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 commit-ish lf)?
26 ('merge' sp commit-ish lf)*
27 (file_change | ls)*
28 lf?;
29 commit_msg ::= data;
30
31 ls ::= 'ls' sp '"' quoted(path) '"' lf;
32
33 file_change ::= file_clr
34 | file_del
35 | file_rnm
36 | file_cpy
37 | file_obm
38 | file_inm;
39 file_clr ::= 'deleteall' lf;
40 file_del ::= 'D' sp path_str lf;
41 file_rnm ::= 'R' sp path_str sp path_str lf;
42 file_cpy ::= 'C' sp path_str sp path_str lf;
43 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
44 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
45 data;
46 note_obm ::= 'N' sp (hexsha1 | idnum) sp commit-ish lf;
47 note_inm ::= 'N' sp 'inline' sp commit-ish lf
48 data;
49
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp commit-ish lf
52 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
53 tag_msg;
54 tag_msg ::= data;
55
56 reset_branch ::= 'reset' sp ref_str lf
57 ('from' sp commit-ish lf)?
58 lf?;
59
60 checkpoint ::= 'checkpoint' lf
61 lf?;
62
63 progress ::= 'progress' sp not_lf* lf
64 lf?;
65
66 # note: the first idnum in a stream should be 1 and subsequent
67 # idnums should not have gaps between values as this will cause
68 # the stream parser to reserve space for the gapped values. An
69 # idnum can be updated in the future to a new object by issuing
70 # a new mark directive with the old idnum.
71 #
72 mark ::= 'mark' sp idnum lf;
73 data ::= (delimited_data | exact_data)
74 lf?;
75
76 # note: delim may be any string but must not contain lf.
77 # data_line may contain any data but must not be exactly
78 # delim.
79 delimited_data ::= 'data' sp '<<' delim lf
80 (data_line lf)*
81 delim lf;
82
83 # note: declen indicates the length of binary_data in bytes.
84 # declen does not include the lf preceding the binary data.
85 #
86 exact_data ::= 'data' sp declen lf
87 binary_data;
88
89 # note: quoted strings are C-style quoting supporting \c for
90 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
91 # is the signed byte value in octal. Note that the only
92 # characters which must actually be escaped to protect the
93 # stream formatting is: \, " and LF. Otherwise these values
94 # are UTF8.
95 #
96 commit-ish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
97 ref_str ::= ref;
98 sha1exp_str ::= sha1exp;
99 tag_str ::= tag;
100 path_str ::= path | '"' quoted(path) '"' ;
101 mode ::= '100644' | '644'
102 | '100755' | '755'
103 | '120000'
104 ;
105
106 declen ::= # unsigned 32 bit value, ascii base10 notation;
107 bigint ::= # unsigned integer value, ascii base10 notation;
108 binary_data ::= # file content, not interpreted;
109
110 when ::= raw_when | rfc2822_when;
111 raw_when ::= ts sp tz;
112 rfc2822_when ::= # Valid RFC 2822 date and time;
113
114 sp ::= # ASCII space character;
115 lf ::= # ASCII newline (LF) character;
116
117 # note: a colon (':') must precede the numerical value assigned to
118 # an idnum. This is to distinguish it from a ref or tag name as
119 # GIT does not permit ':' in ref or tag strings.
120 #
121 idnum ::= ':' bigint;
122 path ::= # GIT style file path, e.g. "a/b/c";
123 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
124 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
125 sha1exp ::= # Any valid GIT SHA1 expression;
126 hexsha1 ::= # SHA1 in hexadecimal format;
127
128 # note: name and email are UTF8 strings, however name must not
129 # contain '<' or lf and email must not contain any of the
130 # following: '<', '>', lf.
131 #
132 name ::= # valid GIT author/committer name;
133 email ::= # valid GIT author/committer email;
134 ts ::= # time since the epoch in seconds, ascii base10 notation;
135 tz ::= # GIT style timezone;
136
137 # note: comments, ls and cat requests may appear anywhere
138 # in the input, except within a data command. Any form
139 # of the data command always escapes the related input
140 # from comment processing.
141 #
142 # In case it is not clear, the '#' that starts the comment
143 # must be the first character on that line (an lf
144 # preceded it).
145 #
146
147 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
148 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
149
150 comment ::= '#' not_lf* lf;
151 not_lf ::= # Any byte that is not ASCII newline (LF);
152*/
153
154#include "builtin.h"
155#include "cache.h"
156#include "object.h"
157#include "blob.h"
158#include "tree.h"
159#include "commit.h"
160#include "delta.h"
161#include "pack.h"
162#include "refs.h"
163#include "csum-file.h"
164#include "quote.h"
165#include "exec_cmd.h"
166#include "dir.h"
167
168#define PACK_ID_BITS 16
169#define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
170#define DEPTH_BITS 13
171#define MAX_DEPTH ((1<<DEPTH_BITS)-1)
172
173/*
174 * We abuse the setuid bit on directories to mean "do not delta".
175 */
176#define NO_DELTA S_ISUID
177
178struct object_entry {
179 struct pack_idx_entry idx;
180 struct object_entry *next;
181 uint32_t type : TYPE_BITS,
182 pack_id : PACK_ID_BITS,
183 depth : DEPTH_BITS;
184};
185
186struct object_entry_pool {
187 struct object_entry_pool *next_pool;
188 struct object_entry *next_free;
189 struct object_entry *end;
190 struct object_entry entries[FLEX_ARRAY]; /* more */
191};
192
193struct mark_set {
194 union {
195 struct object_entry *marked[1024];
196 struct mark_set *sets[1024];
197 } data;
198 unsigned int shift;
199};
200
201struct last_object {
202 struct strbuf data;
203 off_t offset;
204 unsigned int depth;
205 unsigned no_swap : 1;
206};
207
208struct mem_pool {
209 struct mem_pool *next_pool;
210 char *next_free;
211 char *end;
212 uintmax_t space[FLEX_ARRAY]; /* more */
213};
214
215struct atom_str {
216 struct atom_str *next_atom;
217 unsigned short str_len;
218 char str_dat[FLEX_ARRAY]; /* more */
219};
220
221struct tree_content;
222struct tree_entry {
223 struct tree_content *tree;
224 struct atom_str *name;
225 struct tree_entry_ms {
226 uint16_t mode;
227 unsigned char sha1[20];
228 } versions[2];
229};
230
231struct tree_content {
232 unsigned int entry_capacity; /* must match avail_tree_content */
233 unsigned int entry_count;
234 unsigned int delta_depth;
235 struct tree_entry *entries[FLEX_ARRAY]; /* more */
236};
237
238struct avail_tree_content {
239 unsigned int entry_capacity; /* must match tree_content */
240 struct avail_tree_content *next_avail;
241};
242
243struct branch {
244 struct branch *table_next_branch;
245 struct branch *active_next_branch;
246 const char *name;
247 struct tree_entry branch_tree;
248 uintmax_t last_commit;
249 uintmax_t num_notes;
250 unsigned active : 1;
251 unsigned pack_id : PACK_ID_BITS;
252 unsigned char sha1[20];
253};
254
255struct tag {
256 struct tag *next_tag;
257 const char *name;
258 unsigned int pack_id;
259 unsigned char sha1[20];
260};
261
262struct hash_list {
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 struct recent_command *prev;
275 struct recent_command *next;
276 char *buf;
277};
278
279/* Configured limits on output */
280static unsigned long max_depth = 10;
281static off_t max_packsize;
282static int force_update;
283static int pack_compression_level = Z_DEFAULT_COMPRESSION;
284static int pack_compression_seen;
285
286/* Stats and misc. counters */
287static uintmax_t alloc_count;
288static uintmax_t marks_set_count;
289static uintmax_t object_count_by_type[1 << TYPE_BITS];
290static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
291static uintmax_t delta_count_by_type[1 << TYPE_BITS];
292static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
293static unsigned long object_count;
294static unsigned long branch_count;
295static unsigned long branch_load_count;
296static int failure;
297static FILE *pack_edges;
298static unsigned int show_stats = 1;
299static int global_argc;
300static char **global_argv;
301
302/* Memory pools */
303static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
304static size_t total_allocd;
305static struct mem_pool *mem_pool;
306
307/* Atom management */
308static unsigned int atom_table_sz = 4451;
309static unsigned int atom_cnt;
310static struct atom_str **atom_table;
311
312/* The .pack file being generated */
313static struct pack_idx_option pack_idx_opts;
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 import_marks_file_ignore_missing;
329static int relative_marks_paths;
330
331/* Our last blob */
332static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
333
334/* Tree management */
335static unsigned int tree_entry_alloc = 1000;
336static void *avail_tree_entry;
337static unsigned int avail_tree_table_sz = 100;
338static struct avail_tree_content **avail_tree_table;
339static struct strbuf old_tree = STRBUF_INIT;
340static struct strbuf new_tree = STRBUF_INIT;
341
342/* Branch data */
343static unsigned long max_active_branches = 5;
344static unsigned long cur_active_branches;
345static unsigned long branch_table_sz = 1039;
346static struct branch **branch_table;
347static struct branch *active_branches;
348
349/* Tag data */
350static struct tag *first_tag;
351static struct tag *last_tag;
352
353/* Input stream parsing */
354static whenspec_type whenspec = WHENSPEC_RAW;
355static struct strbuf command_buf = STRBUF_INIT;
356static int unread_command_buf;
357static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
358static struct recent_command *cmd_tail = &cmd_hist;
359static struct recent_command *rc_free;
360static unsigned int cmd_save = 100;
361static uintmax_t next_mark;
362static struct strbuf new_data = STRBUF_INIT;
363static int seen_data_command;
364static int require_explicit_termination;
365
366/* Signal handling */
367static volatile sig_atomic_t checkpoint_requested;
368
369/* Where to write output of cat-blob commands */
370static int cat_blob_fd = STDOUT_FILENO;
371
372static void parse_argv(void);
373static void parse_cat_blob(void);
374static void parse_ls(struct branch *b);
375
376static void write_branch_report(FILE *rpt, struct branch *b)
377{
378 fprintf(rpt, "%s:\n", b->name);
379
380 fprintf(rpt, " status :");
381 if (b->active)
382 fputs(" active", rpt);
383 if (b->branch_tree.tree)
384 fputs(" loaded", rpt);
385 if (is_null_sha1(b->branch_tree.versions[1].sha1))
386 fputs(" dirty", rpt);
387 fputc('\n', rpt);
388
389 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
390 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
391 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
392 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
393
394 fputs(" last pack : ", rpt);
395 if (b->pack_id < MAX_PACK_ID)
396 fprintf(rpt, "%u", b->pack_id);
397 fputc('\n', rpt);
398
399 fputc('\n', rpt);
400}
401
402static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
403
404static void write_crash_report(const char *err)
405{
406 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
407 FILE *rpt = fopen(loc, "w");
408 struct branch *b;
409 unsigned long lu;
410 struct recent_command *rc;
411
412 if (!rpt) {
413 error("can't write crash report %s: %s", loc, strerror(errno));
414 return;
415 }
416
417 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
418
419 fprintf(rpt, "fast-import crash report:\n");
420 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
421 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
422 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
423 fputc('\n', rpt);
424
425 fputs("fatal: ", rpt);
426 fputs(err, rpt);
427 fputc('\n', rpt);
428
429 fputc('\n', rpt);
430 fputs("Most Recent Commands Before Crash\n", rpt);
431 fputs("---------------------------------\n", rpt);
432 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
433 if (rc->next == &cmd_hist)
434 fputs("* ", rpt);
435 else
436 fputs(" ", rpt);
437 fputs(rc->buf, rpt);
438 fputc('\n', rpt);
439 }
440
441 fputc('\n', rpt);
442 fputs("Active Branch LRU\n", rpt);
443 fputs("-----------------\n", rpt);
444 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
445 cur_active_branches,
446 max_active_branches);
447 fputc('\n', rpt);
448 fputs(" pos clock name\n", rpt);
449 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
450 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
451 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
452 ++lu, b->last_commit, b->name);
453
454 fputc('\n', rpt);
455 fputs("Inactive Branches\n", rpt);
456 fputs("-----------------\n", rpt);
457 for (lu = 0; lu < branch_table_sz; lu++) {
458 for (b = branch_table[lu]; b; b = b->table_next_branch)
459 write_branch_report(rpt, b);
460 }
461
462 if (first_tag) {
463 struct tag *tg;
464 fputc('\n', rpt);
465 fputs("Annotated Tags\n", rpt);
466 fputs("--------------\n", rpt);
467 for (tg = first_tag; tg; tg = tg->next_tag) {
468 fputs(sha1_to_hex(tg->sha1), rpt);
469 fputc(' ', rpt);
470 fputs(tg->name, rpt);
471 fputc('\n', rpt);
472 }
473 }
474
475 fputc('\n', rpt);
476 fputs("Marks\n", rpt);
477 fputs("-----\n", rpt);
478 if (export_marks_file)
479 fprintf(rpt, " exported to %s\n", export_marks_file);
480 else
481 dump_marks_helper(rpt, 0, marks);
482
483 fputc('\n', rpt);
484 fputs("-------------------\n", rpt);
485 fputs("END OF CRASH REPORT\n", rpt);
486 fclose(rpt);
487}
488
489static void end_packfile(void);
490static void unkeep_all_packs(void);
491static void dump_marks(void);
492
493static NORETURN void die_nicely(const char *err, va_list params)
494{
495 static int zombie;
496 char message[2 * PATH_MAX];
497
498 vsnprintf(message, sizeof(message), err, params);
499 fputs("fatal: ", stderr);
500 fputs(message, stderr);
501 fputc('\n', stderr);
502
503 if (!zombie) {
504 zombie = 1;
505 write_crash_report(message);
506 end_packfile();
507 unkeep_all_packs();
508 dump_marks();
509 }
510 exit(128);
511}
512
513#ifndef SIGUSR1 /* Windows, for example */
514
515static void set_checkpoint_signal(void)
516{
517}
518
519#else
520
521static void checkpoint_signal(int signo)
522{
523 checkpoint_requested = 1;
524}
525
526static void set_checkpoint_signal(void)
527{
528 struct sigaction sa;
529
530 memset(&sa, 0, sizeof(sa));
531 sa.sa_handler = checkpoint_signal;
532 sigemptyset(&sa.sa_mask);
533 sa.sa_flags = SA_RESTART;
534 sigaction(SIGUSR1, &sa, NULL);
535}
536
537#endif
538
539static void alloc_objects(unsigned int cnt)
540{
541 struct object_entry_pool *b;
542
543 b = xmalloc(sizeof(struct object_entry_pool)
544 + cnt * sizeof(struct object_entry));
545 b->next_pool = blocks;
546 b->next_free = b->entries;
547 b->end = b->entries + cnt;
548 blocks = b;
549 alloc_count += cnt;
550}
551
552static struct object_entry *new_object(unsigned char *sha1)
553{
554 struct object_entry *e;
555
556 if (blocks->next_free == blocks->end)
557 alloc_objects(object_entry_alloc);
558
559 e = blocks->next_free++;
560 hashcpy(e->idx.sha1, sha1);
561 return e;
562}
563
564static struct object_entry *find_object(unsigned char *sha1)
565{
566 unsigned int h = sha1[0] << 8 | sha1[1];
567 struct object_entry *e;
568 for (e = object_table[h]; e; e = e->next)
569 if (!hashcmp(sha1, e->idx.sha1))
570 return e;
571 return NULL;
572}
573
574static struct object_entry *insert_object(unsigned char *sha1)
575{
576 unsigned int h = sha1[0] << 8 | sha1[1];
577 struct object_entry *e = object_table[h];
578
579 while (e) {
580 if (!hashcmp(sha1, e->idx.sha1))
581 return e;
582 e = e->next;
583 }
584
585 e = new_object(sha1);
586 e->next = object_table[h];
587 e->idx.offset = 0;
588 object_table[h] = e;
589 return e;
590}
591
592static unsigned int hc_str(const char *s, size_t len)
593{
594 unsigned int r = 0;
595 while (len-- > 0)
596 r = r * 31 + *s++;
597 return r;
598}
599
600static void *pool_alloc(size_t len)
601{
602 struct mem_pool *p;
603 void *r;
604
605 /* round up to a 'uintmax_t' alignment */
606 if (len & (sizeof(uintmax_t) - 1))
607 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
608
609 for (p = mem_pool; p; p = p->next_pool)
610 if ((p->end - p->next_free >= len))
611 break;
612
613 if (!p) {
614 if (len >= (mem_pool_alloc/2)) {
615 total_allocd += len;
616 return xmalloc(len);
617 }
618 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
619 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
620 p->next_pool = mem_pool;
621 p->next_free = (char *) p->space;
622 p->end = p->next_free + mem_pool_alloc;
623 mem_pool = p;
624 }
625
626 r = p->next_free;
627 p->next_free += len;
628 return r;
629}
630
631static void *pool_calloc(size_t count, size_t size)
632{
633 size_t len = count * size;
634 void *r = pool_alloc(len);
635 memset(r, 0, len);
636 return r;
637}
638
639static char *pool_strdup(const char *s)
640{
641 char *r = pool_alloc(strlen(s) + 1);
642 strcpy(r, s);
643 return r;
644}
645
646static void insert_mark(uintmax_t idnum, struct object_entry *oe)
647{
648 struct mark_set *s = marks;
649 while ((idnum >> s->shift) >= 1024) {
650 s = pool_calloc(1, sizeof(struct mark_set));
651 s->shift = marks->shift + 10;
652 s->data.sets[0] = marks;
653 marks = s;
654 }
655 while (s->shift) {
656 uintmax_t i = idnum >> s->shift;
657 idnum -= i << s->shift;
658 if (!s->data.sets[i]) {
659 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
660 s->data.sets[i]->shift = s->shift - 10;
661 }
662 s = s->data.sets[i];
663 }
664 if (!s->data.marked[idnum])
665 marks_set_count++;
666 s->data.marked[idnum] = oe;
667}
668
669static struct object_entry *find_mark(uintmax_t idnum)
670{
671 uintmax_t orig_idnum = idnum;
672 struct mark_set *s = marks;
673 struct object_entry *oe = NULL;
674 if ((idnum >> s->shift) < 1024) {
675 while (s && s->shift) {
676 uintmax_t i = idnum >> s->shift;
677 idnum -= i << s->shift;
678 s = s->data.sets[i];
679 }
680 if (s)
681 oe = s->data.marked[idnum];
682 }
683 if (!oe)
684 die("mark :%" PRIuMAX " not declared", orig_idnum);
685 return oe;
686}
687
688static struct atom_str *to_atom(const char *s, unsigned short len)
689{
690 unsigned int hc = hc_str(s, len) % atom_table_sz;
691 struct atom_str *c;
692
693 for (c = atom_table[hc]; c; c = c->next_atom)
694 if (c->str_len == len && !strncmp(s, c->str_dat, len))
695 return c;
696
697 c = pool_alloc(sizeof(struct atom_str) + len + 1);
698 c->str_len = len;
699 strncpy(c->str_dat, s, len);
700 c->str_dat[len] = 0;
701 c->next_atom = atom_table[hc];
702 atom_table[hc] = c;
703 atom_cnt++;
704 return c;
705}
706
707static struct branch *lookup_branch(const char *name)
708{
709 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
710 struct branch *b;
711
712 for (b = branch_table[hc]; b; b = b->table_next_branch)
713 if (!strcmp(name, b->name))
714 return b;
715 return NULL;
716}
717
718static struct branch *new_branch(const char *name)
719{
720 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
721 struct branch *b = lookup_branch(name);
722
723 if (b)
724 die("Invalid attempt to create duplicate branch: %s", name);
725 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
726 die("Branch name doesn't conform to GIT standards: %s", name);
727
728 b = pool_calloc(1, sizeof(struct branch));
729 b->name = pool_strdup(name);
730 b->table_next_branch = branch_table[hc];
731 b->branch_tree.versions[0].mode = S_IFDIR;
732 b->branch_tree.versions[1].mode = S_IFDIR;
733 b->num_notes = 0;
734 b->active = 0;
735 b->pack_id = MAX_PACK_ID;
736 branch_table[hc] = b;
737 branch_count++;
738 return b;
739}
740
741static unsigned int hc_entries(unsigned int cnt)
742{
743 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
744 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
745}
746
747static struct tree_content *new_tree_content(unsigned int cnt)
748{
749 struct avail_tree_content *f, *l = NULL;
750 struct tree_content *t;
751 unsigned int hc = hc_entries(cnt);
752
753 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
754 if (f->entry_capacity >= cnt)
755 break;
756
757 if (f) {
758 if (l)
759 l->next_avail = f->next_avail;
760 else
761 avail_tree_table[hc] = f->next_avail;
762 } else {
763 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
764 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
765 f->entry_capacity = cnt;
766 }
767
768 t = (struct tree_content*)f;
769 t->entry_count = 0;
770 t->delta_depth = 0;
771 return t;
772}
773
774static void release_tree_entry(struct tree_entry *e);
775static void release_tree_content(struct tree_content *t)
776{
777 struct avail_tree_content *f = (struct avail_tree_content*)t;
778 unsigned int hc = hc_entries(f->entry_capacity);
779 f->next_avail = avail_tree_table[hc];
780 avail_tree_table[hc] = f;
781}
782
783static void release_tree_content_recursive(struct tree_content *t)
784{
785 unsigned int i;
786 for (i = 0; i < t->entry_count; i++)
787 release_tree_entry(t->entries[i]);
788 release_tree_content(t);
789}
790
791static struct tree_content *grow_tree_content(
792 struct tree_content *t,
793 int amt)
794{
795 struct tree_content *r = new_tree_content(t->entry_count + amt);
796 r->entry_count = t->entry_count;
797 r->delta_depth = t->delta_depth;
798 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
799 release_tree_content(t);
800 return r;
801}
802
803static struct tree_entry *new_tree_entry(void)
804{
805 struct tree_entry *e;
806
807 if (!avail_tree_entry) {
808 unsigned int n = tree_entry_alloc;
809 total_allocd += n * sizeof(struct tree_entry);
810 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
811 while (n-- > 1) {
812 *((void**)e) = e + 1;
813 e++;
814 }
815 *((void**)e) = NULL;
816 }
817
818 e = avail_tree_entry;
819 avail_tree_entry = *((void**)e);
820 return e;
821}
822
823static void release_tree_entry(struct tree_entry *e)
824{
825 if (e->tree)
826 release_tree_content_recursive(e->tree);
827 *((void**)e) = avail_tree_entry;
828 avail_tree_entry = e;
829}
830
831static struct tree_content *dup_tree_content(struct tree_content *s)
832{
833 struct tree_content *d;
834 struct tree_entry *a, *b;
835 unsigned int i;
836
837 if (!s)
838 return NULL;
839 d = new_tree_content(s->entry_count);
840 for (i = 0; i < s->entry_count; i++) {
841 a = s->entries[i];
842 b = new_tree_entry();
843 memcpy(b, a, sizeof(*a));
844 if (a->tree && is_null_sha1(b->versions[1].sha1))
845 b->tree = dup_tree_content(a->tree);
846 else
847 b->tree = NULL;
848 d->entries[i] = b;
849 }
850 d->entry_count = s->entry_count;
851 d->delta_depth = s->delta_depth;
852
853 return d;
854}
855
856static void start_packfile(void)
857{
858 static char tmp_file[PATH_MAX];
859 struct packed_git *p;
860 struct pack_header hdr;
861 int pack_fd;
862
863 pack_fd = odb_mkstemp(tmp_file, sizeof(tmp_file),
864 "pack/tmp_pack_XXXXXX");
865 p = xcalloc(1, sizeof(*p) + strlen(tmp_file) + 2);
866 strcpy(p->pack_name, tmp_file);
867 p->pack_fd = pack_fd;
868 p->do_not_close = 1;
869 pack_file = sha1fd(pack_fd, p->pack_name);
870
871 hdr.hdr_signature = htonl(PACK_SIGNATURE);
872 hdr.hdr_version = htonl(2);
873 hdr.hdr_entries = 0;
874 sha1write(pack_file, &hdr, sizeof(hdr));
875
876 pack_data = p;
877 pack_size = sizeof(hdr);
878 object_count = 0;
879
880 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
881 all_packs[pack_id] = p;
882}
883
884static const char *create_index(void)
885{
886 const char *tmpfile;
887 struct pack_idx_entry **idx, **c, **last;
888 struct object_entry *e;
889 struct object_entry_pool *o;
890
891 /* Build the table of object IDs. */
892 idx = xmalloc(object_count * sizeof(*idx));
893 c = idx;
894 for (o = blocks; o; o = o->next_pool)
895 for (e = o->next_free; e-- != o->entries;)
896 if (pack_id == e->pack_id)
897 *c++ = &e->idx;
898 last = idx + object_count;
899 if (c != last)
900 die("internal consistency error creating the index");
901
902 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
903 free(idx);
904 return tmpfile;
905}
906
907static char *keep_pack(const char *curr_index_name)
908{
909 static char name[PATH_MAX];
910 static const char *keep_msg = "fast-import";
911 int keep_fd;
912
913 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
914 if (keep_fd < 0)
915 die_errno("cannot create keep file");
916 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
917 if (close(keep_fd))
918 die_errno("failed to write keep file");
919
920 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
921 get_object_directory(), sha1_to_hex(pack_data->sha1));
922 if (move_temp_to_file(pack_data->pack_name, name))
923 die("cannot store pack file");
924
925 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
926 get_object_directory(), sha1_to_hex(pack_data->sha1));
927 if (move_temp_to_file(curr_index_name, name))
928 die("cannot store index file");
929 free((void *)curr_index_name);
930 return name;
931}
932
933static void unkeep_all_packs(void)
934{
935 static char name[PATH_MAX];
936 int k;
937
938 for (k = 0; k < pack_id; k++) {
939 struct packed_git *p = all_packs[k];
940 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
941 get_object_directory(), sha1_to_hex(p->sha1));
942 unlink_or_warn(name);
943 }
944}
945
946static void end_packfile(void)
947{
948 struct packed_git *old_p = pack_data, *new_p;
949
950 clear_delta_base_cache();
951 if (object_count) {
952 unsigned char cur_pack_sha1[20];
953 char *idx_name;
954 int i;
955 struct branch *b;
956 struct tag *t;
957
958 close_pack_windows(pack_data);
959 sha1close(pack_file, cur_pack_sha1, 0);
960 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
961 pack_data->pack_name, object_count,
962 cur_pack_sha1, pack_size);
963 close(pack_data->pack_fd);
964 idx_name = keep_pack(create_index());
965
966 /* Register the packfile with core git's machinery. */
967 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
968 if (!new_p)
969 die("core git rejected index %s", idx_name);
970 all_packs[pack_id] = new_p;
971 install_packed_git(new_p);
972
973 /* Print the boundary */
974 if (pack_edges) {
975 fprintf(pack_edges, "%s:", new_p->pack_name);
976 for (i = 0; i < branch_table_sz; i++) {
977 for (b = branch_table[i]; b; b = b->table_next_branch) {
978 if (b->pack_id == pack_id)
979 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
980 }
981 }
982 for (t = first_tag; t; t = t->next_tag) {
983 if (t->pack_id == pack_id)
984 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
985 }
986 fputc('\n', pack_edges);
987 fflush(pack_edges);
988 }
989
990 pack_id++;
991 }
992 else {
993 close(old_p->pack_fd);
994 unlink_or_warn(old_p->pack_name);
995 }
996 free(old_p);
997
998 /* We can't carry a delta across packfiles. */
999 strbuf_release(&last_blob.data);
1000 last_blob.offset = 0;
1001 last_blob.depth = 0;
1002}
1003
1004static void cycle_packfile(void)
1005{
1006 end_packfile();
1007 start_packfile();
1008}
1009
1010static int store_object(
1011 enum object_type type,
1012 struct strbuf *dat,
1013 struct last_object *last,
1014 unsigned char *sha1out,
1015 uintmax_t mark)
1016{
1017 void *out, *delta;
1018 struct object_entry *e;
1019 unsigned char hdr[96];
1020 unsigned char sha1[20];
1021 unsigned long hdrlen, deltalen;
1022 git_SHA_CTX c;
1023 git_zstream s;
1024
1025 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1026 (unsigned long)dat->len) + 1;
1027 git_SHA1_Init(&c);
1028 git_SHA1_Update(&c, hdr, hdrlen);
1029 git_SHA1_Update(&c, dat->buf, dat->len);
1030 git_SHA1_Final(sha1, &c);
1031 if (sha1out)
1032 hashcpy(sha1out, sha1);
1033
1034 e = insert_object(sha1);
1035 if (mark)
1036 insert_mark(mark, e);
1037 if (e->idx.offset) {
1038 duplicate_count_by_type[type]++;
1039 return 1;
1040 } else if (find_sha1_pack(sha1, packed_git)) {
1041 e->type = type;
1042 e->pack_id = MAX_PACK_ID;
1043 e->idx.offset = 1; /* just not zero! */
1044 duplicate_count_by_type[type]++;
1045 return 1;
1046 }
1047
1048 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1049 delta_count_attempts_by_type[type]++;
1050 delta = diff_delta(last->data.buf, last->data.len,
1051 dat->buf, dat->len,
1052 &deltalen, dat->len - 20);
1053 } else
1054 delta = NULL;
1055
1056 git_deflate_init(&s, pack_compression_level);
1057 if (delta) {
1058 s.next_in = delta;
1059 s.avail_in = deltalen;
1060 } else {
1061 s.next_in = (void *)dat->buf;
1062 s.avail_in = dat->len;
1063 }
1064 s.avail_out = git_deflate_bound(&s, s.avail_in);
1065 s.next_out = out = xmalloc(s.avail_out);
1066 while (git_deflate(&s, Z_FINISH) == Z_OK)
1067 ; /* nothing */
1068 git_deflate_end(&s);
1069
1070 /* Determine if we should auto-checkpoint. */
1071 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1072 || (pack_size + 60 + s.total_out) < pack_size) {
1073
1074 /* This new object needs to *not* have the current pack_id. */
1075 e->pack_id = pack_id + 1;
1076 cycle_packfile();
1077
1078 /* We cannot carry a delta into the new pack. */
1079 if (delta) {
1080 free(delta);
1081 delta = NULL;
1082
1083 git_deflate_init(&s, pack_compression_level);
1084 s.next_in = (void *)dat->buf;
1085 s.avail_in = dat->len;
1086 s.avail_out = git_deflate_bound(&s, s.avail_in);
1087 s.next_out = out = xrealloc(out, s.avail_out);
1088 while (git_deflate(&s, Z_FINISH) == Z_OK)
1089 ; /* nothing */
1090 git_deflate_end(&s);
1091 }
1092 }
1093
1094 e->type = type;
1095 e->pack_id = pack_id;
1096 e->idx.offset = pack_size;
1097 object_count++;
1098 object_count_by_type[type]++;
1099
1100 crc32_begin(pack_file);
1101
1102 if (delta) {
1103 off_t ofs = e->idx.offset - last->offset;
1104 unsigned pos = sizeof(hdr) - 1;
1105
1106 delta_count_by_type[type]++;
1107 e->depth = last->depth + 1;
1108
1109 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1110 sha1write(pack_file, hdr, hdrlen);
1111 pack_size += hdrlen;
1112
1113 hdr[pos] = ofs & 127;
1114 while (ofs >>= 7)
1115 hdr[--pos] = 128 | (--ofs & 127);
1116 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1117 pack_size += sizeof(hdr) - pos;
1118 } else {
1119 e->depth = 0;
1120 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1121 sha1write(pack_file, hdr, hdrlen);
1122 pack_size += hdrlen;
1123 }
1124
1125 sha1write(pack_file, out, s.total_out);
1126 pack_size += s.total_out;
1127
1128 e->idx.crc32 = crc32_end(pack_file);
1129
1130 free(out);
1131 free(delta);
1132 if (last) {
1133 if (last->no_swap) {
1134 last->data = *dat;
1135 } else {
1136 strbuf_swap(&last->data, dat);
1137 }
1138 last->offset = e->idx.offset;
1139 last->depth = e->depth;
1140 }
1141 return 0;
1142}
1143
1144static void truncate_pack(struct sha1file_checkpoint *checkpoint)
1145{
1146 if (sha1file_truncate(pack_file, checkpoint))
1147 die_errno("cannot truncate pack to skip duplicate");
1148 pack_size = checkpoint->offset;
1149}
1150
1151static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1152{
1153 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1154 unsigned char *in_buf = xmalloc(in_sz);
1155 unsigned char *out_buf = xmalloc(out_sz);
1156 struct object_entry *e;
1157 unsigned char sha1[20];
1158 unsigned long hdrlen;
1159 off_t offset;
1160 git_SHA_CTX c;
1161 git_zstream s;
1162 struct sha1file_checkpoint checkpoint;
1163 int status = Z_OK;
1164
1165 /* Determine if we should auto-checkpoint. */
1166 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1167 || (pack_size + 60 + len) < pack_size)
1168 cycle_packfile();
1169
1170 sha1file_checkpoint(pack_file, &checkpoint);
1171 offset = checkpoint.offset;
1172
1173 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1174 if (out_sz <= hdrlen)
1175 die("impossibly large object header");
1176
1177 git_SHA1_Init(&c);
1178 git_SHA1_Update(&c, out_buf, hdrlen);
1179
1180 crc32_begin(pack_file);
1181
1182 git_deflate_init(&s, pack_compression_level);
1183
1184 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1185 if (out_sz <= hdrlen)
1186 die("impossibly large object header");
1187
1188 s.next_out = out_buf + hdrlen;
1189 s.avail_out = out_sz - hdrlen;
1190
1191 while (status != Z_STREAM_END) {
1192 if (0 < len && !s.avail_in) {
1193 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1194 size_t n = fread(in_buf, 1, cnt, stdin);
1195 if (!n && feof(stdin))
1196 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1197
1198 git_SHA1_Update(&c, in_buf, n);
1199 s.next_in = in_buf;
1200 s.avail_in = n;
1201 len -= n;
1202 }
1203
1204 status = git_deflate(&s, len ? 0 : Z_FINISH);
1205
1206 if (!s.avail_out || status == Z_STREAM_END) {
1207 size_t n = s.next_out - out_buf;
1208 sha1write(pack_file, out_buf, n);
1209 pack_size += n;
1210 s.next_out = out_buf;
1211 s.avail_out = out_sz;
1212 }
1213
1214 switch (status) {
1215 case Z_OK:
1216 case Z_BUF_ERROR:
1217 case Z_STREAM_END:
1218 continue;
1219 default:
1220 die("unexpected deflate failure: %d", status);
1221 }
1222 }
1223 git_deflate_end(&s);
1224 git_SHA1_Final(sha1, &c);
1225
1226 if (sha1out)
1227 hashcpy(sha1out, sha1);
1228
1229 e = insert_object(sha1);
1230
1231 if (mark)
1232 insert_mark(mark, e);
1233
1234 if (e->idx.offset) {
1235 duplicate_count_by_type[OBJ_BLOB]++;
1236 truncate_pack(&checkpoint);
1237
1238 } else if (find_sha1_pack(sha1, packed_git)) {
1239 e->type = OBJ_BLOB;
1240 e->pack_id = MAX_PACK_ID;
1241 e->idx.offset = 1; /* just not zero! */
1242 duplicate_count_by_type[OBJ_BLOB]++;
1243 truncate_pack(&checkpoint);
1244
1245 } else {
1246 e->depth = 0;
1247 e->type = OBJ_BLOB;
1248 e->pack_id = pack_id;
1249 e->idx.offset = offset;
1250 e->idx.crc32 = crc32_end(pack_file);
1251 object_count++;
1252 object_count_by_type[OBJ_BLOB]++;
1253 }
1254
1255 free(in_buf);
1256 free(out_buf);
1257}
1258
1259/* All calls must be guarded by find_object() or find_mark() to
1260 * ensure the 'struct object_entry' passed was written by this
1261 * process instance. We unpack the entry by the offset, avoiding
1262 * the need for the corresponding .idx file. This unpacking rule
1263 * works because we only use OBJ_REF_DELTA within the packfiles
1264 * created by fast-import.
1265 *
1266 * oe must not be NULL. Such an oe usually comes from giving
1267 * an unknown SHA-1 to find_object() or an undefined mark to
1268 * find_mark(). Callers must test for this condition and use
1269 * the standard read_sha1_file() when it happens.
1270 *
1271 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1272 * find_mark(), where the mark was reloaded from an existing marks
1273 * file and is referencing an object that this fast-import process
1274 * instance did not write out to a packfile. Callers must test for
1275 * this condition and use read_sha1_file() instead.
1276 */
1277static void *gfi_unpack_entry(
1278 struct object_entry *oe,
1279 unsigned long *sizep)
1280{
1281 enum object_type type;
1282 struct packed_git *p = all_packs[oe->pack_id];
1283 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1284 /* The object is stored in the packfile we are writing to
1285 * and we have modified it since the last time we scanned
1286 * back to read a previously written object. If an old
1287 * window covered [p->pack_size, p->pack_size + 20) its
1288 * data is stale and is not valid. Closing all windows
1289 * and updating the packfile length ensures we can read
1290 * the newly written data.
1291 */
1292 close_pack_windows(p);
1293 sha1flush(pack_file);
1294
1295 /* We have to offer 20 bytes additional on the end of
1296 * the packfile as the core unpacker code assumes the
1297 * footer is present at the file end and must promise
1298 * at least 20 bytes within any window it maps. But
1299 * we don't actually create the footer here.
1300 */
1301 p->pack_size = pack_size + 20;
1302 }
1303 return unpack_entry(p, oe->idx.offset, &type, sizep);
1304}
1305
1306static const char *get_mode(const char *str, uint16_t *modep)
1307{
1308 unsigned char c;
1309 uint16_t mode = 0;
1310
1311 while ((c = *str++) != ' ') {
1312 if (c < '0' || c > '7')
1313 return NULL;
1314 mode = (mode << 3) + (c - '0');
1315 }
1316 *modep = mode;
1317 return str;
1318}
1319
1320static void load_tree(struct tree_entry *root)
1321{
1322 unsigned char *sha1 = root->versions[1].sha1;
1323 struct object_entry *myoe;
1324 struct tree_content *t;
1325 unsigned long size;
1326 char *buf;
1327 const char *c;
1328
1329 root->tree = t = new_tree_content(8);
1330 if (is_null_sha1(sha1))
1331 return;
1332
1333 myoe = find_object(sha1);
1334 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1335 if (myoe->type != OBJ_TREE)
1336 die("Not a tree: %s", sha1_to_hex(sha1));
1337 t->delta_depth = myoe->depth;
1338 buf = gfi_unpack_entry(myoe, &size);
1339 if (!buf)
1340 die("Can't load tree %s", sha1_to_hex(sha1));
1341 } else {
1342 enum object_type type;
1343 buf = read_sha1_file(sha1, &type, &size);
1344 if (!buf || type != OBJ_TREE)
1345 die("Can't load tree %s", sha1_to_hex(sha1));
1346 }
1347
1348 c = buf;
1349 while (c != (buf + size)) {
1350 struct tree_entry *e = new_tree_entry();
1351
1352 if (t->entry_count == t->entry_capacity)
1353 root->tree = t = grow_tree_content(t, t->entry_count);
1354 t->entries[t->entry_count++] = e;
1355
1356 e->tree = NULL;
1357 c = get_mode(c, &e->versions[1].mode);
1358 if (!c)
1359 die("Corrupt mode in %s", sha1_to_hex(sha1));
1360 e->versions[0].mode = e->versions[1].mode;
1361 e->name = to_atom(c, strlen(c));
1362 c += e->name->str_len + 1;
1363 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1364 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1365 c += 20;
1366 }
1367 free(buf);
1368}
1369
1370static int tecmp0 (const void *_a, const void *_b)
1371{
1372 struct tree_entry *a = *((struct tree_entry**)_a);
1373 struct tree_entry *b = *((struct tree_entry**)_b);
1374 return base_name_compare(
1375 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1376 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1377}
1378
1379static int tecmp1 (const void *_a, const void *_b)
1380{
1381 struct tree_entry *a = *((struct tree_entry**)_a);
1382 struct tree_entry *b = *((struct tree_entry**)_b);
1383 return base_name_compare(
1384 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1385 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1386}
1387
1388static void mktree(struct tree_content *t, int v, struct strbuf *b)
1389{
1390 size_t maxlen = 0;
1391 unsigned int i;
1392
1393 if (!v)
1394 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1395 else
1396 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1397
1398 for (i = 0; i < t->entry_count; i++) {
1399 if (t->entries[i]->versions[v].mode)
1400 maxlen += t->entries[i]->name->str_len + 34;
1401 }
1402
1403 strbuf_reset(b);
1404 strbuf_grow(b, maxlen);
1405 for (i = 0; i < t->entry_count; i++) {
1406 struct tree_entry *e = t->entries[i];
1407 if (!e->versions[v].mode)
1408 continue;
1409 strbuf_addf(b, "%o %s%c",
1410 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1411 e->name->str_dat, '\0');
1412 strbuf_add(b, e->versions[v].sha1, 20);
1413 }
1414}
1415
1416static void store_tree(struct tree_entry *root)
1417{
1418 struct tree_content *t = root->tree;
1419 unsigned int i, j, del;
1420 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1421 struct object_entry *le = NULL;
1422
1423 if (!is_null_sha1(root->versions[1].sha1))
1424 return;
1425
1426 for (i = 0; i < t->entry_count; i++) {
1427 if (t->entries[i]->tree)
1428 store_tree(t->entries[i]);
1429 }
1430
1431 if (!(root->versions[0].mode & NO_DELTA))
1432 le = find_object(root->versions[0].sha1);
1433 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1434 mktree(t, 0, &old_tree);
1435 lo.data = old_tree;
1436 lo.offset = le->idx.offset;
1437 lo.depth = t->delta_depth;
1438 }
1439
1440 mktree(t, 1, &new_tree);
1441 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1442
1443 t->delta_depth = lo.depth;
1444 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1445 struct tree_entry *e = t->entries[i];
1446 if (e->versions[1].mode) {
1447 e->versions[0].mode = e->versions[1].mode;
1448 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1449 t->entries[j++] = e;
1450 } else {
1451 release_tree_entry(e);
1452 del++;
1453 }
1454 }
1455 t->entry_count -= del;
1456}
1457
1458static void tree_content_replace(
1459 struct tree_entry *root,
1460 const unsigned char *sha1,
1461 const uint16_t mode,
1462 struct tree_content *newtree)
1463{
1464 if (!S_ISDIR(mode))
1465 die("Root cannot be a non-directory");
1466 hashclr(root->versions[0].sha1);
1467 hashcpy(root->versions[1].sha1, sha1);
1468 if (root->tree)
1469 release_tree_content_recursive(root->tree);
1470 root->tree = newtree;
1471}
1472
1473static int tree_content_set(
1474 struct tree_entry *root,
1475 const char *p,
1476 const unsigned char *sha1,
1477 const uint16_t mode,
1478 struct tree_content *subtree)
1479{
1480 struct tree_content *t;
1481 const char *slash1;
1482 unsigned int i, n;
1483 struct tree_entry *e;
1484
1485 slash1 = strchrnul(p, '/');
1486 n = slash1 - p;
1487 if (!n)
1488 die("Empty path component found in input");
1489 if (!*slash1 && !S_ISDIR(mode) && subtree)
1490 die("Non-directories cannot have subtrees");
1491
1492 if (!root->tree)
1493 load_tree(root);
1494 t = root->tree;
1495 for (i = 0; i < t->entry_count; i++) {
1496 e = t->entries[i];
1497 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1498 if (!*slash1) {
1499 if (!S_ISDIR(mode)
1500 && e->versions[1].mode == mode
1501 && !hashcmp(e->versions[1].sha1, sha1))
1502 return 0;
1503 e->versions[1].mode = mode;
1504 hashcpy(e->versions[1].sha1, sha1);
1505 if (e->tree)
1506 release_tree_content_recursive(e->tree);
1507 e->tree = subtree;
1508
1509 /*
1510 * We need to leave e->versions[0].sha1 alone
1511 * to avoid modifying the preimage tree used
1512 * when writing out the parent directory.
1513 * But after replacing the subdir with a
1514 * completely different one, it's not a good
1515 * delta base any more, and besides, we've
1516 * thrown away the tree entries needed to
1517 * make a delta against it.
1518 *
1519 * So let's just explicitly disable deltas
1520 * for the subtree.
1521 */
1522 if (S_ISDIR(e->versions[0].mode))
1523 e->versions[0].mode |= NO_DELTA;
1524
1525 hashclr(root->versions[1].sha1);
1526 return 1;
1527 }
1528 if (!S_ISDIR(e->versions[1].mode)) {
1529 e->tree = new_tree_content(8);
1530 e->versions[1].mode = S_IFDIR;
1531 }
1532 if (!e->tree)
1533 load_tree(e);
1534 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1535 hashclr(root->versions[1].sha1);
1536 return 1;
1537 }
1538 return 0;
1539 }
1540 }
1541
1542 if (t->entry_count == t->entry_capacity)
1543 root->tree = t = grow_tree_content(t, t->entry_count);
1544 e = new_tree_entry();
1545 e->name = to_atom(p, n);
1546 e->versions[0].mode = 0;
1547 hashclr(e->versions[0].sha1);
1548 t->entries[t->entry_count++] = e;
1549 if (*slash1) {
1550 e->tree = new_tree_content(8);
1551 e->versions[1].mode = S_IFDIR;
1552 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1553 } else {
1554 e->tree = subtree;
1555 e->versions[1].mode = mode;
1556 hashcpy(e->versions[1].sha1, sha1);
1557 }
1558 hashclr(root->versions[1].sha1);
1559 return 1;
1560}
1561
1562static int tree_content_remove(
1563 struct tree_entry *root,
1564 const char *p,
1565 struct tree_entry *backup_leaf,
1566 int allow_root)
1567{
1568 struct tree_content *t;
1569 const char *slash1;
1570 unsigned int i, n;
1571 struct tree_entry *e;
1572
1573 slash1 = strchrnul(p, '/');
1574 n = slash1 - p;
1575
1576 if (!root->tree)
1577 load_tree(root);
1578
1579 if (!*p && allow_root) {
1580 e = root;
1581 goto del_entry;
1582 }
1583
1584 t = root->tree;
1585 for (i = 0; i < t->entry_count; i++) {
1586 e = t->entries[i];
1587 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1588 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1589 /*
1590 * If p names a file in some subdirectory, and a
1591 * file or symlink matching the name of the
1592 * parent directory of p exists, then p cannot
1593 * exist and need not be deleted.
1594 */
1595 return 1;
1596 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1597 goto del_entry;
1598 if (!e->tree)
1599 load_tree(e);
1600 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1601 for (n = 0; n < e->tree->entry_count; n++) {
1602 if (e->tree->entries[n]->versions[1].mode) {
1603 hashclr(root->versions[1].sha1);
1604 return 1;
1605 }
1606 }
1607 backup_leaf = NULL;
1608 goto del_entry;
1609 }
1610 return 0;
1611 }
1612 }
1613 return 0;
1614
1615del_entry:
1616 if (backup_leaf)
1617 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1618 else if (e->tree)
1619 release_tree_content_recursive(e->tree);
1620 e->tree = NULL;
1621 e->versions[1].mode = 0;
1622 hashclr(e->versions[1].sha1);
1623 hashclr(root->versions[1].sha1);
1624 return 1;
1625}
1626
1627static int tree_content_get(
1628 struct tree_entry *root,
1629 const char *p,
1630 struct tree_entry *leaf,
1631 int allow_root)
1632{
1633 struct tree_content *t;
1634 const char *slash1;
1635 unsigned int i, n;
1636 struct tree_entry *e;
1637
1638 slash1 = strchrnul(p, '/');
1639 n = slash1 - p;
1640 if (!n && !allow_root)
1641 die("Empty path component found in input");
1642
1643 if (!root->tree)
1644 load_tree(root);
1645
1646 if (!n) {
1647 e = root;
1648 goto found_entry;
1649 }
1650
1651 t = root->tree;
1652 for (i = 0; i < t->entry_count; i++) {
1653 e = t->entries[i];
1654 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1655 if (!*slash1)
1656 goto found_entry;
1657 if (!S_ISDIR(e->versions[1].mode))
1658 return 0;
1659 if (!e->tree)
1660 load_tree(e);
1661 return tree_content_get(e, slash1 + 1, leaf, 0);
1662 }
1663 }
1664 return 0;
1665
1666found_entry:
1667 memcpy(leaf, e, sizeof(*leaf));
1668 if (e->tree && is_null_sha1(e->versions[1].sha1))
1669 leaf->tree = dup_tree_content(e->tree);
1670 else
1671 leaf->tree = NULL;
1672 return 1;
1673}
1674
1675static int update_branch(struct branch *b)
1676{
1677 static const char *msg = "fast-import";
1678 struct ref_lock *lock;
1679 unsigned char old_sha1[20];
1680
1681 if (is_null_sha1(b->sha1))
1682 return 0;
1683 if (read_ref(b->name, old_sha1))
1684 hashclr(old_sha1);
1685 lock = lock_any_ref_for_update(b->name, old_sha1, 0, NULL);
1686 if (!lock)
1687 return error("Unable to lock %s", b->name);
1688 if (!force_update && !is_null_sha1(old_sha1)) {
1689 struct commit *old_cmit, *new_cmit;
1690
1691 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1692 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1693 if (!old_cmit || !new_cmit) {
1694 unlock_ref(lock);
1695 return error("Branch %s is missing commits.", b->name);
1696 }
1697
1698 if (!in_merge_bases(old_cmit, new_cmit)) {
1699 unlock_ref(lock);
1700 warning("Not updating %s"
1701 " (new tip %s does not contain %s)",
1702 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1703 return -1;
1704 }
1705 }
1706 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1707 return error("Unable to update %s", b->name);
1708 return 0;
1709}
1710
1711static void dump_branches(void)
1712{
1713 unsigned int i;
1714 struct branch *b;
1715
1716 for (i = 0; i < branch_table_sz; i++) {
1717 for (b = branch_table[i]; b; b = b->table_next_branch)
1718 failure |= update_branch(b);
1719 }
1720}
1721
1722static void dump_tags(void)
1723{
1724 static const char *msg = "fast-import";
1725 struct tag *t;
1726 struct ref_lock *lock;
1727 char ref_name[PATH_MAX];
1728
1729 for (t = first_tag; t; t = t->next_tag) {
1730 sprintf(ref_name, "tags/%s", t->name);
1731 lock = lock_ref_sha1(ref_name, NULL);
1732 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1733 failure |= error("Unable to update %s", ref_name);
1734 }
1735}
1736
1737static void dump_marks_helper(FILE *f,
1738 uintmax_t base,
1739 struct mark_set *m)
1740{
1741 uintmax_t k;
1742 if (m->shift) {
1743 for (k = 0; k < 1024; k++) {
1744 if (m->data.sets[k])
1745 dump_marks_helper(f, base + (k << m->shift),
1746 m->data.sets[k]);
1747 }
1748 } else {
1749 for (k = 0; k < 1024; k++) {
1750 if (m->data.marked[k])
1751 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1752 sha1_to_hex(m->data.marked[k]->idx.sha1));
1753 }
1754 }
1755}
1756
1757static void dump_marks(void)
1758{
1759 static struct lock_file mark_lock;
1760 int mark_fd;
1761 FILE *f;
1762
1763 if (!export_marks_file)
1764 return;
1765
1766 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1767 if (mark_fd < 0) {
1768 failure |= error("Unable to write marks file %s: %s",
1769 export_marks_file, strerror(errno));
1770 return;
1771 }
1772
1773 f = fdopen(mark_fd, "w");
1774 if (!f) {
1775 int saved_errno = errno;
1776 rollback_lock_file(&mark_lock);
1777 failure |= error("Unable to write marks file %s: %s",
1778 export_marks_file, strerror(saved_errno));
1779 return;
1780 }
1781
1782 /*
1783 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1784 * Assign -1 to the lock file descriptor so that commit_lock_file()
1785 * won't try to close() it.
1786 */
1787 mark_lock.fd = -1;
1788
1789 dump_marks_helper(f, 0, marks);
1790 if (ferror(f) || fclose(f)) {
1791 int saved_errno = errno;
1792 rollback_lock_file(&mark_lock);
1793 failure |= error("Unable to write marks file %s: %s",
1794 export_marks_file, strerror(saved_errno));
1795 return;
1796 }
1797
1798 if (commit_lock_file(&mark_lock)) {
1799 int saved_errno = errno;
1800 rollback_lock_file(&mark_lock);
1801 failure |= error("Unable to commit marks file %s: %s",
1802 export_marks_file, strerror(saved_errno));
1803 return;
1804 }
1805}
1806
1807static void read_marks(void)
1808{
1809 char line[512];
1810 FILE *f = fopen(import_marks_file, "r");
1811 if (f)
1812 ;
1813 else if (import_marks_file_ignore_missing && errno == ENOENT)
1814 return; /* Marks file does not exist */
1815 else
1816 die_errno("cannot read '%s'", import_marks_file);
1817 while (fgets(line, sizeof(line), f)) {
1818 uintmax_t mark;
1819 char *end;
1820 unsigned char sha1[20];
1821 struct object_entry *e;
1822
1823 end = strchr(line, '\n');
1824 if (line[0] != ':' || !end)
1825 die("corrupt mark line: %s", line);
1826 *end = 0;
1827 mark = strtoumax(line + 1, &end, 10);
1828 if (!mark || end == line + 1
1829 || *end != ' ' || get_sha1_hex(end + 1, sha1))
1830 die("corrupt mark line: %s", line);
1831 e = find_object(sha1);
1832 if (!e) {
1833 enum object_type type = sha1_object_info(sha1, NULL);
1834 if (type < 0)
1835 die("object not found: %s", sha1_to_hex(sha1));
1836 e = insert_object(sha1);
1837 e->type = type;
1838 e->pack_id = MAX_PACK_ID;
1839 e->idx.offset = 1; /* just not zero! */
1840 }
1841 insert_mark(mark, e);
1842 }
1843 fclose(f);
1844}
1845
1846
1847static int read_next_command(void)
1848{
1849 static int stdin_eof = 0;
1850
1851 if (stdin_eof) {
1852 unread_command_buf = 0;
1853 return EOF;
1854 }
1855
1856 for (;;) {
1857 if (unread_command_buf) {
1858 unread_command_buf = 0;
1859 } else {
1860 struct recent_command *rc;
1861
1862 strbuf_detach(&command_buf, NULL);
1863 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1864 if (stdin_eof)
1865 return EOF;
1866
1867 if (!seen_data_command
1868 && !starts_with(command_buf.buf, "feature ")
1869 && !starts_with(command_buf.buf, "option ")) {
1870 parse_argv();
1871 }
1872
1873 rc = rc_free;
1874 if (rc)
1875 rc_free = rc->next;
1876 else {
1877 rc = cmd_hist.next;
1878 cmd_hist.next = rc->next;
1879 cmd_hist.next->prev = &cmd_hist;
1880 free(rc->buf);
1881 }
1882
1883 rc->buf = command_buf.buf;
1884 rc->prev = cmd_tail;
1885 rc->next = cmd_hist.prev;
1886 rc->prev->next = rc;
1887 cmd_tail = rc;
1888 }
1889 if (starts_with(command_buf.buf, "cat-blob ")) {
1890 parse_cat_blob();
1891 continue;
1892 }
1893 if (command_buf.buf[0] == '#')
1894 continue;
1895 return 0;
1896 }
1897}
1898
1899static void skip_optional_lf(void)
1900{
1901 int term_char = fgetc(stdin);
1902 if (term_char != '\n' && term_char != EOF)
1903 ungetc(term_char, stdin);
1904}
1905
1906static void parse_mark(void)
1907{
1908 if (starts_with(command_buf.buf, "mark :")) {
1909 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1910 read_next_command();
1911 }
1912 else
1913 next_mark = 0;
1914}
1915
1916static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1917{
1918 strbuf_reset(sb);
1919
1920 if (!starts_with(command_buf.buf, "data "))
1921 die("Expected 'data n' command, found: %s", command_buf.buf);
1922
1923 if (starts_with(command_buf.buf + 5, "<<")) {
1924 char *term = xstrdup(command_buf.buf + 5 + 2);
1925 size_t term_len = command_buf.len - 5 - 2;
1926
1927 strbuf_detach(&command_buf, NULL);
1928 for (;;) {
1929 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1930 die("EOF in data (terminator '%s' not found)", term);
1931 if (term_len == command_buf.len
1932 && !strcmp(term, command_buf.buf))
1933 break;
1934 strbuf_addbuf(sb, &command_buf);
1935 strbuf_addch(sb, '\n');
1936 }
1937 free(term);
1938 }
1939 else {
1940 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1941 size_t n = 0, length = (size_t)len;
1942
1943 if (limit && limit < len) {
1944 *len_res = len;
1945 return 0;
1946 }
1947 if (length < len)
1948 die("data is too large to use in this context");
1949
1950 while (n < length) {
1951 size_t s = strbuf_fread(sb, length - n, stdin);
1952 if (!s && feof(stdin))
1953 die("EOF in data (%lu bytes remaining)",
1954 (unsigned long)(length - n));
1955 n += s;
1956 }
1957 }
1958
1959 skip_optional_lf();
1960 return 1;
1961}
1962
1963static int validate_raw_date(const char *src, char *result, int maxlen)
1964{
1965 const char *orig_src = src;
1966 char *endp;
1967 unsigned long num;
1968
1969 errno = 0;
1970
1971 num = strtoul(src, &endp, 10);
1972 /* NEEDSWORK: perhaps check for reasonable values? */
1973 if (errno || endp == src || *endp != ' ')
1974 return -1;
1975
1976 src = endp + 1;
1977 if (*src != '-' && *src != '+')
1978 return -1;
1979
1980 num = strtoul(src + 1, &endp, 10);
1981 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1982 1400 < num)
1983 return -1;
1984
1985 strcpy(result, orig_src);
1986 return 0;
1987}
1988
1989static char *parse_ident(const char *buf)
1990{
1991 const char *ltgt;
1992 size_t name_len;
1993 char *ident;
1994
1995 /* ensure there is a space delimiter even if there is no name */
1996 if (*buf == '<')
1997 --buf;
1998
1999 ltgt = buf + strcspn(buf, "<>");
2000 if (*ltgt != '<')
2001 die("Missing < in ident string: %s", buf);
2002 if (ltgt != buf && ltgt[-1] != ' ')
2003 die("Missing space before < in ident string: %s", buf);
2004 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2005 if (*ltgt != '>')
2006 die("Missing > in ident string: %s", buf);
2007 ltgt++;
2008 if (*ltgt != ' ')
2009 die("Missing space after > in ident string: %s", buf);
2010 ltgt++;
2011 name_len = ltgt - buf;
2012 ident = xmalloc(name_len + 24);
2013 strncpy(ident, buf, name_len);
2014
2015 switch (whenspec) {
2016 case WHENSPEC_RAW:
2017 if (validate_raw_date(ltgt, ident + name_len, 24) < 0)
2018 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2019 break;
2020 case WHENSPEC_RFC2822:
2021 if (parse_date(ltgt, ident + name_len, 24) < 0)
2022 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2023 break;
2024 case WHENSPEC_NOW:
2025 if (strcmp("now", ltgt))
2026 die("Date in ident must be 'now': %s", buf);
2027 datestamp(ident + name_len, 24);
2028 break;
2029 }
2030
2031 return ident;
2032}
2033
2034static void parse_and_store_blob(
2035 struct last_object *last,
2036 unsigned char *sha1out,
2037 uintmax_t mark)
2038{
2039 static struct strbuf buf = STRBUF_INIT;
2040 uintmax_t len;
2041
2042 if (parse_data(&buf, big_file_threshold, &len))
2043 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2044 else {
2045 if (last) {
2046 strbuf_release(&last->data);
2047 last->offset = 0;
2048 last->depth = 0;
2049 }
2050 stream_blob(len, sha1out, mark);
2051 skip_optional_lf();
2052 }
2053}
2054
2055static void parse_new_blob(void)
2056{
2057 read_next_command();
2058 parse_mark();
2059 parse_and_store_blob(&last_blob, NULL, next_mark);
2060}
2061
2062static void unload_one_branch(void)
2063{
2064 while (cur_active_branches
2065 && cur_active_branches >= max_active_branches) {
2066 uintmax_t min_commit = ULONG_MAX;
2067 struct branch *e, *l = NULL, *p = NULL;
2068
2069 for (e = active_branches; e; e = e->active_next_branch) {
2070 if (e->last_commit < min_commit) {
2071 p = l;
2072 min_commit = e->last_commit;
2073 }
2074 l = e;
2075 }
2076
2077 if (p) {
2078 e = p->active_next_branch;
2079 p->active_next_branch = e->active_next_branch;
2080 } else {
2081 e = active_branches;
2082 active_branches = e->active_next_branch;
2083 }
2084 e->active = 0;
2085 e->active_next_branch = NULL;
2086 if (e->branch_tree.tree) {
2087 release_tree_content_recursive(e->branch_tree.tree);
2088 e->branch_tree.tree = NULL;
2089 }
2090 cur_active_branches--;
2091 }
2092}
2093
2094static void load_branch(struct branch *b)
2095{
2096 load_tree(&b->branch_tree);
2097 if (!b->active) {
2098 b->active = 1;
2099 b->active_next_branch = active_branches;
2100 active_branches = b;
2101 cur_active_branches++;
2102 branch_load_count++;
2103 }
2104}
2105
2106static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2107{
2108 unsigned char fanout = 0;
2109 while ((num_notes >>= 8))
2110 fanout++;
2111 return fanout;
2112}
2113
2114static void construct_path_with_fanout(const char *hex_sha1,
2115 unsigned char fanout, char *path)
2116{
2117 unsigned int i = 0, j = 0;
2118 if (fanout >= 20)
2119 die("Too large fanout (%u)", fanout);
2120 while (fanout) {
2121 path[i++] = hex_sha1[j++];
2122 path[i++] = hex_sha1[j++];
2123 path[i++] = '/';
2124 fanout--;
2125 }
2126 memcpy(path + i, hex_sha1 + j, 40 - j);
2127 path[i + 40 - j] = '\0';
2128}
2129
2130static uintmax_t do_change_note_fanout(
2131 struct tree_entry *orig_root, struct tree_entry *root,
2132 char *hex_sha1, unsigned int hex_sha1_len,
2133 char *fullpath, unsigned int fullpath_len,
2134 unsigned char fanout)
2135{
2136 struct tree_content *t = root->tree;
2137 struct tree_entry *e, leaf;
2138 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2139 uintmax_t num_notes = 0;
2140 unsigned char sha1[20];
2141 char realpath[60];
2142
2143 for (i = 0; t && i < t->entry_count; i++) {
2144 e = t->entries[i];
2145 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2146 tmp_fullpath_len = fullpath_len;
2147
2148 /*
2149 * We're interested in EITHER existing note entries (entries
2150 * with exactly 40 hex chars in path, not including directory
2151 * separators), OR directory entries that may contain note
2152 * entries (with < 40 hex chars in path).
2153 * Also, each path component in a note entry must be a multiple
2154 * of 2 chars.
2155 */
2156 if (!e->versions[1].mode ||
2157 tmp_hex_sha1_len > 40 ||
2158 e->name->str_len % 2)
2159 continue;
2160
2161 /* This _may_ be a note entry, or a subdir containing notes */
2162 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2163 e->name->str_len);
2164 if (tmp_fullpath_len)
2165 fullpath[tmp_fullpath_len++] = '/';
2166 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2167 e->name->str_len);
2168 tmp_fullpath_len += e->name->str_len;
2169 fullpath[tmp_fullpath_len] = '\0';
2170
2171 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2172 /* This is a note entry */
2173 if (fanout == 0xff) {
2174 /* Counting mode, no rename */
2175 num_notes++;
2176 continue;
2177 }
2178 construct_path_with_fanout(hex_sha1, fanout, realpath);
2179 if (!strcmp(fullpath, realpath)) {
2180 /* Note entry is in correct location */
2181 num_notes++;
2182 continue;
2183 }
2184
2185 /* Rename fullpath to realpath */
2186 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2187 die("Failed to remove path %s", fullpath);
2188 tree_content_set(orig_root, realpath,
2189 leaf.versions[1].sha1,
2190 leaf.versions[1].mode,
2191 leaf.tree);
2192 } else if (S_ISDIR(e->versions[1].mode)) {
2193 /* This is a subdir that may contain note entries */
2194 if (!e->tree)
2195 load_tree(e);
2196 num_notes += do_change_note_fanout(orig_root, e,
2197 hex_sha1, tmp_hex_sha1_len,
2198 fullpath, tmp_fullpath_len, fanout);
2199 }
2200
2201 /* The above may have reallocated the current tree_content */
2202 t = root->tree;
2203 }
2204 return num_notes;
2205}
2206
2207static uintmax_t change_note_fanout(struct tree_entry *root,
2208 unsigned char fanout)
2209{
2210 char hex_sha1[40], path[60];
2211 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2212}
2213
2214/*
2215 * Given a pointer into a string, parse a mark reference:
2216 *
2217 * idnum ::= ':' bigint;
2218 *
2219 * Return the first character after the value in *endptr.
2220 *
2221 * Complain if the following character is not what is expected,
2222 * either a space or end of the string.
2223 */
2224static uintmax_t parse_mark_ref(const char *p, char **endptr)
2225{
2226 uintmax_t mark;
2227
2228 assert(*p == ':');
2229 p++;
2230 mark = strtoumax(p, endptr, 10);
2231 if (*endptr == p)
2232 die("No value after ':' in mark: %s", command_buf.buf);
2233 return mark;
2234}
2235
2236/*
2237 * Parse the mark reference, and complain if this is not the end of
2238 * the string.
2239 */
2240static uintmax_t parse_mark_ref_eol(const char *p)
2241{
2242 char *end;
2243 uintmax_t mark;
2244
2245 mark = parse_mark_ref(p, &end);
2246 if (*end != '\0')
2247 die("Garbage after mark: %s", command_buf.buf);
2248 return mark;
2249}
2250
2251/*
2252 * Parse the mark reference, demanding a trailing space. Return a
2253 * pointer to the space.
2254 */
2255static uintmax_t parse_mark_ref_space(const char **p)
2256{
2257 uintmax_t mark;
2258 char *end;
2259
2260 mark = parse_mark_ref(*p, &end);
2261 if (*end != ' ')
2262 die("Missing space after mark: %s", command_buf.buf);
2263 *p = end;
2264 return mark;
2265}
2266
2267static void file_change_m(struct branch *b)
2268{
2269 const char *p = command_buf.buf + 2;
2270 static struct strbuf uq = STRBUF_INIT;
2271 const char *endp;
2272 struct object_entry *oe;
2273 unsigned char sha1[20];
2274 uint16_t mode, inline_data = 0;
2275
2276 p = get_mode(p, &mode);
2277 if (!p)
2278 die("Corrupt mode: %s", command_buf.buf);
2279 switch (mode) {
2280 case 0644:
2281 case 0755:
2282 mode |= S_IFREG;
2283 case S_IFREG | 0644:
2284 case S_IFREG | 0755:
2285 case S_IFLNK:
2286 case S_IFDIR:
2287 case S_IFGITLINK:
2288 /* ok */
2289 break;
2290 default:
2291 die("Corrupt mode: %s", command_buf.buf);
2292 }
2293
2294 if (*p == ':') {
2295 oe = find_mark(parse_mark_ref_space(&p));
2296 hashcpy(sha1, oe->idx.sha1);
2297 } else if (starts_with(p, "inline ")) {
2298 inline_data = 1;
2299 oe = NULL; /* not used with inline_data, but makes gcc happy */
2300 p += strlen("inline"); /* advance to space */
2301 } else {
2302 if (get_sha1_hex(p, sha1))
2303 die("Invalid dataref: %s", command_buf.buf);
2304 oe = find_object(sha1);
2305 p += 40;
2306 if (*p != ' ')
2307 die("Missing space after SHA1: %s", command_buf.buf);
2308 }
2309 assert(*p == ' ');
2310 p++; /* skip space */
2311
2312 strbuf_reset(&uq);
2313 if (!unquote_c_style(&uq, p, &endp)) {
2314 if (*endp)
2315 die("Garbage after path in: %s", command_buf.buf);
2316 p = uq.buf;
2317 }
2318
2319 /* Git does not track empty, non-toplevel directories. */
2320 if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
2321 tree_content_remove(&b->branch_tree, p, NULL, 0);
2322 return;
2323 }
2324
2325 if (S_ISGITLINK(mode)) {
2326 if (inline_data)
2327 die("Git links cannot be specified 'inline': %s",
2328 command_buf.buf);
2329 else if (oe) {
2330 if (oe->type != OBJ_COMMIT)
2331 die("Not a commit (actually a %s): %s",
2332 typename(oe->type), command_buf.buf);
2333 }
2334 /*
2335 * Accept the sha1 without checking; it expected to be in
2336 * another repository.
2337 */
2338 } else if (inline_data) {
2339 if (S_ISDIR(mode))
2340 die("Directories cannot be specified 'inline': %s",
2341 command_buf.buf);
2342 if (p != uq.buf) {
2343 strbuf_addstr(&uq, p);
2344 p = uq.buf;
2345 }
2346 read_next_command();
2347 parse_and_store_blob(&last_blob, sha1, 0);
2348 } else {
2349 enum object_type expected = S_ISDIR(mode) ?
2350 OBJ_TREE: OBJ_BLOB;
2351 enum object_type type = oe ? oe->type :
2352 sha1_object_info(sha1, NULL);
2353 if (type < 0)
2354 die("%s not found: %s",
2355 S_ISDIR(mode) ? "Tree" : "Blob",
2356 command_buf.buf);
2357 if (type != expected)
2358 die("Not a %s (actually a %s): %s",
2359 typename(expected), typename(type),
2360 command_buf.buf);
2361 }
2362
2363 if (!*p) {
2364 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2365 return;
2366 }
2367 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2368}
2369
2370static void file_change_d(struct branch *b)
2371{
2372 const char *p = command_buf.buf + 2;
2373 static struct strbuf uq = STRBUF_INIT;
2374 const char *endp;
2375
2376 strbuf_reset(&uq);
2377 if (!unquote_c_style(&uq, p, &endp)) {
2378 if (*endp)
2379 die("Garbage after path in: %s", command_buf.buf);
2380 p = uq.buf;
2381 }
2382 tree_content_remove(&b->branch_tree, p, NULL, 1);
2383}
2384
2385static void file_change_cr(struct branch *b, int rename)
2386{
2387 const char *s, *d;
2388 static struct strbuf s_uq = STRBUF_INIT;
2389 static struct strbuf d_uq = STRBUF_INIT;
2390 const char *endp;
2391 struct tree_entry leaf;
2392
2393 s = command_buf.buf + 2;
2394 strbuf_reset(&s_uq);
2395 if (!unquote_c_style(&s_uq, s, &endp)) {
2396 if (*endp != ' ')
2397 die("Missing space after source: %s", command_buf.buf);
2398 } else {
2399 endp = strchr(s, ' ');
2400 if (!endp)
2401 die("Missing space after source: %s", command_buf.buf);
2402 strbuf_add(&s_uq, s, endp - s);
2403 }
2404 s = s_uq.buf;
2405
2406 endp++;
2407 if (!*endp)
2408 die("Missing dest: %s", command_buf.buf);
2409
2410 d = endp;
2411 strbuf_reset(&d_uq);
2412 if (!unquote_c_style(&d_uq, d, &endp)) {
2413 if (*endp)
2414 die("Garbage after dest in: %s", command_buf.buf);
2415 d = d_uq.buf;
2416 }
2417
2418 memset(&leaf, 0, sizeof(leaf));
2419 if (rename)
2420 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2421 else
2422 tree_content_get(&b->branch_tree, s, &leaf, 1);
2423 if (!leaf.versions[1].mode)
2424 die("Path %s not in branch", s);
2425 if (!*d) { /* C "path/to/subdir" "" */
2426 tree_content_replace(&b->branch_tree,
2427 leaf.versions[1].sha1,
2428 leaf.versions[1].mode,
2429 leaf.tree);
2430 return;
2431 }
2432 tree_content_set(&b->branch_tree, d,
2433 leaf.versions[1].sha1,
2434 leaf.versions[1].mode,
2435 leaf.tree);
2436}
2437
2438static void note_change_n(struct branch *b, unsigned char *old_fanout)
2439{
2440 const char *p = command_buf.buf + 2;
2441 static struct strbuf uq = STRBUF_INIT;
2442 struct object_entry *oe;
2443 struct branch *s;
2444 unsigned char sha1[20], commit_sha1[20];
2445 char path[60];
2446 uint16_t inline_data = 0;
2447 unsigned char new_fanout;
2448
2449 /*
2450 * When loading a branch, we don't traverse its tree to count the real
2451 * number of notes (too expensive to do this for all non-note refs).
2452 * This means that recently loaded notes refs might incorrectly have
2453 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2454 *
2455 * Fix this by traversing the tree and counting the number of notes
2456 * when b->num_notes == 0. If the notes tree is truly empty, the
2457 * calculation should not take long.
2458 */
2459 if (b->num_notes == 0 && *old_fanout == 0) {
2460 /* Invoke change_note_fanout() in "counting mode". */
2461 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2462 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2463 }
2464
2465 /* Now parse the notemodify command. */
2466 /* <dataref> or 'inline' */
2467 if (*p == ':') {
2468 oe = find_mark(parse_mark_ref_space(&p));
2469 hashcpy(sha1, oe->idx.sha1);
2470 } else if (starts_with(p, "inline ")) {
2471 inline_data = 1;
2472 oe = NULL; /* not used with inline_data, but makes gcc happy */
2473 p += strlen("inline"); /* advance to space */
2474 } else {
2475 if (get_sha1_hex(p, sha1))
2476 die("Invalid dataref: %s", command_buf.buf);
2477 oe = find_object(sha1);
2478 p += 40;
2479 if (*p != ' ')
2480 die("Missing space after SHA1: %s", command_buf.buf);
2481 }
2482 assert(*p == ' ');
2483 p++; /* skip space */
2484
2485 /* <commit-ish> */
2486 s = lookup_branch(p);
2487 if (s) {
2488 if (is_null_sha1(s->sha1))
2489 die("Can't add a note on empty branch.");
2490 hashcpy(commit_sha1, s->sha1);
2491 } else if (*p == ':') {
2492 uintmax_t commit_mark = parse_mark_ref_eol(p);
2493 struct object_entry *commit_oe = find_mark(commit_mark);
2494 if (commit_oe->type != OBJ_COMMIT)
2495 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2496 hashcpy(commit_sha1, commit_oe->idx.sha1);
2497 } else if (!get_sha1(p, commit_sha1)) {
2498 unsigned long size;
2499 char *buf = read_object_with_reference(commit_sha1,
2500 commit_type, &size, commit_sha1);
2501 if (!buf || size < 46)
2502 die("Not a valid commit: %s", p);
2503 free(buf);
2504 } else
2505 die("Invalid ref name or SHA1 expression: %s", p);
2506
2507 if (inline_data) {
2508 if (p != uq.buf) {
2509 strbuf_addstr(&uq, p);
2510 p = uq.buf;
2511 }
2512 read_next_command();
2513 parse_and_store_blob(&last_blob, sha1, 0);
2514 } else if (oe) {
2515 if (oe->type != OBJ_BLOB)
2516 die("Not a blob (actually a %s): %s",
2517 typename(oe->type), command_buf.buf);
2518 } else if (!is_null_sha1(sha1)) {
2519 enum object_type type = sha1_object_info(sha1, NULL);
2520 if (type < 0)
2521 die("Blob not found: %s", command_buf.buf);
2522 if (type != OBJ_BLOB)
2523 die("Not a blob (actually a %s): %s",
2524 typename(type), command_buf.buf);
2525 }
2526
2527 construct_path_with_fanout(sha1_to_hex(commit_sha1), *old_fanout, path);
2528 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2529 b->num_notes--;
2530
2531 if (is_null_sha1(sha1))
2532 return; /* nothing to insert */
2533
2534 b->num_notes++;
2535 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2536 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2537 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2538}
2539
2540static void file_change_deleteall(struct branch *b)
2541{
2542 release_tree_content_recursive(b->branch_tree.tree);
2543 hashclr(b->branch_tree.versions[0].sha1);
2544 hashclr(b->branch_tree.versions[1].sha1);
2545 load_tree(&b->branch_tree);
2546 b->num_notes = 0;
2547}
2548
2549static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2550{
2551 if (!buf || size < 46)
2552 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2553 if (memcmp("tree ", buf, 5)
2554 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2555 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2556 hashcpy(b->branch_tree.versions[0].sha1,
2557 b->branch_tree.versions[1].sha1);
2558}
2559
2560static void parse_from_existing(struct branch *b)
2561{
2562 if (is_null_sha1(b->sha1)) {
2563 hashclr(b->branch_tree.versions[0].sha1);
2564 hashclr(b->branch_tree.versions[1].sha1);
2565 } else {
2566 unsigned long size;
2567 char *buf;
2568
2569 buf = read_object_with_reference(b->sha1,
2570 commit_type, &size, b->sha1);
2571 parse_from_commit(b, buf, size);
2572 free(buf);
2573 }
2574}
2575
2576static int parse_from(struct branch *b)
2577{
2578 const char *from;
2579 struct branch *s;
2580
2581 if (!starts_with(command_buf.buf, "from "))
2582 return 0;
2583
2584 if (b->branch_tree.tree) {
2585 release_tree_content_recursive(b->branch_tree.tree);
2586 b->branch_tree.tree = NULL;
2587 }
2588
2589 from = strchr(command_buf.buf, ' ') + 1;
2590 s = lookup_branch(from);
2591 if (b == s)
2592 die("Can't create a branch from itself: %s", b->name);
2593 else if (s) {
2594 unsigned char *t = s->branch_tree.versions[1].sha1;
2595 hashcpy(b->sha1, s->sha1);
2596 hashcpy(b->branch_tree.versions[0].sha1, t);
2597 hashcpy(b->branch_tree.versions[1].sha1, t);
2598 } else if (*from == ':') {
2599 uintmax_t idnum = parse_mark_ref_eol(from);
2600 struct object_entry *oe = find_mark(idnum);
2601 if (oe->type != OBJ_COMMIT)
2602 die("Mark :%" PRIuMAX " not a commit", idnum);
2603 hashcpy(b->sha1, oe->idx.sha1);
2604 if (oe->pack_id != MAX_PACK_ID) {
2605 unsigned long size;
2606 char *buf = gfi_unpack_entry(oe, &size);
2607 parse_from_commit(b, buf, size);
2608 free(buf);
2609 } else
2610 parse_from_existing(b);
2611 } else if (!get_sha1(from, b->sha1))
2612 parse_from_existing(b);
2613 else
2614 die("Invalid ref name or SHA1 expression: %s", from);
2615
2616 read_next_command();
2617 return 1;
2618}
2619
2620static struct hash_list *parse_merge(unsigned int *count)
2621{
2622 struct hash_list *list = NULL, **tail = &list, *n;
2623 const char *from;
2624 struct branch *s;
2625
2626 *count = 0;
2627 while (starts_with(command_buf.buf, "merge ")) {
2628 from = strchr(command_buf.buf, ' ') + 1;
2629 n = xmalloc(sizeof(*n));
2630 s = lookup_branch(from);
2631 if (s)
2632 hashcpy(n->sha1, s->sha1);
2633 else if (*from == ':') {
2634 uintmax_t idnum = parse_mark_ref_eol(from);
2635 struct object_entry *oe = find_mark(idnum);
2636 if (oe->type != OBJ_COMMIT)
2637 die("Mark :%" PRIuMAX " not a commit", idnum);
2638 hashcpy(n->sha1, oe->idx.sha1);
2639 } else if (!get_sha1(from, n->sha1)) {
2640 unsigned long size;
2641 char *buf = read_object_with_reference(n->sha1,
2642 commit_type, &size, n->sha1);
2643 if (!buf || size < 46)
2644 die("Not a valid commit: %s", from);
2645 free(buf);
2646 } else
2647 die("Invalid ref name or SHA1 expression: %s", from);
2648
2649 n->next = NULL;
2650 *tail = n;
2651 tail = &n->next;
2652
2653 (*count)++;
2654 read_next_command();
2655 }
2656 return list;
2657}
2658
2659static void parse_new_commit(void)
2660{
2661 static struct strbuf msg = STRBUF_INIT;
2662 struct branch *b;
2663 char *sp;
2664 char *author = NULL;
2665 char *committer = NULL;
2666 struct hash_list *merge_list = NULL;
2667 unsigned int merge_count;
2668 unsigned char prev_fanout, new_fanout;
2669
2670 /* Obtain the branch name from the rest of our command */
2671 sp = strchr(command_buf.buf, ' ') + 1;
2672 b = lookup_branch(sp);
2673 if (!b)
2674 b = new_branch(sp);
2675
2676 read_next_command();
2677 parse_mark();
2678 if (starts_with(command_buf.buf, "author ")) {
2679 author = parse_ident(command_buf.buf + 7);
2680 read_next_command();
2681 }
2682 if (starts_with(command_buf.buf, "committer ")) {
2683 committer = parse_ident(command_buf.buf + 10);
2684 read_next_command();
2685 }
2686 if (!committer)
2687 die("Expected committer but didn't get one");
2688 parse_data(&msg, 0, NULL);
2689 read_next_command();
2690 parse_from(b);
2691 merge_list = parse_merge(&merge_count);
2692
2693 /* ensure the branch is active/loaded */
2694 if (!b->branch_tree.tree || !max_active_branches) {
2695 unload_one_branch();
2696 load_branch(b);
2697 }
2698
2699 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2700
2701 /* file_change* */
2702 while (command_buf.len > 0) {
2703 if (starts_with(command_buf.buf, "M "))
2704 file_change_m(b);
2705 else if (starts_with(command_buf.buf, "D "))
2706 file_change_d(b);
2707 else if (starts_with(command_buf.buf, "R "))
2708 file_change_cr(b, 1);
2709 else if (starts_with(command_buf.buf, "C "))
2710 file_change_cr(b, 0);
2711 else if (starts_with(command_buf.buf, "N "))
2712 note_change_n(b, &prev_fanout);
2713 else if (!strcmp("deleteall", command_buf.buf))
2714 file_change_deleteall(b);
2715 else if (starts_with(command_buf.buf, "ls "))
2716 parse_ls(b);
2717 else {
2718 unread_command_buf = 1;
2719 break;
2720 }
2721 if (read_next_command() == EOF)
2722 break;
2723 }
2724
2725 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2726 if (new_fanout != prev_fanout)
2727 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2728
2729 /* build the tree and the commit */
2730 store_tree(&b->branch_tree);
2731 hashcpy(b->branch_tree.versions[0].sha1,
2732 b->branch_tree.versions[1].sha1);
2733
2734 strbuf_reset(&new_data);
2735 strbuf_addf(&new_data, "tree %s\n",
2736 sha1_to_hex(b->branch_tree.versions[1].sha1));
2737 if (!is_null_sha1(b->sha1))
2738 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2739 while (merge_list) {
2740 struct hash_list *next = merge_list->next;
2741 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2742 free(merge_list);
2743 merge_list = next;
2744 }
2745 strbuf_addf(&new_data,
2746 "author %s\n"
2747 "committer %s\n"
2748 "\n",
2749 author ? author : committer, committer);
2750 strbuf_addbuf(&new_data, &msg);
2751 free(author);
2752 free(committer);
2753
2754 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2755 b->pack_id = pack_id;
2756 b->last_commit = object_count_by_type[OBJ_COMMIT];
2757}
2758
2759static void parse_new_tag(void)
2760{
2761 static struct strbuf msg = STRBUF_INIT;
2762 char *sp;
2763 const char *from;
2764 char *tagger;
2765 struct branch *s;
2766 struct tag *t;
2767 uintmax_t from_mark = 0;
2768 unsigned char sha1[20];
2769 enum object_type type;
2770
2771 /* Obtain the new tag name from the rest of our command */
2772 sp = strchr(command_buf.buf, ' ') + 1;
2773 t = pool_alloc(sizeof(struct tag));
2774 memset(t, 0, sizeof(struct tag));
2775 t->name = pool_strdup(sp);
2776 if (last_tag)
2777 last_tag->next_tag = t;
2778 else
2779 first_tag = t;
2780 last_tag = t;
2781 read_next_command();
2782
2783 /* from ... */
2784 if (!starts_with(command_buf.buf, "from "))
2785 die("Expected from command, got %s", command_buf.buf);
2786 from = strchr(command_buf.buf, ' ') + 1;
2787 s = lookup_branch(from);
2788 if (s) {
2789 if (is_null_sha1(s->sha1))
2790 die("Can't tag an empty branch.");
2791 hashcpy(sha1, s->sha1);
2792 type = OBJ_COMMIT;
2793 } else if (*from == ':') {
2794 struct object_entry *oe;
2795 from_mark = parse_mark_ref_eol(from);
2796 oe = find_mark(from_mark);
2797 type = oe->type;
2798 hashcpy(sha1, oe->idx.sha1);
2799 } else if (!get_sha1(from, sha1)) {
2800 struct object_entry *oe = find_object(sha1);
2801 if (!oe) {
2802 type = sha1_object_info(sha1, NULL);
2803 if (type < 0)
2804 die("Not a valid object: %s", from);
2805 } else
2806 type = oe->type;
2807 } else
2808 die("Invalid ref name or SHA1 expression: %s", from);
2809 read_next_command();
2810
2811 /* tagger ... */
2812 if (starts_with(command_buf.buf, "tagger ")) {
2813 tagger = parse_ident(command_buf.buf + 7);
2814 read_next_command();
2815 } else
2816 tagger = NULL;
2817
2818 /* tag payload/message */
2819 parse_data(&msg, 0, NULL);
2820
2821 /* build the tag object */
2822 strbuf_reset(&new_data);
2823
2824 strbuf_addf(&new_data,
2825 "object %s\n"
2826 "type %s\n"
2827 "tag %s\n",
2828 sha1_to_hex(sha1), typename(type), t->name);
2829 if (tagger)
2830 strbuf_addf(&new_data,
2831 "tagger %s\n", tagger);
2832 strbuf_addch(&new_data, '\n');
2833 strbuf_addbuf(&new_data, &msg);
2834 free(tagger);
2835
2836 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2837 t->pack_id = MAX_PACK_ID;
2838 else
2839 t->pack_id = pack_id;
2840}
2841
2842static void parse_reset_branch(void)
2843{
2844 struct branch *b;
2845 char *sp;
2846
2847 /* Obtain the branch name from the rest of our command */
2848 sp = strchr(command_buf.buf, ' ') + 1;
2849 b = lookup_branch(sp);
2850 if (b) {
2851 hashclr(b->sha1);
2852 hashclr(b->branch_tree.versions[0].sha1);
2853 hashclr(b->branch_tree.versions[1].sha1);
2854 if (b->branch_tree.tree) {
2855 release_tree_content_recursive(b->branch_tree.tree);
2856 b->branch_tree.tree = NULL;
2857 }
2858 }
2859 else
2860 b = new_branch(sp);
2861 read_next_command();
2862 parse_from(b);
2863 if (command_buf.len > 0)
2864 unread_command_buf = 1;
2865}
2866
2867static void cat_blob_write(const char *buf, unsigned long size)
2868{
2869 if (write_in_full(cat_blob_fd, buf, size) != size)
2870 die_errno("Write to frontend failed");
2871}
2872
2873static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2874{
2875 struct strbuf line = STRBUF_INIT;
2876 unsigned long size;
2877 enum object_type type = 0;
2878 char *buf;
2879
2880 if (!oe || oe->pack_id == MAX_PACK_ID) {
2881 buf = read_sha1_file(sha1, &type, &size);
2882 } else {
2883 type = oe->type;
2884 buf = gfi_unpack_entry(oe, &size);
2885 }
2886
2887 /*
2888 * Output based on batch_one_object() from cat-file.c.
2889 */
2890 if (type <= 0) {
2891 strbuf_reset(&line);
2892 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2893 cat_blob_write(line.buf, line.len);
2894 strbuf_release(&line);
2895 free(buf);
2896 return;
2897 }
2898 if (!buf)
2899 die("Can't read object %s", sha1_to_hex(sha1));
2900 if (type != OBJ_BLOB)
2901 die("Object %s is a %s but a blob was expected.",
2902 sha1_to_hex(sha1), typename(type));
2903 strbuf_reset(&line);
2904 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2905 typename(type), size);
2906 cat_blob_write(line.buf, line.len);
2907 strbuf_release(&line);
2908 cat_blob_write(buf, size);
2909 cat_blob_write("\n", 1);
2910 if (oe && oe->pack_id == pack_id) {
2911 last_blob.offset = oe->idx.offset;
2912 strbuf_attach(&last_blob.data, buf, size, size);
2913 last_blob.depth = oe->depth;
2914 } else
2915 free(buf);
2916}
2917
2918static void parse_cat_blob(void)
2919{
2920 const char *p;
2921 struct object_entry *oe = oe;
2922 unsigned char sha1[20];
2923
2924 /* cat-blob SP <object> LF */
2925 p = command_buf.buf + strlen("cat-blob ");
2926 if (*p == ':') {
2927 oe = find_mark(parse_mark_ref_eol(p));
2928 if (!oe)
2929 die("Unknown mark: %s", command_buf.buf);
2930 hashcpy(sha1, oe->idx.sha1);
2931 } else {
2932 if (get_sha1_hex(p, sha1))
2933 die("Invalid dataref: %s", command_buf.buf);
2934 if (p[40])
2935 die("Garbage after SHA1: %s", command_buf.buf);
2936 oe = find_object(sha1);
2937 }
2938
2939 cat_blob(oe, sha1);
2940}
2941
2942static struct object_entry *dereference(struct object_entry *oe,
2943 unsigned char sha1[20])
2944{
2945 unsigned long size;
2946 char *buf = NULL;
2947 if (!oe) {
2948 enum object_type type = sha1_object_info(sha1, NULL);
2949 if (type < 0)
2950 die("object not found: %s", sha1_to_hex(sha1));
2951 /* cache it! */
2952 oe = insert_object(sha1);
2953 oe->type = type;
2954 oe->pack_id = MAX_PACK_ID;
2955 oe->idx.offset = 1;
2956 }
2957 switch (oe->type) {
2958 case OBJ_TREE: /* easy case. */
2959 return oe;
2960 case OBJ_COMMIT:
2961 case OBJ_TAG:
2962 break;
2963 default:
2964 die("Not a tree-ish: %s", command_buf.buf);
2965 }
2966
2967 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2968 buf = gfi_unpack_entry(oe, &size);
2969 } else {
2970 enum object_type unused;
2971 buf = read_sha1_file(sha1, &unused, &size);
2972 }
2973 if (!buf)
2974 die("Can't load object %s", sha1_to_hex(sha1));
2975
2976 /* Peel one layer. */
2977 switch (oe->type) {
2978 case OBJ_TAG:
2979 if (size < 40 + strlen("object ") ||
2980 get_sha1_hex(buf + strlen("object "), sha1))
2981 die("Invalid SHA1 in tag: %s", command_buf.buf);
2982 break;
2983 case OBJ_COMMIT:
2984 if (size < 40 + strlen("tree ") ||
2985 get_sha1_hex(buf + strlen("tree "), sha1))
2986 die("Invalid SHA1 in commit: %s", command_buf.buf);
2987 }
2988
2989 free(buf);
2990 return find_object(sha1);
2991}
2992
2993static struct object_entry *parse_treeish_dataref(const char **p)
2994{
2995 unsigned char sha1[20];
2996 struct object_entry *e;
2997
2998 if (**p == ':') { /* <mark> */
2999 e = find_mark(parse_mark_ref_space(p));
3000 if (!e)
3001 die("Unknown mark: %s", command_buf.buf);
3002 hashcpy(sha1, e->idx.sha1);
3003 } else { /* <sha1> */
3004 if (get_sha1_hex(*p, sha1))
3005 die("Invalid dataref: %s", command_buf.buf);
3006 e = find_object(sha1);
3007 *p += 40;
3008 }
3009
3010 while (!e || e->type != OBJ_TREE)
3011 e = dereference(e, sha1);
3012 return e;
3013}
3014
3015static void print_ls(int mode, const unsigned char *sha1, const char *path)
3016{
3017 static struct strbuf line = STRBUF_INIT;
3018
3019 /* See show_tree(). */
3020 const char *type =
3021 S_ISGITLINK(mode) ? commit_type :
3022 S_ISDIR(mode) ? tree_type :
3023 blob_type;
3024
3025 if (!mode) {
3026 /* missing SP path LF */
3027 strbuf_reset(&line);
3028 strbuf_addstr(&line, "missing ");
3029 quote_c_style(path, &line, NULL, 0);
3030 strbuf_addch(&line, '\n');
3031 } else {
3032 /* mode SP type SP object_name TAB path LF */
3033 strbuf_reset(&line);
3034 strbuf_addf(&line, "%06o %s %s\t",
3035 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3036 quote_c_style(path, &line, NULL, 0);
3037 strbuf_addch(&line, '\n');
3038 }
3039 cat_blob_write(line.buf, line.len);
3040}
3041
3042static void parse_ls(struct branch *b)
3043{
3044 const char *p;
3045 struct tree_entry *root = NULL;
3046 struct tree_entry leaf = {NULL};
3047
3048 /* ls SP (<tree-ish> SP)? <path> */
3049 p = command_buf.buf + strlen("ls ");
3050 if (*p == '"') {
3051 if (!b)
3052 die("Not in a commit: %s", command_buf.buf);
3053 root = &b->branch_tree;
3054 } else {
3055 struct object_entry *e = parse_treeish_dataref(&p);
3056 root = new_tree_entry();
3057 hashcpy(root->versions[1].sha1, e->idx.sha1);
3058 if (!is_null_sha1(root->versions[1].sha1))
3059 root->versions[1].mode = S_IFDIR;
3060 load_tree(root);
3061 if (*p++ != ' ')
3062 die("Missing space after tree-ish: %s", command_buf.buf);
3063 }
3064 if (*p == '"') {
3065 static struct strbuf uq = STRBUF_INIT;
3066 const char *endp;
3067 strbuf_reset(&uq);
3068 if (unquote_c_style(&uq, p, &endp))
3069 die("Invalid path: %s", command_buf.buf);
3070 if (*endp)
3071 die("Garbage after path in: %s", command_buf.buf);
3072 p = uq.buf;
3073 }
3074 tree_content_get(root, p, &leaf, 1);
3075 /*
3076 * A directory in preparation would have a sha1 of zero
3077 * until it is saved. Save, for simplicity.
3078 */
3079 if (S_ISDIR(leaf.versions[1].mode))
3080 store_tree(&leaf);
3081
3082 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
3083 if (leaf.tree)
3084 release_tree_content_recursive(leaf.tree);
3085 if (!b || root != &b->branch_tree)
3086 release_tree_entry(root);
3087}
3088
3089static void checkpoint(void)
3090{
3091 checkpoint_requested = 0;
3092 if (object_count) {
3093 cycle_packfile();
3094 dump_branches();
3095 dump_tags();
3096 dump_marks();
3097 }
3098}
3099
3100static void parse_checkpoint(void)
3101{
3102 checkpoint_requested = 1;
3103 skip_optional_lf();
3104}
3105
3106static void parse_progress(void)
3107{
3108 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3109 fputc('\n', stdout);
3110 fflush(stdout);
3111 skip_optional_lf();
3112}
3113
3114static char* make_fast_import_path(const char *path)
3115{
3116 struct strbuf abs_path = STRBUF_INIT;
3117
3118 if (!relative_marks_paths || is_absolute_path(path))
3119 return xstrdup(path);
3120 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
3121 return strbuf_detach(&abs_path, NULL);
3122}
3123
3124static void option_import_marks(const char *marks,
3125 int from_stream, int ignore_missing)
3126{
3127 if (import_marks_file) {
3128 if (from_stream)
3129 die("Only one import-marks command allowed per stream");
3130
3131 /* read previous mark file */
3132 if(!import_marks_file_from_stream)
3133 read_marks();
3134 }
3135
3136 import_marks_file = make_fast_import_path(marks);
3137 safe_create_leading_directories_const(import_marks_file);
3138 import_marks_file_from_stream = from_stream;
3139 import_marks_file_ignore_missing = ignore_missing;
3140}
3141
3142static void option_date_format(const char *fmt)
3143{
3144 if (!strcmp(fmt, "raw"))
3145 whenspec = WHENSPEC_RAW;
3146 else if (!strcmp(fmt, "rfc2822"))
3147 whenspec = WHENSPEC_RFC2822;
3148 else if (!strcmp(fmt, "now"))
3149 whenspec = WHENSPEC_NOW;
3150 else
3151 die("unknown --date-format argument %s", fmt);
3152}
3153
3154static unsigned long ulong_arg(const char *option, const char *arg)
3155{
3156 char *endptr;
3157 unsigned long rv = strtoul(arg, &endptr, 0);
3158 if (strchr(arg, '-') || endptr == arg || *endptr)
3159 die("%s: argument must be a non-negative integer", option);
3160 return rv;
3161}
3162
3163static void option_depth(const char *depth)
3164{
3165 max_depth = ulong_arg("--depth", depth);
3166 if (max_depth > MAX_DEPTH)
3167 die("--depth cannot exceed %u", MAX_DEPTH);
3168}
3169
3170static void option_active_branches(const char *branches)
3171{
3172 max_active_branches = ulong_arg("--active-branches", branches);
3173}
3174
3175static void option_export_marks(const char *marks)
3176{
3177 export_marks_file = make_fast_import_path(marks);
3178 safe_create_leading_directories_const(export_marks_file);
3179}
3180
3181static void option_cat_blob_fd(const char *fd)
3182{
3183 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3184 if (n > (unsigned long) INT_MAX)
3185 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3186 cat_blob_fd = (int) n;
3187}
3188
3189static void option_export_pack_edges(const char *edges)
3190{
3191 if (pack_edges)
3192 fclose(pack_edges);
3193 pack_edges = fopen(edges, "a");
3194 if (!pack_edges)
3195 die_errno("Cannot open '%s'", edges);
3196}
3197
3198static int parse_one_option(const char *option)
3199{
3200 if (starts_with(option, "max-pack-size=")) {
3201 unsigned long v;
3202 if (!git_parse_ulong(option + 14, &v))
3203 return 0;
3204 if (v < 8192) {
3205 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3206 v *= 1024 * 1024;
3207 } else if (v < 1024 * 1024) {
3208 warning("minimum max-pack-size is 1 MiB");
3209 v = 1024 * 1024;
3210 }
3211 max_packsize = v;
3212 } else if (starts_with(option, "big-file-threshold=")) {
3213 unsigned long v;
3214 if (!git_parse_ulong(option + 19, &v))
3215 return 0;
3216 big_file_threshold = v;
3217 } else if (starts_with(option, "depth=")) {
3218 option_depth(option + 6);
3219 } else if (starts_with(option, "active-branches=")) {
3220 option_active_branches(option + 16);
3221 } else if (starts_with(option, "export-pack-edges=")) {
3222 option_export_pack_edges(option + 18);
3223 } else if (starts_with(option, "quiet")) {
3224 show_stats = 0;
3225 } else if (starts_with(option, "stats")) {
3226 show_stats = 1;
3227 } else {
3228 return 0;
3229 }
3230
3231 return 1;
3232}
3233
3234static int parse_one_feature(const char *feature, int from_stream)
3235{
3236 if (starts_with(feature, "date-format=")) {
3237 option_date_format(feature + 12);
3238 } else if (starts_with(feature, "import-marks=")) {
3239 option_import_marks(feature + 13, from_stream, 0);
3240 } else if (starts_with(feature, "import-marks-if-exists=")) {
3241 option_import_marks(feature + strlen("import-marks-if-exists="),
3242 from_stream, 1);
3243 } else if (starts_with(feature, "export-marks=")) {
3244 option_export_marks(feature + 13);
3245 } else if (!strcmp(feature, "cat-blob")) {
3246 ; /* Don't die - this feature is supported */
3247 } else if (!strcmp(feature, "relative-marks")) {
3248 relative_marks_paths = 1;
3249 } else if (!strcmp(feature, "no-relative-marks")) {
3250 relative_marks_paths = 0;
3251 } else if (!strcmp(feature, "done")) {
3252 require_explicit_termination = 1;
3253 } else if (!strcmp(feature, "force")) {
3254 force_update = 1;
3255 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3256 ; /* do nothing; we have the feature */
3257 } else {
3258 return 0;
3259 }
3260
3261 return 1;
3262}
3263
3264static void parse_feature(void)
3265{
3266 char *feature = command_buf.buf + 8;
3267
3268 if (seen_data_command)
3269 die("Got feature command '%s' after data command", feature);
3270
3271 if (parse_one_feature(feature, 1))
3272 return;
3273
3274 die("This version of fast-import does not support feature %s.", feature);
3275}
3276
3277static void parse_option(void)
3278{
3279 char *option = command_buf.buf + 11;
3280
3281 if (seen_data_command)
3282 die("Got option command '%s' after data command", option);
3283
3284 if (parse_one_option(option))
3285 return;
3286
3287 die("This version of fast-import does not support option: %s", option);
3288}
3289
3290static int git_pack_config(const char *k, const char *v, void *cb)
3291{
3292 if (!strcmp(k, "pack.depth")) {
3293 max_depth = git_config_int(k, v);
3294 if (max_depth > MAX_DEPTH)
3295 max_depth = MAX_DEPTH;
3296 return 0;
3297 }
3298 if (!strcmp(k, "pack.compression")) {
3299 int level = git_config_int(k, v);
3300 if (level == -1)
3301 level = Z_DEFAULT_COMPRESSION;
3302 else if (level < 0 || level > Z_BEST_COMPRESSION)
3303 die("bad pack compression level %d", level);
3304 pack_compression_level = level;
3305 pack_compression_seen = 1;
3306 return 0;
3307 }
3308 if (!strcmp(k, "pack.indexversion")) {
3309 pack_idx_opts.version = git_config_int(k, v);
3310 if (pack_idx_opts.version > 2)
3311 die("bad pack.indexversion=%"PRIu32,
3312 pack_idx_opts.version);
3313 return 0;
3314 }
3315 if (!strcmp(k, "pack.packsizelimit")) {
3316 max_packsize = git_config_ulong(k, v);
3317 return 0;
3318 }
3319 return git_default_config(k, v, cb);
3320}
3321
3322static const char fast_import_usage[] =
3323"git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3324
3325static void parse_argv(void)
3326{
3327 unsigned int i;
3328
3329 for (i = 1; i < global_argc; i++) {
3330 const char *a = global_argv[i];
3331
3332 if (*a != '-' || !strcmp(a, "--"))
3333 break;
3334
3335 if (parse_one_option(a + 2))
3336 continue;
3337
3338 if (parse_one_feature(a + 2, 0))
3339 continue;
3340
3341 if (starts_with(a + 2, "cat-blob-fd=")) {
3342 option_cat_blob_fd(a + 2 + strlen("cat-blob-fd="));
3343 continue;
3344 }
3345
3346 die("unknown option %s", a);
3347 }
3348 if (i != global_argc)
3349 usage(fast_import_usage);
3350
3351 seen_data_command = 1;
3352 if (import_marks_file)
3353 read_marks();
3354}
3355
3356int main(int argc, char **argv)
3357{
3358 unsigned int i;
3359
3360 git_extract_argv0_path(argv[0]);
3361
3362 git_setup_gettext();
3363
3364 if (argc == 2 && !strcmp(argv[1], "-h"))
3365 usage(fast_import_usage);
3366
3367 setup_git_directory();
3368 reset_pack_idx_option(&pack_idx_opts);
3369 git_config(git_pack_config, NULL);
3370 if (!pack_compression_seen && core_compression_seen)
3371 pack_compression_level = core_compression_level;
3372
3373 alloc_objects(object_entry_alloc);
3374 strbuf_init(&command_buf, 0);
3375 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3376 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3377 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3378 marks = pool_calloc(1, sizeof(struct mark_set));
3379
3380 global_argc = argc;
3381 global_argv = argv;
3382
3383 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3384 for (i = 0; i < (cmd_save - 1); i++)
3385 rc_free[i].next = &rc_free[i + 1];
3386 rc_free[cmd_save - 1].next = NULL;
3387
3388 prepare_packed_git();
3389 start_packfile();
3390 set_die_routine(die_nicely);
3391 set_checkpoint_signal();
3392 while (read_next_command() != EOF) {
3393 if (!strcmp("blob", command_buf.buf))
3394 parse_new_blob();
3395 else if (starts_with(command_buf.buf, "ls "))
3396 parse_ls(NULL);
3397 else if (starts_with(command_buf.buf, "commit "))
3398 parse_new_commit();
3399 else if (starts_with(command_buf.buf, "tag "))
3400 parse_new_tag();
3401 else if (starts_with(command_buf.buf, "reset "))
3402 parse_reset_branch();
3403 else if (!strcmp("checkpoint", command_buf.buf))
3404 parse_checkpoint();
3405 else if (!strcmp("done", command_buf.buf))
3406 break;
3407 else if (starts_with(command_buf.buf, "progress "))
3408 parse_progress();
3409 else if (starts_with(command_buf.buf, "feature "))
3410 parse_feature();
3411 else if (starts_with(command_buf.buf, "option git "))
3412 parse_option();
3413 else if (starts_with(command_buf.buf, "option "))
3414 /* ignore non-git options*/;
3415 else
3416 die("Unsupported command: %s", command_buf.buf);
3417
3418 if (checkpoint_requested)
3419 checkpoint();
3420 }
3421
3422 /* argv hasn't been parsed yet, do so */
3423 if (!seen_data_command)
3424 parse_argv();
3425
3426 if (require_explicit_termination && feof(stdin))
3427 die("stream ends early");
3428
3429 end_packfile();
3430
3431 dump_branches();
3432 dump_tags();
3433 unkeep_all_packs();
3434 dump_marks();
3435
3436 if (pack_edges)
3437 fclose(pack_edges);
3438
3439 if (show_stats) {
3440 uintmax_t total_count = 0, duplicate_count = 0;
3441 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3442 total_count += object_count_by_type[i];
3443 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3444 duplicate_count += duplicate_count_by_type[i];
3445
3446 fprintf(stderr, "%s statistics:\n", argv[0]);
3447 fprintf(stderr, "---------------------------------------------------------------------\n");
3448 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3449 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3450 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB], delta_count_attempts_by_type[OBJ_BLOB]);
3451 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE], delta_count_attempts_by_type[OBJ_TREE]);
3452 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT], delta_count_attempts_by_type[OBJ_COMMIT]);
3453 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas of %10" PRIuMAX" attempts)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG], delta_count_attempts_by_type[OBJ_TAG]);
3454 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3455 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3456 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3457 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3458 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3459 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3460 fprintf(stderr, "---------------------------------------------------------------------\n");
3461 pack_report();
3462 fprintf(stderr, "---------------------------------------------------------------------\n");
3463 fprintf(stderr, "\n");
3464 }
3465
3466 return failure ? 1 : 0;
3467}