fix multiple issues in index-pack
[git/trast.git] / index-pack.c
blobf109a00270af28c8ea04a9ff704b6db70b8ff52e
1 #include "cache.h"
2 #include "delta.h"
3 #include "pack.h"
4 #include "csum-file.h"
5 #include "blob.h"
6 #include "commit.h"
7 #include "tag.h"
8 #include "tree.h"
9 #include "progress.h"
10 #include "fsck.h"
12 static const char index_pack_usage[] =
13 "git index-pack [-v] [-o <index-file>] [{ ---keep | --keep=<msg> }] [--strict] { <pack-file> | --stdin [--fix-thin] [<pack-file>] }";
15 struct object_entry
17 struct pack_idx_entry idx;
18 unsigned long size;
19 unsigned int hdr_size;
20 enum object_type type;
21 enum object_type real_type;
24 union delta_base {
25 unsigned char sha1[20];
26 off_t offset;
29 struct base_data {
30 struct base_data *base;
31 struct base_data *child;
32 struct object_entry *obj;
33 void *data;
34 unsigned long size;
38 * Even if sizeof(union delta_base) == 24 on 64-bit archs, we really want
39 * to memcmp() only the first 20 bytes.
41 #define UNION_BASE_SZ 20
43 #define FLAG_LINK (1u<<20)
44 #define FLAG_CHECKED (1u<<21)
46 struct delta_entry
48 union delta_base base;
49 int obj_no;
52 static struct object_entry *objects;
53 static struct delta_entry *deltas;
54 static struct base_data *base_cache;
55 static size_t base_cache_used;
56 static int nr_objects;
57 static int nr_deltas;
58 static int nr_resolved_deltas;
60 static int from_stdin;
61 static int strict;
62 static int verbose;
64 static struct progress *progress;
66 /* We always read in 4kB chunks. */
67 static unsigned char input_buffer[4096];
68 static unsigned int input_offset, input_len;
69 static off_t consumed_bytes;
70 static git_SHA_CTX input_ctx;
71 static uint32_t input_crc32;
72 static int input_fd, output_fd, pack_fd;
74 static int mark_link(struct object *obj, int type, void *data)
76 if (!obj)
77 return -1;
79 if (type != OBJ_ANY && obj->type != type)
80 die("object type mismatch at %s", sha1_to_hex(obj->sha1));
82 obj->flags |= FLAG_LINK;
83 return 0;
86 /* The content of each linked object must have been checked
87 or it must be already present in the object database */
88 static void check_object(struct object *obj)
90 if (!obj)
91 return;
93 if (!(obj->flags & FLAG_LINK))
94 return;
96 if (!(obj->flags & FLAG_CHECKED)) {
97 unsigned long size;
98 int type = sha1_object_info(obj->sha1, &size);
99 if (type != obj->type || type <= 0)
100 die("object of unexpected type");
101 obj->flags |= FLAG_CHECKED;
102 return;
106 static void check_objects(void)
108 unsigned i, max;
110 max = get_max_object_index();
111 for (i = 0; i < max; i++)
112 check_object(get_indexed_object(i));
116 /* Discard current buffer used content. */
117 static void flush(void)
119 if (input_offset) {
120 if (output_fd >= 0)
121 write_or_die(output_fd, input_buffer, input_offset);
122 git_SHA1_Update(&input_ctx, input_buffer, input_offset);
123 memmove(input_buffer, input_buffer + input_offset, input_len);
124 input_offset = 0;
129 * Make sure at least "min" bytes are available in the buffer, and
130 * return the pointer to the buffer.
132 static void *fill(int min)
134 if (min <= input_len)
135 return input_buffer + input_offset;
136 if (min > sizeof(input_buffer))
137 die("cannot fill %d bytes", min);
138 flush();
139 do {
140 ssize_t ret = xread(input_fd, input_buffer + input_len,
141 sizeof(input_buffer) - input_len);
142 if (ret <= 0) {
143 if (!ret)
144 die("early EOF");
145 die("read error on input: %s", strerror(errno));
147 input_len += ret;
148 if (from_stdin)
149 display_throughput(progress, consumed_bytes + input_len);
150 } while (input_len < min);
151 return input_buffer;
154 static void use(int bytes)
156 if (bytes > input_len)
157 die("used more bytes than were available");
158 input_crc32 = crc32(input_crc32, input_buffer + input_offset, bytes);
159 input_len -= bytes;
160 input_offset += bytes;
162 /* make sure off_t is sufficiently large not to wrap */
163 if (consumed_bytes > consumed_bytes + bytes)
164 die("pack too large for current definition of off_t");
165 consumed_bytes += bytes;
168 static char *open_pack_file(char *pack_name)
170 if (from_stdin) {
171 input_fd = 0;
172 if (!pack_name) {
173 static char tmpfile[PATH_MAX];
174 snprintf(tmpfile, sizeof(tmpfile),
175 "%s/pack/tmp_pack_XXXXXX", get_object_directory());
176 output_fd = xmkstemp(tmpfile);
177 pack_name = xstrdup(tmpfile);
178 } else
179 output_fd = open(pack_name, O_CREAT|O_EXCL|O_RDWR, 0600);
180 if (output_fd < 0)
181 die("unable to create %s: %s\n", pack_name, strerror(errno));
182 pack_fd = output_fd;
183 } else {
184 input_fd = open(pack_name, O_RDONLY);
185 if (input_fd < 0)
186 die("cannot open packfile '%s': %s",
187 pack_name, strerror(errno));
188 output_fd = -1;
189 pack_fd = input_fd;
191 git_SHA1_Init(&input_ctx);
192 return pack_name;
195 static void parse_pack_header(void)
197 struct pack_header *hdr = fill(sizeof(struct pack_header));
199 /* Header consistency check */
200 if (hdr->hdr_signature != htonl(PACK_SIGNATURE))
201 die("pack signature mismatch");
202 if (!pack_version_ok(hdr->hdr_version))
203 die("pack version %"PRIu32" unsupported",
204 ntohl(hdr->hdr_version));
206 nr_objects = ntohl(hdr->hdr_entries);
207 use(sizeof(struct pack_header));
210 static void bad_object(unsigned long offset, const char *format,
211 ...) NORETURN __attribute__((format (printf, 2, 3)));
213 static void bad_object(unsigned long offset, const char *format, ...)
215 va_list params;
216 char buf[1024];
218 va_start(params, format);
219 vsnprintf(buf, sizeof(buf), format, params);
220 va_end(params);
221 die("pack has bad object at offset %lu: %s", offset, buf);
224 static void free_base_data(struct base_data *c)
226 if (c->data) {
227 free(c->data);
228 c->data = NULL;
229 base_cache_used -= c->size;
233 static void prune_base_data(struct base_data *retain)
235 struct base_data *b = base_cache;
236 for (b = base_cache;
237 base_cache_used > delta_base_cache_limit && b;
238 b = b->child) {
239 if (b->data && b != retain)
240 free_base_data(b);
244 static void link_base_data(struct base_data *base, struct base_data *c)
246 if (base)
247 base->child = c;
248 else
249 base_cache = c;
251 c->base = base;
252 c->child = NULL;
253 if (c->data)
254 base_cache_used += c->size;
255 prune_base_data(c);
258 static void unlink_base_data(struct base_data *c)
260 struct base_data *base = c->base;
261 if (base)
262 base->child = NULL;
263 else
264 base_cache = NULL;
265 free_base_data(c);
268 static void *unpack_entry_data(unsigned long offset, unsigned long size)
270 z_stream stream;
271 void *buf = xmalloc(size);
273 memset(&stream, 0, sizeof(stream));
274 stream.next_out = buf;
275 stream.avail_out = size;
276 stream.next_in = fill(1);
277 stream.avail_in = input_len;
278 inflateInit(&stream);
280 for (;;) {
281 int ret = inflate(&stream, 0);
282 use(input_len - stream.avail_in);
283 if (stream.total_out == size && ret == Z_STREAM_END)
284 break;
285 if (ret != Z_OK)
286 bad_object(offset, "inflate returned %d", ret);
287 stream.next_in = fill(1);
288 stream.avail_in = input_len;
290 inflateEnd(&stream);
291 return buf;
294 static void *unpack_raw_entry(struct object_entry *obj, union delta_base *delta_base)
296 unsigned char *p, c;
297 unsigned long size;
298 off_t base_offset;
299 unsigned shift;
300 void *data;
302 obj->idx.offset = consumed_bytes;
303 input_crc32 = crc32(0, Z_NULL, 0);
305 p = fill(1);
306 c = *p;
307 use(1);
308 obj->type = (c >> 4) & 7;
309 size = (c & 15);
310 shift = 4;
311 while (c & 0x80) {
312 p = fill(1);
313 c = *p;
314 use(1);
315 size += (c & 0x7fUL) << shift;
316 shift += 7;
318 obj->size = size;
320 switch (obj->type) {
321 case OBJ_REF_DELTA:
322 hashcpy(delta_base->sha1, fill(20));
323 use(20);
324 break;
325 case OBJ_OFS_DELTA:
326 memset(delta_base, 0, sizeof(*delta_base));
327 p = fill(1);
328 c = *p;
329 use(1);
330 base_offset = c & 127;
331 while (c & 128) {
332 base_offset += 1;
333 if (!base_offset || MSB(base_offset, 7))
334 bad_object(obj->idx.offset, "offset value overflow for delta base object");
335 p = fill(1);
336 c = *p;
337 use(1);
338 base_offset = (base_offset << 7) + (c & 127);
340 delta_base->offset = obj->idx.offset - base_offset;
341 if (delta_base->offset >= obj->idx.offset)
342 bad_object(obj->idx.offset, "delta base offset is out of bound");
343 break;
344 case OBJ_COMMIT:
345 case OBJ_TREE:
346 case OBJ_BLOB:
347 case OBJ_TAG:
348 break;
349 default:
350 bad_object(obj->idx.offset, "unknown object type %d", obj->type);
352 obj->hdr_size = consumed_bytes - obj->idx.offset;
354 data = unpack_entry_data(obj->idx.offset, obj->size);
355 obj->idx.crc32 = input_crc32;
356 return data;
359 static void *get_data_from_pack(struct object_entry *obj)
361 off_t from = obj[0].idx.offset + obj[0].hdr_size;
362 unsigned long len = obj[1].idx.offset - from;
363 unsigned long rdy = 0;
364 unsigned char *src, *data;
365 z_stream stream;
366 int st;
368 src = xmalloc(len);
369 data = src;
370 do {
371 ssize_t n = pread(pack_fd, data + rdy, len - rdy, from + rdy);
372 if (n < 0)
373 die("cannot pread pack file: %s", strerror(errno));
374 if (!n)
375 die("premature end of pack file, %lu bytes missing",
376 len - rdy);
377 rdy += n;
378 } while (rdy < len);
379 data = xmalloc(obj->size);
380 memset(&stream, 0, sizeof(stream));
381 stream.next_out = data;
382 stream.avail_out = obj->size;
383 stream.next_in = src;
384 stream.avail_in = len;
385 inflateInit(&stream);
386 while ((st = inflate(&stream, Z_FINISH)) == Z_OK);
387 inflateEnd(&stream);
388 if (st != Z_STREAM_END || stream.total_out != obj->size)
389 die("serious inflate inconsistency");
390 free(src);
391 return data;
394 static int find_delta(const union delta_base *base)
396 int first = 0, last = nr_deltas;
398 while (first < last) {
399 int next = (first + last) / 2;
400 struct delta_entry *delta = &deltas[next];
401 int cmp;
403 cmp = memcmp(base, &delta->base, UNION_BASE_SZ);
404 if (!cmp)
405 return next;
406 if (cmp < 0) {
407 last = next;
408 continue;
410 first = next+1;
412 return -first-1;
415 static void find_delta_children(const union delta_base *base,
416 int *first_index, int *last_index)
418 int first = find_delta(base);
419 int last = first;
420 int end = nr_deltas - 1;
422 if (first < 0) {
423 *first_index = 0;
424 *last_index = -1;
425 return;
427 while (first > 0 && !memcmp(&deltas[first - 1].base, base, UNION_BASE_SZ))
428 --first;
429 while (last < end && !memcmp(&deltas[last + 1].base, base, UNION_BASE_SZ))
430 ++last;
431 *first_index = first;
432 *last_index = last;
435 static void sha1_object(const void *data, unsigned long size,
436 enum object_type type, unsigned char *sha1)
438 hash_sha1_file(data, size, typename(type), sha1);
439 if (has_sha1_file(sha1)) {
440 void *has_data;
441 enum object_type has_type;
442 unsigned long has_size;
443 has_data = read_sha1_file(sha1, &has_type, &has_size);
444 if (!has_data)
445 die("cannot read existing object %s", sha1_to_hex(sha1));
446 if (size != has_size || type != has_type ||
447 memcmp(data, has_data, size) != 0)
448 die("SHA1 COLLISION FOUND WITH %s !", sha1_to_hex(sha1));
449 free(has_data);
451 if (strict) {
452 if (type == OBJ_BLOB) {
453 struct blob *blob = lookup_blob(sha1);
454 if (blob)
455 blob->object.flags |= FLAG_CHECKED;
456 else
457 die("invalid blob object %s", sha1_to_hex(sha1));
458 } else {
459 struct object *obj;
460 int eaten;
461 void *buf = (void *) data;
464 * we do not need to free the memory here, as the
465 * buf is deleted by the caller.
467 obj = parse_object_buffer(sha1, type, size, buf, &eaten);
468 if (!obj)
469 die("invalid %s", typename(type));
470 if (fsck_object(obj, 1, fsck_error_function))
471 die("Error in object");
472 if (fsck_walk(obj, mark_link, 0))
473 die("Not all child objects of %s are reachable", sha1_to_hex(obj->sha1));
475 if (obj->type == OBJ_TREE) {
476 struct tree *item = (struct tree *) obj;
477 item->buffer = NULL;
479 if (obj->type == OBJ_COMMIT) {
480 struct commit *commit = (struct commit *) obj;
481 commit->buffer = NULL;
483 obj->flags |= FLAG_CHECKED;
488 static void *get_base_data(struct base_data *c)
490 if (!c->data) {
491 struct object_entry *obj = c->obj;
493 if (obj->type == OBJ_REF_DELTA || obj->type == OBJ_OFS_DELTA) {
494 void *base = get_base_data(c->base);
495 void *raw = get_data_from_pack(obj);
496 c->data = patch_delta(
497 base, c->base->size,
498 raw, obj->size,
499 &c->size);
500 free(raw);
501 if (!c->data)
502 bad_object(obj->idx.offset, "failed to apply delta");
503 } else {
504 c->data = get_data_from_pack(obj);
505 c->size = obj->size;
508 base_cache_used += c->size;
509 prune_base_data(c);
511 return c->data;
514 static void resolve_delta(struct object_entry *delta_obj,
515 struct base_data *base, struct base_data *result)
517 void *base_data, *delta_data;
519 delta_obj->real_type = base->obj->real_type;
520 delta_data = get_data_from_pack(delta_obj);
521 base_data = get_base_data(base);
522 result->obj = delta_obj;
523 result->data = patch_delta(base_data, base->size,
524 delta_data, delta_obj->size, &result->size);
525 free(delta_data);
526 if (!result->data)
527 bad_object(delta_obj->idx.offset, "failed to apply delta");
528 sha1_object(result->data, result->size, delta_obj->real_type,
529 delta_obj->idx.sha1);
530 nr_resolved_deltas++;
533 static void find_unresolved_deltas(struct base_data *base,
534 struct base_data *prev_base)
536 int i, ref_first, ref_last, ofs_first, ofs_last;
539 * This is a recursive function. Those brackets should help reducing
540 * stack usage by limiting the scope of the delta_base union.
543 union delta_base base_spec;
545 hashcpy(base_spec.sha1, base->obj->idx.sha1);
546 find_delta_children(&base_spec, &ref_first, &ref_last);
548 memset(&base_spec, 0, sizeof(base_spec));
549 base_spec.offset = base->obj->idx.offset;
550 find_delta_children(&base_spec, &ofs_first, &ofs_last);
553 if (ref_last == -1 && ofs_last == -1)
554 return;
556 link_base_data(prev_base, base);
558 for (i = ref_first; i <= ref_last; i++) {
559 struct object_entry *child = objects + deltas[i].obj_no;
560 if (child->real_type == OBJ_REF_DELTA) {
561 struct base_data result;
562 resolve_delta(child, base, &result);
563 if (i == ref_last && ofs_last == -1)
564 free_base_data(base);
565 find_unresolved_deltas(&result, base);
569 for (i = ofs_first; i <= ofs_last; i++) {
570 struct object_entry *child = objects + deltas[i].obj_no;
571 if (child->real_type == OBJ_OFS_DELTA) {
572 struct base_data result;
573 resolve_delta(child, base, &result);
574 if (i == ofs_last)
575 free_base_data(base);
576 find_unresolved_deltas(&result, base);
580 unlink_base_data(base);
583 static int compare_delta_entry(const void *a, const void *b)
585 const struct delta_entry *delta_a = a;
586 const struct delta_entry *delta_b = b;
587 return memcmp(&delta_a->base, &delta_b->base, UNION_BASE_SZ);
590 /* Parse all objects and return the pack content SHA1 hash */
591 static void parse_pack_objects(unsigned char *sha1)
593 int i;
594 struct delta_entry *delta = deltas;
595 struct stat st;
598 * First pass:
599 * - find locations of all objects;
600 * - calculate SHA1 of all non-delta objects;
601 * - remember base (SHA1 or offset) for all deltas.
603 if (verbose)
604 progress = start_progress(
605 from_stdin ? "Receiving objects" : "Indexing objects",
606 nr_objects);
607 for (i = 0; i < nr_objects; i++) {
608 struct object_entry *obj = &objects[i];
609 void *data = unpack_raw_entry(obj, &delta->base);
610 obj->real_type = obj->type;
611 if (obj->type == OBJ_REF_DELTA || obj->type == OBJ_OFS_DELTA) {
612 nr_deltas++;
613 delta->obj_no = i;
614 delta++;
615 } else
616 sha1_object(data, obj->size, obj->type, obj->idx.sha1);
617 free(data);
618 display_progress(progress, i+1);
620 objects[i].idx.offset = consumed_bytes;
621 stop_progress(&progress);
623 /* Check pack integrity */
624 flush();
625 git_SHA1_Final(sha1, &input_ctx);
626 if (hashcmp(fill(20), sha1))
627 die("pack is corrupted (SHA1 mismatch)");
628 use(20);
630 /* If input_fd is a file, we should have reached its end now. */
631 if (fstat(input_fd, &st))
632 die("cannot fstat packfile: %s", strerror(errno));
633 if (S_ISREG(st.st_mode) &&
634 lseek(input_fd, 0, SEEK_CUR) - input_len != st.st_size)
635 die("pack has junk at the end");
637 if (!nr_deltas)
638 return;
640 /* Sort deltas by base SHA1/offset for fast searching */
641 qsort(deltas, nr_deltas, sizeof(struct delta_entry),
642 compare_delta_entry);
645 * Second pass:
646 * - for all non-delta objects, look if it is used as a base for
647 * deltas;
648 * - if used as a base, uncompress the object and apply all deltas,
649 * recursively checking if the resulting object is used as a base
650 * for some more deltas.
652 if (verbose)
653 progress = start_progress("Resolving deltas", nr_deltas);
654 for (i = 0; i < nr_objects; i++) {
655 struct object_entry *obj = &objects[i];
656 struct base_data base_obj;
658 if (obj->type == OBJ_REF_DELTA || obj->type == OBJ_OFS_DELTA)
659 continue;
660 base_obj.obj = obj;
661 base_obj.data = NULL;
662 find_unresolved_deltas(&base_obj, NULL);
663 display_progress(progress, nr_resolved_deltas);
667 static int write_compressed(struct sha1file *f, void *in, unsigned int size)
669 z_stream stream;
670 unsigned long maxsize;
671 void *out;
673 memset(&stream, 0, sizeof(stream));
674 deflateInit(&stream, zlib_compression_level);
675 maxsize = deflateBound(&stream, size);
676 out = xmalloc(maxsize);
678 /* Compress it */
679 stream.next_in = in;
680 stream.avail_in = size;
681 stream.next_out = out;
682 stream.avail_out = maxsize;
683 while (deflate(&stream, Z_FINISH) == Z_OK);
684 deflateEnd(&stream);
686 size = stream.total_out;
687 sha1write(f, out, size);
688 free(out);
689 return size;
692 static struct object_entry *append_obj_to_pack(struct sha1file *f,
693 const unsigned char *sha1, void *buf,
694 unsigned long size, enum object_type type)
696 struct object_entry *obj = &objects[nr_objects++];
697 unsigned char header[10];
698 unsigned long s = size;
699 int n = 0;
700 unsigned char c = (type << 4) | (s & 15);
701 s >>= 4;
702 while (s) {
703 header[n++] = c | 0x80;
704 c = s & 0x7f;
705 s >>= 7;
707 header[n++] = c;
708 crc32_begin(f);
709 sha1write(f, header, n);
710 obj[0].size = size;
711 obj[0].hdr_size = n;
712 obj[0].type = type;
713 obj[0].real_type = type;
714 obj[1].idx.offset = obj[0].idx.offset + n;
715 obj[1].idx.offset += write_compressed(f, buf, size);
716 obj[0].idx.crc32 = crc32_end(f);
717 sha1flush(f);
718 hashcpy(obj->idx.sha1, sha1);
719 return obj;
722 static int delta_pos_compare(const void *_a, const void *_b)
724 struct delta_entry *a = *(struct delta_entry **)_a;
725 struct delta_entry *b = *(struct delta_entry **)_b;
726 return a->obj_no - b->obj_no;
729 static void fix_unresolved_deltas(struct sha1file *f, int nr_unresolved)
731 struct delta_entry **sorted_by_pos;
732 int i, n = 0;
735 * Since many unresolved deltas may well be themselves base objects
736 * for more unresolved deltas, we really want to include the
737 * smallest number of base objects that would cover as much delta
738 * as possible by picking the
739 * trunc deltas first, allowing for other deltas to resolve without
740 * additional base objects. Since most base objects are to be found
741 * before deltas depending on them, a good heuristic is to start
742 * resolving deltas in the same order as their position in the pack.
744 sorted_by_pos = xmalloc(nr_unresolved * sizeof(*sorted_by_pos));
745 for (i = 0; i < nr_deltas; i++) {
746 if (objects[deltas[i].obj_no].real_type != OBJ_REF_DELTA)
747 continue;
748 sorted_by_pos[n++] = &deltas[i];
750 qsort(sorted_by_pos, n, sizeof(*sorted_by_pos), delta_pos_compare);
752 for (i = 0; i < n; i++) {
753 struct delta_entry *d = sorted_by_pos[i];
754 enum object_type type;
755 struct base_data base_obj;
757 if (objects[d->obj_no].real_type != OBJ_REF_DELTA)
758 continue;
759 base_obj.data = read_sha1_file(d->base.sha1, &type, &base_obj.size);
760 if (!base_obj.data)
761 continue;
763 if (check_sha1_signature(d->base.sha1, base_obj.data,
764 base_obj.size, typename(type)))
765 die("local object %s is corrupt", sha1_to_hex(d->base.sha1));
766 base_obj.obj = append_obj_to_pack(f, d->base.sha1,
767 base_obj.data, base_obj.size, type);
768 find_unresolved_deltas(&base_obj, NULL);
769 display_progress(progress, nr_resolved_deltas);
771 free(sorted_by_pos);
774 static void final(const char *final_pack_name, const char *curr_pack_name,
775 const char *final_index_name, const char *curr_index_name,
776 const char *keep_name, const char *keep_msg,
777 unsigned char *sha1)
779 const char *report = "pack";
780 char name[PATH_MAX];
781 int err;
783 if (!from_stdin) {
784 close(input_fd);
785 } else {
786 fsync_or_die(output_fd, curr_pack_name);
787 err = close(output_fd);
788 if (err)
789 die("error while closing pack file: %s", strerror(errno));
790 chmod(curr_pack_name, 0444);
793 if (keep_msg) {
794 int keep_fd, keep_msg_len = strlen(keep_msg);
795 if (!keep_name) {
796 snprintf(name, sizeof(name), "%s/pack/pack-%s.keep",
797 get_object_directory(), sha1_to_hex(sha1));
798 keep_name = name;
800 keep_fd = open(keep_name, O_RDWR|O_CREAT|O_EXCL, 0600);
801 if (keep_fd < 0) {
802 if (errno != EEXIST)
803 die("cannot write keep file");
804 } else {
805 if (keep_msg_len > 0) {
806 write_or_die(keep_fd, keep_msg, keep_msg_len);
807 write_or_die(keep_fd, "\n", 1);
809 if (close(keep_fd) != 0)
810 die("cannot write keep file");
811 report = "keep";
815 if (final_pack_name != curr_pack_name) {
816 if (!final_pack_name) {
817 snprintf(name, sizeof(name), "%s/pack/pack-%s.pack",
818 get_object_directory(), sha1_to_hex(sha1));
819 final_pack_name = name;
821 if (move_temp_to_file(curr_pack_name, final_pack_name))
822 die("cannot store pack file");
825 chmod(curr_index_name, 0444);
826 if (final_index_name != curr_index_name) {
827 if (!final_index_name) {
828 snprintf(name, sizeof(name), "%s/pack/pack-%s.idx",
829 get_object_directory(), sha1_to_hex(sha1));
830 final_index_name = name;
832 if (move_temp_to_file(curr_index_name, final_index_name))
833 die("cannot store index file");
836 if (!from_stdin) {
837 printf("%s\n", sha1_to_hex(sha1));
838 } else {
839 char buf[48];
840 int len = snprintf(buf, sizeof(buf), "%s\t%s\n",
841 report, sha1_to_hex(sha1));
842 write_or_die(1, buf, len);
845 * Let's just mimic git-unpack-objects here and write
846 * the last part of the input buffer to stdout.
848 while (input_len) {
849 err = xwrite(1, input_buffer + input_offset, input_len);
850 if (err <= 0)
851 break;
852 input_len -= err;
853 input_offset += err;
858 static int git_index_pack_config(const char *k, const char *v, void *cb)
860 if (!strcmp(k, "pack.indexversion")) {
861 pack_idx_default_version = git_config_int(k, v);
862 if (pack_idx_default_version > 2)
863 die("bad pack.indexversion=%"PRIu32,
864 pack_idx_default_version);
865 return 0;
867 return git_default_config(k, v, cb);
870 int main(int argc, char **argv)
872 int i, fix_thin_pack = 0;
873 char *curr_pack, *pack_name = NULL;
874 char *curr_index, *index_name = NULL;
875 const char *keep_name = NULL, *keep_msg = NULL;
876 char *index_name_buf = NULL, *keep_name_buf = NULL;
877 struct pack_idx_entry **idx_objects;
878 unsigned char pack_sha1[20];
879 int nongit = 0;
881 setup_git_directory_gently(&nongit);
882 git_config(git_index_pack_config, NULL);
884 for (i = 1; i < argc; i++) {
885 char *arg = argv[i];
887 if (*arg == '-') {
888 if (!strcmp(arg, "--stdin")) {
889 from_stdin = 1;
890 } else if (!strcmp(arg, "--fix-thin")) {
891 fix_thin_pack = 1;
892 } else if (!strcmp(arg, "--strict")) {
893 strict = 1;
894 } else if (!strcmp(arg, "--keep")) {
895 keep_msg = "";
896 } else if (!prefixcmp(arg, "--keep=")) {
897 keep_msg = arg + 7;
898 } else if (!prefixcmp(arg, "--pack_header=")) {
899 struct pack_header *hdr;
900 char *c;
902 hdr = (struct pack_header *)input_buffer;
903 hdr->hdr_signature = htonl(PACK_SIGNATURE);
904 hdr->hdr_version = htonl(strtoul(arg + 14, &c, 10));
905 if (*c != ',')
906 die("bad %s", arg);
907 hdr->hdr_entries = htonl(strtoul(c + 1, &c, 10));
908 if (*c)
909 die("bad %s", arg);
910 input_len = sizeof(*hdr);
911 } else if (!strcmp(arg, "-v")) {
912 verbose = 1;
913 } else if (!strcmp(arg, "-o")) {
914 if (index_name || (i+1) >= argc)
915 usage(index_pack_usage);
916 index_name = argv[++i];
917 } else if (!prefixcmp(arg, "--index-version=")) {
918 char *c;
919 pack_idx_default_version = strtoul(arg + 16, &c, 10);
920 if (pack_idx_default_version > 2)
921 die("bad %s", arg);
922 if (*c == ',')
923 pack_idx_off32_limit = strtoul(c+1, &c, 0);
924 if (*c || pack_idx_off32_limit & 0x80000000)
925 die("bad %s", arg);
926 } else
927 usage(index_pack_usage);
928 continue;
931 if (pack_name)
932 usage(index_pack_usage);
933 pack_name = arg;
936 if (!pack_name && !from_stdin)
937 usage(index_pack_usage);
938 if (fix_thin_pack && !from_stdin)
939 die("--fix-thin cannot be used without --stdin");
940 if (!index_name && pack_name) {
941 int len = strlen(pack_name);
942 if (!has_extension(pack_name, ".pack"))
943 die("packfile name '%s' does not end with '.pack'",
944 pack_name);
945 index_name_buf = xmalloc(len);
946 memcpy(index_name_buf, pack_name, len - 5);
947 strcpy(index_name_buf + len - 5, ".idx");
948 index_name = index_name_buf;
950 if (keep_msg && !keep_name && pack_name) {
951 int len = strlen(pack_name);
952 if (!has_extension(pack_name, ".pack"))
953 die("packfile name '%s' does not end with '.pack'",
954 pack_name);
955 keep_name_buf = xmalloc(len);
956 memcpy(keep_name_buf, pack_name, len - 5);
957 strcpy(keep_name_buf + len - 5, ".keep");
958 keep_name = keep_name_buf;
961 curr_pack = open_pack_file(pack_name);
962 parse_pack_header();
963 objects = xmalloc((nr_objects + 1) * sizeof(struct object_entry));
964 deltas = xmalloc(nr_objects * sizeof(struct delta_entry));
965 parse_pack_objects(pack_sha1);
966 if (nr_deltas == nr_resolved_deltas) {
967 stop_progress(&progress);
968 /* Flush remaining pack final 20-byte SHA1. */
969 flush();
970 } else {
971 if (fix_thin_pack) {
972 struct sha1file *f;
973 unsigned char read_sha1[20], tail_sha1[20];
974 char msg[48];
975 int nr_unresolved = nr_deltas - nr_resolved_deltas;
976 int nr_objects_initial = nr_objects;
977 if (nr_unresolved <= 0)
978 die("confusion beyond insanity");
979 objects = xrealloc(objects,
980 (nr_objects + nr_unresolved + 1)
981 * sizeof(*objects));
982 f = sha1fd(output_fd, curr_pack);
983 fix_unresolved_deltas(f, nr_unresolved);
984 sprintf(msg, "completed with %d local objects",
985 nr_objects - nr_objects_initial);
986 stop_progress_msg(&progress, msg);
987 sha1close(f, tail_sha1, 0);
988 hashcpy(read_sha1, pack_sha1);
989 fixup_pack_header_footer(output_fd, pack_sha1,
990 curr_pack, nr_objects,
991 read_sha1, consumed_bytes-20);
992 if (hashcmp(read_sha1, tail_sha1) != 0)
993 die("Unexpected tail checksum for %s "
994 "(disk corruption?)", curr_pack);
996 if (nr_deltas != nr_resolved_deltas)
997 die("pack has %d unresolved deltas",
998 nr_deltas - nr_resolved_deltas);
1000 free(deltas);
1001 if (strict)
1002 check_objects();
1004 idx_objects = xmalloc((nr_objects) * sizeof(struct pack_idx_entry *));
1005 for (i = 0; i < nr_objects; i++)
1006 idx_objects[i] = &objects[i].idx;
1007 curr_index = write_idx_file(index_name, idx_objects, nr_objects, pack_sha1);
1008 free(idx_objects);
1010 final(pack_name, curr_pack,
1011 index_name, curr_index,
1012 keep_name, keep_msg,
1013 pack_sha1);
1014 free(objects);
1015 free(index_name_buf);
1016 free(keep_name_buf);
1017 if (pack_name == NULL)
1018 free(curr_pack);
1019 if (index_name == NULL)
1020 free(curr_index);
1022 return 0;