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 object_entry *next;
168 uint32_t offset;
169 uint32_t type : TYPE_BITS,
170 pack_id : PACK_ID_BITS,
171 depth : DEPTH_BITS;
172 unsigned char sha1[20];
173 };
175 struct object_entry_pool
176 {
177 struct object_entry_pool *next_pool;
178 struct object_entry *next_free;
179 struct object_entry *end;
180 struct object_entry entries[FLEX_ARRAY]; /* more */
181 };
183 struct mark_set
184 {
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 {
194 struct strbuf data;
195 uint32_t offset;
196 unsigned int depth;
197 unsigned no_swap : 1;
198 };
200 struct mem_pool
201 {
202 struct mem_pool *next_pool;
203 char *next_free;
204 char *end;
205 uintmax_t space[FLEX_ARRAY]; /* more */
206 };
208 struct atom_str
209 {
210 struct atom_str *next_atom;
211 unsigned short str_len;
212 char str_dat[FLEX_ARRAY]; /* more */
213 };
215 struct tree_content;
216 struct tree_entry
217 {
218 struct tree_content *tree;
219 struct atom_str *name;
220 struct tree_entry_ms
221 {
222 uint16_t mode;
223 unsigned char sha1[20];
224 } versions[2];
225 };
227 struct tree_content
228 {
229 unsigned int entry_capacity; /* must match avail_tree_content */
230 unsigned int entry_count;
231 unsigned int delta_depth;
232 struct tree_entry *entries[FLEX_ARRAY]; /* more */
233 };
235 struct avail_tree_content
236 {
237 unsigned int entry_capacity; /* must match tree_content */
238 struct avail_tree_content *next_avail;
239 };
241 struct branch
242 {
243 struct branch *table_next_branch;
244 struct branch *active_next_branch;
245 const char *name;
246 struct tree_entry branch_tree;
247 uintmax_t last_commit;
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 int force_update;
284 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
285 static int pack_compression_seen;
287 /* Stats and misc. counters */
288 static uintmax_t alloc_count;
289 static uintmax_t marks_set_count;
290 static uintmax_t object_count_by_type[1 << TYPE_BITS];
291 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
292 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
293 static unsigned long object_count;
294 static unsigned long branch_count;
295 static unsigned long branch_load_count;
296 static int failure;
297 static FILE *pack_edges;
298 static unsigned int show_stats = 1;
300 /* Memory pools */
301 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
302 static size_t total_allocd;
303 static struct mem_pool *mem_pool;
305 /* Atom management */
306 static unsigned int atom_table_sz = 4451;
307 static unsigned int atom_cnt;
308 static struct atom_str **atom_table;
310 /* The .pack file being generated */
311 static unsigned int pack_id;
312 static struct packed_git *pack_data;
313 static struct packed_git **all_packs;
314 static unsigned long pack_size;
316 /* Table of objects we've written. */
317 static unsigned int object_entry_alloc = 5000;
318 static struct object_entry_pool *blocks;
319 static struct object_entry *object_table[1 << 16];
320 static struct mark_set *marks;
321 static const char *export_marks_file;
322 static const char *import_marks_file;
324 /* Our last blob */
325 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
327 /* Tree management */
328 static unsigned int tree_entry_alloc = 1000;
329 static void *avail_tree_entry;
330 static unsigned int avail_tree_table_sz = 100;
331 static struct avail_tree_content **avail_tree_table;
332 static struct strbuf old_tree = STRBUF_INIT;
333 static struct strbuf new_tree = STRBUF_INIT;
335 /* Branch data */
336 static unsigned long max_active_branches = 5;
337 static unsigned long cur_active_branches;
338 static unsigned long branch_table_sz = 1039;
339 static struct branch **branch_table;
340 static struct branch *active_branches;
342 /* Tag data */
343 static struct tag *first_tag;
344 static struct tag *last_tag;
346 /* Input stream parsing */
347 static whenspec_type whenspec = WHENSPEC_RAW;
348 static struct strbuf command_buf = STRBUF_INIT;
349 static int unread_command_buf;
350 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
351 static struct recent_command *cmd_tail = &cmd_hist;
352 static struct recent_command *rc_free;
353 static unsigned int cmd_save = 100;
354 static uintmax_t next_mark;
355 static struct strbuf new_data = STRBUF_INIT;
356 static int seen_data_command;
358 static void write_branch_report(FILE *rpt, struct branch *b)
359 {
360 fprintf(rpt, "%s:\n", b->name);
362 fprintf(rpt, " status :");
363 if (b->active)
364 fputs(" active", rpt);
365 if (b->branch_tree.tree)
366 fputs(" loaded", rpt);
367 if (is_null_sha1(b->branch_tree.versions[1].sha1))
368 fputs(" dirty", rpt);
369 fputc('\n', rpt);
371 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
372 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
373 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
374 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
376 fputs(" last pack : ", rpt);
377 if (b->pack_id < MAX_PACK_ID)
378 fprintf(rpt, "%u", b->pack_id);
379 fputc('\n', rpt);
381 fputc('\n', rpt);
382 }
384 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
386 static void write_crash_report(const char *err)
387 {
388 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
389 FILE *rpt = fopen(loc, "w");
390 struct branch *b;
391 unsigned long lu;
392 struct recent_command *rc;
394 if (!rpt) {
395 error("can't write crash report %s: %s", loc, strerror(errno));
396 return;
397 }
399 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
401 fprintf(rpt, "fast-import crash report:\n");
402 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
403 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
404 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
405 fputc('\n', rpt);
407 fputs("fatal: ", rpt);
408 fputs(err, rpt);
409 fputc('\n', rpt);
411 fputc('\n', rpt);
412 fputs("Most Recent Commands Before Crash\n", rpt);
413 fputs("---------------------------------\n", rpt);
414 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
415 if (rc->next == &cmd_hist)
416 fputs("* ", rpt);
417 else
418 fputs(" ", rpt);
419 fputs(rc->buf, rpt);
420 fputc('\n', rpt);
421 }
423 fputc('\n', rpt);
424 fputs("Active Branch LRU\n", rpt);
425 fputs("-----------------\n", rpt);
426 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
427 cur_active_branches,
428 max_active_branches);
429 fputc('\n', rpt);
430 fputs(" pos clock name\n", rpt);
431 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
432 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
433 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
434 ++lu, b->last_commit, b->name);
436 fputc('\n', rpt);
437 fputs("Inactive Branches\n", rpt);
438 fputs("-----------------\n", rpt);
439 for (lu = 0; lu < branch_table_sz; lu++) {
440 for (b = branch_table[lu]; b; b = b->table_next_branch)
441 write_branch_report(rpt, b);
442 }
444 if (first_tag) {
445 struct tag *tg;
446 fputc('\n', rpt);
447 fputs("Annotated Tags\n", rpt);
448 fputs("--------------\n", rpt);
449 for (tg = first_tag; tg; tg = tg->next_tag) {
450 fputs(sha1_to_hex(tg->sha1), rpt);
451 fputc(' ', rpt);
452 fputs(tg->name, rpt);
453 fputc('\n', rpt);
454 }
455 }
457 fputc('\n', rpt);
458 fputs("Marks\n", rpt);
459 fputs("-----\n", rpt);
460 if (export_marks_file)
461 fprintf(rpt, " exported to %s\n", export_marks_file);
462 else
463 dump_marks_helper(rpt, 0, marks);
465 fputc('\n', rpt);
466 fputs("-------------------\n", rpt);
467 fputs("END OF CRASH REPORT\n", rpt);
468 fclose(rpt);
469 }
471 static void end_packfile(void);
472 static void unkeep_all_packs(void);
473 static void dump_marks(void);
475 static NORETURN void die_nicely(const char *err, va_list params)
476 {
477 static int zombie;
478 char message[2 * PATH_MAX];
480 vsnprintf(message, sizeof(message), err, params);
481 fputs("fatal: ", stderr);
482 fputs(message, stderr);
483 fputc('\n', stderr);
485 if (!zombie) {
486 zombie = 1;
487 write_crash_report(message);
488 end_packfile();
489 unkeep_all_packs();
490 dump_marks();
491 }
492 exit(128);
493 }
495 static void alloc_objects(unsigned int cnt)
496 {
497 struct object_entry_pool *b;
499 b = xmalloc(sizeof(struct object_entry_pool)
500 + cnt * sizeof(struct object_entry));
501 b->next_pool = blocks;
502 b->next_free = b->entries;
503 b->end = b->entries + cnt;
504 blocks = b;
505 alloc_count += cnt;
506 }
508 static struct object_entry *new_object(unsigned char *sha1)
509 {
510 struct object_entry *e;
512 if (blocks->next_free == blocks->end)
513 alloc_objects(object_entry_alloc);
515 e = blocks->next_free++;
516 hashcpy(e->sha1, sha1);
517 return e;
518 }
520 static struct object_entry *find_object(unsigned char *sha1)
521 {
522 unsigned int h = sha1[0] << 8 | sha1[1];
523 struct object_entry *e;
524 for (e = object_table[h]; e; e = e->next)
525 if (!hashcmp(sha1, e->sha1))
526 return e;
527 return NULL;
528 }
530 static struct object_entry *insert_object(unsigned char *sha1)
531 {
532 unsigned int h = sha1[0] << 8 | sha1[1];
533 struct object_entry *e = object_table[h];
534 struct object_entry *p = NULL;
536 while (e) {
537 if (!hashcmp(sha1, e->sha1))
538 return e;
539 p = e;
540 e = e->next;
541 }
543 e = new_object(sha1);
544 e->next = NULL;
545 e->offset = 0;
546 if (p)
547 p->next = e;
548 else
549 object_table[h] = e;
550 return e;
551 }
553 static unsigned int hc_str(const char *s, size_t len)
554 {
555 unsigned int r = 0;
556 while (len-- > 0)
557 r = r * 31 + *s++;
558 return r;
559 }
561 static void *pool_alloc(size_t len)
562 {
563 struct mem_pool *p;
564 void *r;
566 /* round up to a 'uintmax_t' alignment */
567 if (len & (sizeof(uintmax_t) - 1))
568 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
570 for (p = mem_pool; p; p = p->next_pool)
571 if ((p->end - p->next_free >= len))
572 break;
574 if (!p) {
575 if (len >= (mem_pool_alloc/2)) {
576 total_allocd += len;
577 return xmalloc(len);
578 }
579 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
580 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
581 p->next_pool = mem_pool;
582 p->next_free = (char *) p->space;
583 p->end = p->next_free + mem_pool_alloc;
584 mem_pool = p;
585 }
587 r = p->next_free;
588 p->next_free += len;
589 return r;
590 }
592 static void *pool_calloc(size_t count, size_t size)
593 {
594 size_t len = count * size;
595 void *r = pool_alloc(len);
596 memset(r, 0, len);
597 return r;
598 }
600 static char *pool_strdup(const char *s)
601 {
602 char *r = pool_alloc(strlen(s) + 1);
603 strcpy(r, s);
604 return r;
605 }
607 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
608 {
609 struct mark_set *s = marks;
610 while ((idnum >> s->shift) >= 1024) {
611 s = pool_calloc(1, sizeof(struct mark_set));
612 s->shift = marks->shift + 10;
613 s->data.sets[0] = marks;
614 marks = s;
615 }
616 while (s->shift) {
617 uintmax_t i = idnum >> s->shift;
618 idnum -= i << s->shift;
619 if (!s->data.sets[i]) {
620 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
621 s->data.sets[i]->shift = s->shift - 10;
622 }
623 s = s->data.sets[i];
624 }
625 if (!s->data.marked[idnum])
626 marks_set_count++;
627 s->data.marked[idnum] = oe;
628 }
630 static struct object_entry *find_mark(uintmax_t idnum)
631 {
632 uintmax_t orig_idnum = idnum;
633 struct mark_set *s = marks;
634 struct object_entry *oe = NULL;
635 if ((idnum >> s->shift) < 1024) {
636 while (s && s->shift) {
637 uintmax_t i = idnum >> s->shift;
638 idnum -= i << s->shift;
639 s = s->data.sets[i];
640 }
641 if (s)
642 oe = s->data.marked[idnum];
643 }
644 if (!oe)
645 die("mark :%" PRIuMAX " not declared", orig_idnum);
646 return oe;
647 }
649 static struct atom_str *to_atom(const char *s, unsigned short len)
650 {
651 unsigned int hc = hc_str(s, len) % atom_table_sz;
652 struct atom_str *c;
654 for (c = atom_table[hc]; c; c = c->next_atom)
655 if (c->str_len == len && !strncmp(s, c->str_dat, len))
656 return c;
658 c = pool_alloc(sizeof(struct atom_str) + len + 1);
659 c->str_len = len;
660 strncpy(c->str_dat, s, len);
661 c->str_dat[len] = 0;
662 c->next_atom = atom_table[hc];
663 atom_table[hc] = c;
664 atom_cnt++;
665 return c;
666 }
668 static struct branch *lookup_branch(const char *name)
669 {
670 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
671 struct branch *b;
673 for (b = branch_table[hc]; b; b = b->table_next_branch)
674 if (!strcmp(name, b->name))
675 return b;
676 return NULL;
677 }
679 static struct branch *new_branch(const char *name)
680 {
681 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
682 struct branch *b = lookup_branch(name);
684 if (b)
685 die("Invalid attempt to create duplicate branch: %s", name);
686 switch (check_ref_format(name)) {
687 case 0: break; /* its valid */
688 case CHECK_REF_FORMAT_ONELEVEL:
689 break; /* valid, but too few '/', allow anyway */
690 default:
691 die("Branch name doesn't conform to GIT standards: %s", name);
692 }
694 b = pool_calloc(1, sizeof(struct branch));
695 b->name = pool_strdup(name);
696 b->table_next_branch = branch_table[hc];
697 b->branch_tree.versions[0].mode = S_IFDIR;
698 b->branch_tree.versions[1].mode = S_IFDIR;
699 b->active = 0;
700 b->pack_id = MAX_PACK_ID;
701 branch_table[hc] = b;
702 branch_count++;
703 return b;
704 }
706 static unsigned int hc_entries(unsigned int cnt)
707 {
708 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
709 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
710 }
712 static struct tree_content *new_tree_content(unsigned int cnt)
713 {
714 struct avail_tree_content *f, *l = NULL;
715 struct tree_content *t;
716 unsigned int hc = hc_entries(cnt);
718 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
719 if (f->entry_capacity >= cnt)
720 break;
722 if (f) {
723 if (l)
724 l->next_avail = f->next_avail;
725 else
726 avail_tree_table[hc] = f->next_avail;
727 } else {
728 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
729 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
730 f->entry_capacity = cnt;
731 }
733 t = (struct tree_content*)f;
734 t->entry_count = 0;
735 t->delta_depth = 0;
736 return t;
737 }
739 static void release_tree_entry(struct tree_entry *e);
740 static void release_tree_content(struct tree_content *t)
741 {
742 struct avail_tree_content *f = (struct avail_tree_content*)t;
743 unsigned int hc = hc_entries(f->entry_capacity);
744 f->next_avail = avail_tree_table[hc];
745 avail_tree_table[hc] = f;
746 }
748 static void release_tree_content_recursive(struct tree_content *t)
749 {
750 unsigned int i;
751 for (i = 0; i < t->entry_count; i++)
752 release_tree_entry(t->entries[i]);
753 release_tree_content(t);
754 }
756 static struct tree_content *grow_tree_content(
757 struct tree_content *t,
758 int amt)
759 {
760 struct tree_content *r = new_tree_content(t->entry_count + amt);
761 r->entry_count = t->entry_count;
762 r->delta_depth = t->delta_depth;
763 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
764 release_tree_content(t);
765 return r;
766 }
768 static struct tree_entry *new_tree_entry(void)
769 {
770 struct tree_entry *e;
772 if (!avail_tree_entry) {
773 unsigned int n = tree_entry_alloc;
774 total_allocd += n * sizeof(struct tree_entry);
775 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
776 while (n-- > 1) {
777 *((void**)e) = e + 1;
778 e++;
779 }
780 *((void**)e) = NULL;
781 }
783 e = avail_tree_entry;
784 avail_tree_entry = *((void**)e);
785 return e;
786 }
788 static void release_tree_entry(struct tree_entry *e)
789 {
790 if (e->tree)
791 release_tree_content_recursive(e->tree);
792 *((void**)e) = avail_tree_entry;
793 avail_tree_entry = e;
794 }
796 static struct tree_content *dup_tree_content(struct tree_content *s)
797 {
798 struct tree_content *d;
799 struct tree_entry *a, *b;
800 unsigned int i;
802 if (!s)
803 return NULL;
804 d = new_tree_content(s->entry_count);
805 for (i = 0; i < s->entry_count; i++) {
806 a = s->entries[i];
807 b = new_tree_entry();
808 memcpy(b, a, sizeof(*a));
809 if (a->tree && is_null_sha1(b->versions[1].sha1))
810 b->tree = dup_tree_content(a->tree);
811 else
812 b->tree = NULL;
813 d->entries[i] = b;
814 }
815 d->entry_count = s->entry_count;
816 d->delta_depth = s->delta_depth;
818 return d;
819 }
821 static void start_packfile(void)
822 {
823 static char tmpfile[PATH_MAX];
824 struct packed_git *p;
825 struct pack_header hdr;
826 int pack_fd;
828 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
829 "pack/tmp_pack_XXXXXX");
830 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
831 strcpy(p->pack_name, tmpfile);
832 p->pack_fd = pack_fd;
834 hdr.hdr_signature = htonl(PACK_SIGNATURE);
835 hdr.hdr_version = htonl(2);
836 hdr.hdr_entries = 0;
837 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
839 pack_data = p;
840 pack_size = sizeof(hdr);
841 object_count = 0;
843 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
844 all_packs[pack_id] = p;
845 }
847 static int oecmp (const void *a_, const void *b_)
848 {
849 struct object_entry *a = *((struct object_entry**)a_);
850 struct object_entry *b = *((struct object_entry**)b_);
851 return hashcmp(a->sha1, b->sha1);
852 }
854 static char *create_index(void)
855 {
856 static char tmpfile[PATH_MAX];
857 git_SHA_CTX ctx;
858 struct sha1file *f;
859 struct object_entry **idx, **c, **last, *e;
860 struct object_entry_pool *o;
861 uint32_t array[256];
862 int i, idx_fd;
864 /* Build the sorted table of object IDs. */
865 idx = xmalloc(object_count * sizeof(struct object_entry*));
866 c = idx;
867 for (o = blocks; o; o = o->next_pool)
868 for (e = o->next_free; e-- != o->entries;)
869 if (pack_id == e->pack_id)
870 *c++ = e;
871 last = idx + object_count;
872 if (c != last)
873 die("internal consistency error creating the index");
874 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
876 /* Generate the fan-out array. */
877 c = idx;
878 for (i = 0; i < 256; i++) {
879 struct object_entry **next = c;
880 while (next < last) {
881 if ((*next)->sha1[0] != i)
882 break;
883 next++;
884 }
885 array[i] = htonl(next - idx);
886 c = next;
887 }
889 idx_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
890 "pack/tmp_idx_XXXXXX");
891 f = sha1fd(idx_fd, tmpfile);
892 sha1write(f, array, 256 * sizeof(int));
893 git_SHA1_Init(&ctx);
894 for (c = idx; c != last; c++) {
895 uint32_t offset = htonl((*c)->offset);
896 sha1write(f, &offset, 4);
897 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
898 git_SHA1_Update(&ctx, (*c)->sha1, 20);
899 }
900 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
901 sha1close(f, NULL, CSUM_FSYNC);
902 free(idx);
903 git_SHA1_Final(pack_data->sha1, &ctx);
904 return tmpfile;
905 }
907 static char *keep_pack(char *curr_index_name)
908 {
909 static char name[PATH_MAX];
910 static const char *keep_msg = "fast-import";
911 int keep_fd;
913 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
914 if (keep_fd < 0)
915 die_errno("cannot create keep file");
916 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
917 if (close(keep_fd))
918 die_errno("failed to write keep file");
920 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
921 get_object_directory(), sha1_to_hex(pack_data->sha1));
922 if (move_temp_to_file(pack_data->pack_name, name))
923 die("cannot store pack file");
925 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
926 get_object_directory(), sha1_to_hex(pack_data->sha1));
927 if (move_temp_to_file(curr_index_name, name))
928 die("cannot store index file");
929 return name;
930 }
932 static void unkeep_all_packs(void)
933 {
934 static char name[PATH_MAX];
935 int k;
937 for (k = 0; k < pack_id; k++) {
938 struct packed_git *p = all_packs[k];
939 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
940 get_object_directory(), sha1_to_hex(p->sha1));
941 unlink_or_warn(name);
942 }
943 }
945 static void end_packfile(void)
946 {
947 struct packed_git *old_p = pack_data, *new_p;
949 clear_delta_base_cache();
950 if (object_count) {
951 char *idx_name;
952 int i;
953 struct branch *b;
954 struct tag *t;
956 close_pack_windows(pack_data);
957 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
958 pack_data->pack_name, object_count,
959 NULL, 0);
960 close(pack_data->pack_fd);
961 idx_name = keep_pack(create_index());
963 /* Register the packfile with core git's machinery. */
964 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
965 if (!new_p)
966 die("core git rejected index %s", idx_name);
967 all_packs[pack_id] = new_p;
968 install_packed_git(new_p);
970 /* Print the boundary */
971 if (pack_edges) {
972 fprintf(pack_edges, "%s:", new_p->pack_name);
973 for (i = 0; i < branch_table_sz; i++) {
974 for (b = branch_table[i]; b; b = b->table_next_branch) {
975 if (b->pack_id == pack_id)
976 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
977 }
978 }
979 for (t = first_tag; t; t = t->next_tag) {
980 if (t->pack_id == pack_id)
981 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
982 }
983 fputc('\n', pack_edges);
984 fflush(pack_edges);
985 }
987 pack_id++;
988 }
989 else {
990 close(old_p->pack_fd);
991 unlink_or_warn(old_p->pack_name);
992 }
993 free(old_p);
995 /* We can't carry a delta across packfiles. */
996 strbuf_release(&last_blob.data);
997 last_blob.offset = 0;
998 last_blob.depth = 0;
999 }
1001 static void cycle_packfile(void)
1002 {
1003 end_packfile();
1004 start_packfile();
1005 }
1007 static size_t encode_header(
1008 enum object_type type,
1009 size_t size,
1010 unsigned char *hdr)
1011 {
1012 int n = 1;
1013 unsigned char c;
1015 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
1016 die("bad type %d", type);
1018 c = (type << 4) | (size & 15);
1019 size >>= 4;
1020 while (size) {
1021 *hdr++ = c | 0x80;
1022 c = size & 0x7f;
1023 size >>= 7;
1024 n++;
1025 }
1026 *hdr = c;
1027 return n;
1028 }
1030 static int store_object(
1031 enum object_type type,
1032 struct strbuf *dat,
1033 struct last_object *last,
1034 unsigned char *sha1out,
1035 uintmax_t mark)
1036 {
1037 void *out, *delta;
1038 struct object_entry *e;
1039 unsigned char hdr[96];
1040 unsigned char sha1[20];
1041 unsigned long hdrlen, deltalen;
1042 git_SHA_CTX c;
1043 z_stream s;
1045 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1046 (unsigned long)dat->len) + 1;
1047 git_SHA1_Init(&c);
1048 git_SHA1_Update(&c, hdr, hdrlen);
1049 git_SHA1_Update(&c, dat->buf, dat->len);
1050 git_SHA1_Final(sha1, &c);
1051 if (sha1out)
1052 hashcpy(sha1out, sha1);
1054 e = insert_object(sha1);
1055 if (mark)
1056 insert_mark(mark, e);
1057 if (e->offset) {
1058 duplicate_count_by_type[type]++;
1059 return 1;
1060 } else if (find_sha1_pack(sha1, packed_git)) {
1061 e->type = type;
1062 e->pack_id = MAX_PACK_ID;
1063 e->offset = 1; /* just not zero! */
1064 duplicate_count_by_type[type]++;
1065 return 1;
1066 }
1068 if (last && last->data.buf && last->depth < max_depth) {
1069 delta = diff_delta(last->data.buf, last->data.len,
1070 dat->buf, dat->len,
1071 &deltalen, 0);
1072 if (delta && deltalen >= dat->len) {
1073 free(delta);
1074 delta = NULL;
1075 }
1076 } else
1077 delta = NULL;
1079 memset(&s, 0, sizeof(s));
1080 deflateInit(&s, pack_compression_level);
1081 if (delta) {
1082 s.next_in = delta;
1083 s.avail_in = deltalen;
1084 } else {
1085 s.next_in = (void *)dat->buf;
1086 s.avail_in = dat->len;
1087 }
1088 s.avail_out = deflateBound(&s, s.avail_in);
1089 s.next_out = out = xmalloc(s.avail_out);
1090 while (deflate(&s, Z_FINISH) == Z_OK)
1091 /* nothing */;
1092 deflateEnd(&s);
1094 /* Determine if we should auto-checkpoint. */
1095 if ((pack_size + 60 + s.total_out) > max_packsize
1096 || (pack_size + 60 + s.total_out) < pack_size) {
1098 /* This new object needs to *not* have the current pack_id. */
1099 e->pack_id = pack_id + 1;
1100 cycle_packfile();
1102 /* We cannot carry a delta into the new pack. */
1103 if (delta) {
1104 free(delta);
1105 delta = NULL;
1107 memset(&s, 0, sizeof(s));
1108 deflateInit(&s, pack_compression_level);
1109 s.next_in = (void *)dat->buf;
1110 s.avail_in = dat->len;
1111 s.avail_out = deflateBound(&s, s.avail_in);
1112 s.next_out = out = xrealloc(out, s.avail_out);
1113 while (deflate(&s, Z_FINISH) == Z_OK)
1114 /* nothing */;
1115 deflateEnd(&s);
1116 }
1117 }
1119 e->type = type;
1120 e->pack_id = pack_id;
1121 e->offset = pack_size;
1122 object_count++;
1123 object_count_by_type[type]++;
1125 if (delta) {
1126 unsigned long ofs = e->offset - last->offset;
1127 unsigned pos = sizeof(hdr) - 1;
1129 delta_count_by_type[type]++;
1130 e->depth = last->depth + 1;
1132 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1133 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1134 pack_size += hdrlen;
1136 hdr[pos] = ofs & 127;
1137 while (ofs >>= 7)
1138 hdr[--pos] = 128 | (--ofs & 127);
1139 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
1140 pack_size += sizeof(hdr) - pos;
1141 } else {
1142 e->depth = 0;
1143 hdrlen = encode_header(type, dat->len, hdr);
1144 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1145 pack_size += hdrlen;
1146 }
1148 write_or_die(pack_data->pack_fd, out, s.total_out);
1149 pack_size += s.total_out;
1151 free(out);
1152 free(delta);
1153 if (last) {
1154 if (last->no_swap) {
1155 last->data = *dat;
1156 } else {
1157 strbuf_swap(&last->data, dat);
1158 }
1159 last->offset = e->offset;
1160 last->depth = e->depth;
1161 }
1162 return 0;
1163 }
1165 /* All calls must be guarded by find_object() or find_mark() to
1166 * ensure the 'struct object_entry' passed was written by this
1167 * process instance. We unpack the entry by the offset, avoiding
1168 * the need for the corresponding .idx file. This unpacking rule
1169 * works because we only use OBJ_REF_DELTA within the packfiles
1170 * created by fast-import.
1171 *
1172 * oe must not be NULL. Such an oe usually comes from giving
1173 * an unknown SHA-1 to find_object() or an undefined mark to
1174 * find_mark(). Callers must test for this condition and use
1175 * the standard read_sha1_file() when it happens.
1176 *
1177 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1178 * find_mark(), where the mark was reloaded from an existing marks
1179 * file and is referencing an object that this fast-import process
1180 * instance did not write out to a packfile. Callers must test for
1181 * this condition and use read_sha1_file() instead.
1182 */
1183 static void *gfi_unpack_entry(
1184 struct object_entry *oe,
1185 unsigned long *sizep)
1186 {
1187 enum object_type type;
1188 struct packed_git *p = all_packs[oe->pack_id];
1189 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1190 /* The object is stored in the packfile we are writing to
1191 * and we have modified it since the last time we scanned
1192 * back to read a previously written object. If an old
1193 * window covered [p->pack_size, p->pack_size + 20) its
1194 * data is stale and is not valid. Closing all windows
1195 * and updating the packfile length ensures we can read
1196 * the newly written data.
1197 */
1198 close_pack_windows(p);
1200 /* We have to offer 20 bytes additional on the end of
1201 * the packfile as the core unpacker code assumes the
1202 * footer is present at the file end and must promise
1203 * at least 20 bytes within any window it maps. But
1204 * we don't actually create the footer here.
1205 */
1206 p->pack_size = pack_size + 20;
1207 }
1208 return unpack_entry(p, oe->offset, &type, sizep);
1209 }
1211 static const char *get_mode(const char *str, uint16_t *modep)
1212 {
1213 unsigned char c;
1214 uint16_t mode = 0;
1216 while ((c = *str++) != ' ') {
1217 if (c < '0' || c > '7')
1218 return NULL;
1219 mode = (mode << 3) + (c - '0');
1220 }
1221 *modep = mode;
1222 return str;
1223 }
1225 static void load_tree(struct tree_entry *root)
1226 {
1227 unsigned char *sha1 = root->versions[1].sha1;
1228 struct object_entry *myoe;
1229 struct tree_content *t;
1230 unsigned long size;
1231 char *buf;
1232 const char *c;
1234 root->tree = t = new_tree_content(8);
1235 if (is_null_sha1(sha1))
1236 return;
1238 myoe = find_object(sha1);
1239 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1240 if (myoe->type != OBJ_TREE)
1241 die("Not a tree: %s", sha1_to_hex(sha1));
1242 t->delta_depth = myoe->depth;
1243 buf = gfi_unpack_entry(myoe, &size);
1244 if (!buf)
1245 die("Can't load tree %s", sha1_to_hex(sha1));
1246 } else {
1247 enum object_type type;
1248 buf = read_sha1_file(sha1, &type, &size);
1249 if (!buf || type != OBJ_TREE)
1250 die("Can't load tree %s", sha1_to_hex(sha1));
1251 }
1253 c = buf;
1254 while (c != (buf + size)) {
1255 struct tree_entry *e = new_tree_entry();
1257 if (t->entry_count == t->entry_capacity)
1258 root->tree = t = grow_tree_content(t, t->entry_count);
1259 t->entries[t->entry_count++] = e;
1261 e->tree = NULL;
1262 c = get_mode(c, &e->versions[1].mode);
1263 if (!c)
1264 die("Corrupt mode in %s", sha1_to_hex(sha1));
1265 e->versions[0].mode = e->versions[1].mode;
1266 e->name = to_atom(c, strlen(c));
1267 c += e->name->str_len + 1;
1268 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1269 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1270 c += 20;
1271 }
1272 free(buf);
1273 }
1275 static int tecmp0 (const void *_a, const void *_b)
1276 {
1277 struct tree_entry *a = *((struct tree_entry**)_a);
1278 struct tree_entry *b = *((struct tree_entry**)_b);
1279 return base_name_compare(
1280 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1281 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1282 }
1284 static int tecmp1 (const void *_a, const void *_b)
1285 {
1286 struct tree_entry *a = *((struct tree_entry**)_a);
1287 struct tree_entry *b = *((struct tree_entry**)_b);
1288 return base_name_compare(
1289 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1290 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1291 }
1293 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1294 {
1295 size_t maxlen = 0;
1296 unsigned int i;
1298 if (!v)
1299 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1300 else
1301 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1303 for (i = 0; i < t->entry_count; i++) {
1304 if (t->entries[i]->versions[v].mode)
1305 maxlen += t->entries[i]->name->str_len + 34;
1306 }
1308 strbuf_reset(b);
1309 strbuf_grow(b, maxlen);
1310 for (i = 0; i < t->entry_count; i++) {
1311 struct tree_entry *e = t->entries[i];
1312 if (!e->versions[v].mode)
1313 continue;
1314 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1315 e->name->str_dat, '\0');
1316 strbuf_add(b, e->versions[v].sha1, 20);
1317 }
1318 }
1320 static void store_tree(struct tree_entry *root)
1321 {
1322 struct tree_content *t = root->tree;
1323 unsigned int i, j, del;
1324 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1325 struct object_entry *le;
1327 if (!is_null_sha1(root->versions[1].sha1))
1328 return;
1330 for (i = 0; i < t->entry_count; i++) {
1331 if (t->entries[i]->tree)
1332 store_tree(t->entries[i]);
1333 }
1335 le = find_object(root->versions[0].sha1);
1336 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1337 mktree(t, 0, &old_tree);
1338 lo.data = old_tree;
1339 lo.offset = le->offset;
1340 lo.depth = t->delta_depth;
1341 }
1343 mktree(t, 1, &new_tree);
1344 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1346 t->delta_depth = lo.depth;
1347 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1348 struct tree_entry *e = t->entries[i];
1349 if (e->versions[1].mode) {
1350 e->versions[0].mode = e->versions[1].mode;
1351 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1352 t->entries[j++] = e;
1353 } else {
1354 release_tree_entry(e);
1355 del++;
1356 }
1357 }
1358 t->entry_count -= del;
1359 }
1361 static int tree_content_set(
1362 struct tree_entry *root,
1363 const char *p,
1364 const unsigned char *sha1,
1365 const uint16_t mode,
1366 struct tree_content *subtree)
1367 {
1368 struct tree_content *t = root->tree;
1369 const char *slash1;
1370 unsigned int i, n;
1371 struct tree_entry *e;
1373 slash1 = strchr(p, '/');
1374 if (slash1)
1375 n = slash1 - p;
1376 else
1377 n = strlen(p);
1378 if (!n)
1379 die("Empty path component found in input");
1380 if (!slash1 && !S_ISDIR(mode) && subtree)
1381 die("Non-directories cannot have subtrees");
1383 for (i = 0; i < t->entry_count; i++) {
1384 e = t->entries[i];
1385 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1386 if (!slash1) {
1387 if (!S_ISDIR(mode)
1388 && e->versions[1].mode == mode
1389 && !hashcmp(e->versions[1].sha1, sha1))
1390 return 0;
1391 e->versions[1].mode = mode;
1392 hashcpy(e->versions[1].sha1, sha1);
1393 if (e->tree)
1394 release_tree_content_recursive(e->tree);
1395 e->tree = subtree;
1396 hashclr(root->versions[1].sha1);
1397 return 1;
1398 }
1399 if (!S_ISDIR(e->versions[1].mode)) {
1400 e->tree = new_tree_content(8);
1401 e->versions[1].mode = S_IFDIR;
1402 }
1403 if (!e->tree)
1404 load_tree(e);
1405 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1406 hashclr(root->versions[1].sha1);
1407 return 1;
1408 }
1409 return 0;
1410 }
1411 }
1413 if (t->entry_count == t->entry_capacity)
1414 root->tree = t = grow_tree_content(t, t->entry_count);
1415 e = new_tree_entry();
1416 e->name = to_atom(p, n);
1417 e->versions[0].mode = 0;
1418 hashclr(e->versions[0].sha1);
1419 t->entries[t->entry_count++] = e;
1420 if (slash1) {
1421 e->tree = new_tree_content(8);
1422 e->versions[1].mode = S_IFDIR;
1423 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1424 } else {
1425 e->tree = subtree;
1426 e->versions[1].mode = mode;
1427 hashcpy(e->versions[1].sha1, sha1);
1428 }
1429 hashclr(root->versions[1].sha1);
1430 return 1;
1431 }
1433 static int tree_content_remove(
1434 struct tree_entry *root,
1435 const char *p,
1436 struct tree_entry *backup_leaf)
1437 {
1438 struct tree_content *t = root->tree;
1439 const char *slash1;
1440 unsigned int i, n;
1441 struct tree_entry *e;
1443 slash1 = strchr(p, '/');
1444 if (slash1)
1445 n = slash1 - p;
1446 else
1447 n = strlen(p);
1449 for (i = 0; i < t->entry_count; i++) {
1450 e = t->entries[i];
1451 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1452 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1453 goto del_entry;
1454 if (!e->tree)
1455 load_tree(e);
1456 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1457 for (n = 0; n < e->tree->entry_count; n++) {
1458 if (e->tree->entries[n]->versions[1].mode) {
1459 hashclr(root->versions[1].sha1);
1460 return 1;
1461 }
1462 }
1463 backup_leaf = NULL;
1464 goto del_entry;
1465 }
1466 return 0;
1467 }
1468 }
1469 return 0;
1471 del_entry:
1472 if (backup_leaf)
1473 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1474 else if (e->tree)
1475 release_tree_content_recursive(e->tree);
1476 e->tree = NULL;
1477 e->versions[1].mode = 0;
1478 hashclr(e->versions[1].sha1);
1479 hashclr(root->versions[1].sha1);
1480 return 1;
1481 }
1483 static int tree_content_get(
1484 struct tree_entry *root,
1485 const char *p,
1486 struct tree_entry *leaf)
1487 {
1488 struct tree_content *t = root->tree;
1489 const char *slash1;
1490 unsigned int i, n;
1491 struct tree_entry *e;
1493 slash1 = strchr(p, '/');
1494 if (slash1)
1495 n = slash1 - p;
1496 else
1497 n = strlen(p);
1499 for (i = 0; i < t->entry_count; i++) {
1500 e = t->entries[i];
1501 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1502 if (!slash1) {
1503 memcpy(leaf, e, sizeof(*leaf));
1504 if (e->tree && is_null_sha1(e->versions[1].sha1))
1505 leaf->tree = dup_tree_content(e->tree);
1506 else
1507 leaf->tree = NULL;
1508 return 1;
1509 }
1510 if (!S_ISDIR(e->versions[1].mode))
1511 return 0;
1512 if (!e->tree)
1513 load_tree(e);
1514 return tree_content_get(e, slash1 + 1, leaf);
1515 }
1516 }
1517 return 0;
1518 }
1520 static int update_branch(struct branch *b)
1521 {
1522 static const char *msg = "fast-import";
1523 struct ref_lock *lock;
1524 unsigned char old_sha1[20];
1526 if (is_null_sha1(b->sha1))
1527 return 0;
1528 if (read_ref(b->name, old_sha1))
1529 hashclr(old_sha1);
1530 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1531 if (!lock)
1532 return error("Unable to lock %s", b->name);
1533 if (!force_update && !is_null_sha1(old_sha1)) {
1534 struct commit *old_cmit, *new_cmit;
1536 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1537 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1538 if (!old_cmit || !new_cmit) {
1539 unlock_ref(lock);
1540 return error("Branch %s is missing commits.", b->name);
1541 }
1543 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1544 unlock_ref(lock);
1545 warning("Not updating %s"
1546 " (new tip %s does not contain %s)",
1547 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1548 return -1;
1549 }
1550 }
1551 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1552 return error("Unable to update %s", b->name);
1553 return 0;
1554 }
1556 static void dump_branches(void)
1557 {
1558 unsigned int i;
1559 struct branch *b;
1561 for (i = 0; i < branch_table_sz; i++) {
1562 for (b = branch_table[i]; b; b = b->table_next_branch)
1563 failure |= update_branch(b);
1564 }
1565 }
1567 static void dump_tags(void)
1568 {
1569 static const char *msg = "fast-import";
1570 struct tag *t;
1571 struct ref_lock *lock;
1572 char ref_name[PATH_MAX];
1574 for (t = first_tag; t; t = t->next_tag) {
1575 sprintf(ref_name, "tags/%s", t->name);
1576 lock = lock_ref_sha1(ref_name, NULL);
1577 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1578 failure |= error("Unable to update %s", ref_name);
1579 }
1580 }
1582 static void dump_marks_helper(FILE *f,
1583 uintmax_t base,
1584 struct mark_set *m)
1585 {
1586 uintmax_t k;
1587 if (m->shift) {
1588 for (k = 0; k < 1024; k++) {
1589 if (m->data.sets[k])
1590 dump_marks_helper(f, (base + k) << m->shift,
1591 m->data.sets[k]);
1592 }
1593 } else {
1594 for (k = 0; k < 1024; k++) {
1595 if (m->data.marked[k])
1596 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1597 sha1_to_hex(m->data.marked[k]->sha1));
1598 }
1599 }
1600 }
1602 static void dump_marks(void)
1603 {
1604 static struct lock_file mark_lock;
1605 int mark_fd;
1606 FILE *f;
1608 if (!export_marks_file)
1609 return;
1611 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1612 if (mark_fd < 0) {
1613 failure |= error("Unable to write marks file %s: %s",
1614 export_marks_file, strerror(errno));
1615 return;
1616 }
1618 f = fdopen(mark_fd, "w");
1619 if (!f) {
1620 int saved_errno = errno;
1621 rollback_lock_file(&mark_lock);
1622 failure |= error("Unable to write marks file %s: %s",
1623 export_marks_file, strerror(saved_errno));
1624 return;
1625 }
1627 /*
1628 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1629 * Assign -1 to the lock file descriptor so that commit_lock_file()
1630 * won't try to close() it.
1631 */
1632 mark_lock.fd = -1;
1634 dump_marks_helper(f, 0, marks);
1635 if (ferror(f) || fclose(f)) {
1636 int saved_errno = errno;
1637 rollback_lock_file(&mark_lock);
1638 failure |= error("Unable to write marks file %s: %s",
1639 export_marks_file, strerror(saved_errno));
1640 return;
1641 }
1643 if (commit_lock_file(&mark_lock)) {
1644 int saved_errno = errno;
1645 rollback_lock_file(&mark_lock);
1646 failure |= error("Unable to commit marks file %s: %s",
1647 export_marks_file, strerror(saved_errno));
1648 return;
1649 }
1650 }
1652 static void read_marks(void)
1653 {
1654 char line[512];
1655 FILE *f = fopen(import_marks_file, "r");
1656 if (!f)
1657 die_errno("cannot read '%s'", import_marks_file);
1658 while (fgets(line, sizeof(line), f)) {
1659 uintmax_t mark;
1660 char *end;
1661 unsigned char sha1[20];
1662 struct object_entry *e;
1664 end = strchr(line, '\n');
1665 if (line[0] != ':' || !end)
1666 die("corrupt mark line: %s", line);
1667 *end = 0;
1668 mark = strtoumax(line + 1, &end, 10);
1669 if (!mark || end == line + 1
1670 || *end != ' ' || get_sha1(end + 1, sha1))
1671 die("corrupt mark line: %s", line);
1672 e = find_object(sha1);
1673 if (!e) {
1674 enum object_type type = sha1_object_info(sha1, NULL);
1675 if (type < 0)
1676 die("object not found: %s", sha1_to_hex(sha1));
1677 e = insert_object(sha1);
1678 e->type = type;
1679 e->pack_id = MAX_PACK_ID;
1680 e->offset = 1; /* just not zero! */
1681 }
1682 insert_mark(mark, e);
1683 }
1684 fclose(f);
1685 }
1688 static int read_next_command(void)
1689 {
1690 static int stdin_eof = 0;
1692 if (stdin_eof) {
1693 unread_command_buf = 0;
1694 return EOF;
1695 }
1697 do {
1698 if (unread_command_buf) {
1699 unread_command_buf = 0;
1700 } else {
1701 struct recent_command *rc;
1703 strbuf_detach(&command_buf, NULL);
1704 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1705 if (stdin_eof)
1706 return EOF;
1708 if (!seen_data_command
1709 && prefixcmp(command_buf.buf, "feature ")) {
1710 seen_data_command = 1;
1711 }
1713 rc = rc_free;
1714 if (rc)
1715 rc_free = rc->next;
1716 else {
1717 rc = cmd_hist.next;
1718 cmd_hist.next = rc->next;
1719 cmd_hist.next->prev = &cmd_hist;
1720 free(rc->buf);
1721 }
1723 rc->buf = command_buf.buf;
1724 rc->prev = cmd_tail;
1725 rc->next = cmd_hist.prev;
1726 rc->prev->next = rc;
1727 cmd_tail = rc;
1728 }
1729 } while (command_buf.buf[0] == '#');
1731 return 0;
1732 }
1734 static void skip_optional_lf(void)
1735 {
1736 int term_char = fgetc(stdin);
1737 if (term_char != '\n' && term_char != EOF)
1738 ungetc(term_char, stdin);
1739 }
1741 static void parse_mark(void)
1742 {
1743 if (!prefixcmp(command_buf.buf, "mark :")) {
1744 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1745 read_next_command();
1746 }
1747 else
1748 next_mark = 0;
1749 }
1751 static void parse_data(struct strbuf *sb)
1752 {
1753 strbuf_reset(sb);
1755 if (prefixcmp(command_buf.buf, "data "))
1756 die("Expected 'data n' command, found: %s", command_buf.buf);
1758 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1759 char *term = xstrdup(command_buf.buf + 5 + 2);
1760 size_t term_len = command_buf.len - 5 - 2;
1762 strbuf_detach(&command_buf, NULL);
1763 for (;;) {
1764 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1765 die("EOF in data (terminator '%s' not found)", term);
1766 if (term_len == command_buf.len
1767 && !strcmp(term, command_buf.buf))
1768 break;
1769 strbuf_addbuf(sb, &command_buf);
1770 strbuf_addch(sb, '\n');
1771 }
1772 free(term);
1773 }
1774 else {
1775 size_t n = 0, length;
1777 length = strtoul(command_buf.buf + 5, NULL, 10);
1779 while (n < length) {
1780 size_t s = strbuf_fread(sb, length - n, stdin);
1781 if (!s && feof(stdin))
1782 die("EOF in data (%lu bytes remaining)",
1783 (unsigned long)(length - n));
1784 n += s;
1785 }
1786 }
1788 skip_optional_lf();
1789 }
1791 static int validate_raw_date(const char *src, char *result, int maxlen)
1792 {
1793 const char *orig_src = src;
1794 char *endp;
1795 unsigned long num;
1797 errno = 0;
1799 num = strtoul(src, &endp, 10);
1800 /* NEEDSWORK: perhaps check for reasonable values? */
1801 if (errno || endp == src || *endp != ' ')
1802 return -1;
1804 src = endp + 1;
1805 if (*src != '-' && *src != '+')
1806 return -1;
1808 num = strtoul(src + 1, &endp, 10);
1809 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1810 1400 < num)
1811 return -1;
1813 strcpy(result, orig_src);
1814 return 0;
1815 }
1817 static char *parse_ident(const char *buf)
1818 {
1819 const char *gt;
1820 size_t name_len;
1821 char *ident;
1823 gt = strrchr(buf, '>');
1824 if (!gt)
1825 die("Missing > in ident string: %s", buf);
1826 gt++;
1827 if (*gt != ' ')
1828 die("Missing space after > in ident string: %s", buf);
1829 gt++;
1830 name_len = gt - buf;
1831 ident = xmalloc(name_len + 24);
1832 strncpy(ident, buf, name_len);
1834 switch (whenspec) {
1835 case WHENSPEC_RAW:
1836 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1837 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1838 break;
1839 case WHENSPEC_RFC2822:
1840 if (parse_date(gt, ident + name_len, 24) < 0)
1841 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1842 break;
1843 case WHENSPEC_NOW:
1844 if (strcmp("now", gt))
1845 die("Date in ident must be 'now': %s", buf);
1846 datestamp(ident + name_len, 24);
1847 break;
1848 }
1850 return ident;
1851 }
1853 static void parse_new_blob(void)
1854 {
1855 static struct strbuf buf = STRBUF_INIT;
1857 read_next_command();
1858 parse_mark();
1859 parse_data(&buf);
1860 store_object(OBJ_BLOB, &buf, &last_blob, NULL, next_mark);
1861 }
1863 static void unload_one_branch(void)
1864 {
1865 while (cur_active_branches
1866 && cur_active_branches >= max_active_branches) {
1867 uintmax_t min_commit = ULONG_MAX;
1868 struct branch *e, *l = NULL, *p = NULL;
1870 for (e = active_branches; e; e = e->active_next_branch) {
1871 if (e->last_commit < min_commit) {
1872 p = l;
1873 min_commit = e->last_commit;
1874 }
1875 l = e;
1876 }
1878 if (p) {
1879 e = p->active_next_branch;
1880 p->active_next_branch = e->active_next_branch;
1881 } else {
1882 e = active_branches;
1883 active_branches = e->active_next_branch;
1884 }
1885 e->active = 0;
1886 e->active_next_branch = NULL;
1887 if (e->branch_tree.tree) {
1888 release_tree_content_recursive(e->branch_tree.tree);
1889 e->branch_tree.tree = NULL;
1890 }
1891 cur_active_branches--;
1892 }
1893 }
1895 static void load_branch(struct branch *b)
1896 {
1897 load_tree(&b->branch_tree);
1898 if (!b->active) {
1899 b->active = 1;
1900 b->active_next_branch = active_branches;
1901 active_branches = b;
1902 cur_active_branches++;
1903 branch_load_count++;
1904 }
1905 }
1907 static void file_change_m(struct branch *b)
1908 {
1909 const char *p = command_buf.buf + 2;
1910 static struct strbuf uq = STRBUF_INIT;
1911 const char *endp;
1912 struct object_entry *oe = oe;
1913 unsigned char sha1[20];
1914 uint16_t mode, inline_data = 0;
1916 p = get_mode(p, &mode);
1917 if (!p)
1918 die("Corrupt mode: %s", command_buf.buf);
1919 switch (mode) {
1920 case 0644:
1921 case 0755:
1922 mode |= S_IFREG;
1923 case S_IFREG | 0644:
1924 case S_IFREG | 0755:
1925 case S_IFLNK:
1926 case S_IFGITLINK:
1927 /* ok */
1928 break;
1929 default:
1930 die("Corrupt mode: %s", command_buf.buf);
1931 }
1933 if (*p == ':') {
1934 char *x;
1935 oe = find_mark(strtoumax(p + 1, &x, 10));
1936 hashcpy(sha1, oe->sha1);
1937 p = x;
1938 } else if (!prefixcmp(p, "inline")) {
1939 inline_data = 1;
1940 p += 6;
1941 } else {
1942 if (get_sha1_hex(p, sha1))
1943 die("Invalid SHA1: %s", command_buf.buf);
1944 oe = find_object(sha1);
1945 p += 40;
1946 }
1947 if (*p++ != ' ')
1948 die("Missing space after SHA1: %s", command_buf.buf);
1950 strbuf_reset(&uq);
1951 if (!unquote_c_style(&uq, p, &endp)) {
1952 if (*endp)
1953 die("Garbage after path in: %s", command_buf.buf);
1954 p = uq.buf;
1955 }
1957 if (S_ISGITLINK(mode)) {
1958 if (inline_data)
1959 die("Git links cannot be specified 'inline': %s",
1960 command_buf.buf);
1961 else if (oe) {
1962 if (oe->type != OBJ_COMMIT)
1963 die("Not a commit (actually a %s): %s",
1964 typename(oe->type), command_buf.buf);
1965 }
1966 /*
1967 * Accept the sha1 without checking; it expected to be in
1968 * another repository.
1969 */
1970 } else if (inline_data) {
1971 static struct strbuf buf = STRBUF_INIT;
1973 if (p != uq.buf) {
1974 strbuf_addstr(&uq, p);
1975 p = uq.buf;
1976 }
1977 read_next_command();
1978 parse_data(&buf);
1979 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
1980 } else if (oe) {
1981 if (oe->type != OBJ_BLOB)
1982 die("Not a blob (actually a %s): %s",
1983 typename(oe->type), command_buf.buf);
1984 } else {
1985 enum object_type type = sha1_object_info(sha1, NULL);
1986 if (type < 0)
1987 die("Blob not found: %s", command_buf.buf);
1988 if (type != OBJ_BLOB)
1989 die("Not a blob (actually a %s): %s",
1990 typename(type), command_buf.buf);
1991 }
1993 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
1994 }
1996 static void file_change_d(struct branch *b)
1997 {
1998 const char *p = command_buf.buf + 2;
1999 static struct strbuf uq = STRBUF_INIT;
2000 const char *endp;
2002 strbuf_reset(&uq);
2003 if (!unquote_c_style(&uq, p, &endp)) {
2004 if (*endp)
2005 die("Garbage after path in: %s", command_buf.buf);
2006 p = uq.buf;
2007 }
2008 tree_content_remove(&b->branch_tree, p, NULL);
2009 }
2011 static void file_change_cr(struct branch *b, int rename)
2012 {
2013 const char *s, *d;
2014 static struct strbuf s_uq = STRBUF_INIT;
2015 static struct strbuf d_uq = STRBUF_INIT;
2016 const char *endp;
2017 struct tree_entry leaf;
2019 s = command_buf.buf + 2;
2020 strbuf_reset(&s_uq);
2021 if (!unquote_c_style(&s_uq, s, &endp)) {
2022 if (*endp != ' ')
2023 die("Missing space after source: %s", command_buf.buf);
2024 } else {
2025 endp = strchr(s, ' ');
2026 if (!endp)
2027 die("Missing space after source: %s", command_buf.buf);
2028 strbuf_add(&s_uq, s, endp - s);
2029 }
2030 s = s_uq.buf;
2032 endp++;
2033 if (!*endp)
2034 die("Missing dest: %s", command_buf.buf);
2036 d = endp;
2037 strbuf_reset(&d_uq);
2038 if (!unquote_c_style(&d_uq, d, &endp)) {
2039 if (*endp)
2040 die("Garbage after dest in: %s", command_buf.buf);
2041 d = d_uq.buf;
2042 }
2044 memset(&leaf, 0, sizeof(leaf));
2045 if (rename)
2046 tree_content_remove(&b->branch_tree, s, &leaf);
2047 else
2048 tree_content_get(&b->branch_tree, s, &leaf);
2049 if (!leaf.versions[1].mode)
2050 die("Path %s not in branch", s);
2051 tree_content_set(&b->branch_tree, d,
2052 leaf.versions[1].sha1,
2053 leaf.versions[1].mode,
2054 leaf.tree);
2055 }
2057 static void note_change_n(struct branch *b)
2058 {
2059 const char *p = command_buf.buf + 2;
2060 static struct strbuf uq = STRBUF_INIT;
2061 struct object_entry *oe = oe;
2062 struct branch *s;
2063 unsigned char sha1[20], commit_sha1[20];
2064 uint16_t inline_data = 0;
2066 /* <dataref> or 'inline' */
2067 if (*p == ':') {
2068 char *x;
2069 oe = find_mark(strtoumax(p + 1, &x, 10));
2070 hashcpy(sha1, oe->sha1);
2071 p = x;
2072 } else if (!prefixcmp(p, "inline")) {
2073 inline_data = 1;
2074 p += 6;
2075 } else {
2076 if (get_sha1_hex(p, sha1))
2077 die("Invalid SHA1: %s", command_buf.buf);
2078 oe = find_object(sha1);
2079 p += 40;
2080 }
2081 if (*p++ != ' ')
2082 die("Missing space after SHA1: %s", command_buf.buf);
2084 /* <committish> */
2085 s = lookup_branch(p);
2086 if (s) {
2087 hashcpy(commit_sha1, s->sha1);
2088 } else if (*p == ':') {
2089 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2090 struct object_entry *commit_oe = find_mark(commit_mark);
2091 if (commit_oe->type != OBJ_COMMIT)
2092 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2093 hashcpy(commit_sha1, commit_oe->sha1);
2094 } else if (!get_sha1(p, commit_sha1)) {
2095 unsigned long size;
2096 char *buf = read_object_with_reference(commit_sha1,
2097 commit_type, &size, commit_sha1);
2098 if (!buf || size < 46)
2099 die("Not a valid commit: %s", p);
2100 free(buf);
2101 } else
2102 die("Invalid ref name or SHA1 expression: %s", p);
2104 if (inline_data) {
2105 static struct strbuf buf = STRBUF_INIT;
2107 if (p != uq.buf) {
2108 strbuf_addstr(&uq, p);
2109 p = uq.buf;
2110 }
2111 read_next_command();
2112 parse_data(&buf);
2113 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
2114 } else if (oe) {
2115 if (oe->type != OBJ_BLOB)
2116 die("Not a blob (actually a %s): %s",
2117 typename(oe->type), command_buf.buf);
2118 } else {
2119 enum object_type type = sha1_object_info(sha1, NULL);
2120 if (type < 0)
2121 die("Blob not found: %s", command_buf.buf);
2122 if (type != OBJ_BLOB)
2123 die("Not a blob (actually a %s): %s",
2124 typename(type), command_buf.buf);
2125 }
2127 tree_content_set(&b->branch_tree, sha1_to_hex(commit_sha1), sha1,
2128 S_IFREG | 0644, NULL);
2129 }
2131 static void file_change_deleteall(struct branch *b)
2132 {
2133 release_tree_content_recursive(b->branch_tree.tree);
2134 hashclr(b->branch_tree.versions[0].sha1);
2135 hashclr(b->branch_tree.versions[1].sha1);
2136 load_tree(&b->branch_tree);
2137 }
2139 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2140 {
2141 if (!buf || size < 46)
2142 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2143 if (memcmp("tree ", buf, 5)
2144 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2145 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2146 hashcpy(b->branch_tree.versions[0].sha1,
2147 b->branch_tree.versions[1].sha1);
2148 }
2150 static void parse_from_existing(struct branch *b)
2151 {
2152 if (is_null_sha1(b->sha1)) {
2153 hashclr(b->branch_tree.versions[0].sha1);
2154 hashclr(b->branch_tree.versions[1].sha1);
2155 } else {
2156 unsigned long size;
2157 char *buf;
2159 buf = read_object_with_reference(b->sha1,
2160 commit_type, &size, b->sha1);
2161 parse_from_commit(b, buf, size);
2162 free(buf);
2163 }
2164 }
2166 static int parse_from(struct branch *b)
2167 {
2168 const char *from;
2169 struct branch *s;
2171 if (prefixcmp(command_buf.buf, "from "))
2172 return 0;
2174 if (b->branch_tree.tree) {
2175 release_tree_content_recursive(b->branch_tree.tree);
2176 b->branch_tree.tree = NULL;
2177 }
2179 from = strchr(command_buf.buf, ' ') + 1;
2180 s = lookup_branch(from);
2181 if (b == s)
2182 die("Can't create a branch from itself: %s", b->name);
2183 else if (s) {
2184 unsigned char *t = s->branch_tree.versions[1].sha1;
2185 hashcpy(b->sha1, s->sha1);
2186 hashcpy(b->branch_tree.versions[0].sha1, t);
2187 hashcpy(b->branch_tree.versions[1].sha1, t);
2188 } else if (*from == ':') {
2189 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2190 struct object_entry *oe = find_mark(idnum);
2191 if (oe->type != OBJ_COMMIT)
2192 die("Mark :%" PRIuMAX " not a commit", idnum);
2193 hashcpy(b->sha1, oe->sha1);
2194 if (oe->pack_id != MAX_PACK_ID) {
2195 unsigned long size;
2196 char *buf = gfi_unpack_entry(oe, &size);
2197 parse_from_commit(b, buf, size);
2198 free(buf);
2199 } else
2200 parse_from_existing(b);
2201 } else if (!get_sha1(from, b->sha1))
2202 parse_from_existing(b);
2203 else
2204 die("Invalid ref name or SHA1 expression: %s", from);
2206 read_next_command();
2207 return 1;
2208 }
2210 static struct hash_list *parse_merge(unsigned int *count)
2211 {
2212 struct hash_list *list = NULL, *n, *e = e;
2213 const char *from;
2214 struct branch *s;
2216 *count = 0;
2217 while (!prefixcmp(command_buf.buf, "merge ")) {
2218 from = strchr(command_buf.buf, ' ') + 1;
2219 n = xmalloc(sizeof(*n));
2220 s = lookup_branch(from);
2221 if (s)
2222 hashcpy(n->sha1, s->sha1);
2223 else if (*from == ':') {
2224 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2225 struct object_entry *oe = find_mark(idnum);
2226 if (oe->type != OBJ_COMMIT)
2227 die("Mark :%" PRIuMAX " not a commit", idnum);
2228 hashcpy(n->sha1, oe->sha1);
2229 } else if (!get_sha1(from, n->sha1)) {
2230 unsigned long size;
2231 char *buf = read_object_with_reference(n->sha1,
2232 commit_type, &size, n->sha1);
2233 if (!buf || size < 46)
2234 die("Not a valid commit: %s", from);
2235 free(buf);
2236 } else
2237 die("Invalid ref name or SHA1 expression: %s", from);
2239 n->next = NULL;
2240 if (list)
2241 e->next = n;
2242 else
2243 list = n;
2244 e = n;
2245 (*count)++;
2246 read_next_command();
2247 }
2248 return list;
2249 }
2251 static void parse_new_commit(void)
2252 {
2253 static struct strbuf msg = STRBUF_INIT;
2254 struct branch *b;
2255 char *sp;
2256 char *author = NULL;
2257 char *committer = NULL;
2258 struct hash_list *merge_list = NULL;
2259 unsigned int merge_count;
2261 /* Obtain the branch name from the rest of our command */
2262 sp = strchr(command_buf.buf, ' ') + 1;
2263 b = lookup_branch(sp);
2264 if (!b)
2265 b = new_branch(sp);
2267 read_next_command();
2268 parse_mark();
2269 if (!prefixcmp(command_buf.buf, "author ")) {
2270 author = parse_ident(command_buf.buf + 7);
2271 read_next_command();
2272 }
2273 if (!prefixcmp(command_buf.buf, "committer ")) {
2274 committer = parse_ident(command_buf.buf + 10);
2275 read_next_command();
2276 }
2277 if (!committer)
2278 die("Expected committer but didn't get one");
2279 parse_data(&msg);
2280 read_next_command();
2281 parse_from(b);
2282 merge_list = parse_merge(&merge_count);
2284 /* ensure the branch is active/loaded */
2285 if (!b->branch_tree.tree || !max_active_branches) {
2286 unload_one_branch();
2287 load_branch(b);
2288 }
2290 /* file_change* */
2291 while (command_buf.len > 0) {
2292 if (!prefixcmp(command_buf.buf, "M "))
2293 file_change_m(b);
2294 else if (!prefixcmp(command_buf.buf, "D "))
2295 file_change_d(b);
2296 else if (!prefixcmp(command_buf.buf, "R "))
2297 file_change_cr(b, 1);
2298 else if (!prefixcmp(command_buf.buf, "C "))
2299 file_change_cr(b, 0);
2300 else if (!prefixcmp(command_buf.buf, "N "))
2301 note_change_n(b);
2302 else if (!strcmp("deleteall", command_buf.buf))
2303 file_change_deleteall(b);
2304 else {
2305 unread_command_buf = 1;
2306 break;
2307 }
2308 if (read_next_command() == EOF)
2309 break;
2310 }
2312 /* build the tree and the commit */
2313 store_tree(&b->branch_tree);
2314 hashcpy(b->branch_tree.versions[0].sha1,
2315 b->branch_tree.versions[1].sha1);
2317 strbuf_reset(&new_data);
2318 strbuf_addf(&new_data, "tree %s\n",
2319 sha1_to_hex(b->branch_tree.versions[1].sha1));
2320 if (!is_null_sha1(b->sha1))
2321 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2322 while (merge_list) {
2323 struct hash_list *next = merge_list->next;
2324 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2325 free(merge_list);
2326 merge_list = next;
2327 }
2328 strbuf_addf(&new_data,
2329 "author %s\n"
2330 "committer %s\n"
2331 "\n",
2332 author ? author : committer, committer);
2333 strbuf_addbuf(&new_data, &msg);
2334 free(author);
2335 free(committer);
2337 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2338 b->pack_id = pack_id;
2339 b->last_commit = object_count_by_type[OBJ_COMMIT];
2340 }
2342 static void parse_new_tag(void)
2343 {
2344 static struct strbuf msg = STRBUF_INIT;
2345 char *sp;
2346 const char *from;
2347 char *tagger;
2348 struct branch *s;
2349 struct tag *t;
2350 uintmax_t from_mark = 0;
2351 unsigned char sha1[20];
2353 /* Obtain the new tag name from the rest of our command */
2354 sp = strchr(command_buf.buf, ' ') + 1;
2355 t = pool_alloc(sizeof(struct tag));
2356 t->next_tag = NULL;
2357 t->name = pool_strdup(sp);
2358 if (last_tag)
2359 last_tag->next_tag = t;
2360 else
2361 first_tag = t;
2362 last_tag = t;
2363 read_next_command();
2365 /* from ... */
2366 if (prefixcmp(command_buf.buf, "from "))
2367 die("Expected from command, got %s", command_buf.buf);
2368 from = strchr(command_buf.buf, ' ') + 1;
2369 s = lookup_branch(from);
2370 if (s) {
2371 hashcpy(sha1, s->sha1);
2372 } else if (*from == ':') {
2373 struct object_entry *oe;
2374 from_mark = strtoumax(from + 1, NULL, 10);
2375 oe = find_mark(from_mark);
2376 if (oe->type != OBJ_COMMIT)
2377 die("Mark :%" PRIuMAX " not a commit", from_mark);
2378 hashcpy(sha1, oe->sha1);
2379 } else if (!get_sha1(from, sha1)) {
2380 unsigned long size;
2381 char *buf;
2383 buf = read_object_with_reference(sha1,
2384 commit_type, &size, sha1);
2385 if (!buf || size < 46)
2386 die("Not a valid commit: %s", from);
2387 free(buf);
2388 } else
2389 die("Invalid ref name or SHA1 expression: %s", from);
2390 read_next_command();
2392 /* tagger ... */
2393 if (!prefixcmp(command_buf.buf, "tagger ")) {
2394 tagger = parse_ident(command_buf.buf + 7);
2395 read_next_command();
2396 } else
2397 tagger = NULL;
2399 /* tag payload/message */
2400 parse_data(&msg);
2402 /* build the tag object */
2403 strbuf_reset(&new_data);
2405 strbuf_addf(&new_data,
2406 "object %s\n"
2407 "type %s\n"
2408 "tag %s\n",
2409 sha1_to_hex(sha1), commit_type, t->name);
2410 if (tagger)
2411 strbuf_addf(&new_data,
2412 "tagger %s\n", tagger);
2413 strbuf_addch(&new_data, '\n');
2414 strbuf_addbuf(&new_data, &msg);
2415 free(tagger);
2417 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2418 t->pack_id = MAX_PACK_ID;
2419 else
2420 t->pack_id = pack_id;
2421 }
2423 static void parse_reset_branch(void)
2424 {
2425 struct branch *b;
2426 char *sp;
2428 /* Obtain the branch name from the rest of our command */
2429 sp = strchr(command_buf.buf, ' ') + 1;
2430 b = lookup_branch(sp);
2431 if (b) {
2432 hashclr(b->sha1);
2433 hashclr(b->branch_tree.versions[0].sha1);
2434 hashclr(b->branch_tree.versions[1].sha1);
2435 if (b->branch_tree.tree) {
2436 release_tree_content_recursive(b->branch_tree.tree);
2437 b->branch_tree.tree = NULL;
2438 }
2439 }
2440 else
2441 b = new_branch(sp);
2442 read_next_command();
2443 parse_from(b);
2444 if (command_buf.len > 0)
2445 unread_command_buf = 1;
2446 }
2448 static void parse_checkpoint(void)
2449 {
2450 if (object_count) {
2451 cycle_packfile();
2452 dump_branches();
2453 dump_tags();
2454 dump_marks();
2455 }
2456 skip_optional_lf();
2457 }
2459 static void parse_progress(void)
2460 {
2461 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2462 fputc('\n', stdout);
2463 fflush(stdout);
2464 skip_optional_lf();
2465 }
2467 static void option_import_marks(const char *marks)
2468 {
2469 import_marks_file = xstrdup(marks);
2470 }
2472 static void option_date_format(const char *fmt)
2473 {
2474 if (!strcmp(fmt, "raw"))
2475 whenspec = WHENSPEC_RAW;
2476 else if (!strcmp(fmt, "rfc2822"))
2477 whenspec = WHENSPEC_RFC2822;
2478 else if (!strcmp(fmt, "now"))
2479 whenspec = WHENSPEC_NOW;
2480 else
2481 die("unknown --date-format argument %s", fmt);
2482 }
2484 static void option_max_pack_size(const char *packsize)
2485 {
2486 max_packsize = strtoumax(packsize, NULL, 0) * 1024 * 1024;
2487 }
2489 static void option_depth(const char *depth)
2490 {
2491 max_depth = strtoul(depth, NULL, 0);
2492 if (max_depth > MAX_DEPTH)
2493 die("--depth cannot exceed %u", MAX_DEPTH);
2494 }
2496 static void option_active_branches(const char *branches)
2497 {
2498 max_active_branches = strtoul(branches, NULL, 0);
2499 }
2501 static void option_export_marks(const char *marks)
2502 {
2503 export_marks_file = xstrdup(marks);
2504 }
2506 static void option_export_pack_edges(const char *edges)
2507 {
2508 if (pack_edges)
2509 fclose(pack_edges);
2510 pack_edges = fopen(edges, "a");
2511 if (!pack_edges)
2512 die_errno("Cannot open '%s'", edges);
2513 }
2515 static void parse_one_option(const char *option)
2516 {
2517 if (!prefixcmp(option, "date-format=")) {
2518 option_date_format(option + 12);
2519 } else if (!prefixcmp(option, "max-pack-size=")) {
2520 option_max_pack_size(option + 14);
2521 } else if (!prefixcmp(option, "depth=")) {
2522 option_depth(option + 6);
2523 } else if (!prefixcmp(option, "active-branches=")) {
2524 option_active_branches(option + 16);
2525 } else if (!prefixcmp(option, "import-marks=")) {
2526 option_import_marks(option + 13);
2527 } else if (!prefixcmp(option, "export-marks=")) {
2528 option_export_marks(option + 13);
2529 } else if (!prefixcmp(option, "export-pack-edges=")) {
2530 option_export_pack_edges(option + 18);
2531 } else if (!prefixcmp(option, "force")) {
2532 force_update = 1;
2533 } else if (!prefixcmp(option, "quiet")) {
2534 show_stats = 0;
2535 } else if (!prefixcmp(option, "stats")) {
2536 show_stats = 1;
2537 } else {
2538 die("Unsupported option: %s", option);
2539 }
2540 }
2542 static int parse_one_feature(const char *feature)
2543 {
2544 if (!prefixcmp(feature, "date-format=")) {
2545 option_date_format(feature + 12);
2546 } else if (!prefixcmp(feature, "import-marks=")) {
2547 option_import_marks(feature + 13);
2548 } else if (!prefixcmp(feature, "export-marks=")) {
2549 option_export_marks(feature + 13);
2550 } else if (!prefixcmp(feature, "force")) {
2551 force_update = 1;
2552 } else {
2553 return 0;
2554 }
2556 return 1;
2557 }
2559 static void parse_feature(void)
2560 {
2561 char *feature = command_buf.buf + 8;
2563 if (seen_data_command)
2564 die("Got feature command '%s' after data command", feature);
2566 if (parse_one_feature(feature))
2567 return;
2569 die("This version of fast-import does not support feature %s.", feature);
2570 }
2572 static int git_pack_config(const char *k, const char *v, void *cb)
2573 {
2574 if (!strcmp(k, "pack.depth")) {
2575 max_depth = git_config_int(k, v);
2576 if (max_depth > MAX_DEPTH)
2577 max_depth = MAX_DEPTH;
2578 return 0;
2579 }
2580 if (!strcmp(k, "pack.compression")) {
2581 int level = git_config_int(k, v);
2582 if (level == -1)
2583 level = Z_DEFAULT_COMPRESSION;
2584 else if (level < 0 || level > Z_BEST_COMPRESSION)
2585 die("bad pack compression level %d", level);
2586 pack_compression_level = level;
2587 pack_compression_seen = 1;
2588 return 0;
2589 }
2590 return git_default_config(k, v, cb);
2591 }
2593 static const char fast_import_usage[] =
2594 "git fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2596 int main(int argc, const char **argv)
2597 {
2598 unsigned int i;
2600 git_extract_argv0_path(argv[0]);
2602 if (argc == 2 && !strcmp(argv[1], "-h"))
2603 usage(fast_import_usage);
2605 setup_git_directory();
2606 git_config(git_pack_config, NULL);
2607 if (!pack_compression_seen && core_compression_seen)
2608 pack_compression_level = core_compression_level;
2610 alloc_objects(object_entry_alloc);
2611 strbuf_init(&command_buf, 0);
2612 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2613 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2614 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2615 marks = pool_calloc(1, sizeof(struct mark_set));
2617 for (i = 1; i < argc; i++) {
2618 const char *a = argv[i];
2620 if (*a != '-' || !strcmp(a, "--"))
2621 break;
2623 parse_one_option(a + 2);
2624 }
2625 if (i != argc)
2626 usage(fast_import_usage);
2627 if (import_marks_file)
2628 read_marks();
2630 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2631 for (i = 0; i < (cmd_save - 1); i++)
2632 rc_free[i].next = &rc_free[i + 1];
2633 rc_free[cmd_save - 1].next = NULL;
2635 prepare_packed_git();
2636 start_packfile();
2637 set_die_routine(die_nicely);
2638 while (read_next_command() != EOF) {
2639 if (!strcmp("blob", command_buf.buf))
2640 parse_new_blob();
2641 else if (!prefixcmp(command_buf.buf, "commit "))
2642 parse_new_commit();
2643 else if (!prefixcmp(command_buf.buf, "tag "))
2644 parse_new_tag();
2645 else if (!prefixcmp(command_buf.buf, "reset "))
2646 parse_reset_branch();
2647 else if (!strcmp("checkpoint", command_buf.buf))
2648 parse_checkpoint();
2649 else if (!prefixcmp(command_buf.buf, "progress "))
2650 parse_progress();
2651 else if (!prefixcmp(command_buf.buf, "feature "))
2652 parse_feature();
2653 else
2654 die("Unsupported command: %s", command_buf.buf);
2655 }
2656 end_packfile();
2658 dump_branches();
2659 dump_tags();
2660 unkeep_all_packs();
2661 dump_marks();
2663 if (pack_edges)
2664 fclose(pack_edges);
2666 if (show_stats) {
2667 uintmax_t total_count = 0, duplicate_count = 0;
2668 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2669 total_count += object_count_by_type[i];
2670 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2671 duplicate_count += duplicate_count_by_type[i];
2673 fprintf(stderr, "%s statistics:\n", argv[0]);
2674 fprintf(stderr, "---------------------------------------------------------------------\n");
2675 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2676 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2677 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]);
2678 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]);
2679 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]);
2680 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]);
2681 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2682 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2683 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2684 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2685 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2686 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2687 fprintf(stderr, "---------------------------------------------------------------------\n");
2688 pack_report();
2689 fprintf(stderr, "---------------------------------------------------------------------\n");
2690 fprintf(stderr, "\n");
2691 }
2693 return failure ? 1 : 0;
2694 }