?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
improve delta long block matching with big files
[git/gitweb.git] / fast-import.c
blobf9bfcc72c87bf79fcb7a5faef01bc0d12aa15420
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 struct object_entry
137 {
138 struct object_entry *next;
139 uint32_t offset;
140 unsigned type : TYPE_BITS;
141 unsigned pack_id : PACK_ID_BITS;
142 unsigned char sha1[20];
143 };
144
145 struct object_entry_pool
146 {
147 struct object_entry_pool *next_pool;
148 struct object_entry *next_free;
149 struct object_entry *end;
150 struct object_entry entries[FLEX_ARRAY]; /* more */
151 };
152
153 struct mark_set
154 {
155 union {
156 struct object_entry *marked[1024];
157 struct mark_set *sets[1024];
158 } data;
159 unsigned int shift;
160 };
161
162 struct last_object
163 {
164 void *data;
165 unsigned long len;
166 uint32_t offset;
167 unsigned int depth;
168 unsigned no_free:1;
169 };
170
171 struct mem_pool
172 {
173 struct mem_pool *next_pool;
174 char *next_free;
175 char *end;
176 char space[FLEX_ARRAY]; /* more */
177 };
178
179 struct atom_str
180 {
181 struct atom_str *next_atom;
182 unsigned short str_len;
183 char str_dat[FLEX_ARRAY]; /* more */
184 };
185
186 struct tree_content;
187 struct tree_entry
188 {
189 struct tree_content *tree;
190 struct atom_str* name;
191 struct tree_entry_ms
192 {
193 uint16_t mode;
194 unsigned char sha1[20];
195 } versions[2];
196 };
197
198 struct tree_content
199 {
200 unsigned int entry_capacity; /* must match avail_tree_content */
201 unsigned int entry_count;
202 unsigned int delta_depth;
203 struct tree_entry *entries[FLEX_ARRAY]; /* more */
204 };
205
206 struct avail_tree_content
207 {
208 unsigned int entry_capacity; /* must match tree_content */
209 struct avail_tree_content *next_avail;
210 };
211
212 struct branch
213 {
214 struct branch *table_next_branch;
215 struct branch *active_next_branch;
216 const char *name;
217 struct tree_entry branch_tree;
218 uintmax_t last_commit;
219 unsigned active : 1;
220 unsigned pack_id : PACK_ID_BITS;
221 unsigned char sha1[20];
222 };
223
224 struct tag
225 {
226 struct tag *next_tag;
227 const char *name;
228 unsigned int pack_id;
229 unsigned char sha1[20];
230 };
231
232 struct dbuf
233 {
234 void *buffer;
235 size_t capacity;
236 };
237
238 struct hash_list
239 {
240 struct hash_list *next;
241 unsigned char sha1[20];
242 };
243
244 typedef enum {
245 WHENSPEC_RAW = 1,
246 WHENSPEC_RFC2822,
247 WHENSPEC_NOW,
248 } whenspec_type;
249
250 /* Configured limits on output */
251 static unsigned long max_depth = 10;
252 static off_t max_packsize = (1LL << 32) - 1;
253 static int force_update;
254
255 /* Stats and misc. counters */
256 static uintmax_t alloc_count;
257 static uintmax_t marks_set_count;
258 static uintmax_t object_count_by_type[1 << TYPE_BITS];
259 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
260 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
261 static unsigned long object_count;
262 static unsigned long branch_count;
263 static unsigned long branch_load_count;
264 static int failure;
265 static FILE *pack_edges;
266
267 /* Memory pools */
268 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
269 static size_t total_allocd;
270 static struct mem_pool *mem_pool;
271
272 /* Atom management */
273 static unsigned int atom_table_sz = 4451;
274 static unsigned int atom_cnt;
275 static struct atom_str **atom_table;
276
277 /* The .pack file being generated */
278 static unsigned int pack_id;
279 static struct packed_git *pack_data;
280 static struct packed_git **all_packs;
281 static unsigned long pack_size;
282
283 /* Table of objects we've written. */
284 static unsigned int object_entry_alloc = 5000;
285 static struct object_entry_pool *blocks;
286 static struct object_entry *object_table[1 << 16];
287 static struct mark_set *marks;
288 static const char* mark_file;
289
290 /* Our last blob */
291 static struct last_object last_blob;
292
293 /* Tree management */
294 static unsigned int tree_entry_alloc = 1000;
295 static void *avail_tree_entry;
296 static unsigned int avail_tree_table_sz = 100;
297 static struct avail_tree_content **avail_tree_table;
298 static struct dbuf old_tree;
299 static struct dbuf new_tree;
300
301 /* Branch data */
302 static unsigned long max_active_branches = 5;
303 static unsigned long cur_active_branches;
304 static unsigned long branch_table_sz = 1039;
305 static struct branch **branch_table;
306 static struct branch *active_branches;
307
308 /* Tag data */
309 static struct tag *first_tag;
310 static struct tag *last_tag;
311
312 /* Input stream parsing */
313 static whenspec_type whenspec = WHENSPEC_RAW;
314 static struct strbuf command_buf;
315 static uintmax_t next_mark;
316 static struct dbuf new_data;
317
318
319 static void alloc_objects(unsigned int cnt)
320 {
321 struct object_entry_pool *b;
322
323 b = xmalloc(sizeof(struct object_entry_pool)
324 + cnt * sizeof(struct object_entry));
325 b->next_pool = blocks;
326 b->next_free = b->entries;
327 b->end = b->entries + cnt;
328 blocks = b;
329 alloc_count += cnt;
330 }
331
332 static struct object_entry *new_object(unsigned char *sha1)
333 {
334 struct object_entry *e;
335
336 if (blocks->next_free == blocks->end)
337 alloc_objects(object_entry_alloc);
338
339 e = blocks->next_free++;
340 hashcpy(e->sha1, sha1);
341 return e;
342 }
343
344 static struct object_entry *find_object(unsigned char *sha1)
345 {
346 unsigned int h = sha1[0] << 8 | sha1[1];
347 struct object_entry *e;
348 for (e = object_table[h]; e; e = e->next)
349 if (!hashcmp(sha1, e->sha1))
350 return e;
351 return NULL;
352 }
353
354 static struct object_entry *insert_object(unsigned char *sha1)
355 {
356 unsigned int h = sha1[0] << 8 | sha1[1];
357 struct object_entry *e = object_table[h];
358 struct object_entry *p = NULL;
359
360 while (e) {
361 if (!hashcmp(sha1, e->sha1))
362 return e;
363 p = e;
364 e = e->next;
365 }
366
367 e = new_object(sha1);
368 e->next = NULL;
369 e->offset = 0;
370 if (p)
371 p->next = e;
372 else
373 object_table[h] = e;
374 return e;
375 }
376
377 static unsigned int hc_str(const char *s, size_t len)
378 {
379 unsigned int r = 0;
380 while (len-- > 0)
381 r = r * 31 + *s++;
382 return r;
383 }
384
385 static void *pool_alloc(size_t len)
386 {
387 struct mem_pool *p;
388 void *r;
389
390 for (p = mem_pool; p; p = p->next_pool)
391 if ((p->end - p->next_free >= len))
392 break;
393
394 if (!p) {
395 if (len >= (mem_pool_alloc/2)) {
396 total_allocd += len;
397 return xmalloc(len);
398 }
399 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
400 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
401 p->next_pool = mem_pool;
402 p->next_free = p->space;
403 p->end = p->next_free + mem_pool_alloc;
404 mem_pool = p;
405 }
406
407 r = p->next_free;
408 /* round out to a pointer alignment */
409 if (len & (sizeof(void*) - 1))
410 len += sizeof(void*) - (len & (sizeof(void*) - 1));
411 p->next_free += len;
412 return r;
413 }
414
415 static void *pool_calloc(size_t count, size_t size)
416 {
417 size_t len = count * size;
418 void *r = pool_alloc(len);
419 memset(r, 0, len);
420 return r;
421 }
422
423 static char *pool_strdup(const char *s)
424 {
425 char *r = pool_alloc(strlen(s) + 1);
426 strcpy(r, s);
427 return r;
428 }
429
430 static void size_dbuf(struct dbuf *b, size_t maxlen)
431 {
432 if (b->buffer) {
433 if (b->capacity >= maxlen)
434 return;
435 free(b->buffer);
436 }
437 b->capacity = ((maxlen / 1024) + 1) * 1024;
438 b->buffer = xmalloc(b->capacity);
439 }
440
441 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
442 {
443 struct mark_set *s = marks;
444 while ((idnum >> s->shift) >= 1024) {
445 s = pool_calloc(1, sizeof(struct mark_set));
446 s->shift = marks->shift + 10;
447 s->data.sets[0] = marks;
448 marks = s;
449 }
450 while (s->shift) {
451 uintmax_t i = idnum >> s->shift;
452 idnum -= i << s->shift;
453 if (!s->data.sets[i]) {
454 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
455 s->data.sets[i]->shift = s->shift - 10;
456 }
457 s = s->data.sets[i];
458 }
459 if (!s->data.marked[idnum])
460 marks_set_count++;
461 s->data.marked[idnum] = oe;
462 }
463
464 static struct object_entry *find_mark(uintmax_t idnum)
465 {
466 uintmax_t orig_idnum = idnum;
467 struct mark_set *s = marks;
468 struct object_entry *oe = NULL;
469 if ((idnum >> s->shift) < 1024) {
470 while (s && s->shift) {
471 uintmax_t i = idnum >> s->shift;
472 idnum -= i << s->shift;
473 s = s->data.sets[i];
474 }
475 if (s)
476 oe = s->data.marked[idnum];
477 }
478 if (!oe)
479 die("mark :%" PRIuMAX " not declared", orig_idnum);
480 return oe;
481 }
482
483 static struct atom_str *to_atom(const char *s, unsigned short len)
484 {
485 unsigned int hc = hc_str(s, len) % atom_table_sz;
486 struct atom_str *c;
487
488 for (c = atom_table[hc]; c; c = c->next_atom)
489 if (c->str_len == len && !strncmp(s, c->str_dat, len))
490 return c;
491
492 c = pool_alloc(sizeof(struct atom_str) + len + 1);
493 c->str_len = len;
494 strncpy(c->str_dat, s, len);
495 c->str_dat[len] = 0;
496 c->next_atom = atom_table[hc];
497 atom_table[hc] = c;
498 atom_cnt++;
499 return c;
500 }
501
502 static struct branch *lookup_branch(const char *name)
503 {
504 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
505 struct branch *b;
506
507 for (b = branch_table[hc]; b; b = b->table_next_branch)
508 if (!strcmp(name, b->name))
509 return b;
510 return NULL;
511 }
512
513 static struct branch *new_branch(const char *name)
514 {
515 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
516 struct branch* b = lookup_branch(name);
517
518 if (b)
519 die("Invalid attempt to create duplicate branch: %s", name);
520 if (check_ref_format(name))
521 die("Branch name doesn't conform to GIT standards: %s", name);
522
523 b = pool_calloc(1, sizeof(struct branch));
524 b->name = pool_strdup(name);
525 b->table_next_branch = branch_table[hc];
526 b->branch_tree.versions[0].mode = S_IFDIR;
527 b->branch_tree.versions[1].mode = S_IFDIR;
528 b->active = 0;
529 b->pack_id = MAX_PACK_ID;
530 branch_table[hc] = b;
531 branch_count++;
532 return b;
533 }
534
535 static unsigned int hc_entries(unsigned int cnt)
536 {
537 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
538 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
539 }
540
541 static struct tree_content *new_tree_content(unsigned int cnt)
542 {
543 struct avail_tree_content *f, *l = NULL;
544 struct tree_content *t;
545 unsigned int hc = hc_entries(cnt);
546
547 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
548 if (f->entry_capacity >= cnt)
549 break;
550
551 if (f) {
552 if (l)
553 l->next_avail = f->next_avail;
554 else
555 avail_tree_table[hc] = f->next_avail;
556 } else {
557 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
558 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
559 f->entry_capacity = cnt;
560 }
561
562 t = (struct tree_content*)f;
563 t->entry_count = 0;
564 t->delta_depth = 0;
565 return t;
566 }
567
568 static void release_tree_entry(struct tree_entry *e);
569 static void release_tree_content(struct tree_content *t)
570 {
571 struct avail_tree_content *f = (struct avail_tree_content*)t;
572 unsigned int hc = hc_entries(f->entry_capacity);
573 f->next_avail = avail_tree_table[hc];
574 avail_tree_table[hc] = f;
575 }
576
577 static void release_tree_content_recursive(struct tree_content *t)
578 {
579 unsigned int i;
580 for (i = 0; i < t->entry_count; i++)
581 release_tree_entry(t->entries[i]);
582 release_tree_content(t);
583 }
584
585 static struct tree_content *grow_tree_content(
586 struct tree_content *t,
587 int amt)
588 {
589 struct tree_content *r = new_tree_content(t->entry_count + amt);
590 r->entry_count = t->entry_count;
591 r->delta_depth = t->delta_depth;
592 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
593 release_tree_content(t);
594 return r;
595 }
596
597 static struct tree_entry *new_tree_entry(void)
598 {
599 struct tree_entry *e;
600
601 if (!avail_tree_entry) {
602 unsigned int n = tree_entry_alloc;
603 total_allocd += n * sizeof(struct tree_entry);
604 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
605 while (n-- > 1) {
606 *((void**)e) = e + 1;
607 e++;
608 }
609 *((void**)e) = NULL;
610 }
611
612 e = avail_tree_entry;
613 avail_tree_entry = *((void**)e);
614 return e;
615 }
616
617 static void release_tree_entry(struct tree_entry *e)
618 {
619 if (e->tree)
620 release_tree_content_recursive(e->tree);
621 *((void**)e) = avail_tree_entry;
622 avail_tree_entry = e;
623 }
624
625 static void start_packfile(void)
626 {
627 static char tmpfile[PATH_MAX];
628 struct packed_git *p;
629 struct pack_header hdr;
630 int pack_fd;
631
632 snprintf(tmpfile, sizeof(tmpfile),
633 "%s/tmp_pack_XXXXXX", get_object_directory());
634 pack_fd = mkstemp(tmpfile);
635 if (pack_fd < 0)
636 die("Can't create %s: %s", tmpfile, strerror(errno));
637 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
638 strcpy(p->pack_name, tmpfile);
639 p->pack_fd = pack_fd;
640
641 hdr.hdr_signature = htonl(PACK_SIGNATURE);
642 hdr.hdr_version = htonl(2);
643 hdr.hdr_entries = 0;
644 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
645
646 pack_data = p;
647 pack_size = sizeof(hdr);
648 object_count = 0;
649
650 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
651 all_packs[pack_id] = p;
652 }
653
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 }
660
661 static char *create_index(void)
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 uint32_t array[256];
669 int i, idx_fd;
670
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);
682
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 }
695
696 snprintf(tmpfile, sizeof(tmpfile),
697 "%s/tmp_idx_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 uint32_t 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 }
716
717 static char *keep_pack(char *curr_index_name)
718 {
719 static char name[PATH_MAX];
720 static const char *keep_msg = "fast-import";
721 int keep_fd;
722
723 chmod(pack_data->pack_name, 0444);
724 chmod(curr_index_name, 0444);
725
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);
733
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");
738
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 }
745
746 static void unkeep_all_packs(void)
747 {
748 static char name[PATH_MAX];
749 int k;
750
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 }
758
759 static void end_packfile(void)
760 {
761 struct packed_git *old_p = pack_data, *new_p;
762
763 if (object_count) {
764 char *idx_name;
765 int i;
766 struct branch *b;
767 struct tag *t;
768
769 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
770 pack_data->pack_name, object_count);
771 close(pack_data->pack_fd);
772 idx_name = keep_pack(create_index());
773
774 /* Register the packfile with core git's machinary. */
775 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
776 if (!new_p)
777 die("core git rejected index %s", idx_name);
778 new_p->windows = old_p->windows;
779 all_packs[pack_id] = new_p;
780 install_packed_git(new_p);
781
782 /* Print the boundary */
783 if (pack_edges) {
784 fprintf(pack_edges, "%s:", new_p->pack_name);
785 for (i = 0; i < branch_table_sz; i++) {
786 for (b = branch_table[i]; b; b = b->table_next_branch) {
787 if (b->pack_id == pack_id)
788 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
789 }
790 }
791 for (t = first_tag; t; t = t->next_tag) {
792 if (t->pack_id == pack_id)
793 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
794 }
795 fputc('\n', pack_edges);
796 fflush(pack_edges);
797 }
798
799 pack_id++;
800 }
801 else
802 unlink(old_p->pack_name);
803 free(old_p);
804
805 /* We can't carry a delta across packfiles. */
806 free(last_blob.data);
807 last_blob.data = NULL;
808 last_blob.len = 0;
809 last_blob.offset = 0;
810 last_blob.depth = 0;
811 }
812
813 static void cycle_packfile(void)
814 {
815 end_packfile();
816 start_packfile();
817 }
818
819 static size_t encode_header(
820 enum object_type type,
821 size_t size,
822 unsigned char *hdr)
823 {
824 int n = 1;
825 unsigned char c;
826
827 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
828 die("bad type %d", type);
829
830 c = (type << 4) | (size & 15);
831 size >>= 4;
832 while (size) {
833 *hdr++ = c | 0x80;
834 c = size & 0x7f;
835 size >>= 7;
836 n++;
837 }
838 *hdr = c;
839 return n;
840 }
841
842 static int store_object(
843 enum object_type type,
844 void *dat,
845 size_t datlen,
846 struct last_object *last,
847 unsigned char *sha1out,
848 uintmax_t mark)
849 {
850 void *out, *delta;
851 struct object_entry *e;
852 unsigned char hdr[96];
853 unsigned char sha1[20];
854 unsigned long hdrlen, deltalen;
855 SHA_CTX c;
856 z_stream s;
857
858 hdrlen = sprintf((char*)hdr,"%s %lu", typename(type),
859 (unsigned long)datlen) + 1;
860 SHA1_Init(&c);
861 SHA1_Update(&c, hdr, hdrlen);
862 SHA1_Update(&c, dat, datlen);
863 SHA1_Final(sha1, &c);
864 if (sha1out)
865 hashcpy(sha1out, sha1);
866
867 e = insert_object(sha1);
868 if (mark)
869 insert_mark(mark, e);
870 if (e->offset) {
871 duplicate_count_by_type[type]++;
872 return 1;
873 } else if (find_sha1_pack(sha1, packed_git)) {
874 e->type = type;
875 e->pack_id = MAX_PACK_ID;
876 e->offset = 1; /* just not zero! */
877 duplicate_count_by_type[type]++;
878 return 1;
879 }
880
881 if (last && last->data && last->depth < max_depth) {
882 delta = diff_delta(last->data, last->len,
883 dat, datlen,
884 &deltalen, 0);
885 if (delta && deltalen >= datlen) {
886 free(delta);
887 delta = NULL;
888 }
889 } else
890 delta = NULL;
891
892 memset(&s, 0, sizeof(s));
893 deflateInit(&s, zlib_compression_level);
894 if (delta) {
895 s.next_in = delta;
896 s.avail_in = deltalen;
897 } else {
898 s.next_in = dat;
899 s.avail_in = datlen;
900 }
901 s.avail_out = deflateBound(&s, s.avail_in);
902 s.next_out = out = xmalloc(s.avail_out);
903 while (deflate(&s, Z_FINISH) == Z_OK)
904 /* nothing */;
905 deflateEnd(&s);
906
907 /* Determine if we should auto-checkpoint. */
908 if ((pack_size + 60 + s.total_out) > max_packsize
909 || (pack_size + 60 + s.total_out) < pack_size) {
910
911 /* This new object needs to *not* have the current pack_id. */
912 e->pack_id = pack_id + 1;
913 cycle_packfile();
914
915 /* We cannot carry a delta into the new pack. */
916 if (delta) {
917 free(delta);
918 delta = NULL;
919
920 memset(&s, 0, sizeof(s));
921 deflateInit(&s, zlib_compression_level);
922 s.next_in = dat;
923 s.avail_in = datlen;
924 s.avail_out = deflateBound(&s, s.avail_in);
925 s.next_out = out = xrealloc(out, s.avail_out);
926 while (deflate(&s, Z_FINISH) == Z_OK)
927 /* nothing */;
928 deflateEnd(&s);
929 }
930 }
931
932 e->type = type;
933 e->pack_id = pack_id;
934 e->offset = pack_size;
935 object_count++;
936 object_count_by_type[type]++;
937
938 if (delta) {
939 unsigned long ofs = e->offset - last->offset;
940 unsigned pos = sizeof(hdr) - 1;
941
942 delta_count_by_type[type]++;
943 last->depth++;
944
945 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
946 write_or_die(pack_data->pack_fd, hdr, hdrlen);
947 pack_size += hdrlen;
948
949 hdr[pos] = ofs & 127;
950 while (ofs >>= 7)
951 hdr[--pos] = 128 | (--ofs & 127);
952 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
953 pack_size += sizeof(hdr) - pos;
954 } else {
955 if (last)
956 last->depth = 0;
957 hdrlen = encode_header(type, datlen, hdr);
958 write_or_die(pack_data->pack_fd, hdr, hdrlen);
959 pack_size += hdrlen;
960 }
961
962 write_or_die(pack_data->pack_fd, out, s.total_out);
963 pack_size += s.total_out;
964
965 free(out);
966 free(delta);
967 if (last) {
968 if (!last->no_free)
969 free(last->data);
970 last->data = dat;
971 last->offset = e->offset;
972 last->len = datlen;
973 }
974 return 0;
975 }
976
977 static void *gfi_unpack_entry(
978 struct object_entry *oe,
979 unsigned long *sizep)
980 {
981 enum object_type type;
982 struct packed_git *p = all_packs[oe->pack_id];
983 if (p == pack_data)
984 p->pack_size = pack_size + 20;
985 return unpack_entry(p, oe->offset, &type, sizep);
986 }
987
988 static const char *get_mode(const char *str, uint16_t *modep)
989 {
990 unsigned char c;
991 uint16_t mode = 0;
992
993 while ((c = *str++) != ' ') {
994 if (c < '0' || c > '7')
995 return NULL;
996 mode = (mode << 3) + (c - '0');
997 }
998 *modep = mode;
999 return str;
1000 }
1001
1002 static void load_tree(struct tree_entry *root)
1003 {
1004 unsigned char* sha1 = root->versions[1].sha1;
1005 struct object_entry *myoe;
1006 struct tree_content *t;
1007 unsigned long size;
1008 char *buf;
1009 const char *c;
1010
1011 root->tree = t = new_tree_content(8);
1012 if (is_null_sha1(sha1))
1013 return;
1014
1015 myoe = find_object(sha1);
1016 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1017 if (myoe->type != OBJ_TREE)
1018 die("Not a tree: %s", sha1_to_hex(sha1));
1019 t->delta_depth = 0;
1020 buf = gfi_unpack_entry(myoe, &size);
1021 } else {
1022 enum object_type type;
1023 buf = read_sha1_file(sha1, &type, &size);
1024 if (!buf || type != OBJ_TREE)
1025 die("Can't load tree %s", sha1_to_hex(sha1));
1026 }
1027
1028 c = buf;
1029 while (c != (buf + size)) {
1030 struct tree_entry *e = new_tree_entry();
1031
1032 if (t->entry_count == t->entry_capacity)
1033 root->tree = t = grow_tree_content(t, t->entry_count);
1034 t->entries[t->entry_count++] = e;
1035
1036 e->tree = NULL;
1037 c = get_mode(c, &e->versions[1].mode);
1038 if (!c)
1039 die("Corrupt mode in %s", sha1_to_hex(sha1));
1040 e->versions[0].mode = e->versions[1].mode;
1041 e->name = to_atom(c, strlen(c));
1042 c += e->name->str_len + 1;
1043 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1044 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1045 c += 20;
1046 }
1047 free(buf);
1048 }
1049
1050 static int tecmp0 (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[0].mode,
1056 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1057 }
1058
1059 static int tecmp1 (const void *_a, const void *_b)
1060 {
1061 struct tree_entry *a = *((struct tree_entry**)_a);
1062 struct tree_entry *b = *((struct tree_entry**)_b);
1063 return base_name_compare(
1064 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1065 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1066 }
1067
1068 static void mktree(struct tree_content *t,
1069 int v,
1070 unsigned long *szp,
1071 struct dbuf *b)
1072 {
1073 size_t maxlen = 0;
1074 unsigned int i;
1075 char *c;
1076
1077 if (!v)
1078 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1079 else
1080 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1081
1082 for (i = 0; i < t->entry_count; i++) {
1083 if (t->entries[i]->versions[v].mode)
1084 maxlen += t->entries[i]->name->str_len + 34;
1085 }
1086
1087 size_dbuf(b, maxlen);
1088 c = b->buffer;
1089 for (i = 0; i < t->entry_count; i++) {
1090 struct tree_entry *e = t->entries[i];
1091 if (!e->versions[v].mode)
1092 continue;
1093 c += sprintf(c, "%o", (unsigned int)e->versions[v].mode);
1094 *c++ = ' ';
1095 strcpy(c, e->name->str_dat);
1096 c += e->name->str_len + 1;
1097 hashcpy((unsigned char*)c, e->versions[v].sha1);
1098 c += 20;
1099 }
1100 *szp = c - (char*)b->buffer;
1101 }
1102
1103 static void store_tree(struct tree_entry *root)
1104 {
1105 struct tree_content *t = root->tree;
1106 unsigned int i, j, del;
1107 unsigned long new_len;
1108 struct last_object lo;
1109 struct object_entry *le;
1110
1111 if (!is_null_sha1(root->versions[1].sha1))
1112 return;
1113
1114 for (i = 0; i < t->entry_count; i++) {
1115 if (t->entries[i]->tree)
1116 store_tree(t->entries[i]);
1117 }
1118
1119 le = find_object(root->versions[0].sha1);
1120 if (!S_ISDIR(root->versions[0].mode)
1121 || !le
1122 || le->pack_id != pack_id) {
1123 lo.data = NULL;
1124 lo.depth = 0;
1125 lo.no_free = 0;
1126 } else {
1127 mktree(t, 0, &lo.len, &old_tree);
1128 lo.data = old_tree.buffer;
1129 lo.offset = le->offset;
1130 lo.depth = t->delta_depth;
1131 lo.no_free = 1;
1132 }
1133
1134 mktree(t, 1, &new_len, &new_tree);
1135 store_object(OBJ_TREE, new_tree.buffer, new_len,
1136 &lo, root->versions[1].sha1, 0);
1137
1138 t->delta_depth = lo.depth;
1139 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1140 struct tree_entry *e = t->entries[i];
1141 if (e->versions[1].mode) {
1142 e->versions[0].mode = e->versions[1].mode;
1143 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1144 t->entries[j++] = e;
1145 } else {
1146 release_tree_entry(e);
1147 del++;
1148 }
1149 }
1150 t->entry_count -= del;
1151 }
1152
1153 static int tree_content_set(
1154 struct tree_entry *root,
1155 const char *p,
1156 const unsigned char *sha1,
1157 const uint16_t mode)
1158 {
1159 struct tree_content *t = root->tree;
1160 const char *slash1;
1161 unsigned int i, n;
1162 struct tree_entry *e;
1163
1164 slash1 = strchr(p, '/');
1165 if (slash1)
1166 n = slash1 - p;
1167 else
1168 n = strlen(p);
1169 if (!n)
1170 die("Empty path component found in input");
1171
1172 for (i = 0; i < t->entry_count; i++) {
1173 e = t->entries[i];
1174 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1175 if (!slash1) {
1176 if (e->versions[1].mode == mode
1177 && !hashcmp(e->versions[1].sha1, sha1))
1178 return 0;
1179 e->versions[1].mode = mode;
1180 hashcpy(e->versions[1].sha1, sha1);
1181 if (e->tree) {
1182 release_tree_content_recursive(e->tree);
1183 e->tree = NULL;
1184 }
1185 hashclr(root->versions[1].sha1);
1186 return 1;
1187 }
1188 if (!S_ISDIR(e->versions[1].mode)) {
1189 e->tree = new_tree_content(8);
1190 e->versions[1].mode = S_IFDIR;
1191 }
1192 if (!e->tree)
1193 load_tree(e);
1194 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
1195 hashclr(root->versions[1].sha1);
1196 return 1;
1197 }
1198 return 0;
1199 }
1200 }
1201
1202 if (t->entry_count == t->entry_capacity)
1203 root->tree = t = grow_tree_content(t, t->entry_count);
1204 e = new_tree_entry();
1205 e->name = to_atom(p, n);
1206 e->versions[0].mode = 0;
1207 hashclr(e->versions[0].sha1);
1208 t->entries[t->entry_count++] = e;
1209 if (slash1) {
1210 e->tree = new_tree_content(8);
1211 e->versions[1].mode = S_IFDIR;
1212 tree_content_set(e, slash1 + 1, sha1, mode);
1213 } else {
1214 e->tree = NULL;
1215 e->versions[1].mode = mode;
1216 hashcpy(e->versions[1].sha1, sha1);
1217 }
1218 hashclr(root->versions[1].sha1);
1219 return 1;
1220 }
1221
1222 static int tree_content_remove(struct tree_entry *root, const char *p)
1223 {
1224 struct tree_content *t = root->tree;
1225 const char *slash1;
1226 unsigned int i, n;
1227 struct tree_entry *e;
1228
1229 slash1 = strchr(p, '/');
1230 if (slash1)
1231 n = slash1 - p;
1232 else
1233 n = strlen(p);
1234
1235 for (i = 0; i < t->entry_count; i++) {
1236 e = t->entries[i];
1237 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1238 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1239 goto del_entry;
1240 if (!e->tree)
1241 load_tree(e);
1242 if (tree_content_remove(e, slash1 + 1)) {
1243 for (n = 0; n < e->tree->entry_count; n++) {
1244 if (e->tree->entries[n]->versions[1].mode) {
1245 hashclr(root->versions[1].sha1);
1246 return 1;
1247 }
1248 }
1249 goto del_entry;
1250 }
1251 return 0;
1252 }
1253 }
1254 return 0;
1255
1256 del_entry:
1257 if (e->tree) {
1258 release_tree_content_recursive(e->tree);
1259 e->tree = NULL;
1260 }
1261 e->versions[1].mode = 0;
1262 hashclr(e->versions[1].sha1);
1263 hashclr(root->versions[1].sha1);
1264 return 1;
1265 }
1266
1267 static int update_branch(struct branch *b)
1268 {
1269 static const char *msg = "fast-import";
1270 struct ref_lock *lock;
1271 unsigned char old_sha1[20];
1272
1273 if (read_ref(b->name, old_sha1))
1274 hashclr(old_sha1);
1275 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1276 if (!lock)
1277 return error("Unable to lock %s", b->name);
1278 if (!force_update && !is_null_sha1(old_sha1)) {
1279 struct commit *old_cmit, *new_cmit;
1280
1281 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1282 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1283 if (!old_cmit || !new_cmit) {
1284 unlock_ref(lock);
1285 return error("Branch %s is missing commits.", b->name);
1286 }
1287
1288 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1289 unlock_ref(lock);
1290 warning("Not updating %s"
1291 " (new tip %s does not contain %s)",
1292 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1293 return -1;
1294 }
1295 }
1296 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1297 return error("Unable to update %s", b->name);
1298 return 0;
1299 }
1300
1301 static void dump_branches(void)
1302 {
1303 unsigned int i;
1304 struct branch *b;
1305
1306 for (i = 0; i < branch_table_sz; i++) {
1307 for (b = branch_table[i]; b; b = b->table_next_branch)
1308 failure |= update_branch(b);
1309 }
1310 }
1311
1312 static void dump_tags(void)
1313 {
1314 static const char *msg = "fast-import";
1315 struct tag *t;
1316 struct ref_lock *lock;
1317 char ref_name[PATH_MAX];
1318
1319 for (t = first_tag; t; t = t->next_tag) {
1320 sprintf(ref_name, "tags/%s", t->name);
1321 lock = lock_ref_sha1(ref_name, NULL);
1322 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1323 failure |= error("Unable to update %s", ref_name);
1324 }
1325 }
1326
1327 static void dump_marks_helper(FILE *f,
1328 uintmax_t base,
1329 struct mark_set *m)
1330 {
1331 uintmax_t k;
1332 if (m->shift) {
1333 for (k = 0; k < 1024; k++) {
1334 if (m->data.sets[k])
1335 dump_marks_helper(f, (base + k) << m->shift,
1336 m->data.sets[k]);
1337 }
1338 } else {
1339 for (k = 0; k < 1024; k++) {
1340 if (m->data.marked[k])
1341 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1342 sha1_to_hex(m->data.marked[k]->sha1));
1343 }
1344 }
1345 }
1346
1347 static void dump_marks(void)
1348 {
1349 static struct lock_file mark_lock;
1350 int mark_fd;
1351 FILE *f;
1352
1353 if (!mark_file)
1354 return;
1355
1356 mark_fd = hold_lock_file_for_update(&mark_lock, mark_file, 0);
1357 if (mark_fd < 0) {
1358 failure |= error("Unable to write marks file %s: %s",
1359 mark_file, strerror(errno));
1360 return;
1361 }
1362
1363 f = fdopen(mark_fd, "w");
1364 if (!f) {
1365 rollback_lock_file(&mark_lock);
1366 failure |= error("Unable to write marks file %s: %s",
1367 mark_file, strerror(errno));
1368 return;
1369 }
1370
1371 dump_marks_helper(f, 0, marks);
1372 fclose(f);
1373 if (commit_lock_file(&mark_lock))
1374 failure |= error("Unable to write marks file %s: %s",
1375 mark_file, strerror(errno));
1376 }
1377
1378 static void read_next_command(void)
1379 {
1380 read_line(&command_buf, stdin, '\n');
1381 }
1382
1383 static void cmd_mark(void)
1384 {
1385 if (!prefixcmp(command_buf.buf, "mark :")) {
1386 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1387 read_next_command();
1388 }
1389 else
1390 next_mark = 0;
1391 }
1392
1393 static void *cmd_data (size_t *size)
1394 {
1395 size_t length;
1396 char *buffer;
1397
1398 if (prefixcmp(command_buf.buf, "data "))
1399 die("Expected 'data n' command, found: %s", command_buf.buf);
1400
1401 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1402 char *term = xstrdup(command_buf.buf + 5 + 2);
1403 size_t sz = 8192, term_len = command_buf.len - 5 - 2;
1404 length = 0;
1405 buffer = xmalloc(sz);
1406 for (;;) {
1407 read_next_command();
1408 if (command_buf.eof)
1409 die("EOF in data (terminator '%s' not found)", term);
1410 if (term_len == command_buf.len
1411 && !strcmp(term, command_buf.buf))
1412 break;
1413 if (sz < (length + command_buf.len)) {
1414 sz = sz * 3 / 2 + 16;
1415 if (sz < (length + command_buf.len))
1416 sz = length + command_buf.len;
1417 buffer = xrealloc(buffer, sz);
1418 }
1419 memcpy(buffer + length,
1420 command_buf.buf,
1421 command_buf.len - 1);
1422 length += command_buf.len - 1;
1423 buffer[length++] = '\n';
1424 }
1425 free(term);
1426 }
1427 else {
1428 size_t n = 0;
1429 length = strtoul(command_buf.buf + 5, NULL, 10);
1430 buffer = xmalloc(length);
1431 while (n < length) {
1432 size_t s = fread(buffer + n, 1, length - n, stdin);
1433 if (!s && feof(stdin))
1434 die("EOF in data (%lu bytes remaining)",
1435 (unsigned long)(length - n));
1436 n += s;
1437 }
1438 }
1439
1440 if (fgetc(stdin) != '\n')
1441 die("An lf did not trail the binary data as expected.");
1442
1443 *size = length;
1444 return buffer;
1445 }
1446
1447 static int validate_raw_date(const char *src, char *result, int maxlen)
1448 {
1449 const char *orig_src = src;
1450 char *endp, sign;
1451
1452 strtoul(src, &endp, 10);
1453 if (endp == src || *endp != ' ')
1454 return -1;
1455
1456 src = endp + 1;
1457 if (*src != '-' && *src != '+')
1458 return -1;
1459 sign = *src;
1460
1461 strtoul(src + 1, &endp, 10);
1462 if (endp == src || *endp || (endp - orig_src) >= maxlen)
1463 return -1;
1464
1465 strcpy(result, orig_src);
1466 return 0;
1467 }
1468
1469 static char *parse_ident(const char *buf)
1470 {
1471 const char *gt;
1472 size_t name_len;
1473 char *ident;
1474
1475 gt = strrchr(buf, '>');
1476 if (!gt)
1477 die("Missing > in ident string: %s", buf);
1478 gt++;
1479 if (*gt != ' ')
1480 die("Missing space after > in ident string: %s", buf);
1481 gt++;
1482 name_len = gt - buf;
1483 ident = xmalloc(name_len + 24);
1484 strncpy(ident, buf, name_len);
1485
1486 switch (whenspec) {
1487 case WHENSPEC_RAW:
1488 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1489 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1490 break;
1491 case WHENSPEC_RFC2822:
1492 if (parse_date(gt, ident + name_len, 24) < 0)
1493 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1494 break;
1495 case WHENSPEC_NOW:
1496 if (strcmp("now", gt))
1497 die("Date in ident must be 'now': %s", buf);
1498 datestamp(ident + name_len, 24);
1499 break;
1500 }
1501
1502 return ident;
1503 }
1504
1505 static void cmd_new_blob(void)
1506 {
1507 size_t l;
1508 void *d;
1509
1510 read_next_command();
1511 cmd_mark();
1512 d = cmd_data(&l);
1513
1514 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1515 free(d);
1516 }
1517
1518 static void unload_one_branch(void)
1519 {
1520 while (cur_active_branches
1521 && cur_active_branches >= max_active_branches) {
1522 uintmax_t min_commit = ULONG_MAX;
1523 struct branch *e, *l = NULL, *p = NULL;
1524
1525 for (e = active_branches; e; e = e->active_next_branch) {
1526 if (e->last_commit < min_commit) {
1527 p = l;
1528 min_commit = e->last_commit;
1529 }
1530 l = e;
1531 }
1532
1533 if (p) {
1534 e = p->active_next_branch;
1535 p->active_next_branch = e->active_next_branch;
1536 } else {
1537 e = active_branches;
1538 active_branches = e->active_next_branch;
1539 }
1540 e->active = 0;
1541 e->active_next_branch = NULL;
1542 if (e->branch_tree.tree) {
1543 release_tree_content_recursive(e->branch_tree.tree);
1544 e->branch_tree.tree = NULL;
1545 }
1546 cur_active_branches--;
1547 }
1548 }
1549
1550 static void load_branch(struct branch *b)
1551 {
1552 load_tree(&b->branch_tree);
1553 if (!b->active) {
1554 b->active = 1;
1555 b->active_next_branch = active_branches;
1556 active_branches = b;
1557 cur_active_branches++;
1558 branch_load_count++;
1559 }
1560 }
1561
1562 static void file_change_m(struct branch *b)
1563 {
1564 const char *p = command_buf.buf + 2;
1565 char *p_uq;
1566 const char *endp;
1567 struct object_entry *oe = oe;
1568 unsigned char sha1[20];
1569 uint16_t mode, inline_data = 0;
1570
1571 p = get_mode(p, &mode);
1572 if (!p)
1573 die("Corrupt mode: %s", command_buf.buf);
1574 switch (mode) {
1575 case S_IFREG | 0644:
1576 case S_IFREG | 0755:
1577 case S_IFLNK:
1578 case 0644:
1579 case 0755:
1580 /* ok */
1581 break;
1582 default:
1583 die("Corrupt mode: %s", command_buf.buf);
1584 }
1585
1586 if (*p == ':') {
1587 char *x;
1588 oe = find_mark(strtoumax(p + 1, &x, 10));
1589 hashcpy(sha1, oe->sha1);
1590 p = x;
1591 } else if (!prefixcmp(p, "inline")) {
1592 inline_data = 1;
1593 p += 6;
1594 } else {
1595 if (get_sha1_hex(p, sha1))
1596 die("Invalid SHA1: %s", command_buf.buf);
1597 oe = find_object(sha1);
1598 p += 40;
1599 }
1600 if (*p++ != ' ')
1601 die("Missing space after SHA1: %s", command_buf.buf);
1602
1603 p_uq = unquote_c_style(p, &endp);
1604 if (p_uq) {
1605 if (*endp)
1606 die("Garbage after path in: %s", command_buf.buf);
1607 p = p_uq;
1608 }
1609
1610 if (inline_data) {
1611 size_t l;
1612 void *d;
1613 if (!p_uq)
1614 p = p_uq = xstrdup(p);
1615 read_next_command();
1616 d = cmd_data(&l);
1617 if (store_object(OBJ_BLOB, d, l, &last_blob, sha1, 0))
1618 free(d);
1619 } else if (oe) {
1620 if (oe->type != OBJ_BLOB)
1621 die("Not a blob (actually a %s): %s",
1622 command_buf.buf, typename(oe->type));
1623 } else {
1624 enum object_type type = sha1_object_info(sha1, NULL);
1625 if (type < 0)
1626 die("Blob not found: %s", command_buf.buf);
1627 if (type != OBJ_BLOB)
1628 die("Not a blob (actually a %s): %s",
1629 typename(type), command_buf.buf);
1630 }
1631
1632 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1633 free(p_uq);
1634 }
1635
1636 static void file_change_d(struct branch *b)
1637 {
1638 const char *p = command_buf.buf + 2;
1639 char *p_uq;
1640 const char *endp;
1641
1642 p_uq = unquote_c_style(p, &endp);
1643 if (p_uq) {
1644 if (*endp)
1645 die("Garbage after path in: %s", command_buf.buf);
1646 p = p_uq;
1647 }
1648 tree_content_remove(&b->branch_tree, p);
1649 free(p_uq);
1650 }
1651
1652 static void file_change_deleteall(struct branch *b)
1653 {
1654 release_tree_content_recursive(b->branch_tree.tree);
1655 hashclr(b->branch_tree.versions[0].sha1);
1656 hashclr(b->branch_tree.versions[1].sha1);
1657 load_tree(&b->branch_tree);
1658 }
1659
1660 static void cmd_from_commit(struct branch *b, char *buf, unsigned long size)
1661 {
1662 if (!buf || size < 46)
1663 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
1664 if (memcmp("tree ", buf, 5)
1665 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1666 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1667 hashcpy(b->branch_tree.versions[0].sha1,
1668 b->branch_tree.versions[1].sha1);
1669 }
1670
1671 static void cmd_from_existing(struct branch *b)
1672 {
1673 if (is_null_sha1(b->sha1)) {
1674 hashclr(b->branch_tree.versions[0].sha1);
1675 hashclr(b->branch_tree.versions[1].sha1);
1676 } else {
1677 unsigned long size;
1678 char *buf;
1679
1680 buf = read_object_with_reference(b->sha1,
1681 commit_type, &size, b->sha1);
1682 cmd_from_commit(b, buf, size);
1683 free(buf);
1684 }
1685 }
1686
1687 static void cmd_from(struct branch *b)
1688 {
1689 const char *from;
1690 struct branch *s;
1691
1692 if (prefixcmp(command_buf.buf, "from "))
1693 return;
1694
1695 if (b->branch_tree.tree) {
1696 release_tree_content_recursive(b->branch_tree.tree);
1697 b->branch_tree.tree = NULL;
1698 }
1699
1700 from = strchr(command_buf.buf, ' ') + 1;
1701 s = lookup_branch(from);
1702 if (b == s)
1703 die("Can't create a branch from itself: %s", b->name);
1704 else if (s) {
1705 unsigned char *t = s->branch_tree.versions[1].sha1;
1706 hashcpy(b->sha1, s->sha1);
1707 hashcpy(b->branch_tree.versions[0].sha1, t);
1708 hashcpy(b->branch_tree.versions[1].sha1, t);
1709 } else if (*from == ':') {
1710 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1711 struct object_entry *oe = find_mark(idnum);
1712 if (oe->type != OBJ_COMMIT)
1713 die("Mark :%" PRIuMAX " not a commit", idnum);
1714 hashcpy(b->sha1, oe->sha1);
1715 if (oe->pack_id != MAX_PACK_ID) {
1716 unsigned long size;
1717 char *buf = gfi_unpack_entry(oe, &size);
1718 cmd_from_commit(b, buf, size);
1719 free(buf);
1720 } else
1721 cmd_from_existing(b);
1722 } else if (!get_sha1(from, b->sha1))
1723 cmd_from_existing(b);
1724 else
1725 die("Invalid ref name or SHA1 expression: %s", from);
1726
1727 read_next_command();
1728 }
1729
1730 static struct hash_list *cmd_merge(unsigned int *count)
1731 {
1732 struct hash_list *list = NULL, *n, *e = e;
1733 const char *from;
1734 struct branch *s;
1735
1736 *count = 0;
1737 while (!prefixcmp(command_buf.buf, "merge ")) {
1738 from = strchr(command_buf.buf, ' ') + 1;
1739 n = xmalloc(sizeof(*n));
1740 s = lookup_branch(from);
1741 if (s)
1742 hashcpy(n->sha1, s->sha1);
1743 else if (*from == ':') {
1744 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1745 struct object_entry *oe = find_mark(idnum);
1746 if (oe->type != OBJ_COMMIT)
1747 die("Mark :%" PRIuMAX " not a commit", idnum);
1748 hashcpy(n->sha1, oe->sha1);
1749 } else if (!get_sha1(from, n->sha1)) {
1750 unsigned long size;
1751 char *buf = read_object_with_reference(n->sha1,
1752 commit_type, &size, n->sha1);
1753 if (!buf || size < 46)
1754 die("Not a valid commit: %s", from);
1755 free(buf);
1756 } else
1757 die("Invalid ref name or SHA1 expression: %s", from);
1758
1759 n->next = NULL;
1760 if (list)
1761 e->next = n;
1762 else
1763 list = n;
1764 e = n;
1765 (*count)++;
1766 read_next_command();
1767 }
1768 return list;
1769 }
1770
1771 static void cmd_new_commit(void)
1772 {
1773 struct branch *b;
1774 void *msg;
1775 size_t msglen;
1776 char *sp;
1777 char *author = NULL;
1778 char *committer = NULL;
1779 struct hash_list *merge_list = NULL;
1780 unsigned int merge_count;
1781
1782 /* Obtain the branch name from the rest of our command */
1783 sp = strchr(command_buf.buf, ' ') + 1;
1784 b = lookup_branch(sp);
1785 if (!b)
1786 b = new_branch(sp);
1787
1788 read_next_command();
1789 cmd_mark();
1790 if (!prefixcmp(command_buf.buf, "author ")) {
1791 author = parse_ident(command_buf.buf + 7);
1792 read_next_command();
1793 }
1794 if (!prefixcmp(command_buf.buf, "committer ")) {
1795 committer = parse_ident(command_buf.buf + 10);
1796 read_next_command();
1797 }
1798 if (!committer)
1799 die("Expected committer but didn't get one");
1800 msg = cmd_data(&msglen);
1801 read_next_command();
1802 cmd_from(b);
1803 merge_list = cmd_merge(&merge_count);
1804
1805 /* ensure the branch is active/loaded */
1806 if (!b->branch_tree.tree || !max_active_branches) {
1807 unload_one_branch();
1808 load_branch(b);
1809 }
1810
1811 /* file_change* */
1812 for (;;) {
1813 if (1 == command_buf.len)
1814 break;
1815 else if (!prefixcmp(command_buf.buf, "M "))
1816 file_change_m(b);
1817 else if (!prefixcmp(command_buf.buf, "D "))
1818 file_change_d(b);
1819 else if (!strcmp("deleteall", command_buf.buf))
1820 file_change_deleteall(b);
1821 else
1822 die("Unsupported file_change: %s", command_buf.buf);
1823 read_next_command();
1824 }
1825
1826 /* build the tree and the commit */
1827 store_tree(&b->branch_tree);
1828 hashcpy(b->branch_tree.versions[0].sha1,
1829 b->branch_tree.versions[1].sha1);
1830 size_dbuf(&new_data, 114 + msglen
1831 + merge_count * 49
1832 + (author
1833 ? strlen(author) + strlen(committer)
1834 : 2 * strlen(committer)));
1835 sp = new_data.buffer;
1836 sp += sprintf(sp, "tree %s\n",
1837 sha1_to_hex(b->branch_tree.versions[1].sha1));
1838 if (!is_null_sha1(b->sha1))
1839 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1840 while (merge_list) {
1841 struct hash_list *next = merge_list->next;
1842 sp += sprintf(sp, "parent %s\n", sha1_to_hex(merge_list->sha1));
1843 free(merge_list);
1844 merge_list = next;
1845 }
1846 sp += sprintf(sp, "author %s\n", author ? author : committer);
1847 sp += sprintf(sp, "committer %s\n", committer);
1848 *sp++ = '\n';
1849 memcpy(sp, msg, msglen);
1850 sp += msglen;
1851 free(author);
1852 free(committer);
1853 free(msg);
1854
1855 if (!store_object(OBJ_COMMIT,
1856 new_data.buffer, sp - (char*)new_data.buffer,
1857 NULL, b->sha1, next_mark))
1858 b->pack_id = pack_id;
1859 b->last_commit = object_count_by_type[OBJ_COMMIT];
1860 }
1861
1862 static void cmd_new_tag(void)
1863 {
1864 char *sp;
1865 const char *from;
1866 char *tagger;
1867 struct branch *s;
1868 void *msg;
1869 size_t msglen;
1870 struct tag *t;
1871 uintmax_t from_mark = 0;
1872 unsigned char sha1[20];
1873
1874 /* Obtain the new tag name from the rest of our command */
1875 sp = strchr(command_buf.buf, ' ') + 1;
1876 t = pool_alloc(sizeof(struct tag));
1877 t->next_tag = NULL;
1878 t->name = pool_strdup(sp);
1879 if (last_tag)
1880 last_tag->next_tag = t;
1881 else
1882 first_tag = t;
1883 last_tag = t;
1884 read_next_command();
1885
1886 /* from ... */
1887 if (prefixcmp(command_buf.buf, "from "))
1888 die("Expected from command, got %s", command_buf.buf);
1889 from = strchr(command_buf.buf, ' ') + 1;
1890 s = lookup_branch(from);
1891 if (s) {
1892 hashcpy(sha1, s->sha1);
1893 } else if (*from == ':') {
1894 struct object_entry *oe;
1895 from_mark = strtoumax(from + 1, NULL, 10);
1896 oe = find_mark(from_mark);
1897 if (oe->type != OBJ_COMMIT)
1898 die("Mark :%" PRIuMAX " not a commit", from_mark);
1899 hashcpy(sha1, oe->sha1);
1900 } else if (!get_sha1(from, sha1)) {
1901 unsigned long size;
1902 char *buf;
1903
1904 buf = read_object_with_reference(sha1,
1905 commit_type, &size, sha1);
1906 if (!buf || size < 46)
1907 die("Not a valid commit: %s", from);
1908 free(buf);
1909 } else
1910 die("Invalid ref name or SHA1 expression: %s", from);
1911 read_next_command();
1912
1913 /* tagger ... */
1914 if (prefixcmp(command_buf.buf, "tagger "))
1915 die("Expected tagger command, got %s", command_buf.buf);
1916 tagger = parse_ident(command_buf.buf + 7);
1917
1918 /* tag payload/message */
1919 read_next_command();
1920 msg = cmd_data(&msglen);
1921
1922 /* build the tag object */
1923 size_dbuf(&new_data, 67+strlen(t->name)+strlen(tagger)+msglen);
1924 sp = new_data.buffer;
1925 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
1926 sp += sprintf(sp, "type %s\n", commit_type);
1927 sp += sprintf(sp, "tag %s\n", t->name);
1928 sp += sprintf(sp, "tagger %s\n", tagger);
1929 *sp++ = '\n';
1930 memcpy(sp, msg, msglen);
1931 sp += msglen;
1932 free(tagger);
1933 free(msg);
1934
1935 if (store_object(OBJ_TAG, new_data.buffer,
1936 sp - (char*)new_data.buffer,
1937 NULL, t->sha1, 0))
1938 t->pack_id = MAX_PACK_ID;
1939 else
1940 t->pack_id = pack_id;
1941 }
1942
1943 static void cmd_reset_branch(void)
1944 {
1945 struct branch *b;
1946 char *sp;
1947
1948 /* Obtain the branch name from the rest of our command */
1949 sp = strchr(command_buf.buf, ' ') + 1;
1950 b = lookup_branch(sp);
1951 if (b) {
1952 hashclr(b->sha1);
1953 hashclr(b->branch_tree.versions[0].sha1);
1954 hashclr(b->branch_tree.versions[1].sha1);
1955 if (b->branch_tree.tree) {
1956 release_tree_content_recursive(b->branch_tree.tree);
1957 b->branch_tree.tree = NULL;
1958 }
1959 }
1960 else
1961 b = new_branch(sp);
1962 read_next_command();
1963 cmd_from(b);
1964 }
1965
1966 static void cmd_checkpoint(void)
1967 {
1968 if (object_count) {
1969 cycle_packfile();
1970 dump_branches();
1971 dump_tags();
1972 dump_marks();
1973 }
1974 read_next_command();
1975 }
1976
1977 static void import_marks(const char *input_file)
1978 {
1979 char line[512];
1980 FILE *f = fopen(input_file, "r");
1981 if (!f)
1982 die("cannot read %s: %s", input_file, strerror(errno));
1983 while (fgets(line, sizeof(line), f)) {
1984 uintmax_t mark;
1985 char *end;
1986 unsigned char sha1[20];
1987 struct object_entry *e;
1988
1989 end = strchr(line, '\n');
1990 if (line[0] != ':' || !end)
1991 die("corrupt mark line: %s", line);
1992 *end = 0;
1993 mark = strtoumax(line + 1, &end, 10);
1994 if (!mark || end == line + 1
1995 || *end != ' ' || get_sha1(end + 1, sha1))
1996 die("corrupt mark line: %s", line);
1997 e = find_object(sha1);
1998 if (!e) {
1999 enum object_type type = sha1_object_info(sha1, NULL);
2000 if (type < 0)
2001 die("object not found: %s", sha1_to_hex(sha1));
2002 e = insert_object(sha1);
2003 e->type = type;
2004 e->pack_id = MAX_PACK_ID;
2005 e->offset = 1; /* just not zero! */
2006 }
2007 insert_mark(mark, e);
2008 }
2009 fclose(f);
2010 }
2011
2012 static const char fast_import_usage[] =
2013 "git-fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2014
2015 int main(int argc, const char **argv)
2016 {
2017 int i, show_stats = 1;
2018
2019 git_config(git_default_config);
2020 alloc_objects(object_entry_alloc);
2021 strbuf_init(&command_buf);
2022 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2023 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2024 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2025 marks = pool_calloc(1, sizeof(struct mark_set));
2026
2027 for (i = 1; i < argc; i++) {
2028 const char *a = argv[i];
2029
2030 if (*a != '-' || !strcmp(a, "--"))
2031 break;
2032 else if (!prefixcmp(a, "--date-format=")) {
2033 const char *fmt = a + 14;
2034 if (!strcmp(fmt, "raw"))
2035 whenspec = WHENSPEC_RAW;
2036 else if (!strcmp(fmt, "rfc2822"))
2037 whenspec = WHENSPEC_RFC2822;
2038 else if (!strcmp(fmt, "now"))
2039 whenspec = WHENSPEC_NOW;
2040 else
2041 die("unknown --date-format argument %s", fmt);
2042 }
2043 else if (!prefixcmp(a, "--max-pack-size="))
2044 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
2045 else if (!prefixcmp(a, "--depth="))
2046 max_depth = strtoul(a + 8, NULL, 0);
2047 else if (!prefixcmp(a, "--active-branches="))
2048 max_active_branches = strtoul(a + 18, NULL, 0);
2049 else if (!prefixcmp(a, "--import-marks="))
2050 import_marks(a + 15);
2051 else if (!prefixcmp(a, "--export-marks="))
2052 mark_file = a + 15;
2053 else if (!prefixcmp(a, "--export-pack-edges=")) {
2054 if (pack_edges)
2055 fclose(pack_edges);
2056 pack_edges = fopen(a + 20, "a");
2057 if (!pack_edges)
2058 die("Cannot open %s: %s", a + 20, strerror(errno));
2059 } else if (!strcmp(a, "--force"))
2060 force_update = 1;
2061 else if (!strcmp(a, "--quiet"))
2062 show_stats = 0;
2063 else if (!strcmp(a, "--stats"))
2064 show_stats = 1;
2065 else
2066 die("unknown option %s", a);
2067 }
2068 if (i != argc)
2069 usage(fast_import_usage);
2070
2071 prepare_packed_git();
2072 start_packfile();
2073 for (;;) {
2074 read_next_command();
2075 if (command_buf.eof)
2076 break;
2077 else if (!strcmp("blob", command_buf.buf))
2078 cmd_new_blob();
2079 else if (!prefixcmp(command_buf.buf, "commit "))
2080 cmd_new_commit();
2081 else if (!prefixcmp(command_buf.buf, "tag "))
2082 cmd_new_tag();
2083 else if (!prefixcmp(command_buf.buf, "reset "))
2084 cmd_reset_branch();
2085 else if (!strcmp("checkpoint", command_buf.buf))
2086 cmd_checkpoint();
2087 else
2088 die("Unsupported command: %s", command_buf.buf);
2089 }
2090 end_packfile();
2091
2092 dump_branches();
2093 dump_tags();
2094 unkeep_all_packs();
2095 dump_marks();
2096
2097 if (pack_edges)
2098 fclose(pack_edges);
2099
2100 if (show_stats) {
2101 uintmax_t total_count = 0, duplicate_count = 0;
2102 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2103 total_count += object_count_by_type[i];
2104 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2105 duplicate_count += duplicate_count_by_type[i];
2106
2107 fprintf(stderr, "%s statistics:\n", argv[0]);
2108 fprintf(stderr, "---------------------------------------------------------------------\n");
2109 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2110 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2111 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
2112 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
2113 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
2114 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
2115 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2116 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2117 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2118 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2119 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2120 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2121 fprintf(stderr, "---------------------------------------------------------------------\n");
2122 pack_report();
2123 fprintf(stderr, "---------------------------------------------------------------------\n");
2124 fprintf(stderr, "\n");
2125 }
2126
2127 return failure ? 1 : 0;
2128 }
Gitweb development nest