Merge branch 'maint'
[git/git.git] / fast-import.c
1 /*
2 (See Documentation/git-fast-import.txt for maintained documentation.)
3 Format of STDIN stream:
4
5 stream ::= cmd*;
6
7 cmd ::= new_blob
8 | new_commit
9 | new_tag
10 | reset_branch
11 | checkpoint
12 | progress
13 ;
14
15 new_blob ::= 'blob' lf
16 mark?
17 file_content;
18 file_content ::= data;
19
20 new_commit ::= 'commit' sp ref_str lf
21 mark?
22 ('author' (sp name)? sp '<' email '>' sp when lf)?
23 'committer' (sp name)? sp '<' email '>' sp when lf
24 commit_msg
25 ('from' sp committish lf)?
26 ('merge' sp committish lf)*
27 file_change*
28 lf?;
29 commit_msg ::= data;
30
31 file_change ::= file_clr
32 | file_del
33 | file_rnm
34 | file_cpy
35 | file_obm
36 | file_inm;
37 file_clr ::= 'deleteall' lf;
38 file_del ::= 'D' sp path_str lf;
39 file_rnm ::= 'R' sp path_str sp path_str lf;
40 file_cpy ::= 'C' sp path_str sp path_str lf;
41 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
42 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
43 data;
44 note_obm ::= 'N' sp (hexsha1 | idnum) sp committish lf;
45 note_inm ::= 'N' sp 'inline' sp committish lf
46 data;
47
48 new_tag ::= 'tag' sp tag_str lf
49 'from' sp committish lf
50 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
51 tag_msg;
52 tag_msg ::= data;
53
54 reset_branch ::= 'reset' sp ref_str lf
55 ('from' sp committish lf)?
56 lf?;
57
58 checkpoint ::= 'checkpoint' lf
59 lf?;
60
61 progress ::= 'progress' sp not_lf* lf
62 lf?;
63
64 # note: the first idnum in a stream should be 1 and subsequent
65 # idnums should not have gaps between values as this will cause
66 # the stream parser to reserve space for the gapped values. An
67 # idnum can be updated in the future to a new object by issuing
68 # a new mark directive with the old idnum.
69 #
70 mark ::= 'mark' sp idnum lf;
71 data ::= (delimited_data | exact_data)
72 lf?;
73
74 # note: delim may be any string but must not contain lf.
75 # data_line may contain any data but must not be exactly
76 # delim.
77 delimited_data ::= 'data' sp '<<' delim lf
78 (data_line lf)*
79 delim lf;
80
81 # note: declen indicates the length of binary_data in bytes.
82 # declen does not include the lf preceding the binary data.
83 #
84 exact_data ::= 'data' sp declen lf
85 binary_data;
86
87 # note: quoted strings are C-style quoting supporting \c for
88 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
89 # is the signed byte value in octal. Note that the only
90 # characters which must actually be escaped to protect the
91 # stream formatting is: \, " and LF. Otherwise these values
92 # are UTF8.
93 #
94 committish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
95 ref_str ::= ref;
96 sha1exp_str ::= sha1exp;
97 tag_str ::= tag;
98 path_str ::= path | '"' quoted(path) '"' ;
99 mode ::= '100644' | '644'
100 | '100755' | '755'
101 | '120000'
102 ;
103
104 declen ::= # unsigned 32 bit value, ascii base10 notation;
105 bigint ::= # unsigned integer value, ascii base10 notation;
106 binary_data ::= # file content, not interpreted;
107
108 when ::= raw_when | rfc2822_when;
109 raw_when ::= ts sp tz;
110 rfc2822_when ::= # Valid RFC 2822 date and time;
111
112 sp ::= # ASCII space character;
113 lf ::= # ASCII newline (LF) character;
114
115 # note: a colon (':') must precede the numerical value assigned to
116 # an idnum. This is to distinguish it from a ref or tag name as
117 # GIT does not permit ':' in ref or tag strings.
118 #
119 idnum ::= ':' bigint;
120 path ::= # GIT style file path, e.g. "a/b/c";
121 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
122 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
123 sha1exp ::= # Any valid GIT SHA1 expression;
124 hexsha1 ::= # SHA1 in hexadecimal format;
125
126 # note: name and email are UTF8 strings, however name must not
127 # contain '<' or lf and email must not contain any of the
128 # following: '<', '>', lf.
129 #
130 name ::= # valid GIT author/committer name;
131 email ::= # valid GIT author/committer email;
132 ts ::= # time since the epoch in seconds, ascii base10 notation;
133 tz ::= # GIT style timezone;
134
135 # note: comments may appear anywhere in the input, except
136 # within a data command. Any form of the data command
137 # always escapes the related input from comment processing.
138 #
139 # In case it is not clear, the '#' that starts the comment
140 # must be the first character on that line (an lf
141 # preceded it).
142 #
143 comment ::= '#' not_lf* lf;
144 not_lf ::= # Any byte that is not ASCII newline (LF);
145 */
146
147 #include "builtin.h"
148 #include "cache.h"
149 #include "object.h"
150 #include "blob.h"
151 #include "tree.h"
152 #include "commit.h"
153 #include "delta.h"
154 #include "pack.h"
155 #include "refs.h"
156 #include "csum-file.h"
157 #include "quote.h"
158 #include "exec_cmd.h"
159
160 #define PACK_ID_BITS 16
161 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
162 #define DEPTH_BITS 13
163 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
164
165 struct object_entry
166 {
167 struct object_entry *next;
168 uint32_t offset;
169 uint32_t type : TYPE_BITS,
170 pack_id : PACK_ID_BITS,
171 depth : DEPTH_BITS;
172 unsigned char sha1[20];
173 };
174
175 struct object_entry_pool
176 {
177 struct object_entry_pool *next_pool;
178 struct object_entry *next_free;
179 struct object_entry *end;
180 struct object_entry entries[FLEX_ARRAY]; /* more */
181 };
182
183 struct mark_set
184 {
185 union {
186 struct object_entry *marked[1024];
187 struct mark_set *sets[1024];
188 } data;
189 unsigned int shift;
190 };
191
192 struct last_object
193 {
194 struct strbuf data;
195 uint32_t offset;
196 unsigned int depth;
197 unsigned no_swap : 1;
198 };
199
200 struct mem_pool
201 {
202 struct mem_pool *next_pool;
203 char *next_free;
204 char *end;
205 uintmax_t space[FLEX_ARRAY]; /* more */
206 };
207
208 struct atom_str
209 {
210 struct atom_str *next_atom;
211 unsigned short str_len;
212 char str_dat[FLEX_ARRAY]; /* more */
213 };
214
215 struct tree_content;
216 struct tree_entry
217 {
218 struct tree_content *tree;
219 struct atom_str *name;
220 struct tree_entry_ms
221 {
222 uint16_t mode;
223 unsigned char sha1[20];
224 } versions[2];
225 };
226
227 struct tree_content
228 {
229 unsigned int entry_capacity; /* must match avail_tree_content */
230 unsigned int entry_count;
231 unsigned int delta_depth;
232 struct tree_entry *entries[FLEX_ARRAY]; /* more */
233 };
234
235 struct avail_tree_content
236 {
237 unsigned int entry_capacity; /* must match tree_content */
238 struct avail_tree_content *next_avail;
239 };
240
241 struct branch
242 {
243 struct branch *table_next_branch;
244 struct branch *active_next_branch;
245 const char *name;
246 struct tree_entry branch_tree;
247 uintmax_t last_commit;
248 unsigned active : 1;
249 unsigned pack_id : PACK_ID_BITS;
250 unsigned char sha1[20];
251 };
252
253 struct tag
254 {
255 struct tag *next_tag;
256 const char *name;
257 unsigned int pack_id;
258 unsigned char sha1[20];
259 };
260
261 struct hash_list
262 {
263 struct hash_list *next;
264 unsigned char sha1[20];
265 };
266
267 typedef enum {
268 WHENSPEC_RAW = 1,
269 WHENSPEC_RFC2822,
270 WHENSPEC_NOW,
271 } whenspec_type;
272
273 struct recent_command
274 {
275 struct recent_command *prev;
276 struct recent_command *next;
277 char *buf;
278 };
279
280 /* Configured limits on output */
281 static unsigned long max_depth = 10;
282 static off_t max_packsize = (1LL << 32) - 1;
283 static int force_update;
284 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
285 static int pack_compression_seen;
286
287 /* Stats and misc. counters */
288 static uintmax_t alloc_count;
289 static uintmax_t marks_set_count;
290 static uintmax_t object_count_by_type[1 << TYPE_BITS];
291 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
292 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
293 static unsigned long object_count;
294 static unsigned long branch_count;
295 static unsigned long branch_load_count;
296 static int failure;
297 static FILE *pack_edges;
298 static unsigned int show_stats = 1;
299 static int global_argc;
300 static const char **global_argv;
301
302 /* Memory pools */
303 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
304 static size_t total_allocd;
305 static struct mem_pool *mem_pool;
306
307 /* Atom management */
308 static unsigned int atom_table_sz = 4451;
309 static unsigned int atom_cnt;
310 static struct atom_str **atom_table;
311
312 /* The .pack file being generated */
313 static unsigned int pack_id;
314 static struct packed_git *pack_data;
315 static struct packed_git **all_packs;
316 static unsigned long pack_size;
317
318 /* Table of objects we've written. */
319 static unsigned int object_entry_alloc = 5000;
320 static struct object_entry_pool *blocks;
321 static struct object_entry *object_table[1 << 16];
322 static struct mark_set *marks;
323 static const char *export_marks_file;
324 static const char *import_marks_file;
325 static int import_marks_file_from_stream;
326 static int relative_marks_paths;
327
328 /* Our last blob */
329 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
330
331 /* Tree management */
332 static unsigned int tree_entry_alloc = 1000;
333 static void *avail_tree_entry;
334 static unsigned int avail_tree_table_sz = 100;
335 static struct avail_tree_content **avail_tree_table;
336 static struct strbuf old_tree = STRBUF_INIT;
337 static struct strbuf new_tree = STRBUF_INIT;
338
339 /* Branch data */
340 static unsigned long max_active_branches = 5;
341 static unsigned long cur_active_branches;
342 static unsigned long branch_table_sz = 1039;
343 static struct branch **branch_table;
344 static struct branch *active_branches;
345
346 /* Tag data */
347 static struct tag *first_tag;
348 static struct tag *last_tag;
349
350 /* Input stream parsing */
351 static whenspec_type whenspec = WHENSPEC_RAW;
352 static struct strbuf command_buf = STRBUF_INIT;
353 static int unread_command_buf;
354 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
355 static struct recent_command *cmd_tail = &cmd_hist;
356 static struct recent_command *rc_free;
357 static unsigned int cmd_save = 100;
358 static uintmax_t next_mark;
359 static struct strbuf new_data = STRBUF_INIT;
360 static int seen_data_command;
361
362 static void parse_argv(void);
363
364 static void write_branch_report(FILE *rpt, struct branch *b)
365 {
366 fprintf(rpt, "%s:\n", b->name);
367
368 fprintf(rpt, " status :");
369 if (b->active)
370 fputs(" active", rpt);
371 if (b->branch_tree.tree)
372 fputs(" loaded", rpt);
373 if (is_null_sha1(b->branch_tree.versions[1].sha1))
374 fputs(" dirty", rpt);
375 fputc('\n', rpt);
376
377 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
378 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
379 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
380 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
381
382 fputs(" last pack : ", rpt);
383 if (b->pack_id < MAX_PACK_ID)
384 fprintf(rpt, "%u", b->pack_id);
385 fputc('\n', rpt);
386
387 fputc('\n', rpt);
388 }
389
390 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
391
392 static void write_crash_report(const char *err)
393 {
394 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
395 FILE *rpt = fopen(loc, "w");
396 struct branch *b;
397 unsigned long lu;
398 struct recent_command *rc;
399
400 if (!rpt) {
401 error("can't write crash report %s: %s", loc, strerror(errno));
402 return;
403 }
404
405 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
406
407 fprintf(rpt, "fast-import crash report:\n");
408 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
409 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
410 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
411 fputc('\n', rpt);
412
413 fputs("fatal: ", rpt);
414 fputs(err, rpt);
415 fputc('\n', rpt);
416
417 fputc('\n', rpt);
418 fputs("Most Recent Commands Before Crash\n", rpt);
419 fputs("---------------------------------\n", rpt);
420 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
421 if (rc->next == &cmd_hist)
422 fputs("* ", rpt);
423 else
424 fputs(" ", rpt);
425 fputs(rc->buf, rpt);
426 fputc('\n', rpt);
427 }
428
429 fputc('\n', rpt);
430 fputs("Active Branch LRU\n", rpt);
431 fputs("-----------------\n", rpt);
432 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
433 cur_active_branches,
434 max_active_branches);
435 fputc('\n', rpt);
436 fputs(" pos clock name\n", rpt);
437 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
438 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
439 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
440 ++lu, b->last_commit, b->name);
441
442 fputc('\n', rpt);
443 fputs("Inactive Branches\n", rpt);
444 fputs("-----------------\n", rpt);
445 for (lu = 0; lu < branch_table_sz; lu++) {
446 for (b = branch_table[lu]; b; b = b->table_next_branch)
447 write_branch_report(rpt, b);
448 }
449
450 if (first_tag) {
451 struct tag *tg;
452 fputc('\n', rpt);
453 fputs("Annotated Tags\n", rpt);
454 fputs("--------------\n", rpt);
455 for (tg = first_tag; tg; tg = tg->next_tag) {
456 fputs(sha1_to_hex(tg->sha1), rpt);
457 fputc(' ', rpt);
458 fputs(tg->name, rpt);
459 fputc('\n', rpt);
460 }
461 }
462
463 fputc('\n', rpt);
464 fputs("Marks\n", rpt);
465 fputs("-----\n", rpt);
466 if (export_marks_file)
467 fprintf(rpt, " exported to %s\n", export_marks_file);
468 else
469 dump_marks_helper(rpt, 0, marks);
470
471 fputc('\n', rpt);
472 fputs("-------------------\n", rpt);
473 fputs("END OF CRASH REPORT\n", rpt);
474 fclose(rpt);
475 }
476
477 static void end_packfile(void);
478 static void unkeep_all_packs(void);
479 static void dump_marks(void);
480
481 static NORETURN void die_nicely(const char *err, va_list params)
482 {
483 static int zombie;
484 char message[2 * PATH_MAX];
485
486 vsnprintf(message, sizeof(message), err, params);
487 fputs("fatal: ", stderr);
488 fputs(message, stderr);
489 fputc('\n', stderr);
490
491 if (!zombie) {
492 zombie = 1;
493 write_crash_report(message);
494 end_packfile();
495 unkeep_all_packs();
496 dump_marks();
497 }
498 exit(128);
499 }
500
501 static void alloc_objects(unsigned int cnt)
502 {
503 struct object_entry_pool *b;
504
505 b = xmalloc(sizeof(struct object_entry_pool)
506 + cnt * sizeof(struct object_entry));
507 b->next_pool = blocks;
508 b->next_free = b->entries;
509 b->end = b->entries + cnt;
510 blocks = b;
511 alloc_count += cnt;
512 }
513
514 static struct object_entry *new_object(unsigned char *sha1)
515 {
516 struct object_entry *e;
517
518 if (blocks->next_free == blocks->end)
519 alloc_objects(object_entry_alloc);
520
521 e = blocks->next_free++;
522 hashcpy(e->sha1, sha1);
523 return e;
524 }
525
526 static struct object_entry *find_object(unsigned char *sha1)
527 {
528 unsigned int h = sha1[0] << 8 | sha1[1];
529 struct object_entry *e;
530 for (e = object_table[h]; e; e = e->next)
531 if (!hashcmp(sha1, e->sha1))
532 return e;
533 return NULL;
534 }
535
536 static struct object_entry *insert_object(unsigned char *sha1)
537 {
538 unsigned int h = sha1[0] << 8 | sha1[1];
539 struct object_entry *e = object_table[h];
540 struct object_entry *p = NULL;
541
542 while (e) {
543 if (!hashcmp(sha1, e->sha1))
544 return e;
545 p = e;
546 e = e->next;
547 }
548
549 e = new_object(sha1);
550 e->next = NULL;
551 e->offset = 0;
552 if (p)
553 p->next = e;
554 else
555 object_table[h] = e;
556 return e;
557 }
558
559 static unsigned int hc_str(const char *s, size_t len)
560 {
561 unsigned int r = 0;
562 while (len-- > 0)
563 r = r * 31 + *s++;
564 return r;
565 }
566
567 static void *pool_alloc(size_t len)
568 {
569 struct mem_pool *p;
570 void *r;
571
572 /* round up to a 'uintmax_t' alignment */
573 if (len & (sizeof(uintmax_t) - 1))
574 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
575
576 for (p = mem_pool; p; p = p->next_pool)
577 if ((p->end - p->next_free >= len))
578 break;
579
580 if (!p) {
581 if (len >= (mem_pool_alloc/2)) {
582 total_allocd += len;
583 return xmalloc(len);
584 }
585 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
586 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
587 p->next_pool = mem_pool;
588 p->next_free = (char *) p->space;
589 p->end = p->next_free + mem_pool_alloc;
590 mem_pool = p;
591 }
592
593 r = p->next_free;
594 p->next_free += len;
595 return r;
596 }
597
598 static void *pool_calloc(size_t count, size_t size)
599 {
600 size_t len = count * size;
601 void *r = pool_alloc(len);
602 memset(r, 0, len);
603 return r;
604 }
605
606 static char *pool_strdup(const char *s)
607 {
608 char *r = pool_alloc(strlen(s) + 1);
609 strcpy(r, s);
610 return r;
611 }
612
613 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
614 {
615 struct mark_set *s = marks;
616 while ((idnum >> s->shift) >= 1024) {
617 s = pool_calloc(1, sizeof(struct mark_set));
618 s->shift = marks->shift + 10;
619 s->data.sets[0] = marks;
620 marks = s;
621 }
622 while (s->shift) {
623 uintmax_t i = idnum >> s->shift;
624 idnum -= i << s->shift;
625 if (!s->data.sets[i]) {
626 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
627 s->data.sets[i]->shift = s->shift - 10;
628 }
629 s = s->data.sets[i];
630 }
631 if (!s->data.marked[idnum])
632 marks_set_count++;
633 s->data.marked[idnum] = oe;
634 }
635
636 static struct object_entry *find_mark(uintmax_t idnum)
637 {
638 uintmax_t orig_idnum = idnum;
639 struct mark_set *s = marks;
640 struct object_entry *oe = NULL;
641 if ((idnum >> s->shift) < 1024) {
642 while (s && s->shift) {
643 uintmax_t i = idnum >> s->shift;
644 idnum -= i << s->shift;
645 s = s->data.sets[i];
646 }
647 if (s)
648 oe = s->data.marked[idnum];
649 }
650 if (!oe)
651 die("mark :%" PRIuMAX " not declared", orig_idnum);
652 return oe;
653 }
654
655 static struct atom_str *to_atom(const char *s, unsigned short len)
656 {
657 unsigned int hc = hc_str(s, len) % atom_table_sz;
658 struct atom_str *c;
659
660 for (c = atom_table[hc]; c; c = c->next_atom)
661 if (c->str_len == len && !strncmp(s, c->str_dat, len))
662 return c;
663
664 c = pool_alloc(sizeof(struct atom_str) + len + 1);
665 c->str_len = len;
666 strncpy(c->str_dat, s, len);
667 c->str_dat[len] = 0;
668 c->next_atom = atom_table[hc];
669 atom_table[hc] = c;
670 atom_cnt++;
671 return c;
672 }
673
674 static struct branch *lookup_branch(const char *name)
675 {
676 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
677 struct branch *b;
678
679 for (b = branch_table[hc]; b; b = b->table_next_branch)
680 if (!strcmp(name, b->name))
681 return b;
682 return NULL;
683 }
684
685 static struct branch *new_branch(const char *name)
686 {
687 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
688 struct branch *b = lookup_branch(name);
689
690 if (b)
691 die("Invalid attempt to create duplicate branch: %s", name);
692 switch (check_ref_format(name)) {
693 case 0: break; /* its valid */
694 case CHECK_REF_FORMAT_ONELEVEL:
695 break; /* valid, but too few '/', allow anyway */
696 default:
697 die("Branch name doesn't conform to GIT standards: %s", name);
698 }
699
700 b = pool_calloc(1, sizeof(struct branch));
701 b->name = pool_strdup(name);
702 b->table_next_branch = branch_table[hc];
703 b->branch_tree.versions[0].mode = S_IFDIR;
704 b->branch_tree.versions[1].mode = S_IFDIR;
705 b->active = 0;
706 b->pack_id = MAX_PACK_ID;
707 branch_table[hc] = b;
708 branch_count++;
709 return b;
710 }
711
712 static unsigned int hc_entries(unsigned int cnt)
713 {
714 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
715 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
716 }
717
718 static struct tree_content *new_tree_content(unsigned int cnt)
719 {
720 struct avail_tree_content *f, *l = NULL;
721 struct tree_content *t;
722 unsigned int hc = hc_entries(cnt);
723
724 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
725 if (f->entry_capacity >= cnt)
726 break;
727
728 if (f) {
729 if (l)
730 l->next_avail = f->next_avail;
731 else
732 avail_tree_table[hc] = f->next_avail;
733 } else {
734 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
735 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
736 f->entry_capacity = cnt;
737 }
738
739 t = (struct tree_content*)f;
740 t->entry_count = 0;
741 t->delta_depth = 0;
742 return t;
743 }
744
745 static void release_tree_entry(struct tree_entry *e);
746 static void release_tree_content(struct tree_content *t)
747 {
748 struct avail_tree_content *f = (struct avail_tree_content*)t;
749 unsigned int hc = hc_entries(f->entry_capacity);
750 f->next_avail = avail_tree_table[hc];
751 avail_tree_table[hc] = f;
752 }
753
754 static void release_tree_content_recursive(struct tree_content *t)
755 {
756 unsigned int i;
757 for (i = 0; i < t->entry_count; i++)
758 release_tree_entry(t->entries[i]);
759 release_tree_content(t);
760 }
761
762 static struct tree_content *grow_tree_content(
763 struct tree_content *t,
764 int amt)
765 {
766 struct tree_content *r = new_tree_content(t->entry_count + amt);
767 r->entry_count = t->entry_count;
768 r->delta_depth = t->delta_depth;
769 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
770 release_tree_content(t);
771 return r;
772 }
773
774 static struct tree_entry *new_tree_entry(void)
775 {
776 struct tree_entry *e;
777
778 if (!avail_tree_entry) {
779 unsigned int n = tree_entry_alloc;
780 total_allocd += n * sizeof(struct tree_entry);
781 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
782 while (n-- > 1) {
783 *((void**)e) = e + 1;
784 e++;
785 }
786 *((void**)e) = NULL;
787 }
788
789 e = avail_tree_entry;
790 avail_tree_entry = *((void**)e);
791 return e;
792 }
793
794 static void release_tree_entry(struct tree_entry *e)
795 {
796 if (e->tree)
797 release_tree_content_recursive(e->tree);
798 *((void**)e) = avail_tree_entry;
799 avail_tree_entry = e;
800 }
801
802 static struct tree_content *dup_tree_content(struct tree_content *s)
803 {
804 struct tree_content *d;
805 struct tree_entry *a, *b;
806 unsigned int i;
807
808 if (!s)
809 return NULL;
810 d = new_tree_content(s->entry_count);
811 for (i = 0; i < s->entry_count; i++) {
812 a = s->entries[i];
813 b = new_tree_entry();
814 memcpy(b, a, sizeof(*a));
815 if (a->tree && is_null_sha1(b->versions[1].sha1))
816 b->tree = dup_tree_content(a->tree);
817 else
818 b->tree = NULL;
819 d->entries[i] = b;
820 }
821 d->entry_count = s->entry_count;
822 d->delta_depth = s->delta_depth;
823
824 return d;
825 }
826
827 static void start_packfile(void)
828 {
829 static char tmpfile[PATH_MAX];
830 struct packed_git *p;
831 struct pack_header hdr;
832 int pack_fd;
833
834 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
835 "pack/tmp_pack_XXXXXX");
836 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
837 strcpy(p->pack_name, tmpfile);
838 p->pack_fd = pack_fd;
839
840 hdr.hdr_signature = htonl(PACK_SIGNATURE);
841 hdr.hdr_version = htonl(2);
842 hdr.hdr_entries = 0;
843 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
844
845 pack_data = p;
846 pack_size = sizeof(hdr);
847 object_count = 0;
848
849 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
850 all_packs[pack_id] = p;
851 }
852
853 static int oecmp (const void *a_, const void *b_)
854 {
855 struct object_entry *a = *((struct object_entry**)a_);
856 struct object_entry *b = *((struct object_entry**)b_);
857 return hashcmp(a->sha1, b->sha1);
858 }
859
860 static char *create_index(void)
861 {
862 static char tmpfile[PATH_MAX];
863 git_SHA_CTX ctx;
864 struct sha1file *f;
865 struct object_entry **idx, **c, **last, *e;
866 struct object_entry_pool *o;
867 uint32_t array[256];
868 int i, idx_fd;
869
870 /* Build the sorted table of object IDs. */
871 idx = xmalloc(object_count * sizeof(struct object_entry*));
872 c = idx;
873 for (o = blocks; o; o = o->next_pool)
874 for (e = o->next_free; e-- != o->entries;)
875 if (pack_id == e->pack_id)
876 *c++ = e;
877 last = idx + object_count;
878 if (c != last)
879 die("internal consistency error creating the index");
880 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
881
882 /* Generate the fan-out array. */
883 c = idx;
884 for (i = 0; i < 256; i++) {
885 struct object_entry **next = c;
886 while (next < last) {
887 if ((*next)->sha1[0] != i)
888 break;
889 next++;
890 }
891 array[i] = htonl(next - idx);
892 c = next;
893 }
894
895 idx_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
896 "pack/tmp_idx_XXXXXX");
897 f = sha1fd(idx_fd, tmpfile);
898 sha1write(f, array, 256 * sizeof(int));
899 git_SHA1_Init(&ctx);
900 for (c = idx; c != last; c++) {
901 uint32_t offset = htonl((*c)->offset);
902 sha1write(f, &offset, 4);
903 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
904 git_SHA1_Update(&ctx, (*c)->sha1, 20);
905 }
906 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
907 sha1close(f, NULL, CSUM_FSYNC);
908 free(idx);
909 git_SHA1_Final(pack_data->sha1, &ctx);
910 return tmpfile;
911 }
912
913 static char *keep_pack(char *curr_index_name)
914 {
915 static char name[PATH_MAX];
916 static const char *keep_msg = "fast-import";
917 int keep_fd;
918
919 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
920 if (keep_fd < 0)
921 die_errno("cannot create keep file");
922 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
923 if (close(keep_fd))
924 die_errno("failed to write keep file");
925
926 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
927 get_object_directory(), sha1_to_hex(pack_data->sha1));
928 if (move_temp_to_file(pack_data->pack_name, name))
929 die("cannot store pack file");
930
931 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
932 get_object_directory(), sha1_to_hex(pack_data->sha1));
933 if (move_temp_to_file(curr_index_name, name))
934 die("cannot store index file");
935 return name;
936 }
937
938 static void unkeep_all_packs(void)
939 {
940 static char name[PATH_MAX];
941 int k;
942
943 for (k = 0; k < pack_id; k++) {
944 struct packed_git *p = all_packs[k];
945 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
946 get_object_directory(), sha1_to_hex(p->sha1));
947 unlink_or_warn(name);
948 }
949 }
950
951 static void end_packfile(void)
952 {
953 struct packed_git *old_p = pack_data, *new_p;
954
955 clear_delta_base_cache();
956 if (object_count) {
957 char *idx_name;
958 int i;
959 struct branch *b;
960 struct tag *t;
961
962 close_pack_windows(pack_data);
963 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
964 pack_data->pack_name, object_count,
965 NULL, 0);
966 close(pack_data->pack_fd);
967 idx_name = keep_pack(create_index());
968
969 /* Register the packfile with core git's machinery. */
970 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
971 if (!new_p)
972 die("core git rejected index %s", idx_name);
973 all_packs[pack_id] = new_p;
974 install_packed_git(new_p);
975
976 /* Print the boundary */
977 if (pack_edges) {
978 fprintf(pack_edges, "%s:", new_p->pack_name);
979 for (i = 0; i < branch_table_sz; i++) {
980 for (b = branch_table[i]; b; b = b->table_next_branch) {
981 if (b->pack_id == pack_id)
982 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
983 }
984 }
985 for (t = first_tag; t; t = t->next_tag) {
986 if (t->pack_id == pack_id)
987 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
988 }
989 fputc('\n', pack_edges);
990 fflush(pack_edges);
991 }
992
993 pack_id++;
994 }
995 else {
996 close(old_p->pack_fd);
997 unlink_or_warn(old_p->pack_name);
998 }
999 free(old_p);
1000
1001 /* We can't carry a delta across packfiles. */
1002 strbuf_release(&last_blob.data);
1003 last_blob.offset = 0;
1004 last_blob.depth = 0;
1005 }
1006
1007 static void cycle_packfile(void)
1008 {
1009 end_packfile();
1010 start_packfile();
1011 }
1012
1013 static size_t encode_header(
1014 enum object_type type,
1015 size_t size,
1016 unsigned char *hdr)
1017 {
1018 int n = 1;
1019 unsigned char c;
1020
1021 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
1022 die("bad type %d", type);
1023
1024 c = (type << 4) | (size & 15);
1025 size >>= 4;
1026 while (size) {
1027 *hdr++ = c | 0x80;
1028 c = size & 0x7f;
1029 size >>= 7;
1030 n++;
1031 }
1032 *hdr = c;
1033 return n;
1034 }
1035
1036 static int store_object(
1037 enum object_type type,
1038 struct strbuf *dat,
1039 struct last_object *last,
1040 unsigned char *sha1out,
1041 uintmax_t mark)
1042 {
1043 void *out, *delta;
1044 struct object_entry *e;
1045 unsigned char hdr[96];
1046 unsigned char sha1[20];
1047 unsigned long hdrlen, deltalen;
1048 git_SHA_CTX c;
1049 z_stream s;
1050
1051 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1052 (unsigned long)dat->len) + 1;
1053 git_SHA1_Init(&c);
1054 git_SHA1_Update(&c, hdr, hdrlen);
1055 git_SHA1_Update(&c, dat->buf, dat->len);
1056 git_SHA1_Final(sha1, &c);
1057 if (sha1out)
1058 hashcpy(sha1out, sha1);
1059
1060 e = insert_object(sha1);
1061 if (mark)
1062 insert_mark(mark, e);
1063 if (e->offset) {
1064 duplicate_count_by_type[type]++;
1065 return 1;
1066 } else if (find_sha1_pack(sha1, packed_git)) {
1067 e->type = type;
1068 e->pack_id = MAX_PACK_ID;
1069 e->offset = 1; /* just not zero! */
1070 duplicate_count_by_type[type]++;
1071 return 1;
1072 }
1073
1074 if (last && last->data.buf && last->depth < max_depth) {
1075 delta = diff_delta(last->data.buf, last->data.len,
1076 dat->buf, dat->len,
1077 &deltalen, 0);
1078 if (delta && deltalen >= dat->len) {
1079 free(delta);
1080 delta = NULL;
1081 }
1082 } else
1083 delta = NULL;
1084
1085 memset(&s, 0, sizeof(s));
1086 deflateInit(&s, pack_compression_level);
1087 if (delta) {
1088 s.next_in = delta;
1089 s.avail_in = deltalen;
1090 } else {
1091 s.next_in = (void *)dat->buf;
1092 s.avail_in = dat->len;
1093 }
1094 s.avail_out = deflateBound(&s, s.avail_in);
1095 s.next_out = out = xmalloc(s.avail_out);
1096 while (deflate(&s, Z_FINISH) == Z_OK)
1097 /* nothing */;
1098 deflateEnd(&s);
1099
1100 /* Determine if we should auto-checkpoint. */
1101 if ((pack_size + 60 + s.total_out) > max_packsize
1102 || (pack_size + 60 + s.total_out) < pack_size) {
1103
1104 /* This new object needs to *not* have the current pack_id. */
1105 e->pack_id = pack_id + 1;
1106 cycle_packfile();
1107
1108 /* We cannot carry a delta into the new pack. */
1109 if (delta) {
1110 free(delta);
1111 delta = NULL;
1112
1113 memset(&s, 0, sizeof(s));
1114 deflateInit(&s, pack_compression_level);
1115 s.next_in = (void *)dat->buf;
1116 s.avail_in = dat->len;
1117 s.avail_out = deflateBound(&s, s.avail_in);
1118 s.next_out = out = xrealloc(out, s.avail_out);
1119 while (deflate(&s, Z_FINISH) == Z_OK)
1120 /* nothing */;
1121 deflateEnd(&s);
1122 }
1123 }
1124
1125 e->type = type;
1126 e->pack_id = pack_id;
1127 e->offset = pack_size;
1128 object_count++;
1129 object_count_by_type[type]++;
1130
1131 if (delta) {
1132 unsigned long ofs = e->offset - last->offset;
1133 unsigned pos = sizeof(hdr) - 1;
1134
1135 delta_count_by_type[type]++;
1136 e->depth = last->depth + 1;
1137
1138 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1139 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1140 pack_size += hdrlen;
1141
1142 hdr[pos] = ofs & 127;
1143 while (ofs >>= 7)
1144 hdr[--pos] = 128 | (--ofs & 127);
1145 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
1146 pack_size += sizeof(hdr) - pos;
1147 } else {
1148 e->depth = 0;
1149 hdrlen = encode_header(type, dat->len, hdr);
1150 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1151 pack_size += hdrlen;
1152 }
1153
1154 write_or_die(pack_data->pack_fd, out, s.total_out);
1155 pack_size += s.total_out;
1156
1157 free(out);
1158 free(delta);
1159 if (last) {
1160 if (last->no_swap) {
1161 last->data = *dat;
1162 } else {
1163 strbuf_swap(&last->data, dat);
1164 }
1165 last->offset = e->offset;
1166 last->depth = e->depth;
1167 }
1168 return 0;
1169 }
1170
1171 /* All calls must be guarded by find_object() or find_mark() to
1172 * ensure the 'struct object_entry' passed was written by this
1173 * process instance. We unpack the entry by the offset, avoiding
1174 * the need for the corresponding .idx file. This unpacking rule
1175 * works because we only use OBJ_REF_DELTA within the packfiles
1176 * created by fast-import.
1177 *
1178 * oe must not be NULL. Such an oe usually comes from giving
1179 * an unknown SHA-1 to find_object() or an undefined mark to
1180 * find_mark(). Callers must test for this condition and use
1181 * the standard read_sha1_file() when it happens.
1182 *
1183 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1184 * find_mark(), where the mark was reloaded from an existing marks
1185 * file and is referencing an object that this fast-import process
1186 * instance did not write out to a packfile. Callers must test for
1187 * this condition and use read_sha1_file() instead.
1188 */
1189 static void *gfi_unpack_entry(
1190 struct object_entry *oe,
1191 unsigned long *sizep)
1192 {
1193 enum object_type type;
1194 struct packed_git *p = all_packs[oe->pack_id];
1195 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1196 /* The object is stored in the packfile we are writing to
1197 * and we have modified it since the last time we scanned
1198 * back to read a previously written object. If an old
1199 * window covered [p->pack_size, p->pack_size + 20) its
1200 * data is stale and is not valid. Closing all windows
1201 * and updating the packfile length ensures we can read
1202 * the newly written data.
1203 */
1204 close_pack_windows(p);
1205
1206 /* We have to offer 20 bytes additional on the end of
1207 * the packfile as the core unpacker code assumes the
1208 * footer is present at the file end and must promise
1209 * at least 20 bytes within any window it maps. But
1210 * we don't actually create the footer here.
1211 */
1212 p->pack_size = pack_size + 20;
1213 }
1214 return unpack_entry(p, oe->offset, &type, sizep);
1215 }
1216
1217 static const char *get_mode(const char *str, uint16_t *modep)
1218 {
1219 unsigned char c;
1220 uint16_t mode = 0;
1221
1222 while ((c = *str++) != ' ') {
1223 if (c < '0' || c > '7')
1224 return NULL;
1225 mode = (mode << 3) + (c - '0');
1226 }
1227 *modep = mode;
1228 return str;
1229 }
1230
1231 static void load_tree(struct tree_entry *root)
1232 {
1233 unsigned char *sha1 = root->versions[1].sha1;
1234 struct object_entry *myoe;
1235 struct tree_content *t;
1236 unsigned long size;
1237 char *buf;
1238 const char *c;
1239
1240 root->tree = t = new_tree_content(8);
1241 if (is_null_sha1(sha1))
1242 return;
1243
1244 myoe = find_object(sha1);
1245 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1246 if (myoe->type != OBJ_TREE)
1247 die("Not a tree: %s", sha1_to_hex(sha1));
1248 t->delta_depth = myoe->depth;
1249 buf = gfi_unpack_entry(myoe, &size);
1250 if (!buf)
1251 die("Can't load tree %s", sha1_to_hex(sha1));
1252 } else {
1253 enum object_type type;
1254 buf = read_sha1_file(sha1, &type, &size);
1255 if (!buf || type != OBJ_TREE)
1256 die("Can't load tree %s", sha1_to_hex(sha1));
1257 }
1258
1259 c = buf;
1260 while (c != (buf + size)) {
1261 struct tree_entry *e = new_tree_entry();
1262
1263 if (t->entry_count == t->entry_capacity)
1264 root->tree = t = grow_tree_content(t, t->entry_count);
1265 t->entries[t->entry_count++] = e;
1266
1267 e->tree = NULL;
1268 c = get_mode(c, &e->versions[1].mode);
1269 if (!c)
1270 die("Corrupt mode in %s", sha1_to_hex(sha1));
1271 e->versions[0].mode = e->versions[1].mode;
1272 e->name = to_atom(c, strlen(c));
1273 c += e->name->str_len + 1;
1274 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1275 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1276 c += 20;
1277 }
1278 free(buf);
1279 }
1280
1281 static int tecmp0 (const void *_a, const void *_b)
1282 {
1283 struct tree_entry *a = *((struct tree_entry**)_a);
1284 struct tree_entry *b = *((struct tree_entry**)_b);
1285 return base_name_compare(
1286 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1287 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1288 }
1289
1290 static int tecmp1 (const void *_a, const void *_b)
1291 {
1292 struct tree_entry *a = *((struct tree_entry**)_a);
1293 struct tree_entry *b = *((struct tree_entry**)_b);
1294 return base_name_compare(
1295 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1296 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1297 }
1298
1299 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1300 {
1301 size_t maxlen = 0;
1302 unsigned int i;
1303
1304 if (!v)
1305 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1306 else
1307 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1308
1309 for (i = 0; i < t->entry_count; i++) {
1310 if (t->entries[i]->versions[v].mode)
1311 maxlen += t->entries[i]->name->str_len + 34;
1312 }
1313
1314 strbuf_reset(b);
1315 strbuf_grow(b, maxlen);
1316 for (i = 0; i < t->entry_count; i++) {
1317 struct tree_entry *e = t->entries[i];
1318 if (!e->versions[v].mode)
1319 continue;
1320 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1321 e->name->str_dat, '\0');
1322 strbuf_add(b, e->versions[v].sha1, 20);
1323 }
1324 }
1325
1326 static void store_tree(struct tree_entry *root)
1327 {
1328 struct tree_content *t = root->tree;
1329 unsigned int i, j, del;
1330 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1331 struct object_entry *le;
1332
1333 if (!is_null_sha1(root->versions[1].sha1))
1334 return;
1335
1336 for (i = 0; i < t->entry_count; i++) {
1337 if (t->entries[i]->tree)
1338 store_tree(t->entries[i]);
1339 }
1340
1341 le = find_object(root->versions[0].sha1);
1342 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1343 mktree(t, 0, &old_tree);
1344 lo.data = old_tree;
1345 lo.offset = le->offset;
1346 lo.depth = t->delta_depth;
1347 }
1348
1349 mktree(t, 1, &new_tree);
1350 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1351
1352 t->delta_depth = lo.depth;
1353 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1354 struct tree_entry *e = t->entries[i];
1355 if (e->versions[1].mode) {
1356 e->versions[0].mode = e->versions[1].mode;
1357 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1358 t->entries[j++] = e;
1359 } else {
1360 release_tree_entry(e);
1361 del++;
1362 }
1363 }
1364 t->entry_count -= del;
1365 }
1366
1367 static int tree_content_set(
1368 struct tree_entry *root,
1369 const char *p,
1370 const unsigned char *sha1,
1371 const uint16_t mode,
1372 struct tree_content *subtree)
1373 {
1374 struct tree_content *t = root->tree;
1375 const char *slash1;
1376 unsigned int i, n;
1377 struct tree_entry *e;
1378
1379 slash1 = strchr(p, '/');
1380 if (slash1)
1381 n = slash1 - p;
1382 else
1383 n = strlen(p);
1384 if (!n)
1385 die("Empty path component found in input");
1386 if (!slash1 && !S_ISDIR(mode) && subtree)
1387 die("Non-directories cannot have subtrees");
1388
1389 for (i = 0; i < t->entry_count; i++) {
1390 e = t->entries[i];
1391 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1392 if (!slash1) {
1393 if (!S_ISDIR(mode)
1394 && e->versions[1].mode == mode
1395 && !hashcmp(e->versions[1].sha1, sha1))
1396 return 0;
1397 e->versions[1].mode = mode;
1398 hashcpy(e->versions[1].sha1, sha1);
1399 if (e->tree)
1400 release_tree_content_recursive(e->tree);
1401 e->tree = subtree;
1402 hashclr(root->versions[1].sha1);
1403 return 1;
1404 }
1405 if (!S_ISDIR(e->versions[1].mode)) {
1406 e->tree = new_tree_content(8);
1407 e->versions[1].mode = S_IFDIR;
1408 }
1409 if (!e->tree)
1410 load_tree(e);
1411 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1412 hashclr(root->versions[1].sha1);
1413 return 1;
1414 }
1415 return 0;
1416 }
1417 }
1418
1419 if (t->entry_count == t->entry_capacity)
1420 root->tree = t = grow_tree_content(t, t->entry_count);
1421 e = new_tree_entry();
1422 e->name = to_atom(p, n);
1423 e->versions[0].mode = 0;
1424 hashclr(e->versions[0].sha1);
1425 t->entries[t->entry_count++] = e;
1426 if (slash1) {
1427 e->tree = new_tree_content(8);
1428 e->versions[1].mode = S_IFDIR;
1429 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1430 } else {
1431 e->tree = subtree;
1432 e->versions[1].mode = mode;
1433 hashcpy(e->versions[1].sha1, sha1);
1434 }
1435 hashclr(root->versions[1].sha1);
1436 return 1;
1437 }
1438
1439 static int tree_content_remove(
1440 struct tree_entry *root,
1441 const char *p,
1442 struct tree_entry *backup_leaf)
1443 {
1444 struct tree_content *t = root->tree;
1445 const char *slash1;
1446 unsigned int i, n;
1447 struct tree_entry *e;
1448
1449 slash1 = strchr(p, '/');
1450 if (slash1)
1451 n = slash1 - p;
1452 else
1453 n = strlen(p);
1454
1455 for (i = 0; i < t->entry_count; i++) {
1456 e = t->entries[i];
1457 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1458 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1459 goto del_entry;
1460 if (!e->tree)
1461 load_tree(e);
1462 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1463 for (n = 0; n < e->tree->entry_count; n++) {
1464 if (e->tree->entries[n]->versions[1].mode) {
1465 hashclr(root->versions[1].sha1);
1466 return 1;
1467 }
1468 }
1469 backup_leaf = NULL;
1470 goto del_entry;
1471 }
1472 return 0;
1473 }
1474 }
1475 return 0;
1476
1477 del_entry:
1478 if (backup_leaf)
1479 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1480 else if (e->tree)
1481 release_tree_content_recursive(e->tree);
1482 e->tree = NULL;
1483 e->versions[1].mode = 0;
1484 hashclr(e->versions[1].sha1);
1485 hashclr(root->versions[1].sha1);
1486 return 1;
1487 }
1488
1489 static int tree_content_get(
1490 struct tree_entry *root,
1491 const char *p,
1492 struct tree_entry *leaf)
1493 {
1494 struct tree_content *t = root->tree;
1495 const char *slash1;
1496 unsigned int i, n;
1497 struct tree_entry *e;
1498
1499 slash1 = strchr(p, '/');
1500 if (slash1)
1501 n = slash1 - p;
1502 else
1503 n = strlen(p);
1504
1505 for (i = 0; i < t->entry_count; i++) {
1506 e = t->entries[i];
1507 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1508 if (!slash1) {
1509 memcpy(leaf, e, sizeof(*leaf));
1510 if (e->tree && is_null_sha1(e->versions[1].sha1))
1511 leaf->tree = dup_tree_content(e->tree);
1512 else
1513 leaf->tree = NULL;
1514 return 1;
1515 }
1516 if (!S_ISDIR(e->versions[1].mode))
1517 return 0;
1518 if (!e->tree)
1519 load_tree(e);
1520 return tree_content_get(e, slash1 + 1, leaf);
1521 }
1522 }
1523 return 0;
1524 }
1525
1526 static int update_branch(struct branch *b)
1527 {
1528 static const char *msg = "fast-import";
1529 struct ref_lock *lock;
1530 unsigned char old_sha1[20];
1531
1532 if (is_null_sha1(b->sha1))
1533 return 0;
1534 if (read_ref(b->name, old_sha1))
1535 hashclr(old_sha1);
1536 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1537 if (!lock)
1538 return error("Unable to lock %s", b->name);
1539 if (!force_update && !is_null_sha1(old_sha1)) {
1540 struct commit *old_cmit, *new_cmit;
1541
1542 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1543 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1544 if (!old_cmit || !new_cmit) {
1545 unlock_ref(lock);
1546 return error("Branch %s is missing commits.", b->name);
1547 }
1548
1549 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1550 unlock_ref(lock);
1551 warning("Not updating %s"
1552 " (new tip %s does not contain %s)",
1553 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1554 return -1;
1555 }
1556 }
1557 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1558 return error("Unable to update %s", b->name);
1559 return 0;
1560 }
1561
1562 static void dump_branches(void)
1563 {
1564 unsigned int i;
1565 struct branch *b;
1566
1567 for (i = 0; i < branch_table_sz; i++) {
1568 for (b = branch_table[i]; b; b = b->table_next_branch)
1569 failure |= update_branch(b);
1570 }
1571 }
1572
1573 static void dump_tags(void)
1574 {
1575 static const char *msg = "fast-import";
1576 struct tag *t;
1577 struct ref_lock *lock;
1578 char ref_name[PATH_MAX];
1579
1580 for (t = first_tag; t; t = t->next_tag) {
1581 sprintf(ref_name, "tags/%s", t->name);
1582 lock = lock_ref_sha1(ref_name, NULL);
1583 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1584 failure |= error("Unable to update %s", ref_name);
1585 }
1586 }
1587
1588 static void dump_marks_helper(FILE *f,
1589 uintmax_t base,
1590 struct mark_set *m)
1591 {
1592 uintmax_t k;
1593 if (m->shift) {
1594 for (k = 0; k < 1024; k++) {
1595 if (m->data.sets[k])
1596 dump_marks_helper(f, (base + k) << m->shift,
1597 m->data.sets[k]);
1598 }
1599 } else {
1600 for (k = 0; k < 1024; k++) {
1601 if (m->data.marked[k])
1602 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1603 sha1_to_hex(m->data.marked[k]->sha1));
1604 }
1605 }
1606 }
1607
1608 static void dump_marks(void)
1609 {
1610 static struct lock_file mark_lock;
1611 int mark_fd;
1612 FILE *f;
1613
1614 if (!export_marks_file)
1615 return;
1616
1617 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1618 if (mark_fd < 0) {
1619 failure |= error("Unable to write marks file %s: %s",
1620 export_marks_file, strerror(errno));
1621 return;
1622 }
1623
1624 f = fdopen(mark_fd, "w");
1625 if (!f) {
1626 int saved_errno = errno;
1627 rollback_lock_file(&mark_lock);
1628 failure |= error("Unable to write marks file %s: %s",
1629 export_marks_file, strerror(saved_errno));
1630 return;
1631 }
1632
1633 /*
1634 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1635 * Assign -1 to the lock file descriptor so that commit_lock_file()
1636 * won't try to close() it.
1637 */
1638 mark_lock.fd = -1;
1639
1640 dump_marks_helper(f, 0, marks);
1641 if (ferror(f) || fclose(f)) {
1642 int saved_errno = errno;
1643 rollback_lock_file(&mark_lock);
1644 failure |= error("Unable to write marks file %s: %s",
1645 export_marks_file, strerror(saved_errno));
1646 return;
1647 }
1648
1649 if (commit_lock_file(&mark_lock)) {
1650 int saved_errno = errno;
1651 rollback_lock_file(&mark_lock);
1652 failure |= error("Unable to commit marks file %s: %s",
1653 export_marks_file, strerror(saved_errno));
1654 return;
1655 }
1656 }
1657
1658 static void read_marks(void)
1659 {
1660 char line[512];
1661 FILE *f = fopen(import_marks_file, "r");
1662 if (!f)
1663 die_errno("cannot read '%s'", import_marks_file);
1664 while (fgets(line, sizeof(line), f)) {
1665 uintmax_t mark;
1666 char *end;
1667 unsigned char sha1[20];
1668 struct object_entry *e;
1669
1670 end = strchr(line, '\n');
1671 if (line[0] != ':' || !end)
1672 die("corrupt mark line: %s", line);
1673 *end = 0;
1674 mark = strtoumax(line + 1, &end, 10);
1675 if (!mark || end == line + 1
1676 || *end != ' ' || get_sha1(end + 1, sha1))
1677 die("corrupt mark line: %s", line);
1678 e = find_object(sha1);
1679 if (!e) {
1680 enum object_type type = sha1_object_info(sha1, NULL);
1681 if (type < 0)
1682 die("object not found: %s", sha1_to_hex(sha1));
1683 e = insert_object(sha1);
1684 e->type = type;
1685 e->pack_id = MAX_PACK_ID;
1686 e->offset = 1; /* just not zero! */
1687 }
1688 insert_mark(mark, e);
1689 }
1690 fclose(f);
1691 }
1692
1693
1694 static int read_next_command(void)
1695 {
1696 static int stdin_eof = 0;
1697
1698 if (stdin_eof) {
1699 unread_command_buf = 0;
1700 return EOF;
1701 }
1702
1703 do {
1704 if (unread_command_buf) {
1705 unread_command_buf = 0;
1706 } else {
1707 struct recent_command *rc;
1708
1709 strbuf_detach(&command_buf, NULL);
1710 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1711 if (stdin_eof)
1712 return EOF;
1713
1714 if (!seen_data_command
1715 && prefixcmp(command_buf.buf, "feature ")
1716 && prefixcmp(command_buf.buf, "option ")) {
1717 parse_argv();
1718 }
1719
1720 rc = rc_free;
1721 if (rc)
1722 rc_free = rc->next;
1723 else {
1724 rc = cmd_hist.next;
1725 cmd_hist.next = rc->next;
1726 cmd_hist.next->prev = &cmd_hist;
1727 free(rc->buf);
1728 }
1729
1730 rc->buf = command_buf.buf;
1731 rc->prev = cmd_tail;
1732 rc->next = cmd_hist.prev;
1733 rc->prev->next = rc;
1734 cmd_tail = rc;
1735 }
1736 } while (command_buf.buf[0] == '#');
1737
1738 return 0;
1739 }
1740
1741 static void skip_optional_lf(void)
1742 {
1743 int term_char = fgetc(stdin);
1744 if (term_char != '\n' && term_char != EOF)
1745 ungetc(term_char, stdin);
1746 }
1747
1748 static void parse_mark(void)
1749 {
1750 if (!prefixcmp(command_buf.buf, "mark :")) {
1751 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1752 read_next_command();
1753 }
1754 else
1755 next_mark = 0;
1756 }
1757
1758 static void parse_data(struct strbuf *sb)
1759 {
1760 strbuf_reset(sb);
1761
1762 if (prefixcmp(command_buf.buf, "data "))
1763 die("Expected 'data n' command, found: %s", command_buf.buf);
1764
1765 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1766 char *term = xstrdup(command_buf.buf + 5 + 2);
1767 size_t term_len = command_buf.len - 5 - 2;
1768
1769 strbuf_detach(&command_buf, NULL);
1770 for (;;) {
1771 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1772 die("EOF in data (terminator '%s' not found)", term);
1773 if (term_len == command_buf.len
1774 && !strcmp(term, command_buf.buf))
1775 break;
1776 strbuf_addbuf(sb, &command_buf);
1777 strbuf_addch(sb, '\n');
1778 }
1779 free(term);
1780 }
1781 else {
1782 size_t n = 0, length;
1783
1784 length = strtoul(command_buf.buf + 5, NULL, 10);
1785
1786 while (n < length) {
1787 size_t s = strbuf_fread(sb, length - n, stdin);
1788 if (!s && feof(stdin))
1789 die("EOF in data (%lu bytes remaining)",
1790 (unsigned long)(length - n));
1791 n += s;
1792 }
1793 }
1794
1795 skip_optional_lf();
1796 }
1797
1798 static int validate_raw_date(const char *src, char *result, int maxlen)
1799 {
1800 const char *orig_src = src;
1801 char *endp;
1802 unsigned long num;
1803
1804 errno = 0;
1805
1806 num = strtoul(src, &endp, 10);
1807 /* NEEDSWORK: perhaps check for reasonable values? */
1808 if (errno || endp == src || *endp != ' ')
1809 return -1;
1810
1811 src = endp + 1;
1812 if (*src != '-' && *src != '+')
1813 return -1;
1814
1815 num = strtoul(src + 1, &endp, 10);
1816 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1817 1400 < num)
1818 return -1;
1819
1820 strcpy(result, orig_src);
1821 return 0;
1822 }
1823
1824 static char *parse_ident(const char *buf)
1825 {
1826 const char *gt;
1827 size_t name_len;
1828 char *ident;
1829
1830 gt = strrchr(buf, '>');
1831 if (!gt)
1832 die("Missing > in ident string: %s", buf);
1833 gt++;
1834 if (*gt != ' ')
1835 die("Missing space after > in ident string: %s", buf);
1836 gt++;
1837 name_len = gt - buf;
1838 ident = xmalloc(name_len + 24);
1839 strncpy(ident, buf, name_len);
1840
1841 switch (whenspec) {
1842 case WHENSPEC_RAW:
1843 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1844 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1845 break;
1846 case WHENSPEC_RFC2822:
1847 if (parse_date(gt, ident + name_len, 24) < 0)
1848 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1849 break;
1850 case WHENSPEC_NOW:
1851 if (strcmp("now", gt))
1852 die("Date in ident must be 'now': %s", buf);
1853 datestamp(ident + name_len, 24);
1854 break;
1855 }
1856
1857 return ident;
1858 }
1859
1860 static void parse_new_blob(void)
1861 {
1862 static struct strbuf buf = STRBUF_INIT;
1863
1864 read_next_command();
1865 parse_mark();
1866 parse_data(&buf);
1867 store_object(OBJ_BLOB, &buf, &last_blob, NULL, next_mark);
1868 }
1869
1870 static void unload_one_branch(void)
1871 {
1872 while (cur_active_branches
1873 && cur_active_branches >= max_active_branches) {
1874 uintmax_t min_commit = ULONG_MAX;
1875 struct branch *e, *l = NULL, *p = NULL;
1876
1877 for (e = active_branches; e; e = e->active_next_branch) {
1878 if (e->last_commit < min_commit) {
1879 p = l;
1880 min_commit = e->last_commit;
1881 }
1882 l = e;
1883 }
1884
1885 if (p) {
1886 e = p->active_next_branch;
1887 p->active_next_branch = e->active_next_branch;
1888 } else {
1889 e = active_branches;
1890 active_branches = e->active_next_branch;
1891 }
1892 e->active = 0;
1893 e->active_next_branch = NULL;
1894 if (e->branch_tree.tree) {
1895 release_tree_content_recursive(e->branch_tree.tree);
1896 e->branch_tree.tree = NULL;
1897 }
1898 cur_active_branches--;
1899 }
1900 }
1901
1902 static void load_branch(struct branch *b)
1903 {
1904 load_tree(&b->branch_tree);
1905 if (!b->active) {
1906 b->active = 1;
1907 b->active_next_branch = active_branches;
1908 active_branches = b;
1909 cur_active_branches++;
1910 branch_load_count++;
1911 }
1912 }
1913
1914 static void file_change_m(struct branch *b)
1915 {
1916 const char *p = command_buf.buf + 2;
1917 static struct strbuf uq = STRBUF_INIT;
1918 const char *endp;
1919 struct object_entry *oe = oe;
1920 unsigned char sha1[20];
1921 uint16_t mode, inline_data = 0;
1922
1923 p = get_mode(p, &mode);
1924 if (!p)
1925 die("Corrupt mode: %s", command_buf.buf);
1926 switch (mode) {
1927 case 0644:
1928 case 0755:
1929 mode |= S_IFREG;
1930 case S_IFREG | 0644:
1931 case S_IFREG | 0755:
1932 case S_IFLNK:
1933 case S_IFGITLINK:
1934 /* ok */
1935 break;
1936 default:
1937 die("Corrupt mode: %s", command_buf.buf);
1938 }
1939
1940 if (*p == ':') {
1941 char *x;
1942 oe = find_mark(strtoumax(p + 1, &x, 10));
1943 hashcpy(sha1, oe->sha1);
1944 p = x;
1945 } else if (!prefixcmp(p, "inline")) {
1946 inline_data = 1;
1947 p += 6;
1948 } else {
1949 if (get_sha1_hex(p, sha1))
1950 die("Invalid SHA1: %s", command_buf.buf);
1951 oe = find_object(sha1);
1952 p += 40;
1953 }
1954 if (*p++ != ' ')
1955 die("Missing space after SHA1: %s", command_buf.buf);
1956
1957 strbuf_reset(&uq);
1958 if (!unquote_c_style(&uq, p, &endp)) {
1959 if (*endp)
1960 die("Garbage after path in: %s", command_buf.buf);
1961 p = uq.buf;
1962 }
1963
1964 if (S_ISGITLINK(mode)) {
1965 if (inline_data)
1966 die("Git links cannot be specified 'inline': %s",
1967 command_buf.buf);
1968 else if (oe) {
1969 if (oe->type != OBJ_COMMIT)
1970 die("Not a commit (actually a %s): %s",
1971 typename(oe->type), command_buf.buf);
1972 }
1973 /*
1974 * Accept the sha1 without checking; it expected to be in
1975 * another repository.
1976 */
1977 } else if (inline_data) {
1978 static struct strbuf buf = STRBUF_INIT;
1979
1980 if (p != uq.buf) {
1981 strbuf_addstr(&uq, p);
1982 p = uq.buf;
1983 }
1984 read_next_command();
1985 parse_data(&buf);
1986 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
1987 } else if (oe) {
1988 if (oe->type != OBJ_BLOB)
1989 die("Not a blob (actually a %s): %s",
1990 typename(oe->type), command_buf.buf);
1991 } else {
1992 enum object_type type = sha1_object_info(sha1, NULL);
1993 if (type < 0)
1994 die("Blob not found: %s", command_buf.buf);
1995 if (type != OBJ_BLOB)
1996 die("Not a blob (actually a %s): %s",
1997 typename(type), command_buf.buf);
1998 }
1999
2000 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2001 }
2002
2003 static void file_change_d(struct branch *b)
2004 {
2005 const char *p = command_buf.buf + 2;
2006 static struct strbuf uq = STRBUF_INIT;
2007 const char *endp;
2008
2009 strbuf_reset(&uq);
2010 if (!unquote_c_style(&uq, p, &endp)) {
2011 if (*endp)
2012 die("Garbage after path in: %s", command_buf.buf);
2013 p = uq.buf;
2014 }
2015 tree_content_remove(&b->branch_tree, p, NULL);
2016 }
2017
2018 static void file_change_cr(struct branch *b, int rename)
2019 {
2020 const char *s, *d;
2021 static struct strbuf s_uq = STRBUF_INIT;
2022 static struct strbuf d_uq = STRBUF_INIT;
2023 const char *endp;
2024 struct tree_entry leaf;
2025
2026 s = command_buf.buf + 2;
2027 strbuf_reset(&s_uq);
2028 if (!unquote_c_style(&s_uq, s, &endp)) {
2029 if (*endp != ' ')
2030 die("Missing space after source: %s", command_buf.buf);
2031 } else {
2032 endp = strchr(s, ' ');
2033 if (!endp)
2034 die("Missing space after source: %s", command_buf.buf);
2035 strbuf_add(&s_uq, s, endp - s);
2036 }
2037 s = s_uq.buf;
2038
2039 endp++;
2040 if (!*endp)
2041 die("Missing dest: %s", command_buf.buf);
2042
2043 d = endp;
2044 strbuf_reset(&d_uq);
2045 if (!unquote_c_style(&d_uq, d, &endp)) {
2046 if (*endp)
2047 die("Garbage after dest in: %s", command_buf.buf);
2048 d = d_uq.buf;
2049 }
2050
2051 memset(&leaf, 0, sizeof(leaf));
2052 if (rename)
2053 tree_content_remove(&b->branch_tree, s, &leaf);
2054 else
2055 tree_content_get(&b->branch_tree, s, &leaf);
2056 if (!leaf.versions[1].mode)
2057 die("Path %s not in branch", s);
2058 tree_content_set(&b->branch_tree, d,
2059 leaf.versions[1].sha1,
2060 leaf.versions[1].mode,
2061 leaf.tree);
2062 }
2063
2064 static void note_change_n(struct branch *b)
2065 {
2066 const char *p = command_buf.buf + 2;
2067 static struct strbuf uq = STRBUF_INIT;
2068 struct object_entry *oe = oe;
2069 struct branch *s;
2070 unsigned char sha1[20], commit_sha1[20];
2071 uint16_t inline_data = 0;
2072
2073 /* <dataref> or 'inline' */
2074 if (*p == ':') {
2075 char *x;
2076 oe = find_mark(strtoumax(p + 1, &x, 10));
2077 hashcpy(sha1, oe->sha1);
2078 p = x;
2079 } else if (!prefixcmp(p, "inline")) {
2080 inline_data = 1;
2081 p += 6;
2082 } else {
2083 if (get_sha1_hex(p, sha1))
2084 die("Invalid SHA1: %s", command_buf.buf);
2085 oe = find_object(sha1);
2086 p += 40;
2087 }
2088 if (*p++ != ' ')
2089 die("Missing space after SHA1: %s", command_buf.buf);
2090
2091 /* <committish> */
2092 s = lookup_branch(p);
2093 if (s) {
2094 hashcpy(commit_sha1, s->sha1);
2095 } else if (*p == ':') {
2096 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2097 struct object_entry *commit_oe = find_mark(commit_mark);
2098 if (commit_oe->type != OBJ_COMMIT)
2099 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2100 hashcpy(commit_sha1, commit_oe->sha1);
2101 } else if (!get_sha1(p, commit_sha1)) {
2102 unsigned long size;
2103 char *buf = read_object_with_reference(commit_sha1,
2104 commit_type, &size, commit_sha1);
2105 if (!buf || size < 46)
2106 die("Not a valid commit: %s", p);
2107 free(buf);
2108 } else
2109 die("Invalid ref name or SHA1 expression: %s", p);
2110
2111 if (inline_data) {
2112 static struct strbuf buf = STRBUF_INIT;
2113
2114 if (p != uq.buf) {
2115 strbuf_addstr(&uq, p);
2116 p = uq.buf;
2117 }
2118 read_next_command();
2119 parse_data(&buf);
2120 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
2121 } else if (oe) {
2122 if (oe->type != OBJ_BLOB)
2123 die("Not a blob (actually a %s): %s",
2124 typename(oe->type), command_buf.buf);
2125 } else {
2126 enum object_type type = sha1_object_info(sha1, NULL);
2127 if (type < 0)
2128 die("Blob not found: %s", command_buf.buf);
2129 if (type != OBJ_BLOB)
2130 die("Not a blob (actually a %s): %s",
2131 typename(type), command_buf.buf);
2132 }
2133
2134 tree_content_set(&b->branch_tree, sha1_to_hex(commit_sha1), sha1,
2135 S_IFREG | 0644, NULL);
2136 }
2137
2138 static void file_change_deleteall(struct branch *b)
2139 {
2140 release_tree_content_recursive(b->branch_tree.tree);
2141 hashclr(b->branch_tree.versions[0].sha1);
2142 hashclr(b->branch_tree.versions[1].sha1);
2143 load_tree(&b->branch_tree);
2144 }
2145
2146 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2147 {
2148 if (!buf || size < 46)
2149 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2150 if (memcmp("tree ", buf, 5)
2151 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2152 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2153 hashcpy(b->branch_tree.versions[0].sha1,
2154 b->branch_tree.versions[1].sha1);
2155 }
2156
2157 static void parse_from_existing(struct branch *b)
2158 {
2159 if (is_null_sha1(b->sha1)) {
2160 hashclr(b->branch_tree.versions[0].sha1);
2161 hashclr(b->branch_tree.versions[1].sha1);
2162 } else {
2163 unsigned long size;
2164 char *buf;
2165
2166 buf = read_object_with_reference(b->sha1,
2167 commit_type, &size, b->sha1);
2168 parse_from_commit(b, buf, size);
2169 free(buf);
2170 }
2171 }
2172
2173 static int parse_from(struct branch *b)
2174 {
2175 const char *from;
2176 struct branch *s;
2177
2178 if (prefixcmp(command_buf.buf, "from "))
2179 return 0;
2180
2181 if (b->branch_tree.tree) {
2182 release_tree_content_recursive(b->branch_tree.tree);
2183 b->branch_tree.tree = NULL;
2184 }
2185
2186 from = strchr(command_buf.buf, ' ') + 1;
2187 s = lookup_branch(from);
2188 if (b == s)
2189 die("Can't create a branch from itself: %s", b->name);
2190 else if (s) {
2191 unsigned char *t = s->branch_tree.versions[1].sha1;
2192 hashcpy(b->sha1, s->sha1);
2193 hashcpy(b->branch_tree.versions[0].sha1, t);
2194 hashcpy(b->branch_tree.versions[1].sha1, t);
2195 } else if (*from == ':') {
2196 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2197 struct object_entry *oe = find_mark(idnum);
2198 if (oe->type != OBJ_COMMIT)
2199 die("Mark :%" PRIuMAX " not a commit", idnum);
2200 hashcpy(b->sha1, oe->sha1);
2201 if (oe->pack_id != MAX_PACK_ID) {
2202 unsigned long size;
2203 char *buf = gfi_unpack_entry(oe, &size);
2204 parse_from_commit(b, buf, size);
2205 free(buf);
2206 } else
2207 parse_from_existing(b);
2208 } else if (!get_sha1(from, b->sha1))
2209 parse_from_existing(b);
2210 else
2211 die("Invalid ref name or SHA1 expression: %s", from);
2212
2213 read_next_command();
2214 return 1;
2215 }
2216
2217 static struct hash_list *parse_merge(unsigned int *count)
2218 {
2219 struct hash_list *list = NULL, *n, *e = e;
2220 const char *from;
2221 struct branch *s;
2222
2223 *count = 0;
2224 while (!prefixcmp(command_buf.buf, "merge ")) {
2225 from = strchr(command_buf.buf, ' ') + 1;
2226 n = xmalloc(sizeof(*n));
2227 s = lookup_branch(from);
2228 if (s)
2229 hashcpy(n->sha1, s->sha1);
2230 else if (*from == ':') {
2231 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2232 struct object_entry *oe = find_mark(idnum);
2233 if (oe->type != OBJ_COMMIT)
2234 die("Mark :%" PRIuMAX " not a commit", idnum);
2235 hashcpy(n->sha1, oe->sha1);
2236 } else if (!get_sha1(from, n->sha1)) {
2237 unsigned long size;
2238 char *buf = read_object_with_reference(n->sha1,
2239 commit_type, &size, n->sha1);
2240 if (!buf || size < 46)
2241 die("Not a valid commit: %s", from);
2242 free(buf);
2243 } else
2244 die("Invalid ref name or SHA1 expression: %s", from);
2245
2246 n->next = NULL;
2247 if (list)
2248 e->next = n;
2249 else
2250 list = n;
2251 e = n;
2252 (*count)++;
2253 read_next_command();
2254 }
2255 return list;
2256 }
2257
2258 static void parse_new_commit(void)
2259 {
2260 static struct strbuf msg = STRBUF_INIT;
2261 struct branch *b;
2262 char *sp;
2263 char *author = NULL;
2264 char *committer = NULL;
2265 struct hash_list *merge_list = NULL;
2266 unsigned int merge_count;
2267
2268 /* Obtain the branch name from the rest of our command */
2269 sp = strchr(command_buf.buf, ' ') + 1;
2270 b = lookup_branch(sp);
2271 if (!b)
2272 b = new_branch(sp);
2273
2274 read_next_command();
2275 parse_mark();
2276 if (!prefixcmp(command_buf.buf, "author ")) {
2277 author = parse_ident(command_buf.buf + 7);
2278 read_next_command();
2279 }
2280 if (!prefixcmp(command_buf.buf, "committer ")) {
2281 committer = parse_ident(command_buf.buf + 10);
2282 read_next_command();
2283 }
2284 if (!committer)
2285 die("Expected committer but didn't get one");
2286 parse_data(&msg);
2287 read_next_command();
2288 parse_from(b);
2289 merge_list = parse_merge(&merge_count);
2290
2291 /* ensure the branch is active/loaded */
2292 if (!b->branch_tree.tree || !max_active_branches) {
2293 unload_one_branch();
2294 load_branch(b);
2295 }
2296
2297 /* file_change* */
2298 while (command_buf.len > 0) {
2299 if (!prefixcmp(command_buf.buf, "M "))
2300 file_change_m(b);
2301 else if (!prefixcmp(command_buf.buf, "D "))
2302 file_change_d(b);
2303 else if (!prefixcmp(command_buf.buf, "R "))
2304 file_change_cr(b, 1);
2305 else if (!prefixcmp(command_buf.buf, "C "))
2306 file_change_cr(b, 0);
2307 else if (!prefixcmp(command_buf.buf, "N "))
2308 note_change_n(b);
2309 else if (!strcmp("deleteall", command_buf.buf))
2310 file_change_deleteall(b);
2311 else {
2312 unread_command_buf = 1;
2313 break;
2314 }
2315 if (read_next_command() == EOF)
2316 break;
2317 }
2318
2319 /* build the tree and the commit */
2320 store_tree(&b->branch_tree);
2321 hashcpy(b->branch_tree.versions[0].sha1,
2322 b->branch_tree.versions[1].sha1);
2323
2324 strbuf_reset(&new_data);
2325 strbuf_addf(&new_data, "tree %s\n",
2326 sha1_to_hex(b->branch_tree.versions[1].sha1));
2327 if (!is_null_sha1(b->sha1))
2328 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2329 while (merge_list) {
2330 struct hash_list *next = merge_list->next;
2331 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2332 free(merge_list);
2333 merge_list = next;
2334 }
2335 strbuf_addf(&new_data,
2336 "author %s\n"
2337 "committer %s\n"
2338 "\n",
2339 author ? author : committer, committer);
2340 strbuf_addbuf(&new_data, &msg);
2341 free(author);
2342 free(committer);
2343
2344 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2345 b->pack_id = pack_id;
2346 b->last_commit = object_count_by_type[OBJ_COMMIT];
2347 }
2348
2349 static void parse_new_tag(void)
2350 {
2351 static struct strbuf msg = STRBUF_INIT;
2352 char *sp;
2353 const char *from;
2354 char *tagger;
2355 struct branch *s;
2356 struct tag *t;
2357 uintmax_t from_mark = 0;
2358 unsigned char sha1[20];
2359 enum object_type type;
2360
2361 /* Obtain the new tag name from the rest of our command */
2362 sp = strchr(command_buf.buf, ' ') + 1;
2363 t = pool_alloc(sizeof(struct tag));
2364 t->next_tag = NULL;
2365 t->name = pool_strdup(sp);
2366 if (last_tag)
2367 last_tag->next_tag = t;
2368 else
2369 first_tag = t;
2370 last_tag = t;
2371 read_next_command();
2372
2373 /* from ... */
2374 if (prefixcmp(command_buf.buf, "from "))
2375 die("Expected from command, got %s", command_buf.buf);
2376 from = strchr(command_buf.buf, ' ') + 1;
2377 s = lookup_branch(from);
2378 if (s) {
2379 hashcpy(sha1, s->sha1);
2380 type = OBJ_COMMIT;
2381 } else if (*from == ':') {
2382 struct object_entry *oe;
2383 from_mark = strtoumax(from + 1, NULL, 10);
2384 oe = find_mark(from_mark);
2385 type = oe->type;
2386 hashcpy(sha1, oe->sha1);
2387 } else if (!get_sha1(from, sha1)) {
2388 unsigned long size;
2389 char *buf;
2390
2391 buf = read_sha1_file(sha1, &type, &size);
2392 if (!buf || size < 46)
2393 die("Not a valid commit: %s", from);
2394 free(buf);
2395 } else
2396 die("Invalid ref name or SHA1 expression: %s", from);
2397 read_next_command();
2398
2399 /* tagger ... */
2400 if (!prefixcmp(command_buf.buf, "tagger ")) {
2401 tagger = parse_ident(command_buf.buf + 7);
2402 read_next_command();
2403 } else
2404 tagger = NULL;
2405
2406 /* tag payload/message */
2407 parse_data(&msg);
2408
2409 /* build the tag object */
2410 strbuf_reset(&new_data);
2411
2412 strbuf_addf(&new_data,
2413 "object %s\n"
2414 "type %s\n"
2415 "tag %s\n",
2416 sha1_to_hex(sha1), typename(type), t->name);
2417 if (tagger)
2418 strbuf_addf(&new_data,
2419 "tagger %s\n", tagger);
2420 strbuf_addch(&new_data, '\n');
2421 strbuf_addbuf(&new_data, &msg);
2422 free(tagger);
2423
2424 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2425 t->pack_id = MAX_PACK_ID;
2426 else
2427 t->pack_id = pack_id;
2428 }
2429
2430 static void parse_reset_branch(void)
2431 {
2432 struct branch *b;
2433 char *sp;
2434
2435 /* Obtain the branch name from the rest of our command */
2436 sp = strchr(command_buf.buf, ' ') + 1;
2437 b = lookup_branch(sp);
2438 if (b) {
2439 hashclr(b->sha1);
2440 hashclr(b->branch_tree.versions[0].sha1);
2441 hashclr(b->branch_tree.versions[1].sha1);
2442 if (b->branch_tree.tree) {
2443 release_tree_content_recursive(b->branch_tree.tree);
2444 b->branch_tree.tree = NULL;
2445 }
2446 }
2447 else
2448 b = new_branch(sp);
2449 read_next_command();
2450 parse_from(b);
2451 if (command_buf.len > 0)
2452 unread_command_buf = 1;
2453 }
2454
2455 static void parse_checkpoint(void)
2456 {
2457 if (object_count) {
2458 cycle_packfile();
2459 dump_branches();
2460 dump_tags();
2461 dump_marks();
2462 }
2463 skip_optional_lf();
2464 }
2465
2466 static void parse_progress(void)
2467 {
2468 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2469 fputc('\n', stdout);
2470 fflush(stdout);
2471 skip_optional_lf();
2472 }
2473
2474 static char* make_fast_import_path(const char *path)
2475 {
2476 struct strbuf abs_path = STRBUF_INIT;
2477
2478 if (!relative_marks_paths || is_absolute_path(path))
2479 return xstrdup(path);
2480 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
2481 return strbuf_detach(&abs_path, NULL);
2482 }
2483
2484 static void option_import_marks(const char *marks, int from_stream)
2485 {
2486 if (import_marks_file) {
2487 if (from_stream)
2488 die("Only one import-marks command allowed per stream");
2489
2490 /* read previous mark file */
2491 if(!import_marks_file_from_stream)
2492 read_marks();
2493 }
2494
2495 import_marks_file = make_fast_import_path(marks);
2496 import_marks_file_from_stream = from_stream;
2497 }
2498
2499 static void option_date_format(const char *fmt)
2500 {
2501 if (!strcmp(fmt, "raw"))
2502 whenspec = WHENSPEC_RAW;
2503 else if (!strcmp(fmt, "rfc2822"))
2504 whenspec = WHENSPEC_RFC2822;
2505 else if (!strcmp(fmt, "now"))
2506 whenspec = WHENSPEC_NOW;
2507 else
2508 die("unknown --date-format argument %s", fmt);
2509 }
2510
2511 static void option_max_pack_size(const char *packsize)
2512 {
2513 max_packsize = strtoumax(packsize, NULL, 0) * 1024 * 1024;
2514 }
2515
2516 static void option_depth(const char *depth)
2517 {
2518 max_depth = strtoul(depth, NULL, 0);
2519 if (max_depth > MAX_DEPTH)
2520 die("--depth cannot exceed %u", MAX_DEPTH);
2521 }
2522
2523 static void option_active_branches(const char *branches)
2524 {
2525 max_active_branches = strtoul(branches, NULL, 0);
2526 }
2527
2528 static void option_export_marks(const char *marks)
2529 {
2530 export_marks_file = make_fast_import_path(marks);
2531 }
2532
2533 static void option_export_pack_edges(const char *edges)
2534 {
2535 if (pack_edges)
2536 fclose(pack_edges);
2537 pack_edges = fopen(edges, "a");
2538 if (!pack_edges)
2539 die_errno("Cannot open '%s'", edges);
2540 }
2541
2542 static int parse_one_option(const char *option)
2543 {
2544 if (!prefixcmp(option, "max-pack-size=")) {
2545 option_max_pack_size(option + 14);
2546 } else if (!prefixcmp(option, "depth=")) {
2547 option_depth(option + 6);
2548 } else if (!prefixcmp(option, "active-branches=")) {
2549 option_active_branches(option + 16);
2550 } else if (!prefixcmp(option, "export-pack-edges=")) {
2551 option_export_pack_edges(option + 18);
2552 } else if (!prefixcmp(option, "quiet")) {
2553 show_stats = 0;
2554 } else if (!prefixcmp(option, "stats")) {
2555 show_stats = 1;
2556 } else {
2557 return 0;
2558 }
2559
2560 return 1;
2561 }
2562
2563 static int parse_one_feature(const char *feature, int from_stream)
2564 {
2565 if (!prefixcmp(feature, "date-format=")) {
2566 option_date_format(feature + 12);
2567 } else if (!prefixcmp(feature, "import-marks=")) {
2568 option_import_marks(feature + 13, from_stream);
2569 } else if (!prefixcmp(feature, "export-marks=")) {
2570 option_export_marks(feature + 13);
2571 } else if (!prefixcmp(feature, "relative-marks")) {
2572 relative_marks_paths = 1;
2573 } else if (!prefixcmp(feature, "no-relative-marks")) {
2574 relative_marks_paths = 0;
2575 } else if (!prefixcmp(feature, "force")) {
2576 force_update = 1;
2577 } else {
2578 return 0;
2579 }
2580
2581 return 1;
2582 }
2583
2584 static void parse_feature(void)
2585 {
2586 char *feature = command_buf.buf + 8;
2587
2588 if (seen_data_command)
2589 die("Got feature command '%s' after data command", feature);
2590
2591 if (parse_one_feature(feature, 1))
2592 return;
2593
2594 die("This version of fast-import does not support feature %s.", feature);
2595 }
2596
2597 static void parse_option(void)
2598 {
2599 char *option = command_buf.buf + 11;
2600
2601 if (seen_data_command)
2602 die("Got option command '%s' after data command", option);
2603
2604 if (parse_one_option(option))
2605 return;
2606
2607 die("This version of fast-import does not support option: %s", option);
2608 }
2609
2610 static int git_pack_config(const char *k, const char *v, void *cb)
2611 {
2612 if (!strcmp(k, "pack.depth")) {
2613 max_depth = git_config_int(k, v);
2614 if (max_depth > MAX_DEPTH)
2615 max_depth = MAX_DEPTH;
2616 return 0;
2617 }
2618 if (!strcmp(k, "pack.compression")) {
2619 int level = git_config_int(k, v);
2620 if (level == -1)
2621 level = Z_DEFAULT_COMPRESSION;
2622 else if (level < 0 || level > Z_BEST_COMPRESSION)
2623 die("bad pack compression level %d", level);
2624 pack_compression_level = level;
2625 pack_compression_seen = 1;
2626 return 0;
2627 }
2628 return git_default_config(k, v, cb);
2629 }
2630
2631 static const char fast_import_usage[] =
2632 "git fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2633
2634 static void parse_argv(void)
2635 {
2636 unsigned int i;
2637
2638 for (i = 1; i < global_argc; i++) {
2639 const char *a = global_argv[i];
2640
2641 if (*a != '-' || !strcmp(a, "--"))
2642 break;
2643
2644 if (parse_one_option(a + 2))
2645 continue;
2646
2647 if (parse_one_feature(a + 2, 0))
2648 continue;
2649
2650 die("unknown option %s", a);
2651 }
2652 if (i != global_argc)
2653 usage(fast_import_usage);
2654
2655 seen_data_command = 1;
2656 if (import_marks_file)
2657 read_marks();
2658 }
2659
2660 int main(int argc, const char **argv)
2661 {
2662 unsigned int i;
2663
2664 git_extract_argv0_path(argv[0]);
2665
2666 if (argc == 2 && !strcmp(argv[1], "-h"))
2667 usage(fast_import_usage);
2668
2669 setup_git_directory();
2670 git_config(git_pack_config, NULL);
2671 if (!pack_compression_seen && core_compression_seen)
2672 pack_compression_level = core_compression_level;
2673
2674 alloc_objects(object_entry_alloc);
2675 strbuf_init(&command_buf, 0);
2676 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2677 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2678 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2679 marks = pool_calloc(1, sizeof(struct mark_set));
2680
2681 global_argc = argc;
2682 global_argv = argv;
2683
2684 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2685 for (i = 0; i < (cmd_save - 1); i++)
2686 rc_free[i].next = &rc_free[i + 1];
2687 rc_free[cmd_save - 1].next = NULL;
2688
2689 prepare_packed_git();
2690 start_packfile();
2691 set_die_routine(die_nicely);
2692 while (read_next_command() != EOF) {
2693 if (!strcmp("blob", command_buf.buf))
2694 parse_new_blob();
2695 else if (!prefixcmp(command_buf.buf, "commit "))
2696 parse_new_commit();
2697 else if (!prefixcmp(command_buf.buf, "tag "))
2698 parse_new_tag();
2699 else if (!prefixcmp(command_buf.buf, "reset "))
2700 parse_reset_branch();
2701 else if (!strcmp("checkpoint", command_buf.buf))
2702 parse_checkpoint();
2703 else if (!prefixcmp(command_buf.buf, "progress "))
2704 parse_progress();
2705 else if (!prefixcmp(command_buf.buf, "feature "))
2706 parse_feature();
2707 else if (!prefixcmp(command_buf.buf, "option git "))
2708 parse_option();
2709 else if (!prefixcmp(command_buf.buf, "option "))
2710 /* ignore non-git options*/;
2711 else
2712 die("Unsupported command: %s", command_buf.buf);
2713 }
2714
2715 /* argv hasn't been parsed yet, do so */
2716 if (!seen_data_command)
2717 parse_argv();
2718
2719 end_packfile();
2720
2721 dump_branches();
2722 dump_tags();
2723 unkeep_all_packs();
2724 dump_marks();
2725
2726 if (pack_edges)
2727 fclose(pack_edges);
2728
2729 if (show_stats) {
2730 uintmax_t total_count = 0, duplicate_count = 0;
2731 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2732 total_count += object_count_by_type[i];
2733 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2734 duplicate_count += duplicate_count_by_type[i];
2735
2736 fprintf(stderr, "%s statistics:\n", argv[0]);
2737 fprintf(stderr, "---------------------------------------------------------------------\n");
2738 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2739 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2740 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]);
2741 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]);
2742 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]);
2743 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]);
2744 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2745 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2746 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2747 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2748 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2749 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2750 fprintf(stderr, "---------------------------------------------------------------------\n");
2751 pack_report();
2752 fprintf(stderr, "---------------------------------------------------------------------\n");
2753 fprintf(stderr, "\n");
2754 }
2755
2756 return failure ? 1 : 0;
2757 }