1 #include "cache.h"
2 #include "object.h"
3 #include "delta.h"
4 #include "pack.h"
5 #include "csum-file.h"
6 #include <sys/time.h>
7 #include <signal.h>
9 static const char pack_usage[] = "git-pack-objects [-q] [--no-reuse-delta] [--non-empty] [--local] [--incremental] [--window=N] [--depth=N] {--stdout | base-name} < object-list";
11 struct object_entry {
12 unsigned char sha1[20];
13 unsigned long size; /* uncompressed size */
14 unsigned long offset; /* offset into the final pack file;
15 * nonzero if already written.
16 */
17 unsigned int depth; /* delta depth */
18 unsigned int delta_limit; /* base adjustment for in-pack delta */
19 unsigned int hash; /* name hint hash */
20 enum object_type type;
21 enum object_type in_pack_type; /* could be delta */
22 unsigned long delta_size; /* delta data size (uncompressed) */
23 struct object_entry *delta; /* delta base object */
24 struct packed_git *in_pack; /* already in pack */
25 unsigned int in_pack_offset;
26 struct object_entry *delta_child; /* delitified objects who bases me */
27 struct object_entry *delta_sibling; /* other deltified objects who
28 * uses the same base as me
29 */
30 };
32 /*
33 * Objects we are going to pack are colected in objects array (dynamically
34 * expanded). nr_objects & nr_alloc controls this array. They are stored
35 * in the order we see -- typically rev-list --objects order that gives us
36 * nice "minimum seek" order.
37 *
38 * sorted-by-sha ans sorted-by-type are arrays of pointers that point at
39 * elements in the objects array. The former is used to build the pack
40 * index (lists object names in the ascending order to help offset lookup),
41 * and the latter is used to group similar things together by try_delta()
42 * heuristics.
43 */
45 static unsigned char object_list_sha1[20];
46 static int non_empty = 0;
47 static int no_reuse_delta = 0;
48 static int local = 0;
49 static int incremental = 0;
50 static struct object_entry **sorted_by_sha, **sorted_by_type;
51 static struct object_entry *objects = NULL;
52 static int nr_objects = 0, nr_alloc = 0;
53 static const char *base_name;
54 static unsigned char pack_file_sha1[20];
55 static int progress = 1;
56 static volatile int progress_update = 0;
58 /*
59 * The object names in objects array are hashed with this hashtable,
60 * to help looking up the entry by object name. Binary search from
61 * sorted_by_sha is also possible but this was easier to code and faster.
62 * This hashtable is built after all the objects are seen.
63 */
64 static int *object_ix = NULL;
65 static int object_ix_hashsz = 0;
67 /*
68 * Pack index for existing packs give us easy access to the offsets into
69 * corresponding pack file where each object's data starts, but the entries
70 * do not store the size of the compressed representation (uncompressed
71 * size is easily available by examining the pack entry header). We build
72 * a hashtable of existing packs (pack_revindex), and keep reverse index
73 * here -- pack index file is sorted by object name mapping to offset; this
74 * pack_revindex[].revindex array is an ordered list of offsets, so if you
75 * know the offset of an object, next offset is where its packed
76 * representation ends.
77 */
78 struct pack_revindex {
79 struct packed_git *p;
80 unsigned long *revindex;
81 } *pack_revindex = NULL;
82 static int pack_revindex_hashsz = 0;
84 /*
85 * stats
86 */
87 static int written = 0;
88 static int written_delta = 0;
89 static int reused = 0;
90 static int reused_delta = 0;
92 static int pack_revindex_ix(struct packed_git *p)
93 {
94 unsigned int ui = (unsigned int) p;
95 int i;
97 ui = ui ^ (ui >> 16); /* defeat structure alignment */
98 i = (int)(ui % pack_revindex_hashsz);
99 while (pack_revindex[i].p) {
100 if (pack_revindex[i].p == p)
101 return i;
102 if (++i == pack_revindex_hashsz)
103 i = 0;
104 }
105 return -1 - i;
106 }
108 static void prepare_pack_ix(void)
109 {
110 int num;
111 struct packed_git *p;
112 for (num = 0, p = packed_git; p; p = p->next)
113 num++;
114 if (!num)
115 return;
116 pack_revindex_hashsz = num * 11;
117 pack_revindex = xcalloc(sizeof(*pack_revindex), pack_revindex_hashsz);
118 for (p = packed_git; p; p = p->next) {
119 num = pack_revindex_ix(p);
120 num = - 1 - num;
121 pack_revindex[num].p = p;
122 }
123 /* revindex elements are lazily initialized */
124 }
126 static int cmp_offset(const void *a_, const void *b_)
127 {
128 unsigned long a = *(unsigned long *) a_;
129 unsigned long b = *(unsigned long *) b_;
130 if (a < b)
131 return -1;
132 else if (a == b)
133 return 0;
134 else
135 return 1;
136 }
138 /*
139 * Ordered list of offsets of objects in the pack.
140 */
141 static void prepare_pack_revindex(struct pack_revindex *rix)
142 {
143 struct packed_git *p = rix->p;
144 int num_ent = num_packed_objects(p);
145 int i;
146 void *index = p->index_base + 256;
148 rix->revindex = xmalloc(sizeof(unsigned long) * (num_ent + 1));
149 for (i = 0; i < num_ent; i++) {
150 long hl = *((long *)(index + 24 * i));
151 rix->revindex[i] = ntohl(hl);
152 }
153 /* This knows the pack format -- the 20-byte trailer
154 * follows immediately after the last object data.
155 */
156 rix->revindex[num_ent] = p->pack_size - 20;
157 qsort(rix->revindex, num_ent, sizeof(unsigned long), cmp_offset);
158 }
160 static unsigned long find_packed_object_size(struct packed_git *p,
161 unsigned long ofs)
162 {
163 int num;
164 int lo, hi;
165 struct pack_revindex *rix;
166 unsigned long *revindex;
167 num = pack_revindex_ix(p);
168 if (num < 0)
169 die("internal error: pack revindex uninitialized");
170 rix = &pack_revindex[num];
171 if (!rix->revindex)
172 prepare_pack_revindex(rix);
173 revindex = rix->revindex;
174 lo = 0;
175 hi = num_packed_objects(p) + 1;
176 do {
177 int mi = (lo + hi) / 2;
178 if (revindex[mi] == ofs) {
179 return revindex[mi+1] - ofs;
180 }
181 else if (ofs < revindex[mi])
182 hi = mi;
183 else
184 lo = mi + 1;
185 } while (lo < hi);
186 die("internal error: pack revindex corrupt");
187 }
189 static void *delta_against(void *buf, unsigned long size, struct object_entry *entry)
190 {
191 unsigned long othersize, delta_size;
192 char type[10];
193 void *otherbuf = read_sha1_file(entry->delta->sha1, type, &othersize);
194 void *delta_buf;
196 if (!otherbuf)
197 die("unable to read %s", sha1_to_hex(entry->delta->sha1));
198 delta_buf = diff_delta(otherbuf, othersize,
199 buf, size, &delta_size, 0);
200 if (!delta_buf || delta_size != entry->delta_size)
201 die("delta size changed");
202 free(buf);
203 free(otherbuf);
204 return delta_buf;
205 }
207 /*
208 * The per-object header is a pretty dense thing, which is
209 * - first byte: low four bits are "size", then three bits of "type",
210 * and the high bit is "size continues".
211 * - each byte afterwards: low seven bits are size continuation,
212 * with the high bit being "size continues"
213 */
214 static int encode_header(enum object_type type, unsigned long size, unsigned char *hdr)
215 {
216 int n = 1;
217 unsigned char c;
219 if (type < OBJ_COMMIT || type > OBJ_DELTA)
220 die("bad type %d", type);
222 c = (type << 4) | (size & 15);
223 size >>= 4;
224 while (size) {
225 *hdr++ = c | 0x80;
226 c = size & 0x7f;
227 size >>= 7;
228 n++;
229 }
230 *hdr = c;
231 return n;
232 }
234 static unsigned long write_object(struct sha1file *f, struct object_entry *entry)
235 {
236 unsigned long size;
237 char type[10];
238 void *buf;
239 unsigned char header[10];
240 unsigned hdrlen, datalen;
241 enum object_type obj_type;
242 int to_reuse = 0;
244 obj_type = entry->type;
245 if (! entry->in_pack)
246 to_reuse = 0; /* can't reuse what we don't have */
247 else if (obj_type == OBJ_DELTA)
248 to_reuse = 1; /* check_object() decided it for us */
249 else if (obj_type != entry->in_pack_type)
250 to_reuse = 0; /* pack has delta which is unusable */
251 else if (entry->delta)
252 to_reuse = 0; /* we want to pack afresh */
253 else
254 to_reuse = 1; /* we have it in-pack undeltified,
255 * and we do not need to deltify it.
256 */
258 if (! to_reuse) {
259 buf = read_sha1_file(entry->sha1, type, &size);
260 if (!buf)
261 die("unable to read %s", sha1_to_hex(entry->sha1));
262 if (size != entry->size)
263 die("object %s size inconsistency (%lu vs %lu)",
264 sha1_to_hex(entry->sha1), size, entry->size);
265 if (entry->delta) {
266 buf = delta_against(buf, size, entry);
267 size = entry->delta_size;
268 obj_type = OBJ_DELTA;
269 }
270 /*
271 * The object header is a byte of 'type' followed by zero or
272 * more bytes of length. For deltas, the 20 bytes of delta
273 * sha1 follows that.
274 */
275 hdrlen = encode_header(obj_type, size, header);
276 sha1write(f, header, hdrlen);
278 if (entry->delta) {
279 sha1write(f, entry->delta, 20);
280 hdrlen += 20;
281 }
282 datalen = sha1write_compressed(f, buf, size);
283 free(buf);
284 }
285 else {
286 struct packed_git *p = entry->in_pack;
287 use_packed_git(p);
289 datalen = find_packed_object_size(p, entry->in_pack_offset);
290 buf = p->pack_base + entry->in_pack_offset;
291 sha1write(f, buf, datalen);
292 unuse_packed_git(p);
293 hdrlen = 0; /* not really */
294 if (obj_type == OBJ_DELTA)
295 reused_delta++;
296 reused++;
297 }
298 if (obj_type == OBJ_DELTA)
299 written_delta++;
300 written++;
301 return hdrlen + datalen;
302 }
304 static unsigned long write_one(struct sha1file *f,
305 struct object_entry *e,
306 unsigned long offset)
307 {
308 if (e->offset)
309 /* offset starts from header size and cannot be zero
310 * if it is written already.
311 */
312 return offset;
313 e->offset = offset;
314 offset += write_object(f, e);
315 /* if we are deltified, write out its base object. */
316 if (e->delta)
317 offset = write_one(f, e->delta, offset);
318 return offset;
319 }
321 static void write_pack_file(void)
322 {
323 int i;
324 struct sha1file *f;
325 unsigned long offset;
326 struct pack_header hdr;
327 unsigned last_percent = 999;
328 int do_progress = 0;
330 if (!base_name)
331 f = sha1fd(1, "<stdout>");
332 else {
333 f = sha1create("%s-%s.%s", base_name,
334 sha1_to_hex(object_list_sha1), "pack");
335 do_progress = progress;
336 }
337 if (do_progress)
338 fprintf(stderr, "Writing %d objects.\n", nr_objects);
340 hdr.hdr_signature = htonl(PACK_SIGNATURE);
341 hdr.hdr_version = htonl(PACK_VERSION);
342 hdr.hdr_entries = htonl(nr_objects);
343 sha1write(f, &hdr, sizeof(hdr));
344 offset = sizeof(hdr);
345 for (i = 0; i < nr_objects; i++) {
346 offset = write_one(f, objects + i, offset);
347 if (do_progress) {
348 unsigned percent = written * 100 / nr_objects;
349 if (progress_update || percent != last_percent) {
350 fprintf(stderr, "%4u%% (%u/%u) done\r",
351 percent, written, nr_objects);
352 progress_update = 0;
353 last_percent = percent;
354 }
355 }
356 }
357 if (do_progress)
358 fputc('\n', stderr);
360 sha1close(f, pack_file_sha1, 1);
361 }
363 static void write_index_file(void)
364 {
365 int i;
366 struct sha1file *f = sha1create("%s-%s.%s", base_name, sha1_to_hex(object_list_sha1), "idx");
367 struct object_entry **list = sorted_by_sha;
368 struct object_entry **last = list + nr_objects;
369 unsigned int array[256];
371 /*
372 * Write the first-level table (the list is sorted,
373 * but we use a 256-entry lookup to be able to avoid
374 * having to do eight extra binary search iterations).
375 */
376 for (i = 0; i < 256; i++) {
377 struct object_entry **next = list;
378 while (next < last) {
379 struct object_entry *entry = *next;
380 if (entry->sha1[0] != i)
381 break;
382 next++;
383 }
384 array[i] = htonl(next - sorted_by_sha);
385 list = next;
386 }
387 sha1write(f, array, 256 * sizeof(int));
389 /*
390 * Write the actual SHA1 entries..
391 */
392 list = sorted_by_sha;
393 for (i = 0; i < nr_objects; i++) {
394 struct object_entry *entry = *list++;
395 unsigned int offset = htonl(entry->offset);
396 sha1write(f, &offset, 4);
397 sha1write(f, entry->sha1, 20);
398 }
399 sha1write(f, pack_file_sha1, 20);
400 sha1close(f, NULL, 1);
401 }
403 static int add_object_entry(unsigned char *sha1, unsigned int hash)
404 {
405 unsigned int idx = nr_objects;
406 struct object_entry *entry;
407 struct packed_git *p;
408 unsigned int found_offset = 0;
409 struct packed_git *found_pack = NULL;
411 for (p = packed_git; p; p = p->next) {
412 struct pack_entry e;
413 if (find_pack_entry_one(sha1, &e, p)) {
414 if (incremental)
415 return 0;
416 if (local && !p->pack_local)
417 return 0;
418 if (!found_pack) {
419 found_offset = e.offset;
420 found_pack = e.p;
421 }
422 }
423 }
425 if (idx >= nr_alloc) {
426 unsigned int needed = (idx + 1024) * 3 / 2;
427 objects = xrealloc(objects, needed * sizeof(*entry));
428 nr_alloc = needed;
429 }
430 entry = objects + idx;
431 memset(entry, 0, sizeof(*entry));
432 memcpy(entry->sha1, sha1, 20);
433 entry->hash = hash;
434 if (found_pack) {
435 entry->in_pack = found_pack;
436 entry->in_pack_offset = found_offset;
437 }
438 nr_objects = idx+1;
439 return 1;
440 }
442 static int locate_object_entry_hash(unsigned char *sha1)
443 {
444 int i;
445 unsigned int ui;
446 memcpy(&ui, sha1, sizeof(unsigned int));
447 i = ui % object_ix_hashsz;
448 while (0 < object_ix[i]) {
449 if (!memcmp(sha1, objects[object_ix[i]-1].sha1, 20))
450 return i;
451 if (++i == object_ix_hashsz)
452 i = 0;
453 }
454 return -1 - i;
455 }
457 static struct object_entry *locate_object_entry(unsigned char *sha1)
458 {
459 int i = locate_object_entry_hash(sha1);
460 if (0 <= i)
461 return &objects[object_ix[i]-1];
462 return NULL;
463 }
465 static void check_object(struct object_entry *entry)
466 {
467 char type[20];
469 if (entry->in_pack) {
470 unsigned char base[20];
471 unsigned long size;
472 struct object_entry *base_entry;
474 /* We want in_pack_type even if we do not reuse delta.
475 * There is no point not reusing non-delta representations.
476 */
477 check_reuse_pack_delta(entry->in_pack,
478 entry->in_pack_offset,
479 base, &size,
480 &entry->in_pack_type);
482 /* Check if it is delta, and the base is also an object
483 * we are going to pack. If so we will reuse the existing
484 * delta.
485 */
486 if (!no_reuse_delta &&
487 entry->in_pack_type == OBJ_DELTA &&
488 (base_entry = locate_object_entry(base))) {
490 /* Depth value does not matter - find_deltas()
491 * will never consider reused delta as the
492 * base object to deltify other objects
493 * against, in order to avoid circular deltas.
494 */
496 /* uncompressed size of the delta data */
497 entry->size = entry->delta_size = size;
498 entry->delta = base_entry;
499 entry->type = OBJ_DELTA;
501 entry->delta_sibling = base_entry->delta_child;
502 base_entry->delta_child = entry;
504 return;
505 }
506 /* Otherwise we would do the usual */
507 }
509 if (sha1_object_info(entry->sha1, type, &entry->size))
510 die("unable to get type of object %s",
511 sha1_to_hex(entry->sha1));
513 if (!strcmp(type, "commit")) {
514 entry->type = OBJ_COMMIT;
515 } else if (!strcmp(type, "tree")) {
516 entry->type = OBJ_TREE;
517 } else if (!strcmp(type, "blob")) {
518 entry->type = OBJ_BLOB;
519 } else if (!strcmp(type, "tag")) {
520 entry->type = OBJ_TAG;
521 } else
522 die("unable to pack object %s of type %s",
523 sha1_to_hex(entry->sha1), type);
524 }
526 static void hash_objects(void)
527 {
528 int i;
529 struct object_entry *oe;
531 object_ix_hashsz = nr_objects * 2;
532 object_ix = xcalloc(sizeof(int), object_ix_hashsz);
533 for (i = 0, oe = objects; i < nr_objects; i++, oe++) {
534 int ix = locate_object_entry_hash(oe->sha1);
535 if (0 <= ix) {
536 error("the same object '%s' added twice",
537 sha1_to_hex(oe->sha1));
538 continue;
539 }
540 ix = -1 - ix;
541 object_ix[ix] = i + 1;
542 }
543 }
545 static unsigned int check_delta_limit(struct object_entry *me, unsigned int n)
546 {
547 struct object_entry *child = me->delta_child;
548 unsigned int m = n;
549 while (child) {
550 unsigned int c = check_delta_limit(child, n + 1);
551 if (m < c)
552 m = c;
553 child = child->delta_sibling;
554 }
555 return m;
556 }
558 static void get_object_details(void)
559 {
560 int i;
561 struct object_entry *entry;
563 hash_objects();
564 prepare_pack_ix();
565 for (i = 0, entry = objects; i < nr_objects; i++, entry++)
566 check_object(entry);
567 for (i = 0, entry = objects; i < nr_objects; i++, entry++)
568 if (!entry->delta && entry->delta_child)
569 entry->delta_limit =
570 check_delta_limit(entry, 1);
571 }
573 typedef int (*entry_sort_t)(const struct object_entry *, const struct object_entry *);
575 static entry_sort_t current_sort;
577 static int sort_comparator(const void *_a, const void *_b)
578 {
579 struct object_entry *a = *(struct object_entry **)_a;
580 struct object_entry *b = *(struct object_entry **)_b;
581 return current_sort(a,b);
582 }
584 static struct object_entry **create_sorted_list(entry_sort_t sort)
585 {
586 struct object_entry **list = xmalloc(nr_objects * sizeof(struct object_entry *));
587 int i;
589 for (i = 0; i < nr_objects; i++)
590 list[i] = objects + i;
591 current_sort = sort;
592 qsort(list, nr_objects, sizeof(struct object_entry *), sort_comparator);
593 return list;
594 }
596 static int sha1_sort(const struct object_entry *a, const struct object_entry *b)
597 {
598 return memcmp(a->sha1, b->sha1, 20);
599 }
601 static int type_size_sort(const struct object_entry *a, const struct object_entry *b)
602 {
603 if (a->type < b->type)
604 return -1;
605 if (a->type > b->type)
606 return 1;
607 if (a->hash < b->hash)
608 return -1;
609 if (a->hash > b->hash)
610 return 1;
611 if (a->size < b->size)
612 return -1;
613 if (a->size > b->size)
614 return 1;
615 return a < b ? -1 : (a > b);
616 }
618 struct unpacked {
619 struct object_entry *entry;
620 void *data;
621 };
623 /*
624 * We search for deltas _backwards_ in a list sorted by type and
625 * by size, so that we see progressively smaller and smaller files.
626 * That's because we prefer deltas to be from the bigger file
627 * to the smaller - deletes are potentially cheaper, but perhaps
628 * more importantly, the bigger file is likely the more recent
629 * one.
630 */
631 static int try_delta(struct unpacked *cur, struct unpacked *old, unsigned max_depth)
632 {
633 struct object_entry *cur_entry = cur->entry;
634 struct object_entry *old_entry = old->entry;
635 unsigned long size, oldsize, delta_size, sizediff;
636 long max_size;
637 void *delta_buf;
639 /* Don't bother doing diffs between different types */
640 if (cur_entry->type != old_entry->type)
641 return -1;
643 /* If the current object is at edge, take the depth the objects
644 * that depend on the current object into account -- otherwise
645 * they would become too deep.
646 */
647 if (cur_entry->delta_child) {
648 if (max_depth <= cur_entry->delta_limit)
649 return 0;
650 max_depth -= cur_entry->delta_limit;
651 }
653 size = cur_entry->size;
654 if (size < 50)
655 return -1;
656 oldsize = old_entry->size;
657 sizediff = oldsize > size ? oldsize - size : size - oldsize;
658 if (sizediff > size / 8)
659 return -1;
660 if (old_entry->depth >= max_depth)
661 return 0;
663 /*
664 * NOTE!
665 *
666 * We always delta from the bigger to the smaller, since that's
667 * more space-efficient (deletes don't have to say _what_ they
668 * delete).
669 */
670 max_size = size / 2 - 20;
671 if (cur_entry->delta)
672 max_size = cur_entry->delta_size-1;
673 if (sizediff >= max_size)
674 return -1;
675 delta_buf = diff_delta(old->data, oldsize,
676 cur->data, size, &delta_size, max_size);
677 if (!delta_buf)
678 return 0;
679 cur_entry->delta = old_entry;
680 cur_entry->delta_size = delta_size;
681 cur_entry->depth = old_entry->depth + 1;
682 free(delta_buf);
683 return 0;
684 }
686 static void progress_interval(int signum)
687 {
688 signal(SIGALRM, progress_interval);
689 progress_update = 1;
690 }
692 static void find_deltas(struct object_entry **list, int window, int depth)
693 {
694 int i, idx;
695 unsigned int array_size = window * sizeof(struct unpacked);
696 struct unpacked *array = xmalloc(array_size);
697 unsigned processed = 0;
698 unsigned last_percent = 999;
700 memset(array, 0, array_size);
701 i = nr_objects;
702 idx = 0;
703 if (progress)
704 fprintf(stderr, "Deltifying %d objects.\n", nr_objects);
706 while (--i >= 0) {
707 struct object_entry *entry = list[i];
708 struct unpacked *n = array + idx;
709 unsigned long size;
710 char type[10];
711 int j;
713 processed++;
714 if (progress) {
715 unsigned percent = processed * 100 / nr_objects;
716 if (percent != last_percent || progress_update) {
717 fprintf(stderr, "%4u%% (%u/%u) done\r",
718 percent, processed, nr_objects);
719 progress_update = 0;
720 last_percent = percent;
721 }
722 }
724 if (entry->delta)
725 /* This happens if we decided to reuse existing
726 * delta from a pack. "!no_reuse_delta &&" is implied.
727 */
728 continue;
730 free(n->data);
731 n->entry = entry;
732 n->data = read_sha1_file(entry->sha1, type, &size);
733 if (size != entry->size)
734 die("object %s inconsistent object length (%lu vs %lu)", sha1_to_hex(entry->sha1), size, entry->size);
736 j = window;
737 while (--j > 0) {
738 unsigned int other_idx = idx + j;
739 struct unpacked *m;
740 if (other_idx >= window)
741 other_idx -= window;
742 m = array + other_idx;
743 if (!m->entry)
744 break;
745 if (try_delta(n, m, depth) < 0)
746 break;
747 }
748 idx++;
749 if (idx >= window)
750 idx = 0;
751 }
753 if (progress)
754 fputc('\n', stderr);
756 for (i = 0; i < window; ++i)
757 free(array[i].data);
758 free(array);
759 }
761 static void prepare_pack(int window, int depth)
762 {
763 get_object_details();
764 sorted_by_type = create_sorted_list(type_size_sort);
765 if (window && depth)
766 find_deltas(sorted_by_type, window+1, depth);
767 }
769 static int reuse_cached_pack(unsigned char *sha1, int pack_to_stdout)
770 {
771 static const char cache[] = "pack-cache/pack-%s.%s";
772 char *cached_pack, *cached_idx;
773 int ifd, ofd, ifd_ix = -1;
775 cached_pack = git_path(cache, sha1_to_hex(sha1), "pack");
776 ifd = open(cached_pack, O_RDONLY);
777 if (ifd < 0)
778 return 0;
780 if (!pack_to_stdout) {
781 cached_idx = git_path(cache, sha1_to_hex(sha1), "idx");
782 ifd_ix = open(cached_idx, O_RDONLY);
783 if (ifd_ix < 0) {
784 close(ifd);
785 return 0;
786 }
787 }
789 if (progress)
790 fprintf(stderr, "Reusing %d objects pack %s\n", nr_objects,
791 sha1_to_hex(sha1));
793 if (pack_to_stdout) {
794 if (copy_fd(ifd, 1))
795 exit(1);
796 close(ifd);
797 }
798 else {
799 char name[PATH_MAX];
800 snprintf(name, sizeof(name),
801 "%s-%s.%s", base_name, sha1_to_hex(sha1), "pack");
802 ofd = open(name, O_CREAT | O_EXCL | O_WRONLY, 0666);
803 if (ofd < 0)
804 die("unable to open %s (%s)", name, strerror(errno));
805 if (copy_fd(ifd, ofd))
806 exit(1);
807 close(ifd);
809 snprintf(name, sizeof(name),
810 "%s-%s.%s", base_name, sha1_to_hex(sha1), "idx");
811 ofd = open(name, O_CREAT | O_EXCL | O_WRONLY, 0666);
812 if (ofd < 0)
813 die("unable to open %s (%s)", name, strerror(errno));
814 if (copy_fd(ifd_ix, ofd))
815 exit(1);
816 close(ifd_ix);
817 puts(sha1_to_hex(sha1));
818 }
820 return 1;
821 }
823 int main(int argc, char **argv)
824 {
825 SHA_CTX ctx;
826 char line[PATH_MAX + 20];
827 int window = 10, depth = 10, pack_to_stdout = 0;
828 struct object_entry **list;
829 int i;
831 setup_git_directory();
833 for (i = 1; i < argc; i++) {
834 const char *arg = argv[i];
836 if (*arg == '-') {
837 if (!strcmp("--non-empty", arg)) {
838 non_empty = 1;
839 continue;
840 }
841 if (!strcmp("--local", arg)) {
842 local = 1;
843 continue;
844 }
845 if (!strcmp("--incremental", arg)) {
846 incremental = 1;
847 continue;
848 }
849 if (!strncmp("--window=", arg, 9)) {
850 char *end;
851 window = strtoul(arg+9, &end, 0);
852 if (!arg[9] || *end)
853 usage(pack_usage);
854 continue;
855 }
856 if (!strncmp("--depth=", arg, 8)) {
857 char *end;
858 depth = strtoul(arg+8, &end, 0);
859 if (!arg[8] || *end)
860 usage(pack_usage);
861 continue;
862 }
863 if (!strcmp("-q", arg)) {
864 progress = 0;
865 continue;
866 }
867 if (!strcmp("--no-reuse-delta", arg)) {
868 no_reuse_delta = 1;
869 continue;
870 }
871 if (!strcmp("--stdout", arg)) {
872 pack_to_stdout = 1;
873 continue;
874 }
875 usage(pack_usage);
876 }
877 if (base_name)
878 usage(pack_usage);
879 base_name = arg;
880 }
882 if (pack_to_stdout != !base_name)
883 usage(pack_usage);
885 prepare_packed_git();
887 if (progress) {
888 struct itimerval v;
889 v.it_interval.tv_sec = 1;
890 v.it_interval.tv_usec = 0;
891 v.it_value = v.it_interval;
892 signal(SIGALRM, progress_interval);
893 setitimer(ITIMER_REAL, &v, NULL);
894 fprintf(stderr, "Generating pack...\n");
895 }
897 while (fgets(line, sizeof(line), stdin) != NULL) {
898 unsigned int hash;
899 char *p;
900 unsigned char sha1[20];
902 if (progress_update) {
903 fprintf(stderr, "Counting objects...%d\r", nr_objects);
904 progress_update = 0;
905 }
906 if (get_sha1_hex(line, sha1))
907 die("expected sha1, got garbage:\n %s", line);
908 hash = 0;
909 p = line+40;
910 while (*p) {
911 unsigned char c = *p++;
912 if (isspace(c))
913 continue;
914 hash = hash * 11 + c;
915 }
916 add_object_entry(sha1, hash);
917 }
918 if (progress)
919 fprintf(stderr, "Done counting %d objects.\n", nr_objects);
920 if (non_empty && !nr_objects)
921 return 0;
923 sorted_by_sha = create_sorted_list(sha1_sort);
924 SHA1_Init(&ctx);
925 list = sorted_by_sha;
926 for (i = 0; i < nr_objects; i++) {
927 struct object_entry *entry = *list++;
928 SHA1_Update(&ctx, entry->sha1, 20);
929 }
930 SHA1_Final(object_list_sha1, &ctx);
932 if (reuse_cached_pack(object_list_sha1, pack_to_stdout))
933 ;
934 else {
935 prepare_pack(window, depth);
936 if (progress && pack_to_stdout) {
937 /* the other end usually displays progress itself */
938 struct itimerval v = {{0,},};
939 setitimer(ITIMER_REAL, &v, NULL);
940 signal(SIGALRM, SIG_IGN );
941 progress_update = 0;
942 }
943 write_pack_file();
944 if (!pack_to_stdout) {
945 write_index_file();
946 puts(sha1_to_hex(object_list_sha1));
947 }
948 }
949 if (progress)
950 fprintf(stderr, "Total %d, written %d (delta %d), reused %d (delta %d)\n",
951 nr_objects, written, written_delta, reused, reused_delta);
952 return 0;
953 }