c72e07a2bb73b74e6be9a038fae4eafd0b88e2b2
1#include "builtin.h"
2#include "cache.h"
3#include "object.h"
4#include "blob.h"
5#include "commit.h"
6#include "tag.h"
7#include "tree.h"
8#include "delta.h"
9#include "pack.h"
10#include "csum-file.h"
11#include "tree-walk.h"
12#include "diff.h"
13#include "revision.h"
14#include "list-objects.h"
15
16static const char pack_usage[] = "\
17git-pack-objects [{ -q | --progress | --all-progress }] \n\
18 [--local] [--incremental] [--window=N] [--depth=N] \n\
19 [--no-reuse-delta] [--delta-base-offset] [--non-empty] \n\
20 [--revs [--unpacked | --all]*] [--reflog] [--stdout | base-name] \n\
21 [<ref-list | <object-list]";
22
23struct object_entry {
24 unsigned char sha1[20];
25 uint32_t crc32; /* crc of raw pack data for this object */
26 off_t offset; /* offset into the final pack file */
27 unsigned long size; /* uncompressed size */
28 unsigned int hash; /* name hint hash */
29 unsigned int depth; /* delta depth */
30 struct packed_git *in_pack; /* already in pack */
31 off_t in_pack_offset;
32 struct object_entry *delta; /* delta base object */
33 struct object_entry *delta_child; /* deltified objects who bases me */
34 struct object_entry *delta_sibling; /* other deltified objects who
35 * uses the same base as me
36 */
37 unsigned long delta_size; /* delta data size (uncompressed) */
38 enum object_type type;
39 enum object_type in_pack_type; /* could be delta */
40 unsigned char in_pack_header_size;
41 unsigned char preferred_base; /* we do not pack this, but is available
42 * to be used as the base objectto delta
43 * objects against.
44 */
45};
46
47/*
48 * Objects we are going to pack are collected in objects array (dynamically
49 * expanded). nr_objects & nr_alloc controls this array. They are stored
50 * in the order we see -- typically rev-list --objects order that gives us
51 * nice "minimum seek" order.
52 */
53static struct object_entry *objects;
54static uint32_t nr_objects, nr_alloc, nr_result;
55
56static int non_empty;
57static int no_reuse_delta;
58static int local;
59static int incremental;
60static int allow_ofs_delta;
61static const char *pack_tmp_name, *idx_tmp_name;
62static char tmpname[PATH_MAX];
63static unsigned char pack_file_sha1[20];
64static int progress = 1;
65static volatile sig_atomic_t progress_update;
66static int window = 10;
67static int pack_to_stdout;
68static int num_preferred_base;
69
70/*
71 * The object names in objects array are hashed with this hashtable,
72 * to help looking up the entry by object name.
73 * This hashtable is built after all the objects are seen.
74 */
75static int *object_ix;
76static int object_ix_hashsz;
77
78/*
79 * Pack index for existing packs give us easy access to the offsets into
80 * corresponding pack file where each object's data starts, but the entries
81 * do not store the size of the compressed representation (uncompressed
82 * size is easily available by examining the pack entry header). It is
83 * also rather expensive to find the sha1 for an object given its offset.
84 *
85 * We build a hashtable of existing packs (pack_revindex), and keep reverse
86 * index here -- pack index file is sorted by object name mapping to offset;
87 * this pack_revindex[].revindex array is a list of offset/index_nr pairs
88 * ordered by offset, so if you know the offset of an object, next offset
89 * is where its packed representation ends and the index_nr can be used to
90 * get the object sha1 from the main index.
91 */
92struct revindex_entry {
93 off_t offset;
94 unsigned int nr;
95};
96struct pack_revindex {
97 struct packed_git *p;
98 struct revindex_entry *revindex;
99};
100static struct pack_revindex *pack_revindex;
101static int pack_revindex_hashsz;
102
103/*
104 * stats
105 */
106static uint32_t written, written_delta;
107static uint32_t reused, reused_delta;
108
109static int pack_revindex_ix(struct packed_git *p)
110{
111 unsigned long ui = (unsigned long)p;
112 int i;
113
114 ui = ui ^ (ui >> 16); /* defeat structure alignment */
115 i = (int)(ui % pack_revindex_hashsz);
116 while (pack_revindex[i].p) {
117 if (pack_revindex[i].p == p)
118 return i;
119 if (++i == pack_revindex_hashsz)
120 i = 0;
121 }
122 return -1 - i;
123}
124
125static void prepare_pack_ix(void)
126{
127 int num;
128 struct packed_git *p;
129 for (num = 0, p = packed_git; p; p = p->next)
130 num++;
131 if (!num)
132 return;
133 pack_revindex_hashsz = num * 11;
134 pack_revindex = xcalloc(sizeof(*pack_revindex), pack_revindex_hashsz);
135 for (p = packed_git; p; p = p->next) {
136 num = pack_revindex_ix(p);
137 num = - 1 - num;
138 pack_revindex[num].p = p;
139 }
140 /* revindex elements are lazily initialized */
141}
142
143static int cmp_offset(const void *a_, const void *b_)
144{
145 const struct revindex_entry *a = a_;
146 const struct revindex_entry *b = b_;
147 return (a->offset < b->offset) ? -1 : (a->offset > b->offset) ? 1 : 0;
148}
149
150/*
151 * Ordered list of offsets of objects in the pack.
152 */
153static void prepare_pack_revindex(struct pack_revindex *rix)
154{
155 struct packed_git *p = rix->p;
156 int num_ent = p->num_objects;
157 int i;
158 const char *index = p->index_data;
159
160 rix->revindex = xmalloc(sizeof(*rix->revindex) * (num_ent + 1));
161 index += 4 * 256;
162
163 if (p->index_version > 1) {
164 const uint32_t *off_32 =
165 (uint32_t *)(index + 8 + p->num_objects * (20 + 4));
166 const uint32_t *off_64 = off_32 + p->num_objects;
167 for (i = 0; i < num_ent; i++) {
168 uint32_t off = ntohl(*off_32++);
169 if (!(off & 0x80000000)) {
170 rix->revindex[i].offset = off;
171 } else {
172 rix->revindex[i].offset =
173 ((uint64_t)ntohl(*off_64++)) << 32;
174 rix->revindex[i].offset |=
175 ntohl(*off_64++);
176 }
177 rix->revindex[i].nr = i;
178 }
179 } else {
180 for (i = 0; i < num_ent; i++) {
181 uint32_t hl = *((uint32_t *)(index + 24 * i));
182 rix->revindex[i].offset = ntohl(hl);
183 rix->revindex[i].nr = i;
184 }
185 }
186
187 /* This knows the pack format -- the 20-byte trailer
188 * follows immediately after the last object data.
189 */
190 rix->revindex[num_ent].offset = p->pack_size - 20;
191 rix->revindex[num_ent].nr = -1;
192 qsort(rix->revindex, num_ent, sizeof(*rix->revindex), cmp_offset);
193}
194
195static struct revindex_entry * find_packed_object(struct packed_git *p,
196 off_t ofs)
197{
198 int num;
199 int lo, hi;
200 struct pack_revindex *rix;
201 struct revindex_entry *revindex;
202 num = pack_revindex_ix(p);
203 if (num < 0)
204 die("internal error: pack revindex uninitialized");
205 rix = &pack_revindex[num];
206 if (!rix->revindex)
207 prepare_pack_revindex(rix);
208 revindex = rix->revindex;
209 lo = 0;
210 hi = p->num_objects + 1;
211 do {
212 int mi = (lo + hi) / 2;
213 if (revindex[mi].offset == ofs) {
214 return revindex + mi;
215 }
216 else if (ofs < revindex[mi].offset)
217 hi = mi;
218 else
219 lo = mi + 1;
220 } while (lo < hi);
221 die("internal error: pack revindex corrupt");
222}
223
224static const unsigned char *find_packed_object_name(struct packed_git *p,
225 off_t ofs)
226{
227 struct revindex_entry *entry = find_packed_object(p, ofs);
228 return nth_packed_object_sha1(p, entry->nr);
229}
230
231static void *delta_against(void *buf, unsigned long size, struct object_entry *entry)
232{
233 unsigned long othersize, delta_size;
234 enum object_type type;
235 void *otherbuf = read_sha1_file(entry->delta->sha1, &type, &othersize);
236 void *delta_buf;
237
238 if (!otherbuf)
239 die("unable to read %s", sha1_to_hex(entry->delta->sha1));
240 delta_buf = diff_delta(otherbuf, othersize,
241 buf, size, &delta_size, 0);
242 if (!delta_buf || delta_size != entry->delta_size)
243 die("delta size changed");
244 free(buf);
245 free(otherbuf);
246 return delta_buf;
247}
248
249/*
250 * The per-object header is a pretty dense thing, which is
251 * - first byte: low four bits are "size", then three bits of "type",
252 * and the high bit is "size continues".
253 * - each byte afterwards: low seven bits are size continuation,
254 * with the high bit being "size continues"
255 */
256static int encode_header(enum object_type type, unsigned long size, unsigned char *hdr)
257{
258 int n = 1;
259 unsigned char c;
260
261 if (type < OBJ_COMMIT || type > OBJ_REF_DELTA)
262 die("bad type %d", type);
263
264 c = (type << 4) | (size & 15);
265 size >>= 4;
266 while (size) {
267 *hdr++ = c | 0x80;
268 c = size & 0x7f;
269 size >>= 7;
270 n++;
271 }
272 *hdr = c;
273 return n;
274}
275
276/*
277 * we are going to reuse the existing object data as is. make
278 * sure it is not corrupt.
279 */
280static int check_pack_inflate(struct packed_git *p,
281 struct pack_window **w_curs,
282 off_t offset,
283 off_t len,
284 unsigned long expect)
285{
286 z_stream stream;
287 unsigned char fakebuf[4096], *in;
288 int st;
289
290 memset(&stream, 0, sizeof(stream));
291 inflateInit(&stream);
292 do {
293 in = use_pack(p, w_curs, offset, &stream.avail_in);
294 stream.next_in = in;
295 stream.next_out = fakebuf;
296 stream.avail_out = sizeof(fakebuf);
297 st = inflate(&stream, Z_FINISH);
298 offset += stream.next_in - in;
299 } while (st == Z_OK || st == Z_BUF_ERROR);
300 inflateEnd(&stream);
301 return (st == Z_STREAM_END &&
302 stream.total_out == expect &&
303 stream.total_in == len) ? 0 : -1;
304}
305
306static int check_pack_crc(struct packed_git *p, struct pack_window **w_curs,
307 off_t offset, off_t len, unsigned int nr)
308{
309 const uint32_t *index_crc;
310 uint32_t data_crc = crc32(0, Z_NULL, 0);
311
312 do {
313 unsigned int avail;
314 void *data = use_pack(p, w_curs, offset, &avail);
315 if (avail > len)
316 avail = len;
317 data_crc = crc32(data_crc, data, avail);
318 offset += avail;
319 len -= avail;
320 } while (len);
321
322 index_crc = p->index_data;
323 index_crc += 2 + 256 + p->num_objects * (20/4) + nr;
324
325 return data_crc != ntohl(*index_crc);
326}
327
328static void copy_pack_data(struct sha1file *f,
329 struct packed_git *p,
330 struct pack_window **w_curs,
331 off_t offset,
332 off_t len)
333{
334 unsigned char *in;
335 unsigned int avail;
336
337 while (len) {
338 in = use_pack(p, w_curs, offset, &avail);
339 if (avail > len)
340 avail = (unsigned int)len;
341 sha1write(f, in, avail);
342 offset += avail;
343 len -= avail;
344 }
345}
346
347static int check_loose_inflate(unsigned char *data, unsigned long len, unsigned long expect)
348{
349 z_stream stream;
350 unsigned char fakebuf[4096];
351 int st;
352
353 memset(&stream, 0, sizeof(stream));
354 stream.next_in = data;
355 stream.avail_in = len;
356 stream.next_out = fakebuf;
357 stream.avail_out = sizeof(fakebuf);
358 inflateInit(&stream);
359
360 while (1) {
361 st = inflate(&stream, Z_FINISH);
362 if (st == Z_STREAM_END || st == Z_OK) {
363 st = (stream.total_out == expect &&
364 stream.total_in == len) ? 0 : -1;
365 break;
366 }
367 if (st != Z_BUF_ERROR) {
368 st = -1;
369 break;
370 }
371 stream.next_out = fakebuf;
372 stream.avail_out = sizeof(fakebuf);
373 }
374 inflateEnd(&stream);
375 return st;
376}
377
378static int revalidate_loose_object(struct object_entry *entry,
379 unsigned char *map,
380 unsigned long mapsize)
381{
382 /* we already know this is a loose object with new type header. */
383 enum object_type type;
384 unsigned long size, used;
385
386 if (pack_to_stdout)
387 return 0;
388
389 used = unpack_object_header_gently(map, mapsize, &type, &size);
390 if (!used)
391 return -1;
392 map += used;
393 mapsize -= used;
394 return check_loose_inflate(map, mapsize, size);
395}
396
397static unsigned long write_object(struct sha1file *f,
398 struct object_entry *entry)
399{
400 unsigned long size;
401 enum object_type type;
402 void *buf;
403 unsigned char header[10];
404 unsigned hdrlen;
405 off_t datalen;
406 enum object_type obj_type;
407 int to_reuse = 0;
408
409 if (!pack_to_stdout)
410 crc32_begin(f);
411
412 obj_type = entry->type;
413 if (! entry->in_pack)
414 to_reuse = 0; /* can't reuse what we don't have */
415 else if (obj_type == OBJ_REF_DELTA || obj_type == OBJ_OFS_DELTA)
416 to_reuse = 1; /* check_object() decided it for us */
417 else if (obj_type != entry->in_pack_type)
418 to_reuse = 0; /* pack has delta which is unusable */
419 else if (entry->delta)
420 to_reuse = 0; /* we want to pack afresh */
421 else
422 to_reuse = 1; /* we have it in-pack undeltified,
423 * and we do not need to deltify it.
424 */
425
426 if (!entry->in_pack && !entry->delta) {
427 unsigned char *map;
428 unsigned long mapsize;
429 map = map_sha1_file(entry->sha1, &mapsize);
430 if (map && !legacy_loose_object(map)) {
431 /* We can copy straight into the pack file */
432 if (revalidate_loose_object(entry, map, mapsize))
433 die("corrupt loose object %s",
434 sha1_to_hex(entry->sha1));
435 sha1write(f, map, mapsize);
436 munmap(map, mapsize);
437 written++;
438 reused++;
439 return mapsize;
440 }
441 if (map)
442 munmap(map, mapsize);
443 }
444
445 if (!to_reuse) {
446 buf = read_sha1_file(entry->sha1, &type, &size);
447 if (!buf)
448 die("unable to read %s", sha1_to_hex(entry->sha1));
449 if (size != entry->size)
450 die("object %s size inconsistency (%lu vs %lu)",
451 sha1_to_hex(entry->sha1), size, entry->size);
452 if (entry->delta) {
453 buf = delta_against(buf, size, entry);
454 size = entry->delta_size;
455 obj_type = (allow_ofs_delta && entry->delta->offset) ?
456 OBJ_OFS_DELTA : OBJ_REF_DELTA;
457 }
458 /*
459 * The object header is a byte of 'type' followed by zero or
460 * more bytes of length.
461 */
462 hdrlen = encode_header(obj_type, size, header);
463 sha1write(f, header, hdrlen);
464
465 if (obj_type == OBJ_OFS_DELTA) {
466 /*
467 * Deltas with relative base contain an additional
468 * encoding of the relative offset for the delta
469 * base from this object's position in the pack.
470 */
471 off_t ofs = entry->offset - entry->delta->offset;
472 unsigned pos = sizeof(header) - 1;
473 header[pos] = ofs & 127;
474 while (ofs >>= 7)
475 header[--pos] = 128 | (--ofs & 127);
476 sha1write(f, header + pos, sizeof(header) - pos);
477 hdrlen += sizeof(header) - pos;
478 } else if (obj_type == OBJ_REF_DELTA) {
479 /*
480 * Deltas with a base reference contain
481 * an additional 20 bytes for the base sha1.
482 */
483 sha1write(f, entry->delta->sha1, 20);
484 hdrlen += 20;
485 }
486 datalen = sha1write_compressed(f, buf, size);
487 free(buf);
488 }
489 else {
490 struct packed_git *p = entry->in_pack;
491 struct pack_window *w_curs = NULL;
492 struct revindex_entry *revidx;
493 off_t offset;
494
495 if (entry->delta) {
496 obj_type = (allow_ofs_delta && entry->delta->offset) ?
497 OBJ_OFS_DELTA : OBJ_REF_DELTA;
498 reused_delta++;
499 }
500 hdrlen = encode_header(obj_type, entry->size, header);
501 sha1write(f, header, hdrlen);
502 if (obj_type == OBJ_OFS_DELTA) {
503 off_t ofs = entry->offset - entry->delta->offset;
504 unsigned pos = sizeof(header) - 1;
505 header[pos] = ofs & 127;
506 while (ofs >>= 7)
507 header[--pos] = 128 | (--ofs & 127);
508 sha1write(f, header + pos, sizeof(header) - pos);
509 hdrlen += sizeof(header) - pos;
510 } else if (obj_type == OBJ_REF_DELTA) {
511 sha1write(f, entry->delta->sha1, 20);
512 hdrlen += 20;
513 }
514
515 offset = entry->in_pack_offset;
516 revidx = find_packed_object(p, offset);
517 datalen = revidx[1].offset - offset;
518 if (!pack_to_stdout && p->index_version > 1 &&
519 check_pack_crc(p, &w_curs, offset, datalen, revidx->nr))
520 die("bad packed object CRC for %s", sha1_to_hex(entry->sha1));
521 offset += entry->in_pack_header_size;
522 datalen -= entry->in_pack_header_size;
523 if (!pack_to_stdout && p->index_version == 1 &&
524 check_pack_inflate(p, &w_curs, offset, datalen, entry->size))
525 die("corrupt packed object for %s", sha1_to_hex(entry->sha1));
526 copy_pack_data(f, p, &w_curs, offset, datalen);
527 unuse_pack(&w_curs);
528 reused++;
529 }
530 if (entry->delta)
531 written_delta++;
532 written++;
533 if (!pack_to_stdout)
534 entry->crc32 = crc32_end(f);
535 return hdrlen + datalen;
536}
537
538static off_t write_one(struct sha1file *f,
539 struct object_entry *e,
540 off_t offset)
541{
542 unsigned long size;
543
544 /* offset is non zero if object is written already. */
545 if (e->offset || e->preferred_base)
546 return offset;
547
548 /* if we are deltified, write out base object first. */
549 if (e->delta)
550 offset = write_one(f, e->delta, offset);
551
552 e->offset = offset;
553 size = write_object(f, e);
554
555 /* make sure off_t is sufficiently large not to wrap */
556 if (offset > offset + size)
557 die("pack too large for current definition of off_t");
558 return offset + size;
559}
560
561static off_t write_pack_file(void)
562{
563 uint32_t i;
564 struct sha1file *f;
565 off_t offset, last_obj_offset = 0;
566 struct pack_header hdr;
567 unsigned last_percent = 999;
568 int do_progress = progress;
569
570 if (pack_to_stdout) {
571 f = sha1fd(1, "<stdout>");
572 do_progress >>= 1;
573 } else {
574 int fd;
575 snprintf(tmpname, sizeof(tmpname), "tmp_pack_XXXXXX");
576 fd = mkstemp(tmpname);
577 if (fd < 0)
578 die("unable to create %s: %s\n", tmpname, strerror(errno));
579 pack_tmp_name = xstrdup(tmpname);
580 f = sha1fd(fd, pack_tmp_name);
581 }
582
583 if (do_progress)
584 fprintf(stderr, "Writing %u objects.\n", nr_result);
585
586 hdr.hdr_signature = htonl(PACK_SIGNATURE);
587 hdr.hdr_version = htonl(PACK_VERSION);
588 hdr.hdr_entries = htonl(nr_result);
589 sha1write(f, &hdr, sizeof(hdr));
590 offset = sizeof(hdr);
591 if (!nr_result)
592 goto done;
593 for (i = 0; i < nr_objects; i++) {
594 last_obj_offset = offset;
595 offset = write_one(f, objects + i, offset);
596 if (do_progress) {
597 unsigned percent = written * 100 / nr_result;
598 if (progress_update || percent != last_percent) {
599 fprintf(stderr, "%4u%% (%u/%u) done\r",
600 percent, written, nr_result);
601 progress_update = 0;
602 last_percent = percent;
603 }
604 }
605 }
606 if (do_progress)
607 fputc('\n', stderr);
608 done:
609 if (written != nr_result)
610 die("wrote %u objects while expecting %u", written, nr_result);
611 sha1close(f, pack_file_sha1, 1);
612
613 return last_obj_offset;
614}
615
616static int sha1_sort(const void *_a, const void *_b)
617{
618 const struct object_entry *a = *(struct object_entry **)_a;
619 const struct object_entry *b = *(struct object_entry **)_b;
620 return hashcmp(a->sha1, b->sha1);
621}
622
623static uint32_t index_default_version = 1;
624static uint32_t index_off32_limit = 0x7fffffff;
625
626static void write_index_file(off_t last_obj_offset, unsigned char *sha1)
627{
628 struct sha1file *f;
629 struct object_entry **sorted_by_sha, **list, **last;
630 uint32_t array[256];
631 uint32_t i, index_version;
632 SHA_CTX ctx;
633 int fd;
634
635 snprintf(tmpname, sizeof(tmpname), "tmp_idx_XXXXXX");
636 fd = mkstemp(tmpname);
637 if (fd < 0)
638 die("unable to create %s: %s\n", tmpname, strerror(errno));
639 idx_tmp_name = xstrdup(tmpname);
640 f = sha1fd(fd, idx_tmp_name);
641
642 if (nr_result) {
643 uint32_t j = 0;
644 sorted_by_sha =
645 xcalloc(nr_result, sizeof(struct object_entry *));
646 for (i = 0; i < nr_objects; i++)
647 if (!objects[i].preferred_base)
648 sorted_by_sha[j++] = objects + i;
649 if (j != nr_result)
650 die("listed %u objects while expecting %u", j, nr_result);
651 qsort(sorted_by_sha, nr_result, sizeof(*sorted_by_sha), sha1_sort);
652 list = sorted_by_sha;
653 last = sorted_by_sha + nr_result;
654 } else
655 sorted_by_sha = list = last = NULL;
656
657 /* if last object's offset is >= 2^31 we should use index V2 */
658 index_version = (last_obj_offset >> 31) ? 2 : index_default_version;
659
660 /* index versions 2 and above need a header */
661 if (index_version >= 2) {
662 struct pack_idx_header hdr;
663 hdr.idx_signature = htonl(PACK_IDX_SIGNATURE);
664 hdr.idx_version = htonl(index_version);
665 sha1write(f, &hdr, sizeof(hdr));
666 }
667
668 /*
669 * Write the first-level table (the list is sorted,
670 * but we use a 256-entry lookup to be able to avoid
671 * having to do eight extra binary search iterations).
672 */
673 for (i = 0; i < 256; i++) {
674 struct object_entry **next = list;
675 while (next < last) {
676 struct object_entry *entry = *next;
677 if (entry->sha1[0] != i)
678 break;
679 next++;
680 }
681 array[i] = htonl(next - sorted_by_sha);
682 list = next;
683 }
684 sha1write(f, array, 256 * 4);
685
686 /* Compute the SHA1 hash of sorted object names. */
687 SHA1_Init(&ctx);
688
689 /* Write the actual SHA1 entries. */
690 list = sorted_by_sha;
691 for (i = 0; i < nr_result; i++) {
692 struct object_entry *entry = *list++;
693 if (index_version < 2) {
694 uint32_t offset = htonl(entry->offset);
695 sha1write(f, &offset, 4);
696 }
697 sha1write(f, entry->sha1, 20);
698 SHA1_Update(&ctx, entry->sha1, 20);
699 }
700
701 if (index_version >= 2) {
702 unsigned int nr_large_offset = 0;
703
704 /* write the crc32 table */
705 list = sorted_by_sha;
706 for (i = 0; i < nr_objects; i++) {
707 struct object_entry *entry = *list++;
708 uint32_t crc32_val = htonl(entry->crc32);
709 sha1write(f, &crc32_val, 4);
710 }
711
712 /* write the 32-bit offset table */
713 list = sorted_by_sha;
714 for (i = 0; i < nr_objects; i++) {
715 struct object_entry *entry = *list++;
716 uint32_t offset = (entry->offset <= index_off32_limit) ?
717 entry->offset : (0x80000000 | nr_large_offset++);
718 offset = htonl(offset);
719 sha1write(f, &offset, 4);
720 }
721
722 /* write the large offset table */
723 list = sorted_by_sha;
724 while (nr_large_offset) {
725 struct object_entry *entry = *list++;
726 uint64_t offset = entry->offset;
727 if (offset > index_off32_limit) {
728 uint32_t split[2];
729 split[0] = htonl(offset >> 32);
730 split[1] = htonl(offset & 0xffffffff);
731 sha1write(f, split, 8);
732 nr_large_offset--;
733 }
734 }
735 }
736
737 sha1write(f, pack_file_sha1, 20);
738 sha1close(f, NULL, 1);
739 free(sorted_by_sha);
740 SHA1_Final(sha1, &ctx);
741}
742
743static int locate_object_entry_hash(const unsigned char *sha1)
744{
745 int i;
746 unsigned int ui;
747 memcpy(&ui, sha1, sizeof(unsigned int));
748 i = ui % object_ix_hashsz;
749 while (0 < object_ix[i]) {
750 if (!hashcmp(sha1, objects[object_ix[i] - 1].sha1))
751 return i;
752 if (++i == object_ix_hashsz)
753 i = 0;
754 }
755 return -1 - i;
756}
757
758static struct object_entry *locate_object_entry(const unsigned char *sha1)
759{
760 int i;
761
762 if (!object_ix_hashsz)
763 return NULL;
764
765 i = locate_object_entry_hash(sha1);
766 if (0 <= i)
767 return &objects[object_ix[i]-1];
768 return NULL;
769}
770
771static void rehash_objects(void)
772{
773 uint32_t i;
774 struct object_entry *oe;
775
776 object_ix_hashsz = nr_objects * 3;
777 if (object_ix_hashsz < 1024)
778 object_ix_hashsz = 1024;
779 object_ix = xrealloc(object_ix, sizeof(int) * object_ix_hashsz);
780 memset(object_ix, 0, sizeof(int) * object_ix_hashsz);
781 for (i = 0, oe = objects; i < nr_objects; i++, oe++) {
782 int ix = locate_object_entry_hash(oe->sha1);
783 if (0 <= ix)
784 continue;
785 ix = -1 - ix;
786 object_ix[ix] = i + 1;
787 }
788}
789
790static unsigned name_hash(const char *name)
791{
792 unsigned char c;
793 unsigned hash = 0;
794
795 /*
796 * This effectively just creates a sortable number from the
797 * last sixteen non-whitespace characters. Last characters
798 * count "most", so things that end in ".c" sort together.
799 */
800 while ((c = *name++) != 0) {
801 if (isspace(c))
802 continue;
803 hash = (hash >> 2) + (c << 24);
804 }
805 return hash;
806}
807
808static int add_object_entry(const unsigned char *sha1, enum object_type type,
809 unsigned hash, int exclude)
810{
811 struct object_entry *entry;
812 struct packed_git *p, *found_pack = NULL;
813 off_t found_offset = 0;
814 int ix;
815
816 ix = nr_objects ? locate_object_entry_hash(sha1) : -1;
817 if (ix >= 0) {
818 if (exclude) {
819 entry = objects + object_ix[ix] - 1;
820 if (!entry->preferred_base)
821 nr_result--;
822 entry->preferred_base = 1;
823 }
824 return 0;
825 }
826
827 for (p = packed_git; p; p = p->next) {
828 off_t offset = find_pack_entry_one(sha1, p);
829 if (offset) {
830 if (!found_pack) {
831 found_offset = offset;
832 found_pack = p;
833 }
834 if (exclude)
835 break;
836 if (incremental)
837 return 0;
838 if (local && !p->pack_local)
839 return 0;
840 }
841 }
842
843 if (nr_objects >= nr_alloc) {
844 nr_alloc = (nr_alloc + 1024) * 3 / 2;
845 objects = xrealloc(objects, nr_alloc * sizeof(*entry));
846 }
847
848 entry = objects + nr_objects++;
849 memset(entry, 0, sizeof(*entry));
850 hashcpy(entry->sha1, sha1);
851 entry->hash = hash;
852 if (type)
853 entry->type = type;
854 if (exclude)
855 entry->preferred_base = 1;
856 else
857 nr_result++;
858 if (found_pack) {
859 entry->in_pack = found_pack;
860 entry->in_pack_offset = found_offset;
861 }
862
863 if (object_ix_hashsz * 3 <= nr_objects * 4)
864 rehash_objects();
865 else
866 object_ix[-1 - ix] = nr_objects;
867
868 if (progress_update) {
869 fprintf(stderr, "Counting objects...%u\r", nr_objects);
870 progress_update = 0;
871 }
872
873 return 1;
874}
875
876struct pbase_tree_cache {
877 unsigned char sha1[20];
878 int ref;
879 int temporary;
880 void *tree_data;
881 unsigned long tree_size;
882};
883
884static struct pbase_tree_cache *(pbase_tree_cache[256]);
885static int pbase_tree_cache_ix(const unsigned char *sha1)
886{
887 return sha1[0] % ARRAY_SIZE(pbase_tree_cache);
888}
889static int pbase_tree_cache_ix_incr(int ix)
890{
891 return (ix+1) % ARRAY_SIZE(pbase_tree_cache);
892}
893
894static struct pbase_tree {
895 struct pbase_tree *next;
896 /* This is a phony "cache" entry; we are not
897 * going to evict it nor find it through _get()
898 * mechanism -- this is for the toplevel node that
899 * would almost always change with any commit.
900 */
901 struct pbase_tree_cache pcache;
902} *pbase_tree;
903
904static struct pbase_tree_cache *pbase_tree_get(const unsigned char *sha1)
905{
906 struct pbase_tree_cache *ent, *nent;
907 void *data;
908 unsigned long size;
909 enum object_type type;
910 int neigh;
911 int my_ix = pbase_tree_cache_ix(sha1);
912 int available_ix = -1;
913
914 /* pbase-tree-cache acts as a limited hashtable.
915 * your object will be found at your index or within a few
916 * slots after that slot if it is cached.
917 */
918 for (neigh = 0; neigh < 8; neigh++) {
919 ent = pbase_tree_cache[my_ix];
920 if (ent && !hashcmp(ent->sha1, sha1)) {
921 ent->ref++;
922 return ent;
923 }
924 else if (((available_ix < 0) && (!ent || !ent->ref)) ||
925 ((0 <= available_ix) &&
926 (!ent && pbase_tree_cache[available_ix])))
927 available_ix = my_ix;
928 if (!ent)
929 break;
930 my_ix = pbase_tree_cache_ix_incr(my_ix);
931 }
932
933 /* Did not find one. Either we got a bogus request or
934 * we need to read and perhaps cache.
935 */
936 data = read_sha1_file(sha1, &type, &size);
937 if (!data)
938 return NULL;
939 if (type != OBJ_TREE) {
940 free(data);
941 return NULL;
942 }
943
944 /* We need to either cache or return a throwaway copy */
945
946 if (available_ix < 0)
947 ent = NULL;
948 else {
949 ent = pbase_tree_cache[available_ix];
950 my_ix = available_ix;
951 }
952
953 if (!ent) {
954 nent = xmalloc(sizeof(*nent));
955 nent->temporary = (available_ix < 0);
956 }
957 else {
958 /* evict and reuse */
959 free(ent->tree_data);
960 nent = ent;
961 }
962 hashcpy(nent->sha1, sha1);
963 nent->tree_data = data;
964 nent->tree_size = size;
965 nent->ref = 1;
966 if (!nent->temporary)
967 pbase_tree_cache[my_ix] = nent;
968 return nent;
969}
970
971static void pbase_tree_put(struct pbase_tree_cache *cache)
972{
973 if (!cache->temporary) {
974 cache->ref--;
975 return;
976 }
977 free(cache->tree_data);
978 free(cache);
979}
980
981static int name_cmp_len(const char *name)
982{
983 int i;
984 for (i = 0; name[i] && name[i] != '\n' && name[i] != '/'; i++)
985 ;
986 return i;
987}
988
989static void add_pbase_object(struct tree_desc *tree,
990 const char *name,
991 int cmplen,
992 const char *fullname)
993{
994 struct name_entry entry;
995 int cmp;
996
997 while (tree_entry(tree,&entry)) {
998 cmp = tree_entry_len(entry.path, entry.sha1) != cmplen ? 1 :
999 memcmp(name, entry.path, cmplen);
1000 if (cmp > 0)
1001 continue;
1002 if (cmp < 0)
1003 return;
1004 if (name[cmplen] != '/') {
1005 unsigned hash = name_hash(fullname);
1006 add_object_entry(entry.sha1,
1007 S_ISDIR(entry.mode) ? OBJ_TREE : OBJ_BLOB,
1008 hash, 1);
1009 return;
1010 }
1011 if (S_ISDIR(entry.mode)) {
1012 struct tree_desc sub;
1013 struct pbase_tree_cache *tree;
1014 const char *down = name+cmplen+1;
1015 int downlen = name_cmp_len(down);
1016
1017 tree = pbase_tree_get(entry.sha1);
1018 if (!tree)
1019 return;
1020 init_tree_desc(&sub, tree->tree_data, tree->tree_size);
1021
1022 add_pbase_object(&sub, down, downlen, fullname);
1023 pbase_tree_put(tree);
1024 }
1025 }
1026}
1027
1028static unsigned *done_pbase_paths;
1029static int done_pbase_paths_num;
1030static int done_pbase_paths_alloc;
1031static int done_pbase_path_pos(unsigned hash)
1032{
1033 int lo = 0;
1034 int hi = done_pbase_paths_num;
1035 while (lo < hi) {
1036 int mi = (hi + lo) / 2;
1037 if (done_pbase_paths[mi] == hash)
1038 return mi;
1039 if (done_pbase_paths[mi] < hash)
1040 hi = mi;
1041 else
1042 lo = mi + 1;
1043 }
1044 return -lo-1;
1045}
1046
1047static int check_pbase_path(unsigned hash)
1048{
1049 int pos = (!done_pbase_paths) ? -1 : done_pbase_path_pos(hash);
1050 if (0 <= pos)
1051 return 1;
1052 pos = -pos - 1;
1053 if (done_pbase_paths_alloc <= done_pbase_paths_num) {
1054 done_pbase_paths_alloc = alloc_nr(done_pbase_paths_alloc);
1055 done_pbase_paths = xrealloc(done_pbase_paths,
1056 done_pbase_paths_alloc *
1057 sizeof(unsigned));
1058 }
1059 done_pbase_paths_num++;
1060 if (pos < done_pbase_paths_num)
1061 memmove(done_pbase_paths + pos + 1,
1062 done_pbase_paths + pos,
1063 (done_pbase_paths_num - pos - 1) * sizeof(unsigned));
1064 done_pbase_paths[pos] = hash;
1065 return 0;
1066}
1067
1068static void add_preferred_base_object(const char *name, unsigned hash)
1069{
1070 struct pbase_tree *it;
1071 int cmplen;
1072
1073 if (!num_preferred_base || check_pbase_path(hash))
1074 return;
1075
1076 cmplen = name_cmp_len(name);
1077 for (it = pbase_tree; it; it = it->next) {
1078 if (cmplen == 0) {
1079 add_object_entry(it->pcache.sha1, OBJ_TREE, 0, 1);
1080 }
1081 else {
1082 struct tree_desc tree;
1083 init_tree_desc(&tree, it->pcache.tree_data, it->pcache.tree_size);
1084 add_pbase_object(&tree, name, cmplen, name);
1085 }
1086 }
1087}
1088
1089static void add_preferred_base(unsigned char *sha1)
1090{
1091 struct pbase_tree *it;
1092 void *data;
1093 unsigned long size;
1094 unsigned char tree_sha1[20];
1095
1096 if (window <= num_preferred_base++)
1097 return;
1098
1099 data = read_object_with_reference(sha1, tree_type, &size, tree_sha1);
1100 if (!data)
1101 return;
1102
1103 for (it = pbase_tree; it; it = it->next) {
1104 if (!hashcmp(it->pcache.sha1, tree_sha1)) {
1105 free(data);
1106 return;
1107 }
1108 }
1109
1110 it = xcalloc(1, sizeof(*it));
1111 it->next = pbase_tree;
1112 pbase_tree = it;
1113
1114 hashcpy(it->pcache.sha1, tree_sha1);
1115 it->pcache.tree_data = data;
1116 it->pcache.tree_size = size;
1117}
1118
1119static void check_object(struct object_entry *entry)
1120{
1121 if (entry->in_pack) {
1122 struct packed_git *p = entry->in_pack;
1123 struct pack_window *w_curs = NULL;
1124 const unsigned char *base_ref = NULL;
1125 struct object_entry *base_entry;
1126 unsigned long used, used_0;
1127 unsigned int avail;
1128 off_t ofs;
1129 unsigned char *buf, c;
1130
1131 buf = use_pack(p, &w_curs, entry->in_pack_offset, &avail);
1132
1133 /*
1134 * We want in_pack_type even if we do not reuse delta.
1135 * There is no point not reusing non-delta representations.
1136 */
1137 used = unpack_object_header_gently(buf, avail,
1138 &entry->in_pack_type,
1139 &entry->size);
1140
1141 /*
1142 * Determine if this is a delta and if so whether we can
1143 * reuse it or not. Otherwise let's find out as cheaply as
1144 * possible what the actual type and size for this object is.
1145 */
1146 switch (entry->in_pack_type) {
1147 default:
1148 /* Not a delta hence we've already got all we need. */
1149 entry->type = entry->in_pack_type;
1150 entry->in_pack_header_size = used;
1151 unuse_pack(&w_curs);
1152 return;
1153 case OBJ_REF_DELTA:
1154 if (!no_reuse_delta && !entry->preferred_base)
1155 base_ref = use_pack(p, &w_curs,
1156 entry->in_pack_offset + used, NULL);
1157 entry->in_pack_header_size = used + 20;
1158 break;
1159 case OBJ_OFS_DELTA:
1160 buf = use_pack(p, &w_curs,
1161 entry->in_pack_offset + used, NULL);
1162 used_0 = 0;
1163 c = buf[used_0++];
1164 ofs = c & 127;
1165 while (c & 128) {
1166 ofs += 1;
1167 if (!ofs || MSB(ofs, 7))
1168 die("delta base offset overflow in pack for %s",
1169 sha1_to_hex(entry->sha1));
1170 c = buf[used_0++];
1171 ofs = (ofs << 7) + (c & 127);
1172 }
1173 if (ofs >= entry->in_pack_offset)
1174 die("delta base offset out of bound for %s",
1175 sha1_to_hex(entry->sha1));
1176 ofs = entry->in_pack_offset - ofs;
1177 if (!no_reuse_delta && !entry->preferred_base)
1178 base_ref = find_packed_object_name(p, ofs);
1179 entry->in_pack_header_size = used + used_0;
1180 break;
1181 }
1182
1183 if (base_ref && (base_entry = locate_object_entry(base_ref))) {
1184 /*
1185 * If base_ref was set above that means we wish to
1186 * reuse delta data, and we even found that base
1187 * in the list of objects we want to pack. Goodie!
1188 *
1189 * Depth value does not matter - find_deltas() will
1190 * never consider reused delta as the base object to
1191 * deltify other objects against, in order to avoid
1192 * circular deltas.
1193 */
1194 entry->type = entry->in_pack_type;
1195 entry->delta = base_entry;
1196 entry->delta_sibling = base_entry->delta_child;
1197 base_entry->delta_child = entry;
1198 unuse_pack(&w_curs);
1199 return;
1200 }
1201
1202 if (entry->type) {
1203 /*
1204 * This must be a delta and we already know what the
1205 * final object type is. Let's extract the actual
1206 * object size from the delta header.
1207 */
1208 entry->size = get_size_from_delta(p, &w_curs,
1209 entry->in_pack_offset + entry->in_pack_header_size);
1210 unuse_pack(&w_curs);
1211 return;
1212 }
1213
1214 /*
1215 * No choice but to fall back to the recursive delta walk
1216 * with sha1_object_info() to find about the object type
1217 * at this point...
1218 */
1219 unuse_pack(&w_curs);
1220 }
1221
1222 entry->type = sha1_object_info(entry->sha1, &entry->size);
1223 if (entry->type < 0)
1224 die("unable to get type of object %s",
1225 sha1_to_hex(entry->sha1));
1226}
1227
1228static int pack_offset_sort(const void *_a, const void *_b)
1229{
1230 const struct object_entry *a = *(struct object_entry **)_a;
1231 const struct object_entry *b = *(struct object_entry **)_b;
1232
1233 /* avoid filesystem trashing with loose objects */
1234 if (!a->in_pack && !b->in_pack)
1235 return hashcmp(a->sha1, b->sha1);
1236
1237 if (a->in_pack < b->in_pack)
1238 return -1;
1239 if (a->in_pack > b->in_pack)
1240 return 1;
1241 return a->in_pack_offset < b->in_pack_offset ? -1 :
1242 (a->in_pack_offset > b->in_pack_offset);
1243}
1244
1245static void get_object_details(void)
1246{
1247 uint32_t i;
1248 struct object_entry **sorted_by_offset;
1249
1250 sorted_by_offset = xcalloc(nr_objects, sizeof(struct object_entry *));
1251 for (i = 0; i < nr_objects; i++)
1252 sorted_by_offset[i] = objects + i;
1253 qsort(sorted_by_offset, nr_objects, sizeof(*sorted_by_offset), pack_offset_sort);
1254
1255 prepare_pack_ix();
1256 for (i = 0; i < nr_objects; i++)
1257 check_object(sorted_by_offset[i]);
1258 free(sorted_by_offset);
1259}
1260
1261static int type_size_sort(const void *_a, const void *_b)
1262{
1263 const struct object_entry *a = *(struct object_entry **)_a;
1264 const struct object_entry *b = *(struct object_entry **)_b;
1265
1266 if (a->type < b->type)
1267 return -1;
1268 if (a->type > b->type)
1269 return 1;
1270 if (a->hash < b->hash)
1271 return -1;
1272 if (a->hash > b->hash)
1273 return 1;
1274 if (a->preferred_base < b->preferred_base)
1275 return -1;
1276 if (a->preferred_base > b->preferred_base)
1277 return 1;
1278 if (a->size < b->size)
1279 return -1;
1280 if (a->size > b->size)
1281 return 1;
1282 return a > b ? -1 : (a < b); /* newest last */
1283}
1284
1285struct unpacked {
1286 struct object_entry *entry;
1287 void *data;
1288 struct delta_index *index;
1289};
1290
1291/*
1292 * We search for deltas _backwards_ in a list sorted by type and
1293 * by size, so that we see progressively smaller and smaller files.
1294 * That's because we prefer deltas to be from the bigger file
1295 * to the smaller - deletes are potentially cheaper, but perhaps
1296 * more importantly, the bigger file is likely the more recent
1297 * one.
1298 */
1299static int try_delta(struct unpacked *trg, struct unpacked *src,
1300 unsigned max_depth)
1301{
1302 struct object_entry *trg_entry = trg->entry;
1303 struct object_entry *src_entry = src->entry;
1304 unsigned long trg_size, src_size, delta_size, sizediff, max_size, sz;
1305 enum object_type type;
1306 void *delta_buf;
1307
1308 /* Don't bother doing diffs between different types */
1309 if (trg_entry->type != src_entry->type)
1310 return -1;
1311
1312 /* We do not compute delta to *create* objects we are not
1313 * going to pack.
1314 */
1315 if (trg_entry->preferred_base)
1316 return -1;
1317
1318 /*
1319 * We do not bother to try a delta that we discarded
1320 * on an earlier try, but only when reusing delta data.
1321 */
1322 if (!no_reuse_delta && trg_entry->in_pack &&
1323 trg_entry->in_pack == src_entry->in_pack &&
1324 trg_entry->in_pack_type != OBJ_REF_DELTA &&
1325 trg_entry->in_pack_type != OBJ_OFS_DELTA)
1326 return 0;
1327
1328 /* Let's not bust the allowed depth. */
1329 if (src_entry->depth >= max_depth)
1330 return 0;
1331
1332 /* Now some size filtering heuristics. */
1333 trg_size = trg_entry->size;
1334 max_size = trg_size/2 - 20;
1335 max_size = max_size * (max_depth - src_entry->depth) / max_depth;
1336 if (max_size == 0)
1337 return 0;
1338 if (trg_entry->delta && trg_entry->delta_size <= max_size)
1339 max_size = trg_entry->delta_size-1;
1340 src_size = src_entry->size;
1341 sizediff = src_size < trg_size ? trg_size - src_size : 0;
1342 if (sizediff >= max_size)
1343 return 0;
1344
1345 /* Load data if not already done */
1346 if (!trg->data) {
1347 trg->data = read_sha1_file(trg_entry->sha1, &type, &sz);
1348 if (sz != trg_size)
1349 die("object %s inconsistent object length (%lu vs %lu)",
1350 sha1_to_hex(trg_entry->sha1), sz, trg_size);
1351 }
1352 if (!src->data) {
1353 src->data = read_sha1_file(src_entry->sha1, &type, &sz);
1354 if (sz != src_size)
1355 die("object %s inconsistent object length (%lu vs %lu)",
1356 sha1_to_hex(src_entry->sha1), sz, src_size);
1357 }
1358 if (!src->index) {
1359 src->index = create_delta_index(src->data, src_size);
1360 if (!src->index)
1361 die("out of memory");
1362 }
1363
1364 delta_buf = create_delta(src->index, trg->data, trg_size, &delta_size, max_size);
1365 if (!delta_buf)
1366 return 0;
1367
1368 trg_entry->delta = src_entry;
1369 trg_entry->delta_size = delta_size;
1370 trg_entry->depth = src_entry->depth + 1;
1371 free(delta_buf);
1372 return 1;
1373}
1374
1375static unsigned int check_delta_limit(struct object_entry *me, unsigned int n)
1376{
1377 struct object_entry *child = me->delta_child;
1378 unsigned int m = n;
1379 while (child) {
1380 unsigned int c = check_delta_limit(child, n + 1);
1381 if (m < c)
1382 m = c;
1383 child = child->delta_sibling;
1384 }
1385 return m;
1386}
1387
1388static void find_deltas(struct object_entry **list, int window, int depth)
1389{
1390 uint32_t i = nr_objects, idx = 0, processed = 0;
1391 unsigned int array_size = window * sizeof(struct unpacked);
1392 struct unpacked *array;
1393 unsigned last_percent = 999;
1394 int max_depth;
1395
1396 if (!nr_objects)
1397 return;
1398 array = xmalloc(array_size);
1399 memset(array, 0, array_size);
1400 if (progress)
1401 fprintf(stderr, "Deltifying %u objects.\n", nr_result);
1402
1403 do {
1404 struct object_entry *entry = list[--i];
1405 struct unpacked *n = array + idx;
1406 int j;
1407
1408 if (!entry->preferred_base)
1409 processed++;
1410
1411 if (progress) {
1412 unsigned percent = processed * 100 / nr_result;
1413 if (percent != last_percent || progress_update) {
1414 fprintf(stderr, "%4u%% (%u/%u) done\r",
1415 percent, processed, nr_result);
1416 progress_update = 0;
1417 last_percent = percent;
1418 }
1419 }
1420
1421 if (entry->delta)
1422 /* This happens if we decided to reuse existing
1423 * delta from a pack. "!no_reuse_delta &&" is implied.
1424 */
1425 continue;
1426
1427 if (entry->size < 50)
1428 continue;
1429 free_delta_index(n->index);
1430 n->index = NULL;
1431 free(n->data);
1432 n->data = NULL;
1433 n->entry = entry;
1434
1435 /*
1436 * If the current object is at pack edge, take the depth the
1437 * objects that depend on the current object into account
1438 * otherwise they would become too deep.
1439 */
1440 max_depth = depth;
1441 if (entry->delta_child) {
1442 max_depth -= check_delta_limit(entry, 0);
1443 if (max_depth <= 0)
1444 goto next;
1445 }
1446
1447 j = window;
1448 while (--j > 0) {
1449 uint32_t other_idx = idx + j;
1450 struct unpacked *m;
1451 if (other_idx >= window)
1452 other_idx -= window;
1453 m = array + other_idx;
1454 if (!m->entry)
1455 break;
1456 if (try_delta(n, m, max_depth) < 0)
1457 break;
1458 }
1459
1460 /* if we made n a delta, and if n is already at max
1461 * depth, leaving it in the window is pointless. we
1462 * should evict it first.
1463 */
1464 if (entry->delta && depth <= entry->depth)
1465 continue;
1466
1467 next:
1468 idx++;
1469 if (idx >= window)
1470 idx = 0;
1471 } while (i > 0);
1472
1473 if (progress)
1474 fputc('\n', stderr);
1475
1476 for (i = 0; i < window; ++i) {
1477 free_delta_index(array[i].index);
1478 free(array[i].data);
1479 }
1480 free(array);
1481}
1482
1483static void prepare_pack(int window, int depth)
1484{
1485 struct object_entry **delta_list;
1486 uint32_t i;
1487
1488 get_object_details();
1489
1490 if (!window || !depth)
1491 return;
1492
1493 delta_list = xmalloc(nr_objects * sizeof(*delta_list));
1494 for (i = 0; i < nr_objects; i++)
1495 delta_list[i] = objects + i;
1496 qsort(delta_list, nr_objects, sizeof(*delta_list), type_size_sort);
1497 find_deltas(delta_list, window+1, depth);
1498 free(delta_list);
1499}
1500
1501static void progress_interval(int signum)
1502{
1503 progress_update = 1;
1504}
1505
1506static void setup_progress_signal(void)
1507{
1508 struct sigaction sa;
1509 struct itimerval v;
1510
1511 memset(&sa, 0, sizeof(sa));
1512 sa.sa_handler = progress_interval;
1513 sigemptyset(&sa.sa_mask);
1514 sa.sa_flags = SA_RESTART;
1515 sigaction(SIGALRM, &sa, NULL);
1516
1517 v.it_interval.tv_sec = 1;
1518 v.it_interval.tv_usec = 0;
1519 v.it_value = v.it_interval;
1520 setitimer(ITIMER_REAL, &v, NULL);
1521}
1522
1523static int git_pack_config(const char *k, const char *v)
1524{
1525 if(!strcmp(k, "pack.window")) {
1526 window = git_config_int(k, v);
1527 return 0;
1528 }
1529 return git_default_config(k, v);
1530}
1531
1532static void read_object_list_from_stdin(void)
1533{
1534 char line[40 + 1 + PATH_MAX + 2];
1535 unsigned char sha1[20];
1536 unsigned hash;
1537
1538 for (;;) {
1539 if (!fgets(line, sizeof(line), stdin)) {
1540 if (feof(stdin))
1541 break;
1542 if (!ferror(stdin))
1543 die("fgets returned NULL, not EOF, not error!");
1544 if (errno != EINTR)
1545 die("fgets: %s", strerror(errno));
1546 clearerr(stdin);
1547 continue;
1548 }
1549 if (line[0] == '-') {
1550 if (get_sha1_hex(line+1, sha1))
1551 die("expected edge sha1, got garbage:\n %s",
1552 line);
1553 add_preferred_base(sha1);
1554 continue;
1555 }
1556 if (get_sha1_hex(line, sha1))
1557 die("expected sha1, got garbage:\n %s", line);
1558
1559 hash = name_hash(line+41);
1560 add_preferred_base_object(line+41, hash);
1561 add_object_entry(sha1, 0, hash, 0);
1562 }
1563}
1564
1565static void show_commit(struct commit *commit)
1566{
1567 add_object_entry(commit->object.sha1, OBJ_COMMIT, 0, 0);
1568}
1569
1570static void show_object(struct object_array_entry *p)
1571{
1572 unsigned hash = name_hash(p->name);
1573 add_preferred_base_object(p->name, hash);
1574 add_object_entry(p->item->sha1, p->item->type, hash, 0);
1575}
1576
1577static void show_edge(struct commit *commit)
1578{
1579 add_preferred_base(commit->object.sha1);
1580}
1581
1582static void get_object_list(int ac, const char **av)
1583{
1584 struct rev_info revs;
1585 char line[1000];
1586 int flags = 0;
1587
1588 init_revisions(&revs, NULL);
1589 save_commit_buffer = 0;
1590 track_object_refs = 0;
1591 setup_revisions(ac, av, &revs, NULL);
1592
1593 while (fgets(line, sizeof(line), stdin) != NULL) {
1594 int len = strlen(line);
1595 if (line[len - 1] == '\n')
1596 line[--len] = 0;
1597 if (!len)
1598 break;
1599 if (*line == '-') {
1600 if (!strcmp(line, "--not")) {
1601 flags ^= UNINTERESTING;
1602 continue;
1603 }
1604 die("not a rev '%s'", line);
1605 }
1606 if (handle_revision_arg(line, &revs, flags, 1))
1607 die("bad revision '%s'", line);
1608 }
1609
1610 prepare_revision_walk(&revs);
1611 mark_edges_uninteresting(revs.commits, &revs, show_edge);
1612 traverse_commit_list(&revs, show_commit, show_object);
1613}
1614
1615int cmd_pack_objects(int argc, const char **argv, const char *prefix)
1616{
1617 int depth = 10;
1618 int use_internal_rev_list = 0;
1619 int thin = 0;
1620 uint32_t i;
1621 off_t last_obj_offset;
1622 const char *base_name = NULL;
1623 const char **rp_av;
1624 int rp_ac_alloc = 64;
1625 int rp_ac;
1626
1627 rp_av = xcalloc(rp_ac_alloc, sizeof(*rp_av));
1628
1629 rp_av[0] = "pack-objects";
1630 rp_av[1] = "--objects"; /* --thin will make it --objects-edge */
1631 rp_ac = 2;
1632
1633 git_config(git_pack_config);
1634
1635 progress = isatty(2);
1636 for (i = 1; i < argc; i++) {
1637 const char *arg = argv[i];
1638
1639 if (*arg != '-')
1640 break;
1641
1642 if (!strcmp("--non-empty", arg)) {
1643 non_empty = 1;
1644 continue;
1645 }
1646 if (!strcmp("--local", arg)) {
1647 local = 1;
1648 continue;
1649 }
1650 if (!strcmp("--incremental", arg)) {
1651 incremental = 1;
1652 continue;
1653 }
1654 if (!prefixcmp(arg, "--window=")) {
1655 char *end;
1656 window = strtoul(arg+9, &end, 0);
1657 if (!arg[9] || *end)
1658 usage(pack_usage);
1659 continue;
1660 }
1661 if (!prefixcmp(arg, "--depth=")) {
1662 char *end;
1663 depth = strtoul(arg+8, &end, 0);
1664 if (!arg[8] || *end)
1665 usage(pack_usage);
1666 continue;
1667 }
1668 if (!strcmp("--progress", arg)) {
1669 progress = 1;
1670 continue;
1671 }
1672 if (!strcmp("--all-progress", arg)) {
1673 progress = 2;
1674 continue;
1675 }
1676 if (!strcmp("-q", arg)) {
1677 progress = 0;
1678 continue;
1679 }
1680 if (!strcmp("--no-reuse-delta", arg)) {
1681 no_reuse_delta = 1;
1682 continue;
1683 }
1684 if (!strcmp("--delta-base-offset", arg)) {
1685 allow_ofs_delta = 1;
1686 continue;
1687 }
1688 if (!strcmp("--stdout", arg)) {
1689 pack_to_stdout = 1;
1690 continue;
1691 }
1692 if (!strcmp("--revs", arg)) {
1693 use_internal_rev_list = 1;
1694 continue;
1695 }
1696 if (!strcmp("--unpacked", arg) ||
1697 !prefixcmp(arg, "--unpacked=") ||
1698 !strcmp("--reflog", arg) ||
1699 !strcmp("--all", arg)) {
1700 use_internal_rev_list = 1;
1701 if (rp_ac >= rp_ac_alloc - 1) {
1702 rp_ac_alloc = alloc_nr(rp_ac_alloc);
1703 rp_av = xrealloc(rp_av,
1704 rp_ac_alloc * sizeof(*rp_av));
1705 }
1706 rp_av[rp_ac++] = arg;
1707 continue;
1708 }
1709 if (!strcmp("--thin", arg)) {
1710 use_internal_rev_list = 1;
1711 thin = 1;
1712 rp_av[1] = "--objects-edge";
1713 continue;
1714 }
1715 if (!prefixcmp(arg, "--index-version=")) {
1716 char *c;
1717 index_default_version = strtoul(arg + 16, &c, 10);
1718 if (index_default_version > 2)
1719 die("bad %s", arg);
1720 if (*c == ',')
1721 index_off32_limit = strtoul(c+1, &c, 0);
1722 if (*c || index_off32_limit & 0x80000000)
1723 die("bad %s", arg);
1724 continue;
1725 }
1726 usage(pack_usage);
1727 }
1728
1729 /* Traditionally "pack-objects [options] base extra" failed;
1730 * we would however want to take refs parameter that would
1731 * have been given to upstream rev-list ourselves, which means
1732 * we somehow want to say what the base name is. So the
1733 * syntax would be:
1734 *
1735 * pack-objects [options] base <refs...>
1736 *
1737 * in other words, we would treat the first non-option as the
1738 * base_name and send everything else to the internal revision
1739 * walker.
1740 */
1741
1742 if (!pack_to_stdout)
1743 base_name = argv[i++];
1744
1745 if (pack_to_stdout != !base_name)
1746 usage(pack_usage);
1747
1748 if (!pack_to_stdout && thin)
1749 die("--thin cannot be used to build an indexable pack.");
1750
1751 prepare_packed_git();
1752
1753 if (progress) {
1754 fprintf(stderr, "Generating pack...\n");
1755 setup_progress_signal();
1756 }
1757
1758 if (!use_internal_rev_list)
1759 read_object_list_from_stdin();
1760 else {
1761 rp_av[rp_ac] = NULL;
1762 get_object_list(rp_ac, rp_av);
1763 }
1764
1765 if (progress)
1766 fprintf(stderr, "Done counting %u objects.\n", nr_objects);
1767 if (non_empty && !nr_result)
1768 return 0;
1769 if (progress && (nr_objects != nr_result))
1770 fprintf(stderr, "Result has %u objects.\n", nr_result);
1771 if (nr_result)
1772 prepare_pack(window, depth);
1773 if (progress == 1 && pack_to_stdout) {
1774 /* the other end usually displays progress itself */
1775 struct itimerval v = {{0,},};
1776 setitimer(ITIMER_REAL, &v, NULL);
1777 signal(SIGALRM, SIG_IGN );
1778 progress_update = 0;
1779 }
1780 last_obj_offset = write_pack_file();
1781 if (!pack_to_stdout) {
1782 unsigned char object_list_sha1[20];
1783 write_index_file(last_obj_offset, object_list_sha1);
1784 snprintf(tmpname, sizeof(tmpname), "%s-%s.pack",
1785 base_name, sha1_to_hex(object_list_sha1));
1786 if (rename(pack_tmp_name, tmpname))
1787 die("unable to rename temporary pack file: %s",
1788 strerror(errno));
1789 snprintf(tmpname, sizeof(tmpname), "%s-%s.idx",
1790 base_name, sha1_to_hex(object_list_sha1));
1791 if (rename(idx_tmp_name, tmpname))
1792 die("unable to rename temporary index file: %s",
1793 strerror(errno));
1794 puts(sha1_to_hex(object_list_sha1));
1795 }
1796 if (progress)
1797 fprintf(stderr, "Total %u (delta %u), reused %u (delta %u)\n",
1798 written, written_delta, reused, reused_delta);
1799 return 0;
1800}