Merge branch 'hv/submodule-update-none'
[git.git] / fast-import.c
blob742e7da6b8b58dd0803d89c6ad6c59589274a31b
1 /*
2 (See Documentation/git-fast-import.txt for maintained documentation.)
3 Format of STDIN stream:
5 stream ::= cmd*;
7 cmd ::= new_blob
8 | new_commit
9 | new_tag
10 | reset_branch
11 | checkpoint
12 | progress
15 new_blob ::= 'blob' lf
16 mark?
17 file_content;
18 file_content ::= data;
20 new_commit ::= 'commit' sp ref_str lf
21 mark?
22 ('author' (sp name)? sp '<' email '>' sp when lf)?
23 'committer' (sp name)? sp '<' email '>' sp when lf
24 commit_msg
25 ('from' sp committish lf)?
26 ('merge' sp committish lf)*
27 (file_change | ls)*
28 lf?;
29 commit_msg ::= data;
31 ls ::= 'ls' sp '"' quoted(path) '"' lf;
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 committish lf;
47 note_inm ::= 'N' sp 'inline' sp committish lf
48 data;
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp committish lf
52 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
53 tag_msg;
54 tag_msg ::= data;
56 reset_branch ::= 'reset' sp ref_str lf
57 ('from' sp committish lf)?
58 lf?;
60 checkpoint ::= 'checkpoint' lf
61 lf?;
63 progress ::= 'progress' sp not_lf* lf
64 lf?;
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.
72 mark ::= 'mark' sp idnum lf;
73 data ::= (delimited_data | exact_data)
74 lf?;
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;
83 # note: declen indicates the length of binary_data in bytes.
84 # declen does not include the lf preceding the binary data.
86 exact_data ::= 'data' sp declen lf
87 binary_data;
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.
96 committish ::= (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'
106 declen ::= # unsigned 32 bit value, ascii base10 notation;
107 bigint ::= # unsigned integer value, ascii base10 notation;
108 binary_data ::= # file content, not interpreted;
110 when ::= raw_when | rfc2822_when;
111 raw_when ::= ts sp tz;
112 rfc2822_when ::= # Valid RFC 2822 date and time;
114 sp ::= # ASCII space character;
115 lf ::= # ASCII newline (LF) character;
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.
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;
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.
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;
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.
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).
147 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
148 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
150 comment ::= '#' not_lf* lf;
151 not_lf ::= # Any byte that is not ASCII newline (LF);
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"
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)
174 * We abuse the setuid bit on directories to mean "do not delta".
176 #define NO_DELTA S_ISUID
178 struct 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;
186 struct 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 */
193 struct mark_set {
194 union {
195 struct object_entry *marked[1024];
196 struct mark_set *sets[1024];
197 } data;
198 unsigned int shift;
201 struct last_object {
202 struct strbuf data;
203 off_t offset;
204 unsigned int depth;
205 unsigned no_swap : 1;
208 struct mem_pool {
209 struct mem_pool *next_pool;
210 char *next_free;
211 char *end;
212 uintmax_t space[FLEX_ARRAY]; /* more */
215 struct atom_str {
216 struct atom_str *next_atom;
217 unsigned short str_len;
218 char str_dat[FLEX_ARRAY]; /* more */
221 struct tree_content;
222 struct 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];
231 struct 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 */
238 struct avail_tree_content {
239 unsigned int entry_capacity; /* must match tree_content */
240 struct avail_tree_content *next_avail;
243 struct 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];
255 struct tag {
256 struct tag *next_tag;
257 const char *name;
258 unsigned int pack_id;
259 unsigned char sha1[20];
262 struct hash_list {
263 struct hash_list *next;
264 unsigned char sha1[20];
267 typedef enum {
268 WHENSPEC_RAW = 1,
269 WHENSPEC_RFC2822,
270 WHENSPEC_NOW
271 } whenspec_type;
273 struct recent_command {
274 struct recent_command *prev;
275 struct recent_command *next;
276 char *buf;
279 /* Configured limits on output */
280 static unsigned long max_depth = 10;
281 static off_t max_packsize;
282 static int force_update;
283 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
284 static int pack_compression_seen;
286 /* Stats and misc. counters */
287 static uintmax_t alloc_count;
288 static uintmax_t marks_set_count;
289 static uintmax_t object_count_by_type[1 << TYPE_BITS];
290 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
291 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
292 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
293 static unsigned long object_count;
294 static unsigned long branch_count;
295 static unsigned long branch_load_count;
296 static int failure;
297 static FILE *pack_edges;
298 static unsigned int show_stats = 1;
299 static int global_argc;
300 static const char **global_argv;
302 /* Memory pools */
303 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
304 static size_t total_allocd;
305 static struct mem_pool *mem_pool;
307 /* Atom management */
308 static unsigned int atom_table_sz = 4451;
309 static unsigned int atom_cnt;
310 static struct atom_str **atom_table;
312 /* The .pack file being generated */
313 static struct pack_idx_option pack_idx_opts;
314 static unsigned int pack_id;
315 static struct sha1file *pack_file;
316 static struct packed_git *pack_data;
317 static struct packed_git **all_packs;
318 static off_t pack_size;
320 /* Table of objects we've written. */
321 static unsigned int object_entry_alloc = 5000;
322 static struct object_entry_pool *blocks;
323 static struct object_entry *object_table[1 << 16];
324 static struct mark_set *marks;
325 static const char *export_marks_file;
326 static const char *import_marks_file;
327 static int import_marks_file_from_stream;
328 static int import_marks_file_ignore_missing;
329 static int relative_marks_paths;
331 /* Our last blob */
332 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
334 /* Tree management */
335 static unsigned int tree_entry_alloc = 1000;
336 static void *avail_tree_entry;
337 static unsigned int avail_tree_table_sz = 100;
338 static struct avail_tree_content **avail_tree_table;
339 static struct strbuf old_tree = STRBUF_INIT;
340 static struct strbuf new_tree = STRBUF_INIT;
342 /* Branch data */
343 static unsigned long max_active_branches = 5;
344 static unsigned long cur_active_branches;
345 static unsigned long branch_table_sz = 1039;
346 static struct branch **branch_table;
347 static struct branch *active_branches;
349 /* Tag data */
350 static struct tag *first_tag;
351 static struct tag *last_tag;
353 /* Input stream parsing */
354 static whenspec_type whenspec = WHENSPEC_RAW;
355 static struct strbuf command_buf = STRBUF_INIT;
356 static int unread_command_buf;
357 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
358 static struct recent_command *cmd_tail = &cmd_hist;
359 static struct recent_command *rc_free;
360 static unsigned int cmd_save = 100;
361 static uintmax_t next_mark;
362 static struct strbuf new_data = STRBUF_INIT;
363 static int seen_data_command;
364 static int require_explicit_termination;
366 /* Signal handling */
367 static volatile sig_atomic_t checkpoint_requested;
369 /* Where to write output of cat-blob commands */
370 static int cat_blob_fd = STDOUT_FILENO;
372 static void parse_argv(void);
373 static void parse_cat_blob(void);
374 static void parse_ls(struct branch *b);
376 static void write_branch_report(FILE *rpt, struct branch *b)
378 fprintf(rpt, "%s:\n", b->name);
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);
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);
394 fputs(" last pack : ", rpt);
395 if (b->pack_id < MAX_PACK_ID)
396 fprintf(rpt, "%u", b->pack_id);
397 fputc('\n', rpt);
399 fputc('\n', rpt);
402 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
404 static void write_crash_report(const char *err)
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;
412 if (!rpt) {
413 error("can't write crash report %s: %s", loc, strerror(errno));
414 return;
417 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
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);
425 fputs("fatal: ", rpt);
426 fputs(err, rpt);
427 fputc('\n', rpt);
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);
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);
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);
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);
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);
483 fputc('\n', rpt);
484 fputs("-------------------\n", rpt);
485 fputs("END OF CRASH REPORT\n", rpt);
486 fclose(rpt);
489 static void end_packfile(void);
490 static void unkeep_all_packs(void);
491 static void dump_marks(void);
493 static NORETURN void die_nicely(const char *err, va_list params)
495 static int zombie;
496 char message[2 * PATH_MAX];
498 vsnprintf(message, sizeof(message), err, params);
499 fputs("fatal: ", stderr);
500 fputs(message, stderr);
501 fputc('\n', stderr);
503 if (!zombie) {
504 zombie = 1;
505 write_crash_report(message);
506 end_packfile();
507 unkeep_all_packs();
508 dump_marks();
510 exit(128);
513 #ifndef SIGUSR1 /* Windows, for example */
515 static void set_checkpoint_signal(void)
519 #else
521 static void checkpoint_signal(int signo)
523 checkpoint_requested = 1;
526 static void set_checkpoint_signal(void)
528 struct sigaction sa;
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);
537 #endif
539 static void alloc_objects(unsigned int cnt)
541 struct object_entry_pool *b;
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;
552 static struct object_entry *new_object(unsigned char *sha1)
554 struct object_entry *e;
556 if (blocks->next_free == blocks->end)
557 alloc_objects(object_entry_alloc);
559 e = blocks->next_free++;
560 hashcpy(e->idx.sha1, sha1);
561 return e;
564 static struct object_entry *find_object(unsigned char *sha1)
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;
574 static struct object_entry *insert_object(unsigned char *sha1)
576 unsigned int h = sha1[0] << 8 | sha1[1];
577 struct object_entry *e = object_table[h];
579 while (e) {
580 if (!hashcmp(sha1, e->idx.sha1))
581 return e;
582 e = e->next;
585 e = new_object(sha1);
586 e->next = object_table[h];
587 e->idx.offset = 0;
588 object_table[h] = e;
589 return e;
592 static unsigned int hc_str(const char *s, size_t len)
594 unsigned int r = 0;
595 while (len-- > 0)
596 r = r * 31 + *s++;
597 return r;
600 static void *pool_alloc(size_t len)
602 struct mem_pool *p;
603 void *r;
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));
609 for (p = mem_pool; p; p = p->next_pool)
610 if ((p->end - p->next_free >= len))
611 break;
613 if (!p) {
614 if (len >= (mem_pool_alloc/2)) {
615 total_allocd += len;
616 return xmalloc(len);
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;
626 r = p->next_free;
627 p->next_free += len;
628 return r;
631 static void *pool_calloc(size_t count, size_t size)
633 size_t len = count * size;
634 void *r = pool_alloc(len);
635 memset(r, 0, len);
636 return r;
639 static char *pool_strdup(const char *s)
641 char *r = pool_alloc(strlen(s) + 1);
642 strcpy(r, s);
643 return r;
646 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
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;
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;
662 s = s->data.sets[i];
664 if (!s->data.marked[idnum])
665 marks_set_count++;
666 s->data.marked[idnum] = oe;
669 static struct object_entry *find_mark(uintmax_t idnum)
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];
680 if (s)
681 oe = s->data.marked[idnum];
683 if (!oe)
684 die("mark :%" PRIuMAX " not declared", orig_idnum);
685 return oe;
688 static struct atom_str *to_atom(const char *s, unsigned short len)
690 unsigned int hc = hc_str(s, len) % atom_table_sz;
691 struct atom_str *c;
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;
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;
707 static struct branch *lookup_branch(const char *name)
709 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
710 struct branch *b;
712 for (b = branch_table[hc]; b; b = b->table_next_branch)
713 if (!strcmp(name, b->name))
714 return b;
715 return NULL;
718 static struct branch *new_branch(const char *name)
720 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
721 struct branch *b = lookup_branch(name);
723 if (b)
724 die("Invalid attempt to create duplicate branch: %s", name);
725 switch (check_ref_format(name)) {
726 case 0: break; /* its valid */
727 case CHECK_REF_FORMAT_ONELEVEL:
728 break; /* valid, but too few '/', allow anyway */
729 default:
730 die("Branch name doesn't conform to GIT standards: %s", name);
733 b = pool_calloc(1, sizeof(struct branch));
734 b->name = pool_strdup(name);
735 b->table_next_branch = branch_table[hc];
736 b->branch_tree.versions[0].mode = S_IFDIR;
737 b->branch_tree.versions[1].mode = S_IFDIR;
738 b->num_notes = 0;
739 b->active = 0;
740 b->pack_id = MAX_PACK_ID;
741 branch_table[hc] = b;
742 branch_count++;
743 return b;
746 static unsigned int hc_entries(unsigned int cnt)
748 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
749 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
752 static struct tree_content *new_tree_content(unsigned int cnt)
754 struct avail_tree_content *f, *l = NULL;
755 struct tree_content *t;
756 unsigned int hc = hc_entries(cnt);
758 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
759 if (f->entry_capacity >= cnt)
760 break;
762 if (f) {
763 if (l)
764 l->next_avail = f->next_avail;
765 else
766 avail_tree_table[hc] = f->next_avail;
767 } else {
768 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
769 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
770 f->entry_capacity = cnt;
773 t = (struct tree_content*)f;
774 t->entry_count = 0;
775 t->delta_depth = 0;
776 return t;
779 static void release_tree_entry(struct tree_entry *e);
780 static void release_tree_content(struct tree_content *t)
782 struct avail_tree_content *f = (struct avail_tree_content*)t;
783 unsigned int hc = hc_entries(f->entry_capacity);
784 f->next_avail = avail_tree_table[hc];
785 avail_tree_table[hc] = f;
788 static void release_tree_content_recursive(struct tree_content *t)
790 unsigned int i;
791 for (i = 0; i < t->entry_count; i++)
792 release_tree_entry(t->entries[i]);
793 release_tree_content(t);
796 static struct tree_content *grow_tree_content(
797 struct tree_content *t,
798 int amt)
800 struct tree_content *r = new_tree_content(t->entry_count + amt);
801 r->entry_count = t->entry_count;
802 r->delta_depth = t->delta_depth;
803 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
804 release_tree_content(t);
805 return r;
808 static struct tree_entry *new_tree_entry(void)
810 struct tree_entry *e;
812 if (!avail_tree_entry) {
813 unsigned int n = tree_entry_alloc;
814 total_allocd += n * sizeof(struct tree_entry);
815 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
816 while (n-- > 1) {
817 *((void**)e) = e + 1;
818 e++;
820 *((void**)e) = NULL;
823 e = avail_tree_entry;
824 avail_tree_entry = *((void**)e);
825 return e;
828 static void release_tree_entry(struct tree_entry *e)
830 if (e->tree)
831 release_tree_content_recursive(e->tree);
832 *((void**)e) = avail_tree_entry;
833 avail_tree_entry = e;
836 static struct tree_content *dup_tree_content(struct tree_content *s)
838 struct tree_content *d;
839 struct tree_entry *a, *b;
840 unsigned int i;
842 if (!s)
843 return NULL;
844 d = new_tree_content(s->entry_count);
845 for (i = 0; i < s->entry_count; i++) {
846 a = s->entries[i];
847 b = new_tree_entry();
848 memcpy(b, a, sizeof(*a));
849 if (a->tree && is_null_sha1(b->versions[1].sha1))
850 b->tree = dup_tree_content(a->tree);
851 else
852 b->tree = NULL;
853 d->entries[i] = b;
855 d->entry_count = s->entry_count;
856 d->delta_depth = s->delta_depth;
858 return d;
861 static void start_packfile(void)
863 static char tmpfile[PATH_MAX];
864 struct packed_git *p;
865 struct pack_header hdr;
866 int pack_fd;
868 pack_fd = odb_mkstemp(tmpfile, sizeof(tmpfile),
869 "pack/tmp_pack_XXXXXX");
870 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
871 strcpy(p->pack_name, tmpfile);
872 p->pack_fd = pack_fd;
873 p->do_not_close = 1;
874 pack_file = sha1fd(pack_fd, p->pack_name);
876 hdr.hdr_signature = htonl(PACK_SIGNATURE);
877 hdr.hdr_version = htonl(2);
878 hdr.hdr_entries = 0;
879 sha1write(pack_file, &hdr, sizeof(hdr));
881 pack_data = p;
882 pack_size = sizeof(hdr);
883 object_count = 0;
885 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
886 all_packs[pack_id] = p;
889 static const char *create_index(void)
891 const char *tmpfile;
892 struct pack_idx_entry **idx, **c, **last;
893 struct object_entry *e;
894 struct object_entry_pool *o;
896 /* Build the table of object IDs. */
897 idx = xmalloc(object_count * sizeof(*idx));
898 c = idx;
899 for (o = blocks; o; o = o->next_pool)
900 for (e = o->next_free; e-- != o->entries;)
901 if (pack_id == e->pack_id)
902 *c++ = &e->idx;
903 last = idx + object_count;
904 if (c != last)
905 die("internal consistency error creating the index");
907 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
908 free(idx);
909 return tmpfile;
912 static char *keep_pack(const char *curr_index_name)
914 static char name[PATH_MAX];
915 static const char *keep_msg = "fast-import";
916 int keep_fd;
918 keep_fd = odb_pack_keep(name, sizeof(name), pack_data->sha1);
919 if (keep_fd < 0)
920 die_errno("cannot create keep file");
921 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
922 if (close(keep_fd))
923 die_errno("failed to write keep file");
925 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
926 get_object_directory(), sha1_to_hex(pack_data->sha1));
927 if (move_temp_to_file(pack_data->pack_name, name))
928 die("cannot store pack file");
930 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
931 get_object_directory(), sha1_to_hex(pack_data->sha1));
932 if (move_temp_to_file(curr_index_name, name))
933 die("cannot store index file");
934 free((void *)curr_index_name);
935 return name;
938 static void unkeep_all_packs(void)
940 static char name[PATH_MAX];
941 int k;
943 for (k = 0; k < pack_id; k++) {
944 struct packed_git *p = all_packs[k];
945 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
946 get_object_directory(), sha1_to_hex(p->sha1));
947 unlink_or_warn(name);
951 static void end_packfile(void)
953 struct packed_git *old_p = pack_data, *new_p;
955 clear_delta_base_cache();
956 if (object_count) {
957 unsigned char cur_pack_sha1[20];
958 char *idx_name;
959 int i;
960 struct branch *b;
961 struct tag *t;
963 close_pack_windows(pack_data);
964 sha1close(pack_file, cur_pack_sha1, 0);
965 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
966 pack_data->pack_name, object_count,
967 cur_pack_sha1, pack_size);
968 close(pack_data->pack_fd);
969 idx_name = keep_pack(create_index());
971 /* Register the packfile with core git's machinery. */
972 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
973 if (!new_p)
974 die("core git rejected index %s", idx_name);
975 all_packs[pack_id] = new_p;
976 install_packed_git(new_p);
978 /* Print the boundary */
979 if (pack_edges) {
980 fprintf(pack_edges, "%s:", new_p->pack_name);
981 for (i = 0; i < branch_table_sz; i++) {
982 for (b = branch_table[i]; b; b = b->table_next_branch) {
983 if (b->pack_id == pack_id)
984 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
987 for (t = first_tag; t; t = t->next_tag) {
988 if (t->pack_id == pack_id)
989 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
991 fputc('\n', pack_edges);
992 fflush(pack_edges);
995 pack_id++;
997 else {
998 close(old_p->pack_fd);
999 unlink_or_warn(old_p->pack_name);
1001 free(old_p);
1003 /* We can't carry a delta across packfiles. */
1004 strbuf_release(&last_blob.data);
1005 last_blob.offset = 0;
1006 last_blob.depth = 0;
1009 static void cycle_packfile(void)
1011 end_packfile();
1012 start_packfile();
1015 static int store_object(
1016 enum object_type type,
1017 struct strbuf *dat,
1018 struct last_object *last,
1019 unsigned char *sha1out,
1020 uintmax_t mark)
1022 void *out, *delta;
1023 struct object_entry *e;
1024 unsigned char hdr[96];
1025 unsigned char sha1[20];
1026 unsigned long hdrlen, deltalen;
1027 git_SHA_CTX c;
1028 git_zstream s;
1030 hdrlen = sprintf((char *)hdr,"%s %lu", typename(type),
1031 (unsigned long)dat->len) + 1;
1032 git_SHA1_Init(&c);
1033 git_SHA1_Update(&c, hdr, hdrlen);
1034 git_SHA1_Update(&c, dat->buf, dat->len);
1035 git_SHA1_Final(sha1, &c);
1036 if (sha1out)
1037 hashcpy(sha1out, sha1);
1039 e = insert_object(sha1);
1040 if (mark)
1041 insert_mark(mark, e);
1042 if (e->idx.offset) {
1043 duplicate_count_by_type[type]++;
1044 return 1;
1045 } else if (find_sha1_pack(sha1, packed_git)) {
1046 e->type = type;
1047 e->pack_id = MAX_PACK_ID;
1048 e->idx.offset = 1; /* just not zero! */
1049 duplicate_count_by_type[type]++;
1050 return 1;
1053 if (last && last->data.buf && last->depth < max_depth && dat->len > 20) {
1054 delta_count_attempts_by_type[type]++;
1055 delta = diff_delta(last->data.buf, last->data.len,
1056 dat->buf, dat->len,
1057 &deltalen, dat->len - 20);
1058 } else
1059 delta = NULL;
1061 memset(&s, 0, sizeof(s));
1062 git_deflate_init(&s, pack_compression_level);
1063 if (delta) {
1064 s.next_in = delta;
1065 s.avail_in = deltalen;
1066 } else {
1067 s.next_in = (void *)dat->buf;
1068 s.avail_in = dat->len;
1070 s.avail_out = git_deflate_bound(&s, s.avail_in);
1071 s.next_out = out = xmalloc(s.avail_out);
1072 while (git_deflate(&s, Z_FINISH) == Z_OK)
1073 ; /* nothing */
1074 git_deflate_end(&s);
1076 /* Determine if we should auto-checkpoint. */
1077 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1078 || (pack_size + 60 + s.total_out) < pack_size) {
1080 /* This new object needs to *not* have the current pack_id. */
1081 e->pack_id = pack_id + 1;
1082 cycle_packfile();
1084 /* We cannot carry a delta into the new pack. */
1085 if (delta) {
1086 free(delta);
1087 delta = NULL;
1089 memset(&s, 0, sizeof(s));
1090 git_deflate_init(&s, pack_compression_level);
1091 s.next_in = (void *)dat->buf;
1092 s.avail_in = dat->len;
1093 s.avail_out = git_deflate_bound(&s, s.avail_in);
1094 s.next_out = out = xrealloc(out, s.avail_out);
1095 while (git_deflate(&s, Z_FINISH) == Z_OK)
1096 ; /* nothing */
1097 git_deflate_end(&s);
1101 e->type = type;
1102 e->pack_id = pack_id;
1103 e->idx.offset = pack_size;
1104 object_count++;
1105 object_count_by_type[type]++;
1107 crc32_begin(pack_file);
1109 if (delta) {
1110 off_t ofs = e->idx.offset - last->offset;
1111 unsigned pos = sizeof(hdr) - 1;
1113 delta_count_by_type[type]++;
1114 e->depth = last->depth + 1;
1116 hdrlen = encode_in_pack_object_header(OBJ_OFS_DELTA, deltalen, hdr);
1117 sha1write(pack_file, hdr, hdrlen);
1118 pack_size += hdrlen;
1120 hdr[pos] = ofs & 127;
1121 while (ofs >>= 7)
1122 hdr[--pos] = 128 | (--ofs & 127);
1123 sha1write(pack_file, hdr + pos, sizeof(hdr) - pos);
1124 pack_size += sizeof(hdr) - pos;
1125 } else {
1126 e->depth = 0;
1127 hdrlen = encode_in_pack_object_header(type, dat->len, hdr);
1128 sha1write(pack_file, hdr, hdrlen);
1129 pack_size += hdrlen;
1132 sha1write(pack_file, out, s.total_out);
1133 pack_size += s.total_out;
1135 e->idx.crc32 = crc32_end(pack_file);
1137 free(out);
1138 free(delta);
1139 if (last) {
1140 if (last->no_swap) {
1141 last->data = *dat;
1142 } else {
1143 strbuf_swap(&last->data, dat);
1145 last->offset = e->idx.offset;
1146 last->depth = e->depth;
1148 return 0;
1151 static void truncate_pack(off_t to, git_SHA_CTX *ctx)
1153 if (ftruncate(pack_data->pack_fd, to)
1154 || lseek(pack_data->pack_fd, to, SEEK_SET) != to)
1155 die_errno("cannot truncate pack to skip duplicate");
1156 pack_size = to;
1158 /* yes this is a layering violation */
1159 pack_file->total = to;
1160 pack_file->offset = 0;
1161 pack_file->ctx = *ctx;
1164 static void stream_blob(uintmax_t len, unsigned char *sha1out, uintmax_t mark)
1166 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1167 unsigned char *in_buf = xmalloc(in_sz);
1168 unsigned char *out_buf = xmalloc(out_sz);
1169 struct object_entry *e;
1170 unsigned char sha1[20];
1171 unsigned long hdrlen;
1172 off_t offset;
1173 git_SHA_CTX c;
1174 git_SHA_CTX pack_file_ctx;
1175 git_zstream s;
1176 int status = Z_OK;
1178 /* Determine if we should auto-checkpoint. */
1179 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1180 || (pack_size + 60 + len) < pack_size)
1181 cycle_packfile();
1183 offset = pack_size;
1185 /* preserve the pack_file SHA1 ctx in case we have to truncate later */
1186 sha1flush(pack_file);
1187 pack_file_ctx = pack_file->ctx;
1189 hdrlen = snprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1190 if (out_sz <= hdrlen)
1191 die("impossibly large object header");
1193 git_SHA1_Init(&c);
1194 git_SHA1_Update(&c, out_buf, hdrlen);
1196 crc32_begin(pack_file);
1198 memset(&s, 0, sizeof(s));
1199 git_deflate_init(&s, pack_compression_level);
1201 hdrlen = encode_in_pack_object_header(OBJ_BLOB, len, out_buf);
1202 if (out_sz <= hdrlen)
1203 die("impossibly large object header");
1205 s.next_out = out_buf + hdrlen;
1206 s.avail_out = out_sz - hdrlen;
1208 while (status != Z_STREAM_END) {
1209 if (0 < len && !s.avail_in) {
1210 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1211 size_t n = fread(in_buf, 1, cnt, stdin);
1212 if (!n && feof(stdin))
1213 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1215 git_SHA1_Update(&c, in_buf, n);
1216 s.next_in = in_buf;
1217 s.avail_in = n;
1218 len -= n;
1221 status = git_deflate(&s, len ? 0 : Z_FINISH);
1223 if (!s.avail_out || status == Z_STREAM_END) {
1224 size_t n = s.next_out - out_buf;
1225 sha1write(pack_file, out_buf, n);
1226 pack_size += n;
1227 s.next_out = out_buf;
1228 s.avail_out = out_sz;
1231 switch (status) {
1232 case Z_OK:
1233 case Z_BUF_ERROR:
1234 case Z_STREAM_END:
1235 continue;
1236 default:
1237 die("unexpected deflate failure: %d", status);
1240 git_deflate_end(&s);
1241 git_SHA1_Final(sha1, &c);
1243 if (sha1out)
1244 hashcpy(sha1out, sha1);
1246 e = insert_object(sha1);
1248 if (mark)
1249 insert_mark(mark, e);
1251 if (e->idx.offset) {
1252 duplicate_count_by_type[OBJ_BLOB]++;
1253 truncate_pack(offset, &pack_file_ctx);
1255 } else if (find_sha1_pack(sha1, packed_git)) {
1256 e->type = OBJ_BLOB;
1257 e->pack_id = MAX_PACK_ID;
1258 e->idx.offset = 1; /* just not zero! */
1259 duplicate_count_by_type[OBJ_BLOB]++;
1260 truncate_pack(offset, &pack_file_ctx);
1262 } else {
1263 e->depth = 0;
1264 e->type = OBJ_BLOB;
1265 e->pack_id = pack_id;
1266 e->idx.offset = offset;
1267 e->idx.crc32 = crc32_end(pack_file);
1268 object_count++;
1269 object_count_by_type[OBJ_BLOB]++;
1272 free(in_buf);
1273 free(out_buf);
1276 /* All calls must be guarded by find_object() or find_mark() to
1277 * ensure the 'struct object_entry' passed was written by this
1278 * process instance. We unpack the entry by the offset, avoiding
1279 * the need for the corresponding .idx file. This unpacking rule
1280 * works because we only use OBJ_REF_DELTA within the packfiles
1281 * created by fast-import.
1283 * oe must not be NULL. Such an oe usually comes from giving
1284 * an unknown SHA-1 to find_object() or an undefined mark to
1285 * find_mark(). Callers must test for this condition and use
1286 * the standard read_sha1_file() when it happens.
1288 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1289 * find_mark(), where the mark was reloaded from an existing marks
1290 * file and is referencing an object that this fast-import process
1291 * instance did not write out to a packfile. Callers must test for
1292 * this condition and use read_sha1_file() instead.
1294 static void *gfi_unpack_entry(
1295 struct object_entry *oe,
1296 unsigned long *sizep)
1298 enum object_type type;
1299 struct packed_git *p = all_packs[oe->pack_id];
1300 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1301 /* The object is stored in the packfile we are writing to
1302 * and we have modified it since the last time we scanned
1303 * back to read a previously written object. If an old
1304 * window covered [p->pack_size, p->pack_size + 20) its
1305 * data is stale and is not valid. Closing all windows
1306 * and updating the packfile length ensures we can read
1307 * the newly written data.
1309 close_pack_windows(p);
1310 sha1flush(pack_file);
1312 /* We have to offer 20 bytes additional on the end of
1313 * the packfile as the core unpacker code assumes the
1314 * footer is present at the file end and must promise
1315 * at least 20 bytes within any window it maps. But
1316 * we don't actually create the footer here.
1318 p->pack_size = pack_size + 20;
1320 return unpack_entry(p, oe->idx.offset, &type, sizep);
1323 static const char *get_mode(const char *str, uint16_t *modep)
1325 unsigned char c;
1326 uint16_t mode = 0;
1328 while ((c = *str++) != ' ') {
1329 if (c < '0' || c > '7')
1330 return NULL;
1331 mode = (mode << 3) + (c - '0');
1333 *modep = mode;
1334 return str;
1337 static void load_tree(struct tree_entry *root)
1339 unsigned char *sha1 = root->versions[1].sha1;
1340 struct object_entry *myoe;
1341 struct tree_content *t;
1342 unsigned long size;
1343 char *buf;
1344 const char *c;
1346 root->tree = t = new_tree_content(8);
1347 if (is_null_sha1(sha1))
1348 return;
1350 myoe = find_object(sha1);
1351 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1352 if (myoe->type != OBJ_TREE)
1353 die("Not a tree: %s", sha1_to_hex(sha1));
1354 t->delta_depth = myoe->depth;
1355 buf = gfi_unpack_entry(myoe, &size);
1356 if (!buf)
1357 die("Can't load tree %s", sha1_to_hex(sha1));
1358 } else {
1359 enum object_type type;
1360 buf = read_sha1_file(sha1, &type, &size);
1361 if (!buf || type != OBJ_TREE)
1362 die("Can't load tree %s", sha1_to_hex(sha1));
1365 c = buf;
1366 while (c != (buf + size)) {
1367 struct tree_entry *e = new_tree_entry();
1369 if (t->entry_count == t->entry_capacity)
1370 root->tree = t = grow_tree_content(t, t->entry_count);
1371 t->entries[t->entry_count++] = e;
1373 e->tree = NULL;
1374 c = get_mode(c, &e->versions[1].mode);
1375 if (!c)
1376 die("Corrupt mode in %s", sha1_to_hex(sha1));
1377 e->versions[0].mode = e->versions[1].mode;
1378 e->name = to_atom(c, strlen(c));
1379 c += e->name->str_len + 1;
1380 hashcpy(e->versions[0].sha1, (unsigned char *)c);
1381 hashcpy(e->versions[1].sha1, (unsigned char *)c);
1382 c += 20;
1384 free(buf);
1387 static int tecmp0 (const void *_a, const void *_b)
1389 struct tree_entry *a = *((struct tree_entry**)_a);
1390 struct tree_entry *b = *((struct tree_entry**)_b);
1391 return base_name_compare(
1392 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1393 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1396 static int tecmp1 (const void *_a, const void *_b)
1398 struct tree_entry *a = *((struct tree_entry**)_a);
1399 struct tree_entry *b = *((struct tree_entry**)_b);
1400 return base_name_compare(
1401 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1402 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1405 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1407 size_t maxlen = 0;
1408 unsigned int i;
1410 if (!v)
1411 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1412 else
1413 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1415 for (i = 0; i < t->entry_count; i++) {
1416 if (t->entries[i]->versions[v].mode)
1417 maxlen += t->entries[i]->name->str_len + 34;
1420 strbuf_reset(b);
1421 strbuf_grow(b, maxlen);
1422 for (i = 0; i < t->entry_count; i++) {
1423 struct tree_entry *e = t->entries[i];
1424 if (!e->versions[v].mode)
1425 continue;
1426 strbuf_addf(b, "%o %s%c",
1427 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1428 e->name->str_dat, '\0');
1429 strbuf_add(b, e->versions[v].sha1, 20);
1433 static void store_tree(struct tree_entry *root)
1435 struct tree_content *t = root->tree;
1436 unsigned int i, j, del;
1437 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1438 struct object_entry *le = NULL;
1440 if (!is_null_sha1(root->versions[1].sha1))
1441 return;
1443 for (i = 0; i < t->entry_count; i++) {
1444 if (t->entries[i]->tree)
1445 store_tree(t->entries[i]);
1448 if (!(root->versions[0].mode & NO_DELTA))
1449 le = find_object(root->versions[0].sha1);
1450 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1451 mktree(t, 0, &old_tree);
1452 lo.data = old_tree;
1453 lo.offset = le->idx.offset;
1454 lo.depth = t->delta_depth;
1457 mktree(t, 1, &new_tree);
1458 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1460 t->delta_depth = lo.depth;
1461 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1462 struct tree_entry *e = t->entries[i];
1463 if (e->versions[1].mode) {
1464 e->versions[0].mode = e->versions[1].mode;
1465 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1466 t->entries[j++] = e;
1467 } else {
1468 release_tree_entry(e);
1469 del++;
1472 t->entry_count -= del;
1475 static void tree_content_replace(
1476 struct tree_entry *root,
1477 const unsigned char *sha1,
1478 const uint16_t mode,
1479 struct tree_content *newtree)
1481 if (!S_ISDIR(mode))
1482 die("Root cannot be a non-directory");
1483 hashclr(root->versions[0].sha1);
1484 hashcpy(root->versions[1].sha1, sha1);
1485 if (root->tree)
1486 release_tree_content_recursive(root->tree);
1487 root->tree = newtree;
1490 static int tree_content_set(
1491 struct tree_entry *root,
1492 const char *p,
1493 const unsigned char *sha1,
1494 const uint16_t mode,
1495 struct tree_content *subtree)
1497 struct tree_content *t;
1498 const char *slash1;
1499 unsigned int i, n;
1500 struct tree_entry *e;
1502 slash1 = strchr(p, '/');
1503 if (slash1)
1504 n = slash1 - p;
1505 else
1506 n = strlen(p);
1507 if (!n)
1508 die("Empty path component found in input");
1509 if (!slash1 && !S_ISDIR(mode) && subtree)
1510 die("Non-directories cannot have subtrees");
1512 if (!root->tree)
1513 load_tree(root);
1514 t = root->tree;
1515 for (i = 0; i < t->entry_count; i++) {
1516 e = t->entries[i];
1517 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1518 if (!slash1) {
1519 if (!S_ISDIR(mode)
1520 && e->versions[1].mode == mode
1521 && !hashcmp(e->versions[1].sha1, sha1))
1522 return 0;
1523 e->versions[1].mode = mode;
1524 hashcpy(e->versions[1].sha1, sha1);
1525 if (e->tree)
1526 release_tree_content_recursive(e->tree);
1527 e->tree = subtree;
1530 * We need to leave e->versions[0].sha1 alone
1531 * to avoid modifying the preimage tree used
1532 * when writing out the parent directory.
1533 * But after replacing the subdir with a
1534 * completely different one, it's not a good
1535 * delta base any more, and besides, we've
1536 * thrown away the tree entries needed to
1537 * make a delta against it.
1539 * So let's just explicitly disable deltas
1540 * for the subtree.
1542 if (S_ISDIR(e->versions[0].mode))
1543 e->versions[0].mode |= NO_DELTA;
1545 hashclr(root->versions[1].sha1);
1546 return 1;
1548 if (!S_ISDIR(e->versions[1].mode)) {
1549 e->tree = new_tree_content(8);
1550 e->versions[1].mode = S_IFDIR;
1552 if (!e->tree)
1553 load_tree(e);
1554 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1555 hashclr(root->versions[1].sha1);
1556 return 1;
1558 return 0;
1562 if (t->entry_count == t->entry_capacity)
1563 root->tree = t = grow_tree_content(t, t->entry_count);
1564 e = new_tree_entry();
1565 e->name = to_atom(p, n);
1566 e->versions[0].mode = 0;
1567 hashclr(e->versions[0].sha1);
1568 t->entries[t->entry_count++] = e;
1569 if (slash1) {
1570 e->tree = new_tree_content(8);
1571 e->versions[1].mode = S_IFDIR;
1572 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1573 } else {
1574 e->tree = subtree;
1575 e->versions[1].mode = mode;
1576 hashcpy(e->versions[1].sha1, sha1);
1578 hashclr(root->versions[1].sha1);
1579 return 1;
1582 static int tree_content_remove(
1583 struct tree_entry *root,
1584 const char *p,
1585 struct tree_entry *backup_leaf)
1587 struct tree_content *t;
1588 const char *slash1;
1589 unsigned int i, n;
1590 struct tree_entry *e;
1592 slash1 = strchr(p, '/');
1593 if (slash1)
1594 n = slash1 - p;
1595 else
1596 n = strlen(p);
1598 if (!root->tree)
1599 load_tree(root);
1600 t = root->tree;
1601 for (i = 0; i < t->entry_count; i++) {
1602 e = t->entries[i];
1603 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1604 if (slash1 && !S_ISDIR(e->versions[1].mode))
1606 * If p names a file in some subdirectory, and a
1607 * file or symlink matching the name of the
1608 * parent directory of p exists, then p cannot
1609 * exist and need not be deleted.
1611 return 1;
1612 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1613 goto del_entry;
1614 if (!e->tree)
1615 load_tree(e);
1616 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1617 for (n = 0; n < e->tree->entry_count; n++) {
1618 if (e->tree->entries[n]->versions[1].mode) {
1619 hashclr(root->versions[1].sha1);
1620 return 1;
1623 backup_leaf = NULL;
1624 goto del_entry;
1626 return 0;
1629 return 0;
1631 del_entry:
1632 if (backup_leaf)
1633 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1634 else if (e->tree)
1635 release_tree_content_recursive(e->tree);
1636 e->tree = NULL;
1637 e->versions[1].mode = 0;
1638 hashclr(e->versions[1].sha1);
1639 hashclr(root->versions[1].sha1);
1640 return 1;
1643 static int tree_content_get(
1644 struct tree_entry *root,
1645 const char *p,
1646 struct tree_entry *leaf)
1648 struct tree_content *t;
1649 const char *slash1;
1650 unsigned int i, n;
1651 struct tree_entry *e;
1653 slash1 = strchr(p, '/');
1654 if (slash1)
1655 n = slash1 - p;
1656 else
1657 n = strlen(p);
1659 if (!root->tree)
1660 load_tree(root);
1661 t = root->tree;
1662 for (i = 0; i < t->entry_count; i++) {
1663 e = t->entries[i];
1664 if (e->name->str_len == n && !strncmp_icase(p, e->name->str_dat, n)) {
1665 if (!slash1) {
1666 memcpy(leaf, e, sizeof(*leaf));
1667 if (e->tree && is_null_sha1(e->versions[1].sha1))
1668 leaf->tree = dup_tree_content(e->tree);
1669 else
1670 leaf->tree = NULL;
1671 return 1;
1673 if (!S_ISDIR(e->versions[1].mode))
1674 return 0;
1675 if (!e->tree)
1676 load_tree(e);
1677 return tree_content_get(e, slash1 + 1, leaf);
1680 return 0;
1683 static int update_branch(struct branch *b)
1685 static const char *msg = "fast-import";
1686 struct ref_lock *lock;
1687 unsigned char old_sha1[20];
1689 if (is_null_sha1(b->sha1))
1690 return 0;
1691 if (read_ref(b->name, old_sha1))
1692 hashclr(old_sha1);
1693 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1694 if (!lock)
1695 return error("Unable to lock %s", b->name);
1696 if (!force_update && !is_null_sha1(old_sha1)) {
1697 struct commit *old_cmit, *new_cmit;
1699 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1700 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1701 if (!old_cmit || !new_cmit) {
1702 unlock_ref(lock);
1703 return error("Branch %s is missing commits.", b->name);
1706 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1707 unlock_ref(lock);
1708 warning("Not updating %s"
1709 " (new tip %s does not contain %s)",
1710 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1711 return -1;
1714 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1715 return error("Unable to update %s", b->name);
1716 return 0;
1719 static void dump_branches(void)
1721 unsigned int i;
1722 struct branch *b;
1724 for (i = 0; i < branch_table_sz; i++) {
1725 for (b = branch_table[i]; b; b = b->table_next_branch)
1726 failure |= update_branch(b);
1730 static void dump_tags(void)
1732 static const char *msg = "fast-import";
1733 struct tag *t;
1734 struct ref_lock *lock;
1735 char ref_name[PATH_MAX];
1737 for (t = first_tag; t; t = t->next_tag) {
1738 sprintf(ref_name, "tags/%s", t->name);
1739 lock = lock_ref_sha1(ref_name, NULL);
1740 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1741 failure |= error("Unable to update %s", ref_name);
1745 static void dump_marks_helper(FILE *f,
1746 uintmax_t base,
1747 struct mark_set *m)
1749 uintmax_t k;
1750 if (m->shift) {
1751 for (k = 0; k < 1024; k++) {
1752 if (m->data.sets[k])
1753 dump_marks_helper(f, base + (k << m->shift),
1754 m->data.sets[k]);
1756 } else {
1757 for (k = 0; k < 1024; k++) {
1758 if (m->data.marked[k])
1759 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1760 sha1_to_hex(m->data.marked[k]->idx.sha1));
1765 static void dump_marks(void)
1767 static struct lock_file mark_lock;
1768 int mark_fd;
1769 FILE *f;
1771 if (!export_marks_file)
1772 return;
1774 mark_fd = hold_lock_file_for_update(&mark_lock, export_marks_file, 0);
1775 if (mark_fd < 0) {
1776 failure |= error("Unable to write marks file %s: %s",
1777 export_marks_file, strerror(errno));
1778 return;
1781 f = fdopen(mark_fd, "w");
1782 if (!f) {
1783 int saved_errno = errno;
1784 rollback_lock_file(&mark_lock);
1785 failure |= error("Unable to write marks file %s: %s",
1786 export_marks_file, strerror(saved_errno));
1787 return;
1791 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1792 * Assign -1 to the lock file descriptor so that commit_lock_file()
1793 * won't try to close() it.
1795 mark_lock.fd = -1;
1797 dump_marks_helper(f, 0, marks);
1798 if (ferror(f) || fclose(f)) {
1799 int saved_errno = errno;
1800 rollback_lock_file(&mark_lock);
1801 failure |= error("Unable to write marks file %s: %s",
1802 export_marks_file, strerror(saved_errno));
1803 return;
1806 if (commit_lock_file(&mark_lock)) {
1807 int saved_errno = errno;
1808 rollback_lock_file(&mark_lock);
1809 failure |= error("Unable to commit marks file %s: %s",
1810 export_marks_file, strerror(saved_errno));
1811 return;
1815 static void read_marks(void)
1817 char line[512];
1818 FILE *f = fopen(import_marks_file, "r");
1819 if (f)
1821 else if (import_marks_file_ignore_missing && errno == ENOENT)
1822 return; /* Marks file does not exist */
1823 else
1824 die_errno("cannot read '%s'", import_marks_file);
1825 while (fgets(line, sizeof(line), f)) {
1826 uintmax_t mark;
1827 char *end;
1828 unsigned char sha1[20];
1829 struct object_entry *e;
1831 end = strchr(line, '\n');
1832 if (line[0] != ':' || !end)
1833 die("corrupt mark line: %s", line);
1834 *end = 0;
1835 mark = strtoumax(line + 1, &end, 10);
1836 if (!mark || end == line + 1
1837 || *end != ' ' || get_sha1(end + 1, sha1))
1838 die("corrupt mark line: %s", line);
1839 e = find_object(sha1);
1840 if (!e) {
1841 enum object_type type = sha1_object_info(sha1, NULL);
1842 if (type < 0)
1843 die("object not found: %s", sha1_to_hex(sha1));
1844 e = insert_object(sha1);
1845 e->type = type;
1846 e->pack_id = MAX_PACK_ID;
1847 e->idx.offset = 1; /* just not zero! */
1849 insert_mark(mark, e);
1851 fclose(f);
1855 static int read_next_command(void)
1857 static int stdin_eof = 0;
1859 if (stdin_eof) {
1860 unread_command_buf = 0;
1861 return EOF;
1864 for (;;) {
1865 if (unread_command_buf) {
1866 unread_command_buf = 0;
1867 } else {
1868 struct recent_command *rc;
1870 strbuf_detach(&command_buf, NULL);
1871 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1872 if (stdin_eof)
1873 return EOF;
1875 if (!seen_data_command
1876 && prefixcmp(command_buf.buf, "feature ")
1877 && prefixcmp(command_buf.buf, "option ")) {
1878 parse_argv();
1881 rc = rc_free;
1882 if (rc)
1883 rc_free = rc->next;
1884 else {
1885 rc = cmd_hist.next;
1886 cmd_hist.next = rc->next;
1887 cmd_hist.next->prev = &cmd_hist;
1888 free(rc->buf);
1891 rc->buf = command_buf.buf;
1892 rc->prev = cmd_tail;
1893 rc->next = cmd_hist.prev;
1894 rc->prev->next = rc;
1895 cmd_tail = rc;
1897 if (!prefixcmp(command_buf.buf, "cat-blob ")) {
1898 parse_cat_blob();
1899 continue;
1901 if (command_buf.buf[0] == '#')
1902 continue;
1903 return 0;
1907 static void skip_optional_lf(void)
1909 int term_char = fgetc(stdin);
1910 if (term_char != '\n' && term_char != EOF)
1911 ungetc(term_char, stdin);
1914 static void parse_mark(void)
1916 if (!prefixcmp(command_buf.buf, "mark :")) {
1917 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1918 read_next_command();
1920 else
1921 next_mark = 0;
1924 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1926 strbuf_reset(sb);
1928 if (prefixcmp(command_buf.buf, "data "))
1929 die("Expected 'data n' command, found: %s", command_buf.buf);
1931 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1932 char *term = xstrdup(command_buf.buf + 5 + 2);
1933 size_t term_len = command_buf.len - 5 - 2;
1935 strbuf_detach(&command_buf, NULL);
1936 for (;;) {
1937 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1938 die("EOF in data (terminator '%s' not found)", term);
1939 if (term_len == command_buf.len
1940 && !strcmp(term, command_buf.buf))
1941 break;
1942 strbuf_addbuf(sb, &command_buf);
1943 strbuf_addch(sb, '\n');
1945 free(term);
1947 else {
1948 uintmax_t len = strtoumax(command_buf.buf + 5, NULL, 10);
1949 size_t n = 0, length = (size_t)len;
1951 if (limit && limit < len) {
1952 *len_res = len;
1953 return 0;
1955 if (length < len)
1956 die("data is too large to use in this context");
1958 while (n < length) {
1959 size_t s = strbuf_fread(sb, length - n, stdin);
1960 if (!s && feof(stdin))
1961 die("EOF in data (%lu bytes remaining)",
1962 (unsigned long)(length - n));
1963 n += s;
1967 skip_optional_lf();
1968 return 1;
1971 static int validate_raw_date(const char *src, char *result, int maxlen)
1973 const char *orig_src = src;
1974 char *endp;
1975 unsigned long num;
1977 errno = 0;
1979 num = strtoul(src, &endp, 10);
1980 /* NEEDSWORK: perhaps check for reasonable values? */
1981 if (errno || endp == src || *endp != ' ')
1982 return -1;
1984 src = endp + 1;
1985 if (*src != '-' && *src != '+')
1986 return -1;
1988 num = strtoul(src + 1, &endp, 10);
1989 if (errno || endp == src + 1 || *endp || (endp - orig_src) >= maxlen ||
1990 1400 < num)
1991 return -1;
1993 strcpy(result, orig_src);
1994 return 0;
1997 static char *parse_ident(const char *buf)
1999 const char *ltgt;
2000 size_t name_len;
2001 char *ident;
2003 /* ensure there is a space delimiter even if there is no name */
2004 if (*buf == '<')
2005 --buf;
2007 ltgt = buf + strcspn(buf, "<>");
2008 if (*ltgt != '<')
2009 die("Missing < in ident string: %s", buf);
2010 if (ltgt != buf && ltgt[-1] != ' ')
2011 die("Missing space before < in ident string: %s", buf);
2012 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2013 if (*ltgt != '>')
2014 die("Missing > in ident string: %s", buf);
2015 ltgt++;
2016 if (*ltgt != ' ')
2017 die("Missing space after > in ident string: %s", buf);
2018 ltgt++;
2019 name_len = ltgt - buf;
2020 ident = xmalloc(name_len + 24);
2021 strncpy(ident, buf, name_len);
2023 switch (whenspec) {
2024 case WHENSPEC_RAW:
2025 if (validate_raw_date(ltgt, ident + name_len, 24) < 0)
2026 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2027 break;
2028 case WHENSPEC_RFC2822:
2029 if (parse_date(ltgt, ident + name_len, 24) < 0)
2030 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2031 break;
2032 case WHENSPEC_NOW:
2033 if (strcmp("now", ltgt))
2034 die("Date in ident must be 'now': %s", buf);
2035 datestamp(ident + name_len, 24);
2036 break;
2039 return ident;
2042 static void parse_and_store_blob(
2043 struct last_object *last,
2044 unsigned char *sha1out,
2045 uintmax_t mark)
2047 static struct strbuf buf = STRBUF_INIT;
2048 uintmax_t len;
2050 if (parse_data(&buf, big_file_threshold, &len))
2051 store_object(OBJ_BLOB, &buf, last, sha1out, mark);
2052 else {
2053 if (last) {
2054 strbuf_release(&last->data);
2055 last->offset = 0;
2056 last->depth = 0;
2058 stream_blob(len, sha1out, mark);
2059 skip_optional_lf();
2063 static void parse_new_blob(void)
2065 read_next_command();
2066 parse_mark();
2067 parse_and_store_blob(&last_blob, NULL, next_mark);
2070 static void unload_one_branch(void)
2072 while (cur_active_branches
2073 && cur_active_branches >= max_active_branches) {
2074 uintmax_t min_commit = ULONG_MAX;
2075 struct branch *e, *l = NULL, *p = NULL;
2077 for (e = active_branches; e; e = e->active_next_branch) {
2078 if (e->last_commit < min_commit) {
2079 p = l;
2080 min_commit = e->last_commit;
2082 l = e;
2085 if (p) {
2086 e = p->active_next_branch;
2087 p->active_next_branch = e->active_next_branch;
2088 } else {
2089 e = active_branches;
2090 active_branches = e->active_next_branch;
2092 e->active = 0;
2093 e->active_next_branch = NULL;
2094 if (e->branch_tree.tree) {
2095 release_tree_content_recursive(e->branch_tree.tree);
2096 e->branch_tree.tree = NULL;
2098 cur_active_branches--;
2102 static void load_branch(struct branch *b)
2104 load_tree(&b->branch_tree);
2105 if (!b->active) {
2106 b->active = 1;
2107 b->active_next_branch = active_branches;
2108 active_branches = b;
2109 cur_active_branches++;
2110 branch_load_count++;
2114 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2116 unsigned char fanout = 0;
2117 while ((num_notes >>= 8))
2118 fanout++;
2119 return fanout;
2122 static void construct_path_with_fanout(const char *hex_sha1,
2123 unsigned char fanout, char *path)
2125 unsigned int i = 0, j = 0;
2126 if (fanout >= 20)
2127 die("Too large fanout (%u)", fanout);
2128 while (fanout) {
2129 path[i++] = hex_sha1[j++];
2130 path[i++] = hex_sha1[j++];
2131 path[i++] = '/';
2132 fanout--;
2134 memcpy(path + i, hex_sha1 + j, 40 - j);
2135 path[i + 40 - j] = '\0';
2138 static uintmax_t do_change_note_fanout(
2139 struct tree_entry *orig_root, struct tree_entry *root,
2140 char *hex_sha1, unsigned int hex_sha1_len,
2141 char *fullpath, unsigned int fullpath_len,
2142 unsigned char fanout)
2144 struct tree_content *t = root->tree;
2145 struct tree_entry *e, leaf;
2146 unsigned int i, tmp_hex_sha1_len, tmp_fullpath_len;
2147 uintmax_t num_notes = 0;
2148 unsigned char sha1[20];
2149 char realpath[60];
2151 for (i = 0; t && i < t->entry_count; i++) {
2152 e = t->entries[i];
2153 tmp_hex_sha1_len = hex_sha1_len + e->name->str_len;
2154 tmp_fullpath_len = fullpath_len;
2157 * We're interested in EITHER existing note entries (entries
2158 * with exactly 40 hex chars in path, not including directory
2159 * separators), OR directory entries that may contain note
2160 * entries (with < 40 hex chars in path).
2161 * Also, each path component in a note entry must be a multiple
2162 * of 2 chars.
2164 if (!e->versions[1].mode ||
2165 tmp_hex_sha1_len > 40 ||
2166 e->name->str_len % 2)
2167 continue;
2169 /* This _may_ be a note entry, or a subdir containing notes */
2170 memcpy(hex_sha1 + hex_sha1_len, e->name->str_dat,
2171 e->name->str_len);
2172 if (tmp_fullpath_len)
2173 fullpath[tmp_fullpath_len++] = '/';
2174 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2175 e->name->str_len);
2176 tmp_fullpath_len += e->name->str_len;
2177 fullpath[tmp_fullpath_len] = '\0';
2179 if (tmp_hex_sha1_len == 40 && !get_sha1_hex(hex_sha1, sha1)) {
2180 /* This is a note entry */
2181 construct_path_with_fanout(hex_sha1, fanout, realpath);
2182 if (!strcmp(fullpath, realpath)) {
2183 /* Note entry is in correct location */
2184 num_notes++;
2185 continue;
2188 /* Rename fullpath to realpath */
2189 if (!tree_content_remove(orig_root, fullpath, &leaf))
2190 die("Failed to remove path %s", fullpath);
2191 tree_content_set(orig_root, realpath,
2192 leaf.versions[1].sha1,
2193 leaf.versions[1].mode,
2194 leaf.tree);
2195 } else if (S_ISDIR(e->versions[1].mode)) {
2196 /* This is a subdir that may contain note entries */
2197 if (!e->tree)
2198 load_tree(e);
2199 num_notes += do_change_note_fanout(orig_root, e,
2200 hex_sha1, tmp_hex_sha1_len,
2201 fullpath, tmp_fullpath_len, fanout);
2204 /* The above may have reallocated the current tree_content */
2205 t = root->tree;
2207 return num_notes;
2210 static uintmax_t change_note_fanout(struct tree_entry *root,
2211 unsigned char fanout)
2213 char hex_sha1[40], path[60];
2214 return do_change_note_fanout(root, root, hex_sha1, 0, path, 0, fanout);
2217 static void file_change_m(struct branch *b)
2219 const char *p = command_buf.buf + 2;
2220 static struct strbuf uq = STRBUF_INIT;
2221 const char *endp;
2222 struct object_entry *oe = oe;
2223 unsigned char sha1[20];
2224 uint16_t mode, inline_data = 0;
2226 p = get_mode(p, &mode);
2227 if (!p)
2228 die("Corrupt mode: %s", command_buf.buf);
2229 switch (mode) {
2230 case 0644:
2231 case 0755:
2232 mode |= S_IFREG;
2233 case S_IFREG | 0644:
2234 case S_IFREG | 0755:
2235 case S_IFLNK:
2236 case S_IFDIR:
2237 case S_IFGITLINK:
2238 /* ok */
2239 break;
2240 default:
2241 die("Corrupt mode: %s", command_buf.buf);
2244 if (*p == ':') {
2245 char *x;
2246 oe = find_mark(strtoumax(p + 1, &x, 10));
2247 hashcpy(sha1, oe->idx.sha1);
2248 p = x;
2249 } else if (!prefixcmp(p, "inline")) {
2250 inline_data = 1;
2251 p += 6;
2252 } else {
2253 if (get_sha1_hex(p, sha1))
2254 die("Invalid SHA1: %s", command_buf.buf);
2255 oe = find_object(sha1);
2256 p += 40;
2258 if (*p++ != ' ')
2259 die("Missing space after SHA1: %s", command_buf.buf);
2261 strbuf_reset(&uq);
2262 if (!unquote_c_style(&uq, p, &endp)) {
2263 if (*endp)
2264 die("Garbage after path in: %s", command_buf.buf);
2265 p = uq.buf;
2268 /* Git does not track empty, non-toplevel directories. */
2269 if (S_ISDIR(mode) && !memcmp(sha1, EMPTY_TREE_SHA1_BIN, 20) && *p) {
2270 tree_content_remove(&b->branch_tree, p, NULL);
2271 return;
2274 if (S_ISGITLINK(mode)) {
2275 if (inline_data)
2276 die("Git links cannot be specified 'inline': %s",
2277 command_buf.buf);
2278 else if (oe) {
2279 if (oe->type != OBJ_COMMIT)
2280 die("Not a commit (actually a %s): %s",
2281 typename(oe->type), command_buf.buf);
2284 * Accept the sha1 without checking; it expected to be in
2285 * another repository.
2287 } else if (inline_data) {
2288 if (S_ISDIR(mode))
2289 die("Directories cannot be specified 'inline': %s",
2290 command_buf.buf);
2291 if (p != uq.buf) {
2292 strbuf_addstr(&uq, p);
2293 p = uq.buf;
2295 read_next_command();
2296 parse_and_store_blob(&last_blob, sha1, 0);
2297 } else {
2298 enum object_type expected = S_ISDIR(mode) ?
2299 OBJ_TREE: OBJ_BLOB;
2300 enum object_type type = oe ? oe->type :
2301 sha1_object_info(sha1, NULL);
2302 if (type < 0)
2303 die("%s not found: %s",
2304 S_ISDIR(mode) ? "Tree" : "Blob",
2305 command_buf.buf);
2306 if (type != expected)
2307 die("Not a %s (actually a %s): %s",
2308 typename(expected), typename(type),
2309 command_buf.buf);
2312 if (!*p) {
2313 tree_content_replace(&b->branch_tree, sha1, mode, NULL);
2314 return;
2316 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
2319 static void file_change_d(struct branch *b)
2321 const char *p = command_buf.buf + 2;
2322 static struct strbuf uq = STRBUF_INIT;
2323 const char *endp;
2325 strbuf_reset(&uq);
2326 if (!unquote_c_style(&uq, p, &endp)) {
2327 if (*endp)
2328 die("Garbage after path in: %s", command_buf.buf);
2329 p = uq.buf;
2331 tree_content_remove(&b->branch_tree, p, NULL);
2334 static void file_change_cr(struct branch *b, int rename)
2336 const char *s, *d;
2337 static struct strbuf s_uq = STRBUF_INIT;
2338 static struct strbuf d_uq = STRBUF_INIT;
2339 const char *endp;
2340 struct tree_entry leaf;
2342 s = command_buf.buf + 2;
2343 strbuf_reset(&s_uq);
2344 if (!unquote_c_style(&s_uq, s, &endp)) {
2345 if (*endp != ' ')
2346 die("Missing space after source: %s", command_buf.buf);
2347 } else {
2348 endp = strchr(s, ' ');
2349 if (!endp)
2350 die("Missing space after source: %s", command_buf.buf);
2351 strbuf_add(&s_uq, s, endp - s);
2353 s = s_uq.buf;
2355 endp++;
2356 if (!*endp)
2357 die("Missing dest: %s", command_buf.buf);
2359 d = endp;
2360 strbuf_reset(&d_uq);
2361 if (!unquote_c_style(&d_uq, d, &endp)) {
2362 if (*endp)
2363 die("Garbage after dest in: %s", command_buf.buf);
2364 d = d_uq.buf;
2367 memset(&leaf, 0, sizeof(leaf));
2368 if (rename)
2369 tree_content_remove(&b->branch_tree, s, &leaf);
2370 else
2371 tree_content_get(&b->branch_tree, s, &leaf);
2372 if (!leaf.versions[1].mode)
2373 die("Path %s not in branch", s);
2374 if (!*d) { /* C "path/to/subdir" "" */
2375 tree_content_replace(&b->branch_tree,
2376 leaf.versions[1].sha1,
2377 leaf.versions[1].mode,
2378 leaf.tree);
2379 return;
2381 tree_content_set(&b->branch_tree, d,
2382 leaf.versions[1].sha1,
2383 leaf.versions[1].mode,
2384 leaf.tree);
2387 static void note_change_n(struct branch *b, unsigned char old_fanout)
2389 const char *p = command_buf.buf + 2;
2390 static struct strbuf uq = STRBUF_INIT;
2391 struct object_entry *oe = oe;
2392 struct branch *s;
2393 unsigned char sha1[20], commit_sha1[20];
2394 char path[60];
2395 uint16_t inline_data = 0;
2396 unsigned char new_fanout;
2398 /* <dataref> or 'inline' */
2399 if (*p == ':') {
2400 char *x;
2401 oe = find_mark(strtoumax(p + 1, &x, 10));
2402 hashcpy(sha1, oe->idx.sha1);
2403 p = x;
2404 } else if (!prefixcmp(p, "inline")) {
2405 inline_data = 1;
2406 p += 6;
2407 } else {
2408 if (get_sha1_hex(p, sha1))
2409 die("Invalid SHA1: %s", command_buf.buf);
2410 oe = find_object(sha1);
2411 p += 40;
2413 if (*p++ != ' ')
2414 die("Missing space after SHA1: %s", command_buf.buf);
2416 /* <committish> */
2417 s = lookup_branch(p);
2418 if (s) {
2419 hashcpy(commit_sha1, s->sha1);
2420 } else if (*p == ':') {
2421 uintmax_t commit_mark = strtoumax(p + 1, NULL, 10);
2422 struct object_entry *commit_oe = find_mark(commit_mark);
2423 if (commit_oe->type != OBJ_COMMIT)
2424 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2425 hashcpy(commit_sha1, commit_oe->idx.sha1);
2426 } else if (!get_sha1(p, commit_sha1)) {
2427 unsigned long size;
2428 char *buf = read_object_with_reference(commit_sha1,
2429 commit_type, &size, commit_sha1);
2430 if (!buf || size < 46)
2431 die("Not a valid commit: %s", p);
2432 free(buf);
2433 } else
2434 die("Invalid ref name or SHA1 expression: %s", p);
2436 if (inline_data) {
2437 if (p != uq.buf) {
2438 strbuf_addstr(&uq, p);
2439 p = uq.buf;
2441 read_next_command();
2442 parse_and_store_blob(&last_blob, sha1, 0);
2443 } else if (oe) {
2444 if (oe->type != OBJ_BLOB)
2445 die("Not a blob (actually a %s): %s",
2446 typename(oe->type), command_buf.buf);
2447 } else if (!is_null_sha1(sha1)) {
2448 enum object_type type = sha1_object_info(sha1, NULL);
2449 if (type < 0)
2450 die("Blob not found: %s", command_buf.buf);
2451 if (type != OBJ_BLOB)
2452 die("Not a blob (actually a %s): %s",
2453 typename(type), command_buf.buf);
2456 construct_path_with_fanout(sha1_to_hex(commit_sha1), old_fanout, path);
2457 if (tree_content_remove(&b->branch_tree, path, NULL))
2458 b->num_notes--;
2460 if (is_null_sha1(sha1))
2461 return; /* nothing to insert */
2463 b->num_notes++;
2464 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2465 construct_path_with_fanout(sha1_to_hex(commit_sha1), new_fanout, path);
2466 tree_content_set(&b->branch_tree, path, sha1, S_IFREG | 0644, NULL);
2469 static void file_change_deleteall(struct branch *b)
2471 release_tree_content_recursive(b->branch_tree.tree);
2472 hashclr(b->branch_tree.versions[0].sha1);
2473 hashclr(b->branch_tree.versions[1].sha1);
2474 load_tree(&b->branch_tree);
2475 b->num_notes = 0;
2478 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2480 if (!buf || size < 46)
2481 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2482 if (memcmp("tree ", buf, 5)
2483 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2484 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2485 hashcpy(b->branch_tree.versions[0].sha1,
2486 b->branch_tree.versions[1].sha1);
2489 static void parse_from_existing(struct branch *b)
2491 if (is_null_sha1(b->sha1)) {
2492 hashclr(b->branch_tree.versions[0].sha1);
2493 hashclr(b->branch_tree.versions[1].sha1);
2494 } else {
2495 unsigned long size;
2496 char *buf;
2498 buf = read_object_with_reference(b->sha1,
2499 commit_type, &size, b->sha1);
2500 parse_from_commit(b, buf, size);
2501 free(buf);
2505 static int parse_from(struct branch *b)
2507 const char *from;
2508 struct branch *s;
2510 if (prefixcmp(command_buf.buf, "from "))
2511 return 0;
2513 if (b->branch_tree.tree) {
2514 release_tree_content_recursive(b->branch_tree.tree);
2515 b->branch_tree.tree = NULL;
2518 from = strchr(command_buf.buf, ' ') + 1;
2519 s = lookup_branch(from);
2520 if (b == s)
2521 die("Can't create a branch from itself: %s", b->name);
2522 else if (s) {
2523 unsigned char *t = s->branch_tree.versions[1].sha1;
2524 hashcpy(b->sha1, s->sha1);
2525 hashcpy(b->branch_tree.versions[0].sha1, t);
2526 hashcpy(b->branch_tree.versions[1].sha1, t);
2527 } else if (*from == ':') {
2528 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2529 struct object_entry *oe = find_mark(idnum);
2530 if (oe->type != OBJ_COMMIT)
2531 die("Mark :%" PRIuMAX " not a commit", idnum);
2532 hashcpy(b->sha1, oe->idx.sha1);
2533 if (oe->pack_id != MAX_PACK_ID) {
2534 unsigned long size;
2535 char *buf = gfi_unpack_entry(oe, &size);
2536 parse_from_commit(b, buf, size);
2537 free(buf);
2538 } else
2539 parse_from_existing(b);
2540 } else if (!get_sha1(from, b->sha1))
2541 parse_from_existing(b);
2542 else
2543 die("Invalid ref name or SHA1 expression: %s", from);
2545 read_next_command();
2546 return 1;
2549 static struct hash_list *parse_merge(unsigned int *count)
2551 struct hash_list *list = NULL, *n, *e = e;
2552 const char *from;
2553 struct branch *s;
2555 *count = 0;
2556 while (!prefixcmp(command_buf.buf, "merge ")) {
2557 from = strchr(command_buf.buf, ' ') + 1;
2558 n = xmalloc(sizeof(*n));
2559 s = lookup_branch(from);
2560 if (s)
2561 hashcpy(n->sha1, s->sha1);
2562 else if (*from == ':') {
2563 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2564 struct object_entry *oe = find_mark(idnum);
2565 if (oe->type != OBJ_COMMIT)
2566 die("Mark :%" PRIuMAX " not a commit", idnum);
2567 hashcpy(n->sha1, oe->idx.sha1);
2568 } else if (!get_sha1(from, n->sha1)) {
2569 unsigned long size;
2570 char *buf = read_object_with_reference(n->sha1,
2571 commit_type, &size, n->sha1);
2572 if (!buf || size < 46)
2573 die("Not a valid commit: %s", from);
2574 free(buf);
2575 } else
2576 die("Invalid ref name or SHA1 expression: %s", from);
2578 n->next = NULL;
2579 if (list)
2580 e->next = n;
2581 else
2582 list = n;
2583 e = n;
2584 (*count)++;
2585 read_next_command();
2587 return list;
2590 static void parse_new_commit(void)
2592 static struct strbuf msg = STRBUF_INIT;
2593 struct branch *b;
2594 char *sp;
2595 char *author = NULL;
2596 char *committer = NULL;
2597 struct hash_list *merge_list = NULL;
2598 unsigned int merge_count;
2599 unsigned char prev_fanout, new_fanout;
2601 /* Obtain the branch name from the rest of our command */
2602 sp = strchr(command_buf.buf, ' ') + 1;
2603 b = lookup_branch(sp);
2604 if (!b)
2605 b = new_branch(sp);
2607 read_next_command();
2608 parse_mark();
2609 if (!prefixcmp(command_buf.buf, "author ")) {
2610 author = parse_ident(command_buf.buf + 7);
2611 read_next_command();
2613 if (!prefixcmp(command_buf.buf, "committer ")) {
2614 committer = parse_ident(command_buf.buf + 10);
2615 read_next_command();
2617 if (!committer)
2618 die("Expected committer but didn't get one");
2619 parse_data(&msg, 0, NULL);
2620 read_next_command();
2621 parse_from(b);
2622 merge_list = parse_merge(&merge_count);
2624 /* ensure the branch is active/loaded */
2625 if (!b->branch_tree.tree || !max_active_branches) {
2626 unload_one_branch();
2627 load_branch(b);
2630 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2632 /* file_change* */
2633 while (command_buf.len > 0) {
2634 if (!prefixcmp(command_buf.buf, "M "))
2635 file_change_m(b);
2636 else if (!prefixcmp(command_buf.buf, "D "))
2637 file_change_d(b);
2638 else if (!prefixcmp(command_buf.buf, "R "))
2639 file_change_cr(b, 1);
2640 else if (!prefixcmp(command_buf.buf, "C "))
2641 file_change_cr(b, 0);
2642 else if (!prefixcmp(command_buf.buf, "N "))
2643 note_change_n(b, prev_fanout);
2644 else if (!strcmp("deleteall", command_buf.buf))
2645 file_change_deleteall(b);
2646 else if (!prefixcmp(command_buf.buf, "ls "))
2647 parse_ls(b);
2648 else {
2649 unread_command_buf = 1;
2650 break;
2652 if (read_next_command() == EOF)
2653 break;
2656 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2657 if (new_fanout != prev_fanout)
2658 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2660 /* build the tree and the commit */
2661 store_tree(&b->branch_tree);
2662 hashcpy(b->branch_tree.versions[0].sha1,
2663 b->branch_tree.versions[1].sha1);
2665 strbuf_reset(&new_data);
2666 strbuf_addf(&new_data, "tree %s\n",
2667 sha1_to_hex(b->branch_tree.versions[1].sha1));
2668 if (!is_null_sha1(b->sha1))
2669 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2670 while (merge_list) {
2671 struct hash_list *next = merge_list->next;
2672 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2673 free(merge_list);
2674 merge_list = next;
2676 strbuf_addf(&new_data,
2677 "author %s\n"
2678 "committer %s\n"
2679 "\n",
2680 author ? author : committer, committer);
2681 strbuf_addbuf(&new_data, &msg);
2682 free(author);
2683 free(committer);
2685 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2686 b->pack_id = pack_id;
2687 b->last_commit = object_count_by_type[OBJ_COMMIT];
2690 static void parse_new_tag(void)
2692 static struct strbuf msg = STRBUF_INIT;
2693 char *sp;
2694 const char *from;
2695 char *tagger;
2696 struct branch *s;
2697 struct tag *t;
2698 uintmax_t from_mark = 0;
2699 unsigned char sha1[20];
2700 enum object_type type;
2702 /* Obtain the new tag name from the rest of our command */
2703 sp = strchr(command_buf.buf, ' ') + 1;
2704 t = pool_alloc(sizeof(struct tag));
2705 t->next_tag = NULL;
2706 t->name = pool_strdup(sp);
2707 if (last_tag)
2708 last_tag->next_tag = t;
2709 else
2710 first_tag = t;
2711 last_tag = t;
2712 read_next_command();
2714 /* from ... */
2715 if (prefixcmp(command_buf.buf, "from "))
2716 die("Expected from command, got %s", command_buf.buf);
2717 from = strchr(command_buf.buf, ' ') + 1;
2718 s = lookup_branch(from);
2719 if (s) {
2720 hashcpy(sha1, s->sha1);
2721 type = OBJ_COMMIT;
2722 } else if (*from == ':') {
2723 struct object_entry *oe;
2724 from_mark = strtoumax(from + 1, NULL, 10);
2725 oe = find_mark(from_mark);
2726 type = oe->type;
2727 hashcpy(sha1, oe->idx.sha1);
2728 } else if (!get_sha1(from, sha1)) {
2729 struct object_entry *oe = find_object(sha1);
2730 if (!oe) {
2731 type = sha1_object_info(sha1, NULL);
2732 if (type < 0)
2733 die("Not a valid object: %s", from);
2734 } else
2735 type = oe->type;
2736 } else
2737 die("Invalid ref name or SHA1 expression: %s", from);
2738 read_next_command();
2740 /* tagger ... */
2741 if (!prefixcmp(command_buf.buf, "tagger ")) {
2742 tagger = parse_ident(command_buf.buf + 7);
2743 read_next_command();
2744 } else
2745 tagger = NULL;
2747 /* tag payload/message */
2748 parse_data(&msg, 0, NULL);
2750 /* build the tag object */
2751 strbuf_reset(&new_data);
2753 strbuf_addf(&new_data,
2754 "object %s\n"
2755 "type %s\n"
2756 "tag %s\n",
2757 sha1_to_hex(sha1), typename(type), t->name);
2758 if (tagger)
2759 strbuf_addf(&new_data,
2760 "tagger %s\n", tagger);
2761 strbuf_addch(&new_data, '\n');
2762 strbuf_addbuf(&new_data, &msg);
2763 free(tagger);
2765 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2766 t->pack_id = MAX_PACK_ID;
2767 else
2768 t->pack_id = pack_id;
2771 static void parse_reset_branch(void)
2773 struct branch *b;
2774 char *sp;
2776 /* Obtain the branch name from the rest of our command */
2777 sp = strchr(command_buf.buf, ' ') + 1;
2778 b = lookup_branch(sp);
2779 if (b) {
2780 hashclr(b->sha1);
2781 hashclr(b->branch_tree.versions[0].sha1);
2782 hashclr(b->branch_tree.versions[1].sha1);
2783 if (b->branch_tree.tree) {
2784 release_tree_content_recursive(b->branch_tree.tree);
2785 b->branch_tree.tree = NULL;
2788 else
2789 b = new_branch(sp);
2790 read_next_command();
2791 parse_from(b);
2792 if (command_buf.len > 0)
2793 unread_command_buf = 1;
2796 static void cat_blob_write(const char *buf, unsigned long size)
2798 if (write_in_full(cat_blob_fd, buf, size) != size)
2799 die_errno("Write to frontend failed");
2802 static void cat_blob(struct object_entry *oe, unsigned char sha1[20])
2804 struct strbuf line = STRBUF_INIT;
2805 unsigned long size;
2806 enum object_type type = 0;
2807 char *buf;
2809 if (!oe || oe->pack_id == MAX_PACK_ID) {
2810 buf = read_sha1_file(sha1, &type, &size);
2811 } else {
2812 type = oe->type;
2813 buf = gfi_unpack_entry(oe, &size);
2817 * Output based on batch_one_object() from cat-file.c.
2819 if (type <= 0) {
2820 strbuf_reset(&line);
2821 strbuf_addf(&line, "%s missing\n", sha1_to_hex(sha1));
2822 cat_blob_write(line.buf, line.len);
2823 strbuf_release(&line);
2824 free(buf);
2825 return;
2827 if (!buf)
2828 die("Can't read object %s", sha1_to_hex(sha1));
2829 if (type != OBJ_BLOB)
2830 die("Object %s is a %s but a blob was expected.",
2831 sha1_to_hex(sha1), typename(type));
2832 strbuf_reset(&line);
2833 strbuf_addf(&line, "%s %s %lu\n", sha1_to_hex(sha1),
2834 typename(type), size);
2835 cat_blob_write(line.buf, line.len);
2836 strbuf_release(&line);
2837 cat_blob_write(buf, size);
2838 cat_blob_write("\n", 1);
2839 if (oe && oe->pack_id == pack_id) {
2840 last_blob.offset = oe->idx.offset;
2841 strbuf_attach(&last_blob.data, buf, size, size);
2842 last_blob.depth = oe->depth;
2843 } else
2844 free(buf);
2847 static void parse_cat_blob(void)
2849 const char *p;
2850 struct object_entry *oe = oe;
2851 unsigned char sha1[20];
2853 /* cat-blob SP <object> LF */
2854 p = command_buf.buf + strlen("cat-blob ");
2855 if (*p == ':') {
2856 char *x;
2857 oe = find_mark(strtoumax(p + 1, &x, 10));
2858 if (x == p + 1)
2859 die("Invalid mark: %s", command_buf.buf);
2860 if (!oe)
2861 die("Unknown mark: %s", command_buf.buf);
2862 if (*x)
2863 die("Garbage after mark: %s", command_buf.buf);
2864 hashcpy(sha1, oe->idx.sha1);
2865 } else {
2866 if (get_sha1_hex(p, sha1))
2867 die("Invalid SHA1: %s", command_buf.buf);
2868 if (p[40])
2869 die("Garbage after SHA1: %s", command_buf.buf);
2870 oe = find_object(sha1);
2873 cat_blob(oe, sha1);
2876 static struct object_entry *dereference(struct object_entry *oe,
2877 unsigned char sha1[20])
2879 unsigned long size;
2880 char *buf = NULL;
2881 if (!oe) {
2882 enum object_type type = sha1_object_info(sha1, NULL);
2883 if (type < 0)
2884 die("object not found: %s", sha1_to_hex(sha1));
2885 /* cache it! */
2886 oe = insert_object(sha1);
2887 oe->type = type;
2888 oe->pack_id = MAX_PACK_ID;
2889 oe->idx.offset = 1;
2891 switch (oe->type) {
2892 case OBJ_TREE: /* easy case. */
2893 return oe;
2894 case OBJ_COMMIT:
2895 case OBJ_TAG:
2896 break;
2897 default:
2898 die("Not a treeish: %s", command_buf.buf);
2901 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2902 buf = gfi_unpack_entry(oe, &size);
2903 } else {
2904 enum object_type unused;
2905 buf = read_sha1_file(sha1, &unused, &size);
2907 if (!buf)
2908 die("Can't load object %s", sha1_to_hex(sha1));
2910 /* Peel one layer. */
2911 switch (oe->type) {
2912 case OBJ_TAG:
2913 if (size < 40 + strlen("object ") ||
2914 get_sha1_hex(buf + strlen("object "), sha1))
2915 die("Invalid SHA1 in tag: %s", command_buf.buf);
2916 break;
2917 case OBJ_COMMIT:
2918 if (size < 40 + strlen("tree ") ||
2919 get_sha1_hex(buf + strlen("tree "), sha1))
2920 die("Invalid SHA1 in commit: %s", command_buf.buf);
2923 free(buf);
2924 return find_object(sha1);
2927 static struct object_entry *parse_treeish_dataref(const char **p)
2929 unsigned char sha1[20];
2930 struct object_entry *e;
2932 if (**p == ':') { /* <mark> */
2933 char *endptr;
2934 e = find_mark(strtoumax(*p + 1, &endptr, 10));
2935 if (endptr == *p + 1)
2936 die("Invalid mark: %s", command_buf.buf);
2937 if (!e)
2938 die("Unknown mark: %s", command_buf.buf);
2939 *p = endptr;
2940 hashcpy(sha1, e->idx.sha1);
2941 } else { /* <sha1> */
2942 if (get_sha1_hex(*p, sha1))
2943 die("Invalid SHA1: %s", command_buf.buf);
2944 e = find_object(sha1);
2945 *p += 40;
2948 while (!e || e->type != OBJ_TREE)
2949 e = dereference(e, sha1);
2950 return e;
2953 static void print_ls(int mode, const unsigned char *sha1, const char *path)
2955 static struct strbuf line = STRBUF_INIT;
2957 /* See show_tree(). */
2958 const char *type =
2959 S_ISGITLINK(mode) ? commit_type :
2960 S_ISDIR(mode) ? tree_type :
2961 blob_type;
2963 if (!mode) {
2964 /* missing SP path LF */
2965 strbuf_reset(&line);
2966 strbuf_addstr(&line, "missing ");
2967 quote_c_style(path, &line, NULL, 0);
2968 strbuf_addch(&line, '\n');
2969 } else {
2970 /* mode SP type SP object_name TAB path LF */
2971 strbuf_reset(&line);
2972 strbuf_addf(&line, "%06o %s %s\t",
2973 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
2974 quote_c_style(path, &line, NULL, 0);
2975 strbuf_addch(&line, '\n');
2977 cat_blob_write(line.buf, line.len);
2980 static void parse_ls(struct branch *b)
2982 const char *p;
2983 struct tree_entry *root = NULL;
2984 struct tree_entry leaf = {NULL};
2986 /* ls SP (<treeish> SP)? <path> */
2987 p = command_buf.buf + strlen("ls ");
2988 if (*p == '"') {
2989 if (!b)
2990 die("Not in a commit: %s", command_buf.buf);
2991 root = &b->branch_tree;
2992 } else {
2993 struct object_entry *e = parse_treeish_dataref(&p);
2994 root = new_tree_entry();
2995 hashcpy(root->versions[1].sha1, e->idx.sha1);
2996 load_tree(root);
2997 if (*p++ != ' ')
2998 die("Missing space after tree-ish: %s", command_buf.buf);
3000 if (*p == '"') {
3001 static struct strbuf uq = STRBUF_INIT;
3002 const char *endp;
3003 strbuf_reset(&uq);
3004 if (unquote_c_style(&uq, p, &endp))
3005 die("Invalid path: %s", command_buf.buf);
3006 if (*endp)
3007 die("Garbage after path in: %s", command_buf.buf);
3008 p = uq.buf;
3010 tree_content_get(root, p, &leaf);
3012 * A directory in preparation would have a sha1 of zero
3013 * until it is saved. Save, for simplicity.
3015 if (S_ISDIR(leaf.versions[1].mode))
3016 store_tree(&leaf);
3018 print_ls(leaf.versions[1].mode, leaf.versions[1].sha1, p);
3019 if (!b || root != &b->branch_tree)
3020 release_tree_entry(root);
3023 static void checkpoint(void)
3025 checkpoint_requested = 0;
3026 if (object_count) {
3027 cycle_packfile();
3028 dump_branches();
3029 dump_tags();
3030 dump_marks();
3034 static void parse_checkpoint(void)
3036 checkpoint_requested = 1;
3037 skip_optional_lf();
3040 static void parse_progress(void)
3042 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3043 fputc('\n', stdout);
3044 fflush(stdout);
3045 skip_optional_lf();
3048 static char* make_fast_import_path(const char *path)
3050 struct strbuf abs_path = STRBUF_INIT;
3052 if (!relative_marks_paths || is_absolute_path(path))
3053 return xstrdup(path);
3054 strbuf_addf(&abs_path, "%s/info/fast-import/%s", get_git_dir(), path);
3055 return strbuf_detach(&abs_path, NULL);
3058 static void option_import_marks(const char *marks,
3059 int from_stream, int ignore_missing)
3061 if (import_marks_file) {
3062 if (from_stream)
3063 die("Only one import-marks command allowed per stream");
3065 /* read previous mark file */
3066 if(!import_marks_file_from_stream)
3067 read_marks();
3070 import_marks_file = make_fast_import_path(marks);
3071 safe_create_leading_directories_const(import_marks_file);
3072 import_marks_file_from_stream = from_stream;
3073 import_marks_file_ignore_missing = ignore_missing;
3076 static void option_date_format(const char *fmt)
3078 if (!strcmp(fmt, "raw"))
3079 whenspec = WHENSPEC_RAW;
3080 else if (!strcmp(fmt, "rfc2822"))
3081 whenspec = WHENSPEC_RFC2822;
3082 else if (!strcmp(fmt, "now"))
3083 whenspec = WHENSPEC_NOW;
3084 else
3085 die("unknown --date-format argument %s", fmt);
3088 static unsigned long ulong_arg(const char *option, const char *arg)
3090 char *endptr;
3091 unsigned long rv = strtoul(arg, &endptr, 0);
3092 if (strchr(arg, '-') || endptr == arg || *endptr)
3093 die("%s: argument must be a non-negative integer", option);
3094 return rv;
3097 static void option_depth(const char *depth)
3099 max_depth = ulong_arg("--depth", depth);
3100 if (max_depth > MAX_DEPTH)
3101 die("--depth cannot exceed %u", MAX_DEPTH);
3104 static void option_active_branches(const char *branches)
3106 max_active_branches = ulong_arg("--active-branches", branches);
3109 static void option_export_marks(const char *marks)
3111 export_marks_file = make_fast_import_path(marks);
3112 safe_create_leading_directories_const(export_marks_file);
3115 static void option_cat_blob_fd(const char *fd)
3117 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3118 if (n > (unsigned long) INT_MAX)
3119 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3120 cat_blob_fd = (int) n;
3123 static void option_export_pack_edges(const char *edges)
3125 if (pack_edges)
3126 fclose(pack_edges);
3127 pack_edges = fopen(edges, "a");
3128 if (!pack_edges)
3129 die_errno("Cannot open '%s'", edges);
3132 static int parse_one_option(const char *option)
3134 if (!prefixcmp(option, "max-pack-size=")) {
3135 unsigned long v;
3136 if (!git_parse_ulong(option + 14, &v))
3137 return 0;
3138 if (v < 8192) {
3139 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3140 v *= 1024 * 1024;
3141 } else if (v < 1024 * 1024) {
3142 warning("minimum max-pack-size is 1 MiB");
3143 v = 1024 * 1024;
3145 max_packsize = v;
3146 } else if (!prefixcmp(option, "big-file-threshold=")) {
3147 unsigned long v;
3148 if (!git_parse_ulong(option + 19, &v))
3149 return 0;
3150 big_file_threshold = v;
3151 } else if (!prefixcmp(option, "depth=")) {
3152 option_depth(option + 6);
3153 } else if (!prefixcmp(option, "active-branches=")) {
3154 option_active_branches(option + 16);
3155 } else if (!prefixcmp(option, "export-pack-edges=")) {
3156 option_export_pack_edges(option + 18);
3157 } else if (!prefixcmp(option, "quiet")) {
3158 show_stats = 0;
3159 } else if (!prefixcmp(option, "stats")) {
3160 show_stats = 1;
3161 } else {
3162 return 0;
3165 return 1;
3168 static int parse_one_feature(const char *feature, int from_stream)
3170 if (!prefixcmp(feature, "date-format=")) {
3171 option_date_format(feature + 12);
3172 } else if (!prefixcmp(feature, "import-marks=")) {
3173 option_import_marks(feature + 13, from_stream, 0);
3174 } else if (!prefixcmp(feature, "import-marks-if-exists=")) {
3175 option_import_marks(feature + strlen("import-marks-if-exists="),
3176 from_stream, 1);
3177 } else if (!prefixcmp(feature, "export-marks=")) {
3178 option_export_marks(feature + 13);
3179 } else if (!strcmp(feature, "cat-blob")) {
3180 ; /* Don't die - this feature is supported */
3181 } else if (!strcmp(feature, "relative-marks")) {
3182 relative_marks_paths = 1;
3183 } else if (!strcmp(feature, "no-relative-marks")) {
3184 relative_marks_paths = 0;
3185 } else if (!strcmp(feature, "done")) {
3186 require_explicit_termination = 1;
3187 } else if (!strcmp(feature, "force")) {
3188 force_update = 1;
3189 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3190 ; /* do nothing; we have the feature */
3191 } else {
3192 return 0;
3195 return 1;
3198 static void parse_feature(void)
3200 char *feature = command_buf.buf + 8;
3202 if (seen_data_command)
3203 die("Got feature command '%s' after data command", feature);
3205 if (parse_one_feature(feature, 1))
3206 return;
3208 die("This version of fast-import does not support feature %s.", feature);
3211 static void parse_option(void)
3213 char *option = command_buf.buf + 11;
3215 if (seen_data_command)
3216 die("Got option command '%s' after data command", option);
3218 if (parse_one_option(option))
3219 return;
3221 die("This version of fast-import does not support option: %s", option);
3224 static int git_pack_config(const char *k, const char *v, void *cb)
3226 if (!strcmp(k, "pack.depth")) {
3227 max_depth = git_config_int(k, v);
3228 if (max_depth > MAX_DEPTH)
3229 max_depth = MAX_DEPTH;
3230 return 0;
3232 if (!strcmp(k, "pack.compression")) {
3233 int level = git_config_int(k, v);
3234 if (level == -1)
3235 level = Z_DEFAULT_COMPRESSION;
3236 else if (level < 0 || level > Z_BEST_COMPRESSION)
3237 die("bad pack compression level %d", level);
3238 pack_compression_level = level;
3239 pack_compression_seen = 1;
3240 return 0;
3242 if (!strcmp(k, "pack.indexversion")) {
3243 pack_idx_opts.version = git_config_int(k, v);
3244 if (pack_idx_opts.version > 2)
3245 die("bad pack.indexversion=%"PRIu32,
3246 pack_idx_opts.version);
3247 return 0;
3249 if (!strcmp(k, "pack.packsizelimit")) {
3250 max_packsize = git_config_ulong(k, v);
3251 return 0;
3253 return git_default_config(k, v, cb);
3256 static const char fast_import_usage[] =
3257 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3259 static void parse_argv(void)
3261 unsigned int i;
3263 for (i = 1; i < global_argc; i++) {
3264 const char *a = global_argv[i];
3266 if (*a != '-' || !strcmp(a, "--"))
3267 break;
3269 if (parse_one_option(a + 2))
3270 continue;
3272 if (parse_one_feature(a + 2, 0))
3273 continue;
3275 if (!prefixcmp(a + 2, "cat-blob-fd=")) {
3276 option_cat_blob_fd(a + 2 + strlen("cat-blob-fd="));
3277 continue;
3280 die("unknown option %s", a);
3282 if (i != global_argc)
3283 usage(fast_import_usage);
3285 seen_data_command = 1;
3286 if (import_marks_file)
3287 read_marks();
3290 int main(int argc, const char **argv)
3292 unsigned int i;
3294 git_extract_argv0_path(argv[0]);
3296 if (argc == 2 && !strcmp(argv[1], "-h"))
3297 usage(fast_import_usage);
3299 setup_git_directory();
3300 reset_pack_idx_option(&pack_idx_opts);
3301 git_config(git_pack_config, NULL);
3302 if (!pack_compression_seen && core_compression_seen)
3303 pack_compression_level = core_compression_level;
3305 alloc_objects(object_entry_alloc);
3306 strbuf_init(&command_buf, 0);
3307 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3308 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3309 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3310 marks = pool_calloc(1, sizeof(struct mark_set));
3312 global_argc = argc;
3313 global_argv = argv;
3315 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
3316 for (i = 0; i < (cmd_save - 1); i++)
3317 rc_free[i].next = &rc_free[i + 1];
3318 rc_free[cmd_save - 1].next = NULL;
3320 prepare_packed_git();
3321 start_packfile();
3322 set_die_routine(die_nicely);
3323 set_checkpoint_signal();
3324 while (read_next_command() != EOF) {
3325 if (!strcmp("blob", command_buf.buf))
3326 parse_new_blob();
3327 else if (!prefixcmp(command_buf.buf, "ls "))
3328 parse_ls(NULL);
3329 else if (!prefixcmp(command_buf.buf, "commit "))
3330 parse_new_commit();
3331 else if (!prefixcmp(command_buf.buf, "tag "))
3332 parse_new_tag();
3333 else if (!prefixcmp(command_buf.buf, "reset "))
3334 parse_reset_branch();
3335 else if (!strcmp("checkpoint", command_buf.buf))
3336 parse_checkpoint();
3337 else if (!strcmp("done", command_buf.buf))
3338 break;
3339 else if (!prefixcmp(command_buf.buf, "progress "))
3340 parse_progress();
3341 else if (!prefixcmp(command_buf.buf, "feature "))
3342 parse_feature();
3343 else if (!prefixcmp(command_buf.buf, "option git "))
3344 parse_option();
3345 else if (!prefixcmp(command_buf.buf, "option "))
3346 /* ignore non-git options*/;
3347 else
3348 die("Unsupported command: %s", command_buf.buf);
3350 if (checkpoint_requested)
3351 checkpoint();
3354 /* argv hasn't been parsed yet, do so */
3355 if (!seen_data_command)
3356 parse_argv();
3358 if (require_explicit_termination && feof(stdin))
3359 die("stream ends early");
3361 end_packfile();
3363 dump_branches();
3364 dump_tags();
3365 unkeep_all_packs();
3366 dump_marks();
3368 if (pack_edges)
3369 fclose(pack_edges);
3371 if (show_stats) {
3372 uintmax_t total_count = 0, duplicate_count = 0;
3373 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3374 total_count += object_count_by_type[i];
3375 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3376 duplicate_count += duplicate_count_by_type[i];
3378 fprintf(stderr, "%s statistics:\n", argv[0]);
3379 fprintf(stderr, "---------------------------------------------------------------------\n");
3380 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3381 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3382 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]);
3383 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]);
3384 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]);
3385 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]);
3386 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3387 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3388 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3389 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
3390 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
3391 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3392 fprintf(stderr, "---------------------------------------------------------------------\n");
3393 pack_report();
3394 fprintf(stderr, "---------------------------------------------------------------------\n");
3395 fprintf(stderr, "\n");
3398 return failure ? 1 : 0;