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 write_crash_report(const char *err)
377 char *loc = git_path("fast_import_crash_%d", getpid());
378 FILE *rpt = fopen(loc, "w");
381 struct recent_command *rc;
384 error("can't write crash report %s: %s", loc, strerror(errno));
388 fprintf(stderr, "fast-import: dumping crash report to %s\n", loc);
390 fprintf(rpt, "fast-import crash report:\n");
391 fprintf(rpt, " fast-import process: %d\n", getpid());
392 fprintf(rpt, " parent process : %d\n", getppid());
393 fprintf(rpt, " at %s\n", show_date(time(NULL), 0, DATE_LOCAL));
396 fputs("fatal: ", rpt);
401 fputs("Most Recent Commands Before Crash\n", rpt);
402 fputs("---------------------------------\n", rpt);
403 for (rc = cmd_hist.next; rc != &cmd_hist; rc = rc->next) {
404 if (rc->next == &cmd_hist)
413 fputs("Active Branch LRU\n", rpt);
414 fputs("-----------------\n", rpt);
415 fprintf(rpt, " active_branches = %lu cur, %lu max\n",
417 max_active_branches);
419 fputs(" pos clock name\n", rpt);
420 fputs(" ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n", rpt);
421 for (b = active_branches, lu = 0; b; b = b->active_next_branch)
422 fprintf(rpt, " %2lu) %6" PRIuMAX" %s\n",
423 ++lu, b->last_commit, b->name);
426 fputs("Inactive Branches\n", rpt);
427 fputs("-----------------\n", rpt);
428 for (lu = 0; lu < branch_table_sz; lu++) {
429 for (b = branch_table[lu]; b; b = b->table_next_branch)
430 write_branch_report(rpt, b);
434 fputs("-------------------\n", rpt);
435 fputs("END OF CRASH REPORT\n", rpt);
439 static NORETURN void die_nicely(const char *err, va_list params)
442 char message[2 * PATH_MAX];
444 vsnprintf(message, sizeof(message), err, params);
445 fputs("fatal: ", stderr);
446 fputs(message, stderr);
451 write_crash_report(message);
456 static void alloc_objects(unsigned int cnt)
458 struct object_entry_pool *b;
460 b = xmalloc(sizeof(struct object_entry_pool)
461 + cnt * sizeof(struct object_entry));
462 b->next_pool = blocks;
463 b->next_free = b->entries;
464 b->end = b->entries + cnt;
469 static struct object_entry *new_object(unsigned char *sha1)
471 struct object_entry *e;
473 if (blocks->next_free == blocks->end)
474 alloc_objects(object_entry_alloc);
476 e = blocks->next_free++;
477 hashcpy(e->sha1, sha1);
481 static struct object_entry *find_object(unsigned char *sha1)
483 unsigned int h = sha1[0] << 8 | sha1[1];
484 struct object_entry *e;
485 for (e = object_table[h]; e; e = e->next)
486 if (!hashcmp(sha1, e->sha1))
491 static struct object_entry *insert_object(unsigned char *sha1)
493 unsigned int h = sha1[0] << 8 | sha1[1];
494 struct object_entry *e = object_table[h];
495 struct object_entry *p = NULL;
498 if (!hashcmp(sha1, e->sha1))
504 e = new_object(sha1);
514 static unsigned int hc_str(const char *s, size_t len)
522 static void *pool_alloc(size_t len)
527 for (p = mem_pool; p; p = p->next_pool)
528 if ((p->end - p->next_free >= len))
532 if (len >= (mem_pool_alloc/2)) {
536 total_allocd += sizeof(struct mem_pool) + mem_pool_alloc;
537 p = xmalloc(sizeof(struct mem_pool) + mem_pool_alloc);
538 p->next_pool = mem_pool;
539 p->next_free = (char *) p->space;
540 p->end = p->next_free + mem_pool_alloc;
545 /* round out to a 'uintmax_t' alignment */
546 if (len & (sizeof(uintmax_t) - 1))
547 len += sizeof(uintmax_t) - (len & (sizeof(uintmax_t) - 1));
552 static void *pool_calloc(size_t count, size_t size)
554 size_t len = count * size;
555 void *r = pool_alloc(len);
560 static char *pool_strdup(const char *s)
562 char *r = pool_alloc(strlen(s) + 1);
567 static void insert_mark(uintmax_t idnum, struct object_entry *oe)
569 struct mark_set *s = marks;
570 while ((idnum >> s->shift) >= 1024) {
571 s = pool_calloc(1, sizeof(struct mark_set));
572 s->shift = marks->shift + 10;
573 s->data.sets[0] = marks;
577 uintmax_t i = idnum >> s->shift;
578 idnum -= i << s->shift;
579 if (!s->data.sets[i]) {
580 s->data.sets[i] = pool_calloc(1, sizeof(struct mark_set));
581 s->data.sets[i]->shift = s->shift - 10;
585 if (!s->data.marked[idnum])
587 s->data.marked[idnum] = oe;
590 static struct object_entry *find_mark(uintmax_t idnum)
592 uintmax_t orig_idnum = idnum;
593 struct mark_set *s = marks;
594 struct object_entry *oe = NULL;
595 if ((idnum >> s->shift) < 1024) {
596 while (s && s->shift) {
597 uintmax_t i = idnum >> s->shift;
598 idnum -= i << s->shift;
602 oe = s->data.marked[idnum];
605 die("mark :%" PRIuMAX " not declared", orig_idnum);
609 static struct atom_str *to_atom(const char *s, unsigned short len)
611 unsigned int hc = hc_str(s, len) % atom_table_sz;
614 for (c = atom_table[hc]; c; c = c->next_atom)
615 if (c->str_len == len && !strncmp(s, c->str_dat, len))
618 c = pool_alloc(sizeof(struct atom_str) + len + 1);
620 strncpy(c->str_dat, s, len);
622 c->next_atom = atom_table[hc];
628 static struct branch *lookup_branch(const char *name)
630 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
633 for (b = branch_table[hc]; b; b = b->table_next_branch)
634 if (!strcmp(name, b->name))
639 static struct branch *new_branch(const char *name)
641 unsigned int hc = hc_str(name, strlen(name)) % branch_table_sz;
642 struct branch* b = lookup_branch(name);
645 die("Invalid attempt to create duplicate branch: %s", name);
646 switch (check_ref_format(name)) {
647 case 0: break; /* its valid */
648 case CHECK_REF_FORMAT_ONELEVEL:
649 break; /* valid, but too few '/', allow anyway */
651 die("Branch name doesn't conform to GIT standards: %s", name);
654 b = pool_calloc(1, sizeof(struct branch));
655 b->name = pool_strdup(name);
656 b->table_next_branch = branch_table[hc];
657 b->branch_tree.versions[0].mode = S_IFDIR;
658 b->branch_tree.versions[1].mode = S_IFDIR;
660 b->pack_id = MAX_PACK_ID;
661 branch_table[hc] = b;
666 static unsigned int hc_entries(unsigned int cnt)
668 cnt = cnt & 7 ? (cnt / 8) + 1 : cnt / 8;
669 return cnt < avail_tree_table_sz ? cnt : avail_tree_table_sz - 1;
672 static struct tree_content *new_tree_content(unsigned int cnt)
674 struct avail_tree_content *f, *l = NULL;
675 struct tree_content *t;
676 unsigned int hc = hc_entries(cnt);
678 for (f = avail_tree_table[hc]; f; l = f, f = f->next_avail)
679 if (f->entry_capacity >= cnt)
684 l->next_avail = f->next_avail;
686 avail_tree_table[hc] = f->next_avail;
688 cnt = cnt & 7 ? ((cnt / 8) + 1) * 8 : cnt;
689 f = pool_alloc(sizeof(*t) + sizeof(t->entries[0]) * cnt);
690 f->entry_capacity = cnt;
693 t = (struct tree_content*)f;
699 static void release_tree_entry(struct tree_entry *e);
700 static void release_tree_content(struct tree_content *t)
702 struct avail_tree_content *f = (struct avail_tree_content*)t;
703 unsigned int hc = hc_entries(f->entry_capacity);
704 f->next_avail = avail_tree_table[hc];
705 avail_tree_table[hc] = f;
708 static void release_tree_content_recursive(struct tree_content *t)
711 for (i = 0; i < t->entry_count; i++)
712 release_tree_entry(t->entries[i]);
713 release_tree_content(t);
716 static struct tree_content *grow_tree_content(
717 struct tree_content *t,
720 struct tree_content *r = new_tree_content(t->entry_count + amt);
721 r->entry_count = t->entry_count;
722 r->delta_depth = t->delta_depth;
723 memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
724 release_tree_content(t);
728 static struct tree_entry *new_tree_entry(void)
730 struct tree_entry *e;
732 if (!avail_tree_entry) {
733 unsigned int n = tree_entry_alloc;
734 total_allocd += n * sizeof(struct tree_entry);
735 avail_tree_entry = e = xmalloc(n * sizeof(struct tree_entry));
737 *((void**)e) = e + 1;
743 e = avail_tree_entry;
744 avail_tree_entry = *((void**)e);
748 static void release_tree_entry(struct tree_entry *e)
751 release_tree_content_recursive(e->tree);
752 *((void**)e) = avail_tree_entry;
753 avail_tree_entry = e;
756 static struct tree_content *dup_tree_content(struct tree_content *s)
758 struct tree_content *d;
759 struct tree_entry *a, *b;
764 d = new_tree_content(s->entry_count);
765 for (i = 0; i < s->entry_count; i++) {
767 b = new_tree_entry();
768 memcpy(b, a, sizeof(*a));
769 if (a->tree && is_null_sha1(b->versions[1].sha1))
770 b->tree = dup_tree_content(a->tree);
775 d->entry_count = s->entry_count;
776 d->delta_depth = s->delta_depth;
781 static void start_packfile(void)
783 static char tmpfile[PATH_MAX];
784 struct packed_git *p;
785 struct pack_header hdr;
788 snprintf(tmpfile, sizeof(tmpfile),
789 "%s/tmp_pack_XXXXXX", get_object_directory());
790 pack_fd = xmkstemp(tmpfile);
791 p = xcalloc(1, sizeof(*p) + strlen(tmpfile) + 2);
792 strcpy(p->pack_name, tmpfile);
793 p->pack_fd = pack_fd;
795 hdr.hdr_signature = htonl(PACK_SIGNATURE);
796 hdr.hdr_version = htonl(2);
798 write_or_die(p->pack_fd, &hdr, sizeof(hdr));
801 pack_size = sizeof(hdr);
804 all_packs = xrealloc(all_packs, sizeof(*all_packs) * (pack_id + 1));
805 all_packs[pack_id] = p;
808 static int oecmp (const void *a_, const void *b_)
810 struct object_entry *a = *((struct object_entry**)a_);
811 struct object_entry *b = *((struct object_entry**)b_);
812 return hashcmp(a->sha1, b->sha1);
815 static char *create_index(void)
817 static char tmpfile[PATH_MAX];
820 struct object_entry **idx, **c, **last, *e;
821 struct object_entry_pool *o;
825 /* Build the sorted table of object IDs. */
826 idx = xmalloc(object_count * sizeof(struct object_entry*));
828 for (o = blocks; o; o = o->next_pool)
829 for (e = o->next_free; e-- != o->entries;)
830 if (pack_id == e->pack_id)
832 last = idx + object_count;
834 die("internal consistency error creating the index");
835 qsort(idx, object_count, sizeof(struct object_entry*), oecmp);
837 /* Generate the fan-out array. */
839 for (i = 0; i < 256; i++) {
840 struct object_entry **next = c;;
841 while (next < last) {
842 if ((*next)->sha1[0] != i)
846 array[i] = htonl(next - idx);
850 snprintf(tmpfile, sizeof(tmpfile),
851 "%s/tmp_idx_XXXXXX", get_object_directory());
852 idx_fd = xmkstemp(tmpfile);
853 f = sha1fd(idx_fd, tmpfile);
854 sha1write(f, array, 256 * sizeof(int));
856 for (c = idx; c != last; c++) {
857 uint32_t offset = htonl((*c)->offset);
858 sha1write(f, &offset, 4);
859 sha1write(f, (*c)->sha1, sizeof((*c)->sha1));
860 SHA1_Update(&ctx, (*c)->sha1, 20);
862 sha1write(f, pack_data->sha1, sizeof(pack_data->sha1));
863 sha1close(f, NULL, 1);
865 SHA1_Final(pack_data->sha1, &ctx);
869 static char *keep_pack(char *curr_index_name)
871 static char name[PATH_MAX];
872 static const char *keep_msg = "fast-import";
875 chmod(pack_data->pack_name, 0444);
876 chmod(curr_index_name, 0444);
878 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
879 get_object_directory(), sha1_to_hex(pack_data->sha1));
880 keep_fd = open(name, O_RDWR|O_CREAT|O_EXCL, 0600);
882 die("cannot create keep file");
883 write_or_die(keep_fd, keep_msg, strlen(keep_msg));
885 die("failed to write keep file");
887 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
888 get_object_directory(), sha1_to_hex(pack_data->sha1));
889 if (move_temp_to_file(pack_data->pack_name, name))
890 die("cannot store pack file");
892 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
893 get_object_directory(), sha1_to_hex(pack_data->sha1));
894 if (move_temp_to_file(curr_index_name, name))
895 die("cannot store index file");
899 static void unkeep_all_packs(void)
901 static char name[PATH_MAX];
904 for (k = 0; k < pack_id; k++) {
905 struct packed_git *p = all_packs[k];
906 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
907 get_object_directory(), sha1_to_hex(p->sha1));
912 static void end_packfile(void)
914 struct packed_git *old_p = pack_data, *new_p;
922 close_pack_windows(pack_data);
923 fixup_pack_header_footer(pack_data->pack_fd, pack_data->sha1,
924 pack_data->pack_name, object_count);
925 close(pack_data->pack_fd);
926 idx_name = keep_pack(create_index());
928 /* Register the packfile with core git's machinary. */
929 new_p = add_packed_git(idx_name, strlen(idx_name), 1);
931 die("core git rejected index %s", idx_name);
932 all_packs[pack_id] = new_p;
933 install_packed_git(new_p);
935 /* Print the boundary */
937 fprintf(pack_edges, "%s:", new_p->pack_name);
938 for (i = 0; i < branch_table_sz; i++) {
939 for (b = branch_table[i]; b; b = b->table_next_branch) {
940 if (b->pack_id == pack_id)
941 fprintf(pack_edges, " %s", sha1_to_hex(b->sha1));
944 for (t = first_tag; t; t = t->next_tag) {
945 if (t->pack_id == pack_id)
946 fprintf(pack_edges, " %s", sha1_to_hex(t->sha1));
948 fputc('\n', pack_edges);
955 unlink(old_p->pack_name);
958 /* We can't carry a delta across packfiles. */
959 strbuf_release(&last_blob.data);
960 last_blob.offset = 0;
964 static void cycle_packfile(void)
970 static size_t encode_header(
971 enum object_type type,
978 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
979 die("bad type %d", type);
981 c = (type << 4) | (size & 15);
993 static int store_object(
994 enum object_type type,
996 struct last_object *last,
997 unsigned char *sha1out,
1001 struct object_entry *e;
1002 unsigned char hdr[96];
1003 unsigned char sha1[20];
1004 unsigned long hdrlen, deltalen;
1008 hdrlen = sprintf((char*)hdr,"%s %lu", typename(type),
1009 (unsigned long)dat->len) + 1;
1011 SHA1_Update(&c, hdr, hdrlen);
1012 SHA1_Update(&c, dat->buf, dat->len);
1013 SHA1_Final(sha1, &c);
1015 hashcpy(sha1out, sha1);
1017 e = insert_object(sha1);
1019 insert_mark(mark, e);
1021 duplicate_count_by_type[type]++;
1023 } else if (find_sha1_pack(sha1, packed_git)) {
1025 e->pack_id = MAX_PACK_ID;
1026 e->offset = 1; /* just not zero! */
1027 duplicate_count_by_type[type]++;
1031 if (last && last->data.buf && last->depth < max_depth) {
1032 delta = diff_delta(last->data.buf, last->data.len,
1035 if (delta && deltalen >= dat->len) {
1042 memset(&s, 0, sizeof(s));
1043 deflateInit(&s, pack_compression_level);
1046 s.avail_in = deltalen;
1048 s.next_in = (void *)dat->buf;
1049 s.avail_in = dat->len;
1051 s.avail_out = deflateBound(&s, s.avail_in);
1052 s.next_out = out = xmalloc(s.avail_out);
1053 while (deflate(&s, Z_FINISH) == Z_OK)
1057 /* Determine if we should auto-checkpoint. */
1058 if ((pack_size + 60 + s.total_out) > max_packsize
1059 || (pack_size + 60 + s.total_out) < pack_size) {
1061 /* This new object needs to *not* have the current pack_id. */
1062 e->pack_id = pack_id + 1;
1065 /* We cannot carry a delta into the new pack. */
1070 memset(&s, 0, sizeof(s));
1071 deflateInit(&s, pack_compression_level);
1072 s.next_in = (void *)dat->buf;
1073 s.avail_in = dat->len;
1074 s.avail_out = deflateBound(&s, s.avail_in);
1075 s.next_out = out = xrealloc(out, s.avail_out);
1076 while (deflate(&s, Z_FINISH) == Z_OK)
1083 e->pack_id = pack_id;
1084 e->offset = pack_size;
1086 object_count_by_type[type]++;
1089 unsigned long ofs = e->offset - last->offset;
1090 unsigned pos = sizeof(hdr) - 1;
1092 delta_count_by_type[type]++;
1093 e->depth = last->depth + 1;
1095 hdrlen = encode_header(OBJ_OFS_DELTA, deltalen, hdr);
1096 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1097 pack_size += hdrlen;
1099 hdr[pos] = ofs & 127;
1101 hdr[--pos] = 128 | (--ofs & 127);
1102 write_or_die(pack_data->pack_fd, hdr + pos, sizeof(hdr) - pos);
1103 pack_size += sizeof(hdr) - pos;
1106 hdrlen = encode_header(type, dat->len, hdr);
1107 write_or_die(pack_data->pack_fd, hdr, hdrlen);
1108 pack_size += hdrlen;
1111 write_or_die(pack_data->pack_fd, out, s.total_out);
1112 pack_size += s.total_out;
1117 if (last->no_swap) {
1120 strbuf_swap(&last->data, dat);
1122 last->offset = e->offset;
1123 last->depth = e->depth;
1128 /* All calls must be guarded by find_object() or find_mark() to
1129 * ensure the 'struct object_entry' passed was written by this
1130 * process instance. We unpack the entry by the offset, avoiding
1131 * the need for the corresponding .idx file. This unpacking rule
1132 * works because we only use OBJ_REF_DELTA within the packfiles
1133 * created by fast-import.
1135 * oe must not be NULL. Such an oe usually comes from giving
1136 * an unknown SHA-1 to find_object() or an undefined mark to
1137 * find_mark(). Callers must test for this condition and use
1138 * the standard read_sha1_file() when it happens.
1140 * oe->pack_id must not be MAX_PACK_ID. Such an oe is usually from
1141 * find_mark(), where the mark was reloaded from an existing marks
1142 * file and is referencing an object that this fast-import process
1143 * instance did not write out to a packfile. Callers must test for
1144 * this condition and use read_sha1_file() instead.
1146 static void *gfi_unpack_entry(
1147 struct object_entry *oe,
1148 unsigned long *sizep)
1150 enum object_type type;
1151 struct packed_git *p = all_packs[oe->pack_id];
1152 if (p == pack_data && p->pack_size < (pack_size + 20)) {
1153 /* The object is stored in the packfile we are writing to
1154 * and we have modified it since the last time we scanned
1155 * back to read a previously written object. If an old
1156 * window covered [p->pack_size, p->pack_size + 20) its
1157 * data is stale and is not valid. Closing all windows
1158 * and updating the packfile length ensures we can read
1159 * the newly written data.
1161 close_pack_windows(p);
1163 /* We have to offer 20 bytes additional on the end of
1164 * the packfile as the core unpacker code assumes the
1165 * footer is present at the file end and must promise
1166 * at least 20 bytes within any window it maps. But
1167 * we don't actually create the footer here.
1169 p->pack_size = pack_size + 20;
1171 return unpack_entry(p, oe->offset, &type, sizep);
1174 static const char *get_mode(const char *str, uint16_t *modep)
1179 while ((c = *str++) != ' ') {
1180 if (c < '0' || c > '7')
1182 mode = (mode << 3) + (c - '0');
1188 static void load_tree(struct tree_entry *root)
1190 unsigned char* sha1 = root->versions[1].sha1;
1191 struct object_entry *myoe;
1192 struct tree_content *t;
1197 root->tree = t = new_tree_content(8);
1198 if (is_null_sha1(sha1))
1201 myoe = find_object(sha1);
1202 if (myoe && myoe->pack_id != MAX_PACK_ID) {
1203 if (myoe->type != OBJ_TREE)
1204 die("Not a tree: %s", sha1_to_hex(sha1));
1205 t->delta_depth = myoe->depth;
1206 buf = gfi_unpack_entry(myoe, &size);
1208 enum object_type type;
1209 buf = read_sha1_file(sha1, &type, &size);
1210 if (!buf || type != OBJ_TREE)
1211 die("Can't load tree %s", sha1_to_hex(sha1));
1215 while (c != (buf + size)) {
1216 struct tree_entry *e = new_tree_entry();
1218 if (t->entry_count == t->entry_capacity)
1219 root->tree = t = grow_tree_content(t, t->entry_count);
1220 t->entries[t->entry_count++] = e;
1223 c = get_mode(c, &e->versions[1].mode);
1225 die("Corrupt mode in %s", sha1_to_hex(sha1));
1226 e->versions[0].mode = e->versions[1].mode;
1227 e->name = to_atom(c, strlen(c));
1228 c += e->name->str_len + 1;
1229 hashcpy(e->versions[0].sha1, (unsigned char*)c);
1230 hashcpy(e->versions[1].sha1, (unsigned char*)c);
1236 static int tecmp0 (const void *_a, const void *_b)
1238 struct tree_entry *a = *((struct tree_entry**)_a);
1239 struct tree_entry *b = *((struct tree_entry**)_b);
1240 return base_name_compare(
1241 a->name->str_dat, a->name->str_len, a->versions[0].mode,
1242 b->name->str_dat, b->name->str_len, b->versions[0].mode);
1245 static int tecmp1 (const void *_a, const void *_b)
1247 struct tree_entry *a = *((struct tree_entry**)_a);
1248 struct tree_entry *b = *((struct tree_entry**)_b);
1249 return base_name_compare(
1250 a->name->str_dat, a->name->str_len, a->versions[1].mode,
1251 b->name->str_dat, b->name->str_len, b->versions[1].mode);
1254 static void mktree(struct tree_content *t, int v, struct strbuf *b)
1260 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp0);
1262 qsort(t->entries,t->entry_count,sizeof(t->entries[0]),tecmp1);
1264 for (i = 0; i < t->entry_count; i++) {
1265 if (t->entries[i]->versions[v].mode)
1266 maxlen += t->entries[i]->name->str_len + 34;
1270 strbuf_grow(b, maxlen);
1271 for (i = 0; i < t->entry_count; i++) {
1272 struct tree_entry *e = t->entries[i];
1273 if (!e->versions[v].mode)
1275 strbuf_addf(b, "%o %s%c", (unsigned int)e->versions[v].mode,
1276 e->name->str_dat, '\0');
1277 strbuf_add(b, e->versions[v].sha1, 20);
1281 static void store_tree(struct tree_entry *root)
1283 struct tree_content *t = root->tree;
1284 unsigned int i, j, del;
1285 struct last_object lo = { STRBUF_INIT, 0, 0, /* no_swap */ 1 };
1286 struct object_entry *le;
1288 if (!is_null_sha1(root->versions[1].sha1))
1291 for (i = 0; i < t->entry_count; i++) {
1292 if (t->entries[i]->tree)
1293 store_tree(t->entries[i]);
1296 le = find_object(root->versions[0].sha1);
1297 if (S_ISDIR(root->versions[0].mode) && le && le->pack_id == pack_id) {
1298 mktree(t, 0, &old_tree);
1300 lo.offset = le->offset;
1301 lo.depth = t->delta_depth;
1304 mktree(t, 1, &new_tree);
1305 store_object(OBJ_TREE, &new_tree, &lo, root->versions[1].sha1, 0);
1307 t->delta_depth = lo.depth;
1308 for (i = 0, j = 0, del = 0; i < t->entry_count; i++) {
1309 struct tree_entry *e = t->entries[i];
1310 if (e->versions[1].mode) {
1311 e->versions[0].mode = e->versions[1].mode;
1312 hashcpy(e->versions[0].sha1, e->versions[1].sha1);
1313 t->entries[j++] = e;
1315 release_tree_entry(e);
1319 t->entry_count -= del;
1322 static int tree_content_set(
1323 struct tree_entry *root,
1325 const unsigned char *sha1,
1326 const uint16_t mode,
1327 struct tree_content *subtree)
1329 struct tree_content *t = root->tree;
1332 struct tree_entry *e;
1334 slash1 = strchr(p, '/');
1340 die("Empty path component found in input");
1341 if (!slash1 && !S_ISDIR(mode) && subtree)
1342 die("Non-directories cannot have subtrees");
1344 for (i = 0; i < t->entry_count; i++) {
1346 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1349 && e->versions[1].mode == mode
1350 && !hashcmp(e->versions[1].sha1, sha1))
1352 e->versions[1].mode = mode;
1353 hashcpy(e->versions[1].sha1, sha1);
1355 release_tree_content_recursive(e->tree);
1357 hashclr(root->versions[1].sha1);
1360 if (!S_ISDIR(e->versions[1].mode)) {
1361 e->tree = new_tree_content(8);
1362 e->versions[1].mode = S_IFDIR;
1366 if (tree_content_set(e, slash1 + 1, sha1, mode, subtree)) {
1367 hashclr(root->versions[1].sha1);
1374 if (t->entry_count == t->entry_capacity)
1375 root->tree = t = grow_tree_content(t, t->entry_count);
1376 e = new_tree_entry();
1377 e->name = to_atom(p, n);
1378 e->versions[0].mode = 0;
1379 hashclr(e->versions[0].sha1);
1380 t->entries[t->entry_count++] = e;
1382 e->tree = new_tree_content(8);
1383 e->versions[1].mode = S_IFDIR;
1384 tree_content_set(e, slash1 + 1, sha1, mode, subtree);
1387 e->versions[1].mode = mode;
1388 hashcpy(e->versions[1].sha1, sha1);
1390 hashclr(root->versions[1].sha1);
1394 static int tree_content_remove(
1395 struct tree_entry *root,
1397 struct tree_entry *backup_leaf)
1399 struct tree_content *t = root->tree;
1402 struct tree_entry *e;
1404 slash1 = strchr(p, '/');
1410 for (i = 0; i < t->entry_count; i++) {
1412 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1413 if (!slash1 || !S_ISDIR(e->versions[1].mode))
1417 if (tree_content_remove(e, slash1 + 1, backup_leaf)) {
1418 for (n = 0; n < e->tree->entry_count; n++) {
1419 if (e->tree->entries[n]->versions[1].mode) {
1420 hashclr(root->versions[1].sha1);
1434 memcpy(backup_leaf, e, sizeof(*backup_leaf));
1436 release_tree_content_recursive(e->tree);
1438 e->versions[1].mode = 0;
1439 hashclr(e->versions[1].sha1);
1440 hashclr(root->versions[1].sha1);
1444 static int tree_content_get(
1445 struct tree_entry *root,
1447 struct tree_entry *leaf)
1449 struct tree_content *t = root->tree;
1452 struct tree_entry *e;
1454 slash1 = strchr(p, '/');
1460 for (i = 0; i < t->entry_count; i++) {
1462 if (e->name->str_len == n && !strncmp(p, e->name->str_dat, n)) {
1464 memcpy(leaf, e, sizeof(*leaf));
1465 if (e->tree && is_null_sha1(e->versions[1].sha1))
1466 leaf->tree = dup_tree_content(e->tree);
1471 if (!S_ISDIR(e->versions[1].mode))
1475 return tree_content_get(e, slash1 + 1, leaf);
1481 static int update_branch(struct branch *b)
1483 static const char *msg = "fast-import";
1484 struct ref_lock *lock;
1485 unsigned char old_sha1[20];
1487 if (read_ref(b->name, old_sha1))
1489 lock = lock_any_ref_for_update(b->name, old_sha1, 0);
1491 return error("Unable to lock %s", b->name);
1492 if (!force_update && !is_null_sha1(old_sha1)) {
1493 struct commit *old_cmit, *new_cmit;
1495 old_cmit = lookup_commit_reference_gently(old_sha1, 0);
1496 new_cmit = lookup_commit_reference_gently(b->sha1, 0);
1497 if (!old_cmit || !new_cmit) {
1499 return error("Branch %s is missing commits.", b->name);
1502 if (!in_merge_bases(old_cmit, &new_cmit, 1)) {
1504 warning("Not updating %s"
1505 " (new tip %s does not contain %s)",
1506 b->name, sha1_to_hex(b->sha1), sha1_to_hex(old_sha1));
1510 if (write_ref_sha1(lock, b->sha1, msg) < 0)
1511 return error("Unable to update %s", b->name);
1515 static void dump_branches(void)
1520 for (i = 0; i < branch_table_sz; i++) {
1521 for (b = branch_table[i]; b; b = b->table_next_branch)
1522 failure |= update_branch(b);
1526 static void dump_tags(void)
1528 static const char *msg = "fast-import";
1530 struct ref_lock *lock;
1531 char ref_name[PATH_MAX];
1533 for (t = first_tag; t; t = t->next_tag) {
1534 sprintf(ref_name, "tags/%s", t->name);
1535 lock = lock_ref_sha1(ref_name, NULL);
1536 if (!lock || write_ref_sha1(lock, t->sha1, msg) < 0)
1537 failure |= error("Unable to update %s", ref_name);
1541 static void dump_marks_helper(FILE *f,
1547 for (k = 0; k < 1024; k++) {
1548 if (m->data.sets[k])
1549 dump_marks_helper(f, (base + k) << m->shift,
1553 for (k = 0; k < 1024; k++) {
1554 if (m->data.marked[k])
1555 fprintf(f, ":%" PRIuMAX " %s\n", base + k,
1556 sha1_to_hex(m->data.marked[k]->sha1));
1561 static void dump_marks(void)
1563 static struct lock_file mark_lock;
1570 mark_fd = hold_lock_file_for_update(&mark_lock, mark_file, 0);
1572 failure |= error("Unable to write marks file %s: %s",
1573 mark_file, strerror(errno));
1577 f = fdopen(mark_fd, "w");
1579 int saved_errno = errno;
1580 rollback_lock_file(&mark_lock);
1581 failure |= error("Unable to write marks file %s: %s",
1582 mark_file, strerror(saved_errno));
1587 * Since the lock file was fdopen()'ed, it should not be close()'ed.
1588 * Assign -1 to the lock file descriptor so that commit_lock_file()
1589 * won't try to close() it.
1593 dump_marks_helper(f, 0, marks);
1594 if (ferror(f) || fclose(f)) {
1595 int saved_errno = errno;
1596 rollback_lock_file(&mark_lock);
1597 failure |= error("Unable to write marks file %s: %s",
1598 mark_file, strerror(saved_errno));
1602 if (commit_lock_file(&mark_lock)) {
1603 int saved_errno = errno;
1604 rollback_lock_file(&mark_lock);
1605 failure |= error("Unable to commit marks file %s: %s",
1606 mark_file, strerror(saved_errno));
1611 static int read_next_command(void)
1613 static int stdin_eof = 0;
1616 unread_command_buf = 0;
1621 if (unread_command_buf) {
1622 unread_command_buf = 0;
1624 struct recent_command *rc;
1626 strbuf_detach(&command_buf, NULL);
1627 stdin_eof = strbuf_getline(&command_buf, stdin, '\n');
1636 cmd_hist.next = rc->next;
1637 cmd_hist.next->prev = &cmd_hist;
1641 rc->buf = command_buf.buf;
1642 rc->prev = cmd_tail;
1643 rc->next = cmd_hist.prev;
1644 rc->prev->next = rc;
1647 } while (command_buf.buf[0] == '#');
1652 static void skip_optional_lf(void)
1654 int term_char = fgetc(stdin);
1655 if (term_char != '\n' && term_char != EOF)
1656 ungetc(term_char, stdin);
1659 static void cmd_mark(void)
1661 if (!prefixcmp(command_buf.buf, "mark :")) {
1662 next_mark = strtoumax(command_buf.buf + 6, NULL, 10);
1663 read_next_command();
1669 static void cmd_data(struct strbuf *sb)
1673 if (prefixcmp(command_buf.buf, "data "))
1674 die("Expected 'data n' command, found: %s", command_buf.buf);
1676 if (!prefixcmp(command_buf.buf + 5, "<<")) {
1677 char *term = xstrdup(command_buf.buf + 5 + 2);
1678 size_t term_len = command_buf.len - 5 - 2;
1680 strbuf_detach(&command_buf, NULL);
1682 if (strbuf_getline(&command_buf, stdin, '\n') == EOF)
1683 die("EOF in data (terminator '%s' not found)", term);
1684 if (term_len == command_buf.len
1685 && !strcmp(term, command_buf.buf))
1687 strbuf_addbuf(sb, &command_buf);
1688 strbuf_addch(sb, '\n');
1693 size_t n = 0, length;
1695 length = strtoul(command_buf.buf + 5, NULL, 10);
1697 while (n < length) {
1698 size_t s = strbuf_fread(sb, length - n, stdin);
1699 if (!s && feof(stdin))
1700 die("EOF in data (%lu bytes remaining)",
1701 (unsigned long)(length - n));
1709 static int validate_raw_date(const char *src, char *result, int maxlen)
1711 const char *orig_src = src;
1714 strtoul(src, &endp, 10);
1715 if (endp == src || *endp != ' ')
1719 if (*src != '-' && *src != '+')
1723 strtoul(src + 1, &endp, 10);
1724 if (endp == src || *endp || (endp - orig_src) >= maxlen)
1727 strcpy(result, orig_src);
1731 static char *parse_ident(const char *buf)
1737 gt = strrchr(buf, '>');
1739 die("Missing > in ident string: %s", buf);
1742 die("Missing space after > in ident string: %s", buf);
1744 name_len = gt - buf;
1745 ident = xmalloc(name_len + 24);
1746 strncpy(ident, buf, name_len);
1750 if (validate_raw_date(gt, ident + name_len, 24) < 0)
1751 die("Invalid raw date \"%s\" in ident: %s", gt, buf);
1753 case WHENSPEC_RFC2822:
1754 if (parse_date(gt, ident + name_len, 24) < 0)
1755 die("Invalid rfc2822 date \"%s\" in ident: %s", gt, buf);
1758 if (strcmp("now", gt))
1759 die("Date in ident must be 'now': %s", buf);
1760 datestamp(ident + name_len, 24);
1767 static void cmd_new_blob(void)
1769 static struct strbuf buf = STRBUF_INIT;
1771 read_next_command();
1774 store_object(OBJ_BLOB, &buf, &last_blob, NULL, next_mark);
1777 static void unload_one_branch(void)
1779 while (cur_active_branches
1780 && cur_active_branches >= max_active_branches) {
1781 uintmax_t min_commit = ULONG_MAX;
1782 struct branch *e, *l = NULL, *p = NULL;
1784 for (e = active_branches; e; e = e->active_next_branch) {
1785 if (e->last_commit < min_commit) {
1787 min_commit = e->last_commit;
1793 e = p->active_next_branch;
1794 p->active_next_branch = e->active_next_branch;
1796 e = active_branches;
1797 active_branches = e->active_next_branch;
1800 e->active_next_branch = NULL;
1801 if (e->branch_tree.tree) {
1802 release_tree_content_recursive(e->branch_tree.tree);
1803 e->branch_tree.tree = NULL;
1805 cur_active_branches--;
1809 static void load_branch(struct branch *b)
1811 load_tree(&b->branch_tree);
1814 b->active_next_branch = active_branches;
1815 active_branches = b;
1816 cur_active_branches++;
1817 branch_load_count++;
1821 static void file_change_m(struct branch *b)
1823 const char *p = command_buf.buf + 2;
1824 static struct strbuf uq = STRBUF_INIT;
1826 struct object_entry *oe = oe;
1827 unsigned char sha1[20];
1828 uint16_t mode, inline_data = 0;
1830 p = get_mode(p, &mode);
1832 die("Corrupt mode: %s", command_buf.buf);
1834 case S_IFREG | 0644:
1835 case S_IFREG | 0755:
1842 die("Corrupt mode: %s", command_buf.buf);
1847 oe = find_mark(strtoumax(p + 1, &x, 10));
1848 hashcpy(sha1, oe->sha1);
1850 } else if (!prefixcmp(p, "inline")) {
1854 if (get_sha1_hex(p, sha1))
1855 die("Invalid SHA1: %s", command_buf.buf);
1856 oe = find_object(sha1);
1860 die("Missing space after SHA1: %s", command_buf.buf);
1863 if (!unquote_c_style(&uq, p, &endp)) {
1865 die("Garbage after path in: %s", command_buf.buf);
1870 static struct strbuf buf = STRBUF_INIT;
1873 strbuf_addstr(&uq, p);
1876 read_next_command();
1878 store_object(OBJ_BLOB, &buf, &last_blob, sha1, 0);
1880 if (oe->type != OBJ_BLOB)
1881 die("Not a blob (actually a %s): %s",
1882 typename(oe->type), command_buf.buf);
1884 enum object_type type = sha1_object_info(sha1, NULL);
1886 die("Blob not found: %s", command_buf.buf);
1887 if (type != OBJ_BLOB)
1888 die("Not a blob (actually a %s): %s",
1889 typename(type), command_buf.buf);
1892 tree_content_set(&b->branch_tree, p, sha1, S_IFREG | mode, NULL);
1895 static void file_change_d(struct branch *b)
1897 const char *p = command_buf.buf + 2;
1898 static struct strbuf uq = STRBUF_INIT;
1902 if (!unquote_c_style(&uq, p, &endp)) {
1904 die("Garbage after path in: %s", command_buf.buf);
1907 tree_content_remove(&b->branch_tree, p, NULL);
1910 static void file_change_cr(struct branch *b, int rename)
1913 static struct strbuf s_uq = STRBUF_INIT;
1914 static struct strbuf d_uq = STRBUF_INIT;
1916 struct tree_entry leaf;
1918 s = command_buf.buf + 2;
1919 strbuf_reset(&s_uq);
1920 if (!unquote_c_style(&s_uq, s, &endp)) {
1922 die("Missing space after source: %s", command_buf.buf);
1924 endp = strchr(s, ' ');
1926 die("Missing space after source: %s", command_buf.buf);
1927 strbuf_add(&s_uq, s, endp - s);
1933 die("Missing dest: %s", command_buf.buf);
1936 strbuf_reset(&d_uq);
1937 if (!unquote_c_style(&d_uq, d, &endp)) {
1939 die("Garbage after dest in: %s", command_buf.buf);
1943 memset(&leaf, 0, sizeof(leaf));
1945 tree_content_remove(&b->branch_tree, s, &leaf);
1947 tree_content_get(&b->branch_tree, s, &leaf);
1948 if (!leaf.versions[1].mode)
1949 die("Path %s not in branch", s);
1950 tree_content_set(&b->branch_tree, d,
1951 leaf.versions[1].sha1,
1952 leaf.versions[1].mode,
1956 static void file_change_deleteall(struct branch *b)
1958 release_tree_content_recursive(b->branch_tree.tree);
1959 hashclr(b->branch_tree.versions[0].sha1);
1960 hashclr(b->branch_tree.versions[1].sha1);
1961 load_tree(&b->branch_tree);
1964 static void cmd_from_commit(struct branch *b, char *buf, unsigned long size)
1966 if (!buf || size < 46)
1967 die("Not a valid commit: %s", sha1_to_hex(b->sha1));
1968 if (memcmp("tree ", buf, 5)
1969 || get_sha1_hex(buf + 5, b->branch_tree.versions[1].sha1))
1970 die("The commit %s is corrupt", sha1_to_hex(b->sha1));
1971 hashcpy(b->branch_tree.versions[0].sha1,
1972 b->branch_tree.versions[1].sha1);
1975 static void cmd_from_existing(struct branch *b)
1977 if (is_null_sha1(b->sha1)) {
1978 hashclr(b->branch_tree.versions[0].sha1);
1979 hashclr(b->branch_tree.versions[1].sha1);
1984 buf = read_object_with_reference(b->sha1,
1985 commit_type, &size, b->sha1);
1986 cmd_from_commit(b, buf, size);
1991 static int cmd_from(struct branch *b)
1996 if (prefixcmp(command_buf.buf, "from "))
1999 if (b->branch_tree.tree) {
2000 release_tree_content_recursive(b->branch_tree.tree);
2001 b->branch_tree.tree = NULL;
2004 from = strchr(command_buf.buf, ' ') + 1;
2005 s = lookup_branch(from);
2007 die("Can't create a branch from itself: %s", b->name);
2009 unsigned char *t = s->branch_tree.versions[1].sha1;
2010 hashcpy(b->sha1, s->sha1);
2011 hashcpy(b->branch_tree.versions[0].sha1, t);
2012 hashcpy(b->branch_tree.versions[1].sha1, t);
2013 } else if (*from == ':') {
2014 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2015 struct object_entry *oe = find_mark(idnum);
2016 if (oe->type != OBJ_COMMIT)
2017 die("Mark :%" PRIuMAX " not a commit", idnum);
2018 hashcpy(b->sha1, oe->sha1);
2019 if (oe->pack_id != MAX_PACK_ID) {
2021 char *buf = gfi_unpack_entry(oe, &size);
2022 cmd_from_commit(b, buf, size);
2025 cmd_from_existing(b);
2026 } else if (!get_sha1(from, b->sha1))
2027 cmd_from_existing(b);
2029 die("Invalid ref name or SHA1 expression: %s", from);
2031 read_next_command();
2035 static struct hash_list *cmd_merge(unsigned int *count)
2037 struct hash_list *list = NULL, *n, *e = e;
2042 while (!prefixcmp(command_buf.buf, "merge ")) {
2043 from = strchr(command_buf.buf, ' ') + 1;
2044 n = xmalloc(sizeof(*n));
2045 s = lookup_branch(from);
2047 hashcpy(n->sha1, s->sha1);
2048 else if (*from == ':') {
2049 uintmax_t idnum = strtoumax(from + 1, NULL, 10);
2050 struct object_entry *oe = find_mark(idnum);
2051 if (oe->type != OBJ_COMMIT)
2052 die("Mark :%" PRIuMAX " not a commit", idnum);
2053 hashcpy(n->sha1, oe->sha1);
2054 } else if (!get_sha1(from, n->sha1)) {
2056 char *buf = read_object_with_reference(n->sha1,
2057 commit_type, &size, n->sha1);
2058 if (!buf || size < 46)
2059 die("Not a valid commit: %s", from);
2062 die("Invalid ref name or SHA1 expression: %s", from);
2071 read_next_command();
2076 static void cmd_new_commit(void)
2078 static struct strbuf msg = STRBUF_INIT;
2081 char *author = NULL;
2082 char *committer = NULL;
2083 struct hash_list *merge_list = NULL;
2084 unsigned int merge_count;
2086 /* Obtain the branch name from the rest of our command */
2087 sp = strchr(command_buf.buf, ' ') + 1;
2088 b = lookup_branch(sp);
2092 read_next_command();
2094 if (!prefixcmp(command_buf.buf, "author ")) {
2095 author = parse_ident(command_buf.buf + 7);
2096 read_next_command();
2098 if (!prefixcmp(command_buf.buf, "committer ")) {
2099 committer = parse_ident(command_buf.buf + 10);
2100 read_next_command();
2103 die("Expected committer but didn't get one");
2105 read_next_command();
2107 merge_list = cmd_merge(&merge_count);
2109 /* ensure the branch is active/loaded */
2110 if (!b->branch_tree.tree || !max_active_branches) {
2111 unload_one_branch();
2116 while (command_buf.len > 0) {
2117 if (!prefixcmp(command_buf.buf, "M "))
2119 else if (!prefixcmp(command_buf.buf, "D "))
2121 else if (!prefixcmp(command_buf.buf, "R "))
2122 file_change_cr(b, 1);
2123 else if (!prefixcmp(command_buf.buf, "C "))
2124 file_change_cr(b, 0);
2125 else if (!strcmp("deleteall", command_buf.buf))
2126 file_change_deleteall(b);
2128 unread_command_buf = 1;
2131 if (read_next_command() == EOF)
2135 /* build the tree and the commit */
2136 store_tree(&b->branch_tree);
2137 hashcpy(b->branch_tree.versions[0].sha1,
2138 b->branch_tree.versions[1].sha1);
2140 strbuf_reset(&new_data);
2141 strbuf_addf(&new_data, "tree %s\n",
2142 sha1_to_hex(b->branch_tree.versions[1].sha1));
2143 if (!is_null_sha1(b->sha1))
2144 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(b->sha1));
2145 while (merge_list) {
2146 struct hash_list *next = merge_list->next;
2147 strbuf_addf(&new_data, "parent %s\n", sha1_to_hex(merge_list->sha1));
2151 strbuf_addf(&new_data,
2155 author ? author : committer, committer);
2156 strbuf_addbuf(&new_data, &msg);
2160 if (!store_object(OBJ_COMMIT, &new_data, NULL, b->sha1, next_mark))
2161 b->pack_id = pack_id;
2162 b->last_commit = object_count_by_type[OBJ_COMMIT];
2165 static void cmd_new_tag(void)
2167 static struct strbuf msg = STRBUF_INIT;
2173 uintmax_t from_mark = 0;
2174 unsigned char sha1[20];
2176 /* Obtain the new tag name from the rest of our command */
2177 sp = strchr(command_buf.buf, ' ') + 1;
2178 t = pool_alloc(sizeof(struct tag));
2180 t->name = pool_strdup(sp);
2182 last_tag->next_tag = t;
2186 read_next_command();
2189 if (prefixcmp(command_buf.buf, "from "))
2190 die("Expected from command, got %s", command_buf.buf);
2191 from = strchr(command_buf.buf, ' ') + 1;
2192 s = lookup_branch(from);
2194 hashcpy(sha1, s->sha1);
2195 } else if (*from == ':') {
2196 struct object_entry *oe;
2197 from_mark = strtoumax(from + 1, NULL, 10);
2198 oe = find_mark(from_mark);
2199 if (oe->type != OBJ_COMMIT)
2200 die("Mark :%" PRIuMAX " not a commit", from_mark);
2201 hashcpy(sha1, oe->sha1);
2202 } else if (!get_sha1(from, sha1)) {
2206 buf = read_object_with_reference(sha1,
2207 commit_type, &size, sha1);
2208 if (!buf || size < 46)
2209 die("Not a valid commit: %s", from);
2212 die("Invalid ref name or SHA1 expression: %s", from);
2213 read_next_command();
2216 if (prefixcmp(command_buf.buf, "tagger "))
2217 die("Expected tagger command, got %s", command_buf.buf);
2218 tagger = parse_ident(command_buf.buf + 7);
2220 /* tag payload/message */
2221 read_next_command();
2224 /* build the tag object */
2225 strbuf_reset(&new_data);
2226 strbuf_addf(&new_data,
2232 sha1_to_hex(sha1), commit_type, t->name, tagger);
2233 strbuf_addbuf(&new_data, &msg);
2236 if (store_object(OBJ_TAG, &new_data, NULL, t->sha1, 0))
2237 t->pack_id = MAX_PACK_ID;
2239 t->pack_id = pack_id;
2242 static void cmd_reset_branch(void)
2247 /* Obtain the branch name from the rest of our command */
2248 sp = strchr(command_buf.buf, ' ') + 1;
2249 b = lookup_branch(sp);
2252 hashclr(b->branch_tree.versions[0].sha1);
2253 hashclr(b->branch_tree.versions[1].sha1);
2254 if (b->branch_tree.tree) {
2255 release_tree_content_recursive(b->branch_tree.tree);
2256 b->branch_tree.tree = NULL;
2261 read_next_command();
2262 if (!cmd_from(b) && command_buf.len > 0)
2263 unread_command_buf = 1;
2266 static void cmd_checkpoint(void)
2277 static void cmd_progress(void)
2279 fwrite(command_buf.buf, 1, command_buf.len, stdout);
2280 fputc('\n', stdout);
2285 static void import_marks(const char *input_file)
2288 FILE *f = fopen(input_file, "r");
2290 die("cannot read %s: %s", input_file, strerror(errno));
2291 while (fgets(line, sizeof(line), f)) {
2294 unsigned char sha1[20];
2295 struct object_entry *e;
2297 end = strchr(line, '\n');
2298 if (line[0] != ':' || !end)
2299 die("corrupt mark line: %s", line);
2301 mark = strtoumax(line + 1, &end, 10);
2302 if (!mark || end == line + 1
2303 || *end != ' ' || get_sha1(end + 1, sha1))
2304 die("corrupt mark line: %s", line);
2305 e = find_object(sha1);
2307 enum object_type type = sha1_object_info(sha1, NULL);
2309 die("object not found: %s", sha1_to_hex(sha1));
2310 e = insert_object(sha1);
2312 e->pack_id = MAX_PACK_ID;
2313 e->offset = 1; /* just not zero! */
2315 insert_mark(mark, e);
2320 static int git_pack_config(const char *k, const char *v)
2322 if (!strcmp(k, "pack.depth")) {
2323 max_depth = git_config_int(k, v);
2324 if (max_depth > MAX_DEPTH)
2325 max_depth = MAX_DEPTH;
2328 if (!strcmp(k, "pack.compression")) {
2329 int level = git_config_int(k, v);
2331 level = Z_DEFAULT_COMPRESSION;
2332 else if (level < 0 || level > Z_BEST_COMPRESSION)
2333 die("bad pack compression level %d", level);
2334 pack_compression_level = level;
2335 pack_compression_seen = 1;
2338 return git_default_config(k, v);
2341 static const char fast_import_usage[] =
2342 "git-fast-import [--date-format=f] [--max-pack-size=n] [--depth=n] [--active-branches=n] [--export-marks=marks.file]";
2344 int main(int argc, const char **argv)
2346 unsigned int i, show_stats = 1;
2348 git_config(git_pack_config);
2349 if (!pack_compression_seen && core_compression_seen)
2350 pack_compression_level = core_compression_level;
2352 alloc_objects(object_entry_alloc);
2353 strbuf_init(&command_buf, 0);
2354 atom_table = xcalloc(atom_table_sz, sizeof(struct atom_str*));
2355 branch_table = xcalloc(branch_table_sz, sizeof(struct branch*));
2356 avail_tree_table = xcalloc(avail_tree_table_sz, sizeof(struct avail_tree_content*));
2357 marks = pool_calloc(1, sizeof(struct mark_set));
2359 for (i = 1; i < argc; i++) {
2360 const char *a = argv[i];
2362 if (*a != '-' || !strcmp(a, "--"))
2364 else if (!prefixcmp(a, "--date-format=")) {
2365 const char *fmt = a + 14;
2366 if (!strcmp(fmt, "raw"))
2367 whenspec = WHENSPEC_RAW;
2368 else if (!strcmp(fmt, "rfc2822"))
2369 whenspec = WHENSPEC_RFC2822;
2370 else if (!strcmp(fmt, "now"))
2371 whenspec = WHENSPEC_NOW;
2373 die("unknown --date-format argument %s", fmt);
2375 else if (!prefixcmp(a, "--max-pack-size="))
2376 max_packsize = strtoumax(a + 16, NULL, 0) * 1024 * 1024;
2377 else if (!prefixcmp(a, "--depth=")) {
2378 max_depth = strtoul(a + 8, NULL, 0);
2379 if (max_depth > MAX_DEPTH)
2380 die("--depth cannot exceed %u", MAX_DEPTH);
2382 else if (!prefixcmp(a, "--active-branches="))
2383 max_active_branches = strtoul(a + 18, NULL, 0);
2384 else if (!prefixcmp(a, "--import-marks="))
2385 import_marks(a + 15);
2386 else if (!prefixcmp(a, "--export-marks="))
2388 else if (!prefixcmp(a, "--export-pack-edges=")) {
2391 pack_edges = fopen(a + 20, "a");
2393 die("Cannot open %s: %s", a + 20, strerror(errno));
2394 } else if (!strcmp(a, "--force"))
2396 else if (!strcmp(a, "--quiet"))
2398 else if (!strcmp(a, "--stats"))
2401 die("unknown option %s", a);
2404 usage(fast_import_usage);
2406 rc_free = pool_alloc(cmd_save * sizeof(*rc_free));
2407 for (i = 0; i < (cmd_save - 1); i++)
2408 rc_free[i].next = &rc_free[i + 1];
2409 rc_free[cmd_save - 1].next = NULL;
2411 prepare_packed_git();
2413 set_die_routine(die_nicely);
2414 while (read_next_command() != EOF) {
2415 if (!strcmp("blob", command_buf.buf))
2417 else if (!prefixcmp(command_buf.buf, "commit "))
2419 else if (!prefixcmp(command_buf.buf, "tag "))
2421 else if (!prefixcmp(command_buf.buf, "reset "))
2423 else if (!strcmp("checkpoint", command_buf.buf))
2425 else if (!prefixcmp(command_buf.buf, "progress "))
2428 die("Unsupported command: %s", command_buf.buf);
2441 uintmax_t total_count = 0, duplicate_count = 0;
2442 for (i = 0; i < ARRAY_SIZE(object_count_by_type); i++)
2443 total_count += object_count_by_type[i];
2444 for (i = 0; i < ARRAY_SIZE(duplicate_count_by_type); i++)
2445 duplicate_count += duplicate_count_by_type[i];
2447 fprintf(stderr, "%s statistics:\n", argv[0]);
2448 fprintf(stderr, "---------------------------------------------------------------------\n");
2449 fprintf(stderr, "Alloc'd objects: %10" PRIuMAX "\n", alloc_count);
2450 fprintf(stderr, "Total objects: %10" PRIuMAX " (%10" PRIuMAX " duplicates )\n", total_count, duplicate_count);
2451 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]);
2452 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]);
2453 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]);
2454 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]);
2455 fprintf(stderr, "Total branches: %10lu (%10lu loads )\n", branch_count, branch_load_count);
2456 fprintf(stderr, " marks: %10" PRIuMAX " (%10" PRIuMAX " unique )\n", (((uintmax_t)1) << marks->shift) * 1024, marks_set_count);
2457 fprintf(stderr, " atoms: %10u\n", atom_cnt);
2458 fprintf(stderr, "Memory total: %10" PRIuMAX " KiB\n", (total_allocd + alloc_count*sizeof(struct object_entry))/1024);
2459 fprintf(stderr, " pools: %10lu KiB\n", (unsigned long)(total_allocd/1024));
2460 fprintf(stderr, " objects: %10" PRIuMAX " KiB\n", (alloc_count*sizeof(struct object_entry))/1024);
2461 fprintf(stderr, "---------------------------------------------------------------------\n");
2463 fprintf(stderr, "---------------------------------------------------------------------\n");
2464 fprintf(stderr, "\n");
2467 return failure ? 1 : 0;