2 Format of STDIN stream:
14 new_blob ::= 'blob' lf
17 file_content ::= data;
19 new_commit ::= 'commit' sp ref_str lf
21 ('author' sp name '<' email '>' when lf)?
22 'committer' sp name '<' email '>' when lf
24 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
25 ('merge' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)*
30 file_change ::= file_clr
36 file_clr ::= 'deleteall' lf;
37 file_del ::= 'D' sp path_str lf;
38 file_rnm ::= 'R' sp path_str sp path_str lf;
39 file_cpy ::= 'C' sp path_str sp path_str lf;
40 file_obm ::= 'M' sp mode sp (hexsha1 | idnum) sp path_str lf;
41 file_inm ::= 'M' sp mode sp 'inline' sp path_str lf
44 new_tag ::= 'tag' sp tag_str lf
45 'from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf
46 'tagger' sp name '<' email '>' when lf
50 reset_branch ::= 'reset' sp ref_str lf
51 ('from' sp (ref_str | hexsha1 | sha1exp_str | idnum) lf)?
54 checkpoint ::= 'checkpoint' lf
57 progress ::= 'progress' sp not_lf* lf
60 # note: the first idnum in a stream should be 1 and subsequent
61 # idnums should not have gaps between values as this will cause
62 # the stream parser to reserve space for the gapped values. An
63 # idnum can be updated in the future to a new object by issuing
64 # a new mark directive with the old idnum.
66 mark ::= 'mark' sp idnum lf;
67 data ::= (delimited_data | exact_data)
70 # note: delim may be any string but must not contain lf.
71 # data_line may contain any data but must not be exactly
73 delimited_data ::= 'data' sp '<<' delim lf
77 # note: declen indicates the length of binary_data in bytes.
78 # declen does not include the lf preceeding the binary data.
80 exact_data ::= 'data' sp declen lf
83 # note: quoted strings are C-style quoting supporting \c for
84 # common escapes of 'c' (e..g \n, \t, \\, \") or \nnn where nnn
85 # is the signed byte value in octal. Note that the only
86 # characters which must actually be escaped to protect the
87 # stream formatting is: \, " and LF. Otherwise these values
91 sha1exp_str ::= sha1exp;
93 path_str ::= path | '"' quoted(path) '"' ;
94 mode ::= '100644' | '644'
99 declen ::= # unsigned 32 bit value, ascii base10 notation;
100 bigint ::= # unsigned integer value, ascii base10 notation;
101 binary_data ::= # file content, not interpreted;
103 when ::= raw_when | rfc2822_when;
104 raw_when ::= ts sp tz;
105 rfc2822_when ::= # Valid RFC 2822 date and time;
107 sp ::= # ASCII space character;
108 lf ::= # ASCII newline (LF) character;
110 # note: a colon (':') must precede the numerical value assigned to
111 # an idnum. This is to distinguish it from a ref or tag name as
112 # GIT does not permit ':' in ref or tag strings.
114 idnum ::= ':' bigint;
115 path ::= # GIT style file path, e.g. "a/b/c";
116 ref ::= # GIT ref name, e.g. "refs/heads/MOZ_GECKO_EXPERIMENT";
117 tag ::= # GIT tag name, e.g. "FIREFOX_1_5";
118 sha1exp ::= # Any valid GIT SHA1 expression;
119 hexsha1 ::= # SHA1 in hexadecimal format;
121 # note: name and email are UTF8 strings, however name must not
122 # contain '<' or lf and email must not contain any of the
123 # following: '<', '>', lf.
125 name ::= # valid GIT author/committer name;
126 email ::= # valid GIT author/committer email;
127 ts ::= # time since the epoch in seconds, ascii base10 notation;
128 tz ::= # GIT style timezone;
130 # note: comments may appear anywhere in the input, except
131 # within a data command. Any form of the data command
132 # always escapes the related input from comment processing.
134 # In case it is not clear, the '#' that starts the comment
135 # must be the first character on that the line (an lf have
138 comment ::= '#' not_lf* lf;
139 not_lf ::= # Any byte that is not ASCII newline (LF);
151 #include "csum-file.h"
154 #define PACK_ID_BITS 16
155 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
156 #define DEPTH_BITS 13
157 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
161 struct object_entry *next;
163 uint32_t type : TYPE_BITS,
164 pack_id : PACK_ID_BITS,
166 unsigned char sha1[20];
169 struct object_entry_pool
171 struct object_entry_pool *next_pool;
172 struct object_entry *next_free;
173 struct object_entry *end;
174 struct object_entry entries[FLEX_ARRAY]; /* more */
180 struct object_entry *marked[1024];
181 struct mark_set *sets[1024];
191 unsigned no_swap : 1;
196 struct mem_pool *next_pool;
199 uintmax_t space[FLEX_ARRAY]; /* more */
204 struct atom_str *next_atom;
205 unsigned short str_len;
206 char str_dat[FLEX_ARRAY]; /* more */
212 struct tree_content *tree;
213 struct atom_str* name;
217 unsigned char sha1[20];
223 unsigned int entry_capacity; /* must match avail_tree_content */
224 unsigned int entry_count;
225 unsigned int delta_depth;
226 struct tree_entry *entries[FLEX_ARRAY]; /* more */
229 struct avail_tree_content
231 unsigned int entry_capacity; /* must match tree_content */
232 struct avail_tree_content *next_avail;
237 struct branch *table_next_branch;
238 struct branch *active_next_branch;
240 struct tree_entry branch_tree;
241 uintmax_t last_commit;
243 unsigned pack_id : PACK_ID_BITS;
244 unsigned char sha1[20];
249 struct tag *next_tag;
251 unsigned int pack_id;
252 unsigned char sha1[20];
257 struct hash_list *next;
258 unsigned char sha1[20];
267 struct recent_command
269 struct recent_command *prev;
270 struct recent_command *next;
274 /* Configured limits on output */
275 static unsigned long max_depth = 10;
276 static off_t max_packsize = (1LL << 32) - 1;
277 static int force_update;
278 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
279 static int pack_compression_seen;
281 /* Stats and misc. counters */
282 static uintmax_t alloc_count;
283 static uintmax_t marks_set_count;
284 static uintmax_t object_count_by_type[1 << TYPE_BITS];
285 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
286 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
287 static unsigned long object_count;
288 static unsigned long branch_count;
289 static unsigned long branch_load_count;
291 static FILE *pack_edges;
294 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
295 static size_t total_allocd;
296 static struct mem_pool *mem_pool;
298 /* Atom management */
299 static unsigned int atom_table_sz = 4451;
300 static unsigned int atom_cnt;
301 static struct atom_str **atom_table;
303 /* The .pack file being generated */
304 static unsigned int pack_id;
305 static struct packed_git *pack_data;
306 static struct packed_git **all_packs;
307 static unsigned long pack_size;
309 /* Table of objects we've written. */
310 static unsigned int object_entry_alloc = 5000;
311 static struct object_entry_pool *blocks;
312 static struct object_entry *object_table[1 << 16];
313 static struct mark_set *marks;
314 static const char* mark_file;
317 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
319 /* Tree management */
320 static unsigned int tree_entry_alloc = 1000;
321 static void *avail_tree_entry;
322 static unsigned int avail_tree_table_sz = 100;
323 static struct avail_tree_content **avail_tree_table;
324 static struct strbuf old_tree = STRBUF_INIT;
325 static struct strbuf new_tree = STRBUF_INIT;
328 static unsigned long max_active_branches = 5;
329 static unsigned long cur_active_branches;
330 static unsigned long branch_table_sz = 1039;
331 static struct branch **branch_table;
332 static struct branch *active_branches;
335 static struct tag *first_tag;
336 static struct tag *last_tag;
338 /* Input stream parsing */
339 static whenspec_type whenspec = WHENSPEC_RAW;
340 static struct strbuf command_buf = STRBUF_INIT;
341 static int unread_command_buf;
342 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
343 static struct recent_command *cmd_tail = &cmd_hist;
344 static struct recent_command *rc_free;
345 static unsigned int cmd_save = 100;
346 static uintmax_t next_mark;
347 static struct strbuf new_data = STRBUF_INIT;
349 static void write_branch_report(FILE *rpt, struct branch *b)
351 fprintf(rpt, "%s:\n", b->name);
353 fprintf(rpt, " status :");
355 fputs(" active", rpt);
356 if (b->branch_tree.tree)
357 fputs(" loaded", rpt);
358 if (is_null_sha1(b->branch_tree.versions[1].sha1))
359 fputs(" dirty", rpt);
362 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
363 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
364 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
365 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
367 fputs(" last pack : ", rpt);
368 if (b->pack_id < MAX_PACK_ID)
369 fprintf(rpt, "%u", b->pack_id);
375 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
377 static void write_crash_report(const char *err)
379 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
380 FILE *rpt = fopen(loc, "w");
383 struct recent_command *rc;
386 error("can't write crash report %s: %s", loc, strerror(errno));
390 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
392 fprintf(rpt, "fast-import crash report:\n");
393 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
394 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
395 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
398 fputs("fatal: ", rpt);
403 fputs("Most Recent Commands Before Crash\n", rpt);
404 fputs("---------------------------------\n", rpt);
405 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
406 if (rc->next == &cmd_hist)
415 fputs("Active Branch LRU\n", rpt);
416 fputs("-----------------\n", rpt);
417 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
419 max_active_branches);
421 fputs(" pos clock name\n", rpt);
422 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
423 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
424 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
425 ++lu, b->last_commit, b->name);
428 fputs("Inactive Branches\n", rpt);
429 fputs("-----------------\n", rpt);
430 for (lu = 0; lu < branch_table_sz; lu++) {
431 for (b = branch_table[lu]; b; b = b->table_next_branch)
432 write_branch_report(rpt, b);
438 fputs("Annotated Tags\n", rpt);
439 fputs("--------------\n", rpt);
440 for (tg = first_tag; tg; tg = tg->next_tag) {
441 fputs(sha1_to_hex(tg->sha1), rpt);
443 fputs(tg->name, rpt);
449 fputs("Marks\n", rpt);
450 fputs("-----\n", rpt);
452 fprintf(rpt, " exported to %s\n", mark_file);
454 dump_marks_helper(rpt, 0, marks);
457 fputs("-------------------\n", rpt);
458 fputs("END OF CRASH REPORT\n", rpt);
462 static void end_packfile(void);
463 static void unkeep_all_packs(void);
464 static void dump_marks(void);
466 static NORETURN void die_nicely(const char *err, va_list params)
469 char message[2 * PATH_MAX];
471 vsnprintf(message, sizeof(message), err, params);
472 fputs("fatal: ", stderr);
473 fputs(message, stderr);
478 write_crash_report(message);
486 static void alloc_objects(unsigned int cnt)
488 struct object_entry_pool *b;
490 b = xmalloc(sizeof(struct object_entry_pool)
491 + cnt * sizeof(struct object_entry));
492 b->next_pool = blocks;
493 b->next_free = b->entries;
494 b->end = b->entries + cnt;
499 static struct object_entry *new_object(unsigned char *sha1)
501 struct object_entry *e;
503 if (blocks->next_free == blocks->end)
504 alloc_objects(object_entry_alloc);
506 e = blocks->next_free++;
507 hashcpy(e->sha1, sha1);
511 static struct object_entry *find_object(unsigned char *sha1)
513 unsigned int h = sha1[0] << 8 | sha1[1];
514 struct object_entry *e;
515 for (e = object_table[h]; e; e = e->next)
516 if (!hashcmp(sha1, e->sha1))
521 static struct object_entry *insert_object(unsigned char *sha1)
523 unsigned int h = sha1[0] << 8 | sha1[1];
524 struct object_entry *e = object_table[h];
525 struct object_entry *p = NULL;
528 if (!hashcmp(sha1, e->sha1))
534 e = new_object(sha1);
544 static unsigned int hc_str(const char *s, size_t len)
552 static void *pool_alloc(size_t len)
557 /* round up to a 'uintmax_t' alignment */
558 if (len & (sizeof(uintmax_t) - 1))
559 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
561 for (p = mem_pool; p; p = p->next_pool)
562 if ((p->end - p->next_free >= len))
566 if (len >= (mem_pool_alloc/2)) {
570 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
571 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
572 p->next_pool = mem_pool;
573 p->next_free = (char *) p->space;
574 p->end = p->next_free + mem_pool_alloc;
583 static void *pool_calloc(size_t count, size_t size)
585 size_t len = count * size;
586 void *r = pool_alloc(len);
591 static char *pool_strdup(const char *s)
593 char *r = pool_alloc(strlen(s) + 1);
598 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
600 struct mark_set *s = marks;
601 while ((idnum >> s->shift) >= 1024) {
602 s = pool_calloc(1, sizeof(struct mark_set));
603 s->shift = marks->shift + 10;
604 s->data.sets[0] = marks;
608 uintmax_t i = idnum >> s->shift;
609 idnum -= i << s->shift;
610 if (!s->data.sets[i]) {
611 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
612 s->data.sets[i]->shift = s->shift - 10;
616 if (!s->data.marked[idnum])
618 s->data.marked[idnum] = oe;
621 static struct object_entry *find_mark(uintmax_t idnum)
623 uintmax_t orig_idnum = idnum;
624 struct mark_set *s = marks;
625 struct object_entry *oe = NULL;
626 if ((idnum >> s->shift) < 1024) {
627 while (s && s->shift) {
628 uintmax_t i = idnum >> s->shift;
629 idnum -= i << s->shift;
633 oe = s->data.marked[idnum];
636 die("mark :%" PRIuMAX " not declared", orig_idnum);
640 static struct atom_str *to_atom(const char *s, unsigned short len)
642 unsigned int hc = hc_str(s, len) % atom_table_sz;
645 for (c = atom_table[hc]; c; c = c->next_atom)
646 if (c->str_len == len && !strncmp(s, c->str_dat, len))
649 c = pool_alloc(sizeof(struct atom_str) + len + 1);
651 strncpy(c->str_dat, s, len);
653 c->next_atom = atom_table[hc];
659 static struct branch *lookup_branch(const char *name)
661 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
664 for (b = branch_table[hc]; b; b = b->table_next_branch)
665 if (!strcmp(name, b->name))
670 static struct branch *new_branch(const char *name)
672 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
673 struct branch* b = lookup_branch(name);
676 die("Invalid attempt to create duplicate branch: %s", name);
677 switch (check_ref_format(name)) {
678 case 0: break; /* its valid */
679 case CHECK_REF_FORMAT_ONELEVEL:
680 break; /* valid, but too few '/', allow anyway */
682 die("Branch name doesn't conform to GIT standards: %s", name);
685 b = pool_calloc(1, sizeof(struct branch));
686 b->name = pool_strdup(name);
687 b->table_next_branch = branch_table[hc];
688 b->branch_tree.versions[0].mode = S_IFDIR;
689 b->branch_tree.versions[1].mode = S_IFDIR;
691 b->pack_id = MAX_PACK_ID;
692 branch_table[hc] = b;
697 static unsigned int hc_entries(unsigned int cnt)
699 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
700 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
703 static struct tree_content *new_tree_content(unsigned int cnt)
705 struct avail_tree_content *f, *l = NULL;
706 struct tree_content *t;
707 unsigned int hc = hc_entries(cnt);
709 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
710 if (f->entry_capacity >= cnt)
715 l->next_avail = f->next_avail;
717 avail_tree_table[hc] = f->next_avail;
719 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
720 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
721 f->entry_capacity = cnt;
724 t = (struct tree_content*)f;
730 static void release_tree_entry(struct tree_entry *e);
731 static void release_tree_content(struct tree_content *t)
733 struct avail_tree_content *f = (struct avail_tree_content*)t;
734 unsigned int hc = hc_entries(f->entry_capacity);
735 f->next_avail = avail_tree_table[hc];
736 avail_tree_table[hc] = f;
739 static void release_tree_content_recursive(struct tree_content *t)
742 for (i = 0; i < t->entry_count; i++)
743 release_tree_entry(t->entries[i]);
744 release_tree_content(t);
747 static struct tree_content *grow_tree_content(
748 struct tree_content *t,
751 struct tree_content *r = new_tree_content(t->entry_count + amt);
752 r->entry_count = t->entry_count;
753 r->delta_depth = t->delta_depth;
754 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
755 release_tree_content(t);
759 static struct tree_entry *new_tree_entry(void)
761 struct tree_entry *e;
763 if (!avail_tree_entry) {
764 unsigned int n = tree_entry_alloc;
765 total_allocd += n * sizeof(struct tree_entry);
766 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
768 *((void**)e) = e + 1;
774 e = avail_tree_entry;
775 avail_tree_entry = *((void**)e);
779 static void release_tree_entry(struct tree_entry *e)
782 release_tree_content_recursive(e->tree);
783 *((void**)e) = avail_tree_entry;
784 avail_tree_entry = e;
787 static struct tree_content *dup_tree_content(struct tree_content *s)
789 struct tree_content *d;
790 struct tree_entry *a, *b;
795 d = new_tree_content(s->entry_count);
796 for (i = 0; i < s->entry_count; i++) {
798 b = new_tree_entry();
799 memcpy(b, a, sizeof(*a));
800 if (a->tree && is_null_sha1(b->versions[1].sha1))
801 b->tree = dup_tree_content(a->tree);
806 d->entry_count = s->entry_count;
807 d->delta_depth = s->delta_depth;
812 static void start_packfile(void)
814 static char tmpfile[PATH_MAX];
815 struct packed_git *p;
816 struct pack_header hdr;
819 snprintf(tmpfile, sizeof(tmpfile),
820 "%s/pack/tmp_pack_XXXXXX", get_object_directory());
821 pack_fd = xmkstemp(tmpfile);
822 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
823 strcpy(p->pack_name, tmpfile);
824 p->pack_fd = pack_fd;
826 hdr.hdr_signature = htonl(PACK_SIGNATURE);
827 hdr.hdr_version = htonl(2);
829 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
832 pack_size = sizeof(hdr);
835 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
836 all_packs[pack_id] = p;
839 static int oecmp (const void *a_, const void *b_)
841 struct object_entry *a = *((struct object_entry**)a_);
842 struct object_entry *b = *((struct object_entry**)b_);
843 return hashcmp(a->sha1, b->sha1);
846 static char *create_index(void)
848 static char tmpfile[PATH_MAX];
851 struct object_entry **idx, **c, **last, *e;
852 struct object_entry_pool *o;
856 /* Build the sorted table of object IDs. */
857 idx = xmalloc(object_count * sizeof(struct object_entry*));
859 for (o = blocks; o; o = o->next_pool)
860 for (e = o->next_free; e-- != o->entries;)
861 if (pack_id == e->pack_id)
863 last = idx + object_count;
865 die("internal consistency error creating the index");
866 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
868 /* Generate the fan-out array. */
870 for (i = 0; i < 256; i++) {
871 struct object_entry **next = c;;
872 while (next < last) {
873 if ((*next)->sha1[0] != i)
877 array[i] = htonl(next - idx);
881 snprintf(tmpfile, sizeof(tmpfile),
882 "%s/pack/tmp_idx_XXXXXX", get_object_directory());
883 idx_fd = xmkstemp(tmpfile);
884 f = sha1fd(idx_fd, tmpfile);
885 sha1write(f, array, 256 * sizeof(int));
887 for (c = idx; c != last; c++) {
888 uint32_t offset = htonl((*c)->offset);
889 sha1write(f, &offset, 4);
890 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
891 git_SHA1_Update(&ctx, (*c)->sha1, 20);
893 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
894 sha1close(f, NULL, CSUM_FSYNC);
896 git_SHA1_Final(pack_data->sha1, &ctx);
900 static char *keep_pack(char *curr_index_name)
902 static char name[PATH_MAX];
903 static const char *keep_msg = "fast-import";
906 chmod(pack_data->pack_name, 0444);
907 chmod(curr_index_name, 0444);
909 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
910 get_object_directory(), sha1_to_hex(pack_data->sha1));
911 keep_fd = open(name, O_RDWR|O_CREAT|O_EXCL, 0600);
913 die("cannot create keep file");
914 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
916 die("failed to write keep file");
918 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
919 get_object_directory(), sha1_to_hex(pack_data->sha1));
920 if (move_temp_to_file(pack_data->pack_name, name))
921 die("cannot store pack file");
923 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
924 get_object_directory(), sha1_to_hex(pack_data->sha1));
925 if (move_temp_to_file(curr_index_name, name))
926 die("cannot store index file");
930 static void unkeep_all_packs(void)
932 static char name[PATH_MAX];
935 for (k = 0; k < pack_id; k++) {
936 struct packed_git *p = all_packs[k];
937 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
938 get_object_directory(), sha1_to_hex(p->sha1));
943 static void end_packfile(void)
945 struct packed_git *old_p = pack_data, *new_p;
953 close_pack_windows(pack_data);
954 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
955 pack_data->pack_name, object_count,
957 close(pack_data->pack_fd);
958 idx_name = keep_pack(create_index());
960 /* Register the packfile with core git's machinary. */
961 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
963 die("core git rejected index %s", idx_name);
964 all_packs[pack_id] = new_p;
965 install_packed_git(new_p);
967 /* Print the boundary */
969 fprintf(pack_edges, "%s:", new_p->pack_name);
970 for (i = 0; i < branch_table_sz; i++) {
971 for (b = branch_table[i]; b; b = b->table_next_branch) {
972 if (b->pack_id == pack_id)
973 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
976 for (t = first_tag; t; t = t->next_tag) {
977 if (t->pack_id == pack_id)
978 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
980 fputc('\n', pack_edges);
987 close(old_p->pack_fd);
988 unlink(old_p->pack_name);
992 /* We can't carry a delta across packfiles. */
993 strbuf_release(&last_blob.data);
994 last_blob.offset = 0;
998 static void cycle_packfile(void)
1004 static size_t encode_header(
1005 enum object_type type,
1012 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
1013 die("bad type %d", type);
1015 c = (type << 4) | (size & 15);
1027 static int store_object(
1028 enum object_type type,
1030 struct last_object *last,
1031 unsigned char *sha1out,
1035 struct object_entry *e;
1036 unsigned char hdr[96];
1037 unsigned char sha1[20];
1038 unsigned long hdrlen, deltalen;
1042 hdrlen = sprintf((char*)hdr,"%s %lu", typename(type),
1043 (unsigned long)dat->len) + 1;
1045 git_SHA1_Update(&c, hdr, hdrlen);
1046 git_SHA1_Update(&c, dat->buf, dat->len);
1047 git_SHA1_Final(sha1, &c);
1049 hashcpy(sha1out, sha1);
1051 e = insert_object(sha1);
1053 insert_mark(mark, e);
1055 duplicate_count_by_type[type]++;
1057 } else if (find_sha1_pack(sha1, packed_git)) {
1059 e->pack_id = MAX_PACK_ID;
1060 e->offset = 1; /* just not zero! */
1061 duplicate_count_by_type[type]++;
1065 if (last && last->data.buf && last->depth < max_depth) {
1066 delta = diff_delta(last->data.buf, last->data.len,
1069 if (delta && deltalen >= dat->len) {
1076 memset(&s, 0, sizeof(s));
1077 deflateInit(&s, pack_compression_level);
1080 s.avail_in = deltalen;
1082 s.next_in = (void *)dat->buf;
1083 s.avail_in = dat->len;
1085 s.avail_out = deflateBound(&s, s.avail_in);
1086 s.next_out = out = xmalloc(s.avail_out);
1087 while (deflate(&s, Z_FINISH) == Z_OK)
1091 /* Determine if we should auto-checkpoint. */
1092 if ((pack_size + 60 + s.total_out) > max_packsize
1093 || (pack_size + 60 + s.total_out) < pack_size) {
1095 /* This new object needs to *not* have the current pack_id. */
1096 e->pack_id = pack_id + 1;
1099 /* We cannot carry a delta into the new pack. */
1104 memset(&s, 0, sizeof(s));
1105 deflateInit(&s, pack_compression_level);
1106 s.next_in = (void *)dat->buf;
1107 s.avail_in = dat->len;
1108 s.avail_out = deflateBound(&s, s.avail_in);
1109 s.next_out = out = xrealloc(out, s.avail_out);
1110 while (deflate(&s, Z_FINISH) == Z_OK)
1117 e->pack_id = pack_id;
1118 e->offset = pack_size;
1120 object_count_by_type[type]++;
1123 unsigned long ofs = e->offset - last->offset;
1124 unsigned pos = sizeof(hdr) - 1;
1126 delta_count_by_type[type]++;
1127 e->depth = last->depth + 1;
1129 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1130 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1131 pack_size += hdrlen;
1133 hdr[pos] = ofs & 127;
1135 hdr[--pos] = 128 | (--ofs & 127);
1136 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
1137 pack_size += sizeof(hdr) - pos;
1140 hdrlen = encode_header(type, dat->len, hdr);
1141 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1142 pack_size += hdrlen;
1145 write_or_die(pack_data->pack_fd, out, s.total_out);
1146 pack_size += s.total_out;
1151 if (last->no_swap) {
1154 strbuf_swap(&last->data, dat);
1156 last->offset = e->offset;
1157 last->depth = e->depth;
1162 /* All calls must be guarded by find_object() or find_mark() to
1163 * ensure the 'struct object_entry' passed was written by this
1164 * process instance. We unpack the entry by the offset, avoiding
1165 * the need for the corresponding .idx file. This unpacking rule
1166 * works because we only use OBJ_REF_DELTA within the packfiles
1167 * created by fast-import.
1169 * oe must not be NULL. Such an oe usually comes from giving
1170 * an unknown SHA-1 to find_object() or an undefined mark to
1171 * find_mark(). Callers must test for this condition and use
1172 * the standard read_sha1_file() when it happens.
1174 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1175 * find_mark(), where the mark was reloaded from an existing marks
1176 * file and is referencing an object that this fast-import process
1177 * instance did not write out to a packfile. Callers must test for
1178 * this condition and use read_sha1_file() instead.
1180 static void *gfi_unpack_entry(
1181 struct object_entry *oe,
1182 unsigned long *sizep)
1184 enum object_type type;
1185 struct packed_git *p = all_packs[oe->pack_id];
1186 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1187 /* The object is stored in the packfile we are writing to
1188 * and we have modified it since the last time we scanned
1189 * back to read a previously written object. If an old
1190 * window covered [p->pack_size, p->pack_size + 20) its
1191 * data is stale and is not valid. Closing all windows
1192 * and updating the packfile length ensures we can read
1193 * the newly written data.
1195 close_pack_windows(p);
1197 /* We have to offer 20 bytes additional on the end of
1198 * the packfile as the core unpacker code assumes the
1199 * footer is present at the file end and must promise
1200 * at least 20 bytes within any window it maps. But
1201 * we don't actually create the footer here.
1203 p->pack_size = pack_size + 20;
1205 return unpack_entry(p, oe->offset, &type, sizep);
1208 static const char *get_mode(const char *str, uint16_t *modep)
1213 while ((c = *str++) != ' ') {
1214 if (c < '0' || c > '7')
1216 mode = (mode << 3) + (c - '0');
1222 static void load_tree(struct tree_entry *root)
1224 unsigned char* sha1 = root->versions[1].sha1;
1225 struct object_entry *myoe;
1226 struct tree_content *t;
1231 root->tree = t = new_tree_content(8);
1232 if (is_null_sha1(sha1))
1235 myoe = find_object(sha1);
1236 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1237 if (myoe->type != OBJ_TREE)
1238 die("Not a tree: %s", sha1_to_hex(sha1));
1239 t->delta_depth = myoe->depth;
1240 buf = gfi_unpack_entry(myoe, &size);
1242 die("Can't load tree %s", sha1_to_hex(sha1));
1244 enum object_type type;
1245 buf = read_sha1_file(sha1, &type, &size);
1246 if (!buf || type != OBJ_TREE)
1247 die("Can't load tree %s", sha1_to_hex(sha1));
1251 while (c != (buf + size)) {
1252 struct tree_entry *e = new_tree_entry();
1254 if (t->entry_count == t->entry_capacity)
1255 root->tree = t = grow_tree_content(t, t->entry_count);
1256 t->entries[t->entry_count++] = e;
1259 c = get_mode(c, &e->versions[1].mode);
1261 die("Corrupt mode in %s", sha1_to_hex(sha1));
1262 e->versions[0].mode = e->versions[1].mode;
1263 e->name = to_atom(c, strlen(c));
1264 c += e->name->str_len + 1;
1265 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1266 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1272 static int tecmp0 (const void *_a, const void *_b)
1274 struct tree_entry *a = *((struct tree_entry**)_a);
1275 struct tree_entry *b = *((struct tree_entry**)_b);
1276 return base_name_compare(
1277 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1278 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1281 static int tecmp1 (const void *_a, const void *_b)
1283 struct tree_entry *a = *((struct tree_entry**)_a);
1284 struct tree_entry *b = *((struct tree_entry**)_b);
1285 return base_name_compare(
1286 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1287 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1290 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1296 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1298 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1300 for (i = 0; i < t->entry_count; i++) {
1301 if (t->entries[i]->versions[v].mode)
1302 maxlen += t->entries[i]->name->str_len + 34;
1306 strbuf_grow(b, maxlen);
1307 for (i = 0; i < t->entry_count; i++) {
1308 struct tree_entry *e = t->entries[i];
1309 if (!e->versions[v].mode)
1311 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1312 e->name->str_dat, '\0');
1313 strbuf_add(b, e->versions[v].sha1, 20);
1317 static void store_tree(struct tree_entry *root)
1319 struct tree_content *t = root->tree;
1320 unsigned int i, j, del;
1321 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1322 struct object_entry *le;
1324 if (!is_null_sha1(root->versions[1].sha1))
1327 for (i = 0; i < t->entry_count; i++) {
1328 if (t->entries[i]->tree)
1329 store_tree(t->entries[i]);
1332 le = find_object(root->versions[0].sha1);
1333 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1334 mktree(t, 0, &old_tree);
1336 lo.offset = le->offset;
1337 lo.depth = t->delta_depth;
1340 mktree(t, 1, &new_tree);
1341 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1343 t->delta_depth = lo.depth;
1344 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1345 struct tree_entry *e = t->entries[i];
1346 if (e->versions[1].mode) {
1347 e->versions[0].mode = e->versions[1].mode;
1348 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1349 t->entries[j++] = e;
1351 release_tree_entry(e);
1355 t->entry_count -= del;
1358 static int tree_content_set(
1359 struct tree_entry *root,
1361 const unsigned char *sha1,
1362 const uint16_t mode,
1363 struct tree_content *subtree)
1365 struct tree_content *t = root->tree;
1368 struct tree_entry *e;
1370 slash1 = strchr(p, '/');
1376 die("Empty path component found in input");
1377 if (!slash1 && !S_ISDIR(mode) && subtree)
1378 die("Non-directories cannot have subtrees");
1380 for (i = 0; i < t->entry_count; i++) {
1382 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1385 && e->versions[1].mode == mode
1386 && !hashcmp(e->versions[1].sha1, sha1))
1388 e->versions[1].mode = mode;
1389 hashcpy(e->versions[1].sha1, sha1);
1391 release_tree_content_recursive(e->tree);
1393 hashclr(root->versions[1].sha1);
1396 if (!S_ISDIR(e->versions[1].mode)) {
1397 e->tree = new_tree_content(8);
1398 e->versions[1].mode = S_IFDIR;
1402 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1403 hashclr(root->versions[1].sha1);
1410 if (t->entry_count == t->entry_capacity)
1411 root->tree = t = grow_tree_content(t, t->entry_count);
1412 e = new_tree_entry();
1413 e->name = to_atom(p, n);
1414 e->versions[0].mode = 0;
1415 hashclr(e->versions[0].sha1);
1416 t->entries[t->entry_count++] = e;
1418 e->tree = new_tree_content(8);
1419 e->versions[1].mode = S_IFDIR;
1420 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1423 e->versions[1].mode = mode;
1424 hashcpy(e->versions[1].sha1, sha1);
1426 hashclr(root->versions[1].sha1);
1430 static int tree_content_remove(
1431 struct tree_entry *root,
1433 struct tree_entry *backup_leaf)
1435 struct tree_content *t = root->tree;
1438 struct tree_entry *e;
1440 slash1 = strchr(p, '/');
1446 for (i = 0; i < t->entry_count; i++) {
1448 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1449 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1453 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1454 for (n = 0; n < e->tree->entry_count; n++) {
1455 if (e->tree->entries[n]->versions[1].mode) {
1456 hashclr(root->versions[1].sha1);
1470 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1472 release_tree_content_recursive(e->tree);
1474 e->versions[1].mode = 0;
1475 hashclr(e->versions[1].sha1);
1476 hashclr(root->versions[1].sha1);
1480 static int tree_content_get(
1481 struct tree_entry *root,
1483 struct tree_entry *leaf)
1485 struct tree_content *t = root->tree;
1488 struct tree_entry *e;
1490 slash1 = strchr(p, '/');
1496 for (i = 0; i < t->entry_count; i++) {
1498 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1500 memcpy(leaf, e, sizeof(*leaf));
1501 if (e->tree && is_null_sha1(e->versions[1].sha1))
1502 leaf->tree = dup_tree_content(e->tree);
1507 if (!S_ISDIR(e->versions[1].mode))
1511 return tree_content_get(e, slash1 + 1, leaf);
1517 static int update_branch(struct branch *b)
1519 static const char *msg = "fast-import";
1520 struct ref_lock *lock;
1521 unsigned char old_sha1[20];
1523 if (is_null_sha1(b->sha1))
1525 if (read_ref(b->name, old_sha1))
1527 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1529 return error("Unable to lock %s", b->name);
1530 if (!force_update && !is_null_sha1(old_sha1)) {
1531 struct commit *old_cmit, *new_cmit;
1533 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1534 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1535 if (!old_cmit || !new_cmit) {
1537 return error("Branch %s is missing commits.", b->name);
1540 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1542 warning("Not updating %s"
1543 " (new tip %s does not contain %s)",
1544 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1548 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1549 return error("Unable to update %s", b->name);
1553 static void dump_branches(void)
1558 for (i = 0; i < branch_table_sz; i++) {
1559 for (b = branch_table[i]; b; b = b->table_next_branch)
1560 failure |= update_branch(b);
1564 static void dump_tags(void)
1566 static const char *msg = "fast-import";
1568 struct ref_lock *lock;
1569 char ref_name[PATH_MAX];
1571 for (t = first_tag; t; t = t->next_tag) {
1572 sprintf(ref_name, "tags/%s", t->name);
1573 lock = lock_ref_sha1(ref_name, NULL);
1574 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1575 failure |= error("Unable to update %s", ref_name);
1579 static void dump_marks_helper(FILE *f,
1585 for (k = 0; k < 1024; k++) {
1586 if (m->data.sets[k])
1587 dump_marks_helper(f, (base + k) << m->shift,
1591 for (k = 0; k < 1024; k++) {
1592 if (m->data.marked[k])
1593 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1594 sha1_to_hex(m->data.marked[k]->sha1));
1599 static void dump_marks(void)
1601 static struct lock_file mark_lock;
1608 mark_fd = hold_lock_file_for_update(&mark_lock, mark_file, 0);
1610 failure |= error("Unable to write marks file %s: %s",
1611 mark_file, strerror(errno));
1615 f = fdopen(mark_fd, "w");
1617 int saved_errno = errno;
1618 rollback_lock_file(&mark_lock);
1619 failure |= error("Unable to write marks file %s: %s",
1620 mark_file, strerror(saved_errno));
1625 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1626 * Assign -1 to the lock file descriptor so that commit_lock_file()
1627 * won't try to close() it.
1631 dump_marks_helper(f, 0, marks);
1632 if (ferror(f) || fclose(f)) {
1633 int saved_errno = errno;
1634 rollback_lock_file(&mark_lock);
1635 failure |= error("Unable to write marks file %s: %s",
1636 mark_file, strerror(saved_errno));
1640 if (commit_lock_file(&mark_lock)) {
1641 int saved_errno = errno;
1642 rollback_lock_file(&mark_lock);
1643 failure |= error("Unable to commit marks file %s: %s",
1644 mark_file, strerror(saved_errno));
1649 static int read_next_command(void)
1651 static int stdin_eof = 0;
1654 unread_command_buf = 0;
1659 if (unread_command_buf) {
1660 unread_command_buf = 0;
1662 struct recent_command *rc;
1664 strbuf_detach(&command_buf, NULL);
1665 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1674 cmd_hist.next = rc->next;
1675 cmd_hist.next->prev = &cmd_hist;
1679 rc->buf = command_buf.buf;
1680 rc->prev = cmd_tail;
1681 rc->next = cmd_hist.prev;
1682 rc->prev->next = rc;
1685 } while (command_buf.buf[0] == '#');
1690 static void skip_optional_lf(void)
1692 int term_char = fgetc(stdin);
1693 if (term_char != '\n' && term_char != EOF)
1694 ungetc(term_char, stdin);
1697 static void parse_mark(void)
1699 if (!prefixcmp(command_buf.buf, "mark :")) {
1700 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1701 read_next_command();
1707 static void parse_data(struct strbuf *sb)
1711 if (prefixcmp(command_buf.buf, "data "))
1712 die("Expected 'data n' command, found: %s", command_buf.buf);
1714 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1715 char *term = xstrdup(command_buf.buf + 5 + 2);
1716 size_t term_len = command_buf.len - 5 - 2;
1718 strbuf_detach(&command_buf, NULL);
1720 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1721 die("EOF in data (terminator '%s' not found)", term);
1722 if (term_len == command_buf.len
1723 && !strcmp(term, command_buf.buf))
1725 strbuf_addbuf(sb, &command_buf);
1726 strbuf_addch(sb, '\n');
1731 size_t n = 0, length;
1733 length = strtoul(command_buf.buf + 5, NULL, 10);
1735 while (n < length) {
1736 size_t s = strbuf_fread(sb, length - n, stdin);
1737 if (!s && feof(stdin))
1738 die("EOF in data (%lu bytes remaining)",
1739 (unsigned long)(length - n));
1747 static int validate_raw_date(const char *src, char *result, int maxlen)
1749 const char *orig_src = src;
1752 strtoul(src, &endp, 10);
1753 if (endp == src || *endp != ' ')
1757 if (*src != '-' && *src != '+')
1761 strtoul(src + 1, &endp, 10);
1762 if (endp == src || *endp || (endp - orig_src) >= maxlen)
1765 strcpy(result, orig_src);
1769 static char *parse_ident(const char *buf)
1775 gt = strrchr(buf, '>');
1777 die("Missing > in ident string: %s", buf);
1780 die("Missing space after > in ident string: %s", buf);
1782 name_len = gt - buf;
1783 ident = xmalloc(name_len + 24);
1784 strncpy(ident, buf, name_len);
1788 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1789 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1791 case WHENSPEC_RFC2822:
1792 if (parse_date(gt, ident + name_len, 24) < 0)
1793 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1796 if (strcmp("now", gt))
1797 die("Date in ident must be 'now': %s", buf);
1798 datestamp(ident + name_len, 24);
1805 static void parse_new_blob(void)
1807 static struct strbuf buf = STRBUF_INIT;
1809 read_next_command();
1812 store_object(OBJ_BLOB, &buf, &last_blob, NULL, next_mark);
1815 static void unload_one_branch(void)
1817 while (cur_active_branches
1818 && cur_active_branches >= max_active_branches) {
1819 uintmax_t min_commit = ULONG_MAX;
1820 struct branch *e, *l = NULL, *p = NULL;
1822 for (e = active_branches; e; e = e->active_next_branch) {
1823 if (e->last_commit < min_commit) {
1825 min_commit = e->last_commit;
1831 e = p->active_next_branch;
1832 p->active_next_branch = e->active_next_branch;
1834 e = active_branches;
1835 active_branches = e->active_next_branch;
1838 e->active_next_branch = NULL;
1839 if (e->branch_tree.tree) {
1840 release_tree_content_recursive(e->branch_tree.tree);
1841 e->branch_tree.tree = NULL;
1843 cur_active_branches--;
1847 static void load_branch(struct branch *b)
1849 load_tree(&b->branch_tree);
1852 b->active_next_branch = active_branches;
1853 active_branches = b;
1854 cur_active_branches++;
1855 branch_load_count++;
1859 static void file_change_m(struct branch *b)
1861 const char *p = command_buf.buf + 2;
1862 static struct strbuf uq = STRBUF_INIT;
1864 struct object_entry *oe = oe;
1865 unsigned char sha1[20];
1866 uint16_t mode, inline_data = 0;
1868 p = get_mode(p, &mode);
1870 die("Corrupt mode: %s", command_buf.buf);
1872 case S_IFREG | 0644:
1873 case S_IFREG | 0755:
1881 die("Corrupt mode: %s", command_buf.buf);
1886 oe = find_mark(strtoumax(p + 1, &x, 10));
1887 hashcpy(sha1, oe->sha1);
1889 } else if (!prefixcmp(p, "inline")) {
1893 if (get_sha1_hex(p, sha1))
1894 die("Invalid SHA1: %s", command_buf.buf);
1895 oe = find_object(sha1);
1899 die("Missing space after SHA1: %s", command_buf.buf);
1902 if (!unquote_c_style(&uq, p, &endp)) {
1904 die("Garbage after path in: %s", command_buf.buf);
1908 if (S_ISGITLINK(mode)) {
1910 die("Git links cannot be specified 'inline': %s",
1913 if (oe->type != OBJ_COMMIT)
1914 die("Not a commit (actually a %s): %s",
1915 typename(oe->type), command_buf.buf);
1918 * Accept the sha1 without checking; it expected to be in
1919 * another repository.
1921 } else if (inline_data) {
1922 static struct strbuf buf = STRBUF_INIT;
1925 strbuf_addstr(&uq, p);
1928 read_next_command();
1930 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
1932 if (oe->type != OBJ_BLOB)
1933 die("Not a blob (actually a %s): %s",
1934 typename(oe->type), command_buf.buf);
1936 enum object_type type = sha1_object_info(sha1, NULL);
1938 die("Blob not found: %s", command_buf.buf);
1939 if (type != OBJ_BLOB)
1940 die("Not a blob (actually a %s): %s",
1941 typename(type), command_buf.buf);
1944 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode, NULL);
1947 static void file_change_d(struct branch *b)
1949 const char *p = command_buf.buf + 2;
1950 static struct strbuf uq = STRBUF_INIT;
1954 if (!unquote_c_style(&uq, p, &endp)) {
1956 die("Garbage after path in: %s", command_buf.buf);
1959 tree_content_remove(&b->branch_tree, p, NULL);
1962 static void file_change_cr(struct branch *b, int rename)
1965 static struct strbuf s_uq = STRBUF_INIT;
1966 static struct strbuf d_uq = STRBUF_INIT;
1968 struct tree_entry leaf;
1970 s = command_buf.buf + 2;
1971 strbuf_reset(&s_uq);
1972 if (!unquote_c_style(&s_uq, s, &endp)) {
1974 die("Missing space after source: %s", command_buf.buf);
1976 endp = strchr(s, ' ');
1978 die("Missing space after source: %s", command_buf.buf);
1979 strbuf_add(&s_uq, s, endp - s);
1985 die("Missing dest: %s", command_buf.buf);
1988 strbuf_reset(&d_uq);
1989 if (!unquote_c_style(&d_uq, d, &endp)) {
1991 die("Garbage after dest in: %s", command_buf.buf);
1995 memset(&leaf, 0, sizeof(leaf));
1997 tree_content_remove(&b->branch_tree, s, &leaf);
1999 tree_content_get(&b->branch_tree, s, &leaf);
2000 if (!leaf.versions[1].mode)
2001 die("Path %s not in branch", s);
2002 tree_content_set(&b->branch_tree, d,
2003 leaf.versions[1].sha1,
2004 leaf.versions[1].mode,
2008 static void file_change_deleteall(struct branch *b)
2010 release_tree_content_recursive(b->branch_tree.tree);
2011 hashclr(b->branch_tree.versions[0].sha1);
2012 hashclr(b->branch_tree.versions[1].sha1);
2013 load_tree(&b->branch_tree);
2016 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2018 if (!buf || size < 46)
2019 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2020 if (memcmp("tree ", buf, 5)
2021 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2022 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2023 hashcpy(b->branch_tree.versions[0].sha1,
2024 b->branch_tree.versions[1].sha1);
2027 static void parse_from_existing(struct branch *b)
2029 if (is_null_sha1(b->sha1)) {
2030 hashclr(b->branch_tree.versions[0].sha1);
2031 hashclr(b->branch_tree.versions[1].sha1);
2036 buf = read_object_with_reference(b->sha1,
2037 commit_type, &size, b->sha1);
2038 parse_from_commit(b, buf, size);
2043 static int parse_from(struct branch *b)
2048 if (prefixcmp(command_buf.buf, "from "))
2051 if (b->branch_tree.tree) {
2052 release_tree_content_recursive(b->branch_tree.tree);
2053 b->branch_tree.tree = NULL;
2056 from = strchr(command_buf.buf, ' ') + 1;
2057 s = lookup_branch(from);
2059 die("Can't create a branch from itself: %s", b->name);
2061 unsigned char *t = s->branch_tree.versions[1].sha1;
2062 hashcpy(b->sha1, s->sha1);
2063 hashcpy(b->branch_tree.versions[0].sha1, t);
2064 hashcpy(b->branch_tree.versions[1].sha1, t);
2065 } else if (*from == ':') {
2066 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2067 struct object_entry *oe = find_mark(idnum);
2068 if (oe->type != OBJ_COMMIT)
2069 die("Mark :%" PRIuMAX " not a commit", idnum);
2070 hashcpy(b->sha1, oe->sha1);
2071 if (oe->pack_id != MAX_PACK_ID) {
2073 char *buf = gfi_unpack_entry(oe, &size);
2074 parse_from_commit(b, buf, size);
2077 parse_from_existing(b);
2078 } else if (!get_sha1(from, b->sha1))
2079 parse_from_existing(b);
2081 die("Invalid ref name or SHA1 expression: %s", from);
2083 read_next_command();
2087 static struct hash_list *parse_merge(unsigned int *count)
2089 struct hash_list *list = NULL, *n, *e = e;
2094 while (!prefixcmp(command_buf.buf, "merge ")) {
2095 from = strchr(command_buf.buf, ' ') + 1;
2096 n = xmalloc(sizeof(*n));
2097 s = lookup_branch(from);
2099 hashcpy(n->sha1, s->sha1);
2100 else if (*from == ':') {
2101 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2102 struct object_entry *oe = find_mark(idnum);
2103 if (oe->type != OBJ_COMMIT)
2104 die("Mark :%" PRIuMAX " not a commit", idnum);
2105 hashcpy(n->sha1, oe->sha1);
2106 } else if (!get_sha1(from, n->sha1)) {
2108 char *buf = read_object_with_reference(n->sha1,
2109 commit_type, &size, n->sha1);
2110 if (!buf || size < 46)
2111 die("Not a valid commit: %s", from);
2114 die("Invalid ref name or SHA1 expression: %s", from);
2123 read_next_command();
2128 static void parse_new_commit(void)
2130 static struct strbuf msg = STRBUF_INIT;
2133 char *author = NULL;
2134 char *committer = NULL;
2135 struct hash_list *merge_list = NULL;
2136 unsigned int merge_count;
2138 /* Obtain the branch name from the rest of our command */
2139 sp = strchr(command_buf.buf, ' ') + 1;
2140 b = lookup_branch(sp);
2144 read_next_command();
2146 if (!prefixcmp(command_buf.buf, "author ")) {
2147 author = parse_ident(command_buf.buf + 7);
2148 read_next_command();
2150 if (!prefixcmp(command_buf.buf, "committer ")) {
2151 committer = parse_ident(command_buf.buf + 10);
2152 read_next_command();
2155 die("Expected committer but didn't get one");
2157 read_next_command();
2159 merge_list = parse_merge(&merge_count);
2161 /* ensure the branch is active/loaded */
2162 if (!b->branch_tree.tree || !max_active_branches) {
2163 unload_one_branch();
2168 while (command_buf.len > 0) {
2169 if (!prefixcmp(command_buf.buf, "M "))
2171 else if (!prefixcmp(command_buf.buf, "D "))
2173 else if (!prefixcmp(command_buf.buf, "R "))
2174 file_change_cr(b, 1);
2175 else if (!prefixcmp(command_buf.buf, "C "))
2176 file_change_cr(b, 0);
2177 else if (!strcmp("deleteall", command_buf.buf))
2178 file_change_deleteall(b);
2180 unread_command_buf = 1;
2183 if (read_next_command() == EOF)
2187 /* build the tree and the commit */
2188 store_tree(&b->branch_tree);
2189 hashcpy(b->branch_tree.versions[0].sha1,
2190 b->branch_tree.versions[1].sha1);
2192 strbuf_reset(&new_data);
2193 strbuf_addf(&new_data, "tree %s\n",
2194 sha1_to_hex(b->branch_tree.versions[1].sha1));
2195 if (!is_null_sha1(b->sha1))
2196 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2197 while (merge_list) {
2198 struct hash_list *next = merge_list->next;
2199 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2203 strbuf_addf(&new_data,
2207 author ? author : committer, committer);
2208 strbuf_addbuf(&new_data, &msg);
2212 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2213 b->pack_id = pack_id;
2214 b->last_commit = object_count_by_type[OBJ_COMMIT];
2217 static void parse_new_tag(void)
2219 static struct strbuf msg = STRBUF_INIT;
2225 uintmax_t from_mark = 0;
2226 unsigned char sha1[20];
2228 /* Obtain the new tag name from the rest of our command */
2229 sp = strchr(command_buf.buf, ' ') + 1;
2230 t = pool_alloc(sizeof(struct tag));
2232 t->name = pool_strdup(sp);
2234 last_tag->next_tag = t;
2238 read_next_command();
2241 if (prefixcmp(command_buf.buf, "from "))
2242 die("Expected from command, got %s", command_buf.buf);
2243 from = strchr(command_buf.buf, ' ') + 1;
2244 s = lookup_branch(from);
2246 hashcpy(sha1, s->sha1);
2247 } else if (*from == ':') {
2248 struct object_entry *oe;
2249 from_mark = strtoumax(from + 1, NULL, 10);
2250 oe = find_mark(from_mark);
2251 if (oe->type != OBJ_COMMIT)
2252 die("Mark :%" PRIuMAX " not a commit", from_mark);
2253 hashcpy(sha1, oe->sha1);
2254 } else if (!get_sha1(from, sha1)) {
2258 buf = read_object_with_reference(sha1,
2259 commit_type, &size, sha1);
2260 if (!buf || size < 46)
2261 die("Not a valid commit: %s", from);
2264 die("Invalid ref name or SHA1 expression: %s", from);
2265 read_next_command();
2268 if (prefixcmp(command_buf.buf, "tagger "))
2269 die("Expected tagger command, got %s", command_buf.buf);
2270 tagger = parse_ident(command_buf.buf + 7);
2272 /* tag payload/message */
2273 read_next_command();
2276 /* build the tag object */
2277 strbuf_reset(&new_data);
2278 strbuf_addf(&new_data,
2284 sha1_to_hex(sha1), commit_type, t->name, tagger);
2285 strbuf_addbuf(&new_data, &msg);
2288 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2289 t->pack_id = MAX_PACK_ID;
2291 t->pack_id = pack_id;
2294 static void parse_reset_branch(void)
2299 /* Obtain the branch name from the rest of our command */
2300 sp = strchr(command_buf.buf, ' ') + 1;
2301 b = lookup_branch(sp);
2304 hashclr(b->branch_tree.versions[0].sha1);
2305 hashclr(b->branch_tree.versions[1].sha1);
2306 if (b->branch_tree.tree) {
2307 release_tree_content_recursive(b->branch_tree.tree);
2308 b->branch_tree.tree = NULL;
2313 read_next_command();
2315 if (command_buf.len > 0)
2316 unread_command_buf = 1;
2319 static void parse_checkpoint(void)
2330 static void parse_progress(void)
2332 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2333 fputc('\n', stdout);
2338 static void import_marks(const char *input_file)
2341 FILE *f = fopen(input_file, "r");
2343 die("cannot read %s: %s", input_file, strerror(errno));
2344 while (fgets(line, sizeof(line), f)) {
2347 unsigned char sha1[20];
2348 struct object_entry *e;
2350 end = strchr(line, '\n');
2351 if (line[0] != ':' || !end)
2352 die("corrupt mark line: %s", line);
2354 mark = strtoumax(line + 1, &end, 10);
2355 if (!mark || end == line + 1
2356 || *end != ' ' || get_sha1(end + 1, sha1))
2357 die("corrupt mark line: %s", line);
2358 e = find_object(sha1);
2360 enum object_type type = sha1_object_info(sha1, NULL);
2362 die("object not found: %s", sha1_to_hex(sha1));
2363 e = insert_object(sha1);
2365 e->pack_id = MAX_PACK_ID;
2366 e->offset = 1; /* just not zero! */
2368 insert_mark(mark, e);
2373 static int git_pack_config(const char *k, const char *v, void *cb)
2375 if (!strcmp(k, "pack.depth")) {
2376 max_depth = git_config_int(k, v);
2377 if (max_depth > MAX_DEPTH)
2378 max_depth = MAX_DEPTH;
2381 if (!strcmp(k, "pack.compression")) {
2382 int level = git_config_int(k, v);
2384 level = Z_DEFAULT_COMPRESSION;
2385 else if (level < 0 || level > Z_BEST_COMPRESSION)
2386 die("bad pack compression level %d", level);
2387 pack_compression_level = level;
2388 pack_compression_seen = 1;
2391 return git_default_config(k, v, cb);
2394 static const char fast_import_usage[] =
2395 "git fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2397 int main(int argc, const char **argv)
2399 unsigned int i, show_stats = 1;
2401 setup_git_directory();
2402 git_config(git_pack_config, NULL);
2403 if (!pack_compression_seen && core_compression_seen)
2404 pack_compression_level = core_compression_level;
2406 alloc_objects(object_entry_alloc);
2407 strbuf_init(&command_buf, 0);
2408 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2409 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2410 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2411 marks = pool_calloc(1, sizeof(struct mark_set));
2413 for (i = 1; i < argc; i++) {
2414 const char *a = argv[i];
2416 if (*a != '-' || !strcmp(a, "--"))
2418 else if (!prefixcmp(a, "--date-format=")) {
2419 const char *fmt = a + 14;
2420 if (!strcmp(fmt, "raw"))
2421 whenspec = WHENSPEC_RAW;
2422 else if (!strcmp(fmt, "rfc2822"))
2423 whenspec = WHENSPEC_RFC2822;
2424 else if (!strcmp(fmt, "now"))
2425 whenspec = WHENSPEC_NOW;
2427 die("unknown --date-format argument %s", fmt);
2429 else if (!prefixcmp(a, "--max-pack-size="))
2430 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
2431 else if (!prefixcmp(a, "--depth=")) {
2432 max_depth = strtoul(a + 8, NULL, 0);
2433 if (max_depth > MAX_DEPTH)
2434 die("--depth cannot exceed %u", MAX_DEPTH);
2436 else if (!prefixcmp(a, "--active-branches="))
2437 max_active_branches = strtoul(a + 18, NULL, 0);
2438 else if (!prefixcmp(a, "--import-marks="))
2439 import_marks(a + 15);
2440 else if (!prefixcmp(a, "--export-marks="))
2442 else if (!prefixcmp(a, "--export-pack-edges=")) {
2445 pack_edges = fopen(a + 20, "a");
2447 die("Cannot open %s: %s", a + 20, strerror(errno));
2448 } else if (!strcmp(a, "--force"))
2450 else if (!strcmp(a, "--quiet"))
2452 else if (!strcmp(a, "--stats"))
2455 die("unknown option %s", a);
2458 usage(fast_import_usage);
2460 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2461 for (i = 0; i < (cmd_save - 1); i++)
2462 rc_free[i].next = &rc_free[i + 1];
2463 rc_free[cmd_save - 1].next = NULL;
2465 prepare_packed_git();
2467 set_die_routine(die_nicely);
2468 while (read_next_command() != EOF) {
2469 if (!strcmp("blob", command_buf.buf))
2471 else if (!prefixcmp(command_buf.buf, "commit "))
2473 else if (!prefixcmp(command_buf.buf, "tag "))
2475 else if (!prefixcmp(command_buf.buf, "reset "))
2476 parse_reset_branch();
2477 else if (!strcmp("checkpoint", command_buf.buf))
2479 else if (!prefixcmp(command_buf.buf, "progress "))
2482 die("Unsupported command: %s", command_buf.buf);
2495 uintmax_t total_count = 0, duplicate_count = 0;
2496 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2497 total_count += object_count_by_type[i];
2498 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2499 duplicate_count += duplicate_count_by_type[i];
2501 fprintf(stderr, "%s statistics:\n", argv[0]);
2502 fprintf(stderr, "---------------------------------------------------------------------\n");
2503 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2504 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2505 fprintf(stderr, " blobs : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_BLOB], duplicate_count_by_type[OBJ_BLOB], delta_count_by_type[OBJ_BLOB]);
2506 fprintf(stderr, " trees : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TREE], duplicate_count_by_type[OBJ_TREE], delta_count_by_type[OBJ_TREE]);
2507 fprintf(stderr, " commits: %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_COMMIT], duplicate_count_by_type[OBJ_COMMIT], delta_count_by_type[OBJ_COMMIT]);
2508 fprintf(stderr, " tags : %10" PRIuMAX " (%10" PRIuMAX " duplicates %10" PRIuMAX " deltas)\n", object_count_by_type[OBJ_TAG], duplicate_count_by_type[OBJ_TAG], delta_count_by_type[OBJ_TAG]);
2509 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2510 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2511 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2512 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2513 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2514 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2515 fprintf(stderr, "---------------------------------------------------------------------\n");
2517 fprintf(stderr, "---------------------------------------------------------------------\n");
2518 fprintf(stderr, "\n");
2521 return failure ? 1 : 0;