1/*
2(See Documentation/git-fast-import.txt for maintained documentation.)
3Format of STDIN stream:
4
5 stream ::= cmd*;
6
7 cmd ::= new_blob
8 | new_commit
9 | new_tag
10 | reset_branch
11 | checkpoint
12 | progress
13 ;
14
15 new_blob ::= 'blob' lf
16 mark?
17 file_content;
18 file_content ::= data;
19
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;
30
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;
47
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;
53
54 reset_branch ::= 'reset' sp ref_str lf
55 ('from' sp committish lf)?
56 lf?;
57
58 checkpoint ::= 'checkpoint' lf
59 lf?;
60
61 progress ::= 'progress' sp not_lf* lf
62 lf?;
63
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?;
73
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;
80
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;
86
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 ;
103
104 declen ::= # unsigned 32 bit value, ascii base10 notation;
105 bigint ::= # unsigned integer value, ascii base10 notation;
106 binary_data ::= # file content, not interpreted;
107
108 when ::= raw_when | rfc2822_when;
109 raw_when ::= ts sp tz;
110 rfc2822_when ::= # Valid RFC 2822 date and time;
111
112 sp ::= # ASCII space character;
113 lf ::= # ASCII newline (LF) character;
114
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;
125
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;
134
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*/
146
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"
159
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)
164
165struct object_entry
166{
167 struct pack_idx_entry idx;
168 struct object_entry *next;
169 uint32_t type : TYPE_BITS,
170 pack_id : PACK_ID_BITS,
171 depth : DEPTH_BITS;
172};
173
174struct object_entry_pool
175{
176 struct object_entry_pool *next_pool;
177 struct object_entry *next_free;
178 struct object_entry *end;
179 struct object_entry entries[FLEX_ARRAY]; /* more */
180};
181
182struct mark_set
183{
184 union {
185 struct object_entry *marked[1024];
186 struct mark_set *sets[1024];
187 } data;
188 unsigned int shift;
189};
190
191struct last_object
192{
193 struct strbuf data;
194 off_t offset;
195 unsigned int depth;
196 unsigned no_swap : 1;
197};
198
199struct mem_pool
200{
201 struct mem_pool *next_pool;
202 char *next_free;
203 char *end;
204 uintmax_t space[FLEX_ARRAY]; /* more */
205};
206
207struct atom_str
208{
209 struct atom_str *next_atom;
210 unsigned short str_len;
211 char str_dat[FLEX_ARRAY]; /* more */
212};
213
214struct tree_content;
215struct tree_entry
216{
217 struct tree_content *tree;
218 struct atom_str *name;
219 struct tree_entry_ms
220 {
221 uint16_t mode;
222 unsigned char sha1[20];
223 } versions[2];
224};
225
226struct tree_content
227{
228 unsigned int entry_capacity; /* must match avail_tree_content */
229 unsigned int entry_count;
230 unsigned int delta_depth;
231 struct tree_entry *entries[FLEX_ARRAY]; /* more */
232};
233
234struct avail_tree_content
235{
236 unsigned int entry_capacity; /* must match tree_content */
237 struct avail_tree_content *next_avail;
238};
239
240struct branch
241{
242 struct branch *table_next_branch;
243 struct branch *active_next_branch;
244 const char *name;
245 struct tree_entry branch_tree;
246 uintmax_t last_commit;
247 uintmax_t num_notes;
248 unsigned active : 1;
249 unsigned pack_id : PACK_ID_BITS;
250 unsigned char sha1[20];
251};
252
253struct tag
254{
255 struct tag *next_tag;
256 const char *name;
257 unsigned int pack_id;
258 unsigned char sha1[20];
259};
260
261struct hash_list
262{
263 struct hash_list *next;
264 unsigned char sha1[20];
265};
266
267typedef enum {
268 WHENSPEC_RAW = 1,
269 WHENSPEC_RFC2822,
270 WHENSPEC_NOW
271} whenspec_type;
272
273struct recent_command
274{
275 struct recent_command *prev;
276 struct recent_command *next;
277 char *buf;
278};
279
280/* Configured limits on output */
281static unsigned long max_depth = 10;
282static off_t max_packsize;
283static uintmax_t big_file_threshold = 512 * 1024 * 1024;
284static int force_update;
285static int pack_compression_level = Z_DEFAULT_COMPRESSION;
286static int pack_compression_seen;
287
288/* Stats and misc. counters */
289static uintmax_t alloc_count;
290static uintmax_t marks_set_count;
291static uintmax_t object_count_by_type[1 << TYPE_BITS];
292static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
293static uintmax_t delta_count_by_type[1 << TYPE_BITS];
294static unsigned long object_count;
295static unsigned long branch_count;
296static unsigned long branch_load_count;
297static int failure;
298static FILE *pack_edges;
299static unsigned int show_stats = 1;
300static int global_argc;
301static const char **global_argv;
302
303/* Memory pools */
304static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
305static size_t total_allocd;
306static struct mem_pool *mem_pool;
307
308/* Atom management */
309static unsigned int atom_table_sz = 4451;
310static unsigned int atom_cnt;
311static struct atom_str **atom_table;
312
313/* The .pack file being generated */
314static unsigned int pack_id;
315static struct sha1file *pack_file;
316static struct packed_git *pack_data;
317static struct packed_git **all_packs;
318static off_t pack_size;
319
320/* Table of objects we've written. */
321static unsigned int object_entry_alloc = 5000;
322static struct object_entry_pool *blocks;
323static struct object_entry *object_table[1 << 16];
324static struct mark_set *marks;
325static const char *export_marks_file;
326static const char *import_marks_file;
327static int import_marks_file_from_stream;
328static int relative_marks_paths;
329
330/* Our last blob */
331static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
332
333/* Tree management */
334static unsigned int tree_entry_alloc = 1000;
335static void *avail_tree_entry;
336static unsigned int avail_tree_table_sz = 100;
337static struct avail_tree_content **avail_tree_table;
338static struct strbuf old_tree = STRBUF_INIT;
339static struct strbuf new_tree = STRBUF_INIT;
340
341/* Branch data */
342static unsigned long max_active_branches = 5;
343static unsigned long cur_active_branches;
344static unsigned long branch_table_sz = 1039;
345static struct branch **branch_table;
346static struct branch *active_branches;
347
348/* Tag data */
349static struct tag *first_tag;
350static struct tag *last_tag;
351
352/* Input stream parsing */
353static whenspec_type whenspec = WHENSPEC_RAW;
354static struct strbuf command_buf = STRBUF_INIT;
355static int unread_command_buf;
356static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
357static struct recent_command *cmd_tail = &cmd_hist;
358static struct recent_command *rc_free;
359static unsigned int cmd_save = 100;
360static uintmax_t next_mark;
361static struct strbuf new_data = STRBUF_INIT;
362static int seen_data_command;
363
364static void parse_argv(void);
365
366static void write_branch_report(FILE *rpt, struct branch *b)
367{
368 fprintf(rpt, "%s:\n", b->name);
369
370 fprintf(rpt, " status :");
371 if (b->active)
372 fputs(" active", rpt);
373 if (b->branch_tree.tree)
374 fputs(" loaded", rpt);
375 if (is_null_sha1(b->branch_tree.versions[1].sha1))
376 fputs(" dirty", rpt);
377 fputc('\n', rpt);
378
379 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
380 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
381 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
382 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
383
384 fputs(" last pack : ", rpt);
385 if (b->pack_id < MAX_PACK_ID)
386 fprintf(rpt, "%u", b->pack_id);
387 fputc('\n', rpt);
388
389 fputc('\n', rpt);
390}
391
392static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
393
394static void write_crash_report(const char *err)
395{
396 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
397 FILE *rpt = fopen(loc, "w");
398 struct branch *b;
399 unsigned long lu;
400 struct recent_command *rc;
401
402 if (!rpt) {
403 error("can't write crash report %s: %s", loc, strerror(errno));
404 return;
405 }
406
407 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
408
409 fprintf(rpt, "fast-import crash report:\n");
410 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
411 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
412 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
413 fputc('\n', rpt);
414
415 fputs("fatal: ", rpt);
416 fputs(err, rpt);
417 fputc('\n', rpt);
418
419 fputc('\n', rpt);
420 fputs("Most Recent Commands Before Crash\n", rpt);
421 fputs("---------------------------------\n", rpt);
422 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
423 if (rc->next == &cmd_hist)
424 fputs("* ", rpt);
425 else
426 fputs(" ", rpt);
427 fputs(rc->buf, rpt);
428 fputc('\n', rpt);
429 }
430
431 fputc('\n', rpt);
432 fputs("Active Branch LRU\n", rpt);
433 fputs("-----------------\n", rpt);
434 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
435 cur_active_branches,
436 max_active_branches);
437 fputc('\n', rpt);
438 fputs(" pos clock name\n", rpt);
439 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
440 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
441 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
442 ++lu, b->last_commit, b->name);
443
444 fputc('\n', rpt);
445 fputs("Inactive Branches\n", rpt);
446 fputs("-----------------\n", rpt);
447 for (lu = 0; lu < branch_table_sz; lu++) {
448 for (b = branch_table[lu]; b; b = b->table_next_branch)
449 write_branch_report(rpt, b);
450 }
451
452 if (first_tag) {
453 struct tag *tg;
454 fputc('\n', rpt);
455 fputs("Annotated Tags\n", rpt);
456 fputs("--------------\n", rpt);
457 for (tg = first_tag; tg; tg = tg->next_tag) {
458 fputs(sha1_to_hex(tg->sha1), rpt);
459 fputc(' ', rpt);
460 fputs(tg->name, rpt);
461 fputc('\n', rpt);
462 }
463 }
464
465 fputc('\n', rpt);
466 fputs("Marks\n", rpt);
467 fputs("-----\n", rpt);
468 if (export_marks_file)
469 fprintf(rpt, " exported to %s\n", export_marks_file);
470 else
471 dump_marks_helper(rpt, 0, marks);
472
473 fputc('\n', rpt);
474 fputs("-------------------\n", rpt);
475 fputs("END OF CRASH REPORT\n", rpt);
476 fclose(rpt);
477}
478
479static void end_packfile(void);
480static void unkeep_all_packs(void);
481static void dump_marks(void);
482
483static NORETURN void die_nicely(const char *err, va_list params)
484{
485 static int zombie;
486 char message[2 * PATH_MAX];
487
488 vsnprintf(message, sizeof(message), err, params);
489 fputs("fatal: ", stderr);
490 fputs(message, stderr);
491 fputc('\n', stderr);
492
493 if (!zombie) {
494 zombie = 1;
495 write_crash_report(message);
496 end_packfile();
497 unkeep_all_packs();
498 dump_marks();
499 }
500 exit(128);
501}
502
503static void alloc_objects(unsigned int cnt)
504{
505 struct object_entry_pool *b;
506
507 b = xmalloc(sizeof(struct object_entry_pool)
508 + cnt * sizeof(struct object_entry));
509 b->next_pool = blocks;
510 b->next_free = b->entries;
511 b->end = b->entries + cnt;
512 blocks = b;
513 alloc_count += cnt;
514}
515
516static struct object_entry *new_object(unsigned char *sha1)
517{
518 struct object_entry *e;
519
520 if (blocks->next_free == blocks->end)
521 alloc_objects(object_entry_alloc);
522
523 e = blocks->next_free++;
524 hashcpy(e->idx.sha1, sha1);
525 return e;
526}
527
528static struct object_entry *find_object(unsigned char *sha1)
529{
530 unsigned int h = sha1[0] << 8 | sha1[1];
531 struct object_entry *e;
532 for (e = object_table[h]; e; e = e->next)
533 if (!hashcmp(sha1, e->idx.sha1))
534 return e;
535 return NULL;
536}
537
538static struct object_entry *insert_object(unsigned char *sha1)
539{
540 unsigned int h = sha1[0] << 8 | sha1[1];
541 struct object_entry *e = object_table[h];
542
543 while (e) {
544 if (!hashcmp(sha1, e->idx.sha1))
545 return e;
546 e = e->next;
547 }
548
549 e = new_object(sha1);
550 e->next = object_table[h];
551 e->idx.offset = 0;
552 object_table[h] = e;
553 return e;
554}
555
556static unsigned int hc_str(const char *s, size_t len)
557{
558 unsigned int r = 0;
559 while (len-- > 0)
560 r = r * 31 + *s++;
561 return r;
562}
563
564static void *pool_alloc(size_t len)
565{
566 struct mem_pool *p;
567 void *r;
568
569 /* round up to a 'uintmax_t' alignment */
570 if (len & (sizeof(uintmax_t) - 1))
571 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
572
573 for (p = mem_pool; p; p = p->next_pool)
574 if ((p->end - p->next_free >= len))
575 break;
576
577 if (!p) {
578 if (len >= (mem_pool_alloc/2)) {
579 total_allocd += len;
580 return xmalloc(len);
581 }
582 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
583 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
584 p->next_pool = mem_pool;
585 p->next_free = (char *) p->space;
586 p->end = p->next_free + mem_pool_alloc;
587 mem_pool = p;
588 }
589
590 r = p->next_free;
591 p->next_free += len;
592 return r;
593}
594
595static void *pool_calloc(size_t count, size_t size)
596{
597 size_t len = count * size;
598 void *r = pool_alloc(len);
599 memset(r, 0, len);
600 return r;
601}
602
603static char *pool_strdup(const char *s)
604{
605 char *r = pool_alloc(strlen(s) + 1);
606 strcpy(r, s);
607 return r;
608}
609
610static void insert_mark(uintmax_t idnum, struct object_entry *oe)
611{
612 struct mark_set *s = marks;
613 while ((idnum >> s->shift) >= 1024) {
614 s = pool_calloc(1, sizeof(struct mark_set));
615 s->shift = marks->shift + 10;
616 s->data.sets[0] = marks;
617 marks = s;
618 }
619 while (s->shift) {
620 uintmax_t i = idnum >> s->shift;
621 idnum -= i << s->shift;
622 if (!s->data.sets[i]) {
623 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
624 s->data.sets[i]->shift = s->shift - 10;
625 }
626 s = s->data.sets[i];
627 }
628 if (!s->data.marked[idnum])
629 marks_set_count++;
630 s->data.marked[idnum] = oe;
631}
632
633static struct object_entry *find_mark(uintmax_t idnum)
634{
635 uintmax_t orig_idnum = idnum;
636 struct mark_set *s = marks;
637 struct object_entry *oe = NULL;
638 if ((idnum >> s->shift) < 1024) {
639 while (s && s->shift) {
640 uintmax_t i = idnum >> s->shift;
641 idnum -= i << s->shift;
642 s = s->data.sets[i];
643 }
644 if (s)
645 oe = s->data.marked[idnum];
646 }
647 if (!oe)
648 die("mark :%" PRIuMAX " not declared", orig_idnum);
649 return oe;
650}
651
652static struct atom_str *to_atom(const char *s, unsigned short len)
653{
654 unsigned int hc = hc_str(s, len) % atom_table_sz;
655 struct atom_str *c;
656
657 for (c = atom_table[hc]; c; c = c->next_atom)
658 if (c->str_len == len && !strncmp(s, c->str_dat, len))
659 return c;
660
661 c = pool_alloc(sizeof(struct atom_str) + len + 1);
662 c->str_len = len;
663 strncpy(c->str_dat, s, len);
664 c->str_dat[len] = 0;
665 c->next_atom = atom_table[hc];
666 atom_table[hc] = c;
667 atom_cnt++;
668 return c;
669}
670
671static struct branch *lookup_branch(const char *name)
672{
673 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
674 struct branch *b;
675
676 for (b = branch_table[hc]; b; b = b->table_next_branch)
677 if (!strcmp(name, b->name))
678 return b;
679 return NULL;
680}
681
682static struct branch *new_branch(const char *name)
683{
684 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
685 struct branch *b = lookup_branch(name);
686
687 if (b)
688 die("Invalid attempt to create duplicate branch: %s", name);
689 switch (check_ref_format(name)) {
690 case 0: break; /* its valid */
691 case CHECK_REF_FORMAT_ONELEVEL:
692 break; /* valid, but too few '/', allow anyway */
693 default:
694 die("Branch name doesn't conform to GIT standards: %s", name);
695 }
696
697 b = pool_calloc(1, sizeof(struct branch));
698 b->name = pool_strdup(name);
699 b->table_next_branch = branch_table[hc];
700 b->branch_tree.versions[0].mode = S_IFDIR;
701 b->branch_tree.versions[1].mode = S_IFDIR;
702 b->num_notes = 0;
703 b->active = 0;
704 b->pack_id = MAX_PACK_ID;
705 branch_table[hc] = b;
706 branch_count++;
707 return b;
708}
709
710static unsigned int hc_entries(unsigned int cnt)
711{
712 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
713 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
714}
715
716static struct tree_content *new_tree_content(unsigned int cnt)
717{
718 struct avail_tree_content *f, *l = NULL;
719 struct tree_content *t;
720 unsigned int hc = hc_entries(cnt);
721
722 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
723 if (f->entry_capacity >= cnt)
724 break;
725
726 if (f) {
727 if (l)
728 l->next_avail = f->next_avail;
729 else
730 avail_tree_table[hc] = f->next_avail;
731 } else {
732 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
733 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
734 f->entry_capacity = cnt;
735 }
736
737 t = (struct tree_content*)f;
738 t->entry_count = 0;
739 t->delta_depth = 0;
740 return t;
741}
742
743static void release_tree_entry(struct tree_entry *e);
744static void release_tree_content(struct tree_content *t)
745{
746 struct avail_tree_content *f = (struct avail_tree_content*)t;
747 unsigned int hc = hc_entries(f->entry_capacity);
748 f->next_avail = avail_tree_table[hc];
749 avail_tree_table[hc] = f;
750}
751
752static void release_tree_content_recursive(struct tree_content *t)
753{
754 unsigned int i;
755 for (i = 0; i < t->entry_count; i++)
756 release_tree_entry(t->entries[i]);
757 release_tree_content(t);
758}
759
760static struct tree_content *grow_tree_content(
761 struct tree_content *t,
762 int amt)
763{
764 struct tree_content *r = new_tree_content(t->entry_count + amt);
765 r->entry_count = t->entry_count;
766 r->delta_depth = t->delta_depth;
767 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
768 release_tree_content(t);
769 return r;
770}
771
772static struct tree_entry *new_tree_entry(void)
773{
774 struct tree_entry *e;
775
776 if (!avail_tree_entry) {
777 unsigned int n = tree_entry_alloc;
778 total_allocd += n * sizeof(struct tree_entry);
779 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
780 while (n-- > 1) {
781 *((void**)e) = e + 1;
782 e++;
783 }
784 *((void**)e) = NULL;
785 }
786
787 e = avail_tree_entry;
788 avail_tree_entry = *((void**)e);
789 return e;
790}
791
792static void release_tree_entry(struct tree_entry *e)
793{
794 if (e->tree)
795 release_tree_content_recursive(e->tree);
796 *((void**)e) = avail_tree_entry;
797 avail_tree_entry = e;
798}
799
800static struct tree_content *dup_tree_content(struct tree_content *s)
801{
802 struct tree_content *d;
803 struct tree_entry *a, *b;
804 unsigned int i;
805
806 if (!s)
807 return NULL;
808 d = new_tree_content(s->entry_count);
809 for (i = 0; i < s->entry_count; i++) {
810 a = s->entries[i];
811 b = new_tree_entry();
812 memcpy(b, a, sizeof(*a));
813 if (a->tree && is_null_sha1(b->versions[1].sha1))
814 b->tree = dup_tree_content(a->tree);
815 else
816 b->tree = NULL;
817 d->entries[i] = b;
818 }
819 d->entry_count = s->entry_count;
820 d->delta_depth = s->delta_depth;
821
822 return d;
823}
824
825static void start_packfile(void)
826{
827 static char tmpfile[PATH_MAX];
828 struct packed_git *p;
829 struct pack_header hdr;
830 int pack_fd;
831
832 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
833 "pack/tmp_pack_XXXXXX");
834 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
835 strcpy(p->pack_name, tmpfile);
836 p->pack_fd = pack_fd;
837 pack_file = sha1fd(pack_fd, p->pack_name);
838
839 hdr.hdr_signature = htonl(PACK_SIGNATURE);
840 hdr.hdr_version = htonl(2);
841 hdr.hdr_entries = 0;
842 sha1write(pack_file, &hdr, sizeof(hdr));
843
844 pack_data = p;
845 pack_size = sizeof(hdr);
846 object_count = 0;
847
848 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
849 all_packs[pack_id] = p;
850}
851
852static const char *create_index(void)
853{
854 const char *tmpfile;
855 struct pack_idx_entry **idx, **c, **last;
856 struct object_entry *e;
857 struct object_entry_pool *o;
858
859 /* Build the table of object IDs. */
860 idx = xmalloc(object_count * sizeof(*idx));
861 c = idx;
862 for (o = blocks; o; o = o->next_pool)
863 for (e = o->next_free; e-- != o->entries;)
864 if (pack_id == e->pack_id)
865 *c++ = &e->idx;
866 last = idx + object_count;
867 if (c != last)
868 die("internal consistency error creating the index");
869
870 tmpfile = write_idx_file(NULL, idx, object_count, pack_data->sha1);
871 free(idx);
872 return tmpfile;
873}
874
875static char *keep_pack(const char *curr_index_name)
876{
877 static char name[PATH_MAX];
878 static const char *keep_msg = "fast-import";
879 int keep_fd;
880
881 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
882 if (keep_fd < 0)
883 die_errno("cannot create keep file");
884 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
885 if (close(keep_fd))
886 die_errno("failed to write keep file");
887
888 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
889 get_object_directory(), sha1_to_hex(pack_data->sha1));
890 if (move_temp_to_file(pack_data->pack_name, name))
891 die("cannot store pack file");
892
893 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
894 get_object_directory(), sha1_to_hex(pack_data->sha1));
895 if (move_temp_to_file(curr_index_name, name))
896 die("cannot store index file");
897 free((void *)curr_index_name);
898 return name;
899}
900
901static void unkeep_all_packs(void)
902{
903 static char name[PATH_MAX];
904 int k;
905
906 for (k = 0; k < pack_id; k++) {
907 struct packed_git *p = all_packs[k];
908 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
909 get_object_directory(), sha1_to_hex(p->sha1));
910 unlink_or_warn(name);
911 }
912}
913
914static void end_packfile(void)
915{
916 struct packed_git *old_p = pack_data, *new_p;
917
918 clear_delta_base_cache();
919 if (object_count) {
920 unsigned char cur_pack_sha1[20];
921 char *idx_name;
922 int i;
923 struct branch *b;
924 struct tag *t;
925
926 close_pack_windows(pack_data);
927 sha1close(pack_file, cur_pack_sha1, 0);
928 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
929 pack_data->pack_name, object_count,
930 cur_pack_sha1, pack_size);
931 close(pack_data->pack_fd);
932 idx_name = keep_pack(create_index());
933
934 /* Register the packfile with core git's machinery. */
935 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
936 if (!new_p)
937 die("core git rejected index %s", idx_name);
938 all_packs[pack_id] = new_p;
939 install_packed_git(new_p);
940
941 /* Print the boundary */
942 if (pack_edges) {
943 fprintf(pack_edges, "%s:", new_p->pack_name);
944 for (i = 0; i < branch_table_sz; i++) {
945 for (b = branch_table[i]; b; b = b->table_next_branch) {
946 if (b->pack_id == pack_id)
947 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
948 }
949 }
950 for (t = first_tag; t; t = t->next_tag) {
951 if (t->pack_id == pack_id)
952 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
953 }
954 fputc('\n', pack_edges);
955 fflush(pack_edges);
956 }
957
958 pack_id++;
959 }
960 else {
961 close(old_p->pack_fd);
962 unlink_or_warn(old_p->pack_name);
963 }
964 free(old_p);
965
966 /* We can't carry a delta across packfiles. */
967 strbuf_release(&last_blob.data);
968 last_blob.offset = 0;
969 last_blob.depth = 0;
970}
971
972static void cycle_packfile(void)
973{
974 end_packfile();
975 start_packfile();
976}
977
978static int store_object(
979 enum object_type type,
980 struct strbuf *dat,
981 struct last_object *last,
982 unsigned char *sha1out,
983 uintmax_t mark)
984{
985 void *out, *delta;
986 struct object_entry *e;
987 unsigned char hdr[96];
988 unsigned char sha1[20];
989 unsigned long hdrlen, deltalen;
990 git_SHA_CTX c;
991 z_stream s;
992
993 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
994 (unsigned long)dat->len) + 1;
995 git_SHA1_Init(&c);
996 git_SHA1_Update(&c, hdr, hdrlen);
997 git_SHA1_Update(&c, dat->buf, dat->len);
998 git_SHA1_Final(sha1, &c);
999 if (sha1out)
1000 hashcpy(sha1out, sha1);
1001
1002 e = insert_object(sha1);
1003 if (mark)
1004 insert_mark(mark, e);
1005 if (e->idx.offset) {
1006 duplicate_count_by_type[type]++;
1007 return 1;
1008 } else if (find_sha1_pack(sha1, packed_git)) {
1009 e->type = type;
1010 e->pack_id = MAX_PACK_ID;
1011 e->idx.offset = 1; /* just not zero! */
1012 duplicate_count_by_type[type]++;
1013 return 1;
1014 }
1015
1016 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1017 delta = diff_delta(last->data.buf, last->data.len,
1018 dat->buf, dat->len,
1019 &deltalen, dat->len - 20);
1020 } else
1021 delta = NULL;
1022
1023 memset(&s, 0, sizeof(s));
1024 deflateInit(&s, pack_compression_level);
1025 if (delta) {
1026 s.next_in = delta;
1027 s.avail_in = deltalen;
1028 } else {
1029 s.next_in = (void *)dat->buf;
1030 s.avail_in = dat->len;
1031 }
1032 s.avail_out = deflateBound(&s, s.avail_in);
1033 s.next_out = out = xmalloc(s.avail_out);
1034 while (deflate(&s, Z_FINISH) == Z_OK)
1035 /* nothing */;
1036 deflateEnd(&s);
1037
1038 /* Determine if we should auto-checkpoint. */
1039 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1040 || (pack_size + 60 + s.total_out) < pack_size) {
1041
1042 /* This new object needs to *not* have the current pack_id. */
1043 e->pack_id = pack_id + 1;
1044 cycle_packfile();
1045
1046 /* We cannot carry a delta into the new pack. */
1047 if (delta) {
1048 free(delta);
1049 delta = NULL;
1050
1051 memset(&s, 0, sizeof(s));
1052 deflateInit(&s, pack_compression_level);
1053 s.next_in = (void *)dat->buf;
1054 s.avail_in = dat->len;
1055 s.avail_out = deflateBound(&s, s.avail_in);
1056 s.next_out = out = xrealloc(out, s.avail_out);
1057 while (deflate(&s, Z_FINISH) == Z_OK)
1058 /* nothing */;
1059 deflateEnd(&s);
1060 }
1061 }
1062
1063 e->type = type;
1064 e->pack_id = pack_id;
1065 e->idx.offset = pack_size;
1066 object_count++;
1067 object_count_by_type[type]++;
1068
1069 crc32_begin(pack_file);
1070
1071 if (delta) {
1072 off_t ofs = e->idx.offset - last->offset;
1073 unsigned pos = sizeof(hdr) - 1;
1074
1075 delta_count_by_type[type]++;
1076 e->depth = last->depth + 1;
1077
1078 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1079 sha1write(pack_file, hdr, hdrlen);
1080 pack_size += hdrlen;
1081
1082 hdr[pos] = ofs & 127;
1083 while (ofs >>= 7)
1084 hdr[--pos] = 128 | (--ofs & 127);
1085 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1086 pack_size += sizeof(hdr) - pos;
1087 } else {
1088 e->depth = 0;
1089 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1090 sha1write(pack_file, hdr, hdrlen);
1091 pack_size += hdrlen;
1092 }
1093
1094 sha1write(pack_file, out, s.total_out);
1095 pack_size += s.total_out;
1096
1097 e->idx.crc32 = crc32_end(pack_file);
1098
1099 free(out);
1100 free(delta);
1101 if (last) {
1102 if (last->no_swap) {
1103 last->data = *dat;
1104 } else {
1105 strbuf_swap(&last->data, dat);
1106 }
1107 last->offset = e->idx.offset;
1108 last->depth = e->depth;
1109 }
1110 return 0;
1111}
1112
1113static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1114{
1115 if (ftruncate(pack_data->pack_fd, to)
1116 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1117 die_errno("cannot truncate pack to skip duplicate");
1118 pack_size = to;
1119
1120 /* yes this is a layering violation */
1121 pack_file->total = to;
1122 pack_file->offset = 0;
1123 pack_file->ctx = *ctx;
1124}
1125
1126static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1127{
1128 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1129 unsigned char *in_buf = xmalloc(in_sz);
1130 unsigned char *out_buf = xmalloc(out_sz);
1131 struct object_entry *e;
1132 unsigned char sha1[20];
1133 unsigned long hdrlen;
1134 off_t offset;
1135 git_SHA_CTX c;
1136 git_SHA_CTX pack_file_ctx;
1137 z_stream s;
1138 int status = Z_OK;
1139
1140 /* Determine if we should auto-checkpoint. */
1141 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1142 || (pack_size + 60 + len) < pack_size)
1143 cycle_packfile();
1144
1145 offset = pack_size;
1146
1147 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1148 sha1flush(pack_file);
1149 pack_file_ctx = pack_file->ctx;
1150
1151 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1152 if (out_sz <= hdrlen)
1153 die("impossibly large object header");
1154
1155 git_SHA1_Init(&c);
1156 git_SHA1_Update(&c, out_buf, hdrlen);
1157
1158 crc32_begin(pack_file);
1159
1160 memset(&s, 0, sizeof(s));
1161 deflateInit(&s, pack_compression_level);
1162
1163 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1164 if (out_sz <= hdrlen)
1165 die("impossibly large object header");
1166
1167 s.next_out = out_buf + hdrlen;
1168 s.avail_out = out_sz - hdrlen;
1169
1170 while (status != Z_STREAM_END) {
1171 if (0 < len && !s.avail_in) {
1172 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1173 size_t n = fread(in_buf, 1, cnt, stdin);
1174 if (!n && feof(stdin))
1175 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1176
1177 git_SHA1_Update(&c, in_buf, n);
1178 s.next_in = in_buf;
1179 s.avail_in = n;
1180 len -= n;
1181 }
1182
1183 status = deflate(&s, len ? 0 : Z_FINISH);
1184
1185 if (!s.avail_out || status == Z_STREAM_END) {
1186 size_t n = s.next_out - out_buf;
1187 sha1write(pack_file, out_buf, n);
1188 pack_size += n;
1189 s.next_out = out_buf;
1190 s.avail_out = out_sz;
1191 }
1192
1193 switch (status) {
1194 case Z_OK:
1195 case Z_BUF_ERROR:
1196 case Z_STREAM_END:
1197 continue;
1198 default:
1199 die("unexpected deflate failure: %d", status);
1200 }
1201 }
1202 deflateEnd(&s);
1203 git_SHA1_Final(sha1, &c);
1204
1205 if (sha1out)
1206 hashcpy(sha1out, sha1);
1207
1208 e = insert_object(sha1);
1209
1210 if (mark)
1211 insert_mark(mark, e);
1212
1213 if (e->idx.offset) {
1214 duplicate_count_by_type[OBJ_BLOB]++;
1215 truncate_pack(offset, &pack_file_ctx);
1216
1217 } else if (find_sha1_pack(sha1, packed_git)) {
1218 e->type = OBJ_BLOB;
1219 e->pack_id = MAX_PACK_ID;
1220 e->idx.offset = 1; /* just not zero! */
1221 duplicate_count_by_type[OBJ_BLOB]++;
1222 truncate_pack(offset, &pack_file_ctx);
1223
1224 } else {
1225 e->depth = 0;
1226 e->type = OBJ_BLOB;
1227 e->pack_id = pack_id;
1228 e->idx.offset = offset;
1229 e->idx.crc32 = crc32_end(pack_file);
1230 object_count++;
1231 object_count_by_type[OBJ_BLOB]++;
1232 }
1233
1234 free(in_buf);
1235 free(out_buf);
1236}
1237
1238/* All calls must be guarded by find_object() or find_mark() to
1239 * ensure the 'struct object_entry' passed was written by this
1240 * process instance. We unpack the entry by the offset, avoiding
1241 * the need for the corresponding .idx file. This unpacking rule
1242 * works because we only use OBJ_REF_DELTA within the packfiles
1243 * created by fast-import.
1244 *
1245 * oe must not be NULL. Such an oe usually comes from giving
1246 * an unknown SHA-1 to find_object() or an undefined mark to
1247 * find_mark(). Callers must test for this condition and use
1248 * the standard read_sha1_file() when it happens.
1249 *
1250 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1251 * find_mark(), where the mark was reloaded from an existing marks
1252 * file and is referencing an object that this fast-import process
1253 * instance did not write out to a packfile. Callers must test for
1254 * this condition and use read_sha1_file() instead.
1255 */
1256static void *gfi_unpack_entry(
1257 struct object_entry *oe,
1258 unsigned long *sizep)
1259{
1260 enum object_type type;
1261 struct packed_git *p = all_packs[oe->pack_id];
1262 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1263 /* The object is stored in the packfile we are writing to
1264 * and we have modified it since the last time we scanned
1265 * back to read a previously written object. If an old
1266 * window covered [p->pack_size, p->pack_size + 20) its
1267 * data is stale and is not valid. Closing all windows
1268 * and updating the packfile length ensures we can read
1269 * the newly written data.
1270 */
1271 close_pack_windows(p);
1272 sha1flush(pack_file);
1273
1274 /* We have to offer 20 bytes additional on the end of
1275 * the packfile as the core unpacker code assumes the
1276 * footer is present at the file end and must promise
1277 * at least 20 bytes within any window it maps. But
1278 * we don't actually create the footer here.
1279 */
1280 p->pack_size = pack_size + 20;
1281 }
1282 return unpack_entry(p, oe->idx.offset, &type, sizep);
1283}
1284
1285static const char *get_mode(const char *str, uint16_t *modep)
1286{
1287 unsigned char c;
1288 uint16_t mode = 0;
1289
1290 while ((c = *str++) != ' ') {
1291 if (c < '0' || c > '7')
1292 return NULL;
1293 mode = (mode << 3) + (c - '0');
1294 }
1295 *modep = mode;
1296 return str;
1297}
1298
1299static void load_tree(struct tree_entry *root)
1300{
1301 unsigned char *sha1 = root->versions[1].sha1;
1302 struct object_entry *myoe;
1303 struct tree_content *t;
1304 unsigned long size;
1305 char *buf;
1306 const char *c;
1307
1308 root->tree = t = new_tree_content(8);
1309 if (is_null_sha1(sha1))
1310 return;
1311
1312 myoe = find_object(sha1);
1313 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1314 if (myoe->type != OBJ_TREE)
1315 die("Not a tree: %s", sha1_to_hex(sha1));
1316 t->delta_depth = myoe->depth;
1317 buf = gfi_unpack_entry(myoe, &size);
1318 if (!buf)
1319 die("Can't load tree %s", sha1_to_hex(sha1));
1320 } else {
1321 enum object_type type;
1322 buf = read_sha1_file(sha1, &type, &size);
1323 if (!buf || type != OBJ_TREE)
1324 die("Can't load tree %s", sha1_to_hex(sha1));
1325 }
1326
1327 c = buf;
1328 while (c != (buf + size)) {
1329 struct tree_entry *e = new_tree_entry();
1330
1331 if (t->entry_count == t->entry_capacity)
1332 root->tree = t = grow_tree_content(t, t->entry_count);
1333 t->entries[t->entry_count++] = e;
1334
1335 e->tree = NULL;
1336 c = get_mode(c, &e->versions[1].mode);
1337 if (!c)
1338 die("Corrupt mode in %s", sha1_to_hex(sha1));
1339 e->versions[0].mode = e->versions[1].mode;
1340 e->name = to_atom(c, strlen(c));
1341 c += e->name->str_len + 1;
1342 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1343 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1344 c += 20;
1345 }
1346 free(buf);
1347}
1348
1349static int tecmp0 (const void *_a, const void *_b)
1350{
1351 struct tree_entry *a = *((struct tree_entry**)_a);
1352 struct tree_entry *b = *((struct tree_entry**)_b);
1353 return base_name_compare(
1354 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1355 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1356}
1357
1358static int tecmp1 (const void *_a, const void *_b)
1359{
1360 struct tree_entry *a = *((struct tree_entry**)_a);
1361 struct tree_entry *b = *((struct tree_entry**)_b);
1362 return base_name_compare(
1363 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1364 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1365}
1366
1367static void mktree(struct tree_content *t, int v, struct strbuf *b)
1368{
1369 size_t maxlen = 0;
1370 unsigned int i;
1371
1372 if (!v)
1373 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1374 else
1375 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1376
1377 for (i = 0; i < t->entry_count; i++) {
1378 if (t->entries[i]->versions[v].mode)
1379 maxlen += t->entries[i]->name->str_len + 34;
1380 }
1381
1382 strbuf_reset(b);
1383 strbuf_grow(b, maxlen);
1384 for (i = 0; i < t->entry_count; i++) {
1385 struct tree_entry *e = t->entries[i];
1386 if (!e->versions[v].mode)
1387 continue;
1388 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1389 e->name->str_dat, '\0');
1390 strbuf_add(b, e->versions[v].sha1, 20);
1391 }
1392}
1393
1394static void store_tree(struct tree_entry *root)
1395{
1396 struct tree_content *t = root->tree;
1397 unsigned int i, j, del;
1398 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1399 struct object_entry *le;
1400
1401 if (!is_null_sha1(root->versions[1].sha1))
1402 return;
1403
1404 for (i = 0; i < t->entry_count; i++) {
1405 if (t->entries[i]->tree)
1406 store_tree(t->entries[i]);
1407 }
1408
1409 le = find_object(root->versions[0].sha1);
1410 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1411 mktree(t, 0, &old_tree);
1412 lo.data = old_tree;
1413 lo.offset = le->idx.offset;
1414 lo.depth = t->delta_depth;
1415 }
1416
1417 mktree(t, 1, &new_tree);
1418 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1419
1420 t->delta_depth = lo.depth;
1421 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1422 struct tree_entry *e = t->entries[i];
1423 if (e->versions[1].mode) {
1424 e->versions[0].mode = e->versions[1].mode;
1425 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1426 t->entries[j++] = e;
1427 } else {
1428 release_tree_entry(e);
1429 del++;
1430 }
1431 }
1432 t->entry_count -= del;
1433}
1434
1435static int tree_content_set(
1436 struct tree_entry *root,
1437 const char *p,
1438 const unsigned char *sha1,
1439 const uint16_t mode,
1440 struct tree_content *subtree)
1441{
1442 struct tree_content *t = root->tree;
1443 const char *slash1;
1444 unsigned int i, n;
1445 struct tree_entry *e;
1446
1447 slash1 = strchr(p, '/');
1448 if (slash1)
1449 n = slash1 - p;
1450 else
1451 n = strlen(p);
1452 if (!n)
1453 die("Empty path component found in input");
1454 if (!slash1 && !S_ISDIR(mode) && subtree)
1455 die("Non-directories cannot have subtrees");
1456
1457 for (i = 0; i < t->entry_count; i++) {
1458 e = t->entries[i];
1459 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1460 if (!slash1) {
1461 if (!S_ISDIR(mode)
1462 && e->versions[1].mode == mode
1463 && !hashcmp(e->versions[1].sha1, sha1))
1464 return 0;
1465 e->versions[1].mode = mode;
1466 hashcpy(e->versions[1].sha1, sha1);
1467 if (e->tree)
1468 release_tree_content_recursive(e->tree);
1469 e->tree = subtree;
1470 hashclr(root->versions[1].sha1);
1471 return 1;
1472 }
1473 if (!S_ISDIR(e->versions[1].mode)) {
1474 e->tree = new_tree_content(8);
1475 e->versions[1].mode = S_IFDIR;
1476 }
1477 if (!e->tree)
1478 load_tree(e);
1479 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1480 hashclr(root->versions[1].sha1);
1481 return 1;
1482 }
1483 return 0;
1484 }
1485 }
1486
1487 if (t->entry_count == t->entry_capacity)
1488 root->tree = t = grow_tree_content(t, t->entry_count);
1489 e = new_tree_entry();
1490 e->name = to_atom(p, n);
1491 e->versions[0].mode = 0;
1492 hashclr(e->versions[0].sha1);
1493 t->entries[t->entry_count++] = e;
1494 if (slash1) {
1495 e->tree = new_tree_content(8);
1496 e->versions[1].mode = S_IFDIR;
1497 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1498 } else {
1499 e->tree = subtree;
1500 e->versions[1].mode = mode;
1501 hashcpy(e->versions[1].sha1, sha1);
1502 }
1503 hashclr(root->versions[1].sha1);
1504 return 1;
1505}
1506
1507static int tree_content_remove(
1508 struct tree_entry *root,
1509 const char *p,
1510 struct tree_entry *backup_leaf)
1511{
1512 struct tree_content *t = root->tree;
1513 const char *slash1;
1514 unsigned int i, n;
1515 struct tree_entry *e;
1516
1517 slash1 = strchr(p, '/');
1518 if (slash1)
1519 n = slash1 - p;
1520 else
1521 n = strlen(p);
1522
1523 for (i = 0; i < t->entry_count; i++) {
1524 e = t->entries[i];
1525 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1526 if (slash1 && !S_ISDIR(e->versions[1].mode))
1527 /*
1528 * If p names a file in some subdirectory, and a
1529 * file or symlink matching the name of the
1530 * parent directory of p exists, then p cannot
1531 * exist and need not be deleted.
1532 */
1533 return 1;
1534 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1535 goto del_entry;
1536 if (!e->tree)
1537 load_tree(e);
1538 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1539 for (n = 0; n < e->tree->entry_count; n++) {
1540 if (e->tree->entries[n]->versions[1].mode) {
1541 hashclr(root->versions[1].sha1);
1542 return 1;
1543 }
1544 }
1545 backup_leaf = NULL;
1546 goto del_entry;
1547 }
1548 return 0;
1549 }
1550 }
1551 return 0;
1552
1553del_entry:
1554 if (backup_leaf)
1555 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1556 else if (e->tree)
1557 release_tree_content_recursive(e->tree);
1558 e->tree = NULL;
1559 e->versions[1].mode = 0;
1560 hashclr(e->versions[1].sha1);
1561 hashclr(root->versions[1].sha1);
1562 return 1;
1563}
1564
1565static int tree_content_get(
1566 struct tree_entry *root,
1567 const char *p,
1568 struct tree_entry *leaf)
1569{
1570 struct tree_content *t = root->tree;
1571 const char *slash1;
1572 unsigned int i, n;
1573 struct tree_entry *e;
1574
1575 slash1 = strchr(p, '/');
1576 if (slash1)
1577 n = slash1 - p;
1578 else
1579 n = strlen(p);
1580
1581 for (i = 0; i < t->entry_count; i++) {
1582 e = t->entries[i];
1583 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1584 if (!slash1) {
1585 memcpy(leaf, e, sizeof(*leaf));
1586 if (e->tree && is_null_sha1(e->versions[1].sha1))
1587 leaf->tree = dup_tree_content(e->tree);
1588 else
1589 leaf->tree = NULL;
1590 return 1;
1591 }
1592 if (!S_ISDIR(e->versions[1].mode))
1593 return 0;
1594 if (!e->tree)
1595 load_tree(e);
1596 return tree_content_get(e, slash1 + 1, leaf);
1597 }
1598 }
1599 return 0;
1600}
1601
1602static int update_branch(struct branch *b)
1603{
1604 static const char *msg = "fast-import";
1605 struct ref_lock *lock;
1606 unsigned char old_sha1[20];
1607
1608 if (is_null_sha1(b->sha1))
1609 return 0;
1610 if (read_ref(b->name, old_sha1))
1611 hashclr(old_sha1);
1612 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1613 if (!lock)
1614 return error("Unable to lock %s", b->name);
1615 if (!force_update && !is_null_sha1(old_sha1)) {
1616 struct commit *old_cmit, *new_cmit;
1617
1618 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1619 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1620 if (!old_cmit || !new_cmit) {
1621 unlock_ref(lock);
1622 return error("Branch %s is missing commits.", b->name);
1623 }
1624
1625 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1626 unlock_ref(lock);
1627 warning("Not updating %s"
1628 " (new tip %s does not contain %s)",
1629 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1630 return -1;
1631 }
1632 }
1633 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1634 return error("Unable to update %s", b->name);
1635 return 0;
1636}
1637
1638static void dump_branches(void)
1639{
1640 unsigned int i;
1641 struct branch *b;
1642
1643 for (i = 0; i < branch_table_sz; i++) {
1644 for (b = branch_table[i]; b; b = b->table_next_branch)
1645 failure |= update_branch(b);
1646 }
1647}
1648
1649static void dump_tags(void)
1650{
1651 static const char *msg = "fast-import";
1652 struct tag *t;
1653 struct ref_lock *lock;
1654 char ref_name[PATH_MAX];
1655
1656 for (t = first_tag; t; t = t->next_tag) {
1657 sprintf(ref_name, "tags/%s", t->name);
1658 lock = lock_ref_sha1(ref_name, NULL);
1659 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1660 failure |= error("Unable to update %s", ref_name);
1661 }
1662}
1663
1664static void dump_marks_helper(FILE *f,
1665 uintmax_t base,
1666 struct mark_set *m)
1667{
1668 uintmax_t k;
1669 if (m->shift) {
1670 for (k = 0; k < 1024; k++) {
1671 if (m->data.sets[k])
1672 dump_marks_helper(f, base + (k << m->shift),
1673 m->data.sets[k]);
1674 }
1675 } else {
1676 for (k = 0; k < 1024; k++) {
1677 if (m->data.marked[k])
1678 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1679 sha1_to_hex(m->data.marked[k]->idx.sha1));
1680 }
1681 }
1682}
1683
1684static void dump_marks(void)
1685{
1686 static struct lock_file mark_lock;
1687 int mark_fd;
1688 FILE *f;
1689
1690 if (!export_marks_file)
1691 return;
1692
1693 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1694 if (mark_fd < 0) {
1695 failure |= error("Unable to write marks file %s: %s",
1696 export_marks_file, strerror(errno));
1697 return;
1698 }
1699
1700 f = fdopen(mark_fd, "w");
1701 if (!f) {
1702 int saved_errno = errno;
1703 rollback_lock_file(&mark_lock);
1704 failure |= error("Unable to write marks file %s: %s",
1705 export_marks_file, strerror(saved_errno));
1706 return;
1707 }
1708
1709 /*
1710 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1711 * Assign -1 to the lock file descriptor so that commit_lock_file()
1712 * won't try to close() it.
1713 */
1714 mark_lock.fd = -1;
1715
1716 dump_marks_helper(f, 0, marks);
1717 if (ferror(f) || fclose(f)) {
1718 int saved_errno = errno;
1719 rollback_lock_file(&mark_lock);
1720 failure |= error("Unable to write marks file %s: %s",
1721 export_marks_file, strerror(saved_errno));
1722 return;
1723 }
1724
1725 if (commit_lock_file(&mark_lock)) {
1726 int saved_errno = errno;
1727 rollback_lock_file(&mark_lock);
1728 failure |= error("Unable to commit marks file %s: %s",
1729 export_marks_file, strerror(saved_errno));
1730 return;
1731 }
1732}
1733
1734static void read_marks(void)
1735{
1736 char line[512];
1737 FILE *f = fopen(import_marks_file, "r");
1738 if (!f)
1739 die_errno("cannot read '%s'", import_marks_file);
1740 while (fgets(line, sizeof(line), f)) {
1741 uintmax_t mark;
1742 char *end;
1743 unsigned char sha1[20];
1744 struct object_entry *e;
1745
1746 end = strchr(line, '\n');
1747 if (line[0] != ':' || !end)
1748 die("corrupt mark line: %s", line);
1749 *end = 0;
1750 mark = strtoumax(line + 1, &end, 10);
1751 if (!mark || end == line + 1
1752 || *end != ' ' || get_sha1(end + 1, sha1))
1753 die("corrupt mark line: %s", line);
1754 e = find_object(sha1);
1755 if (!e) {
1756 enum object_type type = sha1_object_info(sha1, NULL);
1757 if (type < 0)
1758 die("object not found: %s", sha1_to_hex(sha1));
1759 e = insert_object(sha1);
1760 e->type = type;
1761 e->pack_id = MAX_PACK_ID;
1762 e->idx.offset = 1; /* just not zero! */
1763 }
1764 insert_mark(mark, e);
1765 }
1766 fclose(f);
1767}
1768
1769
1770static int read_next_command(void)
1771{
1772 static int stdin_eof = 0;
1773
1774 if (stdin_eof) {
1775 unread_command_buf = 0;
1776 return EOF;
1777 }
1778
1779 do {
1780 if (unread_command_buf) {
1781 unread_command_buf = 0;
1782 } else {
1783 struct recent_command *rc;
1784
1785 strbuf_detach(&command_buf, NULL);
1786 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1787 if (stdin_eof)
1788 return EOF;
1789
1790 if (!seen_data_command
1791 && prefixcmp(command_buf.buf, "feature ")
1792 && prefixcmp(command_buf.buf, "option ")) {
1793 parse_argv();
1794 }
1795
1796 rc = rc_free;
1797 if (rc)
1798 rc_free = rc->next;
1799 else {
1800 rc = cmd_hist.next;
1801 cmd_hist.next = rc->next;
1802 cmd_hist.next->prev = &cmd_hist;
1803 free(rc->buf);
1804 }
1805
1806 rc->buf = command_buf.buf;
1807 rc->prev = cmd_tail;
1808 rc->next = cmd_hist.prev;
1809 rc->prev->next = rc;
1810 cmd_tail = rc;
1811 }
1812 } while (command_buf.buf[0] == '#');
1813
1814 return 0;
1815}
1816
1817static void skip_optional_lf(void)
1818{
1819 int term_char = fgetc(stdin);
1820 if (term_char != '\n' && term_char != EOF)
1821 ungetc(term_char, stdin);
1822}
1823
1824static void parse_mark(void)
1825{
1826 if (!prefixcmp(command_buf.buf, "mark :")) {
1827 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1828 read_next_command();
1829 }
1830 else
1831 next_mark = 0;
1832}
1833
1834static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1835{
1836 strbuf_reset(sb);
1837
1838 if (prefixcmp(command_buf.buf, "data "))
1839 die("Expected 'data n' command, found: %s", command_buf.buf);
1840
1841 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1842 char *term = xstrdup(command_buf.buf + 5 + 2);
1843 size_t term_len = command_buf.len - 5 - 2;
1844
1845 strbuf_detach(&command_buf, NULL);
1846 for (;;) {
1847 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1848 die("EOF in data (terminator '%s' not found)", term);
1849 if (term_len == command_buf.len
1850 && !strcmp(term, command_buf.buf))
1851 break;
1852 strbuf_addbuf(sb, &command_buf);
1853 strbuf_addch(sb, '\n');
1854 }
1855 free(term);
1856 }
1857 else {
1858 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1859 size_t n = 0, length = (size_t)len;
1860
1861 if (limit && limit < len) {
1862 *len_res = len;
1863 return 0;
1864 }
1865 if (length < len)
1866 die("data is too large to use in this context");
1867
1868 while (n < length) {
1869 size_t s = strbuf_fread(sb, length - n, stdin);
1870 if (!s && feof(stdin))
1871 die("EOF in data (%lu bytes remaining)",
1872 (unsigned long)(length - n));
1873 n += s;
1874 }
1875 }
1876
1877 skip_optional_lf();
1878 return 1;
1879}
1880
1881static int validate_raw_date(const char *src, char *result, int maxlen)
1882{
1883 const char *orig_src = src;
1884 char *endp;
1885 unsigned long num;
1886
1887 errno = 0;
1888
1889 num = strtoul(src, &endp, 10);
1890 /* NEEDSWORK: perhaps check for reasonable values? */
1891 if (errno || endp == src || *endp != ' ')
1892 return -1;
1893
1894 src = endp + 1;
1895 if (*src != '-' && *src != '+')
1896 return -1;
1897
1898 num = strtoul(src + 1, &endp, 10);
1899 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1900 1400 < num)
1901 return -1;
1902
1903 strcpy(result, orig_src);
1904 return 0;
1905}
1906
1907static char *parse_ident(const char *buf)
1908{
1909 const char *gt;
1910 size_t name_len;
1911 char *ident;
1912
1913 gt = strrchr(buf, '>');
1914 if (!gt)
1915 die("Missing > in ident string: %s", buf);
1916 gt++;
1917 if (*gt != ' ')
1918 die("Missing space after > in ident string: %s", buf);
1919 gt++;
1920 name_len = gt - buf;
1921 ident = xmalloc(name_len + 24);
1922 strncpy(ident, buf, name_len);
1923
1924 switch (whenspec) {
1925 case WHENSPEC_RAW:
1926 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1927 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1928 break;
1929 case WHENSPEC_RFC2822:
1930 if (parse_date(gt, ident + name_len, 24) < 0)
1931 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1932 break;
1933 case WHENSPEC_NOW:
1934 if (strcmp("now", gt))
1935 die("Date in ident must be 'now': %s", buf);
1936 datestamp(ident + name_len, 24);
1937 break;
1938 }
1939
1940 return ident;
1941}
1942
1943static void parse_and_store_blob(
1944 struct last_object *last,
1945 unsigned char *sha1out,
1946 uintmax_t mark)
1947{
1948 static struct strbuf buf = STRBUF_INIT;
1949 uintmax_t len;
1950
1951 if (parse_data(&buf, big_file_threshold, &len))
1952 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
1953 else {
1954 if (last) {
1955 strbuf_release(&last->data);
1956 last->offset = 0;
1957 last->depth = 0;
1958 }
1959 stream_blob(len, sha1out, mark);
1960 skip_optional_lf();
1961 }
1962}
1963
1964static void parse_new_blob(void)
1965{
1966 read_next_command();
1967 parse_mark();
1968 parse_and_store_blob(&last_blob, NULL, next_mark);
1969}
1970
1971static void unload_one_branch(void)
1972{
1973 while (cur_active_branches
1974 && cur_active_branches >= max_active_branches) {
1975 uintmax_t min_commit = ULONG_MAX;
1976 struct branch *e, *l = NULL, *p = NULL;
1977
1978 for (e = active_branches; e; e = e->active_next_branch) {
1979 if (e->last_commit < min_commit) {
1980 p = l;
1981 min_commit = e->last_commit;
1982 }
1983 l = e;
1984 }
1985
1986 if (p) {
1987 e = p->active_next_branch;
1988 p->active_next_branch = e->active_next_branch;
1989 } else {
1990 e = active_branches;
1991 active_branches = e->active_next_branch;
1992 }
1993 e->active = 0;
1994 e->active_next_branch = NULL;
1995 if (e->branch_tree.tree) {
1996 release_tree_content_recursive(e->branch_tree.tree);
1997 e->branch_tree.tree = NULL;
1998 }
1999 cur_active_branches--;
2000 }
2001}
2002
2003static void load_branch(struct branch *b)
2004{
2005 load_tree(&b->branch_tree);
2006 if (!b->active) {
2007 b->active = 1;
2008 b->active_next_branch = active_branches;
2009 active_branches = b;
2010 cur_active_branches++;
2011 branch_load_count++;
2012 }
2013}
2014
2015static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2016{
2017 unsigned char fanout = 0;
2018 while ((num_notes >>= 8))
2019 fanout++;
2020 return fanout;
2021}
2022
2023static void construct_path_with_fanout(const char *hex_sha1,
2024 unsigned char fanout, char *path)
2025{
2026 unsigned int i = 0, j = 0;
2027 if (fanout >= 20)
2028 die("Too large fanout (%u)", fanout);
2029 while (fanout) {
2030 path[i++] = hex_sha1[j++];
2031 path[i++] = hex_sha1[j++];
2032 path[i++] = '/';
2033 fanout--;
2034 }
2035 memcpy(path + i, hex_sha1 + j, 40 - j);
2036 path[i + 40 - j] = '\0';
2037}
2038
2039static uintmax_t do_change_note_fanout(
2040 struct tree_entry *orig_root, struct tree_entry *root,
2041 char *hex_sha1, unsigned int hex_sha1_len,
2042 char *fullpath, unsigned int fullpath_len,
2043 unsigned char fanout)
2044{
2045 struct tree_content *t = root->tree;
2046 struct tree_entry *e, leaf;
2047 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2048 uintmax_t num_notes = 0;
2049 unsigned char sha1[20];
2050 char realpath[60];
2051
2052 for (i = 0; t && i < t->entry_count; i++) {
2053 e = t->entries[i];
2054 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2055 tmp_fullpath_len = fullpath_len;
2056
2057 /*
2058 * We're interested in EITHER existing note entries (entries
2059 * with exactly 40 hex chars in path, not including directory
2060 * separators), OR directory entries that may contain note
2061 * entries (with < 40 hex chars in path).
2062 * Also, each path component in a note entry must be a multiple
2063 * of 2 chars.
2064 */
2065 if (!e->versions[1].mode ||
2066 tmp_hex_sha1_len > 40 ||
2067 e->name->str_len % 2)
2068 continue;
2069
2070 /* This _may_ be a note entry, or a subdir containing notes */
2071 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2072 e->name->str_len);
2073 if (tmp_fullpath_len)
2074 fullpath[tmp_fullpath_len++] = '/';
2075 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2076 e->name->str_len);
2077 tmp_fullpath_len += e->name->str_len;
2078 fullpath[tmp_fullpath_len] = '\0';
2079
2080 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2081 /* This is a note entry */
2082 construct_path_with_fanout(hex_sha1, fanout, realpath);
2083 if (!strcmp(fullpath, realpath)) {
2084 /* Note entry is in correct location */
2085 num_notes++;
2086 continue;
2087 }
2088
2089 /* Rename fullpath to realpath */
2090 if (!tree_content_remove(orig_root, fullpath, &leaf))
2091 die("Failed to remove path %s", fullpath);
2092 tree_content_set(orig_root, realpath,
2093 leaf.versions[1].sha1,
2094 leaf.versions[1].mode,
2095 leaf.tree);
2096 } else if (S_ISDIR(e->versions[1].mode)) {
2097 /* This is a subdir that may contain note entries */
2098 if (!e->tree)
2099 load_tree(e);
2100 num_notes += do_change_note_fanout(orig_root, e,
2101 hex_sha1, tmp_hex_sha1_len,
2102 fullpath, tmp_fullpath_len, fanout);
2103 }
2104
2105 /* The above may have reallocated the current tree_content */
2106 t = root->tree;
2107 }
2108 return num_notes;
2109}
2110
2111static uintmax_t change_note_fanout(struct tree_entry *root,
2112 unsigned char fanout)
2113{
2114 char hex_sha1[40], path[60];
2115 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2116}
2117
2118static void file_change_m(struct branch *b)
2119{
2120 const char *p = command_buf.buf + 2;
2121 static struct strbuf uq = STRBUF_INIT;
2122 const char *endp;
2123 struct object_entry *oe = oe;
2124 unsigned char sha1[20];
2125 uint16_t mode, inline_data = 0;
2126
2127 p = get_mode(p, &mode);
2128 if (!p)
2129 die("Corrupt mode: %s", command_buf.buf);
2130 switch (mode) {
2131 case 0644:
2132 case 0755:
2133 mode |= S_IFREG;
2134 case S_IFREG | 0644:
2135 case S_IFREG | 0755:
2136 case S_IFLNK:
2137 case S_IFDIR:
2138 case S_IFGITLINK:
2139 /* ok */
2140 break;
2141 default:
2142 die("Corrupt mode: %s", command_buf.buf);
2143 }
2144
2145 if (*p == ':') {
2146 char *x;
2147 oe = find_mark(strtoumax(p + 1, &x, 10));
2148 hashcpy(sha1, oe->idx.sha1);
2149 p = x;
2150 } else if (!prefixcmp(p, "inline")) {
2151 inline_data = 1;
2152 p += 6;
2153 } else {
2154 if (get_sha1_hex(p, sha1))
2155 die("Invalid SHA1: %s", command_buf.buf);
2156 oe = find_object(sha1);
2157 p += 40;
2158 }
2159 if (*p++ != ' ')
2160 die("Missing space after SHA1: %s", command_buf.buf);
2161
2162 strbuf_reset(&uq);
2163 if (!unquote_c_style(&uq, p, &endp)) {
2164 if (*endp)
2165 die("Garbage after path in: %s", command_buf.buf);
2166 p = uq.buf;
2167 }
2168
2169 /* Git does not track empty, non-toplevel directories. */
2170 if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
2171 tree_content_remove(&b->branch_tree, p, NULL);
2172 return;
2173 }
2174
2175 if (S_ISGITLINK(mode)) {
2176 if (inline_data)
2177 die("Git links cannot be specified 'inline': %s",
2178 command_buf.buf);
2179 else if (oe) {
2180 if (oe->type != OBJ_COMMIT)
2181 die("Not a commit (actually a %s): %s",
2182 typename(oe->type), command_buf.buf);
2183 }
2184 /*
2185 * Accept the sha1 without checking; it expected to be in
2186 * another repository.
2187 */
2188 } else if (inline_data) {
2189 if (S_ISDIR(mode))
2190 die("Directories cannot be specified 'inline': %s",
2191 command_buf.buf);
2192 if (p != uq.buf) {
2193 strbuf_addstr(&uq, p);
2194 p = uq.buf;
2195 }
2196 read_next_command();
2197 parse_and_store_blob(&last_blob, sha1, 0);
2198 } else {
2199 enum object_type expected = S_ISDIR(mode) ?
2200 OBJ_TREE: OBJ_BLOB;
2201 enum object_type type = oe ? oe->type :
2202 sha1_object_info(sha1, NULL);
2203 if (type < 0)
2204 die("%s not found: %s",
2205 S_ISDIR(mode) ? "Tree" : "Blob",
2206 command_buf.buf);
2207 if (type != expected)
2208 die("Not a %s (actually a %s): %s",
2209 typename(expected), typename(type),
2210 command_buf.buf);
2211 }
2212
2213 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2214}
2215
2216static void file_change_d(struct branch *b)
2217{
2218 const char *p = command_buf.buf + 2;
2219 static struct strbuf uq = STRBUF_INIT;
2220 const char *endp;
2221
2222 strbuf_reset(&uq);
2223 if (!unquote_c_style(&uq, p, &endp)) {
2224 if (*endp)
2225 die("Garbage after path in: %s", command_buf.buf);
2226 p = uq.buf;
2227 }
2228 tree_content_remove(&b->branch_tree, p, NULL);
2229}
2230
2231static void file_change_cr(struct branch *b, int rename)
2232{
2233 const char *s, *d;
2234 static struct strbuf s_uq = STRBUF_INIT;
2235 static struct strbuf d_uq = STRBUF_INIT;
2236 const char *endp;
2237 struct tree_entry leaf;
2238
2239 s = command_buf.buf + 2;
2240 strbuf_reset(&s_uq);
2241 if (!unquote_c_style(&s_uq, s, &endp)) {
2242 if (*endp != ' ')
2243 die("Missing space after source: %s", command_buf.buf);
2244 } else {
2245 endp = strchr(s, ' ');
2246 if (!endp)
2247 die("Missing space after source: %s", command_buf.buf);
2248 strbuf_add(&s_uq, s, endp - s);
2249 }
2250 s = s_uq.buf;
2251
2252 endp++;
2253 if (!*endp)
2254 die("Missing dest: %s", command_buf.buf);
2255
2256 d = endp;
2257 strbuf_reset(&d_uq);
2258 if (!unquote_c_style(&d_uq, d, &endp)) {
2259 if (*endp)
2260 die("Garbage after dest in: %s", command_buf.buf);
2261 d = d_uq.buf;
2262 }
2263
2264 memset(&leaf, 0, sizeof(leaf));
2265 if (rename)
2266 tree_content_remove(&b->branch_tree, s, &leaf);
2267 else
2268 tree_content_get(&b->branch_tree, s, &leaf);
2269 if (!leaf.versions[1].mode)
2270 die("Path %s not in branch", s);
2271 tree_content_set(&b->branch_tree, d,
2272 leaf.versions[1].sha1,
2273 leaf.versions[1].mode,
2274 leaf.tree);
2275}
2276
2277static void note_change_n(struct branch *b, unsigned char old_fanout)
2278{
2279 const char *p = command_buf.buf + 2;
2280 static struct strbuf uq = STRBUF_INIT;
2281 struct object_entry *oe = oe;
2282 struct branch *s;
2283 unsigned char sha1[20], commit_sha1[20];
2284 char path[60];
2285 uint16_t inline_data = 0;
2286 unsigned char new_fanout;
2287
2288 /* <dataref> or 'inline' */
2289 if (*p == ':') {
2290 char *x;
2291 oe = find_mark(strtoumax(p + 1, &x, 10));
2292 hashcpy(sha1, oe->idx.sha1);
2293 p = x;
2294 } else if (!prefixcmp(p, "inline")) {
2295 inline_data = 1;
2296 p += 6;
2297 } else {
2298 if (get_sha1_hex(p, sha1))
2299 die("Invalid SHA1: %s", command_buf.buf);
2300 oe = find_object(sha1);
2301 p += 40;
2302 }
2303 if (*p++ != ' ')
2304 die("Missing space after SHA1: %s", command_buf.buf);
2305
2306 /* <committish> */
2307 s = lookup_branch(p);
2308 if (s) {
2309 hashcpy(commit_sha1, s->sha1);
2310 } else if (*p == ':') {
2311 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2312 struct object_entry *commit_oe = find_mark(commit_mark);
2313 if (commit_oe->type != OBJ_COMMIT)
2314 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2315 hashcpy(commit_sha1, commit_oe->idx.sha1);
2316 } else if (!get_sha1(p, commit_sha1)) {
2317 unsigned long size;
2318 char *buf = read_object_with_reference(commit_sha1,
2319 commit_type, &size, commit_sha1);
2320 if (!buf || size < 46)
2321 die("Not a valid commit: %s", p);
2322 free(buf);
2323 } else
2324 die("Invalid ref name or SHA1 expression: %s", p);
2325
2326 if (inline_data) {
2327 if (p != uq.buf) {
2328 strbuf_addstr(&uq, p);
2329 p = uq.buf;
2330 }
2331 read_next_command();
2332 parse_and_store_blob(&last_blob, sha1, 0);
2333 } else if (oe) {
2334 if (oe->type != OBJ_BLOB)
2335 die("Not a blob (actually a %s): %s",
2336 typename(oe->type), command_buf.buf);
2337 } else if (!is_null_sha1(sha1)) {
2338 enum object_type type = sha1_object_info(sha1, NULL);
2339 if (type < 0)
2340 die("Blob not found: %s", command_buf.buf);
2341 if (type != OBJ_BLOB)
2342 die("Not a blob (actually a %s): %s",
2343 typename(type), command_buf.buf);
2344 }
2345
2346 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2347 if (tree_content_remove(&b->branch_tree, path, NULL))
2348 b->num_notes--;
2349
2350 if (is_null_sha1(sha1))
2351 return; /* nothing to insert */
2352
2353 b->num_notes++;
2354 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2355 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2356 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2357}
2358
2359static void file_change_deleteall(struct branch *b)
2360{
2361 release_tree_content_recursive(b->branch_tree.tree);
2362 hashclr(b->branch_tree.versions[0].sha1);
2363 hashclr(b->branch_tree.versions[1].sha1);
2364 load_tree(&b->branch_tree);
2365 b->num_notes = 0;
2366}
2367
2368static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2369{
2370 if (!buf || size < 46)
2371 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2372 if (memcmp("tree ", buf, 5)
2373 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2374 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2375 hashcpy(b->branch_tree.versions[0].sha1,
2376 b->branch_tree.versions[1].sha1);
2377}
2378
2379static void parse_from_existing(struct branch *b)
2380{
2381 if (is_null_sha1(b->sha1)) {
2382 hashclr(b->branch_tree.versions[0].sha1);
2383 hashclr(b->branch_tree.versions[1].sha1);
2384 } else {
2385 unsigned long size;
2386 char *buf;
2387
2388 buf = read_object_with_reference(b->sha1,
2389 commit_type, &size, b->sha1);
2390 parse_from_commit(b, buf, size);
2391 free(buf);
2392 }
2393}
2394
2395static int parse_from(struct branch *b)
2396{
2397 const char *from;
2398 struct branch *s;
2399
2400 if (prefixcmp(command_buf.buf, "from "))
2401 return 0;
2402
2403 if (b->branch_tree.tree) {
2404 release_tree_content_recursive(b->branch_tree.tree);
2405 b->branch_tree.tree = NULL;
2406 }
2407
2408 from = strchr(command_buf.buf, ' ') + 1;
2409 s = lookup_branch(from);
2410 if (b == s)
2411 die("Can't create a branch from itself: %s", b->name);
2412 else if (s) {
2413 unsigned char *t = s->branch_tree.versions[1].sha1;
2414 hashcpy(b->sha1, s->sha1);
2415 hashcpy(b->branch_tree.versions[0].sha1, t);
2416 hashcpy(b->branch_tree.versions[1].sha1, t);
2417 } else if (*from == ':') {
2418 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2419 struct object_entry *oe = find_mark(idnum);
2420 if (oe->type != OBJ_COMMIT)
2421 die("Mark :%" PRIuMAX " not a commit", idnum);
2422 hashcpy(b->sha1, oe->idx.sha1);
2423 if (oe->pack_id != MAX_PACK_ID) {
2424 unsigned long size;
2425 char *buf = gfi_unpack_entry(oe, &size);
2426 parse_from_commit(b, buf, size);
2427 free(buf);
2428 } else
2429 parse_from_existing(b);
2430 } else if (!get_sha1(from, b->sha1))
2431 parse_from_existing(b);
2432 else
2433 die("Invalid ref name or SHA1 expression: %s", from);
2434
2435 read_next_command();
2436 return 1;
2437}
2438
2439static struct hash_list *parse_merge(unsigned int *count)
2440{
2441 struct hash_list *list = NULL, *n, *e = e;
2442 const char *from;
2443 struct branch *s;
2444
2445 *count = 0;
2446 while (!prefixcmp(command_buf.buf, "merge ")) {
2447 from = strchr(command_buf.buf, ' ') + 1;
2448 n = xmalloc(sizeof(*n));
2449 s = lookup_branch(from);
2450 if (s)
2451 hashcpy(n->sha1, s->sha1);
2452 else if (*from == ':') {
2453 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2454 struct object_entry *oe = find_mark(idnum);
2455 if (oe->type != OBJ_COMMIT)
2456 die("Mark :%" PRIuMAX " not a commit", idnum);
2457 hashcpy(n->sha1, oe->idx.sha1);
2458 } else if (!get_sha1(from, n->sha1)) {
2459 unsigned long size;
2460 char *buf = read_object_with_reference(n->sha1,
2461 commit_type, &size, n->sha1);
2462 if (!buf || size < 46)
2463 die("Not a valid commit: %s", from);
2464 free(buf);
2465 } else
2466 die("Invalid ref name or SHA1 expression: %s", from);
2467
2468 n->next = NULL;
2469 if (list)
2470 e->next = n;
2471 else
2472 list = n;
2473 e = n;
2474 (*count)++;
2475 read_next_command();
2476 }
2477 return list;
2478}
2479
2480static void parse_new_commit(void)
2481{
2482 static struct strbuf msg = STRBUF_INIT;
2483 struct branch *b;
2484 char *sp;
2485 char *author = NULL;
2486 char *committer = NULL;
2487 struct hash_list *merge_list = NULL;
2488 unsigned int merge_count;
2489 unsigned char prev_fanout, new_fanout;
2490
2491 /* Obtain the branch name from the rest of our command */
2492 sp = strchr(command_buf.buf, ' ') + 1;
2493 b = lookup_branch(sp);
2494 if (!b)
2495 b = new_branch(sp);
2496
2497 read_next_command();
2498 parse_mark();
2499 if (!prefixcmp(command_buf.buf, "author ")) {
2500 author = parse_ident(command_buf.buf + 7);
2501 read_next_command();
2502 }
2503 if (!prefixcmp(command_buf.buf, "committer ")) {
2504 committer = parse_ident(command_buf.buf + 10);
2505 read_next_command();
2506 }
2507 if (!committer)
2508 die("Expected committer but didn't get one");
2509 parse_data(&msg, 0, NULL);
2510 read_next_command();
2511 parse_from(b);
2512 merge_list = parse_merge(&merge_count);
2513
2514 /* ensure the branch is active/loaded */
2515 if (!b->branch_tree.tree || !max_active_branches) {
2516 unload_one_branch();
2517 load_branch(b);
2518 }
2519
2520 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2521
2522 /* file_change* */
2523 while (command_buf.len > 0) {
2524 if (!prefixcmp(command_buf.buf, "M "))
2525 file_change_m(b);
2526 else if (!prefixcmp(command_buf.buf, "D "))
2527 file_change_d(b);
2528 else if (!prefixcmp(command_buf.buf, "R "))
2529 file_change_cr(b, 1);
2530 else if (!prefixcmp(command_buf.buf, "C "))
2531 file_change_cr(b, 0);
2532 else if (!prefixcmp(command_buf.buf, "N "))
2533 note_change_n(b, prev_fanout);
2534 else if (!strcmp("deleteall", command_buf.buf))
2535 file_change_deleteall(b);
2536 else {
2537 unread_command_buf = 1;
2538 break;
2539 }
2540 if (read_next_command() == EOF)
2541 break;
2542 }
2543
2544 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2545 if (new_fanout != prev_fanout)
2546 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2547
2548 /* build the tree and the commit */
2549 store_tree(&b->branch_tree);
2550 hashcpy(b->branch_tree.versions[0].sha1,
2551 b->branch_tree.versions[1].sha1);
2552
2553 strbuf_reset(&new_data);
2554 strbuf_addf(&new_data, "tree %s\n",
2555 sha1_to_hex(b->branch_tree.versions[1].sha1));
2556 if (!is_null_sha1(b->sha1))
2557 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2558 while (merge_list) {
2559 struct hash_list *next = merge_list->next;
2560 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2561 free(merge_list);
2562 merge_list = next;
2563 }
2564 strbuf_addf(&new_data,
2565 "author %s\n"
2566 "committer %s\n"
2567 "\n",
2568 author ? author : committer, committer);
2569 strbuf_addbuf(&new_data, &msg);
2570 free(author);
2571 free(committer);
2572
2573 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2574 b->pack_id = pack_id;
2575 b->last_commit = object_count_by_type[OBJ_COMMIT];
2576}
2577
2578static void parse_new_tag(void)
2579{
2580 static struct strbuf msg = STRBUF_INIT;
2581 char *sp;
2582 const char *from;
2583 char *tagger;
2584 struct branch *s;
2585 struct tag *t;
2586 uintmax_t from_mark = 0;
2587 unsigned char sha1[20];
2588 enum object_type type;
2589
2590 /* Obtain the new tag name from the rest of our command */
2591 sp = strchr(command_buf.buf, ' ') + 1;
2592 t = pool_alloc(sizeof(struct tag));
2593 t->next_tag = NULL;
2594 t->name = pool_strdup(sp);
2595 if (last_tag)
2596 last_tag->next_tag = t;
2597 else
2598 first_tag = t;
2599 last_tag = t;
2600 read_next_command();
2601
2602 /* from ... */
2603 if (prefixcmp(command_buf.buf, "from "))
2604 die("Expected from command, got %s", command_buf.buf);
2605 from = strchr(command_buf.buf, ' ') + 1;
2606 s = lookup_branch(from);
2607 if (s) {
2608 hashcpy(sha1, s->sha1);
2609 type = OBJ_COMMIT;
2610 } else if (*from == ':') {
2611 struct object_entry *oe;
2612 from_mark = strtoumax(from + 1, NULL, 10);
2613 oe = find_mark(from_mark);
2614 type = oe->type;
2615 hashcpy(sha1, oe->idx.sha1);
2616 } else if (!get_sha1(from, sha1)) {
2617 unsigned long size;
2618 char *buf;
2619
2620 buf = read_sha1_file(sha1, &type, &size);
2621 if (!buf || size < 46)
2622 die("Not a valid commit: %s", from);
2623 free(buf);
2624 } else
2625 die("Invalid ref name or SHA1 expression: %s", from);
2626 read_next_command();
2627
2628 /* tagger ... */
2629 if (!prefixcmp(command_buf.buf, "tagger ")) {
2630 tagger = parse_ident(command_buf.buf + 7);
2631 read_next_command();
2632 } else
2633 tagger = NULL;
2634
2635 /* tag payload/message */
2636 parse_data(&msg, 0, NULL);
2637
2638 /* build the tag object */
2639 strbuf_reset(&new_data);
2640
2641 strbuf_addf(&new_data,
2642 "object %s\n"
2643 "type %s\n"
2644 "tag %s\n",
2645 sha1_to_hex(sha1), typename(type), t->name);
2646 if (tagger)
2647 strbuf_addf(&new_data,
2648 "tagger %s\n", tagger);
2649 strbuf_addch(&new_data, '\n');
2650 strbuf_addbuf(&new_data, &msg);
2651 free(tagger);
2652
2653 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2654 t->pack_id = MAX_PACK_ID;
2655 else
2656 t->pack_id = pack_id;
2657}
2658
2659static void parse_reset_branch(void)
2660{
2661 struct branch *b;
2662 char *sp;
2663
2664 /* Obtain the branch name from the rest of our command */
2665 sp = strchr(command_buf.buf, ' ') + 1;
2666 b = lookup_branch(sp);
2667 if (b) {
2668 hashclr(b->sha1);
2669 hashclr(b->branch_tree.versions[0].sha1);
2670 hashclr(b->branch_tree.versions[1].sha1);
2671 if (b->branch_tree.tree) {
2672 release_tree_content_recursive(b->branch_tree.tree);
2673 b->branch_tree.tree = NULL;
2674 }
2675 }
2676 else
2677 b = new_branch(sp);
2678 read_next_command();
2679 parse_from(b);
2680 if (command_buf.len > 0)
2681 unread_command_buf = 1;
2682}
2683
2684static void parse_checkpoint(void)
2685{
2686 if (object_count) {
2687 cycle_packfile();
2688 dump_branches();
2689 dump_tags();
2690 dump_marks();
2691 }
2692 skip_optional_lf();
2693}
2694
2695static void parse_progress(void)
2696{
2697 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2698 fputc('\n', stdout);
2699 fflush(stdout);
2700 skip_optional_lf();
2701}
2702
2703static char* make_fast_import_path(const char *path)
2704{
2705 struct strbuf abs_path = STRBUF_INIT;
2706
2707 if (!relative_marks_paths || is_absolute_path(path))
2708 return xstrdup(path);
2709 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
2710 return strbuf_detach(&abs_path, NULL);
2711}
2712
2713static void option_import_marks(const char *marks, int from_stream)
2714{
2715 if (import_marks_file) {
2716 if (from_stream)
2717 die("Only one import-marks command allowed per stream");
2718
2719 /* read previous mark file */
2720 if(!import_marks_file_from_stream)
2721 read_marks();
2722 }
2723
2724 import_marks_file = make_fast_import_path(marks);
2725 safe_create_leading_directories_const(import_marks_file);
2726 import_marks_file_from_stream = from_stream;
2727}
2728
2729static void option_date_format(const char *fmt)
2730{
2731 if (!strcmp(fmt, "raw"))
2732 whenspec = WHENSPEC_RAW;
2733 else if (!strcmp(fmt, "rfc2822"))
2734 whenspec = WHENSPEC_RFC2822;
2735 else if (!strcmp(fmt, "now"))
2736 whenspec = WHENSPEC_NOW;
2737 else
2738 die("unknown --date-format argument %s", fmt);
2739}
2740
2741static void option_depth(const char *depth)
2742{
2743 max_depth = strtoul(depth, NULL, 0);
2744 if (max_depth > MAX_DEPTH)
2745 die("--depth cannot exceed %u", MAX_DEPTH);
2746}
2747
2748static void option_active_branches(const char *branches)
2749{
2750 max_active_branches = strtoul(branches, NULL, 0);
2751}
2752
2753static void option_export_marks(const char *marks)
2754{
2755 export_marks_file = make_fast_import_path(marks);
2756 safe_create_leading_directories_const(export_marks_file);
2757}
2758
2759static void option_export_pack_edges(const char *edges)
2760{
2761 if (pack_edges)
2762 fclose(pack_edges);
2763 pack_edges = fopen(edges, "a");
2764 if (!pack_edges)
2765 die_errno("Cannot open '%s'", edges);
2766}
2767
2768static int parse_one_option(const char *option)
2769{
2770 if (!prefixcmp(option, "max-pack-size=")) {
2771 unsigned long v;
2772 if (!git_parse_ulong(option + 14, &v))
2773 return 0;
2774 if (v < 8192) {
2775 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
2776 v *= 1024 * 1024;
2777 } else if (v < 1024 * 1024) {
2778 warning("minimum max-pack-size is 1 MiB");
2779 v = 1024 * 1024;
2780 }
2781 max_packsize = v;
2782 } else if (!prefixcmp(option, "big-file-threshold=")) {
2783 unsigned long v;
2784 if (!git_parse_ulong(option + 19, &v))
2785 return 0;
2786 big_file_threshold = v;
2787 } else if (!prefixcmp(option, "depth=")) {
2788 option_depth(option + 6);
2789 } else if (!prefixcmp(option, "active-branches=")) {
2790 option_active_branches(option + 16);
2791 } else if (!prefixcmp(option, "export-pack-edges=")) {
2792 option_export_pack_edges(option + 18);
2793 } else if (!prefixcmp(option, "quiet")) {
2794 show_stats = 0;
2795 } else if (!prefixcmp(option, "stats")) {
2796 show_stats = 1;
2797 } else {
2798 return 0;
2799 }
2800
2801 return 1;
2802}
2803
2804static int parse_one_feature(const char *feature, int from_stream)
2805{
2806 if (!prefixcmp(feature, "date-format=")) {
2807 option_date_format(feature + 12);
2808 } else if (!prefixcmp(feature, "import-marks=")) {
2809 option_import_marks(feature + 13, from_stream);
2810 } else if (!prefixcmp(feature, "export-marks=")) {
2811 option_export_marks(feature + 13);
2812 } else if (!prefixcmp(feature, "relative-marks")) {
2813 relative_marks_paths = 1;
2814 } else if (!prefixcmp(feature, "no-relative-marks")) {
2815 relative_marks_paths = 0;
2816 } else if (!prefixcmp(feature, "force")) {
2817 force_update = 1;
2818 } else if (!strcmp(feature, "notes")) {
2819 ; /* do nothing; we have the feature */
2820 } else {
2821 return 0;
2822 }
2823
2824 return 1;
2825}
2826
2827static void parse_feature(void)
2828{
2829 char *feature = command_buf.buf + 8;
2830
2831 if (seen_data_command)
2832 die("Got feature command '%s' after data command", feature);
2833
2834 if (parse_one_feature(feature, 1))
2835 return;
2836
2837 die("This version of fast-import does not support feature %s.", feature);
2838}
2839
2840static void parse_option(void)
2841{
2842 char *option = command_buf.buf + 11;
2843
2844 if (seen_data_command)
2845 die("Got option command '%s' after data command", option);
2846
2847 if (parse_one_option(option))
2848 return;
2849
2850 die("This version of fast-import does not support option: %s", option);
2851}
2852
2853static int git_pack_config(const char *k, const char *v, void *cb)
2854{
2855 if (!strcmp(k, "pack.depth")) {
2856 max_depth = git_config_int(k, v);
2857 if (max_depth > MAX_DEPTH)
2858 max_depth = MAX_DEPTH;
2859 return 0;
2860 }
2861 if (!strcmp(k, "pack.compression")) {
2862 int level = git_config_int(k, v);
2863 if (level == -1)
2864 level = Z_DEFAULT_COMPRESSION;
2865 else if (level < 0 || level > Z_BEST_COMPRESSION)
2866 die("bad pack compression level %d", level);
2867 pack_compression_level = level;
2868 pack_compression_seen = 1;
2869 return 0;
2870 }
2871 if (!strcmp(k, "pack.indexversion")) {
2872 pack_idx_default_version = git_config_int(k, v);
2873 if (pack_idx_default_version > 2)
2874 die("bad pack.indexversion=%"PRIu32,
2875 pack_idx_default_version);
2876 return 0;
2877 }
2878 if (!strcmp(k, "pack.packsizelimit")) {
2879 max_packsize = git_config_ulong(k, v);
2880 return 0;
2881 }
2882 if (!strcmp(k, "core.bigfilethreshold")) {
2883 long n = git_config_int(k, v);
2884 big_file_threshold = 0 < n ? n : 0;
2885 }
2886 return git_default_config(k, v, cb);
2887}
2888
2889static const char fast_import_usage[] =
2890"git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
2891
2892static void parse_argv(void)
2893{
2894 unsigned int i;
2895
2896 for (i = 1; i < global_argc; i++) {
2897 const char *a = global_argv[i];
2898
2899 if (*a != '-' || !strcmp(a, "--"))
2900 break;
2901
2902 if (parse_one_option(a + 2))
2903 continue;
2904
2905 if (parse_one_feature(a + 2, 0))
2906 continue;
2907
2908 die("unknown option %s", a);
2909 }
2910 if (i != global_argc)
2911 usage(fast_import_usage);
2912
2913 seen_data_command = 1;
2914 if (import_marks_file)
2915 read_marks();
2916}
2917
2918int main(int argc, const char **argv)
2919{
2920 unsigned int i;
2921
2922 git_extract_argv0_path(argv[0]);
2923
2924 if (argc == 2 && !strcmp(argv[1], "-h"))
2925 usage(fast_import_usage);
2926
2927 setup_git_directory();
2928 git_config(git_pack_config, NULL);
2929 if (!pack_compression_seen && core_compression_seen)
2930 pack_compression_level = core_compression_level;
2931
2932 alloc_objects(object_entry_alloc);
2933 strbuf_init(&command_buf, 0);
2934 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2935 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2936 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2937 marks = pool_calloc(1, sizeof(struct mark_set));
2938
2939 global_argc = argc;
2940 global_argv = argv;
2941
2942 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2943 for (i = 0; i < (cmd_save - 1); i++)
2944 rc_free[i].next = &rc_free[i + 1];
2945 rc_free[cmd_save - 1].next = NULL;
2946
2947 prepare_packed_git();
2948 start_packfile();
2949 set_die_routine(die_nicely);
2950 while (read_next_command() != EOF) {
2951 if (!strcmp("blob", command_buf.buf))
2952 parse_new_blob();
2953 else if (!prefixcmp(command_buf.buf, "commit "))
2954 parse_new_commit();
2955 else if (!prefixcmp(command_buf.buf, "tag "))
2956 parse_new_tag();
2957 else if (!prefixcmp(command_buf.buf, "reset "))
2958 parse_reset_branch();
2959 else if (!strcmp("checkpoint", command_buf.buf))
2960 parse_checkpoint();
2961 else if (!prefixcmp(command_buf.buf, "progress "))
2962 parse_progress();
2963 else if (!prefixcmp(command_buf.buf, "feature "))
2964 parse_feature();
2965 else if (!prefixcmp(command_buf.buf, "option git "))
2966 parse_option();
2967 else if (!prefixcmp(command_buf.buf, "option "))
2968 /* ignore non-git options*/;
2969 else
2970 die("Unsupported command: %s", command_buf.buf);
2971 }
2972
2973 /* argv hasn't been parsed yet, do so */
2974 if (!seen_data_command)
2975 parse_argv();
2976
2977 end_packfile();
2978
2979 dump_branches();
2980 dump_tags();
2981 unkeep_all_packs();
2982 dump_marks();
2983
2984 if (pack_edges)
2985 fclose(pack_edges);
2986
2987 if (show_stats) {
2988 uintmax_t total_count = 0, duplicate_count = 0;
2989 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2990 total_count += object_count_by_type[i];
2991 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2992 duplicate_count += duplicate_count_by_type[i];
2993
2994 fprintf(stderr, "%s statistics:\n", argv[0]);
2995 fprintf(stderr, "---------------------------------------------------------------------\n");
2996 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2997 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2998 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]);
2999 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]);
3000 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]);
3001 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]);
3002 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3003 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3004 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3005 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3006 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3007 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3008 fprintf(stderr, "---------------------------------------------------------------------\n");
3009 pack_report();
3010 fprintf(stderr, "---------------------------------------------------------------------\n");
3011 fprintf(stderr, "\n");
3012 }
3013
3014 return failure ? 1 : 0;
3015}