archive.c: make archiver static
[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_%"PRIuMAX, (uintmax_t) 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: %"PRIuMAX"\n", (uintmax_t) getpid());
394 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) 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/pack/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/pack/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, CSUM_FSYNC);
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 NULL, 0);
956 close(pack_data->pack_fd);
957 idx_name = keep_pack(create_index());
958
959 /* Register the packfile with core git's machinary. */
960 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
961 if (!new_p)
962 die("core git rejected index %s", idx_name);
963 all_packs[pack_id] = new_p;
964 install_packed_git(new_p);
965
966 /* Print the boundary */
967 if (pack_edges) {
968 fprintf(pack_edges, "%s:", new_p->pack_name);
969 for (i = 0; i < branch_table_sz; i++) {
970 for (b = branch_table[i]; b; b = b->table_next_branch) {
971 if (b->pack_id == pack_id)
972 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
973 }
974 }
975 for (t = first_tag; t; t = t->next_tag) {
976 if (t->pack_id == pack_id)
977 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
978 }
979 fputc('\n', pack_edges);
980 fflush(pack_edges);
981 }
982
983 pack_id++;
984 }
985 else
986 unlink(old_p->pack_name);
987 free(old_p);
988
989 /* We can't carry a delta across packfiles. */
990 strbuf_release(&last_blob.data);
991 last_blob.offset = 0;
992 last_blob.depth = 0;
993 }
994
995 static void cycle_packfile(void)
996 {
997 end_packfile();
998 start_packfile();
999 }
1000
1001 static size_t encode_header(
1002 enum object_type type,
1003 size_t size,
1004 unsigned char *hdr)
1005 {
1006 int n = 1;
1007 unsigned char c;
1008
1009 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
1010 die("bad type %d", type);
1011
1012 c = (type << 4) | (size & 15);
1013 size >>= 4;
1014 while (size) {
1015 *hdr++ = c | 0x80;
1016 c = size & 0x7f;
1017 size >>= 7;
1018 n++;
1019 }
1020 *hdr = c;
1021 return n;
1022 }
1023
1024 static int store_object(
1025 enum object_type type,
1026 struct strbuf *dat,
1027 struct last_object *last,
1028 unsigned char *sha1out,
1029 uintmax_t mark)
1030 {
1031 void *out, *delta;
1032 struct object_entry *e;
1033 unsigned char hdr[96];
1034 unsigned char sha1[20];
1035 unsigned long hdrlen, deltalen;
1036 SHA_CTX c;
1037 z_stream s;
1038
1039 hdrlen = sprintf((char*)hdr,"%s %lu", typename(type),
1040 (unsigned long)dat->len) + 1;
1041 SHA1_Init(&c);
1042 SHA1_Update(&c, hdr, hdrlen);
1043 SHA1_Update(&c, dat->buf, dat->len);
1044 SHA1_Final(sha1, &c);
1045 if (sha1out)
1046 hashcpy(sha1out, sha1);
1047
1048 e = insert_object(sha1);
1049 if (mark)
1050 insert_mark(mark, e);
1051 if (e->offset) {
1052 duplicate_count_by_type[type]++;
1053 return 1;
1054 } else if (find_sha1_pack(sha1, packed_git)) {
1055 e->type = type;
1056 e->pack_id = MAX_PACK_ID;
1057 e->offset = 1; /* just not zero! */
1058 duplicate_count_by_type[type]++;
1059 return 1;
1060 }
1061
1062 if (last && last->data.buf && last->depth < max_depth) {
1063 delta = diff_delta(last->data.buf, last->data.len,
1064 dat->buf, dat->len,
1065 &deltalen, 0);
1066 if (delta && deltalen >= dat->len) {
1067 free(delta);
1068 delta = NULL;
1069 }
1070 } else
1071 delta = NULL;
1072
1073 memset(&s, 0, sizeof(s));
1074 deflateInit(&s, pack_compression_level);
1075 if (delta) {
1076 s.next_in = delta;
1077 s.avail_in = deltalen;
1078 } else {
1079 s.next_in = (void *)dat->buf;
1080 s.avail_in = dat->len;
1081 }
1082 s.avail_out = deflateBound(&s, s.avail_in);
1083 s.next_out = out = xmalloc(s.avail_out);
1084 while (deflate(&s, Z_FINISH) == Z_OK)
1085 /* nothing */;
1086 deflateEnd(&s);
1087
1088 /* Determine if we should auto-checkpoint. */
1089 if ((pack_size + 60 + s.total_out) > max_packsize
1090 || (pack_size + 60 + s.total_out) < pack_size) {
1091
1092 /* This new object needs to *not* have the current pack_id. */
1093 e->pack_id = pack_id + 1;
1094 cycle_packfile();
1095
1096 /* We cannot carry a delta into the new pack. */
1097 if (delta) {
1098 free(delta);
1099 delta = NULL;
1100
1101 memset(&s, 0, sizeof(s));
1102 deflateInit(&s, pack_compression_level);
1103 s.next_in = (void *)dat->buf;
1104 s.avail_in = dat->len;
1105 s.avail_out = deflateBound(&s, s.avail_in);
1106 s.next_out = out = xrealloc(out, s.avail_out);
1107 while (deflate(&s, Z_FINISH) == Z_OK)
1108 /* nothing */;
1109 deflateEnd(&s);
1110 }
1111 }
1112
1113 e->type = type;
1114 e->pack_id = pack_id;
1115 e->offset = pack_size;
1116 object_count++;
1117 object_count_by_type[type]++;
1118
1119 if (delta) {
1120 unsigned long ofs = e->offset - last->offset;
1121 unsigned pos = sizeof(hdr) - 1;
1122
1123 delta_count_by_type[type]++;
1124 e->depth = last->depth + 1;
1125
1126 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1127 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1128 pack_size += hdrlen;
1129
1130 hdr[pos] = ofs & 127;
1131 while (ofs >>= 7)
1132 hdr[--pos] = 128 | (--ofs & 127);
1133 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
1134 pack_size += sizeof(hdr) - pos;
1135 } else {
1136 e->depth = 0;
1137 hdrlen = encode_header(type, dat->len, hdr);
1138 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1139 pack_size += hdrlen;
1140 }
1141
1142 write_or_die(pack_data->pack_fd, out, s.total_out);
1143 pack_size += s.total_out;
1144
1145 free(out);
1146 free(delta);
1147 if (last) {
1148 if (last->no_swap) {
1149 last->data = *dat;
1150 } else {
1151 strbuf_swap(&last->data, dat);
1152 }
1153 last->offset = e->offset;
1154 last->depth = e->depth;
1155 }
1156 return 0;
1157 }
1158
1159 /* All calls must be guarded by find_object() or find_mark() to
1160 * ensure the 'struct object_entry' passed was written by this
1161 * process instance. We unpack the entry by the offset, avoiding
1162 * the need for the corresponding .idx file. This unpacking rule
1163 * works because we only use OBJ_REF_DELTA within the packfiles
1164 * created by fast-import.
1165 *
1166 * oe must not be NULL. Such an oe usually comes from giving
1167 * an unknown SHA-1 to find_object() or an undefined mark to
1168 * find_mark(). Callers must test for this condition and use
1169 * the standard read_sha1_file() when it happens.
1170 *
1171 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1172 * find_mark(), where the mark was reloaded from an existing marks
1173 * file and is referencing an object that this fast-import process
1174 * instance did not write out to a packfile. Callers must test for
1175 * this condition and use read_sha1_file() instead.
1176 */
1177 static void *gfi_unpack_entry(
1178 struct object_entry *oe,
1179 unsigned long *sizep)
1180 {
1181 enum object_type type;
1182 struct packed_git *p = all_packs[oe->pack_id];
1183 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1184 /* The object is stored in the packfile we are writing to
1185 * and we have modified it since the last time we scanned
1186 * back to read a previously written object. If an old
1187 * window covered [p->pack_size, p->pack_size + 20) its
1188 * data is stale and is not valid. Closing all windows
1189 * and updating the packfile length ensures we can read
1190 * the newly written data.
1191 */
1192 close_pack_windows(p);
1193
1194 /* We have to offer 20 bytes additional on the end of
1195 * the packfile as the core unpacker code assumes the
1196 * footer is present at the file end and must promise
1197 * at least 20 bytes within any window it maps. But
1198 * we don't actually create the footer here.
1199 */
1200 p->pack_size = pack_size + 20;
1201 }
1202 return unpack_entry(p, oe->offset, &type, sizep);
1203 }
1204
1205 static const char *get_mode(const char *str, uint16_t *modep)
1206 {
1207 unsigned char c;
1208 uint16_t mode = 0;
1209
1210 while ((c = *str++) != ' ') {
1211 if (c < '0' || c > '7')
1212 return NULL;
1213 mode = (mode << 3) + (c - '0');
1214 }
1215 *modep = mode;
1216 return str;
1217 }
1218
1219 static void load_tree(struct tree_entry *root)
1220 {
1221 unsigned char* sha1 = root->versions[1].sha1;
1222 struct object_entry *myoe;
1223 struct tree_content *t;
1224 unsigned long size;
1225 char *buf;
1226 const char *c;
1227
1228 root->tree = t = new_tree_content(8);
1229 if (is_null_sha1(sha1))
1230 return;
1231
1232 myoe = find_object(sha1);
1233 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1234 if (myoe->type != OBJ_TREE)
1235 die("Not a tree: %s", sha1_to_hex(sha1));
1236 t->delta_depth = myoe->depth;
1237 buf = gfi_unpack_entry(myoe, &size);
1238 if (!buf)
1239 die("Can't load tree %s", sha1_to_hex(sha1));
1240 } else {
1241 enum object_type type;
1242 buf = read_sha1_file(sha1, &type, &size);
1243 if (!buf || type != OBJ_TREE)
1244 die("Can't load tree %s", sha1_to_hex(sha1));
1245 }
1246
1247 c = buf;
1248 while (c != (buf + size)) {
1249 struct tree_entry *e = new_tree_entry();
1250
1251 if (t->entry_count == t->entry_capacity)
1252 root->tree = t = grow_tree_content(t, t->entry_count);
1253 t->entries[t->entry_count++] = e;
1254
1255 e->tree = NULL;
1256 c = get_mode(c, &e->versions[1].mode);
1257 if (!c)
1258 die("Corrupt mode in %s", sha1_to_hex(sha1));
1259 e->versions[0].mode = e->versions[1].mode;
1260 e->name = to_atom(c, strlen(c));
1261 c += e->name->str_len + 1;
1262 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1263 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1264 c += 20;
1265 }
1266 free(buf);
1267 }
1268
1269 static int tecmp0 (const void *_a, const void *_b)
1270 {
1271 struct tree_entry *a = *((struct tree_entry**)_a);
1272 struct tree_entry *b = *((struct tree_entry**)_b);
1273 return base_name_compare(
1274 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1275 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1276 }
1277
1278 static int tecmp1 (const void *_a, const void *_b)
1279 {
1280 struct tree_entry *a = *((struct tree_entry**)_a);
1281 struct tree_entry *b = *((struct tree_entry**)_b);
1282 return base_name_compare(
1283 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1284 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1285 }
1286
1287 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1288 {
1289 size_t maxlen = 0;
1290 unsigned int i;
1291
1292 if (!v)
1293 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1294 else
1295 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1296
1297 for (i = 0; i < t->entry_count; i++) {
1298 if (t->entries[i]->versions[v].mode)
1299 maxlen += t->entries[i]->name->str_len + 34;
1300 }
1301
1302 strbuf_reset(b);
1303 strbuf_grow(b, maxlen);
1304 for (i = 0; i < t->entry_count; i++) {
1305 struct tree_entry *e = t->entries[i];
1306 if (!e->versions[v].mode)
1307 continue;
1308 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1309 e->name->str_dat, '\0');
1310 strbuf_add(b, e->versions[v].sha1, 20);
1311 }
1312 }
1313
1314 static void store_tree(struct tree_entry *root)
1315 {
1316 struct tree_content *t = root->tree;
1317 unsigned int i, j, del;
1318 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1319 struct object_entry *le;
1320
1321 if (!is_null_sha1(root->versions[1].sha1))
1322 return;
1323
1324 for (i = 0; i < t->entry_count; i++) {
1325 if (t->entries[i]->tree)
1326 store_tree(t->entries[i]);
1327 }
1328
1329 le = find_object(root->versions[0].sha1);
1330 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1331 mktree(t, 0, &old_tree);
1332 lo.data = old_tree;
1333 lo.offset = le->offset;
1334 lo.depth = t->delta_depth;
1335 }
1336
1337 mktree(t, 1, &new_tree);
1338 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1339
1340 t->delta_depth = lo.depth;
1341 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1342 struct tree_entry *e = t->entries[i];
1343 if (e->versions[1].mode) {
1344 e->versions[0].mode = e->versions[1].mode;
1345 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1346 t->entries[j++] = e;
1347 } else {
1348 release_tree_entry(e);
1349 del++;
1350 }
1351 }
1352 t->entry_count -= del;
1353 }
1354
1355 static int tree_content_set(
1356 struct tree_entry *root,
1357 const char *p,
1358 const unsigned char *sha1,
1359 const uint16_t mode,
1360 struct tree_content *subtree)
1361 {
1362 struct tree_content *t = root->tree;
1363 const char *slash1;
1364 unsigned int i, n;
1365 struct tree_entry *e;
1366
1367 slash1 = strchr(p, '/');
1368 if (slash1)
1369 n = slash1 - p;
1370 else
1371 n = strlen(p);
1372 if (!n)
1373 die("Empty path component found in input");
1374 if (!slash1 && !S_ISDIR(mode) && subtree)
1375 die("Non-directories cannot have subtrees");
1376
1377 for (i = 0; i < t->entry_count; i++) {
1378 e = t->entries[i];
1379 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1380 if (!slash1) {
1381 if (!S_ISDIR(mode)
1382 && e->versions[1].mode == mode
1383 && !hashcmp(e->versions[1].sha1, sha1))
1384 return 0;
1385 e->versions[1].mode = mode;
1386 hashcpy(e->versions[1].sha1, sha1);
1387 if (e->tree)
1388 release_tree_content_recursive(e->tree);
1389 e->tree = subtree;
1390 hashclr(root->versions[1].sha1);
1391 return 1;
1392 }
1393 if (!S_ISDIR(e->versions[1].mode)) {
1394 e->tree = new_tree_content(8);
1395 e->versions[1].mode = S_IFDIR;
1396 }
1397 if (!e->tree)
1398 load_tree(e);
1399 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1400 hashclr(root->versions[1].sha1);
1401 return 1;
1402 }
1403 return 0;
1404 }
1405 }
1406
1407 if (t->entry_count == t->entry_capacity)
1408 root->tree = t = grow_tree_content(t, t->entry_count);
1409 e = new_tree_entry();
1410 e->name = to_atom(p, n);
1411 e->versions[0].mode = 0;
1412 hashclr(e->versions[0].sha1);
1413 t->entries[t->entry_count++] = e;
1414 if (slash1) {
1415 e->tree = new_tree_content(8);
1416 e->versions[1].mode = S_IFDIR;
1417 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1418 } else {
1419 e->tree = subtree;
1420 e->versions[1].mode = mode;
1421 hashcpy(e->versions[1].sha1, sha1);
1422 }
1423 hashclr(root->versions[1].sha1);
1424 return 1;
1425 }
1426
1427 static int tree_content_remove(
1428 struct tree_entry *root,
1429 const char *p,
1430 struct tree_entry *backup_leaf)
1431 {
1432 struct tree_content *t = root->tree;
1433 const char *slash1;
1434 unsigned int i, n;
1435 struct tree_entry *e;
1436
1437 slash1 = strchr(p, '/');
1438 if (slash1)
1439 n = slash1 - p;
1440 else
1441 n = strlen(p);
1442
1443 for (i = 0; i < t->entry_count; i++) {
1444 e = t->entries[i];
1445 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1446 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1447 goto del_entry;
1448 if (!e->tree)
1449 load_tree(e);
1450 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1451 for (n = 0; n < e->tree->entry_count; n++) {
1452 if (e->tree->entries[n]->versions[1].mode) {
1453 hashclr(root->versions[1].sha1);
1454 return 1;
1455 }
1456 }
1457 backup_leaf = NULL;
1458 goto del_entry;
1459 }
1460 return 0;
1461 }
1462 }
1463 return 0;
1464
1465 del_entry:
1466 if (backup_leaf)
1467 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1468 else if (e->tree)
1469 release_tree_content_recursive(e->tree);
1470 e->tree = NULL;
1471 e->versions[1].mode = 0;
1472 hashclr(e->versions[1].sha1);
1473 hashclr(root->versions[1].sha1);
1474 return 1;
1475 }
1476
1477 static int tree_content_get(
1478 struct tree_entry *root,
1479 const char *p,
1480 struct tree_entry *leaf)
1481 {
1482 struct tree_content *t = root->tree;
1483 const char *slash1;
1484 unsigned int i, n;
1485 struct tree_entry *e;
1486
1487 slash1 = strchr(p, '/');
1488 if (slash1)
1489 n = slash1 - p;
1490 else
1491 n = strlen(p);
1492
1493 for (i = 0; i < t->entry_count; i++) {
1494 e = t->entries[i];
1495 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1496 if (!slash1) {
1497 memcpy(leaf, e, sizeof(*leaf));
1498 if (e->tree && is_null_sha1(e->versions[1].sha1))
1499 leaf->tree = dup_tree_content(e->tree);
1500 else
1501 leaf->tree = NULL;
1502 return 1;
1503 }
1504 if (!S_ISDIR(e->versions[1].mode))
1505 return 0;
1506 if (!e->tree)
1507 load_tree(e);
1508 return tree_content_get(e, slash1 + 1, leaf);
1509 }
1510 }
1511 return 0;
1512 }
1513
1514 static int update_branch(struct branch *b)
1515 {
1516 static const char *msg = "fast-import";
1517 struct ref_lock *lock;
1518 unsigned char old_sha1[20];
1519
1520 if (is_null_sha1(b->sha1))
1521 return 0;
1522 if (read_ref(b->name, old_sha1))
1523 hashclr(old_sha1);
1524 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1525 if (!lock)
1526 return error("Unable to lock %s", b->name);
1527 if (!force_update && !is_null_sha1(old_sha1)) {
1528 struct commit *old_cmit, *new_cmit;
1529
1530 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1531 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1532 if (!old_cmit || !new_cmit) {
1533 unlock_ref(lock);
1534 return error("Branch %s is missing commits.", b->name);
1535 }
1536
1537 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1538 unlock_ref(lock);
1539 warning("Not updating %s"
1540 " (new tip %s does not contain %s)",
1541 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1542 return -1;
1543 }
1544 }
1545 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1546 return error("Unable to update %s", b->name);
1547 return 0;
1548 }
1549
1550 static void dump_branches(void)
1551 {
1552 unsigned int i;
1553 struct branch *b;
1554
1555 for (i = 0; i < branch_table_sz; i++) {
1556 for (b = branch_table[i]; b; b = b->table_next_branch)
1557 failure |= update_branch(b);
1558 }
1559 }
1560
1561 static void dump_tags(void)
1562 {
1563 static const char *msg = "fast-import";
1564 struct tag *t;
1565 struct ref_lock *lock;
1566 char ref_name[PATH_MAX];
1567
1568 for (t = first_tag; t; t = t->next_tag) {
1569 sprintf(ref_name, "tags/%s", t->name);
1570 lock = lock_ref_sha1(ref_name, NULL);
1571 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1572 failure |= error("Unable to update %s", ref_name);
1573 }
1574 }
1575
1576 static void dump_marks_helper(FILE *f,
1577 uintmax_t base,
1578 struct mark_set *m)
1579 {
1580 uintmax_t k;
1581 if (m->shift) {
1582 for (k = 0; k < 1024; k++) {
1583 if (m->data.sets[k])
1584 dump_marks_helper(f, (base + k) << m->shift,
1585 m->data.sets[k]);
1586 }
1587 } else {
1588 for (k = 0; k < 1024; k++) {
1589 if (m->data.marked[k])
1590 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1591 sha1_to_hex(m->data.marked[k]->sha1));
1592 }
1593 }
1594 }
1595
1596 static void dump_marks(void)
1597 {
1598 static struct lock_file mark_lock;
1599 int mark_fd;
1600 FILE *f;
1601
1602 if (!mark_file)
1603 return;
1604
1605 mark_fd = hold_lock_file_for_update(&mark_lock, mark_file, 0);
1606 if (mark_fd < 0) {
1607 failure |= error("Unable to write marks file %s: %s",
1608 mark_file, strerror(errno));
1609 return;
1610 }
1611
1612 f = fdopen(mark_fd, "w");
1613 if (!f) {
1614 int saved_errno = errno;
1615 rollback_lock_file(&mark_lock);
1616 failure |= error("Unable to write marks file %s: %s",
1617 mark_file, strerror(saved_errno));
1618 return;
1619 }
1620
1621 /*
1622 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1623 * Assign -1 to the lock file descriptor so that commit_lock_file()
1624 * won't try to close() it.
1625 */
1626 mark_lock.fd = -1;
1627
1628 dump_marks_helper(f, 0, marks);
1629 if (ferror(f) || fclose(f)) {
1630 int saved_errno = errno;
1631 rollback_lock_file(&mark_lock);
1632 failure |= error("Unable to write marks file %s: %s",
1633 mark_file, strerror(saved_errno));
1634 return;
1635 }
1636
1637 if (commit_lock_file(&mark_lock)) {
1638 int saved_errno = errno;
1639 rollback_lock_file(&mark_lock);
1640 failure |= error("Unable to commit marks file %s: %s",
1641 mark_file, strerror(saved_errno));
1642 return;
1643 }
1644 }
1645
1646 static int read_next_command(void)
1647 {
1648 static int stdin_eof = 0;
1649
1650 if (stdin_eof) {
1651 unread_command_buf = 0;
1652 return EOF;
1653 }
1654
1655 do {
1656 if (unread_command_buf) {
1657 unread_command_buf = 0;
1658 } else {
1659 struct recent_command *rc;
1660
1661 strbuf_detach(&command_buf, NULL);
1662 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1663 if (stdin_eof)
1664 return EOF;
1665
1666 rc = rc_free;
1667 if (rc)
1668 rc_free = rc->next;
1669 else {
1670 rc = cmd_hist.next;
1671 cmd_hist.next = rc->next;
1672 cmd_hist.next->prev = &cmd_hist;
1673 free(rc->buf);
1674 }
1675
1676 rc->buf = command_buf.buf;
1677 rc->prev = cmd_tail;
1678 rc->next = cmd_hist.prev;
1679 rc->prev->next = rc;
1680 cmd_tail = rc;
1681 }
1682 } while (command_buf.buf[0] == '#');
1683
1684 return 0;
1685 }
1686
1687 static void skip_optional_lf(void)
1688 {
1689 int term_char = fgetc(stdin);
1690 if (term_char != '\n' && term_char != EOF)
1691 ungetc(term_char, stdin);
1692 }
1693
1694 static void parse_mark(void)
1695 {
1696 if (!prefixcmp(command_buf.buf, "mark :")) {
1697 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1698 read_next_command();
1699 }
1700 else
1701 next_mark = 0;
1702 }
1703
1704 static void parse_data(struct strbuf *sb)
1705 {
1706 strbuf_reset(sb);
1707
1708 if (prefixcmp(command_buf.buf, "data "))
1709 die("Expected 'data n' command, found: %s", command_buf.buf);
1710
1711 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1712 char *term = xstrdup(command_buf.buf + 5 + 2);
1713 size_t term_len = command_buf.len - 5 - 2;
1714
1715 strbuf_detach(&command_buf, NULL);
1716 for (;;) {
1717 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1718 die("EOF in data (terminator '%s' not found)", term);
1719 if (term_len == command_buf.len
1720 && !strcmp(term, command_buf.buf))
1721 break;
1722 strbuf_addbuf(sb, &command_buf);
1723 strbuf_addch(sb, '\n');
1724 }
1725 free(term);
1726 }
1727 else {
1728 size_t n = 0, length;
1729
1730 length = strtoul(command_buf.buf + 5, NULL, 10);
1731
1732 while (n < length) {
1733 size_t s = strbuf_fread(sb, length - n, stdin);
1734 if (!s && feof(stdin))
1735 die("EOF in data (%lu bytes remaining)",
1736 (unsigned long)(length - n));
1737 n += s;
1738 }
1739 }
1740
1741 skip_optional_lf();
1742 }
1743
1744 static int validate_raw_date(const char *src, char *result, int maxlen)
1745 {
1746 const char *orig_src = src;
1747 char *endp, sign;
1748
1749 strtoul(src, &endp, 10);
1750 if (endp == src || *endp != ' ')
1751 return -1;
1752
1753 src = endp + 1;
1754 if (*src != '-' && *src != '+')
1755 return -1;
1756 sign = *src;
1757
1758 strtoul(src + 1, &endp, 10);
1759 if (endp == src || *endp || (endp - orig_src) >= maxlen)
1760 return -1;
1761
1762 strcpy(result, orig_src);
1763 return 0;
1764 }
1765
1766 static char *parse_ident(const char *buf)
1767 {
1768 const char *gt;
1769 size_t name_len;
1770 char *ident;
1771
1772 gt = strrchr(buf, '>');
1773 if (!gt)
1774 die("Missing > in ident string: %s", buf);
1775 gt++;
1776 if (*gt != ' ')
1777 die("Missing space after > in ident string: %s", buf);
1778 gt++;
1779 name_len = gt - buf;
1780 ident = xmalloc(name_len + 24);
1781 strncpy(ident, buf, name_len);
1782
1783 switch (whenspec) {
1784 case WHENSPEC_RAW:
1785 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1786 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1787 break;
1788 case WHENSPEC_RFC2822:
1789 if (parse_date(gt, ident + name_len, 24) < 0)
1790 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1791 break;
1792 case WHENSPEC_NOW:
1793 if (strcmp("now", gt))
1794 die("Date in ident must be 'now': %s", buf);
1795 datestamp(ident + name_len, 24);
1796 break;
1797 }
1798
1799 return ident;
1800 }
1801
1802 static void parse_new_blob(void)
1803 {
1804 static struct strbuf buf = STRBUF_INIT;
1805
1806 read_next_command();
1807 parse_mark();
1808 parse_data(&buf);
1809 store_object(OBJ_BLOB, &buf, &last_blob, NULL, next_mark);
1810 }
1811
1812 static void unload_one_branch(void)
1813 {
1814 while (cur_active_branches
1815 && cur_active_branches >= max_active_branches) {
1816 uintmax_t min_commit = ULONG_MAX;
1817 struct branch *e, *l = NULL, *p = NULL;
1818
1819 for (e = active_branches; e; e = e->active_next_branch) {
1820 if (e->last_commit < min_commit) {
1821 p = l;
1822 min_commit = e->last_commit;
1823 }
1824 l = e;
1825 }
1826
1827 if (p) {
1828 e = p->active_next_branch;
1829 p->active_next_branch = e->active_next_branch;
1830 } else {
1831 e = active_branches;
1832 active_branches = e->active_next_branch;
1833 }
1834 e->active = 0;
1835 e->active_next_branch = NULL;
1836 if (e->branch_tree.tree) {
1837 release_tree_content_recursive(e->branch_tree.tree);
1838 e->branch_tree.tree = NULL;
1839 }
1840 cur_active_branches--;
1841 }
1842 }
1843
1844 static void load_branch(struct branch *b)
1845 {
1846 load_tree(&b->branch_tree);
1847 if (!b->active) {
1848 b->active = 1;
1849 b->active_next_branch = active_branches;
1850 active_branches = b;
1851 cur_active_branches++;
1852 branch_load_count++;
1853 }
1854 }
1855
1856 static void file_change_m(struct branch *b)
1857 {
1858 const char *p = command_buf.buf + 2;
1859 static struct strbuf uq = STRBUF_INIT;
1860 const char *endp;
1861 struct object_entry *oe = oe;
1862 unsigned char sha1[20];
1863 uint16_t mode, inline_data = 0;
1864
1865 p = get_mode(p, &mode);
1866 if (!p)
1867 die("Corrupt mode: %s", command_buf.buf);
1868 switch (mode) {
1869 case S_IFREG | 0644:
1870 case S_IFREG | 0755:
1871 case S_IFLNK:
1872 case S_IFGITLINK:
1873 case 0644:
1874 case 0755:
1875 /* ok */
1876 break;
1877 default:
1878 die("Corrupt mode: %s", command_buf.buf);
1879 }
1880
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;
1894 }
1895 if (*p++ != ' ')
1896 die("Missing space after SHA1: %s", command_buf.buf);
1897
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;
1903 }
1904
1905 if (S_ISGITLINK(mode)) {
1906 if (inline_data)
1907 die("Git links cannot be specified 'inline': %s",
1908 command_buf.buf);
1909 else if (oe) {
1910 if (oe->type != OBJ_COMMIT)
1911 die("Not a commit (actually a %s): %s",
1912 typename(oe->type), command_buf.buf);
1913 }
1914 /*
1915 * Accept the sha1 without checking; it expected to be in
1916 * another repository.
1917 */
1918 } else if (inline_data) {
1919 static struct strbuf buf = STRBUF_INIT;
1920
1921 if (p != uq.buf) {
1922 strbuf_addstr(&uq, p);
1923 p = uq.buf;
1924 }
1925 read_next_command();
1926 parse_data(&buf);
1927 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
1928 } else if (oe) {
1929 if (oe->type != OBJ_BLOB)
1930 die("Not a blob (actually a %s): %s",
1931 typename(oe->type), command_buf.buf);
1932 } else {
1933 enum object_type type = sha1_object_info(sha1, NULL);
1934 if (type < 0)
1935 die("Blob not found: %s", command_buf.buf);
1936 if (type != OBJ_BLOB)
1937 die("Not a blob (actually a %s): %s",
1938 typename(type), command_buf.buf);
1939 }
1940
1941 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode, NULL);
1942 }
1943
1944 static void file_change_d(struct branch *b)
1945 {
1946 const char *p = command_buf.buf + 2;
1947 static struct strbuf uq = STRBUF_INIT;
1948 const char *endp;
1949
1950 strbuf_reset(&uq);
1951 if (!unquote_c_style(&uq, p, &endp)) {
1952 if (*endp)
1953 die("Garbage after path in: %s", command_buf.buf);
1954 p = uq.buf;
1955 }
1956 tree_content_remove(&b->branch_tree, p, NULL);
1957 }
1958
1959 static void file_change_cr(struct branch *b, int rename)
1960 {
1961 const char *s, *d;
1962 static struct strbuf s_uq = STRBUF_INIT;
1963 static struct strbuf d_uq = STRBUF_INIT;
1964 const char *endp;
1965 struct tree_entry leaf;
1966
1967 s = command_buf.buf + 2;
1968 strbuf_reset(&s_uq);
1969 if (!unquote_c_style(&s_uq, s, &endp)) {
1970 if (*endp != ' ')
1971 die("Missing space after source: %s", command_buf.buf);
1972 } else {
1973 endp = strchr(s, ' ');
1974 if (!endp)
1975 die("Missing space after source: %s", command_buf.buf);
1976 strbuf_add(&s_uq, s, endp - s);
1977 }
1978 s = s_uq.buf;
1979
1980 endp++;
1981 if (!*endp)
1982 die("Missing dest: %s", command_buf.buf);
1983
1984 d = endp;
1985 strbuf_reset(&d_uq);
1986 if (!unquote_c_style(&d_uq, d, &endp)) {
1987 if (*endp)
1988 die("Garbage after dest in: %s", command_buf.buf);
1989 d = d_uq.buf;
1990 }
1991
1992 memset(&leaf, 0, sizeof(leaf));
1993 if (rename)
1994 tree_content_remove(&b->branch_tree, s, &leaf);
1995 else
1996 tree_content_get(&b->branch_tree, s, &leaf);
1997 if (!leaf.versions[1].mode)
1998 die("Path %s not in branch", s);
1999 tree_content_set(&b->branch_tree, d,
2000 leaf.versions[1].sha1,
2001 leaf.versions[1].mode,
2002 leaf.tree);
2003 }
2004
2005 static void file_change_deleteall(struct branch *b)
2006 {
2007 release_tree_content_recursive(b->branch_tree.tree);
2008 hashclr(b->branch_tree.versions[0].sha1);
2009 hashclr(b->branch_tree.versions[1].sha1);
2010 load_tree(&b->branch_tree);
2011 }
2012
2013 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2014 {
2015 if (!buf || size < 46)
2016 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2017 if (memcmp("tree ", buf, 5)
2018 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2019 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2020 hashcpy(b->branch_tree.versions[0].sha1,
2021 b->branch_tree.versions[1].sha1);
2022 }
2023
2024 static void parse_from_existing(struct branch *b)
2025 {
2026 if (is_null_sha1(b->sha1)) {
2027 hashclr(b->branch_tree.versions[0].sha1);
2028 hashclr(b->branch_tree.versions[1].sha1);
2029 } else {
2030 unsigned long size;
2031 char *buf;
2032
2033 buf = read_object_with_reference(b->sha1,
2034 commit_type, &size, b->sha1);
2035 parse_from_commit(b, buf, size);
2036 free(buf);
2037 }
2038 }
2039
2040 static int parse_from(struct branch *b)
2041 {
2042 const char *from;
2043 struct branch *s;
2044
2045 if (prefixcmp(command_buf.buf, "from "))
2046 return 0;
2047
2048 if (b->branch_tree.tree) {
2049 release_tree_content_recursive(b->branch_tree.tree);
2050 b->branch_tree.tree = NULL;
2051 }
2052
2053 from = strchr(command_buf.buf, ' ') + 1;
2054 s = lookup_branch(from);
2055 if (b == s)
2056 die("Can't create a branch from itself: %s", b->name);
2057 else if (s) {
2058 unsigned char *t = s->branch_tree.versions[1].sha1;
2059 hashcpy(b->sha1, s->sha1);
2060 hashcpy(b->branch_tree.versions[0].sha1, t);
2061 hashcpy(b->branch_tree.versions[1].sha1, t);
2062 } else if (*from == ':') {
2063 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2064 struct object_entry *oe = find_mark(idnum);
2065 if (oe->type != OBJ_COMMIT)
2066 die("Mark :%" PRIuMAX " not a commit", idnum);
2067 hashcpy(b->sha1, oe->sha1);
2068 if (oe->pack_id != MAX_PACK_ID) {
2069 unsigned long size;
2070 char *buf = gfi_unpack_entry(oe, &size);
2071 parse_from_commit(b, buf, size);
2072 free(buf);
2073 } else
2074 parse_from_existing(b);
2075 } else if (!get_sha1(from, b->sha1))
2076 parse_from_existing(b);
2077 else
2078 die("Invalid ref name or SHA1 expression: %s", from);
2079
2080 read_next_command();
2081 return 1;
2082 }
2083
2084 static struct hash_list *parse_merge(unsigned int *count)
2085 {
2086 struct hash_list *list = NULL, *n, *e = e;
2087 const char *from;
2088 struct branch *s;
2089
2090 *count = 0;
2091 while (!prefixcmp(command_buf.buf, "merge ")) {
2092 from = strchr(command_buf.buf, ' ') + 1;
2093 n = xmalloc(sizeof(*n));
2094 s = lookup_branch(from);
2095 if (s)
2096 hashcpy(n->sha1, s->sha1);
2097 else if (*from == ':') {
2098 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2099 struct object_entry *oe = find_mark(idnum);
2100 if (oe->type != OBJ_COMMIT)
2101 die("Mark :%" PRIuMAX " not a commit", idnum);
2102 hashcpy(n->sha1, oe->sha1);
2103 } else if (!get_sha1(from, n->sha1)) {
2104 unsigned long size;
2105 char *buf = read_object_with_reference(n->sha1,
2106 commit_type, &size, n->sha1);
2107 if (!buf || size < 46)
2108 die("Not a valid commit: %s", from);
2109 free(buf);
2110 } else
2111 die("Invalid ref name or SHA1 expression: %s", from);
2112
2113 n->next = NULL;
2114 if (list)
2115 e->next = n;
2116 else
2117 list = n;
2118 e = n;
2119 (*count)++;
2120 read_next_command();
2121 }
2122 return list;
2123 }
2124
2125 static void parse_new_commit(void)
2126 {
2127 static struct strbuf msg = STRBUF_INIT;
2128 struct branch *b;
2129 char *sp;
2130 char *author = NULL;
2131 char *committer = NULL;
2132 struct hash_list *merge_list = NULL;
2133 unsigned int merge_count;
2134
2135 /* Obtain the branch name from the rest of our command */
2136 sp = strchr(command_buf.buf, ' ') + 1;
2137 b = lookup_branch(sp);
2138 if (!b)
2139 b = new_branch(sp);
2140
2141 read_next_command();
2142 parse_mark();
2143 if (!prefixcmp(command_buf.buf, "author ")) {
2144 author = parse_ident(command_buf.buf + 7);
2145 read_next_command();
2146 }
2147 if (!prefixcmp(command_buf.buf, "committer ")) {
2148 committer = parse_ident(command_buf.buf + 10);
2149 read_next_command();
2150 }
2151 if (!committer)
2152 die("Expected committer but didn't get one");
2153 parse_data(&msg);
2154 read_next_command();
2155 parse_from(b);
2156 merge_list = parse_merge(&merge_count);
2157
2158 /* ensure the branch is active/loaded */
2159 if (!b->branch_tree.tree || !max_active_branches) {
2160 unload_one_branch();
2161 load_branch(b);
2162 }
2163
2164 /* file_change* */
2165 while (command_buf.len > 0) {
2166 if (!prefixcmp(command_buf.buf, "M "))
2167 file_change_m(b);
2168 else if (!prefixcmp(command_buf.buf, "D "))
2169 file_change_d(b);
2170 else if (!prefixcmp(command_buf.buf, "R "))
2171 file_change_cr(b, 1);
2172 else if (!prefixcmp(command_buf.buf, "C "))
2173 file_change_cr(b, 0);
2174 else if (!strcmp("deleteall", command_buf.buf))
2175 file_change_deleteall(b);
2176 else {
2177 unread_command_buf = 1;
2178 break;
2179 }
2180 if (read_next_command() == EOF)
2181 break;
2182 }
2183
2184 /* build the tree and the commit */
2185 store_tree(&b->branch_tree);
2186 hashcpy(b->branch_tree.versions[0].sha1,
2187 b->branch_tree.versions[1].sha1);
2188
2189 strbuf_reset(&new_data);
2190 strbuf_addf(&new_data, "tree %s\n",
2191 sha1_to_hex(b->branch_tree.versions[1].sha1));
2192 if (!is_null_sha1(b->sha1))
2193 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2194 while (merge_list) {
2195 struct hash_list *next = merge_list->next;
2196 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2197 free(merge_list);
2198 merge_list = next;
2199 }
2200 strbuf_addf(&new_data,
2201 "author %s\n"
2202 "committer %s\n"
2203 "\n",
2204 author ? author : committer, committer);
2205 strbuf_addbuf(&new_data, &msg);
2206 free(author);
2207 free(committer);
2208
2209 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2210 b->pack_id = pack_id;
2211 b->last_commit = object_count_by_type[OBJ_COMMIT];
2212 }
2213
2214 static void parse_new_tag(void)
2215 {
2216 static struct strbuf msg = STRBUF_INIT;
2217 char *sp;
2218 const char *from;
2219 char *tagger;
2220 struct branch *s;
2221 struct tag *t;
2222 uintmax_t from_mark = 0;
2223 unsigned char sha1[20];
2224
2225 /* Obtain the new tag name from the rest of our command */
2226 sp = strchr(command_buf.buf, ' ') + 1;
2227 t = pool_alloc(sizeof(struct tag));
2228 t->next_tag = NULL;
2229 t->name = pool_strdup(sp);
2230 if (last_tag)
2231 last_tag->next_tag = t;
2232 else
2233 first_tag = t;
2234 last_tag = t;
2235 read_next_command();
2236
2237 /* from ... */
2238 if (prefixcmp(command_buf.buf, "from "))
2239 die("Expected from command, got %s", command_buf.buf);
2240 from = strchr(command_buf.buf, ' ') + 1;
2241 s = lookup_branch(from);
2242 if (s) {
2243 hashcpy(sha1, s->sha1);
2244 } else if (*from == ':') {
2245 struct object_entry *oe;
2246 from_mark = strtoumax(from + 1, NULL, 10);
2247 oe = find_mark(from_mark);
2248 if (oe->type != OBJ_COMMIT)
2249 die("Mark :%" PRIuMAX " not a commit", from_mark);
2250 hashcpy(sha1, oe->sha1);
2251 } else if (!get_sha1(from, sha1)) {
2252 unsigned long size;
2253 char *buf;
2254
2255 buf = read_object_with_reference(sha1,
2256 commit_type, &size, sha1);
2257 if (!buf || size < 46)
2258 die("Not a valid commit: %s", from);
2259 free(buf);
2260 } else
2261 die("Invalid ref name or SHA1 expression: %s", from);
2262 read_next_command();
2263
2264 /* tagger ... */
2265 if (prefixcmp(command_buf.buf, "tagger "))
2266 die("Expected tagger command, got %s", command_buf.buf);
2267 tagger = parse_ident(command_buf.buf + 7);
2268
2269 /* tag payload/message */
2270 read_next_command();
2271 parse_data(&msg);
2272
2273 /* build the tag object */
2274 strbuf_reset(&new_data);
2275 strbuf_addf(&new_data,
2276 "object %s\n"
2277 "type %s\n"
2278 "tag %s\n"
2279 "tagger %s\n"
2280 "\n",
2281 sha1_to_hex(sha1), commit_type, t->name, tagger);
2282 strbuf_addbuf(&new_data, &msg);
2283 free(tagger);
2284
2285 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2286 t->pack_id = MAX_PACK_ID;
2287 else
2288 t->pack_id = pack_id;
2289 }
2290
2291 static void parse_reset_branch(void)
2292 {
2293 struct branch *b;
2294 char *sp;
2295
2296 /* Obtain the branch name from the rest of our command */
2297 sp = strchr(command_buf.buf, ' ') + 1;
2298 b = lookup_branch(sp);
2299 if (b) {
2300 hashclr(b->sha1);
2301 hashclr(b->branch_tree.versions[0].sha1);
2302 hashclr(b->branch_tree.versions[1].sha1);
2303 if (b->branch_tree.tree) {
2304 release_tree_content_recursive(b->branch_tree.tree);
2305 b->branch_tree.tree = NULL;
2306 }
2307 }
2308 else
2309 b = new_branch(sp);
2310 read_next_command();
2311 parse_from(b);
2312 if (command_buf.len > 0)
2313 unread_command_buf = 1;
2314 }
2315
2316 static void parse_checkpoint(void)
2317 {
2318 if (object_count) {
2319 cycle_packfile();
2320 dump_branches();
2321 dump_tags();
2322 dump_marks();
2323 }
2324 skip_optional_lf();
2325 }
2326
2327 static void parse_progress(void)
2328 {
2329 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2330 fputc('\n', stdout);
2331 fflush(stdout);
2332 skip_optional_lf();
2333 }
2334
2335 static void import_marks(const char *input_file)
2336 {
2337 char line[512];
2338 FILE *f = fopen(input_file, "r");
2339 if (!f)
2340 die("cannot read %s: %s", input_file, strerror(errno));
2341 while (fgets(line, sizeof(line), f)) {
2342 uintmax_t mark;
2343 char *end;
2344 unsigned char sha1[20];
2345 struct object_entry *e;
2346
2347 end = strchr(line, '\n');
2348 if (line[0] != ':' || !end)
2349 die("corrupt mark line: %s", line);
2350 *end = 0;
2351 mark = strtoumax(line + 1, &end, 10);
2352 if (!mark || end == line + 1
2353 || *end != ' ' || get_sha1(end + 1, sha1))
2354 die("corrupt mark line: %s", line);
2355 e = find_object(sha1);
2356 if (!e) {
2357 enum object_type type = sha1_object_info(sha1, NULL);
2358 if (type < 0)
2359 die("object not found: %s", sha1_to_hex(sha1));
2360 e = insert_object(sha1);
2361 e->type = type;
2362 e->pack_id = MAX_PACK_ID;
2363 e->offset = 1; /* just not zero! */
2364 }
2365 insert_mark(mark, e);
2366 }
2367 fclose(f);
2368 }
2369
2370 static int git_pack_config(const char *k, const char *v, void *cb)
2371 {
2372 if (!strcmp(k, "pack.depth")) {
2373 max_depth = git_config_int(k, v);
2374 if (max_depth > MAX_DEPTH)
2375 max_depth = MAX_DEPTH;
2376 return 0;
2377 }
2378 if (!strcmp(k, "pack.compression")) {
2379 int level = git_config_int(k, v);
2380 if (level == -1)
2381 level = Z_DEFAULT_COMPRESSION;
2382 else if (level < 0 || level > Z_BEST_COMPRESSION)
2383 die("bad pack compression level %d", level);
2384 pack_compression_level = level;
2385 pack_compression_seen = 1;
2386 return 0;
2387 }
2388 return git_default_config(k, v, cb);
2389 }
2390
2391 static const char fast_import_usage[] =
2392 "git fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2393
2394 int main(int argc, const char **argv)
2395 {
2396 unsigned int i, show_stats = 1;
2397
2398 setup_git_directory();
2399 git_config(git_pack_config, NULL);
2400 if (!pack_compression_seen && core_compression_seen)
2401 pack_compression_level = core_compression_level;
2402
2403 alloc_objects(object_entry_alloc);
2404 strbuf_init(&command_buf, 0);
2405 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2406 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2407 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2408 marks = pool_calloc(1, sizeof(struct mark_set));
2409
2410 for (i = 1; i < argc; i++) {
2411 const char *a = argv[i];
2412
2413 if (*a != '-' || !strcmp(a, "--"))
2414 break;
2415 else if (!prefixcmp(a, "--date-format=")) {
2416 const char *fmt = a + 14;
2417 if (!strcmp(fmt, "raw"))
2418 whenspec = WHENSPEC_RAW;
2419 else if (!strcmp(fmt, "rfc2822"))
2420 whenspec = WHENSPEC_RFC2822;
2421 else if (!strcmp(fmt, "now"))
2422 whenspec = WHENSPEC_NOW;
2423 else
2424 die("unknown --date-format argument %s", fmt);
2425 }
2426 else if (!prefixcmp(a, "--max-pack-size="))
2427 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
2428 else if (!prefixcmp(a, "--depth=")) {
2429 max_depth = strtoul(a + 8, NULL, 0);
2430 if (max_depth > MAX_DEPTH)
2431 die("--depth cannot exceed %u", MAX_DEPTH);
2432 }
2433 else if (!prefixcmp(a, "--active-branches="))
2434 max_active_branches = strtoul(a + 18, NULL, 0);
2435 else if (!prefixcmp(a, "--import-marks="))
2436 import_marks(a + 15);
2437 else if (!prefixcmp(a, "--export-marks="))
2438 mark_file = a + 15;
2439 else if (!prefixcmp(a, "--export-pack-edges=")) {
2440 if (pack_edges)
2441 fclose(pack_edges);
2442 pack_edges = fopen(a + 20, "a");
2443 if (!pack_edges)
2444 die("Cannot open %s: %s", a + 20, strerror(errno));
2445 } else if (!strcmp(a, "--force"))
2446 force_update = 1;
2447 else if (!strcmp(a, "--quiet"))
2448 show_stats = 0;
2449 else if (!strcmp(a, "--stats"))
2450 show_stats = 1;
2451 else
2452 die("unknown option %s", a);
2453 }
2454 if (i != argc)
2455 usage(fast_import_usage);
2456
2457 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2458 for (i = 0; i < (cmd_save - 1); i++)
2459 rc_free[i].next = &rc_free[i + 1];
2460 rc_free[cmd_save - 1].next = NULL;
2461
2462 prepare_packed_git();
2463 start_packfile();
2464 set_die_routine(die_nicely);
2465 while (read_next_command() != EOF) {
2466 if (!strcmp("blob", command_buf.buf))
2467 parse_new_blob();
2468 else if (!prefixcmp(command_buf.buf, "commit "))
2469 parse_new_commit();
2470 else if (!prefixcmp(command_buf.buf, "tag "))
2471 parse_new_tag();
2472 else if (!prefixcmp(command_buf.buf, "reset "))
2473 parse_reset_branch();
2474 else if (!strcmp("checkpoint", command_buf.buf))
2475 parse_checkpoint();
2476 else if (!prefixcmp(command_buf.buf, "progress "))
2477 parse_progress();
2478 else
2479 die("Unsupported command: %s", command_buf.buf);
2480 }
2481 end_packfile();
2482
2483 dump_branches();
2484 dump_tags();
2485 unkeep_all_packs();
2486 dump_marks();
2487
2488 if (pack_edges)
2489 fclose(pack_edges);
2490
2491 if (show_stats) {
2492 uintmax_t total_count = 0, duplicate_count = 0;
2493 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2494 total_count += object_count_by_type[i];
2495 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2496 duplicate_count += duplicate_count_by_type[i];
2497
2498 fprintf(stderr, "%s statistics:\n", argv[0]);
2499 fprintf(stderr, "---------------------------------------------------------------------\n");
2500 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2501 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2502 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]);
2503 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]);
2504 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]);
2505 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]);
2506 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2507 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2508 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2509 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2510 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2511 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2512 fprintf(stderr, "---------------------------------------------------------------------\n");
2513 pack_report();
2514 fprintf(stderr, "---------------------------------------------------------------------\n");
2515 fprintf(stderr, "\n");
2516 }
2517
2518 return failure ? 1 : 0;
2519 }