diff: squelch empty diffs even more
[git.git] / fast-import.c
blob170cccdcebfd3c715c17d7d762556cd3878a5fd8
1 /*
2 Format of STDIN stream:
4 stream ::= cmd*;
6 cmd ::= new_blob
7 | new_commit
8 | new_tag
9 | reset_branch
10 | checkpoint
13 new_blob ::= 'blob' lf
14 mark?
15 file_content;
16 file_content ::= data;
18 new_commit ::= 'commit' sp ref_str lf
19 mark?
20 ('author' sp name '<' email '>' when lf)?
21 'committer' sp name '<' email '>' when lf
22 commit_msg
23 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
24 ('merge' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)*
25 file_change*
26 lf;
27 commit_msg ::= data;
29 file_change ::= file_clr
30 | file_del
31 | file_rnm
32 | file_cpy
33 | file_obm
34 | file_inm;
35 file_clr ::= 'deleteall' lf;
36 file_del ::= 'D' sp path_str lf;
37 file_rnm ::= 'R' sp path_str sp path_str lf;
38 file_cpy ::= 'C' sp path_str sp path_str lf;
39 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
40 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
41 data;
43 new_tag ::= 'tag' sp tag_str lf
44 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
45 'tagger' sp name '<' email '>' when lf
46 tag_msg;
47 tag_msg ::= data;
49 reset_branch ::= 'reset' sp ref_str lf
50 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
51 lf;
53 checkpoint ::= 'checkpoint' lf
54 lf;
56 # note: the first idnum in a stream should be 1 and subsequent
57 # idnums should not have gaps between values as this will cause
58 # the stream parser to reserve space for the gapped values. An
59 # idnum can be updated in the future to a new object by issuing
60 # a new mark directive with the old idnum.
62 mark ::= 'mark' sp idnum lf;
63 data ::= (delimited_data | exact_data)
64 lf;
66 # note: delim may be any string but must not contain lf.
67 # data_line may contain any data but must not be exactly
68 # delim.
69 delimited_data ::= 'data' sp '<<' delim lf
70 (data_line lf)*
71 delim lf;
73 # note: declen indicates the length of binary_data in bytes.
74 # declen does not include the lf preceeding the binary data.
76 exact_data ::= 'data' sp declen lf
77 binary_data;
79 # note: quoted strings are C-style quoting supporting \c for
80 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
81 # is the signed byte value in octal. Note that the only
82 # characters which must actually be escaped to protect the
83 # stream formatting is: \, " and LF. Otherwise these values
84 # are UTF8.
86 ref_str ::= ref;
87 sha1exp_str ::= sha1exp;
88 tag_str ::= tag;
89 path_str ::= path | '"' quoted(path) '"' ;
90 mode ::= '100644' | '644'
91 | '100755' | '755'
92 | '120000'
95 declen ::= # unsigned 32 bit value, ascii base10 notation;
96 bigint ::= # unsigned integer value, ascii base10 notation;
97 binary_data ::= # file content, not interpreted;
99 when ::= raw_when | rfc2822_when;
100 raw_when ::= ts sp tz;
101 rfc2822_when ::= # Valid RFC 2822 date and time;
103 sp ::= # ASCII space character;
104 lf ::= # ASCII newline (LF) character;
106 # note: a colon (':') must precede the numerical value assigned to
107 # an idnum. This is to distinguish it from a ref or tag name as
108 # GIT does not permit ':' in ref or tag strings.
110 idnum ::= ':' bigint;
111 path ::= # GIT style file path, e.g. "a/b/c";
112 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
113 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
114 sha1exp ::= # Any valid GIT SHA1 expression;
115 hexsha1 ::= # SHA1 in hexadecimal format;
117 # note: name and email are UTF8 strings, however name must not
118 # contain '<' or lf and email must not contain any of the
119 # following: '<', '>', lf.
121 name ::= # valid GIT author/committer name;
122 email ::= # valid GIT author/committer email;
123 ts ::= # time since the epoch in seconds, ascii base10 notation;
124 tz ::= # GIT style timezone;
127 #include "builtin.h"
128 #include "cache.h"
129 #include "object.h"
130 #include "blob.h"
131 #include "tree.h"
132 #include "commit.h"
133 #include "delta.h"
134 #include "pack.h"
135 #include "refs.h"
136 #include "csum-file.h"
137 #include "strbuf.h"
138 #include "quote.h"
140 #define PACK_ID_BITS 16
141 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
143 struct object_entry
145 struct object_entry *next;
146 uint32_t offset;
147 unsigned type : TYPE_BITS;
148 unsigned pack_id : PACK_ID_BITS;
149 unsigned char sha1[20];
152 struct object_entry_pool
154 struct object_entry_pool *next_pool;
155 struct object_entry *next_free;
156 struct object_entry *end;
157 struct object_entry entries[FLEX_ARRAY]; /* more */
160 struct mark_set
162 union {
163 struct object_entry *marked[1024];
164 struct mark_set *sets[1024];
165 } data;
166 unsigned int shift;
169 struct last_object
171 void *data;
172 unsigned long len;
173 uint32_t offset;
174 unsigned int depth;
175 unsigned no_free:1;
178 struct mem_pool
180 struct mem_pool *next_pool;
181 char *next_free;
182 char *end;
183 char space[FLEX_ARRAY]; /* more */
186 struct atom_str
188 struct atom_str *next_atom;
189 unsigned short str_len;
190 char str_dat[FLEX_ARRAY]; /* more */
193 struct tree_content;
194 struct tree_entry
196 struct tree_content *tree;
197 struct atom_str* name;
198 struct tree_entry_ms
200 uint16_t mode;
201 unsigned char sha1[20];
202 } versions[2];
205 struct tree_content
207 unsigned int entry_capacity; /* must match avail_tree_content */
208 unsigned int entry_count;
209 unsigned int delta_depth;
210 struct tree_entry *entries[FLEX_ARRAY]; /* more */
213 struct avail_tree_content
215 unsigned int entry_capacity; /* must match tree_content */
216 struct avail_tree_content *next_avail;
219 struct branch
221 struct branch *table_next_branch;
222 struct branch *active_next_branch;
223 const char *name;
224 struct tree_entry branch_tree;
225 uintmax_t last_commit;
226 unsigned active : 1;
227 unsigned pack_id : PACK_ID_BITS;
228 unsigned char sha1[20];
231 struct tag
233 struct tag *next_tag;
234 const char *name;
235 unsigned int pack_id;
236 unsigned char sha1[20];
239 struct dbuf
241 void *buffer;
242 size_t capacity;
245 struct hash_list
247 struct hash_list *next;
248 unsigned char sha1[20];
251 typedef enum {
252 WHENSPEC_RAW = 1,
253 WHENSPEC_RFC2822,
254 WHENSPEC_NOW,
255 } whenspec_type;
257 /* Configured limits on output */
258 static unsigned long max_depth = 10;
259 static off_t max_packsize = (1LL << 32) - 1;
260 static int force_update;
262 /* Stats and misc. counters */
263 static uintmax_t alloc_count;
264 static uintmax_t marks_set_count;
265 static uintmax_t object_count_by_type[1 << TYPE_BITS];
266 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
267 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
268 static unsigned long object_count;
269 static unsigned long branch_count;
270 static unsigned long branch_load_count;
271 static int failure;
272 static FILE *pack_edges;
274 /* Memory pools */
275 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
276 static size_t total_allocd;
277 static struct mem_pool *mem_pool;
279 /* Atom management */
280 static unsigned int atom_table_sz = 4451;
281 static unsigned int atom_cnt;
282 static struct atom_str **atom_table;
284 /* The .pack file being generated */
285 static unsigned int pack_id;
286 static struct packed_git *pack_data;
287 static struct packed_git **all_packs;
288 static unsigned long pack_size;
290 /* Table of objects we've written. */
291 static unsigned int object_entry_alloc = 5000;
292 static struct object_entry_pool *blocks;
293 static struct object_entry *object_table[1 << 16];
294 static struct mark_set *marks;
295 static const char* mark_file;
297 /* Our last blob */
298 static struct last_object last_blob;
300 /* Tree management */
301 static unsigned int tree_entry_alloc = 1000;
302 static void *avail_tree_entry;
303 static unsigned int avail_tree_table_sz = 100;
304 static struct avail_tree_content **avail_tree_table;
305 static struct dbuf old_tree;
306 static struct dbuf new_tree;
308 /* Branch data */
309 static unsigned long max_active_branches = 5;
310 static unsigned long cur_active_branches;
311 static unsigned long branch_table_sz = 1039;
312 static struct branch **branch_table;
313 static struct branch *active_branches;
315 /* Tag data */
316 static struct tag *first_tag;
317 static struct tag *last_tag;
319 /* Input stream parsing */
320 static whenspec_type whenspec = WHENSPEC_RAW;
321 static struct strbuf command_buf;
322 static uintmax_t next_mark;
323 static struct dbuf new_data;
326 static void alloc_objects(unsigned int cnt)
328 struct object_entry_pool *b;
330 b = xmalloc(sizeof(struct object_entry_pool)
331 + cnt * sizeof(struct object_entry));
332 b->next_pool = blocks;
333 b->next_free = b->entries;
334 b->end = b->entries + cnt;
335 blocks = b;
336 alloc_count += cnt;
339 static struct object_entry *new_object(unsigned char *sha1)
341 struct object_entry *e;
343 if (blocks->next_free == blocks->end)
344 alloc_objects(object_entry_alloc);
346 e = blocks->next_free++;
347 hashcpy(e->sha1, sha1);
348 return e;
351 static struct object_entry *find_object(unsigned char *sha1)
353 unsigned int h = sha1[0] << 8 | sha1[1];
354 struct object_entry *e;
355 for (e = object_table[h]; e; e = e->next)
356 if (!hashcmp(sha1, e->sha1))
357 return e;
358 return NULL;
361 static struct object_entry *insert_object(unsigned char *sha1)
363 unsigned int h = sha1[0] << 8 | sha1[1];
364 struct object_entry *e = object_table[h];
365 struct object_entry *p = NULL;
367 while (e) {
368 if (!hashcmp(sha1, e->sha1))
369 return e;
370 p = e;
371 e = e->next;
374 e = new_object(sha1);
375 e->next = NULL;
376 e->offset = 0;
377 if (p)
378 p->next = e;
379 else
380 object_table[h] = e;
381 return e;
384 static unsigned int hc_str(const char *s, size_t len)
386 unsigned int r = 0;
387 while (len-- > 0)
388 r = r * 31 + *s++;
389 return r;
392 static void *pool_alloc(size_t len)
394 struct mem_pool *p;
395 void *r;
397 for (p = mem_pool; p; p = p->next_pool)
398 if ((p->end - p->next_free >= len))
399 break;
401 if (!p) {
402 if (len >= (mem_pool_alloc/2)) {
403 total_allocd += len;
404 return xmalloc(len);
406 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
407 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
408 p->next_pool = mem_pool;
409 p->next_free = p->space;
410 p->end = p->next_free + mem_pool_alloc;
411 mem_pool = p;
414 r = p->next_free;
415 /* round out to a pointer alignment */
416 if (len & (sizeof(void*) - 1))
417 len += sizeof(void*) - (len & (sizeof(void*) - 1));
418 p->next_free += len;
419 return r;
422 static void *pool_calloc(size_t count, size_t size)
424 size_t len = count * size;
425 void *r = pool_alloc(len);
426 memset(r, 0, len);
427 return r;
430 static char *pool_strdup(const char *s)
432 char *r = pool_alloc(strlen(s) + 1);
433 strcpy(r, s);
434 return r;
437 static void size_dbuf(struct dbuf *b, size_t maxlen)
439 if (b->buffer) {
440 if (b->capacity >= maxlen)
441 return;
442 free(b->buffer);
444 b->capacity = ((maxlen / 1024) + 1) * 1024;
445 b->buffer = xmalloc(b->capacity);
448 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
450 struct mark_set *s = marks;
451 while ((idnum >> s->shift) >= 1024) {
452 s = pool_calloc(1, sizeof(struct mark_set));
453 s->shift = marks->shift + 10;
454 s->data.sets[0] = marks;
455 marks = s;
457 while (s->shift) {
458 uintmax_t i = idnum >> s->shift;
459 idnum -= i << s->shift;
460 if (!s->data.sets[i]) {
461 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
462 s->data.sets[i]->shift = s->shift - 10;
464 s = s->data.sets[i];
466 if (!s->data.marked[idnum])
467 marks_set_count++;
468 s->data.marked[idnum] = oe;
471 static struct object_entry *find_mark(uintmax_t idnum)
473 uintmax_t orig_idnum = idnum;
474 struct mark_set *s = marks;
475 struct object_entry *oe = NULL;
476 if ((idnum >> s->shift) < 1024) {
477 while (s && s->shift) {
478 uintmax_t i = idnum >> s->shift;
479 idnum -= i << s->shift;
480 s = s->data.sets[i];
482 if (s)
483 oe = s->data.marked[idnum];
485 if (!oe)
486 die("mark :%" PRIuMAX " not declared", orig_idnum);
487 return oe;
490 static struct atom_str *to_atom(const char *s, unsigned short len)
492 unsigned int hc = hc_str(s, len) % atom_table_sz;
493 struct atom_str *c;
495 for (c = atom_table[hc]; c; c = c->next_atom)
496 if (c->str_len == len && !strncmp(s, c->str_dat, len))
497 return c;
499 c = pool_alloc(sizeof(struct atom_str) + len + 1);
500 c->str_len = len;
501 strncpy(c->str_dat, s, len);
502 c->str_dat[len] = 0;
503 c->next_atom = atom_table[hc];
504 atom_table[hc] = c;
505 atom_cnt++;
506 return c;
509 static struct branch *lookup_branch(const char *name)
511 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
512 struct branch *b;
514 for (b = branch_table[hc]; b; b = b->table_next_branch)
515 if (!strcmp(name, b->name))
516 return b;
517 return NULL;
520 static struct branch *new_branch(const char *name)
522 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
523 struct branch* b = lookup_branch(name);
525 if (b)
526 die("Invalid attempt to create duplicate branch: %s", name);
527 if (check_ref_format(name))
528 die("Branch name doesn't conform to GIT standards: %s", name);
530 b = pool_calloc(1, sizeof(struct branch));
531 b->name = pool_strdup(name);
532 b->table_next_branch = branch_table[hc];
533 b->branch_tree.versions[0].mode = S_IFDIR;
534 b->branch_tree.versions[1].mode = S_IFDIR;
535 b->active = 0;
536 b->pack_id = MAX_PACK_ID;
537 branch_table[hc] = b;
538 branch_count++;
539 return b;
542 static unsigned int hc_entries(unsigned int cnt)
544 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
545 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
548 static struct tree_content *new_tree_content(unsigned int cnt)
550 struct avail_tree_content *f, *l = NULL;
551 struct tree_content *t;
552 unsigned int hc = hc_entries(cnt);
554 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
555 if (f->entry_capacity >= cnt)
556 break;
558 if (f) {
559 if (l)
560 l->next_avail = f->next_avail;
561 else
562 avail_tree_table[hc] = f->next_avail;
563 } else {
564 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
565 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
566 f->entry_capacity = cnt;
569 t = (struct tree_content*)f;
570 t->entry_count = 0;
571 t->delta_depth = 0;
572 return t;
575 static void release_tree_entry(struct tree_entry *e);
576 static void release_tree_content(struct tree_content *t)
578 struct avail_tree_content *f = (struct avail_tree_content*)t;
579 unsigned int hc = hc_entries(f->entry_capacity);
580 f->next_avail = avail_tree_table[hc];
581 avail_tree_table[hc] = f;
584 static void release_tree_content_recursive(struct tree_content *t)
586 unsigned int i;
587 for (i = 0; i < t->entry_count; i++)
588 release_tree_entry(t->entries[i]);
589 release_tree_content(t);
592 static struct tree_content *grow_tree_content(
593 struct tree_content *t,
594 int amt)
596 struct tree_content *r = new_tree_content(t->entry_count + amt);
597 r->entry_count = t->entry_count;
598 r->delta_depth = t->delta_depth;
599 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
600 release_tree_content(t);
601 return r;
604 static struct tree_entry *new_tree_entry(void)
606 struct tree_entry *e;
608 if (!avail_tree_entry) {
609 unsigned int n = tree_entry_alloc;
610 total_allocd += n * sizeof(struct tree_entry);
611 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
612 while (n-- > 1) {
613 *((void**)e) = e + 1;
614 e++;
616 *((void**)e) = NULL;
619 e = avail_tree_entry;
620 avail_tree_entry = *((void**)e);
621 return e;
624 static void release_tree_entry(struct tree_entry *e)
626 if (e->tree)
627 release_tree_content_recursive(e->tree);
628 *((void**)e) = avail_tree_entry;
629 avail_tree_entry = e;
632 static struct tree_content *dup_tree_content(struct tree_content *s)
634 struct tree_content *d;
635 struct tree_entry *a, *b;
636 unsigned int i;
638 if (!s)
639 return NULL;
640 d = new_tree_content(s->entry_count);
641 for (i = 0; i < s->entry_count; i++) {
642 a = s->entries[i];
643 b = new_tree_entry();
644 memcpy(b, a, sizeof(*a));
645 if (a->tree && is_null_sha1(b->versions[1].sha1))
646 b->tree = dup_tree_content(a->tree);
647 else
648 b->tree = NULL;
649 d->entries[i] = b;
651 d->entry_count = s->entry_count;
652 d->delta_depth = s->delta_depth;
654 return d;
657 static void start_packfile(void)
659 static char tmpfile[PATH_MAX];
660 struct packed_git *p;
661 struct pack_header hdr;
662 int pack_fd;
664 snprintf(tmpfile, sizeof(tmpfile),
665 "%s/tmp_pack_XXXXXX", get_object_directory());
666 pack_fd = xmkstemp(tmpfile);
667 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
668 strcpy(p->pack_name, tmpfile);
669 p->pack_fd = pack_fd;
671 hdr.hdr_signature = htonl(PACK_SIGNATURE);
672 hdr.hdr_version = htonl(2);
673 hdr.hdr_entries = 0;
674 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
676 pack_data = p;
677 pack_size = sizeof(hdr);
678 object_count = 0;
680 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
681 all_packs[pack_id] = p;
684 static int oecmp (const void *a_, const void *b_)
686 struct object_entry *a = *((struct object_entry**)a_);
687 struct object_entry *b = *((struct object_entry**)b_);
688 return hashcmp(a->sha1, b->sha1);
691 static char *create_index(void)
693 static char tmpfile[PATH_MAX];
694 SHA_CTX ctx;
695 struct sha1file *f;
696 struct object_entry **idx, **c, **last, *e;
697 struct object_entry_pool *o;
698 uint32_t array[256];
699 int i, idx_fd;
701 /* Build the sorted table of object IDs. */
702 idx = xmalloc(object_count * sizeof(struct object_entry*));
703 c = idx;
704 for (o = blocks; o; o = o->next_pool)
705 for (e = o->next_free; e-- != o->entries;)
706 if (pack_id == e->pack_id)
707 *c++ = e;
708 last = idx + object_count;
709 if (c != last)
710 die("internal consistency error creating the index");
711 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
713 /* Generate the fan-out array. */
714 c = idx;
715 for (i = 0; i < 256; i++) {
716 struct object_entry **next = c;;
717 while (next < last) {
718 if ((*next)->sha1[0] != i)
719 break;
720 next++;
722 array[i] = htonl(next - idx);
723 c = next;
726 snprintf(tmpfile, sizeof(tmpfile),
727 "%s/tmp_idx_XXXXXX", get_object_directory());
728 idx_fd = xmkstemp(tmpfile);
729 f = sha1fd(idx_fd, tmpfile);
730 sha1write(f, array, 256 * sizeof(int));
731 SHA1_Init(&ctx);
732 for (c = idx; c != last; c++) {
733 uint32_t offset = htonl((*c)->offset);
734 sha1write(f, &offset, 4);
735 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
736 SHA1_Update(&ctx, (*c)->sha1, 20);
738 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
739 sha1close(f, NULL, 1);
740 free(idx);
741 SHA1_Final(pack_data->sha1, &ctx);
742 return tmpfile;
745 static char *keep_pack(char *curr_index_name)
747 static char name[PATH_MAX];
748 static const char *keep_msg = "fast-import";
749 int keep_fd;
751 chmod(pack_data->pack_name, 0444);
752 chmod(curr_index_name, 0444);
754 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
755 get_object_directory(), sha1_to_hex(pack_data->sha1));
756 keep_fd = open(name, O_RDWR|O_CREAT|O_EXCL, 0600);
757 if (keep_fd < 0)
758 die("cannot create keep file");
759 write(keep_fd, keep_msg, strlen(keep_msg));
760 close(keep_fd);
762 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
763 get_object_directory(), sha1_to_hex(pack_data->sha1));
764 if (move_temp_to_file(pack_data->pack_name, name))
765 die("cannot store pack file");
767 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
768 get_object_directory(), sha1_to_hex(pack_data->sha1));
769 if (move_temp_to_file(curr_index_name, name))
770 die("cannot store index file");
771 return name;
774 static void unkeep_all_packs(void)
776 static char name[PATH_MAX];
777 int k;
779 for (k = 0; k < pack_id; k++) {
780 struct packed_git *p = all_packs[k];
781 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
782 get_object_directory(), sha1_to_hex(p->sha1));
783 unlink(name);
787 static void end_packfile(void)
789 struct packed_git *old_p = pack_data, *new_p;
791 if (object_count) {
792 char *idx_name;
793 int i;
794 struct branch *b;
795 struct tag *t;
797 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
798 pack_data->pack_name, object_count);
799 close(pack_data->pack_fd);
800 idx_name = keep_pack(create_index());
802 /* Register the packfile with core git's machinary. */
803 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
804 if (!new_p)
805 die("core git rejected index %s", idx_name);
806 new_p->windows = old_p->windows;
807 all_packs[pack_id] = new_p;
808 install_packed_git(new_p);
810 /* Print the boundary */
811 if (pack_edges) {
812 fprintf(pack_edges, "%s:", new_p->pack_name);
813 for (i = 0; i < branch_table_sz; i++) {
814 for (b = branch_table[i]; b; b = b->table_next_branch) {
815 if (b->pack_id == pack_id)
816 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
819 for (t = first_tag; t; t = t->next_tag) {
820 if (t->pack_id == pack_id)
821 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
823 fputc('\n', pack_edges);
824 fflush(pack_edges);
827 pack_id++;
829 else
830 unlink(old_p->pack_name);
831 free(old_p);
833 /* We can't carry a delta across packfiles. */
834 free(last_blob.data);
835 last_blob.data = NULL;
836 last_blob.len = 0;
837 last_blob.offset = 0;
838 last_blob.depth = 0;
841 static void cycle_packfile(void)
843 end_packfile();
844 start_packfile();
847 static size_t encode_header(
848 enum object_type type,
849 size_t size,
850 unsigned char *hdr)
852 int n = 1;
853 unsigned char c;
855 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
856 die("bad type %d", type);
858 c = (type << 4) | (size & 15);
859 size >>= 4;
860 while (size) {
861 *hdr++ = c | 0x80;
862 c = size & 0x7f;
863 size >>= 7;
864 n++;
866 *hdr = c;
867 return n;
870 static int store_object(
871 enum object_type type,
872 void *dat,
873 size_t datlen,
874 struct last_object *last,
875 unsigned char *sha1out,
876 uintmax_t mark)
878 void *out, *delta;
879 struct object_entry *e;
880 unsigned char hdr[96];
881 unsigned char sha1[20];
882 unsigned long hdrlen, deltalen;
883 SHA_CTX c;
884 z_stream s;
886 hdrlen = sprintf((char*)hdr,"%s %lu", typename(type),
887 (unsigned long)datlen) + 1;
888 SHA1_Init(&c);
889 SHA1_Update(&c, hdr, hdrlen);
890 SHA1_Update(&c, dat, datlen);
891 SHA1_Final(sha1, &c);
892 if (sha1out)
893 hashcpy(sha1out, sha1);
895 e = insert_object(sha1);
896 if (mark)
897 insert_mark(mark, e);
898 if (e->offset) {
899 duplicate_count_by_type[type]++;
900 return 1;
901 } else if (find_sha1_pack(sha1, packed_git)) {
902 e->type = type;
903 e->pack_id = MAX_PACK_ID;
904 e->offset = 1; /* just not zero! */
905 duplicate_count_by_type[type]++;
906 return 1;
909 if (last && last->data && last->depth < max_depth) {
910 delta = diff_delta(last->data, last->len,
911 dat, datlen,
912 &deltalen, 0);
913 if (delta && deltalen >= datlen) {
914 free(delta);
915 delta = NULL;
917 } else
918 delta = NULL;
920 memset(&s, 0, sizeof(s));
921 deflateInit(&s, zlib_compression_level);
922 if (delta) {
923 s.next_in = delta;
924 s.avail_in = deltalen;
925 } else {
926 s.next_in = dat;
927 s.avail_in = datlen;
929 s.avail_out = deflateBound(&s, s.avail_in);
930 s.next_out = out = xmalloc(s.avail_out);
931 while (deflate(&s, Z_FINISH) == Z_OK)
932 /* nothing */;
933 deflateEnd(&s);
935 /* Determine if we should auto-checkpoint. */
936 if ((pack_size + 60 + s.total_out) > max_packsize
937 || (pack_size + 60 + s.total_out) < pack_size) {
939 /* This new object needs to *not* have the current pack_id. */
940 e->pack_id = pack_id + 1;
941 cycle_packfile();
943 /* We cannot carry a delta into the new pack. */
944 if (delta) {
945 free(delta);
946 delta = NULL;
948 memset(&s, 0, sizeof(s));
949 deflateInit(&s, zlib_compression_level);
950 s.next_in = dat;
951 s.avail_in = datlen;
952 s.avail_out = deflateBound(&s, s.avail_in);
953 s.next_out = out = xrealloc(out, s.avail_out);
954 while (deflate(&s, Z_FINISH) == Z_OK)
955 /* nothing */;
956 deflateEnd(&s);
960 e->type = type;
961 e->pack_id = pack_id;
962 e->offset = pack_size;
963 object_count++;
964 object_count_by_type[type]++;
966 if (delta) {
967 unsigned long ofs = e->offset - last->offset;
968 unsigned pos = sizeof(hdr) - 1;
970 delta_count_by_type[type]++;
971 last->depth++;
973 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
974 write_or_die(pack_data->pack_fd, hdr, hdrlen);
975 pack_size += hdrlen;
977 hdr[pos] = ofs & 127;
978 while (ofs >>= 7)
979 hdr[--pos] = 128 | (--ofs & 127);
980 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
981 pack_size += sizeof(hdr) - pos;
982 } else {
983 if (last)
984 last->depth = 0;
985 hdrlen = encode_header(type, datlen, hdr);
986 write_or_die(pack_data->pack_fd, hdr, hdrlen);
987 pack_size += hdrlen;
990 write_or_die(pack_data->pack_fd, out, s.total_out);
991 pack_size += s.total_out;
993 free(out);
994 free(delta);
995 if (last) {
996 if (!last->no_free)
997 free(last->data);
998 last->data = dat;
999 last->offset = e->offset;
1000 last->len = datlen;
1002 return 0;
1005 static void *gfi_unpack_entry(
1006 struct object_entry *oe,
1007 unsigned long *sizep)
1009 enum object_type type;
1010 struct packed_git *p = all_packs[oe->pack_id];
1011 if (p == pack_data)
1012 p->pack_size = pack_size + 20;
1013 return unpack_entry(p, oe->offset, &type, sizep);
1016 static const char *get_mode(const char *str, uint16_t *modep)
1018 unsigned char c;
1019 uint16_t mode = 0;
1021 while ((c = *str++) != ' ') {
1022 if (c < '0' || c > '7')
1023 return NULL;
1024 mode = (mode << 3) + (c - '0');
1026 *modep = mode;
1027 return str;
1030 static void load_tree(struct tree_entry *root)
1032 unsigned char* sha1 = root->versions[1].sha1;
1033 struct object_entry *myoe;
1034 struct tree_content *t;
1035 unsigned long size;
1036 char *buf;
1037 const char *c;
1039 root->tree = t = new_tree_content(8);
1040 if (is_null_sha1(sha1))
1041 return;
1043 myoe = find_object(sha1);
1044 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1045 if (myoe->type != OBJ_TREE)
1046 die("Not a tree: %s", sha1_to_hex(sha1));
1047 t->delta_depth = 0;
1048 buf = gfi_unpack_entry(myoe, &size);
1049 } else {
1050 enum object_type type;
1051 buf = read_sha1_file(sha1, &type, &size);
1052 if (!buf || type != OBJ_TREE)
1053 die("Can't load tree %s", sha1_to_hex(sha1));
1056 c = buf;
1057 while (c != (buf + size)) {
1058 struct tree_entry *e = new_tree_entry();
1060 if (t->entry_count == t->entry_capacity)
1061 root->tree = t = grow_tree_content(t, t->entry_count);
1062 t->entries[t->entry_count++] = e;
1064 e->tree = NULL;
1065 c = get_mode(c, &e->versions[1].mode);
1066 if (!c)
1067 die("Corrupt mode in %s", sha1_to_hex(sha1));
1068 e->versions[0].mode = e->versions[1].mode;
1069 e->name = to_atom(c, strlen(c));
1070 c += e->name->str_len + 1;
1071 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1072 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1073 c += 20;
1075 free(buf);
1078 static int tecmp0 (const void *_a, const void *_b)
1080 struct tree_entry *a = *((struct tree_entry**)_a);
1081 struct tree_entry *b = *((struct tree_entry**)_b);
1082 return base_name_compare(
1083 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1084 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1087 static int tecmp1 (const void *_a, const void *_b)
1089 struct tree_entry *a = *((struct tree_entry**)_a);
1090 struct tree_entry *b = *((struct tree_entry**)_b);
1091 return base_name_compare(
1092 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1093 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1096 static void mktree(struct tree_content *t,
1097 int v,
1098 unsigned long *szp,
1099 struct dbuf *b)
1101 size_t maxlen = 0;
1102 unsigned int i;
1103 char *c;
1105 if (!v)
1106 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1107 else
1108 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1110 for (i = 0; i < t->entry_count; i++) {
1111 if (t->entries[i]->versions[v].mode)
1112 maxlen += t->entries[i]->name->str_len + 34;
1115 size_dbuf(b, maxlen);
1116 c = b->buffer;
1117 for (i = 0; i < t->entry_count; i++) {
1118 struct tree_entry *e = t->entries[i];
1119 if (!e->versions[v].mode)
1120 continue;
1121 c += sprintf(c, "%o", (unsigned int)e->versions[v].mode);
1122 *c++ = ' ';
1123 strcpy(c, e->name->str_dat);
1124 c += e->name->str_len + 1;
1125 hashcpy((unsigned char*)c, e->versions[v].sha1);
1126 c += 20;
1128 *szp = c - (char*)b->buffer;
1131 static void store_tree(struct tree_entry *root)
1133 struct tree_content *t = root->tree;
1134 unsigned int i, j, del;
1135 unsigned long new_len;
1136 struct last_object lo;
1137 struct object_entry *le;
1139 if (!is_null_sha1(root->versions[1].sha1))
1140 return;
1142 for (i = 0; i < t->entry_count; i++) {
1143 if (t->entries[i]->tree)
1144 store_tree(t->entries[i]);
1147 le = find_object(root->versions[0].sha1);
1148 if (!S_ISDIR(root->versions[0].mode)
1149 || !le
1150 || le->pack_id != pack_id) {
1151 lo.data = NULL;
1152 lo.depth = 0;
1153 lo.no_free = 0;
1154 } else {
1155 mktree(t, 0, &lo.len, &old_tree);
1156 lo.data = old_tree.buffer;
1157 lo.offset = le->offset;
1158 lo.depth = t->delta_depth;
1159 lo.no_free = 1;
1162 mktree(t, 1, &new_len, &new_tree);
1163 store_object(OBJ_TREE, new_tree.buffer, new_len,
1164 &lo, root->versions[1].sha1, 0);
1166 t->delta_depth = lo.depth;
1167 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1168 struct tree_entry *e = t->entries[i];
1169 if (e->versions[1].mode) {
1170 e->versions[0].mode = e->versions[1].mode;
1171 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1172 t->entries[j++] = e;
1173 } else {
1174 release_tree_entry(e);
1175 del++;
1178 t->entry_count -= del;
1181 static int tree_content_set(
1182 struct tree_entry *root,
1183 const char *p,
1184 const unsigned char *sha1,
1185 const uint16_t mode,
1186 struct tree_content *subtree)
1188 struct tree_content *t = root->tree;
1189 const char *slash1;
1190 unsigned int i, n;
1191 struct tree_entry *e;
1193 slash1 = strchr(p, '/');
1194 if (slash1)
1195 n = slash1 - p;
1196 else
1197 n = strlen(p);
1198 if (!n)
1199 die("Empty path component found in input");
1200 if (!slash1 && !S_ISDIR(mode) && subtree)
1201 die("Non-directories cannot have subtrees");
1203 for (i = 0; i < t->entry_count; i++) {
1204 e = t->entries[i];
1205 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1206 if (!slash1) {
1207 if (!S_ISDIR(mode)
1208 && e->versions[1].mode == mode
1209 && !hashcmp(e->versions[1].sha1, sha1))
1210 return 0;
1211 e->versions[1].mode = mode;
1212 hashcpy(e->versions[1].sha1, sha1);
1213 if (e->tree)
1214 release_tree_content_recursive(e->tree);
1215 e->tree = subtree;
1216 hashclr(root->versions[1].sha1);
1217 return 1;
1219 if (!S_ISDIR(e->versions[1].mode)) {
1220 e->tree = new_tree_content(8);
1221 e->versions[1].mode = S_IFDIR;
1223 if (!e->tree)
1224 load_tree(e);
1225 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1226 hashclr(root->versions[1].sha1);
1227 return 1;
1229 return 0;
1233 if (t->entry_count == t->entry_capacity)
1234 root->tree = t = grow_tree_content(t, t->entry_count);
1235 e = new_tree_entry();
1236 e->name = to_atom(p, n);
1237 e->versions[0].mode = 0;
1238 hashclr(e->versions[0].sha1);
1239 t->entries[t->entry_count++] = e;
1240 if (slash1) {
1241 e->tree = new_tree_content(8);
1242 e->versions[1].mode = S_IFDIR;
1243 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1244 } else {
1245 e->tree = subtree;
1246 e->versions[1].mode = mode;
1247 hashcpy(e->versions[1].sha1, sha1);
1249 hashclr(root->versions[1].sha1);
1250 return 1;
1253 static int tree_content_remove(
1254 struct tree_entry *root,
1255 const char *p,
1256 struct tree_entry *backup_leaf)
1258 struct tree_content *t = root->tree;
1259 const char *slash1;
1260 unsigned int i, n;
1261 struct tree_entry *e;
1263 slash1 = strchr(p, '/');
1264 if (slash1)
1265 n = slash1 - p;
1266 else
1267 n = strlen(p);
1269 for (i = 0; i < t->entry_count; i++) {
1270 e = t->entries[i];
1271 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1272 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1273 goto del_entry;
1274 if (!e->tree)
1275 load_tree(e);
1276 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1277 for (n = 0; n < e->tree->entry_count; n++) {
1278 if (e->tree->entries[n]->versions[1].mode) {
1279 hashclr(root->versions[1].sha1);
1280 return 1;
1283 backup_leaf = NULL;
1284 goto del_entry;
1286 return 0;
1289 return 0;
1291 del_entry:
1292 if (backup_leaf)
1293 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1294 else if (e->tree)
1295 release_tree_content_recursive(e->tree);
1296 e->tree = NULL;
1297 e->versions[1].mode = 0;
1298 hashclr(e->versions[1].sha1);
1299 hashclr(root->versions[1].sha1);
1300 return 1;
1303 static int tree_content_get(
1304 struct tree_entry *root,
1305 const char *p,
1306 struct tree_entry *leaf)
1308 struct tree_content *t = root->tree;
1309 const char *slash1;
1310 unsigned int i, n;
1311 struct tree_entry *e;
1313 slash1 = strchr(p, '/');
1314 if (slash1)
1315 n = slash1 - p;
1316 else
1317 n = strlen(p);
1319 for (i = 0; i < t->entry_count; i++) {
1320 e = t->entries[i];
1321 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1322 if (!slash1) {
1323 memcpy(leaf, e, sizeof(*leaf));
1324 if (e->tree && is_null_sha1(e->versions[1].sha1))
1325 leaf->tree = dup_tree_content(e->tree);
1326 else
1327 leaf->tree = NULL;
1328 return 1;
1330 if (!S_ISDIR(e->versions[1].mode))
1331 return 0;
1332 if (!e->tree)
1333 load_tree(e);
1334 return tree_content_get(e, slash1 + 1, leaf);
1337 return 0;
1340 static int update_branch(struct branch *b)
1342 static const char *msg = "fast-import";
1343 struct ref_lock *lock;
1344 unsigned char old_sha1[20];
1346 if (read_ref(b->name, old_sha1))
1347 hashclr(old_sha1);
1348 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1349 if (!lock)
1350 return error("Unable to lock %s", b->name);
1351 if (!force_update && !is_null_sha1(old_sha1)) {
1352 struct commit *old_cmit, *new_cmit;
1354 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1355 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1356 if (!old_cmit || !new_cmit) {
1357 unlock_ref(lock);
1358 return error("Branch %s is missing commits.", b->name);
1361 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1362 unlock_ref(lock);
1363 warning("Not updating %s"
1364 " (new tip %s does not contain %s)",
1365 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1366 return -1;
1369 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1370 return error("Unable to update %s", b->name);
1371 return 0;
1374 static void dump_branches(void)
1376 unsigned int i;
1377 struct branch *b;
1379 for (i = 0; i < branch_table_sz; i++) {
1380 for (b = branch_table[i]; b; b = b->table_next_branch)
1381 failure |= update_branch(b);
1385 static void dump_tags(void)
1387 static const char *msg = "fast-import";
1388 struct tag *t;
1389 struct ref_lock *lock;
1390 char ref_name[PATH_MAX];
1392 for (t = first_tag; t; t = t->next_tag) {
1393 sprintf(ref_name, "tags/%s", t->name);
1394 lock = lock_ref_sha1(ref_name, NULL);
1395 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1396 failure |= error("Unable to update %s", ref_name);
1400 static void dump_marks_helper(FILE *f,
1401 uintmax_t base,
1402 struct mark_set *m)
1404 uintmax_t k;
1405 if (m->shift) {
1406 for (k = 0; k < 1024; k++) {
1407 if (m->data.sets[k])
1408 dump_marks_helper(f, (base + k) << m->shift,
1409 m->data.sets[k]);
1411 } else {
1412 for (k = 0; k < 1024; k++) {
1413 if (m->data.marked[k])
1414 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1415 sha1_to_hex(m->data.marked[k]->sha1));
1420 static void dump_marks(void)
1422 static struct lock_file mark_lock;
1423 int mark_fd;
1424 FILE *f;
1426 if (!mark_file)
1427 return;
1429 mark_fd = hold_lock_file_for_update(&mark_lock, mark_file, 0);
1430 if (mark_fd < 0) {
1431 failure |= error("Unable to write marks file %s: %s",
1432 mark_file, strerror(errno));
1433 return;
1436 f = fdopen(mark_fd, "w");
1437 if (!f) {
1438 rollback_lock_file(&mark_lock);
1439 failure |= error("Unable to write marks file %s: %s",
1440 mark_file, strerror(errno));
1441 return;
1444 dump_marks_helper(f, 0, marks);
1445 fclose(f);
1446 if (commit_lock_file(&mark_lock))
1447 failure |= error("Unable to write marks file %s: %s",
1448 mark_file, strerror(errno));
1451 static void read_next_command(void)
1453 read_line(&command_buf, stdin, '\n');
1456 static void cmd_mark(void)
1458 if (!prefixcmp(command_buf.buf, "mark :")) {
1459 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1460 read_next_command();
1462 else
1463 next_mark = 0;
1466 static void *cmd_data (size_t *size)
1468 size_t length;
1469 char *buffer;
1471 if (prefixcmp(command_buf.buf, "data "))
1472 die("Expected 'data n' command, found: %s", command_buf.buf);
1474 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1475 char *term = xstrdup(command_buf.buf + 5 + 2);
1476 size_t sz = 8192, term_len = command_buf.len - 5 - 2;
1477 length = 0;
1478 buffer = xmalloc(sz);
1479 for (;;) {
1480 read_next_command();
1481 if (command_buf.eof)
1482 die("EOF in data (terminator '%s' not found)", term);
1483 if (term_len == command_buf.len
1484 && !strcmp(term, command_buf.buf))
1485 break;
1486 if (sz < (length + command_buf.len)) {
1487 sz = sz * 3 / 2 + 16;
1488 if (sz < (length + command_buf.len))
1489 sz = length + command_buf.len;
1490 buffer = xrealloc(buffer, sz);
1492 memcpy(buffer + length,
1493 command_buf.buf,
1494 command_buf.len - 1);
1495 length += command_buf.len - 1;
1496 buffer[length++] = '\n';
1498 free(term);
1500 else {
1501 size_t n = 0;
1502 length = strtoul(command_buf.buf + 5, NULL, 10);
1503 buffer = xmalloc(length);
1504 while (n < length) {
1505 size_t s = fread(buffer + n, 1, length - n, stdin);
1506 if (!s && feof(stdin))
1507 die("EOF in data (%lu bytes remaining)",
1508 (unsigned long)(length - n));
1509 n += s;
1513 if (fgetc(stdin) != '\n')
1514 die("An lf did not trail the binary data as expected.");
1516 *size = length;
1517 return buffer;
1520 static int validate_raw_date(const char *src, char *result, int maxlen)
1522 const char *orig_src = src;
1523 char *endp, sign;
1525 strtoul(src, &endp, 10);
1526 if (endp == src || *endp != ' ')
1527 return -1;
1529 src = endp + 1;
1530 if (*src != '-' && *src != '+')
1531 return -1;
1532 sign = *src;
1534 strtoul(src + 1, &endp, 10);
1535 if (endp == src || *endp || (endp - orig_src) >= maxlen)
1536 return -1;
1538 strcpy(result, orig_src);
1539 return 0;
1542 static char *parse_ident(const char *buf)
1544 const char *gt;
1545 size_t name_len;
1546 char *ident;
1548 gt = strrchr(buf, '>');
1549 if (!gt)
1550 die("Missing > in ident string: %s", buf);
1551 gt++;
1552 if (*gt != ' ')
1553 die("Missing space after > in ident string: %s", buf);
1554 gt++;
1555 name_len = gt - buf;
1556 ident = xmalloc(name_len + 24);
1557 strncpy(ident, buf, name_len);
1559 switch (whenspec) {
1560 case WHENSPEC_RAW:
1561 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1562 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1563 break;
1564 case WHENSPEC_RFC2822:
1565 if (parse_date(gt, ident + name_len, 24) < 0)
1566 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1567 break;
1568 case WHENSPEC_NOW:
1569 if (strcmp("now", gt))
1570 die("Date in ident must be 'now': %s", buf);
1571 datestamp(ident + name_len, 24);
1572 break;
1575 return ident;
1578 static void cmd_new_blob(void)
1580 size_t l;
1581 void *d;
1583 read_next_command();
1584 cmd_mark();
1585 d = cmd_data(&l);
1587 if (store_object(OBJ_BLOB, d, l, &last_blob, NULL, next_mark))
1588 free(d);
1591 static void unload_one_branch(void)
1593 while (cur_active_branches
1594 && cur_active_branches >= max_active_branches) {
1595 uintmax_t min_commit = ULONG_MAX;
1596 struct branch *e, *l = NULL, *p = NULL;
1598 for (e = active_branches; e; e = e->active_next_branch) {
1599 if (e->last_commit < min_commit) {
1600 p = l;
1601 min_commit = e->last_commit;
1603 l = e;
1606 if (p) {
1607 e = p->active_next_branch;
1608 p->active_next_branch = e->active_next_branch;
1609 } else {
1610 e = active_branches;
1611 active_branches = e->active_next_branch;
1613 e->active = 0;
1614 e->active_next_branch = NULL;
1615 if (e->branch_tree.tree) {
1616 release_tree_content_recursive(e->branch_tree.tree);
1617 e->branch_tree.tree = NULL;
1619 cur_active_branches--;
1623 static void load_branch(struct branch *b)
1625 load_tree(&b->branch_tree);
1626 if (!b->active) {
1627 b->active = 1;
1628 b->active_next_branch = active_branches;
1629 active_branches = b;
1630 cur_active_branches++;
1631 branch_load_count++;
1635 static void file_change_m(struct branch *b)
1637 const char *p = command_buf.buf + 2;
1638 char *p_uq;
1639 const char *endp;
1640 struct object_entry *oe = oe;
1641 unsigned char sha1[20];
1642 uint16_t mode, inline_data = 0;
1644 p = get_mode(p, &mode);
1645 if (!p)
1646 die("Corrupt mode: %s", command_buf.buf);
1647 switch (mode) {
1648 case S_IFREG | 0644:
1649 case S_IFREG | 0755:
1650 case S_IFLNK:
1651 case 0644:
1652 case 0755:
1653 /* ok */
1654 break;
1655 default:
1656 die("Corrupt mode: %s", command_buf.buf);
1659 if (*p == ':') {
1660 char *x;
1661 oe = find_mark(strtoumax(p + 1, &x, 10));
1662 hashcpy(sha1, oe->sha1);
1663 p = x;
1664 } else if (!prefixcmp(p, "inline")) {
1665 inline_data = 1;
1666 p += 6;
1667 } else {
1668 if (get_sha1_hex(p, sha1))
1669 die("Invalid SHA1: %s", command_buf.buf);
1670 oe = find_object(sha1);
1671 p += 40;
1673 if (*p++ != ' ')
1674 die("Missing space after SHA1: %s", command_buf.buf);
1676 p_uq = unquote_c_style(p, &endp);
1677 if (p_uq) {
1678 if (*endp)
1679 die("Garbage after path in: %s", command_buf.buf);
1680 p = p_uq;
1683 if (inline_data) {
1684 size_t l;
1685 void *d;
1686 if (!p_uq)
1687 p = p_uq = xstrdup(p);
1688 read_next_command();
1689 d = cmd_data(&l);
1690 if (store_object(OBJ_BLOB, d, l, &last_blob, sha1, 0))
1691 free(d);
1692 } else if (oe) {
1693 if (oe->type != OBJ_BLOB)
1694 die("Not a blob (actually a %s): %s",
1695 command_buf.buf, typename(oe->type));
1696 } else {
1697 enum object_type type = sha1_object_info(sha1, NULL);
1698 if (type < 0)
1699 die("Blob not found: %s", command_buf.buf);
1700 if (type != OBJ_BLOB)
1701 die("Not a blob (actually a %s): %s",
1702 typename(type), command_buf.buf);
1705 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode, NULL);
1706 free(p_uq);
1709 static void file_change_d(struct branch *b)
1711 const char *p = command_buf.buf + 2;
1712 char *p_uq;
1713 const char *endp;
1715 p_uq = unquote_c_style(p, &endp);
1716 if (p_uq) {
1717 if (*endp)
1718 die("Garbage after path in: %s", command_buf.buf);
1719 p = p_uq;
1721 tree_content_remove(&b->branch_tree, p, NULL);
1722 free(p_uq);
1725 static void file_change_cr(struct branch *b, int rename)
1727 const char *s, *d;
1728 char *s_uq, *d_uq;
1729 const char *endp;
1730 struct tree_entry leaf;
1732 s = command_buf.buf + 2;
1733 s_uq = unquote_c_style(s, &endp);
1734 if (s_uq) {
1735 if (*endp != ' ')
1736 die("Missing space after source: %s", command_buf.buf);
1738 else {
1739 endp = strchr(s, ' ');
1740 if (!endp)
1741 die("Missing space after source: %s", command_buf.buf);
1742 s_uq = xmalloc(endp - s + 1);
1743 memcpy(s_uq, s, endp - s);
1744 s_uq[endp - s] = 0;
1746 s = s_uq;
1748 endp++;
1749 if (!*endp)
1750 die("Missing dest: %s", command_buf.buf);
1752 d = endp;
1753 d_uq = unquote_c_style(d, &endp);
1754 if (d_uq) {
1755 if (*endp)
1756 die("Garbage after dest in: %s", command_buf.buf);
1757 d = d_uq;
1760 memset(&leaf, 0, sizeof(leaf));
1761 if (rename)
1762 tree_content_remove(&b->branch_tree, s, &leaf);
1763 else
1764 tree_content_get(&b->branch_tree, s, &leaf);
1765 if (!leaf.versions[1].mode)
1766 die("Path %s not in branch", s);
1767 tree_content_set(&b->branch_tree, d,
1768 leaf.versions[1].sha1,
1769 leaf.versions[1].mode,
1770 leaf.tree);
1772 free(s_uq);
1773 free(d_uq);
1776 static void file_change_deleteall(struct branch *b)
1778 release_tree_content_recursive(b->branch_tree.tree);
1779 hashclr(b->branch_tree.versions[0].sha1);
1780 hashclr(b->branch_tree.versions[1].sha1);
1781 load_tree(&b->branch_tree);
1784 static void cmd_from_commit(struct branch *b, char *buf, unsigned long size)
1786 if (!buf || size < 46)
1787 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
1788 if (memcmp("tree ", buf, 5)
1789 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1790 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1791 hashcpy(b->branch_tree.versions[0].sha1,
1792 b->branch_tree.versions[1].sha1);
1795 static void cmd_from_existing(struct branch *b)
1797 if (is_null_sha1(b->sha1)) {
1798 hashclr(b->branch_tree.versions[0].sha1);
1799 hashclr(b->branch_tree.versions[1].sha1);
1800 } else {
1801 unsigned long size;
1802 char *buf;
1804 buf = read_object_with_reference(b->sha1,
1805 commit_type, &size, b->sha1);
1806 cmd_from_commit(b, buf, size);
1807 free(buf);
1811 static void cmd_from(struct branch *b)
1813 const char *from;
1814 struct branch *s;
1816 if (prefixcmp(command_buf.buf, "from "))
1817 return;
1819 if (b->branch_tree.tree) {
1820 release_tree_content_recursive(b->branch_tree.tree);
1821 b->branch_tree.tree = NULL;
1824 from = strchr(command_buf.buf, ' ') + 1;
1825 s = lookup_branch(from);
1826 if (b == s)
1827 die("Can't create a branch from itself: %s", b->name);
1828 else if (s) {
1829 unsigned char *t = s->branch_tree.versions[1].sha1;
1830 hashcpy(b->sha1, s->sha1);
1831 hashcpy(b->branch_tree.versions[0].sha1, t);
1832 hashcpy(b->branch_tree.versions[1].sha1, t);
1833 } else if (*from == ':') {
1834 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1835 struct object_entry *oe = find_mark(idnum);
1836 if (oe->type != OBJ_COMMIT)
1837 die("Mark :%" PRIuMAX " not a commit", idnum);
1838 hashcpy(b->sha1, oe->sha1);
1839 if (oe->pack_id != MAX_PACK_ID) {
1840 unsigned long size;
1841 char *buf = gfi_unpack_entry(oe, &size);
1842 cmd_from_commit(b, buf, size);
1843 free(buf);
1844 } else
1845 cmd_from_existing(b);
1846 } else if (!get_sha1(from, b->sha1))
1847 cmd_from_existing(b);
1848 else
1849 die("Invalid ref name or SHA1 expression: %s", from);
1851 read_next_command();
1854 static struct hash_list *cmd_merge(unsigned int *count)
1856 struct hash_list *list = NULL, *n, *e = e;
1857 const char *from;
1858 struct branch *s;
1860 *count = 0;
1861 while (!prefixcmp(command_buf.buf, "merge ")) {
1862 from = strchr(command_buf.buf, ' ') + 1;
1863 n = xmalloc(sizeof(*n));
1864 s = lookup_branch(from);
1865 if (s)
1866 hashcpy(n->sha1, s->sha1);
1867 else if (*from == ':') {
1868 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
1869 struct object_entry *oe = find_mark(idnum);
1870 if (oe->type != OBJ_COMMIT)
1871 die("Mark :%" PRIuMAX " not a commit", idnum);
1872 hashcpy(n->sha1, oe->sha1);
1873 } else if (!get_sha1(from, n->sha1)) {
1874 unsigned long size;
1875 char *buf = read_object_with_reference(n->sha1,
1876 commit_type, &size, n->sha1);
1877 if (!buf || size < 46)
1878 die("Not a valid commit: %s", from);
1879 free(buf);
1880 } else
1881 die("Invalid ref name or SHA1 expression: %s", from);
1883 n->next = NULL;
1884 if (list)
1885 e->next = n;
1886 else
1887 list = n;
1888 e = n;
1889 (*count)++;
1890 read_next_command();
1892 return list;
1895 static void cmd_new_commit(void)
1897 struct branch *b;
1898 void *msg;
1899 size_t msglen;
1900 char *sp;
1901 char *author = NULL;
1902 char *committer = NULL;
1903 struct hash_list *merge_list = NULL;
1904 unsigned int merge_count;
1906 /* Obtain the branch name from the rest of our command */
1907 sp = strchr(command_buf.buf, ' ') + 1;
1908 b = lookup_branch(sp);
1909 if (!b)
1910 b = new_branch(sp);
1912 read_next_command();
1913 cmd_mark();
1914 if (!prefixcmp(command_buf.buf, "author ")) {
1915 author = parse_ident(command_buf.buf + 7);
1916 read_next_command();
1918 if (!prefixcmp(command_buf.buf, "committer ")) {
1919 committer = parse_ident(command_buf.buf + 10);
1920 read_next_command();
1922 if (!committer)
1923 die("Expected committer but didn't get one");
1924 msg = cmd_data(&msglen);
1925 read_next_command();
1926 cmd_from(b);
1927 merge_list = cmd_merge(&merge_count);
1929 /* ensure the branch is active/loaded */
1930 if (!b->branch_tree.tree || !max_active_branches) {
1931 unload_one_branch();
1932 load_branch(b);
1935 /* file_change* */
1936 for (;;) {
1937 if (1 == command_buf.len)
1938 break;
1939 else if (!prefixcmp(command_buf.buf, "M "))
1940 file_change_m(b);
1941 else if (!prefixcmp(command_buf.buf, "D "))
1942 file_change_d(b);
1943 else if (!prefixcmp(command_buf.buf, "R "))
1944 file_change_cr(b, 1);
1945 else if (!prefixcmp(command_buf.buf, "C "))
1946 file_change_cr(b, 0);
1947 else if (!strcmp("deleteall", command_buf.buf))
1948 file_change_deleteall(b);
1949 else
1950 die("Unsupported file_change: %s", command_buf.buf);
1951 read_next_command();
1954 /* build the tree and the commit */
1955 store_tree(&b->branch_tree);
1956 hashcpy(b->branch_tree.versions[0].sha1,
1957 b->branch_tree.versions[1].sha1);
1958 size_dbuf(&new_data, 114 + msglen
1959 + merge_count * 49
1960 + (author
1961 ? strlen(author) + strlen(committer)
1962 : 2 * strlen(committer)));
1963 sp = new_data.buffer;
1964 sp += sprintf(sp, "tree %s\n",
1965 sha1_to_hex(b->branch_tree.versions[1].sha1));
1966 if (!is_null_sha1(b->sha1))
1967 sp += sprintf(sp, "parent %s\n", sha1_to_hex(b->sha1));
1968 while (merge_list) {
1969 struct hash_list *next = merge_list->next;
1970 sp += sprintf(sp, "parent %s\n", sha1_to_hex(merge_list->sha1));
1971 free(merge_list);
1972 merge_list = next;
1974 sp += sprintf(sp, "author %s\n", author ? author : committer);
1975 sp += sprintf(sp, "committer %s\n", committer);
1976 *sp++ = '\n';
1977 memcpy(sp, msg, msglen);
1978 sp += msglen;
1979 free(author);
1980 free(committer);
1981 free(msg);
1983 if (!store_object(OBJ_COMMIT,
1984 new_data.buffer, sp - (char*)new_data.buffer,
1985 NULL, b->sha1, next_mark))
1986 b->pack_id = pack_id;
1987 b->last_commit = object_count_by_type[OBJ_COMMIT];
1990 static void cmd_new_tag(void)
1992 char *sp;
1993 const char *from;
1994 char *tagger;
1995 struct branch *s;
1996 void *msg;
1997 size_t msglen;
1998 struct tag *t;
1999 uintmax_t from_mark = 0;
2000 unsigned char sha1[20];
2002 /* Obtain the new tag name from the rest of our command */
2003 sp = strchr(command_buf.buf, ' ') + 1;
2004 t = pool_alloc(sizeof(struct tag));
2005 t->next_tag = NULL;
2006 t->name = pool_strdup(sp);
2007 if (last_tag)
2008 last_tag->next_tag = t;
2009 else
2010 first_tag = t;
2011 last_tag = t;
2012 read_next_command();
2014 /* from ... */
2015 if (prefixcmp(command_buf.buf, "from "))
2016 die("Expected from command, got %s", command_buf.buf);
2017 from = strchr(command_buf.buf, ' ') + 1;
2018 s = lookup_branch(from);
2019 if (s) {
2020 hashcpy(sha1, s->sha1);
2021 } else if (*from == ':') {
2022 struct object_entry *oe;
2023 from_mark = strtoumax(from + 1, NULL, 10);
2024 oe = find_mark(from_mark);
2025 if (oe->type != OBJ_COMMIT)
2026 die("Mark :%" PRIuMAX " not a commit", from_mark);
2027 hashcpy(sha1, oe->sha1);
2028 } else if (!get_sha1(from, sha1)) {
2029 unsigned long size;
2030 char *buf;
2032 buf = read_object_with_reference(sha1,
2033 commit_type, &size, sha1);
2034 if (!buf || size < 46)
2035 die("Not a valid commit: %s", from);
2036 free(buf);
2037 } else
2038 die("Invalid ref name or SHA1 expression: %s", from);
2039 read_next_command();
2041 /* tagger ... */
2042 if (prefixcmp(command_buf.buf, "tagger "))
2043 die("Expected tagger command, got %s", command_buf.buf);
2044 tagger = parse_ident(command_buf.buf + 7);
2046 /* tag payload/message */
2047 read_next_command();
2048 msg = cmd_data(&msglen);
2050 /* build the tag object */
2051 size_dbuf(&new_data, 67+strlen(t->name)+strlen(tagger)+msglen);
2052 sp = new_data.buffer;
2053 sp += sprintf(sp, "object %s\n", sha1_to_hex(sha1));
2054 sp += sprintf(sp, "type %s\n", commit_type);
2055 sp += sprintf(sp, "tag %s\n", t->name);
2056 sp += sprintf(sp, "tagger %s\n", tagger);
2057 *sp++ = '\n';
2058 memcpy(sp, msg, msglen);
2059 sp += msglen;
2060 free(tagger);
2061 free(msg);
2063 if (store_object(OBJ_TAG, new_data.buffer,
2064 sp - (char*)new_data.buffer,
2065 NULL, t->sha1, 0))
2066 t->pack_id = MAX_PACK_ID;
2067 else
2068 t->pack_id = pack_id;
2071 static void cmd_reset_branch(void)
2073 struct branch *b;
2074 char *sp;
2076 /* Obtain the branch name from the rest of our command */
2077 sp = strchr(command_buf.buf, ' ') + 1;
2078 b = lookup_branch(sp);
2079 if (b) {
2080 hashclr(b->sha1);
2081 hashclr(b->branch_tree.versions[0].sha1);
2082 hashclr(b->branch_tree.versions[1].sha1);
2083 if (b->branch_tree.tree) {
2084 release_tree_content_recursive(b->branch_tree.tree);
2085 b->branch_tree.tree = NULL;
2088 else
2089 b = new_branch(sp);
2090 read_next_command();
2091 cmd_from(b);
2094 static void cmd_checkpoint(void)
2096 if (object_count) {
2097 cycle_packfile();
2098 dump_branches();
2099 dump_tags();
2100 dump_marks();
2102 read_next_command();
2105 static void import_marks(const char *input_file)
2107 char line[512];
2108 FILE *f = fopen(input_file, "r");
2109 if (!f)
2110 die("cannot read %s: %s", input_file, strerror(errno));
2111 while (fgets(line, sizeof(line), f)) {
2112 uintmax_t mark;
2113 char *end;
2114 unsigned char sha1[20];
2115 struct object_entry *e;
2117 end = strchr(line, '\n');
2118 if (line[0] != ':' || !end)
2119 die("corrupt mark line: %s", line);
2120 *end = 0;
2121 mark = strtoumax(line + 1, &end, 10);
2122 if (!mark || end == line + 1
2123 || *end != ' ' || get_sha1(end + 1, sha1))
2124 die("corrupt mark line: %s", line);
2125 e = find_object(sha1);
2126 if (!e) {
2127 enum object_type type = sha1_object_info(sha1, NULL);
2128 if (type < 0)
2129 die("object not found: %s", sha1_to_hex(sha1));
2130 e = insert_object(sha1);
2131 e->type = type;
2132 e->pack_id = MAX_PACK_ID;
2133 e->offset = 1; /* just not zero! */
2135 insert_mark(mark, e);
2137 fclose(f);
2140 static const char fast_import_usage[] =
2141 "git-fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2143 int main(int argc, const char **argv)
2145 int i, show_stats = 1;
2147 git_config(git_default_config);
2148 alloc_objects(object_entry_alloc);
2149 strbuf_init(&command_buf);
2150 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2151 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2152 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2153 marks = pool_calloc(1, sizeof(struct mark_set));
2155 for (i = 1; i < argc; i++) {
2156 const char *a = argv[i];
2158 if (*a != '-' || !strcmp(a, "--"))
2159 break;
2160 else if (!prefixcmp(a, "--date-format=")) {
2161 const char *fmt = a + 14;
2162 if (!strcmp(fmt, "raw"))
2163 whenspec = WHENSPEC_RAW;
2164 else if (!strcmp(fmt, "rfc2822"))
2165 whenspec = WHENSPEC_RFC2822;
2166 else if (!strcmp(fmt, "now"))
2167 whenspec = WHENSPEC_NOW;
2168 else
2169 die("unknown --date-format argument %s", fmt);
2171 else if (!prefixcmp(a, "--max-pack-size="))
2172 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
2173 else if (!prefixcmp(a, "--depth="))
2174 max_depth = strtoul(a + 8, NULL, 0);
2175 else if (!prefixcmp(a, "--active-branches="))
2176 max_active_branches = strtoul(a + 18, NULL, 0);
2177 else if (!prefixcmp(a, "--import-marks="))
2178 import_marks(a + 15);
2179 else if (!prefixcmp(a, "--export-marks="))
2180 mark_file = a + 15;
2181 else if (!prefixcmp(a, "--export-pack-edges=")) {
2182 if (pack_edges)
2183 fclose(pack_edges);
2184 pack_edges = fopen(a + 20, "a");
2185 if (!pack_edges)
2186 die("Cannot open %s: %s", a + 20, strerror(errno));
2187 } else if (!strcmp(a, "--force"))
2188 force_update = 1;
2189 else if (!strcmp(a, "--quiet"))
2190 show_stats = 0;
2191 else if (!strcmp(a, "--stats"))
2192 show_stats = 1;
2193 else
2194 die("unknown option %s", a);
2196 if (i != argc)
2197 usage(fast_import_usage);
2199 prepare_packed_git();
2200 start_packfile();
2201 for (;;) {
2202 read_next_command();
2203 if (command_buf.eof)
2204 break;
2205 else if (!strcmp("blob", command_buf.buf))
2206 cmd_new_blob();
2207 else if (!prefixcmp(command_buf.buf, "commit "))
2208 cmd_new_commit();
2209 else if (!prefixcmp(command_buf.buf, "tag "))
2210 cmd_new_tag();
2211 else if (!prefixcmp(command_buf.buf, "reset "))
2212 cmd_reset_branch();
2213 else if (!strcmp("checkpoint", command_buf.buf))
2214 cmd_checkpoint();
2215 else
2216 die("Unsupported command: %s", command_buf.buf);
2218 end_packfile();
2220 dump_branches();
2221 dump_tags();
2222 unkeep_all_packs();
2223 dump_marks();
2225 if (pack_edges)
2226 fclose(pack_edges);
2228 if (show_stats) {
2229 uintmax_t total_count = 0, duplicate_count = 0;
2230 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2231 total_count += object_count_by_type[i];
2232 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2233 duplicate_count += duplicate_count_by_type[i];
2235 fprintf(stderr, "%s statistics:\n", argv[0]);
2236 fprintf(stderr, "---------------------------------------------------------------------\n");
2237 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2238 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2239 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
2240 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
2241 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
2242 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
2243 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2244 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2245 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2246 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2247 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2248 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2249 fprintf(stderr, "---------------------------------------------------------------------\n");
2250 pack_report();
2251 fprintf(stderr, "---------------------------------------------------------------------\n");
2252 fprintf(stderr, "\n");
2255 return failure ? 1 : 0;