Merge branch 'sg/test-bash-version-fix'
[git.git] / fast-import.c
blob69886687ce95fd7dfe5363d0c776e21241b6b830
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 commit-ish lf)?
26 ('merge' sp commit-ish 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 commit-ish lf;
47 note_inm ::= 'N' sp 'inline' sp commit-ish lf
48 data;
50 new_tag ::= 'tag' sp tag_str lf
51 'from' sp commit-ish lf
52 ('tagger' (sp name)? sp '<' email '>' sp when lf)?
53 tag_msg;
54 tag_msg ::= data;
56 reset_branch ::= 'reset' sp ref_str lf
57 ('from' sp commit-ish 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 commit-ish ::= (ref_str | hexsha1 | sha1exp_str | idnum);
97 ref_str ::= ref;
98 sha1exp_str ::= sha1exp;
99 tag_str ::= tag;
100 path_str ::= path | '"' quoted(path) '"' ;
101 mode ::= '100644' | '644'
102 | '100755' | '755'
103 | '120000'
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, get-mark, ls-tree, and cat-blob requests may
138 # appear anywhere in the input, except within a data command. Any
139 # form of the data command always escapes the related input from
140 # 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 get_mark ::= 'get-mark' sp idnum lf;
148 cat_blob ::= 'cat-blob' sp (hexsha1 | idnum) lf;
149 ls_tree ::= 'ls' sp (hexsha1 | idnum) sp path_str lf;
151 comment ::= '#' not_lf* lf;
152 not_lf ::= # Any byte that is not ASCII newline (LF);
155 #include "builtin.h"
156 #include "cache.h"
157 #include "repository.h"
158 #include "config.h"
159 #include "lockfile.h"
160 #include "object.h"
161 #include "blob.h"
162 #include "tree.h"
163 #include "commit.h"
164 #include "delta.h"
165 #include "pack.h"
166 #include "refs.h"
167 #include "csum-file.h"
168 #include "quote.h"
169 #include "dir.h"
170 #include "run-command.h"
171 #include "packfile.h"
172 #include "object-store.h"
173 #include "mem-pool.h"
174 #include "commit-reach.h"
176 #define PACK_ID_BITS 16
177 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
178 #define DEPTH_BITS 13
179 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
182 * We abuse the setuid bit on directories to mean "do not delta".
184 #define NO_DELTA S_ISUID
186 struct object_entry {
187 struct pack_idx_entry idx;
188 struct object_entry *next;
189 uint32_t type : TYPE_BITS,
190 pack_id : PACK_ID_BITS,
191 depth : DEPTH_BITS;
194 struct object_entry_pool {
195 struct object_entry_pool *next_pool;
196 struct object_entry *next_free;
197 struct object_entry *end;
198 struct object_entry entries[FLEX_ARRAY]; /* more */
201 struct mark_set {
202 union {
203 struct object_entry *marked[1024];
204 struct mark_set *sets[1024];
205 } data;
206 unsigned int shift;
209 struct last_object {
210 struct strbuf data;
211 off_t offset;
212 unsigned int depth;
213 unsigned no_swap : 1;
216 struct atom_str {
217 struct atom_str *next_atom;
218 unsigned short str_len;
219 char str_dat[FLEX_ARRAY]; /* more */
222 struct tree_content;
223 struct tree_entry {
224 struct tree_content *tree;
225 struct atom_str *name;
226 struct tree_entry_ms {
227 uint16_t mode;
228 struct object_id oid;
229 } versions[2];
232 struct tree_content {
233 unsigned int entry_capacity; /* must match avail_tree_content */
234 unsigned int entry_count;
235 unsigned int delta_depth;
236 struct tree_entry *entries[FLEX_ARRAY]; /* more */
239 struct avail_tree_content {
240 unsigned int entry_capacity; /* must match tree_content */
241 struct avail_tree_content *next_avail;
244 struct branch {
245 struct branch *table_next_branch;
246 struct branch *active_next_branch;
247 const char *name;
248 struct tree_entry branch_tree;
249 uintmax_t last_commit;
250 uintmax_t num_notes;
251 unsigned active : 1;
252 unsigned delete : 1;
253 unsigned pack_id : PACK_ID_BITS;
254 struct object_id oid;
257 struct tag {
258 struct tag *next_tag;
259 const char *name;
260 unsigned int pack_id;
261 struct object_id oid;
264 struct hash_list {
265 struct hash_list *next;
266 struct object_id oid;
269 typedef enum {
270 WHENSPEC_RAW = 1,
271 WHENSPEC_RFC2822,
272 WHENSPEC_NOW
273 } whenspec_type;
275 struct recent_command {
276 struct recent_command *prev;
277 struct recent_command *next;
278 char *buf;
281 /* Configured limits on output */
282 static unsigned long max_depth = 50;
283 static off_t max_packsize;
284 static int unpack_limit = 100;
285 static int force_update;
287 /* Stats and misc. counters */
288 static uintmax_t alloc_count;
289 static uintmax_t marks_set_count;
290 static uintmax_t object_count_by_type[1 << TYPE_BITS];
291 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
292 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
293 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
294 static unsigned long object_count;
295 static unsigned long branch_count;
296 static unsigned long branch_load_count;
297 static int failure;
298 static FILE *pack_edges;
299 static unsigned int show_stats = 1;
300 static int global_argc;
301 static const char **global_argv;
303 /* Memory pools */
304 static struct mem_pool fi_mem_pool = {NULL, 2*1024*1024 -
305 sizeof(struct mp_block), 0 };
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 hashfile *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 import_marks_file_done;
330 static int relative_marks_paths;
332 /* Our last blob */
333 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
335 /* Tree management */
336 static unsigned int tree_entry_alloc = 1000;
337 static void *avail_tree_entry;
338 static unsigned int avail_tree_table_sz = 100;
339 static struct avail_tree_content **avail_tree_table;
340 static size_t tree_entry_allocd;
341 static struct strbuf old_tree = STRBUF_INIT;
342 static struct strbuf new_tree = STRBUF_INIT;
344 /* Branch data */
345 static unsigned long max_active_branches = 5;
346 static unsigned long cur_active_branches;
347 static unsigned long branch_table_sz = 1039;
348 static struct branch **branch_table;
349 static struct branch *active_branches;
351 /* Tag data */
352 static struct tag *first_tag;
353 static struct tag *last_tag;
355 /* Input stream parsing */
356 static whenspec_type whenspec = WHENSPEC_RAW;
357 static struct strbuf command_buf = STRBUF_INIT;
358 static int unread_command_buf;
359 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
360 static struct recent_command *cmd_tail = &cmd_hist;
361 static struct recent_command *rc_free;
362 static unsigned int cmd_save = 100;
363 static uintmax_t next_mark;
364 static struct strbuf new_data = STRBUF_INIT;
365 static int seen_data_command;
366 static int require_explicit_termination;
368 /* Signal handling */
369 static volatile sig_atomic_t checkpoint_requested;
371 /* Where to write output of cat-blob commands */
372 static int cat_blob_fd = STDOUT_FILENO;
374 static void parse_argv(void);
375 static void parse_get_mark(const char *p);
376 static void parse_cat_blob(const char *p);
377 static void parse_ls(const char *p, struct branch *b);
379 static void write_branch_report(FILE *rpt, struct branch *b)
381 fprintf(rpt, "%s:\n", b->name);
383 fprintf(rpt, " status :");
384 if (b->active)
385 fputs(" active", rpt);
386 if (b->branch_tree.tree)
387 fputs(" loaded", rpt);
388 if (is_null_oid(&b->branch_tree.versions[1].oid))
389 fputs(" dirty", rpt);
390 fputc('\n', rpt);
392 fprintf(rpt, " tip commit : %s\n", oid_to_hex(&b->oid));
393 fprintf(rpt, " old tree : %s\n",
394 oid_to_hex(&b->branch_tree.versions[0].oid));
395 fprintf(rpt, " cur tree : %s\n",
396 oid_to_hex(&b->branch_tree.versions[1].oid));
397 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
399 fputs(" last pack : ", rpt);
400 if (b->pack_id < MAX_PACK_ID)
401 fprintf(rpt, "%u", b->pack_id);
402 fputc('\n', rpt);
404 fputc('\n', rpt);
407 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
409 static void write_crash_report(const char *err)
411 char *loc = git_pathdup("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
412 FILE *rpt = fopen(loc, "w");
413 struct branch *b;
414 unsigned long lu;
415 struct recent_command *rc;
417 if (!rpt) {
418 error_errno("can't write crash report %s", loc);
419 free(loc);
420 return;
423 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
425 fprintf(rpt, "fast-import crash report:\n");
426 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
427 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
428 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_MODE(ISO8601)));
429 fputc('\n', rpt);
431 fputs("fatal: ", rpt);
432 fputs(err, rpt);
433 fputc('\n', rpt);
435 fputc('\n', rpt);
436 fputs("Most Recent Commands Before Crash\n", rpt);
437 fputs("---------------------------------\n", rpt);
438 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
439 if (rc->next == &cmd_hist)
440 fputs("* ", rpt);
441 else
442 fputs(" ", rpt);
443 fputs(rc->buf, rpt);
444 fputc('\n', rpt);
447 fputc('\n', rpt);
448 fputs("Active Branch LRU\n", rpt);
449 fputs("-----------------\n", rpt);
450 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
451 cur_active_branches,
452 max_active_branches);
453 fputc('\n', rpt);
454 fputs(" pos clock name\n", rpt);
455 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
456 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
457 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
458 ++lu, b->last_commit, b->name);
460 fputc('\n', rpt);
461 fputs("Inactive Branches\n", rpt);
462 fputs("-----------------\n", rpt);
463 for (lu = 0; lu < branch_table_sz; lu++) {
464 for (b = branch_table[lu]; b; b = b->table_next_branch)
465 write_branch_report(rpt, b);
468 if (first_tag) {
469 struct tag *tg;
470 fputc('\n', rpt);
471 fputs("Annotated Tags\n", rpt);
472 fputs("--------------\n", rpt);
473 for (tg = first_tag; tg; tg = tg->next_tag) {
474 fputs(oid_to_hex(&tg->oid), rpt);
475 fputc(' ', rpt);
476 fputs(tg->name, rpt);
477 fputc('\n', rpt);
481 fputc('\n', rpt);
482 fputs("Marks\n", rpt);
483 fputs("-----\n", rpt);
484 if (export_marks_file)
485 fprintf(rpt, " exported to %s\n", export_marks_file);
486 else
487 dump_marks_helper(rpt, 0, marks);
489 fputc('\n', rpt);
490 fputs("-------------------\n", rpt);
491 fputs("END OF CRASH REPORT\n", rpt);
492 fclose(rpt);
493 free(loc);
496 static void end_packfile(void);
497 static void unkeep_all_packs(void);
498 static void dump_marks(void);
500 static NORETURN void die_nicely(const char *err, va_list params)
502 static int zombie;
503 char message[2 * PATH_MAX];
505 vsnprintf(message, sizeof(message), err, params);
506 fputs("fatal: ", stderr);
507 fputs(message, stderr);
508 fputc('\n', stderr);
510 if (!zombie) {
511 zombie = 1;
512 write_crash_report(message);
513 end_packfile();
514 unkeep_all_packs();
515 dump_marks();
517 exit(128);
520 #ifndef SIGUSR1 /* Windows, for example */
522 static void set_checkpoint_signal(void)
526 #else
528 static void checkpoint_signal(int signo)
530 checkpoint_requested = 1;
533 static void set_checkpoint_signal(void)
535 struct sigaction sa;
537 memset(&sa, 0, sizeof(sa));
538 sa.sa_handler = checkpoint_signal;
539 sigemptyset(&sa.sa_mask);
540 sa.sa_flags = SA_RESTART;
541 sigaction(SIGUSR1, &sa, NULL);
544 #endif
546 static void alloc_objects(unsigned int cnt)
548 struct object_entry_pool *b;
550 b = xmalloc(sizeof(struct object_entry_pool)
551 + cnt * sizeof(struct object_entry));
552 b->next_pool = blocks;
553 b->next_free = b->entries;
554 b->end = b->entries + cnt;
555 blocks = b;
556 alloc_count += cnt;
559 static struct object_entry *new_object(struct object_id *oid)
561 struct object_entry *e;
563 if (blocks->next_free == blocks->end)
564 alloc_objects(object_entry_alloc);
566 e = blocks->next_free++;
567 oidcpy(&e->idx.oid, oid);
568 return e;
571 static struct object_entry *find_object(struct object_id *oid)
573 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
574 struct object_entry *e;
575 for (e = object_table[h]; e; e = e->next)
576 if (oideq(oid, &e->idx.oid))
577 return e;
578 return NULL;
581 static struct object_entry *insert_object(struct object_id *oid)
583 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
584 struct object_entry *e = object_table[h];
586 while (e) {
587 if (oideq(oid, &e->idx.oid))
588 return e;
589 e = e->next;
592 e = new_object(oid);
593 e->next = object_table[h];
594 e->idx.offset = 0;
595 object_table[h] = e;
596 return e;
599 static void invalidate_pack_id(unsigned int id)
601 unsigned int h;
602 unsigned long lu;
603 struct tag *t;
605 for (h = 0; h < ARRAY_SIZE(object_table); h++) {
606 struct object_entry *e;
608 for (e = object_table[h]; e; e = e->next)
609 if (e->pack_id == id)
610 e->pack_id = MAX_PACK_ID;
613 for (lu = 0; lu < branch_table_sz; lu++) {
614 struct branch *b;
616 for (b = branch_table[lu]; b; b = b->table_next_branch)
617 if (b->pack_id == id)
618 b->pack_id = MAX_PACK_ID;
621 for (t = first_tag; t; t = t->next_tag)
622 if (t->pack_id == id)
623 t->pack_id = MAX_PACK_ID;
626 static unsigned int hc_str(const char *s, size_t len)
628 unsigned int r = 0;
629 while (len-- > 0)
630 r = r * 31 + *s++;
631 return r;
634 static char *pool_strdup(const char *s)
636 size_t len = strlen(s) + 1;
637 char *r = mem_pool_alloc(&fi_mem_pool, len);
638 memcpy(r, s, len);
639 return r;
642 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
644 struct mark_set *s = marks;
645 while ((idnum >> s->shift) >= 1024) {
646 s = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct mark_set));
647 s->shift = marks->shift + 10;
648 s->data.sets[0] = marks;
649 marks = s;
651 while (s->shift) {
652 uintmax_t i = idnum >> s->shift;
653 idnum -= i << s->shift;
654 if (!s->data.sets[i]) {
655 s->data.sets[i] = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct mark_set));
656 s->data.sets[i]->shift = s->shift - 10;
658 s = s->data.sets[i];
660 if (!s->data.marked[idnum])
661 marks_set_count++;
662 s->data.marked[idnum] = oe;
665 static struct object_entry *find_mark(uintmax_t idnum)
667 uintmax_t orig_idnum = idnum;
668 struct mark_set *s = marks;
669 struct object_entry *oe = NULL;
670 if ((idnum >> s->shift) < 1024) {
671 while (s && s->shift) {
672 uintmax_t i = idnum >> s->shift;
673 idnum -= i << s->shift;
674 s = s->data.sets[i];
676 if (s)
677 oe = s->data.marked[idnum];
679 if (!oe)
680 die("mark :%" PRIuMAX " not declared", orig_idnum);
681 return oe;
684 static struct atom_str *to_atom(const char *s, unsigned short len)
686 unsigned int hc = hc_str(s, len) % atom_table_sz;
687 struct atom_str *c;
689 for (c = atom_table[hc]; c; c = c->next_atom)
690 if (c->str_len == len && !strncmp(s, c->str_dat, len))
691 return c;
693 c = mem_pool_alloc(&fi_mem_pool, sizeof(struct atom_str) + len + 1);
694 c->str_len = len;
695 memcpy(c->str_dat, s, len);
696 c->str_dat[len] = 0;
697 c->next_atom = atom_table[hc];
698 atom_table[hc] = c;
699 atom_cnt++;
700 return c;
703 static struct branch *lookup_branch(const char *name)
705 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
706 struct branch *b;
708 for (b = branch_table[hc]; b; b = b->table_next_branch)
709 if (!strcmp(name, b->name))
710 return b;
711 return NULL;
714 static struct branch *new_branch(const char *name)
716 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
717 struct branch *b = lookup_branch(name);
719 if (b)
720 die("Invalid attempt to create duplicate branch: %s", name);
721 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
722 die("Branch name doesn't conform to GIT standards: %s", name);
724 b = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct branch));
725 b->name = pool_strdup(name);
726 b->table_next_branch = branch_table[hc];
727 b->branch_tree.versions[0].mode = S_IFDIR;
728 b->branch_tree.versions[1].mode = S_IFDIR;
729 b->num_notes = 0;
730 b->active = 0;
731 b->pack_id = MAX_PACK_ID;
732 branch_table[hc] = b;
733 branch_count++;
734 return b;
737 static unsigned int hc_entries(unsigned int cnt)
739 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
740 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
743 static struct tree_content *new_tree_content(unsigned int cnt)
745 struct avail_tree_content *f, *l = NULL;
746 struct tree_content *t;
747 unsigned int hc = hc_entries(cnt);
749 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
750 if (f->entry_capacity >= cnt)
751 break;
753 if (f) {
754 if (l)
755 l->next_avail = f->next_avail;
756 else
757 avail_tree_table[hc] = f->next_avail;
758 } else {
759 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
760 f = mem_pool_alloc(&fi_mem_pool, sizeof(*t) + sizeof(t->entries[0]) * cnt);
761 f->entry_capacity = cnt;
764 t = (struct tree_content*)f;
765 t->entry_count = 0;
766 t->delta_depth = 0;
767 return t;
770 static void release_tree_entry(struct tree_entry *e);
771 static void release_tree_content(struct tree_content *t)
773 struct avail_tree_content *f = (struct avail_tree_content*)t;
774 unsigned int hc = hc_entries(f->entry_capacity);
775 f->next_avail = avail_tree_table[hc];
776 avail_tree_table[hc] = f;
779 static void release_tree_content_recursive(struct tree_content *t)
781 unsigned int i;
782 for (i = 0; i < t->entry_count; i++)
783 release_tree_entry(t->entries[i]);
784 release_tree_content(t);
787 static struct tree_content *grow_tree_content(
788 struct tree_content *t,
789 int amt)
791 struct tree_content *r = new_tree_content(t->entry_count + amt);
792 r->entry_count = t->entry_count;
793 r->delta_depth = t->delta_depth;
794 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
795 release_tree_content(t);
796 return r;
799 static struct tree_entry *new_tree_entry(void)
801 struct tree_entry *e;
803 if (!avail_tree_entry) {
804 unsigned int n = tree_entry_alloc;
805 tree_entry_allocd += n * sizeof(struct tree_entry);
806 ALLOC_ARRAY(e, n);
807 avail_tree_entry = e;
808 while (n-- > 1) {
809 *((void**)e) = e + 1;
810 e++;
812 *((void**)e) = NULL;
815 e = avail_tree_entry;
816 avail_tree_entry = *((void**)e);
817 return e;
820 static void release_tree_entry(struct tree_entry *e)
822 if (e->tree)
823 release_tree_content_recursive(e->tree);
824 *((void**)e) = avail_tree_entry;
825 avail_tree_entry = e;
828 static struct tree_content *dup_tree_content(struct tree_content *s)
830 struct tree_content *d;
831 struct tree_entry *a, *b;
832 unsigned int i;
834 if (!s)
835 return NULL;
836 d = new_tree_content(s->entry_count);
837 for (i = 0; i < s->entry_count; i++) {
838 a = s->entries[i];
839 b = new_tree_entry();
840 memcpy(b, a, sizeof(*a));
841 if (a->tree && is_null_oid(&b->versions[1].oid))
842 b->tree = dup_tree_content(a->tree);
843 else
844 b->tree = NULL;
845 d->entries[i] = b;
847 d->entry_count = s->entry_count;
848 d->delta_depth = s->delta_depth;
850 return d;
853 static void start_packfile(void)
855 struct strbuf tmp_file = STRBUF_INIT;
856 struct packed_git *p;
857 struct pack_header hdr;
858 int pack_fd;
860 pack_fd = odb_mkstemp(&tmp_file, "pack/tmp_pack_XXXXXX");
861 FLEX_ALLOC_STR(p, pack_name, tmp_file.buf);
862 strbuf_release(&tmp_file);
864 p->pack_fd = pack_fd;
865 p->do_not_close = 1;
866 pack_file = hashfd(pack_fd, p->pack_name);
868 hdr.hdr_signature = htonl(PACK_SIGNATURE);
869 hdr.hdr_version = htonl(2);
870 hdr.hdr_entries = 0;
871 hashwrite(pack_file, &hdr, sizeof(hdr));
873 pack_data = p;
874 pack_size = sizeof(hdr);
875 object_count = 0;
877 REALLOC_ARRAY(all_packs, pack_id + 1);
878 all_packs[pack_id] = p;
881 static const char *create_index(void)
883 const char *tmpfile;
884 struct pack_idx_entry **idx, **c, **last;
885 struct object_entry *e;
886 struct object_entry_pool *o;
888 /* Build the table of object IDs. */
889 ALLOC_ARRAY(idx, object_count);
890 c = idx;
891 for (o = blocks; o; o = o->next_pool)
892 for (e = o->next_free; e-- != o->entries;)
893 if (pack_id == e->pack_id)
894 *c++ = &e->idx;
895 last = idx + object_count;
896 if (c != last)
897 die("internal consistency error creating the index");
899 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts, pack_data->sha1);
900 free(idx);
901 return tmpfile;
904 static char *keep_pack(const char *curr_index_name)
906 static const char *keep_msg = "fast-import";
907 struct strbuf name = STRBUF_INIT;
908 int keep_fd;
910 odb_pack_name(&name, pack_data->sha1, "keep");
911 keep_fd = odb_pack_keep(name.buf);
912 if (keep_fd < 0)
913 die_errno("cannot create keep file");
914 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
915 if (close(keep_fd))
916 die_errno("failed to write keep file");
918 odb_pack_name(&name, pack_data->sha1, "pack");
919 if (finalize_object_file(pack_data->pack_name, name.buf))
920 die("cannot store pack file");
922 odb_pack_name(&name, pack_data->sha1, "idx");
923 if (finalize_object_file(curr_index_name, name.buf))
924 die("cannot store index file");
925 free((void *)curr_index_name);
926 return strbuf_detach(&name, NULL);
929 static void unkeep_all_packs(void)
931 struct strbuf name = STRBUF_INIT;
932 int k;
934 for (k = 0; k < pack_id; k++) {
935 struct packed_git *p = all_packs[k];
936 odb_pack_name(&name, p->sha1, "keep");
937 unlink_or_warn(name.buf);
939 strbuf_release(&name);
942 static int loosen_small_pack(const struct packed_git *p)
944 struct child_process unpack = CHILD_PROCESS_INIT;
946 if (lseek(p->pack_fd, 0, SEEK_SET) < 0)
947 die_errno("Failed seeking to start of '%s'", p->pack_name);
949 unpack.in = p->pack_fd;
950 unpack.git_cmd = 1;
951 unpack.stdout_to_stderr = 1;
952 argv_array_push(&unpack.args, "unpack-objects");
953 if (!show_stats)
954 argv_array_push(&unpack.args, "-q");
956 return run_command(&unpack);
959 static void end_packfile(void)
961 static int running;
963 if (running || !pack_data)
964 return;
966 running = 1;
967 clear_delta_base_cache();
968 if (object_count) {
969 struct packed_git *new_p;
970 struct object_id cur_pack_oid;
971 char *idx_name;
972 int i;
973 struct branch *b;
974 struct tag *t;
976 close_pack_windows(pack_data);
977 finalize_hashfile(pack_file, cur_pack_oid.hash, 0);
978 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
979 pack_data->pack_name, object_count,
980 cur_pack_oid.hash, pack_size);
982 if (object_count <= unpack_limit) {
983 if (!loosen_small_pack(pack_data)) {
984 invalidate_pack_id(pack_id);
985 goto discard_pack;
989 close(pack_data->pack_fd);
990 idx_name = keep_pack(create_index());
992 /* Register the packfile with core git's machinery. */
993 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
994 if (!new_p)
995 die("core git rejected index %s", idx_name);
996 all_packs[pack_id] = new_p;
997 install_packed_git(the_repository, new_p);
998 free(idx_name);
1000 /* Print the boundary */
1001 if (pack_edges) {
1002 fprintf(pack_edges, "%s:", new_p->pack_name);
1003 for (i = 0; i < branch_table_sz; i++) {
1004 for (b = branch_table[i]; b; b = b->table_next_branch) {
1005 if (b->pack_id == pack_id)
1006 fprintf(pack_edges, " %s",
1007 oid_to_hex(&b->oid));
1010 for (t = first_tag; t; t = t->next_tag) {
1011 if (t->pack_id == pack_id)
1012 fprintf(pack_edges, " %s",
1013 oid_to_hex(&t->oid));
1015 fputc('\n', pack_edges);
1016 fflush(pack_edges);
1019 pack_id++;
1021 else {
1022 discard_pack:
1023 close(pack_data->pack_fd);
1024 unlink_or_warn(pack_data->pack_name);
1026 FREE_AND_NULL(pack_data);
1027 running = 0;
1029 /* We can't carry a delta across packfiles. */
1030 strbuf_release(&last_blob.data);
1031 last_blob.offset = 0;
1032 last_blob.depth = 0;
1035 static void cycle_packfile(void)
1037 end_packfile();
1038 start_packfile();
1041 static int store_object(
1042 enum object_type type,
1043 struct strbuf *dat,
1044 struct last_object *last,
1045 struct object_id *oidout,
1046 uintmax_t mark)
1048 void *out, *delta;
1049 struct object_entry *e;
1050 unsigned char hdr[96];
1051 struct object_id oid;
1052 unsigned long hdrlen, deltalen;
1053 git_hash_ctx c;
1054 git_zstream s;
1056 hdrlen = xsnprintf((char *)hdr, sizeof(hdr), "%s %lu",
1057 type_name(type), (unsigned long)dat->len) + 1;
1058 the_hash_algo->init_fn(&c);
1059 the_hash_algo->update_fn(&c, hdr, hdrlen);
1060 the_hash_algo->update_fn(&c, dat->buf, dat->len);
1061 the_hash_algo->final_fn(oid.hash, &c);
1062 if (oidout)
1063 oidcpy(oidout, &oid);
1065 e = insert_object(&oid);
1066 if (mark)
1067 insert_mark(mark, e);
1068 if (e->idx.offset) {
1069 duplicate_count_by_type[type]++;
1070 return 1;
1071 } else if (find_sha1_pack(oid.hash,
1072 get_all_packs(the_repository))) {
1073 e->type = type;
1074 e->pack_id = MAX_PACK_ID;
1075 e->idx.offset = 1; /* just not zero! */
1076 duplicate_count_by_type[type]++;
1077 return 1;
1080 if (last && last->data.len && last->data.buf && last->depth < max_depth
1081 && dat->len > the_hash_algo->rawsz) {
1083 delta_count_attempts_by_type[type]++;
1084 delta = diff_delta(last->data.buf, last->data.len,
1085 dat->buf, dat->len,
1086 &deltalen, dat->len - the_hash_algo->rawsz);
1087 } else
1088 delta = NULL;
1090 git_deflate_init(&s, pack_compression_level);
1091 if (delta) {
1092 s.next_in = delta;
1093 s.avail_in = deltalen;
1094 } else {
1095 s.next_in = (void *)dat->buf;
1096 s.avail_in = dat->len;
1098 s.avail_out = git_deflate_bound(&s, s.avail_in);
1099 s.next_out = out = xmalloc(s.avail_out);
1100 while (git_deflate(&s, Z_FINISH) == Z_OK)
1101 ; /* nothing */
1102 git_deflate_end(&s);
1104 /* Determine if we should auto-checkpoint. */
1105 if ((max_packsize && (pack_size + 60 + s.total_out) > max_packsize)
1106 || (pack_size + 60 + s.total_out) < pack_size) {
1108 /* This new object needs to *not* have the current pack_id. */
1109 e->pack_id = pack_id + 1;
1110 cycle_packfile();
1112 /* We cannot carry a delta into the new pack. */
1113 if (delta) {
1114 FREE_AND_NULL(delta);
1116 git_deflate_init(&s, pack_compression_level);
1117 s.next_in = (void *)dat->buf;
1118 s.avail_in = dat->len;
1119 s.avail_out = git_deflate_bound(&s, s.avail_in);
1120 s.next_out = out = xrealloc(out, s.avail_out);
1121 while (git_deflate(&s, Z_FINISH) == Z_OK)
1122 ; /* nothing */
1123 git_deflate_end(&s);
1127 e->type = type;
1128 e->pack_id = pack_id;
1129 e->idx.offset = pack_size;
1130 object_count++;
1131 object_count_by_type[type]++;
1133 crc32_begin(pack_file);
1135 if (delta) {
1136 off_t ofs = e->idx.offset - last->offset;
1137 unsigned pos = sizeof(hdr) - 1;
1139 delta_count_by_type[type]++;
1140 e->depth = last->depth + 1;
1142 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1143 OBJ_OFS_DELTA, deltalen);
1144 hashwrite(pack_file, hdr, hdrlen);
1145 pack_size += hdrlen;
1147 hdr[pos] = ofs & 127;
1148 while (ofs >>= 7)
1149 hdr[--pos] = 128 | (--ofs & 127);
1150 hashwrite(pack_file, hdr + pos, sizeof(hdr) - pos);
1151 pack_size += sizeof(hdr) - pos;
1152 } else {
1153 e->depth = 0;
1154 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1155 type, dat->len);
1156 hashwrite(pack_file, hdr, hdrlen);
1157 pack_size += hdrlen;
1160 hashwrite(pack_file, out, s.total_out);
1161 pack_size += s.total_out;
1163 e->idx.crc32 = crc32_end(pack_file);
1165 free(out);
1166 free(delta);
1167 if (last) {
1168 if (last->no_swap) {
1169 last->data = *dat;
1170 } else {
1171 strbuf_swap(&last->data, dat);
1173 last->offset = e->idx.offset;
1174 last->depth = e->depth;
1176 return 0;
1179 static void truncate_pack(struct hashfile_checkpoint *checkpoint)
1181 if (hashfile_truncate(pack_file, checkpoint))
1182 die_errno("cannot truncate pack to skip duplicate");
1183 pack_size = checkpoint->offset;
1186 static void stream_blob(uintmax_t len, struct object_id *oidout, uintmax_t mark)
1188 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1189 unsigned char *in_buf = xmalloc(in_sz);
1190 unsigned char *out_buf = xmalloc(out_sz);
1191 struct object_entry *e;
1192 struct object_id oid;
1193 unsigned long hdrlen;
1194 off_t offset;
1195 git_hash_ctx c;
1196 git_zstream s;
1197 struct hashfile_checkpoint checkpoint;
1198 int status = Z_OK;
1200 /* Determine if we should auto-checkpoint. */
1201 if ((max_packsize && (pack_size + 60 + len) > max_packsize)
1202 || (pack_size + 60 + len) < pack_size)
1203 cycle_packfile();
1205 hashfile_checkpoint(pack_file, &checkpoint);
1206 offset = checkpoint.offset;
1208 hdrlen = xsnprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1210 the_hash_algo->init_fn(&c);
1211 the_hash_algo->update_fn(&c, out_buf, hdrlen);
1213 crc32_begin(pack_file);
1215 git_deflate_init(&s, pack_compression_level);
1217 hdrlen = encode_in_pack_object_header(out_buf, out_sz, OBJ_BLOB, len);
1219 s.next_out = out_buf + hdrlen;
1220 s.avail_out = out_sz - hdrlen;
1222 while (status != Z_STREAM_END) {
1223 if (0 < len && !s.avail_in) {
1224 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1225 size_t n = fread(in_buf, 1, cnt, stdin);
1226 if (!n && feof(stdin))
1227 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1229 the_hash_algo->update_fn(&c, in_buf, n);
1230 s.next_in = in_buf;
1231 s.avail_in = n;
1232 len -= n;
1235 status = git_deflate(&s, len ? 0 : Z_FINISH);
1237 if (!s.avail_out || status == Z_STREAM_END) {
1238 size_t n = s.next_out - out_buf;
1239 hashwrite(pack_file, out_buf, n);
1240 pack_size += n;
1241 s.next_out = out_buf;
1242 s.avail_out = out_sz;
1245 switch (status) {
1246 case Z_OK:
1247 case Z_BUF_ERROR:
1248 case Z_STREAM_END:
1249 continue;
1250 default:
1251 die("unexpected deflate failure: %d", status);
1254 git_deflate_end(&s);
1255 the_hash_algo->final_fn(oid.hash, &c);
1257 if (oidout)
1258 oidcpy(oidout, &oid);
1260 e = insert_object(&oid);
1262 if (mark)
1263 insert_mark(mark, e);
1265 if (e->idx.offset) {
1266 duplicate_count_by_type[OBJ_BLOB]++;
1267 truncate_pack(&checkpoint);
1269 } else if (find_sha1_pack(oid.hash,
1270 get_all_packs(the_repository))) {
1271 e->type = OBJ_BLOB;
1272 e->pack_id = MAX_PACK_ID;
1273 e->idx.offset = 1; /* just not zero! */
1274 duplicate_count_by_type[OBJ_BLOB]++;
1275 truncate_pack(&checkpoint);
1277 } else {
1278 e->depth = 0;
1279 e->type = OBJ_BLOB;
1280 e->pack_id = pack_id;
1281 e->idx.offset = offset;
1282 e->idx.crc32 = crc32_end(pack_file);
1283 object_count++;
1284 object_count_by_type[OBJ_BLOB]++;
1287 free(in_buf);
1288 free(out_buf);
1291 /* All calls must be guarded by find_object() or find_mark() to
1292 * ensure the 'struct object_entry' passed was written by this
1293 * process instance. We unpack the entry by the offset, avoiding
1294 * the need for the corresponding .idx file. This unpacking rule
1295 * works because we only use OBJ_REF_DELTA within the packfiles
1296 * created by fast-import.
1298 * oe must not be NULL. Such an oe usually comes from giving
1299 * an unknown SHA-1 to find_object() or an undefined mark to
1300 * find_mark(). Callers must test for this condition and use
1301 * the standard read_sha1_file() when it happens.
1303 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1304 * find_mark(), where the mark was reloaded from an existing marks
1305 * file and is referencing an object that this fast-import process
1306 * instance did not write out to a packfile. Callers must test for
1307 * this condition and use read_sha1_file() instead.
1309 static void *gfi_unpack_entry(
1310 struct object_entry *oe,
1311 unsigned long *sizep)
1313 enum object_type type;
1314 struct packed_git *p = all_packs[oe->pack_id];
1315 if (p == pack_data && p->pack_size < (pack_size + the_hash_algo->rawsz)) {
1316 /* The object is stored in the packfile we are writing to
1317 * and we have modified it since the last time we scanned
1318 * back to read a previously written object. If an old
1319 * window covered [p->pack_size, p->pack_size + rawsz) its
1320 * data is stale and is not valid. Closing all windows
1321 * and updating the packfile length ensures we can read
1322 * the newly written data.
1324 close_pack_windows(p);
1325 hashflush(pack_file);
1327 /* We have to offer rawsz bytes additional on the end of
1328 * the packfile as the core unpacker code assumes the
1329 * footer is present at the file end and must promise
1330 * at least rawsz bytes within any window it maps. But
1331 * we don't actually create the footer here.
1333 p->pack_size = pack_size + the_hash_algo->rawsz;
1335 return unpack_entry(the_repository, p, oe->idx.offset, &type, sizep);
1338 static const char *get_mode(const char *str, uint16_t *modep)
1340 unsigned char c;
1341 uint16_t mode = 0;
1343 while ((c = *str++) != ' ') {
1344 if (c < '0' || c > '7')
1345 return NULL;
1346 mode = (mode << 3) + (c - '0');
1348 *modep = mode;
1349 return str;
1352 static void load_tree(struct tree_entry *root)
1354 struct object_id *oid = &root->versions[1].oid;
1355 struct object_entry *myoe;
1356 struct tree_content *t;
1357 unsigned long size;
1358 char *buf;
1359 const char *c;
1361 root->tree = t = new_tree_content(8);
1362 if (is_null_oid(oid))
1363 return;
1365 myoe = find_object(oid);
1366 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1367 if (myoe->type != OBJ_TREE)
1368 die("Not a tree: %s", oid_to_hex(oid));
1369 t->delta_depth = myoe->depth;
1370 buf = gfi_unpack_entry(myoe, &size);
1371 if (!buf)
1372 die("Can't load tree %s", oid_to_hex(oid));
1373 } else {
1374 enum object_type type;
1375 buf = read_object_file(oid, &type, &size);
1376 if (!buf || type != OBJ_TREE)
1377 die("Can't load tree %s", oid_to_hex(oid));
1380 c = buf;
1381 while (c != (buf + size)) {
1382 struct tree_entry *e = new_tree_entry();
1384 if (t->entry_count == t->entry_capacity)
1385 root->tree = t = grow_tree_content(t, t->entry_count);
1386 t->entries[t->entry_count++] = e;
1388 e->tree = NULL;
1389 c = get_mode(c, &e->versions[1].mode);
1390 if (!c)
1391 die("Corrupt mode in %s", oid_to_hex(oid));
1392 e->versions[0].mode = e->versions[1].mode;
1393 e->name = to_atom(c, strlen(c));
1394 c += e->name->str_len + 1;
1395 hashcpy(e->versions[0].oid.hash, (unsigned char *)c);
1396 hashcpy(e->versions[1].oid.hash, (unsigned char *)c);
1397 c += GIT_SHA1_RAWSZ;
1399 free(buf);
1402 static int tecmp0 (const void *_a, const void *_b)
1404 struct tree_entry *a = *((struct tree_entry**)_a);
1405 struct tree_entry *b = *((struct tree_entry**)_b);
1406 return base_name_compare(
1407 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1408 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1411 static int tecmp1 (const void *_a, const void *_b)
1413 struct tree_entry *a = *((struct tree_entry**)_a);
1414 struct tree_entry *b = *((struct tree_entry**)_b);
1415 return base_name_compare(
1416 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1417 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1420 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1422 size_t maxlen = 0;
1423 unsigned int i;
1425 if (!v)
1426 QSORT(t->entries, t->entry_count, tecmp0);
1427 else
1428 QSORT(t->entries, t->entry_count, tecmp1);
1430 for (i = 0; i < t->entry_count; i++) {
1431 if (t->entries[i]->versions[v].mode)
1432 maxlen += t->entries[i]->name->str_len + 34;
1435 strbuf_reset(b);
1436 strbuf_grow(b, maxlen);
1437 for (i = 0; i < t->entry_count; i++) {
1438 struct tree_entry *e = t->entries[i];
1439 if (!e->versions[v].mode)
1440 continue;
1441 strbuf_addf(b, "%o %s%c",
1442 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1443 e->name->str_dat, '\0');
1444 strbuf_add(b, e->versions[v].oid.hash, GIT_SHA1_RAWSZ);
1448 static void store_tree(struct tree_entry *root)
1450 struct tree_content *t;
1451 unsigned int i, j, del;
1452 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1453 struct object_entry *le = NULL;
1455 if (!is_null_oid(&root->versions[1].oid))
1456 return;
1458 if (!root->tree)
1459 load_tree(root);
1460 t = root->tree;
1462 for (i = 0; i < t->entry_count; i++) {
1463 if (t->entries[i]->tree)
1464 store_tree(t->entries[i]);
1467 if (!(root->versions[0].mode & NO_DELTA))
1468 le = find_object(&root->versions[0].oid);
1469 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1470 mktree(t, 0, &old_tree);
1471 lo.data = old_tree;
1472 lo.offset = le->idx.offset;
1473 lo.depth = t->delta_depth;
1476 mktree(t, 1, &new_tree);
1477 store_object(OBJ_TREE, &new_tree, &lo, &root->versions[1].oid, 0);
1479 t->delta_depth = lo.depth;
1480 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1481 struct tree_entry *e = t->entries[i];
1482 if (e->versions[1].mode) {
1483 e->versions[0].mode = e->versions[1].mode;
1484 oidcpy(&e->versions[0].oid, &e->versions[1].oid);
1485 t->entries[j++] = e;
1486 } else {
1487 release_tree_entry(e);
1488 del++;
1491 t->entry_count -= del;
1494 static void tree_content_replace(
1495 struct tree_entry *root,
1496 const struct object_id *oid,
1497 const uint16_t mode,
1498 struct tree_content *newtree)
1500 if (!S_ISDIR(mode))
1501 die("Root cannot be a non-directory");
1502 oidclr(&root->versions[0].oid);
1503 oidcpy(&root->versions[1].oid, oid);
1504 if (root->tree)
1505 release_tree_content_recursive(root->tree);
1506 root->tree = newtree;
1509 static int tree_content_set(
1510 struct tree_entry *root,
1511 const char *p,
1512 const struct object_id *oid,
1513 const uint16_t mode,
1514 struct tree_content *subtree)
1516 struct tree_content *t;
1517 const char *slash1;
1518 unsigned int i, n;
1519 struct tree_entry *e;
1521 slash1 = strchrnul(p, '/');
1522 n = slash1 - p;
1523 if (!n)
1524 die("Empty path component found in input");
1525 if (!*slash1 && !S_ISDIR(mode) && subtree)
1526 die("Non-directories cannot have subtrees");
1528 if (!root->tree)
1529 load_tree(root);
1530 t = root->tree;
1531 for (i = 0; i < t->entry_count; i++) {
1532 e = t->entries[i];
1533 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1534 if (!*slash1) {
1535 if (!S_ISDIR(mode)
1536 && e->versions[1].mode == mode
1537 && oideq(&e->versions[1].oid, oid))
1538 return 0;
1539 e->versions[1].mode = mode;
1540 oidcpy(&e->versions[1].oid, oid);
1541 if (e->tree)
1542 release_tree_content_recursive(e->tree);
1543 e->tree = subtree;
1546 * We need to leave e->versions[0].sha1 alone
1547 * to avoid modifying the preimage tree used
1548 * when writing out the parent directory.
1549 * But after replacing the subdir with a
1550 * completely different one, it's not a good
1551 * delta base any more, and besides, we've
1552 * thrown away the tree entries needed to
1553 * make a delta against it.
1555 * So let's just explicitly disable deltas
1556 * for the subtree.
1558 if (S_ISDIR(e->versions[0].mode))
1559 e->versions[0].mode |= NO_DELTA;
1561 oidclr(&root->versions[1].oid);
1562 return 1;
1564 if (!S_ISDIR(e->versions[1].mode)) {
1565 e->tree = new_tree_content(8);
1566 e->versions[1].mode = S_IFDIR;
1568 if (!e->tree)
1569 load_tree(e);
1570 if (tree_content_set(e, slash1 + 1, oid, mode, subtree)) {
1571 oidclr(&root->versions[1].oid);
1572 return 1;
1574 return 0;
1578 if (t->entry_count == t->entry_capacity)
1579 root->tree = t = grow_tree_content(t, t->entry_count);
1580 e = new_tree_entry();
1581 e->name = to_atom(p, n);
1582 e->versions[0].mode = 0;
1583 oidclr(&e->versions[0].oid);
1584 t->entries[t->entry_count++] = e;
1585 if (*slash1) {
1586 e->tree = new_tree_content(8);
1587 e->versions[1].mode = S_IFDIR;
1588 tree_content_set(e, slash1 + 1, oid, mode, subtree);
1589 } else {
1590 e->tree = subtree;
1591 e->versions[1].mode = mode;
1592 oidcpy(&e->versions[1].oid, oid);
1594 oidclr(&root->versions[1].oid);
1595 return 1;
1598 static int tree_content_remove(
1599 struct tree_entry *root,
1600 const char *p,
1601 struct tree_entry *backup_leaf,
1602 int allow_root)
1604 struct tree_content *t;
1605 const char *slash1;
1606 unsigned int i, n;
1607 struct tree_entry *e;
1609 slash1 = strchrnul(p, '/');
1610 n = slash1 - p;
1612 if (!root->tree)
1613 load_tree(root);
1615 if (!*p && allow_root) {
1616 e = root;
1617 goto del_entry;
1620 t = root->tree;
1621 for (i = 0; i < t->entry_count; i++) {
1622 e = t->entries[i];
1623 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1624 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1626 * If p names a file in some subdirectory, and a
1627 * file or symlink matching the name of the
1628 * parent directory of p exists, then p cannot
1629 * exist and need not be deleted.
1631 return 1;
1632 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1633 goto del_entry;
1634 if (!e->tree)
1635 load_tree(e);
1636 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1637 for (n = 0; n < e->tree->entry_count; n++) {
1638 if (e->tree->entries[n]->versions[1].mode) {
1639 oidclr(&root->versions[1].oid);
1640 return 1;
1643 backup_leaf = NULL;
1644 goto del_entry;
1646 return 0;
1649 return 0;
1651 del_entry:
1652 if (backup_leaf)
1653 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1654 else if (e->tree)
1655 release_tree_content_recursive(e->tree);
1656 e->tree = NULL;
1657 e->versions[1].mode = 0;
1658 oidclr(&e->versions[1].oid);
1659 oidclr(&root->versions[1].oid);
1660 return 1;
1663 static int tree_content_get(
1664 struct tree_entry *root,
1665 const char *p,
1666 struct tree_entry *leaf,
1667 int allow_root)
1669 struct tree_content *t;
1670 const char *slash1;
1671 unsigned int i, n;
1672 struct tree_entry *e;
1674 slash1 = strchrnul(p, '/');
1675 n = slash1 - p;
1676 if (!n && !allow_root)
1677 die("Empty path component found in input");
1679 if (!root->tree)
1680 load_tree(root);
1682 if (!n) {
1683 e = root;
1684 goto found_entry;
1687 t = root->tree;
1688 for (i = 0; i < t->entry_count; i++) {
1689 e = t->entries[i];
1690 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1691 if (!*slash1)
1692 goto found_entry;
1693 if (!S_ISDIR(e->versions[1].mode))
1694 return 0;
1695 if (!e->tree)
1696 load_tree(e);
1697 return tree_content_get(e, slash1 + 1, leaf, 0);
1700 return 0;
1702 found_entry:
1703 memcpy(leaf, e, sizeof(*leaf));
1704 if (e->tree && is_null_oid(&e->versions[1].oid))
1705 leaf->tree = dup_tree_content(e->tree);
1706 else
1707 leaf->tree = NULL;
1708 return 1;
1711 static int update_branch(struct branch *b)
1713 static const char *msg = "fast-import";
1714 struct ref_transaction *transaction;
1715 struct object_id old_oid;
1716 struct strbuf err = STRBUF_INIT;
1718 if (is_null_oid(&b->oid)) {
1719 if (b->delete)
1720 delete_ref(NULL, b->name, NULL, 0);
1721 return 0;
1723 if (read_ref(b->name, &old_oid))
1724 oidclr(&old_oid);
1725 if (!force_update && !is_null_oid(&old_oid)) {
1726 struct commit *old_cmit, *new_cmit;
1728 old_cmit = lookup_commit_reference_gently(the_repository,
1729 &old_oid, 0);
1730 new_cmit = lookup_commit_reference_gently(the_repository,
1731 &b->oid, 0);
1732 if (!old_cmit || !new_cmit)
1733 return error("Branch %s is missing commits.", b->name);
1735 if (!in_merge_bases(old_cmit, new_cmit)) {
1736 warning("Not updating %s"
1737 " (new tip %s does not contain %s)",
1738 b->name, oid_to_hex(&b->oid),
1739 oid_to_hex(&old_oid));
1740 return -1;
1743 transaction = ref_transaction_begin(&err);
1744 if (!transaction ||
1745 ref_transaction_update(transaction, b->name, &b->oid, &old_oid,
1746 0, msg, &err) ||
1747 ref_transaction_commit(transaction, &err)) {
1748 ref_transaction_free(transaction);
1749 error("%s", err.buf);
1750 strbuf_release(&err);
1751 return -1;
1753 ref_transaction_free(transaction);
1754 strbuf_release(&err);
1755 return 0;
1758 static void dump_branches(void)
1760 unsigned int i;
1761 struct branch *b;
1763 for (i = 0; i < branch_table_sz; i++) {
1764 for (b = branch_table[i]; b; b = b->table_next_branch)
1765 failure |= update_branch(b);
1769 static void dump_tags(void)
1771 static const char *msg = "fast-import";
1772 struct tag *t;
1773 struct strbuf ref_name = STRBUF_INIT;
1774 struct strbuf err = STRBUF_INIT;
1775 struct ref_transaction *transaction;
1777 transaction = ref_transaction_begin(&err);
1778 if (!transaction) {
1779 failure |= error("%s", err.buf);
1780 goto cleanup;
1782 for (t = first_tag; t; t = t->next_tag) {
1783 strbuf_reset(&ref_name);
1784 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1786 if (ref_transaction_update(transaction, ref_name.buf,
1787 &t->oid, NULL, 0, msg, &err)) {
1788 failure |= error("%s", err.buf);
1789 goto cleanup;
1792 if (ref_transaction_commit(transaction, &err))
1793 failure |= error("%s", err.buf);
1795 cleanup:
1796 ref_transaction_free(transaction);
1797 strbuf_release(&ref_name);
1798 strbuf_release(&err);
1801 static void dump_marks_helper(FILE *f,
1802 uintmax_t base,
1803 struct mark_set *m)
1805 uintmax_t k;
1806 if (m->shift) {
1807 for (k = 0; k < 1024; k++) {
1808 if (m->data.sets[k])
1809 dump_marks_helper(f, base + (k << m->shift),
1810 m->data.sets[k]);
1812 } else {
1813 for (k = 0; k < 1024; k++) {
1814 if (m->data.marked[k])
1815 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1816 oid_to_hex(&m->data.marked[k]->idx.oid));
1821 static void dump_marks(void)
1823 struct lock_file mark_lock = LOCK_INIT;
1824 FILE *f;
1826 if (!export_marks_file || (import_marks_file && !import_marks_file_done))
1827 return;
1829 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1830 failure |= error_errno("Unable to write marks file %s",
1831 export_marks_file);
1832 return;
1835 f = fdopen_lock_file(&mark_lock, "w");
1836 if (!f) {
1837 int saved_errno = errno;
1838 rollback_lock_file(&mark_lock);
1839 failure |= error("Unable to write marks file %s: %s",
1840 export_marks_file, strerror(saved_errno));
1841 return;
1844 dump_marks_helper(f, 0, marks);
1845 if (commit_lock_file(&mark_lock)) {
1846 failure |= error_errno("Unable to write file %s",
1847 export_marks_file);
1848 return;
1852 static void read_marks(void)
1854 char line[512];
1855 FILE *f = fopen(import_marks_file, "r");
1856 if (f)
1858 else if (import_marks_file_ignore_missing && errno == ENOENT)
1859 goto done; /* Marks file does not exist */
1860 else
1861 die_errno("cannot read '%s'", import_marks_file);
1862 while (fgets(line, sizeof(line), f)) {
1863 uintmax_t mark;
1864 char *end;
1865 struct object_id oid;
1866 struct object_entry *e;
1868 end = strchr(line, '\n');
1869 if (line[0] != ':' || !end)
1870 die("corrupt mark line: %s", line);
1871 *end = 0;
1872 mark = strtoumax(line + 1, &end, 10);
1873 if (!mark || end == line + 1
1874 || *end != ' ' || get_oid_hex(end + 1, &oid))
1875 die("corrupt mark line: %s", line);
1876 e = find_object(&oid);
1877 if (!e) {
1878 enum object_type type = oid_object_info(the_repository,
1879 &oid, NULL);
1880 if (type < 0)
1881 die("object not found: %s", oid_to_hex(&oid));
1882 e = insert_object(&oid);
1883 e->type = type;
1884 e->pack_id = MAX_PACK_ID;
1885 e->idx.offset = 1; /* just not zero! */
1887 insert_mark(mark, e);
1889 fclose(f);
1890 done:
1891 import_marks_file_done = 1;
1895 static int read_next_command(void)
1897 static int stdin_eof = 0;
1899 if (stdin_eof) {
1900 unread_command_buf = 0;
1901 return EOF;
1904 for (;;) {
1905 const char *p;
1907 if (unread_command_buf) {
1908 unread_command_buf = 0;
1909 } else {
1910 struct recent_command *rc;
1912 strbuf_detach(&command_buf, NULL);
1913 stdin_eof = strbuf_getline_lf(&command_buf, stdin);
1914 if (stdin_eof)
1915 return EOF;
1917 if (!seen_data_command
1918 && !starts_with(command_buf.buf, "feature ")
1919 && !starts_with(command_buf.buf, "option ")) {
1920 parse_argv();
1923 rc = rc_free;
1924 if (rc)
1925 rc_free = rc->next;
1926 else {
1927 rc = cmd_hist.next;
1928 cmd_hist.next = rc->next;
1929 cmd_hist.next->prev = &cmd_hist;
1930 free(rc->buf);
1933 rc->buf = command_buf.buf;
1934 rc->prev = cmd_tail;
1935 rc->next = cmd_hist.prev;
1936 rc->prev->next = rc;
1937 cmd_tail = rc;
1939 if (skip_prefix(command_buf.buf, "get-mark ", &p)) {
1940 parse_get_mark(p);
1941 continue;
1943 if (skip_prefix(command_buf.buf, "cat-blob ", &p)) {
1944 parse_cat_blob(p);
1945 continue;
1947 if (command_buf.buf[0] == '#')
1948 continue;
1949 return 0;
1953 static void skip_optional_lf(void)
1955 int term_char = fgetc(stdin);
1956 if (term_char != '\n' && term_char != EOF)
1957 ungetc(term_char, stdin);
1960 static void parse_mark(void)
1962 const char *v;
1963 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1964 next_mark = strtoumax(v, NULL, 10);
1965 read_next_command();
1967 else
1968 next_mark = 0;
1971 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1973 const char *data;
1974 strbuf_reset(sb);
1976 if (!skip_prefix(command_buf.buf, "data ", &data))
1977 die("Expected 'data n' command, found: %s", command_buf.buf);
1979 if (skip_prefix(data, "<<", &data)) {
1980 char *term = xstrdup(data);
1981 size_t term_len = command_buf.len - (data - command_buf.buf);
1983 strbuf_detach(&command_buf, NULL);
1984 for (;;) {
1985 if (strbuf_getline_lf(&command_buf, stdin) == EOF)
1986 die("EOF in data (terminator '%s' not found)", term);
1987 if (term_len == command_buf.len
1988 && !strcmp(term, command_buf.buf))
1989 break;
1990 strbuf_addbuf(sb, &command_buf);
1991 strbuf_addch(sb, '\n');
1993 free(term);
1995 else {
1996 uintmax_t len = strtoumax(data, NULL, 10);
1997 size_t n = 0, length = (size_t)len;
1999 if (limit && limit < len) {
2000 *len_res = len;
2001 return 0;
2003 if (length < len)
2004 die("data is too large to use in this context");
2006 while (n < length) {
2007 size_t s = strbuf_fread(sb, length - n, stdin);
2008 if (!s && feof(stdin))
2009 die("EOF in data (%lu bytes remaining)",
2010 (unsigned long)(length - n));
2011 n += s;
2015 skip_optional_lf();
2016 return 1;
2019 static int validate_raw_date(const char *src, struct strbuf *result)
2021 const char *orig_src = src;
2022 char *endp;
2023 unsigned long num;
2025 errno = 0;
2027 num = strtoul(src, &endp, 10);
2028 /* NEEDSWORK: perhaps check for reasonable values? */
2029 if (errno || endp == src || *endp != ' ')
2030 return -1;
2032 src = endp + 1;
2033 if (*src != '-' && *src != '+')
2034 return -1;
2036 num = strtoul(src + 1, &endp, 10);
2037 if (errno || endp == src + 1 || *endp || 1400 < num)
2038 return -1;
2040 strbuf_addstr(result, orig_src);
2041 return 0;
2044 static char *parse_ident(const char *buf)
2046 const char *ltgt;
2047 size_t name_len;
2048 struct strbuf ident = STRBUF_INIT;
2050 /* ensure there is a space delimiter even if there is no name */
2051 if (*buf == '<')
2052 --buf;
2054 ltgt = buf + strcspn(buf, "<>");
2055 if (*ltgt != '<')
2056 die("Missing < in ident string: %s", buf);
2057 if (ltgt != buf && ltgt[-1] != ' ')
2058 die("Missing space before < in ident string: %s", buf);
2059 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
2060 if (*ltgt != '>')
2061 die("Missing > in ident string: %s", buf);
2062 ltgt++;
2063 if (*ltgt != ' ')
2064 die("Missing space after > in ident string: %s", buf);
2065 ltgt++;
2066 name_len = ltgt - buf;
2067 strbuf_add(&ident, buf, name_len);
2069 switch (whenspec) {
2070 case WHENSPEC_RAW:
2071 if (validate_raw_date(ltgt, &ident) < 0)
2072 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
2073 break;
2074 case WHENSPEC_RFC2822:
2075 if (parse_date(ltgt, &ident) < 0)
2076 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
2077 break;
2078 case WHENSPEC_NOW:
2079 if (strcmp("now", ltgt))
2080 die("Date in ident must be 'now': %s", buf);
2081 datestamp(&ident);
2082 break;
2085 return strbuf_detach(&ident, NULL);
2088 static void parse_and_store_blob(
2089 struct last_object *last,
2090 struct object_id *oidout,
2091 uintmax_t mark)
2093 static struct strbuf buf = STRBUF_INIT;
2094 uintmax_t len;
2096 if (parse_data(&buf, big_file_threshold, &len))
2097 store_object(OBJ_BLOB, &buf, last, oidout, mark);
2098 else {
2099 if (last) {
2100 strbuf_release(&last->data);
2101 last->offset = 0;
2102 last->depth = 0;
2104 stream_blob(len, oidout, mark);
2105 skip_optional_lf();
2109 static void parse_new_blob(void)
2111 read_next_command();
2112 parse_mark();
2113 parse_and_store_blob(&last_blob, NULL, next_mark);
2116 static void unload_one_branch(void)
2118 while (cur_active_branches
2119 && cur_active_branches >= max_active_branches) {
2120 uintmax_t min_commit = ULONG_MAX;
2121 struct branch *e, *l = NULL, *p = NULL;
2123 for (e = active_branches; e; e = e->active_next_branch) {
2124 if (e->last_commit < min_commit) {
2125 p = l;
2126 min_commit = e->last_commit;
2128 l = e;
2131 if (p) {
2132 e = p->active_next_branch;
2133 p->active_next_branch = e->active_next_branch;
2134 } else {
2135 e = active_branches;
2136 active_branches = e->active_next_branch;
2138 e->active = 0;
2139 e->active_next_branch = NULL;
2140 if (e->branch_tree.tree) {
2141 release_tree_content_recursive(e->branch_tree.tree);
2142 e->branch_tree.tree = NULL;
2144 cur_active_branches--;
2148 static void load_branch(struct branch *b)
2150 load_tree(&b->branch_tree);
2151 if (!b->active) {
2152 b->active = 1;
2153 b->active_next_branch = active_branches;
2154 active_branches = b;
2155 cur_active_branches++;
2156 branch_load_count++;
2160 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2162 unsigned char fanout = 0;
2163 while ((num_notes >>= 8))
2164 fanout++;
2165 return fanout;
2168 static void construct_path_with_fanout(const char *hex_sha1,
2169 unsigned char fanout, char *path)
2171 unsigned int i = 0, j = 0;
2172 if (fanout >= the_hash_algo->rawsz)
2173 die("Too large fanout (%u)", fanout);
2174 while (fanout) {
2175 path[i++] = hex_sha1[j++];
2176 path[i++] = hex_sha1[j++];
2177 path[i++] = '/';
2178 fanout--;
2180 memcpy(path + i, hex_sha1 + j, the_hash_algo->hexsz - j);
2181 path[i + the_hash_algo->hexsz - j] = '\0';
2184 static uintmax_t do_change_note_fanout(
2185 struct tree_entry *orig_root, struct tree_entry *root,
2186 char *hex_oid, unsigned int hex_oid_len,
2187 char *fullpath, unsigned int fullpath_len,
2188 unsigned char fanout)
2190 struct tree_content *t;
2191 struct tree_entry *e, leaf;
2192 unsigned int i, tmp_hex_oid_len, tmp_fullpath_len;
2193 uintmax_t num_notes = 0;
2194 struct object_id oid;
2195 char realpath[60];
2197 if (!root->tree)
2198 load_tree(root);
2199 t = root->tree;
2201 for (i = 0; t && i < t->entry_count; i++) {
2202 e = t->entries[i];
2203 tmp_hex_oid_len = hex_oid_len + e->name->str_len;
2204 tmp_fullpath_len = fullpath_len;
2207 * We're interested in EITHER existing note entries (entries
2208 * with exactly 40 hex chars in path, not including directory
2209 * separators), OR directory entries that may contain note
2210 * entries (with < 40 hex chars in path).
2211 * Also, each path component in a note entry must be a multiple
2212 * of 2 chars.
2214 if (!e->versions[1].mode ||
2215 tmp_hex_oid_len > GIT_SHA1_HEXSZ ||
2216 e->name->str_len % 2)
2217 continue;
2219 /* This _may_ be a note entry, or a subdir containing notes */
2220 memcpy(hex_oid + hex_oid_len, e->name->str_dat,
2221 e->name->str_len);
2222 if (tmp_fullpath_len)
2223 fullpath[tmp_fullpath_len++] = '/';
2224 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2225 e->name->str_len);
2226 tmp_fullpath_len += e->name->str_len;
2227 fullpath[tmp_fullpath_len] = '\0';
2229 if (tmp_hex_oid_len == GIT_SHA1_HEXSZ && !get_oid_hex(hex_oid, &oid)) {
2230 /* This is a note entry */
2231 if (fanout == 0xff) {
2232 /* Counting mode, no rename */
2233 num_notes++;
2234 continue;
2236 construct_path_with_fanout(hex_oid, fanout, realpath);
2237 if (!strcmp(fullpath, realpath)) {
2238 /* Note entry is in correct location */
2239 num_notes++;
2240 continue;
2243 /* Rename fullpath to realpath */
2244 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2245 die("Failed to remove path %s", fullpath);
2246 tree_content_set(orig_root, realpath,
2247 &leaf.versions[1].oid,
2248 leaf.versions[1].mode,
2249 leaf.tree);
2250 } else if (S_ISDIR(e->versions[1].mode)) {
2251 /* This is a subdir that may contain note entries */
2252 num_notes += do_change_note_fanout(orig_root, e,
2253 hex_oid, tmp_hex_oid_len,
2254 fullpath, tmp_fullpath_len, fanout);
2257 /* The above may have reallocated the current tree_content */
2258 t = root->tree;
2260 return num_notes;
2263 static uintmax_t change_note_fanout(struct tree_entry *root,
2264 unsigned char fanout)
2267 * The size of path is due to one slash between every two hex digits,
2268 * plus the terminating NUL. Note that there is no slash at the end, so
2269 * the number of slashes is one less than half the number of hex
2270 * characters.
2272 char hex_oid[GIT_MAX_HEXSZ], path[GIT_MAX_HEXSZ + (GIT_MAX_HEXSZ / 2) - 1 + 1];
2273 return do_change_note_fanout(root, root, hex_oid, 0, path, 0, fanout);
2277 * Given a pointer into a string, parse a mark reference:
2279 * idnum ::= ':' bigint;
2281 * Return the first character after the value in *endptr.
2283 * Complain if the following character is not what is expected,
2284 * either a space or end of the string.
2286 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2288 uintmax_t mark;
2290 assert(*p == ':');
2291 p++;
2292 mark = strtoumax(p, endptr, 10);
2293 if (*endptr == p)
2294 die("No value after ':' in mark: %s", command_buf.buf);
2295 return mark;
2299 * Parse the mark reference, and complain if this is not the end of
2300 * the string.
2302 static uintmax_t parse_mark_ref_eol(const char *p)
2304 char *end;
2305 uintmax_t mark;
2307 mark = parse_mark_ref(p, &end);
2308 if (*end != '\0')
2309 die("Garbage after mark: %s", command_buf.buf);
2310 return mark;
2314 * Parse the mark reference, demanding a trailing space. Return a
2315 * pointer to the space.
2317 static uintmax_t parse_mark_ref_space(const char **p)
2319 uintmax_t mark;
2320 char *end;
2322 mark = parse_mark_ref(*p, &end);
2323 if (*end++ != ' ')
2324 die("Missing space after mark: %s", command_buf.buf);
2325 *p = end;
2326 return mark;
2329 static void file_change_m(const char *p, struct branch *b)
2331 static struct strbuf uq = STRBUF_INIT;
2332 const char *endp;
2333 struct object_entry *oe;
2334 struct object_id oid;
2335 uint16_t mode, inline_data = 0;
2337 p = get_mode(p, &mode);
2338 if (!p)
2339 die("Corrupt mode: %s", command_buf.buf);
2340 switch (mode) {
2341 case 0644:
2342 case 0755:
2343 mode |= S_IFREG;
2344 case S_IFREG | 0644:
2345 case S_IFREG | 0755:
2346 case S_IFLNK:
2347 case S_IFDIR:
2348 case S_IFGITLINK:
2349 /* ok */
2350 break;
2351 default:
2352 die("Corrupt mode: %s", command_buf.buf);
2355 if (*p == ':') {
2356 oe = find_mark(parse_mark_ref_space(&p));
2357 oidcpy(&oid, &oe->idx.oid);
2358 } else if (skip_prefix(p, "inline ", &p)) {
2359 inline_data = 1;
2360 oe = NULL; /* not used with inline_data, but makes gcc happy */
2361 } else {
2362 if (parse_oid_hex(p, &oid, &p))
2363 die("Invalid dataref: %s", command_buf.buf);
2364 oe = find_object(&oid);
2365 if (*p++ != ' ')
2366 die("Missing space after SHA1: %s", command_buf.buf);
2369 strbuf_reset(&uq);
2370 if (!unquote_c_style(&uq, p, &endp)) {
2371 if (*endp)
2372 die("Garbage after path in: %s", command_buf.buf);
2373 p = uq.buf;
2376 /* Git does not track empty, non-toplevel directories. */
2377 if (S_ISDIR(mode) && is_empty_tree_oid(&oid) && *p) {
2378 tree_content_remove(&b->branch_tree, p, NULL, 0);
2379 return;
2382 if (S_ISGITLINK(mode)) {
2383 if (inline_data)
2384 die("Git links cannot be specified 'inline': %s",
2385 command_buf.buf);
2386 else if (oe) {
2387 if (oe->type != OBJ_COMMIT)
2388 die("Not a commit (actually a %s): %s",
2389 type_name(oe->type), command_buf.buf);
2392 * Accept the sha1 without checking; it expected to be in
2393 * another repository.
2395 } else if (inline_data) {
2396 if (S_ISDIR(mode))
2397 die("Directories cannot be specified 'inline': %s",
2398 command_buf.buf);
2399 if (p != uq.buf) {
2400 strbuf_addstr(&uq, p);
2401 p = uq.buf;
2403 read_next_command();
2404 parse_and_store_blob(&last_blob, &oid, 0);
2405 } else {
2406 enum object_type expected = S_ISDIR(mode) ?
2407 OBJ_TREE: OBJ_BLOB;
2408 enum object_type type = oe ? oe->type :
2409 oid_object_info(the_repository, &oid,
2410 NULL);
2411 if (type < 0)
2412 die("%s not found: %s",
2413 S_ISDIR(mode) ? "Tree" : "Blob",
2414 command_buf.buf);
2415 if (type != expected)
2416 die("Not a %s (actually a %s): %s",
2417 type_name(expected), type_name(type),
2418 command_buf.buf);
2421 if (!*p) {
2422 tree_content_replace(&b->branch_tree, &oid, mode, NULL);
2423 return;
2425 tree_content_set(&b->branch_tree, p, &oid, mode, NULL);
2428 static void file_change_d(const char *p, struct branch *b)
2430 static struct strbuf uq = STRBUF_INIT;
2431 const char *endp;
2433 strbuf_reset(&uq);
2434 if (!unquote_c_style(&uq, p, &endp)) {
2435 if (*endp)
2436 die("Garbage after path in: %s", command_buf.buf);
2437 p = uq.buf;
2439 tree_content_remove(&b->branch_tree, p, NULL, 1);
2442 static void file_change_cr(const char *s, struct branch *b, int rename)
2444 const char *d;
2445 static struct strbuf s_uq = STRBUF_INIT;
2446 static struct strbuf d_uq = STRBUF_INIT;
2447 const char *endp;
2448 struct tree_entry leaf;
2450 strbuf_reset(&s_uq);
2451 if (!unquote_c_style(&s_uq, s, &endp)) {
2452 if (*endp != ' ')
2453 die("Missing space after source: %s", command_buf.buf);
2454 } else {
2455 endp = strchr(s, ' ');
2456 if (!endp)
2457 die("Missing space after source: %s", command_buf.buf);
2458 strbuf_add(&s_uq, s, endp - s);
2460 s = s_uq.buf;
2462 endp++;
2463 if (!*endp)
2464 die("Missing dest: %s", command_buf.buf);
2466 d = endp;
2467 strbuf_reset(&d_uq);
2468 if (!unquote_c_style(&d_uq, d, &endp)) {
2469 if (*endp)
2470 die("Garbage after dest in: %s", command_buf.buf);
2471 d = d_uq.buf;
2474 memset(&leaf, 0, sizeof(leaf));
2475 if (rename)
2476 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2477 else
2478 tree_content_get(&b->branch_tree, s, &leaf, 1);
2479 if (!leaf.versions[1].mode)
2480 die("Path %s not in branch", s);
2481 if (!*d) { /* C "path/to/subdir" "" */
2482 tree_content_replace(&b->branch_tree,
2483 &leaf.versions[1].oid,
2484 leaf.versions[1].mode,
2485 leaf.tree);
2486 return;
2488 tree_content_set(&b->branch_tree, d,
2489 &leaf.versions[1].oid,
2490 leaf.versions[1].mode,
2491 leaf.tree);
2494 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2496 static struct strbuf uq = STRBUF_INIT;
2497 struct object_entry *oe;
2498 struct branch *s;
2499 struct object_id oid, commit_oid;
2500 char path[60];
2501 uint16_t inline_data = 0;
2502 unsigned char new_fanout;
2505 * When loading a branch, we don't traverse its tree to count the real
2506 * number of notes (too expensive to do this for all non-note refs).
2507 * This means that recently loaded notes refs might incorrectly have
2508 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2510 * Fix this by traversing the tree and counting the number of notes
2511 * when b->num_notes == 0. If the notes tree is truly empty, the
2512 * calculation should not take long.
2514 if (b->num_notes == 0 && *old_fanout == 0) {
2515 /* Invoke change_note_fanout() in "counting mode". */
2516 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2517 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2520 /* Now parse the notemodify command. */
2521 /* <dataref> or 'inline' */
2522 if (*p == ':') {
2523 oe = find_mark(parse_mark_ref_space(&p));
2524 oidcpy(&oid, &oe->idx.oid);
2525 } else if (skip_prefix(p, "inline ", &p)) {
2526 inline_data = 1;
2527 oe = NULL; /* not used with inline_data, but makes gcc happy */
2528 } else {
2529 if (parse_oid_hex(p, &oid, &p))
2530 die("Invalid dataref: %s", command_buf.buf);
2531 oe = find_object(&oid);
2532 if (*p++ != ' ')
2533 die("Missing space after SHA1: %s", command_buf.buf);
2536 /* <commit-ish> */
2537 s = lookup_branch(p);
2538 if (s) {
2539 if (is_null_oid(&s->oid))
2540 die("Can't add a note on empty branch.");
2541 oidcpy(&commit_oid, &s->oid);
2542 } else if (*p == ':') {
2543 uintmax_t commit_mark = parse_mark_ref_eol(p);
2544 struct object_entry *commit_oe = find_mark(commit_mark);
2545 if (commit_oe->type != OBJ_COMMIT)
2546 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2547 oidcpy(&commit_oid, &commit_oe->idx.oid);
2548 } else if (!get_oid(p, &commit_oid)) {
2549 unsigned long size;
2550 char *buf = read_object_with_reference(&commit_oid,
2551 commit_type, &size,
2552 &commit_oid);
2553 if (!buf || size < 46)
2554 die("Not a valid commit: %s", p);
2555 free(buf);
2556 } else
2557 die("Invalid ref name or SHA1 expression: %s", p);
2559 if (inline_data) {
2560 if (p != uq.buf) {
2561 strbuf_addstr(&uq, p);
2562 p = uq.buf;
2564 read_next_command();
2565 parse_and_store_blob(&last_blob, &oid, 0);
2566 } else if (oe) {
2567 if (oe->type != OBJ_BLOB)
2568 die("Not a blob (actually a %s): %s",
2569 type_name(oe->type), command_buf.buf);
2570 } else if (!is_null_oid(&oid)) {
2571 enum object_type type = oid_object_info(the_repository, &oid,
2572 NULL);
2573 if (type < 0)
2574 die("Blob not found: %s", command_buf.buf);
2575 if (type != OBJ_BLOB)
2576 die("Not a blob (actually a %s): %s",
2577 type_name(type), command_buf.buf);
2580 construct_path_with_fanout(oid_to_hex(&commit_oid), *old_fanout, path);
2581 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2582 b->num_notes--;
2584 if (is_null_oid(&oid))
2585 return; /* nothing to insert */
2587 b->num_notes++;
2588 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2589 construct_path_with_fanout(oid_to_hex(&commit_oid), new_fanout, path);
2590 tree_content_set(&b->branch_tree, path, &oid, S_IFREG | 0644, NULL);
2593 static void file_change_deleteall(struct branch *b)
2595 release_tree_content_recursive(b->branch_tree.tree);
2596 oidclr(&b->branch_tree.versions[0].oid);
2597 oidclr(&b->branch_tree.versions[1].oid);
2598 load_tree(&b->branch_tree);
2599 b->num_notes = 0;
2602 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2604 if (!buf || size < GIT_SHA1_HEXSZ + 6)
2605 die("Not a valid commit: %s", oid_to_hex(&b->oid));
2606 if (memcmp("tree ", buf, 5)
2607 || get_oid_hex(buf + 5, &b->branch_tree.versions[1].oid))
2608 die("The commit %s is corrupt", oid_to_hex(&b->oid));
2609 oidcpy(&b->branch_tree.versions[0].oid,
2610 &b->branch_tree.versions[1].oid);
2613 static void parse_from_existing(struct branch *b)
2615 if (is_null_oid(&b->oid)) {
2616 oidclr(&b->branch_tree.versions[0].oid);
2617 oidclr(&b->branch_tree.versions[1].oid);
2618 } else {
2619 unsigned long size;
2620 char *buf;
2622 buf = read_object_with_reference(&b->oid, commit_type, &size,
2623 &b->oid);
2624 parse_from_commit(b, buf, size);
2625 free(buf);
2629 static int parse_from(struct branch *b)
2631 const char *from;
2632 struct branch *s;
2633 struct object_id oid;
2635 if (!skip_prefix(command_buf.buf, "from ", &from))
2636 return 0;
2638 oidcpy(&oid, &b->branch_tree.versions[1].oid);
2640 s = lookup_branch(from);
2641 if (b == s)
2642 die("Can't create a branch from itself: %s", b->name);
2643 else if (s) {
2644 struct object_id *t = &s->branch_tree.versions[1].oid;
2645 oidcpy(&b->oid, &s->oid);
2646 oidcpy(&b->branch_tree.versions[0].oid, t);
2647 oidcpy(&b->branch_tree.versions[1].oid, t);
2648 } else if (*from == ':') {
2649 uintmax_t idnum = parse_mark_ref_eol(from);
2650 struct object_entry *oe = find_mark(idnum);
2651 if (oe->type != OBJ_COMMIT)
2652 die("Mark :%" PRIuMAX " not a commit", idnum);
2653 if (!oideq(&b->oid, &oe->idx.oid)) {
2654 oidcpy(&b->oid, &oe->idx.oid);
2655 if (oe->pack_id != MAX_PACK_ID) {
2656 unsigned long size;
2657 char *buf = gfi_unpack_entry(oe, &size);
2658 parse_from_commit(b, buf, size);
2659 free(buf);
2660 } else
2661 parse_from_existing(b);
2663 } else if (!get_oid(from, &b->oid)) {
2664 parse_from_existing(b);
2665 if (is_null_oid(&b->oid))
2666 b->delete = 1;
2668 else
2669 die("Invalid ref name or SHA1 expression: %s", from);
2671 if (b->branch_tree.tree && !oideq(&oid, &b->branch_tree.versions[1].oid)) {
2672 release_tree_content_recursive(b->branch_tree.tree);
2673 b->branch_tree.tree = NULL;
2676 read_next_command();
2677 return 1;
2680 static struct hash_list *parse_merge(unsigned int *count)
2682 struct hash_list *list = NULL, **tail = &list, *n;
2683 const char *from;
2684 struct branch *s;
2686 *count = 0;
2687 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2688 n = xmalloc(sizeof(*n));
2689 s = lookup_branch(from);
2690 if (s)
2691 oidcpy(&n->oid, &s->oid);
2692 else if (*from == ':') {
2693 uintmax_t idnum = parse_mark_ref_eol(from);
2694 struct object_entry *oe = find_mark(idnum);
2695 if (oe->type != OBJ_COMMIT)
2696 die("Mark :%" PRIuMAX " not a commit", idnum);
2697 oidcpy(&n->oid, &oe->idx.oid);
2698 } else if (!get_oid(from, &n->oid)) {
2699 unsigned long size;
2700 char *buf = read_object_with_reference(&n->oid,
2701 commit_type,
2702 &size, &n->oid);
2703 if (!buf || size < 46)
2704 die("Not a valid commit: %s", from);
2705 free(buf);
2706 } else
2707 die("Invalid ref name or SHA1 expression: %s", from);
2709 n->next = NULL;
2710 *tail = n;
2711 tail = &n->next;
2713 (*count)++;
2714 read_next_command();
2716 return list;
2719 static void parse_new_commit(const char *arg)
2721 static struct strbuf msg = STRBUF_INIT;
2722 struct branch *b;
2723 char *author = NULL;
2724 char *committer = NULL;
2725 struct hash_list *merge_list = NULL;
2726 unsigned int merge_count;
2727 unsigned char prev_fanout, new_fanout;
2728 const char *v;
2730 b = lookup_branch(arg);
2731 if (!b)
2732 b = new_branch(arg);
2734 read_next_command();
2735 parse_mark();
2736 if (skip_prefix(command_buf.buf, "author ", &v)) {
2737 author = parse_ident(v);
2738 read_next_command();
2740 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2741 committer = parse_ident(v);
2742 read_next_command();
2744 if (!committer)
2745 die("Expected committer but didn't get one");
2746 parse_data(&msg, 0, NULL);
2747 read_next_command();
2748 parse_from(b);
2749 merge_list = parse_merge(&merge_count);
2751 /* ensure the branch is active/loaded */
2752 if (!b->branch_tree.tree || !max_active_branches) {
2753 unload_one_branch();
2754 load_branch(b);
2757 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2759 /* file_change* */
2760 while (command_buf.len > 0) {
2761 if (skip_prefix(command_buf.buf, "M ", &v))
2762 file_change_m(v, b);
2763 else if (skip_prefix(command_buf.buf, "D ", &v))
2764 file_change_d(v, b);
2765 else if (skip_prefix(command_buf.buf, "R ", &v))
2766 file_change_cr(v, b, 1);
2767 else if (skip_prefix(command_buf.buf, "C ", &v))
2768 file_change_cr(v, b, 0);
2769 else if (skip_prefix(command_buf.buf, "N ", &v))
2770 note_change_n(v, b, &prev_fanout);
2771 else if (!strcmp("deleteall", command_buf.buf))
2772 file_change_deleteall(b);
2773 else if (skip_prefix(command_buf.buf, "ls ", &v))
2774 parse_ls(v, b);
2775 else {
2776 unread_command_buf = 1;
2777 break;
2779 if (read_next_command() == EOF)
2780 break;
2783 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2784 if (new_fanout != prev_fanout)
2785 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2787 /* build the tree and the commit */
2788 store_tree(&b->branch_tree);
2789 oidcpy(&b->branch_tree.versions[0].oid,
2790 &b->branch_tree.versions[1].oid);
2792 strbuf_reset(&new_data);
2793 strbuf_addf(&new_data, "tree %s\n",
2794 oid_to_hex(&b->branch_tree.versions[1].oid));
2795 if (!is_null_oid(&b->oid))
2796 strbuf_addf(&new_data, "parent %s\n",
2797 oid_to_hex(&b->oid));
2798 while (merge_list) {
2799 struct hash_list *next = merge_list->next;
2800 strbuf_addf(&new_data, "parent %s\n",
2801 oid_to_hex(&merge_list->oid));
2802 free(merge_list);
2803 merge_list = next;
2805 strbuf_addf(&new_data,
2806 "author %s\n"
2807 "committer %s\n"
2808 "\n",
2809 author ? author : committer, committer);
2810 strbuf_addbuf(&new_data, &msg);
2811 free(author);
2812 free(committer);
2814 if (!store_object(OBJ_COMMIT, &new_data, NULL, &b->oid, next_mark))
2815 b->pack_id = pack_id;
2816 b->last_commit = object_count_by_type[OBJ_COMMIT];
2819 static void parse_new_tag(const char *arg)
2821 static struct strbuf msg = STRBUF_INIT;
2822 const char *from;
2823 char *tagger;
2824 struct branch *s;
2825 struct tag *t;
2826 uintmax_t from_mark = 0;
2827 struct object_id oid;
2828 enum object_type type;
2829 const char *v;
2831 t = mem_pool_alloc(&fi_mem_pool, sizeof(struct tag));
2832 memset(t, 0, sizeof(struct tag));
2833 t->name = pool_strdup(arg);
2834 if (last_tag)
2835 last_tag->next_tag = t;
2836 else
2837 first_tag = t;
2838 last_tag = t;
2839 read_next_command();
2841 /* from ... */
2842 if (!skip_prefix(command_buf.buf, "from ", &from))
2843 die("Expected from command, got %s", command_buf.buf);
2844 s = lookup_branch(from);
2845 if (s) {
2846 if (is_null_oid(&s->oid))
2847 die("Can't tag an empty branch.");
2848 oidcpy(&oid, &s->oid);
2849 type = OBJ_COMMIT;
2850 } else if (*from == ':') {
2851 struct object_entry *oe;
2852 from_mark = parse_mark_ref_eol(from);
2853 oe = find_mark(from_mark);
2854 type = oe->type;
2855 oidcpy(&oid, &oe->idx.oid);
2856 } else if (!get_oid(from, &oid)) {
2857 struct object_entry *oe = find_object(&oid);
2858 if (!oe) {
2859 type = oid_object_info(the_repository, &oid, NULL);
2860 if (type < 0)
2861 die("Not a valid object: %s", from);
2862 } else
2863 type = oe->type;
2864 } else
2865 die("Invalid ref name or SHA1 expression: %s", from);
2866 read_next_command();
2868 /* tagger ... */
2869 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2870 tagger = parse_ident(v);
2871 read_next_command();
2872 } else
2873 tagger = NULL;
2875 /* tag payload/message */
2876 parse_data(&msg, 0, NULL);
2878 /* build the tag object */
2879 strbuf_reset(&new_data);
2881 strbuf_addf(&new_data,
2882 "object %s\n"
2883 "type %s\n"
2884 "tag %s\n",
2885 oid_to_hex(&oid), type_name(type), t->name);
2886 if (tagger)
2887 strbuf_addf(&new_data,
2888 "tagger %s\n", tagger);
2889 strbuf_addch(&new_data, '\n');
2890 strbuf_addbuf(&new_data, &msg);
2891 free(tagger);
2893 if (store_object(OBJ_TAG, &new_data, NULL, &t->oid, 0))
2894 t->pack_id = MAX_PACK_ID;
2895 else
2896 t->pack_id = pack_id;
2899 static void parse_reset_branch(const char *arg)
2901 struct branch *b;
2903 b = lookup_branch(arg);
2904 if (b) {
2905 oidclr(&b->oid);
2906 oidclr(&b->branch_tree.versions[0].oid);
2907 oidclr(&b->branch_tree.versions[1].oid);
2908 if (b->branch_tree.tree) {
2909 release_tree_content_recursive(b->branch_tree.tree);
2910 b->branch_tree.tree = NULL;
2913 else
2914 b = new_branch(arg);
2915 read_next_command();
2916 parse_from(b);
2917 if (command_buf.len > 0)
2918 unread_command_buf = 1;
2921 static void cat_blob_write(const char *buf, unsigned long size)
2923 if (write_in_full(cat_blob_fd, buf, size) < 0)
2924 die_errno("Write to frontend failed");
2927 static void cat_blob(struct object_entry *oe, struct object_id *oid)
2929 struct strbuf line = STRBUF_INIT;
2930 unsigned long size;
2931 enum object_type type = 0;
2932 char *buf;
2934 if (!oe || oe->pack_id == MAX_PACK_ID) {
2935 buf = read_object_file(oid, &type, &size);
2936 } else {
2937 type = oe->type;
2938 buf = gfi_unpack_entry(oe, &size);
2942 * Output based on batch_one_object() from cat-file.c.
2944 if (type <= 0) {
2945 strbuf_reset(&line);
2946 strbuf_addf(&line, "%s missing\n", oid_to_hex(oid));
2947 cat_blob_write(line.buf, line.len);
2948 strbuf_release(&line);
2949 free(buf);
2950 return;
2952 if (!buf)
2953 die("Can't read object %s", oid_to_hex(oid));
2954 if (type != OBJ_BLOB)
2955 die("Object %s is a %s but a blob was expected.",
2956 oid_to_hex(oid), type_name(type));
2957 strbuf_reset(&line);
2958 strbuf_addf(&line, "%s %s %"PRIuMAX"\n", oid_to_hex(oid),
2959 type_name(type), (uintmax_t)size);
2960 cat_blob_write(line.buf, line.len);
2961 strbuf_release(&line);
2962 cat_blob_write(buf, size);
2963 cat_blob_write("\n", 1);
2964 if (oe && oe->pack_id == pack_id) {
2965 last_blob.offset = oe->idx.offset;
2966 strbuf_attach(&last_blob.data, buf, size, size);
2967 last_blob.depth = oe->depth;
2968 } else
2969 free(buf);
2972 static void parse_get_mark(const char *p)
2974 struct object_entry *oe;
2975 char output[GIT_MAX_HEXSZ + 2];
2977 /* get-mark SP <object> LF */
2978 if (*p != ':')
2979 die("Not a mark: %s", p);
2981 oe = find_mark(parse_mark_ref_eol(p));
2982 if (!oe)
2983 die("Unknown mark: %s", command_buf.buf);
2985 xsnprintf(output, sizeof(output), "%s\n", oid_to_hex(&oe->idx.oid));
2986 cat_blob_write(output, GIT_SHA1_HEXSZ + 1);
2989 static void parse_cat_blob(const char *p)
2991 struct object_entry *oe;
2992 struct object_id oid;
2994 /* cat-blob SP <object> LF */
2995 if (*p == ':') {
2996 oe = find_mark(parse_mark_ref_eol(p));
2997 if (!oe)
2998 die("Unknown mark: %s", command_buf.buf);
2999 oidcpy(&oid, &oe->idx.oid);
3000 } else {
3001 if (parse_oid_hex(p, &oid, &p))
3002 die("Invalid dataref: %s", command_buf.buf);
3003 if (*p)
3004 die("Garbage after SHA1: %s", command_buf.buf);
3005 oe = find_object(&oid);
3008 cat_blob(oe, &oid);
3011 static struct object_entry *dereference(struct object_entry *oe,
3012 struct object_id *oid)
3014 unsigned long size;
3015 char *buf = NULL;
3016 if (!oe) {
3017 enum object_type type = oid_object_info(the_repository, oid,
3018 NULL);
3019 if (type < 0)
3020 die("object not found: %s", oid_to_hex(oid));
3021 /* cache it! */
3022 oe = insert_object(oid);
3023 oe->type = type;
3024 oe->pack_id = MAX_PACK_ID;
3025 oe->idx.offset = 1;
3027 switch (oe->type) {
3028 case OBJ_TREE: /* easy case. */
3029 return oe;
3030 case OBJ_COMMIT:
3031 case OBJ_TAG:
3032 break;
3033 default:
3034 die("Not a tree-ish: %s", command_buf.buf);
3037 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
3038 buf = gfi_unpack_entry(oe, &size);
3039 } else {
3040 enum object_type unused;
3041 buf = read_object_file(oid, &unused, &size);
3043 if (!buf)
3044 die("Can't load object %s", oid_to_hex(oid));
3046 /* Peel one layer. */
3047 switch (oe->type) {
3048 case OBJ_TAG:
3049 if (size < GIT_SHA1_HEXSZ + strlen("object ") ||
3050 get_oid_hex(buf + strlen("object "), oid))
3051 die("Invalid SHA1 in tag: %s", command_buf.buf);
3052 break;
3053 case OBJ_COMMIT:
3054 if (size < GIT_SHA1_HEXSZ + strlen("tree ") ||
3055 get_oid_hex(buf + strlen("tree "), oid))
3056 die("Invalid SHA1 in commit: %s", command_buf.buf);
3059 free(buf);
3060 return find_object(oid);
3063 static struct object_entry *parse_treeish_dataref(const char **p)
3065 struct object_id oid;
3066 struct object_entry *e;
3068 if (**p == ':') { /* <mark> */
3069 e = find_mark(parse_mark_ref_space(p));
3070 if (!e)
3071 die("Unknown mark: %s", command_buf.buf);
3072 oidcpy(&oid, &e->idx.oid);
3073 } else { /* <sha1> */
3074 if (parse_oid_hex(*p, &oid, p))
3075 die("Invalid dataref: %s", command_buf.buf);
3076 e = find_object(&oid);
3077 if (*(*p)++ != ' ')
3078 die("Missing space after tree-ish: %s", command_buf.buf);
3081 while (!e || e->type != OBJ_TREE)
3082 e = dereference(e, &oid);
3083 return e;
3086 static void print_ls(int mode, const unsigned char *sha1, const char *path)
3088 static struct strbuf line = STRBUF_INIT;
3090 /* See show_tree(). */
3091 const char *type =
3092 S_ISGITLINK(mode) ? commit_type :
3093 S_ISDIR(mode) ? tree_type :
3094 blob_type;
3096 if (!mode) {
3097 /* missing SP path LF */
3098 strbuf_reset(&line);
3099 strbuf_addstr(&line, "missing ");
3100 quote_c_style(path, &line, NULL, 0);
3101 strbuf_addch(&line, '\n');
3102 } else {
3103 /* mode SP type SP object_name TAB path LF */
3104 strbuf_reset(&line);
3105 strbuf_addf(&line, "%06o %s %s\t",
3106 mode & ~NO_DELTA, type, sha1_to_hex(sha1));
3107 quote_c_style(path, &line, NULL, 0);
3108 strbuf_addch(&line, '\n');
3110 cat_blob_write(line.buf, line.len);
3113 static void parse_ls(const char *p, struct branch *b)
3115 struct tree_entry *root = NULL;
3116 struct tree_entry leaf = {NULL};
3118 /* ls SP (<tree-ish> SP)? <path> */
3119 if (*p == '"') {
3120 if (!b)
3121 die("Not in a commit: %s", command_buf.buf);
3122 root = &b->branch_tree;
3123 } else {
3124 struct object_entry *e = parse_treeish_dataref(&p);
3125 root = new_tree_entry();
3126 oidcpy(&root->versions[1].oid, &e->idx.oid);
3127 if (!is_null_oid(&root->versions[1].oid))
3128 root->versions[1].mode = S_IFDIR;
3129 load_tree(root);
3131 if (*p == '"') {
3132 static struct strbuf uq = STRBUF_INIT;
3133 const char *endp;
3134 strbuf_reset(&uq);
3135 if (unquote_c_style(&uq, p, &endp))
3136 die("Invalid path: %s", command_buf.buf);
3137 if (*endp)
3138 die("Garbage after path in: %s", command_buf.buf);
3139 p = uq.buf;
3141 tree_content_get(root, p, &leaf, 1);
3143 * A directory in preparation would have a sha1 of zero
3144 * until it is saved. Save, for simplicity.
3146 if (S_ISDIR(leaf.versions[1].mode))
3147 store_tree(&leaf);
3149 print_ls(leaf.versions[1].mode, leaf.versions[1].oid.hash, p);
3150 if (leaf.tree)
3151 release_tree_content_recursive(leaf.tree);
3152 if (!b || root != &b->branch_tree)
3153 release_tree_entry(root);
3156 static void checkpoint(void)
3158 checkpoint_requested = 0;
3159 if (object_count) {
3160 cycle_packfile();
3162 dump_branches();
3163 dump_tags();
3164 dump_marks();
3167 static void parse_checkpoint(void)
3169 checkpoint_requested = 1;
3170 skip_optional_lf();
3173 static void parse_progress(void)
3175 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3176 fputc('\n', stdout);
3177 fflush(stdout);
3178 skip_optional_lf();
3181 static char* make_fast_import_path(const char *path)
3183 if (!relative_marks_paths || is_absolute_path(path))
3184 return xstrdup(path);
3185 return git_pathdup("info/fast-import/%s", path);
3188 static void option_import_marks(const char *marks,
3189 int from_stream, int ignore_missing)
3191 if (import_marks_file) {
3192 if (from_stream)
3193 die("Only one import-marks command allowed per stream");
3195 /* read previous mark file */
3196 if(!import_marks_file_from_stream)
3197 read_marks();
3200 import_marks_file = make_fast_import_path(marks);
3201 safe_create_leading_directories_const(import_marks_file);
3202 import_marks_file_from_stream = from_stream;
3203 import_marks_file_ignore_missing = ignore_missing;
3206 static void option_date_format(const char *fmt)
3208 if (!strcmp(fmt, "raw"))
3209 whenspec = WHENSPEC_RAW;
3210 else if (!strcmp(fmt, "rfc2822"))
3211 whenspec = WHENSPEC_RFC2822;
3212 else if (!strcmp(fmt, "now"))
3213 whenspec = WHENSPEC_NOW;
3214 else
3215 die("unknown --date-format argument %s", fmt);
3218 static unsigned long ulong_arg(const char *option, const char *arg)
3220 char *endptr;
3221 unsigned long rv = strtoul(arg, &endptr, 0);
3222 if (strchr(arg, '-') || endptr == arg || *endptr)
3223 die("%s: argument must be a non-negative integer", option);
3224 return rv;
3227 static void option_depth(const char *depth)
3229 max_depth = ulong_arg("--depth", depth);
3230 if (max_depth > MAX_DEPTH)
3231 die("--depth cannot exceed %u", MAX_DEPTH);
3234 static void option_active_branches(const char *branches)
3236 max_active_branches = ulong_arg("--active-branches", branches);
3239 static void option_export_marks(const char *marks)
3241 export_marks_file = make_fast_import_path(marks);
3242 safe_create_leading_directories_const(export_marks_file);
3245 static void option_cat_blob_fd(const char *fd)
3247 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3248 if (n > (unsigned long) INT_MAX)
3249 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3250 cat_blob_fd = (int) n;
3253 static void option_export_pack_edges(const char *edges)
3255 if (pack_edges)
3256 fclose(pack_edges);
3257 pack_edges = xfopen(edges, "a");
3260 static int parse_one_option(const char *option)
3262 if (skip_prefix(option, "max-pack-size=", &option)) {
3263 unsigned long v;
3264 if (!git_parse_ulong(option, &v))
3265 return 0;
3266 if (v < 8192) {
3267 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3268 v *= 1024 * 1024;
3269 } else if (v < 1024 * 1024) {
3270 warning("minimum max-pack-size is 1 MiB");
3271 v = 1024 * 1024;
3273 max_packsize = v;
3274 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3275 unsigned long v;
3276 if (!git_parse_ulong(option, &v))
3277 return 0;
3278 big_file_threshold = v;
3279 } else if (skip_prefix(option, "depth=", &option)) {
3280 option_depth(option);
3281 } else if (skip_prefix(option, "active-branches=", &option)) {
3282 option_active_branches(option);
3283 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3284 option_export_pack_edges(option);
3285 } else if (starts_with(option, "quiet")) {
3286 show_stats = 0;
3287 } else if (starts_with(option, "stats")) {
3288 show_stats = 1;
3289 } else {
3290 return 0;
3293 return 1;
3296 static int parse_one_feature(const char *feature, int from_stream)
3298 const char *arg;
3300 if (skip_prefix(feature, "date-format=", &arg)) {
3301 option_date_format(arg);
3302 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3303 option_import_marks(arg, from_stream, 0);
3304 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3305 option_import_marks(arg, from_stream, 1);
3306 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3307 option_export_marks(arg);
3308 } else if (!strcmp(feature, "get-mark")) {
3309 ; /* Don't die - this feature is supported */
3310 } else if (!strcmp(feature, "cat-blob")) {
3311 ; /* Don't die - this feature is supported */
3312 } else if (!strcmp(feature, "relative-marks")) {
3313 relative_marks_paths = 1;
3314 } else if (!strcmp(feature, "no-relative-marks")) {
3315 relative_marks_paths = 0;
3316 } else if (!strcmp(feature, "done")) {
3317 require_explicit_termination = 1;
3318 } else if (!strcmp(feature, "force")) {
3319 force_update = 1;
3320 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3321 ; /* do nothing; we have the feature */
3322 } else {
3323 return 0;
3326 return 1;
3329 static void parse_feature(const char *feature)
3331 if (seen_data_command)
3332 die("Got feature command '%s' after data command", feature);
3334 if (parse_one_feature(feature, 1))
3335 return;
3337 die("This version of fast-import does not support feature %s.", feature);
3340 static void parse_option(const char *option)
3342 if (seen_data_command)
3343 die("Got option command '%s' after data command", option);
3345 if (parse_one_option(option))
3346 return;
3348 die("This version of fast-import does not support option: %s", option);
3351 static void git_pack_config(void)
3353 int indexversion_value;
3354 int limit;
3355 unsigned long packsizelimit_value;
3357 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3358 if (max_depth > MAX_DEPTH)
3359 max_depth = MAX_DEPTH;
3361 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3362 pack_idx_opts.version = indexversion_value;
3363 if (pack_idx_opts.version > 2)
3364 git_die_config("pack.indexversion",
3365 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3367 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3368 max_packsize = packsizelimit_value;
3370 if (!git_config_get_int("fastimport.unpacklimit", &limit))
3371 unpack_limit = limit;
3372 else if (!git_config_get_int("transfer.unpacklimit", &limit))
3373 unpack_limit = limit;
3375 git_config(git_default_config, NULL);
3378 static const char fast_import_usage[] =
3379 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3381 static void parse_argv(void)
3383 unsigned int i;
3385 for (i = 1; i < global_argc; i++) {
3386 const char *a = global_argv[i];
3388 if (*a != '-' || !strcmp(a, "--"))
3389 break;
3391 if (!skip_prefix(a, "--", &a))
3392 die("unknown option %s", a);
3394 if (parse_one_option(a))
3395 continue;
3397 if (parse_one_feature(a, 0))
3398 continue;
3400 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3401 option_cat_blob_fd(a);
3402 continue;
3405 die("unknown option --%s", a);
3407 if (i != global_argc)
3408 usage(fast_import_usage);
3410 seen_data_command = 1;
3411 if (import_marks_file)
3412 read_marks();
3415 int cmd_main(int argc, const char **argv)
3417 unsigned int i;
3419 if (argc == 2 && !strcmp(argv[1], "-h"))
3420 usage(fast_import_usage);
3422 setup_git_directory();
3423 reset_pack_idx_option(&pack_idx_opts);
3424 git_pack_config();
3426 alloc_objects(object_entry_alloc);
3427 strbuf_init(&command_buf, 0);
3428 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3429 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3430 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3431 marks = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct mark_set));
3433 global_argc = argc;
3434 global_argv = argv;
3436 rc_free = mem_pool_alloc(&fi_mem_pool, cmd_save * sizeof(*rc_free));
3437 for (i = 0; i < (cmd_save - 1); i++)
3438 rc_free[i].next = &rc_free[i + 1];
3439 rc_free[cmd_save - 1].next = NULL;
3441 start_packfile();
3442 set_die_routine(die_nicely);
3443 set_checkpoint_signal();
3444 while (read_next_command() != EOF) {
3445 const char *v;
3446 if (!strcmp("blob", command_buf.buf))
3447 parse_new_blob();
3448 else if (skip_prefix(command_buf.buf, "ls ", &v))
3449 parse_ls(v, NULL);
3450 else if (skip_prefix(command_buf.buf, "commit ", &v))
3451 parse_new_commit(v);
3452 else if (skip_prefix(command_buf.buf, "tag ", &v))
3453 parse_new_tag(v);
3454 else if (skip_prefix(command_buf.buf, "reset ", &v))
3455 parse_reset_branch(v);
3456 else if (!strcmp("checkpoint", command_buf.buf))
3457 parse_checkpoint();
3458 else if (!strcmp("done", command_buf.buf))
3459 break;
3460 else if (starts_with(command_buf.buf, "progress "))
3461 parse_progress();
3462 else if (skip_prefix(command_buf.buf, "feature ", &v))
3463 parse_feature(v);
3464 else if (skip_prefix(command_buf.buf, "option git ", &v))
3465 parse_option(v);
3466 else if (starts_with(command_buf.buf, "option "))
3467 /* ignore non-git options*/;
3468 else
3469 die("Unsupported command: %s", command_buf.buf);
3471 if (checkpoint_requested)
3472 checkpoint();
3475 /* argv hasn't been parsed yet, do so */
3476 if (!seen_data_command)
3477 parse_argv();
3479 if (require_explicit_termination && feof(stdin))
3480 die("stream ends early");
3482 end_packfile();
3484 dump_branches();
3485 dump_tags();
3486 unkeep_all_packs();
3487 dump_marks();
3489 if (pack_edges)
3490 fclose(pack_edges);
3492 if (show_stats) {
3493 uintmax_t total_count = 0, duplicate_count = 0;
3494 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3495 total_count += object_count_by_type[i];
3496 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3497 duplicate_count += duplicate_count_by_type[i];
3499 fprintf(stderr, "%s statistics:\n", argv[0]);
3500 fprintf(stderr, "---------------------------------------------------------------------\n");
3501 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3502 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3503 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]);
3504 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]);
3505 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]);
3506 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]);
3507 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3508 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3509 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3510 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (tree_entry_allocd + fi_mem_pool.pool_alloc + alloc_count*sizeof(struct object_entry))/1024);
3511 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)((tree_entry_allocd + fi_mem_pool.pool_alloc) /1024));
3512 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3513 fprintf(stderr, "---------------------------------------------------------------------\n");
3514 pack_report();
3515 fprintf(stderr, "---------------------------------------------------------------------\n");
3516 fprintf(stderr, "\n");
3519 return failure ? 1 : 0;