f94f307ee6028ba70edb9d8d1f9014a5dbed275f
1 /*
2 Format of STDIN stream:
4 stream ::= cmd*;
6 cmd ::= new_blob
7 | new_commit
8 | new_tag
9 | reset_branch
10 ;
12 new_blob ::= 'blob' lf
13 mark?
14 file_content;
15 file_content ::= data;
17 new_commit ::= 'commit' sp ref_str lf
18 mark?
19 ('author' sp name '<' email '>' ts tz lf)?
20 'committer' sp name '<' email '>' ts tz lf
21 commit_msg
22 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
23 file_change*
24 lf;
25 commit_msg ::= data;
27 file_change ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf
28 | 'D' sp path_str lf
29 ;
30 mode ::= '644' | '755';
32 new_tag ::= 'tag' sp tag_str lf
33 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
34 'tagger' sp name '<' email '>' ts tz lf
35 tag_msg;
36 tag_msg ::= data;
38 reset_branch ::= 'reset' sp ref_str lf;
40 # note: the first idnum in a stream should be 1 and subsequent
41 # idnums should not have gaps between values as this will cause
42 # the stream parser to reserve space for the gapped values. An
43 # idnum can be updated in the future to a new object by issuing
44 # a new mark directive with the old idnum.
45 #
46 mark ::= 'mark' sp idnum lf;
48 # note: declen indicates the length of binary_data in bytes.
49 # declen does not include the lf preceeding or trailing the
50 # binary data.
51 #
52 data ::= 'data' sp declen lf
53 binary_data
54 lf;
56 # note: quoted strings are C-style quoting supporting \c for
57 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
58 # is the signed byte value in octal. Note that the only
59 # characters which must actually be escaped to protect the
60 # stream formatting is: \, " and LF. Otherwise these values
61 # are UTF8.
62 #
63 ref_str ::= ref | '"' quoted(ref) '"' ;
64 sha1exp_str ::= sha1exp | '"' quoted(sha1exp) '"' ;
65 tag_str ::= tag | '"' quoted(tag) '"' ;
66 path_str ::= path | '"' quoted(path) '"' ;
68 declen ::= # unsigned 32 bit value, ascii base10 notation;
69 binary_data ::= # file content, not interpreted;
71 sp ::= # ASCII space character;
72 lf ::= # ASCII newline (LF) character;
74 # note: a colon (':') must precede the numerical value assigned to
75 # an idnum. This is to distinguish it from a ref or tag name as
76 # GIT does not permit ':' in ref or tag strings.
77 #
78 idnum ::= ':' declen;
79 path ::= # GIT style file path, e.g. "a/b/c";
80 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
81 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
82 sha1exp ::= # Any valid GIT SHA1 expression;
83 hexsha1 ::= # SHA1 in hexadecimal format;
85 # note: name and email are UTF8 strings, however name must not
86 # contain '<' or lf and email must not contain any of the
87 # following: '<', '>', lf.
88 #
89 name ::= # valid GIT author/committer name;
90 email ::= # valid GIT author/committer email;
91 ts ::= # time since the epoch in seconds, ascii base10 notation;
92 tz ::= # GIT style timezone;
93 */
95 #include "builtin.h"
96 #include "cache.h"
97 #include "object.h"
98 #include "blob.h"
99 #include "tree.h"
100 #include "delta.h"
101 #include "pack.h"
102 #include "refs.h"
103 #include "csum-file.h"
104 #include "strbuf.h"
105 #include "quote.h"
107 struct object_entry
108 {
109 struct object_entry *next;
110 enum object_type type;
111 unsigned long offset;
112 unsigned char sha1[20];
113 };
115 struct object_entry_pool
116 {
117 struct object_entry_pool *next_pool;
118 struct object_entry *next_free;
119 struct object_entry *end;
120 struct object_entry entries[FLEX_ARRAY]; /* more */
121 };
123 struct mark_set
124 {
125 int shift;
126 union {
127 struct object_entry *marked[1024];
128 struct mark_set *sets[1024];
129 } data;
130 };
132 struct last_object
133 {
134 void *data;
135 unsigned long len;
136 unsigned int depth;
137 int no_free;
138 unsigned char sha1[20];
139 };
141 struct mem_pool
142 {
143 struct mem_pool *next_pool;
144 char *next_free;
145 char *end;
146 char space[FLEX_ARRAY]; /* more */
147 };
149 struct atom_str
150 {
151 struct atom_str *next_atom;
152 int str_len;
153 char str_dat[FLEX_ARRAY]; /* more */
154 };
156 struct tree_content;
157 struct tree_entry
158 {
159 struct tree_content *tree;
160 struct atom_str* name;
161 struct tree_entry_ms
162 {
163 unsigned int mode;
164 unsigned char sha1[20];
165 } versions[2];
166 };
168 struct tree_content
169 {
170 unsigned int entry_capacity; /* must match avail_tree_content */
171 unsigned int entry_count;
172 unsigned int delta_depth;
173 struct tree_entry *entries[FLEX_ARRAY]; /* more */
174 };
176 struct avail_tree_content
177 {
178 unsigned int entry_capacity; /* must match tree_content */
179 struct avail_tree_content *next_avail;
180 };
182 struct branch
183 {
184 struct branch *table_next_branch;
185 struct branch *active_next_branch;
186 const char *name;
187 unsigned long last_commit;
188 struct tree_entry branch_tree;
189 unsigned char sha1[20];
190 };
192 struct tag
193 {
194 struct tag *next_tag;
195 const char *name;
196 unsigned char sha1[20];
197 };
199 struct dbuf
200 {
201 void *buffer;
202 size_t capacity;
203 };
206 /* Stats and misc. counters */
207 static unsigned long max_depth = 10;
208 static unsigned long alloc_count;
209 static unsigned long branch_count;
210 static unsigned long branch_load_count;
211 static unsigned long remap_count;
212 static unsigned long object_count;
213 static unsigned long duplicate_count;
214 static unsigned long marks_set_count;
215 static unsigned long object_count_by_type[9];
216 static unsigned long duplicate_count_by_type[9];
217 static unsigned long delta_count_by_type[9];
219 /* Memory pools */
220 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
221 static size_t total_allocd;
222 static struct mem_pool *mem_pool;
224 /* Atom management */
225 static unsigned int atom_table_sz = 4451;
226 static unsigned int atom_cnt;
227 static struct atom_str **atom_table;
229 /* The .pack file being generated */
230 static int pack_fd;
231 static unsigned long pack_size;
232 static unsigned char pack_sha1[20];
233 static unsigned char* pack_base;
234 static unsigned long pack_moff;
235 static unsigned long pack_mlen = 128*1024*1024;
236 static unsigned long page_size;
238 /* Table of objects we've written. */
239 static unsigned int object_entry_alloc = 5000;
240 static struct object_entry_pool *blocks;
241 static struct object_entry *object_table[1 << 16];
242 static struct mark_set *marks;
243 static const char* mark_file;
245 /* Our last blob */
246 static struct last_object last_blob;
248 /* Tree management */
249 static unsigned int tree_entry_alloc = 1000;
250 static void *avail_tree_entry;
251 static unsigned int avail_tree_table_sz = 100;
252 static struct avail_tree_content **avail_tree_table;
253 static struct dbuf old_tree;
254 static struct dbuf new_tree;
256 /* Branch data */
257 static unsigned long max_active_branches = 5;
258 static unsigned long cur_active_branches;
259 static unsigned long branch_table_sz = 1039;
260 static struct branch **branch_table;
261 static struct branch *active_branches;
263 /* Tag data */
264 static struct tag *first_tag;
265 static struct tag *last_tag;
267 /* Input stream parsing */
268 static struct strbuf command_buf;
269 static unsigned long next_mark;
270 static struct dbuf new_data;
271 static FILE* branch_log;
274 static void alloc_objects(int cnt)
275 {
276 struct object_entry_pool *b;
278 b = xmalloc(sizeof(struct object_entry_pool)
279 + cnt * sizeof(struct object_entry));
280 b->next_pool = blocks;
281 b->next_free = b->entries;
282 b->end = b->entries + cnt;
283 blocks = b;
284 alloc_count += cnt;
285 }
287 static struct object_entry* new_object(unsigned char *sha1)
288 {
289 struct object_entry *e;
291 if (blocks->next_free == blocks->end)
292 alloc_objects(object_entry_alloc);
294 e = blocks->next_free++;
295 hashcpy(e->sha1, sha1);
296 return e;
297 }
299 static struct object_entry* find_object(unsigned char *sha1)
300 {
301 unsigned int h = sha1[0] << 8 | sha1[1];
302 struct object_entry *e;
303 for (e = object_table[h]; e; e = e->next)
304 if (!hashcmp(sha1, e->sha1))
305 return e;
306 return NULL;
307 }
309 static struct object_entry* insert_object(unsigned char *sha1)
310 {
311 unsigned int h = sha1[0] << 8 | sha1[1];
312 struct object_entry *e = object_table[h];
313 struct object_entry *p = NULL;
315 while (e) {
316 if (!hashcmp(sha1, e->sha1))
317 return e;
318 p = e;
319 e = e->next;
320 }
322 e = new_object(sha1);
323 e->next = NULL;
324 e->offset = 0;
325 if (p)
326 p->next = e;
327 else
328 object_table[h] = e;
329 return e;
330 }
332 static unsigned int hc_str(const char *s, size_t len)
333 {
334 unsigned int r = 0;
335 while (len-- > 0)
336 r = r * 31 + *s++;
337 return r;
338 }
340 static void* pool_alloc(size_t len)
341 {
342 struct mem_pool *p;
343 void *r;
345 for (p = mem_pool; p; p = p->next_pool)
346 if ((p->end - p->next_free >= len))
347 break;
349 if (!p) {
350 if (len >= (mem_pool_alloc/2)) {
351 total_allocd += len;
352 return xmalloc(len);
353 }
354 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
355 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
356 p->next_pool = mem_pool;
357 p->next_free = p->space;
358 p->end = p->next_free + mem_pool_alloc;
359 mem_pool = p;
360 }
362 r = p->next_free;
363 /* round out to a pointer alignment */
364 if (len & (sizeof(void*) - 1))
365 len += sizeof(void*) - (len & (sizeof(void*) - 1));
366 p->next_free += len;
367 return r;
368 }
370 static void* pool_calloc(size_t count, size_t size)
371 {
372 size_t len = count * size;
373 void *r = pool_alloc(len);
374 memset(r, 0, len);
375 return r;
376 }
378 static char* pool_strdup(const char *s)
379 {
380 char *r = pool_alloc(strlen(s) + 1);
381 strcpy(r, s);
382 return r;
383 }
385 static void size_dbuf(struct dbuf *b, size_t maxlen)
386 {
387 if (b->buffer) {
388 if (b->capacity >= maxlen)
389 return;
390 free(b->buffer);
391 }
392 b->capacity = ((maxlen / 1024) + 1) * 1024;
393 b->buffer = xmalloc(b->capacity);
394 }
396 static void insert_mark(unsigned long idnum, struct object_entry *oe)
397 {
398 struct mark_set *s = marks;
399 while ((idnum >> s->shift) >= 1024) {
400 s = pool_calloc(1, sizeof(struct mark_set));
401 s->shift = marks->shift + 10;
402 s->data.sets[0] = marks;
403 marks = s;
404 }
405 while (s->shift) {
406 unsigned long i = idnum >> s->shift;
407 idnum -= i << s->shift;
408 if (!s->data.sets[i]) {
409 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
410 s->data.sets[i]->shift = s->shift - 10;
411 }
412 s = s->data.sets[i];
413 }
414 if (!s->data.marked[idnum])
415 marks_set_count++;
416 s->data.marked[idnum] = oe;
417 }
419 static struct object_entry* find_mark(unsigned long idnum)
420 {
421 unsigned long orig_idnum = idnum;
422 struct mark_set *s = marks;
423 struct object_entry *oe = NULL;
424 if ((idnum >> s->shift) < 1024) {
425 while (s && s->shift) {
426 unsigned long i = idnum >> s->shift;
427 idnum -= i << s->shift;
428 s = s->data.sets[i];
429 }
430 if (s)
431 oe = s->data.marked[idnum];
432 }
433 if (!oe)
434 die("mark :%lu not declared", orig_idnum);
435 return oe;
436 }
438 static struct atom_str* to_atom(const char *s, size_t len)
439 {
440 unsigned int hc = hc_str(s, len) % atom_table_sz;
441 struct atom_str *c;
443 for (c = atom_table[hc]; c; c = c->next_atom)
444 if (c->str_len == len && !strncmp(s, c->str_dat, len))
445 return c;
447 c = pool_alloc(sizeof(struct atom_str) + len + 1);
448 c->str_len = len;
449 strncpy(c->str_dat, s, len);
450 c->str_dat[len] = 0;
451 c->next_atom = atom_table[hc];
452 atom_table[hc] = c;
453 atom_cnt++;
454 return c;
455 }
457 static struct branch* lookup_branch(const char *name)
458 {
459 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
460 struct branch *b;
462 for (b = branch_table[hc]; b; b = b->table_next_branch)
463 if (!strcmp(name, b->name))
464 return b;
465 return NULL;
466 }
468 static struct branch* new_branch(const char *name)
469 {
470 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
471 struct branch* b = lookup_branch(name);
473 if (b)
474 die("Invalid attempt to create duplicate branch: %s", name);
475 if (check_ref_format(name))
476 die("Branch name doesn't conform to GIT standards: %s", name);
478 b = pool_calloc(1, sizeof(struct branch));
479 b->name = pool_strdup(name);
480 b->table_next_branch = branch_table[hc];
481 branch_table[hc] = b;
482 branch_count++;
483 return b;
484 }
486 static unsigned int hc_entries(unsigned int cnt)
487 {
488 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
489 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
490 }
492 static struct tree_content* new_tree_content(unsigned int cnt)
493 {
494 struct avail_tree_content *f, *l = NULL;
495 struct tree_content *t;
496 unsigned int hc = hc_entries(cnt);
498 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
499 if (f->entry_capacity >= cnt)
500 break;
502 if (f) {
503 if (l)
504 l->next_avail = f->next_avail;
505 else
506 avail_tree_table[hc] = f->next_avail;
507 } else {
508 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
509 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
510 f->entry_capacity = cnt;
511 }
513 t = (struct tree_content*)f;
514 t->entry_count = 0;
515 t->delta_depth = 0;
516 return t;
517 }
519 static void release_tree_entry(struct tree_entry *e);
520 static void release_tree_content(struct tree_content *t)
521 {
522 struct avail_tree_content *f = (struct avail_tree_content*)t;
523 unsigned int hc = hc_entries(f->entry_capacity);
524 f->next_avail = avail_tree_table[hc];
525 avail_tree_table[hc] = f;
526 }
528 static void release_tree_content_recursive(struct tree_content *t)
529 {
530 unsigned int i;
531 for (i = 0; i < t->entry_count; i++)
532 release_tree_entry(t->entries[i]);
533 release_tree_content(t);
534 }
536 static struct tree_content* grow_tree_content(
537 struct tree_content *t,
538 int amt)
539 {
540 struct tree_content *r = new_tree_content(t->entry_count + amt);
541 r->entry_count = t->entry_count;
542 r->delta_depth = t->delta_depth;
543 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
544 release_tree_content(t);
545 return r;
546 }
548 static struct tree_entry* new_tree_entry()
549 {
550 struct tree_entry *e;
552 if (!avail_tree_entry) {
553 unsigned int n = tree_entry_alloc;
554 total_allocd += n * sizeof(struct tree_entry);
555 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
556 while (n-- > 1) {
557 *((void**)e) = e + 1;
558 e++;
559 }
560 *((void**)e) = NULL;
561 }
563 e = avail_tree_entry;
564 avail_tree_entry = *((void**)e);
565 return e;
566 }
568 static void release_tree_entry(struct tree_entry *e)
569 {
570 if (e->tree)
571 release_tree_content_recursive(e->tree);
572 *((void**)e) = avail_tree_entry;
573 avail_tree_entry = e;
574 }
576 static void yread(int fd, void *buffer, size_t length)
577 {
578 ssize_t ret = 0;
579 while (ret < length) {
580 ssize_t size = xread(fd, (char *) buffer + ret, length - ret);
581 if (!size)
582 die("Read from descriptor %i: end of stream", fd);
583 if (size < 0)
584 die("Read from descriptor %i: %s", fd, strerror(errno));
585 ret += size;
586 }
587 }
589 static size_t encode_header(
590 enum object_type type,
591 size_t size,
592 unsigned char *hdr)
593 {
594 int n = 1;
595 unsigned char c;
597 if (type < OBJ_COMMIT || type > OBJ_DELTA)
598 die("bad type %d", type);
600 c = (type << 4) | (size & 15);
601 size >>= 4;
602 while (size) {
603 *hdr++ = c | 0x80;
604 c = size & 0x7f;
605 size >>= 7;
606 n++;
607 }
608 *hdr = c;
609 return n;
610 }
612 static int store_object(
613 enum object_type type,
614 void *dat,
615 size_t datlen,
616 struct last_object *last,
617 unsigned char *sha1out,
618 unsigned long mark)
619 {
620 void *out, *delta;
621 struct object_entry *e;
622 unsigned char hdr[96];
623 unsigned char sha1[20];
624 unsigned long hdrlen, deltalen;
625 SHA_CTX c;
626 z_stream s;
628 hdrlen = sprintf((char*)hdr,"%s %lu",type_names[type],datlen) + 1;
629 SHA1_Init(&c);
630 SHA1_Update(&c, hdr, hdrlen);
631 SHA1_Update(&c, dat, datlen);
632 SHA1_Final(sha1, &c);
633 if (sha1out)
634 hashcpy(sha1out, sha1);
636 e = insert_object(sha1);
637 if (mark)
638 insert_mark(mark, e);
639 if (e->offset) {
640 duplicate_count++;
641 duplicate_count_by_type[type]++;
642 return 1;
643 }
644 e->type = type;
645 e->offset = pack_size;
646 object_count++;
647 object_count_by_type[type]++;
649 if (last && last->data && last->depth < max_depth)
650 delta = diff_delta(last->data, last->len,
651 dat, datlen,
652 &deltalen, 0);
653 else
654 delta = 0;
656 memset(&s, 0, sizeof(s));
657 deflateInit(&s, zlib_compression_level);
659 if (delta) {
660 delta_count_by_type[type]++;
661 last->depth++;
662 s.next_in = delta;
663 s.avail_in = deltalen;
664 hdrlen = encode_header(OBJ_DELTA, deltalen, hdr);
665 write_or_die(pack_fd, hdr, hdrlen);
666 write_or_die(pack_fd, last->sha1, sizeof(sha1));
667 pack_size += hdrlen + sizeof(sha1);
668 } else {
669 if (last)
670 last->depth = 0;
671 s.next_in = dat;
672 s.avail_in = datlen;
673 hdrlen = encode_header(type, datlen, hdr);
674 write_or_die(pack_fd, hdr, hdrlen);
675 pack_size += hdrlen;
676 }
678 s.avail_out = deflateBound(&s, s.avail_in);
679 s.next_out = out = xmalloc(s.avail_out);
680 while (deflate(&s, Z_FINISH) == Z_OK)
681 /* nothing */;
682 deflateEnd(&s);
684 write_or_die(pack_fd, out, s.total_out);
685 pack_size += s.total_out;
687 free(out);
688 if (delta)
689 free(delta);
690 if (last) {
691 if (last->data && !last->no_free)
692 free(last->data);
693 last->data = dat;
694 last->len = datlen;
695 hashcpy(last->sha1, sha1);
696 }
697 return 0;
698 }
700 static unsigned char* map_pack(unsigned long offset, unsigned int *left)
701 {
702 if (offset >= pack_size)
703 die("object offset outside of pack file");
704 if (!pack_base
705 || offset < pack_moff
706 || (offset + 20) >= (pack_moff + pack_mlen)) {
707 if (pack_base)
708 munmap(pack_base, pack_mlen);
709 pack_moff = (offset / page_size) * page_size;
710 pack_base = mmap(NULL,pack_mlen,PROT_READ,MAP_SHARED,
711 pack_fd,pack_moff);
712 if (pack_base == MAP_FAILED)
713 die("Failed to map generated pack: %s", strerror(errno));
714 remap_count++;
715 }
716 offset -= pack_moff;
717 if (left)
718 *left = pack_mlen - offset;
719 return pack_base + offset;
720 }
722 static unsigned long unpack_object_header(unsigned long offset,
723 enum object_type *type,
724 unsigned long *sizep)
725 {
726 unsigned shift;
727 unsigned char c;
728 unsigned long size;
730 c = *map_pack(offset++, NULL);
731 *type = (c >> 4) & 7;
732 size = c & 15;
733 shift = 4;
734 while (c & 0x80) {
735 c = *map_pack(offset++, NULL);
736 size += (c & 0x7f) << shift;
737 shift += 7;
738 }
739 *sizep = size;
740 return offset;
741 }
743 static void *unpack_non_delta_entry(unsigned long o, unsigned long sz)
744 {
745 z_stream stream;
746 unsigned char *result;
748 result = xmalloc(sz + 1);
749 result[sz] = 0;
751 memset(&stream, 0, sizeof(stream));
752 stream.next_in = map_pack(o, &stream.avail_in);
753 stream.next_out = result;
754 stream.avail_out = sz;
756 inflateInit(&stream);
757 for (;;) {
758 int st = inflate(&stream, Z_FINISH);
759 if (st == Z_STREAM_END)
760 break;
761 if (st == Z_OK || st == Z_BUF_ERROR) {
762 o = stream.next_in - pack_base + pack_moff;
763 stream.next_in = map_pack(o, &stream.avail_in);
764 continue;
765 }
766 die("Error %i from zlib during inflate.", st);
767 }
768 inflateEnd(&stream);
769 if (stream.total_out != sz)
770 die("Error after inflate: sizes mismatch");
771 return result;
772 }
774 static void *unpack_entry(unsigned long offset,
775 unsigned long *sizep,
776 unsigned int *delta_depth);
778 static void *unpack_delta_entry(unsigned long offset,
779 unsigned long delta_size,
780 unsigned long *sizep,
781 unsigned int *delta_depth)
782 {
783 struct object_entry *base_oe;
784 unsigned char *base_sha1;
785 void *delta_data, *base, *result;
786 unsigned long base_size, result_size;
788 base_sha1 = map_pack(offset, NULL);
789 base_oe = find_object(base_sha1);
790 if (!base_oe)
791 die("I'm broken; I can't find a base I know must be here.");
792 base = unpack_entry(base_oe->offset, &base_size, delta_depth);
793 delta_data = unpack_non_delta_entry(offset + 20, delta_size);
794 result = patch_delta(base, base_size,
795 delta_data, delta_size,
796 &result_size);
797 if (!result)
798 die("failed to apply delta");
799 free(delta_data);
800 free(base);
801 *sizep = result_size;
802 (*delta_depth)++;
803 return result;
804 }
806 static void *unpack_entry(unsigned long offset,
807 unsigned long *sizep,
808 unsigned int *delta_depth)
809 {
810 unsigned long size;
811 enum object_type kind;
813 offset = unpack_object_header(offset, &kind, &size);
814 switch (kind) {
815 case OBJ_DELTA:
816 return unpack_delta_entry(offset, size, sizep, delta_depth);
817 case OBJ_COMMIT:
818 case OBJ_TREE:
819 case OBJ_BLOB:
820 case OBJ_TAG:
821 *sizep = size;
822 *delta_depth = 0;
823 return unpack_non_delta_entry(offset, size);
824 default:
825 die("I created an object I can't read!");
826 }
827 }
829 static const char *get_mode(const char *str, unsigned int *modep)
830 {
831 unsigned char c;
832 unsigned int mode = 0;
834 while ((c = *str++) != ' ') {
835 if (c < '0' || c > '7')
836 return NULL;
837 mode = (mode << 3) + (c - '0');
838 }
839 *modep = mode;
840 return str;
841 }
843 static void load_tree(struct tree_entry *root)
844 {
845 unsigned char* sha1 = root->versions[1].sha1;
846 struct object_entry *myoe;
847 struct tree_content *t;
848 unsigned long size;
849 char *buf;
850 const char *c;
852 root->tree = t = new_tree_content(8);
853 if (is_null_sha1(sha1))
854 return;
856 myoe = find_object(sha1);
857 if (myoe) {
858 if (myoe->type != OBJ_TREE)
859 die("Not a tree: %s", sha1_to_hex(sha1));
860 buf = unpack_entry(myoe->offset, &size, &t->delta_depth);
861 } else {
862 char type[20];
863 buf = read_sha1_file(sha1, type, &size);
864 if (!buf || strcmp(type, tree_type))
865 die("Can't load tree %s", sha1_to_hex(sha1));
866 }
868 c = buf;
869 while (c != (buf + size)) {
870 struct tree_entry *e = new_tree_entry();
872 if (t->entry_count == t->entry_capacity)
873 root->tree = t = grow_tree_content(t, 8);
874 t->entries[t->entry_count++] = e;
876 e->tree = NULL;
877 c = get_mode(c, &e->versions[1].mode);
878 if (!c)
879 die("Corrupt mode in %s", sha1_to_hex(sha1));
880 e->versions[0].mode = e->versions[1].mode;
881 e->name = to_atom(c, strlen(c));
882 c += e->name->str_len + 1;
883 hashcpy(e->versions[0].sha1, (unsigned char*)c);
884 hashcpy(e->versions[1].sha1, (unsigned char*)c);
885 c += 20;
886 }
887 free(buf);
888 }
890 static int tecmp0 (const void *_a, const void *_b)
891 {
892 struct tree_entry *a = *((struct tree_entry**)_a);
893 struct tree_entry *b = *((struct tree_entry**)_b);
894 return base_name_compare(
895 a->name->str_dat, a->name->str_len, a->versions[0].mode,
896 b->name->str_dat, b->name->str_len, b->versions[0].mode);
897 }
899 static int tecmp1 (const void *_a, const void *_b)
900 {
901 struct tree_entry *a = *((struct tree_entry**)_a);
902 struct tree_entry *b = *((struct tree_entry**)_b);
903 return base_name_compare(
904 a->name->str_dat, a->name->str_len, a->versions[1].mode,
905 b->name->str_dat, b->name->str_len, b->versions[1].mode);
906 }
908 static void mktree(struct tree_content *t,
909 int v,
910 unsigned long *szp,
911 struct dbuf *b)
912 {
913 size_t maxlen = 0;
914 unsigned int i;
915 char *c;
917 if (!v)
918 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
919 else
920 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
922 for (i = 0; i < t->entry_count; i++) {
923 if (t->entries[i]->versions[v].mode)
924 maxlen += t->entries[i]->name->str_len + 34;
925 }
927 size_dbuf(b, maxlen);
928 c = b->buffer;
929 for (i = 0; i < t->entry_count; i++) {
930 struct tree_entry *e = t->entries[i];
931 if (!e->versions[v].mode)
932 continue;
933 c += sprintf(c, "%o", e->versions[v].mode);
934 *c++ = ' ';
935 strcpy(c, e->name->str_dat);
936 c += e->name->str_len + 1;
937 hashcpy((unsigned char*)c, e->versions[v].sha1);
938 c += 20;
939 }
940 *szp = c - (char*)b->buffer;
941 }
943 static void store_tree(struct tree_entry *root)
944 {
945 struct tree_content *t = root->tree;
946 unsigned int i, j, del;
947 unsigned long new_len;
948 struct last_object lo;
950 if (!is_null_sha1(root->versions[1].sha1))
951 return;
953 for (i = 0; i < t->entry_count; i++) {
954 if (t->entries[i]->tree)
955 store_tree(t->entries[i]);
956 }
958 if (is_null_sha1(root->versions[0].sha1)
959 || !find_object(root->versions[0].sha1)) {
960 lo.data = NULL;
961 lo.depth = 0;
962 } else {
963 mktree(t, 0, &lo.len, &old_tree);
964 lo.data = old_tree.buffer;
965 lo.depth = t->delta_depth;
966 lo.no_free = 1;
967 hashcpy(lo.sha1, root->versions[0].sha1);
968 }
969 mktree(t, 1, &new_len, &new_tree);
971 store_object(OBJ_TREE, new_tree.buffer, new_len,
972 &lo, root->versions[1].sha1, 0);
974 t->delta_depth = lo.depth;
975 hashcpy(root->versions[0].sha1, root->versions[1].sha1);
976 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
977 struct tree_entry *e = t->entries[i];
978 if (e->versions[1].mode) {
979 e->versions[0].mode = e->versions[1].mode;
980 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
981 t->entries[j++] = e;
982 } else {
983 release_tree_entry(e);
984 del++;
985 }
986 }
987 t->entry_count -= del;
988 }
990 static int tree_content_set(
991 struct tree_entry *root,
992 const char *p,
993 const unsigned char *sha1,
994 const unsigned int mode)
995 {
996 struct tree_content *t = root->tree;
997 const char *slash1;
998 unsigned int i, n;
999 struct tree_entry *e;
1001 slash1 = strchr(p, '/');
1002 if (slash1)
1003 n = slash1 - p;
1004 else
1005 n = strlen(p);
1007 for (i = 0; i < t->entry_count; i++) {
1008 e = t->entries[i];
1009 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1010 if (!slash1) {
1011 if (e->versions[1].mode == mode
1012 && !hashcmp(e->versions[1].sha1, sha1))
1013 return 0;
1014 e->versions[1].mode = mode;
1015 hashcpy(e->versions[1].sha1, sha1);
1016 if (e->tree) {
1017 release_tree_content_recursive(e->tree);
1018 e->tree = NULL;
1019 }
1020 hashclr(root->versions[1].sha1);
1021 return 1;
1022 }
1023 if (!S_ISDIR(e->versions[1].mode)) {
1024 e->tree = new_tree_content(8);
1025 e->versions[1].mode = S_IFDIR;
1026 }
1027 if (!e->tree)
1028 load_tree(e);
1029 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
1030 hashclr(root->versions[1].sha1);
1031 return 1;
1032 }
1033 return 0;
1034 }
1035 }
1037 if (t->entry_count == t->entry_capacity)
1038 root->tree = t = grow_tree_content(t, 8);
1039 e = new_tree_entry();
1040 e->name = to_atom(p, n);
1041 e->versions[0].mode = 0;
1042 hashclr(e->versions[0].sha1);
1043 t->entries[t->entry_count++] = e;
1044 if (slash1) {
1045 e->tree = new_tree_content(8);
1046 e->versions[1].mode = S_IFDIR;
1047 tree_content_set(e, slash1 + 1, sha1, mode);
1048 } else {
1049 e->tree = NULL;
1050 e->versions[1].mode = mode;
1051 hashcpy(e->versions[1].sha1, sha1);
1052 }
1053 hashclr(root->versions[1].sha1);
1054 return 1;
1055 }
1057 static int tree_content_remove(struct tree_entry *root, const char *p)
1058 {
1059 struct tree_content *t = root->tree;
1060 const char *slash1;
1061 unsigned int i, n;
1062 struct tree_entry *e;
1064 slash1 = strchr(p, '/');
1065 if (slash1)
1066 n = slash1 - p;
1067 else
1068 n = strlen(p);
1070 for (i = 0; i < t->entry_count; i++) {
1071 e = t->entries[i];
1072 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1073 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1074 goto del_entry;
1075 if (!e->tree)
1076 load_tree(e);
1077 if (tree_content_remove(e, slash1 + 1)) {
1078 if (!e->tree->entry_count)
1079 goto del_entry;
1080 hashclr(root->versions[1].sha1);
1081 return 1;
1082 }
1083 return 0;
1084 }
1085 }
1086 return 0;
1088 del_entry:
1089 if (e->tree) {
1090 release_tree_content_recursive(e->tree);
1091 e->tree = NULL;
1092 }
1093 e->versions[1].mode = 0;
1094 hashclr(e->versions[1].sha1);
1095 hashclr(root->versions[1].sha1);
1096 return 1;
1097 }
1099 static void init_pack_header()
1100 {
1101 struct pack_header hdr;
1103 hdr.hdr_signature = htonl(PACK_SIGNATURE);
1104 hdr.hdr_version = htonl(2);
1105 hdr.hdr_entries = 0;
1107 write_or_die(pack_fd, &hdr, sizeof(hdr));
1108 pack_size = sizeof(hdr);
1109 }
1111 static void fixup_header_footer()
1112 {
1113 SHA_CTX c;
1114 char hdr[8];
1115 unsigned long cnt;
1116 char *buf;
1117 size_t n;
1119 if (lseek(pack_fd, 0, SEEK_SET) != 0)
1120 die("Failed seeking to start: %s", strerror(errno));
1122 SHA1_Init(&c);
1123 yread(pack_fd, hdr, 8);
1124 SHA1_Update(&c, hdr, 8);
1126 cnt = htonl(object_count);
1127 SHA1_Update(&c, &cnt, 4);
1128 write_or_die(pack_fd, &cnt, 4);
1130 buf = xmalloc(128 * 1024);
1131 for (;;) {
1132 n = xread(pack_fd, buf, 128 * 1024);
1133 if (n <= 0)
1134 break;
1135 SHA1_Update(&c, buf, n);
1136 }
1137 free(buf);
1139 SHA1_Final(pack_sha1, &c);
1140 write_or_die(pack_fd, pack_sha1, sizeof(pack_sha1));
1141 }
1143 static int oecmp (const void *_a, const void *_b)
1144 {
1145 struct object_entry *a = *((struct object_entry**)_a);
1146 struct object_entry *b = *((struct object_entry**)_b);
1147 return hashcmp(a->sha1, b->sha1);
1148 }
1150 static void write_index(const char *idx_name)
1151 {
1152 struct sha1file *f;
1153 struct object_entry **idx, **c, **last;
1154 struct object_entry *e;
1155 struct object_entry_pool *o;
1156 unsigned int array[256];
1157 int i;
1159 /* Build the sorted table of object IDs. */
1160 idx = xmalloc(object_count * sizeof(struct object_entry*));
1161 c = idx;
1162 for (o = blocks; o; o = o->next_pool)
1163 for (e = o->entries; e != o->next_free; e++)
1164 *c++ = e;
1165 last = idx + object_count;
1166 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
1168 /* Generate the fan-out array. */
1169 c = idx;
1170 for (i = 0; i < 256; i++) {
1171 struct object_entry **next = c;;
1172 while (next < last) {
1173 if ((*next)->sha1[0] != i)
1174 break;
1175 next++;
1176 }
1177 array[i] = htonl(next - idx);
1178 c = next;
1179 }
1181 f = sha1create("%s", idx_name);
1182 sha1write(f, array, 256 * sizeof(int));
1183 for (c = idx; c != last; c++) {
1184 unsigned int offset = htonl((*c)->offset);
1185 sha1write(f, &offset, 4);
1186 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
1187 }
1188 sha1write(f, pack_sha1, sizeof(pack_sha1));
1189 sha1close(f, NULL, 1);
1190 free(idx);
1191 }
1193 static void dump_branches()
1194 {
1195 static const char *msg = "fast-import";
1196 unsigned int i;
1197 struct branch *b;
1198 struct ref_lock *lock;
1200 for (i = 0; i < branch_table_sz; i++) {
1201 for (b = branch_table[i]; b; b = b->table_next_branch) {
1202 lock = lock_any_ref_for_update(b->name, NULL, 0);
1203 if (!lock || write_ref_sha1(lock, b->sha1, msg) < 0)
1204 die("Can't write %s", b->name);
1205 }
1206 }
1207 }
1209 static void dump_tags()
1210 {
1211 static const char *msg = "fast-import";
1212 struct tag *t;
1213 struct ref_lock *lock;
1214 char path[PATH_MAX];
1216 for (t = first_tag; t; t = t->next_tag) {
1217 sprintf(path, "refs/tags/%s", t->name);
1218 lock = lock_any_ref_for_update(path, NULL, 0);
1219 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1220 die("Can't write %s", path);
1221 }
1222 }
1224 static void dump_marks_helper(FILE *f,
1225 unsigned long base,
1226 struct mark_set *m)
1227 {
1228 int k;
1229 if (m->shift) {
1230 for (k = 0; k < 1024; k++) {
1231 if (m->data.sets[k])
1232 dump_marks_helper(f, (base + k) << m->shift,
1233 m->data.sets[k]);
1234 }
1235 } else {
1236 for (k = 0; k < 1024; k++) {
1237 if (m->data.marked[k])
1238 fprintf(f, ":%lu %s\n", base + k,
1239 sha1_to_hex(m->data.marked[k]->sha1));
1240 }
1241 }
1242 }
1244 static void dump_marks()
1245 {
1246 if (mark_file)
1247 {
1248 FILE *f = fopen(mark_file, "w");
1249 dump_marks_helper(f, 0, marks);
1250 fclose(f);
1251 }
1252 }
1254 static void read_next_command()
1255 {
1256 read_line(&command_buf, stdin, '\n');
1257 }
1259 static void cmd_mark()
1260 {
1261 if (!strncmp("mark :", command_buf.buf, 6)) {
1262 next_mark = strtoul(command_buf.buf + 6, NULL, 10);
1263 read_next_command();
1264 }
1265 else
1266 next_mark = 0;
1267 }
1269 static void* cmd_data (size_t *size)
1270 {
1271 size_t n = 0;
1272 void *buffer;
1273 size_t length;
1275 if (strncmp("data ", command_buf.buf, 5))
1276 die("Expected 'data n' command, found: %s", command_buf.buf);
1278 length = strtoul(command_buf.buf + 5, NULL, 10);
1279 buffer = xmalloc(length);
1281 while (n < length) {
1282 size_t s = fread((char*)buffer + n, 1, length - n, stdin);
1283 if (!s && feof(stdin))
1284 die("EOF in data (%lu bytes remaining)", length - n);
1285 n += s;
1286 }
1288 if (fgetc(stdin) != '\n')
1289 die("An lf did not trail the binary data as expected.");
1291 *size = length;
1292 return buffer;
1293 }
1295 static void cmd_new_blob()
1296 {
1297 size_t l;
1298 void *d;
1300 read_next_command();
1301 cmd_mark();
1302 d = cmd_data(&l);
1304 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1305 free(d);
1306 }
1308 static void unload_one_branch()
1309 {
1310 while (cur_active_branches
1311 && cur_active_branches >= max_active_branches) {
1312 unsigned long min_commit = ULONG_MAX;
1313 struct branch *e, *l = NULL, *p = NULL;
1315 for (e = active_branches; e; e = e->active_next_branch) {
1316 if (e->last_commit < min_commit) {
1317 p = l;
1318 min_commit = e->last_commit;
1319 }
1320 l = e;
1321 }
1323 if (p) {
1324 e = p->active_next_branch;
1325 p->active_next_branch = e->active_next_branch;
1326 } else {
1327 e = active_branches;
1328 active_branches = e->active_next_branch;
1329 }
1330 e->active_next_branch = NULL;
1331 if (e->branch_tree.tree) {
1332 release_tree_content_recursive(e->branch_tree.tree);
1333 e->branch_tree.tree = NULL;
1334 }
1335 cur_active_branches--;
1336 }
1337 }
1339 static void load_branch(struct branch *b)
1340 {
1341 load_tree(&b->branch_tree);
1342 b->active_next_branch = active_branches;
1343 active_branches = b;
1344 cur_active_branches++;
1345 branch_load_count++;
1346 }
1348 static void file_change_m(struct branch *b)
1349 {
1350 const char *p = command_buf.buf + 2;
1351 char *p_uq;
1352 const char *endp;
1353 struct object_entry *oe;
1354 unsigned char sha1[20];
1355 unsigned int mode;
1356 char type[20];
1358 p = get_mode(p, &mode);
1359 if (!p)
1360 die("Corrupt mode: %s", command_buf.buf);
1361 switch (mode) {
1362 case S_IFREG | 0644:
1363 case S_IFREG | 0755:
1364 case S_IFLNK:
1365 case 0644:
1366 case 0755:
1367 /* ok */
1368 break;
1369 default:
1370 die("Corrupt mode: %s", command_buf.buf);
1371 }
1373 if (*p == ':') {
1374 char *x;
1375 oe = find_mark(strtoul(p + 1, &x, 10));
1376 p = x;
1377 } else {
1378 if (get_sha1_hex(p, sha1))
1379 die("Invalid SHA1: %s", command_buf.buf);
1380 oe = find_object(sha1);
1381 p += 40;
1382 }
1383 if (*p++ != ' ')
1384 die("Missing space after SHA1: %s", command_buf.buf);
1386 p_uq = unquote_c_style(p, &endp);
1387 if (p_uq) {
1388 if (*endp)
1389 die("Garbage after path in: %s", command_buf.buf);
1390 p = p_uq;
1391 }
1393 if (oe) {
1394 if (oe->type != OBJ_BLOB)
1395 die("Not a blob (actually a %s): %s",
1396 command_buf.buf, type_names[oe->type]);
1397 } else {
1398 if (sha1_object_info(sha1, type, NULL))
1399 die("Blob not found: %s", command_buf.buf);
1400 if (strcmp(blob_type, type))
1401 die("Not a blob (actually a %s): %s",
1402 command_buf.buf, type);
1403 }
1405 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1407 if (p_uq)
1408 free(p_uq);
1409 }
1411 static void file_change_d(struct branch *b)
1412 {
1413 const char *p = command_buf.buf + 2;
1414 char *p_uq;
1415 const char *endp;
1417 p_uq = unquote_c_style(p, &endp);
1418 if (p_uq) {
1419 if (*endp)
1420 die("Garbage after path in: %s", command_buf.buf);
1421 p = p_uq;
1422 }
1423 tree_content_remove(&b->branch_tree, p);
1424 if (p_uq)
1425 free(p_uq);
1426 }
1428 static void cmd_from(struct branch *b)
1429 {
1430 const char *from, *endp;
1431 char *str_uq;
1432 struct branch *s;
1434 if (strncmp("from ", command_buf.buf, 5))
1435 return;
1437 if (b->last_commit)
1438 die("Can't reinitailize branch %s", b->name);
1440 from = strchr(command_buf.buf, ' ') + 1;
1441 str_uq = unquote_c_style(from, &endp);
1442 if (str_uq) {
1443 if (*endp)
1444 die("Garbage after string in: %s", command_buf.buf);
1445 from = str_uq;
1446 }
1448 s = lookup_branch(from);
1449 if (b == s)
1450 die("Can't create a branch from itself: %s", b->name);
1451 else if (s) {
1452 unsigned char *t = s->branch_tree.versions[1].sha1;
1453 hashcpy(b->sha1, s->sha1);
1454 hashcpy(b->branch_tree.versions[0].sha1, t);
1455 hashcpy(b->branch_tree.versions[1].sha1, t);
1456 } else if (*from == ':') {
1457 unsigned long idnum = strtoul(from + 1, NULL, 10);
1458 struct object_entry *oe = find_mark(idnum);
1459 unsigned long size;
1460 unsigned int depth;
1461 char *buf;
1462 if (oe->type != OBJ_COMMIT)
1463 die("Mark :%lu not a commit", idnum);
1464 hashcpy(b->sha1, oe->sha1);
1465 buf = unpack_entry(oe->offset, &size, &depth);
1466 if (!buf || size < 46)
1467 die("Not a valid commit: %s", from);
1468 if (memcmp("tree ", buf, 5)
1469 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1470 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1471 free(buf);
1472 hashcpy(b->branch_tree.versions[0].sha1,
1473 b->branch_tree.versions[1].sha1);
1474 } else if (!get_sha1(from, b->sha1)) {
1475 if (is_null_sha1(b->sha1)) {
1476 hashclr(b->branch_tree.versions[0].sha1);
1477 hashclr(b->branch_tree.versions[1].sha1);
1478 } else {
1479 unsigned long size;
1480 char *buf;
1482 buf = read_object_with_reference(b->sha1,
1483 type_names[OBJ_COMMIT], &size, b->sha1);
1484 if (!buf || size < 46)
1485 die("Not a valid commit: %s", from);
1486 if (memcmp("tree ", buf, 5)
1487 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1488 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1489 free(buf);
1490 hashcpy(b->branch_tree.versions[0].sha1,
1491 b->branch_tree.versions[1].sha1);
1492 }
1493 } else
1494 die("Invalid ref name or SHA1 expression: %s", from);
1496 read_next_command();
1497 }
1499 static void cmd_new_commit()
1500 {
1501 struct branch *b;
1502 void *msg;
1503 size_t msglen;
1504 char *str_uq;
1505 const char *endp;
1506 char *sp;
1507 char *author = NULL;
1508 char *committer = NULL;
1510 /* Obtain the branch name from the rest of our command */
1511 sp = strchr(command_buf.buf, ' ') + 1;
1512 str_uq = unquote_c_style(sp, &endp);
1513 if (str_uq) {
1514 if (*endp)
1515 die("Garbage after ref in: %s", command_buf.buf);
1516 sp = str_uq;
1517 }
1518 b = lookup_branch(sp);
1519 if (!b)
1520 b = new_branch(sp);
1521 if (str_uq)
1522 free(str_uq);
1524 read_next_command();
1525 cmd_mark();
1526 if (!strncmp("author ", command_buf.buf, 7)) {
1527 author = strdup(command_buf.buf);
1528 read_next_command();
1529 }
1530 if (!strncmp("committer ", command_buf.buf, 10)) {
1531 committer = strdup(command_buf.buf);
1532 read_next_command();
1533 }
1534 if (!committer)
1535 die("Expected committer but didn't get one");
1536 msg = cmd_data(&msglen);
1537 read_next_command();
1538 cmd_from(b);
1540 /* ensure the branch is active/loaded */
1541 if (!b->branch_tree.tree || !max_active_branches) {
1542 unload_one_branch();
1543 load_branch(b);
1544 }
1546 /* file_change* */
1547 for (;;) {
1548 if (1 == command_buf.len)
1549 break;
1550 else if (!strncmp("M ", command_buf.buf, 2))
1551 file_change_m(b);
1552 else if (!strncmp("D ", command_buf.buf, 2))
1553 file_change_d(b);
1554 else
1555 die("Unsupported file_change: %s", command_buf.buf);
1556 read_next_command();
1557 }
1559 /* build the tree and the commit */
1560 store_tree(&b->branch_tree);
1561 size_dbuf(&new_data, 97 + msglen
1562 + (author
1563 ? strlen(author) + strlen(committer)
1564 : 2 * strlen(committer)));
1565 sp = new_data.buffer;
1566 sp += sprintf(sp, "tree %s\n",
1567 sha1_to_hex(b->branch_tree.versions[1].sha1));
1568 if (!is_null_sha1(b->sha1))
1569 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1570 if (author)
1571 sp += sprintf(sp, "%s\n", author);
1572 else
1573 sp += sprintf(sp, "author %s\n", committer + 10);
1574 sp += sprintf(sp, "%s\n\n", committer);
1575 memcpy(sp, msg, msglen);
1576 sp += msglen;
1577 if (author)
1578 free(author);
1579 free(committer);
1580 free(msg);
1582 store_object(OBJ_COMMIT,
1583 new_data.buffer, sp - (char*)new_data.buffer,
1584 NULL, b->sha1, next_mark);
1585 b->last_commit = object_count_by_type[OBJ_COMMIT];
1587 if (branch_log) {
1588 int need_dq = quote_c_style(b->name, NULL, NULL, 0);
1589 fprintf(branch_log, "commit ");
1590 if (need_dq) {
1591 fputc('"', branch_log);
1592 quote_c_style(b->name, NULL, branch_log, 0);
1593 fputc('"', branch_log);
1594 } else
1595 fprintf(branch_log, "%s", b->name);
1596 fprintf(branch_log," :%lu %s\n",next_mark,sha1_to_hex(b->sha1));
1597 }
1598 }
1600 static void cmd_new_tag()
1601 {
1602 char *str_uq;
1603 const char *endp;
1604 char *sp;
1605 const char *from;
1606 char *tagger;
1607 struct branch *s;
1608 void *msg;
1609 size_t msglen;
1610 struct tag *t;
1611 unsigned long from_mark = 0;
1612 unsigned char sha1[20];
1614 /* Obtain the new tag name from the rest of our command */
1615 sp = strchr(command_buf.buf, ' ') + 1;
1616 str_uq = unquote_c_style(sp, &endp);
1617 if (str_uq) {
1618 if (*endp)
1619 die("Garbage after tag name in: %s", command_buf.buf);
1620 sp = str_uq;
1621 }
1622 t = pool_alloc(sizeof(struct tag));
1623 t->next_tag = NULL;
1624 t->name = pool_strdup(sp);
1625 if (last_tag)
1626 last_tag->next_tag = t;
1627 else
1628 first_tag = t;
1629 last_tag = t;
1630 if (str_uq)
1631 free(str_uq);
1632 read_next_command();
1634 /* from ... */
1635 if (strncmp("from ", command_buf.buf, 5))
1636 die("Expected from command, got %s", command_buf.buf);
1638 from = strchr(command_buf.buf, ' ') + 1;
1639 str_uq = unquote_c_style(from, &endp);
1640 if (str_uq) {
1641 if (*endp)
1642 die("Garbage after string in: %s", command_buf.buf);
1643 from = str_uq;
1644 }
1646 s = lookup_branch(from);
1647 if (s) {
1648 hashcpy(sha1, s->sha1);
1649 } else if (*from == ':') {
1650 from_mark = strtoul(from + 1, NULL, 10);
1651 struct object_entry *oe = find_mark(from_mark);
1652 if (oe->type != OBJ_COMMIT)
1653 die("Mark :%lu not a commit", from_mark);
1654 hashcpy(sha1, oe->sha1);
1655 } else if (!get_sha1(from, sha1)) {
1656 unsigned long size;
1657 char *buf;
1659 buf = read_object_with_reference(sha1,
1660 type_names[OBJ_COMMIT], &size, sha1);
1661 if (!buf || size < 46)
1662 die("Not a valid commit: %s", from);
1663 free(buf);
1664 } else
1665 die("Invalid ref name or SHA1 expression: %s", from);
1667 if (str_uq)
1668 free(str_uq);
1669 read_next_command();
1671 /* tagger ... */
1672 if (strncmp("tagger ", command_buf.buf, 7))
1673 die("Expected tagger command, got %s", command_buf.buf);
1674 tagger = strdup(command_buf.buf);
1676 /* tag payload/message */
1677 read_next_command();
1678 msg = cmd_data(&msglen);
1680 /* build the tag object */
1681 size_dbuf(&new_data, 67+strlen(t->name)+strlen(tagger)+msglen);
1682 sp = new_data.buffer;
1683 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
1684 sp += sprintf(sp, "type %s\n", type_names[OBJ_COMMIT]);
1685 sp += sprintf(sp, "tag %s\n", t->name);
1686 sp += sprintf(sp, "%s\n\n", tagger);
1687 memcpy(sp, msg, msglen);
1688 sp += msglen;
1689 free(tagger);
1690 free(msg);
1692 store_object(OBJ_TAG, new_data.buffer, sp - (char*)new_data.buffer,
1693 NULL, t->sha1, 0);
1695 if (branch_log) {
1696 int need_dq = quote_c_style(t->name, NULL, NULL, 0);
1697 fprintf(branch_log, "tag ");
1698 if (need_dq) {
1699 fputc('"', branch_log);
1700 quote_c_style(t->name, NULL, branch_log, 0);
1701 fputc('"', branch_log);
1702 } else
1703 fprintf(branch_log, "%s", t->name);
1704 fprintf(branch_log," :%lu %s\n",from_mark,sha1_to_hex(t->sha1));
1705 }
1706 }
1708 static void cmd_reset_branch()
1709 {
1710 struct branch *b;
1711 char *str_uq;
1712 const char *endp;
1713 char *sp;
1715 /* Obtain the branch name from the rest of our command */
1716 sp = strchr(command_buf.buf, ' ') + 1;
1717 str_uq = unquote_c_style(sp, &endp);
1718 if (str_uq) {
1719 if (*endp)
1720 die("Garbage after ref in: %s", command_buf.buf);
1721 sp = str_uq;
1722 }
1723 b = lookup_branch(sp);
1724 if (b) {
1725 b->last_commit = 0;
1726 if (b->branch_tree.tree) {
1727 release_tree_content_recursive(b->branch_tree.tree);
1728 b->branch_tree.tree = NULL;
1729 }
1730 }
1731 if (str_uq)
1732 free(str_uq);
1733 }
1735 static const char fast_import_usage[] =
1736 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file] [--branch-log=log] temp.pack";
1738 int main(int argc, const char **argv)
1739 {
1740 const char *base_name;
1741 int i;
1742 unsigned long est_obj_cnt = object_entry_alloc;
1743 char *pack_name;
1744 char *idx_name;
1745 struct stat sb;
1747 setup_ident();
1748 git_config(git_default_config);
1749 page_size = getpagesize();
1751 for (i = 1; i < argc; i++) {
1752 const char *a = argv[i];
1754 if (*a != '-' || !strcmp(a, "--"))
1755 break;
1756 else if (!strncmp(a, "--objects=", 10))
1757 est_obj_cnt = strtoul(a + 10, NULL, 0);
1758 else if (!strncmp(a, "--depth=", 8))
1759 max_depth = strtoul(a + 8, NULL, 0);
1760 else if (!strncmp(a, "--active-branches=", 18))
1761 max_active_branches = strtoul(a + 18, NULL, 0);
1762 else if (!strncmp(a, "--export-marks=", 15))
1763 mark_file = a + 15;
1764 else if (!strncmp(a, "--branch-log=", 13)) {
1765 branch_log = fopen(a + 13, "w");
1766 if (!branch_log)
1767 die("Can't create %s: %s", a + 13, strerror(errno));
1768 }
1769 else
1770 die("unknown option %s", a);
1771 }
1772 if ((i+1) != argc)
1773 usage(fast_import_usage);
1774 base_name = argv[i];
1776 pack_name = xmalloc(strlen(base_name) + 6);
1777 sprintf(pack_name, "%s.pack", base_name);
1778 idx_name = xmalloc(strlen(base_name) + 5);
1779 sprintf(idx_name, "%s.idx", base_name);
1781 pack_fd = open(pack_name, O_RDWR|O_CREAT|O_EXCL, 0666);
1782 if (pack_fd < 0)
1783 die("Can't create %s: %s", pack_name, strerror(errno));
1785 init_pack_header();
1786 alloc_objects(est_obj_cnt);
1787 strbuf_init(&command_buf);
1789 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
1790 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
1791 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
1792 marks = pool_calloc(1, sizeof(struct mark_set));
1794 for (;;) {
1795 read_next_command();
1796 if (command_buf.eof)
1797 break;
1798 else if (!strcmp("blob", command_buf.buf))
1799 cmd_new_blob();
1800 else if (!strncmp("commit ", command_buf.buf, 7))
1801 cmd_new_commit();
1802 else if (!strncmp("tag ", command_buf.buf, 4))
1803 cmd_new_tag();
1804 else if (!strncmp("reset ", command_buf.buf, 6))
1805 cmd_reset_branch();
1806 else
1807 die("Unsupported command: %s", command_buf.buf);
1808 }
1810 fixup_header_footer();
1811 close(pack_fd);
1812 write_index(idx_name);
1813 dump_branches();
1814 dump_tags();
1815 dump_marks();
1816 if (branch_log)
1817 fclose(branch_log);
1819 fprintf(stderr, "%s statistics:\n", argv[0]);
1820 fprintf(stderr, "---------------------------------------------------\n");
1821 fprintf(stderr, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count, alloc_count - est_obj_cnt);
1822 fprintf(stderr, "Total objects: %10lu (%10lu duplicates)\n", object_count, duplicate_count);
1823 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]);
1824 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]);
1825 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]);
1826 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]);
1827 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
1828 fprintf(stderr, " marks: %10u (%10lu unique )\n", (1 << marks->shift) * 1024, marks_set_count);
1829 fprintf(stderr, " atoms: %10u\n", atom_cnt);
1830 fprintf(stderr, "Memory total: %10lu KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
1831 fprintf(stderr, " pools: %10lu KiB\n", total_allocd/1024);
1832 fprintf(stderr, " objects: %10lu KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
1833 fprintf(stderr, "Pack remaps: %10lu\n", remap_count);
1834 fprintf(stderr, "---------------------------------------------------\n");
1836 stat(pack_name, &sb);
1837 fprintf(stderr, "Pack size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1838 stat(idx_name, &sb);
1839 fprintf(stderr, "Index size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1841 fprintf(stderr, "\n");
1843 return 0;
1844 }