Provide git_config with a callback-data parameter
[git/git.git] / fast-import.c
1 /*
2 Format of STDIN stream:
3
4 stream ::= cmd*;
5
6 cmd ::= new_blob
7 | new_commit
8 | new_tag
9 | reset_branch
10 | checkpoint
11 | progress
12 ;
13
14 new_blob ::= 'blob' lf
15 mark?
16 file_content;
17 file_content ::= data;
18
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;
29
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;
43
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;
49
50 reset_branch ::= 'reset' sp ref_str lf
51 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
52 lf?;
53
54 checkpoint ::= 'checkpoint' lf
55 lf?;
56
57 progress ::= 'progress' sp not_lf* lf
58 lf?;
59
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.
65 #
66 mark ::= 'mark' sp idnum lf;
67 data ::= (delimited_data | exact_data)
68 lf?;
69
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;
76
77 # note: declen indicates the length of binary_data in bytes.
78 # declen does not include the lf preceeding the binary data.
79 #
80 exact_data ::= 'data' sp declen lf
81 binary_data;
82
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.
89 #
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'
97 ;
98
99 declen ::= # unsigned 32 bit value, ascii base10 notation;
100 bigint ::= # unsigned integer value, ascii base10 notation;
101 binary_data ::= # file content, not interpreted;
102
103 when ::= raw_when | rfc2822_when;
104 raw_when ::= ts sp tz;
105 rfc2822_when ::= # Valid RFC 2822 date and time;
106
107 sp ::= # ASCII space character;
108 lf ::= # ASCII newline (LF) character;
109
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.
113 #
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;
120
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.
124 #
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;
129
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.
133 #
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).
137 #
138 comment ::= '#' not_lf* lf;
139 not_lf ::= # Any byte that is not ASCII newline (LF);
140 */
141
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"
153
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)
158
159 struct object_entry
160 {
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];
167 };
168
169 struct object_entry_pool
170 {
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 */
175 };
176
177 struct mark_set
178 {
179 union {
180 struct object_entry *marked[1024];
181 struct mark_set *sets[1024];
182 } data;
183 unsigned int shift;
184 };
185
186 struct last_object
187 {
188 struct strbuf data;
189 uint32_t offset;
190 unsigned int depth;
191 unsigned no_swap : 1;
192 };
193
194 struct mem_pool
195 {
196 struct mem_pool *next_pool;
197 char *next_free;
198 char *end;
199 uintmax_t space[FLEX_ARRAY]; /* more */
200 };
201
202 struct atom_str
203 {
204 struct atom_str *next_atom;
205 unsigned short str_len;
206 char str_dat[FLEX_ARRAY]; /* more */
207 };
208
209 struct tree_content;
210 struct tree_entry
211 {
212 struct tree_content *tree;
213 struct atom_str* name;
214 struct tree_entry_ms
215 {
216 uint16_t mode;
217 unsigned char sha1[20];
218 } versions[2];
219 };
220
221 struct tree_content
222 {
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 */
227 };
228
229 struct avail_tree_content
230 {
231 unsigned int entry_capacity; /* must match tree_content */
232 struct avail_tree_content *next_avail;
233 };
234
235 struct branch
236 {
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];
245 };
246
247 struct tag
248 {
249 struct tag *next_tag;
250 const char *name;
251 unsigned int pack_id;
252 unsigned char sha1[20];
253 };
254
255 struct hash_list
256 {
257 struct hash_list *next;
258 unsigned char sha1[20];
259 };
260
261 typedef enum {
262 WHENSPEC_RAW = 1,
263 WHENSPEC_RFC2822,
264 WHENSPEC_NOW,
265 } whenspec_type;
266
267 struct recent_command
268 {
269 struct recent_command *prev;
270 struct recent_command *next;
271 char *buf;
272 };
273
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;
280
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;
292
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;
297
298 /* Atom management */
299 static unsigned int atom_table_sz = 4451;
300 static unsigned int atom_cnt;
301 static struct atom_str **atom_table;
302
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;
308
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;
315
316 /* Our last blob */
317 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
318
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;
326
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;
333
334 /* Tag data */
335 static struct tag *first_tag;
336 static struct tag *last_tag;
337
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;
348
349 static void write_branch_report(FILE *rpt, struct branch *b)
350 {
351 fprintf(rpt, "%s:\n", b->name);
352
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);
361
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);
366
367 fputs(" last pack : ", rpt);
368 if (b->pack_id < MAX_PACK_ID)
369 fprintf(rpt, "%u", b->pack_id);
370 fputc('\n', rpt);
371
372 fputc('\n', rpt);
373 }
374
375 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
376
377 static void write_crash_report(const char *err)
378 {
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;
384
385 if (!rpt) {
386 error("can't write crash report %s: %s", loc, strerror(errno));
387 return;
388 }
389
390 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
391
392 fprintf(rpt, "fast-import crash report:\n");
393 fprintf(rpt, " fast-import process: %d\n", getpid());
394 fprintf(rpt, " parent process : %d\n", getppid());
395 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
396 fputc('\n', rpt);
397
398 fputs("fatal: ", rpt);
399 fputs(err, rpt);
400 fputc('\n', rpt);
401
402 fputc('\n', rpt);
403 fputs("Most Recent Commands Before Crash\n", rpt);
404 fputs("---------------------------------\n", rpt);
405 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
406 if (rc->next == &cmd_hist)
407 fputs("* ", rpt);
408 else
409 fputs(" ", rpt);
410 fputs(rc->buf, rpt);
411 fputc('\n', rpt);
412 }
413
414 fputc('\n', rpt);
415 fputs("Active Branch LRU\n", rpt);
416 fputs("-----------------\n", rpt);
417 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
418 cur_active_branches,
419 max_active_branches);
420 fputc('\n', rpt);
421 fputs(" pos clock name\n", rpt);
422 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
423 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
424 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
425 ++lu, b->last_commit, b->name);
426
427 fputc('\n', rpt);
428 fputs("Inactive Branches\n", rpt);
429 fputs("-----------------\n", rpt);
430 for (lu = 0; lu < branch_table_sz; lu++) {
431 for (b = branch_table[lu]; b; b = b->table_next_branch)
432 write_branch_report(rpt, b);
433 }
434
435 if (first_tag) {
436 struct tag *tg;
437 fputc('\n', rpt);
438 fputs("Annotated Tags\n", rpt);
439 fputs("--------------\n", rpt);
440 for (tg = first_tag; tg; tg = tg->next_tag) {
441 fputs(sha1_to_hex(tg->sha1), rpt);
442 fputc(' ', rpt);
443 fputs(tg->name, rpt);
444 fputc('\n', rpt);
445 }
446 }
447
448 fputc('\n', rpt);
449 fputs("Marks\n", rpt);
450 fputs("-----\n", rpt);
451 if (mark_file)
452 fprintf(rpt, " exported to %s\n", mark_file);
453 else
454 dump_marks_helper(rpt, 0, marks);
455
456 fputc('\n', rpt);
457 fputs("-------------------\n", rpt);
458 fputs("END OF CRASH REPORT\n", rpt);
459 fclose(rpt);
460 }
461
462 static void end_packfile(void);
463 static void unkeep_all_packs(void);
464 static void dump_marks(void);
465
466 static NORETURN void die_nicely(const char *err, va_list params)
467 {
468 static int zombie;
469 char message[2 * PATH_MAX];
470
471 vsnprintf(message, sizeof(message), err, params);
472 fputs("fatal: ", stderr);
473 fputs(message, stderr);
474 fputc('\n', stderr);
475
476 if (!zombie) {
477 zombie = 1;
478 write_crash_report(message);
479 end_packfile();
480 unkeep_all_packs();
481 dump_marks();
482 }
483 exit(128);
484 }
485
486 static void alloc_objects(unsigned int cnt)
487 {
488 struct object_entry_pool *b;
489
490 b = xmalloc(sizeof(struct object_entry_pool)
491 + cnt * sizeof(struct object_entry));
492 b->next_pool = blocks;
493 b->next_free = b->entries;
494 b->end = b->entries + cnt;
495 blocks = b;
496 alloc_count += cnt;
497 }
498
499 static struct object_entry *new_object(unsigned char *sha1)
500 {
501 struct object_entry *e;
502
503 if (blocks->next_free == blocks->end)
504 alloc_objects(object_entry_alloc);
505
506 e = blocks->next_free++;
507 hashcpy(e->sha1, sha1);
508 return e;
509 }
510
511 static struct object_entry *find_object(unsigned char *sha1)
512 {
513 unsigned int h = sha1[0] << 8 | sha1[1];
514 struct object_entry *e;
515 for (e = object_table[h]; e; e = e->next)
516 if (!hashcmp(sha1, e->sha1))
517 return e;
518 return NULL;
519 }
520
521 static struct object_entry *insert_object(unsigned char *sha1)
522 {
523 unsigned int h = sha1[0] << 8 | sha1[1];
524 struct object_entry *e = object_table[h];
525 struct object_entry *p = NULL;
526
527 while (e) {
528 if (!hashcmp(sha1, e->sha1))
529 return e;
530 p = e;
531 e = e->next;
532 }
533
534 e = new_object(sha1);
535 e->next = NULL;
536 e->offset = 0;
537 if (p)
538 p->next = e;
539 else
540 object_table[h] = e;
541 return e;
542 }
543
544 static unsigned int hc_str(const char *s, size_t len)
545 {
546 unsigned int r = 0;
547 while (len-- > 0)
548 r = r * 31 + *s++;
549 return r;
550 }
551
552 static void *pool_alloc(size_t len)
553 {
554 struct mem_pool *p;
555 void *r;
556
557 for (p = mem_pool; p; p = p->next_pool)
558 if ((p->end - p->next_free >= len))
559 break;
560
561 if (!p) {
562 if (len >= (mem_pool_alloc/2)) {
563 total_allocd += len;
564 return xmalloc(len);
565 }
566 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
567 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
568 p->next_pool = mem_pool;
569 p->next_free = (char *) p->space;
570 p->end = p->next_free + mem_pool_alloc;
571 mem_pool = p;
572 }
573
574 r = p->next_free;
575 /* round out to a 'uintmax_t' alignment */
576 if (len & (sizeof(uintmax_t) - 1))
577 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
578 p->next_free += len;
579 return r;
580 }
581
582 static void *pool_calloc(size_t count, size_t size)
583 {
584 size_t len = count * size;
585 void *r = pool_alloc(len);
586 memset(r, 0, len);
587 return r;
588 }
589
590 static char *pool_strdup(const char *s)
591 {
592 char *r = pool_alloc(strlen(s) + 1);
593 strcpy(r, s);
594 return r;
595 }
596
597 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
598 {
599 struct mark_set *s = marks;
600 while ((idnum >> s->shift) >= 1024) {
601 s = pool_calloc(1, sizeof(struct mark_set));
602 s->shift = marks->shift + 10;
603 s->data.sets[0] = marks;
604 marks = s;
605 }
606 while (s->shift) {
607 uintmax_t i = idnum >> s->shift;
608 idnum -= i << s->shift;
609 if (!s->data.sets[i]) {
610 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
611 s->data.sets[i]->shift = s->shift - 10;
612 }
613 s = s->data.sets[i];
614 }
615 if (!s->data.marked[idnum])
616 marks_set_count++;
617 s->data.marked[idnum] = oe;
618 }
619
620 static struct object_entry *find_mark(uintmax_t idnum)
621 {
622 uintmax_t orig_idnum = idnum;
623 struct mark_set *s = marks;
624 struct object_entry *oe = NULL;
625 if ((idnum >> s->shift) < 1024) {
626 while (s && s->shift) {
627 uintmax_t i = idnum >> s->shift;
628 idnum -= i << s->shift;
629 s = s->data.sets[i];
630 }
631 if (s)
632 oe = s->data.marked[idnum];
633 }
634 if (!oe)
635 die("mark :%" PRIuMAX " not declared", orig_idnum);
636 return oe;
637 }
638
639 static struct atom_str *to_atom(const char *s, unsigned short len)
640 {
641 unsigned int hc = hc_str(s, len) % atom_table_sz;
642 struct atom_str *c;
643
644 for (c = atom_table[hc]; c; c = c->next_atom)
645 if (c->str_len == len && !strncmp(s, c->str_dat, len))
646 return c;
647
648 c = pool_alloc(sizeof(struct atom_str) + len + 1);
649 c->str_len = len;
650 strncpy(c->str_dat, s, len);
651 c->str_dat[len] = 0;
652 c->next_atom = atom_table[hc];
653 atom_table[hc] = c;
654 atom_cnt++;
655 return c;
656 }
657
658 static struct branch *lookup_branch(const char *name)
659 {
660 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
661 struct branch *b;
662
663 for (b = branch_table[hc]; b; b = b->table_next_branch)
664 if (!strcmp(name, b->name))
665 return b;
666 return NULL;
667 }
668
669 static struct branch *new_branch(const char *name)
670 {
671 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
672 struct branch* b = lookup_branch(name);
673
674 if (b)
675 die("Invalid attempt to create duplicate branch: %s", name);
676 switch (check_ref_format(name)) {
677 case 0: break; /* its valid */
678 case CHECK_REF_FORMAT_ONELEVEL:
679 break; /* valid, but too few '/', allow anyway */
680 default:
681 die("Branch name doesn't conform to GIT standards: %s", name);
682 }
683
684 b = pool_calloc(1, sizeof(struct branch));
685 b->name = pool_strdup(name);
686 b->table_next_branch = branch_table[hc];
687 b->branch_tree.versions[0].mode = S_IFDIR;
688 b->branch_tree.versions[1].mode = S_IFDIR;
689 b->active = 0;
690 b->pack_id = MAX_PACK_ID;
691 branch_table[hc] = b;
692 branch_count++;
693 return b;
694 }
695
696 static unsigned int hc_entries(unsigned int cnt)
697 {
698 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
699 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
700 }
701
702 static struct tree_content *new_tree_content(unsigned int cnt)
703 {
704 struct avail_tree_content *f, *l = NULL;
705 struct tree_content *t;
706 unsigned int hc = hc_entries(cnt);
707
708 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
709 if (f->entry_capacity >= cnt)
710 break;
711
712 if (f) {
713 if (l)
714 l->next_avail = f->next_avail;
715 else
716 avail_tree_table[hc] = f->next_avail;
717 } else {
718 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
719 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
720 f->entry_capacity = cnt;
721 }
722
723 t = (struct tree_content*)f;
724 t->entry_count = 0;
725 t->delta_depth = 0;
726 return t;
727 }
728
729 static void release_tree_entry(struct tree_entry *e);
730 static void release_tree_content(struct tree_content *t)
731 {
732 struct avail_tree_content *f = (struct avail_tree_content*)t;
733 unsigned int hc = hc_entries(f->entry_capacity);
734 f->next_avail = avail_tree_table[hc];
735 avail_tree_table[hc] = f;
736 }
737
738 static void release_tree_content_recursive(struct tree_content *t)
739 {
740 unsigned int i;
741 for (i = 0; i < t->entry_count; i++)
742 release_tree_entry(t->entries[i]);
743 release_tree_content(t);
744 }
745
746 static struct tree_content *grow_tree_content(
747 struct tree_content *t,
748 int amt)
749 {
750 struct tree_content *r = new_tree_content(t->entry_count + amt);
751 r->entry_count = t->entry_count;
752 r->delta_depth = t->delta_depth;
753 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
754 release_tree_content(t);
755 return r;
756 }
757
758 static struct tree_entry *new_tree_entry(void)
759 {
760 struct tree_entry *e;
761
762 if (!avail_tree_entry) {
763 unsigned int n = tree_entry_alloc;
764 total_allocd += n * sizeof(struct tree_entry);
765 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
766 while (n-- > 1) {
767 *((void**)e) = e + 1;
768 e++;
769 }
770 *((void**)e) = NULL;
771 }
772
773 e = avail_tree_entry;
774 avail_tree_entry = *((void**)e);
775 return e;
776 }
777
778 static void release_tree_entry(struct tree_entry *e)
779 {
780 if (e->tree)
781 release_tree_content_recursive(e->tree);
782 *((void**)e) = avail_tree_entry;
783 avail_tree_entry = e;
784 }
785
786 static struct tree_content *dup_tree_content(struct tree_content *s)
787 {
788 struct tree_content *d;
789 struct tree_entry *a, *b;
790 unsigned int i;
791
792 if (!s)
793 return NULL;
794 d = new_tree_content(s->entry_count);
795 for (i = 0; i < s->entry_count; i++) {
796 a = s->entries[i];
797 b = new_tree_entry();
798 memcpy(b, a, sizeof(*a));
799 if (a->tree && is_null_sha1(b->versions[1].sha1))
800 b->tree = dup_tree_content(a->tree);
801 else
802 b->tree = NULL;
803 d->entries[i] = b;
804 }
805 d->entry_count = s->entry_count;
806 d->delta_depth = s->delta_depth;
807
808 return d;
809 }
810
811 static void start_packfile(void)
812 {
813 static char tmpfile[PATH_MAX];
814 struct packed_git *p;
815 struct pack_header hdr;
816 int pack_fd;
817
818 snprintf(tmpfile, sizeof(tmpfile),
819 "%s/tmp_pack_XXXXXX", get_object_directory());
820 pack_fd = xmkstemp(tmpfile);
821 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
822 strcpy(p->pack_name, tmpfile);
823 p->pack_fd = pack_fd;
824
825 hdr.hdr_signature = htonl(PACK_SIGNATURE);
826 hdr.hdr_version = htonl(2);
827 hdr.hdr_entries = 0;
828 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
829
830 pack_data = p;
831 pack_size = sizeof(hdr);
832 object_count = 0;
833
834 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
835 all_packs[pack_id] = p;
836 }
837
838 static int oecmp (const void *a_, const void *b_)
839 {
840 struct object_entry *a = *((struct object_entry**)a_);
841 struct object_entry *b = *((struct object_entry**)b_);
842 return hashcmp(a->sha1, b->sha1);
843 }
844
845 static char *create_index(void)
846 {
847 static char tmpfile[PATH_MAX];
848 SHA_CTX ctx;
849 struct sha1file *f;
850 struct object_entry **idx, **c, **last, *e;
851 struct object_entry_pool *o;
852 uint32_t array[256];
853 int i, idx_fd;
854
855 /* Build the sorted table of object IDs. */
856 idx = xmalloc(object_count * sizeof(struct object_entry*));
857 c = idx;
858 for (o = blocks; o; o = o->next_pool)
859 for (e = o->next_free; e-- != o->entries;)
860 if (pack_id == e->pack_id)
861 *c++ = e;
862 last = idx + object_count;
863 if (c != last)
864 die("internal consistency error creating the index");
865 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
866
867 /* Generate the fan-out array. */
868 c = idx;
869 for (i = 0; i < 256; i++) {
870 struct object_entry **next = c;;
871 while (next < last) {
872 if ((*next)->sha1[0] != i)
873 break;
874 next++;
875 }
876 array[i] = htonl(next - idx);
877 c = next;
878 }
879
880 snprintf(tmpfile, sizeof(tmpfile),
881 "%s/tmp_idx_XXXXXX", get_object_directory());
882 idx_fd = xmkstemp(tmpfile);
883 f = sha1fd(idx_fd, tmpfile);
884 sha1write(f, array, 256 * sizeof(int));
885 SHA1_Init(&ctx);
886 for (c = idx; c != last; c++) {
887 uint32_t offset = htonl((*c)->offset);
888 sha1write(f, &offset, 4);
889 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
890 SHA1_Update(&ctx, (*c)->sha1, 20);
891 }
892 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
893 sha1close(f, NULL, 1);
894 free(idx);
895 SHA1_Final(pack_data->sha1, &ctx);
896 return tmpfile;
897 }
898
899 static char *keep_pack(char *curr_index_name)
900 {
901 static char name[PATH_MAX];
902 static const char *keep_msg = "fast-import";
903 int keep_fd;
904
905 chmod(pack_data->pack_name, 0444);
906 chmod(curr_index_name, 0444);
907
908 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
909 get_object_directory(), sha1_to_hex(pack_data->sha1));
910 keep_fd = open(name, O_RDWR|O_CREAT|O_EXCL, 0600);
911 if (keep_fd < 0)
912 die("cannot create keep file");
913 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
914 if (close(keep_fd))
915 die("failed to write keep file");
916
917 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
918 get_object_directory(), sha1_to_hex(pack_data->sha1));
919 if (move_temp_to_file(pack_data->pack_name, name))
920 die("cannot store pack file");
921
922 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
923 get_object_directory(), sha1_to_hex(pack_data->sha1));
924 if (move_temp_to_file(curr_index_name, name))
925 die("cannot store index file");
926 return name;
927 }
928
929 static void unkeep_all_packs(void)
930 {
931 static char name[PATH_MAX];
932 int k;
933
934 for (k = 0; k < pack_id; k++) {
935 struct packed_git *p = all_packs[k];
936 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
937 get_object_directory(), sha1_to_hex(p->sha1));
938 unlink(name);
939 }
940 }
941
942 static void end_packfile(void)
943 {
944 struct packed_git *old_p = pack_data, *new_p;
945
946 if (object_count) {
947 char *idx_name;
948 int i;
949 struct branch *b;
950 struct tag *t;
951
952 close_pack_windows(pack_data);
953 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
954 pack_data->pack_name, object_count);
955 close(pack_data->pack_fd);
956 idx_name = keep_pack(create_index());
957
958 /* Register the packfile with core git's machinary. */
959 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
960 if (!new_p)
961 die("core git rejected index %s", idx_name);
962 all_packs[pack_id] = new_p;
963 install_packed_git(new_p);
964
965 /* Print the boundary */
966 if (pack_edges) {
967 fprintf(pack_edges, "%s:", new_p->pack_name);
968 for (i = 0; i < branch_table_sz; i++) {
969 for (b = branch_table[i]; b; b = b->table_next_branch) {
970 if (b->pack_id == pack_id)
971 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
972 }
973 }
974 for (t = first_tag; t; t = t->next_tag) {
975 if (t->pack_id == pack_id)
976 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
977 }
978 fputc('\n', pack_edges);
979 fflush(pack_edges);
980 }
981
982 pack_id++;
983 }
984 else
985 unlink(old_p->pack_name);
986 free(old_p);
987
988 /* We can't carry a delta across packfiles. */
989 strbuf_release(&last_blob.data);
990 last_blob.offset = 0;
991 last_blob.depth = 0;
992 }
993
994 static void cycle_packfile(void)
995 {
996 end_packfile();
997 start_packfile();
998 }
999
1000 static size_t encode_header(
1001 enum object_type type,
1002 size_t size,
1003 unsigned char *hdr)
1004 {
1005 int n = 1;
1006 unsigned char c;
1007
1008 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
1009 die("bad type %d", type);
1010
1011 c = (type << 4) | (size & 15);
1012 size >>= 4;
1013 while (size) {
1014 *hdr++ = c | 0x80;
1015 c = size & 0x7f;
1016 size >>= 7;
1017 n++;
1018 }
1019 *hdr = c;
1020 return n;
1021 }
1022
1023 static int store_object(
1024 enum object_type type,
1025 struct strbuf *dat,
1026 struct last_object *last,
1027 unsigned char *sha1out,
1028 uintmax_t mark)
1029 {
1030 void *out, *delta;
1031 struct object_entry *e;
1032 unsigned char hdr[96];
1033 unsigned char sha1[20];
1034 unsigned long hdrlen, deltalen;
1035 SHA_CTX c;
1036 z_stream s;
1037
1038 hdrlen = sprintf((char*)hdr,"%s %lu", typename(type),
1039 (unsigned long)dat->len) + 1;
1040 SHA1_Init(&c);
1041 SHA1_Update(&c, hdr, hdrlen);
1042 SHA1_Update(&c, dat->buf, dat->len);
1043 SHA1_Final(sha1, &c);
1044 if (sha1out)
1045 hashcpy(sha1out, sha1);
1046
1047 e = insert_object(sha1);
1048 if (mark)
1049 insert_mark(mark, e);
1050 if (e->offset) {
1051 duplicate_count_by_type[type]++;
1052 return 1;
1053 } else if (find_sha1_pack(sha1, packed_git)) {
1054 e->type = type;
1055 e->pack_id = MAX_PACK_ID;
1056 e->offset = 1; /* just not zero! */
1057 duplicate_count_by_type[type]++;
1058 return 1;
1059 }
1060
1061 if (last && last->data.buf && last->depth < max_depth) {
1062 delta = diff_delta(last->data.buf, last->data.len,
1063 dat->buf, dat->len,
1064 &deltalen, 0);
1065 if (delta && deltalen >= dat->len) {
1066 free(delta);
1067 delta = NULL;
1068 }
1069 } else
1070 delta = NULL;
1071
1072 memset(&s, 0, sizeof(s));
1073 deflateInit(&s, pack_compression_level);
1074 if (delta) {
1075 s.next_in = delta;
1076 s.avail_in = deltalen;
1077 } else {
1078 s.next_in = (void *)dat->buf;
1079 s.avail_in = dat->len;
1080 }
1081 s.avail_out = deflateBound(&s, s.avail_in);
1082 s.next_out = out = xmalloc(s.avail_out);
1083 while (deflate(&s, Z_FINISH) == Z_OK)
1084 /* nothing */;
1085 deflateEnd(&s);
1086
1087 /* Determine if we should auto-checkpoint. */
1088 if ((pack_size + 60 + s.total_out) > max_packsize
1089 || (pack_size + 60 + s.total_out) < pack_size) {
1090
1091 /* This new object needs to *not* have the current pack_id. */
1092 e->pack_id = pack_id + 1;
1093 cycle_packfile();
1094
1095 /* We cannot carry a delta into the new pack. */
1096 if (delta) {
1097 free(delta);
1098 delta = NULL;
1099
1100 memset(&s, 0, sizeof(s));
1101 deflateInit(&s, pack_compression_level);
1102 s.next_in = (void *)dat->buf;
1103 s.avail_in = dat->len;
1104 s.avail_out = deflateBound(&s, s.avail_in);
1105 s.next_out = out = xrealloc(out, s.avail_out);
1106 while (deflate(&s, Z_FINISH) == Z_OK)
1107 /* nothing */;
1108 deflateEnd(&s);
1109 }
1110 }
1111
1112 e->type = type;
1113 e->pack_id = pack_id;
1114 e->offset = pack_size;
1115 object_count++;
1116 object_count_by_type[type]++;
1117
1118 if (delta) {
1119 unsigned long ofs = e->offset - last->offset;
1120 unsigned pos = sizeof(hdr) - 1;
1121
1122 delta_count_by_type[type]++;
1123 e->depth = last->depth + 1;
1124
1125 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1126 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1127 pack_size += hdrlen;
1128
1129 hdr[pos] = ofs & 127;
1130 while (ofs >>= 7)
1131 hdr[--pos] = 128 | (--ofs & 127);
1132 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
1133 pack_size += sizeof(hdr) - pos;
1134 } else {
1135 e->depth = 0;
1136 hdrlen = encode_header(type, dat->len, hdr);
1137 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1138 pack_size += hdrlen;
1139 }
1140
1141 write_or_die(pack_data->pack_fd, out, s.total_out);
1142 pack_size += s.total_out;
1143
1144 free(out);
1145 free(delta);
1146 if (last) {
1147 if (last->no_swap) {
1148 last->data = *dat;
1149 } else {
1150 strbuf_swap(&last->data, dat);
1151 }
1152 last->offset = e->offset;
1153 last->depth = e->depth;
1154 }
1155 return 0;
1156 }
1157
1158 /* All calls must be guarded by find_object() or find_mark() to
1159 * ensure the 'struct object_entry' passed was written by this
1160 * process instance. We unpack the entry by the offset, avoiding
1161 * the need for the corresponding .idx file. This unpacking rule
1162 * works because we only use OBJ_REF_DELTA within the packfiles
1163 * created by fast-import.
1164 *
1165 * oe must not be NULL. Such an oe usually comes from giving
1166 * an unknown SHA-1 to find_object() or an undefined mark to
1167 * find_mark(). Callers must test for this condition and use
1168 * the standard read_sha1_file() when it happens.
1169 *
1170 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1171 * find_mark(), where the mark was reloaded from an existing marks
1172 * file and is referencing an object that this fast-import process
1173 * instance did not write out to a packfile. Callers must test for
1174 * this condition and use read_sha1_file() instead.
1175 */
1176 static void *gfi_unpack_entry(
1177 struct object_entry *oe,
1178 unsigned long *sizep)
1179 {
1180 enum object_type type;
1181 struct packed_git *p = all_packs[oe->pack_id];
1182 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1183 /* The object is stored in the packfile we are writing to
1184 * and we have modified it since the last time we scanned
1185 * back to read a previously written object. If an old
1186 * window covered [p->pack_size, p->pack_size + 20) its
1187 * data is stale and is not valid. Closing all windows
1188 * and updating the packfile length ensures we can read
1189 * the newly written data.
1190 */
1191 close_pack_windows(p);
1192
1193 /* We have to offer 20 bytes additional on the end of
1194 * the packfile as the core unpacker code assumes the
1195 * footer is present at the file end and must promise
1196 * at least 20 bytes within any window it maps. But
1197 * we don't actually create the footer here.
1198 */
1199 p->pack_size = pack_size + 20;
1200 }
1201 return unpack_entry(p, oe->offset, &type, sizep);
1202 }
1203
1204 static const char *get_mode(const char *str, uint16_t *modep)
1205 {
1206 unsigned char c;
1207 uint16_t mode = 0;
1208
1209 while ((c = *str++) != ' ') {
1210 if (c < '0' || c > '7')
1211 return NULL;
1212 mode = (mode << 3) + (c - '0');
1213 }
1214 *modep = mode;
1215 return str;
1216 }
1217
1218 static void load_tree(struct tree_entry *root)
1219 {
1220 unsigned char* sha1 = root->versions[1].sha1;
1221 struct object_entry *myoe;
1222 struct tree_content *t;
1223 unsigned long size;
1224 char *buf;
1225 const char *c;
1226
1227 root->tree = t = new_tree_content(8);
1228 if (is_null_sha1(sha1))
1229 return;
1230
1231 myoe = find_object(sha1);
1232 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1233 if (myoe->type != OBJ_TREE)
1234 die("Not a tree: %s", sha1_to_hex(sha1));
1235 t->delta_depth = myoe->depth;
1236 buf = gfi_unpack_entry(myoe, &size);
1237 if (!buf)
1238 die("Can't load tree %s", sha1_to_hex(sha1));
1239 } else {
1240 enum object_type type;
1241 buf = read_sha1_file(sha1, &type, &size);
1242 if (!buf || type != OBJ_TREE)
1243 die("Can't load tree %s", sha1_to_hex(sha1));
1244 }
1245
1246 c = buf;
1247 while (c != (buf + size)) {
1248 struct tree_entry *e = new_tree_entry();
1249
1250 if (t->entry_count == t->entry_capacity)
1251 root->tree = t = grow_tree_content(t, t->entry_count);
1252 t->entries[t->entry_count++] = e;
1253
1254 e->tree = NULL;
1255 c = get_mode(c, &e->versions[1].mode);
1256 if (!c)
1257 die("Corrupt mode in %s", sha1_to_hex(sha1));
1258 e->versions[0].mode = e->versions[1].mode;
1259 e->name = to_atom(c, strlen(c));
1260 c += e->name->str_len + 1;
1261 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1262 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1263 c += 20;
1264 }
1265 free(buf);
1266 }
1267
1268 static int tecmp0 (const void *_a, const void *_b)
1269 {
1270 struct tree_entry *a = *((struct tree_entry**)_a);
1271 struct tree_entry *b = *((struct tree_entry**)_b);
1272 return base_name_compare(
1273 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1274 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1275 }
1276
1277 static int tecmp1 (const void *_a, const void *_b)
1278 {
1279 struct tree_entry *a = *((struct tree_entry**)_a);
1280 struct tree_entry *b = *((struct tree_entry**)_b);
1281 return base_name_compare(
1282 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1283 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1284 }
1285
1286 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1287 {
1288 size_t maxlen = 0;
1289 unsigned int i;
1290
1291 if (!v)
1292 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1293 else
1294 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1295
1296 for (i = 0; i < t->entry_count; i++) {
1297 if (t->entries[i]->versions[v].mode)
1298 maxlen += t->entries[i]->name->str_len + 34;
1299 }
1300
1301 strbuf_reset(b);
1302 strbuf_grow(b, maxlen);
1303 for (i = 0; i < t->entry_count; i++) {
1304 struct tree_entry *e = t->entries[i];
1305 if (!e->versions[v].mode)
1306 continue;
1307 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1308 e->name->str_dat, '\0');
1309 strbuf_add(b, e->versions[v].sha1, 20);
1310 }
1311 }
1312
1313 static void store_tree(struct tree_entry *root)
1314 {
1315 struct tree_content *t = root->tree;
1316 unsigned int i, j, del;
1317 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1318 struct object_entry *le;
1319
1320 if (!is_null_sha1(root->versions[1].sha1))
1321 return;
1322
1323 for (i = 0; i < t->entry_count; i++) {
1324 if (t->entries[i]->tree)
1325 store_tree(t->entries[i]);
1326 }
1327
1328 le = find_object(root->versions[0].sha1);
1329 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1330 mktree(t, 0, &old_tree);
1331 lo.data = old_tree;
1332 lo.offset = le->offset;
1333 lo.depth = t->delta_depth;
1334 }
1335
1336 mktree(t, 1, &new_tree);
1337 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1338
1339 t->delta_depth = lo.depth;
1340 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1341 struct tree_entry *e = t->entries[i];
1342 if (e->versions[1].mode) {
1343 e->versions[0].mode = e->versions[1].mode;
1344 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1345 t->entries[j++] = e;
1346 } else {
1347 release_tree_entry(e);
1348 del++;
1349 }
1350 }
1351 t->entry_count -= del;
1352 }
1353
1354 static int tree_content_set(
1355 struct tree_entry *root,
1356 const char *p,
1357 const unsigned char *sha1,
1358 const uint16_t mode,
1359 struct tree_content *subtree)
1360 {
1361 struct tree_content *t = root->tree;
1362 const char *slash1;
1363 unsigned int i, n;
1364 struct tree_entry *e;
1365
1366 slash1 = strchr(p, '/');
1367 if (slash1)
1368 n = slash1 - p;
1369 else
1370 n = strlen(p);
1371 if (!n)
1372 die("Empty path component found in input");
1373 if (!slash1 && !S_ISDIR(mode) && subtree)
1374 die("Non-directories cannot have subtrees");
1375
1376 for (i = 0; i < t->entry_count; i++) {
1377 e = t->entries[i];
1378 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1379 if (!slash1) {
1380 if (!S_ISDIR(mode)
1381 && e->versions[1].mode == mode
1382 && !hashcmp(e->versions[1].sha1, sha1))
1383 return 0;
1384 e->versions[1].mode = mode;
1385 hashcpy(e->versions[1].sha1, sha1);
1386 if (e->tree)
1387 release_tree_content_recursive(e->tree);
1388 e->tree = subtree;
1389 hashclr(root->versions[1].sha1);
1390 return 1;
1391 }
1392 if (!S_ISDIR(e->versions[1].mode)) {
1393 e->tree = new_tree_content(8);
1394 e->versions[1].mode = S_IFDIR;
1395 }
1396 if (!e->tree)
1397 load_tree(e);
1398 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1399 hashclr(root->versions[1].sha1);
1400 return 1;
1401 }
1402 return 0;
1403 }
1404 }
1405
1406 if (t->entry_count == t->entry_capacity)
1407 root->tree = t = grow_tree_content(t, t->entry_count);
1408 e = new_tree_entry();
1409 e->name = to_atom(p, n);
1410 e->versions[0].mode = 0;
1411 hashclr(e->versions[0].sha1);
1412 t->entries[t->entry_count++] = e;
1413 if (slash1) {
1414 e->tree = new_tree_content(8);
1415 e->versions[1].mode = S_IFDIR;
1416 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1417 } else {
1418 e->tree = subtree;
1419 e->versions[1].mode = mode;
1420 hashcpy(e->versions[1].sha1, sha1);
1421 }
1422 hashclr(root->versions[1].sha1);
1423 return 1;
1424 }
1425
1426 static int tree_content_remove(
1427 struct tree_entry *root,
1428 const char *p,
1429 struct tree_entry *backup_leaf)
1430 {
1431 struct tree_content *t = root->tree;
1432 const char *slash1;
1433 unsigned int i, n;
1434 struct tree_entry *e;
1435
1436 slash1 = strchr(p, '/');
1437 if (slash1)
1438 n = slash1 - p;
1439 else
1440 n = strlen(p);
1441
1442 for (i = 0; i < t->entry_count; i++) {
1443 e = t->entries[i];
1444 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1445 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1446 goto del_entry;
1447 if (!e->tree)
1448 load_tree(e);
1449 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1450 for (n = 0; n < e->tree->entry_count; n++) {
1451 if (e->tree->entries[n]->versions[1].mode) {
1452 hashclr(root->versions[1].sha1);
1453 return 1;
1454 }
1455 }
1456 backup_leaf = NULL;
1457 goto del_entry;
1458 }
1459 return 0;
1460 }
1461 }
1462 return 0;
1463
1464 del_entry:
1465 if (backup_leaf)
1466 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1467 else if (e->tree)
1468 release_tree_content_recursive(e->tree);
1469 e->tree = NULL;
1470 e->versions[1].mode = 0;
1471 hashclr(e->versions[1].sha1);
1472 hashclr(root->versions[1].sha1);
1473 return 1;
1474 }
1475
1476 static int tree_content_get(
1477 struct tree_entry *root,
1478 const char *p,
1479 struct tree_entry *leaf)
1480 {
1481 struct tree_content *t = root->tree;
1482 const char *slash1;
1483 unsigned int i, n;
1484 struct tree_entry *e;
1485
1486 slash1 = strchr(p, '/');
1487 if (slash1)
1488 n = slash1 - p;
1489 else
1490 n = strlen(p);
1491
1492 for (i = 0; i < t->entry_count; i++) {
1493 e = t->entries[i];
1494 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1495 if (!slash1) {
1496 memcpy(leaf, e, sizeof(*leaf));
1497 if (e->tree && is_null_sha1(e->versions[1].sha1))
1498 leaf->tree = dup_tree_content(e->tree);
1499 else
1500 leaf->tree = NULL;
1501 return 1;
1502 }
1503 if (!S_ISDIR(e->versions[1].mode))
1504 return 0;
1505 if (!e->tree)
1506 load_tree(e);
1507 return tree_content_get(e, slash1 + 1, leaf);
1508 }
1509 }
1510 return 0;
1511 }
1512
1513 static int update_branch(struct branch *b)
1514 {
1515 static const char *msg = "fast-import";
1516 struct ref_lock *lock;
1517 unsigned char old_sha1[20];
1518
1519 if (is_null_sha1(b->sha1))
1520 return 0;
1521 if (read_ref(b->name, old_sha1))
1522 hashclr(old_sha1);
1523 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1524 if (!lock)
1525 return error("Unable to lock %s", b->name);
1526 if (!force_update && !is_null_sha1(old_sha1)) {
1527 struct commit *old_cmit, *new_cmit;
1528
1529 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1530 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1531 if (!old_cmit || !new_cmit) {
1532 unlock_ref(lock);
1533 return error("Branch %s is missing commits.", b->name);
1534 }
1535
1536 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1537 unlock_ref(lock);
1538 warning("Not updating %s"
1539 " (new tip %s does not contain %s)",
1540 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1541 return -1;
1542 }
1543 }
1544 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1545 return error("Unable to update %s", b->name);
1546 return 0;
1547 }
1548
1549 static void dump_branches(void)
1550 {
1551 unsigned int i;
1552 struct branch *b;
1553
1554 for (i = 0; i < branch_table_sz; i++) {
1555 for (b = branch_table[i]; b; b = b->table_next_branch)
1556 failure |= update_branch(b);
1557 }
1558 }
1559
1560 static void dump_tags(void)
1561 {
1562 static const char *msg = "fast-import";
1563 struct tag *t;
1564 struct ref_lock *lock;
1565 char ref_name[PATH_MAX];
1566
1567 for (t = first_tag; t; t = t->next_tag) {
1568 sprintf(ref_name, "tags/%s", t->name);
1569 lock = lock_ref_sha1(ref_name, NULL);
1570 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1571 failure |= error("Unable to update %s", ref_name);
1572 }
1573 }
1574
1575 static void dump_marks_helper(FILE *f,
1576 uintmax_t base,
1577 struct mark_set *m)
1578 {
1579 uintmax_t k;
1580 if (m->shift) {
1581 for (k = 0; k < 1024; k++) {
1582 if (m->data.sets[k])
1583 dump_marks_helper(f, (base + k) << m->shift,
1584 m->data.sets[k]);
1585 }
1586 } else {
1587 for (k = 0; k < 1024; k++) {
1588 if (m->data.marked[k])
1589 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1590 sha1_to_hex(m->data.marked[k]->sha1));
1591 }
1592 }
1593 }
1594
1595 static void dump_marks(void)
1596 {
1597 static struct lock_file mark_lock;
1598 int mark_fd;
1599 FILE *f;
1600
1601 if (!mark_file)
1602 return;
1603
1604 mark_fd = hold_lock_file_for_update(&mark_lock, mark_file, 0);
1605 if (mark_fd < 0) {
1606 failure |= error("Unable to write marks file %s: %s",
1607 mark_file, strerror(errno));
1608 return;
1609 }
1610
1611 f = fdopen(mark_fd, "w");
1612 if (!f) {
1613 int saved_errno = errno;
1614 rollback_lock_file(&mark_lock);
1615 failure |= error("Unable to write marks file %s: %s",
1616 mark_file, strerror(saved_errno));
1617 return;
1618 }
1619
1620 /*
1621 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1622 * Assign -1 to the lock file descriptor so that commit_lock_file()
1623 * won't try to close() it.
1624 */
1625 mark_lock.fd = -1;
1626
1627 dump_marks_helper(f, 0, marks);
1628 if (ferror(f) || fclose(f)) {
1629 int saved_errno = errno;
1630 rollback_lock_file(&mark_lock);
1631 failure |= error("Unable to write marks file %s: %s",
1632 mark_file, strerror(saved_errno));
1633 return;
1634 }
1635
1636 if (commit_lock_file(&mark_lock)) {
1637 int saved_errno = errno;
1638 rollback_lock_file(&mark_lock);
1639 failure |= error("Unable to commit marks file %s: %s",
1640 mark_file, strerror(saved_errno));
1641 return;
1642 }
1643 }
1644
1645 static int read_next_command(void)
1646 {
1647 static int stdin_eof = 0;
1648
1649 if (stdin_eof) {
1650 unread_command_buf = 0;
1651 return EOF;
1652 }
1653
1654 do {
1655 if (unread_command_buf) {
1656 unread_command_buf = 0;
1657 } else {
1658 struct recent_command *rc;
1659
1660 strbuf_detach(&command_buf, NULL);
1661 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1662 if (stdin_eof)
1663 return EOF;
1664
1665 rc = rc_free;
1666 if (rc)
1667 rc_free = rc->next;
1668 else {
1669 rc = cmd_hist.next;
1670 cmd_hist.next = rc->next;
1671 cmd_hist.next->prev = &cmd_hist;
1672 free(rc->buf);
1673 }
1674
1675 rc->buf = command_buf.buf;
1676 rc->prev = cmd_tail;
1677 rc->next = cmd_hist.prev;
1678 rc->prev->next = rc;
1679 cmd_tail = rc;
1680 }
1681 } while (command_buf.buf[0] == '#');
1682
1683 return 0;
1684 }
1685
1686 static void skip_optional_lf(void)
1687 {
1688 int term_char = fgetc(stdin);
1689 if (term_char != '\n' && term_char != EOF)
1690 ungetc(term_char, stdin);
1691 }
1692
1693 static void cmd_mark(void)
1694 {
1695 if (!prefixcmp(command_buf.buf, "mark :")) {
1696 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1697 read_next_command();
1698 }
1699 else
1700 next_mark = 0;
1701 }
1702
1703 static void cmd_data(struct strbuf *sb)
1704 {
1705 strbuf_reset(sb);
1706
1707 if (prefixcmp(command_buf.buf, "data "))
1708 die("Expected 'data n' command, found: %s", command_buf.buf);
1709
1710 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1711 char *term = xstrdup(command_buf.buf + 5 + 2);
1712 size_t term_len = command_buf.len - 5 - 2;
1713
1714 strbuf_detach(&command_buf, NULL);
1715 for (;;) {
1716 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1717 die("EOF in data (terminator '%s' not found)", term);
1718 if (term_len == command_buf.len
1719 && !strcmp(term, command_buf.buf))
1720 break;
1721 strbuf_addbuf(sb, &command_buf);
1722 strbuf_addch(sb, '\n');
1723 }
1724 free(term);
1725 }
1726 else {
1727 size_t n = 0, length;
1728
1729 length = strtoul(command_buf.buf + 5, NULL, 10);
1730
1731 while (n < length) {
1732 size_t s = strbuf_fread(sb, length - n, stdin);
1733 if (!s && feof(stdin))
1734 die("EOF in data (%lu bytes remaining)",
1735 (unsigned long)(length - n));
1736 n += s;
1737 }
1738 }
1739
1740 skip_optional_lf();
1741 }
1742
1743 static int validate_raw_date(const char *src, char *result, int maxlen)
1744 {
1745 const char *orig_src = src;
1746 char *endp, sign;
1747
1748 strtoul(src, &endp, 10);
1749 if (endp == src || *endp != ' ')
1750 return -1;
1751
1752 src = endp + 1;
1753 if (*src != '-' && *src != '+')
1754 return -1;
1755 sign = *src;
1756
1757 strtoul(src + 1, &endp, 10);
1758 if (endp == src || *endp || (endp - orig_src) >= maxlen)
1759 return -1;
1760
1761 strcpy(result, orig_src);
1762 return 0;
1763 }
1764
1765 static char *parse_ident(const char *buf)
1766 {
1767 const char *gt;
1768 size_t name_len;
1769 char *ident;
1770
1771 gt = strrchr(buf, '>');
1772 if (!gt)
1773 die("Missing > in ident string: %s", buf);
1774 gt++;
1775 if (*gt != ' ')
1776 die("Missing space after > in ident string: %s", buf);
1777 gt++;
1778 name_len = gt - buf;
1779 ident = xmalloc(name_len + 24);
1780 strncpy(ident, buf, name_len);
1781
1782 switch (whenspec) {
1783 case WHENSPEC_RAW:
1784 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1785 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1786 break;
1787 case WHENSPEC_RFC2822:
1788 if (parse_date(gt, ident + name_len, 24) < 0)
1789 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1790 break;
1791 case WHENSPEC_NOW:
1792 if (strcmp("now", gt))
1793 die("Date in ident must be 'now': %s", buf);
1794 datestamp(ident + name_len, 24);
1795 break;
1796 }
1797
1798 return ident;
1799 }
1800
1801 static void cmd_new_blob(void)
1802 {
1803 static struct strbuf buf = STRBUF_INIT;
1804
1805 read_next_command();
1806 cmd_mark();
1807 cmd_data(&buf);
1808 store_object(OBJ_BLOB, &buf, &last_blob, NULL, next_mark);
1809 }
1810
1811 static void unload_one_branch(void)
1812 {
1813 while (cur_active_branches
1814 && cur_active_branches >= max_active_branches) {
1815 uintmax_t min_commit = ULONG_MAX;
1816 struct branch *e, *l = NULL, *p = NULL;
1817
1818 for (e = active_branches; e; e = e->active_next_branch) {
1819 if (e->last_commit < min_commit) {
1820 p = l;
1821 min_commit = e->last_commit;
1822 }
1823 l = e;
1824 }
1825
1826 if (p) {
1827 e = p->active_next_branch;
1828 p->active_next_branch = e->active_next_branch;
1829 } else {
1830 e = active_branches;
1831 active_branches = e->active_next_branch;
1832 }
1833 e->active = 0;
1834 e->active_next_branch = NULL;
1835 if (e->branch_tree.tree) {
1836 release_tree_content_recursive(e->branch_tree.tree);
1837 e->branch_tree.tree = NULL;
1838 }
1839 cur_active_branches--;
1840 }
1841 }
1842
1843 static void load_branch(struct branch *b)
1844 {
1845 load_tree(&b->branch_tree);
1846 if (!b->active) {
1847 b->active = 1;
1848 b->active_next_branch = active_branches;
1849 active_branches = b;
1850 cur_active_branches++;
1851 branch_load_count++;
1852 }
1853 }
1854
1855 static void file_change_m(struct branch *b)
1856 {
1857 const char *p = command_buf.buf + 2;
1858 static struct strbuf uq = STRBUF_INIT;
1859 const char *endp;
1860 struct object_entry *oe = oe;
1861 unsigned char sha1[20];
1862 uint16_t mode, inline_data = 0;
1863
1864 p = get_mode(p, &mode);
1865 if (!p)
1866 die("Corrupt mode: %s", command_buf.buf);
1867 switch (mode) {
1868 case S_IFREG | 0644:
1869 case S_IFREG | 0755:
1870 case S_IFLNK:
1871 case 0644:
1872 case 0755:
1873 /* ok */
1874 break;
1875 default:
1876 die("Corrupt mode: %s", command_buf.buf);
1877 }
1878
1879 if (*p == ':') {
1880 char *x;
1881 oe = find_mark(strtoumax(p + 1, &x, 10));
1882 hashcpy(sha1, oe->sha1);
1883 p = x;
1884 } else if (!prefixcmp(p, "inline")) {
1885 inline_data = 1;
1886 p += 6;
1887 } else {
1888 if (get_sha1_hex(p, sha1))
1889 die("Invalid SHA1: %s", command_buf.buf);
1890 oe = find_object(sha1);
1891 p += 40;
1892 }
1893 if (*p++ != ' ')
1894 die("Missing space after SHA1: %s", command_buf.buf);
1895
1896 strbuf_reset(&uq);
1897 if (!unquote_c_style(&uq, p, &endp)) {
1898 if (*endp)
1899 die("Garbage after path in: %s", command_buf.buf);
1900 p = uq.buf;
1901 }
1902
1903 if (inline_data) {
1904 static struct strbuf buf = STRBUF_INIT;
1905
1906 if (p != uq.buf) {
1907 strbuf_addstr(&uq, p);
1908 p = uq.buf;
1909 }
1910 read_next_command();
1911 cmd_data(&buf);
1912 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
1913 } else if (oe) {
1914 if (oe->type != OBJ_BLOB)
1915 die("Not a blob (actually a %s): %s",
1916 typename(oe->type), command_buf.buf);
1917 } else {
1918 enum object_type type = sha1_object_info(sha1, NULL);
1919 if (type < 0)
1920 die("Blob not found: %s", command_buf.buf);
1921 if (type != OBJ_BLOB)
1922 die("Not a blob (actually a %s): %s",
1923 typename(type), command_buf.buf);
1924 }
1925
1926 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode, NULL);
1927 }
1928
1929 static void file_change_d(struct branch *b)
1930 {
1931 const char *p = command_buf.buf + 2;
1932 static struct strbuf uq = STRBUF_INIT;
1933 const char *endp;
1934
1935 strbuf_reset(&uq);
1936 if (!unquote_c_style(&uq, p, &endp)) {
1937 if (*endp)
1938 die("Garbage after path in: %s", command_buf.buf);
1939 p = uq.buf;
1940 }
1941 tree_content_remove(&b->branch_tree, p, NULL);
1942 }
1943
1944 static void file_change_cr(struct branch *b, int rename)
1945 {
1946 const char *s, *d;
1947 static struct strbuf s_uq = STRBUF_INIT;
1948 static struct strbuf d_uq = STRBUF_INIT;
1949 const char *endp;
1950 struct tree_entry leaf;
1951
1952 s = command_buf.buf + 2;
1953 strbuf_reset(&s_uq);
1954 if (!unquote_c_style(&s_uq, s, &endp)) {
1955 if (*endp != ' ')
1956 die("Missing space after source: %s", command_buf.buf);
1957 } else {
1958 endp = strchr(s, ' ');
1959 if (!endp)
1960 die("Missing space after source: %s", command_buf.buf);
1961 strbuf_add(&s_uq, s, endp - s);
1962 }
1963 s = s_uq.buf;
1964
1965 endp++;
1966 if (!*endp)
1967 die("Missing dest: %s", command_buf.buf);
1968
1969 d = endp;
1970 strbuf_reset(&d_uq);
1971 if (!unquote_c_style(&d_uq, d, &endp)) {
1972 if (*endp)
1973 die("Garbage after dest in: %s", command_buf.buf);
1974 d = d_uq.buf;
1975 }
1976
1977 memset(&leaf, 0, sizeof(leaf));
1978 if (rename)
1979 tree_content_remove(&b->branch_tree, s, &leaf);
1980 else
1981 tree_content_get(&b->branch_tree, s, &leaf);
1982 if (!leaf.versions[1].mode)
1983 die("Path %s not in branch", s);
1984 tree_content_set(&b->branch_tree, d,
1985 leaf.versions[1].sha1,
1986 leaf.versions[1].mode,
1987 leaf.tree);
1988 }
1989
1990 static void file_change_deleteall(struct branch *b)
1991 {
1992 release_tree_content_recursive(b->branch_tree.tree);
1993 hashclr(b->branch_tree.versions[0].sha1);
1994 hashclr(b->branch_tree.versions[1].sha1);
1995 load_tree(&b->branch_tree);
1996 }
1997
1998 static void cmd_from_commit(struct branch *b, char *buf, unsigned long size)
1999 {
2000 if (!buf || size < 46)
2001 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2002 if (memcmp("tree ", buf, 5)
2003 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2004 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2005 hashcpy(b->branch_tree.versions[0].sha1,
2006 b->branch_tree.versions[1].sha1);
2007 }
2008
2009 static void cmd_from_existing(struct branch *b)
2010 {
2011 if (is_null_sha1(b->sha1)) {
2012 hashclr(b->branch_tree.versions[0].sha1);
2013 hashclr(b->branch_tree.versions[1].sha1);
2014 } else {
2015 unsigned long size;
2016 char *buf;
2017
2018 buf = read_object_with_reference(b->sha1,
2019 commit_type, &size, b->sha1);
2020 cmd_from_commit(b, buf, size);
2021 free(buf);
2022 }
2023 }
2024
2025 static int cmd_from(struct branch *b)
2026 {
2027 const char *from;
2028 struct branch *s;
2029
2030 if (prefixcmp(command_buf.buf, "from "))
2031 return 0;
2032
2033 if (b->branch_tree.tree) {
2034 release_tree_content_recursive(b->branch_tree.tree);
2035 b->branch_tree.tree = NULL;
2036 }
2037
2038 from = strchr(command_buf.buf, ' ') + 1;
2039 s = lookup_branch(from);
2040 if (b == s)
2041 die("Can't create a branch from itself: %s", b->name);
2042 else if (s) {
2043 unsigned char *t = s->branch_tree.versions[1].sha1;
2044 hashcpy(b->sha1, s->sha1);
2045 hashcpy(b->branch_tree.versions[0].sha1, t);
2046 hashcpy(b->branch_tree.versions[1].sha1, t);
2047 } else if (*from == ':') {
2048 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2049 struct object_entry *oe = find_mark(idnum);
2050 if (oe->type != OBJ_COMMIT)
2051 die("Mark :%" PRIuMAX " not a commit", idnum);
2052 hashcpy(b->sha1, oe->sha1);
2053 if (oe->pack_id != MAX_PACK_ID) {
2054 unsigned long size;
2055 char *buf = gfi_unpack_entry(oe, &size);
2056 cmd_from_commit(b, buf, size);
2057 free(buf);
2058 } else
2059 cmd_from_existing(b);
2060 } else if (!get_sha1(from, b->sha1))
2061 cmd_from_existing(b);
2062 else
2063 die("Invalid ref name or SHA1 expression: %s", from);
2064
2065 read_next_command();
2066 return 1;
2067 }
2068
2069 static struct hash_list *cmd_merge(unsigned int *count)
2070 {
2071 struct hash_list *list = NULL, *n, *e = e;
2072 const char *from;
2073 struct branch *s;
2074
2075 *count = 0;
2076 while (!prefixcmp(command_buf.buf, "merge ")) {
2077 from = strchr(command_buf.buf, ' ') + 1;
2078 n = xmalloc(sizeof(*n));
2079 s = lookup_branch(from);
2080 if (s)
2081 hashcpy(n->sha1, s->sha1);
2082 else if (*from == ':') {
2083 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2084 struct object_entry *oe = find_mark(idnum);
2085 if (oe->type != OBJ_COMMIT)
2086 die("Mark :%" PRIuMAX " not a commit", idnum);
2087 hashcpy(n->sha1, oe->sha1);
2088 } else if (!get_sha1(from, n->sha1)) {
2089 unsigned long size;
2090 char *buf = read_object_with_reference(n->sha1,
2091 commit_type, &size, n->sha1);
2092 if (!buf || size < 46)
2093 die("Not a valid commit: %s", from);
2094 free(buf);
2095 } else
2096 die("Invalid ref name or SHA1 expression: %s", from);
2097
2098 n->next = NULL;
2099 if (list)
2100 e->next = n;
2101 else
2102 list = n;
2103 e = n;
2104 (*count)++;
2105 read_next_command();
2106 }
2107 return list;
2108 }
2109
2110 static void cmd_new_commit(void)
2111 {
2112 static struct strbuf msg = STRBUF_INIT;
2113 struct branch *b;
2114 char *sp;
2115 char *author = NULL;
2116 char *committer = NULL;
2117 struct hash_list *merge_list = NULL;
2118 unsigned int merge_count;
2119
2120 /* Obtain the branch name from the rest of our command */
2121 sp = strchr(command_buf.buf, ' ') + 1;
2122 b = lookup_branch(sp);
2123 if (!b)
2124 b = new_branch(sp);
2125
2126 read_next_command();
2127 cmd_mark();
2128 if (!prefixcmp(command_buf.buf, "author ")) {
2129 author = parse_ident(command_buf.buf + 7);
2130 read_next_command();
2131 }
2132 if (!prefixcmp(command_buf.buf, "committer ")) {
2133 committer = parse_ident(command_buf.buf + 10);
2134 read_next_command();
2135 }
2136 if (!committer)
2137 die("Expected committer but didn't get one");
2138 cmd_data(&msg);
2139 read_next_command();
2140 cmd_from(b);
2141 merge_list = cmd_merge(&merge_count);
2142
2143 /* ensure the branch is active/loaded */
2144 if (!b->branch_tree.tree || !max_active_branches) {
2145 unload_one_branch();
2146 load_branch(b);
2147 }
2148
2149 /* file_change* */
2150 while (command_buf.len > 0) {
2151 if (!prefixcmp(command_buf.buf, "M "))
2152 file_change_m(b);
2153 else if (!prefixcmp(command_buf.buf, "D "))
2154 file_change_d(b);
2155 else if (!prefixcmp(command_buf.buf, "R "))
2156 file_change_cr(b, 1);
2157 else if (!prefixcmp(command_buf.buf, "C "))
2158 file_change_cr(b, 0);
2159 else if (!strcmp("deleteall", command_buf.buf))
2160 file_change_deleteall(b);
2161 else {
2162 unread_command_buf = 1;
2163 break;
2164 }
2165 if (read_next_command() == EOF)
2166 break;
2167 }
2168
2169 /* build the tree and the commit */
2170 store_tree(&b->branch_tree);
2171 hashcpy(b->branch_tree.versions[0].sha1,
2172 b->branch_tree.versions[1].sha1);
2173
2174 strbuf_reset(&new_data);
2175 strbuf_addf(&new_data, "tree %s\n",
2176 sha1_to_hex(b->branch_tree.versions[1].sha1));
2177 if (!is_null_sha1(b->sha1))
2178 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2179 while (merge_list) {
2180 struct hash_list *next = merge_list->next;
2181 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2182 free(merge_list);
2183 merge_list = next;
2184 }
2185 strbuf_addf(&new_data,
2186 "author %s\n"
2187 "committer %s\n"
2188 "\n",
2189 author ? author : committer, committer);
2190 strbuf_addbuf(&new_data, &msg);
2191 free(author);
2192 free(committer);
2193
2194 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2195 b->pack_id = pack_id;
2196 b->last_commit = object_count_by_type[OBJ_COMMIT];
2197 }
2198
2199 static void cmd_new_tag(void)
2200 {
2201 static struct strbuf msg = STRBUF_INIT;
2202 char *sp;
2203 const char *from;
2204 char *tagger;
2205 struct branch *s;
2206 struct tag *t;
2207 uintmax_t from_mark = 0;
2208 unsigned char sha1[20];
2209
2210 /* Obtain the new tag name from the rest of our command */
2211 sp = strchr(command_buf.buf, ' ') + 1;
2212 t = pool_alloc(sizeof(struct tag));
2213 t->next_tag = NULL;
2214 t->name = pool_strdup(sp);
2215 if (last_tag)
2216 last_tag->next_tag = t;
2217 else
2218 first_tag = t;
2219 last_tag = t;
2220 read_next_command();
2221
2222 /* from ... */
2223 if (prefixcmp(command_buf.buf, "from "))
2224 die("Expected from command, got %s", command_buf.buf);
2225 from = strchr(command_buf.buf, ' ') + 1;
2226 s = lookup_branch(from);
2227 if (s) {
2228 hashcpy(sha1, s->sha1);
2229 } else if (*from == ':') {
2230 struct object_entry *oe;
2231 from_mark = strtoumax(from + 1, NULL, 10);
2232 oe = find_mark(from_mark);
2233 if (oe->type != OBJ_COMMIT)
2234 die("Mark :%" PRIuMAX " not a commit", from_mark);
2235 hashcpy(sha1, oe->sha1);
2236 } else if (!get_sha1(from, sha1)) {
2237 unsigned long size;
2238 char *buf;
2239
2240 buf = read_object_with_reference(sha1,
2241 commit_type, &size, sha1);
2242 if (!buf || size < 46)
2243 die("Not a valid commit: %s", from);
2244 free(buf);
2245 } else
2246 die("Invalid ref name or SHA1 expression: %s", from);
2247 read_next_command();
2248
2249 /* tagger ... */
2250 if (prefixcmp(command_buf.buf, "tagger "))
2251 die("Expected tagger command, got %s", command_buf.buf);
2252 tagger = parse_ident(command_buf.buf + 7);
2253
2254 /* tag payload/message */
2255 read_next_command();
2256 cmd_data(&msg);
2257
2258 /* build the tag object */
2259 strbuf_reset(&new_data);
2260 strbuf_addf(&new_data,
2261 "object %s\n"
2262 "type %s\n"
2263 "tag %s\n"
2264 "tagger %s\n"
2265 "\n",
2266 sha1_to_hex(sha1), commit_type, t->name, tagger);
2267 strbuf_addbuf(&new_data, &msg);
2268 free(tagger);
2269
2270 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2271 t->pack_id = MAX_PACK_ID;
2272 else
2273 t->pack_id = pack_id;
2274 }
2275
2276 static void cmd_reset_branch(void)
2277 {
2278 struct branch *b;
2279 char *sp;
2280
2281 /* Obtain the branch name from the rest of our command */
2282 sp = strchr(command_buf.buf, ' ') + 1;
2283 b = lookup_branch(sp);
2284 if (b) {
2285 hashclr(b->sha1);
2286 hashclr(b->branch_tree.versions[0].sha1);
2287 hashclr(b->branch_tree.versions[1].sha1);
2288 if (b->branch_tree.tree) {
2289 release_tree_content_recursive(b->branch_tree.tree);
2290 b->branch_tree.tree = NULL;
2291 }
2292 }
2293 else
2294 b = new_branch(sp);
2295 read_next_command();
2296 cmd_from(b);
2297 if (command_buf.len > 0)
2298 unread_command_buf = 1;
2299 }
2300
2301 static void cmd_checkpoint(void)
2302 {
2303 if (object_count) {
2304 cycle_packfile();
2305 dump_branches();
2306 dump_tags();
2307 dump_marks();
2308 }
2309 skip_optional_lf();
2310 }
2311
2312 static void cmd_progress(void)
2313 {
2314 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2315 fputc('\n', stdout);
2316 fflush(stdout);
2317 skip_optional_lf();
2318 }
2319
2320 static void import_marks(const char *input_file)
2321 {
2322 char line[512];
2323 FILE *f = fopen(input_file, "r");
2324 if (!f)
2325 die("cannot read %s: %s", input_file, strerror(errno));
2326 while (fgets(line, sizeof(line), f)) {
2327 uintmax_t mark;
2328 char *end;
2329 unsigned char sha1[20];
2330 struct object_entry *e;
2331
2332 end = strchr(line, '\n');
2333 if (line[0] != ':' || !end)
2334 die("corrupt mark line: %s", line);
2335 *end = 0;
2336 mark = strtoumax(line + 1, &end, 10);
2337 if (!mark || end == line + 1
2338 || *end != ' ' || get_sha1(end + 1, sha1))
2339 die("corrupt mark line: %s", line);
2340 e = find_object(sha1);
2341 if (!e) {
2342 enum object_type type = sha1_object_info(sha1, NULL);
2343 if (type < 0)
2344 die("object not found: %s", sha1_to_hex(sha1));
2345 e = insert_object(sha1);
2346 e->type = type;
2347 e->pack_id = MAX_PACK_ID;
2348 e->offset = 1; /* just not zero! */
2349 }
2350 insert_mark(mark, e);
2351 }
2352 fclose(f);
2353 }
2354
2355 static int git_pack_config(const char *k, const char *v, void *cb)
2356 {
2357 if (!strcmp(k, "pack.depth")) {
2358 max_depth = git_config_int(k, v);
2359 if (max_depth > MAX_DEPTH)
2360 max_depth = MAX_DEPTH;
2361 return 0;
2362 }
2363 if (!strcmp(k, "pack.compression")) {
2364 int level = git_config_int(k, v);
2365 if (level == -1)
2366 level = Z_DEFAULT_COMPRESSION;
2367 else if (level < 0 || level > Z_BEST_COMPRESSION)
2368 die("bad pack compression level %d", level);
2369 pack_compression_level = level;
2370 pack_compression_seen = 1;
2371 return 0;
2372 }
2373 return git_default_config(k, v, cb);
2374 }
2375
2376 static const char fast_import_usage[] =
2377 "git-fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2378
2379 int main(int argc, const char **argv)
2380 {
2381 unsigned int i, show_stats = 1;
2382
2383 setup_git_directory();
2384 git_config(git_pack_config, NULL);
2385 if (!pack_compression_seen && core_compression_seen)
2386 pack_compression_level = core_compression_level;
2387
2388 alloc_objects(object_entry_alloc);
2389 strbuf_init(&command_buf, 0);
2390 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2391 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2392 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2393 marks = pool_calloc(1, sizeof(struct mark_set));
2394
2395 for (i = 1; i < argc; i++) {
2396 const char *a = argv[i];
2397
2398 if (*a != '-' || !strcmp(a, "--"))
2399 break;
2400 else if (!prefixcmp(a, "--date-format=")) {
2401 const char *fmt = a + 14;
2402 if (!strcmp(fmt, "raw"))
2403 whenspec = WHENSPEC_RAW;
2404 else if (!strcmp(fmt, "rfc2822"))
2405 whenspec = WHENSPEC_RFC2822;
2406 else if (!strcmp(fmt, "now"))
2407 whenspec = WHENSPEC_NOW;
2408 else
2409 die("unknown --date-format argument %s", fmt);
2410 }
2411 else if (!prefixcmp(a, "--max-pack-size="))
2412 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
2413 else if (!prefixcmp(a, "--depth=")) {
2414 max_depth = strtoul(a + 8, NULL, 0);
2415 if (max_depth > MAX_DEPTH)
2416 die("--depth cannot exceed %u", MAX_DEPTH);
2417 }
2418 else if (!prefixcmp(a, "--active-branches="))
2419 max_active_branches = strtoul(a + 18, NULL, 0);
2420 else if (!prefixcmp(a, "--import-marks="))
2421 import_marks(a + 15);
2422 else if (!prefixcmp(a, "--export-marks="))
2423 mark_file = a + 15;
2424 else if (!prefixcmp(a, "--export-pack-edges=")) {
2425 if (pack_edges)
2426 fclose(pack_edges);
2427 pack_edges = fopen(a + 20, "a");
2428 if (!pack_edges)
2429 die("Cannot open %s: %s", a + 20, strerror(errno));
2430 } else if (!strcmp(a, "--force"))
2431 force_update = 1;
2432 else if (!strcmp(a, "--quiet"))
2433 show_stats = 0;
2434 else if (!strcmp(a, "--stats"))
2435 show_stats = 1;
2436 else
2437 die("unknown option %s", a);
2438 }
2439 if (i != argc)
2440 usage(fast_import_usage);
2441
2442 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2443 for (i = 0; i < (cmd_save - 1); i++)
2444 rc_free[i].next = &rc_free[i + 1];
2445 rc_free[cmd_save - 1].next = NULL;
2446
2447 prepare_packed_git();
2448 start_packfile();
2449 set_die_routine(die_nicely);
2450 while (read_next_command() != EOF) {
2451 if (!strcmp("blob", command_buf.buf))
2452 cmd_new_blob();
2453 else if (!prefixcmp(command_buf.buf, "commit "))
2454 cmd_new_commit();
2455 else if (!prefixcmp(command_buf.buf, "tag "))
2456 cmd_new_tag();
2457 else if (!prefixcmp(command_buf.buf, "reset "))
2458 cmd_reset_branch();
2459 else if (!strcmp("checkpoint", command_buf.buf))
2460 cmd_checkpoint();
2461 else if (!prefixcmp(command_buf.buf, "progress "))
2462 cmd_progress();
2463 else
2464 die("Unsupported command: %s", command_buf.buf);
2465 }
2466 end_packfile();
2467
2468 dump_branches();
2469 dump_tags();
2470 unkeep_all_packs();
2471 dump_marks();
2472
2473 if (pack_edges)
2474 fclose(pack_edges);
2475
2476 if (show_stats) {
2477 uintmax_t total_count = 0, duplicate_count = 0;
2478 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2479 total_count += object_count_by_type[i];
2480 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2481 duplicate_count += duplicate_count_by_type[i];
2482
2483 fprintf(stderr, "%s statistics:\n", argv[0]);
2484 fprintf(stderr, "---------------------------------------------------------------------\n");
2485 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2486 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2487 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]);
2488 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]);
2489 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]);
2490 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]);
2491 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2492 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2493 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2494 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2495 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2496 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2497 fprintf(stderr, "---------------------------------------------------------------------\n");
2498 pack_report();
2499 fprintf(stderr, "---------------------------------------------------------------------\n");
2500 fprintf(stderr, "\n");
2501 }
2502
2503 return failure ? 1 : 0;
2504 }