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