packfile: allow install_packed_git to handle arbitrary repositories
[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 commit-ish lf)?
26 ('merge' sp commit-ish lf)*
27 (file_change | ls)*
28 lf?;
29 commit_msg ::= data;
30
31 ls ::= 'ls' sp '"' quoted(path) '"' lf;
32
33 file_change ::= file_clr
34 | file_del
35 | file_rnm
36 | file_cpy
37 | file_obm
38 | file_inm;
39 file_clr ::= 'deleteall' lf;
40 file_del ::= 'D' sp path_str lf;
41 file_rnm ::= 'R' sp path_str sp path_str lf;
42 file_cpy ::= 'C' sp path_str sp path_str lf;
43 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
44 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
45 data;
46 note_obm ::= 'N' sp (hexsha1 | idnum) sp commit-ish lf;
47 note_inm ::= 'N' sp 'inline' sp commit-ish lf
48 data;
49
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp commit-ish lf
52 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
53 tag_msg;
54 tag_msg ::= data;
55
56 reset_branch ::= 'reset' sp ref_str lf
57 ('from' sp commit-ish lf)?
58 lf?;
59
60 checkpoint ::= 'checkpoint' lf
61 lf?;
62
63 progress ::= 'progress' sp not_lf* lf
64 lf?;
65
66 # note: the first idnum in a stream should be 1 and subsequent
67 # idnums should not have gaps between values as this will cause
68 # the stream parser to reserve space for the gapped values. An
69 # idnum can be updated in the future to a new object by issuing
70 # a new mark directive with the old idnum.
71 #
72 mark ::= 'mark' sp idnum lf;
73 data ::= (delimited_data | exact_data)
74 lf?;
75
76 # note: delim may be any string but must not contain lf.
77 # data_line may contain any data but must not be exactly
78 # delim.
79 delimited_data ::= 'data' sp '<<' delim lf
80 (data_line lf)*
81 delim lf;
82
83 # note: declen indicates the length of binary_data in bytes.
84 # declen does not include the lf preceding the binary data.
85 #
86 exact_data ::= 'data' sp declen lf
87 binary_data;
88
89 # note: quoted strings are C-style quoting supporting \c for
90 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
91 # is the signed byte value in octal. Note that the only
92 # characters which must actually be escaped to protect the
93 # stream formatting is: \, " and LF. Otherwise these values
94 # are UTF8.
95 #
96 commit-ish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
97 ref_str ::= ref;
98 sha1exp_str ::= sha1exp;
99 tag_str ::= tag;
100 path_str ::= path | '"' quoted(path) '"' ;
101 mode ::= '100644' | '644'
102 | '100755' | '755'
103 | '120000'
104 ;
105
106 declen ::= # unsigned 32 bit value, ascii base10 notation;
107 bigint ::= # unsigned integer value, ascii base10 notation;
108 binary_data ::= # file content, not interpreted;
109
110 when ::= raw_when | rfc2822_when;
111 raw_when ::= ts sp tz;
112 rfc2822_when ::= # Valid RFC 2822 date and time;
113
114 sp ::= # ASCII space character;
115 lf ::= # ASCII newline (LF) character;
116
117 # note: a colon (':') must precede the numerical value assigned to
118 # an idnum. This is to distinguish it from a ref or tag name as
119 # GIT does not permit ':' in ref or tag strings.
120 #
121 idnum ::= ':' bigint;
122 path ::= # GIT style file path, e.g. "a/b/c";
123 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
124 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
125 sha1exp ::= # Any valid GIT SHA1 expression;
126 hexsha1 ::= # SHA1 in hexadecimal format;
127
128 # note: name and email are UTF8 strings, however name must not
129 # contain '<' or lf and email must not contain any of the
130 # following: '<', '>', lf.
131 #
132 name ::= # valid GIT author/committer name;
133 email ::= # valid GIT author/committer email;
134 ts ::= # time since the epoch in seconds, ascii base10 notation;
135 tz ::= # GIT style timezone;
136
137 # note: comments, get-mark, ls-tree, and cat-blob requests may
138 # appear anywhere in the input, except within a data command. Any
139 # form of the data command always escapes the related input from
140 # comment processing.
141 #
142 # In case it is not clear, the '#' that starts the comment
143 # must be the first character on that line (an lf
144 # preceded it).
145 #
146
147 get_mark ::= 'get-mark' sp idnum lf;
148 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
149 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
150
151 comment ::= '#' not_lf* lf;
152 not_lf ::= # Any byte that is not ASCII newline (LF);
153 */
154
155 #include "builtin.h"
156 #include "cache.h"
157 #include "repository.h"
158 #include "config.h"
159 #include "lockfile.h"
160 #include "object.h"
161 #include "blob.h"
162 #include "tree.h"
163 #include "commit.h"
164 #include "delta.h"
165 #include "pack.h"
166 #include "refs.h"
167 #include "csum-file.h"
168 #include "quote.h"
169 #include "dir.h"
170 #include "run-command.h"
171 #include "packfile.h"
172 #include "object-store.h"
173
174 #define PACK_ID_BITS 16
175 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
176 #define DEPTH_BITS 13
177 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
178
179 /*
180 * We abuse the setuid bit on directories to mean "do not delta".
181 */
182 #define NO_DELTA S_ISUID
183
184 struct object_entry {
185 struct pack_idx_entry idx;
186 struct object_entry *next;
187 uint32_t type : TYPE_BITS,
188 pack_id : PACK_ID_BITS,
189 depth : DEPTH_BITS;
190 };
191
192 struct object_entry_pool {
193 struct object_entry_pool *next_pool;
194 struct object_entry *next_free;
195 struct object_entry *end;
196 struct object_entry entries[FLEX_ARRAY]; /* more */
197 };
198
199 struct mark_set {
200 union {
201 struct object_entry *marked[1024];
202 struct mark_set *sets[1024];
203 } data;
204 unsigned int shift;
205 };
206
207 struct last_object {
208 struct strbuf data;
209 off_t offset;
210 unsigned int depth;
211 unsigned no_swap : 1;
212 };
213
214 struct mem_pool {
215 struct mem_pool *next_pool;
216 char *next_free;
217 char *end;
218 uintmax_t space[FLEX_ARRAY]; /* more */
219 };
220
221 struct atom_str {
222 struct atom_str *next_atom;
223 unsigned short str_len;
224 char str_dat[FLEX_ARRAY]; /* more */
225 };
226
227 struct tree_content;
228 struct tree_entry {
229 struct tree_content *tree;
230 struct atom_str *name;
231 struct tree_entry_ms {
232 uint16_t mode;
233 struct object_id oid;
234 } versions[2];
235 };
236
237 struct tree_content {
238 unsigned int entry_capacity; /* must match avail_tree_content */
239 unsigned int entry_count;
240 unsigned int delta_depth;
241 struct tree_entry *entries[FLEX_ARRAY]; /* more */
242 };
243
244 struct avail_tree_content {
245 unsigned int entry_capacity; /* must match tree_content */
246 struct avail_tree_content *next_avail;
247 };
248
249 struct branch {
250 struct branch *table_next_branch;
251 struct branch *active_next_branch;
252 const char *name;
253 struct tree_entry branch_tree;
254 uintmax_t last_commit;
255 uintmax_t num_notes;
256 unsigned active : 1;
257 unsigned delete : 1;
258 unsigned pack_id : PACK_ID_BITS;
259 struct object_id oid;
260 };
261
262 struct tag {
263 struct tag *next_tag;
264 const char *name;
265 unsigned int pack_id;
266 struct object_id oid;
267 };
268
269 struct hash_list {
270 struct hash_list *next;
271 struct object_id oid;
272 };
273
274 typedef enum {
275 WHENSPEC_RAW = 1,
276 WHENSPEC_RFC2822,
277 WHENSPEC_NOW
278 } whenspec_type;
279
280 struct recent_command {
281 struct recent_command *prev;
282 struct recent_command *next;
283 char *buf;
284 };
285
286 /* Configured limits on output */
287 static unsigned long max_depth = 50;
288 static off_t max_packsize;
289 static int unpack_limit = 100;
290 static int force_update;
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 uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
299 static unsigned long object_count;
300 static unsigned long branch_count;
301 static unsigned long branch_load_count;
302 static int failure;
303 static FILE *pack_edges;
304 static unsigned int show_stats = 1;
305 static int global_argc;
306 static const char **global_argv;
307
308 /* Memory pools */
309 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
310 static size_t total_allocd;
311 static struct mem_pool *mem_pool;
312
313 /* Atom management */
314 static unsigned int atom_table_sz = 4451;
315 static unsigned int atom_cnt;
316 static struct atom_str **atom_table;
317
318 /* The .pack file being generated */
319 static struct pack_idx_option pack_idx_opts;
320 static unsigned int pack_id;
321 static struct sha1file *pack_file;
322 static struct packed_git *pack_data;
323 static struct packed_git **all_packs;
324 static off_t pack_size;
325
326 /* Table of objects we've written. */
327 static unsigned int object_entry_alloc = 5000;
328 static struct object_entry_pool *blocks;
329 static struct object_entry *object_table[1 << 16];
330 static struct mark_set *marks;
331 static const char *export_marks_file;
332 static const char *import_marks_file;
333 static int import_marks_file_from_stream;
334 static int import_marks_file_ignore_missing;
335 static int import_marks_file_done;
336 static int relative_marks_paths;
337
338 /* Our last blob */
339 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
340
341 /* Tree management */
342 static unsigned int tree_entry_alloc = 1000;
343 static void *avail_tree_entry;
344 static unsigned int avail_tree_table_sz = 100;
345 static struct avail_tree_content **avail_tree_table;
346 static struct strbuf old_tree = STRBUF_INIT;
347 static struct strbuf new_tree = STRBUF_INIT;
348
349 /* Branch data */
350 static unsigned long max_active_branches = 5;
351 static unsigned long cur_active_branches;
352 static unsigned long branch_table_sz = 1039;
353 static struct branch **branch_table;
354 static struct branch *active_branches;
355
356 /* Tag data */
357 static struct tag *first_tag;
358 static struct tag *last_tag;
359
360 /* Input stream parsing */
361 static whenspec_type whenspec = WHENSPEC_RAW;
362 static struct strbuf command_buf = STRBUF_INIT;
363 static int unread_command_buf;
364 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
365 static struct recent_command *cmd_tail = &cmd_hist;
366 static struct recent_command *rc_free;
367 static unsigned int cmd_save = 100;
368 static uintmax_t next_mark;
369 static struct strbuf new_data = STRBUF_INIT;
370 static int seen_data_command;
371 static int require_explicit_termination;
372
373 /* Signal handling */
374 static volatile sig_atomic_t checkpoint_requested;
375
376 /* Where to write output of cat-blob commands */
377 static int cat_blob_fd = STDOUT_FILENO;
378
379 static void parse_argv(void);
380 static void parse_get_mark(const char *p);
381 static void parse_cat_blob(const char *p);
382 static void parse_ls(const char *p, struct branch *b);
383
384 static void write_branch_report(FILE *rpt, struct branch *b)
385 {
386 fprintf(rpt, "%s:\n", b->name);
387
388 fprintf(rpt, " status :");
389 if (b->active)
390 fputs(" active", rpt);
391 if (b->branch_tree.tree)
392 fputs(" loaded", rpt);
393 if (is_null_oid(&b->branch_tree.versions[1].oid))
394 fputs(" dirty", rpt);
395 fputc('\n', rpt);
396
397 fprintf(rpt, " tip commit : %s\n", oid_to_hex(&b->oid));
398 fprintf(rpt, " old tree : %s\n",
399 oid_to_hex(&b->branch_tree.versions[0].oid));
400 fprintf(rpt, " cur tree : %s\n",
401 oid_to_hex(&b->branch_tree.versions[1].oid));
402 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
403
404 fputs(" last pack : ", rpt);
405 if (b->pack_id < MAX_PACK_ID)
406 fprintf(rpt, "%u", b->pack_id);
407 fputc('\n', rpt);
408
409 fputc('\n', rpt);
410 }
411
412 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
413
414 static void write_crash_report(const char *err)
415 {
416 char *loc = git_pathdup("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
417 FILE *rpt = fopen(loc, "w");
418 struct branch *b;
419 unsigned long lu;
420 struct recent_command *rc;
421
422 if (!rpt) {
423 error_errno("can't write crash report %s", loc);
424 free(loc);
425 return;
426 }
427
428 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
429
430 fprintf(rpt, "fast-import crash report:\n");
431 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
432 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
433 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_MODE(ISO8601)));
434 fputc('\n', rpt);
435
436 fputs("fatal: ", rpt);
437 fputs(err, rpt);
438 fputc('\n', rpt);
439
440 fputc('\n', rpt);
441 fputs("Most Recent Commands Before Crash\n", rpt);
442 fputs("---------------------------------\n", rpt);
443 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
444 if (rc->next == &cmd_hist)
445 fputs("* ", rpt);
446 else
447 fputs(" ", rpt);
448 fputs(rc->buf, rpt);
449 fputc('\n', rpt);
450 }
451
452 fputc('\n', rpt);
453 fputs("Active Branch LRU\n", rpt);
454 fputs("-----------------\n", rpt);
455 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
456 cur_active_branches,
457 max_active_branches);
458 fputc('\n', rpt);
459 fputs(" pos clock name\n", rpt);
460 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
461 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
462 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
463 ++lu, b->last_commit, b->name);
464
465 fputc('\n', rpt);
466 fputs("Inactive Branches\n", rpt);
467 fputs("-----------------\n", rpt);
468 for (lu = 0; lu < branch_table_sz; lu++) {
469 for (b = branch_table[lu]; b; b = b->table_next_branch)
470 write_branch_report(rpt, b);
471 }
472
473 if (first_tag) {
474 struct tag *tg;
475 fputc('\n', rpt);
476 fputs("Annotated Tags\n", rpt);
477 fputs("--------------\n", rpt);
478 for (tg = first_tag; tg; tg = tg->next_tag) {
479 fputs(oid_to_hex(&tg->oid), rpt);
480 fputc(' ', rpt);
481 fputs(tg->name, rpt);
482 fputc('\n', rpt);
483 }
484 }
485
486 fputc('\n', rpt);
487 fputs("Marks\n", rpt);
488 fputs("-----\n", rpt);
489 if (export_marks_file)
490 fprintf(rpt, " exported to %s\n", export_marks_file);
491 else
492 dump_marks_helper(rpt, 0, marks);
493
494 fputc('\n', rpt);
495 fputs("-------------------\n", rpt);
496 fputs("END OF CRASH REPORT\n", rpt);
497 fclose(rpt);
498 free(loc);
499 }
500
501 static void end_packfile(void);
502 static void unkeep_all_packs(void);
503 static void dump_marks(void);
504
505 static NORETURN void die_nicely(const char *err, va_list params)
506 {
507 static int zombie;
508 char message[2 * PATH_MAX];
509
510 vsnprintf(message, sizeof(message), err, params);
511 fputs("fatal: ", stderr);
512 fputs(message, stderr);
513 fputc('\n', stderr);
514
515 if (!zombie) {
516 zombie = 1;
517 write_crash_report(message);
518 end_packfile();
519 unkeep_all_packs();
520 dump_marks();
521 }
522 exit(128);
523 }
524
525 #ifndef SIGUSR1 /* Windows, for example */
526
527 static void set_checkpoint_signal(void)
528 {
529 }
530
531 #else
532
533 static void checkpoint_signal(int signo)
534 {
535 checkpoint_requested = 1;
536 }
537
538 static void set_checkpoint_signal(void)
539 {
540 struct sigaction sa;
541
542 memset(&sa, 0, sizeof(sa));
543 sa.sa_handler = checkpoint_signal;
544 sigemptyset(&sa.sa_mask);
545 sa.sa_flags = SA_RESTART;
546 sigaction(SIGUSR1, &sa, NULL);
547 }
548
549 #endif
550
551 static void alloc_objects(unsigned int cnt)
552 {
553 struct object_entry_pool *b;
554
555 b = xmalloc(sizeof(struct object_entry_pool)
556 + cnt * sizeof(struct object_entry));
557 b->next_pool = blocks;
558 b->next_free = b->entries;
559 b->end = b->entries + cnt;
560 blocks = b;
561 alloc_count += cnt;
562 }
563
564 static struct object_entry *new_object(struct object_id *oid)
565 {
566 struct object_entry *e;
567
568 if (blocks->next_free == blocks->end)
569 alloc_objects(object_entry_alloc);
570
571 e = blocks->next_free++;
572 oidcpy(&e->idx.oid, oid);
573 return e;
574 }
575
576 static struct object_entry *find_object(struct object_id *oid)
577 {
578 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
579 struct object_entry *e;
580 for (e = object_table[h]; e; e = e->next)
581 if (!oidcmp(oid, &e->idx.oid))
582 return e;
583 return NULL;
584 }
585
586 static struct object_entry *insert_object(struct object_id *oid)
587 {
588 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
589 struct object_entry *e = object_table[h];
590
591 while (e) {
592 if (!oidcmp(oid, &e->idx.oid))
593 return e;
594 e = e->next;
595 }
596
597 e = new_object(oid);
598 e->next = object_table[h];
599 e->idx.offset = 0;
600 object_table[h] = e;
601 return e;
602 }
603
604 static void invalidate_pack_id(unsigned int id)
605 {
606 unsigned int h;
607 unsigned long lu;
608 struct tag *t;
609
610 for (h = 0; h < ARRAY_SIZE(object_table); h++) {
611 struct object_entry *e;
612
613 for (e = object_table[h]; e; e = e->next)
614 if (e->pack_id == id)
615 e->pack_id = MAX_PACK_ID;
616 }
617
618 for (lu = 0; lu < branch_table_sz; lu++) {
619 struct branch *b;
620
621 for (b = branch_table[lu]; b; b = b->table_next_branch)
622 if (b->pack_id == id)
623 b->pack_id = MAX_PACK_ID;
624 }
625
626 for (t = first_tag; t; t = t->next_tag)
627 if (t->pack_id == id)
628 t->pack_id = MAX_PACK_ID;
629 }
630
631 static unsigned int hc_str(const char *s, size_t len)
632 {
633 unsigned int r = 0;
634 while (len-- > 0)
635 r = r * 31 + *s++;
636 return r;
637 }
638
639 static void *pool_alloc(size_t len)
640 {
641 struct mem_pool *p;
642 void *r;
643
644 /* round up to a 'uintmax_t' alignment */
645 if (len & (sizeof(uintmax_t) - 1))
646 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
647
648 for (p = mem_pool; p; p = p->next_pool)
649 if ((p->end - p->next_free >= len))
650 break;
651
652 if (!p) {
653 if (len >= (mem_pool_alloc/2)) {
654 total_allocd += len;
655 return xmalloc(len);
656 }
657 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
658 p = xmalloc(st_add(sizeof(struct mem_pool), mem_pool_alloc));
659 p->next_pool = mem_pool;
660 p->next_free = (char *) p->space;
661 p->end = p->next_free + mem_pool_alloc;
662 mem_pool = p;
663 }
664
665 r = p->next_free;
666 p->next_free += len;
667 return r;
668 }
669
670 static void *pool_calloc(size_t count, size_t size)
671 {
672 size_t len = count * size;
673 void *r = pool_alloc(len);
674 memset(r, 0, len);
675 return r;
676 }
677
678 static char *pool_strdup(const char *s)
679 {
680 size_t len = strlen(s) + 1;
681 char *r = pool_alloc(len);
682 memcpy(r, s, len);
683 return r;
684 }
685
686 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
687 {
688 struct mark_set *s = marks;
689 while ((idnum >> s->shift) >= 1024) {
690 s = pool_calloc(1, sizeof(struct mark_set));
691 s->shift = marks->shift + 10;
692 s->data.sets[0] = marks;
693 marks = s;
694 }
695 while (s->shift) {
696 uintmax_t i = idnum >> s->shift;
697 idnum -= i << s->shift;
698 if (!s->data.sets[i]) {
699 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
700 s->data.sets[i]->shift = s->shift - 10;
701 }
702 s = s->data.sets[i];
703 }
704 if (!s->data.marked[idnum])
705 marks_set_count++;
706 s->data.marked[idnum] = oe;
707 }
708
709 static struct object_entry *find_mark(uintmax_t idnum)
710 {
711 uintmax_t orig_idnum = idnum;
712 struct mark_set *s = marks;
713 struct object_entry *oe = NULL;
714 if ((idnum >> s->shift) < 1024) {
715 while (s && s->shift) {
716 uintmax_t i = idnum >> s->shift;
717 idnum -= i << s->shift;
718 s = s->data.sets[i];
719 }
720 if (s)
721 oe = s->data.marked[idnum];
722 }
723 if (!oe)
724 die("mark :%" PRIuMAX " not declared", orig_idnum);
725 return oe;
726 }
727
728 static struct atom_str *to_atom(const char *s, unsigned short len)
729 {
730 unsigned int hc = hc_str(s, len) % atom_table_sz;
731 struct atom_str *c;
732
733 for (c = atom_table[hc]; c; c = c->next_atom)
734 if (c->str_len == len && !strncmp(s, c->str_dat, len))
735 return c;
736
737 c = pool_alloc(sizeof(struct atom_str) + len + 1);
738 c->str_len = len;
739 memcpy(c->str_dat, s, len);
740 c->str_dat[len] = 0;
741 c->next_atom = atom_table[hc];
742 atom_table[hc] = c;
743 atom_cnt++;
744 return c;
745 }
746
747 static struct branch *lookup_branch(const char *name)
748 {
749 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
750 struct branch *b;
751
752 for (b = branch_table[hc]; b; b = b->table_next_branch)
753 if (!strcmp(name, b->name))
754 return b;
755 return NULL;
756 }
757
758 static struct branch *new_branch(const char *name)
759 {
760 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
761 struct branch *b = lookup_branch(name);
762
763 if (b)
764 die("Invalid attempt to create duplicate branch: %s", name);
765 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
766 die("Branch name doesn't conform to GIT standards: %s", name);
767
768 b = pool_calloc(1, sizeof(struct branch));
769 b->name = pool_strdup(name);
770 b->table_next_branch = branch_table[hc];
771 b->branch_tree.versions[0].mode = S_IFDIR;
772 b->branch_tree.versions[1].mode = S_IFDIR;
773 b->num_notes = 0;
774 b->active = 0;
775 b->pack_id = MAX_PACK_ID;
776 branch_table[hc] = b;
777 branch_count++;
778 return b;
779 }
780
781 static unsigned int hc_entries(unsigned int cnt)
782 {
783 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
784 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
785 }
786
787 static struct tree_content *new_tree_content(unsigned int cnt)
788 {
789 struct avail_tree_content *f, *l = NULL;
790 struct tree_content *t;
791 unsigned int hc = hc_entries(cnt);
792
793 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
794 if (f->entry_capacity >= cnt)
795 break;
796
797 if (f) {
798 if (l)
799 l->next_avail = f->next_avail;
800 else
801 avail_tree_table[hc] = f->next_avail;
802 } else {
803 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
804 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
805 f->entry_capacity = cnt;
806 }
807
808 t = (struct tree_content*)f;
809 t->entry_count = 0;
810 t->delta_depth = 0;
811 return t;
812 }
813
814 static void release_tree_entry(struct tree_entry *e);
815 static void release_tree_content(struct tree_content *t)
816 {
817 struct avail_tree_content *f = (struct avail_tree_content*)t;
818 unsigned int hc = hc_entries(f->entry_capacity);
819 f->next_avail = avail_tree_table[hc];
820 avail_tree_table[hc] = f;
821 }
822
823 static void release_tree_content_recursive(struct tree_content *t)
824 {
825 unsigned int i;
826 for (i = 0; i < t->entry_count; i++)
827 release_tree_entry(t->entries[i]);
828 release_tree_content(t);
829 }
830
831 static struct tree_content *grow_tree_content(
832 struct tree_content *t,
833 int amt)
834 {
835 struct tree_content *r = new_tree_content(t->entry_count + amt);
836 r->entry_count = t->entry_count;
837 r->delta_depth = t->delta_depth;
838 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
839 release_tree_content(t);
840 return r;
841 }
842
843 static struct tree_entry *new_tree_entry(void)
844 {
845 struct tree_entry *e;
846
847 if (!avail_tree_entry) {
848 unsigned int n = tree_entry_alloc;
849 total_allocd += n * sizeof(struct tree_entry);
850 ALLOC_ARRAY(e, n);
851 avail_tree_entry = e;
852 while (n-- > 1) {
853 *((void**)e) = e + 1;
854 e++;
855 }
856 *((void**)e) = NULL;
857 }
858
859 e = avail_tree_entry;
860 avail_tree_entry = *((void**)e);
861 return e;
862 }
863
864 static void release_tree_entry(struct tree_entry *e)
865 {
866 if (e->tree)
867 release_tree_content_recursive(e->tree);
868 *((void**)e) = avail_tree_entry;
869 avail_tree_entry = e;
870 }
871
872 static struct tree_content *dup_tree_content(struct tree_content *s)
873 {
874 struct tree_content *d;
875 struct tree_entry *a, *b;
876 unsigned int i;
877
878 if (!s)
879 return NULL;
880 d = new_tree_content(s->entry_count);
881 for (i = 0; i < s->entry_count; i++) {
882 a = s->entries[i];
883 b = new_tree_entry();
884 memcpy(b, a, sizeof(*a));
885 if (a->tree && is_null_oid(&b->versions[1].oid))
886 b->tree = dup_tree_content(a->tree);
887 else
888 b->tree = NULL;
889 d->entries[i] = b;
890 }
891 d->entry_count = s->entry_count;
892 d->delta_depth = s->delta_depth;
893
894 return d;
895 }
896
897 static void start_packfile(void)
898 {
899 struct strbuf tmp_file = STRBUF_INIT;
900 struct packed_git *p;
901 struct pack_header hdr;
902 int pack_fd;
903
904 pack_fd = odb_mkstemp(&tmp_file, "pack/tmp_pack_XXXXXX");
905 FLEX_ALLOC_STR(p, pack_name, tmp_file.buf);
906 strbuf_release(&tmp_file);
907
908 p->pack_fd = pack_fd;
909 p->do_not_close = 1;
910 pack_file = sha1fd(pack_fd, p->pack_name);
911
912 hdr.hdr_signature = htonl(PACK_SIGNATURE);
913 hdr.hdr_version = htonl(2);
914 hdr.hdr_entries = 0;
915 sha1write(pack_file, &hdr, sizeof(hdr));
916
917 pack_data = p;
918 pack_size = sizeof(hdr);
919 object_count = 0;
920
921 REALLOC_ARRAY(all_packs, pack_id + 1);
922 all_packs[pack_id] = p;
923 }
924
925 static const char *create_index(void)
926 {
927 const char *tmpfile;
928 struct pack_idx_entry **idx, **c, **last;
929 struct object_entry *e;
930 struct object_entry_pool *o;
931
932 /* Build the table of object IDs. */
933 ALLOC_ARRAY(idx, object_count);
934 c = idx;
935 for (o = blocks; o; o = o->next_pool)
936 for (e = o->next_free; e-- != o->entries;)
937 if (pack_id == e->pack_id)
938 *c++ = &e->idx;
939 last = idx + object_count;
940 if (c != last)
941 die("internal consistency error creating the index");
942
943 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
944 free(idx);
945 return tmpfile;
946 }
947
948 static char *keep_pack(const char *curr_index_name)
949 {
950 static const char *keep_msg = "fast-import";
951 struct strbuf name = STRBUF_INIT;
952 int keep_fd;
953
954 odb_pack_name(&name, pack_data->sha1, "keep");
955 keep_fd = odb_pack_keep(name.buf);
956 if (keep_fd < 0)
957 die_errno("cannot create keep file");
958 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
959 if (close(keep_fd))
960 die_errno("failed to write keep file");
961
962 odb_pack_name(&name, pack_data->sha1, "pack");
963 if (finalize_object_file(pack_data->pack_name, name.buf))
964 die("cannot store pack file");
965
966 odb_pack_name(&name, pack_data->sha1, "idx");
967 if (finalize_object_file(curr_index_name, name.buf))
968 die("cannot store index file");
969 free((void *)curr_index_name);
970 return strbuf_detach(&name, NULL);
971 }
972
973 static void unkeep_all_packs(void)
974 {
975 struct strbuf name = STRBUF_INIT;
976 int k;
977
978 for (k = 0; k < pack_id; k++) {
979 struct packed_git *p = all_packs[k];
980 odb_pack_name(&name, p->sha1, "keep");
981 unlink_or_warn(name.buf);
982 }
983 strbuf_release(&name);
984 }
985
986 static int loosen_small_pack(const struct packed_git *p)
987 {
988 struct child_process unpack = CHILD_PROCESS_INIT;
989
990 if (lseek(p->pack_fd, 0, SEEK_SET) < 0)
991 die_errno("Failed seeking to start of '%s'", p->pack_name);
992
993 unpack.in = p->pack_fd;
994 unpack.git_cmd = 1;
995 unpack.stdout_to_stderr = 1;
996 argv_array_push(&unpack.args, "unpack-objects");
997 if (!show_stats)
998 argv_array_push(&unpack.args, "-q");
999
1000 return run_command(&unpack);
1001 }
1002
1003 static void end_packfile(void)
1004 {
1005 static int running;
1006
1007 if (running || !pack_data)
1008 return;
1009
1010 running = 1;
1011 clear_delta_base_cache();
1012 if (object_count) {
1013 struct packed_git *new_p;
1014 struct object_id cur_pack_oid;
1015 char *idx_name;
1016 int i;
1017 struct branch *b;
1018 struct tag *t;
1019
1020 close_pack_windows(pack_data);
1021 sha1close(pack_file, cur_pack_oid.hash, 0);
1022 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
1023 pack_data->pack_name, object_count,
1024 cur_pack_oid.hash, pack_size);
1025
1026 if (object_count <= unpack_limit) {
1027 if (!loosen_small_pack(pack_data)) {
1028 invalidate_pack_id(pack_id);
1029 goto discard_pack;
1030 }
1031 }
1032
1033 close(pack_data->pack_fd);
1034 idx_name = keep_pack(create_index());
1035
1036 /* Register the packfile with core git's machinery. */
1037 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
1038 if (!new_p)
1039 die("core git rejected index %s", idx_name);
1040 all_packs[pack_id] = new_p;
1041 install_packed_git(the_repository, new_p);
1042 free(idx_name);
1043
1044 /* Print the boundary */
1045 if (pack_edges) {
1046 fprintf(pack_edges, "%s:", new_p->pack_name);
1047 for (i = 0; i < branch_table_sz; i++) {
1048 for (b = branch_table[i]; b; b = b->table_next_branch) {
1049 if (b->pack_id == pack_id)
1050 fprintf(pack_edges, " %s",
1051 oid_to_hex(&b->oid));
1052 }
1053 }
1054 for (t = first_tag; t; t = t->next_tag) {
1055 if (t->pack_id == pack_id)
1056 fprintf(pack_edges, " %s",
1057 oid_to_hex(&t->oid));
1058 }
1059 fputc('\n', pack_edges);
1060 fflush(pack_edges);
1061 }
1062
1063 pack_id++;
1064 }
1065 else {
1066 discard_pack:
1067 close(pack_data->pack_fd);
1068 unlink_or_warn(pack_data->pack_name);
1069 }
1070 FREE_AND_NULL(pack_data);
1071 running = 0;
1072
1073 /* We can't carry a delta across packfiles. */
1074 strbuf_release(&last_blob.data);
1075 last_blob.offset = 0;
1076 last_blob.depth = 0;
1077 }
1078
1079 static void cycle_packfile(void)
1080 {
1081 end_packfile();
1082 start_packfile();
1083 }
1084
1085 static int store_object(
1086 enum object_type type,
1087 struct strbuf *dat,
1088 struct last_object *last,
1089 struct object_id *oidout,
1090 uintmax_t mark)
1091 {
1092 void *out, *delta;
1093 struct object_entry *e;
1094 unsigned char hdr[96];
1095 struct object_id oid;
1096 unsigned long hdrlen, deltalen;
1097 git_SHA_CTX c;
1098 git_zstream s;
1099
1100 hdrlen = xsnprintf((char *)hdr, sizeof(hdr), "%s %lu",
1101 typename(type), (unsigned long)dat->len) + 1;
1102 git_SHA1_Init(&c);
1103 git_SHA1_Update(&c, hdr, hdrlen);
1104 git_SHA1_Update(&c, dat->buf, dat->len);
1105 git_SHA1_Final(oid.hash, &c);
1106 if (oidout)
1107 oidcpy(oidout, &oid);
1108
1109 e = insert_object(&oid);
1110 if (mark)
1111 insert_mark(mark, e);
1112 if (e->idx.offset) {
1113 duplicate_count_by_type[type]++;
1114 return 1;
1115 } else if (find_sha1_pack(oid.hash,
1116 get_packed_git(the_repository))) {
1117 e->type = type;
1118 e->pack_id = MAX_PACK_ID;
1119 e->idx.offset = 1; /* just not zero! */
1120 duplicate_count_by_type[type]++;
1121 return 1;
1122 }
1123
1124 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1125 delta_count_attempts_by_type[type]++;
1126 delta = diff_delta(last->data.buf, last->data.len,
1127 dat->buf, dat->len,
1128 &deltalen, dat->len - 20);
1129 } else
1130 delta = NULL;
1131
1132 git_deflate_init(&s, pack_compression_level);
1133 if (delta) {
1134 s.next_in = delta;
1135 s.avail_in = deltalen;
1136 } else {
1137 s.next_in = (void *)dat->buf;
1138 s.avail_in = dat->len;
1139 }
1140 s.avail_out = git_deflate_bound(&s, s.avail_in);
1141 s.next_out = out = xmalloc(s.avail_out);
1142 while (git_deflate(&s, Z_FINISH) == Z_OK)
1143 ; /* nothing */
1144 git_deflate_end(&s);
1145
1146 /* Determine if we should auto-checkpoint. */
1147 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1148 || (pack_size + 60 + s.total_out) < pack_size) {
1149
1150 /* This new object needs to *not* have the current pack_id. */
1151 e->pack_id = pack_id + 1;
1152 cycle_packfile();
1153
1154 /* We cannot carry a delta into the new pack. */
1155 if (delta) {
1156 FREE_AND_NULL(delta);
1157
1158 git_deflate_init(&s, pack_compression_level);
1159 s.next_in = (void *)dat->buf;
1160 s.avail_in = dat->len;
1161 s.avail_out = git_deflate_bound(&s, s.avail_in);
1162 s.next_out = out = xrealloc(out, s.avail_out);
1163 while (git_deflate(&s, Z_FINISH) == Z_OK)
1164 ; /* nothing */
1165 git_deflate_end(&s);
1166 }
1167 }
1168
1169 e->type = type;
1170 e->pack_id = pack_id;
1171 e->idx.offset = pack_size;
1172 object_count++;
1173 object_count_by_type[type]++;
1174
1175 crc32_begin(pack_file);
1176
1177 if (delta) {
1178 off_t ofs = e->idx.offset - last->offset;
1179 unsigned pos = sizeof(hdr) - 1;
1180
1181 delta_count_by_type[type]++;
1182 e->depth = last->depth + 1;
1183
1184 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1185 OBJ_OFS_DELTA, deltalen);
1186 sha1write(pack_file, hdr, hdrlen);
1187 pack_size += hdrlen;
1188
1189 hdr[pos] = ofs & 127;
1190 while (ofs >>= 7)
1191 hdr[--pos] = 128 | (--ofs & 127);
1192 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1193 pack_size += sizeof(hdr) - pos;
1194 } else {
1195 e->depth = 0;
1196 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1197 type, dat->len);
1198 sha1write(pack_file, hdr, hdrlen);
1199 pack_size += hdrlen;
1200 }
1201
1202 sha1write(pack_file, out, s.total_out);
1203 pack_size += s.total_out;
1204
1205 e->idx.crc32 = crc32_end(pack_file);
1206
1207 free(out);
1208 free(delta);
1209 if (last) {
1210 if (last->no_swap) {
1211 last->data = *dat;
1212 } else {
1213 strbuf_swap(&last->data, dat);
1214 }
1215 last->offset = e->idx.offset;
1216 last->depth = e->depth;
1217 }
1218 return 0;
1219 }
1220
1221 static void truncate_pack(struct sha1file_checkpoint *checkpoint)
1222 {
1223 if (sha1file_truncate(pack_file, checkpoint))
1224 die_errno("cannot truncate pack to skip duplicate");
1225 pack_size = checkpoint->offset;
1226 }
1227
1228 static void stream_blob(uintmax_t len, struct object_id *oidout, uintmax_t mark)
1229 {
1230 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1231 unsigned char *in_buf = xmalloc(in_sz);
1232 unsigned char *out_buf = xmalloc(out_sz);
1233 struct object_entry *e;
1234 struct object_id oid;
1235 unsigned long hdrlen;
1236 off_t offset;
1237 git_SHA_CTX c;
1238 git_zstream s;
1239 struct sha1file_checkpoint checkpoint;
1240 int status = Z_OK;
1241
1242 /* Determine if we should auto-checkpoint. */
1243 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1244 || (pack_size + 60 + len) < pack_size)
1245 cycle_packfile();
1246
1247 sha1file_checkpoint(pack_file, &checkpoint);
1248 offset = checkpoint.offset;
1249
1250 hdrlen = xsnprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1251
1252 git_SHA1_Init(&c);
1253 git_SHA1_Update(&c, out_buf, hdrlen);
1254
1255 crc32_begin(pack_file);
1256
1257 git_deflate_init(&s, pack_compression_level);
1258
1259 hdrlen = encode_in_pack_object_header(out_buf, out_sz, OBJ_BLOB, len);
1260
1261 s.next_out = out_buf + hdrlen;
1262 s.avail_out = out_sz - hdrlen;
1263
1264 while (status != Z_STREAM_END) {
1265 if (0 < len && !s.avail_in) {
1266 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1267 size_t n = fread(in_buf, 1, cnt, stdin);
1268 if (!n && feof(stdin))
1269 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1270
1271 git_SHA1_Update(&c, in_buf, n);
1272 s.next_in = in_buf;
1273 s.avail_in = n;
1274 len -= n;
1275 }
1276
1277 status = git_deflate(&s, len ? 0 : Z_FINISH);
1278
1279 if (!s.avail_out || status == Z_STREAM_END) {
1280 size_t n = s.next_out - out_buf;
1281 sha1write(pack_file, out_buf, n);
1282 pack_size += n;
1283 s.next_out = out_buf;
1284 s.avail_out = out_sz;
1285 }
1286
1287 switch (status) {
1288 case Z_OK:
1289 case Z_BUF_ERROR:
1290 case Z_STREAM_END:
1291 continue;
1292 default:
1293 die("unexpected deflate failure: %d", status);
1294 }
1295 }
1296 git_deflate_end(&s);
1297 git_SHA1_Final(oid.hash, &c);
1298
1299 if (oidout)
1300 oidcpy(oidout, &oid);
1301
1302 e = insert_object(&oid);
1303
1304 if (mark)
1305 insert_mark(mark, e);
1306
1307 if (e->idx.offset) {
1308 duplicate_count_by_type[OBJ_BLOB]++;
1309 truncate_pack(&checkpoint);
1310
1311 } else if (find_sha1_pack(oid.hash,
1312 get_packed_git(the_repository))) {
1313 e->type = OBJ_BLOB;
1314 e->pack_id = MAX_PACK_ID;
1315 e->idx.offset = 1; /* just not zero! */
1316 duplicate_count_by_type[OBJ_BLOB]++;
1317 truncate_pack(&checkpoint);
1318
1319 } else {
1320 e->depth = 0;
1321 e->type = OBJ_BLOB;
1322 e->pack_id = pack_id;
1323 e->idx.offset = offset;
1324 e->idx.crc32 = crc32_end(pack_file);
1325 object_count++;
1326 object_count_by_type[OBJ_BLOB]++;
1327 }
1328
1329 free(in_buf);
1330 free(out_buf);
1331 }
1332
1333 /* All calls must be guarded by find_object() or find_mark() to
1334 * ensure the 'struct object_entry' passed was written by this
1335 * process instance. We unpack the entry by the offset, avoiding
1336 * the need for the corresponding .idx file. This unpacking rule
1337 * works because we only use OBJ_REF_DELTA within the packfiles
1338 * created by fast-import.
1339 *
1340 * oe must not be NULL. Such an oe usually comes from giving
1341 * an unknown SHA-1 to find_object() or an undefined mark to
1342 * find_mark(). Callers must test for this condition and use
1343 * the standard read_sha1_file() when it happens.
1344 *
1345 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1346 * find_mark(), where the mark was reloaded from an existing marks
1347 * file and is referencing an object that this fast-import process
1348 * instance did not write out to a packfile. Callers must test for
1349 * this condition and use read_sha1_file() instead.
1350 */
1351 static void *gfi_unpack_entry(
1352 struct object_entry *oe,
1353 unsigned long *sizep)
1354 {
1355 enum object_type type;
1356 struct packed_git *p = all_packs[oe->pack_id];
1357 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1358 /* The object is stored in the packfile we are writing to
1359 * and we have modified it since the last time we scanned
1360 * back to read a previously written object. If an old
1361 * window covered [p->pack_size, p->pack_size + 20) its
1362 * data is stale and is not valid. Closing all windows
1363 * and updating the packfile length ensures we can read
1364 * the newly written data.
1365 */
1366 close_pack_windows(p);
1367 sha1flush(pack_file);
1368
1369 /* We have to offer 20 bytes additional on the end of
1370 * the packfile as the core unpacker code assumes the
1371 * footer is present at the file end and must promise
1372 * at least 20 bytes within any window it maps. But
1373 * we don't actually create the footer here.
1374 */
1375 p->pack_size = pack_size + 20;
1376 }
1377 return unpack_entry(p, oe->idx.offset, &type, sizep);
1378 }
1379
1380 static const char *get_mode(const char *str, uint16_t *modep)
1381 {
1382 unsigned char c;
1383 uint16_t mode = 0;
1384
1385 while ((c = *str++) != ' ') {
1386 if (c < '0' || c > '7')
1387 return NULL;
1388 mode = (mode << 3) + (c - '0');
1389 }
1390 *modep = mode;
1391 return str;
1392 }
1393
1394 static void load_tree(struct tree_entry *root)
1395 {
1396 struct object_id *oid = &root->versions[1].oid;
1397 struct object_entry *myoe;
1398 struct tree_content *t;
1399 unsigned long size;
1400 char *buf;
1401 const char *c;
1402
1403 root->tree = t = new_tree_content(8);
1404 if (is_null_oid(oid))
1405 return;
1406
1407 myoe = find_object(oid);
1408 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1409 if (myoe->type != OBJ_TREE)
1410 die("Not a tree: %s", oid_to_hex(oid));
1411 t->delta_depth = myoe->depth;
1412 buf = gfi_unpack_entry(myoe, &size);
1413 if (!buf)
1414 die("Can't load tree %s", oid_to_hex(oid));
1415 } else {
1416 enum object_type type;
1417 buf = read_sha1_file(oid->hash, &type, &size);
1418 if (!buf || type != OBJ_TREE)
1419 die("Can't load tree %s", oid_to_hex(oid));
1420 }
1421
1422 c = buf;
1423 while (c != (buf + size)) {
1424 struct tree_entry *e = new_tree_entry();
1425
1426 if (t->entry_count == t->entry_capacity)
1427 root->tree = t = grow_tree_content(t, t->entry_count);
1428 t->entries[t->entry_count++] = e;
1429
1430 e->tree = NULL;
1431 c = get_mode(c, &e->versions[1].mode);
1432 if (!c)
1433 die("Corrupt mode in %s", oid_to_hex(oid));
1434 e->versions[0].mode = e->versions[1].mode;
1435 e->name = to_atom(c, strlen(c));
1436 c += e->name->str_len + 1;
1437 hashcpy(e->versions[0].oid.hash, (unsigned char *)c);
1438 hashcpy(e->versions[1].oid.hash, (unsigned char *)c);
1439 c += GIT_SHA1_RAWSZ;
1440 }
1441 free(buf);
1442 }
1443
1444 static int tecmp0 (const void *_a, const void *_b)
1445 {
1446 struct tree_entry *a = *((struct tree_entry**)_a);
1447 struct tree_entry *b = *((struct tree_entry**)_b);
1448 return base_name_compare(
1449 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1450 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1451 }
1452
1453 static int tecmp1 (const void *_a, const void *_b)
1454 {
1455 struct tree_entry *a = *((struct tree_entry**)_a);
1456 struct tree_entry *b = *((struct tree_entry**)_b);
1457 return base_name_compare(
1458 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1459 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1460 }
1461
1462 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1463 {
1464 size_t maxlen = 0;
1465 unsigned int i;
1466
1467 if (!v)
1468 QSORT(t->entries, t->entry_count, tecmp0);
1469 else
1470 QSORT(t->entries, t->entry_count, tecmp1);
1471
1472 for (i = 0; i < t->entry_count; i++) {
1473 if (t->entries[i]->versions[v].mode)
1474 maxlen += t->entries[i]->name->str_len + 34;
1475 }
1476
1477 strbuf_reset(b);
1478 strbuf_grow(b, maxlen);
1479 for (i = 0; i < t->entry_count; i++) {
1480 struct tree_entry *e = t->entries[i];
1481 if (!e->versions[v].mode)
1482 continue;
1483 strbuf_addf(b, "%o %s%c",
1484 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1485 e->name->str_dat, '\0');
1486 strbuf_add(b, e->versions[v].oid.hash, GIT_SHA1_RAWSZ);
1487 }
1488 }
1489
1490 static void store_tree(struct tree_entry *root)
1491 {
1492 struct tree_content *t;
1493 unsigned int i, j, del;
1494 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1495 struct object_entry *le = NULL;
1496
1497 if (!is_null_oid(&root->versions[1].oid))
1498 return;
1499
1500 if (!root->tree)
1501 load_tree(root);
1502 t = root->tree;
1503
1504 for (i = 0; i < t->entry_count; i++) {
1505 if (t->entries[i]->tree)
1506 store_tree(t->entries[i]);
1507 }
1508
1509 if (!(root->versions[0].mode & NO_DELTA))
1510 le = find_object(&root->versions[0].oid);
1511 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1512 mktree(t, 0, &old_tree);
1513 lo.data = old_tree;
1514 lo.offset = le->idx.offset;
1515 lo.depth = t->delta_depth;
1516 }
1517
1518 mktree(t, 1, &new_tree);
1519 store_object(OBJ_TREE, &new_tree, &lo, &root->versions[1].oid, 0);
1520
1521 t->delta_depth = lo.depth;
1522 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1523 struct tree_entry *e = t->entries[i];
1524 if (e->versions[1].mode) {
1525 e->versions[0].mode = e->versions[1].mode;
1526 oidcpy(&e->versions[0].oid, &e->versions[1].oid);
1527 t->entries[j++] = e;
1528 } else {
1529 release_tree_entry(e);
1530 del++;
1531 }
1532 }
1533 t->entry_count -= del;
1534 }
1535
1536 static void tree_content_replace(
1537 struct tree_entry *root,
1538 const struct object_id *oid,
1539 const uint16_t mode,
1540 struct tree_content *newtree)
1541 {
1542 if (!S_ISDIR(mode))
1543 die("Root cannot be a non-directory");
1544 oidclr(&root->versions[0].oid);
1545 oidcpy(&root->versions[1].oid, oid);
1546 if (root->tree)
1547 release_tree_content_recursive(root->tree);
1548 root->tree = newtree;
1549 }
1550
1551 static int tree_content_set(
1552 struct tree_entry *root,
1553 const char *p,
1554 const struct object_id *oid,
1555 const uint16_t mode,
1556 struct tree_content *subtree)
1557 {
1558 struct tree_content *t;
1559 const char *slash1;
1560 unsigned int i, n;
1561 struct tree_entry *e;
1562
1563 slash1 = strchrnul(p, '/');
1564 n = slash1 - p;
1565 if (!n)
1566 die("Empty path component found in input");
1567 if (!*slash1 && !S_ISDIR(mode) && subtree)
1568 die("Non-directories cannot have subtrees");
1569
1570 if (!root->tree)
1571 load_tree(root);
1572 t = root->tree;
1573 for (i = 0; i < t->entry_count; i++) {
1574 e = t->entries[i];
1575 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1576 if (!*slash1) {
1577 if (!S_ISDIR(mode)
1578 && e->versions[1].mode == mode
1579 && !oidcmp(&e->versions[1].oid, oid))
1580 return 0;
1581 e->versions[1].mode = mode;
1582 oidcpy(&e->versions[1].oid, oid);
1583 if (e->tree)
1584 release_tree_content_recursive(e->tree);
1585 e->tree = subtree;
1586
1587 /*
1588 * We need to leave e->versions[0].sha1 alone
1589 * to avoid modifying the preimage tree used
1590 * when writing out the parent directory.
1591 * But after replacing the subdir with a
1592 * completely different one, it's not a good
1593 * delta base any more, and besides, we've
1594 * thrown away the tree entries needed to
1595 * make a delta against it.
1596 *
1597 * So let's just explicitly disable deltas
1598 * for the subtree.
1599 */
1600 if (S_ISDIR(e->versions[0].mode))
1601 e->versions[0].mode |= NO_DELTA;
1602
1603 oidclr(&root->versions[1].oid);
1604 return 1;
1605 }
1606 if (!S_ISDIR(e->versions[1].mode)) {
1607 e->tree = new_tree_content(8);
1608 e->versions[1].mode = S_IFDIR;
1609 }
1610 if (!e->tree)
1611 load_tree(e);
1612 if (tree_content_set(e, slash1 + 1, oid, mode, subtree)) {
1613 oidclr(&root->versions[1].oid);
1614 return 1;
1615 }
1616 return 0;
1617 }
1618 }
1619
1620 if (t->entry_count == t->entry_capacity)
1621 root->tree = t = grow_tree_content(t, t->entry_count);
1622 e = new_tree_entry();
1623 e->name = to_atom(p, n);
1624 e->versions[0].mode = 0;
1625 oidclr(&e->versions[0].oid);
1626 t->entries[t->entry_count++] = e;
1627 if (*slash1) {
1628 e->tree = new_tree_content(8);
1629 e->versions[1].mode = S_IFDIR;
1630 tree_content_set(e, slash1 + 1, oid, mode, subtree);
1631 } else {
1632 e->tree = subtree;
1633 e->versions[1].mode = mode;
1634 oidcpy(&e->versions[1].oid, oid);
1635 }
1636 oidclr(&root->versions[1].oid);
1637 return 1;
1638 }
1639
1640 static int tree_content_remove(
1641 struct tree_entry *root,
1642 const char *p,
1643 struct tree_entry *backup_leaf,
1644 int allow_root)
1645 {
1646 struct tree_content *t;
1647 const char *slash1;
1648 unsigned int i, n;
1649 struct tree_entry *e;
1650
1651 slash1 = strchrnul(p, '/');
1652 n = slash1 - p;
1653
1654 if (!root->tree)
1655 load_tree(root);
1656
1657 if (!*p && allow_root) {
1658 e = root;
1659 goto del_entry;
1660 }
1661
1662 t = root->tree;
1663 for (i = 0; i < t->entry_count; i++) {
1664 e = t->entries[i];
1665 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1666 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1667 /*
1668 * If p names a file in some subdirectory, and a
1669 * file or symlink matching the name of the
1670 * parent directory of p exists, then p cannot
1671 * exist and need not be deleted.
1672 */
1673 return 1;
1674 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1675 goto del_entry;
1676 if (!e->tree)
1677 load_tree(e);
1678 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1679 for (n = 0; n < e->tree->entry_count; n++) {
1680 if (e->tree->entries[n]->versions[1].mode) {
1681 oidclr(&root->versions[1].oid);
1682 return 1;
1683 }
1684 }
1685 backup_leaf = NULL;
1686 goto del_entry;
1687 }
1688 return 0;
1689 }
1690 }
1691 return 0;
1692
1693 del_entry:
1694 if (backup_leaf)
1695 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1696 else if (e->tree)
1697 release_tree_content_recursive(e->tree);
1698 e->tree = NULL;
1699 e->versions[1].mode = 0;
1700 oidclr(&e->versions[1].oid);
1701 oidclr(&root->versions[1].oid);
1702 return 1;
1703 }
1704
1705 static int tree_content_get(
1706 struct tree_entry *root,
1707 const char *p,
1708 struct tree_entry *leaf,
1709 int allow_root)
1710 {
1711 struct tree_content *t;
1712 const char *slash1;
1713 unsigned int i, n;
1714 struct tree_entry *e;
1715
1716 slash1 = strchrnul(p, '/');
1717 n = slash1 - p;
1718 if (!n && !allow_root)
1719 die("Empty path component found in input");
1720
1721 if (!root->tree)
1722 load_tree(root);
1723
1724 if (!n) {
1725 e = root;
1726 goto found_entry;
1727 }
1728
1729 t = root->tree;
1730 for (i = 0; i < t->entry_count; i++) {
1731 e = t->entries[i];
1732 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1733 if (!*slash1)
1734 goto found_entry;
1735 if (!S_ISDIR(e->versions[1].mode))
1736 return 0;
1737 if (!e->tree)
1738 load_tree(e);
1739 return tree_content_get(e, slash1 + 1, leaf, 0);
1740 }
1741 }
1742 return 0;
1743
1744 found_entry:
1745 memcpy(leaf, e, sizeof(*leaf));
1746 if (e->tree && is_null_oid(&e->versions[1].oid))
1747 leaf->tree = dup_tree_content(e->tree);
1748 else
1749 leaf->tree = NULL;
1750 return 1;
1751 }
1752
1753 static int update_branch(struct branch *b)
1754 {
1755 static const char *msg = "fast-import";
1756 struct ref_transaction *transaction;
1757 struct object_id old_oid;
1758 struct strbuf err = STRBUF_INIT;
1759
1760 if (is_null_oid(&b->oid)) {
1761 if (b->delete)
1762 delete_ref(NULL, b->name, NULL, 0);
1763 return 0;
1764 }
1765 if (read_ref(b->name, &old_oid))
1766 oidclr(&old_oid);
1767 if (!force_update && !is_null_oid(&old_oid)) {
1768 struct commit *old_cmit, *new_cmit;
1769
1770 old_cmit = lookup_commit_reference_gently(&old_oid, 0);
1771 new_cmit = lookup_commit_reference_gently(&b->oid, 0);
1772 if (!old_cmit || !new_cmit)
1773 return error("Branch %s is missing commits.", b->name);
1774
1775 if (!in_merge_bases(old_cmit, new_cmit)) {
1776 warning("Not updating %s"
1777 " (new tip %s does not contain %s)",
1778 b->name, oid_to_hex(&b->oid),
1779 oid_to_hex(&old_oid));
1780 return -1;
1781 }
1782 }
1783 transaction = ref_transaction_begin(&err);
1784 if (!transaction ||
1785 ref_transaction_update(transaction, b->name, &b->oid, &old_oid,
1786 0, msg, &err) ||
1787 ref_transaction_commit(transaction, &err)) {
1788 ref_transaction_free(transaction);
1789 error("%s", err.buf);
1790 strbuf_release(&err);
1791 return -1;
1792 }
1793 ref_transaction_free(transaction);
1794 strbuf_release(&err);
1795 return 0;
1796 }
1797
1798 static void dump_branches(void)
1799 {
1800 unsigned int i;
1801 struct branch *b;
1802
1803 for (i = 0; i < branch_table_sz; i++) {
1804 for (b = branch_table[i]; b; b = b->table_next_branch)
1805 failure |= update_branch(b);
1806 }
1807 }
1808
1809 static void dump_tags(void)
1810 {
1811 static const char *msg = "fast-import";
1812 struct tag *t;
1813 struct strbuf ref_name = STRBUF_INIT;
1814 struct strbuf err = STRBUF_INIT;
1815 struct ref_transaction *transaction;
1816
1817 transaction = ref_transaction_begin(&err);
1818 if (!transaction) {
1819 failure |= error("%s", err.buf);
1820 goto cleanup;
1821 }
1822 for (t = first_tag; t; t = t->next_tag) {
1823 strbuf_reset(&ref_name);
1824 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1825
1826 if (ref_transaction_update(transaction, ref_name.buf,
1827 &t->oid, NULL, 0, msg, &err)) {
1828 failure |= error("%s", err.buf);
1829 goto cleanup;
1830 }
1831 }
1832 if (ref_transaction_commit(transaction, &err))
1833 failure |= error("%s", err.buf);
1834
1835 cleanup:
1836 ref_transaction_free(transaction);
1837 strbuf_release(&ref_name);
1838 strbuf_release(&err);
1839 }
1840
1841 static void dump_marks_helper(FILE *f,
1842 uintmax_t base,
1843 struct mark_set *m)
1844 {
1845 uintmax_t k;
1846 if (m->shift) {
1847 for (k = 0; k < 1024; k++) {
1848 if (m->data.sets[k])
1849 dump_marks_helper(f, base + (k << m->shift),
1850 m->data.sets[k]);
1851 }
1852 } else {
1853 for (k = 0; k < 1024; k++) {
1854 if (m->data.marked[k])
1855 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1856 oid_to_hex(&m->data.marked[k]->idx.oid));
1857 }
1858 }
1859 }
1860
1861 static void dump_marks(void)
1862 {
1863 static struct lock_file mark_lock;
1864 FILE *f;
1865
1866 if (!export_marks_file || (import_marks_file && !import_marks_file_done))
1867 return;
1868
1869 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1870 failure |= error_errno("Unable to write marks file %s",
1871 export_marks_file);
1872 return;
1873 }
1874
1875 f = fdopen_lock_file(&mark_lock, "w");
1876 if (!f) {
1877 int saved_errno = errno;
1878 rollback_lock_file(&mark_lock);
1879 failure |= error("Unable to write marks file %s: %s",
1880 export_marks_file, strerror(saved_errno));
1881 return;
1882 }
1883
1884 dump_marks_helper(f, 0, marks);
1885 if (commit_lock_file(&mark_lock)) {
1886 failure |= error_errno("Unable to write file %s",
1887 export_marks_file);
1888 return;
1889 }
1890 }
1891
1892 static void read_marks(void)
1893 {
1894 char line[512];
1895 FILE *f = fopen(import_marks_file, "r");
1896 if (f)
1897 ;
1898 else if (import_marks_file_ignore_missing && errno == ENOENT)
1899 goto done; /* Marks file does not exist */
1900 else
1901 die_errno("cannot read '%s'", import_marks_file);
1902 while (fgets(line, sizeof(line), f)) {
1903 uintmax_t mark;
1904 char *end;
1905 struct object_id oid;
1906 struct object_entry *e;
1907
1908 end = strchr(line, '\n');
1909 if (line[0] != ':' || !end)
1910 die("corrupt mark line: %s", line);
1911 *end = 0;
1912 mark = strtoumax(line + 1, &end, 10);
1913 if (!mark || end == line + 1
1914 || *end != ' ' || get_oid_hex(end + 1, &oid))
1915 die("corrupt mark line: %s", line);
1916 e = find_object(&oid);
1917 if (!e) {
1918 enum object_type type = sha1_object_info(oid.hash, NULL);
1919 if (type < 0)
1920 die("object not found: %s", oid_to_hex(&oid));
1921 e = insert_object(&oid);
1922 e->type = type;
1923 e->pack_id = MAX_PACK_ID;
1924 e->idx.offset = 1; /* just not zero! */
1925 }
1926 insert_mark(mark, e);
1927 }
1928 fclose(f);
1929 done:
1930 import_marks_file_done = 1;
1931 }
1932
1933
1934 static int read_next_command(void)
1935 {
1936 static int stdin_eof = 0;
1937
1938 if (stdin_eof) {
1939 unread_command_buf = 0;
1940 return EOF;
1941 }
1942
1943 for (;;) {
1944 const char *p;
1945
1946 if (unread_command_buf) {
1947 unread_command_buf = 0;
1948 } else {
1949 struct recent_command *rc;
1950
1951 strbuf_detach(&command_buf, NULL);
1952 stdin_eof = strbuf_getline_lf(&command_buf, stdin);
1953 if (stdin_eof)
1954 return EOF;
1955
1956 if (!seen_data_command
1957 && !starts_with(command_buf.buf, "feature ")
1958 && !starts_with(command_buf.buf, "option ")) {
1959 parse_argv();
1960 }
1961
1962 rc = rc_free;
1963 if (rc)
1964 rc_free = rc->next;
1965 else {
1966 rc = cmd_hist.next;
1967 cmd_hist.next = rc->next;
1968 cmd_hist.next->prev = &cmd_hist;
1969 free(rc->buf);
1970 }
1971
1972 rc->buf = command_buf.buf;
1973 rc->prev = cmd_tail;
1974 rc->next = cmd_hist.prev;
1975 rc->prev->next = rc;
1976 cmd_tail = rc;
1977 }
1978 if (skip_prefix(command_buf.buf, "get-mark ", &p)) {
1979 parse_get_mark(p);
1980 continue;
1981 }
1982 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1983 parse_cat_blob(p);
1984 continue;
1985 }
1986 if (command_buf.buf[0] == '#')
1987 continue;
1988 return 0;
1989 }
1990 }
1991
1992 static void skip_optional_lf(void)
1993 {
1994 int term_char = fgetc(stdin);
1995 if (term_char != '\n' && term_char != EOF)
1996 ungetc(term_char, stdin);
1997 }
1998
1999 static void parse_mark(void)
2000 {
2001 const char *v;
2002 if (skip_prefix(command_buf.buf, "mark :", &v)) {
2003 next_mark = strtoumax(v, NULL, 10);
2004 read_next_command();
2005 }
2006 else
2007 next_mark = 0;
2008 }
2009
2010 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
2011 {
2012 const char *data;
2013 strbuf_reset(sb);
2014
2015 if (!skip_prefix(command_buf.buf, "data ", &data))
2016 die("Expected 'data n' command, found: %s", command_buf.buf);
2017
2018 if (skip_prefix(data, "<<", &data)) {
2019 char *term = xstrdup(data);
2020 size_t term_len = command_buf.len - (data - command_buf.buf);
2021
2022 strbuf_detach(&command_buf, NULL);
2023 for (;;) {
2024 if (strbuf_getline_lf(&command_buf, stdin) == EOF)
2025 die("EOF in data (terminator '%s' not found)", term);
2026 if (term_len == command_buf.len
2027 && !strcmp(term, command_buf.buf))
2028 break;
2029 strbuf_addbuf(sb, &command_buf);
2030 strbuf_addch(sb, '\n');
2031 }
2032 free(term);
2033 }
2034 else {
2035 uintmax_t len = strtoumax(data, NULL, 10);
2036 size_t n = 0, length = (size_t)len;
2037
2038 if (limit && limit < len) {
2039 *len_res = len;
2040 return 0;
2041 }
2042 if (length < len)
2043 die("data is too large to use in this context");
2044
2045 while (n < length) {
2046 size_t s = strbuf_fread(sb, length - n, stdin);
2047 if (!s && feof(stdin))
2048 die("EOF in data (%lu bytes remaining)",
2049 (unsigned long)(length - n));
2050 n += s;
2051 }
2052 }
2053
2054 skip_optional_lf();
2055 return 1;
2056 }
2057
2058 static int validate_raw_date(const char *src, struct strbuf *result)
2059 {
2060 const char *orig_src = src;
2061 char *endp;
2062 unsigned long num;
2063
2064 errno = 0;
2065
2066 num = strtoul(src, &endp, 10);
2067 /* NEEDSWORK: perhaps check for reasonable values? */
2068 if (errno || endp == src || *endp != ' ')
2069 return -1;
2070
2071 src = endp + 1;
2072 if (*src != '-' && *src != '+')
2073 return -1;
2074
2075 num = strtoul(src + 1, &endp, 10);
2076 if (errno || endp == src + 1 || *endp || 1400 < num)
2077 return -1;
2078
2079 strbuf_addstr(result, orig_src);
2080 return 0;
2081 }
2082
2083 static char *parse_ident(const char *buf)
2084 {
2085 const char *ltgt;
2086 size_t name_len;
2087 struct strbuf ident = STRBUF_INIT;
2088
2089 /* ensure there is a space delimiter even if there is no name */
2090 if (*buf == '<')
2091 --buf;
2092
2093 ltgt = buf + strcspn(buf, "<>");
2094 if (*ltgt != '<')
2095 die("Missing < in ident string: %s", buf);
2096 if (ltgt != buf && ltgt[-1] != ' ')
2097 die("Missing space before < in ident string: %s", buf);
2098 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2099 if (*ltgt != '>')
2100 die("Missing > in ident string: %s", buf);
2101 ltgt++;
2102 if (*ltgt != ' ')
2103 die("Missing space after > in ident string: %s", buf);
2104 ltgt++;
2105 name_len = ltgt - buf;
2106 strbuf_add(&ident, buf, name_len);
2107
2108 switch (whenspec) {
2109 case WHENSPEC_RAW:
2110 if (validate_raw_date(ltgt, &ident) < 0)
2111 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2112 break;
2113 case WHENSPEC_RFC2822:
2114 if (parse_date(ltgt, &ident) < 0)
2115 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2116 break;
2117 case WHENSPEC_NOW:
2118 if (strcmp("now", ltgt))
2119 die("Date in ident must be 'now': %s", buf);
2120 datestamp(&ident);
2121 break;
2122 }
2123
2124 return strbuf_detach(&ident, NULL);
2125 }
2126
2127 static void parse_and_store_blob(
2128 struct last_object *last,
2129 struct object_id *oidout,
2130 uintmax_t mark)
2131 {
2132 static struct strbuf buf = STRBUF_INIT;
2133 uintmax_t len;
2134
2135 if (parse_data(&buf, big_file_threshold, &len))
2136 store_object(OBJ_BLOB, &buf, last, oidout, mark);
2137 else {
2138 if (last) {
2139 strbuf_release(&last->data);
2140 last->offset = 0;
2141 last->depth = 0;
2142 }
2143 stream_blob(len, oidout, mark);
2144 skip_optional_lf();
2145 }
2146 }
2147
2148 static void parse_new_blob(void)
2149 {
2150 read_next_command();
2151 parse_mark();
2152 parse_and_store_blob(&last_blob, NULL, next_mark);
2153 }
2154
2155 static void unload_one_branch(void)
2156 {
2157 while (cur_active_branches
2158 && cur_active_branches >= max_active_branches) {
2159 uintmax_t min_commit = ULONG_MAX;
2160 struct branch *e, *l = NULL, *p = NULL;
2161
2162 for (e = active_branches; e; e = e->active_next_branch) {
2163 if (e->last_commit < min_commit) {
2164 p = l;
2165 min_commit = e->last_commit;
2166 }
2167 l = e;
2168 }
2169
2170 if (p) {
2171 e = p->active_next_branch;
2172 p->active_next_branch = e->active_next_branch;
2173 } else {
2174 e = active_branches;
2175 active_branches = e->active_next_branch;
2176 }
2177 e->active = 0;
2178 e->active_next_branch = NULL;
2179 if (e->branch_tree.tree) {
2180 release_tree_content_recursive(e->branch_tree.tree);
2181 e->branch_tree.tree = NULL;
2182 }
2183 cur_active_branches--;
2184 }
2185 }
2186
2187 static void load_branch(struct branch *b)
2188 {
2189 load_tree(&b->branch_tree);
2190 if (!b->active) {
2191 b->active = 1;
2192 b->active_next_branch = active_branches;
2193 active_branches = b;
2194 cur_active_branches++;
2195 branch_load_count++;
2196 }
2197 }
2198
2199 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2200 {
2201 unsigned char fanout = 0;
2202 while ((num_notes >>= 8))
2203 fanout++;
2204 return fanout;
2205 }
2206
2207 static void construct_path_with_fanout(const char *hex_sha1,
2208 unsigned char fanout, char *path)
2209 {
2210 unsigned int i = 0, j = 0;
2211 if (fanout >= 20)
2212 die("Too large fanout (%u)", fanout);
2213 while (fanout) {
2214 path[i++] = hex_sha1[j++];
2215 path[i++] = hex_sha1[j++];
2216 path[i++] = '/';
2217 fanout--;
2218 }
2219 memcpy(path + i, hex_sha1 + j, GIT_SHA1_HEXSZ - j);
2220 path[i + GIT_SHA1_HEXSZ - j] = '\0';
2221 }
2222
2223 static uintmax_t do_change_note_fanout(
2224 struct tree_entry *orig_root, struct tree_entry *root,
2225 char *hex_oid, unsigned int hex_oid_len,
2226 char *fullpath, unsigned int fullpath_len,
2227 unsigned char fanout)
2228 {
2229 struct tree_content *t;
2230 struct tree_entry *e, leaf;
2231 unsigned int i, tmp_hex_oid_len, tmp_fullpath_len;
2232 uintmax_t num_notes = 0;
2233 struct object_id oid;
2234 char realpath[60];
2235
2236 if (!root->tree)
2237 load_tree(root);
2238 t = root->tree;
2239
2240 for (i = 0; t && i < t->entry_count; i++) {
2241 e = t->entries[i];
2242 tmp_hex_oid_len = hex_oid_len + e->name->str_len;
2243 tmp_fullpath_len = fullpath_len;
2244
2245 /*
2246 * We're interested in EITHER existing note entries (entries
2247 * with exactly 40 hex chars in path, not including directory
2248 * separators), OR directory entries that may contain note
2249 * entries (with < 40 hex chars in path).
2250 * Also, each path component in a note entry must be a multiple
2251 * of 2 chars.
2252 */
2253 if (!e->versions[1].mode ||
2254 tmp_hex_oid_len > GIT_SHA1_HEXSZ ||
2255 e->name->str_len % 2)
2256 continue;
2257
2258 /* This _may_ be a note entry, or a subdir containing notes */
2259 memcpy(hex_oid + hex_oid_len, e->name->str_dat,
2260 e->name->str_len);
2261 if (tmp_fullpath_len)
2262 fullpath[tmp_fullpath_len++] = '/';
2263 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2264 e->name->str_len);
2265 tmp_fullpath_len += e->name->str_len;
2266 fullpath[tmp_fullpath_len] = '\0';
2267
2268 if (tmp_hex_oid_len == GIT_SHA1_HEXSZ && !get_oid_hex(hex_oid, &oid)) {
2269 /* This is a note entry */
2270 if (fanout == 0xff) {
2271 /* Counting mode, no rename */
2272 num_notes++;
2273 continue;
2274 }
2275 construct_path_with_fanout(hex_oid, fanout, realpath);
2276 if (!strcmp(fullpath, realpath)) {
2277 /* Note entry is in correct location */
2278 num_notes++;
2279 continue;
2280 }
2281
2282 /* Rename fullpath to realpath */
2283 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2284 die("Failed to remove path %s", fullpath);
2285 tree_content_set(orig_root, realpath,
2286 &leaf.versions[1].oid,
2287 leaf.versions[1].mode,
2288 leaf.tree);
2289 } else if (S_ISDIR(e->versions[1].mode)) {
2290 /* This is a subdir that may contain note entries */
2291 num_notes += do_change_note_fanout(orig_root, e,
2292 hex_oid, tmp_hex_oid_len,
2293 fullpath, tmp_fullpath_len, fanout);
2294 }
2295
2296 /* The above may have reallocated the current tree_content */
2297 t = root->tree;
2298 }
2299 return num_notes;
2300 }
2301
2302 static uintmax_t change_note_fanout(struct tree_entry *root,
2303 unsigned char fanout)
2304 {
2305 /*
2306 * The size of path is due to one slash between every two hex digits,
2307 * plus the terminating NUL. Note that there is no slash at the end, so
2308 * the number of slashes is one less than half the number of hex
2309 * characters.
2310 */
2311 char hex_oid[GIT_MAX_HEXSZ], path[GIT_MAX_HEXSZ + (GIT_MAX_HEXSZ / 2) - 1 + 1];
2312 return do_change_note_fanout(root, root, hex_oid, 0, path, 0, fanout);
2313 }
2314
2315 /*
2316 * Given a pointer into a string, parse a mark reference:
2317 *
2318 * idnum ::= ':' bigint;
2319 *
2320 * Return the first character after the value in *endptr.
2321 *
2322 * Complain if the following character is not what is expected,
2323 * either a space or end of the string.
2324 */
2325 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2326 {
2327 uintmax_t mark;
2328
2329 assert(*p == ':');
2330 p++;
2331 mark = strtoumax(p, endptr, 10);
2332 if (*endptr == p)
2333 die("No value after ':' in mark: %s", command_buf.buf);
2334 return mark;
2335 }
2336
2337 /*
2338 * Parse the mark reference, and complain if this is not the end of
2339 * the string.
2340 */
2341 static uintmax_t parse_mark_ref_eol(const char *p)
2342 {
2343 char *end;
2344 uintmax_t mark;
2345
2346 mark = parse_mark_ref(p, &end);
2347 if (*end != '\0')
2348 die("Garbage after mark: %s", command_buf.buf);
2349 return mark;
2350 }
2351
2352 /*
2353 * Parse the mark reference, demanding a trailing space. Return a
2354 * pointer to the space.
2355 */
2356 static uintmax_t parse_mark_ref_space(const char **p)
2357 {
2358 uintmax_t mark;
2359 char *end;
2360
2361 mark = parse_mark_ref(*p, &end);
2362 if (*end++ != ' ')
2363 die("Missing space after mark: %s", command_buf.buf);
2364 *p = end;
2365 return mark;
2366 }
2367
2368 static void file_change_m(const char *p, struct branch *b)
2369 {
2370 static struct strbuf uq = STRBUF_INIT;
2371 const char *endp;
2372 struct object_entry *oe;
2373 struct object_id oid;
2374 uint16_t mode, inline_data = 0;
2375
2376 p = get_mode(p, &mode);
2377 if (!p)
2378 die("Corrupt mode: %s", command_buf.buf);
2379 switch (mode) {
2380 case 0644:
2381 case 0755:
2382 mode |= S_IFREG;
2383 case S_IFREG | 0644:
2384 case S_IFREG | 0755:
2385 case S_IFLNK:
2386 case S_IFDIR:
2387 case S_IFGITLINK:
2388 /* ok */
2389 break;
2390 default:
2391 die("Corrupt mode: %s", command_buf.buf);
2392 }
2393
2394 if (*p == ':') {
2395 oe = find_mark(parse_mark_ref_space(&p));
2396 oidcpy(&oid, &oe->idx.oid);
2397 } else if (skip_prefix(p, "inline ", &p)) {
2398 inline_data = 1;
2399 oe = NULL; /* not used with inline_data, but makes gcc happy */
2400 } else {
2401 if (parse_oid_hex(p, &oid, &p))
2402 die("Invalid dataref: %s", command_buf.buf);
2403 oe = find_object(&oid);
2404 if (*p++ != ' ')
2405 die("Missing space after SHA1: %s", command_buf.buf);
2406 }
2407
2408 strbuf_reset(&uq);
2409 if (!unquote_c_style(&uq, p, &endp)) {
2410 if (*endp)
2411 die("Garbage after path in: %s", command_buf.buf);
2412 p = uq.buf;
2413 }
2414
2415 /* Git does not track empty, non-toplevel directories. */
2416 if (S_ISDIR(mode) && is_empty_tree_oid(&oid) && *p) {
2417 tree_content_remove(&b->branch_tree, p, NULL, 0);
2418 return;
2419 }
2420
2421 if (S_ISGITLINK(mode)) {
2422 if (inline_data)
2423 die("Git links cannot be specified 'inline': %s",
2424 command_buf.buf);
2425 else if (oe) {
2426 if (oe->type != OBJ_COMMIT)
2427 die("Not a commit (actually a %s): %s",
2428 typename(oe->type), command_buf.buf);
2429 }
2430 /*
2431 * Accept the sha1 without checking; it expected to be in
2432 * another repository.
2433 */
2434 } else if (inline_data) {
2435 if (S_ISDIR(mode))
2436 die("Directories cannot be specified 'inline': %s",
2437 command_buf.buf);
2438 if (p != uq.buf) {
2439 strbuf_addstr(&uq, p);
2440 p = uq.buf;
2441 }
2442 read_next_command();
2443 parse_and_store_blob(&last_blob, &oid, 0);
2444 } else {
2445 enum object_type expected = S_ISDIR(mode) ?
2446 OBJ_TREE: OBJ_BLOB;
2447 enum object_type type = oe ? oe->type :
2448 sha1_object_info(oid.hash, NULL);
2449 if (type < 0)
2450 die("%s not found: %s",
2451 S_ISDIR(mode) ? "Tree" : "Blob",
2452 command_buf.buf);
2453 if (type != expected)
2454 die("Not a %s (actually a %s): %s",
2455 typename(expected), typename(type),
2456 command_buf.buf);
2457 }
2458
2459 if (!*p) {
2460 tree_content_replace(&b->branch_tree, &oid, mode, NULL);
2461 return;
2462 }
2463 tree_content_set(&b->branch_tree, p, &oid, mode, NULL);
2464 }
2465
2466 static void file_change_d(const char *p, struct branch *b)
2467 {
2468 static struct strbuf uq = STRBUF_INIT;
2469 const char *endp;
2470
2471 strbuf_reset(&uq);
2472 if (!unquote_c_style(&uq, p, &endp)) {
2473 if (*endp)
2474 die("Garbage after path in: %s", command_buf.buf);
2475 p = uq.buf;
2476 }
2477 tree_content_remove(&b->branch_tree, p, NULL, 1);
2478 }
2479
2480 static void file_change_cr(const char *s, struct branch *b, int rename)
2481 {
2482 const char *d;
2483 static struct strbuf s_uq = STRBUF_INIT;
2484 static struct strbuf d_uq = STRBUF_INIT;
2485 const char *endp;
2486 struct tree_entry leaf;
2487
2488 strbuf_reset(&s_uq);
2489 if (!unquote_c_style(&s_uq, s, &endp)) {
2490 if (*endp != ' ')
2491 die("Missing space after source: %s", command_buf.buf);
2492 } else {
2493 endp = strchr(s, ' ');
2494 if (!endp)
2495 die("Missing space after source: %s", command_buf.buf);
2496 strbuf_add(&s_uq, s, endp - s);
2497 }
2498 s = s_uq.buf;
2499
2500 endp++;
2501 if (!*endp)
2502 die("Missing dest: %s", command_buf.buf);
2503
2504 d = endp;
2505 strbuf_reset(&d_uq);
2506 if (!unquote_c_style(&d_uq, d, &endp)) {
2507 if (*endp)
2508 die("Garbage after dest in: %s", command_buf.buf);
2509 d = d_uq.buf;
2510 }
2511
2512 memset(&leaf, 0, sizeof(leaf));
2513 if (rename)
2514 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2515 else
2516 tree_content_get(&b->branch_tree, s, &leaf, 1);
2517 if (!leaf.versions[1].mode)
2518 die("Path %s not in branch", s);
2519 if (!*d) { /* C "path/to/subdir" "" */
2520 tree_content_replace(&b->branch_tree,
2521 &leaf.versions[1].oid,
2522 leaf.versions[1].mode,
2523 leaf.tree);
2524 return;
2525 }
2526 tree_content_set(&b->branch_tree, d,
2527 &leaf.versions[1].oid,
2528 leaf.versions[1].mode,
2529 leaf.tree);
2530 }
2531
2532 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2533 {
2534 static struct strbuf uq = STRBUF_INIT;
2535 struct object_entry *oe;
2536 struct branch *s;
2537 struct object_id oid, commit_oid;
2538 char path[60];
2539 uint16_t inline_data = 0;
2540 unsigned char new_fanout;
2541
2542 /*
2543 * When loading a branch, we don't traverse its tree to count the real
2544 * number of notes (too expensive to do this for all non-note refs).
2545 * This means that recently loaded notes refs might incorrectly have
2546 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2547 *
2548 * Fix this by traversing the tree and counting the number of notes
2549 * when b->num_notes == 0. If the notes tree is truly empty, the
2550 * calculation should not take long.
2551 */
2552 if (b->num_notes == 0 && *old_fanout == 0) {
2553 /* Invoke change_note_fanout() in "counting mode". */
2554 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2555 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2556 }
2557
2558 /* Now parse the notemodify command. */
2559 /* <dataref> or 'inline' */
2560 if (*p == ':') {
2561 oe = find_mark(parse_mark_ref_space(&p));
2562 oidcpy(&oid, &oe->idx.oid);
2563 } else if (skip_prefix(p, "inline ", &p)) {
2564 inline_data = 1;
2565 oe = NULL; /* not used with inline_data, but makes gcc happy */
2566 } else {
2567 if (parse_oid_hex(p, &oid, &p))
2568 die("Invalid dataref: %s", command_buf.buf);
2569 oe = find_object(&oid);
2570 if (*p++ != ' ')
2571 die("Missing space after SHA1: %s", command_buf.buf);
2572 }
2573
2574 /* <commit-ish> */
2575 s = lookup_branch(p);
2576 if (s) {
2577 if (is_null_oid(&s->oid))
2578 die("Can't add a note on empty branch.");
2579 oidcpy(&commit_oid, &s->oid);
2580 } else if (*p == ':') {
2581 uintmax_t commit_mark = parse_mark_ref_eol(p);
2582 struct object_entry *commit_oe = find_mark(commit_mark);
2583 if (commit_oe->type != OBJ_COMMIT)
2584 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2585 oidcpy(&commit_oid, &commit_oe->idx.oid);
2586 } else if (!get_oid(p, &commit_oid)) {
2587 unsigned long size;
2588 char *buf = read_object_with_reference(commit_oid.hash,
2589 commit_type, &size, commit_oid.hash);
2590 if (!buf || size < 46)
2591 die("Not a valid commit: %s", p);
2592 free(buf);
2593 } else
2594 die("Invalid ref name or SHA1 expression: %s", p);
2595
2596 if (inline_data) {
2597 if (p != uq.buf) {
2598 strbuf_addstr(&uq, p);
2599 p = uq.buf;
2600 }
2601 read_next_command();
2602 parse_and_store_blob(&last_blob, &oid, 0);
2603 } else if (oe) {
2604 if (oe->type != OBJ_BLOB)
2605 die("Not a blob (actually a %s): %s",
2606 typename(oe->type), command_buf.buf);
2607 } else if (!is_null_oid(&oid)) {
2608 enum object_type type = sha1_object_info(oid.hash, NULL);
2609 if (type < 0)
2610 die("Blob not found: %s", command_buf.buf);
2611 if (type != OBJ_BLOB)
2612 die("Not a blob (actually a %s): %s",
2613 typename(type), command_buf.buf);
2614 }
2615
2616 construct_path_with_fanout(oid_to_hex(&commit_oid), *old_fanout, path);
2617 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2618 b->num_notes--;
2619
2620 if (is_null_oid(&oid))
2621 return; /* nothing to insert */
2622
2623 b->num_notes++;
2624 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2625 construct_path_with_fanout(oid_to_hex(&commit_oid), new_fanout, path);
2626 tree_content_set(&b->branch_tree, path, &oid, S_IFREG | 0644, NULL);
2627 }
2628
2629 static void file_change_deleteall(struct branch *b)
2630 {
2631 release_tree_content_recursive(b->branch_tree.tree);
2632 oidclr(&b->branch_tree.versions[0].oid);
2633 oidclr(&b->branch_tree.versions[1].oid);
2634 load_tree(&b->branch_tree);
2635 b->num_notes = 0;
2636 }
2637
2638 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2639 {
2640 if (!buf || size < GIT_SHA1_HEXSZ + 6)
2641 die("Not a valid commit: %s", oid_to_hex(&b->oid));
2642 if (memcmp("tree ", buf, 5)
2643 || get_oid_hex(buf + 5, &b->branch_tree.versions[1].oid))
2644 die("The commit %s is corrupt", oid_to_hex(&b->oid));
2645 oidcpy(&b->branch_tree.versions[0].oid,
2646 &b->branch_tree.versions[1].oid);
2647 }
2648
2649 static void parse_from_existing(struct branch *b)
2650 {
2651 if (is_null_oid(&b->oid)) {
2652 oidclr(&b->branch_tree.versions[0].oid);
2653 oidclr(&b->branch_tree.versions[1].oid);
2654 } else {
2655 unsigned long size;
2656 char *buf;
2657
2658 buf = read_object_with_reference(b->oid.hash,
2659 commit_type, &size,
2660 b->oid.hash);
2661 parse_from_commit(b, buf, size);
2662 free(buf);
2663 }
2664 }
2665
2666 static int parse_from(struct branch *b)
2667 {
2668 const char *from;
2669 struct branch *s;
2670 struct object_id oid;
2671
2672 if (!skip_prefix(command_buf.buf, "from ", &from))
2673 return 0;
2674
2675 oidcpy(&oid, &b->branch_tree.versions[1].oid);
2676
2677 s = lookup_branch(from);
2678 if (b == s)
2679 die("Can't create a branch from itself: %s", b->name);
2680 else if (s) {
2681 struct object_id *t = &s->branch_tree.versions[1].oid;
2682 oidcpy(&b->oid, &s->oid);
2683 oidcpy(&b->branch_tree.versions[0].oid, t);
2684 oidcpy(&b->branch_tree.versions[1].oid, t);
2685 } else if (*from == ':') {
2686 uintmax_t idnum = parse_mark_ref_eol(from);
2687 struct object_entry *oe = find_mark(idnum);
2688 if (oe->type != OBJ_COMMIT)
2689 die("Mark :%" PRIuMAX " not a commit", idnum);
2690 if (oidcmp(&b->oid, &oe->idx.oid)) {
2691 oidcpy(&b->oid, &oe->idx.oid);
2692 if (oe->pack_id != MAX_PACK_ID) {
2693 unsigned long size;
2694 char *buf = gfi_unpack_entry(oe, &size);
2695 parse_from_commit(b, buf, size);
2696 free(buf);
2697 } else
2698 parse_from_existing(b);
2699 }
2700 } else if (!get_oid(from, &b->oid)) {
2701 parse_from_existing(b);
2702 if (is_null_oid(&b->oid))
2703 b->delete = 1;
2704 }
2705 else
2706 die("Invalid ref name or SHA1 expression: %s", from);
2707
2708 if (b->branch_tree.tree && oidcmp(&oid, &b->branch_tree.versions[1].oid)) {
2709 release_tree_content_recursive(b->branch_tree.tree);
2710 b->branch_tree.tree = NULL;
2711 }
2712
2713 read_next_command();
2714 return 1;
2715 }
2716
2717 static struct hash_list *parse_merge(unsigned int *count)
2718 {
2719 struct hash_list *list = NULL, **tail = &list, *n;
2720 const char *from;
2721 struct branch *s;
2722
2723 *count = 0;
2724 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2725 n = xmalloc(sizeof(*n));
2726 s = lookup_branch(from);
2727 if (s)
2728 oidcpy(&n->oid, &s->oid);
2729 else if (*from == ':') {
2730 uintmax_t idnum = parse_mark_ref_eol(from);
2731 struct object_entry *oe = find_mark(idnum);
2732 if (oe->type != OBJ_COMMIT)
2733 die("Mark :%" PRIuMAX " not a commit", idnum);
2734 oidcpy(&n->oid, &oe->idx.oid);
2735 } else if (!get_oid(from, &n->oid)) {
2736 unsigned long size;
2737 char *buf = read_object_with_reference(n->oid.hash,
2738 commit_type, &size, n->oid.hash);
2739 if (!buf || size < 46)
2740 die("Not a valid commit: %s", from);
2741 free(buf);
2742 } else
2743 die("Invalid ref name or SHA1 expression: %s", from);
2744
2745 n->next = NULL;
2746 *tail = n;
2747 tail = &n->next;
2748
2749 (*count)++;
2750 read_next_command();
2751 }
2752 return list;
2753 }
2754
2755 static void parse_new_commit(const char *arg)
2756 {
2757 static struct strbuf msg = STRBUF_INIT;
2758 struct branch *b;
2759 char *author = NULL;
2760 char *committer = NULL;
2761 struct hash_list *merge_list = NULL;
2762 unsigned int merge_count;
2763 unsigned char prev_fanout, new_fanout;
2764 const char *v;
2765
2766 b = lookup_branch(arg);
2767 if (!b)
2768 b = new_branch(arg);
2769
2770 read_next_command();
2771 parse_mark();
2772 if (skip_prefix(command_buf.buf, "author ", &v)) {
2773 author = parse_ident(v);
2774 read_next_command();
2775 }
2776 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2777 committer = parse_ident(v);
2778 read_next_command();
2779 }
2780 if (!committer)
2781 die("Expected committer but didn't get one");
2782 parse_data(&msg, 0, NULL);
2783 read_next_command();
2784 parse_from(b);
2785 merge_list = parse_merge(&merge_count);
2786
2787 /* ensure the branch is active/loaded */
2788 if (!b->branch_tree.tree || !max_active_branches) {
2789 unload_one_branch();
2790 load_branch(b);
2791 }
2792
2793 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2794
2795 /* file_change* */
2796 while (command_buf.len > 0) {
2797 if (skip_prefix(command_buf.buf, "M ", &v))
2798 file_change_m(v, b);
2799 else if (skip_prefix(command_buf.buf, "D ", &v))
2800 file_change_d(v, b);
2801 else if (skip_prefix(command_buf.buf, "R ", &v))
2802 file_change_cr(v, b, 1);
2803 else if (skip_prefix(command_buf.buf, "C ", &v))
2804 file_change_cr(v, b, 0);
2805 else if (skip_prefix(command_buf.buf, "N ", &v))
2806 note_change_n(v, b, &prev_fanout);
2807 else if (!strcmp("deleteall", command_buf.buf))
2808 file_change_deleteall(b);
2809 else if (skip_prefix(command_buf.buf, "ls ", &v))
2810 parse_ls(v, b);
2811 else {
2812 unread_command_buf = 1;
2813 break;
2814 }
2815 if (read_next_command() == EOF)
2816 break;
2817 }
2818
2819 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2820 if (new_fanout != prev_fanout)
2821 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2822
2823 /* build the tree and the commit */
2824 store_tree(&b->branch_tree);
2825 oidcpy(&b->branch_tree.versions[0].oid,
2826 &b->branch_tree.versions[1].oid);
2827
2828 strbuf_reset(&new_data);
2829 strbuf_addf(&new_data, "tree %s\n",
2830 oid_to_hex(&b->branch_tree.versions[1].oid));
2831 if (!is_null_oid(&b->oid))
2832 strbuf_addf(&new_data, "parent %s\n",
2833 oid_to_hex(&b->oid));
2834 while (merge_list) {
2835 struct hash_list *next = merge_list->next;
2836 strbuf_addf(&new_data, "parent %s\n",
2837 oid_to_hex(&merge_list->oid));
2838 free(merge_list);
2839 merge_list = next;
2840 }
2841 strbuf_addf(&new_data,
2842 "author %s\n"
2843 "committer %s\n"
2844 "\n",
2845 author ? author : committer, committer);
2846 strbuf_addbuf(&new_data, &msg);
2847 free(author);
2848 free(committer);
2849
2850 if (!store_object(OBJ_COMMIT, &new_data, NULL, &b->oid, next_mark))
2851 b->pack_id = pack_id;
2852 b->last_commit = object_count_by_type[OBJ_COMMIT];
2853 }
2854
2855 static void parse_new_tag(const char *arg)
2856 {
2857 static struct strbuf msg = STRBUF_INIT;
2858 const char *from;
2859 char *tagger;
2860 struct branch *s;
2861 struct tag *t;
2862 uintmax_t from_mark = 0;
2863 struct object_id oid;
2864 enum object_type type;
2865 const char *v;
2866
2867 t = pool_alloc(sizeof(struct tag));
2868 memset(t, 0, sizeof(struct tag));
2869 t->name = pool_strdup(arg);
2870 if (last_tag)
2871 last_tag->next_tag = t;
2872 else
2873 first_tag = t;
2874 last_tag = t;
2875 read_next_command();
2876
2877 /* from ... */
2878 if (!skip_prefix(command_buf.buf, "from ", &from))
2879 die("Expected from command, got %s", command_buf.buf);
2880 s = lookup_branch(from);
2881 if (s) {
2882 if (is_null_oid(&s->oid))
2883 die("Can't tag an empty branch.");
2884 oidcpy(&oid, &s->oid);
2885 type = OBJ_COMMIT;
2886 } else if (*from == ':') {
2887 struct object_entry *oe;
2888 from_mark = parse_mark_ref_eol(from);
2889 oe = find_mark(from_mark);
2890 type = oe->type;
2891 oidcpy(&oid, &oe->idx.oid);
2892 } else if (!get_oid(from, &oid)) {
2893 struct object_entry *oe = find_object(&oid);
2894 if (!oe) {
2895 type = sha1_object_info(oid.hash, NULL);
2896 if (type < 0)
2897 die("Not a valid object: %s", from);
2898 } else
2899 type = oe->type;
2900 } else
2901 die("Invalid ref name or SHA1 expression: %s", from);
2902 read_next_command();
2903
2904 /* tagger ... */
2905 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2906 tagger = parse_ident(v);
2907 read_next_command();
2908 } else
2909 tagger = NULL;
2910
2911 /* tag payload/message */
2912 parse_data(&msg, 0, NULL);
2913
2914 /* build the tag object */
2915 strbuf_reset(&new_data);
2916
2917 strbuf_addf(&new_data,
2918 "object %s\n"
2919 "type %s\n"
2920 "tag %s\n",
2921 oid_to_hex(&oid), typename(type), t->name);
2922 if (tagger)
2923 strbuf_addf(&new_data,
2924 "tagger %s\n", tagger);
2925 strbuf_addch(&new_data, '\n');
2926 strbuf_addbuf(&new_data, &msg);
2927 free(tagger);
2928
2929 if (store_object(OBJ_TAG, &new_data, NULL, &t->oid, 0))
2930 t->pack_id = MAX_PACK_ID;
2931 else
2932 t->pack_id = pack_id;
2933 }
2934
2935 static void parse_reset_branch(const char *arg)
2936 {
2937 struct branch *b;
2938
2939 b = lookup_branch(arg);
2940 if (b) {
2941 oidclr(&b->oid);
2942 oidclr(&b->branch_tree.versions[0].oid);
2943 oidclr(&b->branch_tree.versions[1].oid);
2944 if (b->branch_tree.tree) {
2945 release_tree_content_recursive(b->branch_tree.tree);
2946 b->branch_tree.tree = NULL;
2947 }
2948 }
2949 else
2950 b = new_branch(arg);
2951 read_next_command();
2952 parse_from(b);
2953 if (command_buf.len > 0)
2954 unread_command_buf = 1;
2955 }
2956
2957 static void cat_blob_write(const char *buf, unsigned long size)
2958 {
2959 if (write_in_full(cat_blob_fd, buf, size) < 0)
2960 die_errno("Write to frontend failed");
2961 }
2962
2963 static void cat_blob(struct object_entry *oe, struct object_id *oid)
2964 {
2965 struct strbuf line = STRBUF_INIT;
2966 unsigned long size;
2967 enum object_type type = 0;
2968 char *buf;
2969
2970 if (!oe || oe->pack_id == MAX_PACK_ID) {
2971 buf = read_sha1_file(oid->hash, &type, &size);
2972 } else {
2973 type = oe->type;
2974 buf = gfi_unpack_entry(oe, &size);
2975 }
2976
2977 /*
2978 * Output based on batch_one_object() from cat-file.c.
2979 */
2980 if (type <= 0) {
2981 strbuf_reset(&line);
2982 strbuf_addf(&line, "%s missing\n", oid_to_hex(oid));
2983 cat_blob_write(line.buf, line.len);
2984 strbuf_release(&line);
2985 free(buf);
2986 return;
2987 }
2988 if (!buf)
2989 die("Can't read object %s", oid_to_hex(oid));
2990 if (type != OBJ_BLOB)
2991 die("Object %s is a %s but a blob was expected.",
2992 oid_to_hex(oid), typename(type));
2993 strbuf_reset(&line);
2994 strbuf_addf(&line, "%s %s %lu\n", oid_to_hex(oid),
2995 typename(type), size);
2996 cat_blob_write(line.buf, line.len);
2997 strbuf_release(&line);
2998 cat_blob_write(buf, size);
2999 cat_blob_write("\n", 1);
3000 if (oe && oe->pack_id == pack_id) {
3001 last_blob.offset = oe->idx.offset;
3002 strbuf_attach(&last_blob.data, buf, size, size);
3003 last_blob.depth = oe->depth;
3004 } else
3005 free(buf);
3006 }
3007
3008 static void parse_get_mark(const char *p)
3009 {
3010 struct object_entry *oe = oe;
3011 char output[GIT_MAX_HEXSZ + 2];
3012
3013 /* get-mark SP <object> LF */
3014 if (*p != ':')
3015 die("Not a mark: %s", p);
3016
3017 oe = find_mark(parse_mark_ref_eol(p));
3018 if (!oe)
3019 die("Unknown mark: %s", command_buf.buf);
3020
3021 xsnprintf(output, sizeof(output), "%s\n", oid_to_hex(&oe->idx.oid));
3022 cat_blob_write(output, GIT_SHA1_HEXSZ + 1);
3023 }
3024
3025 static void parse_cat_blob(const char *p)
3026 {
3027 struct object_entry *oe = oe;
3028 struct object_id oid;
3029
3030 /* cat-blob SP <object> LF */
3031 if (*p == ':') {
3032 oe = find_mark(parse_mark_ref_eol(p));
3033 if (!oe)
3034 die("Unknown mark: %s", command_buf.buf);
3035 oidcpy(&oid, &oe->idx.oid);
3036 } else {
3037 if (parse_oid_hex(p, &oid, &p))
3038 die("Invalid dataref: %s", command_buf.buf);
3039 if (*p)
3040 die("Garbage after SHA1: %s", command_buf.buf);
3041 oe = find_object(&oid);
3042 }
3043
3044 cat_blob(oe, &oid);
3045 }
3046
3047 static struct object_entry *dereference(struct object_entry *oe,
3048 struct object_id *oid)
3049 {
3050 unsigned long size;
3051 char *buf = NULL;
3052 if (!oe) {
3053 enum object_type type = sha1_object_info(oid->hash, NULL);
3054 if (type < 0)
3055 die("object not found: %s", oid_to_hex(oid));
3056 /* cache it! */
3057 oe = insert_object(oid);
3058 oe->type = type;
3059 oe->pack_id = MAX_PACK_ID;
3060 oe->idx.offset = 1;
3061 }
3062 switch (oe->type) {
3063 case OBJ_TREE: /* easy case. */
3064 return oe;
3065 case OBJ_COMMIT:
3066 case OBJ_TAG:
3067 break;
3068 default:
3069 die("Not a tree-ish: %s", command_buf.buf);
3070 }
3071
3072 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
3073 buf = gfi_unpack_entry(oe, &size);
3074 } else {
3075 enum object_type unused;
3076 buf = read_sha1_file(oid->hash, &unused, &size);
3077 }
3078 if (!buf)
3079 die("Can't load object %s", oid_to_hex(oid));
3080
3081 /* Peel one layer. */
3082 switch (oe->type) {
3083 case OBJ_TAG:
3084 if (size < GIT_SHA1_HEXSZ + strlen("object ") ||
3085 get_oid_hex(buf + strlen("object "), oid))
3086 die("Invalid SHA1 in tag: %s", command_buf.buf);
3087 break;
3088 case OBJ_COMMIT:
3089 if (size < GIT_SHA1_HEXSZ + strlen("tree ") ||
3090 get_oid_hex(buf + strlen("tree "), oid))
3091 die("Invalid SHA1 in commit: %s", command_buf.buf);
3092 }
3093
3094 free(buf);
3095 return find_object(oid);
3096 }
3097
3098 static struct object_entry *parse_treeish_dataref(const char **p)
3099 {
3100 struct object_id oid;
3101 struct object_entry *e;
3102
3103 if (**p == ':') { /* <mark> */
3104 e = find_mark(parse_mark_ref_space(p));
3105 if (!e)
3106 die("Unknown mark: %s", command_buf.buf);
3107 oidcpy(&oid, &e->idx.oid);
3108 } else { /* <sha1> */
3109 if (parse_oid_hex(*p, &oid, p))
3110 die("Invalid dataref: %s", command_buf.buf);
3111 e = find_object(&oid);
3112 if (*(*p)++ != ' ')
3113 die("Missing space after tree-ish: %s", command_buf.buf);
3114 }
3115
3116 while (!e || e->type != OBJ_TREE)
3117 e = dereference(e, &oid);
3118 return e;
3119 }
3120
3121 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3122 {
3123 static struct strbuf line = STRBUF_INIT;
3124
3125 /* See show_tree(). */
3126 const char *type =
3127 S_ISGITLINK(mode) ? commit_type :
3128 S_ISDIR(mode) ? tree_type :
3129 blob_type;
3130
3131 if (!mode) {
3132 /* missing SP path LF */
3133 strbuf_reset(&line);
3134 strbuf_addstr(&line, "missing ");
3135 quote_c_style(path, &line, NULL, 0);
3136 strbuf_addch(&line, '\n');
3137 } else {
3138 /* mode SP type SP object_name TAB path LF */
3139 strbuf_reset(&line);
3140 strbuf_addf(&line, "%06o %s %s\t",
3141 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3142 quote_c_style(path, &line, NULL, 0);
3143 strbuf_addch(&line, '\n');
3144 }
3145 cat_blob_write(line.buf, line.len);
3146 }
3147
3148 static void parse_ls(const char *p, struct branch *b)
3149 {
3150 struct tree_entry *root = NULL;
3151 struct tree_entry leaf = {NULL};
3152
3153 /* ls SP (<tree-ish> SP)? <path> */
3154 if (*p == '"') {
3155 if (!b)
3156 die("Not in a commit: %s", command_buf.buf);
3157 root = &b->branch_tree;
3158 } else {
3159 struct object_entry *e = parse_treeish_dataref(&p);
3160 root = new_tree_entry();
3161 oidcpy(&root->versions[1].oid, &e->idx.oid);
3162 if (!is_null_oid(&root->versions[1].oid))
3163 root->versions[1].mode = S_IFDIR;
3164 load_tree(root);
3165 }
3166 if (*p == '"') {
3167 static struct strbuf uq = STRBUF_INIT;
3168 const char *endp;
3169 strbuf_reset(&uq);
3170 if (unquote_c_style(&uq, p, &endp))
3171 die("Invalid path: %s", command_buf.buf);
3172 if (*endp)
3173 die("Garbage after path in: %s", command_buf.buf);
3174 p = uq.buf;
3175 }
3176 tree_content_get(root, p, &leaf, 1);
3177 /*
3178 * A directory in preparation would have a sha1 of zero
3179 * until it is saved. Save, for simplicity.
3180 */
3181 if (S_ISDIR(leaf.versions[1].mode))
3182 store_tree(&leaf);
3183
3184 print_ls(leaf.versions[1].mode, leaf.versions[1].oid.hash, p);
3185 if (leaf.tree)
3186 release_tree_content_recursive(leaf.tree);
3187 if (!b || root != &b->branch_tree)
3188 release_tree_entry(root);
3189 }
3190
3191 static void checkpoint(void)
3192 {
3193 checkpoint_requested = 0;
3194 if (object_count) {
3195 cycle_packfile();
3196 }
3197 dump_branches();
3198 dump_tags();
3199 dump_marks();
3200 }
3201
3202 static void parse_checkpoint(void)
3203 {
3204 checkpoint_requested = 1;
3205 skip_optional_lf();
3206 }
3207
3208 static void parse_progress(void)
3209 {
3210 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3211 fputc('\n', stdout);
3212 fflush(stdout);
3213 skip_optional_lf();
3214 }
3215
3216 static char* make_fast_import_path(const char *path)
3217 {
3218 if (!relative_marks_paths || is_absolute_path(path))
3219 return xstrdup(path);
3220 return git_pathdup("info/fast-import/%s", path);
3221 }
3222
3223 static void option_import_marks(const char *marks,
3224 int from_stream, int ignore_missing)
3225 {
3226 if (import_marks_file) {
3227 if (from_stream)
3228 die("Only one import-marks command allowed per stream");
3229
3230 /* read previous mark file */
3231 if(!import_marks_file_from_stream)
3232 read_marks();
3233 }
3234
3235 import_marks_file = make_fast_import_path(marks);
3236 safe_create_leading_directories_const(import_marks_file);
3237 import_marks_file_from_stream = from_stream;
3238 import_marks_file_ignore_missing = ignore_missing;
3239 }
3240
3241 static void option_date_format(const char *fmt)
3242 {
3243 if (!strcmp(fmt, "raw"))
3244 whenspec = WHENSPEC_RAW;
3245 else if (!strcmp(fmt, "rfc2822"))
3246 whenspec = WHENSPEC_RFC2822;
3247 else if (!strcmp(fmt, "now"))
3248 whenspec = WHENSPEC_NOW;
3249 else
3250 die("unknown --date-format argument %s", fmt);
3251 }
3252
3253 static unsigned long ulong_arg(const char *option, const char *arg)
3254 {
3255 char *endptr;
3256 unsigned long rv = strtoul(arg, &endptr, 0);
3257 if (strchr(arg, '-') || endptr == arg || *endptr)
3258 die("%s: argument must be a non-negative integer", option);
3259 return rv;
3260 }
3261
3262 static void option_depth(const char *depth)
3263 {
3264 max_depth = ulong_arg("--depth", depth);
3265 if (max_depth > MAX_DEPTH)
3266 die("--depth cannot exceed %u", MAX_DEPTH);
3267 }
3268
3269 static void option_active_branches(const char *branches)
3270 {
3271 max_active_branches = ulong_arg("--active-branches", branches);
3272 }
3273
3274 static void option_export_marks(const char *marks)
3275 {
3276 export_marks_file = make_fast_import_path(marks);
3277 safe_create_leading_directories_const(export_marks_file);
3278 }
3279
3280 static void option_cat_blob_fd(const char *fd)
3281 {
3282 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3283 if (n > (unsigned long) INT_MAX)
3284 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3285 cat_blob_fd = (int) n;
3286 }
3287
3288 static void option_export_pack_edges(const char *edges)
3289 {
3290 if (pack_edges)
3291 fclose(pack_edges);
3292 pack_edges = xfopen(edges, "a");
3293 }
3294
3295 static int parse_one_option(const char *option)
3296 {
3297 if (skip_prefix(option, "max-pack-size=", &option)) {
3298 unsigned long v;
3299 if (!git_parse_ulong(option, &v))
3300 return 0;
3301 if (v < 8192) {
3302 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3303 v *= 1024 * 1024;
3304 } else if (v < 1024 * 1024) {
3305 warning("minimum max-pack-size is 1 MiB");
3306 v = 1024 * 1024;
3307 }
3308 max_packsize = v;
3309 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3310 unsigned long v;
3311 if (!git_parse_ulong(option, &v))
3312 return 0;
3313 big_file_threshold = v;
3314 } else if (skip_prefix(option, "depth=", &option)) {
3315 option_depth(option);
3316 } else if (skip_prefix(option, "active-branches=", &option)) {
3317 option_active_branches(option);
3318 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3319 option_export_pack_edges(option);
3320 } else if (starts_with(option, "quiet")) {
3321 show_stats = 0;
3322 } else if (starts_with(option, "stats")) {
3323 show_stats = 1;
3324 } else {
3325 return 0;
3326 }
3327
3328 return 1;
3329 }
3330
3331 static int parse_one_feature(const char *feature, int from_stream)
3332 {
3333 const char *arg;
3334
3335 if (skip_prefix(feature, "date-format=", &arg)) {
3336 option_date_format(arg);
3337 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3338 option_import_marks(arg, from_stream, 0);
3339 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3340 option_import_marks(arg, from_stream, 1);
3341 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3342 option_export_marks(arg);
3343 } else if (!strcmp(feature, "get-mark")) {
3344 ; /* Don't die - this feature is supported */
3345 } else if (!strcmp(feature, "cat-blob")) {
3346 ; /* Don't die - this feature is supported */
3347 } else if (!strcmp(feature, "relative-marks")) {
3348 relative_marks_paths = 1;
3349 } else if (!strcmp(feature, "no-relative-marks")) {
3350 relative_marks_paths = 0;
3351 } else if (!strcmp(feature, "done")) {
3352 require_explicit_termination = 1;
3353 } else if (!strcmp(feature, "force")) {
3354 force_update = 1;
3355 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3356 ; /* do nothing; we have the feature */
3357 } else {
3358 return 0;
3359 }
3360
3361 return 1;
3362 }
3363
3364 static void parse_feature(const char *feature)
3365 {
3366 if (seen_data_command)
3367 die("Got feature command '%s' after data command", feature);
3368
3369 if (parse_one_feature(feature, 1))
3370 return;
3371
3372 die("This version of fast-import does not support feature %s.", feature);
3373 }
3374
3375 static void parse_option(const char *option)
3376 {
3377 if (seen_data_command)
3378 die("Got option command '%s' after data command", option);
3379
3380 if (parse_one_option(option))
3381 return;
3382
3383 die("This version of fast-import does not support option: %s", option);
3384 }
3385
3386 static void git_pack_config(void)
3387 {
3388 int indexversion_value;
3389 int limit;
3390 unsigned long packsizelimit_value;
3391
3392 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3393 if (max_depth > MAX_DEPTH)
3394 max_depth = MAX_DEPTH;
3395 }
3396 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3397 pack_idx_opts.version = indexversion_value;
3398 if (pack_idx_opts.version > 2)
3399 git_die_config("pack.indexversion",
3400 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3401 }
3402 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3403 max_packsize = packsizelimit_value;
3404
3405 if (!git_config_get_int("fastimport.unpacklimit", &limit))
3406 unpack_limit = limit;
3407 else if (!git_config_get_int("transfer.unpacklimit", &limit))
3408 unpack_limit = limit;
3409
3410 git_config(git_default_config, NULL);
3411 }
3412
3413 static const char fast_import_usage[] =
3414 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3415
3416 static void parse_argv(void)
3417 {
3418 unsigned int i;
3419
3420 for (i = 1; i < global_argc; i++) {
3421 const char *a = global_argv[i];
3422
3423 if (*a != '-' || !strcmp(a, "--"))
3424 break;
3425
3426 if (!skip_prefix(a, "--", &a))
3427 die("unknown option %s", a);
3428
3429 if (parse_one_option(a))
3430 continue;
3431
3432 if (parse_one_feature(a, 0))
3433 continue;
3434
3435 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3436 option_cat_blob_fd(a);
3437 continue;
3438 }
3439
3440 die("unknown option --%s", a);
3441 }
3442 if (i != global_argc)
3443 usage(fast_import_usage);
3444
3445 seen_data_command = 1;
3446 if (import_marks_file)
3447 read_marks();
3448 }
3449
3450 int cmd_main(int argc, const char **argv)
3451 {
3452 unsigned int i;
3453
3454 if (argc == 2 && !strcmp(argv[1], "-h"))
3455 usage(fast_import_usage);
3456
3457 setup_git_directory();
3458 reset_pack_idx_option(&pack_idx_opts);
3459 git_pack_config();
3460
3461 alloc_objects(object_entry_alloc);
3462 strbuf_init(&command_buf, 0);
3463 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3464 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3465 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3466 marks = pool_calloc(1, sizeof(struct mark_set));
3467
3468 global_argc = argc;
3469 global_argv = argv;
3470
3471 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3472 for (i = 0; i < (cmd_save - 1); i++)
3473 rc_free[i].next = &rc_free[i + 1];
3474 rc_free[cmd_save - 1].next = NULL;
3475
3476 prepare_packed_git();
3477 start_packfile();
3478 set_die_routine(die_nicely);
3479 set_checkpoint_signal();
3480 while (read_next_command() != EOF) {
3481 const char *v;
3482 if (!strcmp("blob", command_buf.buf))
3483 parse_new_blob();
3484 else if (skip_prefix(command_buf.buf, "ls ", &v))
3485 parse_ls(v, NULL);
3486 else if (skip_prefix(command_buf.buf, "commit ", &v))
3487 parse_new_commit(v);
3488 else if (skip_prefix(command_buf.buf, "tag ", &v))
3489 parse_new_tag(v);
3490 else if (skip_prefix(command_buf.buf, "reset ", &v))
3491 parse_reset_branch(v);
3492 else if (!strcmp("checkpoint", command_buf.buf))
3493 parse_checkpoint();
3494 else if (!strcmp("done", command_buf.buf))
3495 break;
3496 else if (starts_with(command_buf.buf, "progress "))
3497 parse_progress();
3498 else if (skip_prefix(command_buf.buf, "feature ", &v))
3499 parse_feature(v);
3500 else if (skip_prefix(command_buf.buf, "option git ", &v))
3501 parse_option(v);
3502 else if (starts_with(command_buf.buf, "option "))
3503 /* ignore non-git options*/;
3504 else
3505 die("Unsupported command: %s", command_buf.buf);
3506
3507 if (checkpoint_requested)
3508 checkpoint();
3509 }
3510
3511 /* argv hasn't been parsed yet, do so */
3512 if (!seen_data_command)
3513 parse_argv();
3514
3515 if (require_explicit_termination && feof(stdin))
3516 die("stream ends early");
3517
3518 end_packfile();
3519
3520 dump_branches();
3521 dump_tags();
3522 unkeep_all_packs();
3523 dump_marks();
3524
3525 if (pack_edges)
3526 fclose(pack_edges);
3527
3528 if (show_stats) {
3529 uintmax_t total_count = 0, duplicate_count = 0;
3530 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3531 total_count += object_count_by_type[i];
3532 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3533 duplicate_count += duplicate_count_by_type[i];
3534
3535 fprintf(stderr, "%s statistics:\n", argv[0]);
3536 fprintf(stderr, "---------------------------------------------------------------------\n");
3537 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3538 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3539 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB], delta_count_attempts_by_type[OBJ_BLOB]);
3540 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE], delta_count_attempts_by_type[OBJ_TREE]);
3541 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT], delta_count_attempts_by_type[OBJ_COMMIT]);
3542 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG], delta_count_attempts_by_type[OBJ_TAG]);
3543 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3544 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3545 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3546 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3547 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3548 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3549 fprintf(stderr, "---------------------------------------------------------------------\n");
3550 pack_report();
3551 fprintf(stderr, "---------------------------------------------------------------------\n");
3552 fprintf(stderr, "\n");
3553 }
3554
3555 return failure ? 1 : 0;
3556 }