?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Documentation/config.txt: default gc.aggressiveWindow is 250, not 10
[git/jnareb-git.git] / fast-import.c
blob74f08bd554ae4cab1dc9ee38272376e5cbec1200
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 pack_idx_entry idx;
168 struct object_entry *next;
169 uint32_t type : TYPE_BITS,
170 pack_id : PACK_ID_BITS,
171 depth : DEPTH_BITS;
172 };
173
174 struct object_entry_pool
175 {
176 struct object_entry_pool *next_pool;
177 struct object_entry *next_free;
178 struct object_entry *end;
179 struct object_entry entries[FLEX_ARRAY]; /* more */
180 };
181
182 struct mark_set
183 {
184 union {
185 struct object_entry *marked[1024];
186 struct mark_set *sets[1024];
187 } data;
188 unsigned int shift;
189 };
190
191 struct last_object
192 {
193 struct strbuf data;
194 off_t offset;
195 unsigned int depth;
196 unsigned no_swap : 1;
197 };
198
199 struct mem_pool
200 {
201 struct mem_pool *next_pool;
202 char *next_free;
203 char *end;
204 uintmax_t space[FLEX_ARRAY]; /* more */
205 };
206
207 struct atom_str
208 {
209 struct atom_str *next_atom;
210 unsigned short str_len;
211 char str_dat[FLEX_ARRAY]; /* more */
212 };
213
214 struct tree_content;
215 struct tree_entry
216 {
217 struct tree_content *tree;
218 struct atom_str *name;
219 struct tree_entry_ms
220 {
221 uint16_t mode;
222 unsigned char sha1[20];
223 } versions[2];
224 };
225
226 struct tree_content
227 {
228 unsigned int entry_capacity; /* must match avail_tree_content */
229 unsigned int entry_count;
230 unsigned int delta_depth;
231 struct tree_entry *entries[FLEX_ARRAY]; /* more */
232 };
233
234 struct avail_tree_content
235 {
236 unsigned int entry_capacity; /* must match tree_content */
237 struct avail_tree_content *next_avail;
238 };
239
240 struct branch
241 {
242 struct branch *table_next_branch;
243 struct branch *active_next_branch;
244 const char *name;
245 struct tree_entry branch_tree;
246 uintmax_t last_commit;
247 uintmax_t num_notes;
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;
283 static uintmax_t big_file_threshold = 512 * 1024 * 1024;
284 static int force_update;
285 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
286 static int pack_compression_seen;
287
288 /* Stats and misc. counters */
289 static uintmax_t alloc_count;
290 static uintmax_t marks_set_count;
291 static uintmax_t object_count_by_type[1 << TYPE_BITS];
292 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
293 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
294 static unsigned long object_count;
295 static unsigned long branch_count;
296 static unsigned long branch_load_count;
297 static int failure;
298 static FILE *pack_edges;
299 static unsigned int show_stats = 1;
300 static int global_argc;
301 static const char **global_argv;
302
303 /* Memory pools */
304 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
305 static size_t total_allocd;
306 static struct mem_pool *mem_pool;
307
308 /* Atom management */
309 static unsigned int atom_table_sz = 4451;
310 static unsigned int atom_cnt;
311 static struct atom_str **atom_table;
312
313 /* The .pack file being generated */
314 static unsigned int pack_id;
315 static struct sha1file *pack_file;
316 static struct packed_git *pack_data;
317 static struct packed_git **all_packs;
318 static off_t pack_size;
319
320 /* Table of objects we've written. */
321 static unsigned int object_entry_alloc = 5000;
322 static struct object_entry_pool *blocks;
323 static struct object_entry *object_table[1 << 16];
324 static struct mark_set *marks;
325 static const char *export_marks_file;
326 static const char *import_marks_file;
327 static int import_marks_file_from_stream;
328 static int relative_marks_paths;
329
330 /* Our last blob */
331 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
332
333 /* Tree management */
334 static unsigned int tree_entry_alloc = 1000;
335 static void *avail_tree_entry;
336 static unsigned int avail_tree_table_sz = 100;
337 static struct avail_tree_content **avail_tree_table;
338 static struct strbuf old_tree = STRBUF_INIT;
339 static struct strbuf new_tree = STRBUF_INIT;
340
341 /* Branch data */
342 static unsigned long max_active_branches = 5;
343 static unsigned long cur_active_branches;
344 static unsigned long branch_table_sz = 1039;
345 static struct branch **branch_table;
346 static struct branch *active_branches;
347
348 /* Tag data */
349 static struct tag *first_tag;
350 static struct tag *last_tag;
351
352 /* Input stream parsing */
353 static whenspec_type whenspec = WHENSPEC_RAW;
354 static struct strbuf command_buf = STRBUF_INIT;
355 static int unread_command_buf;
356 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
357 static struct recent_command *cmd_tail = &cmd_hist;
358 static struct recent_command *rc_free;
359 static unsigned int cmd_save = 100;
360 static uintmax_t next_mark;
361 static struct strbuf new_data = STRBUF_INIT;
362 static int seen_data_command;
363
364 static void parse_argv(void);
365
366 static void write_branch_report(FILE *rpt, struct branch *b)
367 {
368 fprintf(rpt, "%s:\n", b->name);
369
370 fprintf(rpt, " status :");
371 if (b->active)
372 fputs(" active", rpt);
373 if (b->branch_tree.tree)
374 fputs(" loaded", rpt);
375 if (is_null_sha1(b->branch_tree.versions[1].sha1))
376 fputs(" dirty", rpt);
377 fputc('\n', rpt);
378
379 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
380 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
381 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
382 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
383
384 fputs(" last pack : ", rpt);
385 if (b->pack_id < MAX_PACK_ID)
386 fprintf(rpt, "%u", b->pack_id);
387 fputc('\n', rpt);
388
389 fputc('\n', rpt);
390 }
391
392 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
393
394 static void write_crash_report(const char *err)
395 {
396 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
397 FILE *rpt = fopen(loc, "w");
398 struct branch *b;
399 unsigned long lu;
400 struct recent_command *rc;
401
402 if (!rpt) {
403 error("can't write crash report %s: %s", loc, strerror(errno));
404 return;
405 }
406
407 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
408
409 fprintf(rpt, "fast-import crash report:\n");
410 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
411 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
412 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
413 fputc('\n', rpt);
414
415 fputs("fatal: ", rpt);
416 fputs(err, rpt);
417 fputc('\n', rpt);
418
419 fputc('\n', rpt);
420 fputs("Most Recent Commands Before Crash\n", rpt);
421 fputs("---------------------------------\n", rpt);
422 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
423 if (rc->next == &cmd_hist)
424 fputs("* ", rpt);
425 else
426 fputs(" ", rpt);
427 fputs(rc->buf, rpt);
428 fputc('\n', rpt);
429 }
430
431 fputc('\n', rpt);
432 fputs("Active Branch LRU\n", rpt);
433 fputs("-----------------\n", rpt);
434 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
435 cur_active_branches,
436 max_active_branches);
437 fputc('\n', rpt);
438 fputs(" pos clock name\n", rpt);
439 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
440 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
441 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
442 ++lu, b->last_commit, b->name);
443
444 fputc('\n', rpt);
445 fputs("Inactive Branches\n", rpt);
446 fputs("-----------------\n", rpt);
447 for (lu = 0; lu < branch_table_sz; lu++) {
448 for (b = branch_table[lu]; b; b = b->table_next_branch)
449 write_branch_report(rpt, b);
450 }
451
452 if (first_tag) {
453 struct tag *tg;
454 fputc('\n', rpt);
455 fputs("Annotated Tags\n", rpt);
456 fputs("--------------\n", rpt);
457 for (tg = first_tag; tg; tg = tg->next_tag) {
458 fputs(sha1_to_hex(tg->sha1), rpt);
459 fputc(' ', rpt);
460 fputs(tg->name, rpt);
461 fputc('\n', rpt);
462 }
463 }
464
465 fputc('\n', rpt);
466 fputs("Marks\n", rpt);
467 fputs("-----\n", rpt);
468 if (export_marks_file)
469 fprintf(rpt, " exported to %s\n", export_marks_file);
470 else
471 dump_marks_helper(rpt, 0, marks);
472
473 fputc('\n', rpt);
474 fputs("-------------------\n", rpt);
475 fputs("END OF CRASH REPORT\n", rpt);
476 fclose(rpt);
477 }
478
479 static void end_packfile(void);
480 static void unkeep_all_packs(void);
481 static void dump_marks(void);
482
483 static NORETURN void die_nicely(const char *err, va_list params)
484 {
485 static int zombie;
486 char message[2 * PATH_MAX];
487
488 vsnprintf(message, sizeof(message), err, params);
489 fputs("fatal: ", stderr);
490 fputs(message, stderr);
491 fputc('\n', stderr);
492
493 if (!zombie) {
494 zombie = 1;
495 write_crash_report(message);
496 end_packfile();
497 unkeep_all_packs();
498 dump_marks();
499 }
500 exit(128);
501 }
502
503 static void alloc_objects(unsigned int cnt)
504 {
505 struct object_entry_pool *b;
506
507 b = xmalloc(sizeof(struct object_entry_pool)
508 + cnt * sizeof(struct object_entry));
509 b->next_pool = blocks;
510 b->next_free = b->entries;
511 b->end = b->entries + cnt;
512 blocks = b;
513 alloc_count += cnt;
514 }
515
516 static struct object_entry *new_object(unsigned char *sha1)
517 {
518 struct object_entry *e;
519
520 if (blocks->next_free == blocks->end)
521 alloc_objects(object_entry_alloc);
522
523 e = blocks->next_free++;
524 hashcpy(e->idx.sha1, sha1);
525 return e;
526 }
527
528 static struct object_entry *find_object(unsigned char *sha1)
529 {
530 unsigned int h = sha1[0] << 8 | sha1[1];
531 struct object_entry *e;
532 for (e = object_table[h]; e; e = e->next)
533 if (!hashcmp(sha1, e->idx.sha1))
534 return e;
535 return NULL;
536 }
537
538 static struct object_entry *insert_object(unsigned char *sha1)
539 {
540 unsigned int h = sha1[0] << 8 | sha1[1];
541 struct object_entry *e = object_table[h];
542 struct object_entry *p = NULL;
543
544 while (e) {
545 if (!hashcmp(sha1, e->idx.sha1))
546 return e;
547 p = e;
548 e = e->next;
549 }
550
551 e = new_object(sha1);
552 e->next = NULL;
553 e->idx.offset = 0;
554 if (p)
555 p->next = e;
556 else
557 object_table[h] = e;
558 return e;
559 }
560
561 static unsigned int hc_str(const char *s, size_t len)
562 {
563 unsigned int r = 0;
564 while (len-- > 0)
565 r = r * 31 + *s++;
566 return r;
567 }
568
569 static void *pool_alloc(size_t len)
570 {
571 struct mem_pool *p;
572 void *r;
573
574 /* round up to a 'uintmax_t' alignment */
575 if (len & (sizeof(uintmax_t) - 1))
576 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
577
578 for (p = mem_pool; p; p = p->next_pool)
579 if ((p->end - p->next_free >= len))
580 break;
581
582 if (!p) {
583 if (len >= (mem_pool_alloc/2)) {
584 total_allocd += len;
585 return xmalloc(len);
586 }
587 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
588 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
589 p->next_pool = mem_pool;
590 p->next_free = (char *) p->space;
591 p->end = p->next_free + mem_pool_alloc;
592 mem_pool = p;
593 }
594
595 r = p->next_free;
596 p->next_free += len;
597 return r;
598 }
599
600 static void *pool_calloc(size_t count, size_t size)
601 {
602 size_t len = count * size;
603 void *r = pool_alloc(len);
604 memset(r, 0, len);
605 return r;
606 }
607
608 static char *pool_strdup(const char *s)
609 {
610 char *r = pool_alloc(strlen(s) + 1);
611 strcpy(r, s);
612 return r;
613 }
614
615 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
616 {
617 struct mark_set *s = marks;
618 while ((idnum >> s->shift) >= 1024) {
619 s = pool_calloc(1, sizeof(struct mark_set));
620 s->shift = marks->shift + 10;
621 s->data.sets[0] = marks;
622 marks = s;
623 }
624 while (s->shift) {
625 uintmax_t i = idnum >> s->shift;
626 idnum -= i << s->shift;
627 if (!s->data.sets[i]) {
628 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
629 s->data.sets[i]->shift = s->shift - 10;
630 }
631 s = s->data.sets[i];
632 }
633 if (!s->data.marked[idnum])
634 marks_set_count++;
635 s->data.marked[idnum] = oe;
636 }
637
638 static struct object_entry *find_mark(uintmax_t idnum)
639 {
640 uintmax_t orig_idnum = idnum;
641 struct mark_set *s = marks;
642 struct object_entry *oe = NULL;
643 if ((idnum >> s->shift) < 1024) {
644 while (s && s->shift) {
645 uintmax_t i = idnum >> s->shift;
646 idnum -= i << s->shift;
647 s = s->data.sets[i];
648 }
649 if (s)
650 oe = s->data.marked[idnum];
651 }
652 if (!oe)
653 die("mark :%" PRIuMAX " not declared", orig_idnum);
654 return oe;
655 }
656
657 static struct atom_str *to_atom(const char *s, unsigned short len)
658 {
659 unsigned int hc = hc_str(s, len) % atom_table_sz;
660 struct atom_str *c;
661
662 for (c = atom_table[hc]; c; c = c->next_atom)
663 if (c->str_len == len && !strncmp(s, c->str_dat, len))
664 return c;
665
666 c = pool_alloc(sizeof(struct atom_str) + len + 1);
667 c->str_len = len;
668 strncpy(c->str_dat, s, len);
669 c->str_dat[len] = 0;
670 c->next_atom = atom_table[hc];
671 atom_table[hc] = c;
672 atom_cnt++;
673 return c;
674 }
675
676 static struct branch *lookup_branch(const char *name)
677 {
678 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
679 struct branch *b;
680
681 for (b = branch_table[hc]; b; b = b->table_next_branch)
682 if (!strcmp(name, b->name))
683 return b;
684 return NULL;
685 }
686
687 static struct branch *new_branch(const char *name)
688 {
689 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
690 struct branch *b = lookup_branch(name);
691
692 if (b)
693 die("Invalid attempt to create duplicate branch: %s", name);
694 switch (check_ref_format(name)) {
695 case 0: break; /* its valid */
696 case CHECK_REF_FORMAT_ONELEVEL:
697 break; /* valid, but too few '/', allow anyway */
698 default:
699 die("Branch name doesn't conform to GIT standards: %s", name);
700 }
701
702 b = pool_calloc(1, sizeof(struct branch));
703 b->name = pool_strdup(name);
704 b->table_next_branch = branch_table[hc];
705 b->branch_tree.versions[0].mode = S_IFDIR;
706 b->branch_tree.versions[1].mode = S_IFDIR;
707 b->num_notes = 0;
708 b->active = 0;
709 b->pack_id = MAX_PACK_ID;
710 branch_table[hc] = b;
711 branch_count++;
712 return b;
713 }
714
715 static unsigned int hc_entries(unsigned int cnt)
716 {
717 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
718 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
719 }
720
721 static struct tree_content *new_tree_content(unsigned int cnt)
722 {
723 struct avail_tree_content *f, *l = NULL;
724 struct tree_content *t;
725 unsigned int hc = hc_entries(cnt);
726
727 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
728 if (f->entry_capacity >= cnt)
729 break;
730
731 if (f) {
732 if (l)
733 l->next_avail = f->next_avail;
734 else
735 avail_tree_table[hc] = f->next_avail;
736 } else {
737 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
738 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
739 f->entry_capacity = cnt;
740 }
741
742 t = (struct tree_content*)f;
743 t->entry_count = 0;
744 t->delta_depth = 0;
745 return t;
746 }
747
748 static void release_tree_entry(struct tree_entry *e);
749 static void release_tree_content(struct tree_content *t)
750 {
751 struct avail_tree_content *f = (struct avail_tree_content*)t;
752 unsigned int hc = hc_entries(f->entry_capacity);
753 f->next_avail = avail_tree_table[hc];
754 avail_tree_table[hc] = f;
755 }
756
757 static void release_tree_content_recursive(struct tree_content *t)
758 {
759 unsigned int i;
760 for (i = 0; i < t->entry_count; i++)
761 release_tree_entry(t->entries[i]);
762 release_tree_content(t);
763 }
764
765 static struct tree_content *grow_tree_content(
766 struct tree_content *t,
767 int amt)
768 {
769 struct tree_content *r = new_tree_content(t->entry_count + amt);
770 r->entry_count = t->entry_count;
771 r->delta_depth = t->delta_depth;
772 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
773 release_tree_content(t);
774 return r;
775 }
776
777 static struct tree_entry *new_tree_entry(void)
778 {
779 struct tree_entry *e;
780
781 if (!avail_tree_entry) {
782 unsigned int n = tree_entry_alloc;
783 total_allocd += n * sizeof(struct tree_entry);
784 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
785 while (n-- > 1) {
786 *((void**)e) = e + 1;
787 e++;
788 }
789 *((void**)e) = NULL;
790 }
791
792 e = avail_tree_entry;
793 avail_tree_entry = *((void**)e);
794 return e;
795 }
796
797 static void release_tree_entry(struct tree_entry *e)
798 {
799 if (e->tree)
800 release_tree_content_recursive(e->tree);
801 *((void**)e) = avail_tree_entry;
802 avail_tree_entry = e;
803 }
804
805 static struct tree_content *dup_tree_content(struct tree_content *s)
806 {
807 struct tree_content *d;
808 struct tree_entry *a, *b;
809 unsigned int i;
810
811 if (!s)
812 return NULL;
813 d = new_tree_content(s->entry_count);
814 for (i = 0; i < s->entry_count; i++) {
815 a = s->entries[i];
816 b = new_tree_entry();
817 memcpy(b, a, sizeof(*a));
818 if (a->tree && is_null_sha1(b->versions[1].sha1))
819 b->tree = dup_tree_content(a->tree);
820 else
821 b->tree = NULL;
822 d->entries[i] = b;
823 }
824 d->entry_count = s->entry_count;
825 d->delta_depth = s->delta_depth;
826
827 return d;
828 }
829
830 static void start_packfile(void)
831 {
832 static char tmpfile[PATH_MAX];
833 struct packed_git *p;
834 struct pack_header hdr;
835 int pack_fd;
836
837 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
838 "pack/tmp_pack_XXXXXX");
839 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
840 strcpy(p->pack_name, tmpfile);
841 p->pack_fd = pack_fd;
842 pack_file = sha1fd(pack_fd, p->pack_name);
843
844 hdr.hdr_signature = htonl(PACK_SIGNATURE);
845 hdr.hdr_version = htonl(2);
846 hdr.hdr_entries = 0;
847 sha1write(pack_file, &hdr, sizeof(hdr));
848
849 pack_data = p;
850 pack_size = sizeof(hdr);
851 object_count = 0;
852
853 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
854 all_packs[pack_id] = p;
855 }
856
857 static const char *create_index(void)
858 {
859 const char *tmpfile;
860 struct pack_idx_entry **idx, **c, **last;
861 struct object_entry *e;
862 struct object_entry_pool *o;
863
864 /* Build the table of object IDs. */
865 idx = xmalloc(object_count * sizeof(*idx));
866 c = idx;
867 for (o = blocks; o; o = o->next_pool)
868 for (e = o->next_free; e-- != o->entries;)
869 if (pack_id == e->pack_id)
870 *c++ = &e->idx;
871 last = idx + object_count;
872 if (c != last)
873 die("internal consistency error creating the index");
874
875 tmpfile = write_idx_file(NULL, idx, object_count, pack_data->sha1);
876 free(idx);
877 return tmpfile;
878 }
879
880 static char *keep_pack(const char *curr_index_name)
881 {
882 static char name[PATH_MAX];
883 static const char *keep_msg = "fast-import";
884 int keep_fd;
885
886 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
887 if (keep_fd < 0)
888 die_errno("cannot create keep file");
889 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
890 if (close(keep_fd))
891 die_errno("failed to write keep file");
892
893 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
894 get_object_directory(), sha1_to_hex(pack_data->sha1));
895 if (move_temp_to_file(pack_data->pack_name, name))
896 die("cannot store pack file");
897
898 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
899 get_object_directory(), sha1_to_hex(pack_data->sha1));
900 if (move_temp_to_file(curr_index_name, name))
901 die("cannot store index file");
902 free((void *)curr_index_name);
903 return name;
904 }
905
906 static void unkeep_all_packs(void)
907 {
908 static char name[PATH_MAX];
909 int k;
910
911 for (k = 0; k < pack_id; k++) {
912 struct packed_git *p = all_packs[k];
913 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
914 get_object_directory(), sha1_to_hex(p->sha1));
915 unlink_or_warn(name);
916 }
917 }
918
919 static void end_packfile(void)
920 {
921 struct packed_git *old_p = pack_data, *new_p;
922
923 clear_delta_base_cache();
924 if (object_count) {
925 unsigned char cur_pack_sha1[20];
926 char *idx_name;
927 int i;
928 struct branch *b;
929 struct tag *t;
930
931 close_pack_windows(pack_data);
932 sha1close(pack_file, cur_pack_sha1, 0);
933 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
934 pack_data->pack_name, object_count,
935 cur_pack_sha1, pack_size);
936 close(pack_data->pack_fd);
937 idx_name = keep_pack(create_index());
938
939 /* Register the packfile with core git's machinery. */
940 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
941 if (!new_p)
942 die("core git rejected index %s", idx_name);
943 all_packs[pack_id] = new_p;
944 install_packed_git(new_p);
945
946 /* Print the boundary */
947 if (pack_edges) {
948 fprintf(pack_edges, "%s:", new_p->pack_name);
949 for (i = 0; i < branch_table_sz; i++) {
950 for (b = branch_table[i]; b; b = b->table_next_branch) {
951 if (b->pack_id == pack_id)
952 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
953 }
954 }
955 for (t = first_tag; t; t = t->next_tag) {
956 if (t->pack_id == pack_id)
957 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
958 }
959 fputc('\n', pack_edges);
960 fflush(pack_edges);
961 }
962
963 pack_id++;
964 }
965 else {
966 close(old_p->pack_fd);
967 unlink_or_warn(old_p->pack_name);
968 }
969 free(old_p);
970
971 /* We can't carry a delta across packfiles. */
972 strbuf_release(&last_blob.data);
973 last_blob.offset = 0;
974 last_blob.depth = 0;
975 }
976
977 static void cycle_packfile(void)
978 {
979 end_packfile();
980 start_packfile();
981 }
982
983 static size_t encode_header(
984 enum object_type type,
985 uintmax_t size,
986 unsigned char *hdr)
987 {
988 int n = 1;
989 unsigned char c;
990
991 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
992 die("bad type %d", type);
993
994 c = (type << 4) | (size & 15);
995 size >>= 4;
996 while (size) {
997 *hdr++ = c | 0x80;
998 c = size & 0x7f;
999 size >>= 7;
1000 n++;
1001 }
1002 *hdr = c;
1003 return n;
1004 }
1005
1006 static int store_object(
1007 enum object_type type,
1008 struct strbuf *dat,
1009 struct last_object *last,
1010 unsigned char *sha1out,
1011 uintmax_t mark)
1012 {
1013 void *out, *delta;
1014 struct object_entry *e;
1015 unsigned char hdr[96];
1016 unsigned char sha1[20];
1017 unsigned long hdrlen, deltalen;
1018 git_SHA_CTX c;
1019 z_stream s;
1020
1021 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1022 (unsigned long)dat->len) + 1;
1023 git_SHA1_Init(&c);
1024 git_SHA1_Update(&c, hdr, hdrlen);
1025 git_SHA1_Update(&c, dat->buf, dat->len);
1026 git_SHA1_Final(sha1, &c);
1027 if (sha1out)
1028 hashcpy(sha1out, sha1);
1029
1030 e = insert_object(sha1);
1031 if (mark)
1032 insert_mark(mark, e);
1033 if (e->idx.offset) {
1034 duplicate_count_by_type[type]++;
1035 return 1;
1036 } else if (find_sha1_pack(sha1, packed_git)) {
1037 e->type = type;
1038 e->pack_id = MAX_PACK_ID;
1039 e->idx.offset = 1; /* just not zero! */
1040 duplicate_count_by_type[type]++;
1041 return 1;
1042 }
1043
1044 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1045 delta = diff_delta(last->data.buf, last->data.len,
1046 dat->buf, dat->len,
1047 &deltalen, dat->len - 20);
1048 } else
1049 delta = NULL;
1050
1051 memset(&s, 0, sizeof(s));
1052 deflateInit(&s, pack_compression_level);
1053 if (delta) {
1054 s.next_in = delta;
1055 s.avail_in = deltalen;
1056 } else {
1057 s.next_in = (void *)dat->buf;
1058 s.avail_in = dat->len;
1059 }
1060 s.avail_out = deflateBound(&s, s.avail_in);
1061 s.next_out = out = xmalloc(s.avail_out);
1062 while (deflate(&s, Z_FINISH) == Z_OK)
1063 /* nothing */;
1064 deflateEnd(&s);
1065
1066 /* Determine if we should auto-checkpoint. */
1067 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1068 || (pack_size + 60 + s.total_out) < pack_size) {
1069
1070 /* This new object needs to *not* have the current pack_id. */
1071 e->pack_id = pack_id + 1;
1072 cycle_packfile();
1073
1074 /* We cannot carry a delta into the new pack. */
1075 if (delta) {
1076 free(delta);
1077 delta = NULL;
1078
1079 memset(&s, 0, sizeof(s));
1080 deflateInit(&s, pack_compression_level);
1081 s.next_in = (void *)dat->buf;
1082 s.avail_in = dat->len;
1083 s.avail_out = deflateBound(&s, s.avail_in);
1084 s.next_out = out = xrealloc(out, s.avail_out);
1085 while (deflate(&s, Z_FINISH) == Z_OK)
1086 /* nothing */;
1087 deflateEnd(&s);
1088 }
1089 }
1090
1091 e->type = type;
1092 e->pack_id = pack_id;
1093 e->idx.offset = pack_size;
1094 object_count++;
1095 object_count_by_type[type]++;
1096
1097 crc32_begin(pack_file);
1098
1099 if (delta) {
1100 off_t ofs = e->idx.offset - last->offset;
1101 unsigned pos = sizeof(hdr) - 1;
1102
1103 delta_count_by_type[type]++;
1104 e->depth = last->depth + 1;
1105
1106 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1107 sha1write(pack_file, hdr, hdrlen);
1108 pack_size += hdrlen;
1109
1110 hdr[pos] = ofs & 127;
1111 while (ofs >>= 7)
1112 hdr[--pos] = 128 | (--ofs & 127);
1113 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1114 pack_size += sizeof(hdr) - pos;
1115 } else {
1116 e->depth = 0;
1117 hdrlen = encode_header(type, dat->len, hdr);
1118 sha1write(pack_file, hdr, hdrlen);
1119 pack_size += hdrlen;
1120 }
1121
1122 sha1write(pack_file, out, s.total_out);
1123 pack_size += s.total_out;
1124
1125 e->idx.crc32 = crc32_end(pack_file);
1126
1127 free(out);
1128 free(delta);
1129 if (last) {
1130 if (last->no_swap) {
1131 last->data = *dat;
1132 } else {
1133 strbuf_swap(&last->data, dat);
1134 }
1135 last->offset = e->idx.offset;
1136 last->depth = e->depth;
1137 }
1138 return 0;
1139 }
1140
1141 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1142 {
1143 if (ftruncate(pack_data->pack_fd, to)
1144 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1145 die_errno("cannot truncate pack to skip duplicate");
1146 pack_size = to;
1147
1148 /* yes this is a layering violation */
1149 pack_file->total = to;
1150 pack_file->offset = 0;
1151 pack_file->ctx = *ctx;
1152 }
1153
1154 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1155 {
1156 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1157 unsigned char *in_buf = xmalloc(in_sz);
1158 unsigned char *out_buf = xmalloc(out_sz);
1159 struct object_entry *e;
1160 unsigned char sha1[20];
1161 unsigned long hdrlen;
1162 off_t offset;
1163 git_SHA_CTX c;
1164 git_SHA_CTX pack_file_ctx;
1165 z_stream s;
1166 int status = Z_OK;
1167
1168 /* Determine if we should auto-checkpoint. */
1169 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1170 || (pack_size + 60 + len) < pack_size)
1171 cycle_packfile();
1172
1173 offset = pack_size;
1174
1175 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1176 sha1flush(pack_file);
1177 pack_file_ctx = pack_file->ctx;
1178
1179 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1180 if (out_sz <= hdrlen)
1181 die("impossibly large object header");
1182
1183 git_SHA1_Init(&c);
1184 git_SHA1_Update(&c, out_buf, hdrlen);
1185
1186 crc32_begin(pack_file);
1187
1188 memset(&s, 0, sizeof(s));
1189 deflateInit(&s, pack_compression_level);
1190
1191 hdrlen = encode_header(OBJ_BLOB, len, out_buf);
1192 if (out_sz <= hdrlen)
1193 die("impossibly large object header");
1194
1195 s.next_out = out_buf + hdrlen;
1196 s.avail_out = out_sz - hdrlen;
1197
1198 while (status != Z_STREAM_END) {
1199 if (0 < len && !s.avail_in) {
1200 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1201 size_t n = fread(in_buf, 1, cnt, stdin);
1202 if (!n && feof(stdin))
1203 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1204
1205 git_SHA1_Update(&c, in_buf, n);
1206 s.next_in = in_buf;
1207 s.avail_in = n;
1208 len -= n;
1209 }
1210
1211 status = deflate(&s, len ? 0 : Z_FINISH);
1212
1213 if (!s.avail_out || status == Z_STREAM_END) {
1214 size_t n = s.next_out - out_buf;
1215 sha1write(pack_file, out_buf, n);
1216 pack_size += n;
1217 s.next_out = out_buf;
1218 s.avail_out = out_sz;
1219 }
1220
1221 switch (status) {
1222 case Z_OK:
1223 case Z_BUF_ERROR:
1224 case Z_STREAM_END:
1225 continue;
1226 default:
1227 die("unexpected deflate failure: %d", status);
1228 }
1229 }
1230 deflateEnd(&s);
1231 git_SHA1_Final(sha1, &c);
1232
1233 if (sha1out)
1234 hashcpy(sha1out, sha1);
1235
1236 e = insert_object(sha1);
1237
1238 if (mark)
1239 insert_mark(mark, e);
1240
1241 if (e->idx.offset) {
1242 duplicate_count_by_type[OBJ_BLOB]++;
1243 truncate_pack(offset, &pack_file_ctx);
1244
1245 } else if (find_sha1_pack(sha1, packed_git)) {
1246 e->type = OBJ_BLOB;
1247 e->pack_id = MAX_PACK_ID;
1248 e->idx.offset = 1; /* just not zero! */
1249 duplicate_count_by_type[OBJ_BLOB]++;
1250 truncate_pack(offset, &pack_file_ctx);
1251
1252 } else {
1253 e->depth = 0;
1254 e->type = OBJ_BLOB;
1255 e->pack_id = pack_id;
1256 e->idx.offset = offset;
1257 e->idx.crc32 = crc32_end(pack_file);
1258 object_count++;
1259 object_count_by_type[OBJ_BLOB]++;
1260 }
1261
1262 free(in_buf);
1263 free(out_buf);
1264 }
1265
1266 /* All calls must be guarded by find_object() or find_mark() to
1267 * ensure the 'struct object_entry' passed was written by this
1268 * process instance. We unpack the entry by the offset, avoiding
1269 * the need for the corresponding .idx file. This unpacking rule
1270 * works because we only use OBJ_REF_DELTA within the packfiles
1271 * created by fast-import.
1272 *
1273 * oe must not be NULL. Such an oe usually comes from giving
1274 * an unknown SHA-1 to find_object() or an undefined mark to
1275 * find_mark(). Callers must test for this condition and use
1276 * the standard read_sha1_file() when it happens.
1277 *
1278 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1279 * find_mark(), where the mark was reloaded from an existing marks
1280 * file and is referencing an object that this fast-import process
1281 * instance did not write out to a packfile. Callers must test for
1282 * this condition and use read_sha1_file() instead.
1283 */
1284 static void *gfi_unpack_entry(
1285 struct object_entry *oe,
1286 unsigned long *sizep)
1287 {
1288 enum object_type type;
1289 struct packed_git *p = all_packs[oe->pack_id];
1290 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1291 /* The object is stored in the packfile we are writing to
1292 * and we have modified it since the last time we scanned
1293 * back to read a previously written object. If an old
1294 * window covered [p->pack_size, p->pack_size + 20) its
1295 * data is stale and is not valid. Closing all windows
1296 * and updating the packfile length ensures we can read
1297 * the newly written data.
1298 */
1299 close_pack_windows(p);
1300 sha1flush(pack_file);
1301
1302 /* We have to offer 20 bytes additional on the end of
1303 * the packfile as the core unpacker code assumes the
1304 * footer is present at the file end and must promise
1305 * at least 20 bytes within any window it maps. But
1306 * we don't actually create the footer here.
1307 */
1308 p->pack_size = pack_size + 20;
1309 }
1310 return unpack_entry(p, oe->idx.offset, &type, sizep);
1311 }
1312
1313 static const char *get_mode(const char *str, uint16_t *modep)
1314 {
1315 unsigned char c;
1316 uint16_t mode = 0;
1317
1318 while ((c = *str++) != ' ') {
1319 if (c < '0' || c > '7')
1320 return NULL;
1321 mode = (mode << 3) + (c - '0');
1322 }
1323 *modep = mode;
1324 return str;
1325 }
1326
1327 static void load_tree(struct tree_entry *root)
1328 {
1329 unsigned char *sha1 = root->versions[1].sha1;
1330 struct object_entry *myoe;
1331 struct tree_content *t;
1332 unsigned long size;
1333 char *buf;
1334 const char *c;
1335
1336 root->tree = t = new_tree_content(8);
1337 if (is_null_sha1(sha1))
1338 return;
1339
1340 myoe = find_object(sha1);
1341 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1342 if (myoe->type != OBJ_TREE)
1343 die("Not a tree: %s", sha1_to_hex(sha1));
1344 t->delta_depth = myoe->depth;
1345 buf = gfi_unpack_entry(myoe, &size);
1346 if (!buf)
1347 die("Can't load tree %s", sha1_to_hex(sha1));
1348 } else {
1349 enum object_type type;
1350 buf = read_sha1_file(sha1, &type, &size);
1351 if (!buf || type != OBJ_TREE)
1352 die("Can't load tree %s", sha1_to_hex(sha1));
1353 }
1354
1355 c = buf;
1356 while (c != (buf + size)) {
1357 struct tree_entry *e = new_tree_entry();
1358
1359 if (t->entry_count == t->entry_capacity)
1360 root->tree = t = grow_tree_content(t, t->entry_count);
1361 t->entries[t->entry_count++] = e;
1362
1363 e->tree = NULL;
1364 c = get_mode(c, &e->versions[1].mode);
1365 if (!c)
1366 die("Corrupt mode in %s", sha1_to_hex(sha1));
1367 e->versions[0].mode = e->versions[1].mode;
1368 e->name = to_atom(c, strlen(c));
1369 c += e->name->str_len + 1;
1370 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1371 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1372 c += 20;
1373 }
1374 free(buf);
1375 }
1376
1377 static int tecmp0 (const void *_a, const void *_b)
1378 {
1379 struct tree_entry *a = *((struct tree_entry**)_a);
1380 struct tree_entry *b = *((struct tree_entry**)_b);
1381 return base_name_compare(
1382 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1383 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1384 }
1385
1386 static int tecmp1 (const void *_a, const void *_b)
1387 {
1388 struct tree_entry *a = *((struct tree_entry**)_a);
1389 struct tree_entry *b = *((struct tree_entry**)_b);
1390 return base_name_compare(
1391 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1392 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1393 }
1394
1395 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1396 {
1397 size_t maxlen = 0;
1398 unsigned int i;
1399
1400 if (!v)
1401 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1402 else
1403 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1404
1405 for (i = 0; i < t->entry_count; i++) {
1406 if (t->entries[i]->versions[v].mode)
1407 maxlen += t->entries[i]->name->str_len + 34;
1408 }
1409
1410 strbuf_reset(b);
1411 strbuf_grow(b, maxlen);
1412 for (i = 0; i < t->entry_count; i++) {
1413 struct tree_entry *e = t->entries[i];
1414 if (!e->versions[v].mode)
1415 continue;
1416 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1417 e->name->str_dat, '\0');
1418 strbuf_add(b, e->versions[v].sha1, 20);
1419 }
1420 }
1421
1422 static void store_tree(struct tree_entry *root)
1423 {
1424 struct tree_content *t = root->tree;
1425 unsigned int i, j, del;
1426 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1427 struct object_entry *le;
1428
1429 if (!is_null_sha1(root->versions[1].sha1))
1430 return;
1431
1432 for (i = 0; i < t->entry_count; i++) {
1433 if (t->entries[i]->tree)
1434 store_tree(t->entries[i]);
1435 }
1436
1437 le = find_object(root->versions[0].sha1);
1438 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1439 mktree(t, 0, &old_tree);
1440 lo.data = old_tree;
1441 lo.offset = le->idx.offset;
1442 lo.depth = t->delta_depth;
1443 }
1444
1445 mktree(t, 1, &new_tree);
1446 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1447
1448 t->delta_depth = lo.depth;
1449 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1450 struct tree_entry *e = t->entries[i];
1451 if (e->versions[1].mode) {
1452 e->versions[0].mode = e->versions[1].mode;
1453 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1454 t->entries[j++] = e;
1455 } else {
1456 release_tree_entry(e);
1457 del++;
1458 }
1459 }
1460 t->entry_count -= del;
1461 }
1462
1463 static int tree_content_set(
1464 struct tree_entry *root,
1465 const char *p,
1466 const unsigned char *sha1,
1467 const uint16_t mode,
1468 struct tree_content *subtree)
1469 {
1470 struct tree_content *t = root->tree;
1471 const char *slash1;
1472 unsigned int i, n;
1473 struct tree_entry *e;
1474
1475 slash1 = strchr(p, '/');
1476 if (slash1)
1477 n = slash1 - p;
1478 else
1479 n = strlen(p);
1480 if (!n)
1481 die("Empty path component found in input");
1482 if (!slash1 && !S_ISDIR(mode) && subtree)
1483 die("Non-directories cannot have subtrees");
1484
1485 for (i = 0; i < t->entry_count; i++) {
1486 e = t->entries[i];
1487 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1488 if (!slash1) {
1489 if (!S_ISDIR(mode)
1490 && e->versions[1].mode == mode
1491 && !hashcmp(e->versions[1].sha1, sha1))
1492 return 0;
1493 e->versions[1].mode = mode;
1494 hashcpy(e->versions[1].sha1, sha1);
1495 if (e->tree)
1496 release_tree_content_recursive(e->tree);
1497 e->tree = subtree;
1498 hashclr(root->versions[1].sha1);
1499 return 1;
1500 }
1501 if (!S_ISDIR(e->versions[1].mode)) {
1502 e->tree = new_tree_content(8);
1503 e->versions[1].mode = S_IFDIR;
1504 }
1505 if (!e->tree)
1506 load_tree(e);
1507 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1508 hashclr(root->versions[1].sha1);
1509 return 1;
1510 }
1511 return 0;
1512 }
1513 }
1514
1515 if (t->entry_count == t->entry_capacity)
1516 root->tree = t = grow_tree_content(t, t->entry_count);
1517 e = new_tree_entry();
1518 e->name = to_atom(p, n);
1519 e->versions[0].mode = 0;
1520 hashclr(e->versions[0].sha1);
1521 t->entries[t->entry_count++] = e;
1522 if (slash1) {
1523 e->tree = new_tree_content(8);
1524 e->versions[1].mode = S_IFDIR;
1525 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1526 } else {
1527 e->tree = subtree;
1528 e->versions[1].mode = mode;
1529 hashcpy(e->versions[1].sha1, sha1);
1530 }
1531 hashclr(root->versions[1].sha1);
1532 return 1;
1533 }
1534
1535 static int tree_content_remove(
1536 struct tree_entry *root,
1537 const char *p,
1538 struct tree_entry *backup_leaf)
1539 {
1540 struct tree_content *t = root->tree;
1541 const char *slash1;
1542 unsigned int i, n;
1543 struct tree_entry *e;
1544
1545 slash1 = strchr(p, '/');
1546 if (slash1)
1547 n = slash1 - p;
1548 else
1549 n = strlen(p);
1550
1551 for (i = 0; i < t->entry_count; i++) {
1552 e = t->entries[i];
1553 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1554 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1555 goto del_entry;
1556 if (!e->tree)
1557 load_tree(e);
1558 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1559 for (n = 0; n < e->tree->entry_count; n++) {
1560 if (e->tree->entries[n]->versions[1].mode) {
1561 hashclr(root->versions[1].sha1);
1562 return 1;
1563 }
1564 }
1565 backup_leaf = NULL;
1566 goto del_entry;
1567 }
1568 return 0;
1569 }
1570 }
1571 return 0;
1572
1573 del_entry:
1574 if (backup_leaf)
1575 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1576 else if (e->tree)
1577 release_tree_content_recursive(e->tree);
1578 e->tree = NULL;
1579 e->versions[1].mode = 0;
1580 hashclr(e->versions[1].sha1);
1581 hashclr(root->versions[1].sha1);
1582 return 1;
1583 }
1584
1585 static int tree_content_get(
1586 struct tree_entry *root,
1587 const char *p,
1588 struct tree_entry *leaf)
1589 {
1590 struct tree_content *t = root->tree;
1591 const char *slash1;
1592 unsigned int i, n;
1593 struct tree_entry *e;
1594
1595 slash1 = strchr(p, '/');
1596 if (slash1)
1597 n = slash1 - p;
1598 else
1599 n = strlen(p);
1600
1601 for (i = 0; i < t->entry_count; i++) {
1602 e = t->entries[i];
1603 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1604 if (!slash1) {
1605 memcpy(leaf, e, sizeof(*leaf));
1606 if (e->tree && is_null_sha1(e->versions[1].sha1))
1607 leaf->tree = dup_tree_content(e->tree);
1608 else
1609 leaf->tree = NULL;
1610 return 1;
1611 }
1612 if (!S_ISDIR(e->versions[1].mode))
1613 return 0;
1614 if (!e->tree)
1615 load_tree(e);
1616 return tree_content_get(e, slash1 + 1, leaf);
1617 }
1618 }
1619 return 0;
1620 }
1621
1622 static int update_branch(struct branch *b)
1623 {
1624 static const char *msg = "fast-import";
1625 struct ref_lock *lock;
1626 unsigned char old_sha1[20];
1627
1628 if (is_null_sha1(b->sha1))
1629 return 0;
1630 if (read_ref(b->name, old_sha1))
1631 hashclr(old_sha1);
1632 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1633 if (!lock)
1634 return error("Unable to lock %s", b->name);
1635 if (!force_update && !is_null_sha1(old_sha1)) {
1636 struct commit *old_cmit, *new_cmit;
1637
1638 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1639 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1640 if (!old_cmit || !new_cmit) {
1641 unlock_ref(lock);
1642 return error("Branch %s is missing commits.", b->name);
1643 }
1644
1645 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1646 unlock_ref(lock);
1647 warning("Not updating %s"
1648 " (new tip %s does not contain %s)",
1649 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1650 return -1;
1651 }
1652 }
1653 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1654 return error("Unable to update %s", b->name);
1655 return 0;
1656 }
1657
1658 static void dump_branches(void)
1659 {
1660 unsigned int i;
1661 struct branch *b;
1662
1663 for (i = 0; i < branch_table_sz; i++) {
1664 for (b = branch_table[i]; b; b = b->table_next_branch)
1665 failure |= update_branch(b);
1666 }
1667 }
1668
1669 static void dump_tags(void)
1670 {
1671 static const char *msg = "fast-import";
1672 struct tag *t;
1673 struct ref_lock *lock;
1674 char ref_name[PATH_MAX];
1675
1676 for (t = first_tag; t; t = t->next_tag) {
1677 sprintf(ref_name, "tags/%s", t->name);
1678 lock = lock_ref_sha1(ref_name, NULL);
1679 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1680 failure |= error("Unable to update %s", ref_name);
1681 }
1682 }
1683
1684 static void dump_marks_helper(FILE *f,
1685 uintmax_t base,
1686 struct mark_set *m)
1687 {
1688 uintmax_t k;
1689 if (m->shift) {
1690 for (k = 0; k < 1024; k++) {
1691 if (m->data.sets[k])
1692 dump_marks_helper(f, (base + k) << m->shift,
1693 m->data.sets[k]);
1694 }
1695 } else {
1696 for (k = 0; k < 1024; k++) {
1697 if (m->data.marked[k])
1698 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1699 sha1_to_hex(m->data.marked[k]->idx.sha1));
1700 }
1701 }
1702 }
1703
1704 static void dump_marks(void)
1705 {
1706 static struct lock_file mark_lock;
1707 int mark_fd;
1708 FILE *f;
1709
1710 if (!export_marks_file)
1711 return;
1712
1713 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1714 if (mark_fd < 0) {
1715 failure |= error("Unable to write marks file %s: %s",
1716 export_marks_file, strerror(errno));
1717 return;
1718 }
1719
1720 f = fdopen(mark_fd, "w");
1721 if (!f) {
1722 int saved_errno = errno;
1723 rollback_lock_file(&mark_lock);
1724 failure |= error("Unable to write marks file %s: %s",
1725 export_marks_file, strerror(saved_errno));
1726 return;
1727 }
1728
1729 /*
1730 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1731 * Assign -1 to the lock file descriptor so that commit_lock_file()
1732 * won't try to close() it.
1733 */
1734 mark_lock.fd = -1;
1735
1736 dump_marks_helper(f, 0, marks);
1737 if (ferror(f) || fclose(f)) {
1738 int saved_errno = errno;
1739 rollback_lock_file(&mark_lock);
1740 failure |= error("Unable to write marks file %s: %s",
1741 export_marks_file, strerror(saved_errno));
1742 return;
1743 }
1744
1745 if (commit_lock_file(&mark_lock)) {
1746 int saved_errno = errno;
1747 rollback_lock_file(&mark_lock);
1748 failure |= error("Unable to commit marks file %s: %s",
1749 export_marks_file, strerror(saved_errno));
1750 return;
1751 }
1752 }
1753
1754 static void read_marks(void)
1755 {
1756 char line[512];
1757 FILE *f = fopen(import_marks_file, "r");
1758 if (!f)
1759 die_errno("cannot read '%s'", import_marks_file);
1760 while (fgets(line, sizeof(line), f)) {
1761 uintmax_t mark;
1762 char *end;
1763 unsigned char sha1[20];
1764 struct object_entry *e;
1765
1766 end = strchr(line, '\n');
1767 if (line[0] != ':' || !end)
1768 die("corrupt mark line: %s", line);
1769 *end = 0;
1770 mark = strtoumax(line + 1, &end, 10);
1771 if (!mark || end == line + 1
1772 || *end != ' ' || get_sha1(end + 1, sha1))
1773 die("corrupt mark line: %s", line);
1774 e = find_object(sha1);
1775 if (!e) {
1776 enum object_type type = sha1_object_info(sha1, NULL);
1777 if (type < 0)
1778 die("object not found: %s", sha1_to_hex(sha1));
1779 e = insert_object(sha1);
1780 e->type = type;
1781 e->pack_id = MAX_PACK_ID;
1782 e->idx.offset = 1; /* just not zero! */
1783 }
1784 insert_mark(mark, e);
1785 }
1786 fclose(f);
1787 }
1788
1789
1790 static int read_next_command(void)
1791 {
1792 static int stdin_eof = 0;
1793
1794 if (stdin_eof) {
1795 unread_command_buf = 0;
1796 return EOF;
1797 }
1798
1799 do {
1800 if (unread_command_buf) {
1801 unread_command_buf = 0;
1802 } else {
1803 struct recent_command *rc;
1804
1805 strbuf_detach(&command_buf, NULL);
1806 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1807 if (stdin_eof)
1808 return EOF;
1809
1810 if (!seen_data_command
1811 && prefixcmp(command_buf.buf, "feature ")
1812 && prefixcmp(command_buf.buf, "option ")) {
1813 parse_argv();
1814 }
1815
1816 rc = rc_free;
1817 if (rc)
1818 rc_free = rc->next;
1819 else {
1820 rc = cmd_hist.next;
1821 cmd_hist.next = rc->next;
1822 cmd_hist.next->prev = &cmd_hist;
1823 free(rc->buf);
1824 }
1825
1826 rc->buf = command_buf.buf;
1827 rc->prev = cmd_tail;
1828 rc->next = cmd_hist.prev;
1829 rc->prev->next = rc;
1830 cmd_tail = rc;
1831 }
1832 } while (command_buf.buf[0] == '#');
1833
1834 return 0;
1835 }
1836
1837 static void skip_optional_lf(void)
1838 {
1839 int term_char = fgetc(stdin);
1840 if (term_char != '\n' && term_char != EOF)
1841 ungetc(term_char, stdin);
1842 }
1843
1844 static void parse_mark(void)
1845 {
1846 if (!prefixcmp(command_buf.buf, "mark :")) {
1847 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1848 read_next_command();
1849 }
1850 else
1851 next_mark = 0;
1852 }
1853
1854 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1855 {
1856 strbuf_reset(sb);
1857
1858 if (prefixcmp(command_buf.buf, "data "))
1859 die("Expected 'data n' command, found: %s", command_buf.buf);
1860
1861 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1862 char *term = xstrdup(command_buf.buf + 5 + 2);
1863 size_t term_len = command_buf.len - 5 - 2;
1864
1865 strbuf_detach(&command_buf, NULL);
1866 for (;;) {
1867 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1868 die("EOF in data (terminator '%s' not found)", term);
1869 if (term_len == command_buf.len
1870 && !strcmp(term, command_buf.buf))
1871 break;
1872 strbuf_addbuf(sb, &command_buf);
1873 strbuf_addch(sb, '\n');
1874 }
1875 free(term);
1876 }
1877 else {
1878 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1879 size_t n = 0, length = (size_t)len;
1880
1881 if (limit && limit < len) {
1882 *len_res = len;
1883 return 0;
1884 }
1885 if (length < len)
1886 die("data is too large to use in this context");
1887
1888 while (n < length) {
1889 size_t s = strbuf_fread(sb, length - n, stdin);
1890 if (!s && feof(stdin))
1891 die("EOF in data (%lu bytes remaining)",
1892 (unsigned long)(length - n));
1893 n += s;
1894 }
1895 }
1896
1897 skip_optional_lf();
1898 return 1;
1899 }
1900
1901 static int validate_raw_date(const char *src, char *result, int maxlen)
1902 {
1903 const char *orig_src = src;
1904 char *endp;
1905 unsigned long num;
1906
1907 errno = 0;
1908
1909 num = strtoul(src, &endp, 10);
1910 /* NEEDSWORK: perhaps check for reasonable values? */
1911 if (errno || endp == src || *endp != ' ')
1912 return -1;
1913
1914 src = endp + 1;
1915 if (*src != '-' && *src != '+')
1916 return -1;
1917
1918 num = strtoul(src + 1, &endp, 10);
1919 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1920 1400 < num)
1921 return -1;
1922
1923 strcpy(result, orig_src);
1924 return 0;
1925 }
1926
1927 static char *parse_ident(const char *buf)
1928 {
1929 const char *gt;
1930 size_t name_len;
1931 char *ident;
1932
1933 gt = strrchr(buf, '>');
1934 if (!gt)
1935 die("Missing > in ident string: %s", buf);
1936 gt++;
1937 if (*gt != ' ')
1938 die("Missing space after > in ident string: %s", buf);
1939 gt++;
1940 name_len = gt - buf;
1941 ident = xmalloc(name_len + 24);
1942 strncpy(ident, buf, name_len);
1943
1944 switch (whenspec) {
1945 case WHENSPEC_RAW:
1946 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1947 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1948 break;
1949 case WHENSPEC_RFC2822:
1950 if (parse_date(gt, ident + name_len, 24) < 0)
1951 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1952 break;
1953 case WHENSPEC_NOW:
1954 if (strcmp("now", gt))
1955 die("Date in ident must be 'now': %s", buf);
1956 datestamp(ident + name_len, 24);
1957 break;
1958 }
1959
1960 return ident;
1961 }
1962
1963 static void parse_and_store_blob(
1964 struct last_object *last,
1965 unsigned char *sha1out,
1966 uintmax_t mark)
1967 {
1968 static struct strbuf buf = STRBUF_INIT;
1969 uintmax_t len;
1970
1971 if (parse_data(&buf, big_file_threshold, &len))
1972 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
1973 else {
1974 if (last) {
1975 strbuf_release(&last->data);
1976 last->offset = 0;
1977 last->depth = 0;
1978 }
1979 stream_blob(len, sha1out, mark);
1980 skip_optional_lf();
1981 }
1982 }
1983
1984 static void parse_new_blob(void)
1985 {
1986 read_next_command();
1987 parse_mark();
1988 parse_and_store_blob(&last_blob, NULL, next_mark);
1989 }
1990
1991 static void unload_one_branch(void)
1992 {
1993 while (cur_active_branches
1994 && cur_active_branches >= max_active_branches) {
1995 uintmax_t min_commit = ULONG_MAX;
1996 struct branch *e, *l = NULL, *p = NULL;
1997
1998 for (e = active_branches; e; e = e->active_next_branch) {
1999 if (e->last_commit < min_commit) {
2000 p = l;
2001 min_commit = e->last_commit;
2002 }
2003 l = e;
2004 }
2005
2006 if (p) {
2007 e = p->active_next_branch;
2008 p->active_next_branch = e->active_next_branch;
2009 } else {
2010 e = active_branches;
2011 active_branches = e->active_next_branch;
2012 }
2013 e->active = 0;
2014 e->active_next_branch = NULL;
2015 if (e->branch_tree.tree) {
2016 release_tree_content_recursive(e->branch_tree.tree);
2017 e->branch_tree.tree = NULL;
2018 }
2019 cur_active_branches--;
2020 }
2021 }
2022
2023 static void load_branch(struct branch *b)
2024 {
2025 load_tree(&b->branch_tree);
2026 if (!b->active) {
2027 b->active = 1;
2028 b->active_next_branch = active_branches;
2029 active_branches = b;
2030 cur_active_branches++;
2031 branch_load_count++;
2032 }
2033 }
2034
2035 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2036 {
2037 unsigned char fanout = 0;
2038 while ((num_notes >>= 8))
2039 fanout++;
2040 return fanout;
2041 }
2042
2043 static void construct_path_with_fanout(const char *hex_sha1,
2044 unsigned char fanout, char *path)
2045 {
2046 unsigned int i = 0, j = 0;
2047 if (fanout >= 20)
2048 die("Too large fanout (%u)", fanout);
2049 while (fanout) {
2050 path[i++] = hex_sha1[j++];
2051 path[i++] = hex_sha1[j++];
2052 path[i++] = '/';
2053 fanout--;
2054 }
2055 memcpy(path + i, hex_sha1 + j, 40 - j);
2056 path[i + 40 - j] = '\0';
2057 }
2058
2059 static uintmax_t do_change_note_fanout(
2060 struct tree_entry *orig_root, struct tree_entry *root,
2061 char *hex_sha1, unsigned int hex_sha1_len,
2062 char *fullpath, unsigned int fullpath_len,
2063 unsigned char fanout)
2064 {
2065 struct tree_content *t = root->tree;
2066 struct tree_entry *e, leaf;
2067 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2068 uintmax_t num_notes = 0;
2069 unsigned char sha1[20];
2070 char realpath[60];
2071
2072 for (i = 0; t && i < t->entry_count; i++) {
2073 e = t->entries[i];
2074 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2075 tmp_fullpath_len = fullpath_len;
2076
2077 /*
2078 * We're interested in EITHER existing note entries (entries
2079 * with exactly 40 hex chars in path, not including directory
2080 * separators), OR directory entries that may contain note
2081 * entries (with < 40 hex chars in path).
2082 * Also, each path component in a note entry must be a multiple
2083 * of 2 chars.
2084 */
2085 if (!e->versions[1].mode ||
2086 tmp_hex_sha1_len > 40 ||
2087 e->name->str_len % 2)
2088 continue;
2089
2090 /* This _may_ be a note entry, or a subdir containing notes */
2091 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2092 e->name->str_len);
2093 if (tmp_fullpath_len)
2094 fullpath[tmp_fullpath_len++] = '/';
2095 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2096 e->name->str_len);
2097 tmp_fullpath_len += e->name->str_len;
2098 fullpath[tmp_fullpath_len] = '\0';
2099
2100 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2101 /* This is a note entry */
2102 construct_path_with_fanout(hex_sha1, fanout, realpath);
2103 if (!strcmp(fullpath, realpath)) {
2104 /* Note entry is in correct location */
2105 num_notes++;
2106 continue;
2107 }
2108
2109 /* Rename fullpath to realpath */
2110 if (!tree_content_remove(orig_root, fullpath, &leaf))
2111 die("Failed to remove path %s", fullpath);
2112 tree_content_set(orig_root, realpath,
2113 leaf.versions[1].sha1,
2114 leaf.versions[1].mode,
2115 leaf.tree);
2116 } else if (S_ISDIR(e->versions[1].mode)) {
2117 /* This is a subdir that may contain note entries */
2118 if (!e->tree)
2119 load_tree(e);
2120 num_notes += do_change_note_fanout(orig_root, e,
2121 hex_sha1, tmp_hex_sha1_len,
2122 fullpath, tmp_fullpath_len, fanout);
2123 }
2124
2125 /* The above may have reallocated the current tree_content */
2126 t = root->tree;
2127 }
2128 return num_notes;
2129 }
2130
2131 static uintmax_t change_note_fanout(struct tree_entry *root,
2132 unsigned char fanout)
2133 {
2134 char hex_sha1[40], path[60];
2135 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2136 }
2137
2138 static void file_change_m(struct branch *b)
2139 {
2140 const char *p = command_buf.buf + 2;
2141 static struct strbuf uq = STRBUF_INIT;
2142 const char *endp;
2143 struct object_entry *oe = oe;
2144 unsigned char sha1[20];
2145 uint16_t mode, inline_data = 0;
2146
2147 p = get_mode(p, &mode);
2148 if (!p)
2149 die("Corrupt mode: %s", command_buf.buf);
2150 switch (mode) {
2151 case 0644:
2152 case 0755:
2153 mode |= S_IFREG;
2154 case S_IFREG | 0644:
2155 case S_IFREG | 0755:
2156 case S_IFLNK:
2157 case S_IFGITLINK:
2158 /* ok */
2159 break;
2160 default:
2161 die("Corrupt mode: %s", command_buf.buf);
2162 }
2163
2164 if (*p == ':') {
2165 char *x;
2166 oe = find_mark(strtoumax(p + 1, &x, 10));
2167 hashcpy(sha1, oe->idx.sha1);
2168 p = x;
2169 } else if (!prefixcmp(p, "inline")) {
2170 inline_data = 1;
2171 p += 6;
2172 } else {
2173 if (get_sha1_hex(p, sha1))
2174 die("Invalid SHA1: %s", command_buf.buf);
2175 oe = find_object(sha1);
2176 p += 40;
2177 }
2178 if (*p++ != ' ')
2179 die("Missing space after SHA1: %s", command_buf.buf);
2180
2181 strbuf_reset(&uq);
2182 if (!unquote_c_style(&uq, p, &endp)) {
2183 if (*endp)
2184 die("Garbage after path in: %s", command_buf.buf);
2185 p = uq.buf;
2186 }
2187
2188 if (S_ISGITLINK(mode)) {
2189 if (inline_data)
2190 die("Git links cannot be specified 'inline': %s",
2191 command_buf.buf);
2192 else if (oe) {
2193 if (oe->type != OBJ_COMMIT)
2194 die("Not a commit (actually a %s): %s",
2195 typename(oe->type), command_buf.buf);
2196 }
2197 /*
2198 * Accept the sha1 without checking; it expected to be in
2199 * another repository.
2200 */
2201 } else if (inline_data) {
2202 if (p != uq.buf) {
2203 strbuf_addstr(&uq, p);
2204 p = uq.buf;
2205 }
2206 read_next_command();
2207 parse_and_store_blob(&last_blob, sha1, 0);
2208 } else if (oe) {
2209 if (oe->type != OBJ_BLOB)
2210 die("Not a blob (actually a %s): %s",
2211 typename(oe->type), command_buf.buf);
2212 } else {
2213 enum object_type type = sha1_object_info(sha1, NULL);
2214 if (type < 0)
2215 die("Blob not found: %s", command_buf.buf);
2216 if (type != OBJ_BLOB)
2217 die("Not a blob (actually a %s): %s",
2218 typename(type), command_buf.buf);
2219 }
2220
2221 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2222 }
2223
2224 static void file_change_d(struct branch *b)
2225 {
2226 const char *p = command_buf.buf + 2;
2227 static struct strbuf uq = STRBUF_INIT;
2228 const char *endp;
2229
2230 strbuf_reset(&uq);
2231 if (!unquote_c_style(&uq, p, &endp)) {
2232 if (*endp)
2233 die("Garbage after path in: %s", command_buf.buf);
2234 p = uq.buf;
2235 }
2236 tree_content_remove(&b->branch_tree, p, NULL);
2237 }
2238
2239 static void file_change_cr(struct branch *b, int rename)
2240 {
2241 const char *s, *d;
2242 static struct strbuf s_uq = STRBUF_INIT;
2243 static struct strbuf d_uq = STRBUF_INIT;
2244 const char *endp;
2245 struct tree_entry leaf;
2246
2247 s = command_buf.buf + 2;
2248 strbuf_reset(&s_uq);
2249 if (!unquote_c_style(&s_uq, s, &endp)) {
2250 if (*endp != ' ')
2251 die("Missing space after source: %s", command_buf.buf);
2252 } else {
2253 endp = strchr(s, ' ');
2254 if (!endp)
2255 die("Missing space after source: %s", command_buf.buf);
2256 strbuf_add(&s_uq, s, endp - s);
2257 }
2258 s = s_uq.buf;
2259
2260 endp++;
2261 if (!*endp)
2262 die("Missing dest: %s", command_buf.buf);
2263
2264 d = endp;
2265 strbuf_reset(&d_uq);
2266 if (!unquote_c_style(&d_uq, d, &endp)) {
2267 if (*endp)
2268 die("Garbage after dest in: %s", command_buf.buf);
2269 d = d_uq.buf;
2270 }
2271
2272 memset(&leaf, 0, sizeof(leaf));
2273 if (rename)
2274 tree_content_remove(&b->branch_tree, s, &leaf);
2275 else
2276 tree_content_get(&b->branch_tree, s, &leaf);
2277 if (!leaf.versions[1].mode)
2278 die("Path %s not in branch", s);
2279 tree_content_set(&b->branch_tree, d,
2280 leaf.versions[1].sha1,
2281 leaf.versions[1].mode,
2282 leaf.tree);
2283 }
2284
2285 static void note_change_n(struct branch *b, unsigned char old_fanout)
2286 {
2287 const char *p = command_buf.buf + 2;
2288 static struct strbuf uq = STRBUF_INIT;
2289 struct object_entry *oe = oe;
2290 struct branch *s;
2291 unsigned char sha1[20], commit_sha1[20];
2292 char path[60];
2293 uint16_t inline_data = 0;
2294 unsigned char new_fanout;
2295
2296 /* <dataref> or 'inline' */
2297 if (*p == ':') {
2298 char *x;
2299 oe = find_mark(strtoumax(p + 1, &x, 10));
2300 hashcpy(sha1, oe->idx.sha1);
2301 p = x;
2302 } else if (!prefixcmp(p, "inline")) {
2303 inline_data = 1;
2304 p += 6;
2305 } else {
2306 if (get_sha1_hex(p, sha1))
2307 die("Invalid SHA1: %s", command_buf.buf);
2308 oe = find_object(sha1);
2309 p += 40;
2310 }
2311 if (*p++ != ' ')
2312 die("Missing space after SHA1: %s", command_buf.buf);
2313
2314 /* <committish> */
2315 s = lookup_branch(p);
2316 if (s) {
2317 hashcpy(commit_sha1, s->sha1);
2318 } else if (*p == ':') {
2319 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2320 struct object_entry *commit_oe = find_mark(commit_mark);
2321 if (commit_oe->type != OBJ_COMMIT)
2322 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2323 hashcpy(commit_sha1, commit_oe->idx.sha1);
2324 } else if (!get_sha1(p, commit_sha1)) {
2325 unsigned long size;
2326 char *buf = read_object_with_reference(commit_sha1,
2327 commit_type, &size, commit_sha1);
2328 if (!buf || size < 46)
2329 die("Not a valid commit: %s", p);
2330 free(buf);
2331 } else
2332 die("Invalid ref name or SHA1 expression: %s", p);
2333
2334 if (inline_data) {
2335 if (p != uq.buf) {
2336 strbuf_addstr(&uq, p);
2337 p = uq.buf;
2338 }
2339 read_next_command();
2340 parse_and_store_blob(&last_blob, sha1, 0);
2341 } else if (oe) {
2342 if (oe->type != OBJ_BLOB)
2343 die("Not a blob (actually a %s): %s",
2344 typename(oe->type), command_buf.buf);
2345 } else if (!is_null_sha1(sha1)) {
2346 enum object_type type = sha1_object_info(sha1, NULL);
2347 if (type < 0)
2348 die("Blob not found: %s", command_buf.buf);
2349 if (type != OBJ_BLOB)
2350 die("Not a blob (actually a %s): %s",
2351 typename(type), command_buf.buf);
2352 }
2353
2354 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2355 if (tree_content_remove(&b->branch_tree, path, NULL))
2356 b->num_notes--;
2357
2358 if (is_null_sha1(sha1))
2359 return; /* nothing to insert */
2360
2361 b->num_notes++;
2362 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2363 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2364 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2365 }
2366
2367 static void file_change_deleteall(struct branch *b)
2368 {
2369 release_tree_content_recursive(b->branch_tree.tree);
2370 hashclr(b->branch_tree.versions[0].sha1);
2371 hashclr(b->branch_tree.versions[1].sha1);
2372 load_tree(&b->branch_tree);
2373 b->num_notes = 0;
2374 }
2375
2376 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2377 {
2378 if (!buf || size < 46)
2379 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2380 if (memcmp("tree ", buf, 5)
2381 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2382 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2383 hashcpy(b->branch_tree.versions[0].sha1,
2384 b->branch_tree.versions[1].sha1);
2385 }
2386
2387 static void parse_from_existing(struct branch *b)
2388 {
2389 if (is_null_sha1(b->sha1)) {
2390 hashclr(b->branch_tree.versions[0].sha1);
2391 hashclr(b->branch_tree.versions[1].sha1);
2392 } else {
2393 unsigned long size;
2394 char *buf;
2395
2396 buf = read_object_with_reference(b->sha1,
2397 commit_type, &size, b->sha1);
2398 parse_from_commit(b, buf, size);
2399 free(buf);
2400 }
2401 }
2402
2403 static int parse_from(struct branch *b)
2404 {
2405 const char *from;
2406 struct branch *s;
2407
2408 if (prefixcmp(command_buf.buf, "from "))
2409 return 0;
2410
2411 if (b->branch_tree.tree) {
2412 release_tree_content_recursive(b->branch_tree.tree);
2413 b->branch_tree.tree = NULL;
2414 }
2415
2416 from = strchr(command_buf.buf, ' ') + 1;
2417 s = lookup_branch(from);
2418 if (b == s)
2419 die("Can't create a branch from itself: %s", b->name);
2420 else if (s) {
2421 unsigned char *t = s->branch_tree.versions[1].sha1;
2422 hashcpy(b->sha1, s->sha1);
2423 hashcpy(b->branch_tree.versions[0].sha1, t);
2424 hashcpy(b->branch_tree.versions[1].sha1, t);
2425 } else if (*from == ':') {
2426 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2427 struct object_entry *oe = find_mark(idnum);
2428 if (oe->type != OBJ_COMMIT)
2429 die("Mark :%" PRIuMAX " not a commit", idnum);
2430 hashcpy(b->sha1, oe->idx.sha1);
2431 if (oe->pack_id != MAX_PACK_ID) {
2432 unsigned long size;
2433 char *buf = gfi_unpack_entry(oe, &size);
2434 parse_from_commit(b, buf, size);
2435 free(buf);
2436 } else
2437 parse_from_existing(b);
2438 } else if (!get_sha1(from, b->sha1))
2439 parse_from_existing(b);
2440 else
2441 die("Invalid ref name or SHA1 expression: %s", from);
2442
2443 read_next_command();
2444 return 1;
2445 }
2446
2447 static struct hash_list *parse_merge(unsigned int *count)
2448 {
2449 struct hash_list *list = NULL, *n, *e = e;
2450 const char *from;
2451 struct branch *s;
2452
2453 *count = 0;
2454 while (!prefixcmp(command_buf.buf, "merge ")) {
2455 from = strchr(command_buf.buf, ' ') + 1;
2456 n = xmalloc(sizeof(*n));
2457 s = lookup_branch(from);
2458 if (s)
2459 hashcpy(n->sha1, s->sha1);
2460 else if (*from == ':') {
2461 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2462 struct object_entry *oe = find_mark(idnum);
2463 if (oe->type != OBJ_COMMIT)
2464 die("Mark :%" PRIuMAX " not a commit", idnum);
2465 hashcpy(n->sha1, oe->idx.sha1);
2466 } else if (!get_sha1(from, n->sha1)) {
2467 unsigned long size;
2468 char *buf = read_object_with_reference(n->sha1,
2469 commit_type, &size, n->sha1);
2470 if (!buf || size < 46)
2471 die("Not a valid commit: %s", from);
2472 free(buf);
2473 } else
2474 die("Invalid ref name or SHA1 expression: %s", from);
2475
2476 n->next = NULL;
2477 if (list)
2478 e->next = n;
2479 else
2480 list = n;
2481 e = n;
2482 (*count)++;
2483 read_next_command();
2484 }
2485 return list;
2486 }
2487
2488 static void parse_new_commit(void)
2489 {
2490 static struct strbuf msg = STRBUF_INIT;
2491 struct branch *b;
2492 char *sp;
2493 char *author = NULL;
2494 char *committer = NULL;
2495 struct hash_list *merge_list = NULL;
2496 unsigned int merge_count;
2497 unsigned char prev_fanout, new_fanout;
2498
2499 /* Obtain the branch name from the rest of our command */
2500 sp = strchr(command_buf.buf, ' ') + 1;
2501 b = lookup_branch(sp);
2502 if (!b)
2503 b = new_branch(sp);
2504
2505 read_next_command();
2506 parse_mark();
2507 if (!prefixcmp(command_buf.buf, "author ")) {
2508 author = parse_ident(command_buf.buf + 7);
2509 read_next_command();
2510 }
2511 if (!prefixcmp(command_buf.buf, "committer ")) {
2512 committer = parse_ident(command_buf.buf + 10);
2513 read_next_command();
2514 }
2515 if (!committer)
2516 die("Expected committer but didn't get one");
2517 parse_data(&msg, 0, NULL);
2518 read_next_command();
2519 parse_from(b);
2520 merge_list = parse_merge(&merge_count);
2521
2522 /* ensure the branch is active/loaded */
2523 if (!b->branch_tree.tree || !max_active_branches) {
2524 unload_one_branch();
2525 load_branch(b);
2526 }
2527
2528 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2529
2530 /* file_change* */
2531 while (command_buf.len > 0) {
2532 if (!prefixcmp(command_buf.buf, "M "))
2533 file_change_m(b);
2534 else if (!prefixcmp(command_buf.buf, "D "))
2535 file_change_d(b);
2536 else if (!prefixcmp(command_buf.buf, "R "))
2537 file_change_cr(b, 1);
2538 else if (!prefixcmp(command_buf.buf, "C "))
2539 file_change_cr(b, 0);
2540 else if (!prefixcmp(command_buf.buf, "N "))
2541 note_change_n(b, prev_fanout);
2542 else if (!strcmp("deleteall", command_buf.buf))
2543 file_change_deleteall(b);
2544 else {
2545 unread_command_buf = 1;
2546 break;
2547 }
2548 if (read_next_command() == EOF)
2549 break;
2550 }
2551
2552 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2553 if (new_fanout != prev_fanout)
2554 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2555
2556 /* build the tree and the commit */
2557 store_tree(&b->branch_tree);
2558 hashcpy(b->branch_tree.versions[0].sha1,
2559 b->branch_tree.versions[1].sha1);
2560
2561 strbuf_reset(&new_data);
2562 strbuf_addf(&new_data, "tree %s\n",
2563 sha1_to_hex(b->branch_tree.versions[1].sha1));
2564 if (!is_null_sha1(b->sha1))
2565 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2566 while (merge_list) {
2567 struct hash_list *next = merge_list->next;
2568 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2569 free(merge_list);
2570 merge_list = next;
2571 }
2572 strbuf_addf(&new_data,
2573 "author %s\n"
2574 "committer %s\n"
2575 "\n",
2576 author ? author : committer, committer);
2577 strbuf_addbuf(&new_data, &msg);
2578 free(author);
2579 free(committer);
2580
2581 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2582 b->pack_id = pack_id;
2583 b->last_commit = object_count_by_type[OBJ_COMMIT];
2584 }
2585
2586 static void parse_new_tag(void)
2587 {
2588 static struct strbuf msg = STRBUF_INIT;
2589 char *sp;
2590 const char *from;
2591 char *tagger;
2592 struct branch *s;
2593 struct tag *t;
2594 uintmax_t from_mark = 0;
2595 unsigned char sha1[20];
2596 enum object_type type;
2597
2598 /* Obtain the new tag name from the rest of our command */
2599 sp = strchr(command_buf.buf, ' ') + 1;
2600 t = pool_alloc(sizeof(struct tag));
2601 t->next_tag = NULL;
2602 t->name = pool_strdup(sp);
2603 if (last_tag)
2604 last_tag->next_tag = t;
2605 else
2606 first_tag = t;
2607 last_tag = t;
2608 read_next_command();
2609
2610 /* from ... */
2611 if (prefixcmp(command_buf.buf, "from "))
2612 die("Expected from command, got %s", command_buf.buf);
2613 from = strchr(command_buf.buf, ' ') + 1;
2614 s = lookup_branch(from);
2615 if (s) {
2616 hashcpy(sha1, s->sha1);
2617 type = OBJ_COMMIT;
2618 } else if (*from == ':') {
2619 struct object_entry *oe;
2620 from_mark = strtoumax(from + 1, NULL, 10);
2621 oe = find_mark(from_mark);
2622 type = oe->type;
2623 hashcpy(sha1, oe->idx.sha1);
2624 } else if (!get_sha1(from, sha1)) {
2625 unsigned long size;
2626 char *buf;
2627
2628 buf = read_sha1_file(sha1, &type, &size);
2629 if (!buf || size < 46)
2630 die("Not a valid commit: %s", from);
2631 free(buf);
2632 } else
2633 die("Invalid ref name or SHA1 expression: %s", from);
2634 read_next_command();
2635
2636 /* tagger ... */
2637 if (!prefixcmp(command_buf.buf, "tagger ")) {
2638 tagger = parse_ident(command_buf.buf + 7);
2639 read_next_command();
2640 } else
2641 tagger = NULL;
2642
2643 /* tag payload/message */
2644 parse_data(&msg, 0, NULL);
2645
2646 /* build the tag object */
2647 strbuf_reset(&new_data);
2648
2649 strbuf_addf(&new_data,
2650 "object %s\n"
2651 "type %s\n"
2652 "tag %s\n",
2653 sha1_to_hex(sha1), typename(type), t->name);
2654 if (tagger)
2655 strbuf_addf(&new_data,
2656 "tagger %s\n", tagger);
2657 strbuf_addch(&new_data, '\n');
2658 strbuf_addbuf(&new_data, &msg);
2659 free(tagger);
2660
2661 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2662 t->pack_id = MAX_PACK_ID;
2663 else
2664 t->pack_id = pack_id;
2665 }
2666
2667 static void parse_reset_branch(void)
2668 {
2669 struct branch *b;
2670 char *sp;
2671
2672 /* Obtain the branch name from the rest of our command */
2673 sp = strchr(command_buf.buf, ' ') + 1;
2674 b = lookup_branch(sp);
2675 if (b) {
2676 hashclr(b->sha1);
2677 hashclr(b->branch_tree.versions[0].sha1);
2678 hashclr(b->branch_tree.versions[1].sha1);
2679 if (b->branch_tree.tree) {
2680 release_tree_content_recursive(b->branch_tree.tree);
2681 b->branch_tree.tree = NULL;
2682 }
2683 }
2684 else
2685 b = new_branch(sp);
2686 read_next_command();
2687 parse_from(b);
2688 if (command_buf.len > 0)
2689 unread_command_buf = 1;
2690 }
2691
2692 static void parse_checkpoint(void)
2693 {
2694 if (object_count) {
2695 cycle_packfile();
2696 dump_branches();
2697 dump_tags();
2698 dump_marks();
2699 }
2700 skip_optional_lf();
2701 }
2702
2703 static void parse_progress(void)
2704 {
2705 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2706 fputc('\n', stdout);
2707 fflush(stdout);
2708 skip_optional_lf();
2709 }
2710
2711 static char* make_fast_import_path(const char *path)
2712 {
2713 struct strbuf abs_path = STRBUF_INIT;
2714
2715 if (!relative_marks_paths || is_absolute_path(path))
2716 return xstrdup(path);
2717 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
2718 return strbuf_detach(&abs_path, NULL);
2719 }
2720
2721 static void option_import_marks(const char *marks, int from_stream)
2722 {
2723 if (import_marks_file) {
2724 if (from_stream)
2725 die("Only one import-marks command allowed per stream");
2726
2727 /* read previous mark file */
2728 if(!import_marks_file_from_stream)
2729 read_marks();
2730 }
2731
2732 import_marks_file = make_fast_import_path(marks);
2733 import_marks_file_from_stream = from_stream;
2734 }
2735
2736 static void option_date_format(const char *fmt)
2737 {
2738 if (!strcmp(fmt, "raw"))
2739 whenspec = WHENSPEC_RAW;
2740 else if (!strcmp(fmt, "rfc2822"))
2741 whenspec = WHENSPEC_RFC2822;
2742 else if (!strcmp(fmt, "now"))
2743 whenspec = WHENSPEC_NOW;
2744 else
2745 die("unknown --date-format argument %s", fmt);
2746 }
2747
2748 static void option_depth(const char *depth)
2749 {
2750 max_depth = strtoul(depth, NULL, 0);
2751 if (max_depth > MAX_DEPTH)
2752 die("--depth cannot exceed %u", MAX_DEPTH);
2753 }
2754
2755 static void option_active_branches(const char *branches)
2756 {
2757 max_active_branches = strtoul(branches, NULL, 0);
2758 }
2759
2760 static void option_export_marks(const char *marks)
2761 {
2762 export_marks_file = make_fast_import_path(marks);
2763 }
2764
2765 static void option_export_pack_edges(const char *edges)
2766 {
2767 if (pack_edges)
2768 fclose(pack_edges);
2769 pack_edges = fopen(edges, "a");
2770 if (!pack_edges)
2771 die_errno("Cannot open '%s'", edges);
2772 }
2773
2774 static int parse_one_option(const char *option)
2775 {
2776 if (!prefixcmp(option, "max-pack-size=")) {
2777 unsigned long v;
2778 if (!git_parse_ulong(option + 14, &v))
2779 return 0;
2780 if (v < 8192) {
2781 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
2782 v *= 1024 * 1024;
2783 } else if (v < 1024 * 1024) {
2784 warning("minimum max-pack-size is 1 MiB");
2785 v = 1024 * 1024;
2786 }
2787 max_packsize = v;
2788 } else if (!prefixcmp(option, "big-file-threshold=")) {
2789 unsigned long v;
2790 if (!git_parse_ulong(option + 19, &v))
2791 return 0;
2792 big_file_threshold = v;
2793 } else if (!prefixcmp(option, "depth=")) {
2794 option_depth(option + 6);
2795 } else if (!prefixcmp(option, "active-branches=")) {
2796 option_active_branches(option + 16);
2797 } else if (!prefixcmp(option, "export-pack-edges=")) {
2798 option_export_pack_edges(option + 18);
2799 } else if (!prefixcmp(option, "quiet")) {
2800 show_stats = 0;
2801 } else if (!prefixcmp(option, "stats")) {
2802 show_stats = 1;
2803 } else {
2804 return 0;
2805 }
2806
2807 return 1;
2808 }
2809
2810 static int parse_one_feature(const char *feature, int from_stream)
2811 {
2812 if (!prefixcmp(feature, "date-format=")) {
2813 option_date_format(feature + 12);
2814 } else if (!prefixcmp(feature, "import-marks=")) {
2815 option_import_marks(feature + 13, from_stream);
2816 } else if (!prefixcmp(feature, "export-marks=")) {
2817 option_export_marks(feature + 13);
2818 } else if (!prefixcmp(feature, "relative-marks")) {
2819 relative_marks_paths = 1;
2820 } else if (!prefixcmp(feature, "no-relative-marks")) {
2821 relative_marks_paths = 0;
2822 } else if (!prefixcmp(feature, "force")) {
2823 force_update = 1;
2824 } else {
2825 return 0;
2826 }
2827
2828 return 1;
2829 }
2830
2831 static void parse_feature(void)
2832 {
2833 char *feature = command_buf.buf + 8;
2834
2835 if (seen_data_command)
2836 die("Got feature command '%s' after data command", feature);
2837
2838 if (parse_one_feature(feature, 1))
2839 return;
2840
2841 die("This version of fast-import does not support feature %s.", feature);
2842 }
2843
2844 static void parse_option(void)
2845 {
2846 char *option = command_buf.buf + 11;
2847
2848 if (seen_data_command)
2849 die("Got option command '%s' after data command", option);
2850
2851 if (parse_one_option(option))
2852 return;
2853
2854 die("This version of fast-import does not support option: %s", option);
2855 }
2856
2857 static int git_pack_config(const char *k, const char *v, void *cb)
2858 {
2859 if (!strcmp(k, "pack.depth")) {
2860 max_depth = git_config_int(k, v);
2861 if (max_depth > MAX_DEPTH)
2862 max_depth = MAX_DEPTH;
2863 return 0;
2864 }
2865 if (!strcmp(k, "pack.compression")) {
2866 int level = git_config_int(k, v);
2867 if (level == -1)
2868 level = Z_DEFAULT_COMPRESSION;
2869 else if (level < 0 || level > Z_BEST_COMPRESSION)
2870 die("bad pack compression level %d", level);
2871 pack_compression_level = level;
2872 pack_compression_seen = 1;
2873 return 0;
2874 }
2875 if (!strcmp(k, "pack.indexversion")) {
2876 pack_idx_default_version = git_config_int(k, v);
2877 if (pack_idx_default_version > 2)
2878 die("bad pack.indexversion=%"PRIu32,
2879 pack_idx_default_version);
2880 return 0;
2881 }
2882 if (!strcmp(k, "pack.packsizelimit")) {
2883 max_packsize = git_config_ulong(k, v);
2884 return 0;
2885 }
2886 if (!strcmp(k, "core.bigfilethreshold")) {
2887 long n = git_config_int(k, v);
2888 big_file_threshold = 0 < n ? n : 0;
2889 }
2890 return git_default_config(k, v, cb);
2891 }
2892
2893 static const char fast_import_usage[] =
2894 "git fast-import [--date-format=f] [--max-pack-size=n] [--big-file-threshold=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2895
2896 static void parse_argv(void)
2897 {
2898 unsigned int i;
2899
2900 for (i = 1; i < global_argc; i++) {
2901 const char *a = global_argv[i];
2902
2903 if (*a != '-' || !strcmp(a, "--"))
2904 break;
2905
2906 if (parse_one_option(a + 2))
2907 continue;
2908
2909 if (parse_one_feature(a + 2, 0))
2910 continue;
2911
2912 die("unknown option %s", a);
2913 }
2914 if (i != global_argc)
2915 usage(fast_import_usage);
2916
2917 seen_data_command = 1;
2918 if (import_marks_file)
2919 read_marks();
2920 }
2921
2922 int main(int argc, const char **argv)
2923 {
2924 unsigned int i;
2925
2926 git_extract_argv0_path(argv[0]);
2927
2928 if (argc == 2 && !strcmp(argv[1], "-h"))
2929 usage(fast_import_usage);
2930
2931 setup_git_directory();
2932 git_config(git_pack_config, NULL);
2933 if (!pack_compression_seen && core_compression_seen)
2934 pack_compression_level = core_compression_level;
2935
2936 alloc_objects(object_entry_alloc);
2937 strbuf_init(&command_buf, 0);
2938 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2939 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2940 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2941 marks = pool_calloc(1, sizeof(struct mark_set));
2942
2943 global_argc = argc;
2944 global_argv = argv;
2945
2946 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2947 for (i = 0; i < (cmd_save - 1); i++)
2948 rc_free[i].next = &rc_free[i + 1];
2949 rc_free[cmd_save - 1].next = NULL;
2950
2951 prepare_packed_git();
2952 start_packfile();
2953 set_die_routine(die_nicely);
2954 while (read_next_command() != EOF) {
2955 if (!strcmp("blob", command_buf.buf))
2956 parse_new_blob();
2957 else if (!prefixcmp(command_buf.buf, "commit "))
2958 parse_new_commit();
2959 else if (!prefixcmp(command_buf.buf, "tag "))
2960 parse_new_tag();
2961 else if (!prefixcmp(command_buf.buf, "reset "))
2962 parse_reset_branch();
2963 else if (!strcmp("checkpoint", command_buf.buf))
2964 parse_checkpoint();
2965 else if (!prefixcmp(command_buf.buf, "progress "))
2966 parse_progress();
2967 else if (!prefixcmp(command_buf.buf, "feature "))
2968 parse_feature();
2969 else if (!prefixcmp(command_buf.buf, "option git "))
2970 parse_option();
2971 else if (!prefixcmp(command_buf.buf, "option "))
2972 /* ignore non-git options*/;
2973 else
2974 die("Unsupported command: %s", command_buf.buf);
2975 }
2976
2977 /* argv hasn't been parsed yet, do so */
2978 if (!seen_data_command)
2979 parse_argv();
2980
2981 end_packfile();
2982
2983 dump_branches();
2984 dump_tags();
2985 unkeep_all_packs();
2986 dump_marks();
2987
2988 if (pack_edges)
2989 fclose(pack_edges);
2990
2991 if (show_stats) {
2992 uintmax_t total_count = 0, duplicate_count = 0;
2993 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2994 total_count += object_count_by_type[i];
2995 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2996 duplicate_count += duplicate_count_by_type[i];
2997
2998 fprintf(stderr, "%s statistics:\n", argv[0]);
2999 fprintf(stderr, "---------------------------------------------------------------------\n");
3000 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3001 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3002 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]);
3003 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]);
3004 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]);
3005 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]);
3006 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3007 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3008 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3009 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3010 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3011 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3012 fprintf(stderr, "---------------------------------------------------------------------\n");
3013 pack_report();
3014 fprintf(stderr, "---------------------------------------------------------------------\n");
3015 fprintf(stderr, "\n");
3016 }
3017
3018 return failure ? 1 : 0;
3019 }
git with Jakub Narębski modifications