Windows ANSI - hook write()
[git/mingw/4msysgit/peterh.git] / fast-import.c
blob271b93cf764ab08e299bf324a4c9bb047e10d46e
1 /*
2 Format of STDIN stream:
4 stream ::= cmd*;
6 cmd ::= new_blob
7 | new_commit
8 | new_tag
9 | reset_branch
10 | checkpoint
11 | progress
14 new_blob ::= 'blob' lf
15 mark?
16 file_content;
17 file_content ::= data;
19 new_commit ::= 'commit' sp ref_str lf
20 mark?
21 ('author' sp name '<' email '>' when lf)?
22 'committer' sp name '<' email '>' when lf
23 commit_msg
24 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
25 ('merge' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)*
26 file_change*
27 lf?;
28 commit_msg ::= data;
30 file_change ::= file_clr
31 | file_del
32 | file_rnm
33 | file_cpy
34 | file_obm
35 | file_inm;
36 file_clr ::= 'deleteall' lf;
37 file_del ::= 'D' sp path_str lf;
38 file_rnm ::= 'R' sp path_str sp path_str lf;
39 file_cpy ::= 'C' sp path_str sp path_str lf;
40 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
41 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
42 data;
44 new_tag ::= 'tag' sp tag_str lf
45 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
46 'tagger' sp name '<' email '>' when lf
47 tag_msg;
48 tag_msg ::= data;
50 reset_branch ::= 'reset' sp ref_str lf
51 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
52 lf?;
54 checkpoint ::= 'checkpoint' lf
55 lf?;
57 progress ::= 'progress' sp not_lf* lf
58 lf?;
60 # note: the first idnum in a stream should be 1 and subsequent
61 # idnums should not have gaps between values as this will cause
62 # the stream parser to reserve space for the gapped values. An
63 # idnum can be updated in the future to a new object by issuing
64 # a new mark directive with the old idnum.
66 mark ::= 'mark' sp idnum lf;
67 data ::= (delimited_data | exact_data)
68 lf?;
70 # note: delim may be any string but must not contain lf.
71 # data_line may contain any data but must not be exactly
72 # delim.
73 delimited_data ::= 'data' sp '<<' delim lf
74 (data_line lf)*
75 delim lf;
77 # note: declen indicates the length of binary_data in bytes.
78 # declen does not include the lf preceeding the binary data.
80 exact_data ::= 'data' sp declen lf
81 binary_data;
83 # note: quoted strings are C-style quoting supporting \c for
84 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
85 # is the signed byte value in octal. Note that the only
86 # characters which must actually be escaped to protect the
87 # stream formatting is: \, " and LF. Otherwise these values
88 # are UTF8.
90 ref_str ::= ref;
91 sha1exp_str ::= sha1exp;
92 tag_str ::= tag;
93 path_str ::= path | '"' quoted(path) '"' ;
94 mode ::= '100644' | '644'
95 | '100755' | '755'
96 | '120000'
99 declen ::= # unsigned 32 bit value, ascii base10 notation;
100 bigint ::= # unsigned integer value, ascii base10 notation;
101 binary_data ::= # file content, not interpreted;
103 when ::= raw_when | rfc2822_when;
104 raw_when ::= ts sp tz;
105 rfc2822_when ::= # Valid RFC 2822 date and time;
107 sp ::= # ASCII space character;
108 lf ::= # ASCII newline (LF) character;
110 # note: a colon (':') must precede the numerical value assigned to
111 # an idnum. This is to distinguish it from a ref or tag name as
112 # GIT does not permit ':' in ref or tag strings.
114 idnum ::= ':' bigint;
115 path ::= # GIT style file path, e.g. "a/b/c";
116 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
117 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
118 sha1exp ::= # Any valid GIT SHA1 expression;
119 hexsha1 ::= # SHA1 in hexadecimal format;
121 # note: name and email are UTF8 strings, however name must not
122 # contain '<' or lf and email must not contain any of the
123 # following: '<', '>', lf.
125 name ::= # valid GIT author/committer name;
126 email ::= # valid GIT author/committer email;
127 ts ::= # time since the epoch in seconds, ascii base10 notation;
128 tz ::= # GIT style timezone;
130 # note: comments may appear anywhere in the input, except
131 # within a data command. Any form of the data command
132 # always escapes the related input from comment processing.
134 # In case it is not clear, the '#' that starts the comment
135 # must be the first character on that the line (an lf have
136 # preceeded it).
138 comment ::= '#' not_lf* lf;
139 not_lf ::= # Any byte that is not ASCII newline (LF);
142 #include "builtin.h"
143 #include "cache.h"
144 #include "object.h"
145 #include "blob.h"
146 #include "tree.h"
147 #include "commit.h"
148 #include "delta.h"
149 #include "pack.h"
150 #include "refs.h"
151 #include "csum-file.h"
152 #include "quote.h"
154 #define PACK_ID_BITS 16
155 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
156 #define DEPTH_BITS 13
157 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
159 struct object_entry
161 struct object_entry *next;
162 uint32_t offset;
163 uint32_t type : TYPE_BITS,
164 pack_id : PACK_ID_BITS,
165 depth : DEPTH_BITS;
166 unsigned char sha1[20];
169 struct object_entry_pool
171 struct object_entry_pool *next_pool;
172 struct object_entry *next_free;
173 struct object_entry *end;
174 struct object_entry entries[FLEX_ARRAY]; /* more */
177 struct mark_set
179 union {
180 struct object_entry *marked[1024];
181 struct mark_set *sets[1024];
182 } data;
183 unsigned int shift;
186 struct last_object
188 struct strbuf data;
189 uint32_t offset;
190 unsigned int depth;
191 unsigned no_swap : 1;
194 struct mem_pool
196 struct mem_pool *next_pool;
197 char *next_free;
198 char *end;
199 uintmax_t space[FLEX_ARRAY]; /* more */
202 struct atom_str
204 struct atom_str *next_atom;
205 unsigned short str_len;
206 char str_dat[FLEX_ARRAY]; /* more */
209 struct tree_content;
210 struct tree_entry
212 struct tree_content *tree;
213 struct atom_str* name;
214 struct tree_entry_ms
216 uint16_t mode;
217 unsigned char sha1[20];
218 } versions[2];
221 struct tree_content
223 unsigned int entry_capacity; /* must match avail_tree_content */
224 unsigned int entry_count;
225 unsigned int delta_depth;
226 struct tree_entry *entries[FLEX_ARRAY]; /* more */
229 struct avail_tree_content
231 unsigned int entry_capacity; /* must match tree_content */
232 struct avail_tree_content *next_avail;
235 struct branch
237 struct branch *table_next_branch;
238 struct branch *active_next_branch;
239 const char *name;
240 struct tree_entry branch_tree;
241 uintmax_t last_commit;
242 unsigned active : 1;
243 unsigned pack_id : PACK_ID_BITS;
244 unsigned char sha1[20];
247 struct tag
249 struct tag *next_tag;
250 const char *name;
251 unsigned int pack_id;
252 unsigned char sha1[20];
255 struct hash_list
257 struct hash_list *next;
258 unsigned char sha1[20];
261 typedef enum {
262 WHENSPEC_RAW = 1,
263 WHENSPEC_RFC2822,
264 WHENSPEC_NOW,
265 } whenspec_type;
267 struct recent_command
269 struct recent_command *prev;
270 struct recent_command *next;
271 char *buf;
274 /* Configured limits on output */
275 static unsigned long max_depth = 10;
276 static off_t max_packsize = (1LL << 32) - 1;
277 static int force_update;
278 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
279 static int pack_compression_seen;
281 /* Stats and misc. counters */
282 static uintmax_t alloc_count;
283 static uintmax_t marks_set_count;
284 static uintmax_t object_count_by_type[1 << TYPE_BITS];
285 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
286 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
287 static unsigned long object_count;
288 static unsigned long branch_count;
289 static unsigned long branch_load_count;
290 static int failure;
291 static FILE *pack_edges;
293 /* Memory pools */
294 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
295 static size_t total_allocd;
296 static struct mem_pool *mem_pool;
298 /* Atom management */
299 static unsigned int atom_table_sz = 4451;
300 static unsigned int atom_cnt;
301 static struct atom_str **atom_table;
303 /* The .pack file being generated */
304 static unsigned int pack_id;
305 static struct packed_git *pack_data;
306 static struct packed_git **all_packs;
307 static unsigned long pack_size;
309 /* Table of objects we've written. */
310 static unsigned int object_entry_alloc = 5000;
311 static struct object_entry_pool *blocks;
312 static struct object_entry *object_table[1 << 16];
313 static struct mark_set *marks;
314 static const char* mark_file;
316 /* Our last blob */
317 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
319 /* Tree management */
320 static unsigned int tree_entry_alloc = 1000;
321 static void *avail_tree_entry;
322 static unsigned int avail_tree_table_sz = 100;
323 static struct avail_tree_content **avail_tree_table;
324 static struct strbuf old_tree = STRBUF_INIT;
325 static struct strbuf new_tree = STRBUF_INIT;
327 /* Branch data */
328 static unsigned long max_active_branches = 5;
329 static unsigned long cur_active_branches;
330 static unsigned long branch_table_sz = 1039;
331 static struct branch **branch_table;
332 static struct branch *active_branches;
334 /* Tag data */
335 static struct tag *first_tag;
336 static struct tag *last_tag;
338 /* Input stream parsing */
339 static whenspec_type whenspec = WHENSPEC_RAW;
340 static struct strbuf command_buf = STRBUF_INIT;
341 static int unread_command_buf;
342 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
343 static struct recent_command *cmd_tail = &cmd_hist;
344 static struct recent_command *rc_free;
345 static unsigned int cmd_save = 100;
346 static uintmax_t next_mark;
347 static struct strbuf new_data = STRBUF_INIT;
349 static void write_branch_report(FILE *rpt, struct branch *b)
351 fprintf(rpt, "%s:\n", b->name);
353 fprintf(rpt, " status :");
354 if (b->active)
355 fputs(" active", rpt);
356 if (b->branch_tree.tree)
357 fputs(" loaded", rpt);
358 if (is_null_sha1(b->branch_tree.versions[1].sha1))
359 fputs(" dirty", rpt);
360 fputc('\n', rpt);
362 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
363 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
364 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
365 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
367 fputs(" last pack : ", rpt);
368 if (b->pack_id < MAX_PACK_ID)
369 fprintf(rpt, "%u", b->pack_id);
370 fputc('\n', rpt);
372 fputc('\n', rpt);
375 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
377 static void write_crash_report(const char *err)
379 char *loc = git_path("fast_import_crash_%d", getpid());
380 FILE *rpt = fopen(loc, "w");
381 struct branch *b;
382 unsigned long lu;
383 struct recent_command *rc;
385 if (!rpt) {
386 error("can't write crash report %s: %s", loc, strerror(errno));
387 return;
390 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
392 fprintf(rpt, "fast-import crash report:\n");
393 fprintf(rpt, " fast-import process: %d\n", getpid());
394 #ifndef __MINGW32__
395 fprintf(rpt, " parent process : %d\n", getppid());
396 #endif
397 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
398 fputc('\n', rpt);
400 fputs("fatal: ", rpt);
401 fputs(err, rpt);
402 fputc('\n', rpt);
404 fputc('\n', rpt);
405 fputs("Most Recent Commands Before Crash\n", rpt);
406 fputs("---------------------------------\n", rpt);
407 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
408 if (rc->next == &cmd_hist)
409 fputs("* ", rpt);
410 else
411 fputs(" ", rpt);
412 fputs(rc->buf, rpt);
413 fputc('\n', rpt);
416 fputc('\n', rpt);
417 fputs("Active Branch LRU\n", rpt);
418 fputs("-----------------\n", rpt);
419 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
420 cur_active_branches,
421 max_active_branches);
422 fputc('\n', rpt);
423 fputs(" pos clock name\n", rpt);
424 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
425 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
426 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
427 ++lu, b->last_commit, b->name);
429 fputc('\n', rpt);
430 fputs("Inactive Branches\n", rpt);
431 fputs("-----------------\n", rpt);
432 for (lu = 0; lu < branch_table_sz; lu++) {
433 for (b = branch_table[lu]; b; b = b->table_next_branch)
434 write_branch_report(rpt, b);
437 if (first_tag) {
438 struct tag *tg;
439 fputc('\n', rpt);
440 fputs("Annotated Tags\n", rpt);
441 fputs("--------------\n", rpt);
442 for (tg = first_tag; tg; tg = tg->next_tag) {
443 fputs(sha1_to_hex(tg->sha1), rpt);
444 fputc(' ', rpt);
445 fputs(tg->name, rpt);
446 fputc('\n', rpt);
450 fputc('\n', rpt);
451 fputs("Marks\n", rpt);
452 fputs("-----\n", rpt);
453 if (mark_file)
454 fprintf(rpt, " exported to %s\n", mark_file);
455 else
456 dump_marks_helper(rpt, 0, marks);
458 fputc('\n', rpt);
459 fputs("-------------------\n", rpt);
460 fputs("END OF CRASH REPORT\n", rpt);
461 fclose(rpt);
464 static void end_packfile(void);
465 static void unkeep_all_packs(void);
466 static void dump_marks(void);
468 static NORETURN void die_nicely(const char *err, va_list params)
470 static int zombie;
471 char message[2 * PATH_MAX];
473 vsnprintf(message, sizeof(message), err, params);
474 fputs("fatal: ", stderr);
475 fputs(message, stderr);
476 fputc('\n', stderr);
478 if (!zombie) {
479 zombie = 1;
480 write_crash_report(message);
481 end_packfile();
482 unkeep_all_packs();
483 dump_marks();
485 exit(128);
488 static void alloc_objects(unsigned int cnt)
490 struct object_entry_pool *b;
492 b = xmalloc(sizeof(struct object_entry_pool)
493 + cnt * sizeof(struct object_entry));
494 b->next_pool = blocks;
495 b->next_free = b->entries;
496 b->end = b->entries + cnt;
497 blocks = b;
498 alloc_count += cnt;
501 static struct object_entry *new_object(unsigned char *sha1)
503 struct object_entry *e;
505 if (blocks->next_free == blocks->end)
506 alloc_objects(object_entry_alloc);
508 e = blocks->next_free++;
509 hashcpy(e->sha1, sha1);
510 return e;
513 static struct object_entry *find_object(unsigned char *sha1)
515 unsigned int h = sha1[0] << 8 | sha1[1];
516 struct object_entry *e;
517 for (e = object_table[h]; e; e = e->next)
518 if (!hashcmp(sha1, e->sha1))
519 return e;
520 return NULL;
523 static struct object_entry *insert_object(unsigned char *sha1)
525 unsigned int h = sha1[0] << 8 | sha1[1];
526 struct object_entry *e = object_table[h];
527 struct object_entry *p = NULL;
529 while (e) {
530 if (!hashcmp(sha1, e->sha1))
531 return e;
532 p = e;
533 e = e->next;
536 e = new_object(sha1);
537 e->next = NULL;
538 e->offset = 0;
539 if (p)
540 p->next = e;
541 else
542 object_table[h] = e;
543 return e;
546 static unsigned int hc_str(const char *s, size_t len)
548 unsigned int r = 0;
549 while (len-- > 0)
550 r = r * 31 + *s++;
551 return r;
554 static void *pool_alloc(size_t len)
556 struct mem_pool *p;
557 void *r;
559 for (p = mem_pool; p; p = p->next_pool)
560 if ((p->end - p->next_free >= len))
561 break;
563 if (!p) {
564 if (len >= (mem_pool_alloc/2)) {
565 total_allocd += len;
566 return xmalloc(len);
568 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
569 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
570 p->next_pool = mem_pool;
571 p->next_free = (char *) p->space;
572 p->end = p->next_free + mem_pool_alloc;
573 mem_pool = p;
576 r = p->next_free;
577 /* round out to a 'uintmax_t' alignment */
578 if (len & (sizeof(uintmax_t) - 1))
579 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
580 p->next_free += len;
581 return r;
584 static void *pool_calloc(size_t count, size_t size)
586 size_t len = count * size;
587 void *r = pool_alloc(len);
588 memset(r, 0, len);
589 return r;
592 static char *pool_strdup(const char *s)
594 char *r = pool_alloc(strlen(s) + 1);
595 strcpy(r, s);
596 return r;
599 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
601 struct mark_set *s = marks;
602 while ((idnum >> s->shift) >= 1024) {
603 s = pool_calloc(1, sizeof(struct mark_set));
604 s->shift = marks->shift + 10;
605 s->data.sets[0] = marks;
606 marks = s;
608 while (s->shift) {
609 uintmax_t i = idnum >> s->shift;
610 idnum -= i << s->shift;
611 if (!s->data.sets[i]) {
612 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
613 s->data.sets[i]->shift = s->shift - 10;
615 s = s->data.sets[i];
617 if (!s->data.marked[idnum])
618 marks_set_count++;
619 s->data.marked[idnum] = oe;
622 static struct object_entry *find_mark(uintmax_t idnum)
624 uintmax_t orig_idnum = idnum;
625 struct mark_set *s = marks;
626 struct object_entry *oe = NULL;
627 if ((idnum >> s->shift) < 1024) {
628 while (s && s->shift) {
629 uintmax_t i = idnum >> s->shift;
630 idnum -= i << s->shift;
631 s = s->data.sets[i];
633 if (s)
634 oe = s->data.marked[idnum];
636 if (!oe)
637 die("mark :%" PRIuMAX " not declared", orig_idnum);
638 return oe;
641 static struct atom_str *to_atom(const char *s, unsigned short len)
643 unsigned int hc = hc_str(s, len) % atom_table_sz;
644 struct atom_str *c;
646 for (c = atom_table[hc]; c; c = c->next_atom)
647 if (c->str_len == len && !strncmp(s, c->str_dat, len))
648 return c;
650 c = pool_alloc(sizeof(struct atom_str) + len + 1);
651 c->str_len = len;
652 strncpy(c->str_dat, s, len);
653 c->str_dat[len] = 0;
654 c->next_atom = atom_table[hc];
655 atom_table[hc] = c;
656 atom_cnt++;
657 return c;
660 static struct branch *lookup_branch(const char *name)
662 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
663 struct branch *b;
665 for (b = branch_table[hc]; b; b = b->table_next_branch)
666 if (!strcmp(name, b->name))
667 return b;
668 return NULL;
671 static struct branch *new_branch(const char *name)
673 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
674 struct branch* b = lookup_branch(name);
676 if (b)
677 die("Invalid attempt to create duplicate branch: %s", name);
678 switch (check_ref_format(name)) {
679 case 0: break; /* its valid */
680 case CHECK_REF_FORMAT_ONELEVEL:
681 break; /* valid, but too few '/', allow anyway */
682 default:
683 die("Branch name doesn't conform to GIT standards: %s", name);
686 b = pool_calloc(1, sizeof(struct branch));
687 b->name = pool_strdup(name);
688 b->table_next_branch = branch_table[hc];
689 b->branch_tree.versions[0].mode = S_IFDIR;
690 b->branch_tree.versions[1].mode = S_IFDIR;
691 b->active = 0;
692 b->pack_id = MAX_PACK_ID;
693 branch_table[hc] = b;
694 branch_count++;
695 return b;
698 static unsigned int hc_entries(unsigned int cnt)
700 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
701 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
704 static struct tree_content *new_tree_content(unsigned int cnt)
706 struct avail_tree_content *f, *l = NULL;
707 struct tree_content *t;
708 unsigned int hc = hc_entries(cnt);
710 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
711 if (f->entry_capacity >= cnt)
712 break;
714 if (f) {
715 if (l)
716 l->next_avail = f->next_avail;
717 else
718 avail_tree_table[hc] = f->next_avail;
719 } else {
720 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
721 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
722 f->entry_capacity = cnt;
725 t = (struct tree_content*)f;
726 t->entry_count = 0;
727 t->delta_depth = 0;
728 return t;
731 static void release_tree_entry(struct tree_entry *e);
732 static void release_tree_content(struct tree_content *t)
734 struct avail_tree_content *f = (struct avail_tree_content*)t;
735 unsigned int hc = hc_entries(f->entry_capacity);
736 f->next_avail = avail_tree_table[hc];
737 avail_tree_table[hc] = f;
740 static void release_tree_content_recursive(struct tree_content *t)
742 unsigned int i;
743 for (i = 0; i < t->entry_count; i++)
744 release_tree_entry(t->entries[i]);
745 release_tree_content(t);
748 static struct tree_content *grow_tree_content(
749 struct tree_content *t,
750 int amt)
752 struct tree_content *r = new_tree_content(t->entry_count + amt);
753 r->entry_count = t->entry_count;
754 r->delta_depth = t->delta_depth;
755 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
756 release_tree_content(t);
757 return r;
760 static struct tree_entry *new_tree_entry(void)
762 struct tree_entry *e;
764 if (!avail_tree_entry) {
765 unsigned int n = tree_entry_alloc;
766 total_allocd += n * sizeof(struct tree_entry);
767 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
768 while (n-- > 1) {
769 *((void**)e) = e + 1;
770 e++;
772 *((void**)e) = NULL;
775 e = avail_tree_entry;
776 avail_tree_entry = *((void**)e);
777 return e;
780 static void release_tree_entry(struct tree_entry *e)
782 if (e->tree)
783 release_tree_content_recursive(e->tree);
784 *((void**)e) = avail_tree_entry;
785 avail_tree_entry = e;
788 static struct tree_content *dup_tree_content(struct tree_content *s)
790 struct tree_content *d;
791 struct tree_entry *a, *b;
792 unsigned int i;
794 if (!s)
795 return NULL;
796 d = new_tree_content(s->entry_count);
797 for (i = 0; i < s->entry_count; i++) {
798 a = s->entries[i];
799 b = new_tree_entry();
800 memcpy(b, a, sizeof(*a));
801 if (a->tree && is_null_sha1(b->versions[1].sha1))
802 b->tree = dup_tree_content(a->tree);
803 else
804 b->tree = NULL;
805 d->entries[i] = b;
807 d->entry_count = s->entry_count;
808 d->delta_depth = s->delta_depth;
810 return d;
813 static void start_packfile(void)
815 static char tmpfile[PATH_MAX];
816 struct packed_git *p;
817 struct pack_header hdr;
818 int pack_fd;
820 snprintf(tmpfile, sizeof(tmpfile),
821 "%s/tmp_pack_XXXXXX", get_object_directory());
822 pack_fd = xmkstemp(tmpfile);
823 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
824 strcpy(p->pack_name, tmpfile);
825 p->pack_fd = pack_fd;
827 hdr.hdr_signature = htonl(PACK_SIGNATURE);
828 hdr.hdr_version = htonl(2);
829 hdr.hdr_entries = 0;
830 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
832 pack_data = p;
833 pack_size = sizeof(hdr);
834 object_count = 0;
836 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
837 all_packs[pack_id] = p;
840 static int oecmp (const void *a_, const void *b_)
842 struct object_entry *a = *((struct object_entry**)a_);
843 struct object_entry *b = *((struct object_entry**)b_);
844 return hashcmp(a->sha1, b->sha1);
847 static char *create_index(void)
849 static char tmpfile[PATH_MAX];
850 SHA_CTX ctx;
851 struct sha1file *f;
852 struct object_entry **idx, **c, **last, *e;
853 struct object_entry_pool *o;
854 uint32_t array[256];
855 int i, idx_fd;
857 /* Build the sorted table of object IDs. */
858 idx = xmalloc(object_count * sizeof(struct object_entry*));
859 c = idx;
860 for (o = blocks; o; o = o->next_pool)
861 for (e = o->next_free; e-- != o->entries;)
862 if (pack_id == e->pack_id)
863 *c++ = e;
864 last = idx + object_count;
865 if (c != last)
866 die("internal consistency error creating the index");
867 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
869 /* Generate the fan-out array. */
870 c = idx;
871 for (i = 0; i < 256; i++) {
872 struct object_entry **next = c;;
873 while (next < last) {
874 if ((*next)->sha1[0] != i)
875 break;
876 next++;
878 array[i] = htonl(next - idx);
879 c = next;
882 snprintf(tmpfile, sizeof(tmpfile),
883 "%s/tmp_idx_XXXXXX", get_object_directory());
884 idx_fd = xmkstemp(tmpfile);
885 f = sha1fd(idx_fd, tmpfile);
886 sha1write(f, array, 256 * sizeof(int));
887 SHA1_Init(&ctx);
888 for (c = idx; c != last; c++) {
889 uint32_t offset = htonl((*c)->offset);
890 sha1write(f, &offset, 4);
891 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
892 SHA1_Update(&ctx, (*c)->sha1, 20);
894 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
895 sha1close(f, NULL, CSUM_FSYNC);
896 free(idx);
897 SHA1_Final(pack_data->sha1, &ctx);
898 return tmpfile;
901 static char *keep_pack(char *curr_index_name)
903 static char name[PATH_MAX];
904 static const char *keep_msg = "fast-import";
905 int keep_fd;
907 chmod(pack_data->pack_name, 0444);
908 chmod(curr_index_name, 0444);
910 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
911 get_object_directory(), sha1_to_hex(pack_data->sha1));
912 keep_fd = open(name, O_RDWR|O_CREAT|O_EXCL, 0600);
913 if (keep_fd < 0)
914 die("cannot create keep file");
915 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
916 if (close(keep_fd))
917 die("failed to write keep file");
919 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
920 get_object_directory(), sha1_to_hex(pack_data->sha1));
921 if (move_temp_to_file(pack_data->pack_name, name))
922 die("cannot store pack file");
924 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
925 get_object_directory(), sha1_to_hex(pack_data->sha1));
926 if (move_temp_to_file(curr_index_name, name))
927 die("cannot store index file");
928 return name;
931 static void unkeep_all_packs(void)
933 static char name[PATH_MAX];
934 int k;
936 for (k = 0; k < pack_id; k++) {
937 struct packed_git *p = all_packs[k];
938 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
939 get_object_directory(), sha1_to_hex(p->sha1));
940 unlink(name);
944 static void end_packfile(void)
946 struct packed_git *old_p = pack_data, *new_p;
948 if (object_count) {
949 char *idx_name;
950 int i;
951 struct branch *b;
952 struct tag *t;
954 close_pack_windows(pack_data);
955 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
956 pack_data->pack_name, object_count);
957 close(pack_data->pack_fd);
958 idx_name = keep_pack(create_index());
960 /* Register the packfile with core git's machinary. */
961 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
962 if (!new_p)
963 die("core git rejected index %s", idx_name);
964 all_packs[pack_id] = new_p;
965 install_packed_git(new_p);
967 /* Print the boundary */
968 if (pack_edges) {
969 fprintf(pack_edges, "%s:", new_p->pack_name);
970 for (i = 0; i < branch_table_sz; i++) {
971 for (b = branch_table[i]; b; b = b->table_next_branch) {
972 if (b->pack_id == pack_id)
973 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
976 for (t = first_tag; t; t = t->next_tag) {
977 if (t->pack_id == pack_id)
978 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
980 fputc('\n', pack_edges);
981 fflush(pack_edges);
984 pack_id++;
986 else
987 unlink(old_p->pack_name);
988 free(old_p);
990 /* We can't carry a delta across packfiles. */
991 strbuf_release(&last_blob.data);
992 last_blob.offset = 0;
993 last_blob.depth = 0;
996 static void cycle_packfile(void)
998 end_packfile();
999 start_packfile();
1002 static size_t encode_header(
1003 enum object_type type,
1004 size_t size,
1005 unsigned char *hdr)
1007 int n = 1;
1008 unsigned char c;
1010 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
1011 die("bad type %d", type);
1013 c = (type << 4) | (size & 15);
1014 size >>= 4;
1015 while (size) {
1016 *hdr++ = c | 0x80;
1017 c = size & 0x7f;
1018 size >>= 7;
1019 n++;
1021 *hdr = c;
1022 return n;
1025 static int store_object(
1026 enum object_type type,
1027 struct strbuf *dat,
1028 struct last_object *last,
1029 unsigned char *sha1out,
1030 uintmax_t mark)
1032 void *out, *delta;
1033 struct object_entry *e;
1034 unsigned char hdr[96];
1035 unsigned char sha1[20];
1036 unsigned long hdrlen, deltalen;
1037 SHA_CTX c;
1038 z_stream s;
1040 hdrlen = sprintf((char*)hdr,"%s %lu", typename(type),
1041 (unsigned long)dat->len) + 1;
1042 SHA1_Init(&c);
1043 SHA1_Update(&c, hdr, hdrlen);
1044 SHA1_Update(&c, dat->buf, dat->len);
1045 SHA1_Final(sha1, &c);
1046 if (sha1out)
1047 hashcpy(sha1out, sha1);
1049 e = insert_object(sha1);
1050 if (mark)
1051 insert_mark(mark, e);
1052 if (e->offset) {
1053 duplicate_count_by_type[type]++;
1054 return 1;
1055 } else if (find_sha1_pack(sha1, packed_git)) {
1056 e->type = type;
1057 e->pack_id = MAX_PACK_ID;
1058 e->offset = 1; /* just not zero! */
1059 duplicate_count_by_type[type]++;
1060 return 1;
1063 if (last && last->data.buf && last->depth < max_depth) {
1064 delta = diff_delta(last->data.buf, last->data.len,
1065 dat->buf, dat->len,
1066 &deltalen, 0);
1067 if (delta && deltalen >= dat->len) {
1068 free(delta);
1069 delta = NULL;
1071 } else
1072 delta = NULL;
1074 memset(&s, 0, sizeof(s));
1075 deflateInit(&s, pack_compression_level);
1076 if (delta) {
1077 s.next_in = delta;
1078 s.avail_in = deltalen;
1079 } else {
1080 s.next_in = (void *)dat->buf;
1081 s.avail_in = dat->len;
1083 s.avail_out = deflateBound(&s, s.avail_in);
1084 s.next_out = out = xmalloc(s.avail_out);
1085 while (deflate(&s, Z_FINISH) == Z_OK)
1086 /* nothing */;
1087 deflateEnd(&s);
1089 /* Determine if we should auto-checkpoint. */
1090 if ((pack_size + 60 + s.total_out) > max_packsize
1091 || (pack_size + 60 + s.total_out) < pack_size) {
1093 /* This new object needs to *not* have the current pack_id. */
1094 e->pack_id = pack_id + 1;
1095 cycle_packfile();
1097 /* We cannot carry a delta into the new pack. */
1098 if (delta) {
1099 free(delta);
1100 delta = NULL;
1102 memset(&s, 0, sizeof(s));
1103 deflateInit(&s, pack_compression_level);
1104 s.next_in = (void *)dat->buf;
1105 s.avail_in = dat->len;
1106 s.avail_out = deflateBound(&s, s.avail_in);
1107 s.next_out = out = xrealloc(out, s.avail_out);
1108 while (deflate(&s, Z_FINISH) == Z_OK)
1109 /* nothing */;
1110 deflateEnd(&s);
1114 e->type = type;
1115 e->pack_id = pack_id;
1116 e->offset = pack_size;
1117 object_count++;
1118 object_count_by_type[type]++;
1120 if (delta) {
1121 unsigned long ofs = e->offset - last->offset;
1122 unsigned pos = sizeof(hdr) - 1;
1124 delta_count_by_type[type]++;
1125 e->depth = last->depth + 1;
1127 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1128 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1129 pack_size += hdrlen;
1131 hdr[pos] = ofs & 127;
1132 while (ofs >>= 7)
1133 hdr[--pos] = 128 | (--ofs & 127);
1134 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
1135 pack_size += sizeof(hdr) - pos;
1136 } else {
1137 e->depth = 0;
1138 hdrlen = encode_header(type, dat->len, hdr);
1139 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1140 pack_size += hdrlen;
1143 write_or_die(pack_data->pack_fd, out, s.total_out);
1144 pack_size += s.total_out;
1146 free(out);
1147 free(delta);
1148 if (last) {
1149 if (last->no_swap) {
1150 last->data = *dat;
1151 } else {
1152 strbuf_swap(&last->data, dat);
1154 last->offset = e->offset;
1155 last->depth = e->depth;
1157 return 0;
1160 /* All calls must be guarded by find_object() or find_mark() to
1161 * ensure the 'struct object_entry' passed was written by this
1162 * process instance. We unpack the entry by the offset, avoiding
1163 * the need for the corresponding .idx file. This unpacking rule
1164 * works because we only use OBJ_REF_DELTA within the packfiles
1165 * created by fast-import.
1167 * oe must not be NULL. Such an oe usually comes from giving
1168 * an unknown SHA-1 to find_object() or an undefined mark to
1169 * find_mark(). Callers must test for this condition and use
1170 * the standard read_sha1_file() when it happens.
1172 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1173 * find_mark(), where the mark was reloaded from an existing marks
1174 * file and is referencing an object that this fast-import process
1175 * instance did not write out to a packfile. Callers must test for
1176 * this condition and use read_sha1_file() instead.
1178 static void *gfi_unpack_entry(
1179 struct object_entry *oe,
1180 unsigned long *sizep)
1182 enum object_type type;
1183 struct packed_git *p = all_packs[oe->pack_id];
1184 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1185 /* The object is stored in the packfile we are writing to
1186 * and we have modified it since the last time we scanned
1187 * back to read a previously written object. If an old
1188 * window covered [p->pack_size, p->pack_size + 20) its
1189 * data is stale and is not valid. Closing all windows
1190 * and updating the packfile length ensures we can read
1191 * the newly written data.
1193 close_pack_windows(p);
1195 /* We have to offer 20 bytes additional on the end of
1196 * the packfile as the core unpacker code assumes the
1197 * footer is present at the file end and must promise
1198 * at least 20 bytes within any window it maps. But
1199 * we don't actually create the footer here.
1201 p->pack_size = pack_size + 20;
1203 return unpack_entry(p, oe->offset, &type, sizep);
1206 static const char *get_mode(const char *str, uint16_t *modep)
1208 unsigned char c;
1209 uint16_t mode = 0;
1211 while ((c = *str++) != ' ') {
1212 if (c < '0' || c > '7')
1213 return NULL;
1214 mode = (mode << 3) + (c - '0');
1216 *modep = mode;
1217 return str;
1220 static void load_tree(struct tree_entry *root)
1222 unsigned char* sha1 = root->versions[1].sha1;
1223 struct object_entry *myoe;
1224 struct tree_content *t;
1225 unsigned long size;
1226 char *buf;
1227 const char *c;
1229 root->tree = t = new_tree_content(8);
1230 if (is_null_sha1(sha1))
1231 return;
1233 myoe = find_object(sha1);
1234 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1235 if (myoe->type != OBJ_TREE)
1236 die("Not a tree: %s", sha1_to_hex(sha1));
1237 t->delta_depth = myoe->depth;
1238 buf = gfi_unpack_entry(myoe, &size);
1239 if (!buf)
1240 die("Can't load tree %s", sha1_to_hex(sha1));
1241 } else {
1242 enum object_type type;
1243 buf = read_sha1_file(sha1, &type, &size);
1244 if (!buf || type != OBJ_TREE)
1245 die("Can't load tree %s", sha1_to_hex(sha1));
1248 c = buf;
1249 while (c != (buf + size)) {
1250 struct tree_entry *e = new_tree_entry();
1252 if (t->entry_count == t->entry_capacity)
1253 root->tree = t = grow_tree_content(t, t->entry_count);
1254 t->entries[t->entry_count++] = e;
1256 e->tree = NULL;
1257 c = get_mode(c, &e->versions[1].mode);
1258 if (!c)
1259 die("Corrupt mode in %s", sha1_to_hex(sha1));
1260 e->versions[0].mode = e->versions[1].mode;
1261 e->name = to_atom(c, strlen(c));
1262 c += e->name->str_len + 1;
1263 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1264 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1265 c += 20;
1267 free(buf);
1270 static int tecmp0 (const void *_a, const void *_b)
1272 struct tree_entry *a = *((struct tree_entry**)_a);
1273 struct tree_entry *b = *((struct tree_entry**)_b);
1274 return base_name_compare(
1275 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1276 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1279 static int tecmp1 (const void *_a, const void *_b)
1281 struct tree_entry *a = *((struct tree_entry**)_a);
1282 struct tree_entry *b = *((struct tree_entry**)_b);
1283 return base_name_compare(
1284 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1285 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1288 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1290 size_t maxlen = 0;
1291 unsigned int i;
1293 if (!v)
1294 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1295 else
1296 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1298 for (i = 0; i < t->entry_count; i++) {
1299 if (t->entries[i]->versions[v].mode)
1300 maxlen += t->entries[i]->name->str_len + 34;
1303 strbuf_reset(b);
1304 strbuf_grow(b, maxlen);
1305 for (i = 0; i < t->entry_count; i++) {
1306 struct tree_entry *e = t->entries[i];
1307 if (!e->versions[v].mode)
1308 continue;
1309 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1310 e->name->str_dat, '\0');
1311 strbuf_add(b, e->versions[v].sha1, 20);
1315 static void store_tree(struct tree_entry *root)
1317 struct tree_content *t = root->tree;
1318 unsigned int i, j, del;
1319 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1320 struct object_entry *le;
1322 if (!is_null_sha1(root->versions[1].sha1))
1323 return;
1325 for (i = 0; i < t->entry_count; i++) {
1326 if (t->entries[i]->tree)
1327 store_tree(t->entries[i]);
1330 le = find_object(root->versions[0].sha1);
1331 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1332 mktree(t, 0, &old_tree);
1333 lo.data = old_tree;
1334 lo.offset = le->offset;
1335 lo.depth = t->delta_depth;
1338 mktree(t, 1, &new_tree);
1339 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1341 t->delta_depth = lo.depth;
1342 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1343 struct tree_entry *e = t->entries[i];
1344 if (e->versions[1].mode) {
1345 e->versions[0].mode = e->versions[1].mode;
1346 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1347 t->entries[j++] = e;
1348 } else {
1349 release_tree_entry(e);
1350 del++;
1353 t->entry_count -= del;
1356 static int tree_content_set(
1357 struct tree_entry *root,
1358 const char *p,
1359 const unsigned char *sha1,
1360 const uint16_t mode,
1361 struct tree_content *subtree)
1363 struct tree_content *t = root->tree;
1364 const char *slash1;
1365 unsigned int i, n;
1366 struct tree_entry *e;
1368 slash1 = strchr(p, '/');
1369 if (slash1)
1370 n = slash1 - p;
1371 else
1372 n = strlen(p);
1373 if (!n)
1374 die("Empty path component found in input");
1375 if (!slash1 && !S_ISDIR(mode) && subtree)
1376 die("Non-directories cannot have subtrees");
1378 for (i = 0; i < t->entry_count; i++) {
1379 e = t->entries[i];
1380 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1381 if (!slash1) {
1382 if (!S_ISDIR(mode)
1383 && e->versions[1].mode == mode
1384 && !hashcmp(e->versions[1].sha1, sha1))
1385 return 0;
1386 e->versions[1].mode = mode;
1387 hashcpy(e->versions[1].sha1, sha1);
1388 if (e->tree)
1389 release_tree_content_recursive(e->tree);
1390 e->tree = subtree;
1391 hashclr(root->versions[1].sha1);
1392 return 1;
1394 if (!S_ISDIR(e->versions[1].mode)) {
1395 e->tree = new_tree_content(8);
1396 e->versions[1].mode = S_IFDIR;
1398 if (!e->tree)
1399 load_tree(e);
1400 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1401 hashclr(root->versions[1].sha1);
1402 return 1;
1404 return 0;
1408 if (t->entry_count == t->entry_capacity)
1409 root->tree = t = grow_tree_content(t, t->entry_count);
1410 e = new_tree_entry();
1411 e->name = to_atom(p, n);
1412 e->versions[0].mode = 0;
1413 hashclr(e->versions[0].sha1);
1414 t->entries[t->entry_count++] = e;
1415 if (slash1) {
1416 e->tree = new_tree_content(8);
1417 e->versions[1].mode = S_IFDIR;
1418 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1419 } else {
1420 e->tree = subtree;
1421 e->versions[1].mode = mode;
1422 hashcpy(e->versions[1].sha1, sha1);
1424 hashclr(root->versions[1].sha1);
1425 return 1;
1428 static int tree_content_remove(
1429 struct tree_entry *root,
1430 const char *p,
1431 struct tree_entry *backup_leaf)
1433 struct tree_content *t = root->tree;
1434 const char *slash1;
1435 unsigned int i, n;
1436 struct tree_entry *e;
1438 slash1 = strchr(p, '/');
1439 if (slash1)
1440 n = slash1 - p;
1441 else
1442 n = strlen(p);
1444 for (i = 0; i < t->entry_count; i++) {
1445 e = t->entries[i];
1446 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1447 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1448 goto del_entry;
1449 if (!e->tree)
1450 load_tree(e);
1451 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1452 for (n = 0; n < e->tree->entry_count; n++) {
1453 if (e->tree->entries[n]->versions[1].mode) {
1454 hashclr(root->versions[1].sha1);
1455 return 1;
1458 backup_leaf = NULL;
1459 goto del_entry;
1461 return 0;
1464 return 0;
1466 del_entry:
1467 if (backup_leaf)
1468 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1469 else if (e->tree)
1470 release_tree_content_recursive(e->tree);
1471 e->tree = NULL;
1472 e->versions[1].mode = 0;
1473 hashclr(e->versions[1].sha1);
1474 hashclr(root->versions[1].sha1);
1475 return 1;
1478 static int tree_content_get(
1479 struct tree_entry *root,
1480 const char *p,
1481 struct tree_entry *leaf)
1483 struct tree_content *t = root->tree;
1484 const char *slash1;
1485 unsigned int i, n;
1486 struct tree_entry *e;
1488 slash1 = strchr(p, '/');
1489 if (slash1)
1490 n = slash1 - p;
1491 else
1492 n = strlen(p);
1494 for (i = 0; i < t->entry_count; i++) {
1495 e = t->entries[i];
1496 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1497 if (!slash1) {
1498 memcpy(leaf, e, sizeof(*leaf));
1499 if (e->tree && is_null_sha1(e->versions[1].sha1))
1500 leaf->tree = dup_tree_content(e->tree);
1501 else
1502 leaf->tree = NULL;
1503 return 1;
1505 if (!S_ISDIR(e->versions[1].mode))
1506 return 0;
1507 if (!e->tree)
1508 load_tree(e);
1509 return tree_content_get(e, slash1 + 1, leaf);
1512 return 0;
1515 static int update_branch(struct branch *b)
1517 static const char *msg = "fast-import";
1518 struct ref_lock *lock;
1519 unsigned char old_sha1[20];
1521 if (is_null_sha1(b->sha1))
1522 return 0;
1523 if (read_ref(b->name, old_sha1))
1524 hashclr(old_sha1);
1525 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1526 if (!lock)
1527 return error("Unable to lock %s", b->name);
1528 if (!force_update && !is_null_sha1(old_sha1)) {
1529 struct commit *old_cmit, *new_cmit;
1531 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1532 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1533 if (!old_cmit || !new_cmit) {
1534 unlock_ref(lock);
1535 return error("Branch %s is missing commits.", b->name);
1538 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1539 unlock_ref(lock);
1540 warning("Not updating %s"
1541 " (new tip %s does not contain %s)",
1542 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1543 return -1;
1546 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1547 return error("Unable to update %s", b->name);
1548 return 0;
1551 static void dump_branches(void)
1553 unsigned int i;
1554 struct branch *b;
1556 for (i = 0; i < branch_table_sz; i++) {
1557 for (b = branch_table[i]; b; b = b->table_next_branch)
1558 failure |= update_branch(b);
1562 static void dump_tags(void)
1564 static const char *msg = "fast-import";
1565 struct tag *t;
1566 struct ref_lock *lock;
1567 char ref_name[PATH_MAX];
1569 for (t = first_tag; t; t = t->next_tag) {
1570 sprintf(ref_name, "tags/%s", t->name);
1571 lock = lock_ref_sha1(ref_name, NULL);
1572 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1573 failure |= error("Unable to update %s", ref_name);
1577 static void dump_marks_helper(FILE *f,
1578 uintmax_t base,
1579 struct mark_set *m)
1581 uintmax_t k;
1582 if (m->shift) {
1583 for (k = 0; k < 1024; k++) {
1584 if (m->data.sets[k])
1585 dump_marks_helper(f, (base + k) << m->shift,
1586 m->data.sets[k]);
1588 } else {
1589 for (k = 0; k < 1024; k++) {
1590 if (m->data.marked[k])
1591 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1592 sha1_to_hex(m->data.marked[k]->sha1));
1597 static void dump_marks(void)
1599 static struct lock_file mark_lock;
1600 int mark_fd;
1601 FILE *f;
1603 if (!mark_file)
1604 return;
1606 mark_fd = hold_lock_file_for_update(&mark_lock, mark_file, 0);
1607 if (mark_fd < 0) {
1608 failure |= error("Unable to write marks file %s: %s",
1609 mark_file, strerror(errno));
1610 return;
1613 f = fdopen(mark_fd, "w");
1614 if (!f) {
1615 int saved_errno = errno;
1616 rollback_lock_file(&mark_lock);
1617 failure |= error("Unable to write marks file %s: %s",
1618 mark_file, strerror(saved_errno));
1619 return;
1623 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1624 * Assign -1 to the lock file descriptor so that commit_lock_file()
1625 * won't try to close() it.
1627 mark_lock.fd = -1;
1629 dump_marks_helper(f, 0, marks);
1630 if (ferror(f) || fclose(f)) {
1631 int saved_errno = errno;
1632 rollback_lock_file(&mark_lock);
1633 failure |= error("Unable to write marks file %s: %s",
1634 mark_file, strerror(saved_errno));
1635 return;
1638 if (commit_lock_file(&mark_lock)) {
1639 int saved_errno = errno;
1640 rollback_lock_file(&mark_lock);
1641 failure |= error("Unable to commit marks file %s: %s",
1642 mark_file, strerror(saved_errno));
1643 return;
1647 static int read_next_command(void)
1649 static int stdin_eof = 0;
1651 if (stdin_eof) {
1652 unread_command_buf = 0;
1653 return EOF;
1656 do {
1657 if (unread_command_buf) {
1658 unread_command_buf = 0;
1659 } else {
1660 struct recent_command *rc;
1662 strbuf_detach(&command_buf, NULL);
1663 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1664 if (stdin_eof)
1665 return EOF;
1667 rc = rc_free;
1668 if (rc)
1669 rc_free = rc->next;
1670 else {
1671 rc = cmd_hist.next;
1672 cmd_hist.next = rc->next;
1673 cmd_hist.next->prev = &cmd_hist;
1674 free(rc->buf);
1677 rc->buf = command_buf.buf;
1678 rc->prev = cmd_tail;
1679 rc->next = cmd_hist.prev;
1680 rc->prev->next = rc;
1681 cmd_tail = rc;
1683 } while (command_buf.buf[0] == '#');
1685 return 0;
1688 static void skip_optional_lf(void)
1690 int term_char = fgetc(stdin);
1691 if (term_char != '\n' && term_char != EOF)
1692 ungetc(term_char, stdin);
1695 static void parse_mark(void)
1697 if (!prefixcmp(command_buf.buf, "mark :")) {
1698 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1699 read_next_command();
1701 else
1702 next_mark = 0;
1705 static void parse_data(struct strbuf *sb)
1707 strbuf_reset(sb);
1709 if (prefixcmp(command_buf.buf, "data "))
1710 die("Expected 'data n' command, found: %s", command_buf.buf);
1712 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1713 char *term = xstrdup(command_buf.buf + 5 + 2);
1714 size_t term_len = command_buf.len - 5 - 2;
1716 strbuf_detach(&command_buf, NULL);
1717 for (;;) {
1718 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1719 die("EOF in data (terminator '%s' not found)", term);
1720 if (term_len == command_buf.len
1721 && !strcmp(term, command_buf.buf))
1722 break;
1723 strbuf_addbuf(sb, &command_buf);
1724 strbuf_addch(sb, '\n');
1726 free(term);
1728 else {
1729 size_t n = 0, length;
1731 length = strtoul(command_buf.buf + 5, NULL, 10);
1733 while (n < length) {
1734 size_t s = strbuf_fread(sb, length - n, stdin);
1735 if (!s && feof(stdin))
1736 die("EOF in data (%lu bytes remaining)",
1737 (unsigned long)(length - n));
1738 n += s;
1742 skip_optional_lf();
1745 static int validate_raw_date(const char *src, char *result, int maxlen)
1747 const char *orig_src = src;
1748 char *endp, sign;
1750 strtoul(src, &endp, 10);
1751 if (endp == src || *endp != ' ')
1752 return -1;
1754 src = endp + 1;
1755 if (*src != '-' && *src != '+')
1756 return -1;
1757 sign = *src;
1759 strtoul(src + 1, &endp, 10);
1760 if (endp == src || *endp || (endp - orig_src) >= maxlen)
1761 return -1;
1763 strcpy(result, orig_src);
1764 return 0;
1767 static char *parse_ident(const char *buf)
1769 const char *gt;
1770 size_t name_len;
1771 char *ident;
1773 gt = strrchr(buf, '>');
1774 if (!gt)
1775 die("Missing > in ident string: %s", buf);
1776 gt++;
1777 if (*gt != ' ')
1778 die("Missing space after > in ident string: %s", buf);
1779 gt++;
1780 name_len = gt - buf;
1781 ident = xmalloc(name_len + 24);
1782 strncpy(ident, buf, name_len);
1784 switch (whenspec) {
1785 case WHENSPEC_RAW:
1786 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1787 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1788 break;
1789 case WHENSPEC_RFC2822:
1790 if (parse_date(gt, ident + name_len, 24) < 0)
1791 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1792 break;
1793 case WHENSPEC_NOW:
1794 if (strcmp("now", gt))
1795 die("Date in ident must be 'now': %s", buf);
1796 datestamp(ident + name_len, 24);
1797 break;
1800 return ident;
1803 static void parse_new_blob(void)
1805 static struct strbuf buf = STRBUF_INIT;
1807 read_next_command();
1808 parse_mark();
1809 parse_data(&buf);
1810 store_object(OBJ_BLOB, &buf, &last_blob, NULL, next_mark);
1813 static void unload_one_branch(void)
1815 while (cur_active_branches
1816 && cur_active_branches >= max_active_branches) {
1817 uintmax_t min_commit = ULONG_MAX;
1818 struct branch *e, *l = NULL, *p = NULL;
1820 for (e = active_branches; e; e = e->active_next_branch) {
1821 if (e->last_commit < min_commit) {
1822 p = l;
1823 min_commit = e->last_commit;
1825 l = e;
1828 if (p) {
1829 e = p->active_next_branch;
1830 p->active_next_branch = e->active_next_branch;
1831 } else {
1832 e = active_branches;
1833 active_branches = e->active_next_branch;
1835 e->active = 0;
1836 e->active_next_branch = NULL;
1837 if (e->branch_tree.tree) {
1838 release_tree_content_recursive(e->branch_tree.tree);
1839 e->branch_tree.tree = NULL;
1841 cur_active_branches--;
1845 static void load_branch(struct branch *b)
1847 load_tree(&b->branch_tree);
1848 if (!b->active) {
1849 b->active = 1;
1850 b->active_next_branch = active_branches;
1851 active_branches = b;
1852 cur_active_branches++;
1853 branch_load_count++;
1857 static void file_change_m(struct branch *b)
1859 const char *p = command_buf.buf + 2;
1860 static struct strbuf uq = STRBUF_INIT;
1861 const char *endp;
1862 struct object_entry *oe = oe;
1863 unsigned char sha1[20];
1864 uint16_t mode, inline_data = 0;
1866 p = get_mode(p, &mode);
1867 if (!p)
1868 die("Corrupt mode: %s", command_buf.buf);
1869 switch (mode) {
1870 case S_IFREG | 0644:
1871 case S_IFREG | 0755:
1872 case S_IFLNK:
1873 case 0644:
1874 case 0755:
1875 /* ok */
1876 break;
1877 default:
1878 die("Corrupt mode: %s", command_buf.buf);
1881 if (*p == ':') {
1882 char *x;
1883 oe = find_mark(strtoumax(p + 1, &x, 10));
1884 hashcpy(sha1, oe->sha1);
1885 p = x;
1886 } else if (!prefixcmp(p, "inline")) {
1887 inline_data = 1;
1888 p += 6;
1889 } else {
1890 if (get_sha1_hex(p, sha1))
1891 die("Invalid SHA1: %s", command_buf.buf);
1892 oe = find_object(sha1);
1893 p += 40;
1895 if (*p++ != ' ')
1896 die("Missing space after SHA1: %s", command_buf.buf);
1898 strbuf_reset(&uq);
1899 if (!unquote_c_style(&uq, p, &endp)) {
1900 if (*endp)
1901 die("Garbage after path in: %s", command_buf.buf);
1902 p = uq.buf;
1905 if (inline_data) {
1906 static struct strbuf buf = STRBUF_INIT;
1908 if (p != uq.buf) {
1909 strbuf_addstr(&uq, p);
1910 p = uq.buf;
1912 read_next_command();
1913 parse_data(&buf);
1914 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
1915 } else if (oe) {
1916 if (oe->type != OBJ_BLOB)
1917 die("Not a blob (actually a %s): %s",
1918 typename(oe->type), command_buf.buf);
1919 } else {
1920 enum object_type type = sha1_object_info(sha1, NULL);
1921 if (type < 0)
1922 die("Blob not found: %s", command_buf.buf);
1923 if (type != OBJ_BLOB)
1924 die("Not a blob (actually a %s): %s",
1925 typename(type), command_buf.buf);
1928 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode, NULL);
1931 static void file_change_d(struct branch *b)
1933 const char *p = command_buf.buf + 2;
1934 static struct strbuf uq = STRBUF_INIT;
1935 const char *endp;
1937 strbuf_reset(&uq);
1938 if (!unquote_c_style(&uq, p, &endp)) {
1939 if (*endp)
1940 die("Garbage after path in: %s", command_buf.buf);
1941 p = uq.buf;
1943 tree_content_remove(&b->branch_tree, p, NULL);
1946 static void file_change_cr(struct branch *b, int rename)
1948 const char *s, *d;
1949 static struct strbuf s_uq = STRBUF_INIT;
1950 static struct strbuf d_uq = STRBUF_INIT;
1951 const char *endp;
1952 struct tree_entry leaf;
1954 s = command_buf.buf + 2;
1955 strbuf_reset(&s_uq);
1956 if (!unquote_c_style(&s_uq, s, &endp)) {
1957 if (*endp != ' ')
1958 die("Missing space after source: %s", command_buf.buf);
1959 } else {
1960 endp = strchr(s, ' ');
1961 if (!endp)
1962 die("Missing space after source: %s", command_buf.buf);
1963 strbuf_add(&s_uq, s, endp - s);
1965 s = s_uq.buf;
1967 endp++;
1968 if (!*endp)
1969 die("Missing dest: %s", command_buf.buf);
1971 d = endp;
1972 strbuf_reset(&d_uq);
1973 if (!unquote_c_style(&d_uq, d, &endp)) {
1974 if (*endp)
1975 die("Garbage after dest in: %s", command_buf.buf);
1976 d = d_uq.buf;
1979 memset(&leaf, 0, sizeof(leaf));
1980 if (rename)
1981 tree_content_remove(&b->branch_tree, s, &leaf);
1982 else
1983 tree_content_get(&b->branch_tree, s, &leaf);
1984 if (!leaf.versions[1].mode)
1985 die("Path %s not in branch", s);
1986 tree_content_set(&b->branch_tree, d,
1987 leaf.versions[1].sha1,
1988 leaf.versions[1].mode,
1989 leaf.tree);
1992 static void file_change_deleteall(struct branch *b)
1994 release_tree_content_recursive(b->branch_tree.tree);
1995 hashclr(b->branch_tree.versions[0].sha1);
1996 hashclr(b->branch_tree.versions[1].sha1);
1997 load_tree(&b->branch_tree);
2000 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2002 if (!buf || size < 46)
2003 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2004 if (memcmp("tree ", buf, 5)
2005 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2006 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2007 hashcpy(b->branch_tree.versions[0].sha1,
2008 b->branch_tree.versions[1].sha1);
2011 static void parse_from_existing(struct branch *b)
2013 if (is_null_sha1(b->sha1)) {
2014 hashclr(b->branch_tree.versions[0].sha1);
2015 hashclr(b->branch_tree.versions[1].sha1);
2016 } else {
2017 unsigned long size;
2018 char *buf;
2020 buf = read_object_with_reference(b->sha1,
2021 commit_type, &size, b->sha1);
2022 parse_from_commit(b, buf, size);
2023 free(buf);
2027 static int parse_from(struct branch *b)
2029 const char *from;
2030 struct branch *s;
2032 if (prefixcmp(command_buf.buf, "from "))
2033 return 0;
2035 if (b->branch_tree.tree) {
2036 release_tree_content_recursive(b->branch_tree.tree);
2037 b->branch_tree.tree = NULL;
2040 from = strchr(command_buf.buf, ' ') + 1;
2041 s = lookup_branch(from);
2042 if (b == s)
2043 die("Can't create a branch from itself: %s", b->name);
2044 else if (s) {
2045 unsigned char *t = s->branch_tree.versions[1].sha1;
2046 hashcpy(b->sha1, s->sha1);
2047 hashcpy(b->branch_tree.versions[0].sha1, t);
2048 hashcpy(b->branch_tree.versions[1].sha1, t);
2049 } else if (*from == ':') {
2050 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2051 struct object_entry *oe = find_mark(idnum);
2052 if (oe->type != OBJ_COMMIT)
2053 die("Mark :%" PRIuMAX " not a commit", idnum);
2054 hashcpy(b->sha1, oe->sha1);
2055 if (oe->pack_id != MAX_PACK_ID) {
2056 unsigned long size;
2057 char *buf = gfi_unpack_entry(oe, &size);
2058 parse_from_commit(b, buf, size);
2059 free(buf);
2060 } else
2061 parse_from_existing(b);
2062 } else if (!get_sha1(from, b->sha1))
2063 parse_from_existing(b);
2064 else
2065 die("Invalid ref name or SHA1 expression: %s", from);
2067 read_next_command();
2068 return 1;
2071 static struct hash_list *parse_merge(unsigned int *count)
2073 struct hash_list *list = NULL, *n, *e = e;
2074 const char *from;
2075 struct branch *s;
2077 *count = 0;
2078 while (!prefixcmp(command_buf.buf, "merge ")) {
2079 from = strchr(command_buf.buf, ' ') + 1;
2080 n = xmalloc(sizeof(*n));
2081 s = lookup_branch(from);
2082 if (s)
2083 hashcpy(n->sha1, s->sha1);
2084 else if (*from == ':') {
2085 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2086 struct object_entry *oe = find_mark(idnum);
2087 if (oe->type != OBJ_COMMIT)
2088 die("Mark :%" PRIuMAX " not a commit", idnum);
2089 hashcpy(n->sha1, oe->sha1);
2090 } else if (!get_sha1(from, n->sha1)) {
2091 unsigned long size;
2092 char *buf = read_object_with_reference(n->sha1,
2093 commit_type, &size, n->sha1);
2094 if (!buf || size < 46)
2095 die("Not a valid commit: %s", from);
2096 free(buf);
2097 } else
2098 die("Invalid ref name or SHA1 expression: %s", from);
2100 n->next = NULL;
2101 if (list)
2102 e->next = n;
2103 else
2104 list = n;
2105 e = n;
2106 (*count)++;
2107 read_next_command();
2109 return list;
2112 static void parse_new_commit(void)
2114 static struct strbuf msg = STRBUF_INIT;
2115 struct branch *b;
2116 char *sp;
2117 char *author = NULL;
2118 char *committer = NULL;
2119 struct hash_list *merge_list = NULL;
2120 unsigned int merge_count;
2122 /* Obtain the branch name from the rest of our command */
2123 sp = strchr(command_buf.buf, ' ') + 1;
2124 b = lookup_branch(sp);
2125 if (!b)
2126 b = new_branch(sp);
2128 read_next_command();
2129 parse_mark();
2130 if (!prefixcmp(command_buf.buf, "author ")) {
2131 author = parse_ident(command_buf.buf + 7);
2132 read_next_command();
2134 if (!prefixcmp(command_buf.buf, "committer ")) {
2135 committer = parse_ident(command_buf.buf + 10);
2136 read_next_command();
2138 if (!committer)
2139 die("Expected committer but didn't get one");
2140 parse_data(&msg);
2141 read_next_command();
2142 parse_from(b);
2143 merge_list = parse_merge(&merge_count);
2145 /* ensure the branch is active/loaded */
2146 if (!b->branch_tree.tree || !max_active_branches) {
2147 unload_one_branch();
2148 load_branch(b);
2151 /* file_change* */
2152 while (command_buf.len > 0) {
2153 if (!prefixcmp(command_buf.buf, "M "))
2154 file_change_m(b);
2155 else if (!prefixcmp(command_buf.buf, "D "))
2156 file_change_d(b);
2157 else if (!prefixcmp(command_buf.buf, "R "))
2158 file_change_cr(b, 1);
2159 else if (!prefixcmp(command_buf.buf, "C "))
2160 file_change_cr(b, 0);
2161 else if (!strcmp("deleteall", command_buf.buf))
2162 file_change_deleteall(b);
2163 else {
2164 unread_command_buf = 1;
2165 break;
2167 if (read_next_command() == EOF)
2168 break;
2171 /* build the tree and the commit */
2172 store_tree(&b->branch_tree);
2173 hashcpy(b->branch_tree.versions[0].sha1,
2174 b->branch_tree.versions[1].sha1);
2176 strbuf_reset(&new_data);
2177 strbuf_addf(&new_data, "tree %s\n",
2178 sha1_to_hex(b->branch_tree.versions[1].sha1));
2179 if (!is_null_sha1(b->sha1))
2180 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2181 while (merge_list) {
2182 struct hash_list *next = merge_list->next;
2183 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2184 free(merge_list);
2185 merge_list = next;
2187 strbuf_addf(&new_data,
2188 "author %s\n"
2189 "committer %s\n"
2190 "\n",
2191 author ? author : committer, committer);
2192 strbuf_addbuf(&new_data, &msg);
2193 free(author);
2194 free(committer);
2196 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2197 b->pack_id = pack_id;
2198 b->last_commit = object_count_by_type[OBJ_COMMIT];
2201 static void parse_new_tag(void)
2203 static struct strbuf msg = STRBUF_INIT;
2204 char *sp;
2205 const char *from;
2206 char *tagger;
2207 struct branch *s;
2208 struct tag *t;
2209 uintmax_t from_mark = 0;
2210 unsigned char sha1[20];
2212 /* Obtain the new tag name from the rest of our command */
2213 sp = strchr(command_buf.buf, ' ') + 1;
2214 t = pool_alloc(sizeof(struct tag));
2215 t->next_tag = NULL;
2216 t->name = pool_strdup(sp);
2217 if (last_tag)
2218 last_tag->next_tag = t;
2219 else
2220 first_tag = t;
2221 last_tag = t;
2222 read_next_command();
2224 /* from ... */
2225 if (prefixcmp(command_buf.buf, "from "))
2226 die("Expected from command, got %s", command_buf.buf);
2227 from = strchr(command_buf.buf, ' ') + 1;
2228 s = lookup_branch(from);
2229 if (s) {
2230 hashcpy(sha1, s->sha1);
2231 } else if (*from == ':') {
2232 struct object_entry *oe;
2233 from_mark = strtoumax(from + 1, NULL, 10);
2234 oe = find_mark(from_mark);
2235 if (oe->type != OBJ_COMMIT)
2236 die("Mark :%" PRIuMAX " not a commit", from_mark);
2237 hashcpy(sha1, oe->sha1);
2238 } else if (!get_sha1(from, sha1)) {
2239 unsigned long size;
2240 char *buf;
2242 buf = read_object_with_reference(sha1,
2243 commit_type, &size, sha1);
2244 if (!buf || size < 46)
2245 die("Not a valid commit: %s", from);
2246 free(buf);
2247 } else
2248 die("Invalid ref name or SHA1 expression: %s", from);
2249 read_next_command();
2251 /* tagger ... */
2252 if (prefixcmp(command_buf.buf, "tagger "))
2253 die("Expected tagger command, got %s", command_buf.buf);
2254 tagger = parse_ident(command_buf.buf + 7);
2256 /* tag payload/message */
2257 read_next_command();
2258 parse_data(&msg);
2260 /* build the tag object */
2261 strbuf_reset(&new_data);
2262 strbuf_addf(&new_data,
2263 "object %s\n"
2264 "type %s\n"
2265 "tag %s\n"
2266 "tagger %s\n"
2267 "\n",
2268 sha1_to_hex(sha1), commit_type, t->name, tagger);
2269 strbuf_addbuf(&new_data, &msg);
2270 free(tagger);
2272 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2273 t->pack_id = MAX_PACK_ID;
2274 else
2275 t->pack_id = pack_id;
2278 static void parse_reset_branch(void)
2280 struct branch *b;
2281 char *sp;
2283 /* Obtain the branch name from the rest of our command */
2284 sp = strchr(command_buf.buf, ' ') + 1;
2285 b = lookup_branch(sp);
2286 if (b) {
2287 hashclr(b->sha1);
2288 hashclr(b->branch_tree.versions[0].sha1);
2289 hashclr(b->branch_tree.versions[1].sha1);
2290 if (b->branch_tree.tree) {
2291 release_tree_content_recursive(b->branch_tree.tree);
2292 b->branch_tree.tree = NULL;
2295 else
2296 b = new_branch(sp);
2297 read_next_command();
2298 parse_from(b);
2299 if (command_buf.len > 0)
2300 unread_command_buf = 1;
2303 static void parse_checkpoint(void)
2305 if (object_count) {
2306 cycle_packfile();
2307 dump_branches();
2308 dump_tags();
2309 dump_marks();
2311 skip_optional_lf();
2314 static void parse_progress(void)
2316 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2317 fputc('\n', stdout);
2318 fflush(stdout);
2319 skip_optional_lf();
2322 static void import_marks(const char *input_file)
2324 char line[512];
2325 FILE *f = fopen(input_file, "r");
2326 if (!f)
2327 die("cannot read %s: %s", input_file, strerror(errno));
2328 while (fgets(line, sizeof(line), f)) {
2329 uintmax_t mark;
2330 char *end;
2331 unsigned char sha1[20];
2332 struct object_entry *e;
2334 end = strchr(line, '\n');
2335 if (line[0] != ':' || !end)
2336 die("corrupt mark line: %s", line);
2337 *end = 0;
2338 mark = strtoumax(line + 1, &end, 10);
2339 if (!mark || end == line + 1
2340 || *end != ' ' || get_sha1(end + 1, sha1))
2341 die("corrupt mark line: %s", line);
2342 e = find_object(sha1);
2343 if (!e) {
2344 enum object_type type = sha1_object_info(sha1, NULL);
2345 if (type < 0)
2346 die("object not found: %s", sha1_to_hex(sha1));
2347 e = insert_object(sha1);
2348 e->type = type;
2349 e->pack_id = MAX_PACK_ID;
2350 e->offset = 1; /* just not zero! */
2352 insert_mark(mark, e);
2354 fclose(f);
2357 static int git_pack_config(const char *k, const char *v, void *cb)
2359 if (!strcmp(k, "pack.depth")) {
2360 max_depth = git_config_int(k, v);
2361 if (max_depth > MAX_DEPTH)
2362 max_depth = MAX_DEPTH;
2363 return 0;
2365 if (!strcmp(k, "pack.compression")) {
2366 int level = git_config_int(k, v);
2367 if (level == -1)
2368 level = Z_DEFAULT_COMPRESSION;
2369 else if (level < 0 || level > Z_BEST_COMPRESSION)
2370 die("bad pack compression level %d", level);
2371 pack_compression_level = level;
2372 pack_compression_seen = 1;
2373 return 0;
2375 return git_default_config(k, v, cb);
2378 static const char fast_import_usage[] =
2379 "git-fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2381 int main(int argc, const char **argv)
2383 unsigned int i, show_stats = 1;
2385 setup_git_directory();
2386 git_config(git_pack_config, NULL);
2387 if (!pack_compression_seen && core_compression_seen)
2388 pack_compression_level = core_compression_level;
2390 alloc_objects(object_entry_alloc);
2391 strbuf_init(&command_buf, 0);
2392 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2393 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2394 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2395 marks = pool_calloc(1, sizeof(struct mark_set));
2397 for (i = 1; i < argc; i++) {
2398 const char *a = argv[i];
2400 if (*a != '-' || !strcmp(a, "--"))
2401 break;
2402 else if (!prefixcmp(a, "--date-format=")) {
2403 const char *fmt = a + 14;
2404 if (!strcmp(fmt, "raw"))
2405 whenspec = WHENSPEC_RAW;
2406 else if (!strcmp(fmt, "rfc2822"))
2407 whenspec = WHENSPEC_RFC2822;
2408 else if (!strcmp(fmt, "now"))
2409 whenspec = WHENSPEC_NOW;
2410 else
2411 die("unknown --date-format argument %s", fmt);
2413 else if (!prefixcmp(a, "--max-pack-size="))
2414 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
2415 else if (!prefixcmp(a, "--depth=")) {
2416 max_depth = strtoul(a + 8, NULL, 0);
2417 if (max_depth > MAX_DEPTH)
2418 die("--depth cannot exceed %u", MAX_DEPTH);
2420 else if (!prefixcmp(a, "--active-branches="))
2421 max_active_branches = strtoul(a + 18, NULL, 0);
2422 else if (!prefixcmp(a, "--import-marks="))
2423 import_marks(a + 15);
2424 else if (!prefixcmp(a, "--export-marks="))
2425 mark_file = a + 15;
2426 else if (!prefixcmp(a, "--export-pack-edges=")) {
2427 if (pack_edges)
2428 fclose(pack_edges);
2429 pack_edges = fopen(a + 20, "a");
2430 if (!pack_edges)
2431 die("Cannot open %s: %s", a + 20, strerror(errno));
2432 } else if (!strcmp(a, "--force"))
2433 force_update = 1;
2434 else if (!strcmp(a, "--quiet"))
2435 show_stats = 0;
2436 else if (!strcmp(a, "--stats"))
2437 show_stats = 1;
2438 else
2439 die("unknown option %s", a);
2441 if (i != argc)
2442 usage(fast_import_usage);
2444 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2445 for (i = 0; i < (cmd_save - 1); i++)
2446 rc_free[i].next = &rc_free[i + 1];
2447 rc_free[cmd_save - 1].next = NULL;
2449 prepare_packed_git();
2450 start_packfile();
2451 set_die_routine(die_nicely);
2452 while (read_next_command() != EOF) {
2453 if (!strcmp("blob", command_buf.buf))
2454 parse_new_blob();
2455 else if (!prefixcmp(command_buf.buf, "commit "))
2456 parse_new_commit();
2457 else if (!prefixcmp(command_buf.buf, "tag "))
2458 parse_new_tag();
2459 else if (!prefixcmp(command_buf.buf, "reset "))
2460 parse_reset_branch();
2461 else if (!strcmp("checkpoint", command_buf.buf))
2462 parse_checkpoint();
2463 else if (!prefixcmp(command_buf.buf, "progress "))
2464 parse_progress();
2465 else
2466 die("Unsupported command: %s", command_buf.buf);
2468 end_packfile();
2470 dump_branches();
2471 dump_tags();
2472 unkeep_all_packs();
2473 dump_marks();
2475 if (pack_edges)
2476 fclose(pack_edges);
2478 if (show_stats) {
2479 uintmax_t total_count = 0, duplicate_count = 0;
2480 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2481 total_count += object_count_by_type[i];
2482 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2483 duplicate_count += duplicate_count_by_type[i];
2485 fprintf(stderr, "%s statistics:\n", argv[0]);
2486 fprintf(stderr, "---------------------------------------------------------------------\n");
2487 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2488 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2489 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]);
2490 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]);
2491 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]);
2492 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]);
2493 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2494 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2495 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2496 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2497 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2498 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2499 fprintf(stderr, "---------------------------------------------------------------------\n");
2500 pack_report();
2501 fprintf(stderr, "---------------------------------------------------------------------\n");
2502 fprintf(stderr, "\n");
2505 return failure ? 1 : 0;