?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'by/blame-doc-m-c' into maint
[git/raj.git] / fast-import.c
blob309f2c58a2ba431a82a56f757482a5093cc8257d
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 int store_object(
984 enum object_type type,
985 struct strbuf *dat,
986 struct last_object *last,
987 unsigned char *sha1out,
988 uintmax_t mark)
989 {
990 void *out, *delta;
991 struct object_entry *e;
992 unsigned char hdr[96];
993 unsigned char sha1[20];
994 unsigned long hdrlen, deltalen;
995 git_SHA_CTX c;
996 z_stream s;
997
998 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
999 (unsigned long)dat->len) + 1;
1000 git_SHA1_Init(&c);
1001 git_SHA1_Update(&c, hdr, hdrlen);
1002 git_SHA1_Update(&c, dat->buf, dat->len);
1003 git_SHA1_Final(sha1, &c);
1004 if (sha1out)
1005 hashcpy(sha1out, sha1);
1006
1007 e = insert_object(sha1);
1008 if (mark)
1009 insert_mark(mark, e);
1010 if (e->idx.offset) {
1011 duplicate_count_by_type[type]++;
1012 return 1;
1013 } else if (find_sha1_pack(sha1, packed_git)) {
1014 e->type = type;
1015 e->pack_id = MAX_PACK_ID;
1016 e->idx.offset = 1; /* just not zero! */
1017 duplicate_count_by_type[type]++;
1018 return 1;
1019 }
1020
1021 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1022 delta = diff_delta(last->data.buf, last->data.len,
1023 dat->buf, dat->len,
1024 &deltalen, dat->len - 20);
1025 } else
1026 delta = NULL;
1027
1028 memset(&s, 0, sizeof(s));
1029 deflateInit(&s, pack_compression_level);
1030 if (delta) {
1031 s.next_in = delta;
1032 s.avail_in = deltalen;
1033 } else {
1034 s.next_in = (void *)dat->buf;
1035 s.avail_in = dat->len;
1036 }
1037 s.avail_out = deflateBound(&s, s.avail_in);
1038 s.next_out = out = xmalloc(s.avail_out);
1039 while (deflate(&s, Z_FINISH) == Z_OK)
1040 /* nothing */;
1041 deflateEnd(&s);
1042
1043 /* Determine if we should auto-checkpoint. */
1044 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1045 || (pack_size + 60 + s.total_out) < pack_size) {
1046
1047 /* This new object needs to *not* have the current pack_id. */
1048 e->pack_id = pack_id + 1;
1049 cycle_packfile();
1050
1051 /* We cannot carry a delta into the new pack. */
1052 if (delta) {
1053 free(delta);
1054 delta = NULL;
1055
1056 memset(&s, 0, sizeof(s));
1057 deflateInit(&s, pack_compression_level);
1058 s.next_in = (void *)dat->buf;
1059 s.avail_in = dat->len;
1060 s.avail_out = deflateBound(&s, s.avail_in);
1061 s.next_out = out = xrealloc(out, s.avail_out);
1062 while (deflate(&s, Z_FINISH) == Z_OK)
1063 /* nothing */;
1064 deflateEnd(&s);
1065 }
1066 }
1067
1068 e->type = type;
1069 e->pack_id = pack_id;
1070 e->idx.offset = pack_size;
1071 object_count++;
1072 object_count_by_type[type]++;
1073
1074 crc32_begin(pack_file);
1075
1076 if (delta) {
1077 off_t ofs = e->idx.offset - last->offset;
1078 unsigned pos = sizeof(hdr) - 1;
1079
1080 delta_count_by_type[type]++;
1081 e->depth = last->depth + 1;
1082
1083 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1084 sha1write(pack_file, hdr, hdrlen);
1085 pack_size += hdrlen;
1086
1087 hdr[pos] = ofs & 127;
1088 while (ofs >>= 7)
1089 hdr[--pos] = 128 | (--ofs & 127);
1090 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1091 pack_size += sizeof(hdr) - pos;
1092 } else {
1093 e->depth = 0;
1094 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1095 sha1write(pack_file, hdr, hdrlen);
1096 pack_size += hdrlen;
1097 }
1098
1099 sha1write(pack_file, out, s.total_out);
1100 pack_size += s.total_out;
1101
1102 e->idx.crc32 = crc32_end(pack_file);
1103
1104 free(out);
1105 free(delta);
1106 if (last) {
1107 if (last->no_swap) {
1108 last->data = *dat;
1109 } else {
1110 strbuf_swap(&last->data, dat);
1111 }
1112 last->offset = e->idx.offset;
1113 last->depth = e->depth;
1114 }
1115 return 0;
1116 }
1117
1118 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1119 {
1120 if (ftruncate(pack_data->pack_fd, to)
1121 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1122 die_errno("cannot truncate pack to skip duplicate");
1123 pack_size = to;
1124
1125 /* yes this is a layering violation */
1126 pack_file->total = to;
1127 pack_file->offset = 0;
1128 pack_file->ctx = *ctx;
1129 }
1130
1131 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1132 {
1133 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1134 unsigned char *in_buf = xmalloc(in_sz);
1135 unsigned char *out_buf = xmalloc(out_sz);
1136 struct object_entry *e;
1137 unsigned char sha1[20];
1138 unsigned long hdrlen;
1139 off_t offset;
1140 git_SHA_CTX c;
1141 git_SHA_CTX pack_file_ctx;
1142 z_stream s;
1143 int status = Z_OK;
1144
1145 /* Determine if we should auto-checkpoint. */
1146 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1147 || (pack_size + 60 + len) < pack_size)
1148 cycle_packfile();
1149
1150 offset = pack_size;
1151
1152 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1153 sha1flush(pack_file);
1154 pack_file_ctx = pack_file->ctx;
1155
1156 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1157 if (out_sz <= hdrlen)
1158 die("impossibly large object header");
1159
1160 git_SHA1_Init(&c);
1161 git_SHA1_Update(&c, out_buf, hdrlen);
1162
1163 crc32_begin(pack_file);
1164
1165 memset(&s, 0, sizeof(s));
1166 deflateInit(&s, pack_compression_level);
1167
1168 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1169 if (out_sz <= hdrlen)
1170 die("impossibly large object header");
1171
1172 s.next_out = out_buf + hdrlen;
1173 s.avail_out = out_sz - hdrlen;
1174
1175 while (status != Z_STREAM_END) {
1176 if (0 < len && !s.avail_in) {
1177 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1178 size_t n = fread(in_buf, 1, cnt, stdin);
1179 if (!n && feof(stdin))
1180 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1181
1182 git_SHA1_Update(&c, in_buf, n);
1183 s.next_in = in_buf;
1184 s.avail_in = n;
1185 len -= n;
1186 }
1187
1188 status = deflate(&s, len ? 0 : Z_FINISH);
1189
1190 if (!s.avail_out || status == Z_STREAM_END) {
1191 size_t n = s.next_out - out_buf;
1192 sha1write(pack_file, out_buf, n);
1193 pack_size += n;
1194 s.next_out = out_buf;
1195 s.avail_out = out_sz;
1196 }
1197
1198 switch (status) {
1199 case Z_OK:
1200 case Z_BUF_ERROR:
1201 case Z_STREAM_END:
1202 continue;
1203 default:
1204 die("unexpected deflate failure: %d", status);
1205 }
1206 }
1207 deflateEnd(&s);
1208 git_SHA1_Final(sha1, &c);
1209
1210 if (sha1out)
1211 hashcpy(sha1out, sha1);
1212
1213 e = insert_object(sha1);
1214
1215 if (mark)
1216 insert_mark(mark, e);
1217
1218 if (e->idx.offset) {
1219 duplicate_count_by_type[OBJ_BLOB]++;
1220 truncate_pack(offset, &pack_file_ctx);
1221
1222 } else if (find_sha1_pack(sha1, packed_git)) {
1223 e->type = OBJ_BLOB;
1224 e->pack_id = MAX_PACK_ID;
1225 e->idx.offset = 1; /* just not zero! */
1226 duplicate_count_by_type[OBJ_BLOB]++;
1227 truncate_pack(offset, &pack_file_ctx);
1228
1229 } else {
1230 e->depth = 0;
1231 e->type = OBJ_BLOB;
1232 e->pack_id = pack_id;
1233 e->idx.offset = offset;
1234 e->idx.crc32 = crc32_end(pack_file);
1235 object_count++;
1236 object_count_by_type[OBJ_BLOB]++;
1237 }
1238
1239 free(in_buf);
1240 free(out_buf);
1241 }
1242
1243 /* All calls must be guarded by find_object() or find_mark() to
1244 * ensure the 'struct object_entry' passed was written by this
1245 * process instance. We unpack the entry by the offset, avoiding
1246 * the need for the corresponding .idx file. This unpacking rule
1247 * works because we only use OBJ_REF_DELTA within the packfiles
1248 * created by fast-import.
1249 *
1250 * oe must not be NULL. Such an oe usually comes from giving
1251 * an unknown SHA-1 to find_object() or an undefined mark to
1252 * find_mark(). Callers must test for this condition and use
1253 * the standard read_sha1_file() when it happens.
1254 *
1255 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1256 * find_mark(), where the mark was reloaded from an existing marks
1257 * file and is referencing an object that this fast-import process
1258 * instance did not write out to a packfile. Callers must test for
1259 * this condition and use read_sha1_file() instead.
1260 */
1261 static void *gfi_unpack_entry(
1262 struct object_entry *oe,
1263 unsigned long *sizep)
1264 {
1265 enum object_type type;
1266 struct packed_git *p = all_packs[oe->pack_id];
1267 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1268 /* The object is stored in the packfile we are writing to
1269 * and we have modified it since the last time we scanned
1270 * back to read a previously written object. If an old
1271 * window covered [p->pack_size, p->pack_size + 20) its
1272 * data is stale and is not valid. Closing all windows
1273 * and updating the packfile length ensures we can read
1274 * the newly written data.
1275 */
1276 close_pack_windows(p);
1277 sha1flush(pack_file);
1278
1279 /* We have to offer 20 bytes additional on the end of
1280 * the packfile as the core unpacker code assumes the
1281 * footer is present at the file end and must promise
1282 * at least 20 bytes within any window it maps. But
1283 * we don't actually create the footer here.
1284 */
1285 p->pack_size = pack_size + 20;
1286 }
1287 return unpack_entry(p, oe->idx.offset, &type, sizep);
1288 }
1289
1290 static const char *get_mode(const char *str, uint16_t *modep)
1291 {
1292 unsigned char c;
1293 uint16_t mode = 0;
1294
1295 while ((c = *str++) != ' ') {
1296 if (c < '0' || c > '7')
1297 return NULL;
1298 mode = (mode << 3) + (c - '0');
1299 }
1300 *modep = mode;
1301 return str;
1302 }
1303
1304 static void load_tree(struct tree_entry *root)
1305 {
1306 unsigned char *sha1 = root->versions[1].sha1;
1307 struct object_entry *myoe;
1308 struct tree_content *t;
1309 unsigned long size;
1310 char *buf;
1311 const char *c;
1312
1313 root->tree = t = new_tree_content(8);
1314 if (is_null_sha1(sha1))
1315 return;
1316
1317 myoe = find_object(sha1);
1318 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1319 if (myoe->type != OBJ_TREE)
1320 die("Not a tree: %s", sha1_to_hex(sha1));
1321 t->delta_depth = myoe->depth;
1322 buf = gfi_unpack_entry(myoe, &size);
1323 if (!buf)
1324 die("Can't load tree %s", sha1_to_hex(sha1));
1325 } else {
1326 enum object_type type;
1327 buf = read_sha1_file(sha1, &type, &size);
1328 if (!buf || type != OBJ_TREE)
1329 die("Can't load tree %s", sha1_to_hex(sha1));
1330 }
1331
1332 c = buf;
1333 while (c != (buf + size)) {
1334 struct tree_entry *e = new_tree_entry();
1335
1336 if (t->entry_count == t->entry_capacity)
1337 root->tree = t = grow_tree_content(t, t->entry_count);
1338 t->entries[t->entry_count++] = e;
1339
1340 e->tree = NULL;
1341 c = get_mode(c, &e->versions[1].mode);
1342 if (!c)
1343 die("Corrupt mode in %s", sha1_to_hex(sha1));
1344 e->versions[0].mode = e->versions[1].mode;
1345 e->name = to_atom(c, strlen(c));
1346 c += e->name->str_len + 1;
1347 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1348 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1349 c += 20;
1350 }
1351 free(buf);
1352 }
1353
1354 static int tecmp0 (const void *_a, const void *_b)
1355 {
1356 struct tree_entry *a = *((struct tree_entry**)_a);
1357 struct tree_entry *b = *((struct tree_entry**)_b);
1358 return base_name_compare(
1359 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1360 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1361 }
1362
1363 static int tecmp1 (const void *_a, const void *_b)
1364 {
1365 struct tree_entry *a = *((struct tree_entry**)_a);
1366 struct tree_entry *b = *((struct tree_entry**)_b);
1367 return base_name_compare(
1368 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1369 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1370 }
1371
1372 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1373 {
1374 size_t maxlen = 0;
1375 unsigned int i;
1376
1377 if (!v)
1378 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1379 else
1380 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1381
1382 for (i = 0; i < t->entry_count; i++) {
1383 if (t->entries[i]->versions[v].mode)
1384 maxlen += t->entries[i]->name->str_len + 34;
1385 }
1386
1387 strbuf_reset(b);
1388 strbuf_grow(b, maxlen);
1389 for (i = 0; i < t->entry_count; i++) {
1390 struct tree_entry *e = t->entries[i];
1391 if (!e->versions[v].mode)
1392 continue;
1393 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1394 e->name->str_dat, '\0');
1395 strbuf_add(b, e->versions[v].sha1, 20);
1396 }
1397 }
1398
1399 static void store_tree(struct tree_entry *root)
1400 {
1401 struct tree_content *t = root->tree;
1402 unsigned int i, j, del;
1403 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1404 struct object_entry *le;
1405
1406 if (!is_null_sha1(root->versions[1].sha1))
1407 return;
1408
1409 for (i = 0; i < t->entry_count; i++) {
1410 if (t->entries[i]->tree)
1411 store_tree(t->entries[i]);
1412 }
1413
1414 le = find_object(root->versions[0].sha1);
1415 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1416 mktree(t, 0, &old_tree);
1417 lo.data = old_tree;
1418 lo.offset = le->idx.offset;
1419 lo.depth = t->delta_depth;
1420 }
1421
1422 mktree(t, 1, &new_tree);
1423 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1424
1425 t->delta_depth = lo.depth;
1426 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1427 struct tree_entry *e = t->entries[i];
1428 if (e->versions[1].mode) {
1429 e->versions[0].mode = e->versions[1].mode;
1430 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1431 t->entries[j++] = e;
1432 } else {
1433 release_tree_entry(e);
1434 del++;
1435 }
1436 }
1437 t->entry_count -= del;
1438 }
1439
1440 static int tree_content_set(
1441 struct tree_entry *root,
1442 const char *p,
1443 const unsigned char *sha1,
1444 const uint16_t mode,
1445 struct tree_content *subtree)
1446 {
1447 struct tree_content *t = root->tree;
1448 const char *slash1;
1449 unsigned int i, n;
1450 struct tree_entry *e;
1451
1452 slash1 = strchr(p, '/');
1453 if (slash1)
1454 n = slash1 - p;
1455 else
1456 n = strlen(p);
1457 if (!n)
1458 die("Empty path component found in input");
1459 if (!slash1 && !S_ISDIR(mode) && subtree)
1460 die("Non-directories cannot have subtrees");
1461
1462 for (i = 0; i < t->entry_count; i++) {
1463 e = t->entries[i];
1464 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1465 if (!slash1) {
1466 if (!S_ISDIR(mode)
1467 && e->versions[1].mode == mode
1468 && !hashcmp(e->versions[1].sha1, sha1))
1469 return 0;
1470 e->versions[1].mode = mode;
1471 hashcpy(e->versions[1].sha1, sha1);
1472 if (e->tree)
1473 release_tree_content_recursive(e->tree);
1474 e->tree = subtree;
1475 hashclr(root->versions[1].sha1);
1476 return 1;
1477 }
1478 if (!S_ISDIR(e->versions[1].mode)) {
1479 e->tree = new_tree_content(8);
1480 e->versions[1].mode = S_IFDIR;
1481 }
1482 if (!e->tree)
1483 load_tree(e);
1484 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1485 hashclr(root->versions[1].sha1);
1486 return 1;
1487 }
1488 return 0;
1489 }
1490 }
1491
1492 if (t->entry_count == t->entry_capacity)
1493 root->tree = t = grow_tree_content(t, t->entry_count);
1494 e = new_tree_entry();
1495 e->name = to_atom(p, n);
1496 e->versions[0].mode = 0;
1497 hashclr(e->versions[0].sha1);
1498 t->entries[t->entry_count++] = e;
1499 if (slash1) {
1500 e->tree = new_tree_content(8);
1501 e->versions[1].mode = S_IFDIR;
1502 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1503 } else {
1504 e->tree = subtree;
1505 e->versions[1].mode = mode;
1506 hashcpy(e->versions[1].sha1, sha1);
1507 }
1508 hashclr(root->versions[1].sha1);
1509 return 1;
1510 }
1511
1512 static int tree_content_remove(
1513 struct tree_entry *root,
1514 const char *p,
1515 struct tree_entry *backup_leaf)
1516 {
1517 struct tree_content *t = root->tree;
1518 const char *slash1;
1519 unsigned int i, n;
1520 struct tree_entry *e;
1521
1522 slash1 = strchr(p, '/');
1523 if (slash1)
1524 n = slash1 - p;
1525 else
1526 n = strlen(p);
1527
1528 for (i = 0; i < t->entry_count; i++) {
1529 e = t->entries[i];
1530 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1531 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1532 goto del_entry;
1533 if (!e->tree)
1534 load_tree(e);
1535 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1536 for (n = 0; n < e->tree->entry_count; n++) {
1537 if (e->tree->entries[n]->versions[1].mode) {
1538 hashclr(root->versions[1].sha1);
1539 return 1;
1540 }
1541 }
1542 backup_leaf = NULL;
1543 goto del_entry;
1544 }
1545 return 0;
1546 }
1547 }
1548 return 0;
1549
1550 del_entry:
1551 if (backup_leaf)
1552 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1553 else if (e->tree)
1554 release_tree_content_recursive(e->tree);
1555 e->tree = NULL;
1556 e->versions[1].mode = 0;
1557 hashclr(e->versions[1].sha1);
1558 hashclr(root->versions[1].sha1);
1559 return 1;
1560 }
1561
1562 static int tree_content_get(
1563 struct tree_entry *root,
1564 const char *p,
1565 struct tree_entry *leaf)
1566 {
1567 struct tree_content *t = root->tree;
1568 const char *slash1;
1569 unsigned int i, n;
1570 struct tree_entry *e;
1571
1572 slash1 = strchr(p, '/');
1573 if (slash1)
1574 n = slash1 - p;
1575 else
1576 n = strlen(p);
1577
1578 for (i = 0; i < t->entry_count; i++) {
1579 e = t->entries[i];
1580 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1581 if (!slash1) {
1582 memcpy(leaf, e, sizeof(*leaf));
1583 if (e->tree && is_null_sha1(e->versions[1].sha1))
1584 leaf->tree = dup_tree_content(e->tree);
1585 else
1586 leaf->tree = NULL;
1587 return 1;
1588 }
1589 if (!S_ISDIR(e->versions[1].mode))
1590 return 0;
1591 if (!e->tree)
1592 load_tree(e);
1593 return tree_content_get(e, slash1 + 1, leaf);
1594 }
1595 }
1596 return 0;
1597 }
1598
1599 static int update_branch(struct branch *b)
1600 {
1601 static const char *msg = "fast-import";
1602 struct ref_lock *lock;
1603 unsigned char old_sha1[20];
1604
1605 if (is_null_sha1(b->sha1))
1606 return 0;
1607 if (read_ref(b->name, old_sha1))
1608 hashclr(old_sha1);
1609 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1610 if (!lock)
1611 return error("Unable to lock %s", b->name);
1612 if (!force_update && !is_null_sha1(old_sha1)) {
1613 struct commit *old_cmit, *new_cmit;
1614
1615 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1616 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1617 if (!old_cmit || !new_cmit) {
1618 unlock_ref(lock);
1619 return error("Branch %s is missing commits.", b->name);
1620 }
1621
1622 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1623 unlock_ref(lock);
1624 warning("Not updating %s"
1625 " (new tip %s does not contain %s)",
1626 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1627 return -1;
1628 }
1629 }
1630 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1631 return error("Unable to update %s", b->name);
1632 return 0;
1633 }
1634
1635 static void dump_branches(void)
1636 {
1637 unsigned int i;
1638 struct branch *b;
1639
1640 for (i = 0; i < branch_table_sz; i++) {
1641 for (b = branch_table[i]; b; b = b->table_next_branch)
1642 failure |= update_branch(b);
1643 }
1644 }
1645
1646 static void dump_tags(void)
1647 {
1648 static const char *msg = "fast-import";
1649 struct tag *t;
1650 struct ref_lock *lock;
1651 char ref_name[PATH_MAX];
1652
1653 for (t = first_tag; t; t = t->next_tag) {
1654 sprintf(ref_name, "tags/%s", t->name);
1655 lock = lock_ref_sha1(ref_name, NULL);
1656 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1657 failure |= error("Unable to update %s", ref_name);
1658 }
1659 }
1660
1661 static void dump_marks_helper(FILE *f,
1662 uintmax_t base,
1663 struct mark_set *m)
1664 {
1665 uintmax_t k;
1666 if (m->shift) {
1667 for (k = 0; k < 1024; k++) {
1668 if (m->data.sets[k])
1669 dump_marks_helper(f, (base + k) << m->shift,
1670 m->data.sets[k]);
1671 }
1672 } else {
1673 for (k = 0; k < 1024; k++) {
1674 if (m->data.marked[k])
1675 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1676 sha1_to_hex(m->data.marked[k]->idx.sha1));
1677 }
1678 }
1679 }
1680
1681 static void dump_marks(void)
1682 {
1683 static struct lock_file mark_lock;
1684 int mark_fd;
1685 FILE *f;
1686
1687 if (!export_marks_file)
1688 return;
1689
1690 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1691 if (mark_fd < 0) {
1692 failure |= error("Unable to write marks file %s: %s",
1693 export_marks_file, strerror(errno));
1694 return;
1695 }
1696
1697 f = fdopen(mark_fd, "w");
1698 if (!f) {
1699 int saved_errno = errno;
1700 rollback_lock_file(&mark_lock);
1701 failure |= error("Unable to write marks file %s: %s",
1702 export_marks_file, strerror(saved_errno));
1703 return;
1704 }
1705
1706 /*
1707 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1708 * Assign -1 to the lock file descriptor so that commit_lock_file()
1709 * won't try to close() it.
1710 */
1711 mark_lock.fd = -1;
1712
1713 dump_marks_helper(f, 0, marks);
1714 if (ferror(f) || fclose(f)) {
1715 int saved_errno = errno;
1716 rollback_lock_file(&mark_lock);
1717 failure |= error("Unable to write marks file %s: %s",
1718 export_marks_file, strerror(saved_errno));
1719 return;
1720 }
1721
1722 if (commit_lock_file(&mark_lock)) {
1723 int saved_errno = errno;
1724 rollback_lock_file(&mark_lock);
1725 failure |= error("Unable to commit marks file %s: %s",
1726 export_marks_file, strerror(saved_errno));
1727 return;
1728 }
1729 }
1730
1731 static void read_marks(void)
1732 {
1733 char line[512];
1734 FILE *f = fopen(import_marks_file, "r");
1735 if (!f)
1736 die_errno("cannot read '%s'", import_marks_file);
1737 while (fgets(line, sizeof(line), f)) {
1738 uintmax_t mark;
1739 char *end;
1740 unsigned char sha1[20];
1741 struct object_entry *e;
1742
1743 end = strchr(line, '\n');
1744 if (line[0] != ':' || !end)
1745 die("corrupt mark line: %s", line);
1746 *end = 0;
1747 mark = strtoumax(line + 1, &end, 10);
1748 if (!mark || end == line + 1
1749 || *end != ' ' || get_sha1(end + 1, sha1))
1750 die("corrupt mark line: %s", line);
1751 e = find_object(sha1);
1752 if (!e) {
1753 enum object_type type = sha1_object_info(sha1, NULL);
1754 if (type < 0)
1755 die("object not found: %s", sha1_to_hex(sha1));
1756 e = insert_object(sha1);
1757 e->type = type;
1758 e->pack_id = MAX_PACK_ID;
1759 e->idx.offset = 1; /* just not zero! */
1760 }
1761 insert_mark(mark, e);
1762 }
1763 fclose(f);
1764 }
1765
1766
1767 static int read_next_command(void)
1768 {
1769 static int stdin_eof = 0;
1770
1771 if (stdin_eof) {
1772 unread_command_buf = 0;
1773 return EOF;
1774 }
1775
1776 do {
1777 if (unread_command_buf) {
1778 unread_command_buf = 0;
1779 } else {
1780 struct recent_command *rc;
1781
1782 strbuf_detach(&command_buf, NULL);
1783 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1784 if (stdin_eof)
1785 return EOF;
1786
1787 if (!seen_data_command
1788 && prefixcmp(command_buf.buf, "feature ")
1789 && prefixcmp(command_buf.buf, "option ")) {
1790 parse_argv();
1791 }
1792
1793 rc = rc_free;
1794 if (rc)
1795 rc_free = rc->next;
1796 else {
1797 rc = cmd_hist.next;
1798 cmd_hist.next = rc->next;
1799 cmd_hist.next->prev = &cmd_hist;
1800 free(rc->buf);
1801 }
1802
1803 rc->buf = command_buf.buf;
1804 rc->prev = cmd_tail;
1805 rc->next = cmd_hist.prev;
1806 rc->prev->next = rc;
1807 cmd_tail = rc;
1808 }
1809 } while (command_buf.buf[0] == '#');
1810
1811 return 0;
1812 }
1813
1814 static void skip_optional_lf(void)
1815 {
1816 int term_char = fgetc(stdin);
1817 if (term_char != '\n' && term_char != EOF)
1818 ungetc(term_char, stdin);
1819 }
1820
1821 static void parse_mark(void)
1822 {
1823 if (!prefixcmp(command_buf.buf, "mark :")) {
1824 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1825 read_next_command();
1826 }
1827 else
1828 next_mark = 0;
1829 }
1830
1831 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1832 {
1833 strbuf_reset(sb);
1834
1835 if (prefixcmp(command_buf.buf, "data "))
1836 die("Expected 'data n' command, found: %s", command_buf.buf);
1837
1838 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1839 char *term = xstrdup(command_buf.buf + 5 + 2);
1840 size_t term_len = command_buf.len - 5 - 2;
1841
1842 strbuf_detach(&command_buf, NULL);
1843 for (;;) {
1844 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1845 die("EOF in data (terminator '%s' not found)", term);
1846 if (term_len == command_buf.len
1847 && !strcmp(term, command_buf.buf))
1848 break;
1849 strbuf_addbuf(sb, &command_buf);
1850 strbuf_addch(sb, '\n');
1851 }
1852 free(term);
1853 }
1854 else {
1855 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1856 size_t n = 0, length = (size_t)len;
1857
1858 if (limit && limit < len) {
1859 *len_res = len;
1860 return 0;
1861 }
1862 if (length < len)
1863 die("data is too large to use in this context");
1864
1865 while (n < length) {
1866 size_t s = strbuf_fread(sb, length - n, stdin);
1867 if (!s && feof(stdin))
1868 die("EOF in data (%lu bytes remaining)",
1869 (unsigned long)(length - n));
1870 n += s;
1871 }
1872 }
1873
1874 skip_optional_lf();
1875 return 1;
1876 }
1877
1878 static int validate_raw_date(const char *src, char *result, int maxlen)
1879 {
1880 const char *orig_src = src;
1881 char *endp;
1882 unsigned long num;
1883
1884 errno = 0;
1885
1886 num = strtoul(src, &endp, 10);
1887 /* NEEDSWORK: perhaps check for reasonable values? */
1888 if (errno || endp == src || *endp != ' ')
1889 return -1;
1890
1891 src = endp + 1;
1892 if (*src != '-' && *src != '+')
1893 return -1;
1894
1895 num = strtoul(src + 1, &endp, 10);
1896 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1897 1400 < num)
1898 return -1;
1899
1900 strcpy(result, orig_src);
1901 return 0;
1902 }
1903
1904 static char *parse_ident(const char *buf)
1905 {
1906 const char *gt;
1907 size_t name_len;
1908 char *ident;
1909
1910 gt = strrchr(buf, '>');
1911 if (!gt)
1912 die("Missing > in ident string: %s", buf);
1913 gt++;
1914 if (*gt != ' ')
1915 die("Missing space after > in ident string: %s", buf);
1916 gt++;
1917 name_len = gt - buf;
1918 ident = xmalloc(name_len + 24);
1919 strncpy(ident, buf, name_len);
1920
1921 switch (whenspec) {
1922 case WHENSPEC_RAW:
1923 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1924 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1925 break;
1926 case WHENSPEC_RFC2822:
1927 if (parse_date(gt, ident + name_len, 24) < 0)
1928 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1929 break;
1930 case WHENSPEC_NOW:
1931 if (strcmp("now", gt))
1932 die("Date in ident must be 'now': %s", buf);
1933 datestamp(ident + name_len, 24);
1934 break;
1935 }
1936
1937 return ident;
1938 }
1939
1940 static void parse_and_store_blob(
1941 struct last_object *last,
1942 unsigned char *sha1out,
1943 uintmax_t mark)
1944 {
1945 static struct strbuf buf = STRBUF_INIT;
1946 uintmax_t len;
1947
1948 if (parse_data(&buf, big_file_threshold, &len))
1949 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
1950 else {
1951 if (last) {
1952 strbuf_release(&last->data);
1953 last->offset = 0;
1954 last->depth = 0;
1955 }
1956 stream_blob(len, sha1out, mark);
1957 skip_optional_lf();
1958 }
1959 }
1960
1961 static void parse_new_blob(void)
1962 {
1963 read_next_command();
1964 parse_mark();
1965 parse_and_store_blob(&last_blob, NULL, next_mark);
1966 }
1967
1968 static void unload_one_branch(void)
1969 {
1970 while (cur_active_branches
1971 && cur_active_branches >= max_active_branches) {
1972 uintmax_t min_commit = ULONG_MAX;
1973 struct branch *e, *l = NULL, *p = NULL;
1974
1975 for (e = active_branches; e; e = e->active_next_branch) {
1976 if (e->last_commit < min_commit) {
1977 p = l;
1978 min_commit = e->last_commit;
1979 }
1980 l = e;
1981 }
1982
1983 if (p) {
1984 e = p->active_next_branch;
1985 p->active_next_branch = e->active_next_branch;
1986 } else {
1987 e = active_branches;
1988 active_branches = e->active_next_branch;
1989 }
1990 e->active = 0;
1991 e->active_next_branch = NULL;
1992 if (e->branch_tree.tree) {
1993 release_tree_content_recursive(e->branch_tree.tree);
1994 e->branch_tree.tree = NULL;
1995 }
1996 cur_active_branches--;
1997 }
1998 }
1999
2000 static void load_branch(struct branch *b)
2001 {
2002 load_tree(&b->branch_tree);
2003 if (!b->active) {
2004 b->active = 1;
2005 b->active_next_branch = active_branches;
2006 active_branches = b;
2007 cur_active_branches++;
2008 branch_load_count++;
2009 }
2010 }
2011
2012 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2013 {
2014 unsigned char fanout = 0;
2015 while ((num_notes >>= 8))
2016 fanout++;
2017 return fanout;
2018 }
2019
2020 static void construct_path_with_fanout(const char *hex_sha1,
2021 unsigned char fanout, char *path)
2022 {
2023 unsigned int i = 0, j = 0;
2024 if (fanout >= 20)
2025 die("Too large fanout (%u)", fanout);
2026 while (fanout) {
2027 path[i++] = hex_sha1[j++];
2028 path[i++] = hex_sha1[j++];
2029 path[i++] = '/';
2030 fanout--;
2031 }
2032 memcpy(path + i, hex_sha1 + j, 40 - j);
2033 path[i + 40 - j] = '\0';
2034 }
2035
2036 static uintmax_t do_change_note_fanout(
2037 struct tree_entry *orig_root, struct tree_entry *root,
2038 char *hex_sha1, unsigned int hex_sha1_len,
2039 char *fullpath, unsigned int fullpath_len,
2040 unsigned char fanout)
2041 {
2042 struct tree_content *t = root->tree;
2043 struct tree_entry *e, leaf;
2044 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2045 uintmax_t num_notes = 0;
2046 unsigned char sha1[20];
2047 char realpath[60];
2048
2049 for (i = 0; t && i < t->entry_count; i++) {
2050 e = t->entries[i];
2051 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2052 tmp_fullpath_len = fullpath_len;
2053
2054 /*
2055 * We're interested in EITHER existing note entries (entries
2056 * with exactly 40 hex chars in path, not including directory
2057 * separators), OR directory entries that may contain note
2058 * entries (with < 40 hex chars in path).
2059 * Also, each path component in a note entry must be a multiple
2060 * of 2 chars.
2061 */
2062 if (!e->versions[1].mode ||
2063 tmp_hex_sha1_len > 40 ||
2064 e->name->str_len % 2)
2065 continue;
2066
2067 /* This _may_ be a note entry, or a subdir containing notes */
2068 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2069 e->name->str_len);
2070 if (tmp_fullpath_len)
2071 fullpath[tmp_fullpath_len++] = '/';
2072 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2073 e->name->str_len);
2074 tmp_fullpath_len += e->name->str_len;
2075 fullpath[tmp_fullpath_len] = '\0';
2076
2077 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2078 /* This is a note entry */
2079 construct_path_with_fanout(hex_sha1, fanout, realpath);
2080 if (!strcmp(fullpath, realpath)) {
2081 /* Note entry is in correct location */
2082 num_notes++;
2083 continue;
2084 }
2085
2086 /* Rename fullpath to realpath */
2087 if (!tree_content_remove(orig_root, fullpath, &leaf))
2088 die("Failed to remove path %s", fullpath);
2089 tree_content_set(orig_root, realpath,
2090 leaf.versions[1].sha1,
2091 leaf.versions[1].mode,
2092 leaf.tree);
2093 } else if (S_ISDIR(e->versions[1].mode)) {
2094 /* This is a subdir that may contain note entries */
2095 if (!e->tree)
2096 load_tree(e);
2097 num_notes += do_change_note_fanout(orig_root, e,
2098 hex_sha1, tmp_hex_sha1_len,
2099 fullpath, tmp_fullpath_len, fanout);
2100 }
2101
2102 /* The above may have reallocated the current tree_content */
2103 t = root->tree;
2104 }
2105 return num_notes;
2106 }
2107
2108 static uintmax_t change_note_fanout(struct tree_entry *root,
2109 unsigned char fanout)
2110 {
2111 char hex_sha1[40], path[60];
2112 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2113 }
2114
2115 static void file_change_m(struct branch *b)
2116 {
2117 const char *p = command_buf.buf + 2;
2118 static struct strbuf uq = STRBUF_INIT;
2119 const char *endp;
2120 struct object_entry *oe = oe;
2121 unsigned char sha1[20];
2122 uint16_t mode, inline_data = 0;
2123
2124 p = get_mode(p, &mode);
2125 if (!p)
2126 die("Corrupt mode: %s", command_buf.buf);
2127 switch (mode) {
2128 case 0644:
2129 case 0755:
2130 mode |= S_IFREG;
2131 case S_IFREG | 0644:
2132 case S_IFREG | 0755:
2133 case S_IFLNK:
2134 case S_IFGITLINK:
2135 /* ok */
2136 break;
2137 default:
2138 die("Corrupt mode: %s", command_buf.buf);
2139 }
2140
2141 if (*p == ':') {
2142 char *x;
2143 oe = find_mark(strtoumax(p + 1, &x, 10));
2144 hashcpy(sha1, oe->idx.sha1);
2145 p = x;
2146 } else if (!prefixcmp(p, "inline")) {
2147 inline_data = 1;
2148 p += 6;
2149 } else {
2150 if (get_sha1_hex(p, sha1))
2151 die("Invalid SHA1: %s", command_buf.buf);
2152 oe = find_object(sha1);
2153 p += 40;
2154 }
2155 if (*p++ != ' ')
2156 die("Missing space after SHA1: %s", command_buf.buf);
2157
2158 strbuf_reset(&uq);
2159 if (!unquote_c_style(&uq, p, &endp)) {
2160 if (*endp)
2161 die("Garbage after path in: %s", command_buf.buf);
2162 p = uq.buf;
2163 }
2164
2165 if (S_ISGITLINK(mode)) {
2166 if (inline_data)
2167 die("Git links cannot be specified 'inline': %s",
2168 command_buf.buf);
2169 else if (oe) {
2170 if (oe->type != OBJ_COMMIT)
2171 die("Not a commit (actually a %s): %s",
2172 typename(oe->type), command_buf.buf);
2173 }
2174 /*
2175 * Accept the sha1 without checking; it expected to be in
2176 * another repository.
2177 */
2178 } else if (inline_data) {
2179 if (p != uq.buf) {
2180 strbuf_addstr(&uq, p);
2181 p = uq.buf;
2182 }
2183 read_next_command();
2184 parse_and_store_blob(&last_blob, sha1, 0);
2185 } else if (oe) {
2186 if (oe->type != OBJ_BLOB)
2187 die("Not a blob (actually a %s): %s",
2188 typename(oe->type), command_buf.buf);
2189 } else {
2190 enum object_type type = sha1_object_info(sha1, NULL);
2191 if (type < 0)
2192 die("Blob not found: %s", command_buf.buf);
2193 if (type != OBJ_BLOB)
2194 die("Not a blob (actually a %s): %s",
2195 typename(type), command_buf.buf);
2196 }
2197
2198 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2199 }
2200
2201 static void file_change_d(struct branch *b)
2202 {
2203 const char *p = command_buf.buf + 2;
2204 static struct strbuf uq = STRBUF_INIT;
2205 const char *endp;
2206
2207 strbuf_reset(&uq);
2208 if (!unquote_c_style(&uq, p, &endp)) {
2209 if (*endp)
2210 die("Garbage after path in: %s", command_buf.buf);
2211 p = uq.buf;
2212 }
2213 tree_content_remove(&b->branch_tree, p, NULL);
2214 }
2215
2216 static void file_change_cr(struct branch *b, int rename)
2217 {
2218 const char *s, *d;
2219 static struct strbuf s_uq = STRBUF_INIT;
2220 static struct strbuf d_uq = STRBUF_INIT;
2221 const char *endp;
2222 struct tree_entry leaf;
2223
2224 s = command_buf.buf + 2;
2225 strbuf_reset(&s_uq);
2226 if (!unquote_c_style(&s_uq, s, &endp)) {
2227 if (*endp != ' ')
2228 die("Missing space after source: %s", command_buf.buf);
2229 } else {
2230 endp = strchr(s, ' ');
2231 if (!endp)
2232 die("Missing space after source: %s", command_buf.buf);
2233 strbuf_add(&s_uq, s, endp - s);
2234 }
2235 s = s_uq.buf;
2236
2237 endp++;
2238 if (!*endp)
2239 die("Missing dest: %s", command_buf.buf);
2240
2241 d = endp;
2242 strbuf_reset(&d_uq);
2243 if (!unquote_c_style(&d_uq, d, &endp)) {
2244 if (*endp)
2245 die("Garbage after dest in: %s", command_buf.buf);
2246 d = d_uq.buf;
2247 }
2248
2249 memset(&leaf, 0, sizeof(leaf));
2250 if (rename)
2251 tree_content_remove(&b->branch_tree, s, &leaf);
2252 else
2253 tree_content_get(&b->branch_tree, s, &leaf);
2254 if (!leaf.versions[1].mode)
2255 die("Path %s not in branch", s);
2256 tree_content_set(&b->branch_tree, d,
2257 leaf.versions[1].sha1,
2258 leaf.versions[1].mode,
2259 leaf.tree);
2260 }
2261
2262 static void note_change_n(struct branch *b, unsigned char old_fanout)
2263 {
2264 const char *p = command_buf.buf + 2;
2265 static struct strbuf uq = STRBUF_INIT;
2266 struct object_entry *oe = oe;
2267 struct branch *s;
2268 unsigned char sha1[20], commit_sha1[20];
2269 char path[60];
2270 uint16_t inline_data = 0;
2271 unsigned char new_fanout;
2272
2273 /* <dataref> or 'inline' */
2274 if (*p == ':') {
2275 char *x;
2276 oe = find_mark(strtoumax(p + 1, &x, 10));
2277 hashcpy(sha1, oe->idx.sha1);
2278 p = x;
2279 } else if (!prefixcmp(p, "inline")) {
2280 inline_data = 1;
2281 p += 6;
2282 } else {
2283 if (get_sha1_hex(p, sha1))
2284 die("Invalid SHA1: %s", command_buf.buf);
2285 oe = find_object(sha1);
2286 p += 40;
2287 }
2288 if (*p++ != ' ')
2289 die("Missing space after SHA1: %s", command_buf.buf);
2290
2291 /* <committish> */
2292 s = lookup_branch(p);
2293 if (s) {
2294 hashcpy(commit_sha1, s->sha1);
2295 } else if (*p == ':') {
2296 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2297 struct object_entry *commit_oe = find_mark(commit_mark);
2298 if (commit_oe->type != OBJ_COMMIT)
2299 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2300 hashcpy(commit_sha1, commit_oe->idx.sha1);
2301 } else if (!get_sha1(p, commit_sha1)) {
2302 unsigned long size;
2303 char *buf = read_object_with_reference(commit_sha1,
2304 commit_type, &size, commit_sha1);
2305 if (!buf || size < 46)
2306 die("Not a valid commit: %s", p);
2307 free(buf);
2308 } else
2309 die("Invalid ref name or SHA1 expression: %s", p);
2310
2311 if (inline_data) {
2312 if (p != uq.buf) {
2313 strbuf_addstr(&uq, p);
2314 p = uq.buf;
2315 }
2316 read_next_command();
2317 parse_and_store_blob(&last_blob, sha1, 0);
2318 } else if (oe) {
2319 if (oe->type != OBJ_BLOB)
2320 die("Not a blob (actually a %s): %s",
2321 typename(oe->type), command_buf.buf);
2322 } else if (!is_null_sha1(sha1)) {
2323 enum object_type type = sha1_object_info(sha1, NULL);
2324 if (type < 0)
2325 die("Blob not found: %s", command_buf.buf);
2326 if (type != OBJ_BLOB)
2327 die("Not a blob (actually a %s): %s",
2328 typename(type), command_buf.buf);
2329 }
2330
2331 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2332 if (tree_content_remove(&b->branch_tree, path, NULL))
2333 b->num_notes--;
2334
2335 if (is_null_sha1(sha1))
2336 return; /* nothing to insert */
2337
2338 b->num_notes++;
2339 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2340 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2341 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2342 }
2343
2344 static void file_change_deleteall(struct branch *b)
2345 {
2346 release_tree_content_recursive(b->branch_tree.tree);
2347 hashclr(b->branch_tree.versions[0].sha1);
2348 hashclr(b->branch_tree.versions[1].sha1);
2349 load_tree(&b->branch_tree);
2350 b->num_notes = 0;
2351 }
2352
2353 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2354 {
2355 if (!buf || size < 46)
2356 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2357 if (memcmp("tree ", buf, 5)
2358 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2359 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2360 hashcpy(b->branch_tree.versions[0].sha1,
2361 b->branch_tree.versions[1].sha1);
2362 }
2363
2364 static void parse_from_existing(struct branch *b)
2365 {
2366 if (is_null_sha1(b->sha1)) {
2367 hashclr(b->branch_tree.versions[0].sha1);
2368 hashclr(b->branch_tree.versions[1].sha1);
2369 } else {
2370 unsigned long size;
2371 char *buf;
2372
2373 buf = read_object_with_reference(b->sha1,
2374 commit_type, &size, b->sha1);
2375 parse_from_commit(b, buf, size);
2376 free(buf);
2377 }
2378 }
2379
2380 static int parse_from(struct branch *b)
2381 {
2382 const char *from;
2383 struct branch *s;
2384
2385 if (prefixcmp(command_buf.buf, "from "))
2386 return 0;
2387
2388 if (b->branch_tree.tree) {
2389 release_tree_content_recursive(b->branch_tree.tree);
2390 b->branch_tree.tree = NULL;
2391 }
2392
2393 from = strchr(command_buf.buf, ' ') + 1;
2394 s = lookup_branch(from);
2395 if (b == s)
2396 die("Can't create a branch from itself: %s", b->name);
2397 else if (s) {
2398 unsigned char *t = s->branch_tree.versions[1].sha1;
2399 hashcpy(b->sha1, s->sha1);
2400 hashcpy(b->branch_tree.versions[0].sha1, t);
2401 hashcpy(b->branch_tree.versions[1].sha1, t);
2402 } else if (*from == ':') {
2403 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2404 struct object_entry *oe = find_mark(idnum);
2405 if (oe->type != OBJ_COMMIT)
2406 die("Mark :%" PRIuMAX " not a commit", idnum);
2407 hashcpy(b->sha1, oe->idx.sha1);
2408 if (oe->pack_id != MAX_PACK_ID) {
2409 unsigned long size;
2410 char *buf = gfi_unpack_entry(oe, &size);
2411 parse_from_commit(b, buf, size);
2412 free(buf);
2413 } else
2414 parse_from_existing(b);
2415 } else if (!get_sha1(from, b->sha1))
2416 parse_from_existing(b);
2417 else
2418 die("Invalid ref name or SHA1 expression: %s", from);
2419
2420 read_next_command();
2421 return 1;
2422 }
2423
2424 static struct hash_list *parse_merge(unsigned int *count)
2425 {
2426 struct hash_list *list = NULL, *n, *e = e;
2427 const char *from;
2428 struct branch *s;
2429
2430 *count = 0;
2431 while (!prefixcmp(command_buf.buf, "merge ")) {
2432 from = strchr(command_buf.buf, ' ') + 1;
2433 n = xmalloc(sizeof(*n));
2434 s = lookup_branch(from);
2435 if (s)
2436 hashcpy(n->sha1, s->sha1);
2437 else if (*from == ':') {
2438 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2439 struct object_entry *oe = find_mark(idnum);
2440 if (oe->type != OBJ_COMMIT)
2441 die("Mark :%" PRIuMAX " not a commit", idnum);
2442 hashcpy(n->sha1, oe->idx.sha1);
2443 } else if (!get_sha1(from, n->sha1)) {
2444 unsigned long size;
2445 char *buf = read_object_with_reference(n->sha1,
2446 commit_type, &size, n->sha1);
2447 if (!buf || size < 46)
2448 die("Not a valid commit: %s", from);
2449 free(buf);
2450 } else
2451 die("Invalid ref name or SHA1 expression: %s", from);
2452
2453 n->next = NULL;
2454 if (list)
2455 e->next = n;
2456 else
2457 list = n;
2458 e = n;
2459 (*count)++;
2460 read_next_command();
2461 }
2462 return list;
2463 }
2464
2465 static void parse_new_commit(void)
2466 {
2467 static struct strbuf msg = STRBUF_INIT;
2468 struct branch *b;
2469 char *sp;
2470 char *author = NULL;
2471 char *committer = NULL;
2472 struct hash_list *merge_list = NULL;
2473 unsigned int merge_count;
2474 unsigned char prev_fanout, new_fanout;
2475
2476 /* Obtain the branch name from the rest of our command */
2477 sp = strchr(command_buf.buf, ' ') + 1;
2478 b = lookup_branch(sp);
2479 if (!b)
2480 b = new_branch(sp);
2481
2482 read_next_command();
2483 parse_mark();
2484 if (!prefixcmp(command_buf.buf, "author ")) {
2485 author = parse_ident(command_buf.buf + 7);
2486 read_next_command();
2487 }
2488 if (!prefixcmp(command_buf.buf, "committer ")) {
2489 committer = parse_ident(command_buf.buf + 10);
2490 read_next_command();
2491 }
2492 if (!committer)
2493 die("Expected committer but didn't get one");
2494 parse_data(&msg, 0, NULL);
2495 read_next_command();
2496 parse_from(b);
2497 merge_list = parse_merge(&merge_count);
2498
2499 /* ensure the branch is active/loaded */
2500 if (!b->branch_tree.tree || !max_active_branches) {
2501 unload_one_branch();
2502 load_branch(b);
2503 }
2504
2505 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2506
2507 /* file_change* */
2508 while (command_buf.len > 0) {
2509 if (!prefixcmp(command_buf.buf, "M "))
2510 file_change_m(b);
2511 else if (!prefixcmp(command_buf.buf, "D "))
2512 file_change_d(b);
2513 else if (!prefixcmp(command_buf.buf, "R "))
2514 file_change_cr(b, 1);
2515 else if (!prefixcmp(command_buf.buf, "C "))
2516 file_change_cr(b, 0);
2517 else if (!prefixcmp(command_buf.buf, "N "))
2518 note_change_n(b, prev_fanout);
2519 else if (!strcmp("deleteall", command_buf.buf))
2520 file_change_deleteall(b);
2521 else {
2522 unread_command_buf = 1;
2523 break;
2524 }
2525 if (read_next_command() == EOF)
2526 break;
2527 }
2528
2529 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2530 if (new_fanout != prev_fanout)
2531 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2532
2533 /* build the tree and the commit */
2534 store_tree(&b->branch_tree);
2535 hashcpy(b->branch_tree.versions[0].sha1,
2536 b->branch_tree.versions[1].sha1);
2537
2538 strbuf_reset(&new_data);
2539 strbuf_addf(&new_data, "tree %s\n",
2540 sha1_to_hex(b->branch_tree.versions[1].sha1));
2541 if (!is_null_sha1(b->sha1))
2542 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2543 while (merge_list) {
2544 struct hash_list *next = merge_list->next;
2545 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2546 free(merge_list);
2547 merge_list = next;
2548 }
2549 strbuf_addf(&new_data,
2550 "author %s\n"
2551 "committer %s\n"
2552 "\n",
2553 author ? author : committer, committer);
2554 strbuf_addbuf(&new_data, &msg);
2555 free(author);
2556 free(committer);
2557
2558 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2559 b->pack_id = pack_id;
2560 b->last_commit = object_count_by_type[OBJ_COMMIT];
2561 }
2562
2563 static void parse_new_tag(void)
2564 {
2565 static struct strbuf msg = STRBUF_INIT;
2566 char *sp;
2567 const char *from;
2568 char *tagger;
2569 struct branch *s;
2570 struct tag *t;
2571 uintmax_t from_mark = 0;
2572 unsigned char sha1[20];
2573 enum object_type type;
2574
2575 /* Obtain the new tag name from the rest of our command */
2576 sp = strchr(command_buf.buf, ' ') + 1;
2577 t = pool_alloc(sizeof(struct tag));
2578 t->next_tag = NULL;
2579 t->name = pool_strdup(sp);
2580 if (last_tag)
2581 last_tag->next_tag = t;
2582 else
2583 first_tag = t;
2584 last_tag = t;
2585 read_next_command();
2586
2587 /* from ... */
2588 if (prefixcmp(command_buf.buf, "from "))
2589 die("Expected from command, got %s", command_buf.buf);
2590 from = strchr(command_buf.buf, ' ') + 1;
2591 s = lookup_branch(from);
2592 if (s) {
2593 hashcpy(sha1, s->sha1);
2594 type = OBJ_COMMIT;
2595 } else if (*from == ':') {
2596 struct object_entry *oe;
2597 from_mark = strtoumax(from + 1, NULL, 10);
2598 oe = find_mark(from_mark);
2599 type = oe->type;
2600 hashcpy(sha1, oe->idx.sha1);
2601 } else if (!get_sha1(from, sha1)) {
2602 unsigned long size;
2603 char *buf;
2604
2605 buf = read_sha1_file(sha1, &type, &size);
2606 if (!buf || size < 46)
2607 die("Not a valid commit: %s", from);
2608 free(buf);
2609 } else
2610 die("Invalid ref name or SHA1 expression: %s", from);
2611 read_next_command();
2612
2613 /* tagger ... */
2614 if (!prefixcmp(command_buf.buf, "tagger ")) {
2615 tagger = parse_ident(command_buf.buf + 7);
2616 read_next_command();
2617 } else
2618 tagger = NULL;
2619
2620 /* tag payload/message */
2621 parse_data(&msg, 0, NULL);
2622
2623 /* build the tag object */
2624 strbuf_reset(&new_data);
2625
2626 strbuf_addf(&new_data,
2627 "object %s\n"
2628 "type %s\n"
2629 "tag %s\n",
2630 sha1_to_hex(sha1), typename(type), t->name);
2631 if (tagger)
2632 strbuf_addf(&new_data,
2633 "tagger %s\n", tagger);
2634 strbuf_addch(&new_data, '\n');
2635 strbuf_addbuf(&new_data, &msg);
2636 free(tagger);
2637
2638 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2639 t->pack_id = MAX_PACK_ID;
2640 else
2641 t->pack_id = pack_id;
2642 }
2643
2644 static void parse_reset_branch(void)
2645 {
2646 struct branch *b;
2647 char *sp;
2648
2649 /* Obtain the branch name from the rest of our command */
2650 sp = strchr(command_buf.buf, ' ') + 1;
2651 b = lookup_branch(sp);
2652 if (b) {
2653 hashclr(b->sha1);
2654 hashclr(b->branch_tree.versions[0].sha1);
2655 hashclr(b->branch_tree.versions[1].sha1);
2656 if (b->branch_tree.tree) {
2657 release_tree_content_recursive(b->branch_tree.tree);
2658 b->branch_tree.tree = NULL;
2659 }
2660 }
2661 else
2662 b = new_branch(sp);
2663 read_next_command();
2664 parse_from(b);
2665 if (command_buf.len > 0)
2666 unread_command_buf = 1;
2667 }
2668
2669 static void parse_checkpoint(void)
2670 {
2671 if (object_count) {
2672 cycle_packfile();
2673 dump_branches();
2674 dump_tags();
2675 dump_marks();
2676 }
2677 skip_optional_lf();
2678 }
2679
2680 static void parse_progress(void)
2681 {
2682 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2683 fputc('\n', stdout);
2684 fflush(stdout);
2685 skip_optional_lf();
2686 }
2687
2688 static char* make_fast_import_path(const char *path)
2689 {
2690 struct strbuf abs_path = STRBUF_INIT;
2691
2692 if (!relative_marks_paths || is_absolute_path(path))
2693 return xstrdup(path);
2694 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
2695 return strbuf_detach(&abs_path, NULL);
2696 }
2697
2698 static void option_import_marks(const char *marks, int from_stream)
2699 {
2700 if (import_marks_file) {
2701 if (from_stream)
2702 die("Only one import-marks command allowed per stream");
2703
2704 /* read previous mark file */
2705 if(!import_marks_file_from_stream)
2706 read_marks();
2707 }
2708
2709 import_marks_file = make_fast_import_path(marks);
2710 import_marks_file_from_stream = from_stream;
2711 }
2712
2713 static void option_date_format(const char *fmt)
2714 {
2715 if (!strcmp(fmt, "raw"))
2716 whenspec = WHENSPEC_RAW;
2717 else if (!strcmp(fmt, "rfc2822"))
2718 whenspec = WHENSPEC_RFC2822;
2719 else if (!strcmp(fmt, "now"))
2720 whenspec = WHENSPEC_NOW;
2721 else
2722 die("unknown --date-format argument %s", fmt);
2723 }
2724
2725 static void option_depth(const char *depth)
2726 {
2727 max_depth = strtoul(depth, NULL, 0);
2728 if (max_depth > MAX_DEPTH)
2729 die("--depth cannot exceed %u", MAX_DEPTH);
2730 }
2731
2732 static void option_active_branches(const char *branches)
2733 {
2734 max_active_branches = strtoul(branches, NULL, 0);
2735 }
2736
2737 static void option_export_marks(const char *marks)
2738 {
2739 export_marks_file = make_fast_import_path(marks);
2740 }
2741
2742 static void option_export_pack_edges(const char *edges)
2743 {
2744 if (pack_edges)
2745 fclose(pack_edges);
2746 pack_edges = fopen(edges, "a");
2747 if (!pack_edges)
2748 die_errno("Cannot open '%s'", edges);
2749 }
2750
2751 static int parse_one_option(const char *option)
2752 {
2753 if (!prefixcmp(option, "max-pack-size=")) {
2754 unsigned long v;
2755 if (!git_parse_ulong(option + 14, &v))
2756 return 0;
2757 if (v < 8192) {
2758 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
2759 v *= 1024 * 1024;
2760 } else if (v < 1024 * 1024) {
2761 warning("minimum max-pack-size is 1 MiB");
2762 v = 1024 * 1024;
2763 }
2764 max_packsize = v;
2765 } else if (!prefixcmp(option, "big-file-threshold=")) {
2766 unsigned long v;
2767 if (!git_parse_ulong(option + 19, &v))
2768 return 0;
2769 big_file_threshold = v;
2770 } else if (!prefixcmp(option, "depth=")) {
2771 option_depth(option + 6);
2772 } else if (!prefixcmp(option, "active-branches=")) {
2773 option_active_branches(option + 16);
2774 } else if (!prefixcmp(option, "export-pack-edges=")) {
2775 option_export_pack_edges(option + 18);
2776 } else if (!prefixcmp(option, "quiet")) {
2777 show_stats = 0;
2778 } else if (!prefixcmp(option, "stats")) {
2779 show_stats = 1;
2780 } else {
2781 return 0;
2782 }
2783
2784 return 1;
2785 }
2786
2787 static int parse_one_feature(const char *feature, int from_stream)
2788 {
2789 if (!prefixcmp(feature, "date-format=")) {
2790 option_date_format(feature + 12);
2791 } else if (!prefixcmp(feature, "import-marks=")) {
2792 option_import_marks(feature + 13, from_stream);
2793 } else if (!prefixcmp(feature, "export-marks=")) {
2794 option_export_marks(feature + 13);
2795 } else if (!prefixcmp(feature, "relative-marks")) {
2796 relative_marks_paths = 1;
2797 } else if (!prefixcmp(feature, "no-relative-marks")) {
2798 relative_marks_paths = 0;
2799 } else if (!prefixcmp(feature, "force")) {
2800 force_update = 1;
2801 } else {
2802 return 0;
2803 }
2804
2805 return 1;
2806 }
2807
2808 static void parse_feature(void)
2809 {
2810 char *feature = command_buf.buf + 8;
2811
2812 if (seen_data_command)
2813 die("Got feature command '%s' after data command", feature);
2814
2815 if (parse_one_feature(feature, 1))
2816 return;
2817
2818 die("This version of fast-import does not support feature %s.", feature);
2819 }
2820
2821 static void parse_option(void)
2822 {
2823 char *option = command_buf.buf + 11;
2824
2825 if (seen_data_command)
2826 die("Got option command '%s' after data command", option);
2827
2828 if (parse_one_option(option))
2829 return;
2830
2831 die("This version of fast-import does not support option: %s", option);
2832 }
2833
2834 static int git_pack_config(const char *k, const char *v, void *cb)
2835 {
2836 if (!strcmp(k, "pack.depth")) {
2837 max_depth = git_config_int(k, v);
2838 if (max_depth > MAX_DEPTH)
2839 max_depth = MAX_DEPTH;
2840 return 0;
2841 }
2842 if (!strcmp(k, "pack.compression")) {
2843 int level = git_config_int(k, v);
2844 if (level == -1)
2845 level = Z_DEFAULT_COMPRESSION;
2846 else if (level < 0 || level > Z_BEST_COMPRESSION)
2847 die("bad pack compression level %d", level);
2848 pack_compression_level = level;
2849 pack_compression_seen = 1;
2850 return 0;
2851 }
2852 if (!strcmp(k, "pack.indexversion")) {
2853 pack_idx_default_version = git_config_int(k, v);
2854 if (pack_idx_default_version > 2)
2855 die("bad pack.indexversion=%"PRIu32,
2856 pack_idx_default_version);
2857 return 0;
2858 }
2859 if (!strcmp(k, "pack.packsizelimit")) {
2860 max_packsize = git_config_ulong(k, v);
2861 return 0;
2862 }
2863 if (!strcmp(k, "core.bigfilethreshold")) {
2864 long n = git_config_int(k, v);
2865 big_file_threshold = 0 < n ? n : 0;
2866 }
2867 return git_default_config(k, v, cb);
2868 }
2869
2870 static const char fast_import_usage[] =
2871 "git fast-import [--date-format=f] [--max-pack-size=n] [--big-file-threshold=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2872
2873 static void parse_argv(void)
2874 {
2875 unsigned int i;
2876
2877 for (i = 1; i < global_argc; i++) {
2878 const char *a = global_argv[i];
2879
2880 if (*a != '-' || !strcmp(a, "--"))
2881 break;
2882
2883 if (parse_one_option(a + 2))
2884 continue;
2885
2886 if (parse_one_feature(a + 2, 0))
2887 continue;
2888
2889 die("unknown option %s", a);
2890 }
2891 if (i != global_argc)
2892 usage(fast_import_usage);
2893
2894 seen_data_command = 1;
2895 if (import_marks_file)
2896 read_marks();
2897 }
2898
2899 int main(int argc, const char **argv)
2900 {
2901 unsigned int i;
2902
2903 git_extract_argv0_path(argv[0]);
2904
2905 if (argc == 2 && !strcmp(argv[1], "-h"))
2906 usage(fast_import_usage);
2907
2908 setup_git_directory();
2909 git_config(git_pack_config, NULL);
2910 if (!pack_compression_seen && core_compression_seen)
2911 pack_compression_level = core_compression_level;
2912
2913 alloc_objects(object_entry_alloc);
2914 strbuf_init(&command_buf, 0);
2915 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2916 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2917 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2918 marks = pool_calloc(1, sizeof(struct mark_set));
2919
2920 global_argc = argc;
2921 global_argv = argv;
2922
2923 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2924 for (i = 0; i < (cmd_save - 1); i++)
2925 rc_free[i].next = &rc_free[i + 1];
2926 rc_free[cmd_save - 1].next = NULL;
2927
2928 prepare_packed_git();
2929 start_packfile();
2930 set_die_routine(die_nicely);
2931 while (read_next_command() != EOF) {
2932 if (!strcmp("blob", command_buf.buf))
2933 parse_new_blob();
2934 else if (!prefixcmp(command_buf.buf, "commit "))
2935 parse_new_commit();
2936 else if (!prefixcmp(command_buf.buf, "tag "))
2937 parse_new_tag();
2938 else if (!prefixcmp(command_buf.buf, "reset "))
2939 parse_reset_branch();
2940 else if (!strcmp("checkpoint", command_buf.buf))
2941 parse_checkpoint();
2942 else if (!prefixcmp(command_buf.buf, "progress "))
2943 parse_progress();
2944 else if (!prefixcmp(command_buf.buf, "feature "))
2945 parse_feature();
2946 else if (!prefixcmp(command_buf.buf, "option git "))
2947 parse_option();
2948 else if (!prefixcmp(command_buf.buf, "option "))
2949 /* ignore non-git options*/;
2950 else
2951 die("Unsupported command: %s", command_buf.buf);
2952 }
2953
2954 /* argv hasn't been parsed yet, do so */
2955 if (!seen_data_command)
2956 parse_argv();
2957
2958 end_packfile();
2959
2960 dump_branches();
2961 dump_tags();
2962 unkeep_all_packs();
2963 dump_marks();
2964
2965 if (pack_edges)
2966 fclose(pack_edges);
2967
2968 if (show_stats) {
2969 uintmax_t total_count = 0, duplicate_count = 0;
2970 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2971 total_count += object_count_by_type[i];
2972 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2973 duplicate_count += duplicate_count_by_type[i];
2974
2975 fprintf(stderr, "%s statistics:\n", argv[0]);
2976 fprintf(stderr, "---------------------------------------------------------------------\n");
2977 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2978 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2979 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]);
2980 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]);
2981 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]);
2982 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]);
2983 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2984 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2985 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2986 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2987 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2988 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2989 fprintf(stderr, "---------------------------------------------------------------------\n");
2990 pack_report();
2991 fprintf(stderr, "---------------------------------------------------------------------\n");
2992 fprintf(stderr, "\n");
2993 }
2994
2995 return failure ? 1 : 0;
2996 }
ramsay's fork of git