fast-import: allow for multiple --import-marks= arguments
[git/dscho.git] / fast-import.c
blob4c3406e9f01fbd1e9627d92bd2ef050c99ae5b31
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
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.
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.
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.
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'
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.
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.
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.
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).
143 comment ::= '#' not_lf* lf;
144 not_lf ::= # Any byte that is not ASCII newline (LF);
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
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];
175 struct object_entry_pool
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 */
183 struct mark_set
185 union {
186 struct object_entry *marked[1024];
187 struct mark_set *sets[1024];
188 } data;
189 unsigned int shift;
192 struct last_object
194 struct strbuf data;
195 uint32_t offset;
196 unsigned int depth;
197 unsigned no_swap : 1;
200 struct mem_pool
202 struct mem_pool *next_pool;
203 char *next_free;
204 char *end;
205 uintmax_t space[FLEX_ARRAY]; /* more */
208 struct atom_str
210 struct atom_str *next_atom;
211 unsigned short str_len;
212 char str_dat[FLEX_ARRAY]; /* more */
215 struct tree_content;
216 struct tree_entry
218 struct tree_content *tree;
219 struct atom_str *name;
220 struct tree_entry_ms
222 uint16_t mode;
223 unsigned char sha1[20];
224 } versions[2];
227 struct tree_content
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 */
235 struct avail_tree_content
237 unsigned int entry_capacity; /* must match tree_content */
238 struct avail_tree_content *next_avail;
241 struct branch
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];
253 struct tag
255 struct tag *next_tag;
256 const char *name;
257 unsigned int pack_id;
258 unsigned char sha1[20];
261 struct hash_list
263 struct hash_list *next;
264 unsigned char sha1[20];
267 typedef enum {
268 WHENSPEC_RAW = 1,
269 WHENSPEC_RFC2822,
270 WHENSPEC_NOW,
271 } whenspec_type;
273 struct recent_command
275 struct recent_command *prev;
276 struct recent_command *next;
277 char *buf;
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;
299 static int global_argc;
300 static const char **global_argv;
302 /* Memory pools */
303 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
304 static size_t total_allocd;
305 static struct mem_pool *mem_pool;
307 /* Atom management */
308 static unsigned int atom_table_sz = 4451;
309 static unsigned int atom_cnt;
310 static struct atom_str **atom_table;
312 /* The .pack file being generated */
313 static unsigned int pack_id;
314 static struct packed_git *pack_data;
315 static struct packed_git **all_packs;
316 static unsigned long pack_size;
318 /* Table of objects we've written. */
319 static unsigned int object_entry_alloc = 5000;
320 static struct object_entry_pool *blocks;
321 static struct object_entry *object_table[1 << 16];
322 static struct mark_set *marks;
323 static const char *export_marks_file;
324 static const char *import_marks_file;
325 static int import_marks_file_from_stream;
327 /* Our last blob */
328 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
330 /* Tree management */
331 static unsigned int tree_entry_alloc = 1000;
332 static void *avail_tree_entry;
333 static unsigned int avail_tree_table_sz = 100;
334 static struct avail_tree_content **avail_tree_table;
335 static struct strbuf old_tree = STRBUF_INIT;
336 static struct strbuf new_tree = STRBUF_INIT;
338 /* Branch data */
339 static unsigned long max_active_branches = 5;
340 static unsigned long cur_active_branches;
341 static unsigned long branch_table_sz = 1039;
342 static struct branch **branch_table;
343 static struct branch *active_branches;
345 /* Tag data */
346 static struct tag *first_tag;
347 static struct tag *last_tag;
349 /* Input stream parsing */
350 static whenspec_type whenspec = WHENSPEC_RAW;
351 static struct strbuf command_buf = STRBUF_INIT;
352 static int unread_command_buf;
353 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
354 static struct recent_command *cmd_tail = &cmd_hist;
355 static struct recent_command *rc_free;
356 static unsigned int cmd_save = 100;
357 static uintmax_t next_mark;
358 static struct strbuf new_data = STRBUF_INIT;
359 static int seen_data_command;
361 static void parse_argv(void);
363 static void write_branch_report(FILE *rpt, struct branch *b)
365 fprintf(rpt, "%s:\n", b->name);
367 fprintf(rpt, " status :");
368 if (b->active)
369 fputs(" active", rpt);
370 if (b->branch_tree.tree)
371 fputs(" loaded", rpt);
372 if (is_null_sha1(b->branch_tree.versions[1].sha1))
373 fputs(" dirty", rpt);
374 fputc('\n', rpt);
376 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
377 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
378 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
379 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
381 fputs(" last pack : ", rpt);
382 if (b->pack_id < MAX_PACK_ID)
383 fprintf(rpt, "%u", b->pack_id);
384 fputc('\n', rpt);
386 fputc('\n', rpt);
389 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
391 static void write_crash_report(const char *err)
393 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
394 FILE *rpt = fopen(loc, "w");
395 struct branch *b;
396 unsigned long lu;
397 struct recent_command *rc;
399 if (!rpt) {
400 error("can't write crash report %s: %s", loc, strerror(errno));
401 return;
404 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
406 fprintf(rpt, "fast-import crash report:\n");
407 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
408 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
409 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
410 fputc('\n', rpt);
412 fputs("fatal: ", rpt);
413 fputs(err, rpt);
414 fputc('\n', rpt);
416 fputc('\n', rpt);
417 fputs("Most Recent Commands Before Crash\n", rpt);
418 fputs("---------------------------------\n", rpt);
419 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
420 if (rc->next == &cmd_hist)
421 fputs("* ", rpt);
422 else
423 fputs(" ", rpt);
424 fputs(rc->buf, rpt);
425 fputc('\n', rpt);
428 fputc('\n', rpt);
429 fputs("Active Branch LRU\n", rpt);
430 fputs("-----------------\n", rpt);
431 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
432 cur_active_branches,
433 max_active_branches);
434 fputc('\n', rpt);
435 fputs(" pos clock name\n", rpt);
436 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
437 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
438 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
439 ++lu, b->last_commit, b->name);
441 fputc('\n', rpt);
442 fputs("Inactive Branches\n", rpt);
443 fputs("-----------------\n", rpt);
444 for (lu = 0; lu < branch_table_sz; lu++) {
445 for (b = branch_table[lu]; b; b = b->table_next_branch)
446 write_branch_report(rpt, b);
449 if (first_tag) {
450 struct tag *tg;
451 fputc('\n', rpt);
452 fputs("Annotated Tags\n", rpt);
453 fputs("--------------\n", rpt);
454 for (tg = first_tag; tg; tg = tg->next_tag) {
455 fputs(sha1_to_hex(tg->sha1), rpt);
456 fputc(' ', rpt);
457 fputs(tg->name, rpt);
458 fputc('\n', rpt);
462 fputc('\n', rpt);
463 fputs("Marks\n", rpt);
464 fputs("-----\n", rpt);
465 if (export_marks_file)
466 fprintf(rpt, " exported to %s\n", export_marks_file);
467 else
468 dump_marks_helper(rpt, 0, marks);
470 fputc('\n', rpt);
471 fputs("-------------------\n", rpt);
472 fputs("END OF CRASH REPORT\n", rpt);
473 fclose(rpt);
476 static void end_packfile(void);
477 static void unkeep_all_packs(void);
478 static void dump_marks(void);
480 static NORETURN void die_nicely(const char *err, va_list params)
482 static int zombie;
483 char message[2 * PATH_MAX];
485 vsnprintf(message, sizeof(message), err, params);
486 fputs("fatal: ", stderr);
487 fputs(message, stderr);
488 fputc('\n', stderr);
490 if (!zombie) {
491 zombie = 1;
492 write_crash_report(message);
493 end_packfile();
494 unkeep_all_packs();
495 dump_marks();
497 exit(128);
500 static void alloc_objects(unsigned int cnt)
502 struct object_entry_pool *b;
504 b = xmalloc(sizeof(struct object_entry_pool)
505 + cnt * sizeof(struct object_entry));
506 b->next_pool = blocks;
507 b->next_free = b->entries;
508 b->end = b->entries + cnt;
509 blocks = b;
510 alloc_count += cnt;
513 static struct object_entry *new_object(unsigned char *sha1)
515 struct object_entry *e;
517 if (blocks->next_free == blocks->end)
518 alloc_objects(object_entry_alloc);
520 e = blocks->next_free++;
521 hashcpy(e->sha1, sha1);
522 return e;
525 static struct object_entry *find_object(unsigned char *sha1)
527 unsigned int h = sha1[0] << 8 | sha1[1];
528 struct object_entry *e;
529 for (e = object_table[h]; e; e = e->next)
530 if (!hashcmp(sha1, e->sha1))
531 return e;
532 return NULL;
535 static struct object_entry *insert_object(unsigned char *sha1)
537 unsigned int h = sha1[0] << 8 | sha1[1];
538 struct object_entry *e = object_table[h];
539 struct object_entry *p = NULL;
541 while (e) {
542 if (!hashcmp(sha1, e->sha1))
543 return e;
544 p = e;
545 e = e->next;
548 e = new_object(sha1);
549 e->next = NULL;
550 e->offset = 0;
551 if (p)
552 p->next = e;
553 else
554 object_table[h] = e;
555 return e;
558 static unsigned int hc_str(const char *s, size_t len)
560 unsigned int r = 0;
561 while (len-- > 0)
562 r = r * 31 + *s++;
563 return r;
566 static void *pool_alloc(size_t len)
568 struct mem_pool *p;
569 void *r;
571 /* round up to a 'uintmax_t' alignment */
572 if (len & (sizeof(uintmax_t) - 1))
573 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
575 for (p = mem_pool; p; p = p->next_pool)
576 if ((p->end - p->next_free >= len))
577 break;
579 if (!p) {
580 if (len >= (mem_pool_alloc/2)) {
581 total_allocd += len;
582 return xmalloc(len);
584 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
585 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
586 p->next_pool = mem_pool;
587 p->next_free = (char *) p->space;
588 p->end = p->next_free + mem_pool_alloc;
589 mem_pool = p;
592 r = p->next_free;
593 p->next_free += len;
594 return r;
597 static void *pool_calloc(size_t count, size_t size)
599 size_t len = count * size;
600 void *r = pool_alloc(len);
601 memset(r, 0, len);
602 return r;
605 static char *pool_strdup(const char *s)
607 char *r = pool_alloc(strlen(s) + 1);
608 strcpy(r, s);
609 return r;
612 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
614 struct mark_set *s = marks;
615 while ((idnum >> s->shift) >= 1024) {
616 s = pool_calloc(1, sizeof(struct mark_set));
617 s->shift = marks->shift + 10;
618 s->data.sets[0] = marks;
619 marks = s;
621 while (s->shift) {
622 uintmax_t i = idnum >> s->shift;
623 idnum -= i << s->shift;
624 if (!s->data.sets[i]) {
625 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
626 s->data.sets[i]->shift = s->shift - 10;
628 s = s->data.sets[i];
630 if (!s->data.marked[idnum])
631 marks_set_count++;
632 s->data.marked[idnum] = oe;
635 static struct object_entry *find_mark(uintmax_t idnum)
637 uintmax_t orig_idnum = idnum;
638 struct mark_set *s = marks;
639 struct object_entry *oe = NULL;
640 if ((idnum >> s->shift) < 1024) {
641 while (s && s->shift) {
642 uintmax_t i = idnum >> s->shift;
643 idnum -= i << s->shift;
644 s = s->data.sets[i];
646 if (s)
647 oe = s->data.marked[idnum];
649 if (!oe)
650 die("mark :%" PRIuMAX " not declared", orig_idnum);
651 return oe;
654 static struct atom_str *to_atom(const char *s, unsigned short len)
656 unsigned int hc = hc_str(s, len) % atom_table_sz;
657 struct atom_str *c;
659 for (c = atom_table[hc]; c; c = c->next_atom)
660 if (c->str_len == len && !strncmp(s, c->str_dat, len))
661 return c;
663 c = pool_alloc(sizeof(struct atom_str) + len + 1);
664 c->str_len = len;
665 strncpy(c->str_dat, s, len);
666 c->str_dat[len] = 0;
667 c->next_atom = atom_table[hc];
668 atom_table[hc] = c;
669 atom_cnt++;
670 return c;
673 static struct branch *lookup_branch(const char *name)
675 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
676 struct branch *b;
678 for (b = branch_table[hc]; b; b = b->table_next_branch)
679 if (!strcmp(name, b->name))
680 return b;
681 return NULL;
684 static struct branch *new_branch(const char *name)
686 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
687 struct branch *b = lookup_branch(name);
689 if (b)
690 die("Invalid attempt to create duplicate branch: %s", name);
691 switch (check_ref_format(name)) {
692 case 0: break; /* its valid */
693 case CHECK_REF_FORMAT_ONELEVEL:
694 break; /* valid, but too few '/', allow anyway */
695 default:
696 die("Branch name doesn't conform to GIT standards: %s", name);
699 b = pool_calloc(1, sizeof(struct branch));
700 b->name = pool_strdup(name);
701 b->table_next_branch = branch_table[hc];
702 b->branch_tree.versions[0].mode = S_IFDIR;
703 b->branch_tree.versions[1].mode = S_IFDIR;
704 b->active = 0;
705 b->pack_id = MAX_PACK_ID;
706 branch_table[hc] = b;
707 branch_count++;
708 return b;
711 static unsigned int hc_entries(unsigned int cnt)
713 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
714 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
717 static struct tree_content *new_tree_content(unsigned int cnt)
719 struct avail_tree_content *f, *l = NULL;
720 struct tree_content *t;
721 unsigned int hc = hc_entries(cnt);
723 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
724 if (f->entry_capacity >= cnt)
725 break;
727 if (f) {
728 if (l)
729 l->next_avail = f->next_avail;
730 else
731 avail_tree_table[hc] = f->next_avail;
732 } else {
733 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
734 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
735 f->entry_capacity = cnt;
738 t = (struct tree_content*)f;
739 t->entry_count = 0;
740 t->delta_depth = 0;
741 return t;
744 static void release_tree_entry(struct tree_entry *e);
745 static void release_tree_content(struct tree_content *t)
747 struct avail_tree_content *f = (struct avail_tree_content*)t;
748 unsigned int hc = hc_entries(f->entry_capacity);
749 f->next_avail = avail_tree_table[hc];
750 avail_tree_table[hc] = f;
753 static void release_tree_content_recursive(struct tree_content *t)
755 unsigned int i;
756 for (i = 0; i < t->entry_count; i++)
757 release_tree_entry(t->entries[i]);
758 release_tree_content(t);
761 static struct tree_content *grow_tree_content(
762 struct tree_content *t,
763 int amt)
765 struct tree_content *r = new_tree_content(t->entry_count + amt);
766 r->entry_count = t->entry_count;
767 r->delta_depth = t->delta_depth;
768 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
769 release_tree_content(t);
770 return r;
773 static struct tree_entry *new_tree_entry(void)
775 struct tree_entry *e;
777 if (!avail_tree_entry) {
778 unsigned int n = tree_entry_alloc;
779 total_allocd += n * sizeof(struct tree_entry);
780 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
781 while (n-- > 1) {
782 *((void**)e) = e + 1;
783 e++;
785 *((void**)e) = NULL;
788 e = avail_tree_entry;
789 avail_tree_entry = *((void**)e);
790 return e;
793 static void release_tree_entry(struct tree_entry *e)
795 if (e->tree)
796 release_tree_content_recursive(e->tree);
797 *((void**)e) = avail_tree_entry;
798 avail_tree_entry = e;
801 static struct tree_content *dup_tree_content(struct tree_content *s)
803 struct tree_content *d;
804 struct tree_entry *a, *b;
805 unsigned int i;
807 if (!s)
808 return NULL;
809 d = new_tree_content(s->entry_count);
810 for (i = 0; i < s->entry_count; i++) {
811 a = s->entries[i];
812 b = new_tree_entry();
813 memcpy(b, a, sizeof(*a));
814 if (a->tree && is_null_sha1(b->versions[1].sha1))
815 b->tree = dup_tree_content(a->tree);
816 else
817 b->tree = NULL;
818 d->entries[i] = b;
820 d->entry_count = s->entry_count;
821 d->delta_depth = s->delta_depth;
823 return d;
826 static void start_packfile(void)
828 static char tmpfile[PATH_MAX];
829 struct packed_git *p;
830 struct pack_header hdr;
831 int pack_fd;
833 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
834 "pack/tmp_pack_XXXXXX");
835 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
836 strcpy(p->pack_name, tmpfile);
837 p->pack_fd = pack_fd;
839 hdr.hdr_signature = htonl(PACK_SIGNATURE);
840 hdr.hdr_version = htonl(2);
841 hdr.hdr_entries = 0;
842 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
844 pack_data = p;
845 pack_size = sizeof(hdr);
846 object_count = 0;
848 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
849 all_packs[pack_id] = p;
852 static int oecmp (const void *a_, const void *b_)
854 struct object_entry *a = *((struct object_entry**)a_);
855 struct object_entry *b = *((struct object_entry**)b_);
856 return hashcmp(a->sha1, b->sha1);
859 static char *create_index(void)
861 static char tmpfile[PATH_MAX];
862 git_SHA_CTX ctx;
863 struct sha1file *f;
864 struct object_entry **idx, **c, **last, *e;
865 struct object_entry_pool *o;
866 uint32_t array[256];
867 int i, idx_fd;
869 /* Build the sorted table of object IDs. */
870 idx = xmalloc(object_count * sizeof(struct object_entry*));
871 c = idx;
872 for (o = blocks; o; o = o->next_pool)
873 for (e = o->next_free; e-- != o->entries;)
874 if (pack_id == e->pack_id)
875 *c++ = e;
876 last = idx + object_count;
877 if (c != last)
878 die("internal consistency error creating the index");
879 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
881 /* Generate the fan-out array. */
882 c = idx;
883 for (i = 0; i < 256; i++) {
884 struct object_entry **next = c;
885 while (next < last) {
886 if ((*next)->sha1[0] != i)
887 break;
888 next++;
890 array[i] = htonl(next - idx);
891 c = next;
894 idx_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
895 "pack/tmp_idx_XXXXXX");
896 f = sha1fd(idx_fd, tmpfile);
897 sha1write(f, array, 256 * sizeof(int));
898 git_SHA1_Init(&ctx);
899 for (c = idx; c != last; c++) {
900 uint32_t offset = htonl((*c)->offset);
901 sha1write(f, &offset, 4);
902 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
903 git_SHA1_Update(&ctx, (*c)->sha1, 20);
905 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
906 sha1close(f, NULL, CSUM_FSYNC);
907 free(idx);
908 git_SHA1_Final(pack_data->sha1, &ctx);
909 return tmpfile;
912 static char *keep_pack(char *curr_index_name)
914 static char name[PATH_MAX];
915 static const char *keep_msg = "fast-import";
916 int keep_fd;
918 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
919 if (keep_fd < 0)
920 die_errno("cannot create keep file");
921 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
922 if (close(keep_fd))
923 die_errno("failed to write keep file");
925 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
926 get_object_directory(), sha1_to_hex(pack_data->sha1));
927 if (move_temp_to_file(pack_data->pack_name, name))
928 die("cannot store pack file");
930 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
931 get_object_directory(), sha1_to_hex(pack_data->sha1));
932 if (move_temp_to_file(curr_index_name, name))
933 die("cannot store index file");
934 return name;
937 static void unkeep_all_packs(void)
939 static char name[PATH_MAX];
940 int k;
942 for (k = 0; k < pack_id; k++) {
943 struct packed_git *p = all_packs[k];
944 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
945 get_object_directory(), sha1_to_hex(p->sha1));
946 unlink_or_warn(name);
950 static void end_packfile(void)
952 struct packed_git *old_p = pack_data, *new_p;
954 clear_delta_base_cache();
955 if (object_count) {
956 char *idx_name;
957 int i;
958 struct branch *b;
959 struct tag *t;
961 close_pack_windows(pack_data);
962 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
963 pack_data->pack_name, object_count,
964 NULL, 0);
965 close(pack_data->pack_fd);
966 idx_name = keep_pack(create_index());
968 /* Register the packfile with core git's machinery. */
969 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
970 if (!new_p)
971 die("core git rejected index %s", idx_name);
972 all_packs[pack_id] = new_p;
973 install_packed_git(new_p);
975 /* Print the boundary */
976 if (pack_edges) {
977 fprintf(pack_edges, "%s:", new_p->pack_name);
978 for (i = 0; i < branch_table_sz; i++) {
979 for (b = branch_table[i]; b; b = b->table_next_branch) {
980 if (b->pack_id == pack_id)
981 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
984 for (t = first_tag; t; t = t->next_tag) {
985 if (t->pack_id == pack_id)
986 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
988 fputc('\n', pack_edges);
989 fflush(pack_edges);
992 pack_id++;
994 else {
995 close(old_p->pack_fd);
996 unlink_or_warn(old_p->pack_name);
998 free(old_p);
1000 /* We can't carry a delta across packfiles. */
1001 strbuf_release(&last_blob.data);
1002 last_blob.offset = 0;
1003 last_blob.depth = 0;
1006 static void cycle_packfile(void)
1008 end_packfile();
1009 start_packfile();
1012 static size_t encode_header(
1013 enum object_type type,
1014 size_t size,
1015 unsigned char *hdr)
1017 int n = 1;
1018 unsigned char c;
1020 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
1021 die("bad type %d", type);
1023 c = (type << 4) | (size & 15);
1024 size >>= 4;
1025 while (size) {
1026 *hdr++ = c | 0x80;
1027 c = size & 0x7f;
1028 size >>= 7;
1029 n++;
1031 *hdr = c;
1032 return n;
1035 static int store_object(
1036 enum object_type type,
1037 struct strbuf *dat,
1038 struct last_object *last,
1039 unsigned char *sha1out,
1040 uintmax_t mark)
1042 void *out, *delta;
1043 struct object_entry *e;
1044 unsigned char hdr[96];
1045 unsigned char sha1[20];
1046 unsigned long hdrlen, deltalen;
1047 git_SHA_CTX c;
1048 z_stream s;
1050 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1051 (unsigned long)dat->len) + 1;
1052 git_SHA1_Init(&c);
1053 git_SHA1_Update(&c, hdr, hdrlen);
1054 git_SHA1_Update(&c, dat->buf, dat->len);
1055 git_SHA1_Final(sha1, &c);
1056 if (sha1out)
1057 hashcpy(sha1out, sha1);
1059 e = insert_object(sha1);
1060 if (mark)
1061 insert_mark(mark, e);
1062 if (e->offset) {
1063 duplicate_count_by_type[type]++;
1064 return 1;
1065 } else if (find_sha1_pack(sha1, packed_git)) {
1066 e->type = type;
1067 e->pack_id = MAX_PACK_ID;
1068 e->offset = 1; /* just not zero! */
1069 duplicate_count_by_type[type]++;
1070 return 1;
1073 if (last && last->data.buf && last->depth < max_depth) {
1074 delta = diff_delta(last->data.buf, last->data.len,
1075 dat->buf, dat->len,
1076 &deltalen, 0);
1077 if (delta && deltalen >= dat->len) {
1078 free(delta);
1079 delta = NULL;
1081 } else
1082 delta = NULL;
1084 memset(&s, 0, sizeof(s));
1085 deflateInit(&s, pack_compression_level);
1086 if (delta) {
1087 s.next_in = delta;
1088 s.avail_in = deltalen;
1089 } else {
1090 s.next_in = (void *)dat->buf;
1091 s.avail_in = dat->len;
1093 s.avail_out = deflateBound(&s, s.avail_in);
1094 s.next_out = out = xmalloc(s.avail_out);
1095 while (deflate(&s, Z_FINISH) == Z_OK)
1096 /* nothing */;
1097 deflateEnd(&s);
1099 /* Determine if we should auto-checkpoint. */
1100 if ((pack_size + 60 + s.total_out) > max_packsize
1101 || (pack_size + 60 + s.total_out) < pack_size) {
1103 /* This new object needs to *not* have the current pack_id. */
1104 e->pack_id = pack_id + 1;
1105 cycle_packfile();
1107 /* We cannot carry a delta into the new pack. */
1108 if (delta) {
1109 free(delta);
1110 delta = NULL;
1112 memset(&s, 0, sizeof(s));
1113 deflateInit(&s, pack_compression_level);
1114 s.next_in = (void *)dat->buf;
1115 s.avail_in = dat->len;
1116 s.avail_out = deflateBound(&s, s.avail_in);
1117 s.next_out = out = xrealloc(out, s.avail_out);
1118 while (deflate(&s, Z_FINISH) == Z_OK)
1119 /* nothing */;
1120 deflateEnd(&s);
1124 e->type = type;
1125 e->pack_id = pack_id;
1126 e->offset = pack_size;
1127 object_count++;
1128 object_count_by_type[type]++;
1130 if (delta) {
1131 unsigned long ofs = e->offset - last->offset;
1132 unsigned pos = sizeof(hdr) - 1;
1134 delta_count_by_type[type]++;
1135 e->depth = last->depth + 1;
1137 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1138 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1139 pack_size += hdrlen;
1141 hdr[pos] = ofs & 127;
1142 while (ofs >>= 7)
1143 hdr[--pos] = 128 | (--ofs & 127);
1144 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
1145 pack_size += sizeof(hdr) - pos;
1146 } else {
1147 e->depth = 0;
1148 hdrlen = encode_header(type, dat->len, hdr);
1149 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1150 pack_size += hdrlen;
1153 write_or_die(pack_data->pack_fd, out, s.total_out);
1154 pack_size += s.total_out;
1156 free(out);
1157 free(delta);
1158 if (last) {
1159 if (last->no_swap) {
1160 last->data = *dat;
1161 } else {
1162 strbuf_swap(&last->data, dat);
1164 last->offset = e->offset;
1165 last->depth = e->depth;
1167 return 0;
1170 /* All calls must be guarded by find_object() or find_mark() to
1171 * ensure the 'struct object_entry' passed was written by this
1172 * process instance. We unpack the entry by the offset, avoiding
1173 * the need for the corresponding .idx file. This unpacking rule
1174 * works because we only use OBJ_REF_DELTA within the packfiles
1175 * created by fast-import.
1177 * oe must not be NULL. Such an oe usually comes from giving
1178 * an unknown SHA-1 to find_object() or an undefined mark to
1179 * find_mark(). Callers must test for this condition and use
1180 * the standard read_sha1_file() when it happens.
1182 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1183 * find_mark(), where the mark was reloaded from an existing marks
1184 * file and is referencing an object that this fast-import process
1185 * instance did not write out to a packfile. Callers must test for
1186 * this condition and use read_sha1_file() instead.
1188 static void *gfi_unpack_entry(
1189 struct object_entry *oe,
1190 unsigned long *sizep)
1192 enum object_type type;
1193 struct packed_git *p = all_packs[oe->pack_id];
1194 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1195 /* The object is stored in the packfile we are writing to
1196 * and we have modified it since the last time we scanned
1197 * back to read a previously written object. If an old
1198 * window covered [p->pack_size, p->pack_size + 20) its
1199 * data is stale and is not valid. Closing all windows
1200 * and updating the packfile length ensures we can read
1201 * the newly written data.
1203 close_pack_windows(p);
1205 /* We have to offer 20 bytes additional on the end of
1206 * the packfile as the core unpacker code assumes the
1207 * footer is present at the file end and must promise
1208 * at least 20 bytes within any window it maps. But
1209 * we don't actually create the footer here.
1211 p->pack_size = pack_size + 20;
1213 return unpack_entry(p, oe->offset, &type, sizep);
1216 static const char *get_mode(const char *str, uint16_t *modep)
1218 unsigned char c;
1219 uint16_t mode = 0;
1221 while ((c = *str++) != ' ') {
1222 if (c < '0' || c > '7')
1223 return NULL;
1224 mode = (mode << 3) + (c - '0');
1226 *modep = mode;
1227 return str;
1230 static void load_tree(struct tree_entry *root)
1232 unsigned char *sha1 = root->versions[1].sha1;
1233 struct object_entry *myoe;
1234 struct tree_content *t;
1235 unsigned long size;
1236 char *buf;
1237 const char *c;
1239 root->tree = t = new_tree_content(8);
1240 if (is_null_sha1(sha1))
1241 return;
1243 myoe = find_object(sha1);
1244 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1245 if (myoe->type != OBJ_TREE)
1246 die("Not a tree: %s", sha1_to_hex(sha1));
1247 t->delta_depth = myoe->depth;
1248 buf = gfi_unpack_entry(myoe, &size);
1249 if (!buf)
1250 die("Can't load tree %s", sha1_to_hex(sha1));
1251 } else {
1252 enum object_type type;
1253 buf = read_sha1_file(sha1, &type, &size);
1254 if (!buf || type != OBJ_TREE)
1255 die("Can't load tree %s", sha1_to_hex(sha1));
1258 c = buf;
1259 while (c != (buf + size)) {
1260 struct tree_entry *e = new_tree_entry();
1262 if (t->entry_count == t->entry_capacity)
1263 root->tree = t = grow_tree_content(t, t->entry_count);
1264 t->entries[t->entry_count++] = e;
1266 e->tree = NULL;
1267 c = get_mode(c, &e->versions[1].mode);
1268 if (!c)
1269 die("Corrupt mode in %s", sha1_to_hex(sha1));
1270 e->versions[0].mode = e->versions[1].mode;
1271 e->name = to_atom(c, strlen(c));
1272 c += e->name->str_len + 1;
1273 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1274 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1275 c += 20;
1277 free(buf);
1280 static int tecmp0 (const void *_a, const void *_b)
1282 struct tree_entry *a = *((struct tree_entry**)_a);
1283 struct tree_entry *b = *((struct tree_entry**)_b);
1284 return base_name_compare(
1285 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1286 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1289 static int tecmp1 (const void *_a, const void *_b)
1291 struct tree_entry *a = *((struct tree_entry**)_a);
1292 struct tree_entry *b = *((struct tree_entry**)_b);
1293 return base_name_compare(
1294 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1295 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1298 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1300 size_t maxlen = 0;
1301 unsigned int i;
1303 if (!v)
1304 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1305 else
1306 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1308 for (i = 0; i < t->entry_count; i++) {
1309 if (t->entries[i]->versions[v].mode)
1310 maxlen += t->entries[i]->name->str_len + 34;
1313 strbuf_reset(b);
1314 strbuf_grow(b, maxlen);
1315 for (i = 0; i < t->entry_count; i++) {
1316 struct tree_entry *e = t->entries[i];
1317 if (!e->versions[v].mode)
1318 continue;
1319 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1320 e->name->str_dat, '\0');
1321 strbuf_add(b, e->versions[v].sha1, 20);
1325 static void store_tree(struct tree_entry *root)
1327 struct tree_content *t = root->tree;
1328 unsigned int i, j, del;
1329 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1330 struct object_entry *le;
1332 if (!is_null_sha1(root->versions[1].sha1))
1333 return;
1335 for (i = 0; i < t->entry_count; i++) {
1336 if (t->entries[i]->tree)
1337 store_tree(t->entries[i]);
1340 le = find_object(root->versions[0].sha1);
1341 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1342 mktree(t, 0, &old_tree);
1343 lo.data = old_tree;
1344 lo.offset = le->offset;
1345 lo.depth = t->delta_depth;
1348 mktree(t, 1, &new_tree);
1349 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1351 t->delta_depth = lo.depth;
1352 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1353 struct tree_entry *e = t->entries[i];
1354 if (e->versions[1].mode) {
1355 e->versions[0].mode = e->versions[1].mode;
1356 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1357 t->entries[j++] = e;
1358 } else {
1359 release_tree_entry(e);
1360 del++;
1363 t->entry_count -= del;
1366 static int tree_content_set(
1367 struct tree_entry *root,
1368 const char *p,
1369 const unsigned char *sha1,
1370 const uint16_t mode,
1371 struct tree_content *subtree)
1373 struct tree_content *t = root->tree;
1374 const char *slash1;
1375 unsigned int i, n;
1376 struct tree_entry *e;
1378 slash1 = strchr(p, '/');
1379 if (slash1)
1380 n = slash1 - p;
1381 else
1382 n = strlen(p);
1383 if (!n)
1384 die("Empty path component found in input");
1385 if (!slash1 && !S_ISDIR(mode) && subtree)
1386 die("Non-directories cannot have subtrees");
1388 for (i = 0; i < t->entry_count; i++) {
1389 e = t->entries[i];
1390 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1391 if (!slash1) {
1392 if (!S_ISDIR(mode)
1393 && e->versions[1].mode == mode
1394 && !hashcmp(e->versions[1].sha1, sha1))
1395 return 0;
1396 e->versions[1].mode = mode;
1397 hashcpy(e->versions[1].sha1, sha1);
1398 if (e->tree)
1399 release_tree_content_recursive(e->tree);
1400 e->tree = subtree;
1401 hashclr(root->versions[1].sha1);
1402 return 1;
1404 if (!S_ISDIR(e->versions[1].mode)) {
1405 e->tree = new_tree_content(8);
1406 e->versions[1].mode = S_IFDIR;
1408 if (!e->tree)
1409 load_tree(e);
1410 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1411 hashclr(root->versions[1].sha1);
1412 return 1;
1414 return 0;
1418 if (t->entry_count == t->entry_capacity)
1419 root->tree = t = grow_tree_content(t, t->entry_count);
1420 e = new_tree_entry();
1421 e->name = to_atom(p, n);
1422 e->versions[0].mode = 0;
1423 hashclr(e->versions[0].sha1);
1424 t->entries[t->entry_count++] = e;
1425 if (slash1) {
1426 e->tree = new_tree_content(8);
1427 e->versions[1].mode = S_IFDIR;
1428 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1429 } else {
1430 e->tree = subtree;
1431 e->versions[1].mode = mode;
1432 hashcpy(e->versions[1].sha1, sha1);
1434 hashclr(root->versions[1].sha1);
1435 return 1;
1438 static int tree_content_remove(
1439 struct tree_entry *root,
1440 const char *p,
1441 struct tree_entry *backup_leaf)
1443 struct tree_content *t = root->tree;
1444 const char *slash1;
1445 unsigned int i, n;
1446 struct tree_entry *e;
1448 slash1 = strchr(p, '/');
1449 if (slash1)
1450 n = slash1 - p;
1451 else
1452 n = strlen(p);
1454 for (i = 0; i < t->entry_count; i++) {
1455 e = t->entries[i];
1456 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1457 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1458 goto del_entry;
1459 if (!e->tree)
1460 load_tree(e);
1461 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1462 for (n = 0; n < e->tree->entry_count; n++) {
1463 if (e->tree->entries[n]->versions[1].mode) {
1464 hashclr(root->versions[1].sha1);
1465 return 1;
1468 backup_leaf = NULL;
1469 goto del_entry;
1471 return 0;
1474 return 0;
1476 del_entry:
1477 if (backup_leaf)
1478 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1479 else if (e->tree)
1480 release_tree_content_recursive(e->tree);
1481 e->tree = NULL;
1482 e->versions[1].mode = 0;
1483 hashclr(e->versions[1].sha1);
1484 hashclr(root->versions[1].sha1);
1485 return 1;
1488 static int tree_content_get(
1489 struct tree_entry *root,
1490 const char *p,
1491 struct tree_entry *leaf)
1493 struct tree_content *t = root->tree;
1494 const char *slash1;
1495 unsigned int i, n;
1496 struct tree_entry *e;
1498 slash1 = strchr(p, '/');
1499 if (slash1)
1500 n = slash1 - p;
1501 else
1502 n = strlen(p);
1504 for (i = 0; i < t->entry_count; i++) {
1505 e = t->entries[i];
1506 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1507 if (!slash1) {
1508 memcpy(leaf, e, sizeof(*leaf));
1509 if (e->tree && is_null_sha1(e->versions[1].sha1))
1510 leaf->tree = dup_tree_content(e->tree);
1511 else
1512 leaf->tree = NULL;
1513 return 1;
1515 if (!S_ISDIR(e->versions[1].mode))
1516 return 0;
1517 if (!e->tree)
1518 load_tree(e);
1519 return tree_content_get(e, slash1 + 1, leaf);
1522 return 0;
1525 static int update_branch(struct branch *b)
1527 static const char *msg = "fast-import";
1528 struct ref_lock *lock;
1529 unsigned char old_sha1[20];
1531 if (is_null_sha1(b->sha1))
1532 return 0;
1533 if (read_ref(b->name, old_sha1))
1534 hashclr(old_sha1);
1535 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1536 if (!lock)
1537 return error("Unable to lock %s", b->name);
1538 if (!force_update && !is_null_sha1(old_sha1)) {
1539 struct commit *old_cmit, *new_cmit;
1541 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1542 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1543 if (!old_cmit || !new_cmit) {
1544 unlock_ref(lock);
1545 return error("Branch %s is missing commits.", b->name);
1548 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1549 unlock_ref(lock);
1550 warning("Not updating %s"
1551 " (new tip %s does not contain %s)",
1552 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1553 return -1;
1556 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1557 return error("Unable to update %s", b->name);
1558 return 0;
1561 static void dump_branches(void)
1563 unsigned int i;
1564 struct branch *b;
1566 for (i = 0; i < branch_table_sz; i++) {
1567 for (b = branch_table[i]; b; b = b->table_next_branch)
1568 failure |= update_branch(b);
1572 static void dump_tags(void)
1574 static const char *msg = "fast-import";
1575 struct tag *t;
1576 struct ref_lock *lock;
1577 char ref_name[PATH_MAX];
1579 for (t = first_tag; t; t = t->next_tag) {
1580 sprintf(ref_name, "tags/%s", t->name);
1581 lock = lock_ref_sha1(ref_name, NULL);
1582 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1583 failure |= error("Unable to update %s", ref_name);
1587 static void dump_marks_helper(FILE *f,
1588 uintmax_t base,
1589 struct mark_set *m)
1591 uintmax_t k;
1592 if (m->shift) {
1593 for (k = 0; k < 1024; k++) {
1594 if (m->data.sets[k])
1595 dump_marks_helper(f, (base + k) << m->shift,
1596 m->data.sets[k]);
1598 } else {
1599 for (k = 0; k < 1024; k++) {
1600 if (m->data.marked[k])
1601 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1602 sha1_to_hex(m->data.marked[k]->sha1));
1607 static void dump_marks(void)
1609 static struct lock_file mark_lock;
1610 int mark_fd;
1611 FILE *f;
1613 if (!export_marks_file)
1614 return;
1616 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1617 if (mark_fd < 0) {
1618 failure |= error("Unable to write marks file %s: %s",
1619 export_marks_file, strerror(errno));
1620 return;
1623 f = fdopen(mark_fd, "w");
1624 if (!f) {
1625 int saved_errno = errno;
1626 rollback_lock_file(&mark_lock);
1627 failure |= error("Unable to write marks file %s: %s",
1628 export_marks_file, strerror(saved_errno));
1629 return;
1633 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1634 * Assign -1 to the lock file descriptor so that commit_lock_file()
1635 * won't try to close() it.
1637 mark_lock.fd = -1;
1639 dump_marks_helper(f, 0, marks);
1640 if (ferror(f) || fclose(f)) {
1641 int saved_errno = errno;
1642 rollback_lock_file(&mark_lock);
1643 failure |= error("Unable to write marks file %s: %s",
1644 export_marks_file, strerror(saved_errno));
1645 return;
1648 if (commit_lock_file(&mark_lock)) {
1649 int saved_errno = errno;
1650 rollback_lock_file(&mark_lock);
1651 failure |= error("Unable to commit marks file %s: %s",
1652 export_marks_file, strerror(saved_errno));
1653 return;
1657 static void read_marks(void)
1659 char line[512];
1660 FILE *f = fopen(import_marks_file, "r");
1661 if (!f)
1662 die_errno("cannot read '%s'", import_marks_file);
1663 while (fgets(line, sizeof(line), f)) {
1664 uintmax_t mark;
1665 char *end;
1666 unsigned char sha1[20];
1667 struct object_entry *e;
1669 end = strchr(line, '\n');
1670 if (line[0] != ':' || !end)
1671 die("corrupt mark line: %s", line);
1672 *end = 0;
1673 mark = strtoumax(line + 1, &end, 10);
1674 if (!mark || end == line + 1
1675 || *end != ' ' || get_sha1(end + 1, sha1))
1676 die("corrupt mark line: %s", line);
1677 e = find_object(sha1);
1678 if (!e) {
1679 enum object_type type = sha1_object_info(sha1, NULL);
1680 if (type < 0)
1681 die("object not found: %s", sha1_to_hex(sha1));
1682 e = insert_object(sha1);
1683 e->type = type;
1684 e->pack_id = MAX_PACK_ID;
1685 e->offset = 1; /* just not zero! */
1687 insert_mark(mark, e);
1689 fclose(f);
1693 static int read_next_command(void)
1695 static int stdin_eof = 0;
1697 if (stdin_eof) {
1698 unread_command_buf = 0;
1699 return EOF;
1702 do {
1703 if (unread_command_buf) {
1704 unread_command_buf = 0;
1705 } else {
1706 struct recent_command *rc;
1708 strbuf_detach(&command_buf, NULL);
1709 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1710 if (stdin_eof)
1711 return EOF;
1713 if (!seen_data_command
1714 && prefixcmp(command_buf.buf, "feature ")
1715 && prefixcmp(command_buf.buf, "option ")) {
1716 parse_argv();
1719 rc = rc_free;
1720 if (rc)
1721 rc_free = rc->next;
1722 else {
1723 rc = cmd_hist.next;
1724 cmd_hist.next = rc->next;
1725 cmd_hist.next->prev = &cmd_hist;
1726 free(rc->buf);
1729 rc->buf = command_buf.buf;
1730 rc->prev = cmd_tail;
1731 rc->next = cmd_hist.prev;
1732 rc->prev->next = rc;
1733 cmd_tail = rc;
1735 } while (command_buf.buf[0] == '#');
1737 return 0;
1740 static void skip_optional_lf(void)
1742 int term_char = fgetc(stdin);
1743 if (term_char != '\n' && term_char != EOF)
1744 ungetc(term_char, stdin);
1747 static void parse_mark(void)
1749 if (!prefixcmp(command_buf.buf, "mark :")) {
1750 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1751 read_next_command();
1753 else
1754 next_mark = 0;
1757 static void parse_data(struct strbuf *sb)
1759 strbuf_reset(sb);
1761 if (prefixcmp(command_buf.buf, "data "))
1762 die("Expected 'data n' command, found: %s", command_buf.buf);
1764 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1765 char *term = xstrdup(command_buf.buf + 5 + 2);
1766 size_t term_len = command_buf.len - 5 - 2;
1768 strbuf_detach(&command_buf, NULL);
1769 for (;;) {
1770 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1771 die("EOF in data (terminator '%s' not found)", term);
1772 if (term_len == command_buf.len
1773 && !strcmp(term, command_buf.buf))
1774 break;
1775 strbuf_addbuf(sb, &command_buf);
1776 strbuf_addch(sb, '\n');
1778 free(term);
1780 else {
1781 size_t n = 0, length;
1783 length = strtoul(command_buf.buf + 5, NULL, 10);
1785 while (n < length) {
1786 size_t s = strbuf_fread(sb, length - n, stdin);
1787 if (!s && feof(stdin))
1788 die("EOF in data (%lu bytes remaining)",
1789 (unsigned long)(length - n));
1790 n += s;
1794 skip_optional_lf();
1797 static int validate_raw_date(const char *src, char *result, int maxlen)
1799 const char *orig_src = src;
1800 char *endp;
1801 unsigned long num;
1803 errno = 0;
1805 num = strtoul(src, &endp, 10);
1806 /* NEEDSWORK: perhaps check for reasonable values? */
1807 if (errno || endp == src || *endp != ' ')
1808 return -1;
1810 src = endp + 1;
1811 if (*src != '-' && *src != '+')
1812 return -1;
1814 num = strtoul(src + 1, &endp, 10);
1815 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1816 1400 < num)
1817 return -1;
1819 strcpy(result, orig_src);
1820 return 0;
1823 static char *parse_ident(const char *buf)
1825 const char *gt;
1826 size_t name_len;
1827 char *ident;
1829 gt = strrchr(buf, '>');
1830 if (!gt)
1831 die("Missing > in ident string: %s", buf);
1832 gt++;
1833 if (*gt != ' ')
1834 die("Missing space after > in ident string: %s", buf);
1835 gt++;
1836 name_len = gt - buf;
1837 ident = xmalloc(name_len + 24);
1838 strncpy(ident, buf, name_len);
1840 switch (whenspec) {
1841 case WHENSPEC_RAW:
1842 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1843 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1844 break;
1845 case WHENSPEC_RFC2822:
1846 if (parse_date(gt, ident + name_len, 24) < 0)
1847 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1848 break;
1849 case WHENSPEC_NOW:
1850 if (strcmp("now", gt))
1851 die("Date in ident must be 'now': %s", buf);
1852 datestamp(ident + name_len, 24);
1853 break;
1856 return ident;
1859 static void parse_new_blob(void)
1861 static struct strbuf buf = STRBUF_INIT;
1863 read_next_command();
1864 parse_mark();
1865 parse_data(&buf);
1866 store_object(OBJ_BLOB, &buf, &last_blob, NULL, next_mark);
1869 static void unload_one_branch(void)
1871 while (cur_active_branches
1872 && cur_active_branches >= max_active_branches) {
1873 uintmax_t min_commit = ULONG_MAX;
1874 struct branch *e, *l = NULL, *p = NULL;
1876 for (e = active_branches; e; e = e->active_next_branch) {
1877 if (e->last_commit < min_commit) {
1878 p = l;
1879 min_commit = e->last_commit;
1881 l = e;
1884 if (p) {
1885 e = p->active_next_branch;
1886 p->active_next_branch = e->active_next_branch;
1887 } else {
1888 e = active_branches;
1889 active_branches = e->active_next_branch;
1891 e->active = 0;
1892 e->active_next_branch = NULL;
1893 if (e->branch_tree.tree) {
1894 release_tree_content_recursive(e->branch_tree.tree);
1895 e->branch_tree.tree = NULL;
1897 cur_active_branches--;
1901 static void load_branch(struct branch *b)
1903 load_tree(&b->branch_tree);
1904 if (!b->active) {
1905 b->active = 1;
1906 b->active_next_branch = active_branches;
1907 active_branches = b;
1908 cur_active_branches++;
1909 branch_load_count++;
1913 static void file_change_m(struct branch *b)
1915 const char *p = command_buf.buf + 2;
1916 static struct strbuf uq = STRBUF_INIT;
1917 const char *endp;
1918 struct object_entry *oe = oe;
1919 unsigned char sha1[20];
1920 uint16_t mode, inline_data = 0;
1922 p = get_mode(p, &mode);
1923 if (!p)
1924 die("Corrupt mode: %s", command_buf.buf);
1925 switch (mode) {
1926 case 0644:
1927 case 0755:
1928 mode |= S_IFREG;
1929 case S_IFREG | 0644:
1930 case S_IFREG | 0755:
1931 case S_IFLNK:
1932 case S_IFGITLINK:
1933 /* ok */
1934 break;
1935 default:
1936 die("Corrupt mode: %s", command_buf.buf);
1939 if (*p == ':') {
1940 char *x;
1941 oe = find_mark(strtoumax(p + 1, &x, 10));
1942 hashcpy(sha1, oe->sha1);
1943 p = x;
1944 } else if (!prefixcmp(p, "inline")) {
1945 inline_data = 1;
1946 p += 6;
1947 } else {
1948 if (get_sha1_hex(p, sha1))
1949 die("Invalid SHA1: %s", command_buf.buf);
1950 oe = find_object(sha1);
1951 p += 40;
1953 if (*p++ != ' ')
1954 die("Missing space after SHA1: %s", command_buf.buf);
1956 strbuf_reset(&uq);
1957 if (!unquote_c_style(&uq, p, &endp)) {
1958 if (*endp)
1959 die("Garbage after path in: %s", command_buf.buf);
1960 p = uq.buf;
1963 if (S_ISGITLINK(mode)) {
1964 if (inline_data)
1965 die("Git links cannot be specified 'inline': %s",
1966 command_buf.buf);
1967 else if (oe) {
1968 if (oe->type != OBJ_COMMIT)
1969 die("Not a commit (actually a %s): %s",
1970 typename(oe->type), command_buf.buf);
1973 * Accept the sha1 without checking; it expected to be in
1974 * another repository.
1976 } else if (inline_data) {
1977 static struct strbuf buf = STRBUF_INIT;
1979 if (p != uq.buf) {
1980 strbuf_addstr(&uq, p);
1981 p = uq.buf;
1983 read_next_command();
1984 parse_data(&buf);
1985 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
1986 } else if (oe) {
1987 if (oe->type != OBJ_BLOB)
1988 die("Not a blob (actually a %s): %s",
1989 typename(oe->type), command_buf.buf);
1990 } else {
1991 enum object_type type = sha1_object_info(sha1, NULL);
1992 if (type < 0)
1993 die("Blob not found: %s", command_buf.buf);
1994 if (type != OBJ_BLOB)
1995 die("Not a blob (actually a %s): %s",
1996 typename(type), command_buf.buf);
1999 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2002 static void file_change_d(struct branch *b)
2004 const char *p = command_buf.buf + 2;
2005 static struct strbuf uq = STRBUF_INIT;
2006 const char *endp;
2008 strbuf_reset(&uq);
2009 if (!unquote_c_style(&uq, p, &endp)) {
2010 if (*endp)
2011 die("Garbage after path in: %s", command_buf.buf);
2012 p = uq.buf;
2014 tree_content_remove(&b->branch_tree, p, NULL);
2017 static void file_change_cr(struct branch *b, int rename)
2019 const char *s, *d;
2020 static struct strbuf s_uq = STRBUF_INIT;
2021 static struct strbuf d_uq = STRBUF_INIT;
2022 const char *endp;
2023 struct tree_entry leaf;
2025 s = command_buf.buf + 2;
2026 strbuf_reset(&s_uq);
2027 if (!unquote_c_style(&s_uq, s, &endp)) {
2028 if (*endp != ' ')
2029 die("Missing space after source: %s", command_buf.buf);
2030 } else {
2031 endp = strchr(s, ' ');
2032 if (!endp)
2033 die("Missing space after source: %s", command_buf.buf);
2034 strbuf_add(&s_uq, s, endp - s);
2036 s = s_uq.buf;
2038 endp++;
2039 if (!*endp)
2040 die("Missing dest: %s", command_buf.buf);
2042 d = endp;
2043 strbuf_reset(&d_uq);
2044 if (!unquote_c_style(&d_uq, d, &endp)) {
2045 if (*endp)
2046 die("Garbage after dest in: %s", command_buf.buf);
2047 d = d_uq.buf;
2050 memset(&leaf, 0, sizeof(leaf));
2051 if (rename)
2052 tree_content_remove(&b->branch_tree, s, &leaf);
2053 else
2054 tree_content_get(&b->branch_tree, s, &leaf);
2055 if (!leaf.versions[1].mode)
2056 die("Path %s not in branch", s);
2057 tree_content_set(&b->branch_tree, d,
2058 leaf.versions[1].sha1,
2059 leaf.versions[1].mode,
2060 leaf.tree);
2063 static void note_change_n(struct branch *b)
2065 const char *p = command_buf.buf + 2;
2066 static struct strbuf uq = STRBUF_INIT;
2067 struct object_entry *oe = oe;
2068 struct branch *s;
2069 unsigned char sha1[20], commit_sha1[20];
2070 uint16_t inline_data = 0;
2072 /* <dataref> or 'inline' */
2073 if (*p == ':') {
2074 char *x;
2075 oe = find_mark(strtoumax(p + 1, &x, 10));
2076 hashcpy(sha1, oe->sha1);
2077 p = x;
2078 } else if (!prefixcmp(p, "inline")) {
2079 inline_data = 1;
2080 p += 6;
2081 } else {
2082 if (get_sha1_hex(p, sha1))
2083 die("Invalid SHA1: %s", command_buf.buf);
2084 oe = find_object(sha1);
2085 p += 40;
2087 if (*p++ != ' ')
2088 die("Missing space after SHA1: %s", command_buf.buf);
2090 /* <committish> */
2091 s = lookup_branch(p);
2092 if (s) {
2093 hashcpy(commit_sha1, s->sha1);
2094 } else if (*p == ':') {
2095 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2096 struct object_entry *commit_oe = find_mark(commit_mark);
2097 if (commit_oe->type != OBJ_COMMIT)
2098 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2099 hashcpy(commit_sha1, commit_oe->sha1);
2100 } else if (!get_sha1(p, commit_sha1)) {
2101 unsigned long size;
2102 char *buf = read_object_with_reference(commit_sha1,
2103 commit_type, &size, commit_sha1);
2104 if (!buf || size < 46)
2105 die("Not a valid commit: %s", p);
2106 free(buf);
2107 } else
2108 die("Invalid ref name or SHA1 expression: %s", p);
2110 if (inline_data) {
2111 static struct strbuf buf = STRBUF_INIT;
2113 if (p != uq.buf) {
2114 strbuf_addstr(&uq, p);
2115 p = uq.buf;
2117 read_next_command();
2118 parse_data(&buf);
2119 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
2120 } else if (oe) {
2121 if (oe->type != OBJ_BLOB)
2122 die("Not a blob (actually a %s): %s",
2123 typename(oe->type), command_buf.buf);
2124 } else {
2125 enum object_type type = sha1_object_info(sha1, NULL);
2126 if (type < 0)
2127 die("Blob not found: %s", command_buf.buf);
2128 if (type != OBJ_BLOB)
2129 die("Not a blob (actually a %s): %s",
2130 typename(type), command_buf.buf);
2133 tree_content_set(&b->branch_tree, sha1_to_hex(commit_sha1), sha1,
2134 S_IFREG | 0644, NULL);
2137 static void file_change_deleteall(struct branch *b)
2139 release_tree_content_recursive(b->branch_tree.tree);
2140 hashclr(b->branch_tree.versions[0].sha1);
2141 hashclr(b->branch_tree.versions[1].sha1);
2142 load_tree(&b->branch_tree);
2145 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2147 if (!buf || size < 46)
2148 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2149 if (memcmp("tree ", buf, 5)
2150 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2151 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2152 hashcpy(b->branch_tree.versions[0].sha1,
2153 b->branch_tree.versions[1].sha1);
2156 static void parse_from_existing(struct branch *b)
2158 if (is_null_sha1(b->sha1)) {
2159 hashclr(b->branch_tree.versions[0].sha1);
2160 hashclr(b->branch_tree.versions[1].sha1);
2161 } else {
2162 unsigned long size;
2163 char *buf;
2165 buf = read_object_with_reference(b->sha1,
2166 commit_type, &size, b->sha1);
2167 parse_from_commit(b, buf, size);
2168 free(buf);
2172 static int parse_from(struct branch *b)
2174 const char *from;
2175 struct branch *s;
2177 if (prefixcmp(command_buf.buf, "from "))
2178 return 0;
2180 if (b->branch_tree.tree) {
2181 release_tree_content_recursive(b->branch_tree.tree);
2182 b->branch_tree.tree = NULL;
2185 from = strchr(command_buf.buf, ' ') + 1;
2186 s = lookup_branch(from);
2187 if (b == s)
2188 die("Can't create a branch from itself: %s", b->name);
2189 else if (s) {
2190 unsigned char *t = s->branch_tree.versions[1].sha1;
2191 hashcpy(b->sha1, s->sha1);
2192 hashcpy(b->branch_tree.versions[0].sha1, t);
2193 hashcpy(b->branch_tree.versions[1].sha1, t);
2194 } else if (*from == ':') {
2195 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2196 struct object_entry *oe = find_mark(idnum);
2197 if (oe->type != OBJ_COMMIT)
2198 die("Mark :%" PRIuMAX " not a commit", idnum);
2199 hashcpy(b->sha1, oe->sha1);
2200 if (oe->pack_id != MAX_PACK_ID) {
2201 unsigned long size;
2202 char *buf = gfi_unpack_entry(oe, &size);
2203 parse_from_commit(b, buf, size);
2204 free(buf);
2205 } else
2206 parse_from_existing(b);
2207 } else if (!get_sha1(from, b->sha1))
2208 parse_from_existing(b);
2209 else
2210 die("Invalid ref name or SHA1 expression: %s", from);
2212 read_next_command();
2213 return 1;
2216 static struct hash_list *parse_merge(unsigned int *count)
2218 struct hash_list *list = NULL, *n, *e = e;
2219 const char *from;
2220 struct branch *s;
2222 *count = 0;
2223 while (!prefixcmp(command_buf.buf, "merge ")) {
2224 from = strchr(command_buf.buf, ' ') + 1;
2225 n = xmalloc(sizeof(*n));
2226 s = lookup_branch(from);
2227 if (s)
2228 hashcpy(n->sha1, s->sha1);
2229 else if (*from == ':') {
2230 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2231 struct object_entry *oe = find_mark(idnum);
2232 if (oe->type != OBJ_COMMIT)
2233 die("Mark :%" PRIuMAX " not a commit", idnum);
2234 hashcpy(n->sha1, oe->sha1);
2235 } else if (!get_sha1(from, n->sha1)) {
2236 unsigned long size;
2237 char *buf = read_object_with_reference(n->sha1,
2238 commit_type, &size, n->sha1);
2239 if (!buf || size < 46)
2240 die("Not a valid commit: %s", from);
2241 free(buf);
2242 } else
2243 die("Invalid ref name or SHA1 expression: %s", from);
2245 n->next = NULL;
2246 if (list)
2247 e->next = n;
2248 else
2249 list = n;
2250 e = n;
2251 (*count)++;
2252 read_next_command();
2254 return list;
2257 static void parse_new_commit(void)
2259 static struct strbuf msg = STRBUF_INIT;
2260 struct branch *b;
2261 char *sp;
2262 char *author = NULL;
2263 char *committer = NULL;
2264 struct hash_list *merge_list = NULL;
2265 unsigned int merge_count;
2267 /* Obtain the branch name from the rest of our command */
2268 sp = strchr(command_buf.buf, ' ') + 1;
2269 b = lookup_branch(sp);
2270 if (!b)
2271 b = new_branch(sp);
2273 read_next_command();
2274 parse_mark();
2275 if (!prefixcmp(command_buf.buf, "author ")) {
2276 author = parse_ident(command_buf.buf + 7);
2277 read_next_command();
2279 if (!prefixcmp(command_buf.buf, "committer ")) {
2280 committer = parse_ident(command_buf.buf + 10);
2281 read_next_command();
2283 if (!committer)
2284 die("Expected committer but didn't get one");
2285 parse_data(&msg);
2286 read_next_command();
2287 parse_from(b);
2288 merge_list = parse_merge(&merge_count);
2290 /* ensure the branch is active/loaded */
2291 if (!b->branch_tree.tree || !max_active_branches) {
2292 unload_one_branch();
2293 load_branch(b);
2296 /* file_change* */
2297 while (command_buf.len > 0) {
2298 if (!prefixcmp(command_buf.buf, "M "))
2299 file_change_m(b);
2300 else if (!prefixcmp(command_buf.buf, "D "))
2301 file_change_d(b);
2302 else if (!prefixcmp(command_buf.buf, "R "))
2303 file_change_cr(b, 1);
2304 else if (!prefixcmp(command_buf.buf, "C "))
2305 file_change_cr(b, 0);
2306 else if (!prefixcmp(command_buf.buf, "N "))
2307 note_change_n(b);
2308 else if (!strcmp("deleteall", command_buf.buf))
2309 file_change_deleteall(b);
2310 else {
2311 unread_command_buf = 1;
2312 break;
2314 if (read_next_command() == EOF)
2315 break;
2318 /* build the tree and the commit */
2319 store_tree(&b->branch_tree);
2320 hashcpy(b->branch_tree.versions[0].sha1,
2321 b->branch_tree.versions[1].sha1);
2323 strbuf_reset(&new_data);
2324 strbuf_addf(&new_data, "tree %s\n",
2325 sha1_to_hex(b->branch_tree.versions[1].sha1));
2326 if (!is_null_sha1(b->sha1))
2327 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2328 while (merge_list) {
2329 struct hash_list *next = merge_list->next;
2330 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2331 free(merge_list);
2332 merge_list = next;
2334 strbuf_addf(&new_data,
2335 "author %s\n"
2336 "committer %s\n"
2337 "\n",
2338 author ? author : committer, committer);
2339 strbuf_addbuf(&new_data, &msg);
2340 free(author);
2341 free(committer);
2343 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2344 b->pack_id = pack_id;
2345 b->last_commit = object_count_by_type[OBJ_COMMIT];
2348 static void parse_new_tag(void)
2350 static struct strbuf msg = STRBUF_INIT;
2351 char *sp;
2352 const char *from;
2353 char *tagger;
2354 struct branch *s;
2355 struct tag *t;
2356 uintmax_t from_mark = 0;
2357 unsigned char sha1[20];
2359 /* Obtain the new tag name from the rest of our command */
2360 sp = strchr(command_buf.buf, ' ') + 1;
2361 t = pool_alloc(sizeof(struct tag));
2362 t->next_tag = NULL;
2363 t->name = pool_strdup(sp);
2364 if (last_tag)
2365 last_tag->next_tag = t;
2366 else
2367 first_tag = t;
2368 last_tag = t;
2369 read_next_command();
2371 /* from ... */
2372 if (prefixcmp(command_buf.buf, "from "))
2373 die("Expected from command, got %s", command_buf.buf);
2374 from = strchr(command_buf.buf, ' ') + 1;
2375 s = lookup_branch(from);
2376 if (s) {
2377 hashcpy(sha1, s->sha1);
2378 } else if (*from == ':') {
2379 struct object_entry *oe;
2380 from_mark = strtoumax(from + 1, NULL, 10);
2381 oe = find_mark(from_mark);
2382 if (oe->type != OBJ_COMMIT)
2383 die("Mark :%" PRIuMAX " not a commit", from_mark);
2384 hashcpy(sha1, oe->sha1);
2385 } else if (!get_sha1(from, sha1)) {
2386 unsigned long size;
2387 char *buf;
2389 buf = read_object_with_reference(sha1,
2390 commit_type, &size, sha1);
2391 if (!buf || size < 46)
2392 die("Not a valid commit: %s", from);
2393 free(buf);
2394 } else
2395 die("Invalid ref name or SHA1 expression: %s", from);
2396 read_next_command();
2398 /* tagger ... */
2399 if (!prefixcmp(command_buf.buf, "tagger ")) {
2400 tagger = parse_ident(command_buf.buf + 7);
2401 read_next_command();
2402 } else
2403 tagger = NULL;
2405 /* tag payload/message */
2406 parse_data(&msg);
2408 /* build the tag object */
2409 strbuf_reset(&new_data);
2411 strbuf_addf(&new_data,
2412 "object %s\n"
2413 "type %s\n"
2414 "tag %s\n",
2415 sha1_to_hex(sha1), commit_type, t->name);
2416 if (tagger)
2417 strbuf_addf(&new_data,
2418 "tagger %s\n", tagger);
2419 strbuf_addch(&new_data, '\n');
2420 strbuf_addbuf(&new_data, &msg);
2421 free(tagger);
2423 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2424 t->pack_id = MAX_PACK_ID;
2425 else
2426 t->pack_id = pack_id;
2429 static void parse_reset_branch(void)
2431 struct branch *b;
2432 char *sp;
2434 /* Obtain the branch name from the rest of our command */
2435 sp = strchr(command_buf.buf, ' ') + 1;
2436 b = lookup_branch(sp);
2437 if (b) {
2438 hashclr(b->sha1);
2439 hashclr(b->branch_tree.versions[0].sha1);
2440 hashclr(b->branch_tree.versions[1].sha1);
2441 if (b->branch_tree.tree) {
2442 release_tree_content_recursive(b->branch_tree.tree);
2443 b->branch_tree.tree = NULL;
2446 else
2447 b = new_branch(sp);
2448 read_next_command();
2449 parse_from(b);
2450 if (command_buf.len > 0)
2451 unread_command_buf = 1;
2454 static void parse_checkpoint(void)
2456 if (object_count) {
2457 cycle_packfile();
2458 dump_branches();
2459 dump_tags();
2460 dump_marks();
2462 skip_optional_lf();
2465 static void parse_progress(void)
2467 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2468 fputc('\n', stdout);
2469 fflush(stdout);
2470 skip_optional_lf();
2473 static void option_import_marks(const char *marks, int from_stream)
2475 if (import_marks_file) {
2476 if (from_stream)
2477 die("Only one import-marks command allowed per stream");
2479 /* read previous mark file */
2480 if(!import_marks_file_from_stream)
2481 read_marks();
2484 import_marks_file = xstrdup(marks);
2485 import_marks_file_from_stream = from_stream;
2488 static void option_date_format(const char *fmt)
2490 if (!strcmp(fmt, "raw"))
2491 whenspec = WHENSPEC_RAW;
2492 else if (!strcmp(fmt, "rfc2822"))
2493 whenspec = WHENSPEC_RFC2822;
2494 else if (!strcmp(fmt, "now"))
2495 whenspec = WHENSPEC_NOW;
2496 else
2497 die("unknown --date-format argument %s", fmt);
2500 static void option_max_pack_size(const char *packsize)
2502 max_packsize = strtoumax(packsize, NULL, 0) * 1024 * 1024;
2505 static void option_depth(const char *depth)
2507 max_depth = strtoul(depth, NULL, 0);
2508 if (max_depth > MAX_DEPTH)
2509 die("--depth cannot exceed %u", MAX_DEPTH);
2512 static void option_active_branches(const char *branches)
2514 max_active_branches = strtoul(branches, NULL, 0);
2517 static void option_export_marks(const char *marks)
2519 export_marks_file = xstrdup(marks);
2522 static void option_export_pack_edges(const char *edges)
2524 if (pack_edges)
2525 fclose(pack_edges);
2526 pack_edges = fopen(edges, "a");
2527 if (!pack_edges)
2528 die_errno("Cannot open '%s'", edges);
2531 static int parse_one_option(const char *option)
2533 if (!prefixcmp(option, "max-pack-size=")) {
2534 option_max_pack_size(option + 14);
2535 } else if (!prefixcmp(option, "depth=")) {
2536 option_depth(option + 6);
2537 } else if (!prefixcmp(option, "active-branches=")) {
2538 option_active_branches(option + 16);
2539 } else if (!prefixcmp(option, "export-pack-edges=")) {
2540 option_export_pack_edges(option + 18);
2541 } else if (!prefixcmp(option, "quiet")) {
2542 show_stats = 0;
2543 } else if (!prefixcmp(option, "stats")) {
2544 show_stats = 1;
2545 } else {
2546 return 0;
2549 return 1;
2552 static int parse_one_feature(const char *feature, int from_stream)
2554 if (!prefixcmp(feature, "date-format=")) {
2555 option_date_format(feature + 12);
2556 } else if (!prefixcmp(feature, "import-marks=")) {
2557 option_import_marks(feature + 13, from_stream);
2558 } else if (!prefixcmp(feature, "export-marks=")) {
2559 option_export_marks(feature + 13);
2560 } else if (!prefixcmp(feature, "force")) {
2561 force_update = 1;
2562 } else {
2563 return 0;
2566 return 1;
2569 static void parse_feature(void)
2571 char *feature = command_buf.buf + 8;
2573 if (seen_data_command)
2574 die("Got feature command '%s' after data command", feature);
2576 if (parse_one_feature(feature, 1))
2577 return;
2579 die("This version of fast-import does not support feature %s.", feature);
2582 static void parse_option(void)
2584 char *option = command_buf.buf + 11;
2586 if (seen_data_command)
2587 die("Got option command '%s' after data command", option);
2589 if (parse_one_option(option))
2590 return;
2592 die("This version of fast-import does not support option: %s", option);
2595 static int git_pack_config(const char *k, const char *v, void *cb)
2597 if (!strcmp(k, "pack.depth")) {
2598 max_depth = git_config_int(k, v);
2599 if (max_depth > MAX_DEPTH)
2600 max_depth = MAX_DEPTH;
2601 return 0;
2603 if (!strcmp(k, "pack.compression")) {
2604 int level = git_config_int(k, v);
2605 if (level == -1)
2606 level = Z_DEFAULT_COMPRESSION;
2607 else if (level < 0 || level > Z_BEST_COMPRESSION)
2608 die("bad pack compression level %d", level);
2609 pack_compression_level = level;
2610 pack_compression_seen = 1;
2611 return 0;
2613 return git_default_config(k, v, cb);
2616 static const char fast_import_usage[] =
2617 "git fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2619 static void parse_argv(void)
2621 unsigned int i;
2623 for (i = 1; i < global_argc; i++) {
2624 const char *a = global_argv[i];
2626 if (*a != '-' || !strcmp(a, "--"))
2627 break;
2629 if (parse_one_option(a + 2))
2630 continue;
2632 if (parse_one_feature(a + 2, 0))
2633 continue;
2635 die("unknown option %s", a);
2637 if (i != global_argc)
2638 usage(fast_import_usage);
2640 seen_data_command = 1;
2641 if (import_marks_file)
2642 read_marks();
2645 int main(int argc, const char **argv)
2647 unsigned int i;
2649 git_extract_argv0_path(argv[0]);
2651 if (argc == 2 && !strcmp(argv[1], "-h"))
2652 usage(fast_import_usage);
2654 setup_git_directory();
2655 git_config(git_pack_config, NULL);
2656 if (!pack_compression_seen && core_compression_seen)
2657 pack_compression_level = core_compression_level;
2659 alloc_objects(object_entry_alloc);
2660 strbuf_init(&command_buf, 0);
2661 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2662 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2663 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2664 marks = pool_calloc(1, sizeof(struct mark_set));
2666 global_argc = argc;
2667 global_argv = argv;
2669 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2670 for (i = 0; i < (cmd_save - 1); i++)
2671 rc_free[i].next = &rc_free[i + 1];
2672 rc_free[cmd_save - 1].next = NULL;
2674 prepare_packed_git();
2675 start_packfile();
2676 set_die_routine(die_nicely);
2677 while (read_next_command() != EOF) {
2678 if (!strcmp("blob", command_buf.buf))
2679 parse_new_blob();
2680 else if (!prefixcmp(command_buf.buf, "commit "))
2681 parse_new_commit();
2682 else if (!prefixcmp(command_buf.buf, "tag "))
2683 parse_new_tag();
2684 else if (!prefixcmp(command_buf.buf, "reset "))
2685 parse_reset_branch();
2686 else if (!strcmp("checkpoint", command_buf.buf))
2687 parse_checkpoint();
2688 else if (!prefixcmp(command_buf.buf, "progress "))
2689 parse_progress();
2690 else if (!prefixcmp(command_buf.buf, "feature "))
2691 parse_feature();
2692 else if (!prefixcmp(command_buf.buf, "option git "))
2693 parse_option();
2694 else if (!prefixcmp(command_buf.buf, "option "))
2695 /* ignore non-git options*/;
2696 else
2697 die("Unsupported command: %s", command_buf.buf);
2700 /* argv hasn't been parsed yet, do so */
2701 if (!seen_data_command)
2702 parse_argv();
2704 end_packfile();
2706 dump_branches();
2707 dump_tags();
2708 unkeep_all_packs();
2709 dump_marks();
2711 if (pack_edges)
2712 fclose(pack_edges);
2714 if (show_stats) {
2715 uintmax_t total_count = 0, duplicate_count = 0;
2716 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2717 total_count += object_count_by_type[i];
2718 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2719 duplicate_count += duplicate_count_by_type[i];
2721 fprintf(stderr, "%s statistics:\n", argv[0]);
2722 fprintf(stderr, "---------------------------------------------------------------------\n");
2723 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2724 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2725 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]);
2726 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]);
2727 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]);
2728 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]);
2729 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2730 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2731 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2732 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2733 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2734 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2735 fprintf(stderr, "---------------------------------------------------------------------\n");
2736 pack_report();
2737 fprintf(stderr, "---------------------------------------------------------------------\n");
2738 fprintf(stderr, "\n");
2741 return failure ? 1 : 0;