![[tokkee]](http://tokkee.org/images/avatar.png){kind=avatar}
1 /*
2 (See Documentation/git-fast-import.txt for maintained documentation.)
3 Format of STDIN stream:
5 stream ::= cmd*;
7 cmd ::= new_blob
8 | new_commit
9 | new_tag
10 | reset_branch
11 | checkpoint
12 | progress
13 ;
15 new_blob ::= 'blob' lf
16 mark?
17 file_content;
18 file_content ::= data;
20 new_commit ::= 'commit' sp ref_str lf
21 mark?
22 ('author' (sp name)? sp '<' email '>' sp when lf)?
23 'committer' (sp name)? sp '<' email '>' sp when lf
24 commit_msg
25 ('from' sp committish lf)?
26 ('merge' sp committish lf)*
27 file_change*
28 lf?;
29 commit_msg ::= data;
31 file_change ::= file_clr
32 | file_del
33 | file_rnm
34 | file_cpy
35 | file_obm
36 | file_inm;
37 file_clr ::= 'deleteall' lf;
38 file_del ::= 'D' sp path_str lf;
39 file_rnm ::= 'R' sp path_str sp path_str lf;
40 file_cpy ::= 'C' sp path_str sp path_str lf;
41 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
42 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
43 data;
44 note_obm ::= 'N' sp (hexsha1 | idnum) sp committish lf;
45 note_inm ::= 'N' sp 'inline' sp committish lf
46 data;
48 new_tag ::= 'tag' sp tag_str lf
49 'from' sp committish lf
50 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
51 tag_msg;
52 tag_msg ::= data;
54 reset_branch ::= 'reset' sp ref_str lf
55 ('from' sp committish lf)?
56 lf?;
58 checkpoint ::= 'checkpoint' lf
59 lf?;
61 progress ::= 'progress' sp not_lf* lf
62 lf?;
64 # note: the first idnum in a stream should be 1 and subsequent
65 # idnums should not have gaps between values as this will cause
66 # the stream parser to reserve space for the gapped values. An
67 # idnum can be updated in the future to a new object by issuing
68 # a new mark directive with the old idnum.
69 #
70 mark ::= 'mark' sp idnum lf;
71 data ::= (delimited_data | exact_data)
72 lf?;
74 # note: delim may be any string but must not contain lf.
75 # data_line may contain any data but must not be exactly
76 # delim.
77 delimited_data ::= 'data' sp '<<' delim lf
78 (data_line lf)*
79 delim lf;
81 # note: declen indicates the length of binary_data in bytes.
82 # declen does not include the lf preceding the binary data.
83 #
84 exact_data ::= 'data' sp declen lf
85 binary_data;
87 # note: quoted strings are C-style quoting supporting \c for
88 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
89 # is the signed byte value in octal. Note that the only
90 # characters which must actually be escaped to protect the
91 # stream formatting is: \, " and LF. Otherwise these values
92 # are UTF8.
93 #
94 committish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
95 ref_str ::= ref;
96 sha1exp_str ::= sha1exp;
97 tag_str ::= tag;
98 path_str ::= path | '"' quoted(path) '"' ;
99 mode ::= '100644' | '644'
100 | '100755' | '755'
101 | '120000'
102 ;
104 declen ::= # unsigned 32 bit value, ascii base10 notation;
105 bigint ::= # unsigned integer value, ascii base10 notation;
106 binary_data ::= # file content, not interpreted;
108 when ::= raw_when | rfc2822_when;
109 raw_when ::= ts sp tz;
110 rfc2822_when ::= # Valid RFC 2822 date and time;
112 sp ::= # ASCII space character;
113 lf ::= # ASCII newline (LF) character;
115 # note: a colon (':') must precede the numerical value assigned to
116 # an idnum. This is to distinguish it from a ref or tag name as
117 # GIT does not permit ':' in ref or tag strings.
118 #
119 idnum ::= ':' bigint;
120 path ::= # GIT style file path, e.g. "a/b/c";
121 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
122 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
123 sha1exp ::= # Any valid GIT SHA1 expression;
124 hexsha1 ::= # SHA1 in hexadecimal format;
126 # note: name and email are UTF8 strings, however name must not
127 # contain '<' or lf and email must not contain any of the
128 # following: '<', '>', lf.
129 #
130 name ::= # valid GIT author/committer name;
131 email ::= # valid GIT author/committer email;
132 ts ::= # time since the epoch in seconds, ascii base10 notation;
133 tz ::= # GIT style timezone;
135 # note: comments and cat requests may appear anywhere
136 # in the input, except within a data command. Any form
137 # of the data command always escapes the related input
138 # from comment processing.
139 #
140 # In case it is not clear, the '#' that starts the comment
141 # must be the first character on that line (an lf
142 # preceded it).
143 #
144 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
146 comment ::= '#' not_lf* lf;
147 not_lf ::= # Any byte that is not ASCII newline (LF);
148 */
150 #include "builtin.h"
151 #include "cache.h"
152 #include "object.h"
153 #include "blob.h"
154 #include "tree.h"
155 #include "commit.h"
156 #include "delta.h"
157 #include "pack.h"
158 #include "refs.h"
159 #include "csum-file.h"
160 #include "quote.h"
161 #include "exec_cmd.h"
162 #include "dir.h"
164 #define PACK_ID_BITS 16
165 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
166 #define DEPTH_BITS 13
167 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
169 struct object_entry {
170 struct pack_idx_entry idx;
171 struct object_entry *next;
172 uint32_t type : TYPE_BITS,
173 pack_id : PACK_ID_BITS,
174 depth : DEPTH_BITS;
175 };
177 struct object_entry_pool {
178 struct object_entry_pool *next_pool;
179 struct object_entry *next_free;
180 struct object_entry *end;
181 struct object_entry entries[FLEX_ARRAY]; /* more */
182 };
184 struct mark_set {
185 union {
186 struct object_entry *marked[1024];
187 struct mark_set *sets[1024];
188 } data;
189 unsigned int shift;
190 };
192 struct last_object {
193 struct strbuf data;
194 off_t offset;
195 unsigned int depth;
196 unsigned no_swap : 1;
197 };
199 struct mem_pool {
200 struct mem_pool *next_pool;
201 char *next_free;
202 char *end;
203 uintmax_t space[FLEX_ARRAY]; /* more */
204 };
206 struct atom_str {
207 struct atom_str *next_atom;
208 unsigned short str_len;
209 char str_dat[FLEX_ARRAY]; /* more */
210 };
212 struct tree_content;
213 struct tree_entry {
214 struct tree_content *tree;
215 struct atom_str *name;
216 struct tree_entry_ms {
217 uint16_t mode;
218 unsigned char sha1[20];
219 } versions[2];
220 };
222 struct tree_content {
223 unsigned int entry_capacity; /* must match avail_tree_content */
224 unsigned int entry_count;
225 unsigned int delta_depth;
226 struct tree_entry *entries[FLEX_ARRAY]; /* more */
227 };
229 struct avail_tree_content {
230 unsigned int entry_capacity; /* must match tree_content */
231 struct avail_tree_content *next_avail;
232 };
234 struct branch {
235 struct branch *table_next_branch;
236 struct branch *active_next_branch;
237 const char *name;
238 struct tree_entry branch_tree;
239 uintmax_t last_commit;
240 uintmax_t num_notes;
241 unsigned active : 1;
242 unsigned pack_id : PACK_ID_BITS;
243 unsigned char sha1[20];
244 };
246 struct tag {
247 struct tag *next_tag;
248 const char *name;
249 unsigned int pack_id;
250 unsigned char sha1[20];
251 };
253 struct hash_list {
254 struct hash_list *next;
255 unsigned char sha1[20];
256 };
258 typedef enum {
259 WHENSPEC_RAW = 1,
260 WHENSPEC_RFC2822,
261 WHENSPEC_NOW
262 } whenspec_type;
264 struct recent_command {
265 struct recent_command *prev;
266 struct recent_command *next;
267 char *buf;
268 };
270 /* Configured limits on output */
271 static unsigned long max_depth = 10;
272 static off_t max_packsize;
273 static uintmax_t big_file_threshold = 512 * 1024 * 1024;
274 static int force_update;
275 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
276 static int pack_compression_seen;
278 /* Stats and misc. counters */
279 static uintmax_t alloc_count;
280 static uintmax_t marks_set_count;
281 static uintmax_t object_count_by_type[1 << TYPE_BITS];
282 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
283 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
284 static unsigned long object_count;
285 static unsigned long branch_count;
286 static unsigned long branch_load_count;
287 static int failure;
288 static FILE *pack_edges;
289 static unsigned int show_stats = 1;
290 static int global_argc;
291 static const char **global_argv;
293 /* Memory pools */
294 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
295 static size_t total_allocd;
296 static struct mem_pool *mem_pool;
298 /* Atom management */
299 static unsigned int atom_table_sz = 4451;
300 static unsigned int atom_cnt;
301 static struct atom_str **atom_table;
303 /* The .pack file being generated */
304 static unsigned int pack_id;
305 static struct sha1file *pack_file;
306 static struct packed_git *pack_data;
307 static struct packed_git **all_packs;
308 static off_t pack_size;
310 /* Table of objects we've written. */
311 static unsigned int object_entry_alloc = 5000;
312 static struct object_entry_pool *blocks;
313 static struct object_entry *object_table[1 << 16];
314 static struct mark_set *marks;
315 static const char *export_marks_file;
316 static const char *import_marks_file;
317 static int import_marks_file_from_stream;
318 static int relative_marks_paths;
320 /* Our last blob */
321 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
323 /* Tree management */
324 static unsigned int tree_entry_alloc = 1000;
325 static void *avail_tree_entry;
326 static unsigned int avail_tree_table_sz = 100;
327 static struct avail_tree_content **avail_tree_table;
328 static struct strbuf old_tree = STRBUF_INIT;
329 static struct strbuf new_tree = STRBUF_INIT;
331 /* Branch data */
332 static unsigned long max_active_branches = 5;
333 static unsigned long cur_active_branches;
334 static unsigned long branch_table_sz = 1039;
335 static struct branch **branch_table;
336 static struct branch *active_branches;
338 /* Tag data */
339 static struct tag *first_tag;
340 static struct tag *last_tag;
342 /* Input stream parsing */
343 static whenspec_type whenspec = WHENSPEC_RAW;
344 static struct strbuf command_buf = STRBUF_INIT;
345 static int unread_command_buf;
346 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
347 static struct recent_command *cmd_tail = &cmd_hist;
348 static struct recent_command *rc_free;
349 static unsigned int cmd_save = 100;
350 static uintmax_t next_mark;
351 static struct strbuf new_data = STRBUF_INIT;
352 static int seen_data_command;
354 /* Signal handling */
355 static volatile sig_atomic_t checkpoint_requested;
357 /* Where to write output of cat-blob commands */
358 static int cat_blob_fd = STDOUT_FILENO;
360 static void parse_argv(void);
361 static void parse_cat_blob(void);
363 static void write_branch_report(FILE *rpt, struct branch *b)
364 {
365 fprintf(rpt, "%s:\n", b->name);
367 fprintf(rpt, " status :");
368 if (b->active)
369 fputs(" active", rpt);
370 if (b->branch_tree.tree)
371 fputs(" loaded", rpt);
372 if (is_null_sha1(b->branch_tree.versions[1].sha1))
373 fputs(" dirty", rpt);
374 fputc('\n', rpt);
376 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
377 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
378 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
379 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
381 fputs(" last pack : ", rpt);
382 if (b->pack_id < MAX_PACK_ID)
383 fprintf(rpt, "%u", b->pack_id);
384 fputc('\n', rpt);
386 fputc('\n', rpt);
387 }
389 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
391 static void write_crash_report(const char *err)
392 {
393 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
394 FILE *rpt = fopen(loc, "w");
395 struct branch *b;
396 unsigned long lu;
397 struct recent_command *rc;
399 if (!rpt) {
400 error("can't write crash report %s: %s", loc, strerror(errno));
401 return;
402 }
404 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
406 fprintf(rpt, "fast-import crash report:\n");
407 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
408 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
409 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
410 fputc('\n', rpt);
412 fputs("fatal: ", rpt);
413 fputs(err, rpt);
414 fputc('\n', rpt);
416 fputc('\n', rpt);
417 fputs("Most Recent Commands Before Crash\n", rpt);
418 fputs("---------------------------------\n", rpt);
419 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
420 if (rc->next == &cmd_hist)
421 fputs("* ", rpt);
422 else
423 fputs(" ", rpt);
424 fputs(rc->buf, rpt);
425 fputc('\n', rpt);
426 }
428 fputc('\n', rpt);
429 fputs("Active Branch LRU\n", rpt);
430 fputs("-----------------\n", rpt);
431 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
432 cur_active_branches,
433 max_active_branches);
434 fputc('\n', rpt);
435 fputs(" pos clock name\n", rpt);
436 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
437 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
438 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
439 ++lu, b->last_commit, b->name);
441 fputc('\n', rpt);
442 fputs("Inactive Branches\n", rpt);
443 fputs("-----------------\n", rpt);
444 for (lu = 0; lu < branch_table_sz; lu++) {
445 for (b = branch_table[lu]; b; b = b->table_next_branch)
446 write_branch_report(rpt, b);
447 }
449 if (first_tag) {
450 struct tag *tg;
451 fputc('\n', rpt);
452 fputs("Annotated Tags\n", rpt);
453 fputs("--------------\n", rpt);
454 for (tg = first_tag; tg; tg = tg->next_tag) {
455 fputs(sha1_to_hex(tg->sha1), rpt);
456 fputc(' ', rpt);
457 fputs(tg->name, rpt);
458 fputc('\n', rpt);
459 }
460 }
462 fputc('\n', rpt);
463 fputs("Marks\n", rpt);
464 fputs("-----\n", rpt);
465 if (export_marks_file)
466 fprintf(rpt, " exported to %s\n", export_marks_file);
467 else
468 dump_marks_helper(rpt, 0, marks);
470 fputc('\n', rpt);
471 fputs("-------------------\n", rpt);
472 fputs("END OF CRASH REPORT\n", rpt);
473 fclose(rpt);
474 }
476 static void end_packfile(void);
477 static void unkeep_all_packs(void);
478 static void dump_marks(void);
480 static NORETURN void die_nicely(const char *err, va_list params)
481 {
482 static int zombie;
483 char message[2 * PATH_MAX];
485 vsnprintf(message, sizeof(message), err, params);
486 fputs("fatal: ", stderr);
487 fputs(message, stderr);
488 fputc('\n', stderr);
490 if (!zombie) {
491 zombie = 1;
492 write_crash_report(message);
493 end_packfile();
494 unkeep_all_packs();
495 dump_marks();
496 }
497 exit(128);
498 }
500 #ifndef SIGUSR1 /* Windows, for example */
502 static void set_checkpoint_signal(void)
503 {
504 }
506 #else
508 static void checkpoint_signal(int signo)
509 {
510 checkpoint_requested = 1;
511 }
513 static void set_checkpoint_signal(void)
514 {
515 struct sigaction sa;
517 memset(&sa, 0, sizeof(sa));
518 sa.sa_handler = checkpoint_signal;
519 sigemptyset(&sa.sa_mask);
520 sa.sa_flags = SA_RESTART;
521 sigaction(SIGUSR1, &sa, NULL);
522 }
524 #endif
526 static void alloc_objects(unsigned int cnt)
527 {
528 struct object_entry_pool *b;
530 b = xmalloc(sizeof(struct object_entry_pool)
531 + cnt * sizeof(struct object_entry));
532 b->next_pool = blocks;
533 b->next_free = b->entries;
534 b->end = b->entries + cnt;
535 blocks = b;
536 alloc_count += cnt;
537 }
539 static struct object_entry *new_object(unsigned char *sha1)
540 {
541 struct object_entry *e;
543 if (blocks->next_free == blocks->end)
544 alloc_objects(object_entry_alloc);
546 e = blocks->next_free++;
547 hashcpy(e->idx.sha1, sha1);
548 return e;
549 }
551 static struct object_entry *find_object(unsigned char *sha1)
552 {
553 unsigned int h = sha1[0] << 8 | sha1[1];
554 struct object_entry *e;
555 for (e = object_table[h]; e; e = e->next)
556 if (!hashcmp(sha1, e->idx.sha1))
557 return e;
558 return NULL;
559 }
561 static struct object_entry *insert_object(unsigned char *sha1)
562 {
563 unsigned int h = sha1[0] << 8 | sha1[1];
564 struct object_entry *e = object_table[h];
566 while (e) {
567 if (!hashcmp(sha1, e->idx.sha1))
568 return e;
569 e = e->next;
570 }
572 e = new_object(sha1);
573 e->next = object_table[h];
574 e->idx.offset = 0;
575 object_table[h] = e;
576 return e;
577 }
579 static unsigned int hc_str(const char *s, size_t len)
580 {
581 unsigned int r = 0;
582 while (len-- > 0)
583 r = r * 31 + *s++;
584 return r;
585 }
587 static void *pool_alloc(size_t len)
588 {
589 struct mem_pool *p;
590 void *r;
592 /* round up to a 'uintmax_t' alignment */
593 if (len & (sizeof(uintmax_t) - 1))
594 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
596 for (p = mem_pool; p; p = p->next_pool)
597 if ((p->end - p->next_free >= len))
598 break;
600 if (!p) {
601 if (len >= (mem_pool_alloc/2)) {
602 total_allocd += len;
603 return xmalloc(len);
604 }
605 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
606 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
607 p->next_pool = mem_pool;
608 p->next_free = (char *) p->space;
609 p->end = p->next_free + mem_pool_alloc;
610 mem_pool = p;
611 }
613 r = p->next_free;
614 p->next_free += len;
615 return r;
616 }
618 static void *pool_calloc(size_t count, size_t size)
619 {
620 size_t len = count * size;
621 void *r = pool_alloc(len);
622 memset(r, 0, len);
623 return r;
624 }
626 static char *pool_strdup(const char *s)
627 {
628 char *r = pool_alloc(strlen(s) + 1);
629 strcpy(r, s);
630 return r;
631 }
633 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
634 {
635 struct mark_set *s = marks;
636 while ((idnum >> s->shift) >= 1024) {
637 s = pool_calloc(1, sizeof(struct mark_set));
638 s->shift = marks->shift + 10;
639 s->data.sets[0] = marks;
640 marks = s;
641 }
642 while (s->shift) {
643 uintmax_t i = idnum >> s->shift;
644 idnum -= i << s->shift;
645 if (!s->data.sets[i]) {
646 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
647 s->data.sets[i]->shift = s->shift - 10;
648 }
649 s = s->data.sets[i];
650 }
651 if (!s->data.marked[idnum])
652 marks_set_count++;
653 s->data.marked[idnum] = oe;
654 }
656 static struct object_entry *find_mark(uintmax_t idnum)
657 {
658 uintmax_t orig_idnum = idnum;
659 struct mark_set *s = marks;
660 struct object_entry *oe = NULL;
661 if ((idnum >> s->shift) < 1024) {
662 while (s && s->shift) {
663 uintmax_t i = idnum >> s->shift;
664 idnum -= i << s->shift;
665 s = s->data.sets[i];
666 }
667 if (s)
668 oe = s->data.marked[idnum];
669 }
670 if (!oe)
671 die("mark :%" PRIuMAX " not declared", orig_idnum);
672 return oe;
673 }
675 static struct atom_str *to_atom(const char *s, unsigned short len)
676 {
677 unsigned int hc = hc_str(s, len) % atom_table_sz;
678 struct atom_str *c;
680 for (c = atom_table[hc]; c; c = c->next_atom)
681 if (c->str_len == len && !strncmp(s, c->str_dat, len))
682 return c;
684 c = pool_alloc(sizeof(struct atom_str) + len + 1);
685 c->str_len = len;
686 strncpy(c->str_dat, s, len);
687 c->str_dat[len] = 0;
688 c->next_atom = atom_table[hc];
689 atom_table[hc] = c;
690 atom_cnt++;
691 return c;
692 }
694 static struct branch *lookup_branch(const char *name)
695 {
696 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
697 struct branch *b;
699 for (b = branch_table[hc]; b; b = b->table_next_branch)
700 if (!strcmp(name, b->name))
701 return b;
702 return NULL;
703 }
705 static struct branch *new_branch(const char *name)
706 {
707 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
708 struct branch *b = lookup_branch(name);
710 if (b)
711 die("Invalid attempt to create duplicate branch: %s", name);
712 switch (check_ref_format(name)) {
713 case 0: break; /* its valid */
714 case CHECK_REF_FORMAT_ONELEVEL:
715 break; /* valid, but too few '/', allow anyway */
716 default:
717 die("Branch name doesn't conform to GIT standards: %s", name);
718 }
720 b = pool_calloc(1, sizeof(struct branch));
721 b->name = pool_strdup(name);
722 b->table_next_branch = branch_table[hc];
723 b->branch_tree.versions[0].mode = S_IFDIR;
724 b->branch_tree.versions[1].mode = S_IFDIR;
725 b->num_notes = 0;
726 b->active = 0;
727 b->pack_id = MAX_PACK_ID;
728 branch_table[hc] = b;
729 branch_count++;
730 return b;
731 }
733 static unsigned int hc_entries(unsigned int cnt)
734 {
735 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
736 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
737 }
739 static struct tree_content *new_tree_content(unsigned int cnt)
740 {
741 struct avail_tree_content *f, *l = NULL;
742 struct tree_content *t;
743 unsigned int hc = hc_entries(cnt);
745 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
746 if (f->entry_capacity >= cnt)
747 break;
749 if (f) {
750 if (l)
751 l->next_avail = f->next_avail;
752 else
753 avail_tree_table[hc] = f->next_avail;
754 } else {
755 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
756 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
757 f->entry_capacity = cnt;
758 }
760 t = (struct tree_content*)f;
761 t->entry_count = 0;
762 t->delta_depth = 0;
763 return t;
764 }
766 static void release_tree_entry(struct tree_entry *e);
767 static void release_tree_content(struct tree_content *t)
768 {
769 struct avail_tree_content *f = (struct avail_tree_content*)t;
770 unsigned int hc = hc_entries(f->entry_capacity);
771 f->next_avail = avail_tree_table[hc];
772 avail_tree_table[hc] = f;
773 }
775 static void release_tree_content_recursive(struct tree_content *t)
776 {
777 unsigned int i;
778 for (i = 0; i < t->entry_count; i++)
779 release_tree_entry(t->entries[i]);
780 release_tree_content(t);
781 }
783 static struct tree_content *grow_tree_content(
784 struct tree_content *t,
785 int amt)
786 {
787 struct tree_content *r = new_tree_content(t->entry_count + amt);
788 r->entry_count = t->entry_count;
789 r->delta_depth = t->delta_depth;
790 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
791 release_tree_content(t);
792 return r;
793 }
795 static struct tree_entry *new_tree_entry(void)
796 {
797 struct tree_entry *e;
799 if (!avail_tree_entry) {
800 unsigned int n = tree_entry_alloc;
801 total_allocd += n * sizeof(struct tree_entry);
802 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
803 while (n-- > 1) {
804 *((void**)e) = e + 1;
805 e++;
806 }
807 *((void**)e) = NULL;
808 }
810 e = avail_tree_entry;
811 avail_tree_entry = *((void**)e);
812 return e;
813 }
815 static void release_tree_entry(struct tree_entry *e)
816 {
817 if (e->tree)
818 release_tree_content_recursive(e->tree);
819 *((void**)e) = avail_tree_entry;
820 avail_tree_entry = e;
821 }
823 static struct tree_content *dup_tree_content(struct tree_content *s)
824 {
825 struct tree_content *d;
826 struct tree_entry *a, *b;
827 unsigned int i;
829 if (!s)
830 return NULL;
831 d = new_tree_content(s->entry_count);
832 for (i = 0; i < s->entry_count; i++) {
833 a = s->entries[i];
834 b = new_tree_entry();
835 memcpy(b, a, sizeof(*a));
836 if (a->tree && is_null_sha1(b->versions[1].sha1))
837 b->tree = dup_tree_content(a->tree);
838 else
839 b->tree = NULL;
840 d->entries[i] = b;
841 }
842 d->entry_count = s->entry_count;
843 d->delta_depth = s->delta_depth;
845 return d;
846 }
848 static void start_packfile(void)
849 {
850 static char tmpfile[PATH_MAX];
851 struct packed_git *p;
852 struct pack_header hdr;
853 int pack_fd;
855 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
856 "pack/tmp_pack_XXXXXX");
857 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
858 strcpy(p->pack_name, tmpfile);
859 p->pack_fd = pack_fd;
860 pack_file = sha1fd(pack_fd, p->pack_name);
862 hdr.hdr_signature = htonl(PACK_SIGNATURE);
863 hdr.hdr_version = htonl(2);
864 hdr.hdr_entries = 0;
865 sha1write(pack_file, &hdr, sizeof(hdr));
867 pack_data = p;
868 pack_size = sizeof(hdr);
869 object_count = 0;
871 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
872 all_packs[pack_id] = p;
873 }
875 static const char *create_index(void)
876 {
877 const char *tmpfile;
878 struct pack_idx_entry **idx, **c, **last;
879 struct object_entry *e;
880 struct object_entry_pool *o;
882 /* Build the table of object IDs. */
883 idx = xmalloc(object_count * sizeof(*idx));
884 c = idx;
885 for (o = blocks; o; o = o->next_pool)
886 for (e = o->next_free; e-- != o->entries;)
887 if (pack_id == e->pack_id)
888 *c++ = &e->idx;
889 last = idx + object_count;
890 if (c != last)
891 die("internal consistency error creating the index");
893 tmpfile = write_idx_file(NULL, idx, object_count, pack_data->sha1);
894 free(idx);
895 return tmpfile;
896 }
898 static char *keep_pack(const char *curr_index_name)
899 {
900 static char name[PATH_MAX];
901 static const char *keep_msg = "fast-import";
902 int keep_fd;
904 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
905 if (keep_fd < 0)
906 die_errno("cannot create keep file");
907 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
908 if (close(keep_fd))
909 die_errno("failed to write keep file");
911 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
912 get_object_directory(), sha1_to_hex(pack_data->sha1));
913 if (move_temp_to_file(pack_data->pack_name, name))
914 die("cannot store pack file");
916 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
917 get_object_directory(), sha1_to_hex(pack_data->sha1));
918 if (move_temp_to_file(curr_index_name, name))
919 die("cannot store index file");
920 free((void *)curr_index_name);
921 return name;
922 }
924 static void unkeep_all_packs(void)
925 {
926 static char name[PATH_MAX];
927 int k;
929 for (k = 0; k < pack_id; k++) {
930 struct packed_git *p = all_packs[k];
931 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
932 get_object_directory(), sha1_to_hex(p->sha1));
933 unlink_or_warn(name);
934 }
935 }
937 static void end_packfile(void)
938 {
939 struct packed_git *old_p = pack_data, *new_p;
941 clear_delta_base_cache();
942 if (object_count) {
943 unsigned char cur_pack_sha1[20];
944 char *idx_name;
945 int i;
946 struct branch *b;
947 struct tag *t;
949 close_pack_windows(pack_data);
950 sha1close(pack_file, cur_pack_sha1, 0);
951 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
952 pack_data->pack_name, object_count,
953 cur_pack_sha1, pack_size);
954 close(pack_data->pack_fd);
955 idx_name = keep_pack(create_index());
957 /* Register the packfile with core git's machinery. */
958 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
959 if (!new_p)
960 die("core git rejected index %s", idx_name);
961 all_packs[pack_id] = new_p;
962 install_packed_git(new_p);
964 /* Print the boundary */
965 if (pack_edges) {
966 fprintf(pack_edges, "%s:", new_p->pack_name);
967 for (i = 0; i < branch_table_sz; i++) {
968 for (b = branch_table[i]; b; b = b->table_next_branch) {
969 if (b->pack_id == pack_id)
970 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
971 }
972 }
973 for (t = first_tag; t; t = t->next_tag) {
974 if (t->pack_id == pack_id)
975 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
976 }
977 fputc('\n', pack_edges);
978 fflush(pack_edges);
979 }
981 pack_id++;
982 }
983 else {
984 close(old_p->pack_fd);
985 unlink_or_warn(old_p->pack_name);
986 }
987 free(old_p);
989 /* We can't carry a delta across packfiles. */
990 strbuf_release(&last_blob.data);
991 last_blob.offset = 0;
992 last_blob.depth = 0;
993 }
995 static void cycle_packfile(void)
996 {
997 end_packfile();
998 start_packfile();
999 }
1001 static int store_object(
1002 enum object_type type,
1003 struct strbuf *dat,
1004 struct last_object *last,
1005 unsigned char *sha1out,
1006 uintmax_t mark)
1007 {
1008 void *out, *delta;
1009 struct object_entry *e;
1010 unsigned char hdr[96];
1011 unsigned char sha1[20];
1012 unsigned long hdrlen, deltalen;
1013 git_SHA_CTX c;
1014 z_stream s;
1016 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1017 (unsigned long)dat->len) + 1;
1018 git_SHA1_Init(&c);
1019 git_SHA1_Update(&c, hdr, hdrlen);
1020 git_SHA1_Update(&c, dat->buf, dat->len);
1021 git_SHA1_Final(sha1, &c);
1022 if (sha1out)
1023 hashcpy(sha1out, sha1);
1025 e = insert_object(sha1);
1026 if (mark)
1027 insert_mark(mark, e);
1028 if (e->idx.offset) {
1029 duplicate_count_by_type[type]++;
1030 return 1;
1031 } else if (find_sha1_pack(sha1, packed_git)) {
1032 e->type = type;
1033 e->pack_id = MAX_PACK_ID;
1034 e->idx.offset = 1; /* just not zero! */
1035 duplicate_count_by_type[type]++;
1036 return 1;
1037 }
1039 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1040 delta = diff_delta(last->data.buf, last->data.len,
1041 dat->buf, dat->len,
1042 &deltalen, dat->len - 20);
1043 } else
1044 delta = NULL;
1046 memset(&s, 0, sizeof(s));
1047 deflateInit(&s, pack_compression_level);
1048 if (delta) {
1049 s.next_in = delta;
1050 s.avail_in = deltalen;
1051 } else {
1052 s.next_in = (void *)dat->buf;
1053 s.avail_in = dat->len;
1054 }
1055 s.avail_out = deflateBound(&s, s.avail_in);
1056 s.next_out = out = xmalloc(s.avail_out);
1057 while (deflate(&s, Z_FINISH) == Z_OK)
1058 /* nothing */;
1059 deflateEnd(&s);
1061 /* Determine if we should auto-checkpoint. */
1062 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1063 || (pack_size + 60 + s.total_out) < pack_size) {
1065 /* This new object needs to *not* have the current pack_id. */
1066 e->pack_id = pack_id + 1;
1067 cycle_packfile();
1069 /* We cannot carry a delta into the new pack. */
1070 if (delta) {
1071 free(delta);
1072 delta = NULL;
1074 memset(&s, 0, sizeof(s));
1075 deflateInit(&s, pack_compression_level);
1076 s.next_in = (void *)dat->buf;
1077 s.avail_in = dat->len;
1078 s.avail_out = deflateBound(&s, s.avail_in);
1079 s.next_out = out = xrealloc(out, s.avail_out);
1080 while (deflate(&s, Z_FINISH) == Z_OK)
1081 /* nothing */;
1082 deflateEnd(&s);
1083 }
1084 }
1086 e->type = type;
1087 e->pack_id = pack_id;
1088 e->idx.offset = pack_size;
1089 object_count++;
1090 object_count_by_type[type]++;
1092 crc32_begin(pack_file);
1094 if (delta) {
1095 off_t ofs = e->idx.offset - last->offset;
1096 unsigned pos = sizeof(hdr) - 1;
1098 delta_count_by_type[type]++;
1099 e->depth = last->depth + 1;
1101 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1102 sha1write(pack_file, hdr, hdrlen);
1103 pack_size += hdrlen;
1105 hdr[pos] = ofs & 127;
1106 while (ofs >>= 7)
1107 hdr[--pos] = 128 | (--ofs & 127);
1108 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1109 pack_size += sizeof(hdr) - pos;
1110 } else {
1111 e->depth = 0;
1112 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1113 sha1write(pack_file, hdr, hdrlen);
1114 pack_size += hdrlen;
1115 }
1117 sha1write(pack_file, out, s.total_out);
1118 pack_size += s.total_out;
1120 e->idx.crc32 = crc32_end(pack_file);
1122 free(out);
1123 free(delta);
1124 if (last) {
1125 if (last->no_swap) {
1126 last->data = *dat;
1127 } else {
1128 strbuf_swap(&last->data, dat);
1129 }
1130 last->offset = e->idx.offset;
1131 last->depth = e->depth;
1132 }
1133 return 0;
1134 }
1136 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1137 {
1138 if (ftruncate(pack_data->pack_fd, to)
1139 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1140 die_errno("cannot truncate pack to skip duplicate");
1141 pack_size = to;
1143 /* yes this is a layering violation */
1144 pack_file->total = to;
1145 pack_file->offset = 0;
1146 pack_file->ctx = *ctx;
1147 }
1149 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1150 {
1151 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1152 unsigned char *in_buf = xmalloc(in_sz);
1153 unsigned char *out_buf = xmalloc(out_sz);
1154 struct object_entry *e;
1155 unsigned char sha1[20];
1156 unsigned long hdrlen;
1157 off_t offset;
1158 git_SHA_CTX c;
1159 git_SHA_CTX pack_file_ctx;
1160 z_stream s;
1161 int status = Z_OK;
1163 /* Determine if we should auto-checkpoint. */
1164 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1165 || (pack_size + 60 + len) < pack_size)
1166 cycle_packfile();
1168 offset = pack_size;
1170 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1171 sha1flush(pack_file);
1172 pack_file_ctx = pack_file->ctx;
1174 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1175 if (out_sz <= hdrlen)
1176 die("impossibly large object header");
1178 git_SHA1_Init(&c);
1179 git_SHA1_Update(&c, out_buf, hdrlen);
1181 crc32_begin(pack_file);
1183 memset(&s, 0, sizeof(s));
1184 deflateInit(&s, pack_compression_level);
1186 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1187 if (out_sz <= hdrlen)
1188 die("impossibly large object header");
1190 s.next_out = out_buf + hdrlen;
1191 s.avail_out = out_sz - hdrlen;
1193 while (status != Z_STREAM_END) {
1194 if (0 < len && !s.avail_in) {
1195 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1196 size_t n = fread(in_buf, 1, cnt, stdin);
1197 if (!n && feof(stdin))
1198 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1200 git_SHA1_Update(&c, in_buf, n);
1201 s.next_in = in_buf;
1202 s.avail_in = n;
1203 len -= n;
1204 }
1206 status = deflate(&s, len ? 0 : Z_FINISH);
1208 if (!s.avail_out || status == Z_STREAM_END) {
1209 size_t n = s.next_out - out_buf;
1210 sha1write(pack_file, out_buf, n);
1211 pack_size += n;
1212 s.next_out = out_buf;
1213 s.avail_out = out_sz;
1214 }
1216 switch (status) {
1217 case Z_OK:
1218 case Z_BUF_ERROR:
1219 case Z_STREAM_END:
1220 continue;
1221 default:
1222 die("unexpected deflate failure: %d", status);
1223 }
1224 }
1225 deflateEnd(&s);
1226 git_SHA1_Final(sha1, &c);
1228 if (sha1out)
1229 hashcpy(sha1out, sha1);
1231 e = insert_object(sha1);
1233 if (mark)
1234 insert_mark(mark, e);
1236 if (e->idx.offset) {
1237 duplicate_count_by_type[OBJ_BLOB]++;
1238 truncate_pack(offset, &pack_file_ctx);
1240 } else if (find_sha1_pack(sha1, packed_git)) {
1241 e->type = OBJ_BLOB;
1242 e->pack_id = MAX_PACK_ID;
1243 e->idx.offset = 1; /* just not zero! */
1244 duplicate_count_by_type[OBJ_BLOB]++;
1245 truncate_pack(offset, &pack_file_ctx);
1247 } else {
1248 e->depth = 0;
1249 e->type = OBJ_BLOB;
1250 e->pack_id = pack_id;
1251 e->idx.offset = offset;
1252 e->idx.crc32 = crc32_end(pack_file);
1253 object_count++;
1254 object_count_by_type[OBJ_BLOB]++;
1255 }
1257 free(in_buf);
1258 free(out_buf);
1259 }
1261 /* All calls must be guarded by find_object() or find_mark() to
1262 * ensure the 'struct object_entry' passed was written by this
1263 * process instance. We unpack the entry by the offset, avoiding
1264 * the need for the corresponding .idx file. This unpacking rule
1265 * works because we only use OBJ_REF_DELTA within the packfiles
1266 * created by fast-import.
1267 *
1268 * oe must not be NULL. Such an oe usually comes from giving
1269 * an unknown SHA-1 to find_object() or an undefined mark to
1270 * find_mark(). Callers must test for this condition and use
1271 * the standard read_sha1_file() when it happens.
1272 *
1273 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1274 * find_mark(), where the mark was reloaded from an existing marks
1275 * file and is referencing an object that this fast-import process
1276 * instance did not write out to a packfile. Callers must test for
1277 * this condition and use read_sha1_file() instead.
1278 */
1279 static void *gfi_unpack_entry(
1280 struct object_entry *oe,
1281 unsigned long *sizep)
1282 {
1283 enum object_type type;
1284 struct packed_git *p = all_packs[oe->pack_id];
1285 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1286 /* The object is stored in the packfile we are writing to
1287 * and we have modified it since the last time we scanned
1288 * back to read a previously written object. If an old
1289 * window covered [p->pack_size, p->pack_size + 20) its
1290 * data is stale and is not valid. Closing all windows
1291 * and updating the packfile length ensures we can read
1292 * the newly written data.
1293 */
1294 close_pack_windows(p);
1295 sha1flush(pack_file);
1297 /* We have to offer 20 bytes additional on the end of
1298 * the packfile as the core unpacker code assumes the
1299 * footer is present at the file end and must promise
1300 * at least 20 bytes within any window it maps. But
1301 * we don't actually create the footer here.
1302 */
1303 p->pack_size = pack_size + 20;
1304 }
1305 return unpack_entry(p, oe->idx.offset, &type, sizep);
1306 }
1308 static const char *get_mode(const char *str, uint16_t *modep)
1309 {
1310 unsigned char c;
1311 uint16_t mode = 0;
1313 while ((c = *str++) != ' ') {
1314 if (c < '0' || c > '7')
1315 return NULL;
1316 mode = (mode << 3) + (c - '0');
1317 }
1318 *modep = mode;
1319 return str;
1320 }
1322 static void load_tree(struct tree_entry *root)
1323 {
1324 unsigned char *sha1 = root->versions[1].sha1;
1325 struct object_entry *myoe;
1326 struct tree_content *t;
1327 unsigned long size;
1328 char *buf;
1329 const char *c;
1331 root->tree = t = new_tree_content(8);
1332 if (is_null_sha1(sha1))
1333 return;
1335 myoe = find_object(sha1);
1336 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1337 if (myoe->type != OBJ_TREE)
1338 die("Not a tree: %s", sha1_to_hex(sha1));
1339 t->delta_depth = myoe->depth;
1340 buf = gfi_unpack_entry(myoe, &size);
1341 if (!buf)
1342 die("Can't load tree %s", sha1_to_hex(sha1));
1343 } else {
1344 enum object_type type;
1345 buf = read_sha1_file(sha1, &type, &size);
1346 if (!buf || type != OBJ_TREE)
1347 die("Can't load tree %s", sha1_to_hex(sha1));
1348 }
1350 c = buf;
1351 while (c != (buf + size)) {
1352 struct tree_entry *e = new_tree_entry();
1354 if (t->entry_count == t->entry_capacity)
1355 root->tree = t = grow_tree_content(t, t->entry_count);
1356 t->entries[t->entry_count++] = e;
1358 e->tree = NULL;
1359 c = get_mode(c, &e->versions[1].mode);
1360 if (!c)
1361 die("Corrupt mode in %s", sha1_to_hex(sha1));
1362 e->versions[0].mode = e->versions[1].mode;
1363 e->name = to_atom(c, strlen(c));
1364 c += e->name->str_len + 1;
1365 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1366 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1367 c += 20;
1368 }
1369 free(buf);
1370 }
1372 static int tecmp0 (const void *_a, const void *_b)
1373 {
1374 struct tree_entry *a = *((struct tree_entry**)_a);
1375 struct tree_entry *b = *((struct tree_entry**)_b);
1376 return base_name_compare(
1377 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1378 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1379 }
1381 static int tecmp1 (const void *_a, const void *_b)
1382 {
1383 struct tree_entry *a = *((struct tree_entry**)_a);
1384 struct tree_entry *b = *((struct tree_entry**)_b);
1385 return base_name_compare(
1386 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1387 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1388 }
1390 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1391 {
1392 size_t maxlen = 0;
1393 unsigned int i;
1395 if (!v)
1396 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1397 else
1398 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1400 for (i = 0; i < t->entry_count; i++) {
1401 if (t->entries[i]->versions[v].mode)
1402 maxlen += t->entries[i]->name->str_len + 34;
1403 }
1405 strbuf_reset(b);
1406 strbuf_grow(b, maxlen);
1407 for (i = 0; i < t->entry_count; i++) {
1408 struct tree_entry *e = t->entries[i];
1409 if (!e->versions[v].mode)
1410 continue;
1411 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1412 e->name->str_dat, '\0');
1413 strbuf_add(b, e->versions[v].sha1, 20);
1414 }
1415 }
1417 static void store_tree(struct tree_entry *root)
1418 {
1419 struct tree_content *t = root->tree;
1420 unsigned int i, j, del;
1421 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1422 struct object_entry *le;
1424 if (!is_null_sha1(root->versions[1].sha1))
1425 return;
1427 for (i = 0; i < t->entry_count; i++) {
1428 if (t->entries[i]->tree)
1429 store_tree(t->entries[i]);
1430 }
1432 le = find_object(root->versions[0].sha1);
1433 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1434 mktree(t, 0, &old_tree);
1435 lo.data = old_tree;
1436 lo.offset = le->idx.offset;
1437 lo.depth = t->delta_depth;
1438 }
1440 mktree(t, 1, &new_tree);
1441 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1443 t->delta_depth = lo.depth;
1444 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1445 struct tree_entry *e = t->entries[i];
1446 if (e->versions[1].mode) {
1447 e->versions[0].mode = e->versions[1].mode;
1448 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1449 t->entries[j++] = e;
1450 } else {
1451 release_tree_entry(e);
1452 del++;
1453 }
1454 }
1455 t->entry_count -= del;
1456 }
1458 static void tree_content_replace(
1459 struct tree_entry *root,
1460 const unsigned char *sha1,
1461 const uint16_t mode,
1462 struct tree_content *newtree)
1463 {
1464 if (!S_ISDIR(mode))
1465 die("Root cannot be a non-directory");
1466 hashcpy(root->versions[1].sha1, sha1);
1467 if (root->tree)
1468 release_tree_content_recursive(root->tree);
1469 root->tree = newtree;
1470 }
1472 static int tree_content_set(
1473 struct tree_entry *root,
1474 const char *p,
1475 const unsigned char *sha1,
1476 const uint16_t mode,
1477 struct tree_content *subtree)
1478 {
1479 struct tree_content *t;
1480 const char *slash1;
1481 unsigned int i, n;
1482 struct tree_entry *e;
1484 slash1 = strchr(p, '/');
1485 if (slash1)
1486 n = slash1 - p;
1487 else
1488 n = strlen(p);
1489 if (!n)
1490 die("Empty path component found in input");
1491 if (!slash1 && !S_ISDIR(mode) && subtree)
1492 die("Non-directories cannot have subtrees");
1494 if (!root->tree)
1495 load_tree(root);
1496 t = root->tree;
1497 for (i = 0; i < t->entry_count; i++) {
1498 e = t->entries[i];
1499 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1500 if (!slash1) {
1501 if (!S_ISDIR(mode)
1502 && e->versions[1].mode == mode
1503 && !hashcmp(e->versions[1].sha1, sha1))
1504 return 0;
1505 e->versions[1].mode = mode;
1506 hashcpy(e->versions[1].sha1, sha1);
1507 if (e->tree)
1508 release_tree_content_recursive(e->tree);
1509 e->tree = subtree;
1510 hashclr(root->versions[1].sha1);
1511 return 1;
1512 }
1513 if (!S_ISDIR(e->versions[1].mode)) {
1514 e->tree = new_tree_content(8);
1515 e->versions[1].mode = S_IFDIR;
1516 }
1517 if (!e->tree)
1518 load_tree(e);
1519 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1520 hashclr(root->versions[1].sha1);
1521 return 1;
1522 }
1523 return 0;
1524 }
1525 }
1527 if (t->entry_count == t->entry_capacity)
1528 root->tree = t = grow_tree_content(t, t->entry_count);
1529 e = new_tree_entry();
1530 e->name = to_atom(p, n);
1531 e->versions[0].mode = 0;
1532 hashclr(e->versions[0].sha1);
1533 t->entries[t->entry_count++] = e;
1534 if (slash1) {
1535 e->tree = new_tree_content(8);
1536 e->versions[1].mode = S_IFDIR;
1537 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1538 } else {
1539 e->tree = subtree;
1540 e->versions[1].mode = mode;
1541 hashcpy(e->versions[1].sha1, sha1);
1542 }
1543 hashclr(root->versions[1].sha1);
1544 return 1;
1545 }
1547 static int tree_content_remove(
1548 struct tree_entry *root,
1549 const char *p,
1550 struct tree_entry *backup_leaf)
1551 {
1552 struct tree_content *t;
1553 const char *slash1;
1554 unsigned int i, n;
1555 struct tree_entry *e;
1557 slash1 = strchr(p, '/');
1558 if (slash1)
1559 n = slash1 - p;
1560 else
1561 n = strlen(p);
1563 if (!root->tree)
1564 load_tree(root);
1565 t = root->tree;
1566 for (i = 0; i < t->entry_count; i++) {
1567 e = t->entries[i];
1568 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1569 if (slash1 && !S_ISDIR(e->versions[1].mode))
1570 /*
1571 * If p names a file in some subdirectory, and a
1572 * file or symlink matching the name of the
1573 * parent directory of p exists, then p cannot
1574 * exist and need not be deleted.
1575 */
1576 return 1;
1577 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1578 goto del_entry;
1579 if (!e->tree)
1580 load_tree(e);
1581 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1582 for (n = 0; n < e->tree->entry_count; n++) {
1583 if (e->tree->entries[n]->versions[1].mode) {
1584 hashclr(root->versions[1].sha1);
1585 return 1;
1586 }
1587 }
1588 backup_leaf = NULL;
1589 goto del_entry;
1590 }
1591 return 0;
1592 }
1593 }
1594 return 0;
1596 del_entry:
1597 if (backup_leaf)
1598 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1599 else if (e->tree)
1600 release_tree_content_recursive(e->tree);
1601 e->tree = NULL;
1602 e->versions[1].mode = 0;
1603 hashclr(e->versions[1].sha1);
1604 hashclr(root->versions[1].sha1);
1605 return 1;
1606 }
1608 static int tree_content_get(
1609 struct tree_entry *root,
1610 const char *p,
1611 struct tree_entry *leaf)
1612 {
1613 struct tree_content *t;
1614 const char *slash1;
1615 unsigned int i, n;
1616 struct tree_entry *e;
1618 slash1 = strchr(p, '/');
1619 if (slash1)
1620 n = slash1 - p;
1621 else
1622 n = strlen(p);
1624 if (!root->tree)
1625 load_tree(root);
1626 t = root->tree;
1627 for (i = 0; i < t->entry_count; i++) {
1628 e = t->entries[i];
1629 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1630 if (!slash1) {
1631 memcpy(leaf, e, sizeof(*leaf));
1632 if (e->tree && is_null_sha1(e->versions[1].sha1))
1633 leaf->tree = dup_tree_content(e->tree);
1634 else
1635 leaf->tree = NULL;
1636 return 1;
1637 }
1638 if (!S_ISDIR(e->versions[1].mode))
1639 return 0;
1640 if (!e->tree)
1641 load_tree(e);
1642 return tree_content_get(e, slash1 + 1, leaf);
1643 }
1644 }
1645 return 0;
1646 }
1648 static int update_branch(struct branch *b)
1649 {
1650 static const char *msg = "fast-import";
1651 struct ref_lock *lock;
1652 unsigned char old_sha1[20];
1654 if (is_null_sha1(b->sha1))
1655 return 0;
1656 if (read_ref(b->name, old_sha1))
1657 hashclr(old_sha1);
1658 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1659 if (!lock)
1660 return error("Unable to lock %s", b->name);
1661 if (!force_update && !is_null_sha1(old_sha1)) {
1662 struct commit *old_cmit, *new_cmit;
1664 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1665 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1666 if (!old_cmit || !new_cmit) {
1667 unlock_ref(lock);
1668 return error("Branch %s is missing commits.", b->name);
1669 }
1671 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1672 unlock_ref(lock);
1673 warning("Not updating %s"
1674 " (new tip %s does not contain %s)",
1675 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1676 return -1;
1677 }
1678 }
1679 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1680 return error("Unable to update %s", b->name);
1681 return 0;
1682 }
1684 static void dump_branches(void)
1685 {
1686 unsigned int i;
1687 struct branch *b;
1689 for (i = 0; i < branch_table_sz; i++) {
1690 for (b = branch_table[i]; b; b = b->table_next_branch)
1691 failure |= update_branch(b);
1692 }
1693 }
1695 static void dump_tags(void)
1696 {
1697 static const char *msg = "fast-import";
1698 struct tag *t;
1699 struct ref_lock *lock;
1700 char ref_name[PATH_MAX];
1702 for (t = first_tag; t; t = t->next_tag) {
1703 sprintf(ref_name, "tags/%s", t->name);
1704 lock = lock_ref_sha1(ref_name, NULL);
1705 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1706 failure |= error("Unable to update %s", ref_name);
1707 }
1708 }
1710 static void dump_marks_helper(FILE *f,
1711 uintmax_t base,
1712 struct mark_set *m)
1713 {
1714 uintmax_t k;
1715 if (m->shift) {
1716 for (k = 0; k < 1024; k++) {
1717 if (m->data.sets[k])
1718 dump_marks_helper(f, base + (k << m->shift),
1719 m->data.sets[k]);
1720 }
1721 } else {
1722 for (k = 0; k < 1024; k++) {
1723 if (m->data.marked[k])
1724 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1725 sha1_to_hex(m->data.marked[k]->idx.sha1));
1726 }
1727 }
1728 }
1730 static void dump_marks(void)
1731 {
1732 static struct lock_file mark_lock;
1733 int mark_fd;
1734 FILE *f;
1736 if (!export_marks_file)
1737 return;
1739 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1740 if (mark_fd < 0) {
1741 failure |= error("Unable to write marks file %s: %s",
1742 export_marks_file, strerror(errno));
1743 return;
1744 }
1746 f = fdopen(mark_fd, "w");
1747 if (!f) {
1748 int saved_errno = errno;
1749 rollback_lock_file(&mark_lock);
1750 failure |= error("Unable to write marks file %s: %s",
1751 export_marks_file, strerror(saved_errno));
1752 return;
1753 }
1755 /*
1756 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1757 * Assign -1 to the lock file descriptor so that commit_lock_file()
1758 * won't try to close() it.
1759 */
1760 mark_lock.fd = -1;
1762 dump_marks_helper(f, 0, marks);
1763 if (ferror(f) || fclose(f)) {
1764 int saved_errno = errno;
1765 rollback_lock_file(&mark_lock);
1766 failure |= error("Unable to write marks file %s: %s",
1767 export_marks_file, strerror(saved_errno));
1768 return;
1769 }
1771 if (commit_lock_file(&mark_lock)) {
1772 int saved_errno = errno;
1773 rollback_lock_file(&mark_lock);
1774 failure |= error("Unable to commit marks file %s: %s",
1775 export_marks_file, strerror(saved_errno));
1776 return;
1777 }
1778 }
1780 static void read_marks(void)
1781 {
1782 char line[512];
1783 FILE *f = fopen(import_marks_file, "r");
1784 if (!f)
1785 die_errno("cannot read '%s'", import_marks_file);
1786 while (fgets(line, sizeof(line), f)) {
1787 uintmax_t mark;
1788 char *end;
1789 unsigned char sha1[20];
1790 struct object_entry *e;
1792 end = strchr(line, '\n');
1793 if (line[0] != ':' || !end)
1794 die("corrupt mark line: %s", line);
1795 *end = 0;
1796 mark = strtoumax(line + 1, &end, 10);
1797 if (!mark || end == line + 1
1798 || *end != ' ' || get_sha1(end + 1, sha1))
1799 die("corrupt mark line: %s", line);
1800 e = find_object(sha1);
1801 if (!e) {
1802 enum object_type type = sha1_object_info(sha1, NULL);
1803 if (type < 0)
1804 die("object not found: %s", sha1_to_hex(sha1));
1805 e = insert_object(sha1);
1806 e->type = type;
1807 e->pack_id = MAX_PACK_ID;
1808 e->idx.offset = 1; /* just not zero! */
1809 }
1810 insert_mark(mark, e);
1811 }
1812 fclose(f);
1813 }
1816 static int read_next_command(void)
1817 {
1818 static int stdin_eof = 0;
1820 if (stdin_eof) {
1821 unread_command_buf = 0;
1822 return EOF;
1823 }
1825 for (;;) {
1826 if (unread_command_buf) {
1827 unread_command_buf = 0;
1828 } else {
1829 struct recent_command *rc;
1831 strbuf_detach(&command_buf, NULL);
1832 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1833 if (stdin_eof)
1834 return EOF;
1836 if (!seen_data_command
1837 && prefixcmp(command_buf.buf, "feature ")
1838 && prefixcmp(command_buf.buf, "option ")) {
1839 parse_argv();
1840 }
1842 rc = rc_free;
1843 if (rc)
1844 rc_free = rc->next;
1845 else {
1846 rc = cmd_hist.next;
1847 cmd_hist.next = rc->next;
1848 cmd_hist.next->prev = &cmd_hist;
1849 free(rc->buf);
1850 }
1852 rc->buf = command_buf.buf;
1853 rc->prev = cmd_tail;
1854 rc->next = cmd_hist.prev;
1855 rc->prev->next = rc;
1856 cmd_tail = rc;
1857 }
1858 if (!prefixcmp(command_buf.buf, "cat-blob ")) {
1859 parse_cat_blob();
1860 continue;
1861 }
1862 if (command_buf.buf[0] == '#')
1863 continue;
1864 return 0;
1865 }
1866 }
1868 static void skip_optional_lf(void)
1869 {
1870 int term_char = fgetc(stdin);
1871 if (term_char != '\n' && term_char != EOF)
1872 ungetc(term_char, stdin);
1873 }
1875 static void parse_mark(void)
1876 {
1877 if (!prefixcmp(command_buf.buf, "mark :")) {
1878 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1879 read_next_command();
1880 }
1881 else
1882 next_mark = 0;
1883 }
1885 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1886 {
1887 strbuf_reset(sb);
1889 if (prefixcmp(command_buf.buf, "data "))
1890 die("Expected 'data n' command, found: %s", command_buf.buf);
1892 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1893 char *term = xstrdup(command_buf.buf + 5 + 2);
1894 size_t term_len = command_buf.len - 5 - 2;
1896 strbuf_detach(&command_buf, NULL);
1897 for (;;) {
1898 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1899 die("EOF in data (terminator '%s' not found)", term);
1900 if (term_len == command_buf.len
1901 && !strcmp(term, command_buf.buf))
1902 break;
1903 strbuf_addbuf(sb, &command_buf);
1904 strbuf_addch(sb, '\n');
1905 }
1906 free(term);
1907 }
1908 else {
1909 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1910 size_t n = 0, length = (size_t)len;
1912 if (limit && limit < len) {
1913 *len_res = len;
1914 return 0;
1915 }
1916 if (length < len)
1917 die("data is too large to use in this context");
1919 while (n < length) {
1920 size_t s = strbuf_fread(sb, length - n, stdin);
1921 if (!s && feof(stdin))
1922 die("EOF in data (%lu bytes remaining)",
1923 (unsigned long)(length - n));
1924 n += s;
1925 }
1926 }
1928 skip_optional_lf();
1929 return 1;
1930 }
1932 static int validate_raw_date(const char *src, char *result, int maxlen)
1933 {
1934 const char *orig_src = src;
1935 char *endp;
1936 unsigned long num;
1938 errno = 0;
1940 num = strtoul(src, &endp, 10);
1941 /* NEEDSWORK: perhaps check for reasonable values? */
1942 if (errno || endp == src || *endp != ' ')
1943 return -1;
1945 src = endp + 1;
1946 if (*src != '-' && *src != '+')
1947 return -1;
1949 num = strtoul(src + 1, &endp, 10);
1950 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1951 1400 < num)
1952 return -1;
1954 strcpy(result, orig_src);
1955 return 0;
1956 }
1958 static char *parse_ident(const char *buf)
1959 {
1960 const char *gt;
1961 size_t name_len;
1962 char *ident;
1964 gt = strrchr(buf, '>');
1965 if (!gt)
1966 die("Missing > in ident string: %s", buf);
1967 gt++;
1968 if (*gt != ' ')
1969 die("Missing space after > in ident string: %s", buf);
1970 gt++;
1971 name_len = gt - buf;
1972 ident = xmalloc(name_len + 24);
1973 strncpy(ident, buf, name_len);
1975 switch (whenspec) {
1976 case WHENSPEC_RAW:
1977 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1978 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1979 break;
1980 case WHENSPEC_RFC2822:
1981 if (parse_date(gt, ident + name_len, 24) < 0)
1982 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1983 break;
1984 case WHENSPEC_NOW:
1985 if (strcmp("now", gt))
1986 die("Date in ident must be 'now': %s", buf);
1987 datestamp(ident + name_len, 24);
1988 break;
1989 }
1991 return ident;
1992 }
1994 static void parse_and_store_blob(
1995 struct last_object *last,
1996 unsigned char *sha1out,
1997 uintmax_t mark)
1998 {
1999 static struct strbuf buf = STRBUF_INIT;
2000 uintmax_t len;
2002 if (parse_data(&buf, big_file_threshold, &len))
2003 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2004 else {
2005 if (last) {
2006 strbuf_release(&last->data);
2007 last->offset = 0;
2008 last->depth = 0;
2009 }
2010 stream_blob(len, sha1out, mark);
2011 skip_optional_lf();
2012 }
2013 }
2015 static void parse_new_blob(void)
2016 {
2017 read_next_command();
2018 parse_mark();
2019 parse_and_store_blob(&last_blob, NULL, next_mark);
2020 }
2022 static void unload_one_branch(void)
2023 {
2024 while (cur_active_branches
2025 && cur_active_branches >= max_active_branches) {
2026 uintmax_t min_commit = ULONG_MAX;
2027 struct branch *e, *l = NULL, *p = NULL;
2029 for (e = active_branches; e; e = e->active_next_branch) {
2030 if (e->last_commit < min_commit) {
2031 p = l;
2032 min_commit = e->last_commit;
2033 }
2034 l = e;
2035 }
2037 if (p) {
2038 e = p->active_next_branch;
2039 p->active_next_branch = e->active_next_branch;
2040 } else {
2041 e = active_branches;
2042 active_branches = e->active_next_branch;
2043 }
2044 e->active = 0;
2045 e->active_next_branch = NULL;
2046 if (e->branch_tree.tree) {
2047 release_tree_content_recursive(e->branch_tree.tree);
2048 e->branch_tree.tree = NULL;
2049 }
2050 cur_active_branches--;
2051 }
2052 }
2054 static void load_branch(struct branch *b)
2055 {
2056 load_tree(&b->branch_tree);
2057 if (!b->active) {
2058 b->active = 1;
2059 b->active_next_branch = active_branches;
2060 active_branches = b;
2061 cur_active_branches++;
2062 branch_load_count++;
2063 }
2064 }
2066 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2067 {
2068 unsigned char fanout = 0;
2069 while ((num_notes >>= 8))
2070 fanout++;
2071 return fanout;
2072 }
2074 static void construct_path_with_fanout(const char *hex_sha1,
2075 unsigned char fanout, char *path)
2076 {
2077 unsigned int i = 0, j = 0;
2078 if (fanout >= 20)
2079 die("Too large fanout (%u)", fanout);
2080 while (fanout) {
2081 path[i++] = hex_sha1[j++];
2082 path[i++] = hex_sha1[j++];
2083 path[i++] = '/';
2084 fanout--;
2085 }
2086 memcpy(path + i, hex_sha1 + j, 40 - j);
2087 path[i + 40 - j] = '\0';
2088 }
2090 static uintmax_t do_change_note_fanout(
2091 struct tree_entry *orig_root, struct tree_entry *root,
2092 char *hex_sha1, unsigned int hex_sha1_len,
2093 char *fullpath, unsigned int fullpath_len,
2094 unsigned char fanout)
2095 {
2096 struct tree_content *t = root->tree;
2097 struct tree_entry *e, leaf;
2098 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2099 uintmax_t num_notes = 0;
2100 unsigned char sha1[20];
2101 char realpath[60];
2103 for (i = 0; t && i < t->entry_count; i++) {
2104 e = t->entries[i];
2105 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2106 tmp_fullpath_len = fullpath_len;
2108 /*
2109 * We're interested in EITHER existing note entries (entries
2110 * with exactly 40 hex chars in path, not including directory
2111 * separators), OR directory entries that may contain note
2112 * entries (with < 40 hex chars in path).
2113 * Also, each path component in a note entry must be a multiple
2114 * of 2 chars.
2115 */
2116 if (!e->versions[1].mode ||
2117 tmp_hex_sha1_len > 40 ||
2118 e->name->str_len % 2)
2119 continue;
2121 /* This _may_ be a note entry, or a subdir containing notes */
2122 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2123 e->name->str_len);
2124 if (tmp_fullpath_len)
2125 fullpath[tmp_fullpath_len++] = '/';
2126 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2127 e->name->str_len);
2128 tmp_fullpath_len += e->name->str_len;
2129 fullpath[tmp_fullpath_len] = '\0';
2131 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2132 /* This is a note entry */
2133 construct_path_with_fanout(hex_sha1, fanout, realpath);
2134 if (!strcmp(fullpath, realpath)) {
2135 /* Note entry is in correct location */
2136 num_notes++;
2137 continue;
2138 }
2140 /* Rename fullpath to realpath */
2141 if (!tree_content_remove(orig_root, fullpath, &leaf))
2142 die("Failed to remove path %s", fullpath);
2143 tree_content_set(orig_root, realpath,
2144 leaf.versions[1].sha1,
2145 leaf.versions[1].mode,
2146 leaf.tree);
2147 } else if (S_ISDIR(e->versions[1].mode)) {
2148 /* This is a subdir that may contain note entries */
2149 if (!e->tree)
2150 load_tree(e);
2151 num_notes += do_change_note_fanout(orig_root, e,
2152 hex_sha1, tmp_hex_sha1_len,
2153 fullpath, tmp_fullpath_len, fanout);
2154 }
2156 /* The above may have reallocated the current tree_content */
2157 t = root->tree;
2158 }
2159 return num_notes;
2160 }
2162 static uintmax_t change_note_fanout(struct tree_entry *root,
2163 unsigned char fanout)
2164 {
2165 char hex_sha1[40], path[60];
2166 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2167 }
2169 static void file_change_m(struct branch *b)
2170 {
2171 const char *p = command_buf.buf + 2;
2172 static struct strbuf uq = STRBUF_INIT;
2173 const char *endp;
2174 struct object_entry *oe = oe;
2175 unsigned char sha1[20];
2176 uint16_t mode, inline_data = 0;
2178 p = get_mode(p, &mode);
2179 if (!p)
2180 die("Corrupt mode: %s", command_buf.buf);
2181 switch (mode) {
2182 case 0644:
2183 case 0755:
2184 mode |= S_IFREG;
2185 case S_IFREG | 0644:
2186 case S_IFREG | 0755:
2187 case S_IFLNK:
2188 case S_IFDIR:
2189 case S_IFGITLINK:
2190 /* ok */
2191 break;
2192 default:
2193 die("Corrupt mode: %s", command_buf.buf);
2194 }
2196 if (*p == ':') {
2197 char *x;
2198 oe = find_mark(strtoumax(p + 1, &x, 10));
2199 hashcpy(sha1, oe->idx.sha1);
2200 p = x;
2201 } else if (!prefixcmp(p, "inline")) {
2202 inline_data = 1;
2203 p += 6;
2204 } else {
2205 if (get_sha1_hex(p, sha1))
2206 die("Invalid SHA1: %s", command_buf.buf);
2207 oe = find_object(sha1);
2208 p += 40;
2209 }
2210 if (*p++ != ' ')
2211 die("Missing space after SHA1: %s", command_buf.buf);
2213 strbuf_reset(&uq);
2214 if (!unquote_c_style(&uq, p, &endp)) {
2215 if (*endp)
2216 die("Garbage after path in: %s", command_buf.buf);
2217 p = uq.buf;
2218 }
2220 /* Git does not track empty, non-toplevel directories. */
2221 if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
2222 tree_content_remove(&b->branch_tree, p, NULL);
2223 return;
2224 }
2226 if (S_ISGITLINK(mode)) {
2227 if (inline_data)
2228 die("Git links cannot be specified 'inline': %s",
2229 command_buf.buf);
2230 else if (oe) {
2231 if (oe->type != OBJ_COMMIT)
2232 die("Not a commit (actually a %s): %s",
2233 typename(oe->type), command_buf.buf);
2234 }
2235 /*
2236 * Accept the sha1 without checking; it expected to be in
2237 * another repository.
2238 */
2239 } else if (inline_data) {
2240 if (S_ISDIR(mode))
2241 die("Directories cannot be specified 'inline': %s",
2242 command_buf.buf);
2243 if (p != uq.buf) {
2244 strbuf_addstr(&uq, p);
2245 p = uq.buf;
2246 }
2247 read_next_command();
2248 parse_and_store_blob(&last_blob, sha1, 0);
2249 } else {
2250 enum object_type expected = S_ISDIR(mode) ?
2251 OBJ_TREE: OBJ_BLOB;
2252 enum object_type type = oe ? oe->type :
2253 sha1_object_info(sha1, NULL);
2254 if (type < 0)
2255 die("%s not found: %s",
2256 S_ISDIR(mode) ? "Tree" : "Blob",
2257 command_buf.buf);
2258 if (type != expected)
2259 die("Not a %s (actually a %s): %s",
2260 typename(expected), typename(type),
2261 command_buf.buf);
2262 }
2264 if (!*p) {
2265 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2266 return;
2267 }
2268 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2269 }
2271 static void file_change_d(struct branch *b)
2272 {
2273 const char *p = command_buf.buf + 2;
2274 static struct strbuf uq = STRBUF_INIT;
2275 const char *endp;
2277 strbuf_reset(&uq);
2278 if (!unquote_c_style(&uq, p, &endp)) {
2279 if (*endp)
2280 die("Garbage after path in: %s", command_buf.buf);
2281 p = uq.buf;
2282 }
2283 tree_content_remove(&b->branch_tree, p, NULL);
2284 }
2286 static void file_change_cr(struct branch *b, int rename)
2287 {
2288 const char *s, *d;
2289 static struct strbuf s_uq = STRBUF_INIT;
2290 static struct strbuf d_uq = STRBUF_INIT;
2291 const char *endp;
2292 struct tree_entry leaf;
2294 s = command_buf.buf + 2;
2295 strbuf_reset(&s_uq);
2296 if (!unquote_c_style(&s_uq, s, &endp)) {
2297 if (*endp != ' ')
2298 die("Missing space after source: %s", command_buf.buf);
2299 } else {
2300 endp = strchr(s, ' ');
2301 if (!endp)
2302 die("Missing space after source: %s", command_buf.buf);
2303 strbuf_add(&s_uq, s, endp - s);
2304 }
2305 s = s_uq.buf;
2307 endp++;
2308 if (!*endp)
2309 die("Missing dest: %s", command_buf.buf);
2311 d = endp;
2312 strbuf_reset(&d_uq);
2313 if (!unquote_c_style(&d_uq, d, &endp)) {
2314 if (*endp)
2315 die("Garbage after dest in: %s", command_buf.buf);
2316 d = d_uq.buf;
2317 }
2319 memset(&leaf, 0, sizeof(leaf));
2320 if (rename)
2321 tree_content_remove(&b->branch_tree, s, &leaf);
2322 else
2323 tree_content_get(&b->branch_tree, s, &leaf);
2324 if (!leaf.versions[1].mode)
2325 die("Path %s not in branch", s);
2326 if (!*d) { /* C "path/to/subdir" "" */
2327 tree_content_replace(&b->branch_tree,
2328 leaf.versions[1].sha1,
2329 leaf.versions[1].mode,
2330 leaf.tree);
2331 return;
2332 }
2333 tree_content_set(&b->branch_tree, d,
2334 leaf.versions[1].sha1,
2335 leaf.versions[1].mode,
2336 leaf.tree);
2337 }
2339 static void note_change_n(struct branch *b, unsigned char old_fanout)
2340 {
2341 const char *p = command_buf.buf + 2;
2342 static struct strbuf uq = STRBUF_INIT;
2343 struct object_entry *oe = oe;
2344 struct branch *s;
2345 unsigned char sha1[20], commit_sha1[20];
2346 char path[60];
2347 uint16_t inline_data = 0;
2348 unsigned char new_fanout;
2350 /* <dataref> or 'inline' */
2351 if (*p == ':') {
2352 char *x;
2353 oe = find_mark(strtoumax(p + 1, &x, 10));
2354 hashcpy(sha1, oe->idx.sha1);
2355 p = x;
2356 } else if (!prefixcmp(p, "inline")) {
2357 inline_data = 1;
2358 p += 6;
2359 } else {
2360 if (get_sha1_hex(p, sha1))
2361 die("Invalid SHA1: %s", command_buf.buf);
2362 oe = find_object(sha1);
2363 p += 40;
2364 }
2365 if (*p++ != ' ')
2366 die("Missing space after SHA1: %s", command_buf.buf);
2368 /* <committish> */
2369 s = lookup_branch(p);
2370 if (s) {
2371 hashcpy(commit_sha1, s->sha1);
2372 } else if (*p == ':') {
2373 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2374 struct object_entry *commit_oe = find_mark(commit_mark);
2375 if (commit_oe->type != OBJ_COMMIT)
2376 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2377 hashcpy(commit_sha1, commit_oe->idx.sha1);
2378 } else if (!get_sha1(p, commit_sha1)) {
2379 unsigned long size;
2380 char *buf = read_object_with_reference(commit_sha1,
2381 commit_type, &size, commit_sha1);
2382 if (!buf || size < 46)
2383 die("Not a valid commit: %s", p);
2384 free(buf);
2385 } else
2386 die("Invalid ref name or SHA1 expression: %s", p);
2388 if (inline_data) {
2389 if (p != uq.buf) {
2390 strbuf_addstr(&uq, p);
2391 p = uq.buf;
2392 }
2393 read_next_command();
2394 parse_and_store_blob(&last_blob, sha1, 0);
2395 } else if (oe) {
2396 if (oe->type != OBJ_BLOB)
2397 die("Not a blob (actually a %s): %s",
2398 typename(oe->type), command_buf.buf);
2399 } else if (!is_null_sha1(sha1)) {
2400 enum object_type type = sha1_object_info(sha1, NULL);
2401 if (type < 0)
2402 die("Blob not found: %s", command_buf.buf);
2403 if (type != OBJ_BLOB)
2404 die("Not a blob (actually a %s): %s",
2405 typename(type), command_buf.buf);
2406 }
2408 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2409 if (tree_content_remove(&b->branch_tree, path, NULL))
2410 b->num_notes--;
2412 if (is_null_sha1(sha1))
2413 return; /* nothing to insert */
2415 b->num_notes++;
2416 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2417 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2418 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2419 }
2421 static void file_change_deleteall(struct branch *b)
2422 {
2423 release_tree_content_recursive(b->branch_tree.tree);
2424 hashclr(b->branch_tree.versions[0].sha1);
2425 hashclr(b->branch_tree.versions[1].sha1);
2426 load_tree(&b->branch_tree);
2427 b->num_notes = 0;
2428 }
2430 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2431 {
2432 if (!buf || size < 46)
2433 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2434 if (memcmp("tree ", buf, 5)
2435 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2436 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2437 hashcpy(b->branch_tree.versions[0].sha1,
2438 b->branch_tree.versions[1].sha1);
2439 }
2441 static void parse_from_existing(struct branch *b)
2442 {
2443 if (is_null_sha1(b->sha1)) {
2444 hashclr(b->branch_tree.versions[0].sha1);
2445 hashclr(b->branch_tree.versions[1].sha1);
2446 } else {
2447 unsigned long size;
2448 char *buf;
2450 buf = read_object_with_reference(b->sha1,
2451 commit_type, &size, b->sha1);
2452 parse_from_commit(b, buf, size);
2453 free(buf);
2454 }
2455 }
2457 static int parse_from(struct branch *b)
2458 {
2459 const char *from;
2460 struct branch *s;
2462 if (prefixcmp(command_buf.buf, "from "))
2463 return 0;
2465 if (b->branch_tree.tree) {
2466 release_tree_content_recursive(b->branch_tree.tree);
2467 b->branch_tree.tree = NULL;
2468 }
2470 from = strchr(command_buf.buf, ' ') + 1;
2471 s = lookup_branch(from);
2472 if (b == s)
2473 die("Can't create a branch from itself: %s", b->name);
2474 else if (s) {
2475 unsigned char *t = s->branch_tree.versions[1].sha1;
2476 hashcpy(b->sha1, s->sha1);
2477 hashcpy(b->branch_tree.versions[0].sha1, t);
2478 hashcpy(b->branch_tree.versions[1].sha1, t);
2479 } else if (*from == ':') {
2480 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2481 struct object_entry *oe = find_mark(idnum);
2482 if (oe->type != OBJ_COMMIT)
2483 die("Mark :%" PRIuMAX " not a commit", idnum);
2484 hashcpy(b->sha1, oe->idx.sha1);
2485 if (oe->pack_id != MAX_PACK_ID) {
2486 unsigned long size;
2487 char *buf = gfi_unpack_entry(oe, &size);
2488 parse_from_commit(b, buf, size);
2489 free(buf);
2490 } else
2491 parse_from_existing(b);
2492 } else if (!get_sha1(from, b->sha1))
2493 parse_from_existing(b);
2494 else
2495 die("Invalid ref name or SHA1 expression: %s", from);
2497 read_next_command();
2498 return 1;
2499 }
2501 static struct hash_list *parse_merge(unsigned int *count)
2502 {
2503 struct hash_list *list = NULL, *n, *e = e;
2504 const char *from;
2505 struct branch *s;
2507 *count = 0;
2508 while (!prefixcmp(command_buf.buf, "merge ")) {
2509 from = strchr(command_buf.buf, ' ') + 1;
2510 n = xmalloc(sizeof(*n));
2511 s = lookup_branch(from);
2512 if (s)
2513 hashcpy(n->sha1, s->sha1);
2514 else if (*from == ':') {
2515 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2516 struct object_entry *oe = find_mark(idnum);
2517 if (oe->type != OBJ_COMMIT)
2518 die("Mark :%" PRIuMAX " not a commit", idnum);
2519 hashcpy(n->sha1, oe->idx.sha1);
2520 } else if (!get_sha1(from, n->sha1)) {
2521 unsigned long size;
2522 char *buf = read_object_with_reference(n->sha1,
2523 commit_type, &size, n->sha1);
2524 if (!buf || size < 46)
2525 die("Not a valid commit: %s", from);
2526 free(buf);
2527 } else
2528 die("Invalid ref name or SHA1 expression: %s", from);
2530 n->next = NULL;
2531 if (list)
2532 e->next = n;
2533 else
2534 list = n;
2535 e = n;
2536 (*count)++;
2537 read_next_command();
2538 }
2539 return list;
2540 }
2542 static void parse_new_commit(void)
2543 {
2544 static struct strbuf msg = STRBUF_INIT;
2545 struct branch *b;
2546 char *sp;
2547 char *author = NULL;
2548 char *committer = NULL;
2549 struct hash_list *merge_list = NULL;
2550 unsigned int merge_count;
2551 unsigned char prev_fanout, new_fanout;
2553 /* Obtain the branch name from the rest of our command */
2554 sp = strchr(command_buf.buf, ' ') + 1;
2555 b = lookup_branch(sp);
2556 if (!b)
2557 b = new_branch(sp);
2559 read_next_command();
2560 parse_mark();
2561 if (!prefixcmp(command_buf.buf, "author ")) {
2562 author = parse_ident(command_buf.buf + 7);
2563 read_next_command();
2564 }
2565 if (!prefixcmp(command_buf.buf, "committer ")) {
2566 committer = parse_ident(command_buf.buf + 10);
2567 read_next_command();
2568 }
2569 if (!committer)
2570 die("Expected committer but didn't get one");
2571 parse_data(&msg, 0, NULL);
2572 read_next_command();
2573 parse_from(b);
2574 merge_list = parse_merge(&merge_count);
2576 /* ensure the branch is active/loaded */
2577 if (!b->branch_tree.tree || !max_active_branches) {
2578 unload_one_branch();
2579 load_branch(b);
2580 }
2582 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2584 /* file_change* */
2585 while (command_buf.len > 0) {
2586 if (!prefixcmp(command_buf.buf, "M "))
2587 file_change_m(b);
2588 else if (!prefixcmp(command_buf.buf, "D "))
2589 file_change_d(b);
2590 else if (!prefixcmp(command_buf.buf, "R "))
2591 file_change_cr(b, 1);
2592 else if (!prefixcmp(command_buf.buf, "C "))
2593 file_change_cr(b, 0);
2594 else if (!prefixcmp(command_buf.buf, "N "))
2595 note_change_n(b, prev_fanout);
2596 else if (!strcmp("deleteall", command_buf.buf))
2597 file_change_deleteall(b);
2598 else {
2599 unread_command_buf = 1;
2600 break;
2601 }
2602 if (read_next_command() == EOF)
2603 break;
2604 }
2606 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2607 if (new_fanout != prev_fanout)
2608 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2610 /* build the tree and the commit */
2611 store_tree(&b->branch_tree);
2612 hashcpy(b->branch_tree.versions[0].sha1,
2613 b->branch_tree.versions[1].sha1);
2615 strbuf_reset(&new_data);
2616 strbuf_addf(&new_data, "tree %s\n",
2617 sha1_to_hex(b->branch_tree.versions[1].sha1));
2618 if (!is_null_sha1(b->sha1))
2619 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2620 while (merge_list) {
2621 struct hash_list *next = merge_list->next;
2622 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2623 free(merge_list);
2624 merge_list = next;
2625 }
2626 strbuf_addf(&new_data,
2627 "author %s\n"
2628 "committer %s\n"
2629 "\n",
2630 author ? author : committer, committer);
2631 strbuf_addbuf(&new_data, &msg);
2632 free(author);
2633 free(committer);
2635 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2636 b->pack_id = pack_id;
2637 b->last_commit = object_count_by_type[OBJ_COMMIT];
2638 }
2640 static void parse_new_tag(void)
2641 {
2642 static struct strbuf msg = STRBUF_INIT;
2643 char *sp;
2644 const char *from;
2645 char *tagger;
2646 struct branch *s;
2647 struct tag *t;
2648 uintmax_t from_mark = 0;
2649 unsigned char sha1[20];
2650 enum object_type type;
2652 /* Obtain the new tag name from the rest of our command */
2653 sp = strchr(command_buf.buf, ' ') + 1;
2654 t = pool_alloc(sizeof(struct tag));
2655 t->next_tag = NULL;
2656 t->name = pool_strdup(sp);
2657 if (last_tag)
2658 last_tag->next_tag = t;
2659 else
2660 first_tag = t;
2661 last_tag = t;
2662 read_next_command();
2664 /* from ... */
2665 if (prefixcmp(command_buf.buf, "from "))
2666 die("Expected from command, got %s", command_buf.buf);
2667 from = strchr(command_buf.buf, ' ') + 1;
2668 s = lookup_branch(from);
2669 if (s) {
2670 hashcpy(sha1, s->sha1);
2671 type = OBJ_COMMIT;
2672 } else if (*from == ':') {
2673 struct object_entry *oe;
2674 from_mark = strtoumax(from + 1, NULL, 10);
2675 oe = find_mark(from_mark);
2676 type = oe->type;
2677 hashcpy(sha1, oe->idx.sha1);
2678 } else if (!get_sha1(from, sha1)) {
2679 unsigned long size;
2680 char *buf;
2682 buf = read_sha1_file(sha1, &type, &size);
2683 if (!buf || size < 46)
2684 die("Not a valid commit: %s", from);
2685 free(buf);
2686 } else
2687 die("Invalid ref name or SHA1 expression: %s", from);
2688 read_next_command();
2690 /* tagger ... */
2691 if (!prefixcmp(command_buf.buf, "tagger ")) {
2692 tagger = parse_ident(command_buf.buf + 7);
2693 read_next_command();
2694 } else
2695 tagger = NULL;
2697 /* tag payload/message */
2698 parse_data(&msg, 0, NULL);
2700 /* build the tag object */
2701 strbuf_reset(&new_data);
2703 strbuf_addf(&new_data,
2704 "object %s\n"
2705 "type %s\n"
2706 "tag %s\n",
2707 sha1_to_hex(sha1), typename(type), t->name);
2708 if (tagger)
2709 strbuf_addf(&new_data,
2710 "tagger %s\n", tagger);
2711 strbuf_addch(&new_data, '\n');
2712 strbuf_addbuf(&new_data, &msg);
2713 free(tagger);
2715 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2716 t->pack_id = MAX_PACK_ID;
2717 else
2718 t->pack_id = pack_id;
2719 }
2721 static void parse_reset_branch(void)
2722 {
2723 struct branch *b;
2724 char *sp;
2726 /* Obtain the branch name from the rest of our command */
2727 sp = strchr(command_buf.buf, ' ') + 1;
2728 b = lookup_branch(sp);
2729 if (b) {
2730 hashclr(b->sha1);
2731 hashclr(b->branch_tree.versions[0].sha1);
2732 hashclr(b->branch_tree.versions[1].sha1);
2733 if (b->branch_tree.tree) {
2734 release_tree_content_recursive(b->branch_tree.tree);
2735 b->branch_tree.tree = NULL;
2736 }
2737 }
2738 else
2739 b = new_branch(sp);
2740 read_next_command();
2741 parse_from(b);
2742 if (command_buf.len > 0)
2743 unread_command_buf = 1;
2744 }
2746 static void cat_blob_write(const char *buf, unsigned long size)
2747 {
2748 if (write_in_full(cat_blob_fd, buf, size) != size)
2749 die_errno("Write to frontend failed");
2750 }
2752 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2753 {
2754 struct strbuf line = STRBUF_INIT;
2755 unsigned long size;
2756 enum object_type type = 0;
2757 char *buf;
2759 if (!oe || oe->pack_id == MAX_PACK_ID) {
2760 buf = read_sha1_file(sha1, &type, &size);
2761 } else {
2762 type = oe->type;
2763 buf = gfi_unpack_entry(oe, &size);
2764 }
2766 /*
2767 * Output based on batch_one_object() from cat-file.c.
2768 */
2769 if (type <= 0) {
2770 strbuf_reset(&line);
2771 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2772 cat_blob_write(line.buf, line.len);
2773 strbuf_release(&line);
2774 free(buf);
2775 return;
2776 }
2777 if (!buf)
2778 die("Can't read object %s", sha1_to_hex(sha1));
2779 if (type != OBJ_BLOB)
2780 die("Object %s is a %s but a blob was expected.",
2781 sha1_to_hex(sha1), typename(type));
2782 strbuf_reset(&line);
2783 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2784 typename(type), size);
2785 cat_blob_write(line.buf, line.len);
2786 strbuf_release(&line);
2787 cat_blob_write(buf, size);
2788 cat_blob_write("\n", 1);
2789 free(buf);
2790 }
2792 static void parse_cat_blob(void)
2793 {
2794 const char *p;
2795 struct object_entry *oe = oe;
2796 unsigned char sha1[20];
2798 /* cat-blob SP <object> LF */
2799 p = command_buf.buf + strlen("cat-blob ");
2800 if (*p == ':') {
2801 char *x;
2802 oe = find_mark(strtoumax(p + 1, &x, 10));
2803 if (x == p + 1)
2804 die("Invalid mark: %s", command_buf.buf);
2805 if (!oe)
2806 die("Unknown mark: %s", command_buf.buf);
2807 if (*x)
2808 die("Garbage after mark: %s", command_buf.buf);
2809 hashcpy(sha1, oe->idx.sha1);
2810 } else {
2811 if (get_sha1_hex(p, sha1))
2812 die("Invalid SHA1: %s", command_buf.buf);
2813 if (p[40])
2814 die("Garbage after SHA1: %s", command_buf.buf);
2815 oe = find_object(sha1);
2816 }
2818 cat_blob(oe, sha1);
2819 }
2821 static void checkpoint(void)
2822 {
2823 checkpoint_requested = 0;
2824 if (object_count) {
2825 cycle_packfile();
2826 dump_branches();
2827 dump_tags();
2828 dump_marks();
2829 }
2830 }
2832 static void parse_checkpoint(void)
2833 {
2834 checkpoint_requested = 1;
2835 skip_optional_lf();
2836 }
2838 static void parse_progress(void)
2839 {
2840 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2841 fputc('\n', stdout);
2842 fflush(stdout);
2843 skip_optional_lf();
2844 }
2846 static char* make_fast_import_path(const char *path)
2847 {
2848 struct strbuf abs_path = STRBUF_INIT;
2850 if (!relative_marks_paths || is_absolute_path(path))
2851 return xstrdup(path);
2852 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
2853 return strbuf_detach(&abs_path, NULL);
2854 }
2856 static void option_import_marks(const char *marks, int from_stream)
2857 {
2858 if (import_marks_file) {
2859 if (from_stream)
2860 die("Only one import-marks command allowed per stream");
2862 /* read previous mark file */
2863 if(!import_marks_file_from_stream)
2864 read_marks();
2865 }
2867 import_marks_file = make_fast_import_path(marks);
2868 safe_create_leading_directories_const(import_marks_file);
2869 import_marks_file_from_stream = from_stream;
2870 }
2872 static void option_date_format(const char *fmt)
2873 {
2874 if (!strcmp(fmt, "raw"))
2875 whenspec = WHENSPEC_RAW;
2876 else if (!strcmp(fmt, "rfc2822"))
2877 whenspec = WHENSPEC_RFC2822;
2878 else if (!strcmp(fmt, "now"))
2879 whenspec = WHENSPEC_NOW;
2880 else
2881 die("unknown --date-format argument %s", fmt);
2882 }
2884 static unsigned long ulong_arg(const char *option, const char *arg)
2885 {
2886 char *endptr;
2887 unsigned long rv = strtoul(arg, &endptr, 0);
2888 if (strchr(arg, '-') || endptr == arg || *endptr)
2889 die("%s: argument must be a non-negative integer", option);
2890 return rv;
2891 }
2893 static void option_depth(const char *depth)
2894 {
2895 max_depth = ulong_arg("--depth", depth);
2896 if (max_depth > MAX_DEPTH)
2897 die("--depth cannot exceed %u", MAX_DEPTH);
2898 }
2900 static void option_active_branches(const char *branches)
2901 {
2902 max_active_branches = ulong_arg("--active-branches", branches);
2903 }
2905 static void option_export_marks(const char *marks)
2906 {
2907 export_marks_file = make_fast_import_path(marks);
2908 safe_create_leading_directories_const(export_marks_file);
2909 }
2911 static void option_cat_blob_fd(const char *fd)
2912 {
2913 unsigned long n = ulong_arg("--cat-blob-fd", fd);
2914 if (n > (unsigned long) INT_MAX)
2915 die("--cat-blob-fd cannot exceed %d", INT_MAX);
2916 cat_blob_fd = (int) n;
2917 }
2919 static void option_export_pack_edges(const char *edges)
2920 {
2921 if (pack_edges)
2922 fclose(pack_edges);
2923 pack_edges = fopen(edges, "a");
2924 if (!pack_edges)
2925 die_errno("Cannot open '%s'", edges);
2926 }
2928 static int parse_one_option(const char *option)
2929 {
2930 if (!prefixcmp(option, "max-pack-size=")) {
2931 unsigned long v;
2932 if (!git_parse_ulong(option + 14, &v))
2933 return 0;
2934 if (v < 8192) {
2935 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
2936 v *= 1024 * 1024;
2937 } else if (v < 1024 * 1024) {
2938 warning("minimum max-pack-size is 1 MiB");
2939 v = 1024 * 1024;
2940 }
2941 max_packsize = v;
2942 } else if (!prefixcmp(option, "big-file-threshold=")) {
2943 unsigned long v;
2944 if (!git_parse_ulong(option + 19, &v))
2945 return 0;
2946 big_file_threshold = v;
2947 } else if (!prefixcmp(option, "depth=")) {
2948 option_depth(option + 6);
2949 } else if (!prefixcmp(option, "active-branches=")) {
2950 option_active_branches(option + 16);
2951 } else if (!prefixcmp(option, "export-pack-edges=")) {
2952 option_export_pack_edges(option + 18);
2953 } else if (!prefixcmp(option, "quiet")) {
2954 show_stats = 0;
2955 } else if (!prefixcmp(option, "stats")) {
2956 show_stats = 1;
2957 } else {
2958 return 0;
2959 }
2961 return 1;
2962 }
2964 static int parse_one_feature(const char *feature, int from_stream)
2965 {
2966 if (!prefixcmp(feature, "date-format=")) {
2967 option_date_format(feature + 12);
2968 } else if (!prefixcmp(feature, "import-marks=")) {
2969 option_import_marks(feature + 13, from_stream);
2970 } else if (!prefixcmp(feature, "export-marks=")) {
2971 option_export_marks(feature + 13);
2972 } else if (!strcmp(feature, "cat-blob")) {
2973 ; /* Don't die - this feature is supported */
2974 } else if (!prefixcmp(feature, "relative-marks")) {
2975 relative_marks_paths = 1;
2976 } else if (!prefixcmp(feature, "no-relative-marks")) {
2977 relative_marks_paths = 0;
2978 } else if (!prefixcmp(feature, "force")) {
2979 force_update = 1;
2980 } else if (!strcmp(feature, "notes")) {
2981 ; /* do nothing; we have the feature */
2982 } else {
2983 return 0;
2984 }
2986 return 1;
2987 }
2989 static void parse_feature(void)
2990 {
2991 char *feature = command_buf.buf + 8;
2993 if (seen_data_command)
2994 die("Got feature command '%s' after data command", feature);
2996 if (parse_one_feature(feature, 1))
2997 return;
2999 die("This version of fast-import does not support feature %s.", feature);
3000 }
3002 static void parse_option(void)
3003 {
3004 char *option = command_buf.buf + 11;
3006 if (seen_data_command)
3007 die("Got option command '%s' after data command", option);
3009 if (parse_one_option(option))
3010 return;
3012 die("This version of fast-import does not support option: %s", option);
3013 }
3015 static int git_pack_config(const char *k, const char *v, void *cb)
3016 {
3017 if (!strcmp(k, "pack.depth")) {
3018 max_depth = git_config_int(k, v);
3019 if (max_depth > MAX_DEPTH)
3020 max_depth = MAX_DEPTH;
3021 return 0;
3022 }
3023 if (!strcmp(k, "pack.compression")) {
3024 int level = git_config_int(k, v);
3025 if (level == -1)
3026 level = Z_DEFAULT_COMPRESSION;
3027 else if (level < 0 || level > Z_BEST_COMPRESSION)
3028 die("bad pack compression level %d", level);
3029 pack_compression_level = level;
3030 pack_compression_seen = 1;
3031 return 0;
3032 }
3033 if (!strcmp(k, "pack.indexversion")) {
3034 pack_idx_default_version = git_config_int(k, v);
3035 if (pack_idx_default_version > 2)
3036 die("bad pack.indexversion=%"PRIu32,
3037 pack_idx_default_version);
3038 return 0;
3039 }
3040 if (!strcmp(k, "pack.packsizelimit")) {
3041 max_packsize = git_config_ulong(k, v);
3042 return 0;
3043 }
3044 if (!strcmp(k, "core.bigfilethreshold")) {
3045 long n = git_config_int(k, v);
3046 big_file_threshold = 0 < n ? n : 0;
3047 }
3048 return git_default_config(k, v, cb);
3049 }
3051 static const char fast_import_usage[] =
3052 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3054 static void parse_argv(void)
3055 {
3056 unsigned int i;
3058 for (i = 1; i < global_argc; i++) {
3059 const char *a = global_argv[i];
3061 if (*a != '-' || !strcmp(a, "--"))
3062 break;
3064 if (parse_one_option(a + 2))
3065 continue;
3067 if (parse_one_feature(a + 2, 0))
3068 continue;
3070 if (!prefixcmp(a + 2, "cat-blob-fd=")) {
3071 option_cat_blob_fd(a + 2 + strlen("cat-blob-fd="));
3072 continue;
3073 }
3075 die("unknown option %s", a);
3076 }
3077 if (i != global_argc)
3078 usage(fast_import_usage);
3080 seen_data_command = 1;
3081 if (import_marks_file)
3082 read_marks();
3083 }
3085 int main(int argc, const char **argv)
3086 {
3087 unsigned int i;
3089 git_extract_argv0_path(argv[0]);
3091 if (argc == 2 && !strcmp(argv[1], "-h"))
3092 usage(fast_import_usage);
3094 setup_git_directory();
3095 git_config(git_pack_config, NULL);
3096 if (!pack_compression_seen && core_compression_seen)
3097 pack_compression_level = core_compression_level;
3099 alloc_objects(object_entry_alloc);
3100 strbuf_init(&command_buf, 0);
3101 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3102 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3103 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3104 marks = pool_calloc(1, sizeof(struct mark_set));
3106 global_argc = argc;
3107 global_argv = argv;
3109 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3110 for (i = 0; i < (cmd_save - 1); i++)
3111 rc_free[i].next = &rc_free[i + 1];
3112 rc_free[cmd_save - 1].next = NULL;
3114 prepare_packed_git();
3115 start_packfile();
3116 set_die_routine(die_nicely);
3117 set_checkpoint_signal();
3118 while (read_next_command() != EOF) {
3119 if (!strcmp("blob", command_buf.buf))
3120 parse_new_blob();
3121 else if (!prefixcmp(command_buf.buf, "commit "))
3122 parse_new_commit();
3123 else if (!prefixcmp(command_buf.buf, "tag "))
3124 parse_new_tag();
3125 else if (!prefixcmp(command_buf.buf, "reset "))
3126 parse_reset_branch();
3127 else if (!strcmp("checkpoint", command_buf.buf))
3128 parse_checkpoint();
3129 else if (!prefixcmp(command_buf.buf, "progress "))
3130 parse_progress();
3131 else if (!prefixcmp(command_buf.buf, "feature "))
3132 parse_feature();
3133 else if (!prefixcmp(command_buf.buf, "option git "))
3134 parse_option();
3135 else if (!prefixcmp(command_buf.buf, "option "))
3136 /* ignore non-git options*/;
3137 else
3138 die("Unsupported command: %s", command_buf.buf);
3140 if (checkpoint_requested)
3141 checkpoint();
3142 }
3144 /* argv hasn't been parsed yet, do so */
3145 if (!seen_data_command)
3146 parse_argv();
3148 end_packfile();
3150 dump_branches();
3151 dump_tags();
3152 unkeep_all_packs();
3153 dump_marks();
3155 if (pack_edges)
3156 fclose(pack_edges);
3158 if (show_stats) {
3159 uintmax_t total_count = 0, duplicate_count = 0;
3160 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3161 total_count += object_count_by_type[i];
3162 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3163 duplicate_count += duplicate_count_by_type[i];
3165 fprintf(stderr, "%s statistics:\n", argv[0]);
3166 fprintf(stderr, "---------------------------------------------------------------------\n");
3167 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3168 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3169 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
3170 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
3171 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
3172 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
3173 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3174 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3175 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3176 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3177 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3178 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3179 fprintf(stderr, "---------------------------------------------------------------------\n");
3180 pack_report();
3181 fprintf(stderr, "---------------------------------------------------------------------\n");
3182 fprintf(stderr, "\n");
3183 }
3185 return failure ? 1 : 0;
3186 }