Added option to export the marks table when fast-import terminates.
[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
SP
8 | new_tag
9 ;
10
c44cdc7e
SP
11 new_blob ::= 'blob' lf
12 mark?
13 file_content;
14 file_content ::= data;
463acbe1 15
c44cdc7e 16 new_commit ::= 'commit' sp ref_str lf
00e2b884
SP
17 mark?
18 ('author' sp name '<' email '>' ts tz lf)?
19 'committer' sp name '<' email '>' ts tz lf
20 commit_msg
02f3389d 21 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
c44cdc7e
SP
22 file_change*
23 lf;
24 commit_msg ::= data;
463acbe1 25
c44cdc7e
SP
26 file_change ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf
27 | 'D' sp path_str lf
463acbe1 28 ;
c44cdc7e
SP
29 mode ::= '644' | '755';
30
31 new_tag ::= 'tag' sp tag_str lf
32 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
33 'tagger' sp name '<' email '>' ts tz lf
34 tag_msg;
35 tag_msg ::= data;
36
37 # note: the first idnum in a stream should be 1 and subsequent
38 # idnums should not have gaps between values as this will cause
39 # the stream parser to reserve space for the gapped values. An
40 # idnum can be updated in the future to a new object by issuing
41 # a new mark directive with the old idnum.
42 #
43 mark ::= 'mark' sp idnum lf;
44
45 # note: declen indicates the length of binary_data in bytes.
46 # declen does not include the lf preceeding or trailing the
47 # binary data.
48 #
49 data ::= 'data' sp declen lf
50 binary_data
51 lf;
52
53 # note: quoted strings are C-style quoting supporting \c for
54 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
55 # is the signed byte value in octal. Note that the only
56 # characters which must actually be escaped to protect the
57 # stream formatting is: \, " and LF. Otherwise these values
58 # are UTF8.
59 #
60 ref_str ::= ref | '"' quoted(ref) '"' ;
61 sha1exp_str ::= sha1exp | '"' quoted(sha1exp) '"' ;
62 tag_str ::= tag | '"' quoted(tag) '"' ;
63 path_str ::= path | '"' quoted(path) '"' ;
64
65 declen ::= # unsigned 32 bit value, ascii base10 notation;
463acbe1 66 binary_data ::= # file content, not interpreted;
c44cdc7e 67
463acbe1
SP
68 sp ::= # ASCII space character;
69 lf ::= # ASCII newline (LF) character;
c44cdc7e
SP
70
71 # note: a colon (':') must precede the numerical value assigned to
72 # an idnum. This is to distinguish it from a ref or tag name as
73 # GIT does not permit ':' in ref or tag strings.
74 #
75 idnum ::= ':' declen;
76 path ::= # GIT style file path, e.g. "a/b/c";
77 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
78 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
463acbe1
SP
79 sha1exp ::= # Any valid GIT SHA1 expression;
80 hexsha1 ::= # SHA1 in hexadecimal format;
c44cdc7e
SP
81
82 # note: name and email are UTF8 strings, however name must not
83 # contain '<' or lf and email must not contain any of the
84 # following: '<', '>', lf.
85 #
86 name ::= # valid GIT author/committer name;
463acbe1 87 email ::= # valid GIT author/committer email;
c44cdc7e
SP
88 ts ::= # time since the epoch in seconds, ascii base10 notation;
89 tz ::= # GIT style timezone;
463acbe1
SP
90*/
91
db5e523f
SP
92#include "builtin.h"
93#include "cache.h"
94#include "object.h"
95#include "blob.h"
463acbe1 96#include "tree.h"
db5e523f
SP
97#include "delta.h"
98#include "pack.h"
463acbe1 99#include "refs.h"
db5e523f 100#include "csum-file.h"
c44cdc7e
SP
101#include "strbuf.h"
102#include "quote.h"
db5e523f 103
27d6d290
SP
104struct object_entry
105{
106 struct object_entry *next;
7111feed 107 enum object_type type;
27d6d290
SP
108 unsigned long offset;
109 unsigned char sha1[20];
110};
111
463acbe1 112struct object_entry_pool
27d6d290 113{
463acbe1 114 struct object_entry_pool *next_pool;
27d6d290
SP
115 struct object_entry *next_free;
116 struct object_entry *end;
ac47a738 117 struct object_entry entries[FLEX_ARRAY]; /* more */
27d6d290
SP
118};
119
d8397168
SP
120struct mark_set
121{
122 int shift;
123 union {
124 struct object_entry *marked[1024];
125 struct mark_set *sets[1024];
126 } data;
127};
128
ac47a738
SP
129struct last_object
130{
131 void *data;
6bb5b329
SP
132 unsigned int len;
133 unsigned int depth;
ac47a738
SP
134 unsigned char sha1[20];
135};
136
463acbe1
SP
137struct mem_pool
138{
139 struct mem_pool *next_pool;
140 char *next_free;
141 char *end;
142 char space[FLEX_ARRAY]; /* more */
143};
144
145struct atom_str
146{
147 struct atom_str *next_atom;
148 int str_len;
149 char str_dat[FLEX_ARRAY]; /* more */
150};
151
152struct tree_content;
6bb5b329
SP
153struct tree_entry
154{
463acbe1
SP
155 struct tree_content *tree;
156 struct atom_str* name;
157 unsigned int mode;
6bb5b329 158 unsigned char sha1[20];
6bb5b329
SP
159};
160
463acbe1 161struct tree_content
6bb5b329 162{
463acbe1
SP
163 unsigned int entry_capacity; /* must match avail_tree_content */
164 unsigned int entry_count;
165 struct tree_entry *entries[FLEX_ARRAY]; /* more */
166};
167
168struct avail_tree_content
169{
170 unsigned int entry_capacity; /* must match tree_content */
171 struct avail_tree_content *next_avail;
6bb5b329
SP
172};
173
174struct branch
175{
463acbe1
SP
176 struct branch *table_next_branch;
177 struct branch *active_next_branch;
6bb5b329 178 const char *name;
463acbe1
SP
179 unsigned long last_commit;
180 struct tree_entry branch_tree;
181 unsigned char sha1[20];
6bb5b329
SP
182};
183
72303d44
SP
184struct tag
185{
186 struct tag *next_tag;
187 const char *name;
188 unsigned char sha1[20];
189};
190
463acbe1
SP
191
192/* Stats and misc. counters */
d5c57b28 193static unsigned long max_depth = 10;
27d6d290 194static unsigned long alloc_count;
6bb5b329 195static unsigned long branch_count;
d6c7eb2c 196static unsigned long branch_load_count;
41e5257f 197static unsigned long remap_count;
db5e523f 198static unsigned long object_count;
8bcce301 199static unsigned long duplicate_count;
d8397168 200static unsigned long marks_set_count;
6143f064
SP
201static unsigned long object_count_by_type[9];
202static unsigned long duplicate_count_by_type[9];
ac47a738 203
463acbe1
SP
204/* Memory pools */
205static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
206static size_t total_allocd;
207static struct mem_pool *mem_pool;
208
c44cdc7e 209/* Atom management */
463acbe1
SP
210static unsigned int atom_table_sz = 4451;
211static unsigned int atom_cnt;
212static struct atom_str **atom_table;
213
214/* The .pack file being generated */
ac47a738 215static int pack_fd;
41e5257f 216static unsigned long pack_size;
ac47a738 217static unsigned char pack_sha1[20];
41e5257f
SP
218static void* pack_base;
219static size_t pack_mlen;
ac47a738
SP
220
221/* Table of objects we've written. */
463acbe1
SP
222static unsigned int object_entry_alloc = 1000;
223static struct object_entry_pool *blocks;
224static struct object_entry *object_table[1 << 16];
d8397168 225static struct mark_set *marks;
a6a1a831 226static const char* mark_file;
ac47a738
SP
227
228/* Our last blob */
463acbe1
SP
229static struct last_object last_blob;
230
231/* Tree management */
232static unsigned int tree_entry_alloc = 1000;
233static void *avail_tree_entry;
234static unsigned int avail_tree_table_sz = 100;
235static struct avail_tree_content **avail_tree_table;
8bcce301 236
6bb5b329 237/* Branch data */
d5c57b28
SP
238static unsigned long max_active_branches = 5;
239static unsigned long cur_active_branches;
240static unsigned long branch_table_sz = 1039;
463acbe1
SP
241static struct branch **branch_table;
242static struct branch *active_branches;
243
72303d44
SP
244/* Tag data */
245static struct tag *first_tag;
246static struct tag *last_tag;
247
c44cdc7e
SP
248/* Input stream parsing */
249static struct strbuf command_buf;
d8397168 250static unsigned long next_mark;
c44cdc7e 251
6bb5b329 252
27d6d290 253static void alloc_objects(int cnt)
8bcce301 254{
463acbe1 255 struct object_entry_pool *b;
27d6d290 256
463acbe1 257 b = xmalloc(sizeof(struct object_entry_pool)
27d6d290 258 + cnt * sizeof(struct object_entry));
463acbe1 259 b->next_pool = blocks;
27d6d290
SP
260 b->next_free = b->entries;
261 b->end = b->entries + cnt;
262 blocks = b;
263 alloc_count += cnt;
264}
8bcce301 265
27d6d290 266static struct object_entry* new_object(unsigned char *sha1)
8bcce301 267{
27d6d290 268 struct object_entry *e;
8bcce301 269
27d6d290 270 if (blocks->next_free == blocks->end)
463acbe1 271 alloc_objects(object_entry_alloc);
8bcce301 272
27d6d290
SP
273 e = blocks->next_free++;
274 memcpy(e->sha1, sha1, sizeof(e->sha1));
275 return e;
8bcce301
SP
276}
277
463acbe1
SP
278static struct object_entry* find_object(unsigned char *sha1)
279{
280 unsigned int h = sha1[0] << 8 | sha1[1];
281 struct object_entry *e;
282 for (e = object_table[h]; e; e = e->next)
283 if (!memcmp(sha1, e->sha1, sizeof(e->sha1)))
284 return e;
285 return NULL;
286}
287
8bcce301
SP
288static struct object_entry* insert_object(unsigned char *sha1)
289{
290 unsigned int h = sha1[0] << 8 | sha1[1];
27d6d290 291 struct object_entry *e = object_table[h];
463acbe1 292 struct object_entry *p = NULL;
8bcce301
SP
293
294 while (e) {
295 if (!memcmp(sha1, e->sha1, sizeof(e->sha1)))
296 return e;
297 p = e;
298 e = e->next;
299 }
300
301 e = new_object(sha1);
463acbe1 302 e->next = NULL;
8bcce301
SP
303 e->offset = 0;
304 if (p)
305 p->next = e;
306 else
27d6d290 307 object_table[h] = e;
8bcce301
SP
308 return e;
309}
db5e523f 310
463acbe1
SP
311static unsigned int hc_str(const char *s, size_t len)
312{
313 unsigned int r = 0;
314 while (len-- > 0)
315 r = r * 31 + *s++;
316 return r;
317}
318
319static void* pool_alloc(size_t len)
320{
321 struct mem_pool *p;
322 void *r;
323
324 for (p = mem_pool; p; p = p->next_pool)
325 if ((p->end - p->next_free >= len))
326 break;
327
328 if (!p) {
329 if (len >= (mem_pool_alloc/2)) {
330 total_allocd += len;
331 return xmalloc(len);
332 }
333 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
334 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
335 p->next_pool = mem_pool;
336 p->next_free = p->space;
337 p->end = p->next_free + mem_pool_alloc;
338 mem_pool = p;
339 }
340
341 r = p->next_free;
8d8928b0
SP
342 /* round out to a pointer alignment */
343 if (len & (sizeof(void*) - 1))
344 len += sizeof(void*) - (len & (sizeof(void*) - 1));
463acbe1
SP
345 p->next_free += len;
346 return r;
347}
348
349static void* pool_calloc(size_t count, size_t size)
350{
351 size_t len = count * size;
352 void *r = pool_alloc(len);
353 memset(r, 0, len);
354 return r;
355}
356
357static char* pool_strdup(const char *s)
358{
359 char *r = pool_alloc(strlen(s) + 1);
360 strcpy(r, s);
361 return r;
362}
363
d8397168
SP
364static void insert_mark(unsigned long idnum, struct object_entry *oe)
365{
366 struct mark_set *s = marks;
367 while ((idnum >> s->shift) >= 1024) {
368 s = pool_calloc(1, sizeof(struct mark_set));
369 s->shift = marks->shift + 10;
370 s->data.sets[0] = marks;
371 marks = s;
372 }
373 while (s->shift) {
374 unsigned long i = idnum >> s->shift;
375 idnum -= i << s->shift;
376 if (!s->data.sets[i]) {
377 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
378 s->data.sets[i]->shift = s->shift - 10;
379 }
380 s = s->data.sets[i];
381 }
382 if (!s->data.marked[idnum])
383 marks_set_count++;
384 s->data.marked[idnum] = oe;
385}
386
387static struct object_entry* find_mark(unsigned long idnum)
388{
389 unsigned long orig_idnum = idnum;
390 struct mark_set *s = marks;
391 struct object_entry *oe = NULL;
392 if ((idnum >> s->shift) < 1024) {
393 while (s && s->shift) {
394 unsigned long i = idnum >> s->shift;
395 idnum -= i << s->shift;
396 s = s->data.sets[i];
397 }
398 if (s)
399 oe = s->data.marked[idnum];
400 }
401 if (!oe)
402 die("mark :%lu not declared", orig_idnum);
403 return oe;
404}
405
463acbe1
SP
406static struct atom_str* to_atom(const char *s, size_t len)
407{
408 unsigned int hc = hc_str(s, len) % atom_table_sz;
409 struct atom_str *c;
410
411 for (c = atom_table[hc]; c; c = c->next_atom)
412 if (c->str_len == len && !strncmp(s, c->str_dat, len))
413 return c;
414
415 c = pool_alloc(sizeof(struct atom_str) + len + 1);
416 c->str_len = len;
417 strncpy(c->str_dat, s, len);
418 c->str_dat[len] = 0;
419 c->next_atom = atom_table[hc];
420 atom_table[hc] = c;
421 atom_cnt++;
422 return c;
423}
424
425static struct branch* lookup_branch(const char *name)
426{
427 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
428 struct branch *b;
429
430 for (b = branch_table[hc]; b; b = b->table_next_branch)
431 if (!strcmp(name, b->name))
432 return b;
433 return NULL;
434}
435
436static struct branch* new_branch(const char *name)
437{
438 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
439 struct branch* b = lookup_branch(name);
440
441 if (b)
442 die("Invalid attempt to create duplicate branch: %s", name);
c44cdc7e
SP
443 if (check_ref_format(name))
444 die("Branch name doesn't conform to GIT standards: %s", name);
463acbe1
SP
445
446 b = pool_calloc(1, sizeof(struct branch));
447 b->name = pool_strdup(name);
448 b->table_next_branch = branch_table[hc];
449 branch_table[hc] = b;
450 branch_count++;
451 return b;
452}
453
454static unsigned int hc_entries(unsigned int cnt)
455{
456 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
457 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
458}
459
460static struct tree_content* new_tree_content(unsigned int cnt)
461{
462 struct avail_tree_content *f, *l = NULL;
463 struct tree_content *t;
464 unsigned int hc = hc_entries(cnt);
465
466 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
467 if (f->entry_capacity >= cnt)
468 break;
469
470 if (f) {
471 if (l)
472 l->next_avail = f->next_avail;
473 else
474 avail_tree_table[hc] = f->next_avail;
475 } else {
476 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
477 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
478 f->entry_capacity = cnt;
479 }
480
481 t = (struct tree_content*)f;
482 t->entry_count = 0;
483 return t;
484}
485
486static void release_tree_entry(struct tree_entry *e);
487static void release_tree_content(struct tree_content *t)
488{
489 struct avail_tree_content *f = (struct avail_tree_content*)t;
490 unsigned int hc = hc_entries(f->entry_capacity);
afde8dd9
SP
491 f->next_avail = avail_tree_table[hc];
492 avail_tree_table[hc] = f;
493}
494
495static void release_tree_content_recursive(struct tree_content *t)
496{
463acbe1
SP
497 unsigned int i;
498 for (i = 0; i < t->entry_count; i++)
499 release_tree_entry(t->entries[i]);
afde8dd9 500 release_tree_content(t);
463acbe1
SP
501}
502
503static struct tree_content* grow_tree_content(
504 struct tree_content *t,
505 int amt)
506{
507 struct tree_content *r = new_tree_content(t->entry_count + amt);
508 r->entry_count = t->entry_count;
509 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
510 release_tree_content(t);
511 return r;
512}
513
514static struct tree_entry* new_tree_entry()
515{
516 struct tree_entry *e;
517
518 if (!avail_tree_entry) {
519 unsigned int n = tree_entry_alloc;
8435a9cb 520 total_allocd += n * sizeof(struct tree_entry);
463acbe1
SP
521 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
522 while (n--) {
523 *((void**)e) = e + 1;
524 e++;
525 }
526 }
527
528 e = avail_tree_entry;
529 avail_tree_entry = *((void**)e);
530 return e;
531}
532
533static void release_tree_entry(struct tree_entry *e)
534{
535 if (e->tree)
afde8dd9 536 release_tree_content_recursive(e->tree);
463acbe1
SP
537 *((void**)e) = avail_tree_entry;
538 avail_tree_entry = e;
539}
540
541static void yread(int fd, void *buffer, size_t length)
db5e523f
SP
542{
543 ssize_t ret = 0;
544 while (ret < length) {
545 ssize_t size = xread(fd, (char *) buffer + ret, length - ret);
463acbe1
SP
546 if (!size)
547 die("Read from descriptor %i: end of stream", fd);
548 if (size < 0)
549 die("Read from descriptor %i: %s", fd, strerror(errno));
550 ret += size;
551 }
552}
553
463acbe1 554static void ywrite(int fd, void *buffer, size_t length)
db5e523f
SP
555{
556 ssize_t ret = 0;
557 while (ret < length) {
558 ssize_t size = xwrite(fd, (char *) buffer + ret, length - ret);
463acbe1
SP
559 if (!size)
560 die("Write to descriptor %i: end of file", fd);
561 if (size < 0)
562 die("Write to descriptor %i: %s", fd, strerror(errno));
db5e523f
SP
563 ret += size;
564 }
db5e523f
SP
565}
566
c44cdc7e 567static size_t encode_header(
ac47a738 568 enum object_type type,
c44cdc7e 569 size_t size,
ac47a738 570 unsigned char *hdr)
db5e523f
SP
571{
572 int n = 1;
573 unsigned char c;
574
575 if (type < OBJ_COMMIT || type > OBJ_DELTA)
576 die("bad type %d", type);
577
578 c = (type << 4) | (size & 15);
579 size >>= 4;
580 while (size) {
581 *hdr++ = c | 0x80;
582 c = size & 0x7f;
583 size >>= 7;
584 n++;
585 }
586 *hdr = c;
587 return n;
588}
589
ac47a738
SP
590static int store_object(
591 enum object_type type,
592 void *dat,
c44cdc7e 593 size_t datlen,
6bb5b329 594 struct last_object *last,
d8397168
SP
595 unsigned char *sha1out,
596 unsigned long mark)
db5e523f 597{
db5e523f 598 void *out, *delta;
ac47a738
SP
599 struct object_entry *e;
600 unsigned char hdr[96];
601 unsigned char sha1[20];
db5e523f 602 unsigned long hdrlen, deltalen;
ac47a738
SP
603 SHA_CTX c;
604 z_stream s;
605
606 hdrlen = sprintf((char*)hdr,"%s %lu",type_names[type],datlen) + 1;
607 SHA1_Init(&c);
608 SHA1_Update(&c, hdr, hdrlen);
609 SHA1_Update(&c, dat, datlen);
610 SHA1_Final(sha1, &c);
6bb5b329
SP
611 if (sha1out)
612 memcpy(sha1out, sha1, sizeof(sha1));
ac47a738
SP
613
614 e = insert_object(sha1);
d8397168
SP
615 if (mark)
616 insert_mark(mark, e);
ac47a738
SP
617 if (e->offset) {
618 duplicate_count++;
6143f064 619 duplicate_count_by_type[type]++;
463acbe1 620 return 1;
ac47a738 621 }
7111feed 622 e->type = type;
41e5257f 623 e->offset = pack_size;
ac47a738 624 object_count++;
6143f064 625 object_count_by_type[type]++;
db5e523f 626
463acbe1 627 if (last && last->data && last->depth < max_depth)
ac47a738 628 delta = diff_delta(last->data, last->len,
db5e523f
SP
629 dat, datlen,
630 &deltalen, 0);
ac47a738 631 else
db5e523f
SP
632 delta = 0;
633
634 memset(&s, 0, sizeof(s));
635 deflateInit(&s, zlib_compression_level);
636
637 if (delta) {
ac47a738 638 last->depth++;
db5e523f
SP
639 s.next_in = delta;
640 s.avail_in = deltalen;
641 hdrlen = encode_header(OBJ_DELTA, deltalen, hdr);
463acbe1
SP
642 ywrite(pack_fd, hdr, hdrlen);
643 ywrite(pack_fd, last->sha1, sizeof(sha1));
41e5257f 644 pack_size += hdrlen + sizeof(sha1);
db5e523f 645 } else {
463acbe1
SP
646 if (last)
647 last->depth = 0;
db5e523f
SP
648 s.next_in = dat;
649 s.avail_in = datlen;
ac47a738 650 hdrlen = encode_header(type, datlen, hdr);
463acbe1 651 ywrite(pack_fd, hdr, hdrlen);
41e5257f 652 pack_size += hdrlen;
db5e523f
SP
653 }
654
655 s.avail_out = deflateBound(&s, s.avail_in);
656 s.next_out = out = xmalloc(s.avail_out);
657 while (deflate(&s, Z_FINISH) == Z_OK)
658 /* nothing */;
659 deflateEnd(&s);
660
463acbe1 661 ywrite(pack_fd, out, s.total_out);
41e5257f 662 pack_size += s.total_out;
db5e523f
SP
663
664 free(out);
665 if (delta)
666 free(delta);
463acbe1
SP
667 if (last) {
668 if (last->data)
669 free(last->data);
670 last->data = dat;
671 last->len = datlen;
672 memcpy(last->sha1, sha1, sizeof(sha1));
673 }
674 return 0;
675}
676
41e5257f
SP
677static void* map_pack(unsigned long offset)
678{
679 if (offset >= pack_size)
680 die("object offset outside of pack file");
681 if (offset >= pack_mlen) {
682 if (pack_base)
683 munmap(pack_base, pack_mlen);
684 /* round out how much we map to 16 MB units */
685 pack_mlen = pack_size;
686 if (pack_mlen & ((1 << 24) - 1))
687 pack_mlen = ((pack_mlen >> 24) + 1) << 24;
688 pack_base = mmap(NULL,pack_mlen,PROT_READ,MAP_SHARED,pack_fd,0);
689 if (pack_base == MAP_FAILED)
690 die("Failed to map generated pack: %s", strerror(errno));
691 remap_count++;
692 }
693 return (char*)pack_base + offset;
694}
695
696static unsigned long unpack_object_header(unsigned long offset,
697 enum object_type *type,
698 unsigned long *sizep)
699{
700 unsigned shift;
701 unsigned char c;
702 unsigned long size;
703
704 c = *(unsigned char*)map_pack(offset++);
705 *type = (c >> 4) & 7;
706 size = c & 15;
707 shift = 4;
708 while (c & 0x80) {
709 c = *(unsigned char*)map_pack(offset++);
710 size += (c & 0x7f) << shift;
711 shift += 7;
712 }
713 *sizep = size;
714 return offset;
715}
716
717static void *unpack_non_delta_entry(unsigned long o, unsigned long sz)
718{
719 z_stream stream;
720 unsigned char *result;
721
722 result = xmalloc(sz + 1);
723 result[sz] = 0;
724
725 memset(&stream, 0, sizeof(stream));
726 stream.next_in = map_pack(o);
727 stream.avail_in = pack_mlen - o;
728 stream.next_out = result;
729 stream.avail_out = sz;
730
731 inflateInit(&stream);
732 for (;;) {
733 int st = inflate(&stream, Z_FINISH);
734 if (st == Z_STREAM_END)
735 break;
736 if (st == Z_OK) {
737 o = stream.next_in - (unsigned char*)pack_base;
738 stream.next_in = map_pack(o);
739 stream.avail_in = pack_mlen - o;
740 continue;
741 }
742 die("Error from zlib during inflate.");
743 }
744 inflateEnd(&stream);
745 if (stream.total_out != sz)
746 die("Error after inflate: sizes mismatch");
747 return result;
748}
749
750static void *unpack_entry(unsigned long offset, unsigned long *sizep);
751
752static void *unpack_delta_entry(unsigned long offset,
753 unsigned long delta_size,
754 unsigned long *sizep)
755{
756 struct object_entry *base_oe;
757 unsigned char *base_sha1;
758 void *delta_data, *base, *result;
759 unsigned long base_size, result_size;
760
761 base_sha1 = (unsigned char*)map_pack(offset + 20) - 20;
762 base_oe = find_object(base_sha1);
763 if (!base_oe)
764 die("I'm broken; I can't find a base I know must be here.");
765 base = unpack_entry(base_oe->offset, &base_size);
766 delta_data = unpack_non_delta_entry(offset + 20, delta_size);
767 result = patch_delta(base, base_size,
768 delta_data, delta_size,
769 &result_size);
770 if (!result)
771 die("failed to apply delta");
772 free(delta_data);
773 free(base);
774 *sizep = result_size;
775 return result;
776}
777
778static void *unpack_entry(unsigned long offset, unsigned long *sizep)
779{
780 unsigned long size;
781 enum object_type kind;
782
783 offset = unpack_object_header(offset, &kind, &size);
784 switch (kind) {
785 case OBJ_DELTA:
786 return unpack_delta_entry(offset, size, sizep);
787 case OBJ_COMMIT:
788 case OBJ_TREE:
789 case OBJ_BLOB:
790 case OBJ_TAG:
791 *sizep = size;
792 return unpack_non_delta_entry(offset, size);
793 default:
794 die("I created an object I can't read!");
795 }
796}
797
463acbe1
SP
798static const char *get_mode(const char *str, unsigned int *modep)
799{
800 unsigned char c;
801 unsigned int mode = 0;
802
803 while ((c = *str++) != ' ') {
804 if (c < '0' || c > '7')
805 return NULL;
806 mode = (mode << 3) + (c - '0');
807 }
808 *modep = mode;
809 return str;
810}
811
812static void load_tree(struct tree_entry *root)
813{
814 struct object_entry *myoe;
815 struct tree_content *t;
816 unsigned long size;
817 char *buf;
818 const char *c;
463acbe1
SP
819
820 root->tree = t = new_tree_content(8);
821 if (!memcmp(root->sha1, null_sha1, 20))
822 return;
823
824 myoe = find_object(root->sha1);
825 if (myoe) {
41e5257f
SP
826 if (myoe->type != OBJ_TREE)
827 die("Not a tree: %s", sha1_to_hex(root->sha1));
828 buf = unpack_entry(myoe->offset, &size);
463acbe1 829 } else {
41e5257f 830 char type[20];
463acbe1 831 buf = read_sha1_file(root->sha1, type, &size);
00e2b884 832 if (!buf || strcmp(type, tree_type))
41e5257f 833 die("Can't load tree %s", sha1_to_hex(root->sha1));
463acbe1
SP
834 }
835
836 c = buf;
837 while (c != (buf + size)) {
838 struct tree_entry *e = new_tree_entry();
839
840 if (t->entry_count == t->entry_capacity)
841 root->tree = t = grow_tree_content(t, 8);
842 t->entries[t->entry_count++] = e;
843
844 e->tree = NULL;
845 c = get_mode(c, &e->mode);
846 if (!c)
847 die("Corrupt mode in %s", sha1_to_hex(root->sha1));
848 e->name = to_atom(c, strlen(c));
849 c += e->name->str_len + 1;
850 memcpy(e->sha1, c, sizeof(e->sha1));
851 c += 20;
852 }
853 free(buf);
854}
855
856static int tecmp (const void *_a, const void *_b)
857{
858 struct tree_entry *a = *((struct tree_entry**)_a);
859 struct tree_entry *b = *((struct tree_entry**)_b);
860 return base_name_compare(
861 a->name->str_dat, a->name->str_len, a->mode,
862 b->name->str_dat, b->name->str_len, b->mode);
863}
864
865static void store_tree(struct tree_entry *root)
866{
867 struct tree_content *t = root->tree;
868 unsigned int i;
869 size_t maxlen;
870 char *buf, *c;
871
872 if (memcmp(root->sha1, null_sha1, 20))
873 return;
874
875 maxlen = 0;
876 for (i = 0; i < t->entry_count; i++) {
877 maxlen += t->entries[i]->name->str_len + 34;
878 if (t->entries[i]->tree)
879 store_tree(t->entries[i]);
880 }
881
882 qsort(t->entries, t->entry_count, sizeof(t->entries[0]), tecmp);
883 buf = c = xmalloc(maxlen);
884 for (i = 0; i < t->entry_count; i++) {
885 struct tree_entry *e = t->entries[i];
886 c += sprintf(c, "%o", e->mode);
887 *c++ = ' ';
888 strcpy(c, e->name->str_dat);
889 c += e->name->str_len + 1;
890 memcpy(c, e->sha1, 20);
891 c += 20;
892 }
d8397168 893 store_object(OBJ_TREE, buf, c - buf, NULL, root->sha1, 0);
463acbe1
SP
894 free(buf);
895}
896
897static int tree_content_set(
898 struct tree_entry *root,
899 const char *p,
900 const unsigned char *sha1,
901 const unsigned int mode)
902{
903 struct tree_content *t = root->tree;
904 const char *slash1;
905 unsigned int i, n;
906 struct tree_entry *e;
907
908 slash1 = strchr(p, '/');
909 if (slash1)
910 n = slash1 - p;
911 else
912 n = strlen(p);
913
914 for (i = 0; i < t->entry_count; i++) {
915 e = t->entries[i];
916 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
917 if (!slash1) {
918 if (e->mode == mode && !memcmp(e->sha1, sha1, 20))
919 return 0;
920 e->mode = mode;
921 memcpy(e->sha1, sha1, 20);
922 if (e->tree) {
afde8dd9 923 release_tree_content_recursive(e->tree);
463acbe1
SP
924 e->tree = NULL;
925 }
926 memcpy(root->sha1, null_sha1, 20);
927 return 1;
928 }
929 if (!S_ISDIR(e->mode)) {
930 e->tree = new_tree_content(8);
7111feed 931 e->mode = S_IFDIR;
463acbe1
SP
932 }
933 if (!e->tree)
934 load_tree(e);
935 if (tree_content_set(e, slash1 + 1, sha1, mode)) {
936 memcpy(root->sha1, null_sha1, 20);
937 return 1;
938 }
939 return 0;
940 }
941 }
942
943 if (t->entry_count == t->entry_capacity)
944 root->tree = t = grow_tree_content(t, 8);
945 e = new_tree_entry();
946 e->name = to_atom(p, n);
947 t->entries[t->entry_count++] = e;
948 if (slash1) {
949 e->tree = new_tree_content(8);
7111feed 950 e->mode = S_IFDIR;
463acbe1
SP
951 tree_content_set(e, slash1 + 1, sha1, mode);
952 } else {
953 e->tree = NULL;
954 e->mode = mode;
955 memcpy(e->sha1, sha1, 20);
956 }
957 memcpy(root->sha1, null_sha1, 20);
958 return 1;
959}
960
961static int tree_content_remove(struct tree_entry *root, const char *p)
962{
963 struct tree_content *t = root->tree;
964 const char *slash1;
965 unsigned int i, n;
966 struct tree_entry *e;
967
968 slash1 = strchr(p, '/');
969 if (slash1)
970 n = slash1 - p;
971 else
972 n = strlen(p);
973
974 for (i = 0; i < t->entry_count; i++) {
975 e = t->entries[i];
976 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
977 if (!slash1 || !S_ISDIR(e->mode))
978 goto del_entry;
979 if (!e->tree)
980 load_tree(e);
981 if (tree_content_remove(e, slash1 + 1)) {
982 if (!e->tree->entry_count)
983 goto del_entry;
984 memcpy(root->sha1, null_sha1, 20);
985 return 1;
986 }
987 return 0;
988 }
989 }
990 return 0;
991
992del_entry:
993 for (i++; i < t->entry_count; i++)
994 t->entries[i-1] = t->entries[i];
995 t->entry_count--;
996 release_tree_entry(e);
997 memcpy(root->sha1, null_sha1, 20);
ac47a738 998 return 1;
db5e523f
SP
999}
1000
8bcce301 1001static void init_pack_header()
db5e523f 1002{
c90be46a
SP
1003 struct pack_header hdr;
1004
1005 hdr.hdr_signature = htonl(PACK_SIGNATURE);
1006 hdr.hdr_version = htonl(2);
1007 hdr.hdr_entries = 0;
1008
1009 ywrite(pack_fd, &hdr, sizeof(hdr));
41e5257f 1010 pack_size = sizeof(hdr);
db5e523f
SP
1011}
1012
8bcce301 1013static void fixup_header_footer()
db5e523f
SP
1014{
1015 SHA_CTX c;
1016 char hdr[8];
db5e523f
SP
1017 unsigned long cnt;
1018 char *buf;
1019 size_t n;
1020
ac47a738 1021 if (lseek(pack_fd, 0, SEEK_SET) != 0)
db5e523f
SP
1022 die("Failed seeking to start: %s", strerror(errno));
1023
1024 SHA1_Init(&c);
463acbe1 1025 yread(pack_fd, hdr, 8);
db5e523f
SP
1026 SHA1_Update(&c, hdr, 8);
1027
db5e523f
SP
1028 cnt = htonl(object_count);
1029 SHA1_Update(&c, &cnt, 4);
463acbe1 1030 ywrite(pack_fd, &cnt, 4);
db5e523f
SP
1031
1032 buf = xmalloc(128 * 1024);
1033 for (;;) {
ac47a738 1034 n = xread(pack_fd, buf, 128 * 1024);
db5e523f
SP
1035 if (n <= 0)
1036 break;
1037 SHA1_Update(&c, buf, n);
1038 }
1039 free(buf);
1040
ac47a738 1041 SHA1_Final(pack_sha1, &c);
463acbe1 1042 ywrite(pack_fd, pack_sha1, sizeof(pack_sha1));
db5e523f
SP
1043}
1044
8bcce301 1045static int oecmp (const void *_a, const void *_b)
db5e523f 1046{
8bcce301
SP
1047 struct object_entry *a = *((struct object_entry**)_a);
1048 struct object_entry *b = *((struct object_entry**)_b);
1049 return memcmp(a->sha1, b->sha1, sizeof(a->sha1));
1050}
1051
1052static void write_index(const char *idx_name)
1053{
1054 struct sha1file *f;
1055 struct object_entry **idx, **c, **last;
1056 struct object_entry *e;
463acbe1 1057 struct object_entry_pool *o;
8bcce301
SP
1058 unsigned int array[256];
1059 int i;
1060
1061 /* Build the sorted table of object IDs. */
1062 idx = xmalloc(object_count * sizeof(struct object_entry*));
1063 c = idx;
463acbe1 1064 for (o = blocks; o; o = o->next_pool)
27d6d290
SP
1065 for (e = o->entries; e != o->next_free; e++)
1066 *c++ = e;
8bcce301
SP
1067 last = idx + object_count;
1068 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
1069
1070 /* Generate the fan-out array. */
1071 c = idx;
1072 for (i = 0; i < 256; i++) {
1073 struct object_entry **next = c;;
1074 while (next < last) {
1075 if ((*next)->sha1[0] != i)
1076 break;
1077 next++;
1078 }
1079 array[i] = htonl(next - idx);
1080 c = next;
1081 }
1082
1083 f = sha1create("%s", idx_name);
1084 sha1write(f, array, 256 * sizeof(int));
1085 for (c = idx; c != last; c++) {
1086 unsigned int offset = htonl((*c)->offset);
1087 sha1write(f, &offset, 4);
1088 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
1089 }
ac47a738 1090 sha1write(f, pack_sha1, sizeof(pack_sha1));
8bcce301
SP
1091 sha1close(f, NULL, 1);
1092 free(idx);
1093}
1094
463acbe1
SP
1095static void dump_branches()
1096{
1097 static const char *msg = "fast-import";
1098 unsigned int i;
1099 struct branch *b;
1100 struct ref_lock *lock;
1101
1102 for (i = 0; i < branch_table_sz; i++) {
1103 for (b = branch_table[i]; b; b = b->table_next_branch) {
1104 lock = lock_any_ref_for_update(b->name, NULL, 0);
1105 if (!lock || write_ref_sha1(lock, b->sha1, msg) < 0)
1106 die("Can't write %s", b->name);
1107 }
1108 }
1109}
1110
72303d44
SP
1111static void dump_tags()
1112{
1113 static const char *msg = "fast-import";
1114 struct tag *t;
1115 struct ref_lock *lock;
1116 char path[PATH_MAX];
1117
1118 for (t = first_tag; t; t = t->next_tag) {
1119 sprintf(path, "refs/tags/%s", t->name);
1120 lock = lock_any_ref_for_update(path, NULL, 0);
1121 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1122 die("Can't write %s", path);
1123 }
1124}
1125
a6a1a831
SP
1126static void dump_marks_helper(FILE *f,
1127 unsigned long base,
1128 struct mark_set *m)
1129{
1130 int k;
1131 if (m->shift) {
1132 for (k = 0; k < 1024; k++) {
1133 if (m->data.sets[k])
1134 dump_marks_helper(f, (base + k) << m->shift,
1135 m->data.sets[k]);
1136 }
1137 } else {
1138 for (k = 0; k < 1024; k++) {
1139 if (m->data.marked[k])
1140 fprintf(f, "%lu,%s\n", base + k,
1141 sha1_to_hex(m->data.marked[k]->sha1));
1142 }
1143 }
1144}
1145
1146static void dump_marks()
1147{
1148 if (mark_file)
1149 {
1150 FILE *f = fopen(mark_file, "w");
1151 dump_marks_helper(f, 0, marks);
1152 fclose(f);
1153 }
1154}
1155
c44cdc7e
SP
1156static void read_next_command()
1157{
1158 read_line(&command_buf, stdin, '\n');
1159}
1160
1161static void cmd_mark()
1162{
1163 if (!strncmp("mark :", command_buf.buf, 6)) {
d8397168 1164 next_mark = strtoul(command_buf.buf + 6, NULL, 10);
c44cdc7e
SP
1165 read_next_command();
1166 }
1167 else
d8397168 1168 next_mark = 0;
c44cdc7e
SP
1169}
1170
1171static void* cmd_data (size_t *size)
1172{
1173 size_t n = 0;
1174 void *buffer;
1175 size_t length;
1176
1177 if (strncmp("data ", command_buf.buf, 5))
1178 die("Expected 'data n' command, found: %s", command_buf.buf);
1179
1180 length = strtoul(command_buf.buf + 5, NULL, 10);
1181 buffer = xmalloc(length);
1182
1183 while (n < length) {
1184 size_t s = fread((char*)buffer + n, 1, length - n, stdin);
1185 if (!s && feof(stdin))
1186 die("EOF in data (%lu bytes remaining)", length - n);
1187 n += s;
1188 }
1189
1190 if (fgetc(stdin) != '\n')
1191 die("An lf did not trail the binary data as expected.");
1192
1193 *size = length;
1194 return buffer;
1195}
1196
463acbe1 1197static void cmd_new_blob()
6143f064 1198{
d8397168
SP
1199 size_t l;
1200 void *d;
c44cdc7e
SP
1201
1202 read_next_command();
1203 cmd_mark();
d8397168 1204 d = cmd_data(&l);
6143f064 1205
d8397168
SP
1206 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1207 free(d);
6143f064
SP
1208}
1209
463acbe1 1210static void unload_one_branch()
6bb5b329 1211{
41e5257f
SP
1212 while (cur_active_branches
1213 && cur_active_branches >= max_active_branches) {
463acbe1
SP
1214 unsigned long min_commit = ULONG_MAX;
1215 struct branch *e, *l = NULL, *p = NULL;
1216
1217 for (e = active_branches; e; e = e->active_next_branch) {
1218 if (e->last_commit < min_commit) {
1219 p = l;
1220 min_commit = e->last_commit;
1221 }
1222 l = e;
1223 }
1224
1225 if (p) {
1226 e = p->active_next_branch;
1227 p->active_next_branch = e->active_next_branch;
1228 } else {
1229 e = active_branches;
1230 active_branches = e->active_next_branch;
1231 }
1232 e->active_next_branch = NULL;
1233 if (e->branch_tree.tree) {
afde8dd9 1234 release_tree_content_recursive(e->branch_tree.tree);
463acbe1
SP
1235 e->branch_tree.tree = NULL;
1236 }
1237 cur_active_branches--;
6bb5b329 1238 }
6bb5b329
SP
1239}
1240
463acbe1 1241static void load_branch(struct branch *b)
6bb5b329 1242{
463acbe1
SP
1243 load_tree(&b->branch_tree);
1244 b->active_next_branch = active_branches;
1245 active_branches = b;
1246 cur_active_branches++;
d6c7eb2c 1247 branch_load_count++;
6bb5b329
SP
1248}
1249
463acbe1 1250static void file_change_m(struct branch *b)
6bb5b329 1251{
c44cdc7e
SP
1252 const char *p = command_buf.buf + 2;
1253 char *p_uq;
1254 const char *endp;
7111feed 1255 struct object_entry *oe;
463acbe1 1256 unsigned char sha1[20];
c44cdc7e 1257 unsigned int mode;
7111feed 1258 char type[20];
6bb5b329 1259
c44cdc7e
SP
1260 p = get_mode(p, &mode);
1261 if (!p)
1262 die("Corrupt mode: %s", command_buf.buf);
1263 switch (mode) {
1264 case S_IFREG | 0644:
1265 case S_IFREG | 0755:
ace4a9d1 1266 case S_IFLNK:
c44cdc7e
SP
1267 case 0644:
1268 case 0755:
1269 /* ok */
1270 break;
1271 default:
1272 die("Corrupt mode: %s", command_buf.buf);
1273 }
1274
d8397168
SP
1275 if (*p == ':') {
1276 char *x;
1277 oe = find_mark(strtoul(p + 1, &x, 10));
1278 p = x;
1279 } else {
1280 if (get_sha1_hex(p, sha1))
1281 die("Invalid SHA1: %s", command_buf.buf);
1282 oe = find_object(sha1);
1283 p += 40;
1284 }
c44cdc7e
SP
1285 if (*p++ != ' ')
1286 die("Missing space after SHA1: %s", command_buf.buf);
1287
1288 p_uq = unquote_c_style(p, &endp);
1289 if (p_uq) {
1290 if (*endp)
1291 die("Garbage after path in: %s", command_buf.buf);
1292 p = p_uq;
1293 }
6bb5b329 1294
7111feed
SP
1295 if (oe) {
1296 if (oe->type != OBJ_BLOB)
c44cdc7e
SP
1297 die("Not a blob (actually a %s): %s",
1298 command_buf.buf, type_names[oe->type]);
7111feed
SP
1299 } else {
1300 if (sha1_object_info(sha1, type, NULL))
c44cdc7e 1301 die("Blob not found: %s", command_buf.buf);
7111feed 1302 if (strcmp(blob_type, type))
c44cdc7e
SP
1303 die("Not a blob (actually a %s): %s",
1304 command_buf.buf, type);
7111feed 1305 }
6bb5b329 1306
c44cdc7e
SP
1307 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode);
1308
1309 if (p_uq)
1310 free(p_uq);
463acbe1 1311}
6bb5b329 1312
463acbe1
SP
1313static void file_change_d(struct branch *b)
1314{
c44cdc7e
SP
1315 const char *p = command_buf.buf + 2;
1316 char *p_uq;
1317 const char *endp;
1318
1319 p_uq = unquote_c_style(p, &endp);
1320 if (p_uq) {
1321 if (*endp)
1322 die("Garbage after path in: %s", command_buf.buf);
1323 p = p_uq;
1324 }
1325 tree_content_remove(&b->branch_tree, p);
1326 if (p_uq)
1327 free(p_uq);
6bb5b329
SP
1328}
1329
00e2b884
SP
1330static void cmd_from(struct branch *b)
1331{
1332 const char *from, *endp;
1333 char *str_uq;
1334 struct branch *s;
1335
1336 if (strncmp("from ", command_buf.buf, 5))
1337 return;
1338
1339 if (b->last_commit)
1340 die("Can't reinitailize branch %s", b->name);
1341
1342 from = strchr(command_buf.buf, ' ') + 1;
1343 str_uq = unquote_c_style(from, &endp);
1344 if (str_uq) {
1345 if (*endp)
1346 die("Garbage after string in: %s", command_buf.buf);
1347 from = str_uq;
1348 }
1349
1350 s = lookup_branch(from);
1351 if (b == s)
1352 die("Can't create a branch from itself: %s", b->name);
1353 else if (s) {
1354 memcpy(b->sha1, s->sha1, 20);
1355 memcpy(b->branch_tree.sha1, s->branch_tree.sha1, 20);
1356 } else if (*from == ':') {
1357 unsigned long idnum = strtoul(from + 1, NULL, 10);
1358 struct object_entry *oe = find_mark(idnum);
1359 unsigned long size;
1360 char *buf;
1361 if (oe->type != OBJ_COMMIT)
1362 die("Mark :%lu not a commit", idnum);
1363 memcpy(b->sha1, oe->sha1, 20);
1364 buf = unpack_entry(oe->offset, &size);
1365 if (!buf || size < 46)
1366 die("Not a valid commit: %s", from);
1367 if (memcmp("tree ", buf, 5)
1368 || get_sha1_hex(buf + 5, b->branch_tree.sha1))
1369 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1370 free(buf);
1371 } else if (!get_sha1(from, b->sha1)) {
1372 if (!memcmp(b->sha1, null_sha1, 20))
1373 memcpy(b->branch_tree.sha1, null_sha1, 20);
1374 else {
1375 unsigned long size;
1376 char *buf;
1377
1378 buf = read_object_with_reference(b->sha1,
1379 type_names[OBJ_COMMIT], &size, b->sha1);
1380 if (!buf || size < 46)
1381 die("Not a valid commit: %s", from);
1382 if (memcmp("tree ", buf, 5)
1383 || get_sha1_hex(buf + 5, b->branch_tree.sha1))
1384 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1385 free(buf);
1386 }
1387 } else
1388 die("Invalid ref name or SHA1 expression: %s", from);
1389
1390 read_next_command();
1391}
1392
463acbe1 1393static void cmd_new_commit()
6bb5b329 1394{
c44cdc7e
SP
1395 struct branch *b;
1396 void *msg;
1397 size_t msglen;
1398 char *str_uq;
1399 const char *endp;
1400 char *sp;
1401 char *author = NULL;
1402 char *committer = NULL;
1403 char *body;
1404
1405 /* Obtain the branch name from the rest of our command */
1406 sp = strchr(command_buf.buf, ' ') + 1;
1407 str_uq = unquote_c_style(sp, &endp);
1408 if (str_uq) {
1409 if (*endp)
1410 die("Garbage after ref in: %s", command_buf.buf);
1411 sp = str_uq;
1412 }
1413 b = lookup_branch(sp);
463acbe1 1414 if (!b)
00e2b884 1415 b = new_branch(sp);
c44cdc7e
SP
1416 if (str_uq)
1417 free(str_uq);
1418
1419 read_next_command();
1420 cmd_mark();
1421 if (!strncmp("author ", command_buf.buf, 7)) {
1422 author = strdup(command_buf.buf);
1423 read_next_command();
1424 }
1425 if (!strncmp("committer ", command_buf.buf, 10)) {
1426 committer = strdup(command_buf.buf);
1427 read_next_command();
1428 }
1429 if (!committer)
1430 die("Expected committer but didn't get one");
1431 msg = cmd_data(&msglen);
02f3389d
SP
1432 read_next_command();
1433 cmd_from(b);
c44cdc7e
SP
1434
1435 /* ensure the branch is active/loaded */
41e5257f 1436 if (!b->branch_tree.tree || !max_active_branches) {
463acbe1
SP
1437 unload_one_branch();
1438 load_branch(b);
1439 }
6bb5b329 1440
463acbe1
SP
1441 /* file_change* */
1442 for (;;) {
c44cdc7e 1443 if (1 == command_buf.len)
463acbe1 1444 break;
c44cdc7e 1445 else if (!strncmp("M ", command_buf.buf, 2))
463acbe1 1446 file_change_m(b);
c44cdc7e 1447 else if (!strncmp("D ", command_buf.buf, 2))
463acbe1
SP
1448 file_change_d(b);
1449 else
c44cdc7e 1450 die("Unsupported file_change: %s", command_buf.buf);
02f3389d 1451 read_next_command();
6bb5b329 1452 }
6bb5b329 1453
c44cdc7e 1454 /* build the tree and the commit */
463acbe1 1455 store_tree(&b->branch_tree);
c44cdc7e
SP
1456 body = xmalloc(97 + msglen
1457 + (author
1458 ? strlen(author) + strlen(committer)
1459 : 2 * strlen(committer)));
1460 sp = body;
1461 sp += sprintf(sp, "tree %s\n", sha1_to_hex(b->branch_tree.sha1));
1462 if (memcmp(b->sha1, null_sha1, 20))
1463 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1464 if (author)
1465 sp += sprintf(sp, "%s\n", author);
1466 else
1467 sp += sprintf(sp, "author %s\n", committer + 10);
1468 sp += sprintf(sp, "%s\n\n", committer);
1469 memcpy(sp, msg, msglen);
1470 sp += msglen;
1471 if (author)
1472 free(author);
1473 free(committer);
1474 free(msg);
1475
d8397168 1476 store_object(OBJ_COMMIT, body, sp - body, NULL, b->sha1, next_mark);
463acbe1
SP
1477 free(body);
1478 b->last_commit = object_count_by_type[OBJ_COMMIT];
6bb5b329
SP
1479}
1480
72303d44
SP
1481static void cmd_new_tag()
1482{
1483 char *str_uq;
1484 const char *endp;
1485 char *sp;
1486 const char *from;
1487 char *tagger;
1488 struct branch *s;
1489 void *msg;
1490 size_t msglen;
1491 char *body;
1492 struct tag *t;
1493 unsigned char sha1[20];
1494
1495 /* Obtain the new tag name from the rest of our command */
1496 sp = strchr(command_buf.buf, ' ') + 1;
1497 str_uq = unquote_c_style(sp, &endp);
1498 if (str_uq) {
1499 if (*endp)
1500 die("Garbage after tag name in: %s", command_buf.buf);
1501 sp = str_uq;
1502 }
1503 t = pool_alloc(sizeof(struct tag));
1504 t->next_tag = NULL;
1505 t->name = pool_strdup(sp);
1506 if (last_tag)
1507 last_tag->next_tag = t;
1508 else
1509 first_tag = t;
1510 last_tag = t;
1511 if (str_uq)
1512 free(str_uq);
1513 read_next_command();
1514
1515 /* from ... */
1516 if (strncmp("from ", command_buf.buf, 5))
1517 die("Expected from command, got %s", command_buf.buf);
1518
1519 from = strchr(command_buf.buf, ' ') + 1;
1520 str_uq = unquote_c_style(from, &endp);
1521 if (str_uq) {
1522 if (*endp)
1523 die("Garbage after string in: %s", command_buf.buf);
1524 from = str_uq;
1525 }
1526
1527 s = lookup_branch(from);
1528 if (s) {
1529 memcpy(sha1, s->sha1, 20);
1530 } else if (*from == ':') {
1531 unsigned long idnum = strtoul(from + 1, NULL, 10);
1532 struct object_entry *oe = find_mark(idnum);
1533 if (oe->type != OBJ_COMMIT)
1534 die("Mark :%lu not a commit", idnum);
1535 memcpy(sha1, oe->sha1, 20);
1536 } else if (!get_sha1(from, sha1)) {
1537 unsigned long size;
1538 char *buf;
1539
1540 buf = read_object_with_reference(sha1,
1541 type_names[OBJ_COMMIT], &size, sha1);
1542 if (!buf || size < 46)
1543 die("Not a valid commit: %s", from);
1544 free(buf);
1545 } else
1546 die("Invalid ref name or SHA1 expression: %s", from);
1547
1548 if (str_uq)
1549 free(str_uq);
1550 read_next_command();
1551
1552 /* tagger ... */
1553 if (strncmp("tagger ", command_buf.buf, 7))
1554 die("Expected tagger command, got %s", command_buf.buf);
1555 tagger = strdup(command_buf.buf);
1556
1557 /* tag payload/message */
1558 read_next_command();
1559 msg = cmd_data(&msglen);
1560
1561 /* build the tag object */
1562 body = xmalloc(67 + strlen(t->name) + strlen(tagger) + msglen);
1563 sp = body;
1564 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
1565 sp += sprintf(sp, "type %s\n", type_names[OBJ_COMMIT]);
1566 sp += sprintf(sp, "tag %s\n", t->name);
1567 sp += sprintf(sp, "%s\n\n", tagger);
1568 memcpy(sp, msg, msglen);
1569 sp += msglen;
1570 free(tagger);
1571 free(msg);
1572
1573 store_object(OBJ_TAG, body, sp - body, NULL, t->sha1, 0);
1574 free(body);
1575}
1576
d5c57b28 1577static const char fast_import_usage[] =
a6a1a831 1578"git-fast-import [--objects=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file] temp.pack";
d5c57b28 1579
8bcce301
SP
1580int main(int argc, const char **argv)
1581{
d5c57b28
SP
1582 const char *base_name;
1583 int i;
1584 unsigned long est_obj_cnt = 1000;
8bcce301
SP
1585 char *pack_name;
1586 char *idx_name;
6143f064 1587 struct stat sb;
8bcce301 1588
463acbe1
SP
1589 setup_ident();
1590 git_config(git_default_config);
1591
d5c57b28
SP
1592 for (i = 1; i < argc; i++) {
1593 const char *a = argv[i];
1594
1595 if (*a != '-' || !strcmp(a, "--"))
1596 break;
1597 else if (!strncmp(a, "--objects=", 10))
1598 est_obj_cnt = strtoul(a + 10, NULL, 0);
1599 else if (!strncmp(a, "--depth=", 8))
1600 max_depth = strtoul(a + 8, NULL, 0);
1601 else if (!strncmp(a, "--active-branches=", 18))
1602 max_active_branches = strtoul(a + 18, NULL, 0);
a6a1a831
SP
1603 else if (!strncmp(a, "--export-marks=", 15))
1604 mark_file = a + 15;
d5c57b28
SP
1605 else
1606 die("unknown option %s", a);
1607 }
1608 if ((i+1) != argc)
1609 usage(fast_import_usage);
1610 base_name = argv[i];
1611
8bcce301
SP
1612 pack_name = xmalloc(strlen(base_name) + 6);
1613 sprintf(pack_name, "%s.pack", base_name);
1614 idx_name = xmalloc(strlen(base_name) + 5);
1615 sprintf(idx_name, "%s.idx", base_name);
1616
ac47a738
SP
1617 pack_fd = open(pack_name, O_RDWR|O_CREAT|O_EXCL, 0666);
1618 if (pack_fd < 0)
6143f064 1619 die("Can't create %s: %s", pack_name, strerror(errno));
8bcce301 1620
c44cdc7e 1621 init_pack_header();
27d6d290 1622 alloc_objects(est_obj_cnt);
c44cdc7e 1623 strbuf_init(&command_buf);
463acbe1
SP
1624
1625 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
1626 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
1627 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
d8397168 1628 marks = pool_calloc(1, sizeof(struct mark_set));
463acbe1 1629
db5e523f 1630 for (;;) {
c44cdc7e
SP
1631 read_next_command();
1632 if (command_buf.eof)
db5e523f 1633 break;
c44cdc7e
SP
1634 else if (!strcmp("blob", command_buf.buf))
1635 cmd_new_blob();
c44cdc7e
SP
1636 else if (!strncmp("commit ", command_buf.buf, 7))
1637 cmd_new_commit();
72303d44
SP
1638 else if (!strncmp("tag ", command_buf.buf, 4))
1639 cmd_new_tag();
c44cdc7e
SP
1640 else
1641 die("Unsupported command: %s", command_buf.buf);
db5e523f 1642 }
c44cdc7e 1643
db5e523f 1644 fixup_header_footer();
ac47a738 1645 close(pack_fd);
8bcce301 1646 write_index(idx_name);
463acbe1 1647 dump_branches();
72303d44 1648 dump_tags();
a6a1a831 1649 dump_marks();
8bcce301 1650
6143f064
SP
1651 fprintf(stderr, "%s statistics:\n", argv[0]);
1652 fprintf(stderr, "---------------------------------------------------\n");
1653 fprintf(stderr, "Alloc'd objects: %10lu (%10lu overflow )\n", alloc_count, alloc_count - est_obj_cnt);
1654 fprintf(stderr, "Total objects: %10lu (%10lu duplicates)\n", object_count, duplicate_count);
1655 fprintf(stderr, " blobs : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB]);
1656 fprintf(stderr, " trees : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE]);
1657 fprintf(stderr, " commits: %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT]);
1658 fprintf(stderr, " tags : %10lu (%10lu duplicates)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG]);
d6c7eb2c 1659 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
d8397168 1660 fprintf(stderr, " marks: %10u (%10lu unique )\n", (1 << marks->shift) * 1024, marks_set_count);
d6c7eb2c 1661 fprintf(stderr, " atoms: %10u\n", atom_cnt);
7111feed
SP
1662 fprintf(stderr, "Memory total: %10lu KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
1663 fprintf(stderr, " pools: %10lu KiB\n", total_allocd/1024);
1664 fprintf(stderr, " objects: %10lu KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
41e5257f 1665 fprintf(stderr, "Pack remaps: %10lu\n", remap_count);
6143f064
SP
1666 fprintf(stderr, "---------------------------------------------------\n");
1667
1668 stat(pack_name, &sb);
1669 fprintf(stderr, "Pack size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1670 stat(idx_name, &sb);
1671 fprintf(stderr, "Index size: %10lu KiB\n", (unsigned long)(sb.st_size/1024));
1672
1673 fprintf(stderr, "\n");
db5e523f
SP
1674
1675 return 0;
1676}