Optimize index creation on large object sets in fast-import.
[git/git.git] / fast-import.c
CommitLineData
463acbe1
SP
1/*
2Format of STDIN stream:
3
4 stream ::= cmd*;
5
6 cmd ::= new_blob
c44cdc7e 7 | new_commit
463acbe1 8 | new_tag
5fced8dc 9 | reset_branch
463acbe1
SP
10 ;
11
c44cdc7e
SP
12 new_blob ::= 'blob' lf
13 mark?
14 file_content;
15 file_content ::= data;
463acbe1 16
c44cdc7e 17 new_commit ::= 'commit' sp ref_str lf
00e2b884
SP
18 mark?
19 ('author' sp name '<' email '>' ts tz lf)?
20 'committer' sp name '<' email '>' ts tz lf
21 commit_msg
02f3389d 22 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
62b6f483 23 ('merge' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)*
c44cdc7e
SP
24 file_change*
25 lf;
26 commit_msg ::= data;
463acbe1 27
c44cdc7e
SP
28 file_change ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf
29 | 'D' sp path_str lf
463acbe1 30 ;
c44cdc7e
SP
31 mode ::= '644' | '755';
32
33 new_tag ::= 'tag' sp tag_str lf
34 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
35 'tagger' sp name '<' email '>' ts tz lf
36 tag_msg;
37 tag_msg ::= data;
38
9938ffc5
SP
39 reset_branch ::= 'reset' sp ref_str lf
40 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
41 lf;
5fced8dc 42
7bfe6e26
SP
43 checkpoint ::= 'checkpoint' lf
44 lf;
45
c44cdc7e
SP
46 # note: the first idnum in a stream should be 1 and subsequent
47 # idnums should not have gaps between values as this will cause
48 # the stream parser to reserve space for the gapped values. An
49 # idnum can be updated in the future to a new object by issuing
50 # a new mark directive with the old idnum.
51 #
52 mark ::= 'mark' sp idnum lf;
53
54 # note: declen indicates the length of binary_data in bytes.
55 # declen does not include the lf preceeding or trailing the
56 # binary data.
57 #
58 data ::= 'data' sp declen lf
59 binary_data
60 lf;
61
62 # note: quoted strings are C-style quoting supporting \c for
63 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
64 # is the signed byte value in octal. Note that the only
65 # characters which must actually be escaped to protect the
66 # stream formatting is: \, " and LF. Otherwise these values
67 # are UTF8.
68 #
69 ref_str ::= ref | '"' quoted(ref) '"' ;
70 sha1exp_str ::= sha1exp | '"' quoted(sha1exp) '"' ;
71 tag_str ::= tag | '"' quoted(tag) '"' ;
72 path_str ::= path | '"' quoted(path) '"' ;
73
74 declen ::= # unsigned 32 bit value, ascii base10 notation;
463acbe1 75 binary_data ::= # file content, not interpreted;
c44cdc7e 76
463acbe1
SP
77 sp ::= # ASCII space character;
78 lf ::= # ASCII newline (LF) character;
c44cdc7e
SP
79
80 # note: a colon (':') must precede the numerical value assigned to
81 # an idnum. This is to distinguish it from a ref or tag name as
82 # GIT does not permit ':' in ref or tag strings.
83 #
84 idnum ::= ':' declen;
85 path ::= # GIT style file path, e.g. "a/b/c";
86 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
87 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
463acbe1
SP
88 sha1exp ::= # Any valid GIT SHA1 expression;
89 hexsha1 ::= # SHA1 in hexadecimal format;
c44cdc7e
SP
90
91 # note: name and email are UTF8 strings, however name must not
92 # contain '<' or lf and email must not contain any of the
93 # following: '<', '>', lf.
94 #
95 name ::= # valid GIT author/committer name;
463acbe1 96 email ::= # valid GIT author/committer email;
c44cdc7e
SP
97 ts ::= # time since the epoch in seconds, ascii base10 notation;
98 tz ::= # GIT style timezone;
463acbe1
SP
99*/
100
db5e523f
SP
101#include "builtin.h"
102#include "cache.h"
103#include "object.h"
104#include "blob.h"
463acbe1 105#include "tree.h"
db5e523f
SP
106#include "delta.h"
107#include "pack.h"
463acbe1 108#include "refs.h"
db5e523f 109#include "csum-file.h"
c44cdc7e
SP
110#include "strbuf.h"
111#include "quote.h"
db5e523f 112
27d6d290
SP
113struct object_entry
114{
115 struct object_entry *next;
116 unsigned long offset;
03842d8e 117 unsigned type : TYPE_BITS;
7bfe6e26 118 unsigned pack_id : 16;
27d6d290
SP
119 unsigned char sha1[20];
120};
121
463acbe1 122struct object_entry_pool
27d6d290 123{
463acbe1 124 struct object_entry_pool *next_pool;
27d6d290
SP
125 struct object_entry *next_free;
126 struct object_entry *end;
ac47a738 127 struct object_entry entries[FLEX_ARRAY]; /* more */
27d6d290
SP
128};
129
d8397168
SP
130struct mark_set
131{
132 int shift;
133 union {
134 struct object_entry *marked[1024];
135 struct mark_set *sets[1024];
136 } data;
137};
138
ac47a738
SP
139struct last_object
140{
141 void *data;
4cabf858 142 unsigned long len;
d489bc14 143 unsigned long offset;
6bb5b329 144 unsigned int depth;
d489bc14 145 unsigned no_free:1;
ac47a738
SP
146};
147
463acbe1
SP
148struct mem_pool
149{
150 struct mem_pool *next_pool;
151 char *next_free;
152 char *end;
153 char space[FLEX_ARRAY]; /* more */
154};
155
156struct atom_str
157{
158 struct atom_str *next_atom;
159 int str_len;
160 char str_dat[FLEX_ARRAY]; /* more */
161};
162
163struct tree_content;
6bb5b329
SP
164struct tree_entry
165{
463acbe1
SP
166 struct tree_content *tree;
167 struct atom_str* name;
4cabf858
SP
168 struct tree_entry_ms
169 {
170 unsigned int mode;
171 unsigned char sha1[20];
172 } versions[2];
6bb5b329
SP
173};
174
463acbe1 175struct tree_content
6bb5b329 176{
463acbe1
SP
177 unsigned int entry_capacity; /* must match avail_tree_content */
178 unsigned int entry_count;
4cabf858 179 unsigned int delta_depth;
463acbe1
SP
180 struct tree_entry *entries[FLEX_ARRAY]; /* more */
181};
182
183struct avail_tree_content
184{
185 unsigned int entry_capacity; /* must match tree_content */
186 struct avail_tree_content *next_avail;
6bb5b329
SP
187};
188
189struct branch
190{
463acbe1
SP
191 struct branch *table_next_branch;
192 struct branch *active_next_branch;
6bb5b329 193 const char *name;
463acbe1
SP
194 unsigned long last_commit;
195 struct tree_entry branch_tree;
196 unsigned char sha1[20];
6bb5b329
SP
197};
198
72303d44
SP
199struct tag
200{
201 struct tag *next_tag;
202 const char *name;
203 unsigned char sha1[20];
204};
205
e2eb469d
SP
206struct dbuf
207{
208 void *buffer;
209 size_t capacity;
210};
211
62b6f483
SP
212struct hash_list
213{
214 struct hash_list *next;
215 unsigned char sha1[20];
216};
463acbe1
SP
217
218/* Stats and misc. counters */
d5c57b28 219static unsigned long max_depth = 10;
27d6d290 220static unsigned long alloc_count;
6bb5b329 221static unsigned long branch_count;
d6c7eb2c 222static unsigned long branch_load_count;
db5e523f 223static unsigned long object_count;
d8397168 224static unsigned long marks_set_count;
03842d8e
SP
225static unsigned long object_count_by_type[1 << TYPE_BITS];
226static unsigned long duplicate_count_by_type[1 << TYPE_BITS];
227static unsigned long delta_count_by_type[1 << TYPE_BITS];
ac47a738 228
463acbe1
SP
229/* Memory pools */
230static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
231static size_t total_allocd;
232static struct mem_pool *mem_pool;
233
c44cdc7e 234/* Atom management */
463acbe1
SP
235static unsigned int atom_table_sz = 4451;
236static unsigned int atom_cnt;
237static struct atom_str **atom_table;
238
239/* The .pack file being generated */
f70b6534 240static const char *base_name;
7bfe6e26 241static unsigned int pack_id;
f70b6534 242static char *idx_name;
d489bc14 243static struct packed_git *pack_data;
7bfe6e26 244static struct packed_git **all_packs;
ac47a738 245static int pack_fd;
41e5257f 246static unsigned long pack_size;
ac47a738
SP
247static unsigned char pack_sha1[20];
248
249/* Table of objects we've written. */
4cabf858 250static unsigned int object_entry_alloc = 5000;
463acbe1
SP
251static struct object_entry_pool *blocks;
252static struct object_entry *object_table[1 << 16];
d8397168 253static struct mark_set *marks;
a6a1a831 254static const char* mark_file;
ac47a738
SP
255
256/* Our last blob */
463acbe1
SP
257static struct last_object last_blob;
258
259/* Tree management */
260static unsigned int tree_entry_alloc = 1000;
261static void *avail_tree_entry;
262static unsigned int avail_tree_table_sz = 100;
263static struct avail_tree_content **avail_tree_table;
e2eb469d
SP
264static struct dbuf old_tree;
265static struct dbuf new_tree;
8bcce301 266
6bb5b329 267/* Branch data */
d5c57b28
SP
268static unsigned long max_active_branches = 5;
269static unsigned long cur_active_branches;
270static unsigned long branch_table_sz = 1039;
463acbe1
SP
271static struct branch **branch_table;
272static struct branch *active_branches;
273
72303d44
SP
274/* Tag data */
275static struct tag *first_tag;
276static struct tag *last_tag;
277
c44cdc7e
SP
278/* Input stream parsing */
279static struct strbuf command_buf;
d8397168 280static unsigned long next_mark;
243f801d 281static struct dbuf new_data;
264244a0 282static FILE* branch_log;
c44cdc7e 283
6bb5b329 284
03842d8e 285static void alloc_objects(unsigned int cnt)
8bcce301 286{
463acbe1 287 struct object_entry_pool *b;
27d6d290 288
463acbe1 289 b = xmalloc(sizeof(struct object_entry_pool)
27d6d290 290 + cnt * sizeof(struct object_entry));
463acbe1 291 b->next_pool = blocks;
27d6d290
SP
292 b->next_free = b->entries;
293 b->end = b->entries + cnt;
294 blocks = b;
295 alloc_count += cnt;
296}
8bcce301 297
27d6d290 298static struct object_entry* new_object(unsigned char *sha1)
8bcce301 299{
27d6d290 300 struct object_entry *e;
8bcce301 301
27d6d290 302 if (blocks->next_free == blocks->end)
463acbe1 303 alloc_objects(object_entry_alloc);
8bcce301 304
27d6d290 305 e = blocks->next_free++;
7bfe6e26 306 e->pack_id = pack_id;
445b8599 307 hashcpy(e->sha1, sha1);
27d6d290 308 return e;
8bcce301
SP
309}
310
463acbe1
SP
311static struct object_entry* find_object(unsigned char *sha1)
312{
313 unsigned int h = sha1[0] << 8 | sha1[1];
314 struct object_entry *e;
315 for (e = object_table[h]; e; e = e->next)
445b8599 316 if (!hashcmp(sha1, e->sha1))
463acbe1
SP
317 return e;
318 return NULL;
319}
320
8bcce301
SP
321static struct object_entry* insert_object(unsigned char *sha1)
322{
323 unsigned int h = sha1[0] << 8 | sha1[1];
27d6d290 324 struct object_entry *e = object_table[h];
463acbe1 325 struct object_entry *p = NULL;
8bcce301
SP
326
327 while (e) {
445b8599 328 if (!hashcmp(sha1, e->sha1))
8bcce301
SP
329 return e;
330 p = e;
331 e = e->next;
332 }
333
334 e = new_object(sha1);
463acbe1 335 e->next = NULL;
8bcce301
SP
336 e->offset = 0;
337 if (p)
338 p->next = e;
339 else
27d6d290 340 object_table[h] = e;
8bcce301
SP
341 return e;
342}
db5e523f 343
463acbe1
SP
344static unsigned int hc_str(const char *s, size_t len)
345{
346 unsigned int r = 0;
347 while (len-- > 0)
348 r = r * 31 + *s++;
349 return r;
350}
351
352static void* pool_alloc(size_t len)
353{
354 struct mem_pool *p;
355 void *r;
356
357 for (p = mem_pool; p; p = p->next_pool)
358 if ((p->end - p->next_free >= len))
359 break;
360
361 if (!p) {
362 if (len >= (mem_pool_alloc/2)) {
363 total_allocd += len;
364 return xmalloc(len);
365 }
366 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
367 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
368 p->next_pool = mem_pool;
369 p->next_free = p->space;
370 p->end = p->next_free + mem_pool_alloc;
371 mem_pool = p;
372 }
373
374 r = p->next_free;
8d8928b0
SP
375 /* round out to a pointer alignment */
376 if (len & (sizeof(void*) - 1))
377 len += sizeof(void*) - (len & (sizeof(void*) - 1));
463acbe1
SP
378 p->next_free += len;
379 return r;
380}
381
382static void* pool_calloc(size_t count, size_t size)
383{
384 size_t len = count * size;
385 void *r = pool_alloc(len);
386 memset(r, 0, len);
387 return r;
388}
389
390static char* pool_strdup(const char *s)
391{
392 char *r = pool_alloc(strlen(s) + 1);
393 strcpy(r, s);
394 return r;
395}
396
243f801d
SP
397static void size_dbuf(struct dbuf *b, size_t maxlen)
398{
399 if (b->buffer) {
400 if (b->capacity >= maxlen)
401 return;
402 free(b->buffer);
403 }
404 b->capacity = ((maxlen / 1024) + 1) * 1024;
405 b->buffer = xmalloc(b->capacity);
406}
407
d8397168
SP
408static void insert_mark(unsigned long idnum, struct object_entry *oe)
409{
410 struct mark_set *s = marks;
411 while ((idnum >> s->shift) >= 1024) {
412 s = pool_calloc(1, sizeof(struct mark_set));
413 s->shift = marks->shift + 10;
414 s->data.sets[0] = marks;
415 marks = s;
416 }
417 while (s->shift) {
418 unsigned long i = idnum >> s->shift;
419 idnum -= i << s->shift;
420 if (!s->data.sets[i]) {
421 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
422 s->data.sets[i]->shift = s->shift - 10;
423 }
424 s = s->data.sets[i];
425 }
426 if (!s->data.marked[idnum])
427 marks_set_count++;
428 s->data.marked[idnum] = oe;
429}
430
431static struct object_entry* find_mark(unsigned long idnum)
432{
433 unsigned long orig_idnum = idnum;
434 struct mark_set *s = marks;
435 struct object_entry *oe = NULL;
436 if ((idnum >> s->shift) < 1024) {
437 while (s && s->shift) {
438 unsigned long i = idnum >> s->shift;
439 idnum -= i << s->shift;
440 s = s->data.sets[i];
441 }
442 if (s)
443 oe = s->data.marked[idnum];
444 }
445 if (!oe)
446 die("mark :%lu not declared", orig_idnum);
447 return oe;
448}
449
463acbe1
SP
450static struct atom_str* to_atom(const char *s, size_t len)
451{
452 unsigned int hc = hc_str(s, len) % atom_table_sz;
453 struct atom_str *c;
454
455 for (c = atom_table[hc]; c; c = c->next_atom)
456 if (c->str_len == len && !strncmp(s, c->str_dat, len))
457 return c;
458
459 c = pool_alloc(sizeof(struct atom_str) + len + 1);
460 c->str_len = len;
461 strncpy(c->str_dat, s, len);
462 c->str_dat[len] = 0;
463 c->next_atom = atom_table[hc];
464 atom_table[hc] = c;
465 atom_cnt++;
466 return c;
467}
468
469static struct branch* lookup_branch(const char *name)
470{
471 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
472 struct branch *b;
473
474 for (b = branch_table[hc]; b; b = b->table_next_branch)
475 if (!strcmp(name, b->name))
476 return b;
477 return NULL;
478}
479
480static struct branch* new_branch(const char *name)
481{
482 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
483 struct branch* b = lookup_branch(name);
484
485 if (b)
486 die("Invalid attempt to create duplicate branch: %s", name);
c44cdc7e
SP
487 if (check_ref_format(name))
488 die("Branch name doesn't conform to GIT standards: %s", name);
463acbe1
SP
489
490 b = pool_calloc(1, sizeof(struct branch));
491 b->name = pool_strdup(name);
492 b->table_next_branch = branch_table[hc];
8a8c55ea
SP
493 b->branch_tree.versions[0].mode = S_IFDIR;
494 b->branch_tree.versions[1].mode = S_IFDIR;
463acbe1
SP
495 branch_table[hc] = b;
496 branch_count++;
497 return b;
498}
499
500static unsigned int hc_entries(unsigned int cnt)
501{
502 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
503 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
504}
505
506static struct tree_content* new_tree_content(unsigned int cnt)
507{
508 struct avail_tree_content *f, *l = NULL;
509 struct tree_content *t;
510 unsigned int hc = hc_entries(cnt);
511
512 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
513 if (f->entry_capacity >= cnt)
514 break;
515
516 if (f) {
517 if (l)
518 l->next_avail = f->next_avail;
519 else
520 avail_tree_table[hc] = f->next_avail;
521 } else {
522 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
523 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
524 f->entry_capacity = cnt;
525 }
526
527 t = (struct tree_content*)f;
528 t->entry_count = 0;
4cabf858 529 t->delta_depth = 0;
463acbe1
SP
530 return t;
531}
532
533static void release_tree_entry(struct tree_entry *e);
534static void release_tree_content(struct tree_content *t)
535{
536 struct avail_tree_content *f = (struct avail_tree_content*)t;
537 unsigned int hc = hc_entries(f->entry_capacity);
afde8dd9
SP
538 f->next_avail = avail_tree_table[hc];
539 avail_tree_table[hc] = f;
540}
541
542static void release_tree_content_recursive(struct tree_content *t)
543{
463acbe1
SP
544 unsigned int i;
545 for (i = 0; i < t->entry_count; i++)
546 release_tree_entry(t->entries[i]);
afde8dd9 547 release_tree_content(t);
463acbe1
SP
548}
549
550static struct tree_content* grow_tree_content(
551 struct tree_content *t,
552 int amt)
553{
554 struct tree_content *r = new_tree_content(t->entry_count + amt);
555 r->entry_count = t->entry_count;
4cabf858 556 r->delta_depth = t->delta_depth;
463acbe1
SP
557 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
558 release_tree_content(t);
559 return r;
560}
561
562static struct tree_entry* new_tree_entry()
563{
564 struct tree_entry *e;
565
566 if (!avail_tree_entry) {
567 unsigned int n = tree_entry_alloc;
8435a9cb 568 total_allocd += n * sizeof(struct tree_entry);
463acbe1 569 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
2eb26d84 570 while (n-- > 1) {
463acbe1
SP
571 *((void**)e) = e + 1;
572 e++;
573 }
35ef237c 574 *((void**)e) = NULL;
463acbe1
SP
575 }
576
577 e = avail_tree_entry;
578 avail_tree_entry = *((void**)e);
579 return e;
580}
581
582static void release_tree_entry(struct tree_entry *e)
583{
584 if (e->tree)
afde8dd9 585 release_tree_content_recursive(e->tree);
463acbe1
SP
586 *((void**)e) = avail_tree_entry;
587 avail_tree_entry = e;
588}
589
590static void yread(int fd, void *buffer, size_t length)
db5e523f
SP
591{
592 ssize_t ret = 0;
593 while (ret < length) {
594 ssize_t size = xread(fd, (char *) buffer + ret, length - ret);
463acbe1
SP
595 if (!size)
596 die("Read from descriptor %i: end of stream", fd);
597 if (size < 0)
598 die("Read from descriptor %i: %s", fd, strerror(errno));
599 ret += size;
600 }
601}
602
f70b6534
SP
603static void start_packfile()
604{
7bfe6e26 605 struct packed_git *p;
f70b6534
SP
606 struct pack_header hdr;
607
f70b6534 608 idx_name = xmalloc(strlen(base_name) + 11);
7bfe6e26
SP
609 p = xcalloc(1, sizeof(*p) + strlen(base_name) + 13);
610 sprintf(p->pack_name, "%s%5.5i.pack", base_name, pack_id + 1);
611 sprintf(idx_name, "%s%5.5i.idx", base_name, pack_id + 1);
f70b6534 612
7bfe6e26 613 pack_fd = open(p->pack_name, O_RDWR|O_CREAT|O_EXCL, 0666);
f70b6534 614 if (pack_fd < 0)
7bfe6e26
SP
615 die("Can't create %s: %s", p->pack_name, strerror(errno));
616 p->pack_fd = pack_fd;
f70b6534
SP
617
618 hdr.hdr_signature = htonl(PACK_SIGNATURE);
619 hdr.hdr_version = htonl(2);
620 hdr.hdr_entries = 0;
f70b6534 621 write_or_die(pack_fd, &hdr, sizeof(hdr));
7bfe6e26
SP
622
623 pack_data = p;
f70b6534
SP
624 pack_size = sizeof(hdr);
625 object_count = 0;
7bfe6e26
SP
626
627 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
628 all_packs[pack_id] = p;
f70b6534
SP
629}
630
631static void fixup_header_footer()
632{
633 SHA_CTX c;
634 char hdr[8];
635 unsigned long cnt;
636 char *buf;
637
638 if (lseek(pack_fd, 0, SEEK_SET) != 0)
639 die("Failed seeking to start: %s", strerror(errno));
640
641 SHA1_Init(&c);
642 yread(pack_fd, hdr, 8);
643 SHA1_Update(&c, hdr, 8);
644
645 cnt = htonl(object_count);
646 SHA1_Update(&c, &cnt, 4);
647 write_or_die(pack_fd, &cnt, 4);
648
649 buf = xmalloc(128 * 1024);
650 for (;;) {
651 size_t n = xread(pack_fd, buf, 128 * 1024);
652 if (n <= 0)
653 break;
654 SHA1_Update(&c, buf, n);
655 }
656 free(buf);
657
658 SHA1_Final(pack_sha1, &c);
659 write_or_die(pack_fd, pack_sha1, sizeof(pack_sha1));
660}
661
662static int oecmp (const void *a_, const void *b_)
663{
664 struct object_entry *a = *((struct object_entry**)a_);
665 struct object_entry *b = *((struct object_entry**)b_);
666 return hashcmp(a->sha1, b->sha1);
667}
668
669static void write_index(const char *idx_name)
670{
671 struct sha1file *f;
672 struct object_entry **idx, **c, **last, *e;
673 struct object_entry_pool *o;
674 unsigned int array[256];
675 int i;
676
677 /* Build the sorted table of object IDs. */
678 idx = xmalloc(object_count * sizeof(struct object_entry*));
679 c = idx;
680 for (o = blocks; o; o = o->next_pool)
2fce1f3c
SP
681 for (e = o->next_free; e-- != o->entries;) {
682 if (pack_id != e->pack_id)
683 goto sort_index;
684 *c++ = e;
685 }
686sort_index:
f70b6534 687 last = idx + object_count;
2fce1f3c
SP
688 if (c != last)
689 die("internal consistency error creating the index");
f70b6534
SP
690 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
691
692 /* Generate the fan-out array. */
693 c = idx;
694 for (i = 0; i < 256; i++) {
695 struct object_entry **next = c;;
696 while (next < last) {
697 if ((*next)->sha1[0] != i)
698 break;
699 next++;
700 }
701 array[i] = htonl(next - idx);
702 c = next;
703 }
704
705 f = sha1create("%s", idx_name);
706 sha1write(f, array, 256 * sizeof(int));
707 for (c = idx; c != last; c++) {
708 unsigned int offset = htonl((*c)->offset);
709 sha1write(f, &offset, 4);
710 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
711 }
712 sha1write(f, pack_sha1, sizeof(pack_sha1));
713 sha1close(f, NULL, 1);
714 free(idx);
715}
716
717static void end_packfile()
718{
7bfe6e26
SP
719 struct packed_git *old_p = pack_data, *new_p;
720
3e005baf
SP
721 if (object_count) {
722 fixup_header_footer();
723 write_index(idx_name);
724
725 /* Register the packfile with core git's machinary. */
726 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
727 if (!new_p)
728 die("core git rejected index %s", idx_name);
729 new_p->windows = old_p->windows;
730 new_p->pack_fd = old_p->pack_fd;
731 all_packs[pack_id++] = new_p;
732 install_packed_git(new_p);
733 }
734 else {
735 close(pack_fd);
736 unlink(old_p->pack_name);
737 }
7bfe6e26 738 free(old_p);
f70b6534 739 free(idx_name);
7bfe6e26
SP
740
741 /* We can't carry a delta across packfiles. */
742 free(last_blob.data);
743 last_blob.data = NULL;
744 last_blob.len = 0;
745 last_blob.offset = 0;
746 last_blob.depth = 0;
f70b6534
SP
747}
748
c44cdc7e 749static size_t encode_header(
ac47a738 750 enum object_type type,
c44cdc7e 751 size_t size,
ac47a738 752 unsigned char *hdr)
db5e523f
SP
753{
754 int n = 1;
755 unsigned char c;
756
1fcdd62a 757 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
db5e523f
SP
758 die("bad type %d", type);
759
760 c = (type << 4) | (size & 15);
761 size >>= 4;
762 while (size) {
763 *hdr++ = c | 0x80;
764 c = size & 0x7f;
765 size >>= 7;
766 n++;
767 }
768 *hdr = c;
769 return n;
770}
771
ac47a738
SP
772static int store_object(
773 enum object_type type,
774 void *dat,
c44cdc7e 775 size_t datlen,
6bb5b329 776 struct last_object *last,
d8397168
SP
777 unsigned char *sha1out,
778 unsigned long mark)
db5e523f 779{
db5e523f 780 void *out, *delta;
ac47a738
SP
781 struct object_entry *e;
782 unsigned char hdr[96];
783 unsigned char sha1[20];
db5e523f 784 unsigned long hdrlen, deltalen;
ac47a738
SP
785 SHA_CTX c;
786 z_stream s;
787
788 hdrlen = sprintf((char*)hdr,"%s %lu",type_names[type],datlen) + 1;
789 SHA1_Init(&c);
790 SHA1_Update(&c, hdr, hdrlen);
791 SHA1_Update(&c, dat, datlen);
792 SHA1_Final(sha1, &c);
6bb5b329 793 if (sha1out)
445b8599 794 hashcpy(sha1out, sha1);
ac47a738
SP
795
796 e = insert_object(sha1);
d8397168
SP
797 if (mark)
798 insert_mark(mark, e);
ac47a738 799 if (e->offset) {
6143f064 800 duplicate_count_by_type[type]++;
463acbe1 801 return 1;
ac47a738 802 }
7111feed 803 e->type = type;
41e5257f 804 e->offset = pack_size;
ac47a738 805 object_count++;
6143f064 806 object_count_by_type[type]++;
db5e523f 807
463acbe1 808 if (last && last->data && last->depth < max_depth)
ac47a738 809 delta = diff_delta(last->data, last->len,
db5e523f
SP
810 dat, datlen,
811 &deltalen, 0);
ac47a738 812 else
db5e523f
SP
813 delta = 0;
814
815 memset(&s, 0, sizeof(s));
816 deflateInit(&s, zlib_compression_level);
817
818 if (delta) {
d489bc14
SP
819 unsigned long ofs = e->offset - last->offset;
820 unsigned pos = sizeof(hdr) - 1;
821
4cabf858 822 delta_count_by_type[type]++;
ac47a738 823 last->depth++;
db5e523f
SP
824 s.next_in = delta;
825 s.avail_in = deltalen;
d489bc14
SP
826
827 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
23bc886c 828 write_or_die(pack_fd, hdr, hdrlen);
d489bc14
SP
829 pack_size += hdrlen;
830
831 hdr[pos] = ofs & 127;
832 while (ofs >>= 7)
833 hdr[--pos] = 128 | (--ofs & 127);
834 write_or_die(pack_fd, hdr + pos, sizeof(hdr) - pos);
835 pack_size += sizeof(hdr) - pos;
db5e523f 836 } else {
463acbe1
SP
837 if (last)
838 last->depth = 0;
db5e523f
SP
839 s.next_in = dat;
840 s.avail_in = datlen;
ac47a738 841 hdrlen = encode_header(type, datlen, hdr);
23bc886c 842 write_or_die(pack_fd, hdr, hdrlen);
41e5257f 843 pack_size += hdrlen;
db5e523f
SP
844 }
845
846 s.avail_out = deflateBound(&s, s.avail_in);
847 s.next_out = out = xmalloc(s.avail_out);
848 while (deflate(&s, Z_FINISH) == Z_OK)
849 /* nothing */;
850 deflateEnd(&s);
851
23bc886c 852 write_or_die(pack_fd, out, s.total_out);
41e5257f 853 pack_size += s.total_out;
db5e523f
SP
854
855 free(out);
856 if (delta)
857 free(delta);
463acbe1 858 if (last) {
e2eb469d 859 if (last->data && !last->no_free)
463acbe1
SP
860 free(last->data);
861 last->data = dat;
d489bc14 862 last->offset = e->offset;
463acbe1 863 last->len = datlen;
463acbe1
SP
864 }
865 return 0;
866}
867
7bfe6e26
SP
868static void *gfi_unpack_entry(
869 struct object_entry *oe,
870 unsigned long *sizep)
41e5257f 871{
7bfe6e26
SP
872 static char type[20];
873 struct packed_git *p = all_packs[oe->pack_id];
874 if (p == pack_data)
875 p->pack_size = pack_size + 20;
876 return unpack_entry(p, oe->offset, type, sizep);
41e5257f
SP
877}
878
463acbe1
SP
879static const char *get_mode(const char *str, unsigned int *modep)
880{
881 unsigned char c;
882 unsigned int mode = 0;
883
884 while ((c = *str++) != ' ') {
885 if (c < '0' || c > '7')
886 return NULL;
887 mode = (mode << 3) + (c - '0');
888 }
889 *modep = mode;
890 return str;
891}
892
893static void load_tree(struct tree_entry *root)
894{
4cabf858 895 unsigned char* sha1 = root->versions[1].sha1;
463acbe1
SP
896 struct object_entry *myoe;
897 struct tree_content *t;
898 unsigned long size;
899 char *buf;
900 const char *c;
463acbe1
SP
901
902 root->tree = t = new_tree_content(8);
4cabf858 903 if (is_null_sha1(sha1))
463acbe1
SP
904 return;
905
4cabf858 906 myoe = find_object(sha1);
463acbe1 907 if (myoe) {
41e5257f 908 if (myoe->type != OBJ_TREE)
4cabf858 909 die("Not a tree: %s", sha1_to_hex(sha1));
d489bc14 910 t->delta_depth = 0;
7bfe6e26 911 buf = gfi_unpack_entry(myoe, &size);
463acbe1 912 } else {
41e5257f 913 char type[20];
4cabf858 914 buf = read_sha1_file(sha1, type, &size);
00e2b884 915 if (!buf || strcmp(type, tree_type))
4cabf858 916 die("Can't load tree %s", sha1_to_hex(sha1));
463acbe1
SP
917 }
918
919 c = buf;
920 while (c != (buf + size)) {
921 struct tree_entry *e = new_tree_entry();
922
923 if (t->entry_count == t->entry_capacity)
924 root->tree = t = grow_tree_content(t, 8);
925 t->entries[t->entry_count++] = e;
926
927 e->tree = NULL;
4cabf858 928 c = get_mode(c, &e->versions[1].mode);
463acbe1 929 if (!c)
4cabf858
SP
930 die("Corrupt mode in %s", sha1_to_hex(sha1));
931 e->versions[0].mode = e->versions[1].mode;
463acbe1
SP
932 e->name = to_atom(c, strlen(c));
933 c += e->name->str_len + 1;
4cabf858
SP
934 hashcpy(e->versions[0].sha1, (unsigned char*)c);
935 hashcpy(e->versions[1].sha1, (unsigned char*)c);
463acbe1
SP
936 c += 20;
937 }
938 free(buf);
939}
940
4cabf858 941static int tecmp0 (const void *_a, const void *_b)
463acbe1
SP
942{
943 struct tree_entry *a = *((struct tree_entry**)_a);
944 struct tree_entry *b = *((struct tree_entry**)_b);
945 return base_name_compare(
4cabf858
SP
946 a->name->str_dat, a->name->str_len, a->versions[0].mode,
947 b->name->str_dat, b->name->str_len, b->versions[0].mode);
463acbe1
SP
948}
949
4cabf858 950static int tecmp1 (const void *_a, const void *_b)
463acbe1 951{
4cabf858
SP
952 struct tree_entry *a = *((struct tree_entry**)_a);
953 struct tree_entry *b = *((struct tree_entry**)_b);
954 return base_name_compare(
955 a->name->str_dat, a->name->str_len, a->versions[1].mode,
956 b->name->str_dat, b->name->str_len, b->versions[1].mode);
957}
958
e2eb469d
SP
959static void mktree(struct tree_content *t,
960 int v,
961 unsigned long *szp,
962 struct dbuf *b)
4cabf858
SP
963{
964 size_t maxlen = 0;
463acbe1 965 unsigned int i;
e2eb469d 966 char *c;
463acbe1 967
4cabf858
SP
968 if (!v)
969 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
970 else
971 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
463acbe1 972
463acbe1 973 for (i = 0; i < t->entry_count; i++) {
4cabf858
SP
974 if (t->entries[i]->versions[v].mode)
975 maxlen += t->entries[i]->name->str_len + 34;
463acbe1
SP
976 }
977
243f801d 978 size_dbuf(b, maxlen);
e2eb469d 979 c = b->buffer;
463acbe1
SP
980 for (i = 0; i < t->entry_count; i++) {
981 struct tree_entry *e = t->entries[i];
4cabf858
SP
982 if (!e->versions[v].mode)
983 continue;
984 c += sprintf(c, "%o", e->versions[v].mode);
463acbe1
SP
985 *c++ = ' ';
986 strcpy(c, e->name->str_dat);
987 c += e->name->str_len + 1;
4cabf858 988 hashcpy((unsigned char*)c, e->versions[v].sha1);
463acbe1
SP
989 c += 20;
990 }
e2eb469d 991 *szp = c - (char*)b->buffer;
4cabf858
SP
992}
993
994static void store_tree(struct tree_entry *root)
995{
996 struct tree_content *t = root->tree;
997 unsigned int i, j, del;
e2eb469d 998 unsigned long new_len;
4cabf858 999 struct last_object lo;
d489bc14 1000 struct object_entry *le;
4cabf858
SP
1001
1002 if (!is_null_sha1(root->versions[1].sha1))
1003 return;
1004
1005 for (i = 0; i < t->entry_count; i++) {
1006 if (t->entries[i]->tree)
1007 store_tree(t->entries[i]);
1008 }
1009
d489bc14 1010 le = find_object(root->versions[0].sha1);
7bfe6e26
SP
1011 if (!S_ISDIR(root->versions[0].mode)
1012 || !le
1013 || le->pack_id != pack_id) {
4cabf858
SP
1014 lo.data = NULL;
1015 lo.depth = 0;
1016 } else {
e2eb469d
SP
1017 mktree(t, 0, &lo.len, &old_tree);
1018 lo.data = old_tree.buffer;
d489bc14 1019 lo.offset = le->offset;
4cabf858 1020 lo.depth = t->delta_depth;
e2eb469d 1021 lo.no_free = 1;
4cabf858 1022 }
4cabf858 1023
8a8c55ea 1024 mktree(t, 1, &new_len, &new_tree);
e2eb469d 1025 store_object(OBJ_TREE, new_tree.buffer, new_len,
4cabf858 1026 &lo, root->versions[1].sha1, 0);
4cabf858
SP
1027
1028 t->delta_depth = lo.depth;
4cabf858
SP
1029 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1030 struct tree_entry *e = t->entries[i];
1031 if (e->versions[1].mode) {
1032 e->versions[0].mode = e->versions[1].mode;
1033 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1034 t->entries[j++] = e;
1035 } else {
1036 release_tree_entry(e);
1037 del++;
1038 }
1039 }
1040 t->entry_count -= del;
463acbe1
SP
1041}
1042
1043static int tree_content_set(
1044 struct tree_entry *root,
1045 const char *p,
1046 const unsigned char *sha1,
1047 const unsigned int mode)
1048{
1049 struct tree_content *t = root->tree;
1050 const char *slash1;
1051 unsigned int i, n;
1052 struct tree_entry *e;
1053
1054 slash1 = strchr(p, '/');
1055 if (slash1)
1056 n = slash1 - p;
1057 else
1058 n = strlen(p);
1059
1060 for (i = 0; i < t->entry_count; i++) {
1061 e = t->entries[i];
1062 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1063 if (!slash1) {
4cabf858
SP
1064 if (e->versions[1].mode == mode
1065 && !hashcmp(e->versions[1].sha1, sha1))
463acbe1 1066 return 0;
4cabf858
SP
1067 e->versions[1].mode = mode;
1068 hashcpy(e->versions[1].sha1, sha1);
463acbe1 1069 if (e->tree) {
afde8dd9 1070 release_tree_content_recursive(e->tree);
463acbe1
SP
1071 e->tree = NULL;
1072 }
4cabf858 1073 hashclr(root->versions[1].sha1);
463acbe1
SP
1074 return 1;
1075 }
4cabf858 1076 if (!S_ISDIR(e->versions[1].mode)) {
463acbe1 1077 e->tree = new_tree_content(8);
4cabf858 1078 e->versions[1].mode = S_IFDIR;
463acbe1
SP
1079 }
1080 if (!e->tree)
1081 load_tree(e);
1082 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
4cabf858 1083 hashclr(root->versions[1].sha1);
463acbe1
SP
1084 return 1;
1085 }
1086 return 0;
1087 }
1088 }
1089
1090 if (t->entry_count == t->entry_capacity)
1091 root->tree = t = grow_tree_content(t, 8);
1092 e = new_tree_entry();
1093 e->name = to_atom(p, n);
4cabf858
SP
1094 e->versions[0].mode = 0;
1095 hashclr(e->versions[0].sha1);
463acbe1
SP
1096 t->entries[t->entry_count++] = e;
1097 if (slash1) {
1098 e->tree = new_tree_content(8);
4cabf858 1099 e->versions[1].mode = S_IFDIR;
463acbe1
SP
1100 tree_content_set(e, slash1 + 1, sha1, mode);
1101 } else {
1102 e->tree = NULL;
4cabf858
SP
1103 e->versions[1].mode = mode;
1104 hashcpy(e->versions[1].sha1, sha1);
463acbe1 1105 }
4cabf858 1106 hashclr(root->versions[1].sha1);
463acbe1
SP
1107 return 1;
1108}
1109
1110static int tree_content_remove(struct tree_entry *root, const char *p)
1111{
1112 struct tree_content *t = root->tree;
1113 const char *slash1;
1114 unsigned int i, n;
1115 struct tree_entry *e;
1116
1117 slash1 = strchr(p, '/');
1118 if (slash1)
1119 n = slash1 - p;
1120 else
1121 n = strlen(p);
1122
1123 for (i = 0; i < t->entry_count; i++) {
1124 e = t->entries[i];
1125 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
4cabf858 1126 if (!slash1 || !S_ISDIR(e->versions[1].mode))
463acbe1
SP
1127 goto del_entry;
1128 if (!e->tree)
1129 load_tree(e);
1130 if (tree_content_remove(e, slash1 + 1)) {
b54d6422
SP
1131 for (n = 0; n < e->tree->entry_count; n++) {
1132 if (e->tree->entries[n]->versions[1].mode) {
1133 hashclr(root->versions[1].sha1);
1134 return 1;
1135 }
1136 }
1137 goto del_entry;
463acbe1
SP
1138 }
1139 return 0;
1140 }
1141 }
1142 return 0;
1143
1144del_entry:
4cabf858
SP
1145 if (e->tree) {
1146 release_tree_content_recursive(e->tree);
1147 e->tree = NULL;
1148 }
1149 e->versions[1].mode = 0;
1150 hashclr(e->versions[1].sha1);
1151 hashclr(root->versions[1].sha1);
ac47a738 1152 return 1;
db5e523f
SP
1153}
1154
463acbe1
SP
1155static void dump_branches()
1156{
1157 static const char *msg = "fast-import";
1158 unsigned int i;
1159 struct branch *b;
1160 struct ref_lock *lock;
1161
1162 for (i = 0; i < branch_table_sz; i++) {
1163 for (b = branch_table[i]; b; b = b->table_next_branch) {
1fcdd62a 1164 lock = lock_any_ref_for_update(b->name, NULL);
463acbe1
SP
1165 if (!lock || write_ref_sha1(lock, b->sha1, msg) < 0)
1166 die("Can't write %s", b->name);
1167 }
1168 }
1169}
1170
72303d44
SP
1171static void dump_tags()
1172{
1173 static const char *msg = "fast-import";
1174 struct tag *t;
1175 struct ref_lock *lock;
1176 char path[PATH_MAX];
1177
1178 for (t = first_tag; t; t = t->next_tag) {
1179 sprintf(path, "refs/tags/%s", t->name);
1fcdd62a 1180 lock = lock_any_ref_for_update(path, NULL);
72303d44
SP
1181 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1182 die("Can't write %s", path);
1183 }
1184}
1185
a6a1a831
SP
1186static void dump_marks_helper(FILE *f,
1187 unsigned long base,
1188 struct mark_set *m)
1189{
1190 int k;
1191 if (m->shift) {
1192 for (k = 0; k < 1024; k++) {
1193 if (m->data.sets[k])
1194 dump_marks_helper(f, (base + k) << m->shift,
1195 m->data.sets[k]);
1196 }
1197 } else {
1198 for (k = 0; k < 1024; k++) {
1199 if (m->data.marked[k])
264244a0 1200 fprintf(f, ":%lu %s\n", base + k,
a6a1a831
SP
1201 sha1_to_hex(m->data.marked[k]->sha1));
1202 }
1203 }
1204}
1205
1206static void dump_marks()
1207{
1208 if (mark_file)
1209 {
1210 FILE *f = fopen(mark_file, "w");
1211 dump_marks_helper(f, 0, marks);
1212 fclose(f);
1213 }
1214}
1215
c44cdc7e
SP
1216static void read_next_command()
1217{
1218 read_line(&command_buf, stdin, '\n');
1219}
1220
1221static void cmd_mark()
1222{
1223 if (!strncmp("mark :", command_buf.buf, 6)) {
d8397168 1224 next_mark = strtoul(command_buf.buf + 6, NULL, 10);
c44cdc7e
SP
1225 read_next_command();
1226 }
1227 else
d8397168 1228 next_mark = 0;
c44cdc7e
SP
1229}
1230
1231static void* cmd_data (size_t *size)
1232{
1233 size_t n = 0;
1234 void *buffer;
1235 size_t length;
1236
1237 if (strncmp("data ", command_buf.buf, 5))
1238 die("Expected 'data n' command, found: %s", command_buf.buf);
1239
1240 length = strtoul(command_buf.buf + 5, NULL, 10);
1241 buffer = xmalloc(length);
1242
1243 while (n < length) {
1244 size_t s = fread((char*)buffer + n, 1, length - n, stdin);
1245 if (!s && feof(stdin))
1246 die("EOF in data (%lu bytes remaining)", length - n);
1247 n += s;
1248 }
1249
1250 if (fgetc(stdin) != '\n')
1251 die("An lf did not trail the binary data as expected.");
1252
1253 *size = length;
1254 return buffer;
1255}
1256
463acbe1 1257static void cmd_new_blob()
6143f064 1258{
d8397168
SP
1259 size_t l;
1260 void *d;
c44cdc7e
SP
1261
1262 read_next_command();
1263 cmd_mark();
d8397168 1264 d = cmd_data(&l);
6143f064 1265
d8397168
SP
1266 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1267 free(d);
6143f064
SP
1268}
1269
463acbe1 1270static void unload_one_branch()
6bb5b329 1271{
41e5257f
SP
1272 while (cur_active_branches
1273 && cur_active_branches >= max_active_branches) {
463acbe1
SP
1274 unsigned long min_commit = ULONG_MAX;
1275 struct branch *e, *l = NULL, *p = NULL;
1276
1277 for (e = active_branches; e; e = e->active_next_branch) {
1278 if (e->last_commit < min_commit) {
1279 p = l;
1280 min_commit = e->last_commit;
1281 }
1282 l = e;
1283 }
1284
1285 if (p) {
1286 e = p->active_next_branch;
1287 p->active_next_branch = e->active_next_branch;
1288 } else {
1289 e = active_branches;
1290 active_branches = e->active_next_branch;
1291 }
1292 e->active_next_branch = NULL;
1293 if (e->branch_tree.tree) {
afde8dd9 1294 release_tree_content_recursive(e->branch_tree.tree);
463acbe1
SP
1295 e->branch_tree.tree = NULL;
1296 }
1297 cur_active_branches--;
6bb5b329 1298 }
6bb5b329
SP
1299}
1300
463acbe1 1301static void load_branch(struct branch *b)
6bb5b329 1302{
463acbe1
SP
1303 load_tree(&b->branch_tree);
1304 b->active_next_branch = active_branches;
1305 active_branches = b;
1306 cur_active_branches++;
d6c7eb2c 1307 branch_load_count++;
6bb5b329
SP
1308}
1309
463acbe1 1310static void file_change_m(struct branch *b)
6bb5b329 1311{
c44cdc7e
SP
1312 const char *p = command_buf.buf + 2;
1313 char *p_uq;
1314 const char *endp;
7111feed 1315 struct object_entry *oe;
463acbe1 1316 unsigned char sha1[20];
c44cdc7e 1317 unsigned int mode;
7111feed 1318 char type[20];
6bb5b329 1319
c44cdc7e
SP
1320 p = get_mode(p, &mode);
1321 if (!p)
1322 die("Corrupt mode: %s", command_buf.buf);
1323 switch (mode) {
1324 case S_IFREG | 0644:
1325 case S_IFREG | 0755:
ace4a9d1 1326 case S_IFLNK:
c44cdc7e
SP
1327 case 0644:
1328 case 0755:
1329 /* ok */
1330 break;
1331 default:
1332 die("Corrupt mode: %s", command_buf.buf);
1333 }
1334
d8397168
SP
1335 if (*p == ':') {
1336 char *x;
1337 oe = find_mark(strtoul(p + 1, &x, 10));
cacbdd0a 1338 hashcpy(sha1, oe->sha1);
d8397168
SP
1339 p = x;
1340 } else {
1341 if (get_sha1_hex(p, sha1))
1342 die("Invalid SHA1: %s", command_buf.buf);
1343 oe = find_object(sha1);
1344 p += 40;
1345 }
c44cdc7e
SP
1346 if (*p++ != ' ')
1347 die("Missing space after SHA1: %s", command_buf.buf);
1348
1349 p_uq = unquote_c_style(p, &endp);
1350 if (p_uq) {
1351 if (*endp)
1352 die("Garbage after path in: %s", command_buf.buf);
1353 p = p_uq;
1354 }
6bb5b329 1355
7111feed
SP
1356 if (oe) {
1357 if (oe->type != OBJ_BLOB)
c44cdc7e
SP
1358 die("Not a blob (actually a %s): %s",
1359 command_buf.buf, type_names[oe->type]);
7111feed
SP
1360 } else {
1361 if (sha1_object_info(sha1, type, NULL))
c44cdc7e 1362 die("Blob not found: %s", command_buf.buf);
7111feed 1363 if (strcmp(blob_type, type))
c44cdc7e
SP
1364 die("Not a blob (actually a %s): %s",
1365 command_buf.buf, type);
7111feed 1366 }
6bb5b329 1367
c44cdc7e
SP
1368 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1369
1370 if (p_uq)
1371 free(p_uq);
463acbe1 1372}
6bb5b329 1373
463acbe1
SP
1374static void file_change_d(struct branch *b)
1375{
c44cdc7e
SP
1376 const char *p = command_buf.buf + 2;
1377 char *p_uq;
1378 const char *endp;
1379
1380 p_uq = unquote_c_style(p, &endp);
1381 if (p_uq) {
1382 if (*endp)
1383 die("Garbage after path in: %s", command_buf.buf);
1384 p = p_uq;
1385 }
1386 tree_content_remove(&b->branch_tree, p);
1387 if (p_uq)
1388 free(p_uq);
6bb5b329
SP
1389}
1390
00e2b884
SP
1391static void cmd_from(struct branch *b)
1392{
1393 const char *from, *endp;
1394 char *str_uq;
1395 struct branch *s;
1396
1397 if (strncmp("from ", command_buf.buf, 5))
1398 return;
1399
1400 if (b->last_commit)
1401 die("Can't reinitailize branch %s", b->name);
1402
1403 from = strchr(command_buf.buf, ' ') + 1;
1404 str_uq = unquote_c_style(from, &endp);
1405 if (str_uq) {
1406 if (*endp)
1407 die("Garbage after string in: %s", command_buf.buf);
1408 from = str_uq;
1409 }
1410
1411 s = lookup_branch(from);
1412 if (b == s)
1413 die("Can't create a branch from itself: %s", b->name);
1414 else if (s) {
4cabf858 1415 unsigned char *t = s->branch_tree.versions[1].sha1;
445b8599 1416 hashcpy(b->sha1, s->sha1);
4cabf858
SP
1417 hashcpy(b->branch_tree.versions[0].sha1, t);
1418 hashcpy(b->branch_tree.versions[1].sha1, t);
00e2b884
SP
1419 } else if (*from == ':') {
1420 unsigned long idnum = strtoul(from + 1, NULL, 10);
1421 struct object_entry *oe = find_mark(idnum);
1422 unsigned long size;
1423 char *buf;
1424 if (oe->type != OBJ_COMMIT)
1425 die("Mark :%lu not a commit", idnum);
445b8599 1426 hashcpy(b->sha1, oe->sha1);
7bfe6e26 1427 buf = gfi_unpack_entry(oe, &size);
00e2b884
SP
1428 if (!buf || size < 46)
1429 die("Not a valid commit: %s", from);
1430 if (memcmp("tree ", buf, 5)
4cabf858 1431 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
00e2b884
SP
1432 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1433 free(buf);
4cabf858
SP
1434 hashcpy(b->branch_tree.versions[0].sha1,
1435 b->branch_tree.versions[1].sha1);
00e2b884 1436 } else if (!get_sha1(from, b->sha1)) {
4cabf858
SP
1437 if (is_null_sha1(b->sha1)) {
1438 hashclr(b->branch_tree.versions[0].sha1);
1439 hashclr(b->branch_tree.versions[1].sha1);
1440 } else {
00e2b884
SP
1441 unsigned long size;
1442 char *buf;
1443
1444 buf = read_object_with_reference(b->sha1,
1445 type_names[OBJ_COMMIT], &size, b->sha1);
1446 if (!buf || size < 46)
1447 die("Not a valid commit: %s", from);
1448 if (memcmp("tree ", buf, 5)
4cabf858 1449 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
00e2b884
SP
1450 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1451 free(buf);
4cabf858
SP
1452 hashcpy(b->branch_tree.versions[0].sha1,
1453 b->branch_tree.versions[1].sha1);
00e2b884
SP
1454 }
1455 } else
1456 die("Invalid ref name or SHA1 expression: %s", from);
1457
1458 read_next_command();
1459}
1460
62b6f483
SP
1461static struct hash_list* cmd_merge(unsigned int *count)
1462{
1463 struct hash_list *list = NULL, *n, *e;
1464 const char *from, *endp;
1465 char *str_uq;
1466 struct branch *s;
1467
1468 *count = 0;
1469 while (!strncmp("merge ", command_buf.buf, 6)) {
1470 from = strchr(command_buf.buf, ' ') + 1;
1471 str_uq = unquote_c_style(from, &endp);
1472 if (str_uq) {
1473 if (*endp)
1474 die("Garbage after string in: %s", command_buf.buf);
1475 from = str_uq;
1476 }
1477
1478 n = xmalloc(sizeof(*n));
1479 s = lookup_branch(from);
1480 if (s)
1481 hashcpy(n->sha1, s->sha1);
1482 else if (*from == ':') {
1483 unsigned long idnum = strtoul(from + 1, NULL, 10);
1484 struct object_entry *oe = find_mark(idnum);
1485 if (oe->type != OBJ_COMMIT)
1486 die("Mark :%lu not a commit", idnum);
1487 hashcpy(n->sha1, oe->sha1);
1488 } else if (get_sha1(from, n->sha1))
1489 die("Invalid ref name or SHA1 expression: %s", from);
1490
1491 n->next = NULL;
1492 if (list)
1493 e->next = n;
1494 else
1495 list = n;
1496 e = n;
1497 *count++;
1498 read_next_command();
1499 }
1500 return list;
1501}
1502
463acbe1 1503static void cmd_new_commit()
6bb5b329 1504{
c44cdc7e
SP
1505 struct branch *b;
1506 void *msg;
1507 size_t msglen;
1508 char *str_uq;
1509 const char *endp;
1510 char *sp;
1511 char *author = NULL;
1512 char *committer = NULL;
62b6f483
SP
1513 struct hash_list *merge_list = NULL;
1514 unsigned int merge_count;
c44cdc7e
SP
1515
1516 /* Obtain the branch name from the rest of our command */
1517 sp = strchr(command_buf.buf, ' ') + 1;
1518 str_uq = unquote_c_style(sp, &endp);
1519 if (str_uq) {
1520 if (*endp)
1521 die("Garbage after ref in: %s", command_buf.buf);
1522 sp = str_uq;
1523 }
1524 b = lookup_branch(sp);
463acbe1 1525 if (!b)
00e2b884 1526 b = new_branch(sp);
c44cdc7e
SP
1527 if (str_uq)
1528 free(str_uq);
1529
1530 read_next_command();
1531 cmd_mark();
1532 if (!strncmp("author ", command_buf.buf, 7)) {
1533 author = strdup(command_buf.buf);
1534 read_next_command();
1535 }
1536 if (!strncmp("committer ", command_buf.buf, 10)) {
1537 committer = strdup(command_buf.buf);
1538 read_next_command();
1539 }
1540 if (!committer)
1541 die("Expected committer but didn't get one");
1542 msg = cmd_data(&msglen);
02f3389d
SP
1543 read_next_command();
1544 cmd_from(b);
62b6f483 1545 merge_list = cmd_merge(&merge_count);
c44cdc7e
SP
1546
1547 /* ensure the branch is active/loaded */
41e5257f 1548 if (!b->branch_tree.tree || !max_active_branches) {
463acbe1
SP
1549 unload_one_branch();
1550 load_branch(b);
1551 }
6bb5b329 1552
463acbe1
SP
1553 /* file_change* */
1554 for (;;) {
c44cdc7e 1555 if (1 == command_buf.len)
463acbe1 1556 break;
c44cdc7e 1557 else if (!strncmp("M ", command_buf.buf, 2))
463acbe1 1558 file_change_m(b);
c44cdc7e 1559 else if (!strncmp("D ", command_buf.buf, 2))
463acbe1
SP
1560 file_change_d(b);
1561 else
c44cdc7e 1562 die("Unsupported file_change: %s", command_buf.buf);
02f3389d 1563 read_next_command();
6bb5b329 1564 }
6bb5b329 1565
c44cdc7e 1566 /* build the tree and the commit */
463acbe1 1567 store_tree(&b->branch_tree);
8a8c55ea
SP
1568 hashcpy(b->branch_tree.versions[0].sha1,
1569 b->branch_tree.versions[1].sha1);
243f801d 1570 size_dbuf(&new_data, 97 + msglen
62b6f483 1571 + merge_count * 49
c44cdc7e
SP
1572 + (author
1573 ? strlen(author) + strlen(committer)
1574 : 2 * strlen(committer)));
243f801d 1575 sp = new_data.buffer;
4cabf858
SP
1576 sp += sprintf(sp, "tree %s\n",
1577 sha1_to_hex(b->branch_tree.versions[1].sha1));
445b8599 1578 if (!is_null_sha1(b->sha1))
c44cdc7e 1579 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
62b6f483
SP
1580 while (merge_list) {
1581 struct hash_list *next = merge_list->next;
1582 sp += sprintf(sp, "parent %s\n", sha1_to_hex(merge_list->sha1));
1583 free(merge_list);
1584 merge_list = next;
1585 }
c44cdc7e
SP
1586 if (author)
1587 sp += sprintf(sp, "%s\n", author);
1588 else
1589 sp += sprintf(sp, "author %s\n", committer + 10);
1590 sp += sprintf(sp, "%s\n\n", committer);
1591 memcpy(sp, msg, msglen);
1592 sp += msglen;
1593 if (author)
1594 free(author);
1595 free(committer);
1596 free(msg);
1597
243f801d
SP
1598 store_object(OBJ_COMMIT,
1599 new_data.buffer, sp - (char*)new_data.buffer,
1600 NULL, b->sha1, next_mark);
463acbe1 1601 b->last_commit = object_count_by_type[OBJ_COMMIT];
264244a0
SP
1602
1603 if (branch_log) {
1604 int need_dq = quote_c_style(b->name, NULL, NULL, 0);
1605 fprintf(branch_log, "commit ");
1606 if (need_dq) {
1607 fputc('"', branch_log);
1608 quote_c_style(b->name, NULL, branch_log, 0);
1609 fputc('"', branch_log);
1610 } else
1611 fprintf(branch_log, "%s", b->name);
1612 fprintf(branch_log," :%lu %s\n",next_mark,sha1_to_hex(b->sha1));
1613 }
6bb5b329
SP
1614}
1615
72303d44
SP
1616static void cmd_new_tag()
1617{
1618 char *str_uq;
1619 const char *endp;
1620 char *sp;
1621 const char *from;
1622 char *tagger;
1623 struct branch *s;
1624 void *msg;
1625 size_t msglen;
72303d44 1626 struct tag *t;
264244a0 1627 unsigned long from_mark = 0;
72303d44
SP
1628 unsigned char sha1[20];
1629
1630 /* Obtain the new tag name from the rest of our command */
1631 sp = strchr(command_buf.buf, ' ') + 1;
1632 str_uq = unquote_c_style(sp, &endp);
1633 if (str_uq) {
1634 if (*endp)
1635 die("Garbage after tag name in: %s", command_buf.buf);
1636 sp = str_uq;
1637 }
1638 t = pool_alloc(sizeof(struct tag));
1639 t->next_tag = NULL;
1640 t->name = pool_strdup(sp);
1641 if (last_tag)
1642 last_tag->next_tag = t;
1643 else
1644 first_tag = t;
1645 last_tag = t;
1646 if (str_uq)
1647 free(str_uq);
1648 read_next_command();
1649
1650 /* from ... */
1651 if (strncmp("from ", command_buf.buf, 5))
1652 die("Expected from command, got %s", command_buf.buf);
1653
1654 from = strchr(command_buf.buf, ' ') + 1;
1655 str_uq = unquote_c_style(from, &endp);
1656 if (str_uq) {
1657 if (*endp)
1658 die("Garbage after string in: %s", command_buf.buf);
1659 from = str_uq;
1660 }
1661
1662 s = lookup_branch(from);
1663 if (s) {
445b8599 1664 hashcpy(sha1, s->sha1);
72303d44 1665 } else if (*from == ':') {
264244a0
SP
1666 from_mark = strtoul(from + 1, NULL, 10);
1667 struct object_entry *oe = find_mark(from_mark);
72303d44 1668 if (oe->type != OBJ_COMMIT)
264244a0 1669 die("Mark :%lu not a commit", from_mark);
445b8599 1670 hashcpy(sha1, oe->sha1);
72303d44
SP
1671 } else if (!get_sha1(from, sha1)) {
1672 unsigned long size;
1673 char *buf;
1674
1675 buf = read_object_with_reference(sha1,
1676 type_names[OBJ_COMMIT], &size, sha1);
1677 if (!buf || size < 46)
1678 die("Not a valid commit: %s", from);
1679 free(buf);
1680 } else
1681 die("Invalid ref name or SHA1 expression: %s", from);
1682
1683 if (str_uq)
1684 free(str_uq);
1685 read_next_command();
1686
1687 /* tagger ... */
1688 if (strncmp("tagger ", command_buf.buf, 7))
1689 die("Expected tagger command, got %s", command_buf.buf);
1690 tagger = strdup(command_buf.buf);
1691
1692 /* tag payload/message */
1693 read_next_command();
1694 msg = cmd_data(&msglen);
1695
1696 /* build the tag object */
243f801d
SP
1697 size_dbuf(&new_data, 67+strlen(t->name)+strlen(tagger)+msglen);
1698 sp = new_data.buffer;
72303d44
SP
1699 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
1700 sp += sprintf(sp, "type %s\n", type_names[OBJ_COMMIT]);
1701 sp += sprintf(sp, "tag %s\n", t->name);
1702 sp += sprintf(sp, "%s\n\n", tagger);
1703 memcpy(sp, msg, msglen);
1704 sp += msglen;
1705 free(tagger);
1706 free(msg);
1707
243f801d
SP
1708 store_object(OBJ_TAG, new_data.buffer, sp - (char*)new_data.buffer,
1709 NULL, t->sha1, 0);
264244a0
SP
1710
1711 if (branch_log) {
1712 int need_dq = quote_c_style(t->name, NULL, NULL, 0);
1713 fprintf(branch_log, "tag ");
1714 if (need_dq) {
1715 fputc('"', branch_log);
1716 quote_c_style(t->name, NULL, branch_log, 0);
1717 fputc('"', branch_log);
1718 } else
1719 fprintf(branch_log, "%s", t->name);
1720 fprintf(branch_log," :%lu %s\n",from_mark,sha1_to_hex(t->sha1));
1721 }
72303d44
SP
1722}
1723
5fced8dc
SP
1724static void cmd_reset_branch()
1725{
1726 struct branch *b;
1727 char *str_uq;
1728 const char *endp;
1729 char *sp;
1730
1731 /* Obtain the branch name from the rest of our command */
1732 sp = strchr(command_buf.buf, ' ') + 1;
1733 str_uq = unquote_c_style(sp, &endp);
1734 if (str_uq) {
1735 if (*endp)
1736 die("Garbage after ref in: %s", command_buf.buf);
1737 sp = str_uq;
1738 }
1739 b = lookup_branch(sp);
1740 if (b) {
1741 b->last_commit = 0;
1742 if (b->branch_tree.tree) {
1743 release_tree_content_recursive(b->branch_tree.tree);
1744 b->branch_tree.tree = NULL;
1745 }
1746 }
9938ffc5
SP
1747 else
1748 b = new_branch(sp);
5fced8dc
SP
1749 if (str_uq)
1750 free(str_uq);
9938ffc5
SP
1751 read_next_command();
1752 cmd_from(b);
5fced8dc
SP
1753}
1754
7bfe6e26
SP
1755static void cmd_checkpoint()
1756{
1757 if (object_count) {
1758 end_packfile();
1759 start_packfile();
1760 }
1761 read_next_command();
1762}
1763
d5c57b28 1764static const char fast_import_usage[] =
264244a0 1765"git-fast-import [--objects=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file] [--branch-log=log] temp.pack";
d5c57b28 1766
8bcce301
SP
1767int main(int argc, const char **argv)
1768{
d5c57b28 1769 int i;
243f801d 1770 unsigned long est_obj_cnt = object_entry_alloc;
80144727 1771 unsigned long duplicate_count;
8bcce301 1772
463acbe1
SP
1773 setup_ident();
1774 git_config(git_default_config);
1775
d5c57b28
SP
1776 for (i = 1; i < argc; i++) {
1777 const char *a = argv[i];
1778
1779 if (*a != '-' || !strcmp(a, "--"))
1780 break;
1781 else if (!strncmp(a, "--objects=", 10))
1782 est_obj_cnt = strtoul(a + 10, NULL, 0);
1783 else if (!strncmp(a, "--depth=", 8))
1784 max_depth = strtoul(a + 8, NULL, 0);
1785 else if (!strncmp(a, "--active-branches=", 18))
1786 max_active_branches = strtoul(a + 18, NULL, 0);
a6a1a831
SP
1787 else if (!strncmp(a, "--export-marks=", 15))
1788 mark_file = a + 15;
264244a0
SP
1789 else if (!strncmp(a, "--branch-log=", 13)) {
1790 branch_log = fopen(a + 13, "w");
1791 if (!branch_log)
1792 die("Can't create %s: %s", a + 13, strerror(errno));
1793 }
d5c57b28
SP
1794 else
1795 die("unknown option %s", a);
1796 }
1797 if ((i+1) != argc)
1798 usage(fast_import_usage);
1799 base_name = argv[i];
1800
27d6d290 1801 alloc_objects(est_obj_cnt);
c44cdc7e 1802 strbuf_init(&command_buf);
463acbe1
SP
1803
1804 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
1805 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
1806 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
d8397168 1807 marks = pool_calloc(1, sizeof(struct mark_set));
463acbe1 1808
f70b6534 1809 start_packfile();
db5e523f 1810 for (;;) {
c44cdc7e
SP
1811 read_next_command();
1812 if (command_buf.eof)
db5e523f 1813 break;
c44cdc7e
SP
1814 else if (!strcmp("blob", command_buf.buf))
1815 cmd_new_blob();
c44cdc7e
SP
1816 else if (!strncmp("commit ", command_buf.buf, 7))
1817 cmd_new_commit();
72303d44
SP
1818 else if (!strncmp("tag ", command_buf.buf, 4))
1819 cmd_new_tag();
5fced8dc
SP
1820 else if (!strncmp("reset ", command_buf.buf, 6))
1821 cmd_reset_branch();
7bfe6e26
SP
1822 else if (!strcmp("checkpoint", command_buf.buf))
1823 cmd_checkpoint();
c44cdc7e
SP
1824 else
1825 die("Unsupported command: %s", command_buf.buf);
db5e523f 1826 }
f70b6534 1827 end_packfile();
c44cdc7e 1828
463acbe1 1829 dump_branches();
72303d44 1830 dump_tags();
a6a1a831 1831 dump_marks();
08d7e892
SP
1832 if (branch_log)
1833 fclose(branch_log);
8bcce301 1834
80144727
SP
1835 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
1836 duplicate_count += duplicate_count_by_type[i];
1837
6143f064 1838 fprintf(stderr, "%s statistics:\n", argv[0]);
8a8c55ea 1839 fprintf(stderr, "---------------------------------------------------------------------\n");
6143f064 1840 fprintf(stderr, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count, alloc_count - est_obj_cnt);
8a8c55ea 1841 fprintf(stderr, "Total objects: %10lu (%10lu duplicates )\n", object_count, duplicate_count);
4cabf858
SP
1842 fprintf(stderr, " blobs : %10lu (%10lu duplicates %10lu deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
1843 fprintf(stderr, " trees : %10lu (%10lu duplicates %10lu deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
1844 fprintf(stderr, " commits: %10lu (%10lu duplicates %10lu deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
1845 fprintf(stderr, " tags : %10lu (%10lu duplicates %10lu deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
d6c7eb2c 1846 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
d8397168 1847 fprintf(stderr, " marks: %10u (%10lu unique )\n", (1 << marks->shift) * 1024, marks_set_count);
d6c7eb2c 1848 fprintf(stderr, " atoms: %10u\n", atom_cnt);
7111feed
SP
1849 fprintf(stderr, "Memory total: %10lu KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
1850 fprintf(stderr, " pools: %10lu KiB\n", total_allocd/1024);
1851 fprintf(stderr, " objects: %10lu KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
8a8c55ea 1852 fprintf(stderr, "---------------------------------------------------------------------\n");
7bfe6e26
SP
1853 pack_report();
1854 fprintf(stderr, "---------------------------------------------------------------------\n");
6143f064 1855 fprintf(stderr, "\n");
db5e523f
SP
1856
1857 return 0;
1858}