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"
153 #include "exec_cmd.h"
155 #define PACK_ID_BITS 16
156 #define MAX_PACK_ID ((1<<PACK_ID_BITS)-1)
157 #define DEPTH_BITS 13
158 #define MAX_DEPTH ((1<<DEPTH_BITS)-1)
162 struct object_entry *next;
164 uint32_t type : TYPE_BITS,
165 pack_id : PACK_ID_BITS,
167 unsigned char sha1[20];
170 struct object_entry_pool
172 struct object_entry_pool *next_pool;
173 struct object_entry *next_free;
174 struct object_entry *end;
175 struct object_entry entries[FLEX_ARRAY]; /* more */
181 struct object_entry *marked[1024];
182 struct mark_set *sets[1024];
192 unsigned no_swap : 1;
197 struct mem_pool *next_pool;
200 uintmax_t space[FLEX_ARRAY]; /* more */
205 struct atom_str *next_atom;
206 unsigned short str_len;
207 char str_dat[FLEX_ARRAY]; /* more */
213 struct tree_content *tree;
214 struct atom_str* name;
218 unsigned char sha1[20];
224 unsigned int entry_capacity; /* must match avail_tree_content */
225 unsigned int entry_count;
226 unsigned int delta_depth;
227 struct tree_entry *entries[FLEX_ARRAY]; /* more */
230 struct avail_tree_content
232 unsigned int entry_capacity; /* must match tree_content */
233 struct avail_tree_content *next_avail;
238 struct branch *table_next_branch;
239 struct branch *active_next_branch;
241 struct tree_entry branch_tree;
242 uintmax_t last_commit;
244 unsigned pack_id : PACK_ID_BITS;
245 unsigned char sha1[20];
250 struct tag *next_tag;
252 unsigned int pack_id;
253 unsigned char sha1[20];
258 struct hash_list *next;
259 unsigned char sha1[20];
268 struct recent_command
270 struct recent_command *prev;
271 struct recent_command *next;
275 /* Configured limits on output */
276 static unsigned long max_depth = 10;
277 static off_t max_packsize = (1LL << 32) - 1;
278 static int force_update;
279 static int pack_compression_level = Z_DEFAULT_COMPRESSION;
280 static int pack_compression_seen;
282 /* Stats and misc. counters */
283 static uintmax_t alloc_count;
284 static uintmax_t marks_set_count;
285 static uintmax_t object_count_by_type[1 << TYPE_BITS];
286 static uintmax_t duplicate_count_by_type[1 << TYPE_BITS];
287 static uintmax_t delta_count_by_type[1 << TYPE_BITS];
288 static unsigned long object_count;
289 static unsigned long branch_count;
290 static unsigned long branch_load_count;
292 static FILE *pack_edges;
295 static size_t mem_pool_alloc = 2*1024*1024 - sizeof(struct mem_pool);
296 static size_t total_allocd;
297 static struct mem_pool *mem_pool;
299 /* Atom management */
300 static unsigned int atom_table_sz = 4451;
301 static unsigned int atom_cnt;
302 static struct atom_str **atom_table;
304 /* The .pack file being generated */
305 static unsigned int pack_id;
306 static struct packed_git *pack_data;
307 static struct packed_git **all_packs;
308 static unsigned long pack_size;
310 /* Table of objects we've written. */
311 static unsigned int object_entry_alloc = 5000;
312 static struct object_entry_pool *blocks;
313 static struct object_entry *object_table[1 << 16];
314 static struct mark_set *marks;
315 static const char* mark_file;
318 static struct last_object last_blob = { STRBUF_INIT, 0, 0, 0 };
320 /* Tree management */
321 static unsigned int tree_entry_alloc = 1000;
322 static void *avail_tree_entry;
323 static unsigned int avail_tree_table_sz = 100;
324 static struct avail_tree_content **avail_tree_table;
325 static struct strbuf old_tree = STRBUF_INIT;
326 static struct strbuf new_tree = STRBUF_INIT;
329 static unsigned long max_active_branches = 5;
330 static unsigned long cur_active_branches;
331 static unsigned long branch_table_sz = 1039;
332 static struct branch **branch_table;
333 static struct branch *active_branches;
336 static struct tag *first_tag;
337 static struct tag *last_tag;
339 /* Input stream parsing */
340 static whenspec_type whenspec = WHENSPEC_RAW;
341 static struct strbuf command_buf = STRBUF_INIT;
342 static int unread_command_buf;
343 static struct recent_command cmd_hist = {&cmd_hist, &cmd_hist, NULL};
344 static struct recent_command *cmd_tail = &cmd_hist;
345 static struct recent_command *rc_free;
346 static unsigned int cmd_save = 100;
347 static uintmax_t next_mark;
348 static struct strbuf new_data = STRBUF_INIT;
350 static void write_branch_report(FILE *rpt, struct branch *b)
352 fprintf(rpt, "%s:\n", b->name);
354 fprintf(rpt, " status :");
356 fputs(" active", rpt);
357 if (b->branch_tree.tree)
358 fputs(" loaded", rpt);
359 if (is_null_sha1(b->branch_tree.versions[1].sha1))
360 fputs(" dirty", rpt);
363 fprintf(rpt, " tip commit : %s\n", sha1_to_hex(b->sha1));
364 fprintf(rpt, " old tree : %s\n", sha1_to_hex(b->branch_tree.versions[0].sha1));
365 fprintf(rpt, " cur tree : %s\n", sha1_to_hex(b->branch_tree.versions[1].sha1));
366 fprintf(rpt, " commit clock: %" PRIuMAX "\n", b->last_commit);
368 fputs(" last pack : ", rpt);
369 if (b->pack_id < MAX_PACK_ID)
370 fprintf(rpt, "%u", b->pack_id);
376 static void dump_marks_helper(FILE *, uintmax_t, struct mark_set *);
378 static void write_crash_report(const char *err)
380 char *loc = git_path("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid());
381 FILE *rpt = fopen(loc, "w");
384 struct recent_command *rc;
387 error("can't write crash report %s: %s", loc, strerror(errno));
391 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
393 fprintf(rpt, "fast-import crash report:\n");
394 fprintf(rpt, " fast-import process: %"PRIuMAX"\n", (uintmax_t) getpid());
395 fprintf(rpt, " parent process : %"PRIuMAX"\n", (uintmax_t) getppid());
396 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
399 fputs("fatal: ", rpt);
404 fputs("Most Recent Commands Before Crash\n", rpt);
405 fputs("---------------------------------\n", rpt);
406 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
407 if (rc->next == &cmd_hist)
416 fputs("Active Branch LRU\n", rpt);
417 fputs("-----------------\n", rpt);
418 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
420 max_active_branches);
422 fputs(" pos clock name\n", rpt);
423 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
424 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
425 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
426 ++lu, b->last_commit, b->name);
429 fputs("Inactive Branches\n", rpt);
430 fputs("-----------------\n", rpt);
431 for (lu = 0; lu < branch_table_sz; lu++) {
432 for (b = branch_table[lu]; b; b = b->table_next_branch)
433 write_branch_report(rpt, b);
439 fputs("Annotated Tags\n", rpt);
440 fputs("--------------\n", rpt);
441 for (tg = first_tag; tg; tg = tg->next_tag) {
442 fputs(sha1_to_hex(tg->sha1), rpt);
444 fputs(tg->name, rpt);
450 fputs("Marks\n", rpt);
451 fputs("-----\n", rpt);
453 fprintf(rpt, " exported to %s\n", mark_file);
455 dump_marks_helper(rpt, 0, marks);
458 fputs("-------------------\n", rpt);
459 fputs("END OF CRASH REPORT\n", rpt);
463 static void end_packfile(void);
464 static void unkeep_all_packs(void);
465 static void dump_marks(void);
467 static NORETURN void die_nicely(const char *err, va_list params)
470 char message[2 * PATH_MAX];
472 vsnprintf(message, sizeof(message), err, params);
473 fputs("fatal: ", stderr);
474 fputs(message, stderr);
479 write_crash_report(message);
487 static void alloc_objects(unsigned int cnt)
489 struct object_entry_pool *b;
491 b = xmalloc(sizeof(struct object_entry_pool)
492 + cnt * sizeof(struct object_entry));
493 b->next_pool = blocks;
494 b->next_free = b->entries;
495 b->end = b->entries + cnt;
500 static struct object_entry *new_object(unsigned char *sha1)
502 struct object_entry *e;
504 if (blocks->next_free == blocks->end)
505 alloc_objects(object_entry_alloc);
507 e = blocks->next_free++;
508 hashcpy(e->sha1, sha1);
512 static struct object_entry *find_object(unsigned char *sha1)
514 unsigned int h = sha1[0] << 8 | sha1[1];
515 struct object_entry *e;
516 for (e = object_table[h]; e; e = e->next)
517 if (!hashcmp(sha1, e->sha1))
522 static struct object_entry *insert_object(unsigned char *sha1)
524 unsigned int h = sha1[0] << 8 | sha1[1];
525 struct object_entry *e = object_table[h];
526 struct object_entry *p = NULL;
529 if (!hashcmp(sha1, e->sha1))
535 e = new_object(sha1);
545 static unsigned int hc_str(const char *s, size_t len)
553 static void *pool_alloc(size_t len)
558 /* round up to a 'uintmax_t' alignment */
559 if (len & (sizeof(uintmax_t) - 1))
560 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
562 for (p = mem_pool; p; p = p->next_pool)
563 if ((p->end - p->next_free >= len))
567 if (len >= (mem_pool_alloc/2)) {
571 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
572 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
573 p->next_pool = mem_pool;
574 p->next_free = (char *) p->space;
575 p->end = p->next_free + mem_pool_alloc;
584 static void *pool_calloc(size_t count, size_t size)
586 size_t len = count * size;
587 void *r = pool_alloc(len);
592 static char *pool_strdup(const char *s)
594 char *r = pool_alloc(strlen(s) + 1);
599 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
601 struct mark_set *s = marks;
602 while ((idnum >> s->shift) >= 1024) {
603 s = pool_calloc(1, sizeof(struct mark_set));
604 s->shift = marks->shift + 10;
605 s->data.sets[0] = marks;
609 uintmax_t i = idnum >> s->shift;
610 idnum -= i << s->shift;
611 if (!s->data.sets[i]) {
612 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
613 s->data.sets[i]->shift = s->shift - 10;
617 if (!s->data.marked[idnum])
619 s->data.marked[idnum] = oe;
622 static struct object_entry *find_mark(uintmax_t idnum)
624 uintmax_t orig_idnum = idnum;
625 struct mark_set *s = marks;
626 struct object_entry *oe = NULL;
627 if ((idnum >> s->shift) < 1024) {
628 while (s && s->shift) {
629 uintmax_t i = idnum >> s->shift;
630 idnum -= i << s->shift;
634 oe = s->data.marked[idnum];
637 die("mark :%" PRIuMAX " not declared", orig_idnum);
641 static struct atom_str *to_atom(const char *s, unsigned short len)
643 unsigned int hc = hc_str(s, len) % atom_table_sz;
646 for (c = atom_table[hc]; c; c = c->next_atom)
647 if (c->str_len == len && !strncmp(s, c->str_dat, len))
650 c = pool_alloc(sizeof(struct atom_str) + len + 1);
652 strncpy(c->str_dat, s, len);
654 c->next_atom = atom_table[hc];
660 static struct branch *lookup_branch(const char *name)
662 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
665 for (b = branch_table[hc]; b; b = b->table_next_branch)
666 if (!strcmp(name, b->name))
671 static struct branch *new_branch(const char *name)
673 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
674 struct branch* b = lookup_branch(name);
677 die("Invalid attempt to create duplicate branch: %s", name);
678 switch (check_ref_format(name)) {
679 case 0: break; /* its valid */
680 case CHECK_REF_FORMAT_ONELEVEL:
681 break; /* valid, but too few '/', allow anyway */
683 die("Branch name doesn't conform to GIT standards: %s", name);
686 b = pool_calloc(1, sizeof(struct branch));
687 b->name = pool_strdup(name);
688 b->table_next_branch = branch_table[hc];
689 b->branch_tree.versions[0].mode = S_IFDIR;
690 b->branch_tree.versions[1].mode = S_IFDIR;
692 b->pack_id = MAX_PACK_ID;
693 branch_table[hc] = b;
698 static unsigned int hc_entries(unsigned int cnt)
700 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
701 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
704 static struct tree_content *new_tree_content(unsigned int cnt)
706 struct avail_tree_content *f, *l = NULL;
707 struct tree_content *t;
708 unsigned int hc = hc_entries(cnt);
710 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
711 if (f->entry_capacity >= cnt)
716 l->next_avail = f->next_avail;
718 avail_tree_table[hc] = f->next_avail;
720 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
721 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
722 f->entry_capacity = cnt;
725 t = (struct tree_content*)f;
731 static void release_tree_entry(struct tree_entry *e);
732 static void release_tree_content(struct tree_content *t)
734 struct avail_tree_content *f = (struct avail_tree_content*)t;
735 unsigned int hc = hc_entries(f->entry_capacity);
736 f->next_avail = avail_tree_table[hc];
737 avail_tree_table[hc] = f;
740 static void release_tree_content_recursive(struct tree_content *t)
743 for (i = 0; i < t->entry_count; i++)
744 release_tree_entry(t->entries[i]);
745 release_tree_content(t);
748 static struct tree_content *grow_tree_content(
749 struct tree_content *t,
752 struct tree_content *r = new_tree_content(t->entry_count + amt);
753 r->entry_count = t->entry_count;
754 r->delta_depth = t->delta_depth;
755 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
756 release_tree_content(t);
760 static struct tree_entry *new_tree_entry(void)
762 struct tree_entry *e;
764 if (!avail_tree_entry) {
765 unsigned int n = tree_entry_alloc;
766 total_allocd += n * sizeof(struct tree_entry);
767 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
769 *((void**)e) = e + 1;
775 e = avail_tree_entry;
776 avail_tree_entry = *((void**)e);
780 static void release_tree_entry(struct tree_entry *e)
783 release_tree_content_recursive(e->tree);
784 *((void**)e) = avail_tree_entry;
785 avail_tree_entry = e;
788 static struct tree_content *dup_tree_content(struct tree_content *s)
790 struct tree_content *d;
791 struct tree_entry *a, *b;
796 d = new_tree_content(s->entry_count);
797 for (i = 0; i < s->entry_count; i++) {
799 b = new_tree_entry();
800 memcpy(b, a, sizeof(*a));
801 if (a->tree && is_null_sha1(b->versions[1].sha1))
802 b->tree = dup_tree_content(a->tree);
807 d->entry_count = s->entry_count;
808 d->delta_depth = s->delta_depth;
813 static void start_packfile(void)
815 static char tmpfile[PATH_MAX];
816 struct packed_git *p;
817 struct pack_header hdr;
820 snprintf(tmpfile, sizeof(tmpfile),
821 "%s/pack/tmp_pack_XXXXXX", get_object_directory());
822 pack_fd = xmkstemp(tmpfile);
823 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
824 strcpy(p->pack_name, tmpfile);
825 p->pack_fd = pack_fd;
827 hdr.hdr_signature = htonl(PACK_SIGNATURE);
828 hdr.hdr_version = htonl(2);
830 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
833 pack_size = sizeof(hdr);
836 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
837 all_packs[pack_id] = p;
840 static int oecmp (const void *a_, const void *b_)
842 struct object_entry *a = *((struct object_entry**)a_);
843 struct object_entry *b = *((struct object_entry**)b_);
844 return hashcmp(a->sha1, b->sha1);
847 static char *create_index(void)
849 static char tmpfile[PATH_MAX];
852 struct object_entry **idx, **c, **last, *e;
853 struct object_entry_pool *o;
857 /* Build the sorted table of object IDs. */
858 idx = xmalloc(object_count * sizeof(struct object_entry*));
860 for (o = blocks; o; o = o->next_pool)
861 for (e = o->next_free; e-- != o->entries;)
862 if (pack_id == e->pack_id)
864 last = idx + object_count;
866 die("internal consistency error creating the index");
867 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
869 /* Generate the fan-out array. */
871 for (i = 0; i < 256; i++) {
872 struct object_entry **next = c;;
873 while (next < last) {
874 if ((*next)->sha1[0] != i)
878 array[i] = htonl(next - idx);
882 snprintf(tmpfile, sizeof(tmpfile),
883 "%s/pack/tmp_idx_XXXXXX", get_object_directory());
884 idx_fd = xmkstemp(tmpfile);
885 f = sha1fd(idx_fd, tmpfile);
886 sha1write(f, array, 256 * sizeof(int));
888 for (c = idx; c != last; c++) {
889 uint32_t offset = htonl((*c)->offset);
890 sha1write(f, &offset, 4);
891 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
892 git_SHA1_Update(&ctx, (*c)->sha1, 20);
894 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
895 sha1close(f, NULL, CSUM_FSYNC);
897 git_SHA1_Final(pack_data->sha1, &ctx);
901 static char *keep_pack(char *curr_index_name)
903 static char name[PATH_MAX];
904 static const char *keep_msg = "fast-import";
907 chmod(pack_data->pack_name, 0444);
908 chmod(curr_index_name, 0444);
910 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
911 get_object_directory(), sha1_to_hex(pack_data->sha1));
912 keep_fd = open(name, O_RDWR|O_CREAT|O_EXCL, 0600);
914 die("cannot create keep file");
915 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
917 die("failed to write keep file");
919 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
920 get_object_directory(), sha1_to_hex(pack_data->sha1));
921 if (move_temp_to_file(pack_data->pack_name, name))
922 die("cannot store pack file");
924 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
925 get_object_directory(), sha1_to_hex(pack_data->sha1));
926 if (move_temp_to_file(curr_index_name, name))
927 die("cannot store index file");
931 static void unkeep_all_packs(void)
933 static char name[PATH_MAX];
936 for (k = 0; k < pack_id; k++) {
937 struct packed_git *p = all_packs[k];
938 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
939 get_object_directory(), sha1_to_hex(p->sha1));
944 static void end_packfile(void)
946 struct packed_git *old_p = pack_data, *new_p;
954 close_pack_windows(pack_data);
955 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
956 pack_data->pack_name, object_count,
958 close(pack_data->pack_fd);
959 idx_name = keep_pack(create_index());
961 /* Register the packfile with core git's machinary. */
962 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
964 die("core git rejected index %s", idx_name);
965 all_packs[pack_id] = new_p;
966 install_packed_git(new_p);
968 /* Print the boundary */
970 fprintf(pack_edges, "%s:", new_p->pack_name);
971 for (i = 0; i < branch_table_sz; i++) {
972 for (b = branch_table[i]; b; b = b->table_next_branch) {
973 if (b->pack_id == pack_id)
974 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
977 for (t = first_tag; t; t = t->next_tag) {
978 if (t->pack_id == pack_id)
979 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
981 fputc('\n', pack_edges);
988 close(old_p->pack_fd);
989 unlink(old_p->pack_name);
993 /* We can't carry a delta across packfiles. */
994 strbuf_release(&last_blob.data);
995 last_blob.offset = 0;
999 static void cycle_packfile(void)
1005 static size_t encode_header(
1006 enum object_type type,
1013 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
1014 die("bad type %d", type);
1016 c = (type << 4) | (size & 15);
1028 static int store_object(
1029 enum object_type type,
1031 struct last_object *last,
1032 unsigned char *sha1out,
1036 struct object_entry *e;
1037 unsigned char hdr[96];
1038 unsigned char sha1[20];
1039 unsigned long hdrlen, deltalen;
1043 hdrlen = sprintf((char*)hdr,"%s %lu", typename(type),
1044 (unsigned long)dat->len) + 1;
1046 git_SHA1_Update(&c, hdr, hdrlen);
1047 git_SHA1_Update(&c, dat->buf, dat->len);
1048 git_SHA1_Final(sha1, &c);
1050 hashcpy(sha1out, sha1);
1052 e = insert_object(sha1);
1054 insert_mark(mark, e);
1056 duplicate_count_by_type[type]++;
1058 } else if (find_sha1_pack(sha1, packed_git)) {
1060 e->pack_id = MAX_PACK_ID;
1061 e->offset = 1; /* just not zero! */
1062 duplicate_count_by_type[type]++;
1066 if (last && last->data.buf && last->depth < max_depth) {
1067 delta = diff_delta(last->data.buf, last->data.len,
1070 if (delta && deltalen >= dat->len) {
1077 memset(&s, 0, sizeof(s));
1078 deflateInit(&s, pack_compression_level);
1081 s.avail_in = deltalen;
1083 s.next_in = (void *)dat->buf;
1084 s.avail_in = dat->len;
1086 s.avail_out = deflateBound(&s, s.avail_in);
1087 s.next_out = out = xmalloc(s.avail_out);
1088 while (deflate(&s, Z_FINISH) == Z_OK)
1092 /* Determine if we should auto-checkpoint. */
1093 if ((pack_size + 60 + s.total_out) > max_packsize
1094 || (pack_size + 60 + s.total_out) < pack_size) {
1096 /* This new object needs to *not* have the current pack_id. */
1097 e->pack_id = pack_id + 1;
1100 /* We cannot carry a delta into the new pack. */
1105 memset(&s, 0, sizeof(s));
1106 deflateInit(&s, pack_compression_level);
1107 s.next_in = (void *)dat->buf;
1108 s.avail_in = dat->len;
1109 s.avail_out = deflateBound(&s, s.avail_in);
1110 s.next_out = out = xrealloc(out, s.avail_out);
1111 while (deflate(&s, Z_FINISH) == Z_OK)
1118 e->pack_id = pack_id;
1119 e->offset = pack_size;
1121 object_count_by_type[type]++;
1124 unsigned long ofs = e->offset - last->offset;
1125 unsigned pos = sizeof(hdr) - 1;
1127 delta_count_by_type[type]++;
1128 e->depth = last->depth + 1;
1130 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1131 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1132 pack_size += hdrlen;
1134 hdr[pos] = ofs & 127;
1136 hdr[--pos] = 128 | (--ofs & 127);
1137 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
1138 pack_size += sizeof(hdr) - pos;
1141 hdrlen = encode_header(type, dat->len, hdr);
1142 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1143 pack_size += hdrlen;
1146 write_or_die(pack_data->pack_fd, out, s.total_out);
1147 pack_size += s.total_out;
1152 if (last->no_swap) {
1155 strbuf_swap(&last->data, dat);
1157 last->offset = e->offset;
1158 last->depth = e->depth;
1163 /* All calls must be guarded by find_object() or find_mark() to
1164 * ensure the 'struct object_entry' passed was written by this
1165 * process instance. We unpack the entry by the offset, avoiding
1166 * the need for the corresponding .idx file. This unpacking rule
1167 * works because we only use OBJ_REF_DELTA within the packfiles
1168 * created by fast-import.
1170 * oe must not be NULL. Such an oe usually comes from giving
1171 * an unknown SHA-1 to find_object() or an undefined mark to
1172 * find_mark(). Callers must test for this condition and use
1173 * the standard read_sha1_file() when it happens.
1175 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1176 * find_mark(), where the mark was reloaded from an existing marks
1177 * file and is referencing an object that this fast-import process
1178 * instance did not write out to a packfile. Callers must test for
1179 * this condition and use read_sha1_file() instead.
1181 static void *gfi_unpack_entry(
1182 struct object_entry *oe,
1183 unsigned long *sizep)
1185 enum object_type type;
1186 struct packed_git *p = all_packs[oe->pack_id];
1187 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1188 /* The object is stored in the packfile we are writing to
1189 * and we have modified it since the last time we scanned
1190 * back to read a previously written object. If an old
1191 * window covered [p->pack_size, p->pack_size + 20) its
1192 * data is stale and is not valid. Closing all windows
1193 * and updating the packfile length ensures we can read
1194 * the newly written data.
1196 close_pack_windows(p);
1198 /* We have to offer 20 bytes additional on the end of
1199 * the packfile as the core unpacker code assumes the
1200 * footer is present at the file end and must promise
1201 * at least 20 bytes within any window it maps. But
1202 * we don't actually create the footer here.
1204 p->pack_size = pack_size + 20;
1206 return unpack_entry(p, oe->offset, &type, sizep);
1209 static const char *get_mode(const char *str, uint16_t *modep)
1214 while ((c = *str++) != ' ') {
1215 if (c < '0' || c > '7')
1217 mode = (mode << 3) + (c - '0');
1223 static void load_tree(struct tree_entry *root)
1225 unsigned char* sha1 = root->versions[1].sha1;
1226 struct object_entry *myoe;
1227 struct tree_content *t;
1232 root->tree = t = new_tree_content(8);
1233 if (is_null_sha1(sha1))
1236 myoe = find_object(sha1);
1237 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1238 if (myoe->type != OBJ_TREE)
1239 die("Not a tree: %s", sha1_to_hex(sha1));
1240 t->delta_depth = myoe->depth;
1241 buf = gfi_unpack_entry(myoe, &size);
1243 die("Can't load tree %s", sha1_to_hex(sha1));
1245 enum object_type type;
1246 buf = read_sha1_file(sha1, &type, &size);
1247 if (!buf || type != OBJ_TREE)
1248 die("Can't load tree %s", sha1_to_hex(sha1));
1252 while (c != (buf + size)) {
1253 struct tree_entry *e = new_tree_entry();
1255 if (t->entry_count == t->entry_capacity)
1256 root->tree = t = grow_tree_content(t, t->entry_count);
1257 t->entries[t->entry_count++] = e;
1260 c = get_mode(c, &e->versions[1].mode);
1262 die("Corrupt mode in %s", sha1_to_hex(sha1));
1263 e->versions[0].mode = e->versions[1].mode;
1264 e->name = to_atom(c, strlen(c));
1265 c += e->name->str_len + 1;
1266 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1267 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1273 static int tecmp0 (const void *_a, const void *_b)
1275 struct tree_entry *a = *((struct tree_entry**)_a);
1276 struct tree_entry *b = *((struct tree_entry**)_b);
1277 return base_name_compare(
1278 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1279 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1282 static int tecmp1 (const void *_a, const void *_b)
1284 struct tree_entry *a = *((struct tree_entry**)_a);
1285 struct tree_entry *b = *((struct tree_entry**)_b);
1286 return base_name_compare(
1287 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1288 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1291 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1297 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1299 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1301 for (i = 0; i < t->entry_count; i++) {
1302 if (t->entries[i]->versions[v].mode)
1303 maxlen += t->entries[i]->name->str_len + 34;
1307 strbuf_grow(b, maxlen);
1308 for (i = 0; i < t->entry_count; i++) {
1309 struct tree_entry *e = t->entries[i];
1310 if (!e->versions[v].mode)
1312 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1313 e->name->str_dat, '\0');
1314 strbuf_add(b, e->versions[v].sha1, 20);
1318 static void store_tree(struct tree_entry *root)
1320 struct tree_content *t = root->tree;
1321 unsigned int i, j, del;
1322 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1323 struct object_entry *le;
1325 if (!is_null_sha1(root->versions[1].sha1))
1328 for (i = 0; i < t->entry_count; i++) {
1329 if (t->entries[i]->tree)
1330 store_tree(t->entries[i]);
1333 le = find_object(root->versions[0].sha1);
1334 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1335 mktree(t, 0, &old_tree);
1337 lo.offset = le->offset;
1338 lo.depth = t->delta_depth;
1341 mktree(t, 1, &new_tree);
1342 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1344 t->delta_depth = lo.depth;
1345 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1346 struct tree_entry *e = t->entries[i];
1347 if (e->versions[1].mode) {
1348 e->versions[0].mode = e->versions[1].mode;
1349 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1350 t->entries[j++] = e;
1352 release_tree_entry(e);
1356 t->entry_count -= del;
1359 static int tree_content_set(
1360 struct tree_entry *root,
1362 const unsigned char *sha1,
1363 const uint16_t mode,
1364 struct tree_content *subtree)
1366 struct tree_content *t = root->tree;
1369 struct tree_entry *e;
1371 slash1 = strchr(p, '/');
1377 die("Empty path component found in input");
1378 if (!slash1 && !S_ISDIR(mode) && subtree)
1379 die("Non-directories cannot have subtrees");
1381 for (i = 0; i < t->entry_count; i++) {
1383 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1386 && e->versions[1].mode == mode
1387 && !hashcmp(e->versions[1].sha1, sha1))
1389 e->versions[1].mode = mode;
1390 hashcpy(e->versions[1].sha1, sha1);
1392 release_tree_content_recursive(e->tree);
1394 hashclr(root->versions[1].sha1);
1397 if (!S_ISDIR(e->versions[1].mode)) {
1398 e->tree = new_tree_content(8);
1399 e->versions[1].mode = S_IFDIR;
1403 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1404 hashclr(root->versions[1].sha1);
1411 if (t->entry_count == t->entry_capacity)
1412 root->tree = t = grow_tree_content(t, t->entry_count);
1413 e = new_tree_entry();
1414 e->name = to_atom(p, n);
1415 e->versions[0].mode = 0;
1416 hashclr(e->versions[0].sha1);
1417 t->entries[t->entry_count++] = e;
1419 e->tree = new_tree_content(8);
1420 e->versions[1].mode = S_IFDIR;
1421 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1424 e->versions[1].mode = mode;
1425 hashcpy(e->versions[1].sha1, sha1);
1427 hashclr(root->versions[1].sha1);
1431 static int tree_content_remove(
1432 struct tree_entry *root,
1434 struct tree_entry *backup_leaf)
1436 struct tree_content *t = root->tree;
1439 struct tree_entry *e;
1441 slash1 = strchr(p, '/');
1447 for (i = 0; i < t->entry_count; i++) {
1449 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1450 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1454 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1455 for (n = 0; n < e->tree->entry_count; n++) {
1456 if (e->tree->entries[n]->versions[1].mode) {
1457 hashclr(root->versions[1].sha1);
1471 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1473 release_tree_content_recursive(e->tree);
1475 e->versions[1].mode = 0;
1476 hashclr(e->versions[1].sha1);
1477 hashclr(root->versions[1].sha1);
1481 static int tree_content_get(
1482 struct tree_entry *root,
1484 struct tree_entry *leaf)
1486 struct tree_content *t = root->tree;
1489 struct tree_entry *e;
1491 slash1 = strchr(p, '/');
1497 for (i = 0; i < t->entry_count; i++) {
1499 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1501 memcpy(leaf, e, sizeof(*leaf));
1502 if (e->tree && is_null_sha1(e->versions[1].sha1))
1503 leaf->tree = dup_tree_content(e->tree);
1508 if (!S_ISDIR(e->versions[1].mode))
1512 return tree_content_get(e, slash1 + 1, leaf);
1518 static int update_branch(struct branch *b)
1520 static const char *msg = "fast-import";
1521 struct ref_lock *lock;
1522 unsigned char old_sha1[20];
1524 if (is_null_sha1(b->sha1))
1526 if (read_ref(b->name, old_sha1))
1528 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1530 return error("Unable to lock %s", b->name);
1531 if (!force_update && !is_null_sha1(old_sha1)) {
1532 struct commit *old_cmit, *new_cmit;
1534 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1535 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1536 if (!old_cmit || !new_cmit) {
1538 return error("Branch %s is missing commits.", b->name);
1541 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1543 warning("Not updating %s"
1544 " (new tip %s does not contain %s)",
1545 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1549 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1550 return error("Unable to update %s", b->name);
1554 static void dump_branches(void)
1559 for (i = 0; i < branch_table_sz; i++) {
1560 for (b = branch_table[i]; b; b = b->table_next_branch)
1561 failure |= update_branch(b);
1565 static void dump_tags(void)
1567 static const char *msg = "fast-import";
1569 struct ref_lock *lock;
1570 char ref_name[PATH_MAX];
1572 for (t = first_tag; t; t = t->next_tag) {
1573 sprintf(ref_name, "tags/%s", t->name);
1574 lock = lock_ref_sha1(ref_name, NULL);
1575 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1576 failure |= error("Unable to update %s", ref_name);
1580 static void dump_marks_helper(FILE *f,
1586 for (k = 0; k < 1024; k++) {
1587 if (m->data.sets[k])
1588 dump_marks_helper(f, (base + k) << m->shift,
1592 for (k = 0; k < 1024; k++) {
1593 if (m->data.marked[k])
1594 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1595 sha1_to_hex(m->data.marked[k]->sha1));
1600 static void dump_marks(void)
1602 static struct lock_file mark_lock;
1609 mark_fd = hold_lock_file_for_update(&mark_lock, mark_file, 0);
1611 failure |= error("Unable to write marks file %s: %s",
1612 mark_file, strerror(errno));
1616 f = fdopen(mark_fd, "w");
1618 int saved_errno = errno;
1619 rollback_lock_file(&mark_lock);
1620 failure |= error("Unable to write marks file %s: %s",
1621 mark_file, strerror(saved_errno));
1626 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1627 * Assign -1 to the lock file descriptor so that commit_lock_file()
1628 * won't try to close() it.
1632 dump_marks_helper(f, 0, marks);
1633 if (ferror(f) || fclose(f)) {
1634 int saved_errno = errno;
1635 rollback_lock_file(&mark_lock);
1636 failure |= error("Unable to write marks file %s: %s",
1637 mark_file, strerror(saved_errno));
1641 if (commit_lock_file(&mark_lock)) {
1642 int saved_errno = errno;
1643 rollback_lock_file(&mark_lock);
1644 failure |= error("Unable to commit marks file %s: %s",
1645 mark_file, strerror(saved_errno));
1650 static int read_next_command(void)
1652 static int stdin_eof = 0;
1655 unread_command_buf = 0;
1660 if (unread_command_buf) {
1661 unread_command_buf = 0;
1663 struct recent_command *rc;
1665 strbuf_detach(&command_buf, NULL);
1666 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1675 cmd_hist.next = rc->next;
1676 cmd_hist.next->prev = &cmd_hist;
1680 rc->buf = command_buf.buf;
1681 rc->prev = cmd_tail;
1682 rc->next = cmd_hist.prev;
1683 rc->prev->next = rc;
1686 } while (command_buf.buf[0] == '#');
1691 static void skip_optional_lf(void)
1693 int term_char = fgetc(stdin);
1694 if (term_char != '\n' && term_char != EOF)
1695 ungetc(term_char, stdin);
1698 static void parse_mark(void)
1700 if (!prefixcmp(command_buf.buf, "mark :")) {
1701 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1702 read_next_command();
1708 static void parse_data(struct strbuf *sb)
1712 if (prefixcmp(command_buf.buf, "data "))
1713 die("Expected 'data n' command, found: %s", command_buf.buf);
1715 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1716 char *term = xstrdup(command_buf.buf + 5 + 2);
1717 size_t term_len = command_buf.len - 5 - 2;
1719 strbuf_detach(&command_buf, NULL);
1721 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1722 die("EOF in data (terminator '%s' not found)", term);
1723 if (term_len == command_buf.len
1724 && !strcmp(term, command_buf.buf))
1726 strbuf_addbuf(sb, &command_buf);
1727 strbuf_addch(sb, '\n');
1732 size_t n = 0, length;
1734 length = strtoul(command_buf.buf + 5, NULL, 10);
1736 while (n < length) {
1737 size_t s = strbuf_fread(sb, length - n, stdin);
1738 if (!s && feof(stdin))
1739 die("EOF in data (%lu bytes remaining)",
1740 (unsigned long)(length - n));
1748 static int validate_raw_date(const char *src, char *result, int maxlen)
1750 const char *orig_src = src;
1756 date = strtoul(src, &endp, 10);
1757 if (errno || endp == src || *endp != ' ')
1761 if (*src != '-' && *src != '+')
1765 date = strtoul(src + 1, &endp, 10);
1766 if (errno || endp == src || *endp || (endp - orig_src) >= maxlen)
1769 strcpy(result, orig_src);
1773 static char *parse_ident(const char *buf)
1779 gt = strrchr(buf, '>');
1781 die("Missing > in ident string: %s", buf);
1784 die("Missing space after > in ident string: %s", buf);
1786 name_len = gt - buf;
1787 ident = xmalloc(name_len + 24);
1788 strncpy(ident, buf, name_len);
1792 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1793 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1795 case WHENSPEC_RFC2822:
1796 if (parse_date(gt, ident + name_len, 24) < 0)
1797 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1800 if (strcmp("now", gt))
1801 die("Date in ident must be 'now': %s", buf);
1802 datestamp(ident + name_len, 24);
1809 static void parse_new_blob(void)
1811 static struct strbuf buf = STRBUF_INIT;
1813 read_next_command();
1816 store_object(OBJ_BLOB, &buf, &last_blob, NULL, next_mark);
1819 static void unload_one_branch(void)
1821 while (cur_active_branches
1822 && cur_active_branches >= max_active_branches) {
1823 uintmax_t min_commit = ULONG_MAX;
1824 struct branch *e, *l = NULL, *p = NULL;
1826 for (e = active_branches; e; e = e->active_next_branch) {
1827 if (e->last_commit < min_commit) {
1829 min_commit = e->last_commit;
1835 e = p->active_next_branch;
1836 p->active_next_branch = e->active_next_branch;
1838 e = active_branches;
1839 active_branches = e->active_next_branch;
1842 e->active_next_branch = NULL;
1843 if (e->branch_tree.tree) {
1844 release_tree_content_recursive(e->branch_tree.tree);
1845 e->branch_tree.tree = NULL;
1847 cur_active_branches--;
1851 static void load_branch(struct branch *b)
1853 load_tree(&b->branch_tree);
1856 b->active_next_branch = active_branches;
1857 active_branches = b;
1858 cur_active_branches++;
1859 branch_load_count++;
1863 static void file_change_m(struct branch *b)
1865 const char *p = command_buf.buf + 2;
1866 static struct strbuf uq = STRBUF_INIT;
1868 struct object_entry *oe = oe;
1869 unsigned char sha1[20];
1870 uint16_t mode, inline_data = 0;
1872 p = get_mode(p, &mode);
1874 die("Corrupt mode: %s", command_buf.buf);
1879 case S_IFREG | 0644:
1880 case S_IFREG | 0755:
1886 die("Corrupt mode: %s", command_buf.buf);
1891 oe = find_mark(strtoumax(p + 1, &x, 10));
1892 hashcpy(sha1, oe->sha1);
1894 } else if (!prefixcmp(p, "inline")) {
1898 if (get_sha1_hex(p, sha1))
1899 die("Invalid SHA1: %s", command_buf.buf);
1900 oe = find_object(sha1);
1904 die("Missing space after SHA1: %s", command_buf.buf);
1907 if (!unquote_c_style(&uq, p, &endp)) {
1909 die("Garbage after path in: %s", command_buf.buf);
1913 if (S_ISGITLINK(mode)) {
1915 die("Git links cannot be specified 'inline': %s",
1918 if (oe->type != OBJ_COMMIT)
1919 die("Not a commit (actually a %s): %s",
1920 typename(oe->type), command_buf.buf);
1923 * Accept the sha1 without checking; it expected to be in
1924 * another repository.
1926 } else if (inline_data) {
1927 static struct strbuf buf = STRBUF_INIT;
1930 strbuf_addstr(&uq, p);
1933 read_next_command();
1935 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
1937 if (oe->type != OBJ_BLOB)
1938 die("Not a blob (actually a %s): %s",
1939 typename(oe->type), command_buf.buf);
1941 enum object_type type = sha1_object_info(sha1, NULL);
1943 die("Blob not found: %s", command_buf.buf);
1944 if (type != OBJ_BLOB)
1945 die("Not a blob (actually a %s): %s",
1946 typename(type), command_buf.buf);
1949 tree_content_set(&b->branch_tree, p, sha1, mode, NULL);
1952 static void file_change_d(struct branch *b)
1954 const char *p = command_buf.buf + 2;
1955 static struct strbuf uq = STRBUF_INIT;
1959 if (!unquote_c_style(&uq, p, &endp)) {
1961 die("Garbage after path in: %s", command_buf.buf);
1964 tree_content_remove(&b->branch_tree, p, NULL);
1967 static void file_change_cr(struct branch *b, int rename)
1970 static struct strbuf s_uq = STRBUF_INIT;
1971 static struct strbuf d_uq = STRBUF_INIT;
1973 struct tree_entry leaf;
1975 s = command_buf.buf + 2;
1976 strbuf_reset(&s_uq);
1977 if (!unquote_c_style(&s_uq, s, &endp)) {
1979 die("Missing space after source: %s", command_buf.buf);
1981 endp = strchr(s, ' ');
1983 die("Missing space after source: %s", command_buf.buf);
1984 strbuf_add(&s_uq, s, endp - s);
1990 die("Missing dest: %s", command_buf.buf);
1993 strbuf_reset(&d_uq);
1994 if (!unquote_c_style(&d_uq, d, &endp)) {
1996 die("Garbage after dest in: %s", command_buf.buf);
2000 memset(&leaf, 0, sizeof(leaf));
2002 tree_content_remove(&b->branch_tree, s, &leaf);
2004 tree_content_get(&b->branch_tree, s, &leaf);
2005 if (!leaf.versions[1].mode)
2006 die("Path %s not in branch", s);
2007 tree_content_set(&b->branch_tree, d,
2008 leaf.versions[1].sha1,
2009 leaf.versions[1].mode,
2013 static void file_change_deleteall(struct branch *b)
2015 release_tree_content_recursive(b->branch_tree.tree);
2016 hashclr(b->branch_tree.versions[0].sha1);
2017 hashclr(b->branch_tree.versions[1].sha1);
2018 load_tree(&b->branch_tree);
2021 static void parse_from_commit(struct branch *b, char *buf, unsigned long size)
2023 if (!buf || size < 46)
2024 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
2025 if (memcmp("tree ", buf, 5)
2026 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
2027 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
2028 hashcpy(b->branch_tree.versions[0].sha1,
2029 b->branch_tree.versions[1].sha1);
2032 static void parse_from_existing(struct branch *b)
2034 if (is_null_sha1(b->sha1)) {
2035 hashclr(b->branch_tree.versions[0].sha1);
2036 hashclr(b->branch_tree.versions[1].sha1);
2041 buf = read_object_with_reference(b->sha1,
2042 commit_type, &size, b->sha1);
2043 parse_from_commit(b, buf, size);
2048 static int parse_from(struct branch *b)
2053 if (prefixcmp(command_buf.buf, "from "))
2056 if (b->branch_tree.tree) {
2057 release_tree_content_recursive(b->branch_tree.tree);
2058 b->branch_tree.tree = NULL;
2061 from = strchr(command_buf.buf, ' ') + 1;
2062 s = lookup_branch(from);
2064 die("Can't create a branch from itself: %s", b->name);
2066 unsigned char *t = s->branch_tree.versions[1].sha1;
2067 hashcpy(b->sha1, s->sha1);
2068 hashcpy(b->branch_tree.versions[0].sha1, t);
2069 hashcpy(b->branch_tree.versions[1].sha1, t);
2070 } else if (*from == ':') {
2071 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2072 struct object_entry *oe = find_mark(idnum);
2073 if (oe->type != OBJ_COMMIT)
2074 die("Mark :%" PRIuMAX " not a commit", idnum);
2075 hashcpy(b->sha1, oe->sha1);
2076 if (oe->pack_id != MAX_PACK_ID) {
2078 char *buf = gfi_unpack_entry(oe, &size);
2079 parse_from_commit(b, buf, size);
2082 parse_from_existing(b);
2083 } else if (!get_sha1(from, b->sha1))
2084 parse_from_existing(b);
2086 die("Invalid ref name or SHA1 expression: %s", from);
2088 read_next_command();
2092 static struct hash_list *parse_merge(unsigned int *count)
2094 struct hash_list *list = NULL, *n, *e = e;
2099 while (!prefixcmp(command_buf.buf, "merge ")) {
2100 from = strchr(command_buf.buf, ' ') + 1;
2101 n = xmalloc(sizeof(*n));
2102 s = lookup_branch(from);
2104 hashcpy(n->sha1, s->sha1);
2105 else if (*from == ':') {
2106 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2107 struct object_entry *oe = find_mark(idnum);
2108 if (oe->type != OBJ_COMMIT)
2109 die("Mark :%" PRIuMAX " not a commit", idnum);
2110 hashcpy(n->sha1, oe->sha1);
2111 } else if (!get_sha1(from, n->sha1)) {
2113 char *buf = read_object_with_reference(n->sha1,
2114 commit_type, &size, n->sha1);
2115 if (!buf || size < 46)
2116 die("Not a valid commit: %s", from);
2119 die("Invalid ref name or SHA1 expression: %s", from);
2128 read_next_command();
2133 static void parse_new_commit(void)
2135 static struct strbuf msg = STRBUF_INIT;
2138 char *author = NULL;
2139 char *committer = NULL;
2140 struct hash_list *merge_list = NULL;
2141 unsigned int merge_count;
2143 /* Obtain the branch name from the rest of our command */
2144 sp = strchr(command_buf.buf, ' ') + 1;
2145 b = lookup_branch(sp);
2149 read_next_command();
2151 if (!prefixcmp(command_buf.buf, "author ")) {
2152 author = parse_ident(command_buf.buf + 7);
2153 read_next_command();
2155 if (!prefixcmp(command_buf.buf, "committer ")) {
2156 committer = parse_ident(command_buf.buf + 10);
2157 read_next_command();
2160 die("Expected committer but didn't get one");
2162 read_next_command();
2164 merge_list = parse_merge(&merge_count);
2166 /* ensure the branch is active/loaded */
2167 if (!b->branch_tree.tree || !max_active_branches) {
2168 unload_one_branch();
2173 while (command_buf.len > 0) {
2174 if (!prefixcmp(command_buf.buf, "M "))
2176 else if (!prefixcmp(command_buf.buf, "D "))
2178 else if (!prefixcmp(command_buf.buf, "R "))
2179 file_change_cr(b, 1);
2180 else if (!prefixcmp(command_buf.buf, "C "))
2181 file_change_cr(b, 0);
2182 else if (!strcmp("deleteall", command_buf.buf))
2183 file_change_deleteall(b);
2185 unread_command_buf = 1;
2188 if (read_next_command() == EOF)
2192 /* build the tree and the commit */
2193 store_tree(&b->branch_tree);
2194 hashcpy(b->branch_tree.versions[0].sha1,
2195 b->branch_tree.versions[1].sha1);
2197 strbuf_reset(&new_data);
2198 strbuf_addf(&new_data, "tree %s\n",
2199 sha1_to_hex(b->branch_tree.versions[1].sha1));
2200 if (!is_null_sha1(b->sha1))
2201 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2202 while (merge_list) {
2203 struct hash_list *next = merge_list->next;
2204 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2208 strbuf_addf(&new_data,
2212 author ? author : committer, committer);
2213 strbuf_addbuf(&new_data, &msg);
2217 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2218 b->pack_id = pack_id;
2219 b->last_commit = object_count_by_type[OBJ_COMMIT];
2222 static void parse_new_tag(void)
2224 static struct strbuf msg = STRBUF_INIT;
2230 uintmax_t from_mark = 0;
2231 unsigned char sha1[20];
2233 /* Obtain the new tag name from the rest of our command */
2234 sp = strchr(command_buf.buf, ' ') + 1;
2235 t = pool_alloc(sizeof(struct tag));
2237 t->name = pool_strdup(sp);
2239 last_tag->next_tag = t;
2243 read_next_command();
2246 if (prefixcmp(command_buf.buf, "from "))
2247 die("Expected from command, got %s", command_buf.buf);
2248 from = strchr(command_buf.buf, ' ') + 1;
2249 s = lookup_branch(from);
2251 hashcpy(sha1, s->sha1);
2252 } else if (*from == ':') {
2253 struct object_entry *oe;
2254 from_mark = strtoumax(from + 1, NULL, 10);
2255 oe = find_mark(from_mark);
2256 if (oe->type != OBJ_COMMIT)
2257 die("Mark :%" PRIuMAX " not a commit", from_mark);
2258 hashcpy(sha1, oe->sha1);
2259 } else if (!get_sha1(from, sha1)) {
2263 buf = read_object_with_reference(sha1,
2264 commit_type, &size, sha1);
2265 if (!buf || size < 46)
2266 die("Not a valid commit: %s", from);
2269 die("Invalid ref name or SHA1 expression: %s", from);
2270 read_next_command();
2273 if (!prefixcmp(command_buf.buf, "tagger ")) {
2274 tagger = parse_ident(command_buf.buf + 7);
2275 read_next_command();
2279 /* tag payload/message */
2282 /* build the tag object */
2283 strbuf_reset(&new_data);
2285 strbuf_addf(&new_data,
2289 sha1_to_hex(sha1), commit_type, t->name);
2291 strbuf_addf(&new_data,
2292 "tagger %s\n", tagger);
2293 strbuf_addch(&new_data, '\n');
2294 strbuf_addbuf(&new_data, &msg);
2297 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2298 t->pack_id = MAX_PACK_ID;
2300 t->pack_id = pack_id;
2303 static void parse_reset_branch(void)
2308 /* Obtain the branch name from the rest of our command */
2309 sp = strchr(command_buf.buf, ' ') + 1;
2310 b = lookup_branch(sp);
2313 hashclr(b->branch_tree.versions[0].sha1);
2314 hashclr(b->branch_tree.versions[1].sha1);
2315 if (b->branch_tree.tree) {
2316 release_tree_content_recursive(b->branch_tree.tree);
2317 b->branch_tree.tree = NULL;
2322 read_next_command();
2324 if (command_buf.len > 0)
2325 unread_command_buf = 1;
2328 static void parse_checkpoint(void)
2339 static void parse_progress(void)
2341 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2342 fputc('\n', stdout);
2347 static void import_marks(const char *input_file)
2350 FILE *f = fopen(input_file, "r");
2352 die("cannot read %s: %s", input_file, strerror(errno));
2353 while (fgets(line, sizeof(line), f)) {
2356 unsigned char sha1[20];
2357 struct object_entry *e;
2359 end = strchr(line, '\n');
2360 if (line[0] != ':' || !end)
2361 die("corrupt mark line: %s", line);
2363 mark = strtoumax(line + 1, &end, 10);
2364 if (!mark || end == line + 1
2365 || *end != ' ' || get_sha1(end + 1, sha1))
2366 die("corrupt mark line: %s", line);
2367 e = find_object(sha1);
2369 enum object_type type = sha1_object_info(sha1, NULL);
2371 die("object not found: %s", sha1_to_hex(sha1));
2372 e = insert_object(sha1);
2374 e->pack_id = MAX_PACK_ID;
2375 e->offset = 1; /* just not zero! */
2377 insert_mark(mark, e);
2382 static int git_pack_config(const char *k, const char *v, void *cb)
2384 if (!strcmp(k, "pack.depth")) {
2385 max_depth = git_config_int(k, v);
2386 if (max_depth > MAX_DEPTH)
2387 max_depth = MAX_DEPTH;
2390 if (!strcmp(k, "pack.compression")) {
2391 int level = git_config_int(k, v);
2393 level = Z_DEFAULT_COMPRESSION;
2394 else if (level < 0 || level > Z_BEST_COMPRESSION)
2395 die("bad pack compression level %d", level);
2396 pack_compression_level = level;
2397 pack_compression_seen = 1;
2400 return git_default_config(k, v, cb);
2403 static const char fast_import_usage[] =
2404 "git fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2406 int main(int argc, const char **argv)
2408 unsigned int i, show_stats = 1;
2410 git_extract_argv0_path(argv[0]);
2412 setup_git_directory();
2413 git_config(git_pack_config, NULL);
2414 if (!pack_compression_seen && core_compression_seen)
2415 pack_compression_level = core_compression_level;
2417 alloc_objects(object_entry_alloc);
2418 strbuf_init(&command_buf, 0);
2419 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2420 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2421 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2422 marks = pool_calloc(1, sizeof(struct mark_set));
2424 for (i = 1; i < argc; i++) {
2425 const char *a = argv[i];
2427 if (*a != '-' || !strcmp(a, "--"))
2429 else if (!prefixcmp(a, "--date-format=")) {
2430 const char *fmt = a + 14;
2431 if (!strcmp(fmt, "raw"))
2432 whenspec = WHENSPEC_RAW;
2433 else if (!strcmp(fmt, "rfc2822"))
2434 whenspec = WHENSPEC_RFC2822;
2435 else if (!strcmp(fmt, "now"))
2436 whenspec = WHENSPEC_NOW;
2438 die("unknown --date-format argument %s", fmt);
2440 else if (!prefixcmp(a, "--max-pack-size="))
2441 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
2442 else if (!prefixcmp(a, "--depth=")) {
2443 max_depth = strtoul(a + 8, NULL, 0);
2444 if (max_depth > MAX_DEPTH)
2445 die("--depth cannot exceed %u", MAX_DEPTH);
2447 else if (!prefixcmp(a, "--active-branches="))
2448 max_active_branches = strtoul(a + 18, NULL, 0);
2449 else if (!prefixcmp(a, "--import-marks="))
2450 import_marks(a + 15);
2451 else if (!prefixcmp(a, "--export-marks="))
2453 else if (!prefixcmp(a, "--export-pack-edges=")) {
2456 pack_edges = fopen(a + 20, "a");
2458 die("Cannot open %s: %s", a + 20, strerror(errno));
2459 } else if (!strcmp(a, "--force"))
2461 else if (!strcmp(a, "--quiet"))
2463 else if (!strcmp(a, "--stats"))
2466 die("unknown option %s", a);
2469 usage(fast_import_usage);
2471 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2472 for (i = 0; i < (cmd_save - 1); i++)
2473 rc_free[i].next = &rc_free[i + 1];
2474 rc_free[cmd_save - 1].next = NULL;
2476 prepare_packed_git();
2478 set_die_routine(die_nicely);
2479 while (read_next_command() != EOF) {
2480 if (!strcmp("blob", command_buf.buf))
2482 else if (!prefixcmp(command_buf.buf, "commit "))
2484 else if (!prefixcmp(command_buf.buf, "tag "))
2486 else if (!prefixcmp(command_buf.buf, "reset "))
2487 parse_reset_branch();
2488 else if (!strcmp("checkpoint", command_buf.buf))
2490 else if (!prefixcmp(command_buf.buf, "progress "))
2493 die("Unsupported command: %s", command_buf.buf);
2506 uintmax_t total_count = 0, duplicate_count = 0;
2507 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2508 total_count += object_count_by_type[i];
2509 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2510 duplicate_count += duplicate_count_by_type[i];
2512 fprintf(stderr, "%s statistics:\n", argv[0]);
2513 fprintf(stderr, "---------------------------------------------------------------------\n");
2514 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2515 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2516 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]);
2517 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]);
2518 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]);
2519 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]);
2520 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2521 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2522 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2523 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2524 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2525 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2526 fprintf(stderr, "---------------------------------------------------------------------\n");
2528 fprintf(stderr, "---------------------------------------------------------------------\n");
2529 fprintf(stderr, "\n");
2532 return failure ? 1 : 0;