![[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 p->do_not_close = 1;
861 pack_file = sha1fd(pack_fd, p->pack_name);
863 hdr.hdr_signature = htonl(PACK_SIGNATURE);
864 hdr.hdr_version = htonl(2);
865 hdr.hdr_entries = 0;
866 sha1write(pack_file, &hdr, sizeof(hdr));
868 pack_data = p;
869 pack_size = sizeof(hdr);
870 object_count = 0;
872 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
873 all_packs[pack_id] = p;
874 }
876 static const char *create_index(void)
877 {
878 const char *tmpfile;
879 struct pack_idx_entry **idx, **c, **last;
880 struct object_entry *e;
881 struct object_entry_pool *o;
883 /* Build the table of object IDs. */
884 idx = xmalloc(object_count * sizeof(*idx));
885 c = idx;
886 for (o = blocks; o; o = o->next_pool)
887 for (e = o->next_free; e-- != o->entries;)
888 if (pack_id == e->pack_id)
889 *c++ = &e->idx;
890 last = idx + object_count;
891 if (c != last)
892 die("internal consistency error creating the index");
894 tmpfile = write_idx_file(NULL, idx, object_count, pack_data->sha1);
895 free(idx);
896 return tmpfile;
897 }
899 static char *keep_pack(const char *curr_index_name)
900 {
901 static char name[PATH_MAX];
902 static const char *keep_msg = "fast-import";
903 int keep_fd;
905 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
906 if (keep_fd < 0)
907 die_errno("cannot create keep file");
908 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
909 if (close(keep_fd))
910 die_errno("failed to write keep file");
912 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
913 get_object_directory(), sha1_to_hex(pack_data->sha1));
914 if (move_temp_to_file(pack_data->pack_name, name))
915 die("cannot store pack file");
917 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
918 get_object_directory(), sha1_to_hex(pack_data->sha1));
919 if (move_temp_to_file(curr_index_name, name))
920 die("cannot store index file");
921 free((void *)curr_index_name);
922 return name;
923 }
925 static void unkeep_all_packs(void)
926 {
927 static char name[PATH_MAX];
928 int k;
930 for (k = 0; k < pack_id; k++) {
931 struct packed_git *p = all_packs[k];
932 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
933 get_object_directory(), sha1_to_hex(p->sha1));
934 unlink_or_warn(name);
935 }
936 }
938 static void end_packfile(void)
939 {
940 struct packed_git *old_p = pack_data, *new_p;
942 clear_delta_base_cache();
943 if (object_count) {
944 unsigned char cur_pack_sha1[20];
945 char *idx_name;
946 int i;
947 struct branch *b;
948 struct tag *t;
950 close_pack_windows(pack_data);
951 sha1close(pack_file, cur_pack_sha1, 0);
952 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
953 pack_data->pack_name, object_count,
954 cur_pack_sha1, pack_size);
955 close(pack_data->pack_fd);
956 idx_name = keep_pack(create_index());
958 /* Register the packfile with core git's machinery. */
959 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
960 if (!new_p)
961 die("core git rejected index %s", idx_name);
962 all_packs[pack_id] = new_p;
963 install_packed_git(new_p);
965 /* Print the boundary */
966 if (pack_edges) {
967 fprintf(pack_edges, "%s:", new_p->pack_name);
968 for (i = 0; i < branch_table_sz; i++) {
969 for (b = branch_table[i]; b; b = b->table_next_branch) {
970 if (b->pack_id == pack_id)
971 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
972 }
973 }
974 for (t = first_tag; t; t = t->next_tag) {
975 if (t->pack_id == pack_id)
976 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
977 }
978 fputc('\n', pack_edges);
979 fflush(pack_edges);
980 }
982 pack_id++;
983 }
984 else {
985 close(old_p->pack_fd);
986 unlink_or_warn(old_p->pack_name);
987 }
988 free(old_p);
990 /* We can't carry a delta across packfiles. */
991 strbuf_release(&last_blob.data);
992 last_blob.offset = 0;
993 last_blob.depth = 0;
994 }
996 static void cycle_packfile(void)
997 {
998 end_packfile();
999 start_packfile();
1000 }
1002 static int store_object(
1003 enum object_type type,
1004 struct strbuf *dat,
1005 struct last_object *last,
1006 unsigned char *sha1out,
1007 uintmax_t mark)
1008 {
1009 void *out, *delta;
1010 struct object_entry *e;
1011 unsigned char hdr[96];
1012 unsigned char sha1[20];
1013 unsigned long hdrlen, deltalen;
1014 git_SHA_CTX c;
1015 z_stream s;
1017 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1018 (unsigned long)dat->len) + 1;
1019 git_SHA1_Init(&c);
1020 git_SHA1_Update(&c, hdr, hdrlen);
1021 git_SHA1_Update(&c, dat->buf, dat->len);
1022 git_SHA1_Final(sha1, &c);
1023 if (sha1out)
1024 hashcpy(sha1out, sha1);
1026 e = insert_object(sha1);
1027 if (mark)
1028 insert_mark(mark, e);
1029 if (e->idx.offset) {
1030 duplicate_count_by_type[type]++;
1031 return 1;
1032 } else if (find_sha1_pack(sha1, packed_git)) {
1033 e->type = type;
1034 e->pack_id = MAX_PACK_ID;
1035 e->idx.offset = 1; /* just not zero! */
1036 duplicate_count_by_type[type]++;
1037 return 1;
1038 }
1040 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1041 delta = diff_delta(last->data.buf, last->data.len,
1042 dat->buf, dat->len,
1043 &deltalen, dat->len - 20);
1044 } else
1045 delta = NULL;
1047 memset(&s, 0, sizeof(s));
1048 deflateInit(&s, pack_compression_level);
1049 if (delta) {
1050 s.next_in = delta;
1051 s.avail_in = deltalen;
1052 } else {
1053 s.next_in = (void *)dat->buf;
1054 s.avail_in = dat->len;
1055 }
1056 s.avail_out = deflateBound(&s, s.avail_in);
1057 s.next_out = out = xmalloc(s.avail_out);
1058 while (deflate(&s, Z_FINISH) == Z_OK)
1059 /* nothing */;
1060 deflateEnd(&s);
1062 /* Determine if we should auto-checkpoint. */
1063 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1064 || (pack_size + 60 + s.total_out) < pack_size) {
1066 /* This new object needs to *not* have the current pack_id. */
1067 e->pack_id = pack_id + 1;
1068 cycle_packfile();
1070 /* We cannot carry a delta into the new pack. */
1071 if (delta) {
1072 free(delta);
1073 delta = NULL;
1075 memset(&s, 0, sizeof(s));
1076 deflateInit(&s, pack_compression_level);
1077 s.next_in = (void *)dat->buf;
1078 s.avail_in = dat->len;
1079 s.avail_out = deflateBound(&s, s.avail_in);
1080 s.next_out = out = xrealloc(out, s.avail_out);
1081 while (deflate(&s, Z_FINISH) == Z_OK)
1082 /* nothing */;
1083 deflateEnd(&s);
1084 }
1085 }
1087 e->type = type;
1088 e->pack_id = pack_id;
1089 e->idx.offset = pack_size;
1090 object_count++;
1091 object_count_by_type[type]++;
1093 crc32_begin(pack_file);
1095 if (delta) {
1096 off_t ofs = e->idx.offset - last->offset;
1097 unsigned pos = sizeof(hdr) - 1;
1099 delta_count_by_type[type]++;
1100 e->depth = last->depth + 1;
1102 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1103 sha1write(pack_file, hdr, hdrlen);
1104 pack_size += hdrlen;
1106 hdr[pos] = ofs & 127;
1107 while (ofs >>= 7)
1108 hdr[--pos] = 128 | (--ofs & 127);
1109 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1110 pack_size += sizeof(hdr) - pos;
1111 } else {
1112 e->depth = 0;
1113 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1114 sha1write(pack_file, hdr, hdrlen);
1115 pack_size += hdrlen;
1116 }
1118 sha1write(pack_file, out, s.total_out);
1119 pack_size += s.total_out;
1121 e->idx.crc32 = crc32_end(pack_file);
1123 free(out);
1124 free(delta);
1125 if (last) {
1126 if (last->no_swap) {
1127 last->data = *dat;
1128 } else {
1129 strbuf_swap(&last->data, dat);
1130 }
1131 last->offset = e->idx.offset;
1132 last->depth = e->depth;
1133 }
1134 return 0;
1135 }
1137 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1138 {
1139 if (ftruncate(pack_data->pack_fd, to)
1140 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1141 die_errno("cannot truncate pack to skip duplicate");
1142 pack_size = to;
1144 /* yes this is a layering violation */
1145 pack_file->total = to;
1146 pack_file->offset = 0;
1147 pack_file->ctx = *ctx;
1148 }
1150 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1151 {
1152 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1153 unsigned char *in_buf = xmalloc(in_sz);
1154 unsigned char *out_buf = xmalloc(out_sz);
1155 struct object_entry *e;
1156 unsigned char sha1[20];
1157 unsigned long hdrlen;
1158 off_t offset;
1159 git_SHA_CTX c;
1160 git_SHA_CTX pack_file_ctx;
1161 z_stream s;
1162 int status = Z_OK;
1164 /* Determine if we should auto-checkpoint. */
1165 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1166 || (pack_size + 60 + len) < pack_size)
1167 cycle_packfile();
1169 offset = pack_size;
1171 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1172 sha1flush(pack_file);
1173 pack_file_ctx = pack_file->ctx;
1175 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1176 if (out_sz <= hdrlen)
1177 die("impossibly large object header");
1179 git_SHA1_Init(&c);
1180 git_SHA1_Update(&c, out_buf, hdrlen);
1182 crc32_begin(pack_file);
1184 memset(&s, 0, sizeof(s));
1185 deflateInit(&s, pack_compression_level);
1187 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1188 if (out_sz <= hdrlen)
1189 die("impossibly large object header");
1191 s.next_out = out_buf + hdrlen;
1192 s.avail_out = out_sz - hdrlen;
1194 while (status != Z_STREAM_END) {
1195 if (0 < len && !s.avail_in) {
1196 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1197 size_t n = fread(in_buf, 1, cnt, stdin);
1198 if (!n && feof(stdin))
1199 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1201 git_SHA1_Update(&c, in_buf, n);
1202 s.next_in = in_buf;
1203 s.avail_in = n;
1204 len -= n;
1205 }
1207 status = deflate(&s, len ? 0 : Z_FINISH);
1209 if (!s.avail_out || status == Z_STREAM_END) {
1210 size_t n = s.next_out - out_buf;
1211 sha1write(pack_file, out_buf, n);
1212 pack_size += n;
1213 s.next_out = out_buf;
1214 s.avail_out = out_sz;
1215 }
1217 switch (status) {
1218 case Z_OK:
1219 case Z_BUF_ERROR:
1220 case Z_STREAM_END:
1221 continue;
1222 default:
1223 die("unexpected deflate failure: %d", status);
1224 }
1225 }
1226 deflateEnd(&s);
1227 git_SHA1_Final(sha1, &c);
1229 if (sha1out)
1230 hashcpy(sha1out, sha1);
1232 e = insert_object(sha1);
1234 if (mark)
1235 insert_mark(mark, e);
1237 if (e->idx.offset) {
1238 duplicate_count_by_type[OBJ_BLOB]++;
1239 truncate_pack(offset, &pack_file_ctx);
1241 } else if (find_sha1_pack(sha1, packed_git)) {
1242 e->type = OBJ_BLOB;
1243 e->pack_id = MAX_PACK_ID;
1244 e->idx.offset = 1; /* just not zero! */
1245 duplicate_count_by_type[OBJ_BLOB]++;
1246 truncate_pack(offset, &pack_file_ctx);
1248 } else {
1249 e->depth = 0;
1250 e->type = OBJ_BLOB;
1251 e->pack_id = pack_id;
1252 e->idx.offset = offset;
1253 e->idx.crc32 = crc32_end(pack_file);
1254 object_count++;
1255 object_count_by_type[OBJ_BLOB]++;
1256 }
1258 free(in_buf);
1259 free(out_buf);
1260 }
1262 /* All calls must be guarded by find_object() or find_mark() to
1263 * ensure the 'struct object_entry' passed was written by this
1264 * process instance. We unpack the entry by the offset, avoiding
1265 * the need for the corresponding .idx file. This unpacking rule
1266 * works because we only use OBJ_REF_DELTA within the packfiles
1267 * created by fast-import.
1268 *
1269 * oe must not be NULL. Such an oe usually comes from giving
1270 * an unknown SHA-1 to find_object() or an undefined mark to
1271 * find_mark(). Callers must test for this condition and use
1272 * the standard read_sha1_file() when it happens.
1273 *
1274 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1275 * find_mark(), where the mark was reloaded from an existing marks
1276 * file and is referencing an object that this fast-import process
1277 * instance did not write out to a packfile. Callers must test for
1278 * this condition and use read_sha1_file() instead.
1279 */
1280 static void *gfi_unpack_entry(
1281 struct object_entry *oe,
1282 unsigned long *sizep)
1283 {
1284 enum object_type type;
1285 struct packed_git *p = all_packs[oe->pack_id];
1286 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1287 /* The object is stored in the packfile we are writing to
1288 * and we have modified it since the last time we scanned
1289 * back to read a previously written object. If an old
1290 * window covered [p->pack_size, p->pack_size + 20) its
1291 * data is stale and is not valid. Closing all windows
1292 * and updating the packfile length ensures we can read
1293 * the newly written data.
1294 */
1295 close_pack_windows(p);
1296 sha1flush(pack_file);
1298 /* We have to offer 20 bytes additional on the end of
1299 * the packfile as the core unpacker code assumes the
1300 * footer is present at the file end and must promise
1301 * at least 20 bytes within any window it maps. But
1302 * we don't actually create the footer here.
1303 */
1304 p->pack_size = pack_size + 20;
1305 }
1306 return unpack_entry(p, oe->idx.offset, &type, sizep);
1307 }
1309 static const char *get_mode(const char *str, uint16_t *modep)
1310 {
1311 unsigned char c;
1312 uint16_t mode = 0;
1314 while ((c = *str++) != ' ') {
1315 if (c < '0' || c > '7')
1316 return NULL;
1317 mode = (mode << 3) + (c - '0');
1318 }
1319 *modep = mode;
1320 return str;
1321 }
1323 static void load_tree(struct tree_entry *root)
1324 {
1325 unsigned char *sha1 = root->versions[1].sha1;
1326 struct object_entry *myoe;
1327 struct tree_content *t;
1328 unsigned long size;
1329 char *buf;
1330 const char *c;
1332 root->tree = t = new_tree_content(8);
1333 if (is_null_sha1(sha1))
1334 return;
1336 myoe = find_object(sha1);
1337 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1338 if (myoe->type != OBJ_TREE)
1339 die("Not a tree: %s", sha1_to_hex(sha1));
1340 t->delta_depth = myoe->depth;
1341 buf = gfi_unpack_entry(myoe, &size);
1342 if (!buf)
1343 die("Can't load tree %s", sha1_to_hex(sha1));
1344 } else {
1345 enum object_type type;
1346 buf = read_sha1_file(sha1, &type, &size);
1347 if (!buf || type != OBJ_TREE)
1348 die("Can't load tree %s", sha1_to_hex(sha1));
1349 }
1351 c = buf;
1352 while (c != (buf + size)) {
1353 struct tree_entry *e = new_tree_entry();
1355 if (t->entry_count == t->entry_capacity)
1356 root->tree = t = grow_tree_content(t, t->entry_count);
1357 t->entries[t->entry_count++] = e;
1359 e->tree = NULL;
1360 c = get_mode(c, &e->versions[1].mode);
1361 if (!c)
1362 die("Corrupt mode in %s", sha1_to_hex(sha1));
1363 e->versions[0].mode = e->versions[1].mode;
1364 e->name = to_atom(c, strlen(c));
1365 c += e->name->str_len + 1;
1366 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1367 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1368 c += 20;
1369 }
1370 free(buf);
1371 }
1373 static int tecmp0 (const void *_a, const void *_b)
1374 {
1375 struct tree_entry *a = *((struct tree_entry**)_a);
1376 struct tree_entry *b = *((struct tree_entry**)_b);
1377 return base_name_compare(
1378 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1379 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1380 }
1382 static int tecmp1 (const void *_a, const void *_b)
1383 {
1384 struct tree_entry *a = *((struct tree_entry**)_a);
1385 struct tree_entry *b = *((struct tree_entry**)_b);
1386 return base_name_compare(
1387 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1388 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1389 }
1391 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1392 {
1393 size_t maxlen = 0;
1394 unsigned int i;
1396 if (!v)
1397 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1398 else
1399 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1401 for (i = 0; i < t->entry_count; i++) {
1402 if (t->entries[i]->versions[v].mode)
1403 maxlen += t->entries[i]->name->str_len + 34;
1404 }
1406 strbuf_reset(b);
1407 strbuf_grow(b, maxlen);
1408 for (i = 0; i < t->entry_count; i++) {
1409 struct tree_entry *e = t->entries[i];
1410 if (!e->versions[v].mode)
1411 continue;
1412 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1413 e->name->str_dat, '\0');
1414 strbuf_add(b, e->versions[v].sha1, 20);
1415 }
1416 }
1418 static void store_tree(struct tree_entry *root)
1419 {
1420 struct tree_content *t = root->tree;
1421 unsigned int i, j, del;
1422 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1423 struct object_entry *le;
1425 if (!is_null_sha1(root->versions[1].sha1))
1426 return;
1428 for (i = 0; i < t->entry_count; i++) {
1429 if (t->entries[i]->tree)
1430 store_tree(t->entries[i]);
1431 }
1433 le = find_object(root->versions[0].sha1);
1434 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1435 mktree(t, 0, &old_tree);
1436 lo.data = old_tree;
1437 lo.offset = le->idx.offset;
1438 lo.depth = t->delta_depth;
1439 }
1441 mktree(t, 1, &new_tree);
1442 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1444 t->delta_depth = lo.depth;
1445 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1446 struct tree_entry *e = t->entries[i];
1447 if (e->versions[1].mode) {
1448 e->versions[0].mode = e->versions[1].mode;
1449 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1450 t->entries[j++] = e;
1451 } else {
1452 release_tree_entry(e);
1453 del++;
1454 }
1455 }
1456 t->entry_count -= del;
1457 }
1459 static void tree_content_replace(
1460 struct tree_entry *root,
1461 const unsigned char *sha1,
1462 const uint16_t mode,
1463 struct tree_content *newtree)
1464 {
1465 if (!S_ISDIR(mode))
1466 die("Root cannot be a non-directory");
1467 hashcpy(root->versions[1].sha1, sha1);
1468 if (root->tree)
1469 release_tree_content_recursive(root->tree);
1470 root->tree = newtree;
1471 }
1473 static int tree_content_set(
1474 struct tree_entry *root,
1475 const char *p,
1476 const unsigned char *sha1,
1477 const uint16_t mode,
1478 struct tree_content *subtree)
1479 {
1480 struct tree_content *t;
1481 const char *slash1;
1482 unsigned int i, n;
1483 struct tree_entry *e;
1485 slash1 = strchr(p, '/');
1486 if (slash1)
1487 n = slash1 - p;
1488 else
1489 n = strlen(p);
1490 if (!n)
1491 die("Empty path component found in input");
1492 if (!slash1 && !S_ISDIR(mode) && subtree)
1493 die("Non-directories cannot have subtrees");
1495 if (!root->tree)
1496 load_tree(root);
1497 t = root->tree;
1498 for (i = 0; i < t->entry_count; i++) {
1499 e = t->entries[i];
1500 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1501 if (!slash1) {
1502 if (!S_ISDIR(mode)
1503 && e->versions[1].mode == mode
1504 && !hashcmp(e->versions[1].sha1, sha1))
1505 return 0;
1506 e->versions[1].mode = mode;
1507 hashcpy(e->versions[1].sha1, sha1);
1508 if (e->tree)
1509 release_tree_content_recursive(e->tree);
1510 e->tree = subtree;
1511 hashclr(root->versions[1].sha1);
1512 return 1;
1513 }
1514 if (!S_ISDIR(e->versions[1].mode)) {
1515 e->tree = new_tree_content(8);
1516 e->versions[1].mode = S_IFDIR;
1517 }
1518 if (!e->tree)
1519 load_tree(e);
1520 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1521 hashclr(root->versions[1].sha1);
1522 return 1;
1523 }
1524 return 0;
1525 }
1526 }
1528 if (t->entry_count == t->entry_capacity)
1529 root->tree = t = grow_tree_content(t, t->entry_count);
1530 e = new_tree_entry();
1531 e->name = to_atom(p, n);
1532 e->versions[0].mode = 0;
1533 hashclr(e->versions[0].sha1);
1534 t->entries[t->entry_count++] = e;
1535 if (slash1) {
1536 e->tree = new_tree_content(8);
1537 e->versions[1].mode = S_IFDIR;
1538 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1539 } else {
1540 e->tree = subtree;
1541 e->versions[1].mode = mode;
1542 hashcpy(e->versions[1].sha1, sha1);
1543 }
1544 hashclr(root->versions[1].sha1);
1545 return 1;
1546 }
1548 static int tree_content_remove(
1549 struct tree_entry *root,
1550 const char *p,
1551 struct tree_entry *backup_leaf)
1552 {
1553 struct tree_content *t;
1554 const char *slash1;
1555 unsigned int i, n;
1556 struct tree_entry *e;
1558 slash1 = strchr(p, '/');
1559 if (slash1)
1560 n = slash1 - p;
1561 else
1562 n = strlen(p);
1564 if (!root->tree)
1565 load_tree(root);
1566 t = root->tree;
1567 for (i = 0; i < t->entry_count; i++) {
1568 e = t->entries[i];
1569 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1570 if (slash1 && !S_ISDIR(e->versions[1].mode))
1571 /*
1572 * If p names a file in some subdirectory, and a
1573 * file or symlink matching the name of the
1574 * parent directory of p exists, then p cannot
1575 * exist and need not be deleted.
1576 */
1577 return 1;
1578 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1579 goto del_entry;
1580 if (!e->tree)
1581 load_tree(e);
1582 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1583 for (n = 0; n < e->tree->entry_count; n++) {
1584 if (e->tree->entries[n]->versions[1].mode) {
1585 hashclr(root->versions[1].sha1);
1586 return 1;
1587 }
1588 }
1589 backup_leaf = NULL;
1590 goto del_entry;
1591 }
1592 return 0;
1593 }
1594 }
1595 return 0;
1597 del_entry:
1598 if (backup_leaf)
1599 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1600 else if (e->tree)
1601 release_tree_content_recursive(e->tree);
1602 e->tree = NULL;
1603 e->versions[1].mode = 0;
1604 hashclr(e->versions[1].sha1);
1605 hashclr(root->versions[1].sha1);
1606 return 1;
1607 }
1609 static int tree_content_get(
1610 struct tree_entry *root,
1611 const char *p,
1612 struct tree_entry *leaf)
1613 {
1614 struct tree_content *t;
1615 const char *slash1;
1616 unsigned int i, n;
1617 struct tree_entry *e;
1619 slash1 = strchr(p, '/');
1620 if (slash1)
1621 n = slash1 - p;
1622 else
1623 n = strlen(p);
1625 if (!root->tree)
1626 load_tree(root);
1627 t = root->tree;
1628 for (i = 0; i < t->entry_count; i++) {
1629 e = t->entries[i];
1630 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1631 if (!slash1) {
1632 memcpy(leaf, e, sizeof(*leaf));
1633 if (e->tree && is_null_sha1(e->versions[1].sha1))
1634 leaf->tree = dup_tree_content(e->tree);
1635 else
1636 leaf->tree = NULL;
1637 return 1;
1638 }
1639 if (!S_ISDIR(e->versions[1].mode))
1640 return 0;
1641 if (!e->tree)
1642 load_tree(e);
1643 return tree_content_get(e, slash1 + 1, leaf);
1644 }
1645 }
1646 return 0;
1647 }
1649 static int update_branch(struct branch *b)
1650 {
1651 static const char *msg = "fast-import";
1652 struct ref_lock *lock;
1653 unsigned char old_sha1[20];
1655 if (is_null_sha1(b->sha1))
1656 return 0;
1657 if (read_ref(b->name, old_sha1))
1658 hashclr(old_sha1);
1659 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1660 if (!lock)
1661 return error("Unable to lock %s", b->name);
1662 if (!force_update && !is_null_sha1(old_sha1)) {
1663 struct commit *old_cmit, *new_cmit;
1665 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1666 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1667 if (!old_cmit || !new_cmit) {
1668 unlock_ref(lock);
1669 return error("Branch %s is missing commits.", b->name);
1670 }
1672 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1673 unlock_ref(lock);
1674 warning("Not updating %s"
1675 " (new tip %s does not contain %s)",
1676 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1677 return -1;
1678 }
1679 }
1680 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1681 return error("Unable to update %s", b->name);
1682 return 0;
1683 }
1685 static void dump_branches(void)
1686 {
1687 unsigned int i;
1688 struct branch *b;
1690 for (i = 0; i < branch_table_sz; i++) {
1691 for (b = branch_table[i]; b; b = b->table_next_branch)
1692 failure |= update_branch(b);
1693 }
1694 }
1696 static void dump_tags(void)
1697 {
1698 static const char *msg = "fast-import";
1699 struct tag *t;
1700 struct ref_lock *lock;
1701 char ref_name[PATH_MAX];
1703 for (t = first_tag; t; t = t->next_tag) {
1704 sprintf(ref_name, "tags/%s", t->name);
1705 lock = lock_ref_sha1(ref_name, NULL);
1706 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1707 failure |= error("Unable to update %s", ref_name);
1708 }
1709 }
1711 static void dump_marks_helper(FILE *f,
1712 uintmax_t base,
1713 struct mark_set *m)
1714 {
1715 uintmax_t k;
1716 if (m->shift) {
1717 for (k = 0; k < 1024; k++) {
1718 if (m->data.sets[k])
1719 dump_marks_helper(f, base + (k << m->shift),
1720 m->data.sets[k]);
1721 }
1722 } else {
1723 for (k = 0; k < 1024; k++) {
1724 if (m->data.marked[k])
1725 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1726 sha1_to_hex(m->data.marked[k]->idx.sha1));
1727 }
1728 }
1729 }
1731 static void dump_marks(void)
1732 {
1733 static struct lock_file mark_lock;
1734 int mark_fd;
1735 FILE *f;
1737 if (!export_marks_file)
1738 return;
1740 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1741 if (mark_fd < 0) {
1742 failure |= error("Unable to write marks file %s: %s",
1743 export_marks_file, strerror(errno));
1744 return;
1745 }
1747 f = fdopen(mark_fd, "w");
1748 if (!f) {
1749 int saved_errno = errno;
1750 rollback_lock_file(&mark_lock);
1751 failure |= error("Unable to write marks file %s: %s",
1752 export_marks_file, strerror(saved_errno));
1753 return;
1754 }
1756 /*
1757 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1758 * Assign -1 to the lock file descriptor so that commit_lock_file()
1759 * won't try to close() it.
1760 */
1761 mark_lock.fd = -1;
1763 dump_marks_helper(f, 0, marks);
1764 if (ferror(f) || fclose(f)) {
1765 int saved_errno = errno;
1766 rollback_lock_file(&mark_lock);
1767 failure |= error("Unable to write marks file %s: %s",
1768 export_marks_file, strerror(saved_errno));
1769 return;
1770 }
1772 if (commit_lock_file(&mark_lock)) {
1773 int saved_errno = errno;
1774 rollback_lock_file(&mark_lock);
1775 failure |= error("Unable to commit marks file %s: %s",
1776 export_marks_file, strerror(saved_errno));
1777 return;
1778 }
1779 }
1781 static void read_marks(void)
1782 {
1783 char line[512];
1784 FILE *f = fopen(import_marks_file, "r");
1785 if (!f)
1786 die_errno("cannot read '%s'", import_marks_file);
1787 while (fgets(line, sizeof(line), f)) {
1788 uintmax_t mark;
1789 char *end;
1790 unsigned char sha1[20];
1791 struct object_entry *e;
1793 end = strchr(line, '\n');
1794 if (line[0] != ':' || !end)
1795 die("corrupt mark line: %s", line);
1796 *end = 0;
1797 mark = strtoumax(line + 1, &end, 10);
1798 if (!mark || end == line + 1
1799 || *end != ' ' || get_sha1(end + 1, sha1))
1800 die("corrupt mark line: %s", line);
1801 e = find_object(sha1);
1802 if (!e) {
1803 enum object_type type = sha1_object_info(sha1, NULL);
1804 if (type < 0)
1805 die("object not found: %s", sha1_to_hex(sha1));
1806 e = insert_object(sha1);
1807 e->type = type;
1808 e->pack_id = MAX_PACK_ID;
1809 e->idx.offset = 1; /* just not zero! */
1810 }
1811 insert_mark(mark, e);
1812 }
1813 fclose(f);
1814 }
1817 static int read_next_command(void)
1818 {
1819 static int stdin_eof = 0;
1821 if (stdin_eof) {
1822 unread_command_buf = 0;
1823 return EOF;
1824 }
1826 for (;;) {
1827 if (unread_command_buf) {
1828 unread_command_buf = 0;
1829 } else {
1830 struct recent_command *rc;
1832 strbuf_detach(&command_buf, NULL);
1833 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1834 if (stdin_eof)
1835 return EOF;
1837 if (!seen_data_command
1838 && prefixcmp(command_buf.buf, "feature ")
1839 && prefixcmp(command_buf.buf, "option ")) {
1840 parse_argv();
1841 }
1843 rc = rc_free;
1844 if (rc)
1845 rc_free = rc->next;
1846 else {
1847 rc = cmd_hist.next;
1848 cmd_hist.next = rc->next;
1849 cmd_hist.next->prev = &cmd_hist;
1850 free(rc->buf);
1851 }
1853 rc->buf = command_buf.buf;
1854 rc->prev = cmd_tail;
1855 rc->next = cmd_hist.prev;
1856 rc->prev->next = rc;
1857 cmd_tail = rc;
1858 }
1859 if (!prefixcmp(command_buf.buf, "cat-blob ")) {
1860 parse_cat_blob();
1861 continue;
1862 }
1863 if (command_buf.buf[0] == '#')
1864 continue;
1865 return 0;
1866 }
1867 }
1869 static void skip_optional_lf(void)
1870 {
1871 int term_char = fgetc(stdin);
1872 if (term_char != '\n' && term_char != EOF)
1873 ungetc(term_char, stdin);
1874 }
1876 static void parse_mark(void)
1877 {
1878 if (!prefixcmp(command_buf.buf, "mark :")) {
1879 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1880 read_next_command();
1881 }
1882 else
1883 next_mark = 0;
1884 }
1886 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1887 {
1888 strbuf_reset(sb);
1890 if (prefixcmp(command_buf.buf, "data "))
1891 die("Expected 'data n' command, found: %s", command_buf.buf);
1893 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1894 char *term = xstrdup(command_buf.buf + 5 + 2);
1895 size_t term_len = command_buf.len - 5 - 2;
1897 strbuf_detach(&command_buf, NULL);
1898 for (;;) {
1899 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1900 die("EOF in data (terminator '%s' not found)", term);
1901 if (term_len == command_buf.len
1902 && !strcmp(term, command_buf.buf))
1903 break;
1904 strbuf_addbuf(sb, &command_buf);
1905 strbuf_addch(sb, '\n');
1906 }
1907 free(term);
1908 }
1909 else {
1910 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1911 size_t n = 0, length = (size_t)len;
1913 if (limit && limit < len) {
1914 *len_res = len;
1915 return 0;
1916 }
1917 if (length < len)
1918 die("data is too large to use in this context");
1920 while (n < length) {
1921 size_t s = strbuf_fread(sb, length - n, stdin);
1922 if (!s && feof(stdin))
1923 die("EOF in data (%lu bytes remaining)",
1924 (unsigned long)(length - n));
1925 n += s;
1926 }
1927 }
1929 skip_optional_lf();
1930 return 1;
1931 }
1933 static int validate_raw_date(const char *src, char *result, int maxlen)
1934 {
1935 const char *orig_src = src;
1936 char *endp;
1937 unsigned long num;
1939 errno = 0;
1941 num = strtoul(src, &endp, 10);
1942 /* NEEDSWORK: perhaps check for reasonable values? */
1943 if (errno || endp == src || *endp != ' ')
1944 return -1;
1946 src = endp + 1;
1947 if (*src != '-' && *src != '+')
1948 return -1;
1950 num = strtoul(src + 1, &endp, 10);
1951 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1952 1400 < num)
1953 return -1;
1955 strcpy(result, orig_src);
1956 return 0;
1957 }
1959 static char *parse_ident(const char *buf)
1960 {
1961 const char *gt;
1962 size_t name_len;
1963 char *ident;
1965 gt = strrchr(buf, '>');
1966 if (!gt)
1967 die("Missing > in ident string: %s", buf);
1968 gt++;
1969 if (*gt != ' ')
1970 die("Missing space after > in ident string: %s", buf);
1971 gt++;
1972 name_len = gt - buf;
1973 ident = xmalloc(name_len + 24);
1974 strncpy(ident, buf, name_len);
1976 switch (whenspec) {
1977 case WHENSPEC_RAW:
1978 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1979 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1980 break;
1981 case WHENSPEC_RFC2822:
1982 if (parse_date(gt, ident + name_len, 24) < 0)
1983 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1984 break;
1985 case WHENSPEC_NOW:
1986 if (strcmp("now", gt))
1987 die("Date in ident must be 'now': %s", buf);
1988 datestamp(ident + name_len, 24);
1989 break;
1990 }
1992 return ident;
1993 }
1995 static void parse_and_store_blob(
1996 struct last_object *last,
1997 unsigned char *sha1out,
1998 uintmax_t mark)
1999 {
2000 static struct strbuf buf = STRBUF_INIT;
2001 uintmax_t len;
2003 if (parse_data(&buf, big_file_threshold, &len))
2004 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2005 else {
2006 if (last) {
2007 strbuf_release(&last->data);
2008 last->offset = 0;
2009 last->depth = 0;
2010 }
2011 stream_blob(len, sha1out, mark);
2012 skip_optional_lf();
2013 }
2014 }
2016 static void parse_new_blob(void)
2017 {
2018 read_next_command();
2019 parse_mark();
2020 parse_and_store_blob(&last_blob, NULL, next_mark);
2021 }
2023 static void unload_one_branch(void)
2024 {
2025 while (cur_active_branches
2026 && cur_active_branches >= max_active_branches) {
2027 uintmax_t min_commit = ULONG_MAX;
2028 struct branch *e, *l = NULL, *p = NULL;
2030 for (e = active_branches; e; e = e->active_next_branch) {
2031 if (e->last_commit < min_commit) {
2032 p = l;
2033 min_commit = e->last_commit;
2034 }
2035 l = e;
2036 }
2038 if (p) {
2039 e = p->active_next_branch;
2040 p->active_next_branch = e->active_next_branch;
2041 } else {
2042 e = active_branches;
2043 active_branches = e->active_next_branch;
2044 }
2045 e->active = 0;
2046 e->active_next_branch = NULL;
2047 if (e->branch_tree.tree) {
2048 release_tree_content_recursive(e->branch_tree.tree);
2049 e->branch_tree.tree = NULL;
2050 }
2051 cur_active_branches--;
2052 }
2053 }
2055 static void load_branch(struct branch *b)
2056 {
2057 load_tree(&b->branch_tree);
2058 if (!b->active) {
2059 b->active = 1;
2060 b->active_next_branch = active_branches;
2061 active_branches = b;
2062 cur_active_branches++;
2063 branch_load_count++;
2064 }
2065 }
2067 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2068 {
2069 unsigned char fanout = 0;
2070 while ((num_notes >>= 8))
2071 fanout++;
2072 return fanout;
2073 }
2075 static void construct_path_with_fanout(const char *hex_sha1,
2076 unsigned char fanout, char *path)
2077 {
2078 unsigned int i = 0, j = 0;
2079 if (fanout >= 20)
2080 die("Too large fanout (%u)", fanout);
2081 while (fanout) {
2082 path[i++] = hex_sha1[j++];
2083 path[i++] = hex_sha1[j++];
2084 path[i++] = '/';
2085 fanout--;
2086 }
2087 memcpy(path + i, hex_sha1 + j, 40 - j);
2088 path[i + 40 - j] = '\0';
2089 }
2091 static uintmax_t do_change_note_fanout(
2092 struct tree_entry *orig_root, struct tree_entry *root,
2093 char *hex_sha1, unsigned int hex_sha1_len,
2094 char *fullpath, unsigned int fullpath_len,
2095 unsigned char fanout)
2096 {
2097 struct tree_content *t = root->tree;
2098 struct tree_entry *e, leaf;
2099 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2100 uintmax_t num_notes = 0;
2101 unsigned char sha1[20];
2102 char realpath[60];
2104 for (i = 0; t && i < t->entry_count; i++) {
2105 e = t->entries[i];
2106 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2107 tmp_fullpath_len = fullpath_len;
2109 /*
2110 * We're interested in EITHER existing note entries (entries
2111 * with exactly 40 hex chars in path, not including directory
2112 * separators), OR directory entries that may contain note
2113 * entries (with < 40 hex chars in path).
2114 * Also, each path component in a note entry must be a multiple
2115 * of 2 chars.
2116 */
2117 if (!e->versions[1].mode ||
2118 tmp_hex_sha1_len > 40 ||
2119 e->name->str_len % 2)
2120 continue;
2122 /* This _may_ be a note entry, or a subdir containing notes */
2123 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2124 e->name->str_len);
2125 if (tmp_fullpath_len)
2126 fullpath[tmp_fullpath_len++] = '/';
2127 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2128 e->name->str_len);
2129 tmp_fullpath_len += e->name->str_len;
2130 fullpath[tmp_fullpath_len] = '\0';
2132 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2133 /* This is a note entry */
2134 construct_path_with_fanout(hex_sha1, fanout, realpath);
2135 if (!strcmp(fullpath, realpath)) {
2136 /* Note entry is in correct location */
2137 num_notes++;
2138 continue;
2139 }
2141 /* Rename fullpath to realpath */
2142 if (!tree_content_remove(orig_root, fullpath, &leaf))
2143 die("Failed to remove path %s", fullpath);
2144 tree_content_set(orig_root, realpath,
2145 leaf.versions[1].sha1,
2146 leaf.versions[1].mode,
2147 leaf.tree);
2148 } else if (S_ISDIR(e->versions[1].mode)) {
2149 /* This is a subdir that may contain note entries */
2150 if (!e->tree)
2151 load_tree(e);
2152 num_notes += do_change_note_fanout(orig_root, e,
2153 hex_sha1, tmp_hex_sha1_len,
2154 fullpath, tmp_fullpath_len, fanout);
2155 }
2157 /* The above may have reallocated the current tree_content */
2158 t = root->tree;
2159 }
2160 return num_notes;
2161 }
2163 static uintmax_t change_note_fanout(struct tree_entry *root,
2164 unsigned char fanout)
2165 {
2166 char hex_sha1[40], path[60];
2167 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2168 }
2170 static void file_change_m(struct branch *b)
2171 {
2172 const char *p = command_buf.buf + 2;
2173 static struct strbuf uq = STRBUF_INIT;
2174 const char *endp;
2175 struct object_entry *oe = oe;
2176 unsigned char sha1[20];
2177 uint16_t mode, inline_data = 0;
2179 p = get_mode(p, &mode);
2180 if (!p)
2181 die("Corrupt mode: %s", command_buf.buf);
2182 switch (mode) {
2183 case 0644:
2184 case 0755:
2185 mode |= S_IFREG;
2186 case S_IFREG | 0644:
2187 case S_IFREG | 0755:
2188 case S_IFLNK:
2189 case S_IFDIR:
2190 case S_IFGITLINK:
2191 /* ok */
2192 break;
2193 default:
2194 die("Corrupt mode: %s", command_buf.buf);
2195 }
2197 if (*p == ':') {
2198 char *x;
2199 oe = find_mark(strtoumax(p + 1, &x, 10));
2200 hashcpy(sha1, oe->idx.sha1);
2201 p = x;
2202 } else if (!prefixcmp(p, "inline")) {
2203 inline_data = 1;
2204 p += 6;
2205 } else {
2206 if (get_sha1_hex(p, sha1))
2207 die("Invalid SHA1: %s", command_buf.buf);
2208 oe = find_object(sha1);
2209 p += 40;
2210 }
2211 if (*p++ != ' ')
2212 die("Missing space after SHA1: %s", command_buf.buf);
2214 strbuf_reset(&uq);
2215 if (!unquote_c_style(&uq, p, &endp)) {
2216 if (*endp)
2217 die("Garbage after path in: %s", command_buf.buf);
2218 p = uq.buf;
2219 }
2221 /* Git does not track empty, non-toplevel directories. */
2222 if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
2223 tree_content_remove(&b->branch_tree, p, NULL);
2224 return;
2225 }
2227 if (S_ISGITLINK(mode)) {
2228 if (inline_data)
2229 die("Git links cannot be specified 'inline': %s",
2230 command_buf.buf);
2231 else if (oe) {
2232 if (oe->type != OBJ_COMMIT)
2233 die("Not a commit (actually a %s): %s",
2234 typename(oe->type), command_buf.buf);
2235 }
2236 /*
2237 * Accept the sha1 without checking; it expected to be in
2238 * another repository.
2239 */
2240 } else if (inline_data) {
2241 if (S_ISDIR(mode))
2242 die("Directories cannot be specified 'inline': %s",
2243 command_buf.buf);
2244 if (p != uq.buf) {
2245 strbuf_addstr(&uq, p);
2246 p = uq.buf;
2247 }
2248 read_next_command();
2249 parse_and_store_blob(&last_blob, sha1, 0);
2250 } else {
2251 enum object_type expected = S_ISDIR(mode) ?
2252 OBJ_TREE: OBJ_BLOB;
2253 enum object_type type = oe ? oe->type :
2254 sha1_object_info(sha1, NULL);
2255 if (type < 0)
2256 die("%s not found: %s",
2257 S_ISDIR(mode) ? "Tree" : "Blob",
2258 command_buf.buf);
2259 if (type != expected)
2260 die("Not a %s (actually a %s): %s",
2261 typename(expected), typename(type),
2262 command_buf.buf);
2263 }
2265 if (!*p) {
2266 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2267 return;
2268 }
2269 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2270 }
2272 static void file_change_d(struct branch *b)
2273 {
2274 const char *p = command_buf.buf + 2;
2275 static struct strbuf uq = STRBUF_INIT;
2276 const char *endp;
2278 strbuf_reset(&uq);
2279 if (!unquote_c_style(&uq, p, &endp)) {
2280 if (*endp)
2281 die("Garbage after path in: %s", command_buf.buf);
2282 p = uq.buf;
2283 }
2284 tree_content_remove(&b->branch_tree, p, NULL);
2285 }
2287 static void file_change_cr(struct branch *b, int rename)
2288 {
2289 const char *s, *d;
2290 static struct strbuf s_uq = STRBUF_INIT;
2291 static struct strbuf d_uq = STRBUF_INIT;
2292 const char *endp;
2293 struct tree_entry leaf;
2295 s = command_buf.buf + 2;
2296 strbuf_reset(&s_uq);
2297 if (!unquote_c_style(&s_uq, s, &endp)) {
2298 if (*endp != ' ')
2299 die("Missing space after source: %s", command_buf.buf);
2300 } else {
2301 endp = strchr(s, ' ');
2302 if (!endp)
2303 die("Missing space after source: %s", command_buf.buf);
2304 strbuf_add(&s_uq, s, endp - s);
2305 }
2306 s = s_uq.buf;
2308 endp++;
2309 if (!*endp)
2310 die("Missing dest: %s", command_buf.buf);
2312 d = endp;
2313 strbuf_reset(&d_uq);
2314 if (!unquote_c_style(&d_uq, d, &endp)) {
2315 if (*endp)
2316 die("Garbage after dest in: %s", command_buf.buf);
2317 d = d_uq.buf;
2318 }
2320 memset(&leaf, 0, sizeof(leaf));
2321 if (rename)
2322 tree_content_remove(&b->branch_tree, s, &leaf);
2323 else
2324 tree_content_get(&b->branch_tree, s, &leaf);
2325 if (!leaf.versions[1].mode)
2326 die("Path %s not in branch", s);
2327 if (!*d) { /* C "path/to/subdir" "" */
2328 tree_content_replace(&b->branch_tree,
2329 leaf.versions[1].sha1,
2330 leaf.versions[1].mode,
2331 leaf.tree);
2332 return;
2333 }
2334 tree_content_set(&b->branch_tree, d,
2335 leaf.versions[1].sha1,
2336 leaf.versions[1].mode,
2337 leaf.tree);
2338 }
2340 static void note_change_n(struct branch *b, unsigned char old_fanout)
2341 {
2342 const char *p = command_buf.buf + 2;
2343 static struct strbuf uq = STRBUF_INIT;
2344 struct object_entry *oe = oe;
2345 struct branch *s;
2346 unsigned char sha1[20], commit_sha1[20];
2347 char path[60];
2348 uint16_t inline_data = 0;
2349 unsigned char new_fanout;
2351 /* <dataref> or 'inline' */
2352 if (*p == ':') {
2353 char *x;
2354 oe = find_mark(strtoumax(p + 1, &x, 10));
2355 hashcpy(sha1, oe->idx.sha1);
2356 p = x;
2357 } else if (!prefixcmp(p, "inline")) {
2358 inline_data = 1;
2359 p += 6;
2360 } else {
2361 if (get_sha1_hex(p, sha1))
2362 die("Invalid SHA1: %s", command_buf.buf);
2363 oe = find_object(sha1);
2364 p += 40;
2365 }
2366 if (*p++ != ' ')
2367 die("Missing space after SHA1: %s", command_buf.buf);
2369 /* <committish> */
2370 s = lookup_branch(p);
2371 if (s) {
2372 hashcpy(commit_sha1, s->sha1);
2373 } else if (*p == ':') {
2374 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2375 struct object_entry *commit_oe = find_mark(commit_mark);
2376 if (commit_oe->type != OBJ_COMMIT)
2377 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2378 hashcpy(commit_sha1, commit_oe->idx.sha1);
2379 } else if (!get_sha1(p, commit_sha1)) {
2380 unsigned long size;
2381 char *buf = read_object_with_reference(commit_sha1,
2382 commit_type, &size, commit_sha1);
2383 if (!buf || size < 46)
2384 die("Not a valid commit: %s", p);
2385 free(buf);
2386 } else
2387 die("Invalid ref name or SHA1 expression: %s", p);
2389 if (inline_data) {
2390 if (p != uq.buf) {
2391 strbuf_addstr(&uq, p);
2392 p = uq.buf;
2393 }
2394 read_next_command();
2395 parse_and_store_blob(&last_blob, sha1, 0);
2396 } else if (oe) {
2397 if (oe->type != OBJ_BLOB)
2398 die("Not a blob (actually a %s): %s",
2399 typename(oe->type), command_buf.buf);
2400 } else if (!is_null_sha1(sha1)) {
2401 enum object_type type = sha1_object_info(sha1, NULL);
2402 if (type < 0)
2403 die("Blob not found: %s", command_buf.buf);
2404 if (type != OBJ_BLOB)
2405 die("Not a blob (actually a %s): %s",
2406 typename(type), command_buf.buf);
2407 }
2409 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2410 if (tree_content_remove(&b->branch_tree, path, NULL))
2411 b->num_notes--;
2413 if (is_null_sha1(sha1))
2414 return; /* nothing to insert */
2416 b->num_notes++;
2417 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2418 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2419 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2420 }
2422 static void file_change_deleteall(struct branch *b)
2423 {
2424 release_tree_content_recursive(b->branch_tree.tree);
2425 hashclr(b->branch_tree.versions[0].sha1);
2426 hashclr(b->branch_tree.versions[1].sha1);
2427 load_tree(&b->branch_tree);
2428 b->num_notes = 0;
2429 }
2431 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2432 {
2433 if (!buf || size < 46)
2434 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2435 if (memcmp("tree ", buf, 5)
2436 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2437 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2438 hashcpy(b->branch_tree.versions[0].sha1,
2439 b->branch_tree.versions[1].sha1);
2440 }
2442 static void parse_from_existing(struct branch *b)
2443 {
2444 if (is_null_sha1(b->sha1)) {
2445 hashclr(b->branch_tree.versions[0].sha1);
2446 hashclr(b->branch_tree.versions[1].sha1);
2447 } else {
2448 unsigned long size;
2449 char *buf;
2451 buf = read_object_with_reference(b->sha1,
2452 commit_type, &size, b->sha1);
2453 parse_from_commit(b, buf, size);
2454 free(buf);
2455 }
2456 }
2458 static int parse_from(struct branch *b)
2459 {
2460 const char *from;
2461 struct branch *s;
2463 if (prefixcmp(command_buf.buf, "from "))
2464 return 0;
2466 if (b->branch_tree.tree) {
2467 release_tree_content_recursive(b->branch_tree.tree);
2468 b->branch_tree.tree = NULL;
2469 }
2471 from = strchr(command_buf.buf, ' ') + 1;
2472 s = lookup_branch(from);
2473 if (b == s)
2474 die("Can't create a branch from itself: %s", b->name);
2475 else if (s) {
2476 unsigned char *t = s->branch_tree.versions[1].sha1;
2477 hashcpy(b->sha1, s->sha1);
2478 hashcpy(b->branch_tree.versions[0].sha1, t);
2479 hashcpy(b->branch_tree.versions[1].sha1, t);
2480 } else if (*from == ':') {
2481 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2482 struct object_entry *oe = find_mark(idnum);
2483 if (oe->type != OBJ_COMMIT)
2484 die("Mark :%" PRIuMAX " not a commit", idnum);
2485 hashcpy(b->sha1, oe->idx.sha1);
2486 if (oe->pack_id != MAX_PACK_ID) {
2487 unsigned long size;
2488 char *buf = gfi_unpack_entry(oe, &size);
2489 parse_from_commit(b, buf, size);
2490 free(buf);
2491 } else
2492 parse_from_existing(b);
2493 } else if (!get_sha1(from, b->sha1))
2494 parse_from_existing(b);
2495 else
2496 die("Invalid ref name or SHA1 expression: %s", from);
2498 read_next_command();
2499 return 1;
2500 }
2502 static struct hash_list *parse_merge(unsigned int *count)
2503 {
2504 struct hash_list *list = NULL, *n, *e = e;
2505 const char *from;
2506 struct branch *s;
2508 *count = 0;
2509 while (!prefixcmp(command_buf.buf, "merge ")) {
2510 from = strchr(command_buf.buf, ' ') + 1;
2511 n = xmalloc(sizeof(*n));
2512 s = lookup_branch(from);
2513 if (s)
2514 hashcpy(n->sha1, s->sha1);
2515 else if (*from == ':') {
2516 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2517 struct object_entry *oe = find_mark(idnum);
2518 if (oe->type != OBJ_COMMIT)
2519 die("Mark :%" PRIuMAX " not a commit", idnum);
2520 hashcpy(n->sha1, oe->idx.sha1);
2521 } else if (!get_sha1(from, n->sha1)) {
2522 unsigned long size;
2523 char *buf = read_object_with_reference(n->sha1,
2524 commit_type, &size, n->sha1);
2525 if (!buf || size < 46)
2526 die("Not a valid commit: %s", from);
2527 free(buf);
2528 } else
2529 die("Invalid ref name or SHA1 expression: %s", from);
2531 n->next = NULL;
2532 if (list)
2533 e->next = n;
2534 else
2535 list = n;
2536 e = n;
2537 (*count)++;
2538 read_next_command();
2539 }
2540 return list;
2541 }
2543 static void parse_new_commit(void)
2544 {
2545 static struct strbuf msg = STRBUF_INIT;
2546 struct branch *b;
2547 char *sp;
2548 char *author = NULL;
2549 char *committer = NULL;
2550 struct hash_list *merge_list = NULL;
2551 unsigned int merge_count;
2552 unsigned char prev_fanout, new_fanout;
2554 /* Obtain the branch name from the rest of our command */
2555 sp = strchr(command_buf.buf, ' ') + 1;
2556 b = lookup_branch(sp);
2557 if (!b)
2558 b = new_branch(sp);
2560 read_next_command();
2561 parse_mark();
2562 if (!prefixcmp(command_buf.buf, "author ")) {
2563 author = parse_ident(command_buf.buf + 7);
2564 read_next_command();
2565 }
2566 if (!prefixcmp(command_buf.buf, "committer ")) {
2567 committer = parse_ident(command_buf.buf + 10);
2568 read_next_command();
2569 }
2570 if (!committer)
2571 die("Expected committer but didn't get one");
2572 parse_data(&msg, 0, NULL);
2573 read_next_command();
2574 parse_from(b);
2575 merge_list = parse_merge(&merge_count);
2577 /* ensure the branch is active/loaded */
2578 if (!b->branch_tree.tree || !max_active_branches) {
2579 unload_one_branch();
2580 load_branch(b);
2581 }
2583 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2585 /* file_change* */
2586 while (command_buf.len > 0) {
2587 if (!prefixcmp(command_buf.buf, "M "))
2588 file_change_m(b);
2589 else if (!prefixcmp(command_buf.buf, "D "))
2590 file_change_d(b);
2591 else if (!prefixcmp(command_buf.buf, "R "))
2592 file_change_cr(b, 1);
2593 else if (!prefixcmp(command_buf.buf, "C "))
2594 file_change_cr(b, 0);
2595 else if (!prefixcmp(command_buf.buf, "N "))
2596 note_change_n(b, prev_fanout);
2597 else if (!strcmp("deleteall", command_buf.buf))
2598 file_change_deleteall(b);
2599 else {
2600 unread_command_buf = 1;
2601 break;
2602 }
2603 if (read_next_command() == EOF)
2604 break;
2605 }
2607 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2608 if (new_fanout != prev_fanout)
2609 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2611 /* build the tree and the commit */
2612 store_tree(&b->branch_tree);
2613 hashcpy(b->branch_tree.versions[0].sha1,
2614 b->branch_tree.versions[1].sha1);
2616 strbuf_reset(&new_data);
2617 strbuf_addf(&new_data, "tree %s\n",
2618 sha1_to_hex(b->branch_tree.versions[1].sha1));
2619 if (!is_null_sha1(b->sha1))
2620 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2621 while (merge_list) {
2622 struct hash_list *next = merge_list->next;
2623 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2624 free(merge_list);
2625 merge_list = next;
2626 }
2627 strbuf_addf(&new_data,
2628 "author %s\n"
2629 "committer %s\n"
2630 "\n",
2631 author ? author : committer, committer);
2632 strbuf_addbuf(&new_data, &msg);
2633 free(author);
2634 free(committer);
2636 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2637 b->pack_id = pack_id;
2638 b->last_commit = object_count_by_type[OBJ_COMMIT];
2639 }
2641 static void parse_new_tag(void)
2642 {
2643 static struct strbuf msg = STRBUF_INIT;
2644 char *sp;
2645 const char *from;
2646 char *tagger;
2647 struct branch *s;
2648 struct tag *t;
2649 uintmax_t from_mark = 0;
2650 unsigned char sha1[20];
2651 enum object_type type;
2653 /* Obtain the new tag name from the rest of our command */
2654 sp = strchr(command_buf.buf, ' ') + 1;
2655 t = pool_alloc(sizeof(struct tag));
2656 t->next_tag = NULL;
2657 t->name = pool_strdup(sp);
2658 if (last_tag)
2659 last_tag->next_tag = t;
2660 else
2661 first_tag = t;
2662 last_tag = t;
2663 read_next_command();
2665 /* from ... */
2666 if (prefixcmp(command_buf.buf, "from "))
2667 die("Expected from command, got %s", command_buf.buf);
2668 from = strchr(command_buf.buf, ' ') + 1;
2669 s = lookup_branch(from);
2670 if (s) {
2671 hashcpy(sha1, s->sha1);
2672 type = OBJ_COMMIT;
2673 } else if (*from == ':') {
2674 struct object_entry *oe;
2675 from_mark = strtoumax(from + 1, NULL, 10);
2676 oe = find_mark(from_mark);
2677 type = oe->type;
2678 hashcpy(sha1, oe->idx.sha1);
2679 } else if (!get_sha1(from, sha1)) {
2680 unsigned long size;
2681 char *buf;
2683 buf = read_sha1_file(sha1, &type, &size);
2684 if (!buf || size < 46)
2685 die("Not a valid commit: %s", from);
2686 free(buf);
2687 } else
2688 die("Invalid ref name or SHA1 expression: %s", from);
2689 read_next_command();
2691 /* tagger ... */
2692 if (!prefixcmp(command_buf.buf, "tagger ")) {
2693 tagger = parse_ident(command_buf.buf + 7);
2694 read_next_command();
2695 } else
2696 tagger = NULL;
2698 /* tag payload/message */
2699 parse_data(&msg, 0, NULL);
2701 /* build the tag object */
2702 strbuf_reset(&new_data);
2704 strbuf_addf(&new_data,
2705 "object %s\n"
2706 "type %s\n"
2707 "tag %s\n",
2708 sha1_to_hex(sha1), typename(type), t->name);
2709 if (tagger)
2710 strbuf_addf(&new_data,
2711 "tagger %s\n", tagger);
2712 strbuf_addch(&new_data, '\n');
2713 strbuf_addbuf(&new_data, &msg);
2714 free(tagger);
2716 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2717 t->pack_id = MAX_PACK_ID;
2718 else
2719 t->pack_id = pack_id;
2720 }
2722 static void parse_reset_branch(void)
2723 {
2724 struct branch *b;
2725 char *sp;
2727 /* Obtain the branch name from the rest of our command */
2728 sp = strchr(command_buf.buf, ' ') + 1;
2729 b = lookup_branch(sp);
2730 if (b) {
2731 hashclr(b->sha1);
2732 hashclr(b->branch_tree.versions[0].sha1);
2733 hashclr(b->branch_tree.versions[1].sha1);
2734 if (b->branch_tree.tree) {
2735 release_tree_content_recursive(b->branch_tree.tree);
2736 b->branch_tree.tree = NULL;
2737 }
2738 }
2739 else
2740 b = new_branch(sp);
2741 read_next_command();
2742 parse_from(b);
2743 if (command_buf.len > 0)
2744 unread_command_buf = 1;
2745 }
2747 static void cat_blob_write(const char *buf, unsigned long size)
2748 {
2749 if (write_in_full(cat_blob_fd, buf, size) != size)
2750 die_errno("Write to frontend failed");
2751 }
2753 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2754 {
2755 struct strbuf line = STRBUF_INIT;
2756 unsigned long size;
2757 enum object_type type = 0;
2758 char *buf;
2760 if (!oe || oe->pack_id == MAX_PACK_ID) {
2761 buf = read_sha1_file(sha1, &type, &size);
2762 } else {
2763 type = oe->type;
2764 buf = gfi_unpack_entry(oe, &size);
2765 }
2767 /*
2768 * Output based on batch_one_object() from cat-file.c.
2769 */
2770 if (type <= 0) {
2771 strbuf_reset(&line);
2772 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2773 cat_blob_write(line.buf, line.len);
2774 strbuf_release(&line);
2775 free(buf);
2776 return;
2777 }
2778 if (!buf)
2779 die("Can't read object %s", sha1_to_hex(sha1));
2780 if (type != OBJ_BLOB)
2781 die("Object %s is a %s but a blob was expected.",
2782 sha1_to_hex(sha1), typename(type));
2783 strbuf_reset(&line);
2784 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2785 typename(type), size);
2786 cat_blob_write(line.buf, line.len);
2787 strbuf_release(&line);
2788 cat_blob_write(buf, size);
2789 cat_blob_write("\n", 1);
2790 free(buf);
2791 }
2793 static void parse_cat_blob(void)
2794 {
2795 const char *p;
2796 struct object_entry *oe = oe;
2797 unsigned char sha1[20];
2799 /* cat-blob SP <object> LF */
2800 p = command_buf.buf + strlen("cat-blob ");
2801 if (*p == ':') {
2802 char *x;
2803 oe = find_mark(strtoumax(p + 1, &x, 10));
2804 if (x == p + 1)
2805 die("Invalid mark: %s", command_buf.buf);
2806 if (!oe)
2807 die("Unknown mark: %s", command_buf.buf);
2808 if (*x)
2809 die("Garbage after mark: %s", command_buf.buf);
2810 hashcpy(sha1, oe->idx.sha1);
2811 } else {
2812 if (get_sha1_hex(p, sha1))
2813 die("Invalid SHA1: %s", command_buf.buf);
2814 if (p[40])
2815 die("Garbage after SHA1: %s", command_buf.buf);
2816 oe = find_object(sha1);
2817 }
2819 cat_blob(oe, sha1);
2820 }
2822 static void checkpoint(void)
2823 {
2824 checkpoint_requested = 0;
2825 if (object_count) {
2826 cycle_packfile();
2827 dump_branches();
2828 dump_tags();
2829 dump_marks();
2830 }
2831 }
2833 static void parse_checkpoint(void)
2834 {
2835 checkpoint_requested = 1;
2836 skip_optional_lf();
2837 }
2839 static void parse_progress(void)
2840 {
2841 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2842 fputc('\n', stdout);
2843 fflush(stdout);
2844 skip_optional_lf();
2845 }
2847 static char* make_fast_import_path(const char *path)
2848 {
2849 struct strbuf abs_path = STRBUF_INIT;
2851 if (!relative_marks_paths || is_absolute_path(path))
2852 return xstrdup(path);
2853 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
2854 return strbuf_detach(&abs_path, NULL);
2855 }
2857 static void option_import_marks(const char *marks, int from_stream)
2858 {
2859 if (import_marks_file) {
2860 if (from_stream)
2861 die("Only one import-marks command allowed per stream");
2863 /* read previous mark file */
2864 if(!import_marks_file_from_stream)
2865 read_marks();
2866 }
2868 import_marks_file = make_fast_import_path(marks);
2869 safe_create_leading_directories_const(import_marks_file);
2870 import_marks_file_from_stream = from_stream;
2871 }
2873 static void option_date_format(const char *fmt)
2874 {
2875 if (!strcmp(fmt, "raw"))
2876 whenspec = WHENSPEC_RAW;
2877 else if (!strcmp(fmt, "rfc2822"))
2878 whenspec = WHENSPEC_RFC2822;
2879 else if (!strcmp(fmt, "now"))
2880 whenspec = WHENSPEC_NOW;
2881 else
2882 die("unknown --date-format argument %s", fmt);
2883 }
2885 static unsigned long ulong_arg(const char *option, const char *arg)
2886 {
2887 char *endptr;
2888 unsigned long rv = strtoul(arg, &endptr, 0);
2889 if (strchr(arg, '-') || endptr == arg || *endptr)
2890 die("%s: argument must be a non-negative integer", option);
2891 return rv;
2892 }
2894 static void option_depth(const char *depth)
2895 {
2896 max_depth = ulong_arg("--depth", depth);
2897 if (max_depth > MAX_DEPTH)
2898 die("--depth cannot exceed %u", MAX_DEPTH);
2899 }
2901 static void option_active_branches(const char *branches)
2902 {
2903 max_active_branches = ulong_arg("--active-branches", branches);
2904 }
2906 static void option_export_marks(const char *marks)
2907 {
2908 export_marks_file = make_fast_import_path(marks);
2909 safe_create_leading_directories_const(export_marks_file);
2910 }
2912 static void option_cat_blob_fd(const char *fd)
2913 {
2914 unsigned long n = ulong_arg("--cat-blob-fd", fd);
2915 if (n > (unsigned long) INT_MAX)
2916 die("--cat-blob-fd cannot exceed %d", INT_MAX);
2917 cat_blob_fd = (int) n;
2918 }
2920 static void option_export_pack_edges(const char *edges)
2921 {
2922 if (pack_edges)
2923 fclose(pack_edges);
2924 pack_edges = fopen(edges, "a");
2925 if (!pack_edges)
2926 die_errno("Cannot open '%s'", edges);
2927 }
2929 static int parse_one_option(const char *option)
2930 {
2931 if (!prefixcmp(option, "max-pack-size=")) {
2932 unsigned long v;
2933 if (!git_parse_ulong(option + 14, &v))
2934 return 0;
2935 if (v < 8192) {
2936 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
2937 v *= 1024 * 1024;
2938 } else if (v < 1024 * 1024) {
2939 warning("minimum max-pack-size is 1 MiB");
2940 v = 1024 * 1024;
2941 }
2942 max_packsize = v;
2943 } else if (!prefixcmp(option, "big-file-threshold=")) {
2944 unsigned long v;
2945 if (!git_parse_ulong(option + 19, &v))
2946 return 0;
2947 big_file_threshold = v;
2948 } else if (!prefixcmp(option, "depth=")) {
2949 option_depth(option + 6);
2950 } else if (!prefixcmp(option, "active-branches=")) {
2951 option_active_branches(option + 16);
2952 } else if (!prefixcmp(option, "export-pack-edges=")) {
2953 option_export_pack_edges(option + 18);
2954 } else if (!prefixcmp(option, "quiet")) {
2955 show_stats = 0;
2956 } else if (!prefixcmp(option, "stats")) {
2957 show_stats = 1;
2958 } else {
2959 return 0;
2960 }
2962 return 1;
2963 }
2965 static int parse_one_feature(const char *feature, int from_stream)
2966 {
2967 if (!prefixcmp(feature, "date-format=")) {
2968 option_date_format(feature + 12);
2969 } else if (!prefixcmp(feature, "import-marks=")) {
2970 option_import_marks(feature + 13, from_stream);
2971 } else if (!prefixcmp(feature, "export-marks=")) {
2972 option_export_marks(feature + 13);
2973 } else if (!strcmp(feature, "cat-blob")) {
2974 ; /* Don't die - this feature is supported */
2975 } else if (!prefixcmp(feature, "relative-marks")) {
2976 relative_marks_paths = 1;
2977 } else if (!prefixcmp(feature, "no-relative-marks")) {
2978 relative_marks_paths = 0;
2979 } else if (!prefixcmp(feature, "force")) {
2980 force_update = 1;
2981 } else if (!strcmp(feature, "notes")) {
2982 ; /* do nothing; we have the feature */
2983 } else {
2984 return 0;
2985 }
2987 return 1;
2988 }
2990 static void parse_feature(void)
2991 {
2992 char *feature = command_buf.buf + 8;
2994 if (seen_data_command)
2995 die("Got feature command '%s' after data command", feature);
2997 if (parse_one_feature(feature, 1))
2998 return;
3000 die("This version of fast-import does not support feature %s.", feature);
3001 }
3003 static void parse_option(void)
3004 {
3005 char *option = command_buf.buf + 11;
3007 if (seen_data_command)
3008 die("Got option command '%s' after data command", option);
3010 if (parse_one_option(option))
3011 return;
3013 die("This version of fast-import does not support option: %s", option);
3014 }
3016 static int git_pack_config(const char *k, const char *v, void *cb)
3017 {
3018 if (!strcmp(k, "pack.depth")) {
3019 max_depth = git_config_int(k, v);
3020 if (max_depth > MAX_DEPTH)
3021 max_depth = MAX_DEPTH;
3022 return 0;
3023 }
3024 if (!strcmp(k, "pack.compression")) {
3025 int level = git_config_int(k, v);
3026 if (level == -1)
3027 level = Z_DEFAULT_COMPRESSION;
3028 else if (level < 0 || level > Z_BEST_COMPRESSION)
3029 die("bad pack compression level %d", level);
3030 pack_compression_level = level;
3031 pack_compression_seen = 1;
3032 return 0;
3033 }
3034 if (!strcmp(k, "pack.indexversion")) {
3035 pack_idx_default_version = git_config_int(k, v);
3036 if (pack_idx_default_version > 2)
3037 die("bad pack.indexversion=%"PRIu32,
3038 pack_idx_default_version);
3039 return 0;
3040 }
3041 if (!strcmp(k, "pack.packsizelimit")) {
3042 max_packsize = git_config_ulong(k, v);
3043 return 0;
3044 }
3045 if (!strcmp(k, "core.bigfilethreshold")) {
3046 long n = git_config_int(k, v);
3047 big_file_threshold = 0 < n ? n : 0;
3048 }
3049 return git_default_config(k, v, cb);
3050 }
3052 static const char fast_import_usage[] =
3053 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3055 static void parse_argv(void)
3056 {
3057 unsigned int i;
3059 for (i = 1; i < global_argc; i++) {
3060 const char *a = global_argv[i];
3062 if (*a != '-' || !strcmp(a, "--"))
3063 break;
3065 if (parse_one_option(a + 2))
3066 continue;
3068 if (parse_one_feature(a + 2, 0))
3069 continue;
3071 if (!prefixcmp(a + 2, "cat-blob-fd=")) {
3072 option_cat_blob_fd(a + 2 + strlen("cat-blob-fd="));
3073 continue;
3074 }
3076 die("unknown option %s", a);
3077 }
3078 if (i != global_argc)
3079 usage(fast_import_usage);
3081 seen_data_command = 1;
3082 if (import_marks_file)
3083 read_marks();
3084 }
3086 int main(int argc, const char **argv)
3087 {
3088 unsigned int i;
3090 git_extract_argv0_path(argv[0]);
3092 if (argc == 2 && !strcmp(argv[1], "-h"))
3093 usage(fast_import_usage);
3095 setup_git_directory();
3096 git_config(git_pack_config, NULL);
3097 if (!pack_compression_seen && core_compression_seen)
3098 pack_compression_level = core_compression_level;
3100 alloc_objects(object_entry_alloc);
3101 strbuf_init(&command_buf, 0);
3102 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3103 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3104 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3105 marks = pool_calloc(1, sizeof(struct mark_set));
3107 global_argc = argc;
3108 global_argv = argv;
3110 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3111 for (i = 0; i < (cmd_save - 1); i++)
3112 rc_free[i].next = &rc_free[i + 1];
3113 rc_free[cmd_save - 1].next = NULL;
3115 prepare_packed_git();
3116 start_packfile();
3117 set_die_routine(die_nicely);
3118 set_checkpoint_signal();
3119 while (read_next_command() != EOF) {
3120 if (!strcmp("blob", command_buf.buf))
3121 parse_new_blob();
3122 else if (!prefixcmp(command_buf.buf, "commit "))
3123 parse_new_commit();
3124 else if (!prefixcmp(command_buf.buf, "tag "))
3125 parse_new_tag();
3126 else if (!prefixcmp(command_buf.buf, "reset "))
3127 parse_reset_branch();
3128 else if (!strcmp("checkpoint", command_buf.buf))
3129 parse_checkpoint();
3130 else if (!prefixcmp(command_buf.buf, "progress "))
3131 parse_progress();
3132 else if (!prefixcmp(command_buf.buf, "feature "))
3133 parse_feature();
3134 else if (!prefixcmp(command_buf.buf, "option git "))
3135 parse_option();
3136 else if (!prefixcmp(command_buf.buf, "option "))
3137 /* ignore non-git options*/;
3138 else
3139 die("Unsupported command: %s", command_buf.buf);
3141 if (checkpoint_requested)
3142 checkpoint();
3143 }
3145 /* argv hasn't been parsed yet, do so */
3146 if (!seen_data_command)
3147 parse_argv();
3149 end_packfile();
3151 dump_branches();
3152 dump_tags();
3153 unkeep_all_packs();
3154 dump_marks();
3156 if (pack_edges)
3157 fclose(pack_edges);
3159 if (show_stats) {
3160 uintmax_t total_count = 0, duplicate_count = 0;
3161 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3162 total_count += object_count_by_type[i];
3163 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3164 duplicate_count += duplicate_count_by_type[i];
3166 fprintf(stderr, "%s statistics:\n", argv[0]);
3167 fprintf(stderr, "---------------------------------------------------------------------\n");
3168 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3169 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3170 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]);
3171 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]);
3172 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]);
3173 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]);
3174 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3175 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3176 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3177 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3178 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3179 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3180 fprintf(stderr, "---------------------------------------------------------------------\n");
3181 pack_report();
3182 fprintf(stderr, "---------------------------------------------------------------------\n");
3183 fprintf(stderr, "\n");
3184 }
3186 return failure ? 1 : 0;
3187 }