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 int pack_id;
197 unsigned char sha1[20];
198 };
200 struct tag
201 {
202 struct tag *next_tag;
203 const char *name;
204 unsigned int pack_id;
205 unsigned char sha1[20];
206 };
208 struct dbuf
209 {
210 void *buffer;
211 size_t capacity;
212 };
214 struct hash_list
215 {
216 struct hash_list *next;
217 unsigned char sha1[20];
218 };
220 /* Configured limits on output */
221 static unsigned long max_depth = 10;
222 static unsigned long max_packsize = (1LL << 32) - 1;
223 static uintmax_t max_objects = -1;
225 /* Stats and misc. counters */
226 static uintmax_t alloc_count;
227 static uintmax_t marks_set_count;
228 static uintmax_t object_count_by_type[1 << TYPE_BITS];
229 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
230 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
231 static unsigned long object_count;
232 static unsigned long branch_count;
233 static unsigned long branch_load_count;
235 /* Memory pools */
236 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
237 static size_t total_allocd;
238 static struct mem_pool *mem_pool;
240 /* Atom management */
241 static unsigned int atom_table_sz = 4451;
242 static unsigned int atom_cnt;
243 static struct atom_str **atom_table;
245 /* The .pack file being generated */
246 static unsigned int pack_id;
247 static struct packed_git *pack_data;
248 static struct packed_git **all_packs;
249 static unsigned long pack_size;
251 /* Table of objects we've written. */
252 static unsigned int object_entry_alloc = 5000;
253 static struct object_entry_pool *blocks;
254 static struct object_entry *object_table[1 << 16];
255 static struct mark_set *marks;
256 static const char* mark_file;
258 /* Our last blob */
259 static struct last_object last_blob;
261 /* Tree management */
262 static unsigned int tree_entry_alloc = 1000;
263 static void *avail_tree_entry;
264 static unsigned int avail_tree_table_sz = 100;
265 static struct avail_tree_content **avail_tree_table;
266 static struct dbuf old_tree;
267 static struct dbuf new_tree;
269 /* Branch data */
270 static unsigned long max_active_branches = 5;
271 static unsigned long cur_active_branches;
272 static unsigned long branch_table_sz = 1039;
273 static struct branch **branch_table;
274 static struct branch *active_branches;
276 /* Tag data */
277 static struct tag *first_tag;
278 static struct tag *last_tag;
280 /* Input stream parsing */
281 static struct strbuf command_buf;
282 static uintmax_t next_mark;
283 static struct dbuf new_data;
284 static FILE* branch_log;
287 static void alloc_objects(unsigned int cnt)
288 {
289 struct object_entry_pool *b;
291 b = xmalloc(sizeof(struct object_entry_pool)
292 + cnt * sizeof(struct object_entry));
293 b->next_pool = blocks;
294 b->next_free = b->entries;
295 b->end = b->entries + cnt;
296 blocks = b;
297 alloc_count += cnt;
298 }
300 static struct object_entry* new_object(unsigned char *sha1)
301 {
302 struct object_entry *e;
304 if (blocks->next_free == blocks->end)
305 alloc_objects(object_entry_alloc);
307 e = blocks->next_free++;
308 hashcpy(e->sha1, sha1);
309 return e;
310 }
312 static struct object_entry* find_object(unsigned char *sha1)
313 {
314 unsigned int h = sha1[0] << 8 | sha1[1];
315 struct object_entry *e;
316 for (e = object_table[h]; e; e = e->next)
317 if (!hashcmp(sha1, e->sha1))
318 return e;
319 return NULL;
320 }
322 static struct object_entry* insert_object(unsigned char *sha1)
323 {
324 unsigned int h = sha1[0] << 8 | sha1[1];
325 struct object_entry *e = object_table[h];
326 struct object_entry *p = NULL;
328 while (e) {
329 if (!hashcmp(sha1, e->sha1))
330 return e;
331 p = e;
332 e = e->next;
333 }
335 e = new_object(sha1);
336 e->next = NULL;
337 e->offset = 0;
338 if (p)
339 p->next = e;
340 else
341 object_table[h] = e;
342 return e;
343 }
345 static unsigned int hc_str(const char *s, size_t len)
346 {
347 unsigned int r = 0;
348 while (len-- > 0)
349 r = r * 31 + *s++;
350 return r;
351 }
353 static void* pool_alloc(size_t len)
354 {
355 struct mem_pool *p;
356 void *r;
358 for (p = mem_pool; p; p = p->next_pool)
359 if ((p->end - p->next_free >= len))
360 break;
362 if (!p) {
363 if (len >= (mem_pool_alloc/2)) {
364 total_allocd += len;
365 return xmalloc(len);
366 }
367 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
368 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
369 p->next_pool = mem_pool;
370 p->next_free = p->space;
371 p->end = p->next_free + mem_pool_alloc;
372 mem_pool = p;
373 }
375 r = p->next_free;
376 /* round out to a pointer alignment */
377 if (len & (sizeof(void*) - 1))
378 len += sizeof(void*) - (len & (sizeof(void*) - 1));
379 p->next_free += len;
380 return r;
381 }
383 static void* pool_calloc(size_t count, size_t size)
384 {
385 size_t len = count * size;
386 void *r = pool_alloc(len);
387 memset(r, 0, len);
388 return r;
389 }
391 static char* pool_strdup(const char *s)
392 {
393 char *r = pool_alloc(strlen(s) + 1);
394 strcpy(r, s);
395 return r;
396 }
398 static void size_dbuf(struct dbuf *b, size_t maxlen)
399 {
400 if (b->buffer) {
401 if (b->capacity >= maxlen)
402 return;
403 free(b->buffer);
404 }
405 b->capacity = ((maxlen / 1024) + 1) * 1024;
406 b->buffer = xmalloc(b->capacity);
407 }
409 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
410 {
411 struct mark_set *s = marks;
412 while ((idnum >> s->shift) >= 1024) {
413 s = pool_calloc(1, sizeof(struct mark_set));
414 s->shift = marks->shift + 10;
415 s->data.sets[0] = marks;
416 marks = s;
417 }
418 while (s->shift) {
419 uintmax_t i = idnum >> s->shift;
420 idnum -= i << s->shift;
421 if (!s->data.sets[i]) {
422 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
423 s->data.sets[i]->shift = s->shift - 10;
424 }
425 s = s->data.sets[i];
426 }
427 if (!s->data.marked[idnum])
428 marks_set_count++;
429 s->data.marked[idnum] = oe;
430 }
432 static struct object_entry* find_mark(uintmax_t idnum)
433 {
434 uintmax_t orig_idnum = idnum;
435 struct mark_set *s = marks;
436 struct object_entry *oe = NULL;
437 if ((idnum >> s->shift) < 1024) {
438 while (s && s->shift) {
439 uintmax_t i = idnum >> s->shift;
440 idnum -= i << s->shift;
441 s = s->data.sets[i];
442 }
443 if (s)
444 oe = s->data.marked[idnum];
445 }
446 if (!oe)
447 die("mark :%ju not declared", orig_idnum);
448 return oe;
449 }
451 static struct atom_str* to_atom(const char *s, size_t len)
452 {
453 unsigned int hc = hc_str(s, len) % atom_table_sz;
454 struct atom_str *c;
456 for (c = atom_table[hc]; c; c = c->next_atom)
457 if (c->str_len == len && !strncmp(s, c->str_dat, len))
458 return c;
460 c = pool_alloc(sizeof(struct atom_str) + len + 1);
461 c->str_len = len;
462 strncpy(c->str_dat, s, len);
463 c->str_dat[len] = 0;
464 c->next_atom = atom_table[hc];
465 atom_table[hc] = c;
466 atom_cnt++;
467 return c;
468 }
470 static struct branch* lookup_branch(const char *name)
471 {
472 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
473 struct branch *b;
475 for (b = branch_table[hc]; b; b = b->table_next_branch)
476 if (!strcmp(name, b->name))
477 return b;
478 return NULL;
479 }
481 static struct branch* new_branch(const char *name)
482 {
483 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
484 struct branch* b = lookup_branch(name);
486 if (b)
487 die("Invalid attempt to create duplicate branch: %s", name);
488 if (check_ref_format(name))
489 die("Branch name doesn't conform to GIT standards: %s", name);
491 b = pool_calloc(1, sizeof(struct branch));
492 b->name = pool_strdup(name);
493 b->table_next_branch = branch_table[hc];
494 b->branch_tree.versions[0].mode = S_IFDIR;
495 b->branch_tree.versions[1].mode = S_IFDIR;
496 branch_table[hc] = b;
497 branch_count++;
498 return b;
499 }
501 static unsigned int hc_entries(unsigned int cnt)
502 {
503 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
504 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
505 }
507 static struct tree_content* new_tree_content(unsigned int cnt)
508 {
509 struct avail_tree_content *f, *l = NULL;
510 struct tree_content *t;
511 unsigned int hc = hc_entries(cnt);
513 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
514 if (f->entry_capacity >= cnt)
515 break;
517 if (f) {
518 if (l)
519 l->next_avail = f->next_avail;
520 else
521 avail_tree_table[hc] = f->next_avail;
522 } else {
523 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
524 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
525 f->entry_capacity = cnt;
526 }
528 t = (struct tree_content*)f;
529 t->entry_count = 0;
530 t->delta_depth = 0;
531 return t;
532 }
534 static void release_tree_entry(struct tree_entry *e);
535 static void release_tree_content(struct tree_content *t)
536 {
537 struct avail_tree_content *f = (struct avail_tree_content*)t;
538 unsigned int hc = hc_entries(f->entry_capacity);
539 f->next_avail = avail_tree_table[hc];
540 avail_tree_table[hc] = f;
541 }
543 static void release_tree_content_recursive(struct tree_content *t)
544 {
545 unsigned int i;
546 for (i = 0; i < t->entry_count; i++)
547 release_tree_entry(t->entries[i]);
548 release_tree_content(t);
549 }
551 static struct tree_content* grow_tree_content(
552 struct tree_content *t,
553 int amt)
554 {
555 struct tree_content *r = new_tree_content(t->entry_count + amt);
556 r->entry_count = t->entry_count;
557 r->delta_depth = t->delta_depth;
558 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
559 release_tree_content(t);
560 return r;
561 }
563 static struct tree_entry* new_tree_entry()
564 {
565 struct tree_entry *e;
567 if (!avail_tree_entry) {
568 unsigned int n = tree_entry_alloc;
569 total_allocd += n * sizeof(struct tree_entry);
570 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
571 while (n-- > 1) {
572 *((void**)e) = e + 1;
573 e++;
574 }
575 *((void**)e) = NULL;
576 }
578 e = avail_tree_entry;
579 avail_tree_entry = *((void**)e);
580 return e;
581 }
583 static void release_tree_entry(struct tree_entry *e)
584 {
585 if (e->tree)
586 release_tree_content_recursive(e->tree);
587 *((void**)e) = avail_tree_entry;
588 avail_tree_entry = e;
589 }
591 static void start_packfile()
592 {
593 static char tmpfile[PATH_MAX];
594 struct packed_git *p;
595 struct pack_header hdr;
596 int pack_fd;
598 snprintf(tmpfile, sizeof(tmpfile),
599 "%s/pack_XXXXXX", get_object_directory());
600 pack_fd = mkstemp(tmpfile);
601 if (pack_fd < 0)
602 die("Can't create %s: %s", tmpfile, strerror(errno));
603 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
604 strcpy(p->pack_name, tmpfile);
605 p->pack_fd = pack_fd;
607 hdr.hdr_signature = htonl(PACK_SIGNATURE);
608 hdr.hdr_version = htonl(2);
609 hdr.hdr_entries = 0;
610 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
612 pack_data = p;
613 pack_size = sizeof(hdr);
614 object_count = 0;
616 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
617 all_packs[pack_id] = p;
618 }
620 static void fixup_header_footer()
621 {
622 int pack_fd = pack_data->pack_fd;
623 SHA_CTX c;
624 char hdr[8];
625 unsigned long cnt;
626 char *buf;
628 if (lseek(pack_fd, 0, SEEK_SET) != 0)
629 die("Failed seeking to start: %s", strerror(errno));
631 SHA1_Init(&c);
632 if (read_in_full(pack_fd, hdr, 8) != 8)
633 die("Unable to reread header of %s", pack_data->pack_name);
634 SHA1_Update(&c, hdr, 8);
636 cnt = htonl(object_count);
637 SHA1_Update(&c, &cnt, 4);
638 write_or_die(pack_fd, &cnt, 4);
640 buf = xmalloc(128 * 1024);
641 for (;;) {
642 size_t n = xread(pack_fd, buf, 128 * 1024);
643 if (n <= 0)
644 break;
645 SHA1_Update(&c, buf, n);
646 }
647 free(buf);
649 SHA1_Final(pack_data->sha1, &c);
650 write_or_die(pack_fd, pack_data->sha1, sizeof(pack_data->sha1));
651 close(pack_fd);
652 }
654 static int oecmp (const void *a_, const void *b_)
655 {
656 struct object_entry *a = *((struct object_entry**)a_);
657 struct object_entry *b = *((struct object_entry**)b_);
658 return hashcmp(a->sha1, b->sha1);
659 }
661 static char* create_index()
662 {
663 static char tmpfile[PATH_MAX];
664 SHA_CTX ctx;
665 struct sha1file *f;
666 struct object_entry **idx, **c, **last, *e;
667 struct object_entry_pool *o;
668 unsigned int array[256];
669 int i, idx_fd;
671 /* Build the sorted table of object IDs. */
672 idx = xmalloc(object_count * sizeof(struct object_entry*));
673 c = idx;
674 for (o = blocks; o; o = o->next_pool)
675 for (e = o->next_free; e-- != o->entries;)
676 if (pack_id == e->pack_id)
677 *c++ = e;
678 last = idx + object_count;
679 if (c != last)
680 die("internal consistency error creating the index");
681 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
683 /* Generate the fan-out array. */
684 c = idx;
685 for (i = 0; i < 256; i++) {
686 struct object_entry **next = c;;
687 while (next < last) {
688 if ((*next)->sha1[0] != i)
689 break;
690 next++;
691 }
692 array[i] = htonl(next - idx);
693 c = next;
694 }
696 snprintf(tmpfile, sizeof(tmpfile),
697 "%s/index_XXXXXX", get_object_directory());
698 idx_fd = mkstemp(tmpfile);
699 if (idx_fd < 0)
700 die("Can't create %s: %s", tmpfile, strerror(errno));
701 f = sha1fd(idx_fd, tmpfile);
702 sha1write(f, array, 256 * sizeof(int));
703 SHA1_Init(&ctx);
704 for (c = idx; c != last; c++) {
705 unsigned int offset = htonl((*c)->offset);
706 sha1write(f, &offset, 4);
707 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
708 SHA1_Update(&ctx, (*c)->sha1, 20);
709 }
710 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
711 sha1close(f, NULL, 1);
712 free(idx);
713 SHA1_Final(pack_data->sha1, &ctx);
714 return tmpfile;
715 }
717 static char* keep_pack(char *curr_index_name)
718 {
719 static char name[PATH_MAX];
720 static char *keep_msg = "fast-import";
721 int keep_fd;
723 chmod(pack_data->pack_name, 0444);
724 chmod(curr_index_name, 0444);
726 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
727 get_object_directory(), sha1_to_hex(pack_data->sha1));
728 keep_fd = open(name, O_RDWR|O_CREAT|O_EXCL, 0600);
729 if (keep_fd < 0)
730 die("cannot create keep file");
731 write(keep_fd, keep_msg, strlen(keep_msg));
732 close(keep_fd);
734 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
735 get_object_directory(), sha1_to_hex(pack_data->sha1));
736 if (move_temp_to_file(pack_data->pack_name, name))
737 die("cannot store pack file");
739 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
740 get_object_directory(), sha1_to_hex(pack_data->sha1));
741 if (move_temp_to_file(curr_index_name, name))
742 die("cannot store index file");
743 return name;
744 }
746 static void unkeep_all_packs()
747 {
748 static char name[PATH_MAX];
749 int k;
751 for (k = 0; k < pack_id; k++) {
752 struct packed_git *p = all_packs[k];
753 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
754 get_object_directory(), sha1_to_hex(p->sha1));
755 unlink(name);
756 }
757 }
759 static void end_packfile()
760 {
761 struct packed_git *old_p = pack_data, *new_p;
763 if (object_count) {
764 char *idx_name;
765 int i;
766 struct branch *b;
767 struct tag *t;
769 fixup_header_footer();
770 idx_name = keep_pack(create_index());
772 /* Register the packfile with core git's machinary. */
773 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
774 if (!new_p)
775 die("core git rejected index %s", idx_name);
776 new_p->windows = old_p->windows;
777 all_packs[pack_id] = new_p;
778 install_packed_git(new_p);
780 /* Print the boundary */
781 fprintf(stdout, "%s:", new_p->pack_name);
782 for (i = 0; i < branch_table_sz; i++) {
783 for (b = branch_table[i]; b; b = b->table_next_branch) {
784 if (b->pack_id == pack_id)
785 fprintf(stdout, " %s", sha1_to_hex(b->sha1));
786 }
787 }
788 for (t = first_tag; t; t = t->next_tag) {
789 if (t->pack_id == pack_id)
790 fprintf(stdout, " %s", sha1_to_hex(t->sha1));
791 }
792 fputc('\n', stdout);
794 pack_id++;
795 }
796 else
797 unlink(old_p->pack_name);
798 free(old_p);
800 /* We can't carry a delta across packfiles. */
801 free(last_blob.data);
802 last_blob.data = NULL;
803 last_blob.len = 0;
804 last_blob.offset = 0;
805 last_blob.depth = 0;
806 }
808 static void checkpoint()
809 {
810 end_packfile();
811 start_packfile();
812 }
814 static size_t encode_header(
815 enum object_type type,
816 size_t size,
817 unsigned char *hdr)
818 {
819 int n = 1;
820 unsigned char c;
822 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
823 die("bad type %d", type);
825 c = (type << 4) | (size & 15);
826 size >>= 4;
827 while (size) {
828 *hdr++ = c | 0x80;
829 c = size & 0x7f;
830 size >>= 7;
831 n++;
832 }
833 *hdr = c;
834 return n;
835 }
837 static int store_object(
838 enum object_type type,
839 void *dat,
840 size_t datlen,
841 struct last_object *last,
842 unsigned char *sha1out,
843 uintmax_t mark)
844 {
845 void *out, *delta;
846 struct object_entry *e;
847 unsigned char hdr[96];
848 unsigned char sha1[20];
849 unsigned long hdrlen, deltalen;
850 SHA_CTX c;
851 z_stream s;
853 hdrlen = sprintf((char*)hdr,"%s %lu",type_names[type],datlen) + 1;
854 SHA1_Init(&c);
855 SHA1_Update(&c, hdr, hdrlen);
856 SHA1_Update(&c, dat, datlen);
857 SHA1_Final(sha1, &c);
858 if (sha1out)
859 hashcpy(sha1out, sha1);
861 e = insert_object(sha1);
862 if (mark)
863 insert_mark(mark, e);
864 if (e->offset) {
865 duplicate_count_by_type[type]++;
866 return 1;
867 }
869 if (last && last->data && last->depth < max_depth) {
870 delta = diff_delta(last->data, last->len,
871 dat, datlen,
872 &deltalen, 0);
873 if (delta && deltalen >= datlen) {
874 free(delta);
875 delta = NULL;
876 }
877 } else
878 delta = NULL;
880 memset(&s, 0, sizeof(s));
881 deflateInit(&s, zlib_compression_level);
882 if (delta) {
883 s.next_in = delta;
884 s.avail_in = deltalen;
885 } else {
886 s.next_in = dat;
887 s.avail_in = datlen;
888 }
889 s.avail_out = deflateBound(&s, s.avail_in);
890 s.next_out = out = xmalloc(s.avail_out);
891 while (deflate(&s, Z_FINISH) == Z_OK)
892 /* nothing */;
893 deflateEnd(&s);
895 /* Determine if we should auto-checkpoint. */
896 if ((object_count + 1) > max_objects
897 || (object_count + 1) < object_count
898 || (pack_size + 60 + s.total_out) > max_packsize
899 || (pack_size + 60 + s.total_out) < pack_size) {
901 /* This new object needs to *not* have the current pack_id. */
902 e->pack_id = pack_id + 1;
903 checkpoint();
905 /* We cannot carry a delta into the new pack. */
906 if (delta) {
907 free(delta);
908 delta = NULL;
910 memset(&s, 0, sizeof(s));
911 deflateInit(&s, zlib_compression_level);
912 s.next_in = dat;
913 s.avail_in = datlen;
914 s.avail_out = deflateBound(&s, s.avail_in);
915 s.next_out = out = xrealloc(out, s.avail_out);
916 while (deflate(&s, Z_FINISH) == Z_OK)
917 /* nothing */;
918 deflateEnd(&s);
919 }
920 }
922 e->type = type;
923 e->pack_id = pack_id;
924 e->offset = pack_size;
925 object_count++;
926 object_count_by_type[type]++;
928 if (delta) {
929 unsigned long ofs = e->offset - last->offset;
930 unsigned pos = sizeof(hdr) - 1;
932 delta_count_by_type[type]++;
933 last->depth++;
935 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
936 write_or_die(pack_data->pack_fd, hdr, hdrlen);
937 pack_size += hdrlen;
939 hdr[pos] = ofs & 127;
940 while (ofs >>= 7)
941 hdr[--pos] = 128 | (--ofs & 127);
942 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
943 pack_size += sizeof(hdr) - pos;
944 } else {
945 if (last)
946 last->depth = 0;
947 hdrlen = encode_header(type, datlen, hdr);
948 write_or_die(pack_data->pack_fd, hdr, hdrlen);
949 pack_size += hdrlen;
950 }
952 write_or_die(pack_data->pack_fd, out, s.total_out);
953 pack_size += s.total_out;
955 free(out);
956 if (delta)
957 free(delta);
958 if (last) {
959 if (last->data && !last->no_free)
960 free(last->data);
961 last->data = dat;
962 last->offset = e->offset;
963 last->len = datlen;
964 }
965 return 0;
966 }
968 static void *gfi_unpack_entry(
969 struct object_entry *oe,
970 unsigned long *sizep)
971 {
972 static char type[20];
973 struct packed_git *p = all_packs[oe->pack_id];
974 if (p == pack_data)
975 p->pack_size = pack_size + 20;
976 return unpack_entry(p, oe->offset, type, sizep);
977 }
979 static const char *get_mode(const char *str, unsigned int *modep)
980 {
981 unsigned char c;
982 unsigned int mode = 0;
984 while ((c = *str++) != ' ') {
985 if (c < '0' || c > '7')
986 return NULL;
987 mode = (mode << 3) + (c - '0');
988 }
989 *modep = mode;
990 return str;
991 }
993 static void load_tree(struct tree_entry *root)
994 {
995 unsigned char* sha1 = root->versions[1].sha1;
996 struct object_entry *myoe;
997 struct tree_content *t;
998 unsigned long size;
999 char *buf;
1000 const char *c;
1002 root->tree = t = new_tree_content(8);
1003 if (is_null_sha1(sha1))
1004 return;
1006 myoe = find_object(sha1);
1007 if (myoe) {
1008 if (myoe->type != OBJ_TREE)
1009 die("Not a tree: %s", sha1_to_hex(sha1));
1010 t->delta_depth = 0;
1011 buf = gfi_unpack_entry(myoe, &size);
1012 } else {
1013 char type[20];
1014 buf = read_sha1_file(sha1, type, &size);
1015 if (!buf || strcmp(type, tree_type))
1016 die("Can't load tree %s", sha1_to_hex(sha1));
1017 }
1019 c = buf;
1020 while (c != (buf + size)) {
1021 struct tree_entry *e = new_tree_entry();
1023 if (t->entry_count == t->entry_capacity)
1024 root->tree = t = grow_tree_content(t, 8);
1025 t->entries[t->entry_count++] = e;
1027 e->tree = NULL;
1028 c = get_mode(c, &e->versions[1].mode);
1029 if (!c)
1030 die("Corrupt mode in %s", sha1_to_hex(sha1));
1031 e->versions[0].mode = e->versions[1].mode;
1032 e->name = to_atom(c, strlen(c));
1033 c += e->name->str_len + 1;
1034 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1035 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1036 c += 20;
1037 }
1038 free(buf);
1039 }
1041 static int tecmp0 (const void *_a, const void *_b)
1042 {
1043 struct tree_entry *a = *((struct tree_entry**)_a);
1044 struct tree_entry *b = *((struct tree_entry**)_b);
1045 return base_name_compare(
1046 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1047 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1048 }
1050 static int tecmp1 (const void *_a, const void *_b)
1051 {
1052 struct tree_entry *a = *((struct tree_entry**)_a);
1053 struct tree_entry *b = *((struct tree_entry**)_b);
1054 return base_name_compare(
1055 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1056 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1057 }
1059 static void mktree(struct tree_content *t,
1060 int v,
1061 unsigned long *szp,
1062 struct dbuf *b)
1063 {
1064 size_t maxlen = 0;
1065 unsigned int i;
1066 char *c;
1068 if (!v)
1069 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1070 else
1071 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1073 for (i = 0; i < t->entry_count; i++) {
1074 if (t->entries[i]->versions[v].mode)
1075 maxlen += t->entries[i]->name->str_len + 34;
1076 }
1078 size_dbuf(b, maxlen);
1079 c = b->buffer;
1080 for (i = 0; i < t->entry_count; i++) {
1081 struct tree_entry *e = t->entries[i];
1082 if (!e->versions[v].mode)
1083 continue;
1084 c += sprintf(c, "%o", e->versions[v].mode);
1085 *c++ = ' ';
1086 strcpy(c, e->name->str_dat);
1087 c += e->name->str_len + 1;
1088 hashcpy((unsigned char*)c, e->versions[v].sha1);
1089 c += 20;
1090 }
1091 *szp = c - (char*)b->buffer;
1092 }
1094 static void store_tree(struct tree_entry *root)
1095 {
1096 struct tree_content *t = root->tree;
1097 unsigned int i, j, del;
1098 unsigned long new_len;
1099 struct last_object lo;
1100 struct object_entry *le;
1102 if (!is_null_sha1(root->versions[1].sha1))
1103 return;
1105 for (i = 0; i < t->entry_count; i++) {
1106 if (t->entries[i]->tree)
1107 store_tree(t->entries[i]);
1108 }
1110 le = find_object(root->versions[0].sha1);
1111 if (!S_ISDIR(root->versions[0].mode)
1112 || !le
1113 || le->pack_id != pack_id) {
1114 lo.data = NULL;
1115 lo.depth = 0;
1116 } else {
1117 mktree(t, 0, &lo.len, &old_tree);
1118 lo.data = old_tree.buffer;
1119 lo.offset = le->offset;
1120 lo.depth = t->delta_depth;
1121 lo.no_free = 1;
1122 }
1124 mktree(t, 1, &new_len, &new_tree);
1125 store_object(OBJ_TREE, new_tree.buffer, new_len,
1126 &lo, root->versions[1].sha1, 0);
1128 t->delta_depth = lo.depth;
1129 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1130 struct tree_entry *e = t->entries[i];
1131 if (e->versions[1].mode) {
1132 e->versions[0].mode = e->versions[1].mode;
1133 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1134 t->entries[j++] = e;
1135 } else {
1136 release_tree_entry(e);
1137 del++;
1138 }
1139 }
1140 t->entry_count -= del;
1141 }
1143 static int tree_content_set(
1144 struct tree_entry *root,
1145 const char *p,
1146 const unsigned char *sha1,
1147 const unsigned int mode)
1148 {
1149 struct tree_content *t = root->tree;
1150 const char *slash1;
1151 unsigned int i, n;
1152 struct tree_entry *e;
1154 slash1 = strchr(p, '/');
1155 if (slash1)
1156 n = slash1 - p;
1157 else
1158 n = strlen(p);
1160 for (i = 0; i < t->entry_count; i++) {
1161 e = t->entries[i];
1162 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1163 if (!slash1) {
1164 if (e->versions[1].mode == mode
1165 && !hashcmp(e->versions[1].sha1, sha1))
1166 return 0;
1167 e->versions[1].mode = mode;
1168 hashcpy(e->versions[1].sha1, sha1);
1169 if (e->tree) {
1170 release_tree_content_recursive(e->tree);
1171 e->tree = NULL;
1172 }
1173 hashclr(root->versions[1].sha1);
1174 return 1;
1175 }
1176 if (!S_ISDIR(e->versions[1].mode)) {
1177 e->tree = new_tree_content(8);
1178 e->versions[1].mode = S_IFDIR;
1179 }
1180 if (!e->tree)
1181 load_tree(e);
1182 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
1183 hashclr(root->versions[1].sha1);
1184 return 1;
1185 }
1186 return 0;
1187 }
1188 }
1190 if (t->entry_count == t->entry_capacity)
1191 root->tree = t = grow_tree_content(t, 8);
1192 e = new_tree_entry();
1193 e->name = to_atom(p, n);
1194 e->versions[0].mode = 0;
1195 hashclr(e->versions[0].sha1);
1196 t->entries[t->entry_count++] = e;
1197 if (slash1) {
1198 e->tree = new_tree_content(8);
1199 e->versions[1].mode = S_IFDIR;
1200 tree_content_set(e, slash1 + 1, sha1, mode);
1201 } else {
1202 e->tree = NULL;
1203 e->versions[1].mode = mode;
1204 hashcpy(e->versions[1].sha1, sha1);
1205 }
1206 hashclr(root->versions[1].sha1);
1207 return 1;
1208 }
1210 static int tree_content_remove(struct tree_entry *root, const char *p)
1211 {
1212 struct tree_content *t = root->tree;
1213 const char *slash1;
1214 unsigned int i, n;
1215 struct tree_entry *e;
1217 slash1 = strchr(p, '/');
1218 if (slash1)
1219 n = slash1 - p;
1220 else
1221 n = strlen(p);
1223 for (i = 0; i < t->entry_count; i++) {
1224 e = t->entries[i];
1225 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1226 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1227 goto del_entry;
1228 if (!e->tree)
1229 load_tree(e);
1230 if (tree_content_remove(e, slash1 + 1)) {
1231 for (n = 0; n < e->tree->entry_count; n++) {
1232 if (e->tree->entries[n]->versions[1].mode) {
1233 hashclr(root->versions[1].sha1);
1234 return 1;
1235 }
1236 }
1237 goto del_entry;
1238 }
1239 return 0;
1240 }
1241 }
1242 return 0;
1244 del_entry:
1245 if (e->tree) {
1246 release_tree_content_recursive(e->tree);
1247 e->tree = NULL;
1248 }
1249 e->versions[1].mode = 0;
1250 hashclr(e->versions[1].sha1);
1251 hashclr(root->versions[1].sha1);
1252 return 1;
1253 }
1255 static void dump_branches()
1256 {
1257 static const char *msg = "fast-import";
1258 unsigned int i;
1259 struct branch *b;
1260 struct ref_lock *lock;
1262 for (i = 0; i < branch_table_sz; i++) {
1263 for (b = branch_table[i]; b; b = b->table_next_branch) {
1264 lock = lock_any_ref_for_update(b->name, NULL);
1265 if (!lock || write_ref_sha1(lock, b->sha1, msg) < 0)
1266 die("Can't write %s", b->name);
1267 }
1268 }
1269 }
1271 static void dump_tags()
1272 {
1273 static const char *msg = "fast-import";
1274 struct tag *t;
1275 struct ref_lock *lock;
1276 char path[PATH_MAX];
1278 for (t = first_tag; t; t = t->next_tag) {
1279 sprintf(path, "refs/tags/%s", t->name);
1280 lock = lock_any_ref_for_update(path, NULL);
1281 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1282 die("Can't write %s", path);
1283 }
1284 }
1286 static void dump_marks_helper(FILE *f,
1287 uintmax_t base,
1288 struct mark_set *m)
1289 {
1290 uintmax_t k;
1291 if (m->shift) {
1292 for (k = 0; k < 1024; k++) {
1293 if (m->data.sets[k])
1294 dump_marks_helper(f, (base + k) << m->shift,
1295 m->data.sets[k]);
1296 }
1297 } else {
1298 for (k = 0; k < 1024; k++) {
1299 if (m->data.marked[k])
1300 fprintf(f, ":%ju %s\n", base + k,
1301 sha1_to_hex(m->data.marked[k]->sha1));
1302 }
1303 }
1304 }
1306 static void dump_marks()
1307 {
1308 if (mark_file)
1309 {
1310 FILE *f = fopen(mark_file, "w");
1311 dump_marks_helper(f, 0, marks);
1312 fclose(f);
1313 }
1314 }
1316 static void read_next_command()
1317 {
1318 read_line(&command_buf, stdin, '\n');
1319 }
1321 static void cmd_mark()
1322 {
1323 if (!strncmp("mark :", command_buf.buf, 6)) {
1324 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1325 read_next_command();
1326 }
1327 else
1328 next_mark = 0;
1329 }
1331 static void* cmd_data (size_t *size)
1332 {
1333 size_t n = 0;
1334 void *buffer;
1335 size_t length;
1337 if (strncmp("data ", command_buf.buf, 5))
1338 die("Expected 'data n' command, found: %s", command_buf.buf);
1340 length = strtoul(command_buf.buf + 5, NULL, 10);
1341 buffer = xmalloc(length);
1343 while (n < length) {
1344 size_t s = fread((char*)buffer + n, 1, length - n, stdin);
1345 if (!s && feof(stdin))
1346 die("EOF in data (%lu bytes remaining)", length - n);
1347 n += s;
1348 }
1350 if (fgetc(stdin) != '\n')
1351 die("An lf did not trail the binary data as expected.");
1353 *size = length;
1354 return buffer;
1355 }
1357 static void cmd_new_blob()
1358 {
1359 size_t l;
1360 void *d;
1362 read_next_command();
1363 cmd_mark();
1364 d = cmd_data(&l);
1366 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1367 free(d);
1368 }
1370 static void unload_one_branch()
1371 {
1372 while (cur_active_branches
1373 && cur_active_branches >= max_active_branches) {
1374 unsigned long min_commit = ULONG_MAX;
1375 struct branch *e, *l = NULL, *p = NULL;
1377 for (e = active_branches; e; e = e->active_next_branch) {
1378 if (e->last_commit < min_commit) {
1379 p = l;
1380 min_commit = e->last_commit;
1381 }
1382 l = e;
1383 }
1385 if (p) {
1386 e = p->active_next_branch;
1387 p->active_next_branch = e->active_next_branch;
1388 } else {
1389 e = active_branches;
1390 active_branches = e->active_next_branch;
1391 }
1392 e->active_next_branch = NULL;
1393 if (e->branch_tree.tree) {
1394 release_tree_content_recursive(e->branch_tree.tree);
1395 e->branch_tree.tree = NULL;
1396 }
1397 cur_active_branches--;
1398 }
1399 }
1401 static void load_branch(struct branch *b)
1402 {
1403 load_tree(&b->branch_tree);
1404 b->active_next_branch = active_branches;
1405 active_branches = b;
1406 cur_active_branches++;
1407 branch_load_count++;
1408 }
1410 static void file_change_m(struct branch *b)
1411 {
1412 const char *p = command_buf.buf + 2;
1413 char *p_uq;
1414 const char *endp;
1415 struct object_entry *oe;
1416 unsigned char sha1[20];
1417 unsigned int mode;
1418 char type[20];
1420 p = get_mode(p, &mode);
1421 if (!p)
1422 die("Corrupt mode: %s", command_buf.buf);
1423 switch (mode) {
1424 case S_IFREG | 0644:
1425 case S_IFREG | 0755:
1426 case S_IFLNK:
1427 case 0644:
1428 case 0755:
1429 /* ok */
1430 break;
1431 default:
1432 die("Corrupt mode: %s", command_buf.buf);
1433 }
1435 if (*p == ':') {
1436 char *x;
1437 oe = find_mark(strtoumax(p + 1, &x, 10));
1438 hashcpy(sha1, oe->sha1);
1439 p = x;
1440 } else {
1441 if (get_sha1_hex(p, sha1))
1442 die("Invalid SHA1: %s", command_buf.buf);
1443 oe = find_object(sha1);
1444 p += 40;
1445 }
1446 if (*p++ != ' ')
1447 die("Missing space after SHA1: %s", command_buf.buf);
1449 p_uq = unquote_c_style(p, &endp);
1450 if (p_uq) {
1451 if (*endp)
1452 die("Garbage after path in: %s", command_buf.buf);
1453 p = p_uq;
1454 }
1456 if (oe) {
1457 if (oe->type != OBJ_BLOB)
1458 die("Not a blob (actually a %s): %s",
1459 command_buf.buf, type_names[oe->type]);
1460 } else {
1461 if (sha1_object_info(sha1, type, NULL))
1462 die("Blob not found: %s", command_buf.buf);
1463 if (strcmp(blob_type, type))
1464 die("Not a blob (actually a %s): %s",
1465 command_buf.buf, type);
1466 }
1468 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1470 if (p_uq)
1471 free(p_uq);
1472 }
1474 static void file_change_d(struct branch *b)
1475 {
1476 const char *p = command_buf.buf + 2;
1477 char *p_uq;
1478 const char *endp;
1480 p_uq = unquote_c_style(p, &endp);
1481 if (p_uq) {
1482 if (*endp)
1483 die("Garbage after path in: %s", command_buf.buf);
1484 p = p_uq;
1485 }
1486 tree_content_remove(&b->branch_tree, p);
1487 if (p_uq)
1488 free(p_uq);
1489 }
1491 static void cmd_from(struct branch *b)
1492 {
1493 const char *from, *endp;
1494 char *str_uq;
1495 struct branch *s;
1497 if (strncmp("from ", command_buf.buf, 5))
1498 return;
1500 if (b->last_commit)
1501 die("Can't reinitailize branch %s", b->name);
1503 from = strchr(command_buf.buf, ' ') + 1;
1504 str_uq = unquote_c_style(from, &endp);
1505 if (str_uq) {
1506 if (*endp)
1507 die("Garbage after string in: %s", command_buf.buf);
1508 from = str_uq;
1509 }
1511 s = lookup_branch(from);
1512 if (b == s)
1513 die("Can't create a branch from itself: %s", b->name);
1514 else if (s) {
1515 unsigned char *t = s->branch_tree.versions[1].sha1;
1516 hashcpy(b->sha1, s->sha1);
1517 hashcpy(b->branch_tree.versions[0].sha1, t);
1518 hashcpy(b->branch_tree.versions[1].sha1, t);
1519 } else if (*from == ':') {
1520 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1521 struct object_entry *oe = find_mark(idnum);
1522 unsigned long size;
1523 char *buf;
1524 if (oe->type != OBJ_COMMIT)
1525 die("Mark :%ju not a commit", idnum);
1526 hashcpy(b->sha1, oe->sha1);
1527 buf = gfi_unpack_entry(oe, &size);
1528 if (!buf || size < 46)
1529 die("Not a valid commit: %s", from);
1530 if (memcmp("tree ", buf, 5)
1531 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1532 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1533 free(buf);
1534 hashcpy(b->branch_tree.versions[0].sha1,
1535 b->branch_tree.versions[1].sha1);
1536 } else if (!get_sha1(from, b->sha1)) {
1537 if (is_null_sha1(b->sha1)) {
1538 hashclr(b->branch_tree.versions[0].sha1);
1539 hashclr(b->branch_tree.versions[1].sha1);
1540 } else {
1541 unsigned long size;
1542 char *buf;
1544 buf = read_object_with_reference(b->sha1,
1545 type_names[OBJ_COMMIT], &size, b->sha1);
1546 if (!buf || size < 46)
1547 die("Not a valid commit: %s", from);
1548 if (memcmp("tree ", buf, 5)
1549 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1550 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1551 free(buf);
1552 hashcpy(b->branch_tree.versions[0].sha1,
1553 b->branch_tree.versions[1].sha1);
1554 }
1555 } else
1556 die("Invalid ref name or SHA1 expression: %s", from);
1558 read_next_command();
1559 }
1561 static struct hash_list* cmd_merge(unsigned int *count)
1562 {
1563 struct hash_list *list = NULL, *n, *e;
1564 const char *from, *endp;
1565 char *str_uq;
1566 struct branch *s;
1568 *count = 0;
1569 while (!strncmp("merge ", command_buf.buf, 6)) {
1570 from = strchr(command_buf.buf, ' ') + 1;
1571 str_uq = unquote_c_style(from, &endp);
1572 if (str_uq) {
1573 if (*endp)
1574 die("Garbage after string in: %s", command_buf.buf);
1575 from = str_uq;
1576 }
1578 n = xmalloc(sizeof(*n));
1579 s = lookup_branch(from);
1580 if (s)
1581 hashcpy(n->sha1, s->sha1);
1582 else if (*from == ':') {
1583 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1584 struct object_entry *oe = find_mark(idnum);
1585 if (oe->type != OBJ_COMMIT)
1586 die("Mark :%ju not a commit", idnum);
1587 hashcpy(n->sha1, oe->sha1);
1588 } else if (get_sha1(from, n->sha1))
1589 die("Invalid ref name or SHA1 expression: %s", from);
1591 n->next = NULL;
1592 if (list)
1593 e->next = n;
1594 else
1595 list = n;
1596 e = n;
1597 *count++;
1598 read_next_command();
1599 }
1600 return list;
1601 }
1603 static void cmd_new_commit()
1604 {
1605 struct branch *b;
1606 void *msg;
1607 size_t msglen;
1608 char *str_uq;
1609 const char *endp;
1610 char *sp;
1611 char *author = NULL;
1612 char *committer = NULL;
1613 struct hash_list *merge_list = NULL;
1614 unsigned int merge_count;
1616 /* Obtain the branch name from the rest of our command */
1617 sp = strchr(command_buf.buf, ' ') + 1;
1618 str_uq = unquote_c_style(sp, &endp);
1619 if (str_uq) {
1620 if (*endp)
1621 die("Garbage after ref in: %s", command_buf.buf);
1622 sp = str_uq;
1623 }
1624 b = lookup_branch(sp);
1625 if (!b)
1626 b = new_branch(sp);
1627 if (str_uq)
1628 free(str_uq);
1630 read_next_command();
1631 cmd_mark();
1632 if (!strncmp("author ", command_buf.buf, 7)) {
1633 author = strdup(command_buf.buf);
1634 read_next_command();
1635 }
1636 if (!strncmp("committer ", command_buf.buf, 10)) {
1637 committer = strdup(command_buf.buf);
1638 read_next_command();
1639 }
1640 if (!committer)
1641 die("Expected committer but didn't get one");
1642 msg = cmd_data(&msglen);
1643 read_next_command();
1644 cmd_from(b);
1645 merge_list = cmd_merge(&merge_count);
1647 /* ensure the branch is active/loaded */
1648 if (!b->branch_tree.tree || !max_active_branches) {
1649 unload_one_branch();
1650 load_branch(b);
1651 }
1653 /* file_change* */
1654 for (;;) {
1655 if (1 == command_buf.len)
1656 break;
1657 else if (!strncmp("M ", command_buf.buf, 2))
1658 file_change_m(b);
1659 else if (!strncmp("D ", command_buf.buf, 2))
1660 file_change_d(b);
1661 else
1662 die("Unsupported file_change: %s", command_buf.buf);
1663 read_next_command();
1664 }
1666 /* build the tree and the commit */
1667 store_tree(&b->branch_tree);
1668 hashcpy(b->branch_tree.versions[0].sha1,
1669 b->branch_tree.versions[1].sha1);
1670 size_dbuf(&new_data, 97 + msglen
1671 + merge_count * 49
1672 + (author
1673 ? strlen(author) + strlen(committer)
1674 : 2 * strlen(committer)));
1675 sp = new_data.buffer;
1676 sp += sprintf(sp, "tree %s\n",
1677 sha1_to_hex(b->branch_tree.versions[1].sha1));
1678 if (!is_null_sha1(b->sha1))
1679 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1680 while (merge_list) {
1681 struct hash_list *next = merge_list->next;
1682 sp += sprintf(sp, "parent %s\n", sha1_to_hex(merge_list->sha1));
1683 free(merge_list);
1684 merge_list = next;
1685 }
1686 if (author)
1687 sp += sprintf(sp, "%s\n", author);
1688 else
1689 sp += sprintf(sp, "author %s\n", committer + 10);
1690 sp += sprintf(sp, "%s\n\n", committer);
1691 memcpy(sp, msg, msglen);
1692 sp += msglen;
1693 if (author)
1694 free(author);
1695 free(committer);
1696 free(msg);
1698 store_object(OBJ_COMMIT,
1699 new_data.buffer, sp - (char*)new_data.buffer,
1700 NULL, b->sha1, next_mark);
1701 b->last_commit = object_count_by_type[OBJ_COMMIT];
1702 b->pack_id = pack_id;
1704 if (branch_log) {
1705 int need_dq = quote_c_style(b->name, NULL, NULL, 0);
1706 fprintf(branch_log, "commit ");
1707 if (need_dq) {
1708 fputc('"', branch_log);
1709 quote_c_style(b->name, NULL, branch_log, 0);
1710 fputc('"', branch_log);
1711 } else
1712 fprintf(branch_log, "%s", b->name);
1713 fprintf(branch_log," :%ju %s\n",next_mark,sha1_to_hex(b->sha1));
1714 }
1715 }
1717 static void cmd_new_tag()
1718 {
1719 char *str_uq;
1720 const char *endp;
1721 char *sp;
1722 const char *from;
1723 char *tagger;
1724 struct branch *s;
1725 void *msg;
1726 size_t msglen;
1727 struct tag *t;
1728 uintmax_t from_mark = 0;
1729 unsigned char sha1[20];
1731 /* Obtain the new tag name from the rest of our command */
1732 sp = strchr(command_buf.buf, ' ') + 1;
1733 str_uq = unquote_c_style(sp, &endp);
1734 if (str_uq) {
1735 if (*endp)
1736 die("Garbage after tag name in: %s", command_buf.buf);
1737 sp = str_uq;
1738 }
1739 t = pool_alloc(sizeof(struct tag));
1740 t->next_tag = NULL;
1741 t->name = pool_strdup(sp);
1742 if (last_tag)
1743 last_tag->next_tag = t;
1744 else
1745 first_tag = t;
1746 last_tag = t;
1747 if (str_uq)
1748 free(str_uq);
1749 read_next_command();
1751 /* from ... */
1752 if (strncmp("from ", command_buf.buf, 5))
1753 die("Expected from command, got %s", command_buf.buf);
1755 from = strchr(command_buf.buf, ' ') + 1;
1756 str_uq = unquote_c_style(from, &endp);
1757 if (str_uq) {
1758 if (*endp)
1759 die("Garbage after string in: %s", command_buf.buf);
1760 from = str_uq;
1761 }
1763 s = lookup_branch(from);
1764 if (s) {
1765 hashcpy(sha1, s->sha1);
1766 } else if (*from == ':') {
1767 from_mark = strtoumax(from + 1, NULL, 10);
1768 struct object_entry *oe = find_mark(from_mark);
1769 if (oe->type != OBJ_COMMIT)
1770 die("Mark :%ju not a commit", from_mark);
1771 hashcpy(sha1, oe->sha1);
1772 } else if (!get_sha1(from, sha1)) {
1773 unsigned long size;
1774 char *buf;
1776 buf = read_object_with_reference(sha1,
1777 type_names[OBJ_COMMIT], &size, sha1);
1778 if (!buf || size < 46)
1779 die("Not a valid commit: %s", from);
1780 free(buf);
1781 } else
1782 die("Invalid ref name or SHA1 expression: %s", from);
1784 if (str_uq)
1785 free(str_uq);
1786 read_next_command();
1788 /* tagger ... */
1789 if (strncmp("tagger ", command_buf.buf, 7))
1790 die("Expected tagger command, got %s", command_buf.buf);
1791 tagger = strdup(command_buf.buf);
1793 /* tag payload/message */
1794 read_next_command();
1795 msg = cmd_data(&msglen);
1797 /* build the tag object */
1798 size_dbuf(&new_data, 67+strlen(t->name)+strlen(tagger)+msglen);
1799 sp = new_data.buffer;
1800 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
1801 sp += sprintf(sp, "type %s\n", type_names[OBJ_COMMIT]);
1802 sp += sprintf(sp, "tag %s\n", t->name);
1803 sp += sprintf(sp, "%s\n\n", tagger);
1804 memcpy(sp, msg, msglen);
1805 sp += msglen;
1806 free(tagger);
1807 free(msg);
1809 store_object(OBJ_TAG, new_data.buffer, sp - (char*)new_data.buffer,
1810 NULL, t->sha1, 0);
1811 t->pack_id = pack_id;
1813 if (branch_log) {
1814 int need_dq = quote_c_style(t->name, NULL, NULL, 0);
1815 fprintf(branch_log, "tag ");
1816 if (need_dq) {
1817 fputc('"', branch_log);
1818 quote_c_style(t->name, NULL, branch_log, 0);
1819 fputc('"', branch_log);
1820 } else
1821 fprintf(branch_log, "%s", t->name);
1822 fprintf(branch_log," :%ju %s\n",from_mark,sha1_to_hex(t->sha1));
1823 }
1824 }
1826 static void cmd_reset_branch()
1827 {
1828 struct branch *b;
1829 char *str_uq;
1830 const char *endp;
1831 char *sp;
1833 /* Obtain the branch name from the rest of our command */
1834 sp = strchr(command_buf.buf, ' ') + 1;
1835 str_uq = unquote_c_style(sp, &endp);
1836 if (str_uq) {
1837 if (*endp)
1838 die("Garbage after ref in: %s", command_buf.buf);
1839 sp = str_uq;
1840 }
1841 b = lookup_branch(sp);
1842 if (b) {
1843 b->last_commit = 0;
1844 if (b->branch_tree.tree) {
1845 release_tree_content_recursive(b->branch_tree.tree);
1846 b->branch_tree.tree = NULL;
1847 }
1848 }
1849 else
1850 b = new_branch(sp);
1851 if (str_uq)
1852 free(str_uq);
1853 read_next_command();
1854 cmd_from(b);
1855 }
1857 static void cmd_checkpoint()
1858 {
1859 if (object_count)
1860 checkpoint();
1861 read_next_command();
1862 }
1864 static const char fast_import_usage[] =
1865 "git-fast-import [--objects=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file] [--branch-log=log]";
1867 int main(int argc, const char **argv)
1868 {
1869 int i;
1870 uintmax_t est_obj_cnt = object_entry_alloc;
1871 uintmax_t total_count, duplicate_count;
1873 setup_ident();
1874 git_config(git_default_config);
1876 for (i = 1; i < argc; i++) {
1877 const char *a = argv[i];
1879 if (*a != '-' || !strcmp(a, "--"))
1880 break;
1881 else if (!strncmp(a, "--objects=", 10))
1882 est_obj_cnt = strtoumax(a + 10, NULL, 0);
1883 else if (!strncmp(a, "--max-objects-per-pack=", 23))
1884 max_objects = strtoumax(a + 23, NULL, 0);
1885 else if (!strncmp(a, "--max-pack-size=", 16))
1886 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
1887 else if (!strncmp(a, "--depth=", 8))
1888 max_depth = strtoul(a + 8, NULL, 0);
1889 else if (!strncmp(a, "--active-branches=", 18))
1890 max_active_branches = strtoul(a + 18, NULL, 0);
1891 else if (!strncmp(a, "--export-marks=", 15))
1892 mark_file = a + 15;
1893 else if (!strncmp(a, "--branch-log=", 13)) {
1894 branch_log = fopen(a + 13, "w");
1895 if (!branch_log)
1896 die("Can't create %s: %s", a + 13, strerror(errno));
1897 }
1898 else
1899 die("unknown option %s", a);
1900 }
1901 if (i != argc)
1902 usage(fast_import_usage);
1904 alloc_objects(est_obj_cnt);
1905 strbuf_init(&command_buf);
1907 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
1908 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
1909 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
1910 marks = pool_calloc(1, sizeof(struct mark_set));
1912 start_packfile();
1913 for (;;) {
1914 read_next_command();
1915 if (command_buf.eof)
1916 break;
1917 else if (!strcmp("blob", command_buf.buf))
1918 cmd_new_blob();
1919 else if (!strncmp("commit ", command_buf.buf, 7))
1920 cmd_new_commit();
1921 else if (!strncmp("tag ", command_buf.buf, 4))
1922 cmd_new_tag();
1923 else if (!strncmp("reset ", command_buf.buf, 6))
1924 cmd_reset_branch();
1925 else if (!strcmp("checkpoint", command_buf.buf))
1926 cmd_checkpoint();
1927 else
1928 die("Unsupported command: %s", command_buf.buf);
1929 }
1930 end_packfile();
1932 dump_branches();
1933 dump_tags();
1934 unkeep_all_packs();
1935 dump_marks();
1936 if (branch_log)
1937 fclose(branch_log);
1939 total_count = 0;
1940 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
1941 total_count += object_count_by_type[i];
1942 duplicate_count = 0;
1943 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
1944 duplicate_count += duplicate_count_by_type[i];
1946 fprintf(stderr, "%s statistics:\n", argv[0]);
1947 fprintf(stderr, "---------------------------------------------------------------------\n");
1948 fprintf(stderr, "Alloc'd objects: %10ju (%10ju overflow )\n", alloc_count, alloc_count - est_obj_cnt);
1949 fprintf(stderr, "Total objects: %10ju (%10ju duplicates )\n", total_count, duplicate_count);
1950 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]);
1951 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]);
1952 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]);
1953 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]);
1954 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
1955 fprintf(stderr, " marks: %10ju (%10ju unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
1956 fprintf(stderr, " atoms: %10u\n", atom_cnt);
1957 fprintf(stderr, "Memory total: %10ju KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
1958 fprintf(stderr, " pools: %10lu KiB\n", total_allocd/1024);
1959 fprintf(stderr, " objects: %10ju KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
1960 fprintf(stderr, "---------------------------------------------------------------------\n");
1961 pack_report();
1962 fprintf(stderr, "---------------------------------------------------------------------\n");
1963 fprintf(stderr, "\n");
1965 return 0;
1966 }