Merge branch 'master' of github.com:git/git into git-po-master
[git.git] / fast-import.c
blobb8b65a801cc1f902f7131590816459d2d1e0faa0
1 #include "builtin.h"
2 #include "cache.h"
3 #include "repository.h"
4 #include "config.h"
5 #include "lockfile.h"
6 #include "object.h"
7 #include "blob.h"
8 #include "tree.h"
9 #include "commit.h"
10 #include "delta.h"
11 #include "pack.h"
12 #include "refs.h"
13 #include "csum-file.h"
14 #include "quote.h"
15 #include "dir.h"
16 #include "run-command.h"
17 #include "packfile.h"
18 #include "object-store.h"
19 #include "mem-pool.h"
20 #include "commit-reach.h"
22 #define PACK_ID_BITS 16
23 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
24 #define DEPTH_BITS 13
25 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
28 * We abuse the setuid bit on directories to mean "do not delta".
30 #define NO_DELTA S_ISUID
33 * The amount of additional space required in order to write an object into the
34 * current pack. This is the hash lengths at the end of the pack, plus the
35 * length of one object ID.
37 #define PACK_SIZE_THRESHOLD (the_hash_algo->rawsz * 3)
39 struct object_entry {
40 struct pack_idx_entry idx;
41 struct object_entry *next;
42 uint32_t type : TYPE_BITS,
43 pack_id : PACK_ID_BITS,
44 depth : DEPTH_BITS;
47 struct object_entry_pool {
48 struct object_entry_pool *next_pool;
49 struct object_entry *next_free;
50 struct object_entry *end;
51 struct object_entry entries[FLEX_ARRAY]; /* more */
54 struct mark_set {
55 union {
56 struct object_entry *marked[1024];
57 struct mark_set *sets[1024];
58 } data;
59 unsigned int shift;
62 struct last_object {
63 struct strbuf data;
64 off_t offset;
65 unsigned int depth;
66 unsigned no_swap : 1;
69 struct atom_str {
70 struct atom_str *next_atom;
71 unsigned short str_len;
72 char str_dat[FLEX_ARRAY]; /* more */
75 struct tree_content;
76 struct tree_entry {
77 struct tree_content *tree;
78 struct atom_str *name;
79 struct tree_entry_ms {
80 uint16_t mode;
81 struct object_id oid;
82 } versions[2];
85 struct tree_content {
86 unsigned int entry_capacity; /* must match avail_tree_content */
87 unsigned int entry_count;
88 unsigned int delta_depth;
89 struct tree_entry *entries[FLEX_ARRAY]; /* more */
92 struct avail_tree_content {
93 unsigned int entry_capacity; /* must match tree_content */
94 struct avail_tree_content *next_avail;
97 struct branch {
98 struct branch *table_next_branch;
99 struct branch *active_next_branch;
100 const char *name;
101 struct tree_entry branch_tree;
102 uintmax_t last_commit;
103 uintmax_t num_notes;
104 unsigned active : 1;
105 unsigned delete : 1;
106 unsigned pack_id : PACK_ID_BITS;
107 struct object_id oid;
110 struct tag {
111 struct tag *next_tag;
112 const char *name;
113 unsigned int pack_id;
114 struct object_id oid;
117 struct hash_list {
118 struct hash_list *next;
119 struct object_id oid;
122 typedef enum {
123 WHENSPEC_RAW = 1,
124 WHENSPEC_RFC2822,
125 WHENSPEC_NOW
126 } whenspec_type;
128 struct recent_command {
129 struct recent_command *prev;
130 struct recent_command *next;
131 char *buf;
134 /* Configured limits on output */
135 static unsigned long max_depth = 50;
136 static off_t max_packsize;
137 static int unpack_limit = 100;
138 static int force_update;
140 /* Stats and misc. counters */
141 static uintmax_t alloc_count;
142 static uintmax_t marks_set_count;
143 static uintmax_t object_count_by_type[1 << TYPE_BITS];
144 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
145 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
146 static uintmax_t delta_count_attempts_by_type[1 << TYPE_BITS];
147 static unsigned long object_count;
148 static unsigned long branch_count;
149 static unsigned long branch_load_count;
150 static int failure;
151 static FILE *pack_edges;
152 static unsigned int show_stats = 1;
153 static int global_argc;
154 static const char **global_argv;
156 /* Memory pools */
157 static struct mem_pool fi_mem_pool = {NULL, 2*1024*1024 -
158 sizeof(struct mp_block), 0 };
160 /* Atom management */
161 static unsigned int atom_table_sz = 4451;
162 static unsigned int atom_cnt;
163 static struct atom_str **atom_table;
165 /* The .pack file being generated */
166 static struct pack_idx_option pack_idx_opts;
167 static unsigned int pack_id;
168 static struct hashfile *pack_file;
169 static struct packed_git *pack_data;
170 static struct packed_git **all_packs;
171 static off_t pack_size;
173 /* Table of objects we've written. */
174 static unsigned int object_entry_alloc = 5000;
175 static struct object_entry_pool *blocks;
176 static struct object_entry *object_table[1 << 16];
177 static struct mark_set *marks;
178 static const char *export_marks_file;
179 static const char *import_marks_file;
180 static int import_marks_file_from_stream;
181 static int import_marks_file_ignore_missing;
182 static int import_marks_file_done;
183 static int relative_marks_paths;
185 /* Our last blob */
186 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
188 /* Tree management */
189 static unsigned int tree_entry_alloc = 1000;
190 static void *avail_tree_entry;
191 static unsigned int avail_tree_table_sz = 100;
192 static struct avail_tree_content **avail_tree_table;
193 static size_t tree_entry_allocd;
194 static struct strbuf old_tree = STRBUF_INIT;
195 static struct strbuf new_tree = STRBUF_INIT;
197 /* Branch data */
198 static unsigned long max_active_branches = 5;
199 static unsigned long cur_active_branches;
200 static unsigned long branch_table_sz = 1039;
201 static struct branch **branch_table;
202 static struct branch *active_branches;
204 /* Tag data */
205 static struct tag *first_tag;
206 static struct tag *last_tag;
208 /* Input stream parsing */
209 static whenspec_type whenspec = WHENSPEC_RAW;
210 static struct strbuf command_buf = STRBUF_INIT;
211 static int unread_command_buf;
212 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
213 static struct recent_command *cmd_tail = &cmd_hist;
214 static struct recent_command *rc_free;
215 static unsigned int cmd_save = 100;
216 static uintmax_t next_mark;
217 static struct strbuf new_data = STRBUF_INIT;
218 static int seen_data_command;
219 static int require_explicit_termination;
220 static int allow_unsafe_features;
222 /* Signal handling */
223 static volatile sig_atomic_t checkpoint_requested;
225 /* Where to write output of cat-blob commands */
226 static int cat_blob_fd = STDOUT_FILENO;
228 static void parse_argv(void);
229 static void parse_get_mark(const char *p);
230 static void parse_cat_blob(const char *p);
231 static void parse_ls(const char *p, struct branch *b);
233 static void write_branch_report(FILE *rpt, struct branch *b)
235 fprintf(rpt, "%s:\n", b->name);
237 fprintf(rpt, " status :");
238 if (b->active)
239 fputs(" active", rpt);
240 if (b->branch_tree.tree)
241 fputs(" loaded", rpt);
242 if (is_null_oid(&b->branch_tree.versions[1].oid))
243 fputs(" dirty", rpt);
244 fputc('\n', rpt);
246 fprintf(rpt, " tip commit : %s\n", oid_to_hex(&b->oid));
247 fprintf(rpt, " old tree : %s\n",
248 oid_to_hex(&b->branch_tree.versions[0].oid));
249 fprintf(rpt, " cur tree : %s\n",
250 oid_to_hex(&b->branch_tree.versions[1].oid));
251 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
253 fputs(" last pack : ", rpt);
254 if (b->pack_id < MAX_PACK_ID)
255 fprintf(rpt, "%u", b->pack_id);
256 fputc('\n', rpt);
258 fputc('\n', rpt);
261 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
263 static void write_crash_report(const char *err)
265 char *loc = git_pathdup("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
266 FILE *rpt = fopen(loc, "w");
267 struct branch *b;
268 unsigned long lu;
269 struct recent_command *rc;
271 if (!rpt) {
272 error_errno("can't write crash report %s", loc);
273 free(loc);
274 return;
277 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
279 fprintf(rpt, "fast-import crash report:\n");
280 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
281 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
282 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_MODE(ISO8601)));
283 fputc('\n', rpt);
285 fputs("fatal: ", rpt);
286 fputs(err, rpt);
287 fputc('\n', rpt);
289 fputc('\n', rpt);
290 fputs("Most Recent Commands Before Crash\n", rpt);
291 fputs("---------------------------------\n", rpt);
292 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
293 if (rc->next == &cmd_hist)
294 fputs("* ", rpt);
295 else
296 fputs(" ", rpt);
297 fputs(rc->buf, rpt);
298 fputc('\n', rpt);
301 fputc('\n', rpt);
302 fputs("Active Branch LRU\n", rpt);
303 fputs("-----------------\n", rpt);
304 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
305 cur_active_branches,
306 max_active_branches);
307 fputc('\n', rpt);
308 fputs(" pos clock name\n", rpt);
309 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
310 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
311 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
312 ++lu, b->last_commit, b->name);
314 fputc('\n', rpt);
315 fputs("Inactive Branches\n", rpt);
316 fputs("-----------------\n", rpt);
317 for (lu = 0; lu < branch_table_sz; lu++) {
318 for (b = branch_table[lu]; b; b = b->table_next_branch)
319 write_branch_report(rpt, b);
322 if (first_tag) {
323 struct tag *tg;
324 fputc('\n', rpt);
325 fputs("Annotated Tags\n", rpt);
326 fputs("--------------\n", rpt);
327 for (tg = first_tag; tg; tg = tg->next_tag) {
328 fputs(oid_to_hex(&tg->oid), rpt);
329 fputc(' ', rpt);
330 fputs(tg->name, rpt);
331 fputc('\n', rpt);
335 fputc('\n', rpt);
336 fputs("Marks\n", rpt);
337 fputs("-----\n", rpt);
338 if (export_marks_file)
339 fprintf(rpt, " exported to %s\n", export_marks_file);
340 else
341 dump_marks_helper(rpt, 0, marks);
343 fputc('\n', rpt);
344 fputs("-------------------\n", rpt);
345 fputs("END OF CRASH REPORT\n", rpt);
346 fclose(rpt);
347 free(loc);
350 static void end_packfile(void);
351 static void unkeep_all_packs(void);
352 static void dump_marks(void);
354 static NORETURN void die_nicely(const char *err, va_list params)
356 static int zombie;
357 char message[2 * PATH_MAX];
359 vsnprintf(message, sizeof(message), err, params);
360 fputs("fatal: ", stderr);
361 fputs(message, stderr);
362 fputc('\n', stderr);
364 if (!zombie) {
365 zombie = 1;
366 write_crash_report(message);
367 end_packfile();
368 unkeep_all_packs();
369 dump_marks();
371 exit(128);
374 #ifndef SIGUSR1 /* Windows, for example */
376 static void set_checkpoint_signal(void)
380 #else
382 static void checkpoint_signal(int signo)
384 checkpoint_requested = 1;
387 static void set_checkpoint_signal(void)
389 struct sigaction sa;
391 memset(&sa, 0, sizeof(sa));
392 sa.sa_handler = checkpoint_signal;
393 sigemptyset(&sa.sa_mask);
394 sa.sa_flags = SA_RESTART;
395 sigaction(SIGUSR1, &sa, NULL);
398 #endif
400 static void alloc_objects(unsigned int cnt)
402 struct object_entry_pool *b;
404 b = xmalloc(sizeof(struct object_entry_pool)
405 + cnt * sizeof(struct object_entry));
406 b->next_pool = blocks;
407 b->next_free = b->entries;
408 b->end = b->entries + cnt;
409 blocks = b;
410 alloc_count += cnt;
413 static struct object_entry *new_object(struct object_id *oid)
415 struct object_entry *e;
417 if (blocks->next_free == blocks->end)
418 alloc_objects(object_entry_alloc);
420 e = blocks->next_free++;
421 oidcpy(&e->idx.oid, oid);
422 return e;
425 static struct object_entry *find_object(struct object_id *oid)
427 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
428 struct object_entry *e;
429 for (e = object_table[h]; e; e = e->next)
430 if (oideq(oid, &e->idx.oid))
431 return e;
432 return NULL;
435 static struct object_entry *insert_object(struct object_id *oid)
437 unsigned int h = oid->hash[0] << 8 | oid->hash[1];
438 struct object_entry *e = object_table[h];
440 while (e) {
441 if (oideq(oid, &e->idx.oid))
442 return e;
443 e = e->next;
446 e = new_object(oid);
447 e->next = object_table[h];
448 e->idx.offset = 0;
449 object_table[h] = e;
450 return e;
453 static void invalidate_pack_id(unsigned int id)
455 unsigned int h;
456 unsigned long lu;
457 struct tag *t;
459 for (h = 0; h < ARRAY_SIZE(object_table); h++) {
460 struct object_entry *e;
462 for (e = object_table[h]; e; e = e->next)
463 if (e->pack_id == id)
464 e->pack_id = MAX_PACK_ID;
467 for (lu = 0; lu < branch_table_sz; lu++) {
468 struct branch *b;
470 for (b = branch_table[lu]; b; b = b->table_next_branch)
471 if (b->pack_id == id)
472 b->pack_id = MAX_PACK_ID;
475 for (t = first_tag; t; t = t->next_tag)
476 if (t->pack_id == id)
477 t->pack_id = MAX_PACK_ID;
480 static unsigned int hc_str(const char *s, size_t len)
482 unsigned int r = 0;
483 while (len-- > 0)
484 r = r * 31 + *s++;
485 return r;
488 static char *pool_strdup(const char *s)
490 size_t len = strlen(s) + 1;
491 char *r = mem_pool_alloc(&fi_mem_pool, len);
492 memcpy(r, s, len);
493 return r;
496 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
498 struct mark_set *s = marks;
499 while ((idnum >> s->shift) >= 1024) {
500 s = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct mark_set));
501 s->shift = marks->shift + 10;
502 s->data.sets[0] = marks;
503 marks = s;
505 while (s->shift) {
506 uintmax_t i = idnum >> s->shift;
507 idnum -= i << s->shift;
508 if (!s->data.sets[i]) {
509 s->data.sets[i] = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct mark_set));
510 s->data.sets[i]->shift = s->shift - 10;
512 s = s->data.sets[i];
514 if (!s->data.marked[idnum])
515 marks_set_count++;
516 s->data.marked[idnum] = oe;
519 static struct object_entry *find_mark(uintmax_t idnum)
521 uintmax_t orig_idnum = idnum;
522 struct mark_set *s = marks;
523 struct object_entry *oe = NULL;
524 if ((idnum >> s->shift) < 1024) {
525 while (s && s->shift) {
526 uintmax_t i = idnum >> s->shift;
527 idnum -= i << s->shift;
528 s = s->data.sets[i];
530 if (s)
531 oe = s->data.marked[idnum];
533 if (!oe)
534 die("mark :%" PRIuMAX " not declared", orig_idnum);
535 return oe;
538 static struct atom_str *to_atom(const char *s, unsigned short len)
540 unsigned int hc = hc_str(s, len) % atom_table_sz;
541 struct atom_str *c;
543 for (c = atom_table[hc]; c; c = c->next_atom)
544 if (c->str_len == len && !strncmp(s, c->str_dat, len))
545 return c;
547 c = mem_pool_alloc(&fi_mem_pool, sizeof(struct atom_str) + len + 1);
548 c->str_len = len;
549 memcpy(c->str_dat, s, len);
550 c->str_dat[len] = 0;
551 c->next_atom = atom_table[hc];
552 atom_table[hc] = c;
553 atom_cnt++;
554 return c;
557 static struct branch *lookup_branch(const char *name)
559 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
560 struct branch *b;
562 for (b = branch_table[hc]; b; b = b->table_next_branch)
563 if (!strcmp(name, b->name))
564 return b;
565 return NULL;
568 static struct branch *new_branch(const char *name)
570 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
571 struct branch *b = lookup_branch(name);
573 if (b)
574 die("Invalid attempt to create duplicate branch: %s", name);
575 if (check_refname_format(name, REFNAME_ALLOW_ONELEVEL))
576 die("Branch name doesn't conform to GIT standards: %s", name);
578 b = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct branch));
579 b->name = pool_strdup(name);
580 b->table_next_branch = branch_table[hc];
581 b->branch_tree.versions[0].mode = S_IFDIR;
582 b->branch_tree.versions[1].mode = S_IFDIR;
583 b->num_notes = 0;
584 b->active = 0;
585 b->pack_id = MAX_PACK_ID;
586 branch_table[hc] = b;
587 branch_count++;
588 return b;
591 static unsigned int hc_entries(unsigned int cnt)
593 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
594 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
597 static struct tree_content *new_tree_content(unsigned int cnt)
599 struct avail_tree_content *f, *l = NULL;
600 struct tree_content *t;
601 unsigned int hc = hc_entries(cnt);
603 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
604 if (f->entry_capacity >= cnt)
605 break;
607 if (f) {
608 if (l)
609 l->next_avail = f->next_avail;
610 else
611 avail_tree_table[hc] = f->next_avail;
612 } else {
613 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
614 f = mem_pool_alloc(&fi_mem_pool, sizeof(*t) + sizeof(t->entries[0]) * cnt);
615 f->entry_capacity = cnt;
618 t = (struct tree_content*)f;
619 t->entry_count = 0;
620 t->delta_depth = 0;
621 return t;
624 static void release_tree_entry(struct tree_entry *e);
625 static void release_tree_content(struct tree_content *t)
627 struct avail_tree_content *f = (struct avail_tree_content*)t;
628 unsigned int hc = hc_entries(f->entry_capacity);
629 f->next_avail = avail_tree_table[hc];
630 avail_tree_table[hc] = f;
633 static void release_tree_content_recursive(struct tree_content *t)
635 unsigned int i;
636 for (i = 0; i < t->entry_count; i++)
637 release_tree_entry(t->entries[i]);
638 release_tree_content(t);
641 static struct tree_content *grow_tree_content(
642 struct tree_content *t,
643 int amt)
645 struct tree_content *r = new_tree_content(t->entry_count + amt);
646 r->entry_count = t->entry_count;
647 r->delta_depth = t->delta_depth;
648 COPY_ARRAY(r->entries, t->entries, t->entry_count);
649 release_tree_content(t);
650 return r;
653 static struct tree_entry *new_tree_entry(void)
655 struct tree_entry *e;
657 if (!avail_tree_entry) {
658 unsigned int n = tree_entry_alloc;
659 tree_entry_allocd += n * sizeof(struct tree_entry);
660 ALLOC_ARRAY(e, n);
661 avail_tree_entry = e;
662 while (n-- > 1) {
663 *((void**)e) = e + 1;
664 e++;
666 *((void**)e) = NULL;
669 e = avail_tree_entry;
670 avail_tree_entry = *((void**)e);
671 return e;
674 static void release_tree_entry(struct tree_entry *e)
676 if (e->tree)
677 release_tree_content_recursive(e->tree);
678 *((void**)e) = avail_tree_entry;
679 avail_tree_entry = e;
682 static struct tree_content *dup_tree_content(struct tree_content *s)
684 struct tree_content *d;
685 struct tree_entry *a, *b;
686 unsigned int i;
688 if (!s)
689 return NULL;
690 d = new_tree_content(s->entry_count);
691 for (i = 0; i < s->entry_count; i++) {
692 a = s->entries[i];
693 b = new_tree_entry();
694 memcpy(b, a, sizeof(*a));
695 if (a->tree && is_null_oid(&b->versions[1].oid))
696 b->tree = dup_tree_content(a->tree);
697 else
698 b->tree = NULL;
699 d->entries[i] = b;
701 d->entry_count = s->entry_count;
702 d->delta_depth = s->delta_depth;
704 return d;
707 static void start_packfile(void)
709 struct strbuf tmp_file = STRBUF_INIT;
710 struct packed_git *p;
711 struct pack_header hdr;
712 int pack_fd;
714 pack_fd = odb_mkstemp(&tmp_file, "pack/tmp_pack_XXXXXX");
715 FLEX_ALLOC_STR(p, pack_name, tmp_file.buf);
716 strbuf_release(&tmp_file);
718 p->pack_fd = pack_fd;
719 p->do_not_close = 1;
720 pack_file = hashfd(pack_fd, p->pack_name);
722 hdr.hdr_signature = htonl(PACK_SIGNATURE);
723 hdr.hdr_version = htonl(2);
724 hdr.hdr_entries = 0;
725 hashwrite(pack_file, &hdr, sizeof(hdr));
727 pack_data = p;
728 pack_size = sizeof(hdr);
729 object_count = 0;
731 REALLOC_ARRAY(all_packs, pack_id + 1);
732 all_packs[pack_id] = p;
735 static const char *create_index(void)
737 const char *tmpfile;
738 struct pack_idx_entry **idx, **c, **last;
739 struct object_entry *e;
740 struct object_entry_pool *o;
742 /* Build the table of object IDs. */
743 ALLOC_ARRAY(idx, object_count);
744 c = idx;
745 for (o = blocks; o; o = o->next_pool)
746 for (e = o->next_free; e-- != o->entries;)
747 if (pack_id == e->pack_id)
748 *c++ = &e->idx;
749 last = idx + object_count;
750 if (c != last)
751 die("internal consistency error creating the index");
753 tmpfile = write_idx_file(NULL, idx, object_count, &pack_idx_opts,
754 pack_data->hash);
755 free(idx);
756 return tmpfile;
759 static char *keep_pack(const char *curr_index_name)
761 static const char *keep_msg = "fast-import";
762 struct strbuf name = STRBUF_INIT;
763 int keep_fd;
765 odb_pack_name(&name, pack_data->hash, "keep");
766 keep_fd = odb_pack_keep(name.buf);
767 if (keep_fd < 0)
768 die_errno("cannot create keep file");
769 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
770 if (close(keep_fd))
771 die_errno("failed to write keep file");
773 odb_pack_name(&name, pack_data->hash, "pack");
774 if (finalize_object_file(pack_data->pack_name, name.buf))
775 die("cannot store pack file");
777 odb_pack_name(&name, pack_data->hash, "idx");
778 if (finalize_object_file(curr_index_name, name.buf))
779 die("cannot store index file");
780 free((void *)curr_index_name);
781 return strbuf_detach(&name, NULL);
784 static void unkeep_all_packs(void)
786 struct strbuf name = STRBUF_INIT;
787 int k;
789 for (k = 0; k < pack_id; k++) {
790 struct packed_git *p = all_packs[k];
791 odb_pack_name(&name, p->hash, "keep");
792 unlink_or_warn(name.buf);
794 strbuf_release(&name);
797 static int loosen_small_pack(const struct packed_git *p)
799 struct child_process unpack = CHILD_PROCESS_INIT;
801 if (lseek(p->pack_fd, 0, SEEK_SET) < 0)
802 die_errno("Failed seeking to start of '%s'", p->pack_name);
804 unpack.in = p->pack_fd;
805 unpack.git_cmd = 1;
806 unpack.stdout_to_stderr = 1;
807 argv_array_push(&unpack.args, "unpack-objects");
808 if (!show_stats)
809 argv_array_push(&unpack.args, "-q");
811 return run_command(&unpack);
814 static void end_packfile(void)
816 static int running;
818 if (running || !pack_data)
819 return;
821 running = 1;
822 clear_delta_base_cache();
823 if (object_count) {
824 struct packed_git *new_p;
825 struct object_id cur_pack_oid;
826 char *idx_name;
827 int i;
828 struct branch *b;
829 struct tag *t;
831 close_pack_windows(pack_data);
832 finalize_hashfile(pack_file, cur_pack_oid.hash, 0);
833 fixup_pack_header_footer(pack_data->pack_fd, pack_data->hash,
834 pack_data->pack_name, object_count,
835 cur_pack_oid.hash, pack_size);
837 if (object_count <= unpack_limit) {
838 if (!loosen_small_pack(pack_data)) {
839 invalidate_pack_id(pack_id);
840 goto discard_pack;
844 close(pack_data->pack_fd);
845 idx_name = keep_pack(create_index());
847 /* Register the packfile with core git's machinery. */
848 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
849 if (!new_p)
850 die("core git rejected index %s", idx_name);
851 all_packs[pack_id] = new_p;
852 install_packed_git(the_repository, new_p);
853 free(idx_name);
855 /* Print the boundary */
856 if (pack_edges) {
857 fprintf(pack_edges, "%s:", new_p->pack_name);
858 for (i = 0; i < branch_table_sz; i++) {
859 for (b = branch_table[i]; b; b = b->table_next_branch) {
860 if (b->pack_id == pack_id)
861 fprintf(pack_edges, " %s",
862 oid_to_hex(&b->oid));
865 for (t = first_tag; t; t = t->next_tag) {
866 if (t->pack_id == pack_id)
867 fprintf(pack_edges, " %s",
868 oid_to_hex(&t->oid));
870 fputc('\n', pack_edges);
871 fflush(pack_edges);
874 pack_id++;
876 else {
877 discard_pack:
878 close(pack_data->pack_fd);
879 unlink_or_warn(pack_data->pack_name);
881 FREE_AND_NULL(pack_data);
882 running = 0;
884 /* We can't carry a delta across packfiles. */
885 strbuf_release(&last_blob.data);
886 last_blob.offset = 0;
887 last_blob.depth = 0;
890 static void cycle_packfile(void)
892 end_packfile();
893 start_packfile();
896 static int store_object(
897 enum object_type type,
898 struct strbuf *dat,
899 struct last_object *last,
900 struct object_id *oidout,
901 uintmax_t mark)
903 void *out, *delta;
904 struct object_entry *e;
905 unsigned char hdr[96];
906 struct object_id oid;
907 unsigned long hdrlen, deltalen;
908 git_hash_ctx c;
909 git_zstream s;
911 hdrlen = xsnprintf((char *)hdr, sizeof(hdr), "%s %lu",
912 type_name(type), (unsigned long)dat->len) + 1;
913 the_hash_algo->init_fn(&c);
914 the_hash_algo->update_fn(&c, hdr, hdrlen);
915 the_hash_algo->update_fn(&c, dat->buf, dat->len);
916 the_hash_algo->final_fn(oid.hash, &c);
917 if (oidout)
918 oidcpy(oidout, &oid);
920 e = insert_object(&oid);
921 if (mark)
922 insert_mark(mark, e);
923 if (e->idx.offset) {
924 duplicate_count_by_type[type]++;
925 return 1;
926 } else if (find_sha1_pack(oid.hash,
927 get_all_packs(the_repository))) {
928 e->type = type;
929 e->pack_id = MAX_PACK_ID;
930 e->idx.offset = 1; /* just not zero! */
931 duplicate_count_by_type[type]++;
932 return 1;
935 if (last && last->data.len && last->data.buf && last->depth < max_depth
936 && dat->len > the_hash_algo->rawsz) {
938 delta_count_attempts_by_type[type]++;
939 delta = diff_delta(last->data.buf, last->data.len,
940 dat->buf, dat->len,
941 &deltalen, dat->len - the_hash_algo->rawsz);
942 } else
943 delta = NULL;
945 git_deflate_init(&s, pack_compression_level);
946 if (delta) {
947 s.next_in = delta;
948 s.avail_in = deltalen;
949 } else {
950 s.next_in = (void *)dat->buf;
951 s.avail_in = dat->len;
953 s.avail_out = git_deflate_bound(&s, s.avail_in);
954 s.next_out = out = xmalloc(s.avail_out);
955 while (git_deflate(&s, Z_FINISH) == Z_OK)
956 ; /* nothing */
957 git_deflate_end(&s);
959 /* Determine if we should auto-checkpoint. */
960 if ((max_packsize
961 && (pack_size + PACK_SIZE_THRESHOLD + s.total_out) > max_packsize)
962 || (pack_size + PACK_SIZE_THRESHOLD + s.total_out) < pack_size) {
964 /* This new object needs to *not* have the current pack_id. */
965 e->pack_id = pack_id + 1;
966 cycle_packfile();
968 /* We cannot carry a delta into the new pack. */
969 if (delta) {
970 FREE_AND_NULL(delta);
972 git_deflate_init(&s, pack_compression_level);
973 s.next_in = (void *)dat->buf;
974 s.avail_in = dat->len;
975 s.avail_out = git_deflate_bound(&s, s.avail_in);
976 s.next_out = out = xrealloc(out, s.avail_out);
977 while (git_deflate(&s, Z_FINISH) == Z_OK)
978 ; /* nothing */
979 git_deflate_end(&s);
983 e->type = type;
984 e->pack_id = pack_id;
985 e->idx.offset = pack_size;
986 object_count++;
987 object_count_by_type[type]++;
989 crc32_begin(pack_file);
991 if (delta) {
992 off_t ofs = e->idx.offset - last->offset;
993 unsigned pos = sizeof(hdr) - 1;
995 delta_count_by_type[type]++;
996 e->depth = last->depth + 1;
998 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
999 OBJ_OFS_DELTA, deltalen);
1000 hashwrite(pack_file, hdr, hdrlen);
1001 pack_size += hdrlen;
1003 hdr[pos] = ofs & 127;
1004 while (ofs >>= 7)
1005 hdr[--pos] = 128 | (--ofs & 127);
1006 hashwrite(pack_file, hdr + pos, sizeof(hdr) - pos);
1007 pack_size += sizeof(hdr) - pos;
1008 } else {
1009 e->depth = 0;
1010 hdrlen = encode_in_pack_object_header(hdr, sizeof(hdr),
1011 type, dat->len);
1012 hashwrite(pack_file, hdr, hdrlen);
1013 pack_size += hdrlen;
1016 hashwrite(pack_file, out, s.total_out);
1017 pack_size += s.total_out;
1019 e->idx.crc32 = crc32_end(pack_file);
1021 free(out);
1022 free(delta);
1023 if (last) {
1024 if (last->no_swap) {
1025 last->data = *dat;
1026 } else {
1027 strbuf_swap(&last->data, dat);
1029 last->offset = e->idx.offset;
1030 last->depth = e->depth;
1032 return 0;
1035 static void truncate_pack(struct hashfile_checkpoint *checkpoint)
1037 if (hashfile_truncate(pack_file, checkpoint))
1038 die_errno("cannot truncate pack to skip duplicate");
1039 pack_size = checkpoint->offset;
1042 static void stream_blob(uintmax_t len, struct object_id *oidout, uintmax_t mark)
1044 size_t in_sz = 64 * 1024, out_sz = 64 * 1024;
1045 unsigned char *in_buf = xmalloc(in_sz);
1046 unsigned char *out_buf = xmalloc(out_sz);
1047 struct object_entry *e;
1048 struct object_id oid;
1049 unsigned long hdrlen;
1050 off_t offset;
1051 git_hash_ctx c;
1052 git_zstream s;
1053 struct hashfile_checkpoint checkpoint;
1054 int status = Z_OK;
1056 /* Determine if we should auto-checkpoint. */
1057 if ((max_packsize
1058 && (pack_size + PACK_SIZE_THRESHOLD + len) > max_packsize)
1059 || (pack_size + PACK_SIZE_THRESHOLD + len) < pack_size)
1060 cycle_packfile();
1062 hashfile_checkpoint(pack_file, &checkpoint);
1063 offset = checkpoint.offset;
1065 hdrlen = xsnprintf((char *)out_buf, out_sz, "blob %" PRIuMAX, len) + 1;
1067 the_hash_algo->init_fn(&c);
1068 the_hash_algo->update_fn(&c, out_buf, hdrlen);
1070 crc32_begin(pack_file);
1072 git_deflate_init(&s, pack_compression_level);
1074 hdrlen = encode_in_pack_object_header(out_buf, out_sz, OBJ_BLOB, len);
1076 s.next_out = out_buf + hdrlen;
1077 s.avail_out = out_sz - hdrlen;
1079 while (status != Z_STREAM_END) {
1080 if (0 < len && !s.avail_in) {
1081 size_t cnt = in_sz < len ? in_sz : (size_t)len;
1082 size_t n = fread(in_buf, 1, cnt, stdin);
1083 if (!n && feof(stdin))
1084 die("EOF in data (%" PRIuMAX " bytes remaining)", len);
1086 the_hash_algo->update_fn(&c, in_buf, n);
1087 s.next_in = in_buf;
1088 s.avail_in = n;
1089 len -= n;
1092 status = git_deflate(&s, len ? 0 : Z_FINISH);
1094 if (!s.avail_out || status == Z_STREAM_END) {
1095 size_t n = s.next_out - out_buf;
1096 hashwrite(pack_file, out_buf, n);
1097 pack_size += n;
1098 s.next_out = out_buf;
1099 s.avail_out = out_sz;
1102 switch (status) {
1103 case Z_OK:
1104 case Z_BUF_ERROR:
1105 case Z_STREAM_END:
1106 continue;
1107 default:
1108 die("unexpected deflate failure: %d", status);
1111 git_deflate_end(&s);
1112 the_hash_algo->final_fn(oid.hash, &c);
1114 if (oidout)
1115 oidcpy(oidout, &oid);
1117 e = insert_object(&oid);
1119 if (mark)
1120 insert_mark(mark, e);
1122 if (e->idx.offset) {
1123 duplicate_count_by_type[OBJ_BLOB]++;
1124 truncate_pack(&checkpoint);
1126 } else if (find_sha1_pack(oid.hash,
1127 get_all_packs(the_repository))) {
1128 e->type = OBJ_BLOB;
1129 e->pack_id = MAX_PACK_ID;
1130 e->idx.offset = 1; /* just not zero! */
1131 duplicate_count_by_type[OBJ_BLOB]++;
1132 truncate_pack(&checkpoint);
1134 } else {
1135 e->depth = 0;
1136 e->type = OBJ_BLOB;
1137 e->pack_id = pack_id;
1138 e->idx.offset = offset;
1139 e->idx.crc32 = crc32_end(pack_file);
1140 object_count++;
1141 object_count_by_type[OBJ_BLOB]++;
1144 free(in_buf);
1145 free(out_buf);
1148 /* All calls must be guarded by find_object() or find_mark() to
1149 * ensure the 'struct object_entry' passed was written by this
1150 * process instance. We unpack the entry by the offset, avoiding
1151 * the need for the corresponding .idx file. This unpacking rule
1152 * works because we only use OBJ_REF_DELTA within the packfiles
1153 * created by fast-import.
1155 * oe must not be NULL. Such an oe usually comes from giving
1156 * an unknown SHA-1 to find_object() or an undefined mark to
1157 * find_mark(). Callers must test for this condition and use
1158 * the standard read_sha1_file() when it happens.
1160 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1161 * find_mark(), where the mark was reloaded from an existing marks
1162 * file and is referencing an object that this fast-import process
1163 * instance did not write out to a packfile. Callers must test for
1164 * this condition and use read_sha1_file() instead.
1166 static void *gfi_unpack_entry(
1167 struct object_entry *oe,
1168 unsigned long *sizep)
1170 enum object_type type;
1171 struct packed_git *p = all_packs[oe->pack_id];
1172 if (p == pack_data && p->pack_size < (pack_size + the_hash_algo->rawsz)) {
1173 /* The object is stored in the packfile we are writing to
1174 * and we have modified it since the last time we scanned
1175 * back to read a previously written object. If an old
1176 * window covered [p->pack_size, p->pack_size + rawsz) its
1177 * data is stale and is not valid. Closing all windows
1178 * and updating the packfile length ensures we can read
1179 * the newly written data.
1181 close_pack_windows(p);
1182 hashflush(pack_file);
1184 /* We have to offer rawsz bytes additional on the end of
1185 * the packfile as the core unpacker code assumes the
1186 * footer is present at the file end and must promise
1187 * at least rawsz bytes within any window it maps. But
1188 * we don't actually create the footer here.
1190 p->pack_size = pack_size + the_hash_algo->rawsz;
1192 return unpack_entry(the_repository, p, oe->idx.offset, &type, sizep);
1195 static const char *get_mode(const char *str, uint16_t *modep)
1197 unsigned char c;
1198 uint16_t mode = 0;
1200 while ((c = *str++) != ' ') {
1201 if (c < '0' || c > '7')
1202 return NULL;
1203 mode = (mode << 3) + (c - '0');
1205 *modep = mode;
1206 return str;
1209 static void load_tree(struct tree_entry *root)
1211 struct object_id *oid = &root->versions[1].oid;
1212 struct object_entry *myoe;
1213 struct tree_content *t;
1214 unsigned long size;
1215 char *buf;
1216 const char *c;
1218 root->tree = t = new_tree_content(8);
1219 if (is_null_oid(oid))
1220 return;
1222 myoe = find_object(oid);
1223 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1224 if (myoe->type != OBJ_TREE)
1225 die("Not a tree: %s", oid_to_hex(oid));
1226 t->delta_depth = myoe->depth;
1227 buf = gfi_unpack_entry(myoe, &size);
1228 if (!buf)
1229 die("Can't load tree %s", oid_to_hex(oid));
1230 } else {
1231 enum object_type type;
1232 buf = read_object_file(oid, &type, &size);
1233 if (!buf || type != OBJ_TREE)
1234 die("Can't load tree %s", oid_to_hex(oid));
1237 c = buf;
1238 while (c != (buf + size)) {
1239 struct tree_entry *e = new_tree_entry();
1241 if (t->entry_count == t->entry_capacity)
1242 root->tree = t = grow_tree_content(t, t->entry_count);
1243 t->entries[t->entry_count++] = e;
1245 e->tree = NULL;
1246 c = get_mode(c, &e->versions[1].mode);
1247 if (!c)
1248 die("Corrupt mode in %s", oid_to_hex(oid));
1249 e->versions[0].mode = e->versions[1].mode;
1250 e->name = to_atom(c, strlen(c));
1251 c += e->name->str_len + 1;
1252 hashcpy(e->versions[0].oid.hash, (unsigned char *)c);
1253 hashcpy(e->versions[1].oid.hash, (unsigned char *)c);
1254 c += the_hash_algo->rawsz;
1256 free(buf);
1259 static int tecmp0 (const void *_a, const void *_b)
1261 struct tree_entry *a = *((struct tree_entry**)_a);
1262 struct tree_entry *b = *((struct tree_entry**)_b);
1263 return base_name_compare(
1264 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1265 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1268 static int tecmp1 (const void *_a, const void *_b)
1270 struct tree_entry *a = *((struct tree_entry**)_a);
1271 struct tree_entry *b = *((struct tree_entry**)_b);
1272 return base_name_compare(
1273 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1274 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1277 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1279 size_t maxlen = 0;
1280 unsigned int i;
1282 if (!v)
1283 QSORT(t->entries, t->entry_count, tecmp0);
1284 else
1285 QSORT(t->entries, t->entry_count, tecmp1);
1287 for (i = 0; i < t->entry_count; i++) {
1288 if (t->entries[i]->versions[v].mode)
1289 maxlen += t->entries[i]->name->str_len + 34;
1292 strbuf_reset(b);
1293 strbuf_grow(b, maxlen);
1294 for (i = 0; i < t->entry_count; i++) {
1295 struct tree_entry *e = t->entries[i];
1296 if (!e->versions[v].mode)
1297 continue;
1298 strbuf_addf(b, "%o %s%c",
1299 (unsigned int)(e->versions[v].mode & ~NO_DELTA),
1300 e->name->str_dat, '\0');
1301 strbuf_add(b, e->versions[v].oid.hash, the_hash_algo->rawsz);
1305 static void store_tree(struct tree_entry *root)
1307 struct tree_content *t;
1308 unsigned int i, j, del;
1309 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1310 struct object_entry *le = NULL;
1312 if (!is_null_oid(&root->versions[1].oid))
1313 return;
1315 if (!root->tree)
1316 load_tree(root);
1317 t = root->tree;
1319 for (i = 0; i < t->entry_count; i++) {
1320 if (t->entries[i]->tree)
1321 store_tree(t->entries[i]);
1324 if (!(root->versions[0].mode & NO_DELTA))
1325 le = find_object(&root->versions[0].oid);
1326 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1327 mktree(t, 0, &old_tree);
1328 lo.data = old_tree;
1329 lo.offset = le->idx.offset;
1330 lo.depth = t->delta_depth;
1333 mktree(t, 1, &new_tree);
1334 store_object(OBJ_TREE, &new_tree, &lo, &root->versions[1].oid, 0);
1336 t->delta_depth = lo.depth;
1337 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1338 struct tree_entry *e = t->entries[i];
1339 if (e->versions[1].mode) {
1340 e->versions[0].mode = e->versions[1].mode;
1341 oidcpy(&e->versions[0].oid, &e->versions[1].oid);
1342 t->entries[j++] = e;
1343 } else {
1344 release_tree_entry(e);
1345 del++;
1348 t->entry_count -= del;
1351 static void tree_content_replace(
1352 struct tree_entry *root,
1353 const struct object_id *oid,
1354 const uint16_t mode,
1355 struct tree_content *newtree)
1357 if (!S_ISDIR(mode))
1358 die("Root cannot be a non-directory");
1359 oidclr(&root->versions[0].oid);
1360 oidcpy(&root->versions[1].oid, oid);
1361 if (root->tree)
1362 release_tree_content_recursive(root->tree);
1363 root->tree = newtree;
1366 static int tree_content_set(
1367 struct tree_entry *root,
1368 const char *p,
1369 const struct object_id *oid,
1370 const uint16_t mode,
1371 struct tree_content *subtree)
1373 struct tree_content *t;
1374 const char *slash1;
1375 unsigned int i, n;
1376 struct tree_entry *e;
1378 slash1 = strchrnul(p, '/');
1379 n = slash1 - p;
1380 if (!n)
1381 die("Empty path component found in input");
1382 if (!*slash1 && !S_ISDIR(mode) && subtree)
1383 die("Non-directories cannot have subtrees");
1385 if (!root->tree)
1386 load_tree(root);
1387 t = root->tree;
1388 for (i = 0; i < t->entry_count; i++) {
1389 e = t->entries[i];
1390 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1391 if (!*slash1) {
1392 if (!S_ISDIR(mode)
1393 && e->versions[1].mode == mode
1394 && oideq(&e->versions[1].oid, oid))
1395 return 0;
1396 e->versions[1].mode = mode;
1397 oidcpy(&e->versions[1].oid, oid);
1398 if (e->tree)
1399 release_tree_content_recursive(e->tree);
1400 e->tree = subtree;
1403 * We need to leave e->versions[0].sha1 alone
1404 * to avoid modifying the preimage tree used
1405 * when writing out the parent directory.
1406 * But after replacing the subdir with a
1407 * completely different one, it's not a good
1408 * delta base any more, and besides, we've
1409 * thrown away the tree entries needed to
1410 * make a delta against it.
1412 * So let's just explicitly disable deltas
1413 * for the subtree.
1415 if (S_ISDIR(e->versions[0].mode))
1416 e->versions[0].mode |= NO_DELTA;
1418 oidclr(&root->versions[1].oid);
1419 return 1;
1421 if (!S_ISDIR(e->versions[1].mode)) {
1422 e->tree = new_tree_content(8);
1423 e->versions[1].mode = S_IFDIR;
1425 if (!e->tree)
1426 load_tree(e);
1427 if (tree_content_set(e, slash1 + 1, oid, mode, subtree)) {
1428 oidclr(&root->versions[1].oid);
1429 return 1;
1431 return 0;
1435 if (t->entry_count == t->entry_capacity)
1436 root->tree = t = grow_tree_content(t, t->entry_count);
1437 e = new_tree_entry();
1438 e->name = to_atom(p, n);
1439 e->versions[0].mode = 0;
1440 oidclr(&e->versions[0].oid);
1441 t->entries[t->entry_count++] = e;
1442 if (*slash1) {
1443 e->tree = new_tree_content(8);
1444 e->versions[1].mode = S_IFDIR;
1445 tree_content_set(e, slash1 + 1, oid, mode, subtree);
1446 } else {
1447 e->tree = subtree;
1448 e->versions[1].mode = mode;
1449 oidcpy(&e->versions[1].oid, oid);
1451 oidclr(&root->versions[1].oid);
1452 return 1;
1455 static int tree_content_remove(
1456 struct tree_entry *root,
1457 const char *p,
1458 struct tree_entry *backup_leaf,
1459 int allow_root)
1461 struct tree_content *t;
1462 const char *slash1;
1463 unsigned int i, n;
1464 struct tree_entry *e;
1466 slash1 = strchrnul(p, '/');
1467 n = slash1 - p;
1469 if (!root->tree)
1470 load_tree(root);
1472 if (!*p && allow_root) {
1473 e = root;
1474 goto del_entry;
1477 t = root->tree;
1478 for (i = 0; i < t->entry_count; i++) {
1479 e = t->entries[i];
1480 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1481 if (*slash1 && !S_ISDIR(e->versions[1].mode))
1483 * If p names a file in some subdirectory, and a
1484 * file or symlink matching the name of the
1485 * parent directory of p exists, then p cannot
1486 * exist and need not be deleted.
1488 return 1;
1489 if (!*slash1 || !S_ISDIR(e->versions[1].mode))
1490 goto del_entry;
1491 if (!e->tree)
1492 load_tree(e);
1493 if (tree_content_remove(e, slash1 + 1, backup_leaf, 0)) {
1494 for (n = 0; n < e->tree->entry_count; n++) {
1495 if (e->tree->entries[n]->versions[1].mode) {
1496 oidclr(&root->versions[1].oid);
1497 return 1;
1500 backup_leaf = NULL;
1501 goto del_entry;
1503 return 0;
1506 return 0;
1508 del_entry:
1509 if (backup_leaf)
1510 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1511 else if (e->tree)
1512 release_tree_content_recursive(e->tree);
1513 e->tree = NULL;
1514 e->versions[1].mode = 0;
1515 oidclr(&e->versions[1].oid);
1516 oidclr(&root->versions[1].oid);
1517 return 1;
1520 static int tree_content_get(
1521 struct tree_entry *root,
1522 const char *p,
1523 struct tree_entry *leaf,
1524 int allow_root)
1526 struct tree_content *t;
1527 const char *slash1;
1528 unsigned int i, n;
1529 struct tree_entry *e;
1531 slash1 = strchrnul(p, '/');
1532 n = slash1 - p;
1533 if (!n && !allow_root)
1534 die("Empty path component found in input");
1536 if (!root->tree)
1537 load_tree(root);
1539 if (!n) {
1540 e = root;
1541 goto found_entry;
1544 t = root->tree;
1545 for (i = 0; i < t->entry_count; i++) {
1546 e = t->entries[i];
1547 if (e->name->str_len == n && !fspathncmp(p, e->name->str_dat, n)) {
1548 if (!*slash1)
1549 goto found_entry;
1550 if (!S_ISDIR(e->versions[1].mode))
1551 return 0;
1552 if (!e->tree)
1553 load_tree(e);
1554 return tree_content_get(e, slash1 + 1, leaf, 0);
1557 return 0;
1559 found_entry:
1560 memcpy(leaf, e, sizeof(*leaf));
1561 if (e->tree && is_null_oid(&e->versions[1].oid))
1562 leaf->tree = dup_tree_content(e->tree);
1563 else
1564 leaf->tree = NULL;
1565 return 1;
1568 static int update_branch(struct branch *b)
1570 static const char *msg = "fast-import";
1571 struct ref_transaction *transaction;
1572 struct object_id old_oid;
1573 struct strbuf err = STRBUF_INIT;
1575 if (is_null_oid(&b->oid)) {
1576 if (b->delete)
1577 delete_ref(NULL, b->name, NULL, 0);
1578 return 0;
1580 if (read_ref(b->name, &old_oid))
1581 oidclr(&old_oid);
1582 if (!force_update && !is_null_oid(&old_oid)) {
1583 struct commit *old_cmit, *new_cmit;
1585 old_cmit = lookup_commit_reference_gently(the_repository,
1586 &old_oid, 0);
1587 new_cmit = lookup_commit_reference_gently(the_repository,
1588 &b->oid, 0);
1589 if (!old_cmit || !new_cmit)
1590 return error("Branch %s is missing commits.", b->name);
1592 if (!in_merge_bases(old_cmit, new_cmit)) {
1593 warning("Not updating %s"
1594 " (new tip %s does not contain %s)",
1595 b->name, oid_to_hex(&b->oid),
1596 oid_to_hex(&old_oid));
1597 return -1;
1600 transaction = ref_transaction_begin(&err);
1601 if (!transaction ||
1602 ref_transaction_update(transaction, b->name, &b->oid, &old_oid,
1603 0, msg, &err) ||
1604 ref_transaction_commit(transaction, &err)) {
1605 ref_transaction_free(transaction);
1606 error("%s", err.buf);
1607 strbuf_release(&err);
1608 return -1;
1610 ref_transaction_free(transaction);
1611 strbuf_release(&err);
1612 return 0;
1615 static void dump_branches(void)
1617 unsigned int i;
1618 struct branch *b;
1620 for (i = 0; i < branch_table_sz; i++) {
1621 for (b = branch_table[i]; b; b = b->table_next_branch)
1622 failure |= update_branch(b);
1626 static void dump_tags(void)
1628 static const char *msg = "fast-import";
1629 struct tag *t;
1630 struct strbuf ref_name = STRBUF_INIT;
1631 struct strbuf err = STRBUF_INIT;
1632 struct ref_transaction *transaction;
1634 transaction = ref_transaction_begin(&err);
1635 if (!transaction) {
1636 failure |= error("%s", err.buf);
1637 goto cleanup;
1639 for (t = first_tag; t; t = t->next_tag) {
1640 strbuf_reset(&ref_name);
1641 strbuf_addf(&ref_name, "refs/tags/%s", t->name);
1643 if (ref_transaction_update(transaction, ref_name.buf,
1644 &t->oid, NULL, 0, msg, &err)) {
1645 failure |= error("%s", err.buf);
1646 goto cleanup;
1649 if (ref_transaction_commit(transaction, &err))
1650 failure |= error("%s", err.buf);
1652 cleanup:
1653 ref_transaction_free(transaction);
1654 strbuf_release(&ref_name);
1655 strbuf_release(&err);
1658 static void dump_marks_helper(FILE *f,
1659 uintmax_t base,
1660 struct mark_set *m)
1662 uintmax_t k;
1663 if (m->shift) {
1664 for (k = 0; k < 1024; k++) {
1665 if (m->data.sets[k])
1666 dump_marks_helper(f, base + (k << m->shift),
1667 m->data.sets[k]);
1669 } else {
1670 for (k = 0; k < 1024; k++) {
1671 if (m->data.marked[k])
1672 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1673 oid_to_hex(&m->data.marked[k]->idx.oid));
1678 static void dump_marks(void)
1680 struct lock_file mark_lock = LOCK_INIT;
1681 FILE *f;
1683 if (!export_marks_file || (import_marks_file && !import_marks_file_done))
1684 return;
1686 if (safe_create_leading_directories_const(export_marks_file)) {
1687 failure |= error_errno("unable to create leading directories of %s",
1688 export_marks_file);
1689 return;
1692 if (hold_lock_file_for_update(&mark_lock, export_marks_file, 0) < 0) {
1693 failure |= error_errno("Unable to write marks file %s",
1694 export_marks_file);
1695 return;
1698 f = fdopen_lock_file(&mark_lock, "w");
1699 if (!f) {
1700 int saved_errno = errno;
1701 rollback_lock_file(&mark_lock);
1702 failure |= error("Unable to write marks file %s: %s",
1703 export_marks_file, strerror(saved_errno));
1704 return;
1707 dump_marks_helper(f, 0, marks);
1708 if (commit_lock_file(&mark_lock)) {
1709 failure |= error_errno("Unable to write file %s",
1710 export_marks_file);
1711 return;
1715 static void read_marks(void)
1717 char line[512];
1718 FILE *f = fopen(import_marks_file, "r");
1719 if (f)
1721 else if (import_marks_file_ignore_missing && errno == ENOENT)
1722 goto done; /* Marks file does not exist */
1723 else
1724 die_errno("cannot read '%s'", import_marks_file);
1725 while (fgets(line, sizeof(line), f)) {
1726 uintmax_t mark;
1727 char *end;
1728 struct object_id oid;
1729 struct object_entry *e;
1731 end = strchr(line, '\n');
1732 if (line[0] != ':' || !end)
1733 die("corrupt mark line: %s", line);
1734 *end = 0;
1735 mark = strtoumax(line + 1, &end, 10);
1736 if (!mark || end == line + 1
1737 || *end != ' ' || get_oid_hex(end + 1, &oid))
1738 die("corrupt mark line: %s", line);
1739 e = find_object(&oid);
1740 if (!e) {
1741 enum object_type type = oid_object_info(the_repository,
1742 &oid, NULL);
1743 if (type < 0)
1744 die("object not found: %s", oid_to_hex(&oid));
1745 e = insert_object(&oid);
1746 e->type = type;
1747 e->pack_id = MAX_PACK_ID;
1748 e->idx.offset = 1; /* just not zero! */
1750 insert_mark(mark, e);
1752 fclose(f);
1753 done:
1754 import_marks_file_done = 1;
1758 static int read_next_command(void)
1760 static int stdin_eof = 0;
1762 if (stdin_eof) {
1763 unread_command_buf = 0;
1764 return EOF;
1767 for (;;) {
1768 if (unread_command_buf) {
1769 unread_command_buf = 0;
1770 } else {
1771 struct recent_command *rc;
1773 stdin_eof = strbuf_getline_lf(&command_buf, stdin);
1774 if (stdin_eof)
1775 return EOF;
1777 if (!seen_data_command
1778 && !starts_with(command_buf.buf, "feature ")
1779 && !starts_with(command_buf.buf, "option ")) {
1780 parse_argv();
1783 rc = rc_free;
1784 if (rc)
1785 rc_free = rc->next;
1786 else {
1787 rc = cmd_hist.next;
1788 cmd_hist.next = rc->next;
1789 cmd_hist.next->prev = &cmd_hist;
1790 free(rc->buf);
1793 rc->buf = xstrdup(command_buf.buf);
1794 rc->prev = cmd_tail;
1795 rc->next = cmd_hist.prev;
1796 rc->prev->next = rc;
1797 cmd_tail = rc;
1799 if (command_buf.buf[0] == '#')
1800 continue;
1801 return 0;
1805 static void skip_optional_lf(void)
1807 int term_char = fgetc(stdin);
1808 if (term_char != '\n' && term_char != EOF)
1809 ungetc(term_char, stdin);
1812 static void parse_mark(void)
1814 const char *v;
1815 if (skip_prefix(command_buf.buf, "mark :", &v)) {
1816 next_mark = strtoumax(v, NULL, 10);
1817 read_next_command();
1819 else
1820 next_mark = 0;
1823 static void parse_original_identifier(void)
1825 const char *v;
1826 if (skip_prefix(command_buf.buf, "original-oid ", &v))
1827 read_next_command();
1830 static int parse_data(struct strbuf *sb, uintmax_t limit, uintmax_t *len_res)
1832 const char *data;
1833 strbuf_reset(sb);
1835 if (!skip_prefix(command_buf.buf, "data ", &data))
1836 die("Expected 'data n' command, found: %s", command_buf.buf);
1838 if (skip_prefix(data, "<<", &data)) {
1839 char *term = xstrdup(data);
1840 size_t term_len = command_buf.len - (data - command_buf.buf);
1842 for (;;) {
1843 if (strbuf_getline_lf(&command_buf, stdin) == EOF)
1844 die("EOF in data (terminator '%s' not found)", term);
1845 if (term_len == command_buf.len
1846 && !strcmp(term, command_buf.buf))
1847 break;
1848 strbuf_addbuf(sb, &command_buf);
1849 strbuf_addch(sb, '\n');
1851 free(term);
1853 else {
1854 uintmax_t len = strtoumax(data, NULL, 10);
1855 size_t n = 0, length = (size_t)len;
1857 if (limit && limit < len) {
1858 *len_res = len;
1859 return 0;
1861 if (length < len)
1862 die("data is too large to use in this context");
1864 while (n < length) {
1865 size_t s = strbuf_fread(sb, length - n, stdin);
1866 if (!s && feof(stdin))
1867 die("EOF in data (%lu bytes remaining)",
1868 (unsigned long)(length - n));
1869 n += s;
1873 skip_optional_lf();
1874 return 1;
1877 static int validate_raw_date(const char *src, struct strbuf *result)
1879 const char *orig_src = src;
1880 char *endp;
1881 unsigned long num;
1883 errno = 0;
1885 num = strtoul(src, &endp, 10);
1886 /* NEEDSWORK: perhaps check for reasonable values? */
1887 if (errno || endp == src || *endp != ' ')
1888 return -1;
1890 src = endp + 1;
1891 if (*src != '-' && *src != '+')
1892 return -1;
1894 num = strtoul(src + 1, &endp, 10);
1895 if (errno || endp == src + 1 || *endp || 1400 < num)
1896 return -1;
1898 strbuf_addstr(result, orig_src);
1899 return 0;
1902 static char *parse_ident(const char *buf)
1904 const char *ltgt;
1905 size_t name_len;
1906 struct strbuf ident = STRBUF_INIT;
1908 /* ensure there is a space delimiter even if there is no name */
1909 if (*buf == '<')
1910 --buf;
1912 ltgt = buf + strcspn(buf, "<>");
1913 if (*ltgt != '<')
1914 die("Missing < in ident string: %s", buf);
1915 if (ltgt != buf && ltgt[-1] != ' ')
1916 die("Missing space before < in ident string: %s", buf);
1917 ltgt = ltgt + 1 + strcspn(ltgt + 1, "<>");
1918 if (*ltgt != '>')
1919 die("Missing > in ident string: %s", buf);
1920 ltgt++;
1921 if (*ltgt != ' ')
1922 die("Missing space after > in ident string: %s", buf);
1923 ltgt++;
1924 name_len = ltgt - buf;
1925 strbuf_add(&ident, buf, name_len);
1927 switch (whenspec) {
1928 case WHENSPEC_RAW:
1929 if (validate_raw_date(ltgt, &ident) < 0)
1930 die("Invalid raw date \"%s\" in ident: %s", ltgt, buf);
1931 break;
1932 case WHENSPEC_RFC2822:
1933 if (parse_date(ltgt, &ident) < 0)
1934 die("Invalid rfc2822 date \"%s\" in ident: %s", ltgt, buf);
1935 break;
1936 case WHENSPEC_NOW:
1937 if (strcmp("now", ltgt))
1938 die("Date in ident must be 'now': %s", buf);
1939 datestamp(&ident);
1940 break;
1943 return strbuf_detach(&ident, NULL);
1946 static void parse_and_store_blob(
1947 struct last_object *last,
1948 struct object_id *oidout,
1949 uintmax_t mark)
1951 static struct strbuf buf = STRBUF_INIT;
1952 uintmax_t len;
1954 if (parse_data(&buf, big_file_threshold, &len))
1955 store_object(OBJ_BLOB, &buf, last, oidout, mark);
1956 else {
1957 if (last) {
1958 strbuf_release(&last->data);
1959 last->offset = 0;
1960 last->depth = 0;
1962 stream_blob(len, oidout, mark);
1963 skip_optional_lf();
1967 static void parse_new_blob(void)
1969 read_next_command();
1970 parse_mark();
1971 parse_original_identifier();
1972 parse_and_store_blob(&last_blob, NULL, next_mark);
1975 static void unload_one_branch(void)
1977 while (cur_active_branches
1978 && cur_active_branches >= max_active_branches) {
1979 uintmax_t min_commit = ULONG_MAX;
1980 struct branch *e, *l = NULL, *p = NULL;
1982 for (e = active_branches; e; e = e->active_next_branch) {
1983 if (e->last_commit < min_commit) {
1984 p = l;
1985 min_commit = e->last_commit;
1987 l = e;
1990 if (p) {
1991 e = p->active_next_branch;
1992 p->active_next_branch = e->active_next_branch;
1993 } else {
1994 e = active_branches;
1995 active_branches = e->active_next_branch;
1997 e->active = 0;
1998 e->active_next_branch = NULL;
1999 if (e->branch_tree.tree) {
2000 release_tree_content_recursive(e->branch_tree.tree);
2001 e->branch_tree.tree = NULL;
2003 cur_active_branches--;
2007 static void load_branch(struct branch *b)
2009 load_tree(&b->branch_tree);
2010 if (!b->active) {
2011 b->active = 1;
2012 b->active_next_branch = active_branches;
2013 active_branches = b;
2014 cur_active_branches++;
2015 branch_load_count++;
2019 static unsigned char convert_num_notes_to_fanout(uintmax_t num_notes)
2021 unsigned char fanout = 0;
2022 while ((num_notes >>= 8))
2023 fanout++;
2024 return fanout;
2027 static void construct_path_with_fanout(const char *hex_sha1,
2028 unsigned char fanout, char *path)
2030 unsigned int i = 0, j = 0;
2031 if (fanout >= the_hash_algo->rawsz)
2032 die("Too large fanout (%u)", fanout);
2033 while (fanout) {
2034 path[i++] = hex_sha1[j++];
2035 path[i++] = hex_sha1[j++];
2036 path[i++] = '/';
2037 fanout--;
2039 memcpy(path + i, hex_sha1 + j, the_hash_algo->hexsz - j);
2040 path[i + the_hash_algo->hexsz - j] = '\0';
2043 static uintmax_t do_change_note_fanout(
2044 struct tree_entry *orig_root, struct tree_entry *root,
2045 char *hex_oid, unsigned int hex_oid_len,
2046 char *fullpath, unsigned int fullpath_len,
2047 unsigned char fanout)
2049 struct tree_content *t;
2050 struct tree_entry *e, leaf;
2051 unsigned int i, tmp_hex_oid_len, tmp_fullpath_len;
2052 uintmax_t num_notes = 0;
2053 struct object_id oid;
2054 /* hex oid + '/' between each pair of hex digits + NUL */
2055 char realpath[GIT_MAX_HEXSZ + ((GIT_MAX_HEXSZ / 2) - 1) + 1];
2056 const unsigned hexsz = the_hash_algo->hexsz;
2058 if (!root->tree)
2059 load_tree(root);
2060 t = root->tree;
2062 for (i = 0; t && i < t->entry_count; i++) {
2063 e = t->entries[i];
2064 tmp_hex_oid_len = hex_oid_len + e->name->str_len;
2065 tmp_fullpath_len = fullpath_len;
2068 * We're interested in EITHER existing note entries (entries
2069 * with exactly 40 hex chars in path, not including directory
2070 * separators), OR directory entries that may contain note
2071 * entries (with < 40 hex chars in path).
2072 * Also, each path component in a note entry must be a multiple
2073 * of 2 chars.
2075 if (!e->versions[1].mode ||
2076 tmp_hex_oid_len > hexsz ||
2077 e->name->str_len % 2)
2078 continue;
2080 /* This _may_ be a note entry, or a subdir containing notes */
2081 memcpy(hex_oid + hex_oid_len, e->name->str_dat,
2082 e->name->str_len);
2083 if (tmp_fullpath_len)
2084 fullpath[tmp_fullpath_len++] = '/';
2085 memcpy(fullpath + tmp_fullpath_len, e->name->str_dat,
2086 e->name->str_len);
2087 tmp_fullpath_len += e->name->str_len;
2088 fullpath[tmp_fullpath_len] = '\0';
2090 if (tmp_hex_oid_len == hexsz && !get_oid_hex(hex_oid, &oid)) {
2091 /* This is a note entry */
2092 if (fanout == 0xff) {
2093 /* Counting mode, no rename */
2094 num_notes++;
2095 continue;
2097 construct_path_with_fanout(hex_oid, fanout, realpath);
2098 if (!strcmp(fullpath, realpath)) {
2099 /* Note entry is in correct location */
2100 num_notes++;
2101 continue;
2104 /* Rename fullpath to realpath */
2105 if (!tree_content_remove(orig_root, fullpath, &leaf, 0))
2106 die("Failed to remove path %s", fullpath);
2107 tree_content_set(orig_root, realpath,
2108 &leaf.versions[1].oid,
2109 leaf.versions[1].mode,
2110 leaf.tree);
2111 } else if (S_ISDIR(e->versions[1].mode)) {
2112 /* This is a subdir that may contain note entries */
2113 num_notes += do_change_note_fanout(orig_root, e,
2114 hex_oid, tmp_hex_oid_len,
2115 fullpath, tmp_fullpath_len, fanout);
2118 /* The above may have reallocated the current tree_content */
2119 t = root->tree;
2121 return num_notes;
2124 static uintmax_t change_note_fanout(struct tree_entry *root,
2125 unsigned char fanout)
2128 * The size of path is due to one slash between every two hex digits,
2129 * plus the terminating NUL. Note that there is no slash at the end, so
2130 * the number of slashes is one less than half the number of hex
2131 * characters.
2133 char hex_oid[GIT_MAX_HEXSZ], path[GIT_MAX_HEXSZ + (GIT_MAX_HEXSZ / 2) - 1 + 1];
2134 return do_change_note_fanout(root, root, hex_oid, 0, path, 0, fanout);
2138 * Given a pointer into a string, parse a mark reference:
2140 * idnum ::= ':' bigint;
2142 * Return the first character after the value in *endptr.
2144 * Complain if the following character is not what is expected,
2145 * either a space or end of the string.
2147 static uintmax_t parse_mark_ref(const char *p, char **endptr)
2149 uintmax_t mark;
2151 assert(*p == ':');
2152 p++;
2153 mark = strtoumax(p, endptr, 10);
2154 if (*endptr == p)
2155 die("No value after ':' in mark: %s", command_buf.buf);
2156 return mark;
2160 * Parse the mark reference, and complain if this is not the end of
2161 * the string.
2163 static uintmax_t parse_mark_ref_eol(const char *p)
2165 char *end;
2166 uintmax_t mark;
2168 mark = parse_mark_ref(p, &end);
2169 if (*end != '\0')
2170 die("Garbage after mark: %s", command_buf.buf);
2171 return mark;
2175 * Parse the mark reference, demanding a trailing space. Return a
2176 * pointer to the space.
2178 static uintmax_t parse_mark_ref_space(const char **p)
2180 uintmax_t mark;
2181 char *end;
2183 mark = parse_mark_ref(*p, &end);
2184 if (*end++ != ' ')
2185 die("Missing space after mark: %s", command_buf.buf);
2186 *p = end;
2187 return mark;
2190 static void file_change_m(const char *p, struct branch *b)
2192 static struct strbuf uq = STRBUF_INIT;
2193 const char *endp;
2194 struct object_entry *oe;
2195 struct object_id oid;
2196 uint16_t mode, inline_data = 0;
2198 p = get_mode(p, &mode);
2199 if (!p)
2200 die("Corrupt mode: %s", command_buf.buf);
2201 switch (mode) {
2202 case 0644:
2203 case 0755:
2204 mode |= S_IFREG;
2205 case S_IFREG | 0644:
2206 case S_IFREG | 0755:
2207 case S_IFLNK:
2208 case S_IFDIR:
2209 case S_IFGITLINK:
2210 /* ok */
2211 break;
2212 default:
2213 die("Corrupt mode: %s", command_buf.buf);
2216 if (*p == ':') {
2217 oe = find_mark(parse_mark_ref_space(&p));
2218 oidcpy(&oid, &oe->idx.oid);
2219 } else if (skip_prefix(p, "inline ", &p)) {
2220 inline_data = 1;
2221 oe = NULL; /* not used with inline_data, but makes gcc happy */
2222 } else {
2223 if (parse_oid_hex(p, &oid, &p))
2224 die("Invalid dataref: %s", command_buf.buf);
2225 oe = find_object(&oid);
2226 if (*p++ != ' ')
2227 die("Missing space after SHA1: %s", command_buf.buf);
2230 strbuf_reset(&uq);
2231 if (!unquote_c_style(&uq, p, &endp)) {
2232 if (*endp)
2233 die("Garbage after path in: %s", command_buf.buf);
2234 p = uq.buf;
2237 /* Git does not track empty, non-toplevel directories. */
2238 if (S_ISDIR(mode) && is_empty_tree_oid(&oid) && *p) {
2239 tree_content_remove(&b->branch_tree, p, NULL, 0);
2240 return;
2243 if (S_ISGITLINK(mode)) {
2244 if (inline_data)
2245 die("Git links cannot be specified 'inline': %s",
2246 command_buf.buf);
2247 else if (oe) {
2248 if (oe->type != OBJ_COMMIT)
2249 die("Not a commit (actually a %s): %s",
2250 type_name(oe->type), command_buf.buf);
2253 * Accept the sha1 without checking; it expected to be in
2254 * another repository.
2256 } else if (inline_data) {
2257 if (S_ISDIR(mode))
2258 die("Directories cannot be specified 'inline': %s",
2259 command_buf.buf);
2260 if (p != uq.buf) {
2261 strbuf_addstr(&uq, p);
2262 p = uq.buf;
2264 while (read_next_command() != EOF) {
2265 const char *v;
2266 if (skip_prefix(command_buf.buf, "cat-blob ", &v))
2267 parse_cat_blob(v);
2268 else {
2269 parse_and_store_blob(&last_blob, &oid, 0);
2270 break;
2273 } else {
2274 enum object_type expected = S_ISDIR(mode) ?
2275 OBJ_TREE: OBJ_BLOB;
2276 enum object_type type = oe ? oe->type :
2277 oid_object_info(the_repository, &oid,
2278 NULL);
2279 if (type < 0)
2280 die("%s not found: %s",
2281 S_ISDIR(mode) ? "Tree" : "Blob",
2282 command_buf.buf);
2283 if (type != expected)
2284 die("Not a %s (actually a %s): %s",
2285 type_name(expected), type_name(type),
2286 command_buf.buf);
2289 if (!*p) {
2290 tree_content_replace(&b->branch_tree, &oid, mode, NULL);
2291 return;
2293 tree_content_set(&b->branch_tree, p, &oid, mode, NULL);
2296 static void file_change_d(const char *p, struct branch *b)
2298 static struct strbuf uq = STRBUF_INIT;
2299 const char *endp;
2301 strbuf_reset(&uq);
2302 if (!unquote_c_style(&uq, p, &endp)) {
2303 if (*endp)
2304 die("Garbage after path in: %s", command_buf.buf);
2305 p = uq.buf;
2307 tree_content_remove(&b->branch_tree, p, NULL, 1);
2310 static void file_change_cr(const char *s, struct branch *b, int rename)
2312 const char *d;
2313 static struct strbuf s_uq = STRBUF_INIT;
2314 static struct strbuf d_uq = STRBUF_INIT;
2315 const char *endp;
2316 struct tree_entry leaf;
2318 strbuf_reset(&s_uq);
2319 if (!unquote_c_style(&s_uq, s, &endp)) {
2320 if (*endp != ' ')
2321 die("Missing space after source: %s", command_buf.buf);
2322 } else {
2323 endp = strchr(s, ' ');
2324 if (!endp)
2325 die("Missing space after source: %s", command_buf.buf);
2326 strbuf_add(&s_uq, s, endp - s);
2328 s = s_uq.buf;
2330 endp++;
2331 if (!*endp)
2332 die("Missing dest: %s", command_buf.buf);
2334 d = endp;
2335 strbuf_reset(&d_uq);
2336 if (!unquote_c_style(&d_uq, d, &endp)) {
2337 if (*endp)
2338 die("Garbage after dest in: %s", command_buf.buf);
2339 d = d_uq.buf;
2342 memset(&leaf, 0, sizeof(leaf));
2343 if (rename)
2344 tree_content_remove(&b->branch_tree, s, &leaf, 1);
2345 else
2346 tree_content_get(&b->branch_tree, s, &leaf, 1);
2347 if (!leaf.versions[1].mode)
2348 die("Path %s not in branch", s);
2349 if (!*d) { /* C "path/to/subdir" "" */
2350 tree_content_replace(&b->branch_tree,
2351 &leaf.versions[1].oid,
2352 leaf.versions[1].mode,
2353 leaf.tree);
2354 return;
2356 tree_content_set(&b->branch_tree, d,
2357 &leaf.versions[1].oid,
2358 leaf.versions[1].mode,
2359 leaf.tree);
2362 static void note_change_n(const char *p, struct branch *b, unsigned char *old_fanout)
2364 static struct strbuf uq = STRBUF_INIT;
2365 struct object_entry *oe;
2366 struct branch *s;
2367 struct object_id oid, commit_oid;
2368 char path[GIT_MAX_RAWSZ * 3];
2369 uint16_t inline_data = 0;
2370 unsigned char new_fanout;
2373 * When loading a branch, we don't traverse its tree to count the real
2374 * number of notes (too expensive to do this for all non-note refs).
2375 * This means that recently loaded notes refs might incorrectly have
2376 * b->num_notes == 0, and consequently, old_fanout might be wrong.
2378 * Fix this by traversing the tree and counting the number of notes
2379 * when b->num_notes == 0. If the notes tree is truly empty, the
2380 * calculation should not take long.
2382 if (b->num_notes == 0 && *old_fanout == 0) {
2383 /* Invoke change_note_fanout() in "counting mode". */
2384 b->num_notes = change_note_fanout(&b->branch_tree, 0xff);
2385 *old_fanout = convert_num_notes_to_fanout(b->num_notes);
2388 /* Now parse the notemodify command. */
2389 /* <dataref> or 'inline' */
2390 if (*p == ':') {
2391 oe = find_mark(parse_mark_ref_space(&p));
2392 oidcpy(&oid, &oe->idx.oid);
2393 } else if (skip_prefix(p, "inline ", &p)) {
2394 inline_data = 1;
2395 oe = NULL; /* not used with inline_data, but makes gcc happy */
2396 } else {
2397 if (parse_oid_hex(p, &oid, &p))
2398 die("Invalid dataref: %s", command_buf.buf);
2399 oe = find_object(&oid);
2400 if (*p++ != ' ')
2401 die("Missing space after SHA1: %s", command_buf.buf);
2404 /* <commit-ish> */
2405 s = lookup_branch(p);
2406 if (s) {
2407 if (is_null_oid(&s->oid))
2408 die("Can't add a note on empty branch.");
2409 oidcpy(&commit_oid, &s->oid);
2410 } else if (*p == ':') {
2411 uintmax_t commit_mark = parse_mark_ref_eol(p);
2412 struct object_entry *commit_oe = find_mark(commit_mark);
2413 if (commit_oe->type != OBJ_COMMIT)
2414 die("Mark :%" PRIuMAX " not a commit", commit_mark);
2415 oidcpy(&commit_oid, &commit_oe->idx.oid);
2416 } else if (!get_oid(p, &commit_oid)) {
2417 unsigned long size;
2418 char *buf = read_object_with_reference(the_repository,
2419 &commit_oid,
2420 commit_type, &size,
2421 &commit_oid);
2422 if (!buf || size < the_hash_algo->hexsz + 6)
2423 die("Not a valid commit: %s", p);
2424 free(buf);
2425 } else
2426 die("Invalid ref name or SHA1 expression: %s", p);
2428 if (inline_data) {
2429 if (p != uq.buf) {
2430 strbuf_addstr(&uq, p);
2431 p = uq.buf;
2433 read_next_command();
2434 parse_and_store_blob(&last_blob, &oid, 0);
2435 } else if (oe) {
2436 if (oe->type != OBJ_BLOB)
2437 die("Not a blob (actually a %s): %s",
2438 type_name(oe->type), command_buf.buf);
2439 } else if (!is_null_oid(&oid)) {
2440 enum object_type type = oid_object_info(the_repository, &oid,
2441 NULL);
2442 if (type < 0)
2443 die("Blob not found: %s", command_buf.buf);
2444 if (type != OBJ_BLOB)
2445 die("Not a blob (actually a %s): %s",
2446 type_name(type), command_buf.buf);
2449 construct_path_with_fanout(oid_to_hex(&commit_oid), *old_fanout, path);
2450 if (tree_content_remove(&b->branch_tree, path, NULL, 0))
2451 b->num_notes--;
2453 if (is_null_oid(&oid))
2454 return; /* nothing to insert */
2456 b->num_notes++;
2457 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2458 construct_path_with_fanout(oid_to_hex(&commit_oid), new_fanout, path);
2459 tree_content_set(&b->branch_tree, path, &oid, S_IFREG | 0644, NULL);
2462 static void file_change_deleteall(struct branch *b)
2464 release_tree_content_recursive(b->branch_tree.tree);
2465 oidclr(&b->branch_tree.versions[0].oid);
2466 oidclr(&b->branch_tree.versions[1].oid);
2467 load_tree(&b->branch_tree);
2468 b->num_notes = 0;
2471 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2473 if (!buf || size < the_hash_algo->hexsz + 6)
2474 die("Not a valid commit: %s", oid_to_hex(&b->oid));
2475 if (memcmp("tree ", buf, 5)
2476 || get_oid_hex(buf + 5, &b->branch_tree.versions[1].oid))
2477 die("The commit %s is corrupt", oid_to_hex(&b->oid));
2478 oidcpy(&b->branch_tree.versions[0].oid,
2479 &b->branch_tree.versions[1].oid);
2482 static void parse_from_existing(struct branch *b)
2484 if (is_null_oid(&b->oid)) {
2485 oidclr(&b->branch_tree.versions[0].oid);
2486 oidclr(&b->branch_tree.versions[1].oid);
2487 } else {
2488 unsigned long size;
2489 char *buf;
2491 buf = read_object_with_reference(the_repository,
2492 &b->oid, commit_type, &size,
2493 &b->oid);
2494 parse_from_commit(b, buf, size);
2495 free(buf);
2499 static int parse_objectish(struct branch *b, const char *objectish)
2501 struct branch *s;
2502 struct object_id oid;
2504 oidcpy(&oid, &b->branch_tree.versions[1].oid);
2506 s = lookup_branch(objectish);
2507 if (b == s)
2508 die("Can't create a branch from itself: %s", b->name);
2509 else if (s) {
2510 struct object_id *t = &s->branch_tree.versions[1].oid;
2511 oidcpy(&b->oid, &s->oid);
2512 oidcpy(&b->branch_tree.versions[0].oid, t);
2513 oidcpy(&b->branch_tree.versions[1].oid, t);
2514 } else if (*objectish == ':') {
2515 uintmax_t idnum = parse_mark_ref_eol(objectish);
2516 struct object_entry *oe = find_mark(idnum);
2517 if (oe->type != OBJ_COMMIT)
2518 die("Mark :%" PRIuMAX " not a commit", idnum);
2519 if (!oideq(&b->oid, &oe->idx.oid)) {
2520 oidcpy(&b->oid, &oe->idx.oid);
2521 if (oe->pack_id != MAX_PACK_ID) {
2522 unsigned long size;
2523 char *buf = gfi_unpack_entry(oe, &size);
2524 parse_from_commit(b, buf, size);
2525 free(buf);
2526 } else
2527 parse_from_existing(b);
2529 } else if (!get_oid(objectish, &b->oid)) {
2530 parse_from_existing(b);
2531 if (is_null_oid(&b->oid))
2532 b->delete = 1;
2534 else
2535 die("Invalid ref name or SHA1 expression: %s", objectish);
2537 if (b->branch_tree.tree && !oideq(&oid, &b->branch_tree.versions[1].oid)) {
2538 release_tree_content_recursive(b->branch_tree.tree);
2539 b->branch_tree.tree = NULL;
2542 read_next_command();
2543 return 1;
2546 static int parse_from(struct branch *b)
2548 const char *from;
2550 if (!skip_prefix(command_buf.buf, "from ", &from))
2551 return 0;
2553 return parse_objectish(b, from);
2556 static int parse_objectish_with_prefix(struct branch *b, const char *prefix)
2558 const char *base;
2560 if (!skip_prefix(command_buf.buf, prefix, &base))
2561 return 0;
2563 return parse_objectish(b, base);
2566 static struct hash_list *parse_merge(unsigned int *count)
2568 struct hash_list *list = NULL, **tail = &list, *n;
2569 const char *from;
2570 struct branch *s;
2572 *count = 0;
2573 while (skip_prefix(command_buf.buf, "merge ", &from)) {
2574 n = xmalloc(sizeof(*n));
2575 s = lookup_branch(from);
2576 if (s)
2577 oidcpy(&n->oid, &s->oid);
2578 else if (*from == ':') {
2579 uintmax_t idnum = parse_mark_ref_eol(from);
2580 struct object_entry *oe = find_mark(idnum);
2581 if (oe->type != OBJ_COMMIT)
2582 die("Mark :%" PRIuMAX " not a commit", idnum);
2583 oidcpy(&n->oid, &oe->idx.oid);
2584 } else if (!get_oid(from, &n->oid)) {
2585 unsigned long size;
2586 char *buf = read_object_with_reference(the_repository,
2587 &n->oid,
2588 commit_type,
2589 &size, &n->oid);
2590 if (!buf || size < the_hash_algo->hexsz + 6)
2591 die("Not a valid commit: %s", from);
2592 free(buf);
2593 } else
2594 die("Invalid ref name or SHA1 expression: %s", from);
2596 n->next = NULL;
2597 *tail = n;
2598 tail = &n->next;
2600 (*count)++;
2601 read_next_command();
2603 return list;
2606 static void parse_new_commit(const char *arg)
2608 static struct strbuf msg = STRBUF_INIT;
2609 struct branch *b;
2610 char *author = NULL;
2611 char *committer = NULL;
2612 char *encoding = NULL;
2613 struct hash_list *merge_list = NULL;
2614 unsigned int merge_count;
2615 unsigned char prev_fanout, new_fanout;
2616 const char *v;
2618 b = lookup_branch(arg);
2619 if (!b)
2620 b = new_branch(arg);
2622 read_next_command();
2623 parse_mark();
2624 parse_original_identifier();
2625 if (skip_prefix(command_buf.buf, "author ", &v)) {
2626 author = parse_ident(v);
2627 read_next_command();
2629 if (skip_prefix(command_buf.buf, "committer ", &v)) {
2630 committer = parse_ident(v);
2631 read_next_command();
2633 if (!committer)
2634 die("Expected committer but didn't get one");
2635 if (skip_prefix(command_buf.buf, "encoding ", &v)) {
2636 encoding = xstrdup(v);
2637 read_next_command();
2639 parse_data(&msg, 0, NULL);
2640 read_next_command();
2641 parse_from(b);
2642 merge_list = parse_merge(&merge_count);
2644 /* ensure the branch is active/loaded */
2645 if (!b->branch_tree.tree || !max_active_branches) {
2646 unload_one_branch();
2647 load_branch(b);
2650 prev_fanout = convert_num_notes_to_fanout(b->num_notes);
2652 /* file_change* */
2653 while (command_buf.len > 0) {
2654 if (skip_prefix(command_buf.buf, "M ", &v))
2655 file_change_m(v, b);
2656 else if (skip_prefix(command_buf.buf, "D ", &v))
2657 file_change_d(v, b);
2658 else if (skip_prefix(command_buf.buf, "R ", &v))
2659 file_change_cr(v, b, 1);
2660 else if (skip_prefix(command_buf.buf, "C ", &v))
2661 file_change_cr(v, b, 0);
2662 else if (skip_prefix(command_buf.buf, "N ", &v))
2663 note_change_n(v, b, &prev_fanout);
2664 else if (!strcmp("deleteall", command_buf.buf))
2665 file_change_deleteall(b);
2666 else if (skip_prefix(command_buf.buf, "ls ", &v))
2667 parse_ls(v, b);
2668 else if (skip_prefix(command_buf.buf, "cat-blob ", &v))
2669 parse_cat_blob(v);
2670 else {
2671 unread_command_buf = 1;
2672 break;
2674 if (read_next_command() == EOF)
2675 break;
2678 new_fanout = convert_num_notes_to_fanout(b->num_notes);
2679 if (new_fanout != prev_fanout)
2680 b->num_notes = change_note_fanout(&b->branch_tree, new_fanout);
2682 /* build the tree and the commit */
2683 store_tree(&b->branch_tree);
2684 oidcpy(&b->branch_tree.versions[0].oid,
2685 &b->branch_tree.versions[1].oid);
2687 strbuf_reset(&new_data);
2688 strbuf_addf(&new_data, "tree %s\n",
2689 oid_to_hex(&b->branch_tree.versions[1].oid));
2690 if (!is_null_oid(&b->oid))
2691 strbuf_addf(&new_data, "parent %s\n",
2692 oid_to_hex(&b->oid));
2693 while (merge_list) {
2694 struct hash_list *next = merge_list->next;
2695 strbuf_addf(&new_data, "parent %s\n",
2696 oid_to_hex(&merge_list->oid));
2697 free(merge_list);
2698 merge_list = next;
2700 strbuf_addf(&new_data,
2701 "author %s\n"
2702 "committer %s\n",
2703 author ? author : committer, committer);
2704 if (encoding)
2705 strbuf_addf(&new_data,
2706 "encoding %s\n",
2707 encoding);
2708 strbuf_addch(&new_data, '\n');
2709 strbuf_addbuf(&new_data, &msg);
2710 free(author);
2711 free(committer);
2712 free(encoding);
2714 if (!store_object(OBJ_COMMIT, &new_data, NULL, &b->oid, next_mark))
2715 b->pack_id = pack_id;
2716 b->last_commit = object_count_by_type[OBJ_COMMIT];
2719 static void parse_new_tag(const char *arg)
2721 static struct strbuf msg = STRBUF_INIT;
2722 const char *from;
2723 char *tagger;
2724 struct branch *s;
2725 struct tag *t;
2726 uintmax_t from_mark = 0;
2727 struct object_id oid;
2728 enum object_type type;
2729 const char *v;
2731 t = mem_pool_alloc(&fi_mem_pool, sizeof(struct tag));
2732 memset(t, 0, sizeof(struct tag));
2733 t->name = pool_strdup(arg);
2734 if (last_tag)
2735 last_tag->next_tag = t;
2736 else
2737 first_tag = t;
2738 last_tag = t;
2739 read_next_command();
2740 parse_mark();
2742 /* from ... */
2743 if (!skip_prefix(command_buf.buf, "from ", &from))
2744 die("Expected from command, got %s", command_buf.buf);
2745 s = lookup_branch(from);
2746 if (s) {
2747 if (is_null_oid(&s->oid))
2748 die("Can't tag an empty branch.");
2749 oidcpy(&oid, &s->oid);
2750 type = OBJ_COMMIT;
2751 } else if (*from == ':') {
2752 struct object_entry *oe;
2753 from_mark = parse_mark_ref_eol(from);
2754 oe = find_mark(from_mark);
2755 type = oe->type;
2756 oidcpy(&oid, &oe->idx.oid);
2757 } else if (!get_oid(from, &oid)) {
2758 struct object_entry *oe = find_object(&oid);
2759 if (!oe) {
2760 type = oid_object_info(the_repository, &oid, NULL);
2761 if (type < 0)
2762 die("Not a valid object: %s", from);
2763 } else
2764 type = oe->type;
2765 } else
2766 die("Invalid ref name or SHA1 expression: %s", from);
2767 read_next_command();
2769 /* original-oid ... */
2770 parse_original_identifier();
2772 /* tagger ... */
2773 if (skip_prefix(command_buf.buf, "tagger ", &v)) {
2774 tagger = parse_ident(v);
2775 read_next_command();
2776 } else
2777 tagger = NULL;
2779 /* tag payload/message */
2780 parse_data(&msg, 0, NULL);
2782 /* build the tag object */
2783 strbuf_reset(&new_data);
2785 strbuf_addf(&new_data,
2786 "object %s\n"
2787 "type %s\n"
2788 "tag %s\n",
2789 oid_to_hex(&oid), type_name(type), t->name);
2790 if (tagger)
2791 strbuf_addf(&new_data,
2792 "tagger %s\n", tagger);
2793 strbuf_addch(&new_data, '\n');
2794 strbuf_addbuf(&new_data, &msg);
2795 free(tagger);
2797 if (store_object(OBJ_TAG, &new_data, NULL, &t->oid, next_mark))
2798 t->pack_id = MAX_PACK_ID;
2799 else
2800 t->pack_id = pack_id;
2803 static void parse_reset_branch(const char *arg)
2805 struct branch *b;
2806 const char *tag_name;
2808 b = lookup_branch(arg);
2809 if (b) {
2810 oidclr(&b->oid);
2811 oidclr(&b->branch_tree.versions[0].oid);
2812 oidclr(&b->branch_tree.versions[1].oid);
2813 if (b->branch_tree.tree) {
2814 release_tree_content_recursive(b->branch_tree.tree);
2815 b->branch_tree.tree = NULL;
2818 else
2819 b = new_branch(arg);
2820 read_next_command();
2821 parse_from(b);
2822 if (b->delete && skip_prefix(b->name, "refs/tags/", &tag_name)) {
2824 * Elsewhere, we call dump_branches() before dump_tags(),
2825 * and dump_branches() will handle ref deletions first, so
2826 * in order to make sure the deletion actually takes effect,
2827 * we need to remove the tag from our list of tags to update.
2829 * NEEDSWORK: replace list of tags with hashmap for faster
2830 * deletion?
2832 struct tag *t, *prev = NULL;
2833 for (t = first_tag; t; t = t->next_tag) {
2834 if (!strcmp(t->name, tag_name))
2835 break;
2836 prev = t;
2838 if (t) {
2839 if (prev)
2840 prev->next_tag = t->next_tag;
2841 else
2842 first_tag = t->next_tag;
2843 if (!t->next_tag)
2844 last_tag = prev;
2845 /* There is no mem_pool_free(t) function to call. */
2848 if (command_buf.len > 0)
2849 unread_command_buf = 1;
2852 static void cat_blob_write(const char *buf, unsigned long size)
2854 if (write_in_full(cat_blob_fd, buf, size) < 0)
2855 die_errno("Write to frontend failed");
2858 static void cat_blob(struct object_entry *oe, struct object_id *oid)
2860 struct strbuf line = STRBUF_INIT;
2861 unsigned long size;
2862 enum object_type type = 0;
2863 char *buf;
2865 if (!oe || oe->pack_id == MAX_PACK_ID) {
2866 buf = read_object_file(oid, &type, &size);
2867 } else {
2868 type = oe->type;
2869 buf = gfi_unpack_entry(oe, &size);
2873 * Output based on batch_one_object() from cat-file.c.
2875 if (type <= 0) {
2876 strbuf_reset(&line);
2877 strbuf_addf(&line, "%s missing\n", oid_to_hex(oid));
2878 cat_blob_write(line.buf, line.len);
2879 strbuf_release(&line);
2880 free(buf);
2881 return;
2883 if (!buf)
2884 die("Can't read object %s", oid_to_hex(oid));
2885 if (type != OBJ_BLOB)
2886 die("Object %s is a %s but a blob was expected.",
2887 oid_to_hex(oid), type_name(type));
2888 strbuf_reset(&line);
2889 strbuf_addf(&line, "%s %s %"PRIuMAX"\n", oid_to_hex(oid),
2890 type_name(type), (uintmax_t)size);
2891 cat_blob_write(line.buf, line.len);
2892 strbuf_release(&line);
2893 cat_blob_write(buf, size);
2894 cat_blob_write("\n", 1);
2895 if (oe && oe->pack_id == pack_id) {
2896 last_blob.offset = oe->idx.offset;
2897 strbuf_attach(&last_blob.data, buf, size, size);
2898 last_blob.depth = oe->depth;
2899 } else
2900 free(buf);
2903 static void parse_get_mark(const char *p)
2905 struct object_entry *oe;
2906 char output[GIT_MAX_HEXSZ + 2];
2908 /* get-mark SP <object> LF */
2909 if (*p != ':')
2910 die("Not a mark: %s", p);
2912 oe = find_mark(parse_mark_ref_eol(p));
2913 if (!oe)
2914 die("Unknown mark: %s", command_buf.buf);
2916 xsnprintf(output, sizeof(output), "%s\n", oid_to_hex(&oe->idx.oid));
2917 cat_blob_write(output, the_hash_algo->hexsz + 1);
2920 static void parse_cat_blob(const char *p)
2922 struct object_entry *oe;
2923 struct object_id oid;
2925 /* cat-blob SP <object> LF */
2926 if (*p == ':') {
2927 oe = find_mark(parse_mark_ref_eol(p));
2928 if (!oe)
2929 die("Unknown mark: %s", command_buf.buf);
2930 oidcpy(&oid, &oe->idx.oid);
2931 } else {
2932 if (parse_oid_hex(p, &oid, &p))
2933 die("Invalid dataref: %s", command_buf.buf);
2934 if (*p)
2935 die("Garbage after SHA1: %s", command_buf.buf);
2936 oe = find_object(&oid);
2939 cat_blob(oe, &oid);
2942 static struct object_entry *dereference(struct object_entry *oe,
2943 struct object_id *oid)
2945 unsigned long size;
2946 char *buf = NULL;
2947 const unsigned hexsz = the_hash_algo->hexsz;
2949 if (!oe) {
2950 enum object_type type = oid_object_info(the_repository, oid,
2951 NULL);
2952 if (type < 0)
2953 die("object not found: %s", oid_to_hex(oid));
2954 /* cache it! */
2955 oe = insert_object(oid);
2956 oe->type = type;
2957 oe->pack_id = MAX_PACK_ID;
2958 oe->idx.offset = 1;
2960 switch (oe->type) {
2961 case OBJ_TREE: /* easy case. */
2962 return oe;
2963 case OBJ_COMMIT:
2964 case OBJ_TAG:
2965 break;
2966 default:
2967 die("Not a tree-ish: %s", command_buf.buf);
2970 if (oe->pack_id != MAX_PACK_ID) { /* in a pack being written */
2971 buf = gfi_unpack_entry(oe, &size);
2972 } else {
2973 enum object_type unused;
2974 buf = read_object_file(oid, &unused, &size);
2976 if (!buf)
2977 die("Can't load object %s", oid_to_hex(oid));
2979 /* Peel one layer. */
2980 switch (oe->type) {
2981 case OBJ_TAG:
2982 if (size < hexsz + strlen("object ") ||
2983 get_oid_hex(buf + strlen("object "), oid))
2984 die("Invalid SHA1 in tag: %s", command_buf.buf);
2985 break;
2986 case OBJ_COMMIT:
2987 if (size < hexsz + strlen("tree ") ||
2988 get_oid_hex(buf + strlen("tree "), oid))
2989 die("Invalid SHA1 in commit: %s", command_buf.buf);
2992 free(buf);
2993 return find_object(oid);
2996 static struct object_entry *parse_treeish_dataref(const char **p)
2998 struct object_id oid;
2999 struct object_entry *e;
3001 if (**p == ':') { /* <mark> */
3002 e = find_mark(parse_mark_ref_space(p));
3003 if (!e)
3004 die("Unknown mark: %s", command_buf.buf);
3005 oidcpy(&oid, &e->idx.oid);
3006 } else { /* <sha1> */
3007 if (parse_oid_hex(*p, &oid, p))
3008 die("Invalid dataref: %s", command_buf.buf);
3009 e = find_object(&oid);
3010 if (*(*p)++ != ' ')
3011 die("Missing space after tree-ish: %s", command_buf.buf);
3014 while (!e || e->type != OBJ_TREE)
3015 e = dereference(e, &oid);
3016 return e;
3019 static void print_ls(int mode, const unsigned char *hash, const char *path)
3021 static struct strbuf line = STRBUF_INIT;
3023 /* See show_tree(). */
3024 const char *type =
3025 S_ISGITLINK(mode) ? commit_type :
3026 S_ISDIR(mode) ? tree_type :
3027 blob_type;
3029 if (!mode) {
3030 /* missing SP path LF */
3031 strbuf_reset(&line);
3032 strbuf_addstr(&line, "missing ");
3033 quote_c_style(path, &line, NULL, 0);
3034 strbuf_addch(&line, '\n');
3035 } else {
3036 /* mode SP type SP object_name TAB path LF */
3037 strbuf_reset(&line);
3038 strbuf_addf(&line, "%06o %s %s\t",
3039 mode & ~NO_DELTA, type, hash_to_hex(hash));
3040 quote_c_style(path, &line, NULL, 0);
3041 strbuf_addch(&line, '\n');
3043 cat_blob_write(line.buf, line.len);
3046 static void parse_ls(const char *p, struct branch *b)
3048 struct tree_entry *root = NULL;
3049 struct tree_entry leaf = {NULL};
3051 /* ls SP (<tree-ish> SP)? <path> */
3052 if (*p == '"') {
3053 if (!b)
3054 die("Not in a commit: %s", command_buf.buf);
3055 root = &b->branch_tree;
3056 } else {
3057 struct object_entry *e = parse_treeish_dataref(&p);
3058 root = new_tree_entry();
3059 oidcpy(&root->versions[1].oid, &e->idx.oid);
3060 if (!is_null_oid(&root->versions[1].oid))
3061 root->versions[1].mode = S_IFDIR;
3062 load_tree(root);
3064 if (*p == '"') {
3065 static struct strbuf uq = STRBUF_INIT;
3066 const char *endp;
3067 strbuf_reset(&uq);
3068 if (unquote_c_style(&uq, p, &endp))
3069 die("Invalid path: %s", command_buf.buf);
3070 if (*endp)
3071 die("Garbage after path in: %s", command_buf.buf);
3072 p = uq.buf;
3074 tree_content_get(root, p, &leaf, 1);
3076 * A directory in preparation would have a sha1 of zero
3077 * until it is saved. Save, for simplicity.
3079 if (S_ISDIR(leaf.versions[1].mode))
3080 store_tree(&leaf);
3082 print_ls(leaf.versions[1].mode, leaf.versions[1].oid.hash, p);
3083 if (leaf.tree)
3084 release_tree_content_recursive(leaf.tree);
3085 if (!b || root != &b->branch_tree)
3086 release_tree_entry(root);
3089 static void checkpoint(void)
3091 checkpoint_requested = 0;
3092 if (object_count) {
3093 cycle_packfile();
3095 dump_branches();
3096 dump_tags();
3097 dump_marks();
3100 static void parse_checkpoint(void)
3102 checkpoint_requested = 1;
3103 skip_optional_lf();
3106 static void parse_progress(void)
3108 fwrite(command_buf.buf, 1, command_buf.len, stdout);
3109 fputc('\n', stdout);
3110 fflush(stdout);
3111 skip_optional_lf();
3114 static void parse_alias(void)
3116 struct object_entry *e;
3117 struct branch b;
3119 skip_optional_lf();
3120 read_next_command();
3122 /* mark ... */
3123 parse_mark();
3124 if (!next_mark)
3125 die(_("Expected 'mark' command, got %s"), command_buf.buf);
3127 /* to ... */
3128 memset(&b, 0, sizeof(b));
3129 if (!parse_objectish_with_prefix(&b, "to "))
3130 die(_("Expected 'to' command, got %s"), command_buf.buf);
3131 e = find_object(&b.oid);
3132 assert(e);
3133 insert_mark(next_mark, e);
3136 static char* make_fast_import_path(const char *path)
3138 if (!relative_marks_paths || is_absolute_path(path))
3139 return xstrdup(path);
3140 return git_pathdup("info/fast-import/%s", path);
3143 static void option_import_marks(const char *marks,
3144 int from_stream, int ignore_missing)
3146 if (import_marks_file) {
3147 if (from_stream)
3148 die("Only one import-marks command allowed per stream");
3150 /* read previous mark file */
3151 if(!import_marks_file_from_stream)
3152 read_marks();
3155 import_marks_file = make_fast_import_path(marks);
3156 import_marks_file_from_stream = from_stream;
3157 import_marks_file_ignore_missing = ignore_missing;
3160 static void option_date_format(const char *fmt)
3162 if (!strcmp(fmt, "raw"))
3163 whenspec = WHENSPEC_RAW;
3164 else if (!strcmp(fmt, "rfc2822"))
3165 whenspec = WHENSPEC_RFC2822;
3166 else if (!strcmp(fmt, "now"))
3167 whenspec = WHENSPEC_NOW;
3168 else
3169 die("unknown --date-format argument %s", fmt);
3172 static unsigned long ulong_arg(const char *option, const char *arg)
3174 char *endptr;
3175 unsigned long rv = strtoul(arg, &endptr, 0);
3176 if (strchr(arg, '-') || endptr == arg || *endptr)
3177 die("%s: argument must be a non-negative integer", option);
3178 return rv;
3181 static void option_depth(const char *depth)
3183 max_depth = ulong_arg("--depth", depth);
3184 if (max_depth > MAX_DEPTH)
3185 die("--depth cannot exceed %u", MAX_DEPTH);
3188 static void option_active_branches(const char *branches)
3190 max_active_branches = ulong_arg("--active-branches", branches);
3193 static void option_export_marks(const char *marks)
3195 export_marks_file = make_fast_import_path(marks);
3198 static void option_cat_blob_fd(const char *fd)
3200 unsigned long n = ulong_arg("--cat-blob-fd", fd);
3201 if (n > (unsigned long) INT_MAX)
3202 die("--cat-blob-fd cannot exceed %d", INT_MAX);
3203 cat_blob_fd = (int) n;
3206 static void option_export_pack_edges(const char *edges)
3208 if (pack_edges)
3209 fclose(pack_edges);
3210 pack_edges = xfopen(edges, "a");
3213 static int parse_one_option(const char *option)
3215 if (skip_prefix(option, "max-pack-size=", &option)) {
3216 unsigned long v;
3217 if (!git_parse_ulong(option, &v))
3218 return 0;
3219 if (v < 8192) {
3220 warning("max-pack-size is now in bytes, assuming --max-pack-size=%lum", v);
3221 v *= 1024 * 1024;
3222 } else if (v < 1024 * 1024) {
3223 warning("minimum max-pack-size is 1 MiB");
3224 v = 1024 * 1024;
3226 max_packsize = v;
3227 } else if (skip_prefix(option, "big-file-threshold=", &option)) {
3228 unsigned long v;
3229 if (!git_parse_ulong(option, &v))
3230 return 0;
3231 big_file_threshold = v;
3232 } else if (skip_prefix(option, "depth=", &option)) {
3233 option_depth(option);
3234 } else if (skip_prefix(option, "active-branches=", &option)) {
3235 option_active_branches(option);
3236 } else if (skip_prefix(option, "export-pack-edges=", &option)) {
3237 option_export_pack_edges(option);
3238 } else if (!strcmp(option, "quiet")) {
3239 show_stats = 0;
3240 } else if (!strcmp(option, "stats")) {
3241 show_stats = 1;
3242 } else if (!strcmp(option, "allow-unsafe-features")) {
3243 ; /* already handled during early option parsing */
3244 } else {
3245 return 0;
3248 return 1;
3251 static void check_unsafe_feature(const char *feature, int from_stream)
3253 if (from_stream && !allow_unsafe_features)
3254 die(_("feature '%s' forbidden in input without --allow-unsafe-features"),
3255 feature);
3258 static int parse_one_feature(const char *feature, int from_stream)
3260 const char *arg;
3262 if (skip_prefix(feature, "date-format=", &arg)) {
3263 option_date_format(arg);
3264 } else if (skip_prefix(feature, "import-marks=", &arg)) {
3265 check_unsafe_feature("import-marks", from_stream);
3266 option_import_marks(arg, from_stream, 0);
3267 } else if (skip_prefix(feature, "import-marks-if-exists=", &arg)) {
3268 check_unsafe_feature("import-marks-if-exists", from_stream);
3269 option_import_marks(arg, from_stream, 1);
3270 } else if (skip_prefix(feature, "export-marks=", &arg)) {
3271 check_unsafe_feature(feature, from_stream);
3272 option_export_marks(arg);
3273 } else if (!strcmp(feature, "alias")) {
3274 ; /* Don't die - this feature is supported */
3275 } else if (!strcmp(feature, "get-mark")) {
3276 ; /* Don't die - this feature is supported */
3277 } else if (!strcmp(feature, "cat-blob")) {
3278 ; /* Don't die - this feature is supported */
3279 } else if (!strcmp(feature, "relative-marks")) {
3280 relative_marks_paths = 1;
3281 } else if (!strcmp(feature, "no-relative-marks")) {
3282 relative_marks_paths = 0;
3283 } else if (!strcmp(feature, "done")) {
3284 require_explicit_termination = 1;
3285 } else if (!strcmp(feature, "force")) {
3286 force_update = 1;
3287 } else if (!strcmp(feature, "notes") || !strcmp(feature, "ls")) {
3288 ; /* do nothing; we have the feature */
3289 } else {
3290 return 0;
3293 return 1;
3296 static void parse_feature(const char *feature)
3298 if (seen_data_command)
3299 die("Got feature command '%s' after data command", feature);
3301 if (parse_one_feature(feature, 1))
3302 return;
3304 die("This version of fast-import does not support feature %s.", feature);
3307 static void parse_option(const char *option)
3309 if (seen_data_command)
3310 die("Got option command '%s' after data command", option);
3312 if (parse_one_option(option))
3313 return;
3315 die("This version of fast-import does not support option: %s", option);
3318 static void git_pack_config(void)
3320 int indexversion_value;
3321 int limit;
3322 unsigned long packsizelimit_value;
3324 if (!git_config_get_ulong("pack.depth", &max_depth)) {
3325 if (max_depth > MAX_DEPTH)
3326 max_depth = MAX_DEPTH;
3328 if (!git_config_get_int("pack.indexversion", &indexversion_value)) {
3329 pack_idx_opts.version = indexversion_value;
3330 if (pack_idx_opts.version > 2)
3331 git_die_config("pack.indexversion",
3332 "bad pack.indexversion=%"PRIu32, pack_idx_opts.version);
3334 if (!git_config_get_ulong("pack.packsizelimit", &packsizelimit_value))
3335 max_packsize = packsizelimit_value;
3337 if (!git_config_get_int("fastimport.unpacklimit", &limit))
3338 unpack_limit = limit;
3339 else if (!git_config_get_int("transfer.unpacklimit", &limit))
3340 unpack_limit = limit;
3342 git_config(git_default_config, NULL);
3345 static const char fast_import_usage[] =
3346 "git fast-import [--date-format=<f>] [--max-pack-size=<n>] [--big-file-threshold=<n>] [--depth=<n>] [--active-branches=<n>] [--export-marks=<marks.file>]";
3348 static void parse_argv(void)
3350 unsigned int i;
3352 for (i = 1; i < global_argc; i++) {
3353 const char *a = global_argv[i];
3355 if (*a != '-' || !strcmp(a, "--"))
3356 break;
3358 if (!skip_prefix(a, "--", &a))
3359 die("unknown option %s", a);
3361 if (parse_one_option(a))
3362 continue;
3364 if (parse_one_feature(a, 0))
3365 continue;
3367 if (skip_prefix(a, "cat-blob-fd=", &a)) {
3368 option_cat_blob_fd(a);
3369 continue;
3372 die("unknown option --%s", a);
3374 if (i != global_argc)
3375 usage(fast_import_usage);
3377 seen_data_command = 1;
3378 if (import_marks_file)
3379 read_marks();
3382 int cmd_main(int argc, const char **argv)
3384 unsigned int i;
3386 if (argc == 2 && !strcmp(argv[1], "-h"))
3387 usage(fast_import_usage);
3389 setup_git_directory();
3390 reset_pack_idx_option(&pack_idx_opts);
3391 git_pack_config();
3393 alloc_objects(object_entry_alloc);
3394 strbuf_init(&command_buf, 0);
3395 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
3396 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
3397 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
3398 marks = mem_pool_calloc(&fi_mem_pool, 1, sizeof(struct mark_set));
3401 * We don't parse most options until after we've seen the set of
3402 * "feature" lines at the start of the stream (which allows the command
3403 * line to override stream data). But we must do an early parse of any
3404 * command-line options that impact how we interpret the feature lines.
3406 for (i = 1; i < argc; i++) {
3407 const char *arg = argv[i];
3408 if (*arg != '-' || !strcmp(arg, "--"))
3409 break;
3410 if (!strcmp(arg, "--allow-unsafe-features"))
3411 allow_unsafe_features = 1;
3414 global_argc = argc;
3415 global_argv = argv;
3417 rc_free = mem_pool_alloc(&fi_mem_pool, cmd_save * sizeof(*rc_free));
3418 for (i = 0; i < (cmd_save - 1); i++)
3419 rc_free[i].next = &rc_free[i + 1];
3420 rc_free[cmd_save - 1].next = NULL;
3422 start_packfile();
3423 set_die_routine(die_nicely);
3424 set_checkpoint_signal();
3425 while (read_next_command() != EOF) {
3426 const char *v;
3427 if (!strcmp("blob", command_buf.buf))
3428 parse_new_blob();
3429 else if (skip_prefix(command_buf.buf, "commit ", &v))
3430 parse_new_commit(v);
3431 else if (skip_prefix(command_buf.buf, "tag ", &v))
3432 parse_new_tag(v);
3433 else if (skip_prefix(command_buf.buf, "reset ", &v))
3434 parse_reset_branch(v);
3435 else if (skip_prefix(command_buf.buf, "ls ", &v))
3436 parse_ls(v, NULL);
3437 else if (skip_prefix(command_buf.buf, "cat-blob ", &v))
3438 parse_cat_blob(v);
3439 else if (skip_prefix(command_buf.buf, "get-mark ", &v))
3440 parse_get_mark(v);
3441 else if (!strcmp("checkpoint", command_buf.buf))
3442 parse_checkpoint();
3443 else if (!strcmp("done", command_buf.buf))
3444 break;
3445 else if (!strcmp("alias", command_buf.buf))
3446 parse_alias();
3447 else if (starts_with(command_buf.buf, "progress "))
3448 parse_progress();
3449 else if (skip_prefix(command_buf.buf, "feature ", &v))
3450 parse_feature(v);
3451 else if (skip_prefix(command_buf.buf, "option git ", &v))
3452 parse_option(v);
3453 else if (starts_with(command_buf.buf, "option "))
3454 /* ignore non-git options*/;
3455 else
3456 die("Unsupported command: %s", command_buf.buf);
3458 if (checkpoint_requested)
3459 checkpoint();
3462 /* argv hasn't been parsed yet, do so */
3463 if (!seen_data_command)
3464 parse_argv();
3466 if (require_explicit_termination && feof(stdin))
3467 die("stream ends early");
3469 end_packfile();
3471 dump_branches();
3472 dump_tags();
3473 unkeep_all_packs();
3474 dump_marks();
3476 if (pack_edges)
3477 fclose(pack_edges);
3479 if (show_stats) {
3480 uintmax_t total_count = 0, duplicate_count = 0;
3481 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
3482 total_count += object_count_by_type[i];
3483 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
3484 duplicate_count += duplicate_count_by_type[i];
3486 fprintf(stderr, "%s statistics:\n", argv[0]);
3487 fprintf(stderr, "---------------------------------------------------------------------\n");
3488 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
3489 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
3490 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]);
3491 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]);
3492 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]);
3493 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]);
3494 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
3495 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
3496 fprintf(stderr, " atoms: %10u\n", atom_cnt);
3497 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (tree_entry_allocd + fi_mem_pool.pool_alloc + alloc_count*sizeof(struct object_entry))/1024);
3498 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)((tree_entry_allocd + fi_mem_pool.pool_alloc) /1024));
3499 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
3500 fprintf(stderr, "---------------------------------------------------------------------\n");
3501 pack_report();
3502 fprintf(stderr, "---------------------------------------------------------------------\n");
3503 fprintf(stderr, "\n");
3506 return failure ? 1 : 0;