3886a1b46415ddbf66e5c2be401ad7fd2c012363
[git/git.git] / fast-import.c
1 /*
2 (See Documentation/git-fast-import.txt for maintained documentation.)
3 Format of STDIN stream:
4
5 stream ::= cmd*;
6
7 cmd ::= new_blob
8 | new_commit
9 | new_tag
10 | reset_branch
11 | checkpoint
12 | progress
13 ;
14
15 new_blob ::= 'blob' lf
16 mark?
17 file_content;
18 file_content ::= data;
19
20 new_commit ::= 'commit' sp ref_str lf
21 mark?
22 ('author' (sp name)? sp '<' email '>' sp when lf)?
23 'committer' (sp name)? sp '<' email '>' sp when lf
24 commit_msg
25 ('from' sp committish lf)?
26 ('merge' sp committish lf)*
27 file_change*
28 lf?;
29 commit_msg ::= data;
30
31 file_change ::= file_clr
32 | file_del
33 | file_rnm
34 | file_cpy
35 | file_obm
36 | file_inm;
37 file_clr ::= 'deleteall' lf;
38 file_del ::= 'D' sp path_str lf;
39 file_rnm ::= 'R' sp path_str sp path_str lf;
40 file_cpy ::= 'C' sp path_str sp path_str lf;
41 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
42 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
43 data;
44 note_obm ::= 'N' sp (hexsha1 | idnum) sp committish lf;
45 note_inm ::= 'N' sp 'inline' sp committish lf
46 data;
47
48 new_tag ::= 'tag' sp tag_str lf
49 'from' sp committish lf
50 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
51 tag_msg;
52 tag_msg ::= data;
53
54 reset_branch ::= 'reset' sp ref_str lf
55 ('from' sp committish lf)?
56 lf?;
57
58 checkpoint ::= 'checkpoint' lf
59 lf?;
60
61 progress ::= 'progress' sp not_lf* lf
62 lf?;
63
64 # note: the first idnum in a stream should be 1 and subsequent
65 # idnums should not have gaps between values as this will cause
66 # the stream parser to reserve space for the gapped values. An
67 # idnum can be updated in the future to a new object by issuing
68 # a new mark directive with the old idnum.
69 #
70 mark ::= 'mark' sp idnum lf;
71 data ::= (delimited_data | exact_data)
72 lf?;
73
74 # note: delim may be any string but must not contain lf.
75 # data_line may contain any data but must not be exactly
76 # delim.
77 delimited_data ::= 'data' sp '<<' delim lf
78 (data_line lf)*
79 delim lf;
80
81 # note: declen indicates the length of binary_data in bytes.
82 # declen does not include the lf preceding the binary data.
83 #
84 exact_data ::= 'data' sp declen lf
85 binary_data;
86
87 # note: quoted strings are C-style quoting supporting \c for
88 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
89 # is the signed byte value in octal. Note that the only
90 # characters which must actually be escaped to protect the
91 # stream formatting is: \, " and LF. Otherwise these values
92 # are UTF8.
93 #
94 committish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
95 ref_str ::= ref;
96 sha1exp_str ::= sha1exp;
97 tag_str ::= tag;
98 path_str ::= path | '"' quoted(path) '"' ;
99 mode ::= '100644' | '644'
100 | '100755' | '755'
101 | '120000'
102 ;
103
104 declen ::= # unsigned 32 bit value, ascii base10 notation;
105 bigint ::= # unsigned integer value, ascii base10 notation;
106 binary_data ::= # file content, not interpreted;
107
108 when ::= raw_when | rfc2822_when;
109 raw_when ::= ts sp tz;
110 rfc2822_when ::= # Valid RFC 2822 date and time;
111
112 sp ::= # ASCII space character;
113 lf ::= # ASCII newline (LF) character;
114
115 # note: a colon (':') must precede the numerical value assigned to
116 # an idnum. This is to distinguish it from a ref or tag name as
117 # GIT does not permit ':' in ref or tag strings.
118 #
119 idnum ::= ':' bigint;
120 path ::= # GIT style file path, e.g. "a/b/c";
121 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
122 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
123 sha1exp ::= # Any valid GIT SHA1 expression;
124 hexsha1 ::= # SHA1 in hexadecimal format;
125
126 # note: name and email are UTF8 strings, however name must not
127 # contain '<' or lf and email must not contain any of the
128 # following: '<', '>', lf.
129 #
130 name ::= # valid GIT author/committer name;
131 email ::= # valid GIT author/committer email;
132 ts ::= # time since the epoch in seconds, ascii base10 notation;
133 tz ::= # GIT style timezone;
134
135 # note: comments and cat requests may appear anywhere
136 # in the input, except within a data command. Any form
137 # of the data command always escapes the related input
138 # from comment processing.
139 #
140 # In case it is not clear, the '#' that starts the comment
141 # must be the first character on that line (an lf
142 # preceded it).
143 #
144 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
145
146 comment ::= '#' not_lf* lf;
147 not_lf ::= # Any byte that is not ASCII newline (LF);
148 */
149
150 #include "builtin.h"
151 #include "cache.h"
152 #include "object.h"
153 #include "blob.h"
154 #include "tree.h"
155 #include "commit.h"
156 #include "delta.h"
157 #include "pack.h"
158 #include "refs.h"
159 #include "csum-file.h"
160 #include "quote.h"
161 #include "exec_cmd.h"
162 #include "dir.h"
163
164 #define PACK_ID_BITS 16
165 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
166 #define DEPTH_BITS 13
167 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
168
169 struct object_entry
170 {
171 struct pack_idx_entry idx;
172 struct object_entry *next;
173 uint32_t type : TYPE_BITS,
174 pack_id : PACK_ID_BITS,
175 depth : DEPTH_BITS;
176 };
177
178 struct object_entry_pool
179 {
180 struct object_entry_pool *next_pool;
181 struct object_entry *next_free;
182 struct object_entry *end;
183 struct object_entry entries[FLEX_ARRAY]; /* more */
184 };
185
186 struct mark_set
187 {
188 union {
189 struct object_entry *marked[1024];
190 struct mark_set *sets[1024];
191 } data;
192 unsigned int shift;
193 };
194
195 struct last_object
196 {
197 struct strbuf data;
198 off_t offset;
199 unsigned int depth;
200 unsigned no_swap : 1;
201 };
202
203 struct mem_pool
204 {
205 struct mem_pool *next_pool;
206 char *next_free;
207 char *end;
208 uintmax_t space[FLEX_ARRAY]; /* more */
209 };
210
211 struct atom_str
212 {
213 struct atom_str *next_atom;
214 unsigned short str_len;
215 char str_dat[FLEX_ARRAY]; /* more */
216 };
217
218 struct tree_content;
219 struct tree_entry
220 {
221 struct tree_content *tree;
222 struct atom_str *name;
223 struct tree_entry_ms
224 {
225 uint16_t mode;
226 unsigned char sha1[20];
227 } versions[2];
228 };
229
230 struct tree_content
231 {
232 unsigned int entry_capacity; /* must match avail_tree_content */
233 unsigned int entry_count;
234 unsigned int delta_depth;
235 struct tree_entry *entries[FLEX_ARRAY]; /* more */
236 };
237
238 struct avail_tree_content
239 {
240 unsigned int entry_capacity; /* must match tree_content */
241 struct avail_tree_content *next_avail;
242 };
243
244 struct branch
245 {
246 struct branch *table_next_branch;
247 struct branch *active_next_branch;
248 const char *name;
249 struct tree_entry branch_tree;
250 uintmax_t last_commit;
251 uintmax_t num_notes;
252 unsigned active : 1;
253 unsigned pack_id : PACK_ID_BITS;
254 unsigned char sha1[20];
255 };
256
257 struct tag
258 {
259 struct tag *next_tag;
260 const char *name;
261 unsigned int pack_id;
262 unsigned char sha1[20];
263 };
264
265 struct hash_list
266 {
267 struct hash_list *next;
268 unsigned char sha1[20];
269 };
270
271 typedef enum {
272 WHENSPEC_RAW = 1,
273 WHENSPEC_RFC2822,
274 WHENSPEC_NOW
275 } whenspec_type;
276
277 struct recent_command
278 {
279 struct recent_command *prev;
280 struct recent_command *next;
281 char *buf;
282 };
283
284 /* Configured limits on output */
285 static unsigned long max_depth = 10;
286 static off_t max_packsize;
287 static uintmax_t big_file_threshold = 512 * 1024 * 1024;
288 static int force_update;
289 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
290 static int pack_compression_seen;
291
292 /* Stats and misc. counters */
293 static uintmax_t alloc_count;
294 static uintmax_t marks_set_count;
295 static uintmax_t object_count_by_type[1 << TYPE_BITS];
296 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
297 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
298 static unsigned long object_count;
299 static unsigned long branch_count;
300 static unsigned long branch_load_count;
301 static int failure;
302 static FILE *pack_edges;
303 static unsigned int show_stats = 1;
304 static int global_argc;
305 static const char **global_argv;
306
307 /* Memory pools */
308 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
309 static size_t total_allocd;
310 static struct mem_pool *mem_pool;
311
312 /* Atom management */
313 static unsigned int atom_table_sz = 4451;
314 static unsigned int atom_cnt;
315 static struct atom_str **atom_table;
316
317 /* The .pack file being generated */
318 static unsigned int pack_id;
319 static struct sha1file *pack_file;
320 static struct packed_git *pack_data;
321 static struct packed_git **all_packs;
322 static off_t pack_size;
323
324 /* Table of objects we've written. */
325 static unsigned int object_entry_alloc = 5000;
326 static struct object_entry_pool *blocks;
327 static struct object_entry *object_table[1 << 16];
328 static struct mark_set *marks;
329 static const char *export_marks_file;
330 static const char *import_marks_file;
331 static int import_marks_file_from_stream;
332 static int import_marks_file_ignore_missing;
333 static int relative_marks_paths;
334
335 /* Our last blob */
336 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
337
338 /* Tree management */
339 static unsigned int tree_entry_alloc = 1000;
340 static void *avail_tree_entry;
341 static unsigned int avail_tree_table_sz = 100;
342 static struct avail_tree_content **avail_tree_table;
343 static struct strbuf old_tree = STRBUF_INIT;
344 static struct strbuf new_tree = STRBUF_INIT;
345
346 /* Branch data */
347 static unsigned long max_active_branches = 5;
348 static unsigned long cur_active_branches;
349 static unsigned long branch_table_sz = 1039;
350 static struct branch **branch_table;
351 static struct branch *active_branches;
352
353 /* Tag data */
354 static struct tag *first_tag;
355 static struct tag *last_tag;
356
357 /* Input stream parsing */
358 static whenspec_type whenspec = WHENSPEC_RAW;
359 static struct strbuf command_buf = STRBUF_INIT;
360 static int unread_command_buf;
361 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
362 static struct recent_command *cmd_tail = &cmd_hist;
363 static struct recent_command *rc_free;
364 static unsigned int cmd_save = 100;
365 static uintmax_t next_mark;
366 static struct strbuf new_data = STRBUF_INIT;
367 static int seen_data_command;
368
369 /* Signal handling */
370 static volatile sig_atomic_t checkpoint_requested;
371
372 /* Where to write output of cat-blob commands */
373 static int cat_blob_fd = STDOUT_FILENO;
374
375 static void parse_argv(void);
376 static void parse_cat_blob(void);
377
378 static void write_branch_report(FILE *rpt, struct branch *b)
379 {
380 fprintf(rpt, "%s:\n", b->name);
381
382 fprintf(rpt, " status :");
383 if (b->active)
384 fputs(" active", rpt);
385 if (b->branch_tree.tree)
386 fputs(" loaded", rpt);
387 if (is_null_sha1(b->branch_tree.versions[1].sha1))
388 fputs(" dirty", rpt);
389 fputc('\n', rpt);
390
391 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
392 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
393 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
394 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
395
396 fputs(" last pack : ", rpt);
397 if (b->pack_id < MAX_PACK_ID)
398 fprintf(rpt, "%u", b->pack_id);
399 fputc('\n', rpt);
400
401 fputc('\n', rpt);
402 }
403
404 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
405
406 static void write_crash_report(const char *err)
407 {
408 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
409 FILE *rpt = fopen(loc, "w");
410 struct branch *b;
411 unsigned long lu;
412 struct recent_command *rc;
413
414 if (!rpt) {
415 error("can't write crash report %s: %s", loc, strerror(errno));
416 return;
417 }
418
419 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
420
421 fprintf(rpt, "fast-import crash report:\n");
422 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
423 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
424 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
425 fputc('\n', rpt);
426
427 fputs("fatal: ", rpt);
428 fputs(err, rpt);
429 fputc('\n', rpt);
430
431 fputc('\n', rpt);
432 fputs("Most Recent Commands Before Crash\n", rpt);
433 fputs("---------------------------------\n", rpt);
434 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
435 if (rc->next == &cmd_hist)
436 fputs("* ", rpt);
437 else
438 fputs(" ", rpt);
439 fputs(rc->buf, rpt);
440 fputc('\n', rpt);
441 }
442
443 fputc('\n', rpt);
444 fputs("Active Branch LRU\n", rpt);
445 fputs("-----------------\n", rpt);
446 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
447 cur_active_branches,
448 max_active_branches);
449 fputc('\n', rpt);
450 fputs(" pos clock name\n", rpt);
451 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
452 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
453 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
454 ++lu, b->last_commit, b->name);
455
456 fputc('\n', rpt);
457 fputs("Inactive Branches\n", rpt);
458 fputs("-----------------\n", rpt);
459 for (lu = 0; lu < branch_table_sz; lu++) {
460 for (b = branch_table[lu]; b; b = b->table_next_branch)
461 write_branch_report(rpt, b);
462 }
463
464 if (first_tag) {
465 struct tag *tg;
466 fputc('\n', rpt);
467 fputs("Annotated Tags\n", rpt);
468 fputs("--------------\n", rpt);
469 for (tg = first_tag; tg; tg = tg->next_tag) {
470 fputs(sha1_to_hex(tg->sha1), rpt);
471 fputc(' ', rpt);
472 fputs(tg->name, rpt);
473 fputc('\n', rpt);
474 }
475 }
476
477 fputc('\n', rpt);
478 fputs("Marks\n", rpt);
479 fputs("-----\n", rpt);
480 if (export_marks_file)
481 fprintf(rpt, " exported to %s\n", export_marks_file);
482 else
483 dump_marks_helper(rpt, 0, marks);
484
485 fputc('\n', rpt);
486 fputs("-------------------\n", rpt);
487 fputs("END OF CRASH REPORT\n", rpt);
488 fclose(rpt);
489 }
490
491 static void end_packfile(void);
492 static void unkeep_all_packs(void);
493 static void dump_marks(void);
494
495 static NORETURN void die_nicely(const char *err, va_list params)
496 {
497 static int zombie;
498 char message[2 * PATH_MAX];
499
500 vsnprintf(message, sizeof(message), err, params);
501 fputs("fatal: ", stderr);
502 fputs(message, stderr);
503 fputc('\n', stderr);
504
505 if (!zombie) {
506 zombie = 1;
507 write_crash_report(message);
508 end_packfile();
509 unkeep_all_packs();
510 dump_marks();
511 }
512 exit(128);
513 }
514
515 #ifndef SIGUSR1 /* Windows, for example */
516
517 static void set_checkpoint_signal(void)
518 {
519 }
520
521 #else
522
523 static void checkpoint_signal(int signo)
524 {
525 checkpoint_requested = 1;
526 }
527
528 static void set_checkpoint_signal(void)
529 {
530 struct sigaction sa;
531
532 memset(&sa, 0, sizeof(sa));
533 sa.sa_handler = checkpoint_signal;
534 sigemptyset(&sa.sa_mask);
535 sa.sa_flags = SA_RESTART;
536 sigaction(SIGUSR1, &sa, NULL);
537 }
538
539 #endif
540
541 static void alloc_objects(unsigned int cnt)
542 {
543 struct object_entry_pool *b;
544
545 b = xmalloc(sizeof(struct object_entry_pool)
546 + cnt * sizeof(struct object_entry));
547 b->next_pool = blocks;
548 b->next_free = b->entries;
549 b->end = b->entries + cnt;
550 blocks = b;
551 alloc_count += cnt;
552 }
553
554 static struct object_entry *new_object(unsigned char *sha1)
555 {
556 struct object_entry *e;
557
558 if (blocks->next_free == blocks->end)
559 alloc_objects(object_entry_alloc);
560
561 e = blocks->next_free++;
562 hashcpy(e->idx.sha1, sha1);
563 return e;
564 }
565
566 static struct object_entry *find_object(unsigned char *sha1)
567 {
568 unsigned int h = sha1[0] << 8 | sha1[1];
569 struct object_entry *e;
570 for (e = object_table[h]; e; e = e->next)
571 if (!hashcmp(sha1, e->idx.sha1))
572 return e;
573 return NULL;
574 }
575
576 static struct object_entry *insert_object(unsigned char *sha1)
577 {
578 unsigned int h = sha1[0] << 8 | sha1[1];
579 struct object_entry *e = object_table[h];
580
581 while (e) {
582 if (!hashcmp(sha1, e->idx.sha1))
583 return e;
584 e = e->next;
585 }
586
587 e = new_object(sha1);
588 e->next = object_table[h];
589 e->idx.offset = 0;
590 object_table[h] = e;
591 return e;
592 }
593
594 static unsigned int hc_str(const char *s, size_t len)
595 {
596 unsigned int r = 0;
597 while (len-- > 0)
598 r = r * 31 + *s++;
599 return r;
600 }
601
602 static void *pool_alloc(size_t len)
603 {
604 struct mem_pool *p;
605 void *r;
606
607 /* round up to a 'uintmax_t' alignment */
608 if (len & (sizeof(uintmax_t) - 1))
609 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
610
611 for (p = mem_pool; p; p = p->next_pool)
612 if ((p->end - p->next_free >= len))
613 break;
614
615 if (!p) {
616 if (len >= (mem_pool_alloc/2)) {
617 total_allocd += len;
618 return xmalloc(len);
619 }
620 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
621 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
622 p->next_pool = mem_pool;
623 p->next_free = (char *) p->space;
624 p->end = p->next_free + mem_pool_alloc;
625 mem_pool = p;
626 }
627
628 r = p->next_free;
629 p->next_free += len;
630 return r;
631 }
632
633 static void *pool_calloc(size_t count, size_t size)
634 {
635 size_t len = count * size;
636 void *r = pool_alloc(len);
637 memset(r, 0, len);
638 return r;
639 }
640
641 static char *pool_strdup(const char *s)
642 {
643 char *r = pool_alloc(strlen(s) + 1);
644 strcpy(r, s);
645 return r;
646 }
647
648 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
649 {
650 struct mark_set *s = marks;
651 while ((idnum >> s->shift) >= 1024) {
652 s = pool_calloc(1, sizeof(struct mark_set));
653 s->shift = marks->shift + 10;
654 s->data.sets[0] = marks;
655 marks = s;
656 }
657 while (s->shift) {
658 uintmax_t i = idnum >> s->shift;
659 idnum -= i << s->shift;
660 if (!s->data.sets[i]) {
661 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
662 s->data.sets[i]->shift = s->shift - 10;
663 }
664 s = s->data.sets[i];
665 }
666 if (!s->data.marked[idnum])
667 marks_set_count++;
668 s->data.marked[idnum] = oe;
669 }
670
671 static struct object_entry *find_mark(uintmax_t idnum)
672 {
673 uintmax_t orig_idnum = idnum;
674 struct mark_set *s = marks;
675 struct object_entry *oe = NULL;
676 if ((idnum >> s->shift) < 1024) {
677 while (s && s->shift) {
678 uintmax_t i = idnum >> s->shift;
679 idnum -= i << s->shift;
680 s = s->data.sets[i];
681 }
682 if (s)
683 oe = s->data.marked[idnum];
684 }
685 if (!oe)
686 die("mark :%" PRIuMAX " not declared", orig_idnum);
687 return oe;
688 }
689
690 static struct atom_str *to_atom(const char *s, unsigned short len)
691 {
692 unsigned int hc = hc_str(s, len) % atom_table_sz;
693 struct atom_str *c;
694
695 for (c = atom_table[hc]; c; c = c->next_atom)
696 if (c->str_len == len && !strncmp(s, c->str_dat, len))
697 return c;
698
699 c = pool_alloc(sizeof(struct atom_str) + len + 1);
700 c->str_len = len;
701 strncpy(c->str_dat, s, len);
702 c->str_dat[len] = 0;
703 c->next_atom = atom_table[hc];
704 atom_table[hc] = c;
705 atom_cnt++;
706 return c;
707 }
708
709 static struct branch *lookup_branch(const char *name)
710 {
711 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
712 struct branch *b;
713
714 for (b = branch_table[hc]; b; b = b->table_next_branch)
715 if (!strcmp(name, b->name))
716 return b;
717 return NULL;
718 }
719
720 static struct branch *new_branch(const char *name)
721 {
722 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
723 struct branch *b = lookup_branch(name);
724
725 if (b)
726 die("Invalid attempt to create duplicate branch: %s", name);
727 switch (check_ref_format(name)) {
728 case 0: break; /* its valid */
729 case CHECK_REF_FORMAT_ONELEVEL:
730 break; /* valid, but too few '/', allow anyway */
731 default:
732 die("Branch name doesn't conform to GIT standards: %s", name);
733 }
734
735 b = pool_calloc(1, sizeof(struct branch));
736 b->name = pool_strdup(name);
737 b->table_next_branch = branch_table[hc];
738 b->branch_tree.versions[0].mode = S_IFDIR;
739 b->branch_tree.versions[1].mode = S_IFDIR;
740 b->num_notes = 0;
741 b->active = 0;
742 b->pack_id = MAX_PACK_ID;
743 branch_table[hc] = b;
744 branch_count++;
745 return b;
746 }
747
748 static unsigned int hc_entries(unsigned int cnt)
749 {
750 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
751 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
752 }
753
754 static struct tree_content *new_tree_content(unsigned int cnt)
755 {
756 struct avail_tree_content *f, *l = NULL;
757 struct tree_content *t;
758 unsigned int hc = hc_entries(cnt);
759
760 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
761 if (f->entry_capacity >= cnt)
762 break;
763
764 if (f) {
765 if (l)
766 l->next_avail = f->next_avail;
767 else
768 avail_tree_table[hc] = f->next_avail;
769 } else {
770 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
771 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
772 f->entry_capacity = cnt;
773 }
774
775 t = (struct tree_content*)f;
776 t->entry_count = 0;
777 t->delta_depth = 0;
778 return t;
779 }
780
781 static void release_tree_entry(struct tree_entry *e);
782 static void release_tree_content(struct tree_content *t)
783 {
784 struct avail_tree_content *f = (struct avail_tree_content*)t;
785 unsigned int hc = hc_entries(f->entry_capacity);
786 f->next_avail = avail_tree_table[hc];
787 avail_tree_table[hc] = f;
788 }
789
790 static void release_tree_content_recursive(struct tree_content *t)
791 {
792 unsigned int i;
793 for (i = 0; i < t->entry_count; i++)
794 release_tree_entry(t->entries[i]);
795 release_tree_content(t);
796 }
797
798 static struct tree_content *grow_tree_content(
799 struct tree_content *t,
800 int amt)
801 {
802 struct tree_content *r = new_tree_content(t->entry_count + amt);
803 r->entry_count = t->entry_count;
804 r->delta_depth = t->delta_depth;
805 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
806 release_tree_content(t);
807 return r;
808 }
809
810 static struct tree_entry *new_tree_entry(void)
811 {
812 struct tree_entry *e;
813
814 if (!avail_tree_entry) {
815 unsigned int n = tree_entry_alloc;
816 total_allocd += n * sizeof(struct tree_entry);
817 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
818 while (n-- > 1) {
819 *((void**)e) = e + 1;
820 e++;
821 }
822 *((void**)e) = NULL;
823 }
824
825 e = avail_tree_entry;
826 avail_tree_entry = *((void**)e);
827 return e;
828 }
829
830 static void release_tree_entry(struct tree_entry *e)
831 {
832 if (e->tree)
833 release_tree_content_recursive(e->tree);
834 *((void**)e) = avail_tree_entry;
835 avail_tree_entry = e;
836 }
837
838 static struct tree_content *dup_tree_content(struct tree_content *s)
839 {
840 struct tree_content *d;
841 struct tree_entry *a, *b;
842 unsigned int i;
843
844 if (!s)
845 return NULL;
846 d = new_tree_content(s->entry_count);
847 for (i = 0; i < s->entry_count; i++) {
848 a = s->entries[i];
849 b = new_tree_entry();
850 memcpy(b, a, sizeof(*a));
851 if (a->tree && is_null_sha1(b->versions[1].sha1))
852 b->tree = dup_tree_content(a->tree);
853 else
854 b->tree = NULL;
855 d->entries[i] = b;
856 }
857 d->entry_count = s->entry_count;
858 d->delta_depth = s->delta_depth;
859
860 return d;
861 }
862
863 static void start_packfile(void)
864 {
865 static char tmpfile[PATH_MAX];
866 struct packed_git *p;
867 struct pack_header hdr;
868 int pack_fd;
869
870 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
871 "pack/tmp_pack_XXXXXX");
872 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
873 strcpy(p->pack_name, tmpfile);
874 p->pack_fd = pack_fd;
875 pack_file = sha1fd(pack_fd, p->pack_name);
876
877 hdr.hdr_signature = htonl(PACK_SIGNATURE);
878 hdr.hdr_version = htonl(2);
879 hdr.hdr_entries = 0;
880 sha1write(pack_file, &hdr, sizeof(hdr));
881
882 pack_data = p;
883 pack_size = sizeof(hdr);
884 object_count = 0;
885
886 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
887 all_packs[pack_id] = p;
888 }
889
890 static const char *create_index(void)
891 {
892 const char *tmpfile;
893 struct pack_idx_entry **idx, **c, **last;
894 struct object_entry *e;
895 struct object_entry_pool *o;
896
897 /* Build the table of object IDs. */
898 idx = xmalloc(object_count * sizeof(*idx));
899 c = idx;
900 for (o = blocks; o; o = o->next_pool)
901 for (e = o->next_free; e-- != o->entries;)
902 if (pack_id == e->pack_id)
903 *c++ = &e->idx;
904 last = idx + object_count;
905 if (c != last)
906 die("internal consistency error creating the index");
907
908 tmpfile = write_idx_file(NULL, idx, object_count, pack_data->sha1);
909 free(idx);
910 return tmpfile;
911 }
912
913 static char *keep_pack(const char *curr_index_name)
914 {
915 static char name[PATH_MAX];
916 static const char *keep_msg = "fast-import";
917 int keep_fd;
918
919 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
920 if (keep_fd < 0)
921 die_errno("cannot create keep file");
922 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
923 if (close(keep_fd))
924 die_errno("failed to write keep file");
925
926 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
927 get_object_directory(), sha1_to_hex(pack_data->sha1));
928 if (move_temp_to_file(pack_data->pack_name, name))
929 die("cannot store pack file");
930
931 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
932 get_object_directory(), sha1_to_hex(pack_data->sha1));
933 if (move_temp_to_file(curr_index_name, name))
934 die("cannot store index file");
935 free((void *)curr_index_name);
936 return name;
937 }
938
939 static void unkeep_all_packs(void)
940 {
941 static char name[PATH_MAX];
942 int k;
943
944 for (k = 0; k < pack_id; k++) {
945 struct packed_git *p = all_packs[k];
946 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
947 get_object_directory(), sha1_to_hex(p->sha1));
948 unlink_or_warn(name);
949 }
950 }
951
952 static void end_packfile(void)
953 {
954 struct packed_git *old_p = pack_data, *new_p;
955
956 clear_delta_base_cache();
957 if (object_count) {
958 unsigned char cur_pack_sha1[20];
959 char *idx_name;
960 int i;
961 struct branch *b;
962 struct tag *t;
963
964 close_pack_windows(pack_data);
965 sha1close(pack_file, cur_pack_sha1, 0);
966 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
967 pack_data->pack_name, object_count,
968 cur_pack_sha1, pack_size);
969 close(pack_data->pack_fd);
970 idx_name = keep_pack(create_index());
971
972 /* Register the packfile with core git's machinery. */
973 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
974 if (!new_p)
975 die("core git rejected index %s", idx_name);
976 all_packs[pack_id] = new_p;
977 install_packed_git(new_p);
978
979 /* Print the boundary */
980 if (pack_edges) {
981 fprintf(pack_edges, "%s:", new_p->pack_name);
982 for (i = 0; i < branch_table_sz; i++) {
983 for (b = branch_table[i]; b; b = b->table_next_branch) {
984 if (b->pack_id == pack_id)
985 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
986 }
987 }
988 for (t = first_tag; t; t = t->next_tag) {
989 if (t->pack_id == pack_id)
990 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
991 }
992 fputc('\n', pack_edges);
993 fflush(pack_edges);
994 }
995
996 pack_id++;
997 }
998 else {
999 close(old_p->pack_fd);
1000 unlink_or_warn(old_p->pack_name);
1001 }
1002 free(old_p);
1003
1004 /* We can't carry a delta across packfiles. */
1005 strbuf_release(&last_blob.data);
1006 last_blob.offset = 0;
1007 last_blob.depth = 0;
1008 }
1009
1010 static void cycle_packfile(void)
1011 {
1012 end_packfile();
1013 start_packfile();
1014 }
1015
1016 static int store_object(
1017 enum object_type type,
1018 struct strbuf *dat,
1019 struct last_object *last,
1020 unsigned char *sha1out,
1021 uintmax_t mark)
1022 {
1023 void *out, *delta;
1024 struct object_entry *e;
1025 unsigned char hdr[96];
1026 unsigned char sha1[20];
1027 unsigned long hdrlen, deltalen;
1028 git_SHA_CTX c;
1029 z_stream s;
1030
1031 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1032 (unsigned long)dat->len) + 1;
1033 git_SHA1_Init(&c);
1034 git_SHA1_Update(&c, hdr, hdrlen);
1035 git_SHA1_Update(&c, dat->buf, dat->len);
1036 git_SHA1_Final(sha1, &c);
1037 if (sha1out)
1038 hashcpy(sha1out, sha1);
1039
1040 e = insert_object(sha1);
1041 if (mark)
1042 insert_mark(mark, e);
1043 if (e->idx.offset) {
1044 duplicate_count_by_type[type]++;
1045 return 1;
1046 } else if (find_sha1_pack(sha1, packed_git)) {
1047 e->type = type;
1048 e->pack_id = MAX_PACK_ID;
1049 e->idx.offset = 1; /* just not zero! */
1050 duplicate_count_by_type[type]++;
1051 return 1;
1052 }
1053
1054 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1055 delta = diff_delta(last->data.buf, last->data.len,
1056 dat->buf, dat->len,
1057 &deltalen, dat->len - 20);
1058 } else
1059 delta = NULL;
1060
1061 memset(&s, 0, sizeof(s));
1062 deflateInit(&s, pack_compression_level);
1063 if (delta) {
1064 s.next_in = delta;
1065 s.avail_in = deltalen;
1066 } else {
1067 s.next_in = (void *)dat->buf;
1068 s.avail_in = dat->len;
1069 }
1070 s.avail_out = deflateBound(&s, s.avail_in);
1071 s.next_out = out = xmalloc(s.avail_out);
1072 while (deflate(&s, Z_FINISH) == Z_OK)
1073 /* nothing */;
1074 deflateEnd(&s);
1075
1076 /* Determine if we should auto-checkpoint. */
1077 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1078 || (pack_size + 60 + s.total_out) < pack_size) {
1079
1080 /* This new object needs to *not* have the current pack_id. */
1081 e->pack_id = pack_id + 1;
1082 cycle_packfile();
1083
1084 /* We cannot carry a delta into the new pack. */
1085 if (delta) {
1086 free(delta);
1087 delta = NULL;
1088
1089 memset(&s, 0, sizeof(s));
1090 deflateInit(&s, pack_compression_level);
1091 s.next_in = (void *)dat->buf;
1092 s.avail_in = dat->len;
1093 s.avail_out = deflateBound(&s, s.avail_in);
1094 s.next_out = out = xrealloc(out, s.avail_out);
1095 while (deflate(&s, Z_FINISH) == Z_OK)
1096 /* nothing */;
1097 deflateEnd(&s);
1098 }
1099 }
1100
1101 e->type = type;
1102 e->pack_id = pack_id;
1103 e->idx.offset = pack_size;
1104 object_count++;
1105 object_count_by_type[type]++;
1106
1107 crc32_begin(pack_file);
1108
1109 if (delta) {
1110 off_t ofs = e->idx.offset - last->offset;
1111 unsigned pos = sizeof(hdr) - 1;
1112
1113 delta_count_by_type[type]++;
1114 e->depth = last->depth + 1;
1115
1116 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1117 sha1write(pack_file, hdr, hdrlen);
1118 pack_size += hdrlen;
1119
1120 hdr[pos] = ofs & 127;
1121 while (ofs >>= 7)
1122 hdr[--pos] = 128 | (--ofs & 127);
1123 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1124 pack_size += sizeof(hdr) - pos;
1125 } else {
1126 e->depth = 0;
1127 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1128 sha1write(pack_file, hdr, hdrlen);
1129 pack_size += hdrlen;
1130 }
1131
1132 sha1write(pack_file, out, s.total_out);
1133 pack_size += s.total_out;
1134
1135 e->idx.crc32 = crc32_end(pack_file);
1136
1137 free(out);
1138 free(delta);
1139 if (last) {
1140 if (last->no_swap) {
1141 last->data = *dat;
1142 } else {
1143 strbuf_swap(&last->data, dat);
1144 }
1145 last->offset = e->idx.offset;
1146 last->depth = e->depth;
1147 }
1148 return 0;
1149 }
1150
1151 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1152 {
1153 if (ftruncate(pack_data->pack_fd, to)
1154 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1155 die_errno("cannot truncate pack to skip duplicate");
1156 pack_size = to;
1157
1158 /* yes this is a layering violation */
1159 pack_file->total = to;
1160 pack_file->offset = 0;
1161 pack_file->ctx = *ctx;
1162 }
1163
1164 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1165 {
1166 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1167 unsigned char *in_buf = xmalloc(in_sz);
1168 unsigned char *out_buf = xmalloc(out_sz);
1169 struct object_entry *e;
1170 unsigned char sha1[20];
1171 unsigned long hdrlen;
1172 off_t offset;
1173 git_SHA_CTX c;
1174 git_SHA_CTX pack_file_ctx;
1175 z_stream s;
1176 int status = Z_OK;
1177
1178 /* Determine if we should auto-checkpoint. */
1179 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1180 || (pack_size + 60 + len) < pack_size)
1181 cycle_packfile();
1182
1183 offset = pack_size;
1184
1185 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1186 sha1flush(pack_file);
1187 pack_file_ctx = pack_file->ctx;
1188
1189 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1190 if (out_sz <= hdrlen)
1191 die("impossibly large object header");
1192
1193 git_SHA1_Init(&c);
1194 git_SHA1_Update(&c, out_buf, hdrlen);
1195
1196 crc32_begin(pack_file);
1197
1198 memset(&s, 0, sizeof(s));
1199 deflateInit(&s, pack_compression_level);
1200
1201 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1202 if (out_sz <= hdrlen)
1203 die("impossibly large object header");
1204
1205 s.next_out = out_buf + hdrlen;
1206 s.avail_out = out_sz - hdrlen;
1207
1208 while (status != Z_STREAM_END) {
1209 if (0 < len && !s.avail_in) {
1210 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1211 size_t n = fread(in_buf, 1, cnt, stdin);
1212 if (!n && feof(stdin))
1213 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1214
1215 git_SHA1_Update(&c, in_buf, n);
1216 s.next_in = in_buf;
1217 s.avail_in = n;
1218 len -= n;
1219 }
1220
1221 status = deflate(&s, len ? 0 : Z_FINISH);
1222
1223 if (!s.avail_out || status == Z_STREAM_END) {
1224 size_t n = s.next_out - out_buf;
1225 sha1write(pack_file, out_buf, n);
1226 pack_size += n;
1227 s.next_out = out_buf;
1228 s.avail_out = out_sz;
1229 }
1230
1231 switch (status) {
1232 case Z_OK:
1233 case Z_BUF_ERROR:
1234 case Z_STREAM_END:
1235 continue;
1236 default:
1237 die("unexpected deflate failure: %d", status);
1238 }
1239 }
1240 deflateEnd(&s);
1241 git_SHA1_Final(sha1, &c);
1242
1243 if (sha1out)
1244 hashcpy(sha1out, sha1);
1245
1246 e = insert_object(sha1);
1247
1248 if (mark)
1249 insert_mark(mark, e);
1250
1251 if (e->idx.offset) {
1252 duplicate_count_by_type[OBJ_BLOB]++;
1253 truncate_pack(offset, &pack_file_ctx);
1254
1255 } else if (find_sha1_pack(sha1, packed_git)) {
1256 e->type = OBJ_BLOB;
1257 e->pack_id = MAX_PACK_ID;
1258 e->idx.offset = 1; /* just not zero! */
1259 duplicate_count_by_type[OBJ_BLOB]++;
1260 truncate_pack(offset, &pack_file_ctx);
1261
1262 } else {
1263 e->depth = 0;
1264 e->type = OBJ_BLOB;
1265 e->pack_id = pack_id;
1266 e->idx.offset = offset;
1267 e->idx.crc32 = crc32_end(pack_file);
1268 object_count++;
1269 object_count_by_type[OBJ_BLOB]++;
1270 }
1271
1272 free(in_buf);
1273 free(out_buf);
1274 }
1275
1276 /* All calls must be guarded by find_object() or find_mark() to
1277 * ensure the 'struct object_entry' passed was written by this
1278 * process instance. We unpack the entry by the offset, avoiding
1279 * the need for the corresponding .idx file. This unpacking rule
1280 * works because we only use OBJ_REF_DELTA within the packfiles
1281 * created by fast-import.
1282 *
1283 * oe must not be NULL. Such an oe usually comes from giving
1284 * an unknown SHA-1 to find_object() or an undefined mark to
1285 * find_mark(). Callers must test for this condition and use
1286 * the standard read_sha1_file() when it happens.
1287 *
1288 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1289 * find_mark(), where the mark was reloaded from an existing marks
1290 * file and is referencing an object that this fast-import process
1291 * instance did not write out to a packfile. Callers must test for
1292 * this condition and use read_sha1_file() instead.
1293 */
1294 static void *gfi_unpack_entry(
1295 struct object_entry *oe,
1296 unsigned long *sizep)
1297 {
1298 enum object_type type;
1299 struct packed_git *p = all_packs[oe->pack_id];
1300 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1301 /* The object is stored in the packfile we are writing to
1302 * and we have modified it since the last time we scanned
1303 * back to read a previously written object. If an old
1304 * window covered [p->pack_size, p->pack_size + 20) its
1305 * data is stale and is not valid. Closing all windows
1306 * and updating the packfile length ensures we can read
1307 * the newly written data.
1308 */
1309 close_pack_windows(p);
1310 sha1flush(pack_file);
1311
1312 /* We have to offer 20 bytes additional on the end of
1313 * the packfile as the core unpacker code assumes the
1314 * footer is present at the file end and must promise
1315 * at least 20 bytes within any window it maps. But
1316 * we don't actually create the footer here.
1317 */
1318 p->pack_size = pack_size + 20;
1319 }
1320 return unpack_entry(p, oe->idx.offset, &type, sizep);
1321 }
1322
1323 static const char *get_mode(const char *str, uint16_t *modep)
1324 {
1325 unsigned char c;
1326 uint16_t mode = 0;
1327
1328 while ((c = *str++) != ' ') {
1329 if (c < '0' || c > '7')
1330 return NULL;
1331 mode = (mode << 3) + (c - '0');
1332 }
1333 *modep = mode;
1334 return str;
1335 }
1336
1337 static void load_tree(struct tree_entry *root)
1338 {
1339 unsigned char *sha1 = root->versions[1].sha1;
1340 struct object_entry *myoe;
1341 struct tree_content *t;
1342 unsigned long size;
1343 char *buf;
1344 const char *c;
1345
1346 root->tree = t = new_tree_content(8);
1347 if (is_null_sha1(sha1))
1348 return;
1349
1350 myoe = find_object(sha1);
1351 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1352 if (myoe->type != OBJ_TREE)
1353 die("Not a tree: %s", sha1_to_hex(sha1));
1354 t->delta_depth = myoe->depth;
1355 buf = gfi_unpack_entry(myoe, &size);
1356 if (!buf)
1357 die("Can't load tree %s", sha1_to_hex(sha1));
1358 } else {
1359 enum object_type type;
1360 buf = read_sha1_file(sha1, &type, &size);
1361 if (!buf || type != OBJ_TREE)
1362 die("Can't load tree %s", sha1_to_hex(sha1));
1363 }
1364
1365 c = buf;
1366 while (c != (buf + size)) {
1367 struct tree_entry *e = new_tree_entry();
1368
1369 if (t->entry_count == t->entry_capacity)
1370 root->tree = t = grow_tree_content(t, t->entry_count);
1371 t->entries[t->entry_count++] = e;
1372
1373 e->tree = NULL;
1374 c = get_mode(c, &e->versions[1].mode);
1375 if (!c)
1376 die("Corrupt mode in %s", sha1_to_hex(sha1));
1377 e->versions[0].mode = e->versions[1].mode;
1378 e->name = to_atom(c, strlen(c));
1379 c += e->name->str_len + 1;
1380 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1381 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1382 c += 20;
1383 }
1384 free(buf);
1385 }
1386
1387 static int tecmp0 (const void *_a, const void *_b)
1388 {
1389 struct tree_entry *a = *((struct tree_entry**)_a);
1390 struct tree_entry *b = *((struct tree_entry**)_b);
1391 return base_name_compare(
1392 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1393 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1394 }
1395
1396 static int tecmp1 (const void *_a, const void *_b)
1397 {
1398 struct tree_entry *a = *((struct tree_entry**)_a);
1399 struct tree_entry *b = *((struct tree_entry**)_b);
1400 return base_name_compare(
1401 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1402 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1403 }
1404
1405 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1406 {
1407 size_t maxlen = 0;
1408 unsigned int i;
1409
1410 if (!v)
1411 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1412 else
1413 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1414
1415 for (i = 0; i < t->entry_count; i++) {
1416 if (t->entries[i]->versions[v].mode)
1417 maxlen += t->entries[i]->name->str_len + 34;
1418 }
1419
1420 strbuf_reset(b);
1421 strbuf_grow(b, maxlen);
1422 for (i = 0; i < t->entry_count; i++) {
1423 struct tree_entry *e = t->entries[i];
1424 if (!e->versions[v].mode)
1425 continue;
1426 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1427 e->name->str_dat, '\0');
1428 strbuf_add(b, e->versions[v].sha1, 20);
1429 }
1430 }
1431
1432 static void store_tree(struct tree_entry *root)
1433 {
1434 struct tree_content *t = root->tree;
1435 unsigned int i, j, del;
1436 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1437 struct object_entry *le;
1438
1439 if (!is_null_sha1(root->versions[1].sha1))
1440 return;
1441
1442 for (i = 0; i < t->entry_count; i++) {
1443 if (t->entries[i]->tree)
1444 store_tree(t->entries[i]);
1445 }
1446
1447 le = find_object(root->versions[0].sha1);
1448 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1449 mktree(t, 0, &old_tree);
1450 lo.data = old_tree;
1451 lo.offset = le->idx.offset;
1452 lo.depth = t->delta_depth;
1453 }
1454
1455 mktree(t, 1, &new_tree);
1456 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1457
1458 t->delta_depth = lo.depth;
1459 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1460 struct tree_entry *e = t->entries[i];
1461 if (e->versions[1].mode) {
1462 e->versions[0].mode = e->versions[1].mode;
1463 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1464 t->entries[j++] = e;
1465 } else {
1466 release_tree_entry(e);
1467 del++;
1468 }
1469 }
1470 t->entry_count -= del;
1471 }
1472
1473 static void tree_content_replace(
1474 struct tree_entry *root,
1475 const unsigned char *sha1,
1476 const uint16_t mode,
1477 struct tree_content *newtree)
1478 {
1479 if (!S_ISDIR(mode))
1480 die("Root cannot be a non-directory");
1481 hashcpy(root->versions[1].sha1, sha1);
1482 if (root->tree)
1483 release_tree_content_recursive(root->tree);
1484 root->tree = newtree;
1485 }
1486
1487 static int tree_content_set(
1488 struct tree_entry *root,
1489 const char *p,
1490 const unsigned char *sha1,
1491 const uint16_t mode,
1492 struct tree_content *subtree)
1493 {
1494 struct tree_content *t;
1495 const char *slash1;
1496 unsigned int i, n;
1497 struct tree_entry *e;
1498
1499 slash1 = strchr(p, '/');
1500 if (slash1)
1501 n = slash1 - p;
1502 else
1503 n = strlen(p);
1504 if (!n)
1505 die("Empty path component found in input");
1506 if (!slash1 && !S_ISDIR(mode) && subtree)
1507 die("Non-directories cannot have subtrees");
1508
1509 if (!root->tree)
1510 load_tree(root);
1511 t = root->tree;
1512 for (i = 0; i < t->entry_count; i++) {
1513 e = t->entries[i];
1514 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1515 if (!slash1) {
1516 if (!S_ISDIR(mode)
1517 && e->versions[1].mode == mode
1518 && !hashcmp(e->versions[1].sha1, sha1))
1519 return 0;
1520 e->versions[1].mode = mode;
1521 hashcpy(e->versions[1].sha1, sha1);
1522 if (e->tree)
1523 release_tree_content_recursive(e->tree);
1524 e->tree = subtree;
1525 hashclr(root->versions[1].sha1);
1526 return 1;
1527 }
1528 if (!S_ISDIR(e->versions[1].mode)) {
1529 e->tree = new_tree_content(8);
1530 e->versions[1].mode = S_IFDIR;
1531 }
1532 if (!e->tree)
1533 load_tree(e);
1534 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1535 hashclr(root->versions[1].sha1);
1536 return 1;
1537 }
1538 return 0;
1539 }
1540 }
1541
1542 if (t->entry_count == t->entry_capacity)
1543 root->tree = t = grow_tree_content(t, t->entry_count);
1544 e = new_tree_entry();
1545 e->name = to_atom(p, n);
1546 e->versions[0].mode = 0;
1547 hashclr(e->versions[0].sha1);
1548 t->entries[t->entry_count++] = e;
1549 if (slash1) {
1550 e->tree = new_tree_content(8);
1551 e->versions[1].mode = S_IFDIR;
1552 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1553 } else {
1554 e->tree = subtree;
1555 e->versions[1].mode = mode;
1556 hashcpy(e->versions[1].sha1, sha1);
1557 }
1558 hashclr(root->versions[1].sha1);
1559 return 1;
1560 }
1561
1562 static int tree_content_remove(
1563 struct tree_entry *root,
1564 const char *p,
1565 struct tree_entry *backup_leaf)
1566 {
1567 struct tree_content *t;
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 if (!root->tree)
1579 load_tree(root);
1580 t = root->tree;
1581 for (i = 0; i < t->entry_count; i++) {
1582 e = t->entries[i];
1583 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1584 if (slash1 && !S_ISDIR(e->versions[1].mode))
1585 /*
1586 * If p names a file in some subdirectory, and a
1587 * file or symlink matching the name of the
1588 * parent directory of p exists, then p cannot
1589 * exist and need not be deleted.
1590 */
1591 return 1;
1592 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1593 goto del_entry;
1594 if (!e->tree)
1595 load_tree(e);
1596 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1597 for (n = 0; n < e->tree->entry_count; n++) {
1598 if (e->tree->entries[n]->versions[1].mode) {
1599 hashclr(root->versions[1].sha1);
1600 return 1;
1601 }
1602 }
1603 backup_leaf = NULL;
1604 goto del_entry;
1605 }
1606 return 0;
1607 }
1608 }
1609 return 0;
1610
1611 del_entry:
1612 if (backup_leaf)
1613 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1614 else if (e->tree)
1615 release_tree_content_recursive(e->tree);
1616 e->tree = NULL;
1617 e->versions[1].mode = 0;
1618 hashclr(e->versions[1].sha1);
1619 hashclr(root->versions[1].sha1);
1620 return 1;
1621 }
1622
1623 static int tree_content_get(
1624 struct tree_entry *root,
1625 const char *p,
1626 struct tree_entry *leaf)
1627 {
1628 struct tree_content *t;
1629 const char *slash1;
1630 unsigned int i, n;
1631 struct tree_entry *e;
1632
1633 slash1 = strchr(p, '/');
1634 if (slash1)
1635 n = slash1 - p;
1636 else
1637 n = strlen(p);
1638
1639 if (!root->tree)
1640 load_tree(root);
1641 t = root->tree;
1642 for (i = 0; i < t->entry_count; i++) {
1643 e = t->entries[i];
1644 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1645 if (!slash1) {
1646 memcpy(leaf, e, sizeof(*leaf));
1647 if (e->tree && is_null_sha1(e->versions[1].sha1))
1648 leaf->tree = dup_tree_content(e->tree);
1649 else
1650 leaf->tree = NULL;
1651 return 1;
1652 }
1653 if (!S_ISDIR(e->versions[1].mode))
1654 return 0;
1655 if (!e->tree)
1656 load_tree(e);
1657 return tree_content_get(e, slash1 + 1, leaf);
1658 }
1659 }
1660 return 0;
1661 }
1662
1663 static int update_branch(struct branch *b)
1664 {
1665 static const char *msg = "fast-import";
1666 struct ref_lock *lock;
1667 unsigned char old_sha1[20];
1668
1669 if (is_null_sha1(b->sha1))
1670 return 0;
1671 if (read_ref(b->name, old_sha1))
1672 hashclr(old_sha1);
1673 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1674 if (!lock)
1675 return error("Unable to lock %s", b->name);
1676 if (!force_update && !is_null_sha1(old_sha1)) {
1677 struct commit *old_cmit, *new_cmit;
1678
1679 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1680 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1681 if (!old_cmit || !new_cmit) {
1682 unlock_ref(lock);
1683 return error("Branch %s is missing commits.", b->name);
1684 }
1685
1686 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1687 unlock_ref(lock);
1688 warning("Not updating %s"
1689 " (new tip %s does not contain %s)",
1690 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1691 return -1;
1692 }
1693 }
1694 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1695 return error("Unable to update %s", b->name);
1696 return 0;
1697 }
1698
1699 static void dump_branches(void)
1700 {
1701 unsigned int i;
1702 struct branch *b;
1703
1704 for (i = 0; i < branch_table_sz; i++) {
1705 for (b = branch_table[i]; b; b = b->table_next_branch)
1706 failure |= update_branch(b);
1707 }
1708 }
1709
1710 static void dump_tags(void)
1711 {
1712 static const char *msg = "fast-import";
1713 struct tag *t;
1714 struct ref_lock *lock;
1715 char ref_name[PATH_MAX];
1716
1717 for (t = first_tag; t; t = t->next_tag) {
1718 sprintf(ref_name, "tags/%s", t->name);
1719 lock = lock_ref_sha1(ref_name, NULL);
1720 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1721 failure |= error("Unable to update %s", ref_name);
1722 }
1723 }
1724
1725 static void dump_marks_helper(FILE *f,
1726 uintmax_t base,
1727 struct mark_set *m)
1728 {
1729 uintmax_t k;
1730 if (m->shift) {
1731 for (k = 0; k < 1024; k++) {
1732 if (m->data.sets[k])
1733 dump_marks_helper(f, base + (k << m->shift),
1734 m->data.sets[k]);
1735 }
1736 } else {
1737 for (k = 0; k < 1024; k++) {
1738 if (m->data.marked[k])
1739 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1740 sha1_to_hex(m->data.marked[k]->idx.sha1));
1741 }
1742 }
1743 }
1744
1745 static void dump_marks(void)
1746 {
1747 static struct lock_file mark_lock;
1748 int mark_fd;
1749 FILE *f;
1750
1751 if (!export_marks_file)
1752 return;
1753
1754 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1755 if (mark_fd < 0) {
1756 failure |= error("Unable to write marks file %s: %s",
1757 export_marks_file, strerror(errno));
1758 return;
1759 }
1760
1761 f = fdopen(mark_fd, "w");
1762 if (!f) {
1763 int saved_errno = errno;
1764 rollback_lock_file(&mark_lock);
1765 failure |= error("Unable to write marks file %s: %s",
1766 export_marks_file, strerror(saved_errno));
1767 return;
1768 }
1769
1770 /*
1771 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1772 * Assign -1 to the lock file descriptor so that commit_lock_file()
1773 * won't try to close() it.
1774 */
1775 mark_lock.fd = -1;
1776
1777 dump_marks_helper(f, 0, marks);
1778 if (ferror(f) || fclose(f)) {
1779 int saved_errno = errno;
1780 rollback_lock_file(&mark_lock);
1781 failure |= error("Unable to write marks file %s: %s",
1782 export_marks_file, strerror(saved_errno));
1783 return;
1784 }
1785
1786 if (commit_lock_file(&mark_lock)) {
1787 int saved_errno = errno;
1788 rollback_lock_file(&mark_lock);
1789 failure |= error("Unable to commit marks file %s: %s",
1790 export_marks_file, strerror(saved_errno));
1791 return;
1792 }
1793 }
1794
1795 static void read_marks(void)
1796 {
1797 char line[512];
1798 FILE *f = fopen(import_marks_file, "r");
1799 if (f)
1800 ;
1801 else if (import_marks_file_ignore_missing && errno == ENOENT)
1802 return; /* Marks file does not exist */
1803 else
1804 die_errno("cannot read '%s'", import_marks_file);
1805 while (fgets(line, sizeof(line), f)) {
1806 uintmax_t mark;
1807 char *end;
1808 unsigned char sha1[20];
1809 struct object_entry *e;
1810
1811 end = strchr(line, '\n');
1812 if (line[0] != ':' || !end)
1813 die("corrupt mark line: %s", line);
1814 *end = 0;
1815 mark = strtoumax(line + 1, &end, 10);
1816 if (!mark || end == line + 1
1817 || *end != ' ' || get_sha1(end + 1, sha1))
1818 die("corrupt mark line: %s", line);
1819 e = find_object(sha1);
1820 if (!e) {
1821 enum object_type type = sha1_object_info(sha1, NULL);
1822 if (type < 0)
1823 die("object not found: %s", sha1_to_hex(sha1));
1824 e = insert_object(sha1);
1825 e->type = type;
1826 e->pack_id = MAX_PACK_ID;
1827 e->idx.offset = 1; /* just not zero! */
1828 }
1829 insert_mark(mark, e);
1830 }
1831 fclose(f);
1832 }
1833
1834
1835 static int read_next_command(void)
1836 {
1837 static int stdin_eof = 0;
1838
1839 if (stdin_eof) {
1840 unread_command_buf = 0;
1841 return EOF;
1842 }
1843
1844 for (;;) {
1845 if (unread_command_buf) {
1846 unread_command_buf = 0;
1847 } else {
1848 struct recent_command *rc;
1849
1850 strbuf_detach(&command_buf, NULL);
1851 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1852 if (stdin_eof)
1853 return EOF;
1854
1855 if (!seen_data_command
1856 && prefixcmp(command_buf.buf, "feature ")
1857 && prefixcmp(command_buf.buf, "option ")) {
1858 parse_argv();
1859 }
1860
1861 rc = rc_free;
1862 if (rc)
1863 rc_free = rc->next;
1864 else {
1865 rc = cmd_hist.next;
1866 cmd_hist.next = rc->next;
1867 cmd_hist.next->prev = &cmd_hist;
1868 free(rc->buf);
1869 }
1870
1871 rc->buf = command_buf.buf;
1872 rc->prev = cmd_tail;
1873 rc->next = cmd_hist.prev;
1874 rc->prev->next = rc;
1875 cmd_tail = rc;
1876 }
1877 if (!prefixcmp(command_buf.buf, "cat-blob ")) {
1878 parse_cat_blob();
1879 continue;
1880 }
1881 if (command_buf.buf[0] == '#')
1882 continue;
1883 return 0;
1884 }
1885 }
1886
1887 static void skip_optional_lf(void)
1888 {
1889 int term_char = fgetc(stdin);
1890 if (term_char != '\n' && term_char != EOF)
1891 ungetc(term_char, stdin);
1892 }
1893
1894 static void parse_mark(void)
1895 {
1896 if (!prefixcmp(command_buf.buf, "mark :")) {
1897 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1898 read_next_command();
1899 }
1900 else
1901 next_mark = 0;
1902 }
1903
1904 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1905 {
1906 strbuf_reset(sb);
1907
1908 if (prefixcmp(command_buf.buf, "data "))
1909 die("Expected 'data n' command, found: %s", command_buf.buf);
1910
1911 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1912 char *term = xstrdup(command_buf.buf + 5 + 2);
1913 size_t term_len = command_buf.len - 5 - 2;
1914
1915 strbuf_detach(&command_buf, NULL);
1916 for (;;) {
1917 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1918 die("EOF in data (terminator '%s' not found)", term);
1919 if (term_len == command_buf.len
1920 && !strcmp(term, command_buf.buf))
1921 break;
1922 strbuf_addbuf(sb, &command_buf);
1923 strbuf_addch(sb, '\n');
1924 }
1925 free(term);
1926 }
1927 else {
1928 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1929 size_t n = 0, length = (size_t)len;
1930
1931 if (limit && limit < len) {
1932 *len_res = len;
1933 return 0;
1934 }
1935 if (length < len)
1936 die("data is too large to use in this context");
1937
1938 while (n < length) {
1939 size_t s = strbuf_fread(sb, length - n, stdin);
1940 if (!s && feof(stdin))
1941 die("EOF in data (%lu bytes remaining)",
1942 (unsigned long)(length - n));
1943 n += s;
1944 }
1945 }
1946
1947 skip_optional_lf();
1948 return 1;
1949 }
1950
1951 static int validate_raw_date(const char *src, char *result, int maxlen)
1952 {
1953 const char *orig_src = src;
1954 char *endp;
1955 unsigned long num;
1956
1957 errno = 0;
1958
1959 num = strtoul(src, &endp, 10);
1960 /* NEEDSWORK: perhaps check for reasonable values? */
1961 if (errno || endp == src || *endp != ' ')
1962 return -1;
1963
1964 src = endp + 1;
1965 if (*src != '-' && *src != '+')
1966 return -1;
1967
1968 num = strtoul(src + 1, &endp, 10);
1969 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1970 1400 < num)
1971 return -1;
1972
1973 strcpy(result, orig_src);
1974 return 0;
1975 }
1976
1977 static char *parse_ident(const char *buf)
1978 {
1979 const char *gt;
1980 size_t name_len;
1981 char *ident;
1982
1983 gt = strrchr(buf, '>');
1984 if (!gt)
1985 die("Missing > in ident string: %s", buf);
1986 gt++;
1987 if (*gt != ' ')
1988 die("Missing space after > in ident string: %s", buf);
1989 gt++;
1990 name_len = gt - buf;
1991 ident = xmalloc(name_len + 24);
1992 strncpy(ident, buf, name_len);
1993
1994 switch (whenspec) {
1995 case WHENSPEC_RAW:
1996 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1997 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1998 break;
1999 case WHENSPEC_RFC2822:
2000 if (parse_date(gt, ident + name_len, 24) < 0)
2001 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
2002 break;
2003 case WHENSPEC_NOW:
2004 if (strcmp("now", gt))
2005 die("Date in ident must be 'now': %s", buf);
2006 datestamp(ident + name_len, 24);
2007 break;
2008 }
2009
2010 return ident;
2011 }
2012
2013 static void parse_and_store_blob(
2014 struct last_object *last,
2015 unsigned char *sha1out,
2016 uintmax_t mark)
2017 {
2018 static struct strbuf buf = STRBUF_INIT;
2019 uintmax_t len;
2020
2021 if (parse_data(&buf, big_file_threshold, &len))
2022 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2023 else {
2024 if (last) {
2025 strbuf_release(&last->data);
2026 last->offset = 0;
2027 last->depth = 0;
2028 }
2029 stream_blob(len, sha1out, mark);
2030 skip_optional_lf();
2031 }
2032 }
2033
2034 static void parse_new_blob(void)
2035 {
2036 read_next_command();
2037 parse_mark();
2038 parse_and_store_blob(&last_blob, NULL, next_mark);
2039 }
2040
2041 static void unload_one_branch(void)
2042 {
2043 while (cur_active_branches
2044 && cur_active_branches >= max_active_branches) {
2045 uintmax_t min_commit = ULONG_MAX;
2046 struct branch *e, *l = NULL, *p = NULL;
2047
2048 for (e = active_branches; e; e = e->active_next_branch) {
2049 if (e->last_commit < min_commit) {
2050 p = l;
2051 min_commit = e->last_commit;
2052 }
2053 l = e;
2054 }
2055
2056 if (p) {
2057 e = p->active_next_branch;
2058 p->active_next_branch = e->active_next_branch;
2059 } else {
2060 e = active_branches;
2061 active_branches = e->active_next_branch;
2062 }
2063 e->active = 0;
2064 e->active_next_branch = NULL;
2065 if (e->branch_tree.tree) {
2066 release_tree_content_recursive(e->branch_tree.tree);
2067 e->branch_tree.tree = NULL;
2068 }
2069 cur_active_branches--;
2070 }
2071 }
2072
2073 static void load_branch(struct branch *b)
2074 {
2075 load_tree(&b->branch_tree);
2076 if (!b->active) {
2077 b->active = 1;
2078 b->active_next_branch = active_branches;
2079 active_branches = b;
2080 cur_active_branches++;
2081 branch_load_count++;
2082 }
2083 }
2084
2085 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2086 {
2087 unsigned char fanout = 0;
2088 while ((num_notes >>= 8))
2089 fanout++;
2090 return fanout;
2091 }
2092
2093 static void construct_path_with_fanout(const char *hex_sha1,
2094 unsigned char fanout, char *path)
2095 {
2096 unsigned int i = 0, j = 0;
2097 if (fanout >= 20)
2098 die("Too large fanout (%u)", fanout);
2099 while (fanout) {
2100 path[i++] = hex_sha1[j++];
2101 path[i++] = hex_sha1[j++];
2102 path[i++] = '/';
2103 fanout--;
2104 }
2105 memcpy(path + i, hex_sha1 + j, 40 - j);
2106 path[i + 40 - j] = '\0';
2107 }
2108
2109 static uintmax_t do_change_note_fanout(
2110 struct tree_entry *orig_root, struct tree_entry *root,
2111 char *hex_sha1, unsigned int hex_sha1_len,
2112 char *fullpath, unsigned int fullpath_len,
2113 unsigned char fanout)
2114 {
2115 struct tree_content *t = root->tree;
2116 struct tree_entry *e, leaf;
2117 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2118 uintmax_t num_notes = 0;
2119 unsigned char sha1[20];
2120 char realpath[60];
2121
2122 for (i = 0; t && i < t->entry_count; i++) {
2123 e = t->entries[i];
2124 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2125 tmp_fullpath_len = fullpath_len;
2126
2127 /*
2128 * We're interested in EITHER existing note entries (entries
2129 * with exactly 40 hex chars in path, not including directory
2130 * separators), OR directory entries that may contain note
2131 * entries (with < 40 hex chars in path).
2132 * Also, each path component in a note entry must be a multiple
2133 * of 2 chars.
2134 */
2135 if (!e->versions[1].mode ||
2136 tmp_hex_sha1_len > 40 ||
2137 e->name->str_len % 2)
2138 continue;
2139
2140 /* This _may_ be a note entry, or a subdir containing notes */
2141 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2142 e->name->str_len);
2143 if (tmp_fullpath_len)
2144 fullpath[tmp_fullpath_len++] = '/';
2145 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2146 e->name->str_len);
2147 tmp_fullpath_len += e->name->str_len;
2148 fullpath[tmp_fullpath_len] = '\0';
2149
2150 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2151 /* This is a note entry */
2152 construct_path_with_fanout(hex_sha1, fanout, realpath);
2153 if (!strcmp(fullpath, realpath)) {
2154 /* Note entry is in correct location */
2155 num_notes++;
2156 continue;
2157 }
2158
2159 /* Rename fullpath to realpath */
2160 if (!tree_content_remove(orig_root, fullpath, &leaf))
2161 die("Failed to remove path %s", fullpath);
2162 tree_content_set(orig_root, realpath,
2163 leaf.versions[1].sha1,
2164 leaf.versions[1].mode,
2165 leaf.tree);
2166 } else if (S_ISDIR(e->versions[1].mode)) {
2167 /* This is a subdir that may contain note entries */
2168 if (!e->tree)
2169 load_tree(e);
2170 num_notes += do_change_note_fanout(orig_root, e,
2171 hex_sha1, tmp_hex_sha1_len,
2172 fullpath, tmp_fullpath_len, fanout);
2173 }
2174
2175 /* The above may have reallocated the current tree_content */
2176 t = root->tree;
2177 }
2178 return num_notes;
2179 }
2180
2181 static uintmax_t change_note_fanout(struct tree_entry *root,
2182 unsigned char fanout)
2183 {
2184 char hex_sha1[40], path[60];
2185 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2186 }
2187
2188 static void file_change_m(struct branch *b)
2189 {
2190 const char *p = command_buf.buf + 2;
2191 static struct strbuf uq = STRBUF_INIT;
2192 const char *endp;
2193 struct object_entry *oe = oe;
2194 unsigned char sha1[20];
2195 uint16_t mode, inline_data = 0;
2196
2197 p = get_mode(p, &mode);
2198 if (!p)
2199 die("Corrupt mode: %s", command_buf.buf);
2200 switch (mode) {
2201 case 0644:
2202 case 0755:
2203 mode |= S_IFREG;
2204 case S_IFREG | 0644:
2205 case S_IFREG | 0755:
2206 case S_IFLNK:
2207 case S_IFDIR:
2208 case S_IFGITLINK:
2209 /* ok */
2210 break;
2211 default:
2212 die("Corrupt mode: %s", command_buf.buf);
2213 }
2214
2215 if (*p == ':') {
2216 char *x;
2217 oe = find_mark(strtoumax(p + 1, &x, 10));
2218 hashcpy(sha1, oe->idx.sha1);
2219 p = x;
2220 } else if (!prefixcmp(p, "inline")) {
2221 inline_data = 1;
2222 p += 6;
2223 } else {
2224 if (get_sha1_hex(p, sha1))
2225 die("Invalid SHA1: %s", command_buf.buf);
2226 oe = find_object(sha1);
2227 p += 40;
2228 }
2229 if (*p++ != ' ')
2230 die("Missing space after SHA1: %s", command_buf.buf);
2231
2232 strbuf_reset(&uq);
2233 if (!unquote_c_style(&uq, p, &endp)) {
2234 if (*endp)
2235 die("Garbage after path in: %s", command_buf.buf);
2236 p = uq.buf;
2237 }
2238
2239 /* Git does not track empty, non-toplevel directories. */
2240 if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
2241 tree_content_remove(&b->branch_tree, p, NULL);
2242 return;
2243 }
2244
2245 if (S_ISGITLINK(mode)) {
2246 if (inline_data)
2247 die("Git links cannot be specified 'inline': %s",
2248 command_buf.buf);
2249 else if (oe) {
2250 if (oe->type != OBJ_COMMIT)
2251 die("Not a commit (actually a %s): %s",
2252 typename(oe->type), command_buf.buf);
2253 }
2254 /*
2255 * Accept the sha1 without checking; it expected to be in
2256 * another repository.
2257 */
2258 } else if (inline_data) {
2259 if (S_ISDIR(mode))
2260 die("Directories cannot be specified 'inline': %s",
2261 command_buf.buf);
2262 if (p != uq.buf) {
2263 strbuf_addstr(&uq, p);
2264 p = uq.buf;
2265 }
2266 read_next_command();
2267 parse_and_store_blob(&last_blob, sha1, 0);
2268 } else {
2269 enum object_type expected = S_ISDIR(mode) ?
2270 OBJ_TREE: OBJ_BLOB;
2271 enum object_type type = oe ? oe->type :
2272 sha1_object_info(sha1, NULL);
2273 if (type < 0)
2274 die("%s not found: %s",
2275 S_ISDIR(mode) ? "Tree" : "Blob",
2276 command_buf.buf);
2277 if (type != expected)
2278 die("Not a %s (actually a %s): %s",
2279 typename(expected), typename(type),
2280 command_buf.buf);
2281 }
2282
2283 if (!*p) {
2284 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2285 return;
2286 }
2287 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2288 }
2289
2290 static void file_change_d(struct branch *b)
2291 {
2292 const char *p = command_buf.buf + 2;
2293 static struct strbuf uq = STRBUF_INIT;
2294 const char *endp;
2295
2296 strbuf_reset(&uq);
2297 if (!unquote_c_style(&uq, p, &endp)) {
2298 if (*endp)
2299 die("Garbage after path in: %s", command_buf.buf);
2300 p = uq.buf;
2301 }
2302 tree_content_remove(&b->branch_tree, p, NULL);
2303 }
2304
2305 static void file_change_cr(struct branch *b, int rename)
2306 {
2307 const char *s, *d;
2308 static struct strbuf s_uq = STRBUF_INIT;
2309 static struct strbuf d_uq = STRBUF_INIT;
2310 const char *endp;
2311 struct tree_entry leaf;
2312
2313 s = command_buf.buf + 2;
2314 strbuf_reset(&s_uq);
2315 if (!unquote_c_style(&s_uq, s, &endp)) {
2316 if (*endp != ' ')
2317 die("Missing space after source: %s", command_buf.buf);
2318 } else {
2319 endp = strchr(s, ' ');
2320 if (!endp)
2321 die("Missing space after source: %s", command_buf.buf);
2322 strbuf_add(&s_uq, s, endp - s);
2323 }
2324 s = s_uq.buf;
2325
2326 endp++;
2327 if (!*endp)
2328 die("Missing dest: %s", command_buf.buf);
2329
2330 d = endp;
2331 strbuf_reset(&d_uq);
2332 if (!unquote_c_style(&d_uq, d, &endp)) {
2333 if (*endp)
2334 die("Garbage after dest in: %s", command_buf.buf);
2335 d = d_uq.buf;
2336 }
2337
2338 memset(&leaf, 0, sizeof(leaf));
2339 if (rename)
2340 tree_content_remove(&b->branch_tree, s, &leaf);
2341 else
2342 tree_content_get(&b->branch_tree, s, &leaf);
2343 if (!leaf.versions[1].mode)
2344 die("Path %s not in branch", s);
2345 if (!*d) { /* C "path/to/subdir" "" */
2346 tree_content_replace(&b->branch_tree,
2347 leaf.versions[1].sha1,
2348 leaf.versions[1].mode,
2349 leaf.tree);
2350 return;
2351 }
2352 tree_content_set(&b->branch_tree, d,
2353 leaf.versions[1].sha1,
2354 leaf.versions[1].mode,
2355 leaf.tree);
2356 }
2357
2358 static void note_change_n(struct branch *b, unsigned char old_fanout)
2359 {
2360 const char *p = command_buf.buf + 2;
2361 static struct strbuf uq = STRBUF_INIT;
2362 struct object_entry *oe = oe;
2363 struct branch *s;
2364 unsigned char sha1[20], commit_sha1[20];
2365 char path[60];
2366 uint16_t inline_data = 0;
2367 unsigned char new_fanout;
2368
2369 /* <dataref> or 'inline' */
2370 if (*p == ':') {
2371 char *x;
2372 oe = find_mark(strtoumax(p + 1, &x, 10));
2373 hashcpy(sha1, oe->idx.sha1);
2374 p = x;
2375 } else if (!prefixcmp(p, "inline")) {
2376 inline_data = 1;
2377 p += 6;
2378 } else {
2379 if (get_sha1_hex(p, sha1))
2380 die("Invalid SHA1: %s", command_buf.buf);
2381 oe = find_object(sha1);
2382 p += 40;
2383 }
2384 if (*p++ != ' ')
2385 die("Missing space after SHA1: %s", command_buf.buf);
2386
2387 /* <committish> */
2388 s = lookup_branch(p);
2389 if (s) {
2390 hashcpy(commit_sha1, s->sha1);
2391 } else if (*p == ':') {
2392 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2393 struct object_entry *commit_oe = find_mark(commit_mark);
2394 if (commit_oe->type != OBJ_COMMIT)
2395 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2396 hashcpy(commit_sha1, commit_oe->idx.sha1);
2397 } else if (!get_sha1(p, commit_sha1)) {
2398 unsigned long size;
2399 char *buf = read_object_with_reference(commit_sha1,
2400 commit_type, &size, commit_sha1);
2401 if (!buf || size < 46)
2402 die("Not a valid commit: %s", p);
2403 free(buf);
2404 } else
2405 die("Invalid ref name or SHA1 expression: %s", p);
2406
2407 if (inline_data) {
2408 if (p != uq.buf) {
2409 strbuf_addstr(&uq, p);
2410 p = uq.buf;
2411 }
2412 read_next_command();
2413 parse_and_store_blob(&last_blob, sha1, 0);
2414 } else if (oe) {
2415 if (oe->type != OBJ_BLOB)
2416 die("Not a blob (actually a %s): %s",
2417 typename(oe->type), command_buf.buf);
2418 } else if (!is_null_sha1(sha1)) {
2419 enum object_type type = sha1_object_info(sha1, NULL);
2420 if (type < 0)
2421 die("Blob not found: %s", command_buf.buf);
2422 if (type != OBJ_BLOB)
2423 die("Not a blob (actually a %s): %s",
2424 typename(type), command_buf.buf);
2425 }
2426
2427 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2428 if (tree_content_remove(&b->branch_tree, path, NULL))
2429 b->num_notes--;
2430
2431 if (is_null_sha1(sha1))
2432 return; /* nothing to insert */
2433
2434 b->num_notes++;
2435 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2436 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2437 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2438 }
2439
2440 static void file_change_deleteall(struct branch *b)
2441 {
2442 release_tree_content_recursive(b->branch_tree.tree);
2443 hashclr(b->branch_tree.versions[0].sha1);
2444 hashclr(b->branch_tree.versions[1].sha1);
2445 load_tree(&b->branch_tree);
2446 b->num_notes = 0;
2447 }
2448
2449 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2450 {
2451 if (!buf || size < 46)
2452 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2453 if (memcmp("tree ", buf, 5)
2454 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2455 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2456 hashcpy(b->branch_tree.versions[0].sha1,
2457 b->branch_tree.versions[1].sha1);
2458 }
2459
2460 static void parse_from_existing(struct branch *b)
2461 {
2462 if (is_null_sha1(b->sha1)) {
2463 hashclr(b->branch_tree.versions[0].sha1);
2464 hashclr(b->branch_tree.versions[1].sha1);
2465 } else {
2466 unsigned long size;
2467 char *buf;
2468
2469 buf = read_object_with_reference(b->sha1,
2470 commit_type, &size, b->sha1);
2471 parse_from_commit(b, buf, size);
2472 free(buf);
2473 }
2474 }
2475
2476 static int parse_from(struct branch *b)
2477 {
2478 const char *from;
2479 struct branch *s;
2480
2481 if (prefixcmp(command_buf.buf, "from "))
2482 return 0;
2483
2484 if (b->branch_tree.tree) {
2485 release_tree_content_recursive(b->branch_tree.tree);
2486 b->branch_tree.tree = NULL;
2487 }
2488
2489 from = strchr(command_buf.buf, ' ') + 1;
2490 s = lookup_branch(from);
2491 if (b == s)
2492 die("Can't create a branch from itself: %s", b->name);
2493 else if (s) {
2494 unsigned char *t = s->branch_tree.versions[1].sha1;
2495 hashcpy(b->sha1, s->sha1);
2496 hashcpy(b->branch_tree.versions[0].sha1, t);
2497 hashcpy(b->branch_tree.versions[1].sha1, t);
2498 } else if (*from == ':') {
2499 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2500 struct object_entry *oe = find_mark(idnum);
2501 if (oe->type != OBJ_COMMIT)
2502 die("Mark :%" PRIuMAX " not a commit", idnum);
2503 hashcpy(b->sha1, oe->idx.sha1);
2504 if (oe->pack_id != MAX_PACK_ID) {
2505 unsigned long size;
2506 char *buf = gfi_unpack_entry(oe, &size);
2507 parse_from_commit(b, buf, size);
2508 free(buf);
2509 } else
2510 parse_from_existing(b);
2511 } else if (!get_sha1(from, b->sha1))
2512 parse_from_existing(b);
2513 else
2514 die("Invalid ref name or SHA1 expression: %s", from);
2515
2516 read_next_command();
2517 return 1;
2518 }
2519
2520 static struct hash_list *parse_merge(unsigned int *count)
2521 {
2522 struct hash_list *list = NULL, *n, *e = e;
2523 const char *from;
2524 struct branch *s;
2525
2526 *count = 0;
2527 while (!prefixcmp(command_buf.buf, "merge ")) {
2528 from = strchr(command_buf.buf, ' ') + 1;
2529 n = xmalloc(sizeof(*n));
2530 s = lookup_branch(from);
2531 if (s)
2532 hashcpy(n->sha1, s->sha1);
2533 else if (*from == ':') {
2534 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2535 struct object_entry *oe = find_mark(idnum);
2536 if (oe->type != OBJ_COMMIT)
2537 die("Mark :%" PRIuMAX " not a commit", idnum);
2538 hashcpy(n->sha1, oe->idx.sha1);
2539 } else if (!get_sha1(from, n->sha1)) {
2540 unsigned long size;
2541 char *buf = read_object_with_reference(n->sha1,
2542 commit_type, &size, n->sha1);
2543 if (!buf || size < 46)
2544 die("Not a valid commit: %s", from);
2545 free(buf);
2546 } else
2547 die("Invalid ref name or SHA1 expression: %s", from);
2548
2549 n->next = NULL;
2550 if (list)
2551 e->next = n;
2552 else
2553 list = n;
2554 e = n;
2555 (*count)++;
2556 read_next_command();
2557 }
2558 return list;
2559 }
2560
2561 static void parse_new_commit(void)
2562 {
2563 static struct strbuf msg = STRBUF_INIT;
2564 struct branch *b;
2565 char *sp;
2566 char *author = NULL;
2567 char *committer = NULL;
2568 struct hash_list *merge_list = NULL;
2569 unsigned int merge_count;
2570 unsigned char prev_fanout, new_fanout;
2571
2572 /* Obtain the branch name from the rest of our command */
2573 sp = strchr(command_buf.buf, ' ') + 1;
2574 b = lookup_branch(sp);
2575 if (!b)
2576 b = new_branch(sp);
2577
2578 read_next_command();
2579 parse_mark();
2580 if (!prefixcmp(command_buf.buf, "author ")) {
2581 author = parse_ident(command_buf.buf + 7);
2582 read_next_command();
2583 }
2584 if (!prefixcmp(command_buf.buf, "committer ")) {
2585 committer = parse_ident(command_buf.buf + 10);
2586 read_next_command();
2587 }
2588 if (!committer)
2589 die("Expected committer but didn't get one");
2590 parse_data(&msg, 0, NULL);
2591 read_next_command();
2592 parse_from(b);
2593 merge_list = parse_merge(&merge_count);
2594
2595 /* ensure the branch is active/loaded */
2596 if (!b->branch_tree.tree || !max_active_branches) {
2597 unload_one_branch();
2598 load_branch(b);
2599 }
2600
2601 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2602
2603 /* file_change* */
2604 while (command_buf.len > 0) {
2605 if (!prefixcmp(command_buf.buf, "M "))
2606 file_change_m(b);
2607 else if (!prefixcmp(command_buf.buf, "D "))
2608 file_change_d(b);
2609 else if (!prefixcmp(command_buf.buf, "R "))
2610 file_change_cr(b, 1);
2611 else if (!prefixcmp(command_buf.buf, "C "))
2612 file_change_cr(b, 0);
2613 else if (!prefixcmp(command_buf.buf, "N "))
2614 note_change_n(b, prev_fanout);
2615 else if (!strcmp("deleteall", command_buf.buf))
2616 file_change_deleteall(b);
2617 else {
2618 unread_command_buf = 1;
2619 break;
2620 }
2621 if (read_next_command() == EOF)
2622 break;
2623 }
2624
2625 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2626 if (new_fanout != prev_fanout)
2627 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2628
2629 /* build the tree and the commit */
2630 store_tree(&b->branch_tree);
2631 hashcpy(b->branch_tree.versions[0].sha1,
2632 b->branch_tree.versions[1].sha1);
2633
2634 strbuf_reset(&new_data);
2635 strbuf_addf(&new_data, "tree %s\n",
2636 sha1_to_hex(b->branch_tree.versions[1].sha1));
2637 if (!is_null_sha1(b->sha1))
2638 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2639 while (merge_list) {
2640 struct hash_list *next = merge_list->next;
2641 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2642 free(merge_list);
2643 merge_list = next;
2644 }
2645 strbuf_addf(&new_data,
2646 "author %s\n"
2647 "committer %s\n"
2648 "\n",
2649 author ? author : committer, committer);
2650 strbuf_addbuf(&new_data, &msg);
2651 free(author);
2652 free(committer);
2653
2654 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2655 b->pack_id = pack_id;
2656 b->last_commit = object_count_by_type[OBJ_COMMIT];
2657 }
2658
2659 static void parse_new_tag(void)
2660 {
2661 static struct strbuf msg = STRBUF_INIT;
2662 char *sp;
2663 const char *from;
2664 char *tagger;
2665 struct branch *s;
2666 struct tag *t;
2667 uintmax_t from_mark = 0;
2668 unsigned char sha1[20];
2669 enum object_type type;
2670
2671 /* Obtain the new tag name from the rest of our command */
2672 sp = strchr(command_buf.buf, ' ') + 1;
2673 t = pool_alloc(sizeof(struct tag));
2674 t->next_tag = NULL;
2675 t->name = pool_strdup(sp);
2676 if (last_tag)
2677 last_tag->next_tag = t;
2678 else
2679 first_tag = t;
2680 last_tag = t;
2681 read_next_command();
2682
2683 /* from ... */
2684 if (prefixcmp(command_buf.buf, "from "))
2685 die("Expected from command, got %s", command_buf.buf);
2686 from = strchr(command_buf.buf, ' ') + 1;
2687 s = lookup_branch(from);
2688 if (s) {
2689 hashcpy(sha1, s->sha1);
2690 type = OBJ_COMMIT;
2691 } else if (*from == ':') {
2692 struct object_entry *oe;
2693 from_mark = strtoumax(from + 1, NULL, 10);
2694 oe = find_mark(from_mark);
2695 type = oe->type;
2696 hashcpy(sha1, oe->idx.sha1);
2697 } else if (!get_sha1(from, sha1)) {
2698 unsigned long size;
2699 char *buf;
2700
2701 buf = read_sha1_file(sha1, &type, &size);
2702 if (!buf || size < 46)
2703 die("Not a valid commit: %s", from);
2704 free(buf);
2705 } else
2706 die("Invalid ref name or SHA1 expression: %s", from);
2707 read_next_command();
2708
2709 /* tagger ... */
2710 if (!prefixcmp(command_buf.buf, "tagger ")) {
2711 tagger = parse_ident(command_buf.buf + 7);
2712 read_next_command();
2713 } else
2714 tagger = NULL;
2715
2716 /* tag payload/message */
2717 parse_data(&msg, 0, NULL);
2718
2719 /* build the tag object */
2720 strbuf_reset(&new_data);
2721
2722 strbuf_addf(&new_data,
2723 "object %s\n"
2724 "type %s\n"
2725 "tag %s\n",
2726 sha1_to_hex(sha1), typename(type), t->name);
2727 if (tagger)
2728 strbuf_addf(&new_data,
2729 "tagger %s\n", tagger);
2730 strbuf_addch(&new_data, '\n');
2731 strbuf_addbuf(&new_data, &msg);
2732 free(tagger);
2733
2734 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2735 t->pack_id = MAX_PACK_ID;
2736 else
2737 t->pack_id = pack_id;
2738 }
2739
2740 static void parse_reset_branch(void)
2741 {
2742 struct branch *b;
2743 char *sp;
2744
2745 /* Obtain the branch name from the rest of our command */
2746 sp = strchr(command_buf.buf, ' ') + 1;
2747 b = lookup_branch(sp);
2748 if (b) {
2749 hashclr(b->sha1);
2750 hashclr(b->branch_tree.versions[0].sha1);
2751 hashclr(b->branch_tree.versions[1].sha1);
2752 if (b->branch_tree.tree) {
2753 release_tree_content_recursive(b->branch_tree.tree);
2754 b->branch_tree.tree = NULL;
2755 }
2756 }
2757 else
2758 b = new_branch(sp);
2759 read_next_command();
2760 parse_from(b);
2761 if (command_buf.len > 0)
2762 unread_command_buf = 1;
2763 }
2764
2765 static void cat_blob_write(const char *buf, unsigned long size)
2766 {
2767 if (write_in_full(cat_blob_fd, buf, size) != size)
2768 die_errno("Write to frontend failed");
2769 }
2770
2771 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2772 {
2773 struct strbuf line = STRBUF_INIT;
2774 unsigned long size;
2775 enum object_type type = 0;
2776 char *buf;
2777
2778 if (!oe || oe->pack_id == MAX_PACK_ID) {
2779 buf = read_sha1_file(sha1, &type, &size);
2780 } else {
2781 type = oe->type;
2782 buf = gfi_unpack_entry(oe, &size);
2783 }
2784
2785 /*
2786 * Output based on batch_one_object() from cat-file.c.
2787 */
2788 if (type <= 0) {
2789 strbuf_reset(&line);
2790 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2791 cat_blob_write(line.buf, line.len);
2792 strbuf_release(&line);
2793 free(buf);
2794 return;
2795 }
2796 if (!buf)
2797 die("Can't read object %s", sha1_to_hex(sha1));
2798 if (type != OBJ_BLOB)
2799 die("Object %s is a %s but a blob was expected.",
2800 sha1_to_hex(sha1), typename(type));
2801 strbuf_reset(&line);
2802 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2803 typename(type), size);
2804 cat_blob_write(line.buf, line.len);
2805 strbuf_release(&line);
2806 cat_blob_write(buf, size);
2807 cat_blob_write("\n", 1);
2808 free(buf);
2809 }
2810
2811 static void parse_cat_blob(void)
2812 {
2813 const char *p;
2814 struct object_entry *oe = oe;
2815 unsigned char sha1[20];
2816
2817 /* cat-blob SP <object> LF */
2818 p = command_buf.buf + strlen("cat-blob ");
2819 if (*p == ':') {
2820 char *x;
2821 oe = find_mark(strtoumax(p + 1, &x, 10));
2822 if (x == p + 1)
2823 die("Invalid mark: %s", command_buf.buf);
2824 if (!oe)
2825 die("Unknown mark: %s", command_buf.buf);
2826 if (*x)
2827 die("Garbage after mark: %s", command_buf.buf);
2828 hashcpy(sha1, oe->idx.sha1);
2829 } else {
2830 if (get_sha1_hex(p, sha1))
2831 die("Invalid SHA1: %s", command_buf.buf);
2832 if (p[40])
2833 die("Garbage after SHA1: %s", command_buf.buf);
2834 oe = find_object(sha1);
2835 }
2836
2837 cat_blob(oe, sha1);
2838 }
2839
2840 static void checkpoint(void)
2841 {
2842 checkpoint_requested = 0;
2843 if (object_count) {
2844 cycle_packfile();
2845 dump_branches();
2846 dump_tags();
2847 dump_marks();
2848 }
2849 }
2850
2851 static void parse_checkpoint(void)
2852 {
2853 checkpoint_requested = 1;
2854 skip_optional_lf();
2855 }
2856
2857 static void parse_progress(void)
2858 {
2859 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2860 fputc('\n', stdout);
2861 fflush(stdout);
2862 skip_optional_lf();
2863 }
2864
2865 static char* make_fast_import_path(const char *path)
2866 {
2867 struct strbuf abs_path = STRBUF_INIT;
2868
2869 if (!relative_marks_paths || is_absolute_path(path))
2870 return xstrdup(path);
2871 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
2872 return strbuf_detach(&abs_path, NULL);
2873 }
2874
2875 static void option_import_marks(const char *marks,
2876 int from_stream, int ignore_missing)
2877 {
2878 if (import_marks_file) {
2879 if (from_stream)
2880 die("Only one import-marks command allowed per stream");
2881
2882 /* read previous mark file */
2883 if(!import_marks_file_from_stream)
2884 read_marks();
2885 }
2886
2887 import_marks_file = make_fast_import_path(marks);
2888 safe_create_leading_directories_const(import_marks_file);
2889 import_marks_file_from_stream = from_stream;
2890 import_marks_file_ignore_missing = ignore_missing;
2891 }
2892
2893 static void option_date_format(const char *fmt)
2894 {
2895 if (!strcmp(fmt, "raw"))
2896 whenspec = WHENSPEC_RAW;
2897 else if (!strcmp(fmt, "rfc2822"))
2898 whenspec = WHENSPEC_RFC2822;
2899 else if (!strcmp(fmt, "now"))
2900 whenspec = WHENSPEC_NOW;
2901 else
2902 die("unknown --date-format argument %s", fmt);
2903 }
2904
2905 static unsigned long ulong_arg(const char *option, const char *arg)
2906 {
2907 char *endptr;
2908 unsigned long rv = strtoul(arg, &endptr, 0);
2909 if (strchr(arg, '-') || endptr == arg || *endptr)
2910 die("%s: argument must be a non-negative integer", option);
2911 return rv;
2912 }
2913
2914 static void option_depth(const char *depth)
2915 {
2916 max_depth = ulong_arg("--depth", depth);
2917 if (max_depth > MAX_DEPTH)
2918 die("--depth cannot exceed %u", MAX_DEPTH);
2919 }
2920
2921 static void option_active_branches(const char *branches)
2922 {
2923 max_active_branches = ulong_arg("--active-branches", branches);
2924 }
2925
2926 static void option_export_marks(const char *marks)
2927 {
2928 export_marks_file = make_fast_import_path(marks);
2929 safe_create_leading_directories_const(export_marks_file);
2930 }
2931
2932 static void option_cat_blob_fd(const char *fd)
2933 {
2934 unsigned long n = ulong_arg("--cat-blob-fd", fd);
2935 if (n > (unsigned long) INT_MAX)
2936 die("--cat-blob-fd cannot exceed %d", INT_MAX);
2937 cat_blob_fd = (int) n;
2938 }
2939
2940 static void option_export_pack_edges(const char *edges)
2941 {
2942 if (pack_edges)
2943 fclose(pack_edges);
2944 pack_edges = fopen(edges, "a");
2945 if (!pack_edges)
2946 die_errno("Cannot open '%s'", edges);
2947 }
2948
2949 static int parse_one_option(const char *option)
2950 {
2951 if (!prefixcmp(option, "max-pack-size=")) {
2952 unsigned long v;
2953 if (!git_parse_ulong(option + 14, &v))
2954 return 0;
2955 if (v < 8192) {
2956 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
2957 v *= 1024 * 1024;
2958 } else if (v < 1024 * 1024) {
2959 warning("minimum max-pack-size is 1 MiB");
2960 v = 1024 * 1024;
2961 }
2962 max_packsize = v;
2963 } else if (!prefixcmp(option, "big-file-threshold=")) {
2964 unsigned long v;
2965 if (!git_parse_ulong(option + 19, &v))
2966 return 0;
2967 big_file_threshold = v;
2968 } else if (!prefixcmp(option, "depth=")) {
2969 option_depth(option + 6);
2970 } else if (!prefixcmp(option, "active-branches=")) {
2971 option_active_branches(option + 16);
2972 } else if (!prefixcmp(option, "export-pack-edges=")) {
2973 option_export_pack_edges(option + 18);
2974 } else if (!prefixcmp(option, "quiet")) {
2975 show_stats = 0;
2976 } else if (!prefixcmp(option, "stats")) {
2977 show_stats = 1;
2978 } else {
2979 return 0;
2980 }
2981
2982 return 1;
2983 }
2984
2985 static int parse_one_feature(const char *feature, int from_stream)
2986 {
2987 if (!prefixcmp(feature, "date-format=")) {
2988 option_date_format(feature + 12);
2989 } else if (!prefixcmp(feature, "import-marks=")) {
2990 option_import_marks(feature + 13, from_stream, 0);
2991 } else if (!prefixcmp(feature, "import-marks-if-exists=")) {
2992 option_import_marks(feature + strlen("import-marks-if-exists="),
2993 from_stream, 1);
2994 } else if (!prefixcmp(feature, "export-marks=")) {
2995 option_export_marks(feature + 13);
2996 } else if (!strcmp(feature, "cat-blob")) {
2997 ; /* Don't die - this feature is supported */
2998 } else if (!prefixcmp(feature, "relative-marks")) {
2999 relative_marks_paths = 1;
3000 } else if (!prefixcmp(feature, "no-relative-marks")) {
3001 relative_marks_paths = 0;
3002 } else if (!prefixcmp(feature, "force")) {
3003 force_update = 1;
3004 } else if (!strcmp(feature, "notes")) {
3005 ; /* do nothing; we have the feature */
3006 } else {
3007 return 0;
3008 }
3009
3010 return 1;
3011 }
3012
3013 static void parse_feature(void)
3014 {
3015 char *feature = command_buf.buf + 8;
3016
3017 if (seen_data_command)
3018 die("Got feature command '%s' after data command", feature);
3019
3020 if (parse_one_feature(feature, 1))
3021 return;
3022
3023 die("This version of fast-import does not support feature %s.", feature);
3024 }
3025
3026 static void parse_option(void)
3027 {
3028 char *option = command_buf.buf + 11;
3029
3030 if (seen_data_command)
3031 die("Got option command '%s' after data command", option);
3032
3033 if (parse_one_option(option))
3034 return;
3035
3036 die("This version of fast-import does not support option: %s", option);
3037 }
3038
3039 static int git_pack_config(const char *k, const char *v, void *cb)
3040 {
3041 if (!strcmp(k, "pack.depth")) {
3042 max_depth = git_config_int(k, v);
3043 if (max_depth > MAX_DEPTH)
3044 max_depth = MAX_DEPTH;
3045 return 0;
3046 }
3047 if (!strcmp(k, "pack.compression")) {
3048 int level = git_config_int(k, v);
3049 if (level == -1)
3050 level = Z_DEFAULT_COMPRESSION;
3051 else if (level < 0 || level > Z_BEST_COMPRESSION)
3052 die("bad pack compression level %d", level);
3053 pack_compression_level = level;
3054 pack_compression_seen = 1;
3055 return 0;
3056 }
3057 if (!strcmp(k, "pack.indexversion")) {
3058 pack_idx_default_version = git_config_int(k, v);
3059 if (pack_idx_default_version > 2)
3060 die("bad pack.indexversion=%"PRIu32,
3061 pack_idx_default_version);
3062 return 0;
3063 }
3064 if (!strcmp(k, "pack.packsizelimit")) {
3065 max_packsize = git_config_ulong(k, v);
3066 return 0;
3067 }
3068 if (!strcmp(k, "core.bigfilethreshold")) {
3069 long n = git_config_int(k, v);
3070 big_file_threshold = 0 < n ? n : 0;
3071 }
3072 return git_default_config(k, v, cb);
3073 }
3074
3075 static const char fast_import_usage[] =
3076 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3077
3078 static void parse_argv(void)
3079 {
3080 unsigned int i;
3081
3082 for (i = 1; i < global_argc; i++) {
3083 const char *a = global_argv[i];
3084
3085 if (*a != '-' || !strcmp(a, "--"))
3086 break;
3087
3088 if (parse_one_option(a + 2))
3089 continue;
3090
3091 if (parse_one_feature(a + 2, 0))
3092 continue;
3093
3094 if (!prefixcmp(a + 2, "cat-blob-fd=")) {
3095 option_cat_blob_fd(a + 2 + strlen("cat-blob-fd="));
3096 continue;
3097 }
3098
3099 die("unknown option %s", a);
3100 }
3101 if (i != global_argc)
3102 usage(fast_import_usage);
3103
3104 seen_data_command = 1;
3105 if (import_marks_file)
3106 read_marks();
3107 }
3108
3109 int main(int argc, const char **argv)
3110 {
3111 unsigned int i;
3112
3113 git_extract_argv0_path(argv[0]);
3114
3115 if (argc == 2 && !strcmp(argv[1], "-h"))
3116 usage(fast_import_usage);
3117
3118 setup_git_directory();
3119 git_config(git_pack_config, NULL);
3120 if (!pack_compression_seen && core_compression_seen)
3121 pack_compression_level = core_compression_level;
3122
3123 alloc_objects(object_entry_alloc);
3124 strbuf_init(&command_buf, 0);
3125 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3126 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3127 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3128 marks = pool_calloc(1, sizeof(struct mark_set));
3129
3130 global_argc = argc;
3131 global_argv = argv;
3132
3133 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3134 for (i = 0; i < (cmd_save - 1); i++)
3135 rc_free[i].next = &rc_free[i + 1];
3136 rc_free[cmd_save - 1].next = NULL;
3137
3138 prepare_packed_git();
3139 start_packfile();
3140 set_die_routine(die_nicely);
3141 set_checkpoint_signal();
3142 while (read_next_command() != EOF) {
3143 if (!strcmp("blob", command_buf.buf))
3144 parse_new_blob();
3145 else if (!prefixcmp(command_buf.buf, "commit "))
3146 parse_new_commit();
3147 else if (!prefixcmp(command_buf.buf, "tag "))
3148 parse_new_tag();
3149 else if (!prefixcmp(command_buf.buf, "reset "))
3150 parse_reset_branch();
3151 else if (!strcmp("checkpoint", command_buf.buf))
3152 parse_checkpoint();
3153 else if (!prefixcmp(command_buf.buf, "progress "))
3154 parse_progress();
3155 else if (!prefixcmp(command_buf.buf, "feature "))
3156 parse_feature();
3157 else if (!prefixcmp(command_buf.buf, "option git "))
3158 parse_option();
3159 else if (!prefixcmp(command_buf.buf, "option "))
3160 /* ignore non-git options*/;
3161 else
3162 die("Unsupported command: %s", command_buf.buf);
3163
3164 if (checkpoint_requested)
3165 checkpoint();
3166 }
3167
3168 /* argv hasn't been parsed yet, do so */
3169 if (!seen_data_command)
3170 parse_argv();
3171
3172 end_packfile();
3173
3174 dump_branches();
3175 dump_tags();
3176 unkeep_all_packs();
3177 dump_marks();
3178
3179 if (pack_edges)
3180 fclose(pack_edges);
3181
3182 if (show_stats) {
3183 uintmax_t total_count = 0, duplicate_count = 0;
3184 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3185 total_count += object_count_by_type[i];
3186 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3187 duplicate_count += duplicate_count_by_type[i];
3188
3189 fprintf(stderr, "%s statistics:\n", argv[0]);
3190 fprintf(stderr, "---------------------------------------------------------------------\n");
3191 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3192 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3193 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]);
3194 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]);
3195 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]);
3196 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]);
3197 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3198 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3199 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3200 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3201 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3202 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3203 fprintf(stderr, "---------------------------------------------------------------------\n");
3204 pack_report();
3205 fprintf(stderr, "---------------------------------------------------------------------\n");
3206 fprintf(stderr, "\n");
3207 }
3208
3209 return failure ? 1 : 0;
3210 }