 0);
1772 else if (!strncmp(a, "--depth=", 8))
1773 max_depth = strtoul(a + 8, NULL, 0);
1774 else if (!strncmp(a, "--active-branches=", 18))
1775 max_active_branches = strtoul(a + 18, NULL, 0);
1776 else if (!strncmp(a, "--export-marks=", 15))
1778 else if (!strncmp(a, "--branch-log=", 13)) {
1779 branch_log = fopen(a + 13, "w");
1781 die("Can't create %s: %s", a + 13, strerror(errno));
1784 die("unknown option %s", a);
1787 usage(fast_import_usage);
1788 base_name = argv[i];
1790 alloc_objects(est_obj_cnt);
1791 strbuf_init(&command_buf);
1793 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
1794 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
1795 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
1796 marks = pool_calloc(1, sizeof(struct mark_set));
1800 read_next_command();
1801 if (command_buf.eof)
1803 else if (!strcmp("blob", command_buf.buf))
1805 else if (!strncmp("commit ", command_buf.buf, 7))
1807 else if (!strncmp("tag ", command_buf.buf, 4))
1809 else if (!strncmp("reset ", command_buf.buf, 6))
1811 else if (!strcmp("checkpoint", command_buf.buf))
1814 die("Unsupported command: %s", command_buf.buf);
1824 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
1825 duplicate_count += duplicate_count_by_type[i];
1827 fprintf(stderr, "%s statistics:\n", argv[0]);
1828 fprintf(stderr, "---------------------------------------------------------------------\n");
1829 fprintf(stderr, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count, alloc_count - est_obj_cnt);
1830 fprintf(stderr, "Total objects: %10lu (%10lu duplicates )\n", object_count, duplicate_count);
1831 fprintf(stderr, " blobs : %10lu (%10lu duplicates %10lu deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
1832 fprintf(stderr, " trees : %10lu (%10lu duplicates %10lu deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
1833 fprintf(stderr, " commits: %10lu (%10lu duplicates %10lu deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
1834 fprintf(stderr, " tags : %10lu (%10lu duplicates %10lu deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
1835 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
1836 fprintf(stderr, " marks: %10u (%10lu unique )\n", (1 << marks->shift) * 1024, marks_set_count);
1837 fprintf(stderr, " atoms: %10u\n", atom_cnt);
1838 fprintf(stderr, "Memory total: %10lu KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
1839 fprintf(stderr, " pools: %10lu KiB\n", total_allocd/1024);
1840 fprintf(stderr, " objects: %10lu KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
1841 fprintf(stderr, "---------------------------------------------------------------------\n");
1843 fprintf(stderr, "---------------------------------------------------------------------\n");
1844 fprintf(stderr, "\n");