Merge branch 'maint'
[git/git.git] / fast-import.c
1 /*
2 Format of STDIN stream:
3
4 stream ::= cmd*;
5
6 cmd ::= new_blob
7 | new_commit
8 | new_tag
9 | reset_branch
10 | checkpoint
11 ;
12
13 new_blob ::= 'blob' lf
14 mark?
15 file_content;
16 file_content ::= data;
17
18 new_commit ::= 'commit' sp ref_str lf
19 mark?
20 ('author' sp name '<' email '>' when lf)?
21 'committer' sp name '<' email '>' when lf
22 commit_msg
23 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
24 ('merge' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)*
25 file_change*
26 lf;
27 commit_msg ::= data;
28
29 file_change ::= file_clr | file_del | file_obm | file_inm;
30 file_clr ::= 'deleteall' lf;
31 file_del ::= 'D' sp path_str lf;
32 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
33 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
34 data;
35
36 new_tag ::= 'tag' sp tag_str lf
37 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
38 'tagger' sp name '<' email '>' when lf
39 tag_msg;
40 tag_msg ::= data;
41
42 reset_branch ::= 'reset' sp ref_str lf
43 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
44 lf;
45
46 checkpoint ::= 'checkpoint' lf
47 lf;
48
49 # note: the first idnum in a stream should be 1 and subsequent
50 # idnums should not have gaps between values as this will cause
51 # the stream parser to reserve space for the gapped values. An
52 # idnum can be updated in the future to a new object by issuing
53 # a new mark directive with the old idnum.
54 #
55 mark ::= 'mark' sp idnum lf;
56 data ::= (delimited_data | exact_data)
57 lf;
58
59 # note: delim may be any string but must not contain lf.
60 # data_line may contain any data but must not be exactly
61 # delim.
62 delimited_data ::= 'data' sp '<<' delim lf
63 (data_line lf)*
64 delim lf;
65
66 # note: declen indicates the length of binary_data in bytes.
67 # declen does not include the lf preceeding the binary data.
68 #
69 exact_data ::= 'data' sp declen lf
70 binary_data;
71
72 # note: quoted strings are C-style quoting supporting \c for
73 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
74 # is the signed byte value in octal. Note that the only
75 # characters which must actually be escaped to protect the
76 # stream formatting is: \, " and LF. Otherwise these values
77 # are UTF8.
78 #
79 ref_str ::= ref;
80 sha1exp_str ::= sha1exp;
81 tag_str ::= tag;
82 path_str ::= path | '"' quoted(path) '"' ;
83 mode ::= '100644' | '644'
84 | '100755' | '755'
85 | '120000'
86 ;
87
88 declen ::= # unsigned 32 bit value, ascii base10 notation;
89 bigint ::= # unsigned integer value, ascii base10 notation;
90 binary_data ::= # file content, not interpreted;
91
92 when ::= raw_when | rfc2822_when;
93 raw_when ::= ts sp tz;
94 rfc2822_when ::= # Valid RFC 2822 date and time;
95
96 sp ::= # ASCII space character;
97 lf ::= # ASCII newline (LF) character;
98
99 # note: a colon (':') must precede the numerical value assigned to
100 # an idnum. This is to distinguish it from a ref or tag name as
101 # GIT does not permit ':' in ref or tag strings.
102 #
103 idnum ::= ':' bigint;
104 path ::= # GIT style file path, e.g. "a/b/c";
105 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
106 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
107 sha1exp ::= # Any valid GIT SHA1 expression;
108 hexsha1 ::= # SHA1 in hexadecimal format;
109
110 # note: name and email are UTF8 strings, however name must not
111 # contain '<' or lf and email must not contain any of the
112 # following: '<', '>', lf.
113 #
114 name ::= # valid GIT author/committer name;
115 email ::= # valid GIT author/committer email;
116 ts ::= # time since the epoch in seconds, ascii base10 notation;
117 tz ::= # GIT style timezone;
118 */
119
120 #include "builtin.h"
121 #include "cache.h"
122 #include "object.h"
123 #include "blob.h"
124 #include "tree.h"
125 #include "commit.h"
126 #include "delta.h"
127 #include "pack.h"
128 #include "refs.h"
129 #include "csum-file.h"
130 #include "strbuf.h"
131 #include "quote.h"
132
133 #define PACK_ID_BITS 16
134 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
135
136 #if !defined(NO_C99_FORMAT)
137 #define UM_FMT "%ju"
138 #define UM10_FMT "%10ju"
139 #else
140 /* Assumes unsigned long long exists. */
141 #define UM_FMT "%llu"
142 #define UM10_FMT "%10llu"
143 #endif
144
145 struct object_entry
146 {
147 struct object_entry *next;
148 uint32_t offset;
149 unsigned type : TYPE_BITS;
150 unsigned pack_id : PACK_ID_BITS;
151 unsigned char sha1[20];
152 };
153
154 struct object_entry_pool
155 {
156 struct object_entry_pool *next_pool;
157 struct object_entry *next_free;
158 struct object_entry *end;
159 struct object_entry entries[FLEX_ARRAY]; /* more */
160 };
161
162 struct mark_set
163 {
164 union {
165 struct object_entry *marked[1024];
166 struct mark_set *sets[1024];
167 } data;
168 unsigned int shift;
169 };
170
171 struct last_object
172 {
173 void *data;
174 unsigned long len;
175 uint32_t offset;
176 unsigned int depth;
177 unsigned no_free:1;
178 };
179
180 struct mem_pool
181 {
182 struct mem_pool *next_pool;
183 char *next_free;
184 char *end;
185 char space[FLEX_ARRAY]; /* more */
186 };
187
188 struct atom_str
189 {
190 struct atom_str *next_atom;
191 unsigned short str_len;
192 char str_dat[FLEX_ARRAY]; /* more */
193 };
194
195 struct tree_content;
196 struct tree_entry
197 {
198 struct tree_content *tree;
199 struct atom_str* name;
200 struct tree_entry_ms
201 {
202 uint16_t mode;
203 unsigned char sha1[20];
204 } versions[2];
205 };
206
207 struct tree_content
208 {
209 unsigned int entry_capacity; /* must match avail_tree_content */
210 unsigned int entry_count;
211 unsigned int delta_depth;
212 struct tree_entry *entries[FLEX_ARRAY]; /* more */
213 };
214
215 struct avail_tree_content
216 {
217 unsigned int entry_capacity; /* must match tree_content */
218 struct avail_tree_content *next_avail;
219 };
220
221 struct branch
222 {
223 struct branch *table_next_branch;
224 struct branch *active_next_branch;
225 const char *name;
226 struct tree_entry branch_tree;
227 uintmax_t last_commit;
228 unsigned int pack_id;
229 unsigned char sha1[20];
230 };
231
232 struct tag
233 {
234 struct tag *next_tag;
235 const char *name;
236 unsigned int pack_id;
237 unsigned char sha1[20];
238 };
239
240 struct dbuf
241 {
242 void *buffer;
243 size_t capacity;
244 };
245
246 struct hash_list
247 {
248 struct hash_list *next;
249 unsigned char sha1[20];
250 };
251
252 typedef enum {
253 WHENSPEC_RAW = 1,
254 WHENSPEC_RFC2822,
255 WHENSPEC_NOW,
256 } whenspec_type;
257
258 /* Configured limits on output */
259 static unsigned long max_depth = 10;
260 static unsigned long max_packsize = (1LL << 32) - 1;
261 static int force_update;
262
263 /* Stats and misc. counters */
264 static uintmax_t alloc_count;
265 static uintmax_t marks_set_count;
266 static uintmax_t object_count_by_type[1 << TYPE_BITS];
267 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
268 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
269 static unsigned long object_count;
270 static unsigned long branch_count;
271 static unsigned long branch_load_count;
272 static int failure;
273 static FILE *pack_edges;
274
275 /* Memory pools */
276 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
277 static size_t total_allocd;
278 static struct mem_pool *mem_pool;
279
280 /* Atom management */
281 static unsigned int atom_table_sz = 4451;
282 static unsigned int atom_cnt;
283 static struct atom_str **atom_table;
284
285 /* The .pack file being generated */
286 static unsigned int pack_id;
287 static struct packed_git *pack_data;
288 static struct packed_git **all_packs;
289 static unsigned long pack_size;
290
291 /* Table of objects we've written. */
292 static unsigned int object_entry_alloc = 5000;
293 static struct object_entry_pool *blocks;
294 static struct object_entry *object_table[1 << 16];
295 static struct mark_set *marks;
296 static const char* mark_file;
297
298 /* Our last blob */
299 static struct last_object last_blob;
300
301 /* Tree management */
302 static unsigned int tree_entry_alloc = 1000;
303 static void *avail_tree_entry;
304 static unsigned int avail_tree_table_sz = 100;
305 static struct avail_tree_content **avail_tree_table;
306 static struct dbuf old_tree;
307 static struct dbuf new_tree;
308
309 /* Branch data */
310 static unsigned long max_active_branches = 5;
311 static unsigned long cur_active_branches;
312 static unsigned long branch_table_sz = 1039;
313 static struct branch **branch_table;
314 static struct branch *active_branches;
315
316 /* Tag data */
317 static struct tag *first_tag;
318 static struct tag *last_tag;
319
320 /* Input stream parsing */
321 static whenspec_type whenspec = WHENSPEC_RAW;
322 static struct strbuf command_buf;
323 static uintmax_t next_mark;
324 static struct dbuf new_data;
325
326
327 static void alloc_objects(unsigned int cnt)
328 {
329 struct object_entry_pool *b;
330
331 b = xmalloc(sizeof(struct object_entry_pool)
332 + cnt * sizeof(struct object_entry));
333 b->next_pool = blocks;
334 b->next_free = b->entries;
335 b->end = b->entries + cnt;
336 blocks = b;
337 alloc_count += cnt;
338 }
339
340 static struct object_entry *new_object(unsigned char *sha1)
341 {
342 struct object_entry *e;
343
344 if (blocks->next_free == blocks->end)
345 alloc_objects(object_entry_alloc);
346
347 e = blocks->next_free++;
348 hashcpy(e->sha1, sha1);
349 return e;
350 }
351
352 static struct object_entry *find_object(unsigned char *sha1)
353 {
354 unsigned int h = sha1[0] << 8 | sha1[1];
355 struct object_entry *e;
356 for (e = object_table[h]; e; e = e->next)
357 if (!hashcmp(sha1, e->sha1))
358 return e;
359 return NULL;
360 }
361
362 static struct object_entry *insert_object(unsigned char *sha1)
363 {
364 unsigned int h = sha1[0] << 8 | sha1[1];
365 struct object_entry *e = object_table[h];
366 struct object_entry *p = NULL;
367
368 while (e) {
369 if (!hashcmp(sha1, e->sha1))
370 return e;
371 p = e;
372 e = e->next;
373 }
374
375 e = new_object(sha1);
376 e->next = NULL;
377 e->offset = 0;
378 if (p)
379 p->next = e;
380 else
381 object_table[h] = e;
382 return e;
383 }
384
385 static unsigned int hc_str(const char *s, size_t len)
386 {
387 unsigned int r = 0;
388 while (len-- > 0)
389 r = r * 31 + *s++;
390 return r;
391 }
392
393 static void *pool_alloc(size_t len)
394 {
395 struct mem_pool *p;
396 void *r;
397
398 for (p = mem_pool; p; p = p->next_pool)
399 if ((p->end - p->next_free >= len))
400 break;
401
402 if (!p) {
403 if (len >= (mem_pool_alloc/2)) {
404 total_allocd += len;
405 return xmalloc(len);
406 }
407 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
408 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
409 p->next_pool = mem_pool;
410 p->next_free = p->space;
411 p->end = p->next_free + mem_pool_alloc;
412 mem_pool = p;
413 }
414
415 r = p->next_free;
416 /* round out to a pointer alignment */
417 if (len & (sizeof(void*) - 1))
418 len += sizeof(void*) - (len & (sizeof(void*) - 1));
419 p->next_free += len;
420 return r;
421 }
422
423 static void *pool_calloc(size_t count, size_t size)
424 {
425 size_t len = count * size;
426 void *r = pool_alloc(len);
427 memset(r, 0, len);
428 return r;
429 }
430
431 static char *pool_strdup(const char *s)
432 {
433 char *r = pool_alloc(strlen(s) + 1);
434 strcpy(r, s);
435 return r;
436 }
437
438 static void size_dbuf(struct dbuf *b, size_t maxlen)
439 {
440 if (b->buffer) {
441 if (b->capacity >= maxlen)
442 return;
443 free(b->buffer);
444 }
445 b->capacity = ((maxlen / 1024) + 1) * 1024;
446 b->buffer = xmalloc(b->capacity);
447 }
448
449 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
450 {
451 struct mark_set *s = marks;
452 while ((idnum >> s->shift) >= 1024) {
453 s = pool_calloc(1, sizeof(struct mark_set));
454 s->shift = marks->shift + 10;
455 s->data.sets[0] = marks;
456 marks = s;
457 }
458 while (s->shift) {
459 uintmax_t i = idnum >> s->shift;
460 idnum -= i << s->shift;
461 if (!s->data.sets[i]) {
462 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
463 s->data.sets[i]->shift = s->shift - 10;
464 }
465 s = s->data.sets[i];
466 }
467 if (!s->data.marked[idnum])
468 marks_set_count++;
469 s->data.marked[idnum] = oe;
470 }
471
472 static struct object_entry *find_mark(uintmax_t idnum)
473 {
474 uintmax_t orig_idnum = idnum;
475 struct mark_set *s = marks;
476 struct object_entry *oe = NULL;
477 if ((idnum >> s->shift) < 1024) {
478 while (s && s->shift) {
479 uintmax_t i = idnum >> s->shift;
480 idnum -= i << s->shift;
481 s = s->data.sets[i];
482 }
483 if (s)
484 oe = s->data.marked[idnum];
485 }
486 if (!oe)
487 die("mark :" UM_FMT " not declared", orig_idnum);
488 return oe;
489 }
490
491 static struct atom_str *to_atom(const char *s, unsigned short len)
492 {
493 unsigned int hc = hc_str(s, len) % atom_table_sz;
494 struct atom_str *c;
495
496 for (c = atom_table[hc]; c; c = c->next_atom)
497 if (c->str_len == len && !strncmp(s, c->str_dat, len))
498 return c;
499
500 c = pool_alloc(sizeof(struct atom_str) + len + 1);
501 c->str_len = len;
502 strncpy(c->str_dat, s, len);
503 c->str_dat[len] = 0;
504 c->next_atom = atom_table[hc];
505 atom_table[hc] = c;
506 atom_cnt++;
507 return c;
508 }
509
510 static struct branch *lookup_branch(const char *name)
511 {
512 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
513 struct branch *b;
514
515 for (b = branch_table[hc]; b; b = b->table_next_branch)
516 if (!strcmp(name, b->name))
517 return b;
518 return NULL;
519 }
520
521 static struct branch *new_branch(const char *name)
522 {
523 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
524 struct branch* b = lookup_branch(name);
525
526 if (b)
527 die("Invalid attempt to create duplicate branch: %s", name);
528 if (check_ref_format(name))
529 die("Branch name doesn't conform to GIT standards: %s", name);
530
531 b = pool_calloc(1, sizeof(struct branch));
532 b->name = pool_strdup(name);
533 b->table_next_branch = branch_table[hc];
534 b->branch_tree.versions[0].mode = S_IFDIR;
535 b->branch_tree.versions[1].mode = S_IFDIR;
536 b->pack_id = MAX_PACK_ID;
537 branch_table[hc] = b;
538 branch_count++;
539 return b;
540 }
541
542 static unsigned int hc_entries(unsigned int cnt)
543 {
544 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
545 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
546 }
547
548 static struct tree_content *new_tree_content(unsigned int cnt)
549 {
550 struct avail_tree_content *f, *l = NULL;
551 struct tree_content *t;
552 unsigned int hc = hc_entries(cnt);
553
554 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
555 if (f->entry_capacity >= cnt)
556 break;
557
558 if (f) {
559 if (l)
560 l->next_avail = f->next_avail;
561 else
562 avail_tree_table[hc] = f->next_avail;
563 } else {
564 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
565 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
566 f->entry_capacity = cnt;
567 }
568
569 t = (struct tree_content*)f;
570 t->entry_count = 0;
571 t->delta_depth = 0;
572 return t;
573 }
574
575 static void release_tree_entry(struct tree_entry *e);
576 static void release_tree_content(struct tree_content *t)
577 {
578 struct avail_tree_content *f = (struct avail_tree_content*)t;
579 unsigned int hc = hc_entries(f->entry_capacity);
580 f->next_avail = avail_tree_table[hc];
581 avail_tree_table[hc] = f;
582 }
583
584 static void release_tree_content_recursive(struct tree_content *t)
585 {
586 unsigned int i;
587 for (i = 0; i < t->entry_count; i++)
588 release_tree_entry(t->entries[i]);
589 release_tree_content(t);
590 }
591
592 static struct tree_content *grow_tree_content(
593 struct tree_content *t,
594 int amt)
595 {
596 struct tree_content *r = new_tree_content(t->entry_count + amt);
597 r->entry_count = t->entry_count;
598 r->delta_depth = t->delta_depth;
599 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
600 release_tree_content(t);
601 return r;
602 }
603
604 static struct tree_entry *new_tree_entry(void)
605 {
606 struct tree_entry *e;
607
608 if (!avail_tree_entry) {
609 unsigned int n = tree_entry_alloc;
610 total_allocd += n * sizeof(struct tree_entry);
611 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
612 while (n-- > 1) {
613 *((void**)e) = e + 1;
614 e++;
615 }
616 *((void**)e) = NULL;
617 }
618
619 e = avail_tree_entry;
620 avail_tree_entry = *((void**)e);
621 return e;
622 }
623
624 static void release_tree_entry(struct tree_entry *e)
625 {
626 if (e->tree)
627 release_tree_content_recursive(e->tree);
628 *((void**)e) = avail_tree_entry;
629 avail_tree_entry = e;
630 }
631
632 static void start_packfile(void)
633 {
634 static char tmpfile[PATH_MAX];
635 struct packed_git *p;
636 struct pack_header hdr;
637 int pack_fd;
638
639 snprintf(tmpfile, sizeof(tmpfile),
640 "%s/pack_XXXXXX", get_object_directory());
641 pack_fd = mkstemp(tmpfile);
642 if (pack_fd < 0)
643 die("Can't create %s: %s", tmpfile, strerror(errno));
644 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
645 strcpy(p->pack_name, tmpfile);
646 p->pack_fd = pack_fd;
647
648 hdr.hdr_signature = htonl(PACK_SIGNATURE);
649 hdr.hdr_version = htonl(2);
650 hdr.hdr_entries = 0;
651 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
652
653 pack_data = p;
654 pack_size = sizeof(hdr);
655 object_count = 0;
656
657 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
658 all_packs[pack_id] = p;
659 }
660
661 static void fixup_header_footer(void)
662 {
663 static const int buf_sz = 128 * 1024;
664 int pack_fd = pack_data->pack_fd;
665 SHA_CTX c;
666 struct pack_header hdr;
667 char *buf;
668
669 if (lseek(pack_fd, 0, SEEK_SET) != 0)
670 die("Failed seeking to start: %s", strerror(errno));
671 if (read_in_full(pack_fd, &hdr, sizeof(hdr)) != sizeof(hdr))
672 die("Unable to reread header of %s", pack_data->pack_name);
673 if (lseek(pack_fd, 0, SEEK_SET) != 0)
674 die("Failed seeking to start: %s", strerror(errno));
675 hdr.hdr_entries = htonl(object_count);
676 write_or_die(pack_fd, &hdr, sizeof(hdr));
677
678 SHA1_Init(&c);
679 SHA1_Update(&c, &hdr, sizeof(hdr));
680
681 buf = xmalloc(buf_sz);
682 for (;;) {
683 size_t n = xread(pack_fd, buf, buf_sz);
684 if (!n)
685 break;
686 if (n < 0)
687 die("Failed to checksum %s", pack_data->pack_name);
688 SHA1_Update(&c, buf, n);
689 }
690 free(buf);
691
692 SHA1_Final(pack_data->sha1, &c);
693 write_or_die(pack_fd, pack_data->sha1, sizeof(pack_data->sha1));
694 close(pack_fd);
695 }
696
697 static int oecmp (const void *a_, const void *b_)
698 {
699 struct object_entry *a = *((struct object_entry**)a_);
700 struct object_entry *b = *((struct object_entry**)b_);
701 return hashcmp(a->sha1, b->sha1);
702 }
703
704 static char *create_index(void)
705 {
706 static char tmpfile[PATH_MAX];
707 SHA_CTX ctx;
708 struct sha1file *f;
709 struct object_entry **idx, **c, **last, *e;
710 struct object_entry_pool *o;
711 uint32_t array[256];
712 int i, idx_fd;
713
714 /* Build the sorted table of object IDs. */
715 idx = xmalloc(object_count * sizeof(struct object_entry*));
716 c = idx;
717 for (o = blocks; o; o = o->next_pool)
718 for (e = o->next_free; e-- != o->entries;)
719 if (pack_id == e->pack_id)
720 *c++ = e;
721 last = idx + object_count;
722 if (c != last)
723 die("internal consistency error creating the index");
724 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
725
726 /* Generate the fan-out array. */
727 c = idx;
728 for (i = 0; i < 256; i++) {
729 struct object_entry **next = c;;
730 while (next < last) {
731 if ((*next)->sha1[0] != i)
732 break;
733 next++;
734 }
735 array[i] = htonl(next - idx);
736 c = next;
737 }
738
739 snprintf(tmpfile, sizeof(tmpfile),
740 "%s/index_XXXXXX", get_object_directory());
741 idx_fd = mkstemp(tmpfile);
742 if (idx_fd < 0)
743 die("Can't create %s: %s", tmpfile, strerror(errno));
744 f = sha1fd(idx_fd, tmpfile);
745 sha1write(f, array, 256 * sizeof(int));
746 SHA1_Init(&ctx);
747 for (c = idx; c != last; c++) {
748 uint32_t offset = htonl((*c)->offset);
749 sha1write(f, &offset, 4);
750 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
751 SHA1_Update(&ctx, (*c)->sha1, 20);
752 }
753 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
754 sha1close(f, NULL, 1);
755 free(idx);
756 SHA1_Final(pack_data->sha1, &ctx);
757 return tmpfile;
758 }
759
760 static char *keep_pack(char *curr_index_name)
761 {
762 static char name[PATH_MAX];
763 static char *keep_msg = "fast-import";
764 int keep_fd;
765
766 chmod(pack_data->pack_name, 0444);
767 chmod(curr_index_name, 0444);
768
769 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
770 get_object_directory(), sha1_to_hex(pack_data->sha1));
771 keep_fd = open(name, O_RDWR|O_CREAT|O_EXCL, 0600);
772 if (keep_fd < 0)
773 die("cannot create keep file");
774 write(keep_fd, keep_msg, strlen(keep_msg));
775 close(keep_fd);
776
777 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
778 get_object_directory(), sha1_to_hex(pack_data->sha1));
779 if (move_temp_to_file(pack_data->pack_name, name))
780 die("cannot store pack file");
781
782 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
783 get_object_directory(), sha1_to_hex(pack_data->sha1));
784 if (move_temp_to_file(curr_index_name, name))
785 die("cannot store index file");
786 return name;
787 }
788
789 static void unkeep_all_packs(void)
790 {
791 static char name[PATH_MAX];
792 int k;
793
794 for (k = 0; k < pack_id; k++) {
795 struct packed_git *p = all_packs[k];
796 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
797 get_object_directory(), sha1_to_hex(p->sha1));
798 unlink(name);
799 }
800 }
801
802 static void end_packfile(void)
803 {
804 struct packed_git *old_p = pack_data, *new_p;
805
806 if (object_count) {
807 char *idx_name;
808 int i;
809 struct branch *b;
810 struct tag *t;
811
812 fixup_header_footer();
813 idx_name = keep_pack(create_index());
814
815 /* Register the packfile with core git's machinary. */
816 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
817 if (!new_p)
818 die("core git rejected index %s", idx_name);
819 new_p->windows = old_p->windows;
820 all_packs[pack_id] = new_p;
821 install_packed_git(new_p);
822
823 /* Print the boundary */
824 if (pack_edges) {
825 fprintf(pack_edges, "%s:", new_p->pack_name);
826 for (i = 0; i < branch_table_sz; i++) {
827 for (b = branch_table[i]; b; b = b->table_next_branch) {
828 if (b->pack_id == pack_id)
829 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
830 }
831 }
832 for (t = first_tag; t; t = t->next_tag) {
833 if (t->pack_id == pack_id)
834 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
835 }
836 fputc('\n', pack_edges);
837 fflush(pack_edges);
838 }
839
840 pack_id++;
841 }
842 else
843 unlink(old_p->pack_name);
844 free(old_p);
845
846 /* We can't carry a delta across packfiles. */
847 free(last_blob.data);
848 last_blob.data = NULL;
849 last_blob.len = 0;
850 last_blob.offset = 0;
851 last_blob.depth = 0;
852 }
853
854 static void cycle_packfile(void)
855 {
856 end_packfile();
857 start_packfile();
858 }
859
860 static size_t encode_header(
861 enum object_type type,
862 size_t size,
863 unsigned char *hdr)
864 {
865 int n = 1;
866 unsigned char c;
867
868 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
869 die("bad type %d", type);
870
871 c = (type << 4) | (size & 15);
872 size >>= 4;
873 while (size) {
874 *hdr++ = c | 0x80;
875 c = size & 0x7f;
876 size >>= 7;
877 n++;
878 }
879 *hdr = c;
880 return n;
881 }
882
883 static int store_object(
884 enum object_type type,
885 void *dat,
886 size_t datlen,
887 struct last_object *last,
888 unsigned char *sha1out,
889 uintmax_t mark)
890 {
891 void *out, *delta;
892 struct object_entry *e;
893 unsigned char hdr[96];
894 unsigned char sha1[20];
895 unsigned long hdrlen, deltalen;
896 SHA_CTX c;
897 z_stream s;
898
899 hdrlen = sprintf((char*)hdr,"%s %lu", type_names[type],
900 (unsigned long)datlen) + 1;
901 SHA1_Init(&c);
902 SHA1_Update(&c, hdr, hdrlen);
903 SHA1_Update(&c, dat, datlen);
904 SHA1_Final(sha1, &c);
905 if (sha1out)
906 hashcpy(sha1out, sha1);
907
908 e = insert_object(sha1);
909 if (mark)
910 insert_mark(mark, e);
911 if (e->offset) {
912 duplicate_count_by_type[type]++;
913 return 1;
914 }
915
916 if (last && last->data && last->depth < max_depth) {
917 delta = diff_delta(last->data, last->len,
918 dat, datlen,
919 &deltalen, 0);
920 if (delta && deltalen >= datlen) {
921 free(delta);
922 delta = NULL;
923 }
924 } else
925 delta = NULL;
926
927 memset(&s, 0, sizeof(s));
928 deflateInit(&s, zlib_compression_level);
929 if (delta) {
930 s.next_in = delta;
931 s.avail_in = deltalen;
932 } else {
933 s.next_in = dat;
934 s.avail_in = datlen;
935 }
936 s.avail_out = deflateBound(&s, s.avail_in);
937 s.next_out = out = xmalloc(s.avail_out);
938 while (deflate(&s, Z_FINISH) == Z_OK)
939 /* nothing */;
940 deflateEnd(&s);
941
942 /* Determine if we should auto-checkpoint. */
943 if ((pack_size + 60 + s.total_out) > max_packsize
944 || (pack_size + 60 + s.total_out) < pack_size) {
945
946 /* This new object needs to *not* have the current pack_id. */
947 e->pack_id = pack_id + 1;
948 cycle_packfile();
949
950 /* We cannot carry a delta into the new pack. */
951 if (delta) {
952 free(delta);
953 delta = NULL;
954
955 memset(&s, 0, sizeof(s));
956 deflateInit(&s, zlib_compression_level);
957 s.next_in = dat;
958 s.avail_in = datlen;
959 s.avail_out = deflateBound(&s, s.avail_in);
960 s.next_out = out = xrealloc(out, s.avail_out);
961 while (deflate(&s, Z_FINISH) == Z_OK)
962 /* nothing */;
963 deflateEnd(&s);
964 }
965 }
966
967 e->type = type;
968 e->pack_id = pack_id;
969 e->offset = pack_size;
970 object_count++;
971 object_count_by_type[type]++;
972
973 if (delta) {
974 unsigned long ofs = e->offset - last->offset;
975 unsigned pos = sizeof(hdr) - 1;
976
977 delta_count_by_type[type]++;
978 last->depth++;
979
980 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
981 write_or_die(pack_data->pack_fd, hdr, hdrlen);
982 pack_size += hdrlen;
983
984 hdr[pos] = ofs & 127;
985 while (ofs >>= 7)
986 hdr[--pos] = 128 | (--ofs & 127);
987 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
988 pack_size += sizeof(hdr) - pos;
989 } else {
990 if (last)
991 last->depth = 0;
992 hdrlen = encode_header(type, datlen, hdr);
993 write_or_die(pack_data->pack_fd, hdr, hdrlen);
994 pack_size += hdrlen;
995 }
996
997 write_or_die(pack_data->pack_fd, out, s.total_out);
998 pack_size += s.total_out;
999
1000 free(out);
1001 free(delta);
1002 if (last) {
1003 if (!last->no_free)
1004 free(last->data);
1005 last->data = dat;
1006 last->offset = e->offset;
1007 last->len = datlen;
1008 }
1009 return 0;
1010 }
1011
1012 static void *gfi_unpack_entry(
1013 struct object_entry *oe,
1014 unsigned long *sizep)
1015 {
1016 static char type[20];
1017 struct packed_git *p = all_packs[oe->pack_id];
1018 if (p == pack_data)
1019 p->pack_size = pack_size + 20;
1020 return unpack_entry(p, oe->offset, type, sizep);
1021 }
1022
1023 static const char *get_mode(const char *str, uint16_t *modep)
1024 {
1025 unsigned char c;
1026 uint16_t mode = 0;
1027
1028 while ((c = *str++) != ' ') {
1029 if (c < '0' || c > '7')
1030 return NULL;
1031 mode = (mode << 3) + (c - '0');
1032 }
1033 *modep = mode;
1034 return str;
1035 }
1036
1037 static void load_tree(struct tree_entry *root)
1038 {
1039 unsigned char* sha1 = root->versions[1].sha1;
1040 struct object_entry *myoe;
1041 struct tree_content *t;
1042 unsigned long size;
1043 char *buf;
1044 const char *c;
1045
1046 root->tree = t = new_tree_content(8);
1047 if (is_null_sha1(sha1))
1048 return;
1049
1050 myoe = find_object(sha1);
1051 if (myoe) {
1052 if (myoe->type != OBJ_TREE)
1053 die("Not a tree: %s", sha1_to_hex(sha1));
1054 t->delta_depth = 0;
1055 buf = gfi_unpack_entry(myoe, &size);
1056 } else {
1057 char type[20];
1058 buf = read_sha1_file(sha1, type, &size);
1059 if (!buf || strcmp(type, tree_type))
1060 die("Can't load tree %s", sha1_to_hex(sha1));
1061 }
1062
1063 c = buf;
1064 while (c != (buf + size)) {
1065 struct tree_entry *e = new_tree_entry();
1066
1067 if (t->entry_count == t->entry_capacity)
1068 root->tree = t = grow_tree_content(t, 8);
1069 t->entries[t->entry_count++] = e;
1070
1071 e->tree = NULL;
1072 c = get_mode(c, &e->versions[1].mode);
1073 if (!c)
1074 die("Corrupt mode in %s", sha1_to_hex(sha1));
1075 e->versions[0].mode = e->versions[1].mode;
1076 e->name = to_atom(c, (unsigned short)strlen(c));
1077 c += e->name->str_len + 1;
1078 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1079 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1080 c += 20;
1081 }
1082 free(buf);
1083 }
1084
1085 static int tecmp0 (const void *_a, const void *_b)
1086 {
1087 struct tree_entry *a = *((struct tree_entry**)_a);
1088 struct tree_entry *b = *((struct tree_entry**)_b);
1089 return base_name_compare(
1090 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1091 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1092 }
1093
1094 static int tecmp1 (const void *_a, const void *_b)
1095 {
1096 struct tree_entry *a = *((struct tree_entry**)_a);
1097 struct tree_entry *b = *((struct tree_entry**)_b);
1098 return base_name_compare(
1099 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1100 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1101 }
1102
1103 static void mktree(struct tree_content *t,
1104 int v,
1105 unsigned long *szp,
1106 struct dbuf *b)
1107 {
1108 size_t maxlen = 0;
1109 unsigned int i;
1110 char *c;
1111
1112 if (!v)
1113 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1114 else
1115 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1116
1117 for (i = 0; i < t->entry_count; i++) {
1118 if (t->entries[i]->versions[v].mode)
1119 maxlen += t->entries[i]->name->str_len + 34;
1120 }
1121
1122 size_dbuf(b, maxlen);
1123 c = b->buffer;
1124 for (i = 0; i < t->entry_count; i++) {
1125 struct tree_entry *e = t->entries[i];
1126 if (!e->versions[v].mode)
1127 continue;
1128 c += sprintf(c, "%o", (unsigned int)e->versions[v].mode);
1129 *c++ = ' ';
1130 strcpy(c, e->name->str_dat);
1131 c += e->name->str_len + 1;
1132 hashcpy((unsigned char*)c, e->versions[v].sha1);
1133 c += 20;
1134 }
1135 *szp = c - (char*)b->buffer;
1136 }
1137
1138 static void store_tree(struct tree_entry *root)
1139 {
1140 struct tree_content *t = root->tree;
1141 unsigned int i, j, del;
1142 unsigned long new_len;
1143 struct last_object lo;
1144 struct object_entry *le;
1145
1146 if (!is_null_sha1(root->versions[1].sha1))
1147 return;
1148
1149 for (i = 0; i < t->entry_count; i++) {
1150 if (t->entries[i]->tree)
1151 store_tree(t->entries[i]);
1152 }
1153
1154 le = find_object(root->versions[0].sha1);
1155 if (!S_ISDIR(root->versions[0].mode)
1156 || !le
1157 || le->pack_id != pack_id) {
1158 lo.data = NULL;
1159 lo.depth = 0;
1160 } else {
1161 mktree(t, 0, &lo.len, &old_tree);
1162 lo.data = old_tree.buffer;
1163 lo.offset = le->offset;
1164 lo.depth = t->delta_depth;
1165 lo.no_free = 1;
1166 }
1167
1168 mktree(t, 1, &new_len, &new_tree);
1169 store_object(OBJ_TREE, new_tree.buffer, new_len,
1170 &lo, root->versions[1].sha1, 0);
1171
1172 t->delta_depth = lo.depth;
1173 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1174 struct tree_entry *e = t->entries[i];
1175 if (e->versions[1].mode) {
1176 e->versions[0].mode = e->versions[1].mode;
1177 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1178 t->entries[j++] = e;
1179 } else {
1180 release_tree_entry(e);
1181 del++;
1182 }
1183 }
1184 t->entry_count -= del;
1185 }
1186
1187 static int tree_content_set(
1188 struct tree_entry *root,
1189 const char *p,
1190 const unsigned char *sha1,
1191 const uint16_t mode)
1192 {
1193 struct tree_content *t = root->tree;
1194 const char *slash1;
1195 unsigned int i, n;
1196 struct tree_entry *e;
1197
1198 slash1 = strchr(p, '/');
1199 if (slash1)
1200 n = slash1 - p;
1201 else
1202 n = strlen(p);
1203
1204 for (i = 0; i < t->entry_count; i++) {
1205 e = t->entries[i];
1206 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1207 if (!slash1) {
1208 if (e->versions[1].mode == mode
1209 && !hashcmp(e->versions[1].sha1, sha1))
1210 return 0;
1211 e->versions[1].mode = mode;
1212 hashcpy(e->versions[1].sha1, sha1);
1213 if (e->tree) {
1214 release_tree_content_recursive(e->tree);
1215 e->tree = NULL;
1216 }
1217 hashclr(root->versions[1].sha1);
1218 return 1;
1219 }
1220 if (!S_ISDIR(e->versions[1].mode)) {
1221 e->tree = new_tree_content(8);
1222 e->versions[1].mode = S_IFDIR;
1223 }
1224 if (!e->tree)
1225 load_tree(e);
1226 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
1227 hashclr(root->versions[1].sha1);
1228 return 1;
1229 }
1230 return 0;
1231 }
1232 }
1233
1234 if (t->entry_count == t->entry_capacity)
1235 root->tree = t = grow_tree_content(t, 8);
1236 e = new_tree_entry();
1237 e->name = to_atom(p, (unsigned short)n);
1238 e->versions[0].mode = 0;
1239 hashclr(e->versions[0].sha1);
1240 t->entries[t->entry_count++] = e;
1241 if (slash1) {
1242 e->tree = new_tree_content(8);
1243 e->versions[1].mode = S_IFDIR;
1244 tree_content_set(e, slash1 + 1, sha1, mode);
1245 } else {
1246 e->tree = NULL;
1247 e->versions[1].mode = mode;
1248 hashcpy(e->versions[1].sha1, sha1);
1249 }
1250 hashclr(root->versions[1].sha1);
1251 return 1;
1252 }
1253
1254 static int tree_content_remove(struct tree_entry *root, const char *p)
1255 {
1256 struct tree_content *t = root->tree;
1257 const char *slash1;
1258 unsigned int i, n;
1259 struct tree_entry *e;
1260
1261 slash1 = strchr(p, '/');
1262 if (slash1)
1263 n = slash1 - p;
1264 else
1265 n = strlen(p);
1266
1267 for (i = 0; i < t->entry_count; i++) {
1268 e = t->entries[i];
1269 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1270 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1271 goto del_entry;
1272 if (!e->tree)
1273 load_tree(e);
1274 if (tree_content_remove(e, slash1 + 1)) {
1275 for (n = 0; n < e->tree->entry_count; n++) {
1276 if (e->tree->entries[n]->versions[1].mode) {
1277 hashclr(root->versions[1].sha1);
1278 return 1;
1279 }
1280 }
1281 goto del_entry;
1282 }
1283 return 0;
1284 }
1285 }
1286 return 0;
1287
1288 del_entry:
1289 if (e->tree) {
1290 release_tree_content_recursive(e->tree);
1291 e->tree = NULL;
1292 }
1293 e->versions[1].mode = 0;
1294 hashclr(e->versions[1].sha1);
1295 hashclr(root->versions[1].sha1);
1296 return 1;
1297 }
1298
1299 static int update_branch(struct branch *b)
1300 {
1301 static const char *msg = "fast-import";
1302 struct ref_lock *lock;
1303 unsigned char old_sha1[20];
1304
1305 if (read_ref(b->name, old_sha1))
1306 hashclr(old_sha1);
1307 lock = lock_any_ref_for_update(b->name, old_sha1);
1308 if (!lock)
1309 return error("Unable to lock %s", b->name);
1310 if (!force_update && !is_null_sha1(old_sha1)) {
1311 struct commit *old_cmit, *new_cmit;
1312
1313 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1314 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1315 if (!old_cmit || !new_cmit) {
1316 unlock_ref(lock);
1317 return error("Branch %s is missing commits.", b->name);
1318 }
1319
1320 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1321 unlock_ref(lock);
1322 warn("Not updating %s"
1323 " (new tip %s does not contain %s)",
1324 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1325 return -1;
1326 }
1327 }
1328 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1329 return error("Unable to update %s", b->name);
1330 return 0;
1331 }
1332
1333 static void dump_branches(void)
1334 {
1335 unsigned int i;
1336 struct branch *b;
1337
1338 for (i = 0; i < branch_table_sz; i++) {
1339 for (b = branch_table[i]; b; b = b->table_next_branch)
1340 failure |= update_branch(b);
1341 }
1342 }
1343
1344 static void dump_tags(void)
1345 {
1346 static const char *msg = "fast-import";
1347 struct tag *t;
1348 struct ref_lock *lock;
1349 char ref_name[PATH_MAX];
1350
1351 for (t = first_tag; t; t = t->next_tag) {
1352 sprintf(ref_name, "tags/%s", t->name);
1353 lock = lock_ref_sha1(ref_name, NULL);
1354 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1355 failure |= error("Unable to update %s", ref_name);
1356 }
1357 }
1358
1359 static void dump_marks_helper(FILE *f,
1360 uintmax_t base,
1361 struct mark_set *m)
1362 {
1363 uintmax_t k;
1364 if (m->shift) {
1365 for (k = 0; k < 1024; k++) {
1366 if (m->data.sets[k])
1367 dump_marks_helper(f, (base + k) << m->shift,
1368 m->data.sets[k]);
1369 }
1370 } else {
1371 for (k = 0; k < 1024; k++) {
1372 if (m->data.marked[k])
1373 fprintf(f, ":" UM_FMT " %s\n", base + k,
1374 sha1_to_hex(m->data.marked[k]->sha1));
1375 }
1376 }
1377 }
1378
1379 static void dump_marks(void)
1380 {
1381 if (mark_file)
1382 {
1383 FILE *f = fopen(mark_file, "w");
1384 if (f) {
1385 dump_marks_helper(f, 0, marks);
1386 fclose(f);
1387 } else
1388 failure |= error("Unable to write marks file %s: %s",
1389 mark_file, strerror(errno));
1390 }
1391 }
1392
1393 static void read_next_command(void)
1394 {
1395 read_line(&command_buf, stdin, '\n');
1396 }
1397
1398 static void cmd_mark(void)
1399 {
1400 if (!strncmp("mark :", command_buf.buf, 6)) {
1401 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1402 read_next_command();
1403 }
1404 else
1405 next_mark = 0;
1406 }
1407
1408 static void *cmd_data (size_t *size)
1409 {
1410 size_t length;
1411 char *buffer;
1412
1413 if (strncmp("data ", command_buf.buf, 5))
1414 die("Expected 'data n' command, found: %s", command_buf.buf);
1415
1416 if (!strncmp("<<", command_buf.buf + 5, 2)) {
1417 char *term = xstrdup(command_buf.buf + 5 + 2);
1418 size_t sz = 8192, term_len = command_buf.len - 5 - 2;
1419 length = 0;
1420 buffer = xmalloc(sz);
1421 for (;;) {
1422 read_next_command();
1423 if (command_buf.eof)
1424 die("EOF in data (terminator '%s' not found)", term);
1425 if (term_len == command_buf.len
1426 && !strcmp(term, command_buf.buf))
1427 break;
1428 if (sz < (length + command_buf.len)) {
1429 sz = sz * 3 / 2 + 16;
1430 if (sz < (length + command_buf.len))
1431 sz = length + command_buf.len;
1432 buffer = xrealloc(buffer, sz);
1433 }
1434 memcpy(buffer + length,
1435 command_buf.buf,
1436 command_buf.len - 1);
1437 length += command_buf.len - 1;
1438 buffer[length++] = '\n';
1439 }
1440 free(term);
1441 }
1442 else {
1443 size_t n = 0;
1444 length = strtoul(command_buf.buf + 5, NULL, 10);
1445 buffer = xmalloc(length);
1446 while (n < length) {
1447 size_t s = fread(buffer + n, 1, length - n, stdin);
1448 if (!s && feof(stdin))
1449 die("EOF in data (%lu bytes remaining)",
1450 (unsigned long)(length - n));
1451 n += s;
1452 }
1453 }
1454
1455 if (fgetc(stdin) != '\n')
1456 die("An lf did not trail the binary data as expected.");
1457
1458 *size = length;
1459 return buffer;
1460 }
1461
1462 static int validate_raw_date(const char *src, char *result, int maxlen)
1463 {
1464 const char *orig_src = src;
1465 char *endp, sign;
1466
1467 strtoul(src, &endp, 10);
1468 if (endp == src || *endp != ' ')
1469 return -1;
1470
1471 src = endp + 1;
1472 if (*src != '-' && *src != '+')
1473 return -1;
1474 sign = *src;
1475
1476 strtoul(src + 1, &endp, 10);
1477 if (endp == src || *endp || (endp - orig_src) >= maxlen)
1478 return -1;
1479
1480 strcpy(result, orig_src);
1481 return 0;
1482 }
1483
1484 static char *parse_ident(const char *buf)
1485 {
1486 const char *gt;
1487 size_t name_len;
1488 char *ident;
1489
1490 gt = strrchr(buf, '>');
1491 if (!gt)
1492 die("Missing > in ident string: %s", buf);
1493 gt++;
1494 if (*gt != ' ')
1495 die("Missing space after > in ident string: %s", buf);
1496 gt++;
1497 name_len = gt - buf;
1498 ident = xmalloc(name_len + 24);
1499 strncpy(ident, buf, name_len);
1500
1501 switch (whenspec) {
1502 case WHENSPEC_RAW:
1503 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1504 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1505 break;
1506 case WHENSPEC_RFC2822:
1507 if (parse_date(gt, ident + name_len, 24) < 0)
1508 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1509 break;
1510 case WHENSPEC_NOW:
1511 if (strcmp("now", gt))
1512 die("Date in ident must be 'now': %s", buf);
1513 datestamp(ident + name_len, 24);
1514 break;
1515 }
1516
1517 return ident;
1518 }
1519
1520 static void cmd_new_blob(void)
1521 {
1522 size_t l;
1523 void *d;
1524
1525 read_next_command();
1526 cmd_mark();
1527 d = cmd_data(&l);
1528
1529 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1530 free(d);
1531 }
1532
1533 static void unload_one_branch(void)
1534 {
1535 while (cur_active_branches
1536 && cur_active_branches >= max_active_branches) {
1537 unsigned long min_commit = ULONG_MAX;
1538 struct branch *e, *l = NULL, *p = NULL;
1539
1540 for (e = active_branches; e; e = e->active_next_branch) {
1541 if (e->last_commit < min_commit) {
1542 p = l;
1543 min_commit = e->last_commit;
1544 }
1545 l = e;
1546 }
1547
1548 if (p) {
1549 e = p->active_next_branch;
1550 p->active_next_branch = e->active_next_branch;
1551 } else {
1552 e = active_branches;
1553 active_branches = e->active_next_branch;
1554 }
1555 e->active_next_branch = NULL;
1556 if (e->branch_tree.tree) {
1557 release_tree_content_recursive(e->branch_tree.tree);
1558 e->branch_tree.tree = NULL;
1559 }
1560 cur_active_branches--;
1561 }
1562 }
1563
1564 static void load_branch(struct branch *b)
1565 {
1566 load_tree(&b->branch_tree);
1567 b->active_next_branch = active_branches;
1568 active_branches = b;
1569 cur_active_branches++;
1570 branch_load_count++;
1571 }
1572
1573 static void file_change_m(struct branch *b)
1574 {
1575 const char *p = command_buf.buf + 2;
1576 char *p_uq;
1577 const char *endp;
1578 struct object_entry *oe = oe;
1579 unsigned char sha1[20];
1580 uint16_t mode, inline_data = 0;
1581 char type[20];
1582
1583 p = get_mode(p, &mode);
1584 if (!p)
1585 die("Corrupt mode: %s", command_buf.buf);
1586 switch (mode) {
1587 case S_IFREG | 0644:
1588 case S_IFREG | 0755:
1589 case S_IFLNK:
1590 case 0644:
1591 case 0755:
1592 /* ok */
1593 break;
1594 default:
1595 die("Corrupt mode: %s", command_buf.buf);
1596 }
1597
1598 if (*p == ':') {
1599 char *x;
1600 oe = find_mark(strtoumax(p + 1, &x, 10));
1601 hashcpy(sha1, oe->sha1);
1602 p = x;
1603 } else if (!strncmp("inline", p, 6)) {
1604 inline_data = 1;
1605 p += 6;
1606 } else {
1607 if (get_sha1_hex(p, sha1))
1608 die("Invalid SHA1: %s", command_buf.buf);
1609 oe = find_object(sha1);
1610 p += 40;
1611 }
1612 if (*p++ != ' ')
1613 die("Missing space after SHA1: %s", command_buf.buf);
1614
1615 p_uq = unquote_c_style(p, &endp);
1616 if (p_uq) {
1617 if (*endp)
1618 die("Garbage after path in: %s", command_buf.buf);
1619 p = p_uq;
1620 }
1621
1622 if (inline_data) {
1623 size_t l;
1624 void *d;
1625 if (!p_uq)
1626 p = p_uq = xstrdup(p);
1627 read_next_command();
1628 d = cmd_data(&l);
1629 if (store_object(OBJ_BLOB, d, l, &last_blob, sha1, 0))
1630 free(d);
1631 } else if (oe) {
1632 if (oe->type != OBJ_BLOB)
1633 die("Not a blob (actually a %s): %s",
1634 command_buf.buf, type_names[oe->type]);
1635 } else {
1636 if (sha1_object_info(sha1, type, NULL))
1637 die("Blob not found: %s", command_buf.buf);
1638 if (strcmp(blob_type, type))
1639 die("Not a blob (actually a %s): %s",
1640 command_buf.buf, type);
1641 }
1642
1643 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1644 free(p_uq);
1645 }
1646
1647 static void file_change_d(struct branch *b)
1648 {
1649 const char *p = command_buf.buf + 2;
1650 char *p_uq;
1651 const char *endp;
1652
1653 p_uq = unquote_c_style(p, &endp);
1654 if (p_uq) {
1655 if (*endp)
1656 die("Garbage after path in: %s", command_buf.buf);
1657 p = p_uq;
1658 }
1659 tree_content_remove(&b->branch_tree, p);
1660 free(p_uq);
1661 }
1662
1663 static void file_change_deleteall(struct branch *b)
1664 {
1665 release_tree_content_recursive(b->branch_tree.tree);
1666 hashclr(b->branch_tree.versions[0].sha1);
1667 hashclr(b->branch_tree.versions[1].sha1);
1668 load_tree(&b->branch_tree);
1669 }
1670
1671 static void cmd_from(struct branch *b)
1672 {
1673 const char *from;
1674 struct branch *s;
1675
1676 if (strncmp("from ", command_buf.buf, 5))
1677 return;
1678
1679 if (b->branch_tree.tree) {
1680 release_tree_content_recursive(b->branch_tree.tree);
1681 b->branch_tree.tree = NULL;
1682 }
1683
1684 from = strchr(command_buf.buf, ' ') + 1;
1685 s = lookup_branch(from);
1686 if (b == s)
1687 die("Can't create a branch from itself: %s", b->name);
1688 else if (s) {
1689 unsigned char *t = s->branch_tree.versions[1].sha1;
1690 hashcpy(b->sha1, s->sha1);
1691 hashcpy(b->branch_tree.versions[0].sha1, t);
1692 hashcpy(b->branch_tree.versions[1].sha1, t);
1693 } else if (*from == ':') {
1694 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1695 struct object_entry *oe = find_mark(idnum);
1696 unsigned long size;
1697 char *buf;
1698 if (oe->type != OBJ_COMMIT)
1699 die("Mark :" UM_FMT " not a commit", idnum);
1700 hashcpy(b->sha1, oe->sha1);
1701 buf = gfi_unpack_entry(oe, &size);
1702 if (!buf || size < 46)
1703 die("Not a valid commit: %s", from);
1704 if (memcmp("tree ", buf, 5)
1705 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1706 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1707 free(buf);
1708 hashcpy(b->branch_tree.versions[0].sha1,
1709 b->branch_tree.versions[1].sha1);
1710 } else if (!get_sha1(from, b->sha1)) {
1711 if (is_null_sha1(b->sha1)) {
1712 hashclr(b->branch_tree.versions[0].sha1);
1713 hashclr(b->branch_tree.versions[1].sha1);
1714 } else {
1715 unsigned long size;
1716 char *buf;
1717
1718 buf = read_object_with_reference(b->sha1,
1719 type_names[OBJ_COMMIT], &size, b->sha1);
1720 if (!buf || size < 46)
1721 die("Not a valid commit: %s", from);
1722 if (memcmp("tree ", buf, 5)
1723 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1724 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1725 free(buf);
1726 hashcpy(b->branch_tree.versions[0].sha1,
1727 b->branch_tree.versions[1].sha1);
1728 }
1729 } else
1730 die("Invalid ref name or SHA1 expression: %s", from);
1731
1732 read_next_command();
1733 }
1734
1735 static struct hash_list *cmd_merge(unsigned int *count)
1736 {
1737 struct hash_list *list = NULL, *n, *e = e;
1738 const char *from;
1739 struct branch *s;
1740
1741 *count = 0;
1742 while (!strncmp("merge ", command_buf.buf, 6)) {
1743 from = strchr(command_buf.buf, ' ') + 1;
1744 n = xmalloc(sizeof(*n));
1745 s = lookup_branch(from);
1746 if (s)
1747 hashcpy(n->sha1, s->sha1);
1748 else if (*from == ':') {
1749 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1750 struct object_entry *oe = find_mark(idnum);
1751 if (oe->type != OBJ_COMMIT)
1752 die("Mark :" UM_FMT " not a commit", idnum);
1753 hashcpy(n->sha1, oe->sha1);
1754 } else if (get_sha1(from, n->sha1))
1755 die("Invalid ref name or SHA1 expression: %s", from);
1756
1757 n->next = NULL;
1758 if (list)
1759 e->next = n;
1760 else
1761 list = n;
1762 e = n;
1763 (*count)++;
1764 read_next_command();
1765 }
1766 return list;
1767 }
1768
1769 static void cmd_new_commit(void)
1770 {
1771 struct branch *b;
1772 void *msg;
1773 size_t msglen;
1774 char *sp;
1775 char *author = NULL;
1776 char *committer = NULL;
1777 struct hash_list *merge_list = NULL;
1778 unsigned int merge_count;
1779
1780 /* Obtain the branch name from the rest of our command */
1781 sp = strchr(command_buf.buf, ' ') + 1;
1782 b = lookup_branch(sp);
1783 if (!b)
1784 b = new_branch(sp);
1785
1786 read_next_command();
1787 cmd_mark();
1788 if (!strncmp("author ", command_buf.buf, 7)) {
1789 author = parse_ident(command_buf.buf + 7);
1790 read_next_command();
1791 }
1792 if (!strncmp("committer ", command_buf.buf, 10)) {
1793 committer = parse_ident(command_buf.buf + 10);
1794 read_next_command();
1795 }
1796 if (!committer)
1797 die("Expected committer but didn't get one");
1798 msg = cmd_data(&msglen);
1799 read_next_command();
1800 cmd_from(b);
1801 merge_list = cmd_merge(&merge_count);
1802
1803 /* ensure the branch is active/loaded */
1804 if (!b->branch_tree.tree || !max_active_branches) {
1805 unload_one_branch();
1806 load_branch(b);
1807 }
1808
1809 /* file_change* */
1810 for (;;) {
1811 if (1 == command_buf.len)
1812 break;
1813 else if (!strncmp("M ", command_buf.buf, 2))
1814 file_change_m(b);
1815 else if (!strncmp("D ", command_buf.buf, 2))
1816 file_change_d(b);
1817 else if (!strcmp("deleteall", command_buf.buf))
1818 file_change_deleteall(b);
1819 else
1820 die("Unsupported file_change: %s", command_buf.buf);
1821 read_next_command();
1822 }
1823
1824 /* build the tree and the commit */
1825 store_tree(&b->branch_tree);
1826 hashcpy(b->branch_tree.versions[0].sha1,
1827 b->branch_tree.versions[1].sha1);
1828 size_dbuf(&new_data, 114 + msglen
1829 + merge_count * 49
1830 + (author
1831 ? strlen(author) + strlen(committer)
1832 : 2 * strlen(committer)));
1833 sp = new_data.buffer;
1834 sp += sprintf(sp, "tree %s\n",
1835 sha1_to_hex(b->branch_tree.versions[1].sha1));
1836 if (!is_null_sha1(b->sha1))
1837 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1838 while (merge_list) {
1839 struct hash_list *next = merge_list->next;
1840 sp += sprintf(sp, "parent %s\n", sha1_to_hex(merge_list->sha1));
1841 free(merge_list);
1842 merge_list = next;
1843 }
1844 sp += sprintf(sp, "author %s\n", author ? author : committer);
1845 sp += sprintf(sp, "committer %s\n", committer);
1846 *sp++ = '\n';
1847 memcpy(sp, msg, msglen);
1848 sp += msglen;
1849 free(author);
1850 free(committer);
1851 free(msg);
1852
1853 if (!store_object(OBJ_COMMIT,
1854 new_data.buffer, sp - (char*)new_data.buffer,
1855 NULL, b->sha1, next_mark))
1856 b->pack_id = pack_id;
1857 b->last_commit = object_count_by_type[OBJ_COMMIT];
1858 }
1859
1860 static void cmd_new_tag(void)
1861 {
1862 char *sp;
1863 const char *from;
1864 char *tagger;
1865 struct branch *s;
1866 void *msg;
1867 size_t msglen;
1868 struct tag *t;
1869 uintmax_t from_mark = 0;
1870 unsigned char sha1[20];
1871
1872 /* Obtain the new tag name from the rest of our command */
1873 sp = strchr(command_buf.buf, ' ') + 1;
1874 t = pool_alloc(sizeof(struct tag));
1875 t->next_tag = NULL;
1876 t->name = pool_strdup(sp);
1877 if (last_tag)
1878 last_tag->next_tag = t;
1879 else
1880 first_tag = t;
1881 last_tag = t;
1882 read_next_command();
1883
1884 /* from ... */
1885 if (strncmp("from ", command_buf.buf, 5))
1886 die("Expected from command, got %s", command_buf.buf);
1887 from = strchr(command_buf.buf, ' ') + 1;
1888 s = lookup_branch(from);
1889 if (s) {
1890 hashcpy(sha1, s->sha1);
1891 } else if (*from == ':') {
1892 struct object_entry *oe;
1893 from_mark = strtoumax(from + 1, NULL, 10);
1894 oe = find_mark(from_mark);
1895 if (oe->type != OBJ_COMMIT)
1896 die("Mark :" UM_FMT " not a commit", from_mark);
1897 hashcpy(sha1, oe->sha1);
1898 } else if (!get_sha1(from, sha1)) {
1899 unsigned long size;
1900 char *buf;
1901
1902 buf = read_object_with_reference(sha1,
1903 type_names[OBJ_COMMIT], &size, sha1);
1904 if (!buf || size < 46)
1905 die("Not a valid commit: %s", from);
1906 free(buf);
1907 } else
1908 die("Invalid ref name or SHA1 expression: %s", from);
1909 read_next_command();
1910
1911 /* tagger ... */
1912 if (strncmp("tagger ", command_buf.buf, 7))
1913 die("Expected tagger command, got %s", command_buf.buf);
1914 tagger = parse_ident(command_buf.buf + 7);
1915
1916 /* tag payload/message */
1917 read_next_command();
1918 msg = cmd_data(&msglen);
1919
1920 /* build the tag object */
1921 size_dbuf(&new_data, 67+strlen(t->name)+strlen(tagger)+msglen);
1922 sp = new_data.buffer;
1923 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
1924 sp += sprintf(sp, "type %s\n", type_names[OBJ_COMMIT]);
1925 sp += sprintf(sp, "tag %s\n", t->name);
1926 sp += sprintf(sp, "tagger %s\n", tagger);
1927 *sp++ = '\n';
1928 memcpy(sp, msg, msglen);
1929 sp += msglen;
1930 free(tagger);
1931 free(msg);
1932
1933 if (store_object(OBJ_TAG, new_data.buffer,
1934 sp - (char*)new_data.buffer,
1935 NULL, t->sha1, 0))
1936 t->pack_id = MAX_PACK_ID;
1937 else
1938 t->pack_id = pack_id;
1939 }
1940
1941 static void cmd_reset_branch(void)
1942 {
1943 struct branch *b;
1944 char *sp;
1945
1946 /* Obtain the branch name from the rest of our command */
1947 sp = strchr(command_buf.buf, ' ') + 1;
1948 b = lookup_branch(sp);
1949 if (b) {
1950 hashclr(b->sha1);
1951 hashclr(b->branch_tree.versions[0].sha1);
1952 hashclr(b->branch_tree.versions[1].sha1);
1953 if (b->branch_tree.tree) {
1954 release_tree_content_recursive(b->branch_tree.tree);
1955 b->branch_tree.tree = NULL;
1956 }
1957 }
1958 else
1959 b = new_branch(sp);
1960 read_next_command();
1961 cmd_from(b);
1962 }
1963
1964 static void cmd_checkpoint(void)
1965 {
1966 if (object_count) {
1967 cycle_packfile();
1968 dump_branches();
1969 dump_tags();
1970 dump_marks();
1971 }
1972 read_next_command();
1973 }
1974
1975 static const char fast_import_usage[] =
1976 "git-fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
1977
1978 int main(int argc, const char **argv)
1979 {
1980 int i, show_stats = 1;
1981
1982 git_config(git_default_config);
1983
1984 for (i = 1; i < argc; i++) {
1985 const char *a = argv[i];
1986
1987 if (*a != '-' || !strcmp(a, "--"))
1988 break;
1989 else if (!strncmp(a, "--date-format=", 14)) {
1990 const char *fmt = a + 14;
1991 if (!strcmp(fmt, "raw"))
1992 whenspec = WHENSPEC_RAW;
1993 else if (!strcmp(fmt, "rfc2822"))
1994 whenspec = WHENSPEC_RFC2822;
1995 else if (!strcmp(fmt, "now"))
1996 whenspec = WHENSPEC_NOW;
1997 else
1998 die("unknown --date-format argument %s", fmt);
1999 }
2000 else if (!strncmp(a, "--max-pack-size=", 16))
2001 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
2002 else if (!strncmp(a, "--depth=", 8))
2003 max_depth = strtoul(a + 8, NULL, 0);
2004 else if (!strncmp(a, "--active-branches=", 18))
2005 max_active_branches = strtoul(a + 18, NULL, 0);
2006 else if (!strncmp(a, "--export-marks=", 15))
2007 mark_file = a + 15;
2008 else if (!strncmp(a, "--export-pack-edges=", 20)) {
2009 if (pack_edges)
2010 fclose(pack_edges);
2011 pack_edges = fopen(a + 20, "a");
2012 if (!pack_edges)
2013 die("Cannot open %s: %s", a + 20, strerror(errno));
2014 } else if (!strcmp(a, "--force"))
2015 force_update = 1;
2016 else if (!strcmp(a, "--quiet"))
2017 show_stats = 0;
2018 else if (!strcmp(a, "--stats"))
2019 show_stats = 1;
2020 else
2021 die("unknown option %s", a);
2022 }
2023 if (i != argc)
2024 usage(fast_import_usage);
2025
2026 alloc_objects(object_entry_alloc);
2027 strbuf_init(&command_buf);
2028
2029 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2030 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2031 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2032 marks = pool_calloc(1, sizeof(struct mark_set));
2033
2034 start_packfile();
2035 for (;;) {
2036 read_next_command();
2037 if (command_buf.eof)
2038 break;
2039 else if (!strcmp("blob", command_buf.buf))
2040 cmd_new_blob();
2041 else if (!strncmp("commit ", command_buf.buf, 7))
2042 cmd_new_commit();
2043 else if (!strncmp("tag ", command_buf.buf, 4))
2044 cmd_new_tag();
2045 else if (!strncmp("reset ", command_buf.buf, 6))
2046 cmd_reset_branch();
2047 else if (!strcmp("checkpoint", command_buf.buf))
2048 cmd_checkpoint();
2049 else
2050 die("Unsupported command: %s", command_buf.buf);
2051 }
2052 end_packfile();
2053
2054 dump_branches();
2055 dump_tags();
2056 unkeep_all_packs();
2057 dump_marks();
2058
2059 if (pack_edges)
2060 fclose(pack_edges);
2061
2062 if (show_stats) {
2063 uintmax_t total_count = 0, duplicate_count = 0;
2064 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2065 total_count += object_count_by_type[i];
2066 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2067 duplicate_count += duplicate_count_by_type[i];
2068
2069 fprintf(stderr, "%s statistics:\n", argv[0]);
2070 fprintf(stderr, "---------------------------------------------------------------------\n");
2071 fprintf(stderr, "Alloc'd objects: " UM10_FMT "\n", alloc_count);
2072 fprintf(stderr, "Total objects: " UM10_FMT " (" UM10_FMT " duplicates )\n", total_count, duplicate_count);
2073 fprintf(stderr, " blobs : " UM10_FMT " (" UM10_FMT " duplicates " UM10_FMT " deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
2074 fprintf(stderr, " trees : " UM10_FMT " (" UM10_FMT " duplicates " UM10_FMT " deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
2075 fprintf(stderr, " commits: " UM10_FMT " (" UM10_FMT " duplicates " UM10_FMT " deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
2076 fprintf(stderr, " tags : " UM10_FMT " (" UM10_FMT " duplicates " UM10_FMT " deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
2077 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2078 fprintf(stderr, " marks: " UM10_FMT " (" UM10_FMT " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2079 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2080 fprintf(stderr, "Memory total: " UM10_FMT " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2081 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2082 fprintf(stderr, " objects: " UM10_FMT " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2083 fprintf(stderr, "---------------------------------------------------------------------\n");
2084 pack_report();
2085 fprintf(stderr, "---------------------------------------------------------------------\n");
2086 fprintf(stderr, "\n");
2087 }
2088
2089 return failure ? 1 : 0;
2090 }