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 may appear anywhere in the input, except
136 # within a data command. Any form of the data command
137 # always escapes the related input from comment processing.
138 #
139 # In case it is not clear, the '#' that starts the comment
140 # must be the first character on that line (an lf
141 # preceded it).
142 #
143 comment ::= '#' not_lf* lf;
144 not_lf ::= # Any byte that is not ASCII newline (LF);
145 */
147 #include "builtin.h"
148 #include "cache.h"
149 #include "object.h"
150 #include "blob.h"
151 #include "tree.h"
152 #include "commit.h"
153 #include "delta.h"
154 #include "pack.h"
155 #include "refs.h"
156 #include "csum-file.h"
157 #include "quote.h"
158 #include "exec_cmd.h"
160 #define PACK_ID_BITS 16
161 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
162 #define DEPTH_BITS 13
163 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
165 struct object_entry
166 {
167 struct pack_idx_entry idx;
168 struct object_entry *next;
169 uint32_t type : TYPE_BITS,
170 pack_id : PACK_ID_BITS,
171 depth : DEPTH_BITS;
172 };
174 struct object_entry_pool
175 {
176 struct object_entry_pool *next_pool;
177 struct object_entry *next_free;
178 struct object_entry *end;
179 struct object_entry entries[FLEX_ARRAY]; /* more */
180 };
182 struct mark_set
183 {
184 union {
185 struct object_entry *marked[1024];
186 struct mark_set *sets[1024];
187 } data;
188 unsigned int shift;
189 };
191 struct last_object
192 {
193 struct strbuf data;
194 uint32_t offset;
195 unsigned int depth;
196 unsigned no_swap : 1;
197 };
199 struct mem_pool
200 {
201 struct mem_pool *next_pool;
202 char *next_free;
203 char *end;
204 uintmax_t space[FLEX_ARRAY]; /* more */
205 };
207 struct atom_str
208 {
209 struct atom_str *next_atom;
210 unsigned short str_len;
211 char str_dat[FLEX_ARRAY]; /* more */
212 };
214 struct tree_content;
215 struct tree_entry
216 {
217 struct tree_content *tree;
218 struct atom_str *name;
219 struct tree_entry_ms
220 {
221 uint16_t mode;
222 unsigned char sha1[20];
223 } versions[2];
224 };
226 struct tree_content
227 {
228 unsigned int entry_capacity; /* must match avail_tree_content */
229 unsigned int entry_count;
230 unsigned int delta_depth;
231 struct tree_entry *entries[FLEX_ARRAY]; /* more */
232 };
234 struct avail_tree_content
235 {
236 unsigned int entry_capacity; /* must match tree_content */
237 struct avail_tree_content *next_avail;
238 };
240 struct branch
241 {
242 struct branch *table_next_branch;
243 struct branch *active_next_branch;
244 const char *name;
245 struct tree_entry branch_tree;
246 uintmax_t last_commit;
247 uintmax_t num_notes;
248 unsigned active : 1;
249 unsigned pack_id : PACK_ID_BITS;
250 unsigned char sha1[20];
251 };
253 struct tag
254 {
255 struct tag *next_tag;
256 const char *name;
257 unsigned int pack_id;
258 unsigned char sha1[20];
259 };
261 struct hash_list
262 {
263 struct hash_list *next;
264 unsigned char sha1[20];
265 };
267 typedef enum {
268 WHENSPEC_RAW = 1,
269 WHENSPEC_RFC2822,
270 WHENSPEC_NOW,
271 } whenspec_type;
273 struct recent_command
274 {
275 struct recent_command *prev;
276 struct recent_command *next;
277 char *buf;
278 };
280 /* Configured limits on output */
281 static unsigned long max_depth = 10;
282 static off_t max_packsize = (1LL << 32) - 1;
283 static uintmax_t big_file_threshold = 512 * 1024 * 1024;
284 static int force_update;
285 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
286 static int pack_compression_seen;
288 /* Stats and misc. counters */
289 static uintmax_t alloc_count;
290 static uintmax_t marks_set_count;
291 static uintmax_t object_count_by_type[1 << TYPE_BITS];
292 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
293 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
294 static unsigned long object_count;
295 static unsigned long branch_count;
296 static unsigned long branch_load_count;
297 static int failure;
298 static FILE *pack_edges;
299 static unsigned int show_stats = 1;
300 static int global_argc;
301 static const char **global_argv;
303 /* Memory pools */
304 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
305 static size_t total_allocd;
306 static struct mem_pool *mem_pool;
308 /* Atom management */
309 static unsigned int atom_table_sz = 4451;
310 static unsigned int atom_cnt;
311 static struct atom_str **atom_table;
313 /* The .pack file being generated */
314 static unsigned int pack_id;
315 static struct sha1file *pack_file;
316 static struct packed_git *pack_data;
317 static struct packed_git **all_packs;
318 static unsigned long pack_size;
320 /* Table of objects we've written. */
321 static unsigned int object_entry_alloc = 5000;
322 static struct object_entry_pool *blocks;
323 static struct object_entry *object_table[1 << 16];
324 static struct mark_set *marks;
325 static const char *export_marks_file;
326 static const char *import_marks_file;
327 static int import_marks_file_from_stream;
328 static int relative_marks_paths;
330 /* Our last blob */
331 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
333 /* Tree management */
334 static unsigned int tree_entry_alloc = 1000;
335 static void *avail_tree_entry;
336 static unsigned int avail_tree_table_sz = 100;
337 static struct avail_tree_content **avail_tree_table;
338 static struct strbuf old_tree = STRBUF_INIT;
339 static struct strbuf new_tree = STRBUF_INIT;
341 /* Branch data */
342 static unsigned long max_active_branches = 5;
343 static unsigned long cur_active_branches;
344 static unsigned long branch_table_sz = 1039;
345 static struct branch **branch_table;
346 static struct branch *active_branches;
348 /* Tag data */
349 static struct tag *first_tag;
350 static struct tag *last_tag;
352 /* Input stream parsing */
353 static whenspec_type whenspec = WHENSPEC_RAW;
354 static struct strbuf command_buf = STRBUF_INIT;
355 static int unread_command_buf;
356 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
357 static struct recent_command *cmd_tail = &cmd_hist;
358 static struct recent_command *rc_free;
359 static unsigned int cmd_save = 100;
360 static uintmax_t next_mark;
361 static struct strbuf new_data = STRBUF_INIT;
362 static int seen_data_command;
364 static void parse_argv(void);
366 static void write_branch_report(FILE *rpt, struct branch *b)
367 {
368 fprintf(rpt, "%s:\n", b->name);
370 fprintf(rpt, " status :");
371 if (b->active)
372 fputs(" active", rpt);
373 if (b->branch_tree.tree)
374 fputs(" loaded", rpt);
375 if (is_null_sha1(b->branch_tree.versions[1].sha1))
376 fputs(" dirty", rpt);
377 fputc('\n', rpt);
379 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
380 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
381 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
382 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
384 fputs(" last pack : ", rpt);
385 if (b->pack_id < MAX_PACK_ID)
386 fprintf(rpt, "%u", b->pack_id);
387 fputc('\n', rpt);
389 fputc('\n', rpt);
390 }
392 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
394 static void write_crash_report(const char *err)
395 {
396 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
397 FILE *rpt = fopen(loc, "w");
398 struct branch *b;
399 unsigned long lu;
400 struct recent_command *rc;
402 if (!rpt) {
403 error("can't write crash report %s: %s", loc, strerror(errno));
404 return;
405 }
407 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
409 fprintf(rpt, "fast-import crash report:\n");
410 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
411 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
412 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
413 fputc('\n', rpt);
415 fputs("fatal: ", rpt);
416 fputs(err, rpt);
417 fputc('\n', rpt);
419 fputc('\n', rpt);
420 fputs("Most Recent Commands Before Crash\n", rpt);
421 fputs("---------------------------------\n", rpt);
422 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
423 if (rc->next == &cmd_hist)
424 fputs("* ", rpt);
425 else
426 fputs(" ", rpt);
427 fputs(rc->buf, rpt);
428 fputc('\n', rpt);
429 }
431 fputc('\n', rpt);
432 fputs("Active Branch LRU\n", rpt);
433 fputs("-----------------\n", rpt);
434 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
435 cur_active_branches,
436 max_active_branches);
437 fputc('\n', rpt);
438 fputs(" pos clock name\n", rpt);
439 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
440 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
441 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
442 ++lu, b->last_commit, b->name);
444 fputc('\n', rpt);
445 fputs("Inactive Branches\n", rpt);
446 fputs("-----------------\n", rpt);
447 for (lu = 0; lu < branch_table_sz; lu++) {
448 for (b = branch_table[lu]; b; b = b->table_next_branch)
449 write_branch_report(rpt, b);
450 }
452 if (first_tag) {
453 struct tag *tg;
454 fputc('\n', rpt);
455 fputs("Annotated Tags\n", rpt);
456 fputs("--------------\n", rpt);
457 for (tg = first_tag; tg; tg = tg->next_tag) {
458 fputs(sha1_to_hex(tg->sha1), rpt);
459 fputc(' ', rpt);
460 fputs(tg->name, rpt);
461 fputc('\n', rpt);
462 }
463 }
465 fputc('\n', rpt);
466 fputs("Marks\n", rpt);
467 fputs("-----\n", rpt);
468 if (export_marks_file)
469 fprintf(rpt, " exported to %s\n", export_marks_file);
470 else
471 dump_marks_helper(rpt, 0, marks);
473 fputc('\n', rpt);
474 fputs("-------------------\n", rpt);
475 fputs("END OF CRASH REPORT\n", rpt);
476 fclose(rpt);
477 }
479 static void end_packfile(void);
480 static void unkeep_all_packs(void);
481 static void dump_marks(void);
483 static NORETURN void die_nicely(const char *err, va_list params)
484 {
485 static int zombie;
486 char message[2 * PATH_MAX];
488 vsnprintf(message, sizeof(message), err, params);
489 fputs("fatal: ", stderr);
490 fputs(message, stderr);
491 fputc('\n', stderr);
493 if (!zombie) {
494 zombie = 1;
495 write_crash_report(message);
496 end_packfile();
497 unkeep_all_packs();
498 dump_marks();
499 }
500 exit(128);
501 }
503 static void alloc_objects(unsigned int cnt)
504 {
505 struct object_entry_pool *b;
507 b = xmalloc(sizeof(struct object_entry_pool)
508 + cnt * sizeof(struct object_entry));
509 b->next_pool = blocks;
510 b->next_free = b->entries;
511 b->end = b->entries + cnt;
512 blocks = b;
513 alloc_count += cnt;
514 }
516 static struct object_entry *new_object(unsigned char *sha1)
517 {
518 struct object_entry *e;
520 if (blocks->next_free == blocks->end)
521 alloc_objects(object_entry_alloc);
523 e = blocks->next_free++;
524 hashcpy(e->idx.sha1, sha1);
525 return e;
526 }
528 static struct object_entry *find_object(unsigned char *sha1)
529 {
530 unsigned int h = sha1[0] << 8 | sha1[1];
531 struct object_entry *e;
532 for (e = object_table[h]; e; e = e->next)
533 if (!hashcmp(sha1, e->idx.sha1))
534 return e;
535 return NULL;
536 }
538 static struct object_entry *insert_object(unsigned char *sha1)
539 {
540 unsigned int h = sha1[0] << 8 | sha1[1];
541 struct object_entry *e = object_table[h];
542 struct object_entry *p = NULL;
544 while (e) {
545 if (!hashcmp(sha1, e->idx.sha1))
546 return e;
547 p = e;
548 e = e->next;
549 }
551 e = new_object(sha1);
552 e->next = NULL;
553 e->idx.offset = 0;
554 if (p)
555 p->next = e;
556 else
557 object_table[h] = e;
558 return e;
559 }
561 static unsigned int hc_str(const char *s, size_t len)
562 {
563 unsigned int r = 0;
564 while (len-- > 0)
565 r = r * 31 + *s++;
566 return r;
567 }
569 static void *pool_alloc(size_t len)
570 {
571 struct mem_pool *p;
572 void *r;
574 /* round up to a 'uintmax_t' alignment */
575 if (len & (sizeof(uintmax_t) - 1))
576 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
578 for (p = mem_pool; p; p = p->next_pool)
579 if ((p->end - p->next_free >= len))
580 break;
582 if (!p) {
583 if (len >= (mem_pool_alloc/2)) {
584 total_allocd += len;
585 return xmalloc(len);
586 }
587 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
588 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
589 p->next_pool = mem_pool;
590 p->next_free = (char *) p->space;
591 p->end = p->next_free + mem_pool_alloc;
592 mem_pool = p;
593 }
595 r = p->next_free;
596 p->next_free += len;
597 return r;
598 }
600 static void *pool_calloc(size_t count, size_t size)
601 {
602 size_t len = count * size;
603 void *r = pool_alloc(len);
604 memset(r, 0, len);
605 return r;
606 }
608 static char *pool_strdup(const char *s)
609 {
610 char *r = pool_alloc(strlen(s) + 1);
611 strcpy(r, s);
612 return r;
613 }
615 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
616 {
617 struct mark_set *s = marks;
618 while ((idnum >> s->shift) >= 1024) {
619 s = pool_calloc(1, sizeof(struct mark_set));
620 s->shift = marks->shift + 10;
621 s->data.sets[0] = marks;
622 marks = s;
623 }
624 while (s->shift) {
625 uintmax_t i = idnum >> s->shift;
626 idnum -= i << s->shift;
627 if (!s->data.sets[i]) {
628 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
629 s->data.sets[i]->shift = s->shift - 10;
630 }
631 s = s->data.sets[i];
632 }
633 if (!s->data.marked[idnum])
634 marks_set_count++;
635 s->data.marked[idnum] = oe;
636 }
638 static struct object_entry *find_mark(uintmax_t idnum)
639 {
640 uintmax_t orig_idnum = idnum;
641 struct mark_set *s = marks;
642 struct object_entry *oe = NULL;
643 if ((idnum >> s->shift) < 1024) {
644 while (s && s->shift) {
645 uintmax_t i = idnum >> s->shift;
646 idnum -= i << s->shift;
647 s = s->data.sets[i];
648 }
649 if (s)
650 oe = s->data.marked[idnum];
651 }
652 if (!oe)
653 die("mark :%" PRIuMAX " not declared", orig_idnum);
654 return oe;
655 }
657 static struct atom_str *to_atom(const char *s, unsigned short len)
658 {
659 unsigned int hc = hc_str(s, len) % atom_table_sz;
660 struct atom_str *c;
662 for (c = atom_table[hc]; c; c = c->next_atom)
663 if (c->str_len == len && !strncmp(s, c->str_dat, len))
664 return c;
666 c = pool_alloc(sizeof(struct atom_str) + len + 1);
667 c->str_len = len;
668 strncpy(c->str_dat, s, len);
669 c->str_dat[len] = 0;
670 c->next_atom = atom_table[hc];
671 atom_table[hc] = c;
672 atom_cnt++;
673 return c;
674 }
676 static struct branch *lookup_branch(const char *name)
677 {
678 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
679 struct branch *b;
681 for (b = branch_table[hc]; b; b = b->table_next_branch)
682 if (!strcmp(name, b->name))
683 return b;
684 return NULL;
685 }
687 static struct branch *new_branch(const char *name)
688 {
689 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
690 struct branch *b = lookup_branch(name);
692 if (b)
693 die("Invalid attempt to create duplicate branch: %s", name);
694 switch (check_ref_format(name)) {
695 case 0: break; /* its valid */
696 case CHECK_REF_FORMAT_ONELEVEL:
697 break; /* valid, but too few '/', allow anyway */
698 default:
699 die("Branch name doesn't conform to GIT standards: %s", name);
700 }
702 b = pool_calloc(1, sizeof(struct branch));
703 b->name = pool_strdup(name);
704 b->table_next_branch = branch_table[hc];
705 b->branch_tree.versions[0].mode = S_IFDIR;
706 b->branch_tree.versions[1].mode = S_IFDIR;
707 b->num_notes = 0;
708 b->active = 0;
709 b->pack_id = MAX_PACK_ID;
710 branch_table[hc] = b;
711 branch_count++;
712 return b;
713 }
715 static unsigned int hc_entries(unsigned int cnt)
716 {
717 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
718 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
719 }
721 static struct tree_content *new_tree_content(unsigned int cnt)
722 {
723 struct avail_tree_content *f, *l = NULL;
724 struct tree_content *t;
725 unsigned int hc = hc_entries(cnt);
727 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
728 if (f->entry_capacity >= cnt)
729 break;
731 if (f) {
732 if (l)
733 l->next_avail = f->next_avail;
734 else
735 avail_tree_table[hc] = f->next_avail;
736 } else {
737 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
738 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
739 f->entry_capacity = cnt;
740 }
742 t = (struct tree_content*)f;
743 t->entry_count = 0;
744 t->delta_depth = 0;
745 return t;
746 }
748 static void release_tree_entry(struct tree_entry *e);
749 static void release_tree_content(struct tree_content *t)
750 {
751 struct avail_tree_content *f = (struct avail_tree_content*)t;
752 unsigned int hc = hc_entries(f->entry_capacity);
753 f->next_avail = avail_tree_table[hc];
754 avail_tree_table[hc] = f;
755 }
757 static void release_tree_content_recursive(struct tree_content *t)
758 {
759 unsigned int i;
760 for (i = 0; i < t->entry_count; i++)
761 release_tree_entry(t->entries[i]);
762 release_tree_content(t);
763 }
765 static struct tree_content *grow_tree_content(
766 struct tree_content *t,
767 int amt)
768 {
769 struct tree_content *r = new_tree_content(t->entry_count + amt);
770 r->entry_count = t->entry_count;
771 r->delta_depth = t->delta_depth;
772 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
773 release_tree_content(t);
774 return r;
775 }
777 static struct tree_entry *new_tree_entry(void)
778 {
779 struct tree_entry *e;
781 if (!avail_tree_entry) {
782 unsigned int n = tree_entry_alloc;
783 total_allocd += n * sizeof(struct tree_entry);
784 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
785 while (n-- > 1) {
786 *((void**)e) = e + 1;
787 e++;
788 }
789 *((void**)e) = NULL;
790 }
792 e = avail_tree_entry;
793 avail_tree_entry = *((void**)e);
794 return e;
795 }
797 static void release_tree_entry(struct tree_entry *e)
798 {
799 if (e->tree)
800 release_tree_content_recursive(e->tree);
801 *((void**)e) = avail_tree_entry;
802 avail_tree_entry = e;
803 }
805 static struct tree_content *dup_tree_content(struct tree_content *s)
806 {
807 struct tree_content *d;
808 struct tree_entry *a, *b;
809 unsigned int i;
811 if (!s)
812 return NULL;
813 d = new_tree_content(s->entry_count);
814 for (i = 0; i < s->entry_count; i++) {
815 a = s->entries[i];
816 b = new_tree_entry();
817 memcpy(b, a, sizeof(*a));
818 if (a->tree && is_null_sha1(b->versions[1].sha1))
819 b->tree = dup_tree_content(a->tree);
820 else
821 b->tree = NULL;
822 d->entries[i] = b;
823 }
824 d->entry_count = s->entry_count;
825 d->delta_depth = s->delta_depth;
827 return d;
828 }
830 static void start_packfile(void)
831 {
832 static char tmpfile[PATH_MAX];
833 struct packed_git *p;
834 struct pack_header hdr;
835 int pack_fd;
837 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
838 "pack/tmp_pack_XXXXXX");
839 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
840 strcpy(p->pack_name, tmpfile);
841 p->pack_fd = pack_fd;
842 pack_file = sha1fd(pack_fd, p->pack_name);
844 hdr.hdr_signature = htonl(PACK_SIGNATURE);
845 hdr.hdr_version = htonl(2);
846 hdr.hdr_entries = 0;
847 sha1write(pack_file, &hdr, sizeof(hdr));
849 pack_data = p;
850 pack_size = sizeof(hdr);
851 object_count = 0;
853 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
854 all_packs[pack_id] = p;
855 }
857 static int oecmp (const void *a_, const void *b_)
858 {
859 struct object_entry *a = *((struct object_entry**)a_);
860 struct object_entry *b = *((struct object_entry**)b_);
861 return hashcmp(a->idx.sha1, b->idx.sha1);
862 }
864 static char *create_index(void)
865 {
866 static char tmpfile[PATH_MAX];
867 git_SHA_CTX ctx;
868 struct sha1file *f;
869 struct object_entry **idx, **c, **last, *e;
870 struct object_entry_pool *o;
871 uint32_t array[256];
872 int i, idx_fd;
874 /* Build the sorted table of object IDs. */
875 idx = xmalloc(object_count * sizeof(struct object_entry*));
876 c = idx;
877 for (o = blocks; o; o = o->next_pool)
878 for (e = o->next_free; e-- != o->entries;)
879 if (pack_id == e->pack_id)
880 *c++ = e;
881 last = idx + object_count;
882 if (c != last)
883 die("internal consistency error creating the index");
884 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
886 /* Generate the fan-out array. */
887 c = idx;
888 for (i = 0; i < 256; i++) {
889 struct object_entry **next = c;
890 while (next < last) {
891 if ((*next)->idx.sha1[0] != i)
892 break;
893 next++;
894 }
895 array[i] = htonl(next - idx);
896 c = next;
897 }
899 idx_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
900 "pack/tmp_idx_XXXXXX");
901 f = sha1fd(idx_fd, tmpfile);
902 sha1write(f, array, 256 * sizeof(int));
903 git_SHA1_Init(&ctx);
904 for (c = idx; c != last; c++) {
905 uint32_t offset = htonl((*c)->idx.offset);
906 sha1write(f, &offset, 4);
907 sha1write(f, (*c)->idx.sha1, sizeof((*c)->idx.sha1));
908 git_SHA1_Update(&ctx, (*c)->idx.sha1, 20);
909 }
910 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
911 sha1close(f, NULL, CSUM_FSYNC);
912 free(idx);
913 git_SHA1_Final(pack_data->sha1, &ctx);
914 return tmpfile;
915 }
917 static char *keep_pack(char *curr_index_name)
918 {
919 static char name[PATH_MAX];
920 static const char *keep_msg = "fast-import";
921 int keep_fd;
923 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
924 if (keep_fd < 0)
925 die_errno("cannot create keep file");
926 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
927 if (close(keep_fd))
928 die_errno("failed to write keep file");
930 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
931 get_object_directory(), sha1_to_hex(pack_data->sha1));
932 if (move_temp_to_file(pack_data->pack_name, name))
933 die("cannot store pack file");
935 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
936 get_object_directory(), sha1_to_hex(pack_data->sha1));
937 if (move_temp_to_file(curr_index_name, name))
938 die("cannot store index file");
939 return name;
940 }
942 static void unkeep_all_packs(void)
943 {
944 static char name[PATH_MAX];
945 int k;
947 for (k = 0; k < pack_id; k++) {
948 struct packed_git *p = all_packs[k];
949 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
950 get_object_directory(), sha1_to_hex(p->sha1));
951 unlink_or_warn(name);
952 }
953 }
955 static void end_packfile(void)
956 {
957 struct packed_git *old_p = pack_data, *new_p;
959 clear_delta_base_cache();
960 if (object_count) {
961 unsigned char cur_pack_sha1[20];
962 char *idx_name;
963 int i;
964 struct branch *b;
965 struct tag *t;
967 close_pack_windows(pack_data);
968 sha1close(pack_file, cur_pack_sha1, 0);
969 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
970 pack_data->pack_name, object_count,
971 cur_pack_sha1, pack_size);
972 close(pack_data->pack_fd);
973 idx_name = keep_pack(create_index());
975 /* Register the packfile with core git's machinery. */
976 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
977 if (!new_p)
978 die("core git rejected index %s", idx_name);
979 all_packs[pack_id] = new_p;
980 install_packed_git(new_p);
982 /* Print the boundary */
983 if (pack_edges) {
984 fprintf(pack_edges, "%s:", new_p->pack_name);
985 for (i = 0; i < branch_table_sz; i++) {
986 for (b = branch_table[i]; b; b = b->table_next_branch) {
987 if (b->pack_id == pack_id)
988 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
989 }
990 }
991 for (t = first_tag; t; t = t->next_tag) {
992 if (t->pack_id == pack_id)
993 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
994 }
995 fputc('\n', pack_edges);
996 fflush(pack_edges);
997 }
999 pack_id++;
1000 }
1001 else {
1002 close(old_p->pack_fd);
1003 unlink_or_warn(old_p->pack_name);
1004 }
1005 free(old_p);
1007 /* We can't carry a delta across packfiles. */
1008 strbuf_release(&last_blob.data);
1009 last_blob.offset = 0;
1010 last_blob.depth = 0;
1011 }
1013 static void cycle_packfile(void)
1014 {
1015 end_packfile();
1016 start_packfile();
1017 }
1019 static size_t encode_header(
1020 enum object_type type,
1021 uintmax_t size,
1022 unsigned char *hdr)
1023 {
1024 int n = 1;
1025 unsigned char c;
1027 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
1028 die("bad type %d", type);
1030 c = (type << 4) | (size & 15);
1031 size >>= 4;
1032 while (size) {
1033 *hdr++ = c | 0x80;
1034 c = size & 0x7f;
1035 size >>= 7;
1036 n++;
1037 }
1038 *hdr = c;
1039 return n;
1040 }
1042 static int store_object(
1043 enum object_type type,
1044 struct strbuf *dat,
1045 struct last_object *last,
1046 unsigned char *sha1out,
1047 uintmax_t mark)
1048 {
1049 void *out, *delta;
1050 struct object_entry *e;
1051 unsigned char hdr[96];
1052 unsigned char sha1[20];
1053 unsigned long hdrlen, deltalen;
1054 git_SHA_CTX c;
1055 z_stream s;
1057 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1058 (unsigned long)dat->len) + 1;
1059 git_SHA1_Init(&c);
1060 git_SHA1_Update(&c, hdr, hdrlen);
1061 git_SHA1_Update(&c, dat->buf, dat->len);
1062 git_SHA1_Final(sha1, &c);
1063 if (sha1out)
1064 hashcpy(sha1out, sha1);
1066 e = insert_object(sha1);
1067 if (mark)
1068 insert_mark(mark, e);
1069 if (e->idx.offset) {
1070 duplicate_count_by_type[type]++;
1071 return 1;
1072 } else if (find_sha1_pack(sha1, packed_git)) {
1073 e->type = type;
1074 e->pack_id = MAX_PACK_ID;
1075 e->idx.offset = 1; /* just not zero! */
1076 duplicate_count_by_type[type]++;
1077 return 1;
1078 }
1080 if (last && last->data.buf && last->depth < max_depth) {
1081 delta = diff_delta(last->data.buf, last->data.len,
1082 dat->buf, dat->len,
1083 &deltalen, 0);
1084 if (delta && deltalen >= dat->len) {
1085 free(delta);
1086 delta = NULL;
1087 }
1088 } else
1089 delta = NULL;
1091 memset(&s, 0, sizeof(s));
1092 deflateInit(&s, pack_compression_level);
1093 if (delta) {
1094 s.next_in = delta;
1095 s.avail_in = deltalen;
1096 } else {
1097 s.next_in = (void *)dat->buf;
1098 s.avail_in = dat->len;
1099 }
1100 s.avail_out = deflateBound(&s, s.avail_in);
1101 s.next_out = out = xmalloc(s.avail_out);
1102 while (deflate(&s, Z_FINISH) == Z_OK)
1103 /* nothing */;
1104 deflateEnd(&s);
1106 /* Determine if we should auto-checkpoint. */
1107 if ((pack_size + 60 + s.total_out) > max_packsize
1108 || (pack_size + 60 + s.total_out) < pack_size) {
1110 /* This new object needs to *not* have the current pack_id. */
1111 e->pack_id = pack_id + 1;
1112 cycle_packfile();
1114 /* We cannot carry a delta into the new pack. */
1115 if (delta) {
1116 free(delta);
1117 delta = NULL;
1119 memset(&s, 0, sizeof(s));
1120 deflateInit(&s, pack_compression_level);
1121 s.next_in = (void *)dat->buf;
1122 s.avail_in = dat->len;
1123 s.avail_out = deflateBound(&s, s.avail_in);
1124 s.next_out = out = xrealloc(out, s.avail_out);
1125 while (deflate(&s, Z_FINISH) == Z_OK)
1126 /* nothing */;
1127 deflateEnd(&s);
1128 }
1129 }
1131 e->type = type;
1132 e->pack_id = pack_id;
1133 e->idx.offset = pack_size;
1134 object_count++;
1135 object_count_by_type[type]++;
1137 if (delta) {
1138 unsigned long ofs = e->idx.offset - last->offset;
1139 unsigned pos = sizeof(hdr) - 1;
1141 delta_count_by_type[type]++;
1142 e->depth = last->depth + 1;
1144 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1145 sha1write(pack_file, hdr, hdrlen);
1146 pack_size += hdrlen;
1148 hdr[pos] = ofs & 127;
1149 while (ofs >>= 7)
1150 hdr[--pos] = 128 | (--ofs & 127);
1151 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1152 pack_size += sizeof(hdr) - pos;
1153 } else {
1154 e->depth = 0;
1155 hdrlen = encode_header(type, dat->len, hdr);
1156 sha1write(pack_file, hdr, hdrlen);
1157 pack_size += hdrlen;
1158 }
1160 sha1write(pack_file, out, s.total_out);
1161 pack_size += s.total_out;
1163 free(out);
1164 free(delta);
1165 if (last) {
1166 if (last->no_swap) {
1167 last->data = *dat;
1168 } else {
1169 strbuf_swap(&last->data, dat);
1170 }
1171 last->offset = e->idx.offset;
1172 last->depth = e->depth;
1173 }
1174 return 0;
1175 }
1177 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1178 {
1179 if (ftruncate(pack_data->pack_fd, to)
1180 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1181 die_errno("cannot truncate pack to skip duplicate");
1182 pack_size = to;
1184 /* yes this is a layering violation */
1185 pack_file->total = to;
1186 pack_file->offset = 0;
1187 pack_file->ctx = *ctx;
1188 }
1190 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1191 {
1192 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1193 unsigned char *in_buf = xmalloc(in_sz);
1194 unsigned char *out_buf = xmalloc(out_sz);
1195 struct object_entry *e;
1196 unsigned char sha1[20];
1197 unsigned long hdrlen;
1198 off_t offset;
1199 git_SHA_CTX c;
1200 git_SHA_CTX pack_file_ctx;
1201 z_stream s;
1202 int status = Z_OK;
1204 /* Determine if we should auto-checkpoint. */
1205 if ((pack_size + 60 + len) > max_packsize
1206 || (pack_size + 60 + len) < pack_size)
1207 cycle_packfile();
1209 offset = pack_size;
1211 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1212 sha1flush(pack_file);
1213 pack_file_ctx = pack_file->ctx;
1215 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1216 if (out_sz <= hdrlen)
1217 die("impossibly large object header");
1219 git_SHA1_Init(&c);
1220 git_SHA1_Update(&c, out_buf, hdrlen);
1222 memset(&s, 0, sizeof(s));
1223 deflateInit(&s, pack_compression_level);
1225 hdrlen = encode_header(OBJ_BLOB, len, out_buf);
1226 if (out_sz <= hdrlen)
1227 die("impossibly large object header");
1229 s.next_out = out_buf + hdrlen;
1230 s.avail_out = out_sz - hdrlen;
1232 while (status != Z_STREAM_END) {
1233 if (0 < len && !s.avail_in) {
1234 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1235 size_t n = fread(in_buf, 1, cnt, stdin);
1236 if (!n && feof(stdin))
1237 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1239 git_SHA1_Update(&c, in_buf, n);
1240 s.next_in = in_buf;
1241 s.avail_in = n;
1242 len -= n;
1243 }
1245 status = deflate(&s, len ? 0 : Z_FINISH);
1247 if (!s.avail_out || status == Z_STREAM_END) {
1248 size_t n = s.next_out - out_buf;
1249 sha1write(pack_file, out_buf, n);
1250 pack_size += n;
1251 s.next_out = out_buf;
1252 s.avail_out = out_sz;
1253 }
1255 switch (status) {
1256 case Z_OK:
1257 case Z_BUF_ERROR:
1258 case Z_STREAM_END:
1259 continue;
1260 default:
1261 die("unexpected deflate failure: %d", status);
1262 }
1263 }
1264 deflateEnd(&s);
1265 git_SHA1_Final(sha1, &c);
1267 if (sha1out)
1268 hashcpy(sha1out, sha1);
1270 e = insert_object(sha1);
1272 if (mark)
1273 insert_mark(mark, e);
1275 if (e->idx.offset) {
1276 duplicate_count_by_type[OBJ_BLOB]++;
1277 truncate_pack(offset, &pack_file_ctx);
1279 } else if (find_sha1_pack(sha1, packed_git)) {
1280 e->type = OBJ_BLOB;
1281 e->pack_id = MAX_PACK_ID;
1282 e->idx.offset = 1; /* just not zero! */
1283 duplicate_count_by_type[OBJ_BLOB]++;
1284 truncate_pack(offset, &pack_file_ctx);
1286 } else {
1287 e->depth = 0;
1288 e->type = OBJ_BLOB;
1289 e->pack_id = pack_id;
1290 e->idx.offset = offset;
1291 object_count++;
1292 object_count_by_type[OBJ_BLOB]++;
1293 }
1295 free(in_buf);
1296 free(out_buf);
1297 }
1299 /* All calls must be guarded by find_object() or find_mark() to
1300 * ensure the 'struct object_entry' passed was written by this
1301 * process instance. We unpack the entry by the offset, avoiding
1302 * the need for the corresponding .idx file. This unpacking rule
1303 * works because we only use OBJ_REF_DELTA within the packfiles
1304 * created by fast-import.
1305 *
1306 * oe must not be NULL. Such an oe usually comes from giving
1307 * an unknown SHA-1 to find_object() or an undefined mark to
1308 * find_mark(). Callers must test for this condition and use
1309 * the standard read_sha1_file() when it happens.
1310 *
1311 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1312 * find_mark(), where the mark was reloaded from an existing marks
1313 * file and is referencing an object that this fast-import process
1314 * instance did not write out to a packfile. Callers must test for
1315 * this condition and use read_sha1_file() instead.
1316 */
1317 static void *gfi_unpack_entry(
1318 struct object_entry *oe,
1319 unsigned long *sizep)
1320 {
1321 enum object_type type;
1322 struct packed_git *p = all_packs[oe->pack_id];
1323 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1324 /* The object is stored in the packfile we are writing to
1325 * and we have modified it since the last time we scanned
1326 * back to read a previously written object. If an old
1327 * window covered [p->pack_size, p->pack_size + 20) its
1328 * data is stale and is not valid. Closing all windows
1329 * and updating the packfile length ensures we can read
1330 * the newly written data.
1331 */
1332 close_pack_windows(p);
1333 sha1flush(pack_file);
1335 /* We have to offer 20 bytes additional on the end of
1336 * the packfile as the core unpacker code assumes the
1337 * footer is present at the file end and must promise
1338 * at least 20 bytes within any window it maps. But
1339 * we don't actually create the footer here.
1340 */
1341 p->pack_size = pack_size + 20;
1342 }
1343 return unpack_entry(p, oe->idx.offset, &type, sizep);
1344 }
1346 static const char *get_mode(const char *str, uint16_t *modep)
1347 {
1348 unsigned char c;
1349 uint16_t mode = 0;
1351 while ((c = *str++) != ' ') {
1352 if (c < '0' || c > '7')
1353 return NULL;
1354 mode = (mode << 3) + (c - '0');
1355 }
1356 *modep = mode;
1357 return str;
1358 }
1360 static void load_tree(struct tree_entry *root)
1361 {
1362 unsigned char *sha1 = root->versions[1].sha1;
1363 struct object_entry *myoe;
1364 struct tree_content *t;
1365 unsigned long size;
1366 char *buf;
1367 const char *c;
1369 root->tree = t = new_tree_content(8);
1370 if (is_null_sha1(sha1))
1371 return;
1373 myoe = find_object(sha1);
1374 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1375 if (myoe->type != OBJ_TREE)
1376 die("Not a tree: %s", sha1_to_hex(sha1));
1377 t->delta_depth = myoe->depth;
1378 buf = gfi_unpack_entry(myoe, &size);
1379 if (!buf)
1380 die("Can't load tree %s", sha1_to_hex(sha1));
1381 } else {
1382 enum object_type type;
1383 buf = read_sha1_file(sha1, &type, &size);
1384 if (!buf || type != OBJ_TREE)
1385 die("Can't load tree %s", sha1_to_hex(sha1));
1386 }
1388 c = buf;
1389 while (c != (buf + size)) {
1390 struct tree_entry *e = new_tree_entry();
1392 if (t->entry_count == t->entry_capacity)
1393 root->tree = t = grow_tree_content(t, t->entry_count);
1394 t->entries[t->entry_count++] = e;
1396 e->tree = NULL;
1397 c = get_mode(c, &e->versions[1].mode);
1398 if (!c)
1399 die("Corrupt mode in %s", sha1_to_hex(sha1));
1400 e->versions[0].mode = e->versions[1].mode;
1401 e->name = to_atom(c, strlen(c));
1402 c += e->name->str_len + 1;
1403 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1404 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1405 c += 20;
1406 }
1407 free(buf);
1408 }
1410 static int tecmp0 (const void *_a, const void *_b)
1411 {
1412 struct tree_entry *a = *((struct tree_entry**)_a);
1413 struct tree_entry *b = *((struct tree_entry**)_b);
1414 return base_name_compare(
1415 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1416 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1417 }
1419 static int tecmp1 (const void *_a, const void *_b)
1420 {
1421 struct tree_entry *a = *((struct tree_entry**)_a);
1422 struct tree_entry *b = *((struct tree_entry**)_b);
1423 return base_name_compare(
1424 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1425 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1426 }
1428 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1429 {
1430 size_t maxlen = 0;
1431 unsigned int i;
1433 if (!v)
1434 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1435 else
1436 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1438 for (i = 0; i < t->entry_count; i++) {
1439 if (t->entries[i]->versions[v].mode)
1440 maxlen += t->entries[i]->name->str_len + 34;
1441 }
1443 strbuf_reset(b);
1444 strbuf_grow(b, maxlen);
1445 for (i = 0; i < t->entry_count; i++) {
1446 struct tree_entry *e = t->entries[i];
1447 if (!e->versions[v].mode)
1448 continue;
1449 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1450 e->name->str_dat, '\0');
1451 strbuf_add(b, e->versions[v].sha1, 20);
1452 }
1453 }
1455 static void store_tree(struct tree_entry *root)
1456 {
1457 struct tree_content *t = root->tree;
1458 unsigned int i, j, del;
1459 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1460 struct object_entry *le;
1462 if (!is_null_sha1(root->versions[1].sha1))
1463 return;
1465 for (i = 0; i < t->entry_count; i++) {
1466 if (t->entries[i]->tree)
1467 store_tree(t->entries[i]);
1468 }
1470 le = find_object(root->versions[0].sha1);
1471 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1472 mktree(t, 0, &old_tree);
1473 lo.data = old_tree;
1474 lo.offset = le->idx.offset;
1475 lo.depth = t->delta_depth;
1476 }
1478 mktree(t, 1, &new_tree);
1479 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1481 t->delta_depth = lo.depth;
1482 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1483 struct tree_entry *e = t->entries[i];
1484 if (e->versions[1].mode) {
1485 e->versions[0].mode = e->versions[1].mode;
1486 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1487 t->entries[j++] = e;
1488 } else {
1489 release_tree_entry(e);
1490 del++;
1491 }
1492 }
1493 t->entry_count -= del;
1494 }
1496 static int tree_content_set(
1497 struct tree_entry *root,
1498 const char *p,
1499 const unsigned char *sha1,
1500 const uint16_t mode,
1501 struct tree_content *subtree)
1502 {
1503 struct tree_content *t = root->tree;
1504 const char *slash1;
1505 unsigned int i, n;
1506 struct tree_entry *e;
1508 slash1 = strchr(p, '/');
1509 if (slash1)
1510 n = slash1 - p;
1511 else
1512 n = strlen(p);
1513 if (!n)
1514 die("Empty path component found in input");
1515 if (!slash1 && !S_ISDIR(mode) && subtree)
1516 die("Non-directories cannot have subtrees");
1518 for (i = 0; i < t->entry_count; i++) {
1519 e = t->entries[i];
1520 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1521 if (!slash1) {
1522 if (!S_ISDIR(mode)
1523 && e->versions[1].mode == mode
1524 && !hashcmp(e->versions[1].sha1, sha1))
1525 return 0;
1526 e->versions[1].mode = mode;
1527 hashcpy(e->versions[1].sha1, sha1);
1528 if (e->tree)
1529 release_tree_content_recursive(e->tree);
1530 e->tree = subtree;
1531 hashclr(root->versions[1].sha1);
1532 return 1;
1533 }
1534 if (!S_ISDIR(e->versions[1].mode)) {
1535 e->tree = new_tree_content(8);
1536 e->versions[1].mode = S_IFDIR;
1537 }
1538 if (!e->tree)
1539 load_tree(e);
1540 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1541 hashclr(root->versions[1].sha1);
1542 return 1;
1543 }
1544 return 0;
1545 }
1546 }
1548 if (t->entry_count == t->entry_capacity)
1549 root->tree = t = grow_tree_content(t, t->entry_count);
1550 e = new_tree_entry();
1551 e->name = to_atom(p, n);
1552 e->versions[0].mode = 0;
1553 hashclr(e->versions[0].sha1);
1554 t->entries[t->entry_count++] = e;
1555 if (slash1) {
1556 e->tree = new_tree_content(8);
1557 e->versions[1].mode = S_IFDIR;
1558 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1559 } else {
1560 e->tree = subtree;
1561 e->versions[1].mode = mode;
1562 hashcpy(e->versions[1].sha1, sha1);
1563 }
1564 hashclr(root->versions[1].sha1);
1565 return 1;
1566 }
1568 static int tree_content_remove(
1569 struct tree_entry *root,
1570 const char *p,
1571 struct tree_entry *backup_leaf)
1572 {
1573 struct tree_content *t = root->tree;
1574 const char *slash1;
1575 unsigned int i, n;
1576 struct tree_entry *e;
1578 slash1 = strchr(p, '/');
1579 if (slash1)
1580 n = slash1 - p;
1581 else
1582 n = strlen(p);
1584 for (i = 0; i < t->entry_count; i++) {
1585 e = t->entries[i];
1586 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1587 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1588 goto del_entry;
1589 if (!e->tree)
1590 load_tree(e);
1591 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1592 for (n = 0; n < e->tree->entry_count; n++) {
1593 if (e->tree->entries[n]->versions[1].mode) {
1594 hashclr(root->versions[1].sha1);
1595 return 1;
1596 }
1597 }
1598 backup_leaf = NULL;
1599 goto del_entry;
1600 }
1601 return 0;
1602 }
1603 }
1604 return 0;
1606 del_entry:
1607 if (backup_leaf)
1608 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1609 else if (e->tree)
1610 release_tree_content_recursive(e->tree);
1611 e->tree = NULL;
1612 e->versions[1].mode = 0;
1613 hashclr(e->versions[1].sha1);
1614 hashclr(root->versions[1].sha1);
1615 return 1;
1616 }
1618 static int tree_content_get(
1619 struct tree_entry *root,
1620 const char *p,
1621 struct tree_entry *leaf)
1622 {
1623 struct tree_content *t = root->tree;
1624 const char *slash1;
1625 unsigned int i, n;
1626 struct tree_entry *e;
1628 slash1 = strchr(p, '/');
1629 if (slash1)
1630 n = slash1 - p;
1631 else
1632 n = strlen(p);
1634 for (i = 0; i < t->entry_count; i++) {
1635 e = t->entries[i];
1636 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1637 if (!slash1) {
1638 memcpy(leaf, e, sizeof(*leaf));
1639 if (e->tree && is_null_sha1(e->versions[1].sha1))
1640 leaf->tree = dup_tree_content(e->tree);
1641 else
1642 leaf->tree = NULL;
1643 return 1;
1644 }
1645 if (!S_ISDIR(e->versions[1].mode))
1646 return 0;
1647 if (!e->tree)
1648 load_tree(e);
1649 return tree_content_get(e, slash1 + 1, leaf);
1650 }
1651 }
1652 return 0;
1653 }
1655 static int update_branch(struct branch *b)
1656 {
1657 static const char *msg = "fast-import";
1658 struct ref_lock *lock;
1659 unsigned char old_sha1[20];
1661 if (is_null_sha1(b->sha1))
1662 return 0;
1663 if (read_ref(b->name, old_sha1))
1664 hashclr(old_sha1);
1665 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1666 if (!lock)
1667 return error("Unable to lock %s", b->name);
1668 if (!force_update && !is_null_sha1(old_sha1)) {
1669 struct commit *old_cmit, *new_cmit;
1671 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1672 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1673 if (!old_cmit || !new_cmit) {
1674 unlock_ref(lock);
1675 return error("Branch %s is missing commits.", b->name);
1676 }
1678 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1679 unlock_ref(lock);
1680 warning("Not updating %s"
1681 " (new tip %s does not contain %s)",
1682 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1683 return -1;
1684 }
1685 }
1686 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1687 return error("Unable to update %s", b->name);
1688 return 0;
1689 }
1691 static void dump_branches(void)
1692 {
1693 unsigned int i;
1694 struct branch *b;
1696 for (i = 0; i < branch_table_sz; i++) {
1697 for (b = branch_table[i]; b; b = b->table_next_branch)
1698 failure |= update_branch(b);
1699 }
1700 }
1702 static void dump_tags(void)
1703 {
1704 static const char *msg = "fast-import";
1705 struct tag *t;
1706 struct ref_lock *lock;
1707 char ref_name[PATH_MAX];
1709 for (t = first_tag; t; t = t->next_tag) {
1710 sprintf(ref_name, "tags/%s", t->name);
1711 lock = lock_ref_sha1(ref_name, NULL);
1712 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1713 failure |= error("Unable to update %s", ref_name);
1714 }
1715 }
1717 static void dump_marks_helper(FILE *f,
1718 uintmax_t base,
1719 struct mark_set *m)
1720 {
1721 uintmax_t k;
1722 if (m->shift) {
1723 for (k = 0; k < 1024; k++) {
1724 if (m->data.sets[k])
1725 dump_marks_helper(f, (base + k) << m->shift,
1726 m->data.sets[k]);
1727 }
1728 } else {
1729 for (k = 0; k < 1024; k++) {
1730 if (m->data.marked[k])
1731 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1732 sha1_to_hex(m->data.marked[k]->idx.sha1));
1733 }
1734 }
1735 }
1737 static void dump_marks(void)
1738 {
1739 static struct lock_file mark_lock;
1740 int mark_fd;
1741 FILE *f;
1743 if (!export_marks_file)
1744 return;
1746 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1747 if (mark_fd < 0) {
1748 failure |= error("Unable to write marks file %s: %s",
1749 export_marks_file, strerror(errno));
1750 return;
1751 }
1753 f = fdopen(mark_fd, "w");
1754 if (!f) {
1755 int saved_errno = errno;
1756 rollback_lock_file(&mark_lock);
1757 failure |= error("Unable to write marks file %s: %s",
1758 export_marks_file, strerror(saved_errno));
1759 return;
1760 }
1762 /*
1763 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1764 * Assign -1 to the lock file descriptor so that commit_lock_file()
1765 * won't try to close() it.
1766 */
1767 mark_lock.fd = -1;
1769 dump_marks_helper(f, 0, marks);
1770 if (ferror(f) || fclose(f)) {
1771 int saved_errno = errno;
1772 rollback_lock_file(&mark_lock);
1773 failure |= error("Unable to write marks file %s: %s",
1774 export_marks_file, strerror(saved_errno));
1775 return;
1776 }
1778 if (commit_lock_file(&mark_lock)) {
1779 int saved_errno = errno;
1780 rollback_lock_file(&mark_lock);
1781 failure |= error("Unable to commit marks file %s: %s",
1782 export_marks_file, strerror(saved_errno));
1783 return;
1784 }
1785 }
1787 static void read_marks(void)
1788 {
1789 char line[512];
1790 FILE *f = fopen(import_marks_file, "r");
1791 if (!f)
1792 die_errno("cannot read '%s'", import_marks_file);
1793 while (fgets(line, sizeof(line), f)) {
1794 uintmax_t mark;
1795 char *end;
1796 unsigned char sha1[20];
1797 struct object_entry *e;
1799 end = strchr(line, '\n');
1800 if (line[0] != ':' || !end)
1801 die("corrupt mark line: %s", line);
1802 *end = 0;
1803 mark = strtoumax(line + 1, &end, 10);
1804 if (!mark || end == line + 1
1805 || *end != ' ' || get_sha1(end + 1, sha1))
1806 die("corrupt mark line: %s", line);
1807 e = find_object(sha1);
1808 if (!e) {
1809 enum object_type type = sha1_object_info(sha1, NULL);
1810 if (type < 0)
1811 die("object not found: %s", sha1_to_hex(sha1));
1812 e = insert_object(sha1);
1813 e->type = type;
1814 e->pack_id = MAX_PACK_ID;
1815 e->idx.offset = 1; /* just not zero! */
1816 }
1817 insert_mark(mark, e);
1818 }
1819 fclose(f);
1820 }
1823 static int read_next_command(void)
1824 {
1825 static int stdin_eof = 0;
1827 if (stdin_eof) {
1828 unread_command_buf = 0;
1829 return EOF;
1830 }
1832 do {
1833 if (unread_command_buf) {
1834 unread_command_buf = 0;
1835 } else {
1836 struct recent_command *rc;
1838 strbuf_detach(&command_buf, NULL);
1839 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1840 if (stdin_eof)
1841 return EOF;
1843 if (!seen_data_command
1844 && prefixcmp(command_buf.buf, "feature ")
1845 && prefixcmp(command_buf.buf, "option ")) {
1846 parse_argv();
1847 }
1849 rc = rc_free;
1850 if (rc)
1851 rc_free = rc->next;
1852 else {
1853 rc = cmd_hist.next;
1854 cmd_hist.next = rc->next;
1855 cmd_hist.next->prev = &cmd_hist;
1856 free(rc->buf);
1857 }
1859 rc->buf = command_buf.buf;
1860 rc->prev = cmd_tail;
1861 rc->next = cmd_hist.prev;
1862 rc->prev->next = rc;
1863 cmd_tail = rc;
1864 }
1865 } while (command_buf.buf[0] == '#');
1867 return 0;
1868 }
1870 static void skip_optional_lf(void)
1871 {
1872 int term_char = fgetc(stdin);
1873 if (term_char != '\n' && term_char != EOF)
1874 ungetc(term_char, stdin);
1875 }
1877 static void parse_mark(void)
1878 {
1879 if (!prefixcmp(command_buf.buf, "mark :")) {
1880 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1881 read_next_command();
1882 }
1883 else
1884 next_mark = 0;
1885 }
1887 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1888 {
1889 strbuf_reset(sb);
1891 if (prefixcmp(command_buf.buf, "data "))
1892 die("Expected 'data n' command, found: %s", command_buf.buf);
1894 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1895 char *term = xstrdup(command_buf.buf + 5 + 2);
1896 size_t term_len = command_buf.len - 5 - 2;
1898 strbuf_detach(&command_buf, NULL);
1899 for (;;) {
1900 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1901 die("EOF in data (terminator '%s' not found)", term);
1902 if (term_len == command_buf.len
1903 && !strcmp(term, command_buf.buf))
1904 break;
1905 strbuf_addbuf(sb, &command_buf);
1906 strbuf_addch(sb, '\n');
1907 }
1908 free(term);
1909 }
1910 else {
1911 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1912 size_t n = 0, length = (size_t)len;
1914 if (limit && limit < len) {
1915 *len_res = len;
1916 return 0;
1917 }
1918 if (length < len)
1919 die("data is too large to use in this context");
1921 while (n < length) {
1922 size_t s = strbuf_fread(sb, length - n, stdin);
1923 if (!s && feof(stdin))
1924 die("EOF in data (%lu bytes remaining)",
1925 (unsigned long)(length - n));
1926 n += s;
1927 }
1928 }
1930 skip_optional_lf();
1931 return 1;
1932 }
1934 static int validate_raw_date(const char *src, char *result, int maxlen)
1935 {
1936 const char *orig_src = src;
1937 char *endp;
1938 unsigned long num;
1940 errno = 0;
1942 num = strtoul(src, &endp, 10);
1943 /* NEEDSWORK: perhaps check for reasonable values? */
1944 if (errno || endp == src || *endp != ' ')
1945 return -1;
1947 src = endp + 1;
1948 if (*src != '-' && *src != '+')
1949 return -1;
1951 num = strtoul(src + 1, &endp, 10);
1952 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1953 1400 < num)
1954 return -1;
1956 strcpy(result, orig_src);
1957 return 0;
1958 }
1960 static char *parse_ident(const char *buf)
1961 {
1962 const char *gt;
1963 size_t name_len;
1964 char *ident;
1966 gt = strrchr(buf, '>');
1967 if (!gt)
1968 die("Missing > in ident string: %s", buf);
1969 gt++;
1970 if (*gt != ' ')
1971 die("Missing space after > in ident string: %s", buf);
1972 gt++;
1973 name_len = gt - buf;
1974 ident = xmalloc(name_len + 24);
1975 strncpy(ident, buf, name_len);
1977 switch (whenspec) {
1978 case WHENSPEC_RAW:
1979 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1980 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1981 break;
1982 case WHENSPEC_RFC2822:
1983 if (parse_date(gt, ident + name_len, 24) < 0)
1984 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1985 break;
1986 case WHENSPEC_NOW:
1987 if (strcmp("now", gt))
1988 die("Date in ident must be 'now': %s", buf);
1989 datestamp(ident + name_len, 24);
1990 break;
1991 }
1993 return ident;
1994 }
1996 static void parse_and_store_blob(
1997 struct last_object *last,
1998 unsigned char *sha1out,
1999 uintmax_t mark)
2000 {
2001 static struct strbuf buf = STRBUF_INIT;
2002 uintmax_t len;
2004 if (parse_data(&buf, big_file_threshold, &len))
2005 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2006 else {
2007 if (last) {
2008 strbuf_release(&last->data);
2009 last->offset = 0;
2010 last->depth = 0;
2011 }
2012 stream_blob(len, sha1out, mark);
2013 skip_optional_lf();
2014 }
2015 }
2017 static void parse_new_blob(void)
2018 {
2019 read_next_command();
2020 parse_mark();
2021 parse_and_store_blob(&last_blob, NULL, next_mark);
2022 }
2024 static void unload_one_branch(void)
2025 {
2026 while (cur_active_branches
2027 && cur_active_branches >= max_active_branches) {
2028 uintmax_t min_commit = ULONG_MAX;
2029 struct branch *e, *l = NULL, *p = NULL;
2031 for (e = active_branches; e; e = e->active_next_branch) {
2032 if (e->last_commit < min_commit) {
2033 p = l;
2034 min_commit = e->last_commit;
2035 }
2036 l = e;
2037 }
2039 if (p) {
2040 e = p->active_next_branch;
2041 p->active_next_branch = e->active_next_branch;
2042 } else {
2043 e = active_branches;
2044 active_branches = e->active_next_branch;
2045 }
2046 e->active = 0;
2047 e->active_next_branch = NULL;
2048 if (e->branch_tree.tree) {
2049 release_tree_content_recursive(e->branch_tree.tree);
2050 e->branch_tree.tree = NULL;
2051 }
2052 cur_active_branches--;
2053 }
2054 }
2056 static void load_branch(struct branch *b)
2057 {
2058 load_tree(&b->branch_tree);
2059 if (!b->active) {
2060 b->active = 1;
2061 b->active_next_branch = active_branches;
2062 active_branches = b;
2063 cur_active_branches++;
2064 branch_load_count++;
2065 }
2066 }
2068 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2069 {
2070 unsigned char fanout = 0;
2071 while ((num_notes >>= 8))
2072 fanout++;
2073 return fanout;
2074 }
2076 static void construct_path_with_fanout(const char *hex_sha1,
2077 unsigned char fanout, char *path)
2078 {
2079 unsigned int i = 0, j = 0;
2080 if (fanout >= 20)
2081 die("Too large fanout (%u)", fanout);
2082 while (fanout) {
2083 path[i++] = hex_sha1[j++];
2084 path[i++] = hex_sha1[j++];
2085 path[i++] = '/';
2086 fanout--;
2087 }
2088 memcpy(path + i, hex_sha1 + j, 40 - j);
2089 path[i + 40 - j] = '\0';
2090 }
2092 static uintmax_t do_change_note_fanout(
2093 struct tree_entry *orig_root, struct tree_entry *root,
2094 char *hex_sha1, unsigned int hex_sha1_len,
2095 char *fullpath, unsigned int fullpath_len,
2096 unsigned char fanout)
2097 {
2098 struct tree_content *t = root->tree;
2099 struct tree_entry *e, leaf;
2100 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2101 uintmax_t num_notes = 0;
2102 unsigned char sha1[20];
2103 char realpath[60];
2105 for (i = 0; t && i < t->entry_count; i++) {
2106 e = t->entries[i];
2107 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2108 tmp_fullpath_len = fullpath_len;
2110 /*
2111 * We're interested in EITHER existing note entries (entries
2112 * with exactly 40 hex chars in path, not including directory
2113 * separators), OR directory entries that may contain note
2114 * entries (with < 40 hex chars in path).
2115 * Also, each path component in a note entry must be a multiple
2116 * of 2 chars.
2117 */
2118 if (!e->versions[1].mode ||
2119 tmp_hex_sha1_len > 40 ||
2120 e->name->str_len % 2)
2121 continue;
2123 /* This _may_ be a note entry, or a subdir containing notes */
2124 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2125 e->name->str_len);
2126 if (tmp_fullpath_len)
2127 fullpath[tmp_fullpath_len++] = '/';
2128 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2129 e->name->str_len);
2130 tmp_fullpath_len += e->name->str_len;
2131 fullpath[tmp_fullpath_len] = '\0';
2133 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2134 /* This is a note entry */
2135 construct_path_with_fanout(hex_sha1, fanout, realpath);
2136 if (!strcmp(fullpath, realpath)) {
2137 /* Note entry is in correct location */
2138 num_notes++;
2139 continue;
2140 }
2142 /* Rename fullpath to realpath */
2143 if (!tree_content_remove(orig_root, fullpath, &leaf))
2144 die("Failed to remove path %s", fullpath);
2145 tree_content_set(orig_root, realpath,
2146 leaf.versions[1].sha1,
2147 leaf.versions[1].mode,
2148 leaf.tree);
2149 } else if (S_ISDIR(e->versions[1].mode)) {
2150 /* This is a subdir that may contain note entries */
2151 if (!e->tree)
2152 load_tree(e);
2153 num_notes += do_change_note_fanout(orig_root, e,
2154 hex_sha1, tmp_hex_sha1_len,
2155 fullpath, tmp_fullpath_len, fanout);
2156 }
2158 /* The above may have reallocated the current tree_content */
2159 t = root->tree;
2160 }
2161 return num_notes;
2162 }
2164 static uintmax_t change_note_fanout(struct tree_entry *root,
2165 unsigned char fanout)
2166 {
2167 char hex_sha1[40], path[60];
2168 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2169 }
2171 static void file_change_m(struct branch *b)
2172 {
2173 const char *p = command_buf.buf + 2;
2174 static struct strbuf uq = STRBUF_INIT;
2175 const char *endp;
2176 struct object_entry *oe = oe;
2177 unsigned char sha1[20];
2178 uint16_t mode, inline_data = 0;
2180 p = get_mode(p, &mode);
2181 if (!p)
2182 die("Corrupt mode: %s", command_buf.buf);
2183 switch (mode) {
2184 case 0644:
2185 case 0755:
2186 mode |= S_IFREG;
2187 case S_IFREG | 0644:
2188 case S_IFREG | 0755:
2189 case S_IFLNK:
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 if (S_ISGITLINK(mode)) {
2222 if (inline_data)
2223 die("Git links cannot be specified 'inline': %s",
2224 command_buf.buf);
2225 else if (oe) {
2226 if (oe->type != OBJ_COMMIT)
2227 die("Not a commit (actually a %s): %s",
2228 typename(oe->type), command_buf.buf);
2229 }
2230 /*
2231 * Accept the sha1 without checking; it expected to be in
2232 * another repository.
2233 */
2234 } else if (inline_data) {
2235 if (p != uq.buf) {
2236 strbuf_addstr(&uq, p);
2237 p = uq.buf;
2238 }
2239 read_next_command();
2240 parse_and_store_blob(&last_blob, sha1, 0);
2241 } else if (oe) {
2242 if (oe->type != OBJ_BLOB)
2243 die("Not a blob (actually a %s): %s",
2244 typename(oe->type), command_buf.buf);
2245 } else {
2246 enum object_type type = sha1_object_info(sha1, NULL);
2247 if (type < 0)
2248 die("Blob not found: %s", command_buf.buf);
2249 if (type != OBJ_BLOB)
2250 die("Not a blob (actually a %s): %s",
2251 typename(type), command_buf.buf);
2252 }
2254 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2255 }
2257 static void file_change_d(struct branch *b)
2258 {
2259 const char *p = command_buf.buf + 2;
2260 static struct strbuf uq = STRBUF_INIT;
2261 const char *endp;
2263 strbuf_reset(&uq);
2264 if (!unquote_c_style(&uq, p, &endp)) {
2265 if (*endp)
2266 die("Garbage after path in: %s", command_buf.buf);
2267 p = uq.buf;
2268 }
2269 tree_content_remove(&b->branch_tree, p, NULL);
2270 }
2272 static void file_change_cr(struct branch *b, int rename)
2273 {
2274 const char *s, *d;
2275 static struct strbuf s_uq = STRBUF_INIT;
2276 static struct strbuf d_uq = STRBUF_INIT;
2277 const char *endp;
2278 struct tree_entry leaf;
2280 s = command_buf.buf + 2;
2281 strbuf_reset(&s_uq);
2282 if (!unquote_c_style(&s_uq, s, &endp)) {
2283 if (*endp != ' ')
2284 die("Missing space after source: %s", command_buf.buf);
2285 } else {
2286 endp = strchr(s, ' ');
2287 if (!endp)
2288 die("Missing space after source: %s", command_buf.buf);
2289 strbuf_add(&s_uq, s, endp - s);
2290 }
2291 s = s_uq.buf;
2293 endp++;
2294 if (!*endp)
2295 die("Missing dest: %s", command_buf.buf);
2297 d = endp;
2298 strbuf_reset(&d_uq);
2299 if (!unquote_c_style(&d_uq, d, &endp)) {
2300 if (*endp)
2301 die("Garbage after dest in: %s", command_buf.buf);
2302 d = d_uq.buf;
2303 }
2305 memset(&leaf, 0, sizeof(leaf));
2306 if (rename)
2307 tree_content_remove(&b->branch_tree, s, &leaf);
2308 else
2309 tree_content_get(&b->branch_tree, s, &leaf);
2310 if (!leaf.versions[1].mode)
2311 die("Path %s not in branch", s);
2312 tree_content_set(&b->branch_tree, d,
2313 leaf.versions[1].sha1,
2314 leaf.versions[1].mode,
2315 leaf.tree);
2316 }
2318 static void note_change_n(struct branch *b, unsigned char old_fanout)
2319 {
2320 const char *p = command_buf.buf + 2;
2321 static struct strbuf uq = STRBUF_INIT;
2322 struct object_entry *oe = oe;
2323 struct branch *s;
2324 unsigned char sha1[20], commit_sha1[20];
2325 char path[60];
2326 uint16_t inline_data = 0;
2327 unsigned char new_fanout;
2329 /* <dataref> or 'inline' */
2330 if (*p == ':') {
2331 char *x;
2332 oe = find_mark(strtoumax(p + 1, &x, 10));
2333 hashcpy(sha1, oe->idx.sha1);
2334 p = x;
2335 } else if (!prefixcmp(p, "inline")) {
2336 inline_data = 1;
2337 p += 6;
2338 } else {
2339 if (get_sha1_hex(p, sha1))
2340 die("Invalid SHA1: %s", command_buf.buf);
2341 oe = find_object(sha1);
2342 p += 40;
2343 }
2344 if (*p++ != ' ')
2345 die("Missing space after SHA1: %s", command_buf.buf);
2347 /* <committish> */
2348 s = lookup_branch(p);
2349 if (s) {
2350 hashcpy(commit_sha1, s->sha1);
2351 } else if (*p == ':') {
2352 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2353 struct object_entry *commit_oe = find_mark(commit_mark);
2354 if (commit_oe->type != OBJ_COMMIT)
2355 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2356 hashcpy(commit_sha1, commit_oe->idx.sha1);
2357 } else if (!get_sha1(p, commit_sha1)) {
2358 unsigned long size;
2359 char *buf = read_object_with_reference(commit_sha1,
2360 commit_type, &size, commit_sha1);
2361 if (!buf || size < 46)
2362 die("Not a valid commit: %s", p);
2363 free(buf);
2364 } else
2365 die("Invalid ref name or SHA1 expression: %s", p);
2367 if (inline_data) {
2368 if (p != uq.buf) {
2369 strbuf_addstr(&uq, p);
2370 p = uq.buf;
2371 }
2372 read_next_command();
2373 parse_and_store_blob(&last_blob, sha1, 0);
2374 } else if (oe) {
2375 if (oe->type != OBJ_BLOB)
2376 die("Not a blob (actually a %s): %s",
2377 typename(oe->type), command_buf.buf);
2378 } else if (!is_null_sha1(sha1)) {
2379 enum object_type type = sha1_object_info(sha1, NULL);
2380 if (type < 0)
2381 die("Blob not found: %s", command_buf.buf);
2382 if (type != OBJ_BLOB)
2383 die("Not a blob (actually a %s): %s",
2384 typename(type), command_buf.buf);
2385 }
2387 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2388 if (tree_content_remove(&b->branch_tree, path, NULL))
2389 b->num_notes--;
2391 if (is_null_sha1(sha1))
2392 return; /* nothing to insert */
2394 b->num_notes++;
2395 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2396 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2397 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2398 }
2400 static void file_change_deleteall(struct branch *b)
2401 {
2402 release_tree_content_recursive(b->branch_tree.tree);
2403 hashclr(b->branch_tree.versions[0].sha1);
2404 hashclr(b->branch_tree.versions[1].sha1);
2405 load_tree(&b->branch_tree);
2406 b->num_notes = 0;
2407 }
2409 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2410 {
2411 if (!buf || size < 46)
2412 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2413 if (memcmp("tree ", buf, 5)
2414 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2415 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2416 hashcpy(b->branch_tree.versions[0].sha1,
2417 b->branch_tree.versions[1].sha1);
2418 }
2420 static void parse_from_existing(struct branch *b)
2421 {
2422 if (is_null_sha1(b->sha1)) {
2423 hashclr(b->branch_tree.versions[0].sha1);
2424 hashclr(b->branch_tree.versions[1].sha1);
2425 } else {
2426 unsigned long size;
2427 char *buf;
2429 buf = read_object_with_reference(b->sha1,
2430 commit_type, &size, b->sha1);
2431 parse_from_commit(b, buf, size);
2432 free(buf);
2433 }
2434 }
2436 static int parse_from(struct branch *b)
2437 {
2438 const char *from;
2439 struct branch *s;
2441 if (prefixcmp(command_buf.buf, "from "))
2442 return 0;
2444 if (b->branch_tree.tree) {
2445 release_tree_content_recursive(b->branch_tree.tree);
2446 b->branch_tree.tree = NULL;
2447 }
2449 from = strchr(command_buf.buf, ' ') + 1;
2450 s = lookup_branch(from);
2451 if (b == s)
2452 die("Can't create a branch from itself: %s", b->name);
2453 else if (s) {
2454 unsigned char *t = s->branch_tree.versions[1].sha1;
2455 hashcpy(b->sha1, s->sha1);
2456 hashcpy(b->branch_tree.versions[0].sha1, t);
2457 hashcpy(b->branch_tree.versions[1].sha1, t);
2458 } else if (*from == ':') {
2459 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2460 struct object_entry *oe = find_mark(idnum);
2461 if (oe->type != OBJ_COMMIT)
2462 die("Mark :%" PRIuMAX " not a commit", idnum);
2463 hashcpy(b->sha1, oe->idx.sha1);
2464 if (oe->pack_id != MAX_PACK_ID) {
2465 unsigned long size;
2466 char *buf = gfi_unpack_entry(oe, &size);
2467 parse_from_commit(b, buf, size);
2468 free(buf);
2469 } else
2470 parse_from_existing(b);
2471 } else if (!get_sha1(from, b->sha1))
2472 parse_from_existing(b);
2473 else
2474 die("Invalid ref name or SHA1 expression: %s", from);
2476 read_next_command();
2477 return 1;
2478 }
2480 static struct hash_list *parse_merge(unsigned int *count)
2481 {
2482 struct hash_list *list = NULL, *n, *e = e;
2483 const char *from;
2484 struct branch *s;
2486 *count = 0;
2487 while (!prefixcmp(command_buf.buf, "merge ")) {
2488 from = strchr(command_buf.buf, ' ') + 1;
2489 n = xmalloc(sizeof(*n));
2490 s = lookup_branch(from);
2491 if (s)
2492 hashcpy(n->sha1, s->sha1);
2493 else if (*from == ':') {
2494 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2495 struct object_entry *oe = find_mark(idnum);
2496 if (oe->type != OBJ_COMMIT)
2497 die("Mark :%" PRIuMAX " not a commit", idnum);
2498 hashcpy(n->sha1, oe->idx.sha1);
2499 } else if (!get_sha1(from, n->sha1)) {
2500 unsigned long size;
2501 char *buf = read_object_with_reference(n->sha1,
2502 commit_type, &size, n->sha1);
2503 if (!buf || size < 46)
2504 die("Not a valid commit: %s", from);
2505 free(buf);
2506 } else
2507 die("Invalid ref name or SHA1 expression: %s", from);
2509 n->next = NULL;
2510 if (list)
2511 e->next = n;
2512 else
2513 list = n;
2514 e = n;
2515 (*count)++;
2516 read_next_command();
2517 }
2518 return list;
2519 }
2521 static void parse_new_commit(void)
2522 {
2523 static struct strbuf msg = STRBUF_INIT;
2524 struct branch *b;
2525 char *sp;
2526 char *author = NULL;
2527 char *committer = NULL;
2528 struct hash_list *merge_list = NULL;
2529 unsigned int merge_count;
2530 unsigned char prev_fanout, new_fanout;
2532 /* Obtain the branch name from the rest of our command */
2533 sp = strchr(command_buf.buf, ' ') + 1;
2534 b = lookup_branch(sp);
2535 if (!b)
2536 b = new_branch(sp);
2538 read_next_command();
2539 parse_mark();
2540 if (!prefixcmp(command_buf.buf, "author ")) {
2541 author = parse_ident(command_buf.buf + 7);
2542 read_next_command();
2543 }
2544 if (!prefixcmp(command_buf.buf, "committer ")) {
2545 committer = parse_ident(command_buf.buf + 10);
2546 read_next_command();
2547 }
2548 if (!committer)
2549 die("Expected committer but didn't get one");
2550 parse_data(&msg, 0, NULL);
2551 read_next_command();
2552 parse_from(b);
2553 merge_list = parse_merge(&merge_count);
2555 /* ensure the branch is active/loaded */
2556 if (!b->branch_tree.tree || !max_active_branches) {
2557 unload_one_branch();
2558 load_branch(b);
2559 }
2561 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2563 /* file_change* */
2564 while (command_buf.len > 0) {
2565 if (!prefixcmp(command_buf.buf, "M "))
2566 file_change_m(b);
2567 else if (!prefixcmp(command_buf.buf, "D "))
2568 file_change_d(b);
2569 else if (!prefixcmp(command_buf.buf, "R "))
2570 file_change_cr(b, 1);
2571 else if (!prefixcmp(command_buf.buf, "C "))
2572 file_change_cr(b, 0);
2573 else if (!prefixcmp(command_buf.buf, "N "))
2574 note_change_n(b, prev_fanout);
2575 else if (!strcmp("deleteall", command_buf.buf))
2576 file_change_deleteall(b);
2577 else {
2578 unread_command_buf = 1;
2579 break;
2580 }
2581 if (read_next_command() == EOF)
2582 break;
2583 }
2585 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2586 if (new_fanout != prev_fanout)
2587 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2589 /* build the tree and the commit */
2590 store_tree(&b->branch_tree);
2591 hashcpy(b->branch_tree.versions[0].sha1,
2592 b->branch_tree.versions[1].sha1);
2594 strbuf_reset(&new_data);
2595 strbuf_addf(&new_data, "tree %s\n",
2596 sha1_to_hex(b->branch_tree.versions[1].sha1));
2597 if (!is_null_sha1(b->sha1))
2598 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2599 while (merge_list) {
2600 struct hash_list *next = merge_list->next;
2601 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2602 free(merge_list);
2603 merge_list = next;
2604 }
2605 strbuf_addf(&new_data,
2606 "author %s\n"
2607 "committer %s\n"
2608 "\n",
2609 author ? author : committer, committer);
2610 strbuf_addbuf(&new_data, &msg);
2611 free(author);
2612 free(committer);
2614 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2615 b->pack_id = pack_id;
2616 b->last_commit = object_count_by_type[OBJ_COMMIT];
2617 }
2619 static void parse_new_tag(void)
2620 {
2621 static struct strbuf msg = STRBUF_INIT;
2622 char *sp;
2623 const char *from;
2624 char *tagger;
2625 struct branch *s;
2626 struct tag *t;
2627 uintmax_t from_mark = 0;
2628 unsigned char sha1[20];
2629 enum object_type type;
2631 /* Obtain the new tag name from the rest of our command */
2632 sp = strchr(command_buf.buf, ' ') + 1;
2633 t = pool_alloc(sizeof(struct tag));
2634 t->next_tag = NULL;
2635 t->name = pool_strdup(sp);
2636 if (last_tag)
2637 last_tag->next_tag = t;
2638 else
2639 first_tag = t;
2640 last_tag = t;
2641 read_next_command();
2643 /* from ... */
2644 if (prefixcmp(command_buf.buf, "from "))
2645 die("Expected from command, got %s", command_buf.buf);
2646 from = strchr(command_buf.buf, ' ') + 1;
2647 s = lookup_branch(from);
2648 if (s) {
2649 hashcpy(sha1, s->sha1);
2650 type = OBJ_COMMIT;
2651 } else if (*from == ':') {
2652 struct object_entry *oe;
2653 from_mark = strtoumax(from + 1, NULL, 10);
2654 oe = find_mark(from_mark);
2655 type = oe->type;
2656 hashcpy(sha1, oe->idx.sha1);
2657 } else if (!get_sha1(from, sha1)) {
2658 unsigned long size;
2659 char *buf;
2661 buf = read_sha1_file(sha1, &type, &size);
2662 if (!buf || size < 46)
2663 die("Not a valid commit: %s", from);
2664 free(buf);
2665 } else
2666 die("Invalid ref name or SHA1 expression: %s", from);
2667 read_next_command();
2669 /* tagger ... */
2670 if (!prefixcmp(command_buf.buf, "tagger ")) {
2671 tagger = parse_ident(command_buf.buf + 7);
2672 read_next_command();
2673 } else
2674 tagger = NULL;
2676 /* tag payload/message */
2677 parse_data(&msg, 0, NULL);
2679 /* build the tag object */
2680 strbuf_reset(&new_data);
2682 strbuf_addf(&new_data,
2683 "object %s\n"
2684 "type %s\n"
2685 "tag %s\n",
2686 sha1_to_hex(sha1), typename(type), t->name);
2687 if (tagger)
2688 strbuf_addf(&new_data,
2689 "tagger %s\n", tagger);
2690 strbuf_addch(&new_data, '\n');
2691 strbuf_addbuf(&new_data, &msg);
2692 free(tagger);
2694 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2695 t->pack_id = MAX_PACK_ID;
2696 else
2697 t->pack_id = pack_id;
2698 }
2700 static void parse_reset_branch(void)
2701 {
2702 struct branch *b;
2703 char *sp;
2705 /* Obtain the branch name from the rest of our command */
2706 sp = strchr(command_buf.buf, ' ') + 1;
2707 b = lookup_branch(sp);
2708 if (b) {
2709 hashclr(b->sha1);
2710 hashclr(b->branch_tree.versions[0].sha1);
2711 hashclr(b->branch_tree.versions[1].sha1);
2712 if (b->branch_tree.tree) {
2713 release_tree_content_recursive(b->branch_tree.tree);
2714 b->branch_tree.tree = NULL;
2715 }
2716 }
2717 else
2718 b = new_branch(sp);
2719 read_next_command();
2720 parse_from(b);
2721 if (command_buf.len > 0)
2722 unread_command_buf = 1;
2723 }
2725 static void parse_checkpoint(void)
2726 {
2727 if (object_count) {
2728 cycle_packfile();
2729 dump_branches();
2730 dump_tags();
2731 dump_marks();
2732 }
2733 skip_optional_lf();
2734 }
2736 static void parse_progress(void)
2737 {
2738 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2739 fputc('\n', stdout);
2740 fflush(stdout);
2741 skip_optional_lf();
2742 }
2744 static char* make_fast_import_path(const char *path)
2745 {
2746 struct strbuf abs_path = STRBUF_INIT;
2748 if (!relative_marks_paths || is_absolute_path(path))
2749 return xstrdup(path);
2750 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
2751 return strbuf_detach(&abs_path, NULL);
2752 }
2754 static void option_import_marks(const char *marks, int from_stream)
2755 {
2756 if (import_marks_file) {
2757 if (from_stream)
2758 die("Only one import-marks command allowed per stream");
2760 /* read previous mark file */
2761 if(!import_marks_file_from_stream)
2762 read_marks();
2763 }
2765 import_marks_file = make_fast_import_path(marks);
2766 import_marks_file_from_stream = from_stream;
2767 }
2769 static void option_date_format(const char *fmt)
2770 {
2771 if (!strcmp(fmt, "raw"))
2772 whenspec = WHENSPEC_RAW;
2773 else if (!strcmp(fmt, "rfc2822"))
2774 whenspec = WHENSPEC_RFC2822;
2775 else if (!strcmp(fmt, "now"))
2776 whenspec = WHENSPEC_NOW;
2777 else
2778 die("unknown --date-format argument %s", fmt);
2779 }
2781 static void option_depth(const char *depth)
2782 {
2783 max_depth = strtoul(depth, NULL, 0);
2784 if (max_depth > MAX_DEPTH)
2785 die("--depth cannot exceed %u", MAX_DEPTH);
2786 }
2788 static void option_active_branches(const char *branches)
2789 {
2790 max_active_branches = strtoul(branches, NULL, 0);
2791 }
2793 static void option_export_marks(const char *marks)
2794 {
2795 export_marks_file = make_fast_import_path(marks);
2796 }
2798 static void option_export_pack_edges(const char *edges)
2799 {
2800 if (pack_edges)
2801 fclose(pack_edges);
2802 pack_edges = fopen(edges, "a");
2803 if (!pack_edges)
2804 die_errno("Cannot open '%s'", edges);
2805 }
2807 static int parse_one_option(const char *option)
2808 {
2809 if (!prefixcmp(option, "max-pack-size=")) {
2810 unsigned long v;
2811 if (!git_parse_ulong(option + 14, &v))
2812 return 0;
2813 if (v < 8192) {
2814 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
2815 v *= 1024 * 1024;
2816 } else if (v < 1024 * 1024) {
2817 warning("minimum max-pack-size is 1 MiB");
2818 v = 1024 * 1024;
2819 }
2820 max_packsize = v;
2821 } else if (!prefixcmp(option, "big-file-threshold=")) {
2822 unsigned long v;
2823 if (!git_parse_ulong(option + 19, &v))
2824 return 0;
2825 big_file_threshold = v;
2826 } else if (!prefixcmp(option, "depth=")) {
2827 option_depth(option + 6);
2828 } else if (!prefixcmp(option, "active-branches=")) {
2829 option_active_branches(option + 16);
2830 } else if (!prefixcmp(option, "export-pack-edges=")) {
2831 option_export_pack_edges(option + 18);
2832 } else if (!prefixcmp(option, "quiet")) {
2833 show_stats = 0;
2834 } else if (!prefixcmp(option, "stats")) {
2835 show_stats = 1;
2836 } else {
2837 return 0;
2838 }
2840 return 1;
2841 }
2843 static int parse_one_feature(const char *feature, int from_stream)
2844 {
2845 if (!prefixcmp(feature, "date-format=")) {
2846 option_date_format(feature + 12);
2847 } else if (!prefixcmp(feature, "import-marks=")) {
2848 option_import_marks(feature + 13, from_stream);
2849 } else if (!prefixcmp(feature, "export-marks=")) {
2850 option_export_marks(feature + 13);
2851 } else if (!prefixcmp(feature, "relative-marks")) {
2852 relative_marks_paths = 1;
2853 } else if (!prefixcmp(feature, "no-relative-marks")) {
2854 relative_marks_paths = 0;
2855 } else if (!prefixcmp(feature, "force")) {
2856 force_update = 1;
2857 } else {
2858 return 0;
2859 }
2861 return 1;
2862 }
2864 static void parse_feature(void)
2865 {
2866 char *feature = command_buf.buf + 8;
2868 if (seen_data_command)
2869 die("Got feature command '%s' after data command", feature);
2871 if (parse_one_feature(feature, 1))
2872 return;
2874 die("This version of fast-import does not support feature %s.", feature);
2875 }
2877 static void parse_option(void)
2878 {
2879 char *option = command_buf.buf + 11;
2881 if (seen_data_command)
2882 die("Got option command '%s' after data command", option);
2884 if (parse_one_option(option))
2885 return;
2887 die("This version of fast-import does not support option: %s", option);
2888 }
2890 static int git_pack_config(const char *k, const char *v, void *cb)
2891 {
2892 if (!strcmp(k, "pack.depth")) {
2893 max_depth = git_config_int(k, v);
2894 if (max_depth > MAX_DEPTH)
2895 max_depth = MAX_DEPTH;
2896 return 0;
2897 }
2898 if (!strcmp(k, "pack.compression")) {
2899 int level = git_config_int(k, v);
2900 if (level == -1)
2901 level = Z_DEFAULT_COMPRESSION;
2902 else if (level < 0 || level > Z_BEST_COMPRESSION)
2903 die("bad pack compression level %d", level);
2904 pack_compression_level = level;
2905 pack_compression_seen = 1;
2906 return 0;
2907 }
2908 if (!strcmp(k, "core.bigfilethreshold")) {
2909 long n = git_config_int(k, v);
2910 big_file_threshold = 0 < n ? n : 0;
2911 }
2912 return git_default_config(k, v, cb);
2913 }
2915 static const char fast_import_usage[] =
2916 "git fast-import [--date-format=f] [--max-pack-size=n] [--big-file-threshold=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2918 static void parse_argv(void)
2919 {
2920 unsigned int i;
2922 for (i = 1; i < global_argc; i++) {
2923 const char *a = global_argv[i];
2925 if (*a != '-' || !strcmp(a, "--"))
2926 break;
2928 if (parse_one_option(a + 2))
2929 continue;
2931 if (parse_one_feature(a + 2, 0))
2932 continue;
2934 die("unknown option %s", a);
2935 }
2936 if (i != global_argc)
2937 usage(fast_import_usage);
2939 seen_data_command = 1;
2940 if (import_marks_file)
2941 read_marks();
2942 }
2944 int main(int argc, const char **argv)
2945 {
2946 unsigned int i;
2948 git_extract_argv0_path(argv[0]);
2950 if (argc == 2 && !strcmp(argv[1], "-h"))
2951 usage(fast_import_usage);
2953 setup_git_directory();
2954 git_config(git_pack_config, NULL);
2955 if (!pack_compression_seen && core_compression_seen)
2956 pack_compression_level = core_compression_level;
2958 alloc_objects(object_entry_alloc);
2959 strbuf_init(&command_buf, 0);
2960 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2961 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2962 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2963 marks = pool_calloc(1, sizeof(struct mark_set));
2965 global_argc = argc;
2966 global_argv = argv;
2968 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2969 for (i = 0; i < (cmd_save - 1); i++)
2970 rc_free[i].next = &rc_free[i + 1];
2971 rc_free[cmd_save - 1].next = NULL;
2973 prepare_packed_git();
2974 start_packfile();
2975 set_die_routine(die_nicely);
2976 while (read_next_command() != EOF) {
2977 if (!strcmp("blob", command_buf.buf))
2978 parse_new_blob();
2979 else if (!prefixcmp(command_buf.buf, "commit "))
2980 parse_new_commit();
2981 else if (!prefixcmp(command_buf.buf, "tag "))
2982 parse_new_tag();
2983 else if (!prefixcmp(command_buf.buf, "reset "))
2984 parse_reset_branch();
2985 else if (!strcmp("checkpoint", command_buf.buf))
2986 parse_checkpoint();
2987 else if (!prefixcmp(command_buf.buf, "progress "))
2988 parse_progress();
2989 else if (!prefixcmp(command_buf.buf, "feature "))
2990 parse_feature();
2991 else if (!prefixcmp(command_buf.buf, "option git "))
2992 parse_option();
2993 else if (!prefixcmp(command_buf.buf, "option "))
2994 /* ignore non-git options*/;
2995 else
2996 die("Unsupported command: %s", command_buf.buf);
2997 }
2999 /* argv hasn't been parsed yet, do so */
3000 if (!seen_data_command)
3001 parse_argv();
3003 end_packfile();
3005 dump_branches();
3006 dump_tags();
3007 unkeep_all_packs();
3008 dump_marks();
3010 if (pack_edges)
3011 fclose(pack_edges);
3013 if (show_stats) {
3014 uintmax_t total_count = 0, duplicate_count = 0;
3015 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3016 total_count += object_count_by_type[i];
3017 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3018 duplicate_count += duplicate_count_by_type[i];
3020 fprintf(stderr, "%s statistics:\n", argv[0]);
3021 fprintf(stderr, "---------------------------------------------------------------------\n");
3022 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3023 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3024 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]);
3025 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]);
3026 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]);
3027 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]);
3028 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3029 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3030 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3031 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3032 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3033 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3034 fprintf(stderr, "---------------------------------------------------------------------\n");
3035 pack_report();
3036 fprintf(stderr, "---------------------------------------------------------------------\n");
3037 fprintf(stderr, "\n");
3038 }
3040 return failure ? 1 : 0;
3041 }