ebffa7c904bc892dc086aa8a53257253764d2453
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 ('merge' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)*
24 file_change*
25 lf;
26 commit_msg ::= data;
28 file_change ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf
29 | 'D' sp path_str lf
30 ;
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
36 tag_msg;
37 tag_msg ::= data;
39 reset_branch ::= 'reset' sp ref_str lf
40 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
41 lf;
43 checkpoint ::= 'checkpoint' lf
44 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.
51 #
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
56 # binary data.
57 #
58 data ::= 'data' sp declen lf
59 binary_data
60 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
67 # are UTF8.
68 #
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.
83 #
84 idnum ::= ':' declen;
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.
94 #
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;
99 */
101 #include "builtin.h"
102 #include "cache.h"
103 #include "object.h"
104 #include "blob.h"
105 #include "tree.h"
106 #include "delta.h"
107 #include "pack.h"
108 #include "refs.h"
109 #include "csum-file.h"
110 #include "strbuf.h"
111 #include "quote.h"
113 struct object_entry
114 {
115 struct object_entry *next;
116 unsigned long offset;
117 unsigned type : TYPE_BITS;
118 unsigned pack_id : 16;
119 unsigned char sha1[20];
120 };
122 struct object_entry_pool
123 {
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 */
128 };
130 struct mark_set
131 {
132 int shift;
133 union {
134 struct object_entry *marked[1024];
135 struct mark_set *sets[1024];
136 } data;
137 };
139 struct last_object
140 {
141 void *data;
142 unsigned long len;
143 unsigned long offset;
144 unsigned int depth;
145 unsigned no_free:1;
146 };
148 struct mem_pool
149 {
150 struct mem_pool *next_pool;
151 char *next_free;
152 char *end;
153 char space[FLEX_ARRAY]; /* more */
154 };
156 struct atom_str
157 {
158 struct atom_str *next_atom;
159 int str_len;
160 char str_dat[FLEX_ARRAY]; /* more */
161 };
163 struct tree_content;
164 struct tree_entry
165 {
166 struct tree_content *tree;
167 struct atom_str* name;
168 struct tree_entry_ms
169 {
170 unsigned int mode;
171 unsigned char sha1[20];
172 } versions[2];
173 };
175 struct tree_content
176 {
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 */
181 };
183 struct avail_tree_content
184 {
185 unsigned int entry_capacity; /* must match tree_content */
186 struct avail_tree_content *next_avail;
187 };
189 struct branch
190 {
191 struct branch *table_next_branch;
192 struct branch *active_next_branch;
193 const char *name;
194 unsigned long last_commit;
195 struct tree_entry branch_tree;
196 unsigned char sha1[20];
197 };
199 struct tag
200 {
201 struct tag *next_tag;
202 const char *name;
203 unsigned char sha1[20];
204 };
206 struct dbuf
207 {
208 void *buffer;
209 size_t capacity;
210 };
212 struct hash_list
213 {
214 struct hash_list *next;
215 unsigned char sha1[20];
216 };
218 /* Configured limits on output */
219 static unsigned long max_depth = 10;
220 static unsigned long max_packsize = -1;
221 static uintmax_t max_objects = -1;
223 /* Stats and misc. counters */
224 static uintmax_t alloc_count;
225 static uintmax_t object_count;
226 static uintmax_t marks_set_count;
227 static uintmax_t object_count_by_type[1 << TYPE_BITS];
228 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
229 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
230 static unsigned long branch_count;
231 static unsigned long branch_load_count;
233 /* Memory pools */
234 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
235 static size_t total_allocd;
236 static struct mem_pool *mem_pool;
238 /* Atom management */
239 static unsigned int atom_table_sz = 4451;
240 static unsigned int atom_cnt;
241 static struct atom_str **atom_table;
243 /* The .pack file being generated */
244 static const char *base_name;
245 static unsigned int pack_id;
246 static char *idx_name;
247 static struct packed_git *pack_data;
248 static struct packed_git **all_packs;
249 static int pack_fd;
250 static unsigned long pack_size;
251 static unsigned char pack_sha1[20];
253 /* Table of objects we've written. */
254 static unsigned int object_entry_alloc = 5000;
255 static struct object_entry_pool *blocks;
256 static struct object_entry *object_table[1 << 16];
257 static struct mark_set *marks;
258 static const char* mark_file;
260 /* Our last blob */
261 static struct last_object last_blob;
263 /* Tree management */
264 static unsigned int tree_entry_alloc = 1000;
265 static void *avail_tree_entry;
266 static unsigned int avail_tree_table_sz = 100;
267 static struct avail_tree_content **avail_tree_table;
268 static struct dbuf old_tree;
269 static struct dbuf new_tree;
271 /* Branch data */
272 static unsigned long max_active_branches = 5;
273 static unsigned long cur_active_branches;
274 static unsigned long branch_table_sz = 1039;
275 static struct branch **branch_table;
276 static struct branch *active_branches;
278 /* Tag data */
279 static struct tag *first_tag;
280 static struct tag *last_tag;
282 /* Input stream parsing */
283 static struct strbuf command_buf;
284 static uintmax_t next_mark;
285 static struct dbuf new_data;
286 static FILE* branch_log;
289 static void alloc_objects(unsigned int cnt)
290 {
291 struct object_entry_pool *b;
293 b = xmalloc(sizeof(struct object_entry_pool)
294 + cnt * sizeof(struct object_entry));
295 b->next_pool = blocks;
296 b->next_free = b->entries;
297 b->end = b->entries + cnt;
298 blocks = b;
299 alloc_count += cnt;
300 }
302 static struct object_entry* new_object(unsigned char *sha1)
303 {
304 struct object_entry *e;
306 if (blocks->next_free == blocks->end)
307 alloc_objects(object_entry_alloc);
309 e = blocks->next_free++;
310 hashcpy(e->sha1, sha1);
311 return e;
312 }
314 static struct object_entry* find_object(unsigned char *sha1)
315 {
316 unsigned int h = sha1[0] << 8 | sha1[1];
317 struct object_entry *e;
318 for (e = object_table[h]; e; e = e->next)
319 if (!hashcmp(sha1, e->sha1))
320 return e;
321 return NULL;
322 }
324 static struct object_entry* insert_object(unsigned char *sha1)
325 {
326 unsigned int h = sha1[0] << 8 | sha1[1];
327 struct object_entry *e = object_table[h];
328 struct object_entry *p = NULL;
330 while (e) {
331 if (!hashcmp(sha1, e->sha1))
332 return e;
333 p = e;
334 e = e->next;
335 }
337 e = new_object(sha1);
338 e->next = NULL;
339 e->offset = 0;
340 if (p)
341 p->next = e;
342 else
343 object_table[h] = e;
344 return e;
345 }
347 static unsigned int hc_str(const char *s, size_t len)
348 {
349 unsigned int r = 0;
350 while (len-- > 0)
351 r = r * 31 + *s++;
352 return r;
353 }
355 static void* pool_alloc(size_t len)
356 {
357 struct mem_pool *p;
358 void *r;
360 for (p = mem_pool; p; p = p->next_pool)
361 if ((p->end - p->next_free >= len))
362 break;
364 if (!p) {
365 if (len >= (mem_pool_alloc/2)) {
366 total_allocd += len;
367 return xmalloc(len);
368 }
369 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
370 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
371 p->next_pool = mem_pool;
372 p->next_free = p->space;
373 p->end = p->next_free + mem_pool_alloc;
374 mem_pool = p;
375 }
377 r = p->next_free;
378 /* round out to a pointer alignment */
379 if (len & (sizeof(void*) - 1))
380 len += sizeof(void*) - (len & (sizeof(void*) - 1));
381 p->next_free += len;
382 return r;
383 }
385 static void* pool_calloc(size_t count, size_t size)
386 {
387 size_t len = count * size;
388 void *r = pool_alloc(len);
389 memset(r, 0, len);
390 return r;
391 }
393 static char* pool_strdup(const char *s)
394 {
395 char *r = pool_alloc(strlen(s) + 1);
396 strcpy(r, s);
397 return r;
398 }
400 static void size_dbuf(struct dbuf *b, size_t maxlen)
401 {
402 if (b->buffer) {
403 if (b->capacity >= maxlen)
404 return;
405 free(b->buffer);
406 }
407 b->capacity = ((maxlen / 1024) + 1) * 1024;
408 b->buffer = xmalloc(b->capacity);
409 }
411 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
412 {
413 struct mark_set *s = marks;
414 while ((idnum >> s->shift) >= 1024) {
415 s = pool_calloc(1, sizeof(struct mark_set));
416 s->shift = marks->shift + 10;
417 s->data.sets[0] = marks;
418 marks = s;
419 }
420 while (s->shift) {
421 uintmax_t i = idnum >> s->shift;
422 idnum -= i << s->shift;
423 if (!s->data.sets[i]) {
424 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
425 s->data.sets[i]->shift = s->shift - 10;
426 }
427 s = s->data.sets[i];
428 }
429 if (!s->data.marked[idnum])
430 marks_set_count++;
431 s->data.marked[idnum] = oe;
432 }
434 static struct object_entry* find_mark(uintmax_t idnum)
435 {
436 uintmax_t orig_idnum = idnum;
437 struct mark_set *s = marks;
438 struct object_entry *oe = NULL;
439 if ((idnum >> s->shift) < 1024) {
440 while (s && s->shift) {
441 uintmax_t i = idnum >> s->shift;
442 idnum -= i << s->shift;
443 s = s->data.sets[i];
444 }
445 if (s)
446 oe = s->data.marked[idnum];
447 }
448 if (!oe)
449 die("mark :%ju not declared", orig_idnum);
450 return oe;
451 }
453 static struct atom_str* to_atom(const char *s, size_t len)
454 {
455 unsigned int hc = hc_str(s, len) % atom_table_sz;
456 struct atom_str *c;
458 for (c = atom_table[hc]; c; c = c->next_atom)
459 if (c->str_len == len && !strncmp(s, c->str_dat, len))
460 return c;
462 c = pool_alloc(sizeof(struct atom_str) + len + 1);
463 c->str_len = len;
464 strncpy(c->str_dat, s, len);
465 c->str_dat[len] = 0;
466 c->next_atom = atom_table[hc];
467 atom_table[hc] = c;
468 atom_cnt++;
469 return c;
470 }
472 static struct branch* lookup_branch(const char *name)
473 {
474 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
475 struct branch *b;
477 for (b = branch_table[hc]; b; b = b->table_next_branch)
478 if (!strcmp(name, b->name))
479 return b;
480 return NULL;
481 }
483 static struct branch* new_branch(const char *name)
484 {
485 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
486 struct branch* b = lookup_branch(name);
488 if (b)
489 die("Invalid attempt to create duplicate branch: %s", name);
490 if (check_ref_format(name))
491 die("Branch name doesn't conform to GIT standards: %s", name);
493 b = pool_calloc(1, sizeof(struct branch));
494 b->name = pool_strdup(name);
495 b->table_next_branch = branch_table[hc];
496 b->branch_tree.versions[0].mode = S_IFDIR;
497 b->branch_tree.versions[1].mode = S_IFDIR;
498 branch_table[hc] = b;
499 branch_count++;
500 return b;
501 }
503 static unsigned int hc_entries(unsigned int cnt)
504 {
505 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
506 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
507 }
509 static struct tree_content* new_tree_content(unsigned int cnt)
510 {
511 struct avail_tree_content *f, *l = NULL;
512 struct tree_content *t;
513 unsigned int hc = hc_entries(cnt);
515 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
516 if (f->entry_capacity >= cnt)
517 break;
519 if (f) {
520 if (l)
521 l->next_avail = f->next_avail;
522 else
523 avail_tree_table[hc] = f->next_avail;
524 } else {
525 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
526 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
527 f->entry_capacity = cnt;
528 }
530 t = (struct tree_content*)f;
531 t->entry_count = 0;
532 t->delta_depth = 0;
533 return t;
534 }
536 static void release_tree_entry(struct tree_entry *e);
537 static void release_tree_content(struct tree_content *t)
538 {
539 struct avail_tree_content *f = (struct avail_tree_content*)t;
540 unsigned int hc = hc_entries(f->entry_capacity);
541 f->next_avail = avail_tree_table[hc];
542 avail_tree_table[hc] = f;
543 }
545 static void release_tree_content_recursive(struct tree_content *t)
546 {
547 unsigned int i;
548 for (i = 0; i < t->entry_count; i++)
549 release_tree_entry(t->entries[i]);
550 release_tree_content(t);
551 }
553 static struct tree_content* grow_tree_content(
554 struct tree_content *t,
555 int amt)
556 {
557 struct tree_content *r = new_tree_content(t->entry_count + amt);
558 r->entry_count = t->entry_count;
559 r->delta_depth = t->delta_depth;
560 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
561 release_tree_content(t);
562 return r;
563 }
565 static struct tree_entry* new_tree_entry()
566 {
567 struct tree_entry *e;
569 if (!avail_tree_entry) {
570 unsigned int n = tree_entry_alloc;
571 total_allocd += n * sizeof(struct tree_entry);
572 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
573 while (n-- > 1) {
574 *((void**)e) = e + 1;
575 e++;
576 }
577 *((void**)e) = NULL;
578 }
580 e = avail_tree_entry;
581 avail_tree_entry = *((void**)e);
582 return e;
583 }
585 static void release_tree_entry(struct tree_entry *e)
586 {
587 if (e->tree)
588 release_tree_content_recursive(e->tree);
589 *((void**)e) = avail_tree_entry;
590 avail_tree_entry = e;
591 }
593 static void yread(int fd, void *buffer, size_t length)
594 {
595 ssize_t ret = 0;
596 while (ret < length) {
597 ssize_t size = xread(fd, (char *) buffer + ret, length - ret);
598 if (!size)
599 die("Read from descriptor %i: end of stream", fd);
600 if (size < 0)
601 die("Read from descriptor %i: %s", fd, strerror(errno));
602 ret += size;
603 }
604 }
606 static void start_packfile()
607 {
608 struct packed_git *p;
609 struct pack_header hdr;
611 idx_name = xmalloc(strlen(base_name) + 11);
612 p = xcalloc(1, sizeof(*p) + strlen(base_name) + 13);
613 sprintf(p->pack_name, "%s%5.5i.pack", base_name, pack_id + 1);
614 sprintf(idx_name, "%s%5.5i.idx", base_name, pack_id + 1);
616 pack_fd = open(p->pack_name, O_RDWR|O_CREAT|O_EXCL, 0666);
617 if (pack_fd < 0)
618 die("Can't create %s: %s", p->pack_name, strerror(errno));
619 p->pack_fd = pack_fd;
621 hdr.hdr_signature = htonl(PACK_SIGNATURE);
622 hdr.hdr_version = htonl(2);
623 hdr.hdr_entries = 0;
624 write_or_die(pack_fd, &hdr, sizeof(hdr));
626 pack_data = p;
627 pack_size = sizeof(hdr);
628 object_count = 0;
630 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
631 all_packs[pack_id] = p;
632 }
634 static void fixup_header_footer()
635 {
636 SHA_CTX c;
637 char hdr[8];
638 unsigned long cnt;
639 char *buf;
641 if (lseek(pack_fd, 0, SEEK_SET) != 0)
642 die("Failed seeking to start: %s", strerror(errno));
644 SHA1_Init(&c);
645 yread(pack_fd, hdr, 8);
646 SHA1_Update(&c, hdr, 8);
648 cnt = htonl(object_count);
649 SHA1_Update(&c, &cnt, 4);
650 write_or_die(pack_fd, &cnt, 4);
652 buf = xmalloc(128 * 1024);
653 for (;;) {
654 size_t n = xread(pack_fd, buf, 128 * 1024);
655 if (n <= 0)
656 break;
657 SHA1_Update(&c, buf, n);
658 }
659 free(buf);
661 SHA1_Final(pack_sha1, &c);
662 write_or_die(pack_fd, pack_sha1, sizeof(pack_sha1));
663 }
665 static int oecmp (const void *a_, const void *b_)
666 {
667 struct object_entry *a = *((struct object_entry**)a_);
668 struct object_entry *b = *((struct object_entry**)b_);
669 return hashcmp(a->sha1, b->sha1);
670 }
672 static void write_index(const char *idx_name)
673 {
674 struct sha1file *f;
675 struct object_entry **idx, **c, **last, *e;
676 struct object_entry_pool *o;
677 unsigned int array[256];
678 int i;
680 /* Build the sorted table of object IDs. */
681 idx = xmalloc(object_count * sizeof(struct object_entry*));
682 c = idx;
683 for (o = blocks; o; o = o->next_pool)
684 for (e = o->next_free; e-- != o->entries;)
685 if (pack_id == e->pack_id)
686 *c++ = e;
687 last = idx + object_count;
688 if (c != last)
689 die("internal consistency error creating the index");
690 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
692 /* Generate the fan-out array. */
693 c = idx;
694 for (i = 0; i < 256; i++) {
695 struct object_entry **next = c;;
696 while (next < last) {
697 if ((*next)->sha1[0] != i)
698 break;
699 next++;
700 }
701 array[i] = htonl(next - idx);
702 c = next;
703 }
705 f = sha1create("%s", idx_name);
706 sha1write(f, array, 256 * sizeof(int));
707 for (c = idx; c != last; c++) {
708 unsigned int offset = htonl((*c)->offset);
709 sha1write(f, &offset, 4);
710 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
711 }
712 sha1write(f, pack_sha1, sizeof(pack_sha1));
713 sha1close(f, NULL, 1);
714 free(idx);
715 }
717 static void end_packfile()
718 {
719 struct packed_git *old_p = pack_data, *new_p;
721 if (object_count) {
722 fixup_header_footer();
723 write_index(idx_name);
724 fprintf(stdout, "%s\n", old_p->pack_name);
725 fflush(stdout);
727 /* Register the packfile with core git's machinary. */
728 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
729 if (!new_p)
730 die("core git rejected index %s", idx_name);
731 new_p->windows = old_p->windows;
732 new_p->pack_fd = old_p->pack_fd;
733 all_packs[pack_id++] = new_p;
734 install_packed_git(new_p);
735 }
736 else {
737 close(pack_fd);
738 unlink(old_p->pack_name);
739 }
740 free(old_p);
741 free(idx_name);
743 /* We can't carry a delta across packfiles. */
744 free(last_blob.data);
745 last_blob.data = NULL;
746 last_blob.len = 0;
747 last_blob.offset = 0;
748 last_blob.depth = 0;
749 }
751 static void checkpoint()
752 {
753 end_packfile();
754 start_packfile();
755 }
757 static size_t encode_header(
758 enum object_type type,
759 size_t size,
760 unsigned char *hdr)
761 {
762 int n = 1;
763 unsigned char c;
765 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
766 die("bad type %d", type);
768 c = (type << 4) | (size & 15);
769 size >>= 4;
770 while (size) {
771 *hdr++ = c | 0x80;
772 c = size & 0x7f;
773 size >>= 7;
774 n++;
775 }
776 *hdr = c;
777 return n;
778 }
780 static int store_object(
781 enum object_type type,
782 void *dat,
783 size_t datlen,
784 struct last_object *last,
785 unsigned char *sha1out,
786 uintmax_t mark)
787 {
788 void *out, *delta;
789 struct object_entry *e;
790 unsigned char hdr[96];
791 unsigned char sha1[20];
792 unsigned long hdrlen, deltalen;
793 SHA_CTX c;
794 z_stream s;
796 hdrlen = sprintf((char*)hdr,"%s %lu",type_names[type],datlen) + 1;
797 SHA1_Init(&c);
798 SHA1_Update(&c, hdr, hdrlen);
799 SHA1_Update(&c, dat, datlen);
800 SHA1_Final(sha1, &c);
801 if (sha1out)
802 hashcpy(sha1out, sha1);
804 e = insert_object(sha1);
805 if (mark)
806 insert_mark(mark, e);
807 if (e->offset) {
808 duplicate_count_by_type[type]++;
809 return 1;
810 }
812 if (last && last->data && last->depth < max_depth) {
813 delta = diff_delta(last->data, last->len,
814 dat, datlen,
815 &deltalen, 0);
816 if (delta && deltalen >= datlen) {
817 free(delta);
818 delta = NULL;
819 }
820 } else
821 delta = NULL;
823 memset(&s, 0, sizeof(s));
824 deflateInit(&s, zlib_compression_level);
825 if (delta) {
826 s.next_in = delta;
827 s.avail_in = deltalen;
828 } else {
829 s.next_in = dat;
830 s.avail_in = datlen;
831 }
832 s.avail_out = deflateBound(&s, s.avail_in);
833 s.next_out = out = xmalloc(s.avail_out);
834 while (deflate(&s, Z_FINISH) == Z_OK)
835 /* nothing */;
836 deflateEnd(&s);
838 /* Determine if we should auto-checkpoint. */
839 if ((object_count + 1) > max_objects
840 || (object_count + 1) < object_count
841 || (pack_size + 60 + s.total_out) > max_packsize
842 || (pack_size + 60 + s.total_out) < pack_size) {
844 /* This new object needs to *not* have the current pack_id. */
845 e->pack_id = pack_id + 1;
846 checkpoint();
848 /* We cannot carry a delta into the new pack. */
849 if (delta) {
850 free(delta);
851 delta = NULL;
853 memset(&s, 0, sizeof(s));
854 deflateInit(&s, zlib_compression_level);
855 s.next_in = dat;
856 s.avail_in = datlen;
857 s.avail_out = deflateBound(&s, s.avail_in);
858 s.next_out = out = xrealloc(out, s.avail_out);
859 while (deflate(&s, Z_FINISH) == Z_OK)
860 /* nothing */;
861 deflateEnd(&s);
862 }
863 }
865 e->type = type;
866 e->pack_id = pack_id;
867 e->offset = pack_size;
868 object_count++;
869 object_count_by_type[type]++;
871 if (delta) {
872 unsigned long ofs = e->offset - last->offset;
873 unsigned pos = sizeof(hdr) - 1;
875 delta_count_by_type[type]++;
876 last->depth++;
878 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
879 write_or_die(pack_fd, hdr, hdrlen);
880 pack_size += hdrlen;
882 hdr[pos] = ofs & 127;
883 while (ofs >>= 7)
884 hdr[--pos] = 128 | (--ofs & 127);
885 write_or_die(pack_fd, hdr + pos, sizeof(hdr) - pos);
886 pack_size += sizeof(hdr) - pos;
887 } else {
888 if (last)
889 last->depth = 0;
890 hdrlen = encode_header(type, datlen, hdr);
891 write_or_die(pack_fd, hdr, hdrlen);
892 pack_size += hdrlen;
893 }
895 write_or_die(pack_fd, out, s.total_out);
896 pack_size += s.total_out;
898 free(out);
899 if (delta)
900 free(delta);
901 if (last) {
902 if (last->data && !last->no_free)
903 free(last->data);
904 last->data = dat;
905 last->offset = e->offset;
906 last->len = datlen;
907 }
908 return 0;
909 }
911 static void *gfi_unpack_entry(
912 struct object_entry *oe,
913 unsigned long *sizep)
914 {
915 static char type[20];
916 struct packed_git *p = all_packs[oe->pack_id];
917 if (p == pack_data)
918 p->pack_size = pack_size + 20;
919 return unpack_entry(p, oe->offset, type, sizep);
920 }
922 static const char *get_mode(const char *str, unsigned int *modep)
923 {
924 unsigned char c;
925 unsigned int mode = 0;
927 while ((c = *str++) != ' ') {
928 if (c < '0' || c > '7')
929 return NULL;
930 mode = (mode << 3) + (c - '0');
931 }
932 *modep = mode;
933 return str;
934 }
936 static void load_tree(struct tree_entry *root)
937 {
938 unsigned char* sha1 = root->versions[1].sha1;
939 struct object_entry *myoe;
940 struct tree_content *t;
941 unsigned long size;
942 char *buf;
943 const char *c;
945 root->tree = t = new_tree_content(8);
946 if (is_null_sha1(sha1))
947 return;
949 myoe = find_object(sha1);
950 if (myoe) {
951 if (myoe->type != OBJ_TREE)
952 die("Not a tree: %s", sha1_to_hex(sha1));
953 t->delta_depth = 0;
954 buf = gfi_unpack_entry(myoe, &size);
955 } else {
956 char type[20];
957 buf = read_sha1_file(sha1, type, &size);
958 if (!buf || strcmp(type, tree_type))
959 die("Can't load tree %s", sha1_to_hex(sha1));
960 }
962 c = buf;
963 while (c != (buf + size)) {
964 struct tree_entry *e = new_tree_entry();
966 if (t->entry_count == t->entry_capacity)
967 root->tree = t = grow_tree_content(t, 8);
968 t->entries[t->entry_count++] = e;
970 e->tree = NULL;
971 c = get_mode(c, &e->versions[1].mode);
972 if (!c)
973 die("Corrupt mode in %s", sha1_to_hex(sha1));
974 e->versions[0].mode = e->versions[1].mode;
975 e->name = to_atom(c, strlen(c));
976 c += e->name->str_len + 1;
977 hashcpy(e->versions[0].sha1, (unsigned char*)c);
978 hashcpy(e->versions[1].sha1, (unsigned char*)c);
979 c += 20;
980 }
981 free(buf);
982 }
984 static int tecmp0 (const void *_a, const void *_b)
985 {
986 struct tree_entry *a = *((struct tree_entry**)_a);
987 struct tree_entry *b = *((struct tree_entry**)_b);
988 return base_name_compare(
989 a->name->str_dat, a->name->str_len, a->versions[0].mode,
990 b->name->str_dat, b->name->str_len, b->versions[0].mode);
991 }
993 static int tecmp1 (const void *_a, const void *_b)
994 {
995 struct tree_entry *a = *((struct tree_entry**)_a);
996 struct tree_entry *b = *((struct tree_entry**)_b);
997 return base_name_compare(
998 a->name->str_dat, a->name->str_len, a->versions[1].mode,
999 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1000 }
1002 static void mktree(struct tree_content *t,
1003 int v,
1004 unsigned long *szp,
1005 struct dbuf *b)
1006 {
1007 size_t maxlen = 0;
1008 unsigned int i;
1009 char *c;
1011 if (!v)
1012 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1013 else
1014 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1016 for (i = 0; i < t->entry_count; i++) {
1017 if (t->entries[i]->versions[v].mode)
1018 maxlen += t->entries[i]->name->str_len + 34;
1019 }
1021 size_dbuf(b, maxlen);
1022 c = b->buffer;
1023 for (i = 0; i < t->entry_count; i++) {
1024 struct tree_entry *e = t->entries[i];
1025 if (!e->versions[v].mode)
1026 continue;
1027 c += sprintf(c, "%o", e->versions[v].mode);
1028 *c++ = ' ';
1029 strcpy(c, e->name->str_dat);
1030 c += e->name->str_len + 1;
1031 hashcpy((unsigned char*)c, e->versions[v].sha1);
1032 c += 20;
1033 }
1034 *szp = c - (char*)b->buffer;
1035 }
1037 static void store_tree(struct tree_entry *root)
1038 {
1039 struct tree_content *t = root->tree;
1040 unsigned int i, j, del;
1041 unsigned long new_len;
1042 struct last_object lo;
1043 struct object_entry *le;
1045 if (!is_null_sha1(root->versions[1].sha1))
1046 return;
1048 for (i = 0; i < t->entry_count; i++) {
1049 if (t->entries[i]->tree)
1050 store_tree(t->entries[i]);
1051 }
1053 le = find_object(root->versions[0].sha1);
1054 if (!S_ISDIR(root->versions[0].mode)
1055 || !le
1056 || le->pack_id != pack_id) {
1057 lo.data = NULL;
1058 lo.depth = 0;
1059 } else {
1060 mktree(t, 0, &lo.len, &old_tree);
1061 lo.data = old_tree.buffer;
1062 lo.offset = le->offset;
1063 lo.depth = t->delta_depth;
1064 lo.no_free = 1;
1065 }
1067 mktree(t, 1, &new_len, &new_tree);
1068 store_object(OBJ_TREE, new_tree.buffer, new_len,
1069 &lo, root->versions[1].sha1, 0);
1071 t->delta_depth = lo.depth;
1072 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1073 struct tree_entry *e = t->entries[i];
1074 if (e->versions[1].mode) {
1075 e->versions[0].mode = e->versions[1].mode;
1076 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1077 t->entries[j++] = e;
1078 } else {
1079 release_tree_entry(e);
1080 del++;
1081 }
1082 }
1083 t->entry_count -= del;
1084 }
1086 static int tree_content_set(
1087 struct tree_entry *root,
1088 const char *p,
1089 const unsigned char *sha1,
1090 const unsigned int mode)
1091 {
1092 struct tree_content *t = root->tree;
1093 const char *slash1;
1094 unsigned int i, n;
1095 struct tree_entry *e;
1097 slash1 = strchr(p, '/');
1098 if (slash1)
1099 n = slash1 - p;
1100 else
1101 n = strlen(p);
1103 for (i = 0; i < t->entry_count; i++) {
1104 e = t->entries[i];
1105 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1106 if (!slash1) {
1107 if (e->versions[1].mode == mode
1108 && !hashcmp(e->versions[1].sha1, sha1))
1109 return 0;
1110 e->versions[1].mode = mode;
1111 hashcpy(e->versions[1].sha1, sha1);
1112 if (e->tree) {
1113 release_tree_content_recursive(e->tree);
1114 e->tree = NULL;
1115 }
1116 hashclr(root->versions[1].sha1);
1117 return 1;
1118 }
1119 if (!S_ISDIR(e->versions[1].mode)) {
1120 e->tree = new_tree_content(8);
1121 e->versions[1].mode = S_IFDIR;
1122 }
1123 if (!e->tree)
1124 load_tree(e);
1125 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
1126 hashclr(root->versions[1].sha1);
1127 return 1;
1128 }
1129 return 0;
1130 }
1131 }
1133 if (t->entry_count == t->entry_capacity)
1134 root->tree = t = grow_tree_content(t, 8);
1135 e = new_tree_entry();
1136 e->name = to_atom(p, n);
1137 e->versions[0].mode = 0;
1138 hashclr(e->versions[0].sha1);
1139 t->entries[t->entry_count++] = e;
1140 if (slash1) {
1141 e->tree = new_tree_content(8);
1142 e->versions[1].mode = S_IFDIR;
1143 tree_content_set(e, slash1 + 1, sha1, mode);
1144 } else {
1145 e->tree = NULL;
1146 e->versions[1].mode = mode;
1147 hashcpy(e->versions[1].sha1, sha1);
1148 }
1149 hashclr(root->versions[1].sha1);
1150 return 1;
1151 }
1153 static int tree_content_remove(struct tree_entry *root, const char *p)
1154 {
1155 struct tree_content *t = root->tree;
1156 const char *slash1;
1157 unsigned int i, n;
1158 struct tree_entry *e;
1160 slash1 = strchr(p, '/');
1161 if (slash1)
1162 n = slash1 - p;
1163 else
1164 n = strlen(p);
1166 for (i = 0; i < t->entry_count; i++) {
1167 e = t->entries[i];
1168 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1169 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1170 goto del_entry;
1171 if (!e->tree)
1172 load_tree(e);
1173 if (tree_content_remove(e, slash1 + 1)) {
1174 for (n = 0; n < e->tree->entry_count; n++) {
1175 if (e->tree->entries[n]->versions[1].mode) {
1176 hashclr(root->versions[1].sha1);
1177 return 1;
1178 }
1179 }
1180 goto del_entry;
1181 }
1182 return 0;
1183 }
1184 }
1185 return 0;
1187 del_entry:
1188 if (e->tree) {
1189 release_tree_content_recursive(e->tree);
1190 e->tree = NULL;
1191 }
1192 e->versions[1].mode = 0;
1193 hashclr(e->versions[1].sha1);
1194 hashclr(root->versions[1].sha1);
1195 return 1;
1196 }
1198 static void dump_branches()
1199 {
1200 static const char *msg = "fast-import";
1201 unsigned int i;
1202 struct branch *b;
1203 struct ref_lock *lock;
1205 for (i = 0; i < branch_table_sz; i++) {
1206 for (b = branch_table[i]; b; b = b->table_next_branch) {
1207 lock = lock_any_ref_for_update(b->name, NULL);
1208 if (!lock || write_ref_sha1(lock, b->sha1, msg) < 0)
1209 die("Can't write %s", b->name);
1210 }
1211 }
1212 }
1214 static void dump_tags()
1215 {
1216 static const char *msg = "fast-import";
1217 struct tag *t;
1218 struct ref_lock *lock;
1219 char path[PATH_MAX];
1221 for (t = first_tag; t; t = t->next_tag) {
1222 sprintf(path, "refs/tags/%s", t->name);
1223 lock = lock_any_ref_for_update(path, NULL);
1224 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1225 die("Can't write %s", path);
1226 }
1227 }
1229 static void dump_marks_helper(FILE *f,
1230 uintmax_t base,
1231 struct mark_set *m)
1232 {
1233 uintmax_t k;
1234 if (m->shift) {
1235 for (k = 0; k < 1024; k++) {
1236 if (m->data.sets[k])
1237 dump_marks_helper(f, (base + k) << m->shift,
1238 m->data.sets[k]);
1239 }
1240 } else {
1241 for (k = 0; k < 1024; k++) {
1242 if (m->data.marked[k])
1243 fprintf(f, ":%ju %s\n", base + k,
1244 sha1_to_hex(m->data.marked[k]->sha1));
1245 }
1246 }
1247 }
1249 static void dump_marks()
1250 {
1251 if (mark_file)
1252 {
1253 FILE *f = fopen(mark_file, "w");
1254 dump_marks_helper(f, 0, marks);
1255 fclose(f);
1256 }
1257 }
1259 static void read_next_command()
1260 {
1261 read_line(&command_buf, stdin, '\n');
1262 }
1264 static void cmd_mark()
1265 {
1266 if (!strncmp("mark :", command_buf.buf, 6)) {
1267 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1268 read_next_command();
1269 }
1270 else
1271 next_mark = 0;
1272 }
1274 static void* cmd_data (size_t *size)
1275 {
1276 size_t n = 0;
1277 void *buffer;
1278 size_t length;
1280 if (strncmp("data ", command_buf.buf, 5))
1281 die("Expected 'data n' command, found: %s", command_buf.buf);
1283 length = strtoul(command_buf.buf + 5, NULL, 10);
1284 buffer = xmalloc(length);
1286 while (n < length) {
1287 size_t s = fread((char*)buffer + n, 1, length - n, stdin);
1288 if (!s && feof(stdin))
1289 die("EOF in data (%lu bytes remaining)", length - n);
1290 n += s;
1291 }
1293 if (fgetc(stdin) != '\n')
1294 die("An lf did not trail the binary data as expected.");
1296 *size = length;
1297 return buffer;
1298 }
1300 static void cmd_new_blob()
1301 {
1302 size_t l;
1303 void *d;
1305 read_next_command();
1306 cmd_mark();
1307 d = cmd_data(&l);
1309 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1310 free(d);
1311 }
1313 static void unload_one_branch()
1314 {
1315 while (cur_active_branches
1316 && cur_active_branches >= max_active_branches) {
1317 unsigned long min_commit = ULONG_MAX;
1318 struct branch *e, *l = NULL, *p = NULL;
1320 for (e = active_branches; e; e = e->active_next_branch) {
1321 if (e->last_commit < min_commit) {
1322 p = l;
1323 min_commit = e->last_commit;
1324 }
1325 l = e;
1326 }
1328 if (p) {
1329 e = p->active_next_branch;
1330 p->active_next_branch = e->active_next_branch;
1331 } else {
1332 e = active_branches;
1333 active_branches = e->active_next_branch;
1334 }
1335 e->active_next_branch = NULL;
1336 if (e->branch_tree.tree) {
1337 release_tree_content_recursive(e->branch_tree.tree);
1338 e->branch_tree.tree = NULL;
1339 }
1340 cur_active_branches--;
1341 }
1342 }
1344 static void load_branch(struct branch *b)
1345 {
1346 load_tree(&b->branch_tree);
1347 b->active_next_branch = active_branches;
1348 active_branches = b;
1349 cur_active_branches++;
1350 branch_load_count++;
1351 }
1353 static void file_change_m(struct branch *b)
1354 {
1355 const char *p = command_buf.buf + 2;
1356 char *p_uq;
1357 const char *endp;
1358 struct object_entry *oe;
1359 unsigned char sha1[20];
1360 unsigned int mode;
1361 char type[20];
1363 p = get_mode(p, &mode);
1364 if (!p)
1365 die("Corrupt mode: %s", command_buf.buf);
1366 switch (mode) {
1367 case S_IFREG | 0644:
1368 case S_IFREG | 0755:
1369 case S_IFLNK:
1370 case 0644:
1371 case 0755:
1372 /* ok */
1373 break;
1374 default:
1375 die("Corrupt mode: %s", command_buf.buf);
1376 }
1378 if (*p == ':') {
1379 char *x;
1380 oe = find_mark(strtoumax(p + 1, &x, 10));
1381 hashcpy(sha1, oe->sha1);
1382 p = x;
1383 } else {
1384 if (get_sha1_hex(p, sha1))
1385 die("Invalid SHA1: %s", command_buf.buf);
1386 oe = find_object(sha1);
1387 p += 40;
1388 }
1389 if (*p++ != ' ')
1390 die("Missing space after SHA1: %s", command_buf.buf);
1392 p_uq = unquote_c_style(p, &endp);
1393 if (p_uq) {
1394 if (*endp)
1395 die("Garbage after path in: %s", command_buf.buf);
1396 p = p_uq;
1397 }
1399 if (oe) {
1400 if (oe->type != OBJ_BLOB)
1401 die("Not a blob (actually a %s): %s",
1402 command_buf.buf, type_names[oe->type]);
1403 } else {
1404 if (sha1_object_info(sha1, type, NULL))
1405 die("Blob not found: %s", command_buf.buf);
1406 if (strcmp(blob_type, type))
1407 die("Not a blob (actually a %s): %s",
1408 command_buf.buf, type);
1409 }
1411 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1413 if (p_uq)
1414 free(p_uq);
1415 }
1417 static void file_change_d(struct branch *b)
1418 {
1419 const char *p = command_buf.buf + 2;
1420 char *p_uq;
1421 const char *endp;
1423 p_uq = unquote_c_style(p, &endp);
1424 if (p_uq) {
1425 if (*endp)
1426 die("Garbage after path in: %s", command_buf.buf);
1427 p = p_uq;
1428 }
1429 tree_content_remove(&b->branch_tree, p);
1430 if (p_uq)
1431 free(p_uq);
1432 }
1434 static void cmd_from(struct branch *b)
1435 {
1436 const char *from, *endp;
1437 char *str_uq;
1438 struct branch *s;
1440 if (strncmp("from ", command_buf.buf, 5))
1441 return;
1443 if (b->last_commit)
1444 die("Can't reinitailize branch %s", b->name);
1446 from = strchr(command_buf.buf, ' ') + 1;
1447 str_uq = unquote_c_style(from, &endp);
1448 if (str_uq) {
1449 if (*endp)
1450 die("Garbage after string in: %s", command_buf.buf);
1451 from = str_uq;
1452 }
1454 s = lookup_branch(from);
1455 if (b == s)
1456 die("Can't create a branch from itself: %s", b->name);
1457 else if (s) {
1458 unsigned char *t = s->branch_tree.versions[1].sha1;
1459 hashcpy(b->sha1, s->sha1);
1460 hashcpy(b->branch_tree.versions[0].sha1, t);
1461 hashcpy(b->branch_tree.versions[1].sha1, t);
1462 } else if (*from == ':') {
1463 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1464 struct object_entry *oe = find_mark(idnum);
1465 unsigned long size;
1466 char *buf;
1467 if (oe->type != OBJ_COMMIT)
1468 die("Mark :%ju not a commit", idnum);
1469 hashcpy(b->sha1, oe->sha1);
1470 buf = gfi_unpack_entry(oe, &size);
1471 if (!buf || size < 46)
1472 die("Not a valid commit: %s", from);
1473 if (memcmp("tree ", buf, 5)
1474 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1475 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1476 free(buf);
1477 hashcpy(b->branch_tree.versions[0].sha1,
1478 b->branch_tree.versions[1].sha1);
1479 } else if (!get_sha1(from, b->sha1)) {
1480 if (is_null_sha1(b->sha1)) {
1481 hashclr(b->branch_tree.versions[0].sha1);
1482 hashclr(b->branch_tree.versions[1].sha1);
1483 } else {
1484 unsigned long size;
1485 char *buf;
1487 buf = read_object_with_reference(b->sha1,
1488 type_names[OBJ_COMMIT], &size, b->sha1);
1489 if (!buf || size < 46)
1490 die("Not a valid commit: %s", from);
1491 if (memcmp("tree ", buf, 5)
1492 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1493 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1494 free(buf);
1495 hashcpy(b->branch_tree.versions[0].sha1,
1496 b->branch_tree.versions[1].sha1);
1497 }
1498 } else
1499 die("Invalid ref name or SHA1 expression: %s", from);
1501 read_next_command();
1502 }
1504 static struct hash_list* cmd_merge(unsigned int *count)
1505 {
1506 struct hash_list *list = NULL, *n, *e;
1507 const char *from, *endp;
1508 char *str_uq;
1509 struct branch *s;
1511 *count = 0;
1512 while (!strncmp("merge ", command_buf.buf, 6)) {
1513 from = strchr(command_buf.buf, ' ') + 1;
1514 str_uq = unquote_c_style(from, &endp);
1515 if (str_uq) {
1516 if (*endp)
1517 die("Garbage after string in: %s", command_buf.buf);
1518 from = str_uq;
1519 }
1521 n = xmalloc(sizeof(*n));
1522 s = lookup_branch(from);
1523 if (s)
1524 hashcpy(n->sha1, s->sha1);
1525 else if (*from == ':') {
1526 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1527 struct object_entry *oe = find_mark(idnum);
1528 if (oe->type != OBJ_COMMIT)
1529 die("Mark :%ju not a commit", idnum);
1530 hashcpy(n->sha1, oe->sha1);
1531 } else if (get_sha1(from, n->sha1))
1532 die("Invalid ref name or SHA1 expression: %s", from);
1534 n->next = NULL;
1535 if (list)
1536 e->next = n;
1537 else
1538 list = n;
1539 e = n;
1540 *count++;
1541 read_next_command();
1542 }
1543 return list;
1544 }
1546 static void cmd_new_commit()
1547 {
1548 struct branch *b;
1549 void *msg;
1550 size_t msglen;
1551 char *str_uq;
1552 const char *endp;
1553 char *sp;
1554 char *author = NULL;
1555 char *committer = NULL;
1556 struct hash_list *merge_list = NULL;
1557 unsigned int merge_count;
1559 /* Obtain the branch name from the rest of our command */
1560 sp = strchr(command_buf.buf, ' ') + 1;
1561 str_uq = unquote_c_style(sp, &endp);
1562 if (str_uq) {
1563 if (*endp)
1564 die("Garbage after ref in: %s", command_buf.buf);
1565 sp = str_uq;
1566 }
1567 b = lookup_branch(sp);
1568 if (!b)
1569 b = new_branch(sp);
1570 if (str_uq)
1571 free(str_uq);
1573 read_next_command();
1574 cmd_mark();
1575 if (!strncmp("author ", command_buf.buf, 7)) {
1576 author = strdup(command_buf.buf);
1577 read_next_command();
1578 }
1579 if (!strncmp("committer ", command_buf.buf, 10)) {
1580 committer = strdup(command_buf.buf);
1581 read_next_command();
1582 }
1583 if (!committer)
1584 die("Expected committer but didn't get one");
1585 msg = cmd_data(&msglen);
1586 read_next_command();
1587 cmd_from(b);
1588 merge_list = cmd_merge(&merge_count);
1590 /* ensure the branch is active/loaded */
1591 if (!b->branch_tree.tree || !max_active_branches) {
1592 unload_one_branch();
1593 load_branch(b);
1594 }
1596 /* file_change* */
1597 for (;;) {
1598 if (1 == command_buf.len)
1599 break;
1600 else if (!strncmp("M ", command_buf.buf, 2))
1601 file_change_m(b);
1602 else if (!strncmp("D ", command_buf.buf, 2))
1603 file_change_d(b);
1604 else
1605 die("Unsupported file_change: %s", command_buf.buf);
1606 read_next_command();
1607 }
1609 /* build the tree and the commit */
1610 store_tree(&b->branch_tree);
1611 hashcpy(b->branch_tree.versions[0].sha1,
1612 b->branch_tree.versions[1].sha1);
1613 size_dbuf(&new_data, 97 + msglen
1614 + merge_count * 49
1615 + (author
1616 ? strlen(author) + strlen(committer)
1617 : 2 * strlen(committer)));
1618 sp = new_data.buffer;
1619 sp += sprintf(sp, "tree %s\n",
1620 sha1_to_hex(b->branch_tree.versions[1].sha1));
1621 if (!is_null_sha1(b->sha1))
1622 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1623 while (merge_list) {
1624 struct hash_list *next = merge_list->next;
1625 sp += sprintf(sp, "parent %s\n", sha1_to_hex(merge_list->sha1));
1626 free(merge_list);
1627 merge_list = next;
1628 }
1629 if (author)
1630 sp += sprintf(sp, "%s\n", author);
1631 else
1632 sp += sprintf(sp, "author %s\n", committer + 10);
1633 sp += sprintf(sp, "%s\n\n", committer);
1634 memcpy(sp, msg, msglen);
1635 sp += msglen;
1636 if (author)
1637 free(author);
1638 free(committer);
1639 free(msg);
1641 store_object(OBJ_COMMIT,
1642 new_data.buffer, sp - (char*)new_data.buffer,
1643 NULL, b->sha1, next_mark);
1644 b->last_commit = object_count_by_type[OBJ_COMMIT];
1646 if (branch_log) {
1647 int need_dq = quote_c_style(b->name, NULL, NULL, 0);
1648 fprintf(branch_log, "commit ");
1649 if (need_dq) {
1650 fputc('"', branch_log);
1651 quote_c_style(b->name, NULL, branch_log, 0);
1652 fputc('"', branch_log);
1653 } else
1654 fprintf(branch_log, "%s", b->name);
1655 fprintf(branch_log," :%ju %s\n",next_mark,sha1_to_hex(b->sha1));
1656 }
1657 }
1659 static void cmd_new_tag()
1660 {
1661 char *str_uq;
1662 const char *endp;
1663 char *sp;
1664 const char *from;
1665 char *tagger;
1666 struct branch *s;
1667 void *msg;
1668 size_t msglen;
1669 struct tag *t;
1670 uintmax_t from_mark = 0;
1671 unsigned char sha1[20];
1673 /* Obtain the new tag name from the rest of our command */
1674 sp = strchr(command_buf.buf, ' ') + 1;
1675 str_uq = unquote_c_style(sp, &endp);
1676 if (str_uq) {
1677 if (*endp)
1678 die("Garbage after tag name in: %s", command_buf.buf);
1679 sp = str_uq;
1680 }
1681 t = pool_alloc(sizeof(struct tag));
1682 t->next_tag = NULL;
1683 t->name = pool_strdup(sp);
1684 if (last_tag)
1685 last_tag->next_tag = t;
1686 else
1687 first_tag = t;
1688 last_tag = t;
1689 if (str_uq)
1690 free(str_uq);
1691 read_next_command();
1693 /* from ... */
1694 if (strncmp("from ", command_buf.buf, 5))
1695 die("Expected from command, got %s", command_buf.buf);
1697 from = strchr(command_buf.buf, ' ') + 1;
1698 str_uq = unquote_c_style(from, &endp);
1699 if (str_uq) {
1700 if (*endp)
1701 die("Garbage after string in: %s", command_buf.buf);
1702 from = str_uq;
1703 }
1705 s = lookup_branch(from);
1706 if (s) {
1707 hashcpy(sha1, s->sha1);
1708 } else if (*from == ':') {
1709 from_mark = strtoumax(from + 1, NULL, 10);
1710 struct object_entry *oe = find_mark(from_mark);
1711 if (oe->type != OBJ_COMMIT)
1712 die("Mark :%ju not a commit", from_mark);
1713 hashcpy(sha1, oe->sha1);
1714 } else if (!get_sha1(from, sha1)) {
1715 unsigned long size;
1716 char *buf;
1718 buf = read_object_with_reference(sha1,
1719 type_names[OBJ_COMMIT], &size, sha1);
1720 if (!buf || size < 46)
1721 die("Not a valid commit: %s", from);
1722 free(buf);
1723 } else
1724 die("Invalid ref name or SHA1 expression: %s", from);
1726 if (str_uq)
1727 free(str_uq);
1728 read_next_command();
1730 /* tagger ... */
1731 if (strncmp("tagger ", command_buf.buf, 7))
1732 die("Expected tagger command, got %s", command_buf.buf);
1733 tagger = strdup(command_buf.buf);
1735 /* tag payload/message */
1736 read_next_command();
1737 msg = cmd_data(&msglen);
1739 /* build the tag object */
1740 size_dbuf(&new_data, 67+strlen(t->name)+strlen(tagger)+msglen);
1741 sp = new_data.buffer;
1742 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
1743 sp += sprintf(sp, "type %s\n", type_names[OBJ_COMMIT]);
1744 sp += sprintf(sp, "tag %s\n", t->name);
1745 sp += sprintf(sp, "%s\n\n", tagger);
1746 memcpy(sp, msg, msglen);
1747 sp += msglen;
1748 free(tagger);
1749 free(msg);
1751 store_object(OBJ_TAG, new_data.buffer, sp - (char*)new_data.buffer,
1752 NULL, t->sha1, 0);
1754 if (branch_log) {
1755 int need_dq = quote_c_style(t->name, NULL, NULL, 0);
1756 fprintf(branch_log, "tag ");
1757 if (need_dq) {
1758 fputc('"', branch_log);
1759 quote_c_style(t->name, NULL, branch_log, 0);
1760 fputc('"', branch_log);
1761 } else
1762 fprintf(branch_log, "%s", t->name);
1763 fprintf(branch_log," :%ju %s\n",from_mark,sha1_to_hex(t->sha1));
1764 }
1765 }
1767 static void cmd_reset_branch()
1768 {
1769 struct branch *b;
1770 char *str_uq;
1771 const char *endp;
1772 char *sp;
1774 /* Obtain the branch name from the rest of our command */
1775 sp = strchr(command_buf.buf, ' ') + 1;
1776 str_uq = unquote_c_style(sp, &endp);
1777 if (str_uq) {
1778 if (*endp)
1779 die("Garbage after ref in: %s", command_buf.buf);
1780 sp = str_uq;
1781 }
1782 b = lookup_branch(sp);
1783 if (b) {
1784 b->last_commit = 0;
1785 if (b->branch_tree.tree) {
1786 release_tree_content_recursive(b->branch_tree.tree);
1787 b->branch_tree.tree = NULL;
1788 }
1789 }
1790 else
1791 b = new_branch(sp);
1792 if (str_uq)
1793 free(str_uq);
1794 read_next_command();
1795 cmd_from(b);
1796 }
1798 static void cmd_checkpoint()
1799 {
1800 if (object_count)
1801 checkpoint();
1802 read_next_command();
1803 }
1805 static const char fast_import_usage[] =
1806 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file] [--branch-log=log] temp.pack";
1808 int main(int argc, const char **argv)
1809 {
1810 int i;
1811 uintmax_t est_obj_cnt = object_entry_alloc;
1812 uintmax_t duplicate_count;
1814 setup_ident();
1815 git_config(git_default_config);
1817 for (i = 1; i < argc; i++) {
1818 const char *a = argv[i];
1820 if (*a != '-' || !strcmp(a, "--"))
1821 break;
1822 else if (!strncmp(a, "--objects=", 10))
1823 est_obj_cnt = strtoumax(a + 10, NULL, 0);
1824 else if (!strncmp(a, "--max-objects-per-pack=", 23))
1825 max_objects = strtoumax(a + 23, NULL, 0);
1826 else if (!strncmp(a, "--max-pack-size=", 16))
1827 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
1828 else if (!strncmp(a, "--depth=", 8))
1829 max_depth = strtoul(a + 8, NULL, 0);
1830 else if (!strncmp(a, "--active-branches=", 18))
1831 max_active_branches = strtoul(a + 18, NULL, 0);
1832 else if (!strncmp(a, "--export-marks=", 15))
1833 mark_file = a + 15;
1834 else if (!strncmp(a, "--branch-log=", 13)) {
1835 branch_log = fopen(a + 13, "w");
1836 if (!branch_log)
1837 die("Can't create %s: %s", a + 13, strerror(errno));
1838 }
1839 else
1840 die("unknown option %s", a);
1841 }
1842 if ((i+1) != argc)
1843 usage(fast_import_usage);
1844 base_name = argv[i];
1846 alloc_objects(est_obj_cnt);
1847 strbuf_init(&command_buf);
1849 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
1850 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
1851 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
1852 marks = pool_calloc(1, sizeof(struct mark_set));
1854 start_packfile();
1855 for (;;) {
1856 read_next_command();
1857 if (command_buf.eof)
1858 break;
1859 else if (!strcmp("blob", command_buf.buf))
1860 cmd_new_blob();
1861 else if (!strncmp("commit ", command_buf.buf, 7))
1862 cmd_new_commit();
1863 else if (!strncmp("tag ", command_buf.buf, 4))
1864 cmd_new_tag();
1865 else if (!strncmp("reset ", command_buf.buf, 6))
1866 cmd_reset_branch();
1867 else if (!strcmp("checkpoint", command_buf.buf))
1868 cmd_checkpoint();
1869 else
1870 die("Unsupported command: %s", command_buf.buf);
1871 }
1872 end_packfile();
1874 dump_branches();
1875 dump_tags();
1876 dump_marks();
1877 if (branch_log)
1878 fclose(branch_log);
1880 duplicate_count = 0;
1881 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
1882 duplicate_count += duplicate_count_by_type[i];
1884 fprintf(stderr, "%s statistics:\n", argv[0]);
1885 fprintf(stderr, "---------------------------------------------------------------------\n");
1886 fprintf(stderr, "Alloc'd objects: %10ju (%10ju overflow )\n", alloc_count, alloc_count - est_obj_cnt);
1887 fprintf(stderr, "Total objects: %10ju (%10ju duplicates )\n", object_count, duplicate_count);
1888 fprintf(stderr, " blobs : %10ju (%10ju duplicates %10ju deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
1889 fprintf(stderr, " trees : %10ju (%10ju duplicates %10ju deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
1890 fprintf(stderr, " commits: %10ju (%10ju duplicates %10ju deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
1891 fprintf(stderr, " tags : %10ju (%10ju duplicates %10ju deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
1892 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
1893 fprintf(stderr, " marks: %10ju (%10ju unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
1894 fprintf(stderr, " atoms: %10u\n", atom_cnt);
1895 fprintf(stderr, "Memory total: %10ju KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
1896 fprintf(stderr, " pools: %10lu KiB\n", total_allocd/1024);
1897 fprintf(stderr, " objects: %10ju KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
1898 fprintf(stderr, "---------------------------------------------------------------------\n");
1899 pack_report();
1900 fprintf(stderr, "---------------------------------------------------------------------\n");
1901 fprintf(stderr, "\n");
1903 return 0;
1904 }