]>
Commit | Line | Data |
---|---|---|
5d4a6003 | 1 | #include "builtin.h" |
c323ac7d | 2 | #include "cache.h" |
a80d72db | 3 | #include "repository.h" |
b2141fc1 | 4 | #include "config.h" |
a74db82e | 5 | #include "attr.h" |
c323ac7d | 6 | #include "object.h" |
8e440259 PE |
7 | #include "blob.h" |
8 | #include "commit.h" | |
9 | #include "tag.h" | |
10 | #include "tree.h" | |
c323ac7d | 11 | #include "delta.h" |
a733cb60 | 12 | #include "pack.h" |
3449f8c4 | 13 | #include "pack-revindex.h" |
c38138cd | 14 | #include "csum-file.h" |
1b0c7174 | 15 | #include "tree-walk.h" |
b5d97e6b JH |
16 | #include "diff.h" |
17 | #include "revision.h" | |
18 | #include "list-objects.h" | |
9535ce73 JH |
19 | #include "list-objects-filter.h" |
20 | #include "list-objects-filter-options.h" | |
2834bc27 | 21 | #include "pack-objects.h" |
96a02f8f | 22 | #include "progress.h" |
f0a24aa5 | 23 | #include "refs.h" |
cf2ba13a | 24 | #include "streaming.h" |
93749194 | 25 | #include "thread-utils.h" |
6b8fda2d | 26 | #include "pack-bitmap.h" |
28b8a730 | 27 | #include "delta-islands.h" |
abcb8655 JK |
28 | #include "reachable.h" |
29 | #include "sha1-array.h" | |
edfbb2aa | 30 | #include "argv-array.h" |
ec2dd32c | 31 | #include "list.h" |
0317f455 | 32 | #include "packfile.h" |
a80d72db | 33 | #include "object-store.h" |
ed7e5fc3 | 34 | #include "dir.h" |
6a22d521 | 35 | #include "midx.h" |
8ecce684 | 36 | |
43fa44fa | 37 | #define IN_PACK(obj) oe_in_pack(&to_pack, obj) |
ac77d0c3 NTND |
38 | #define SIZE(obj) oe_size(&to_pack, obj) |
39 | #define SET_SIZE(obj,size) oe_set_size(&to_pack, obj, size) | |
0aca34e8 | 40 | #define DELTA_SIZE(obj) oe_delta_size(&to_pack, obj) |
898eba5e NTND |
41 | #define DELTA(obj) oe_delta(&to_pack, obj) |
42 | #define DELTA_CHILD(obj) oe_delta_child(&to_pack, obj) | |
43 | #define DELTA_SIBLING(obj) oe_delta_sibling(&to_pack, obj) | |
44 | #define SET_DELTA(obj, val) oe_set_delta(&to_pack, obj, val) | |
6a1e32d5 | 45 | #define SET_DELTA_EXT(obj, oid) oe_set_delta_ext(&to_pack, obj, oid) |
0aca34e8 | 46 | #define SET_DELTA_SIZE(obj, val) oe_set_delta_size(&to_pack, obj, val) |
898eba5e NTND |
47 | #define SET_DELTA_CHILD(obj, val) oe_set_delta_child(&to_pack, obj, val) |
48 | #define SET_DELTA_SIBLING(obj, val) oe_set_delta_sibling(&to_pack, obj, val) | |
43fa44fa | 49 | |
99fb6e04 | 50 | static const char *pack_usage[] = { |
b8c1d275 AH |
51 | N_("git pack-objects --stdout [<options>...] [< <ref-list> | < <object-list>]"), |
52 | N_("git pack-objects [<options>...] <base-name> [< <ref-list> | < <object-list>]"), | |
99fb6e04 NTND |
53 | NULL |
54 | }; | |
c323ac7d | 55 | |
3f9ac8d2 | 56 | /* |
2834bc27 VM |
57 | * Objects we are going to pack are collected in the `to_pack` structure. |
58 | * It contains an array (dynamically expanded) of the object data, and a map | |
59 | * that can resolve SHA1s to their position in the array. | |
3f9ac8d2 | 60 | */ |
2834bc27 VM |
61 | static struct packing_data to_pack; |
62 | ||
79814f42 | 63 | static struct pack_idx_entry **written_list; |
5af05043 | 64 | static uint32_t nr_result, nr_written, nr_seen; |
6a1e32d5 | 65 | static struct bitmap_index *bitmap_git; |
28b8a730 | 66 | static uint32_t write_layer; |
3f9ac8d2 | 67 | |
96f1e58f | 68 | static int non_empty; |
a7de7130 | 69 | static int reuse_delta = 1, reuse_object = 1; |
e5e9714a | 70 | static int keep_unreachable, unpack_unreachable, include_tag; |
dddbad72 | 71 | static timestamp_t unpack_unreachable_expiration; |
e26a8c47 | 72 | static int pack_loose_unreachable; |
96f1e58f | 73 | static int local; |
56dfeb62 | 74 | static int have_non_local_packs; |
96f1e58f | 75 | static int incremental; |
ed7e5fc3 NTND |
76 | static int ignore_packed_keep_on_disk; |
77 | static int ignore_packed_keep_in_core; | |
be6b1914 | 78 | static int allow_ofs_delta; |
ebcfb379 | 79 | static struct pack_idx_option pack_idx_opts; |
d01fb92f | 80 | static const char *base_name; |
024701f1 | 81 | static int progress = 1; |
4812a93a | 82 | static int window = 10; |
568508e7 | 83 | static unsigned long pack_size_limit; |
618e613a | 84 | static int depth = 50; |
43cc2b42 | 85 | static int delta_search_threads; |
df6d6101 | 86 | static int pack_to_stdout; |
6a1e32d5 | 87 | static int thin; |
8d1d8f83 | 88 | static int num_preferred_base; |
dc6a0757 | 89 | static struct progress *progress_state; |
c323ac7d | 90 | |
6b8fda2d VM |
91 | static struct packed_git *reuse_packfile; |
92 | static uint32_t reuse_packfile_objects; | |
93 | static off_t reuse_packfile_offset; | |
94 | ||
645c432d KS |
95 | static int use_bitmap_index_default = 1; |
96 | static int use_bitmap_index = -1; | |
7cc8f971 | 97 | static int write_bitmap_index; |
ae4f07fb | 98 | static uint16_t write_bitmap_options; |
6b8fda2d | 99 | |
0c16cd49 JT |
100 | static int exclude_promisor_objects; |
101 | ||
28b8a730 JK |
102 | static int use_delta_islands; |
103 | ||
074b2eea | 104 | static unsigned long delta_cache_size = 0; |
9806f5a7 | 105 | static unsigned long max_delta_cache_size = DEFAULT_DELTA_CACHE_SIZE; |
e3dfddb3 | 106 | static unsigned long cache_max_small_delta_size = 1000; |
074b2eea | 107 | |
a97773ce BD |
108 | static unsigned long window_memory_limit = 0; |
109 | ||
9535ce73 JH |
110 | static struct list_objects_filter_options filter_options; |
111 | ||
112 | enum missing_action { | |
0c16cd49 JT |
113 | MA_ERROR = 0, /* fail if any missing objects are encountered */ |
114 | MA_ALLOW_ANY, /* silently allow ALL missing objects */ | |
115 | MA_ALLOW_PROMISOR, /* silently allow all missing PROMISOR objects */ | |
9535ce73 JH |
116 | }; |
117 | static enum missing_action arg_missing_action; | |
118 | static show_object_fn fn_show_object; | |
119 | ||
3f9ac8d2 JH |
120 | /* |
121 | * stats | |
122 | */ | |
7cadf491 SP |
123 | static uint32_t written, written_delta; |
124 | static uint32_t reused, reused_delta; | |
3f9ac8d2 | 125 | |
7cc8f971 VM |
126 | /* |
127 | * Indexed commits | |
128 | */ | |
129 | static struct commit **indexed_commits; | |
130 | static unsigned int indexed_commits_nr; | |
131 | static unsigned int indexed_commits_alloc; | |
132 | ||
133 | static void index_commit_for_bitmap(struct commit *commit) | |
134 | { | |
135 | if (indexed_commits_nr >= indexed_commits_alloc) { | |
136 | indexed_commits_alloc = (indexed_commits_alloc + 32) * 2; | |
2756ca43 | 137 | REALLOC_ARRAY(indexed_commits, indexed_commits_alloc); |
7cc8f971 VM |
138 | } |
139 | ||
140 | indexed_commits[indexed_commits_nr++] = commit; | |
141 | } | |
780e6e73 | 142 | |
3613f9b4 | 143 | static void *get_delta(struct object_entry *entry) |
c323ac7d | 144 | { |
3613f9b4 NP |
145 | unsigned long size, base_size, delta_size; |
146 | void *buf, *base_buf, *delta_buf; | |
21666f1a | 147 | enum object_type type; |
c323ac7d | 148 | |
b4f5aca4 | 149 | buf = read_object_file(&entry->idx.oid, &type, &size); |
3613f9b4 | 150 | if (!buf) |
f616db6a | 151 | die(_("unable to read %s"), oid_to_hex(&entry->idx.oid)); |
898eba5e NTND |
152 | base_buf = read_object_file(&DELTA(entry)->idx.oid, &type, |
153 | &base_size); | |
3613f9b4 | 154 | if (!base_buf) |
e6a492b7 | 155 | die("unable to read %s", |
898eba5e | 156 | oid_to_hex(&DELTA(entry)->idx.oid)); |
3613f9b4 | 157 | delta_buf = diff_delta(base_buf, base_size, |
dcde55bc | 158 | buf, size, &delta_size, 0); |
1a07e59c NTND |
159 | /* |
160 | * We succesfully computed this delta once but dropped it for | |
161 | * memory reasons. Something is very wrong if this time we | |
162 | * recompute and create a different delta. | |
163 | */ | |
0aca34e8 | 164 | if (!delta_buf || delta_size != DELTA_SIZE(entry)) |
1a07e59c | 165 | BUG("delta size changed"); |
3613f9b4 NP |
166 | free(buf); |
167 | free(base_buf); | |
c323ac7d LT |
168 | return delta_buf; |
169 | } | |
170 | ||
30ebb40a NP |
171 | static unsigned long do_compress(void **pptr, unsigned long size) |
172 | { | |
ef49a7a0 | 173 | git_zstream stream; |
30ebb40a NP |
174 | void *in, *out; |
175 | unsigned long maxsize; | |
176 | ||
55bb5c91 | 177 | git_deflate_init(&stream, pack_compression_level); |
225a6f10 | 178 | maxsize = git_deflate_bound(&stream, size); |
30ebb40a NP |
179 | |
180 | in = *pptr; | |
181 | out = xmalloc(maxsize); | |
182 | *pptr = out; | |
183 | ||
184 | stream.next_in = in; | |
185 | stream.avail_in = size; | |
186 | stream.next_out = out; | |
187 | stream.avail_out = maxsize; | |
55bb5c91 | 188 | while (git_deflate(&stream, Z_FINISH) == Z_OK) |
30ebb40a | 189 | ; /* nothing */ |
55bb5c91 | 190 | git_deflate_end(&stream); |
30ebb40a NP |
191 | |
192 | free(in); | |
193 | return stream.total_out; | |
194 | } | |
195 | ||
98a3beab | 196 | static unsigned long write_large_blob_data(struct git_istream *st, struct hashfile *f, |
188960b4 | 197 | const struct object_id *oid) |
cf2ba13a NTND |
198 | { |
199 | git_zstream stream; | |
200 | unsigned char ibuf[1024 * 16]; | |
201 | unsigned char obuf[1024 * 16]; | |
202 | unsigned long olen = 0; | |
203 | ||
cf2ba13a NTND |
204 | git_deflate_init(&stream, pack_compression_level); |
205 | ||
206 | for (;;) { | |
207 | ssize_t readlen; | |
208 | int zret = Z_OK; | |
209 | readlen = read_istream(st, ibuf, sizeof(ibuf)); | |
210 | if (readlen == -1) | |
188960b4 | 211 | die(_("unable to read %s"), oid_to_hex(oid)); |
cf2ba13a NTND |
212 | |
213 | stream.next_in = ibuf; | |
214 | stream.avail_in = readlen; | |
215 | while ((stream.avail_in || readlen == 0) && | |
216 | (zret == Z_OK || zret == Z_BUF_ERROR)) { | |
217 | stream.next_out = obuf; | |
218 | stream.avail_out = sizeof(obuf); | |
219 | zret = git_deflate(&stream, readlen ? 0 : Z_FINISH); | |
98a3beab | 220 | hashwrite(f, obuf, stream.next_out - obuf); |
cf2ba13a NTND |
221 | olen += stream.next_out - obuf; |
222 | } | |
223 | if (stream.avail_in) | |
224 | die(_("deflate error (%d)"), zret); | |
225 | if (readlen == 0) { | |
226 | if (zret != Z_STREAM_END) | |
227 | die(_("deflate error (%d)"), zret); | |
228 | break; | |
229 | } | |
230 | } | |
231 | git_deflate_end(&stream); | |
232 | return olen; | |
233 | } | |
234 | ||
780e6e73 NP |
235 | /* |
236 | * we are going to reuse the existing object data as is. make | |
237 | * sure it is not corrupt. | |
238 | */ | |
079afb18 SP |
239 | static int check_pack_inflate(struct packed_git *p, |
240 | struct pack_window **w_curs, | |
6777a59f SP |
241 | off_t offset, |
242 | off_t len, | |
079afb18 SP |
243 | unsigned long expect) |
244 | { | |
ef49a7a0 | 245 | git_zstream stream; |
079afb18 SP |
246 | unsigned char fakebuf[4096], *in; |
247 | int st; | |
248 | ||
249 | memset(&stream, 0, sizeof(stream)); | |
39c68542 | 250 | git_inflate_init(&stream); |
079afb18 SP |
251 | do { |
252 | in = use_pack(p, w_curs, offset, &stream.avail_in); | |
253 | stream.next_in = in; | |
254 | stream.next_out = fakebuf; | |
255 | stream.avail_out = sizeof(fakebuf); | |
39c68542 | 256 | st = git_inflate(&stream, Z_FINISH); |
079afb18 SP |
257 | offset += stream.next_in - in; |
258 | } while (st == Z_OK || st == Z_BUF_ERROR); | |
39c68542 | 259 | git_inflate_end(&stream); |
079afb18 SP |
260 | return (st == Z_STREAM_END && |
261 | stream.total_out == expect && | |
262 | stream.total_in == len) ? 0 : -1; | |
263 | } | |
264 | ||
98a3beab | 265 | static void copy_pack_data(struct hashfile *f, |
079afb18 SP |
266 | struct packed_git *p, |
267 | struct pack_window **w_curs, | |
6777a59f SP |
268 | off_t offset, |
269 | off_t len) | |
079afb18 SP |
270 | { |
271 | unsigned char *in; | |
ef49a7a0 | 272 | unsigned long avail; |
079afb18 SP |
273 | |
274 | while (len) { | |
275 | in = use_pack(p, w_curs, offset, &avail); | |
276 | if (avail > len) | |
ef49a7a0 | 277 | avail = (unsigned long)len; |
98a3beab | 278 | hashwrite(f, in, avail); |
079afb18 SP |
279 | offset += avail; |
280 | len -= avail; | |
281 | } | |
282 | } | |
283 | ||
1b4bb16b | 284 | /* Return 0 if we will bust the pack-size limit */ |
98a3beab | 285 | static unsigned long write_no_reuse_object(struct hashfile *f, struct object_entry *entry, |
c9018b03 | 286 | unsigned long limit, int usable_delta) |
c323ac7d | 287 | { |
c9018b03 | 288 | unsigned long size, datalen; |
2c5e2865 JK |
289 | unsigned char header[MAX_PACK_OBJECT_HEADER], |
290 | dheader[MAX_PACK_OBJECT_HEADER]; | |
6777a59f | 291 | unsigned hdrlen; |
2c5ef824 | 292 | enum object_type type; |
c9018b03 | 293 | void *buf; |
cf2ba13a | 294 | struct git_istream *st = NULL; |
41179100 | 295 | const unsigned hashsz = the_hash_algo->rawsz; |
c9018b03 NTND |
296 | |
297 | if (!usable_delta) { | |
fd9b1bae | 298 | if (oe_type(entry) == OBJ_BLOB && |
ac77d0c3 | 299 | oe_size_greater_than(&to_pack, entry, big_file_threshold) && |
ef7b5195 | 300 | (st = open_istream(&entry->idx.oid, &type, &size, NULL)) != NULL) |
cf2ba13a NTND |
301 | buf = NULL; |
302 | else { | |
b4f5aca4 | 303 | buf = read_object_file(&entry->idx.oid, &type, &size); |
cf2ba13a | 304 | if (!buf) |
e6a492b7 | 305 | die(_("unable to read %s"), |
306 | oid_to_hex(&entry->idx.oid)); | |
cf2ba13a | 307 | } |
c9018b03 NTND |
308 | /* |
309 | * make sure no cached delta data remains from a | |
310 | * previous attempt before a pack split occurred. | |
311 | */ | |
6a83d902 | 312 | FREE_AND_NULL(entry->delta_data); |
c9018b03 NTND |
313 | entry->z_delta_size = 0; |
314 | } else if (entry->delta_data) { | |
0aca34e8 | 315 | size = DELTA_SIZE(entry); |
c9018b03 NTND |
316 | buf = entry->delta_data; |
317 | entry->delta_data = NULL; | |
898eba5e | 318 | type = (allow_ofs_delta && DELTA(entry)->idx.offset) ? |
c9018b03 NTND |
319 | OBJ_OFS_DELTA : OBJ_REF_DELTA; |
320 | } else { | |
321 | buf = get_delta(entry); | |
0aca34e8 | 322 | size = DELTA_SIZE(entry); |
898eba5e | 323 | type = (allow_ofs_delta && DELTA(entry)->idx.offset) ? |
c9018b03 NTND |
324 | OBJ_OFS_DELTA : OBJ_REF_DELTA; |
325 | } | |
326 | ||
cf2ba13a NTND |
327 | if (st) /* large blob case, just assume we don't compress well */ |
328 | datalen = size; | |
329 | else if (entry->z_delta_size) | |
c9018b03 NTND |
330 | datalen = entry->z_delta_size; |
331 | else | |
332 | datalen = do_compress(&buf, size); | |
333 | ||
334 | /* | |
335 | * The object header is a byte of 'type' followed by zero or | |
336 | * more bytes of length. | |
337 | */ | |
7202a6fa JK |
338 | hdrlen = encode_in_pack_object_header(header, sizeof(header), |
339 | type, size); | |
c9018b03 NTND |
340 | |
341 | if (type == OBJ_OFS_DELTA) { | |
342 | /* | |
343 | * Deltas with relative base contain an additional | |
344 | * encoding of the relative offset for the delta | |
345 | * base from this object's position in the pack. | |
346 | */ | |
898eba5e | 347 | off_t ofs = entry->idx.offset - DELTA(entry)->idx.offset; |
c9018b03 NTND |
348 | unsigned pos = sizeof(dheader) - 1; |
349 | dheader[pos] = ofs & 127; | |
350 | while (ofs >>= 7) | |
351 | dheader[--pos] = 128 | (--ofs & 127); | |
41179100 | 352 | if (limit && hdrlen + sizeof(dheader) - pos + datalen + hashsz >= limit) { |
cf2ba13a NTND |
353 | if (st) |
354 | close_istream(st); | |
c9018b03 NTND |
355 | free(buf); |
356 | return 0; | |
357 | } | |
98a3beab | 358 | hashwrite(f, header, hdrlen); |
359 | hashwrite(f, dheader + pos, sizeof(dheader) - pos); | |
c9018b03 NTND |
360 | hdrlen += sizeof(dheader) - pos; |
361 | } else if (type == OBJ_REF_DELTA) { | |
362 | /* | |
363 | * Deltas with a base reference contain | |
41179100 | 364 | * additional bytes for the base object ID. |
c9018b03 | 365 | */ |
41179100 | 366 | if (limit && hdrlen + hashsz + datalen + hashsz >= limit) { |
cf2ba13a NTND |
367 | if (st) |
368 | close_istream(st); | |
c9018b03 NTND |
369 | free(buf); |
370 | return 0; | |
371 | } | |
98a3beab | 372 | hashwrite(f, header, hdrlen); |
42c8ce1c | 373 | hashwrite(f, DELTA(entry)->idx.oid.hash, hashsz); |
41179100 | 374 | hdrlen += hashsz; |
c9018b03 | 375 | } else { |
41179100 | 376 | if (limit && hdrlen + datalen + hashsz >= limit) { |
cf2ba13a NTND |
377 | if (st) |
378 | close_istream(st); | |
c9018b03 NTND |
379 | free(buf); |
380 | return 0; | |
381 | } | |
98a3beab | 382 | hashwrite(f, header, hdrlen); |
c9018b03 | 383 | } |
cf2ba13a | 384 | if (st) { |
188960b4 | 385 | datalen = write_large_blob_data(st, f, &entry->idx.oid); |
cf2ba13a NTND |
386 | close_istream(st); |
387 | } else { | |
98a3beab | 388 | hashwrite(f, buf, datalen); |
cf2ba13a NTND |
389 | free(buf); |
390 | } | |
c9018b03 NTND |
391 | |
392 | return hdrlen + datalen; | |
393 | } | |
394 | ||
395 | /* Return 0 if we will bust the pack-size limit */ | |
98a3beab | 396 | static off_t write_reuse_object(struct hashfile *f, struct object_entry *entry, |
af92a645 | 397 | unsigned long limit, int usable_delta) |
c9018b03 | 398 | { |
43fa44fa | 399 | struct packed_git *p = IN_PACK(entry); |
c9018b03 NTND |
400 | struct pack_window *w_curs = NULL; |
401 | struct revindex_entry *revidx; | |
402 | off_t offset; | |
fd9b1bae | 403 | enum object_type type = oe_type(entry); |
211c61c6 | 404 | off_t datalen; |
2c5e2865 JK |
405 | unsigned char header[MAX_PACK_OBJECT_HEADER], |
406 | dheader[MAX_PACK_OBJECT_HEADER]; | |
c9018b03 | 407 | unsigned hdrlen; |
41179100 | 408 | const unsigned hashsz = the_hash_algo->rawsz; |
ac77d0c3 | 409 | unsigned long entry_size = SIZE(entry); |
c9018b03 | 410 | |
898eba5e NTND |
411 | if (DELTA(entry)) |
412 | type = (allow_ofs_delta && DELTA(entry)->idx.offset) ? | |
c9018b03 | 413 | OBJ_OFS_DELTA : OBJ_REF_DELTA; |
7202a6fa | 414 | hdrlen = encode_in_pack_object_header(header, sizeof(header), |
ac77d0c3 | 415 | type, entry_size); |
c9018b03 NTND |
416 | |
417 | offset = entry->in_pack_offset; | |
418 | revidx = find_pack_revindex(p, offset); | |
419 | datalen = revidx[1].offset - offset; | |
420 | if (!pack_to_stdout && p->index_version > 1 && | |
421 | check_pack_crc(p, &w_curs, offset, datalen, revidx->nr)) { | |
f616db6a | 422 | error(_("bad packed object CRC for %s"), |
e6a492b7 | 423 | oid_to_hex(&entry->idx.oid)); |
c9018b03 NTND |
424 | unuse_pack(&w_curs); |
425 | return write_no_reuse_object(f, entry, limit, usable_delta); | |
426 | } | |
427 | ||
428 | offset += entry->in_pack_header_size; | |
429 | datalen -= entry->in_pack_header_size; | |
430 | ||
431 | if (!pack_to_stdout && p->index_version == 1 && | |
ac77d0c3 | 432 | check_pack_inflate(p, &w_curs, offset, datalen, entry_size)) { |
f616db6a | 433 | error(_("corrupt packed object for %s"), |
e6a492b7 | 434 | oid_to_hex(&entry->idx.oid)); |
c9018b03 NTND |
435 | unuse_pack(&w_curs); |
436 | return write_no_reuse_object(f, entry, limit, usable_delta); | |
437 | } | |
438 | ||
439 | if (type == OBJ_OFS_DELTA) { | |
898eba5e | 440 | off_t ofs = entry->idx.offset - DELTA(entry)->idx.offset; |
c9018b03 NTND |
441 | unsigned pos = sizeof(dheader) - 1; |
442 | dheader[pos] = ofs & 127; | |
443 | while (ofs >>= 7) | |
444 | dheader[--pos] = 128 | (--ofs & 127); | |
41179100 | 445 | if (limit && hdrlen + sizeof(dheader) - pos + datalen + hashsz >= limit) { |
c9018b03 NTND |
446 | unuse_pack(&w_curs); |
447 | return 0; | |
448 | } | |
98a3beab | 449 | hashwrite(f, header, hdrlen); |
450 | hashwrite(f, dheader + pos, sizeof(dheader) - pos); | |
c9018b03 NTND |
451 | hdrlen += sizeof(dheader) - pos; |
452 | reused_delta++; | |
453 | } else if (type == OBJ_REF_DELTA) { | |
41179100 | 454 | if (limit && hdrlen + hashsz + datalen + hashsz >= limit) { |
c9018b03 NTND |
455 | unuse_pack(&w_curs); |
456 | return 0; | |
457 | } | |
98a3beab | 458 | hashwrite(f, header, hdrlen); |
42c8ce1c | 459 | hashwrite(f, DELTA(entry)->idx.oid.hash, hashsz); |
41179100 | 460 | hdrlen += hashsz; |
c9018b03 NTND |
461 | reused_delta++; |
462 | } else { | |
41179100 | 463 | if (limit && hdrlen + datalen + hashsz >= limit) { |
c9018b03 NTND |
464 | unuse_pack(&w_curs); |
465 | return 0; | |
466 | } | |
98a3beab | 467 | hashwrite(f, header, hdrlen); |
c9018b03 NTND |
468 | } |
469 | copy_pack_data(f, p, &w_curs, offset, datalen); | |
470 | unuse_pack(&w_curs); | |
471 | reused++; | |
472 | return hdrlen + datalen; | |
473 | } | |
474 | ||
475 | /* Return 0 if we will bust the pack-size limit */ | |
98a3beab | 476 | static off_t write_object(struct hashfile *f, |
af92a645 NTND |
477 | struct object_entry *entry, |
478 | off_t write_offset) | |
c9018b03 | 479 | { |
af92a645 NTND |
480 | unsigned long limit; |
481 | off_t len; | |
2c5ef824 | 482 | int usable_delta, to_reuse; |
c323ac7d | 483 | |
78d1e84f NP |
484 | if (!pack_to_stdout) |
485 | crc32_begin(f); | |
486 | ||
a2430dde | 487 | /* apply size limit if limited packsize and not first object */ |
a1e4760f NP |
488 | if (!pack_size_limit || !nr_written) |
489 | limit = 0; | |
490 | else if (pack_size_limit <= write_offset) | |
491 | /* | |
492 | * the earlier object did not fit the limit; avoid | |
493 | * mistaking this with unlimited (i.e. limit = 0). | |
494 | */ | |
495 | limit = 1; | |
496 | else | |
497 | limit = pack_size_limit - write_offset; | |
2c5ef824 | 498 | |
898eba5e | 499 | if (!DELTA(entry)) |
2c5ef824 NP |
500 | usable_delta = 0; /* no delta */ |
501 | else if (!pack_size_limit) | |
502 | usable_delta = 1; /* unlimited packfile */ | |
898eba5e | 503 | else if (DELTA(entry)->idx.offset == (off_t)-1) |
2c5ef824 | 504 | usable_delta = 0; /* base was written to another pack */ |
898eba5e | 505 | else if (DELTA(entry)->idx.offset) |
2c5ef824 NP |
506 | usable_delta = 1; /* base already exists in this pack */ |
507 | else | |
508 | usable_delta = 0; /* base could end up in another pack */ | |
509 | ||
a7de7130 | 510 | if (!reuse_object) |
fa736f72 | 511 | to_reuse = 0; /* explicit */ |
43fa44fa | 512 | else if (!IN_PACK(entry)) |
ab7cd7bb | 513 | to_reuse = 0; /* can't reuse what we don't have */ |
fd9b1bae NTND |
514 | else if (oe_type(entry) == OBJ_REF_DELTA || |
515 | oe_type(entry) == OBJ_OFS_DELTA) | |
17b08f2c DH |
516 | /* check_object() decided it for us ... */ |
517 | to_reuse = usable_delta; | |
518 | /* ... but pack split may override that */ | |
fd9b1bae | 519 | else if (oe_type(entry) != entry->in_pack_type) |
ab7cd7bb | 520 | to_reuse = 0; /* pack has delta which is unusable */ |
898eba5e | 521 | else if (DELTA(entry)) |
ab7cd7bb JH |
522 | to_reuse = 0; /* we want to pack afresh */ |
523 | else | |
524 | to_reuse = 1; /* we have it in-pack undeltified, | |
525 | * and we do not need to deltify it. | |
526 | */ | |
527 | ||
c9018b03 NTND |
528 | if (!to_reuse) |
529 | len = write_no_reuse_object(f, entry, limit, usable_delta); | |
530 | else | |
531 | len = write_reuse_object(f, entry, limit, usable_delta); | |
532 | if (!len) | |
533 | return 0; | |
64bd76b1 | 534 | |
17b08f2c | 535 | if (usable_delta) |
ab7cd7bb | 536 | written_delta++; |
3f9ac8d2 | 537 | written++; |
78d1e84f | 538 | if (!pack_to_stdout) |
aa7e44bf | 539 | entry->idx.crc32 = crc32_end(f); |
c9018b03 | 540 | return len; |
c323ac7d LT |
541 | } |
542 | ||
f63c79db JH |
543 | enum write_one_status { |
544 | WRITE_ONE_SKIP = -1, /* already written */ | |
545 | WRITE_ONE_BREAK = 0, /* writing this will bust the limit; not written */ | |
546 | WRITE_ONE_WRITTEN = 1, /* normal */ | |
547 | WRITE_ONE_RECURSIVE = 2 /* already scheduled to be written */ | |
548 | }; | |
549 | ||
98a3beab | 550 | static enum write_one_status write_one(struct hashfile *f, |
f63c79db JH |
551 | struct object_entry *e, |
552 | off_t *offset) | |
9d5ab962 | 553 | { |
af92a645 | 554 | off_t size; |
f63c79db | 555 | int recursing; |
d7dd0223 | 556 | |
f63c79db JH |
557 | /* |
558 | * we set offset to 1 (which is an impossible value) to mark | |
559 | * the fact that this object is involved in "write its base | |
560 | * first before writing a deltified object" recursion. | |
561 | */ | |
562 | recursing = (e->idx.offset == 1); | |
563 | if (recursing) { | |
f616db6a | 564 | warning(_("recursive delta detected for object %s"), |
e6a492b7 | 565 | oid_to_hex(&e->idx.oid)); |
f63c79db JH |
566 | return WRITE_ONE_RECURSIVE; |
567 | } else if (e->idx.offset || e->preferred_base) { | |
568 | /* offset is non zero if object is written already. */ | |
569 | return WRITE_ONE_SKIP; | |
570 | } | |
d7dd0223 | 571 | |
720c9f7b | 572 | /* if we are deltified, write out base object first. */ |
898eba5e | 573 | if (DELTA(e)) { |
f63c79db | 574 | e->idx.offset = 1; /* now recurse */ |
898eba5e | 575 | switch (write_one(f, DELTA(e), offset)) { |
f63c79db JH |
576 | case WRITE_ONE_RECURSIVE: |
577 | /* we cannot depend on this one */ | |
898eba5e | 578 | SET_DELTA(e, NULL); |
f63c79db JH |
579 | break; |
580 | default: | |
581 | break; | |
582 | case WRITE_ONE_BREAK: | |
583 | e->idx.offset = recursing; | |
584 | return WRITE_ONE_BREAK; | |
585 | } | |
586 | } | |
d7dd0223 | 587 | |
6ed7f25e NP |
588 | e->idx.offset = *offset; |
589 | size = write_object(f, e, *offset); | |
17b08f2c | 590 | if (!size) { |
f63c79db JH |
591 | e->idx.offset = recursing; |
592 | return WRITE_ONE_BREAK; | |
17b08f2c | 593 | } |
79814f42 | 594 | written_list[nr_written++] = &e->idx; |
d7dd0223 NP |
595 | |
596 | /* make sure off_t is sufficiently large not to wrap */ | |
c03c8315 | 597 | if (signed_add_overflows(*offset, size)) |
f616db6a | 598 | die(_("pack too large for current definition of off_t")); |
6ed7f25e | 599 | *offset += size; |
f63c79db | 600 | return WRITE_ONE_WRITTEN; |
9d5ab962 JH |
601 | } |
602 | ||
d155254c | 603 | static int mark_tagged(const char *path, const struct object_id *oid, int flag, |
1b4bb16b JH |
604 | void *cb_data) |
605 | { | |
188960b4 | 606 | struct object_id peeled; |
d155254c | 607 | struct object_entry *entry = packlist_find(&to_pack, oid->hash, NULL); |
1b4bb16b JH |
608 | |
609 | if (entry) | |
610 | entry->tagged = 1; | |
b420d909 | 611 | if (!peel_ref(path, &peeled)) { |
188960b4 | 612 | entry = packlist_find(&to_pack, peeled.hash, NULL); |
1b4bb16b JH |
613 | if (entry) |
614 | entry->tagged = 1; | |
615 | } | |
616 | return 0; | |
617 | } | |
618 | ||
be126818 | 619 | static inline void add_to_write_order(struct object_entry **wo, |
92bef1a1 | 620 | unsigned int *endp, |
1b4bb16b JH |
621 | struct object_entry *e) |
622 | { | |
fe0ac2fb | 623 | if (e->filled || oe_layer(&to_pack, e) != write_layer) |
1b4bb16b JH |
624 | return; |
625 | wo[(*endp)++] = e; | |
626 | e->filled = 1; | |
627 | } | |
628 | ||
629 | static void add_descendants_to_write_order(struct object_entry **wo, | |
92bef1a1 | 630 | unsigned int *endp, |
1b4bb16b JH |
631 | struct object_entry *e) |
632 | { | |
f380872f DM |
633 | int add_to_order = 1; |
634 | while (e) { | |
635 | if (add_to_order) { | |
636 | struct object_entry *s; | |
637 | /* add this node... */ | |
638 | add_to_write_order(wo, endp, e); | |
639 | /* all its siblings... */ | |
898eba5e | 640 | for (s = DELTA_SIBLING(e); s; s = DELTA_SIBLING(s)) { |
f380872f DM |
641 | add_to_write_order(wo, endp, s); |
642 | } | |
643 | } | |
644 | /* drop down a level to add left subtree nodes if possible */ | |
898eba5e | 645 | if (DELTA_CHILD(e)) { |
f380872f | 646 | add_to_order = 1; |
898eba5e | 647 | e = DELTA_CHILD(e); |
f380872f DM |
648 | } else { |
649 | add_to_order = 0; | |
650 | /* our sibling might have some children, it is next */ | |
898eba5e NTND |
651 | if (DELTA_SIBLING(e)) { |
652 | e = DELTA_SIBLING(e); | |
f380872f DM |
653 | continue; |
654 | } | |
655 | /* go back to our parent node */ | |
898eba5e NTND |
656 | e = DELTA(e); |
657 | while (e && !DELTA_SIBLING(e)) { | |
f380872f DM |
658 | /* we're on the right side of a subtree, keep |
659 | * going up until we can go right again */ | |
898eba5e | 660 | e = DELTA(e); |
f380872f DM |
661 | } |
662 | if (!e) { | |
663 | /* done- we hit our original root node */ | |
664 | return; | |
665 | } | |
666 | /* pass it off to sibling at this level */ | |
898eba5e | 667 | e = DELTA_SIBLING(e); |
f380872f DM |
668 | } |
669 | }; | |
1b4bb16b JH |
670 | } |
671 | ||
672 | static void add_family_to_write_order(struct object_entry **wo, | |
92bef1a1 | 673 | unsigned int *endp, |
1b4bb16b JH |
674 | struct object_entry *e) |
675 | { | |
676 | struct object_entry *root; | |
677 | ||
898eba5e | 678 | for (root = e; DELTA(root); root = DELTA(root)) |
1b4bb16b | 679 | ; /* nothing */ |
1b4bb16b JH |
680 | add_descendants_to_write_order(wo, endp, root); |
681 | } | |
682 | ||
f64ba53e | 683 | static void compute_layer_order(struct object_entry **wo, unsigned int *wo_end) |
1b4bb16b | 684 | { |
f64ba53e | 685 | unsigned int i, last_untagged; |
2834bc27 | 686 | struct object_entry *objects = to_pack.objects; |
1b4bb16b | 687 | |
2834bc27 | 688 | for (i = 0; i < to_pack.nr_objects; i++) { |
1b4bb16b JH |
689 | if (objects[i].tagged) |
690 | break; | |
f64ba53e | 691 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b | 692 | } |
38d4debb | 693 | last_untagged = i; |
1b4bb16b JH |
694 | |
695 | /* | |
696 | * Then fill all the tagged tips. | |
697 | */ | |
2834bc27 | 698 | for (; i < to_pack.nr_objects; i++) { |
1b4bb16b | 699 | if (objects[i].tagged) |
f64ba53e | 700 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b JH |
701 | } |
702 | ||
703 | /* | |
704 | * And then all remaining commits and tags. | |
705 | */ | |
2834bc27 | 706 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
fd9b1bae NTND |
707 | if (oe_type(&objects[i]) != OBJ_COMMIT && |
708 | oe_type(&objects[i]) != OBJ_TAG) | |
1b4bb16b | 709 | continue; |
f64ba53e | 710 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b JH |
711 | } |
712 | ||
713 | /* | |
714 | * And then all the trees. | |
715 | */ | |
2834bc27 | 716 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
fd9b1bae | 717 | if (oe_type(&objects[i]) != OBJ_TREE) |
1b4bb16b | 718 | continue; |
f64ba53e | 719 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b JH |
720 | } |
721 | ||
722 | /* | |
723 | * Finally all the rest in really tight order | |
724 | */ | |
2834bc27 | 725 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
fe0ac2fb | 726 | if (!objects[i].filled && oe_layer(&to_pack, &objects[i]) == write_layer) |
f64ba53e | 727 | add_family_to_write_order(wo, wo_end, &objects[i]); |
38d4debb | 728 | } |
f64ba53e CC |
729 | } |
730 | ||
731 | static struct object_entry **compute_write_order(void) | |
732 | { | |
28b8a730 | 733 | uint32_t max_layers = 1; |
f64ba53e CC |
734 | unsigned int i, wo_end; |
735 | ||
736 | struct object_entry **wo; | |
737 | struct object_entry *objects = to_pack.objects; | |
738 | ||
739 | for (i = 0; i < to_pack.nr_objects; i++) { | |
740 | objects[i].tagged = 0; | |
741 | objects[i].filled = 0; | |
742 | SET_DELTA_CHILD(&objects[i], NULL); | |
743 | SET_DELTA_SIBLING(&objects[i], NULL); | |
744 | } | |
745 | ||
746 | /* | |
747 | * Fully connect delta_child/delta_sibling network. | |
748 | * Make sure delta_sibling is sorted in the original | |
749 | * recency order. | |
750 | */ | |
751 | for (i = to_pack.nr_objects; i > 0;) { | |
752 | struct object_entry *e = &objects[--i]; | |
753 | if (!DELTA(e)) | |
754 | continue; | |
755 | /* Mark me as the first child */ | |
756 | e->delta_sibling_idx = DELTA(e)->delta_child_idx; | |
757 | SET_DELTA_CHILD(DELTA(e), e); | |
38d4debb DM |
758 | } |
759 | ||
f64ba53e CC |
760 | /* |
761 | * Mark objects that are at the tip of tags. | |
762 | */ | |
763 | for_each_tag_ref(mark_tagged, NULL); | |
764 | ||
28b8a730 JK |
765 | if (use_delta_islands) |
766 | max_layers = compute_pack_layers(&to_pack); | |
767 | ||
f64ba53e CC |
768 | ALLOC_ARRAY(wo, to_pack.nr_objects); |
769 | wo_end = 0; | |
770 | ||
28b8a730 JK |
771 | for (; write_layer < max_layers; ++write_layer) |
772 | compute_layer_order(wo, &wo_end); | |
38d4debb | 773 | |
2834bc27 | 774 | if (wo_end != to_pack.nr_objects) |
f616db6a NTND |
775 | die(_("ordered %u objects, expected %"PRIu32), |
776 | wo_end, to_pack.nr_objects); | |
1b4bb16b JH |
777 | |
778 | return wo; | |
779 | } | |
780 | ||
98a3beab | 781 | static off_t write_reused_pack(struct hashfile *f) |
6b8fda2d VM |
782 | { |
783 | unsigned char buffer[8192]; | |
657673f1 | 784 | off_t to_write, total; |
6b8fda2d VM |
785 | int fd; |
786 | ||
787 | if (!is_pack_valid(reuse_packfile)) | |
f616db6a | 788 | die(_("packfile is invalid: %s"), reuse_packfile->pack_name); |
6b8fda2d | 789 | |
a5436b57 | 790 | fd = git_open(reuse_packfile->pack_name); |
6b8fda2d | 791 | if (fd < 0) |
f616db6a | 792 | die_errno(_("unable to open packfile for reuse: %s"), |
6b8fda2d VM |
793 | reuse_packfile->pack_name); |
794 | ||
795 | if (lseek(fd, sizeof(struct pack_header), SEEK_SET) == -1) | |
f616db6a | 796 | die_errno(_("unable to seek in reused packfile")); |
6b8fda2d VM |
797 | |
798 | if (reuse_packfile_offset < 0) | |
41179100 | 799 | reuse_packfile_offset = reuse_packfile->pack_size - the_hash_algo->rawsz; |
6b8fda2d | 800 | |
657673f1 | 801 | total = to_write = reuse_packfile_offset - sizeof(struct pack_header); |
6b8fda2d VM |
802 | |
803 | while (to_write) { | |
804 | int read_pack = xread(fd, buffer, sizeof(buffer)); | |
805 | ||
806 | if (read_pack <= 0) | |
f616db6a | 807 | die_errno(_("unable to read from reused packfile")); |
6b8fda2d VM |
808 | |
809 | if (read_pack > to_write) | |
810 | read_pack = to_write; | |
811 | ||
98a3beab | 812 | hashwrite(f, buffer, read_pack); |
6b8fda2d | 813 | to_write -= read_pack; |
657673f1 JK |
814 | |
815 | /* | |
816 | * We don't know the actual number of objects written, | |
817 | * only how many bytes written, how many bytes total, and | |
818 | * how many objects total. So we can fake it by pretending all | |
819 | * objects we are writing are the same size. This gives us a | |
820 | * smooth progress meter, and at the end it matches the true | |
821 | * answer. | |
822 | */ | |
823 | written = reuse_packfile_objects * | |
824 | (((double)(total - to_write)) / total); | |
825 | display_progress(progress_state, written); | |
6b8fda2d VM |
826 | } |
827 | ||
828 | close(fd); | |
657673f1 JK |
829 | written = reuse_packfile_objects; |
830 | display_progress(progress_state, written); | |
6b8fda2d VM |
831 | return reuse_packfile_offset - sizeof(struct pack_header); |
832 | } | |
833 | ||
9cea46cd EW |
834 | static const char no_split_warning[] = N_( |
835 | "disabling bitmap writing, packs are split due to pack.packSizeLimit" | |
836 | ); | |
837 | ||
d01fb92f | 838 | static void write_pack_file(void) |
c323ac7d | 839 | { |
ebe27b13 | 840 | uint32_t i = 0, j; |
98a3beab | 841 | struct hashfile *f; |
6ed7f25e | 842 | off_t offset; |
ebe27b13 | 843 | uint32_t nr_remaining = nr_result; |
f746bae8 | 844 | time_t last_mtime = 0; |
1b4bb16b | 845 | struct object_entry **write_order; |
81a216a5 | 846 | |
bcd7954e | 847 | if (progress > pack_to_stdout) |
754dbc43 | 848 | progress_state = start_progress(_("Writing objects"), nr_result); |
b32fa95f | 849 | ALLOC_ARRAY(written_list, to_pack.nr_objects); |
1b4bb16b | 850 | write_order = compute_write_order(); |
183bdb2c | 851 | |
ebe27b13 | 852 | do { |
188960b4 | 853 | struct object_id oid; |
7ba502c4 | 854 | char *pack_tmp_name = NULL; |
aa7e44bf | 855 | |
cdf9db3c | 856 | if (pack_to_stdout) |
98a3beab | 857 | f = hashfd_throughput(1, "<stdout>", progress_state); |
cdf9db3c JH |
858 | else |
859 | f = create_tmp_packfile(&pack_tmp_name); | |
ebe27b13 | 860 | |
c0ad4657 | 861 | offset = write_pack_header(f, nr_remaining); |
6b8fda2d VM |
862 | |
863 | if (reuse_packfile) { | |
864 | off_t packfile_size; | |
865 | assert(pack_to_stdout); | |
866 | ||
867 | packfile_size = write_reused_pack(f); | |
868 | offset += packfile_size; | |
869 | } | |
870 | ||
ebe27b13 | 871 | nr_written = 0; |
2834bc27 | 872 | for (; i < to_pack.nr_objects; i++) { |
1b4bb16b | 873 | struct object_entry *e = write_order[i]; |
cddec4f8 | 874 | if (write_one(f, e, &offset) == WRITE_ONE_BREAK) |
720c9f7b NP |
875 | break; |
876 | display_progress(progress_state, written); | |
877 | } | |
c553ca25 | 878 | |
ebe27b13 DH |
879 | /* |
880 | * Did we write the wrong # entries in the header? | |
881 | * If so, rewrite it like in fast-import | |
882 | */ | |
54352bb2 | 883 | if (pack_to_stdout) { |
cfe83216 | 884 | finalize_hashfile(f, oid.hash, CSUM_HASH_IN_STREAM | CSUM_CLOSE); |
54352bb2 | 885 | } else if (nr_written == nr_remaining) { |
cfe83216 | 886 | finalize_hashfile(f, oid.hash, CSUM_HASH_IN_STREAM | CSUM_FSYNC | CSUM_CLOSE); |
ebe27b13 | 887 | } else { |
f2af9f5e | 888 | int fd = finalize_hashfile(f, oid.hash, 0); |
188960b4 | 889 | fixup_pack_header_footer(fd, oid.hash, pack_tmp_name, |
890 | nr_written, oid.hash, offset); | |
7ba502c4 | 891 | close(fd); |
9cea46cd EW |
892 | if (write_bitmap_index) { |
893 | warning(_(no_split_warning)); | |
894 | write_bitmap_index = 0; | |
895 | } | |
ebe27b13 DH |
896 | } |
897 | ||
898 | if (!pack_to_stdout) { | |
f746bae8 | 899 | struct stat st; |
58892711 | 900 | struct strbuf tmpname = STRBUF_INIT; |
d01fb92f | 901 | |
f746bae8 NP |
902 | /* |
903 | * Packs are runtime accessed in their mtime | |
904 | * order since newer packs are more likely to contain | |
905 | * younger objects. So if we are creating multiple | |
906 | * packs then we should modify the mtime of later ones | |
907 | * to preserve this property. | |
908 | */ | |
0e990530 | 909 | if (stat(pack_tmp_name, &st) < 0) { |
f616db6a | 910 | warning_errno(_("failed to stat %s"), pack_tmp_name); |
f746bae8 NP |
911 | } else if (!last_mtime) { |
912 | last_mtime = st.st_mtime; | |
913 | } else { | |
914 | struct utimbuf utb; | |
915 | utb.actime = st.st_atime; | |
916 | utb.modtime = --last_mtime; | |
0e990530 | 917 | if (utime(pack_tmp_name, &utb) < 0) |
f616db6a | 918 | warning_errno(_("failed utime() on %s"), pack_tmp_name); |
f746bae8 NP |
919 | } |
920 | ||
58892711 | 921 | strbuf_addf(&tmpname, "%s-", base_name); |
7cc8f971 VM |
922 | |
923 | if (write_bitmap_index) { | |
188960b4 | 924 | bitmap_writer_set_checksum(oid.hash); |
06af3bba NTND |
925 | bitmap_writer_build_type_index( |
926 | &to_pack, written_list, nr_written); | |
7cc8f971 VM |
927 | } |
928 | ||
58892711 | 929 | finish_tmp_packfile(&tmpname, pack_tmp_name, |
0e990530 | 930 | written_list, nr_written, |
188960b4 | 931 | &pack_idx_opts, oid.hash); |
7cc8f971 VM |
932 | |
933 | if (write_bitmap_index) { | |
188960b4 | 934 | strbuf_addf(&tmpname, "%s.bitmap", oid_to_hex(&oid)); |
7cc8f971 VM |
935 | |
936 | stop_progress(&progress_state); | |
937 | ||
938 | bitmap_writer_show_progress(progress); | |
939 | bitmap_writer_reuse_bitmaps(&to_pack); | |
940 | bitmap_writer_select_commits(indexed_commits, indexed_commits_nr, -1); | |
941 | bitmap_writer_build(&to_pack); | |
ae4f07fb | 942 | bitmap_writer_finish(written_list, nr_written, |
58892711 | 943 | tmpname.buf, write_bitmap_options); |
7cc8f971 VM |
944 | write_bitmap_index = 0; |
945 | } | |
946 | ||
58892711 | 947 | strbuf_release(&tmpname); |
7ba502c4 | 948 | free(pack_tmp_name); |
188960b4 | 949 | puts(oid_to_hex(&oid)); |
ebe27b13 DH |
950 | } |
951 | ||
952 | /* mark written objects as written to previous pack */ | |
953 | for (j = 0; j < nr_written; j++) { | |
79814f42 | 954 | written_list[j]->offset = (off_t)-1; |
ebe27b13 DH |
955 | } |
956 | nr_remaining -= nr_written; | |
2834bc27 | 957 | } while (nr_remaining && i < to_pack.nr_objects); |
ebe27b13 DH |
958 | |
959 | free(written_list); | |
1b4bb16b | 960 | free(write_order); |
4d4fcc54 | 961 | stop_progress(&progress_state); |
67c08ce1 | 962 | if (written != nr_result) |
f616db6a NTND |
963 | die(_("wrote %"PRIu32" objects while expecting %"PRIu32), |
964 | written, nr_result); | |
c323ac7d LT |
965 | } |
966 | ||
a74db82e JH |
967 | static int no_try_delta(const char *path) |
968 | { | |
2aef63d3 | 969 | static struct attr_check *check; |
a74db82e | 970 | |
2aef63d3 JH |
971 | if (!check) |
972 | check = attr_check_initl("delta", NULL); | |
d64324cb | 973 | git_check_attr(&the_index, path, check); |
2aef63d3 | 974 | if (ATTR_FALSE(check->items[0].value)) |
a74db82e JH |
975 | return 1; |
976 | return 0; | |
977 | } | |
978 | ||
ce2bc424 JK |
979 | /* |
980 | * When adding an object, check whether we have already added it | |
981 | * to our packing list. If so, we can skip. However, if we are | |
982 | * being asked to excludei t, but the previous mention was to include | |
983 | * it, make sure to adjust its flags and tweak our numbers accordingly. | |
984 | * | |
985 | * As an optimization, we pass out the index position where we would have | |
986 | * found the item, since that saves us from having to look it up again a | |
987 | * few lines later when we want to add the new entry. | |
988 | */ | |
188960b4 | 989 | static int have_duplicate_entry(const struct object_id *oid, |
ce2bc424 JK |
990 | int exclude, |
991 | uint32_t *index_pos) | |
c323ac7d | 992 | { |
c323ac7d | 993 | struct object_entry *entry; |
29b734e4 | 994 | |
188960b4 | 995 | entry = packlist_find(&to_pack, oid->hash, index_pos); |
ce2bc424 | 996 | if (!entry) |
29b734e4 | 997 | return 0; |
ce2bc424 JK |
998 | |
999 | if (exclude) { | |
1000 | if (!entry->preferred_base) | |
1001 | nr_result--; | |
1002 | entry->preferred_base = 1; | |
29b734e4 | 1003 | } |
c323ac7d | 1004 | |
ce2bc424 JK |
1005 | return 1; |
1006 | } | |
1007 | ||
702d1b95 KS |
1008 | static int want_found_object(int exclude, struct packed_git *p) |
1009 | { | |
1010 | if (exclude) | |
1011 | return 1; | |
1012 | if (incremental) | |
1013 | return 0; | |
1014 | ||
1015 | /* | |
1016 | * When asked to do --local (do not include an object that appears in a | |
1017 | * pack we borrow from elsewhere) or --honor-pack-keep (do not include | |
1018 | * an object that appears in a pack marked with .keep), finding a pack | |
1019 | * that matches the criteria is sufficient for us to decide to omit it. | |
1020 | * However, even if this pack does not satisfy the criteria, we need to | |
1021 | * make sure no copy of this object appears in _any_ pack that makes us | |
1022 | * to omit the object, so we need to check all the packs. | |
1023 | * | |
1024 | * We can however first check whether these options can possible matter; | |
1025 | * if they do not matter we know we want the object in generated pack. | |
1026 | * Otherwise, we signal "-1" at the end to tell the caller that we do | |
1027 | * not know either way, and it needs to check more packs. | |
1028 | */ | |
ed7e5fc3 NTND |
1029 | if (!ignore_packed_keep_on_disk && |
1030 | !ignore_packed_keep_in_core && | |
702d1b95 KS |
1031 | (!local || !have_non_local_packs)) |
1032 | return 1; | |
1033 | ||
1034 | if (local && !p->pack_local) | |
1035 | return 0; | |
ed7e5fc3 NTND |
1036 | if (p->pack_local && |
1037 | ((ignore_packed_keep_on_disk && p->pack_keep) || | |
1038 | (ignore_packed_keep_in_core && p->pack_keep_in_core))) | |
702d1b95 KS |
1039 | return 0; |
1040 | ||
1041 | /* we don't know yet; keep looking for more packs */ | |
1042 | return -1; | |
1043 | } | |
1044 | ||
ce2bc424 JK |
1045 | /* |
1046 | * Check whether we want the object in the pack (e.g., we do not want | |
1047 | * objects found in non-local stores if the "--local" option was used). | |
1048 | * | |
702d1b95 KS |
1049 | * If the caller already knows an existing pack it wants to take the object |
1050 | * from, that is passed in *found_pack and *found_offset; otherwise this | |
1051 | * function finds if there is any pack that has the object and returns the pack | |
1052 | * and its offset in these variables. | |
ce2bc424 | 1053 | */ |
188960b4 | 1054 | static int want_object_in_pack(const struct object_id *oid, |
ce2bc424 JK |
1055 | int exclude, |
1056 | struct packed_git **found_pack, | |
1057 | off_t *found_offset) | |
1058 | { | |
702d1b95 | 1059 | int want; |
8865859d | 1060 | struct list_head *pos; |
6a22d521 | 1061 | struct multi_pack_index *m; |
ce2bc424 | 1062 | |
6862ebbf | 1063 | if (!exclude && local && has_loose_object_nonlocal(oid)) |
daae0625 BC |
1064 | return 0; |
1065 | ||
702d1b95 KS |
1066 | /* |
1067 | * If we already know the pack object lives in, start checks from that | |
1068 | * pack - in the usual case when neither --local was given nor .keep files | |
1069 | * are present we will determine the answer right now. | |
1070 | */ | |
1071 | if (*found_pack) { | |
1072 | want = want_found_object(exclude, *found_pack); | |
1073 | if (want != -1) | |
1074 | return want; | |
1075 | } | |
6a22d521 DS |
1076 | |
1077 | for (m = get_multi_pack_index(the_repository); m; m = m->next) { | |
1078 | struct pack_entry e; | |
1079 | if (fill_midx_entry(oid, &e, m)) { | |
1080 | struct packed_git *p = e.p; | |
1081 | off_t offset; | |
1082 | ||
1083 | if (p == *found_pack) | |
1084 | offset = *found_offset; | |
1085 | else | |
1086 | offset = find_pack_entry_one(oid->hash, p); | |
1087 | ||
1088 | if (offset) { | |
1089 | if (!*found_pack) { | |
1090 | if (!is_pack_valid(p)) | |
1091 | continue; | |
1092 | *found_offset = offset; | |
1093 | *found_pack = p; | |
1094 | } | |
1095 | want = want_found_object(exclude, p); | |
1096 | if (want != -1) | |
1097 | return want; | |
1098 | } | |
1099 | } | |
1100 | } | |
1101 | ||
a80d72db | 1102 | list_for_each(pos, get_packed_git_mru(the_repository)) { |
ec2dd32c | 1103 | struct packed_git *p = list_entry(pos, struct packed_git, mru); |
702d1b95 KS |
1104 | off_t offset; |
1105 | ||
1106 | if (p == *found_pack) | |
1107 | offset = *found_offset; | |
1108 | else | |
188960b4 | 1109 | offset = find_pack_entry_one(oid->hash, p); |
702d1b95 | 1110 | |
5c49c116 | 1111 | if (offset) { |
ce2bc424 | 1112 | if (!*found_pack) { |
319b678a | 1113 | if (!is_pack_valid(p)) |
4c080182 | 1114 | continue; |
ce2bc424 JK |
1115 | *found_offset = offset; |
1116 | *found_pack = p; | |
64560374 | 1117 | } |
702d1b95 | 1118 | want = want_found_object(exclude, p); |
e6e24c94 | 1119 | if (!exclude && want > 0) |
a80d72db SB |
1120 | list_move(&p->mru, |
1121 | get_packed_git_mru(the_repository)); | |
702d1b95 KS |
1122 | if (want != -1) |
1123 | return want; | |
64560374 LT |
1124 | } |
1125 | } | |
eb019375 | 1126 | |
ce2bc424 JK |
1127 | return 1; |
1128 | } | |
1129 | ||
188960b4 | 1130 | static void create_object_entry(const struct object_id *oid, |
ce2bc424 JK |
1131 | enum object_type type, |
1132 | uint32_t hash, | |
1133 | int exclude, | |
1134 | int no_try_delta, | |
1135 | uint32_t index_pos, | |
1136 | struct packed_git *found_pack, | |
1137 | off_t found_offset) | |
1138 | { | |
1139 | struct object_entry *entry; | |
29b734e4 | 1140 | |
188960b4 | 1141 | entry = packlist_alloc(&to_pack, oid->hash, index_pos); |
27225f2e | 1142 | entry->hash = hash; |
fd9b1bae | 1143 | oe_set_type(entry, type); |
29b734e4 NP |
1144 | if (exclude) |
1145 | entry->preferred_base = 1; | |
81a216a5 NP |
1146 | else |
1147 | nr_result++; | |
29b734e4 | 1148 | if (found_pack) { |
43fa44fa | 1149 | oe_set_in_pack(&to_pack, entry, found_pack); |
29b734e4 NP |
1150 | entry->in_pack_offset = found_offset; |
1151 | } | |
7a979d99 | 1152 | |
ce2bc424 JK |
1153 | entry->no_try_delta = no_try_delta; |
1154 | } | |
29b734e4 | 1155 | |
373c67da JK |
1156 | static const char no_closure_warning[] = N_( |
1157 | "disabling bitmap writing, as some objects are not being packed" | |
1158 | ); | |
1159 | ||
188960b4 | 1160 | static int add_object_entry(const struct object_id *oid, enum object_type type, |
ce2bc424 JK |
1161 | const char *name, int exclude) |
1162 | { | |
702d1b95 KS |
1163 | struct packed_git *found_pack = NULL; |
1164 | off_t found_offset = 0; | |
ce2bc424 | 1165 | uint32_t index_pos; |
7a979d99 | 1166 | |
5af05043 NTND |
1167 | display_progress(progress_state, ++nr_seen); |
1168 | ||
188960b4 | 1169 | if (have_duplicate_entry(oid, exclude, &index_pos)) |
ce2bc424 | 1170 | return 0; |
a74db82e | 1171 | |
188960b4 | 1172 | if (!want_object_in_pack(oid, exclude, &found_pack, &found_offset)) { |
373c67da JK |
1173 | /* The pack is missing an object, so it will not have closure */ |
1174 | if (write_bitmap_index) { | |
1175 | warning(_(no_closure_warning)); | |
1176 | write_bitmap_index = 0; | |
1177 | } | |
ce2bc424 | 1178 | return 0; |
373c67da | 1179 | } |
29b734e4 | 1180 | |
188960b4 | 1181 | create_object_entry(oid, type, pack_name_hash(name), |
ce2bc424 JK |
1182 | exclude, name && no_try_delta(name), |
1183 | index_pos, found_pack, found_offset); | |
29b734e4 | 1184 | return 1; |
c323ac7d LT |
1185 | } |
1186 | ||
20664967 | 1187 | static int add_object_entry_from_bitmap(const struct object_id *oid, |
6b8fda2d VM |
1188 | enum object_type type, |
1189 | int flags, uint32_t name_hash, | |
1190 | struct packed_git *pack, off_t offset) | |
1191 | { | |
1192 | uint32_t index_pos; | |
1193 | ||
5af05043 NTND |
1194 | display_progress(progress_state, ++nr_seen); |
1195 | ||
188960b4 | 1196 | if (have_duplicate_entry(oid, 0, &index_pos)) |
6b8fda2d VM |
1197 | return 0; |
1198 | ||
188960b4 | 1199 | if (!want_object_in_pack(oid, 0, &pack, &offset)) |
702d1b95 KS |
1200 | return 0; |
1201 | ||
188960b4 | 1202 | create_object_entry(oid, type, name_hash, 0, 0, index_pos, pack, offset); |
29b734e4 | 1203 | return 1; |
c323ac7d LT |
1204 | } |
1205 | ||
5379a5c5 | 1206 | struct pbase_tree_cache { |
188960b4 | 1207 | struct object_id oid; |
5379a5c5 JH |
1208 | int ref; |
1209 | int temporary; | |
1210 | void *tree_data; | |
1211 | unsigned long tree_size; | |
1212 | }; | |
1213 | ||
1214 | static struct pbase_tree_cache *(pbase_tree_cache[256]); | |
188960b4 | 1215 | static int pbase_tree_cache_ix(const struct object_id *oid) |
5379a5c5 | 1216 | { |
188960b4 | 1217 | return oid->hash[0] % ARRAY_SIZE(pbase_tree_cache); |
5379a5c5 JH |
1218 | } |
1219 | static int pbase_tree_cache_ix_incr(int ix) | |
1220 | { | |
1221 | return (ix+1) % ARRAY_SIZE(pbase_tree_cache); | |
1222 | } | |
1223 | ||
1224 | static struct pbase_tree { | |
1225 | struct pbase_tree *next; | |
1226 | /* This is a phony "cache" entry; we are not | |
01689909 | 1227 | * going to evict it or find it through _get() |
5379a5c5 JH |
1228 | * mechanism -- this is for the toplevel node that |
1229 | * would almost always change with any commit. | |
1230 | */ | |
1231 | struct pbase_tree_cache pcache; | |
1232 | } *pbase_tree; | |
1233 | ||
188960b4 | 1234 | static struct pbase_tree_cache *pbase_tree_get(const struct object_id *oid) |
5379a5c5 JH |
1235 | { |
1236 | struct pbase_tree_cache *ent, *nent; | |
1237 | void *data; | |
1238 | unsigned long size; | |
21666f1a | 1239 | enum object_type type; |
5379a5c5 | 1240 | int neigh; |
188960b4 | 1241 | int my_ix = pbase_tree_cache_ix(oid); |
5379a5c5 JH |
1242 | int available_ix = -1; |
1243 | ||
1244 | /* pbase-tree-cache acts as a limited hashtable. | |
1245 | * your object will be found at your index or within a few | |
1246 | * slots after that slot if it is cached. | |
1247 | */ | |
1248 | for (neigh = 0; neigh < 8; neigh++) { | |
1249 | ent = pbase_tree_cache[my_ix]; | |
4a7e27e9 | 1250 | if (ent && oideq(&ent->oid, oid)) { |
5379a5c5 JH |
1251 | ent->ref++; |
1252 | return ent; | |
1253 | } | |
1254 | else if (((available_ix < 0) && (!ent || !ent->ref)) || | |
1255 | ((0 <= available_ix) && | |
1256 | (!ent && pbase_tree_cache[available_ix]))) | |
1257 | available_ix = my_ix; | |
1258 | if (!ent) | |
1259 | break; | |
1260 | my_ix = pbase_tree_cache_ix_incr(my_ix); | |
1261 | } | |
1262 | ||
1263 | /* Did not find one. Either we got a bogus request or | |
1264 | * we need to read and perhaps cache. | |
1265 | */ | |
b4f5aca4 | 1266 | data = read_object_file(oid, &type, &size); |
5379a5c5 JH |
1267 | if (!data) |
1268 | return NULL; | |
21666f1a | 1269 | if (type != OBJ_TREE) { |
5379a5c5 JH |
1270 | free(data); |
1271 | return NULL; | |
1272 | } | |
1273 | ||
1274 | /* We need to either cache or return a throwaway copy */ | |
1275 | ||
1276 | if (available_ix < 0) | |
1277 | ent = NULL; | |
1278 | else { | |
1279 | ent = pbase_tree_cache[available_ix]; | |
1280 | my_ix = available_ix; | |
1281 | } | |
1282 | ||
1283 | if (!ent) { | |
1284 | nent = xmalloc(sizeof(*nent)); | |
1285 | nent->temporary = (available_ix < 0); | |
1286 | } | |
1287 | else { | |
1288 | /* evict and reuse */ | |
1289 | free(ent->tree_data); | |
1290 | nent = ent; | |
1291 | } | |
188960b4 | 1292 | oidcpy(&nent->oid, oid); |
5379a5c5 JH |
1293 | nent->tree_data = data; |
1294 | nent->tree_size = size; | |
1295 | nent->ref = 1; | |
1296 | if (!nent->temporary) | |
1297 | pbase_tree_cache[my_ix] = nent; | |
1298 | return nent; | |
1299 | } | |
1300 | ||
1301 | static void pbase_tree_put(struct pbase_tree_cache *cache) | |
1302 | { | |
1303 | if (!cache->temporary) { | |
1304 | cache->ref--; | |
1305 | return; | |
1306 | } | |
1307 | free(cache->tree_data); | |
1308 | free(cache); | |
1309 | } | |
1310 | ||
1311 | static int name_cmp_len(const char *name) | |
1312 | { | |
1313 | int i; | |
1314 | for (i = 0; name[i] && name[i] != '\n' && name[i] != '/'; i++) | |
1315 | ; | |
1316 | return i; | |
1317 | } | |
1318 | ||
1319 | static void add_pbase_object(struct tree_desc *tree, | |
5379a5c5 | 1320 | const char *name, |
ce0bd642 LT |
1321 | int cmplen, |
1322 | const char *fullname) | |
3f9ac8d2 | 1323 | { |
4c068a98 | 1324 | struct name_entry entry; |
8a5a8d6c | 1325 | int cmp; |
4c068a98 LT |
1326 | |
1327 | while (tree_entry(tree,&entry)) { | |
1211be6b LT |
1328 | if (S_ISGITLINK(entry.mode)) |
1329 | continue; | |
0de16337 | 1330 | cmp = tree_entry_len(&entry) != cmplen ? 1 : |
8a5a8d6c NP |
1331 | memcmp(name, entry.path, cmplen); |
1332 | if (cmp > 0) | |
7a979d99 | 1333 | continue; |
8a5a8d6c NP |
1334 | if (cmp < 0) |
1335 | return; | |
5379a5c5 | 1336 | if (name[cmplen] != '/') { |
188960b4 | 1337 | add_object_entry(entry.oid, |
4d1012c3 | 1338 | object_type(entry.mode), |
bc32fed5 | 1339 | fullname, 1); |
5379a5c5 JH |
1340 | return; |
1341 | } | |
8a5a8d6c | 1342 | if (S_ISDIR(entry.mode)) { |
7a979d99 | 1343 | struct tree_desc sub; |
5379a5c5 JH |
1344 | struct pbase_tree_cache *tree; |
1345 | const char *down = name+cmplen+1; | |
1346 | int downlen = name_cmp_len(down); | |
1347 | ||
188960b4 | 1348 | tree = pbase_tree_get(entry.oid); |
5379a5c5 JH |
1349 | if (!tree) |
1350 | return; | |
6fda5e51 | 1351 | init_tree_desc(&sub, tree->tree_data, tree->tree_size); |
5379a5c5 | 1352 | |
ce0bd642 | 1353 | add_pbase_object(&sub, down, downlen, fullname); |
5379a5c5 JH |
1354 | pbase_tree_put(tree); |
1355 | } | |
1356 | } | |
1357 | } | |
1d6b38cc | 1358 | |
5379a5c5 JH |
1359 | static unsigned *done_pbase_paths; |
1360 | static int done_pbase_paths_num; | |
1361 | static int done_pbase_paths_alloc; | |
1362 | static int done_pbase_path_pos(unsigned hash) | |
1363 | { | |
1364 | int lo = 0; | |
1365 | int hi = done_pbase_paths_num; | |
1366 | while (lo < hi) { | |
19716b21 | 1367 | int mi = lo + (hi - lo) / 2; |
5379a5c5 JH |
1368 | if (done_pbase_paths[mi] == hash) |
1369 | return mi; | |
1370 | if (done_pbase_paths[mi] < hash) | |
1371 | hi = mi; | |
1372 | else | |
1373 | lo = mi + 1; | |
1374 | } | |
1375 | return -lo-1; | |
1376 | } | |
1377 | ||
1378 | static int check_pbase_path(unsigned hash) | |
1379 | { | |
c7b07805 | 1380 | int pos = done_pbase_path_pos(hash); |
5379a5c5 JH |
1381 | if (0 <= pos) |
1382 | return 1; | |
1383 | pos = -pos - 1; | |
25e19407 DD |
1384 | ALLOC_GROW(done_pbase_paths, |
1385 | done_pbase_paths_num + 1, | |
1386 | done_pbase_paths_alloc); | |
5379a5c5 JH |
1387 | done_pbase_paths_num++; |
1388 | if (pos < done_pbase_paths_num) | |
f331ab9d RS |
1389 | MOVE_ARRAY(done_pbase_paths + pos + 1, done_pbase_paths + pos, |
1390 | done_pbase_paths_num - pos - 1); | |
5379a5c5 JH |
1391 | done_pbase_paths[pos] = hash; |
1392 | return 0; | |
1393 | } | |
1394 | ||
bc32fed5 | 1395 | static void add_preferred_base_object(const char *name) |
5379a5c5 JH |
1396 | { |
1397 | struct pbase_tree *it; | |
8a5a8d6c | 1398 | int cmplen; |
68fb36eb | 1399 | unsigned hash = pack_name_hash(name); |
5379a5c5 | 1400 | |
8a5a8d6c | 1401 | if (!num_preferred_base || check_pbase_path(hash)) |
5379a5c5 JH |
1402 | return; |
1403 | ||
8a5a8d6c | 1404 | cmplen = name_cmp_len(name); |
5379a5c5 JH |
1405 | for (it = pbase_tree; it; it = it->next) { |
1406 | if (cmplen == 0) { | |
188960b4 | 1407 | add_object_entry(&it->pcache.oid, OBJ_TREE, NULL, 1); |
5379a5c5 JH |
1408 | } |
1409 | else { | |
1410 | struct tree_desc tree; | |
6fda5e51 | 1411 | init_tree_desc(&tree, it->pcache.tree_data, it->pcache.tree_size); |
ce0bd642 | 1412 | add_pbase_object(&tree, name, cmplen, name); |
7a979d99 | 1413 | } |
3f9ac8d2 | 1414 | } |
3f9ac8d2 JH |
1415 | } |
1416 | ||
188960b4 | 1417 | static void add_preferred_base(struct object_id *oid) |
3f9ac8d2 | 1418 | { |
5379a5c5 JH |
1419 | struct pbase_tree *it; |
1420 | void *data; | |
1421 | unsigned long size; | |
188960b4 | 1422 | struct object_id tree_oid; |
1d6b38cc | 1423 | |
8d1d8f83 JH |
1424 | if (window <= num_preferred_base++) |
1425 | return; | |
1426 | ||
02f0547e | 1427 | data = read_object_with_reference(oid, tree_type, &size, &tree_oid); |
5379a5c5 | 1428 | if (!data) |
7a979d99 | 1429 | return; |
5379a5c5 JH |
1430 | |
1431 | for (it = pbase_tree; it; it = it->next) { | |
4a7e27e9 | 1432 | if (oideq(&it->pcache.oid, &tree_oid)) { |
5379a5c5 JH |
1433 | free(data); |
1434 | return; | |
1435 | } | |
1436 | } | |
1437 | ||
1438 | it = xcalloc(1, sizeof(*it)); | |
1439 | it->next = pbase_tree; | |
1440 | pbase_tree = it; | |
1441 | ||
188960b4 | 1442 | oidcpy(&it->pcache.oid, &tree_oid); |
5379a5c5 JH |
1443 | it->pcache.tree_data = data; |
1444 | it->pcache.tree_size = size; | |
3f9ac8d2 JH |
1445 | } |
1446 | ||
0ef95f72 NP |
1447 | static void cleanup_preferred_base(void) |
1448 | { | |
1449 | struct pbase_tree *it; | |
1450 | unsigned i; | |
1451 | ||
1452 | it = pbase_tree; | |
1453 | pbase_tree = NULL; | |
1454 | while (it) { | |
095b3b2c BW |
1455 | struct pbase_tree *tmp = it; |
1456 | it = tmp->next; | |
1457 | free(tmp->pcache.tree_data); | |
1458 | free(tmp); | |
0ef95f72 NP |
1459 | } |
1460 | ||
1461 | for (i = 0; i < ARRAY_SIZE(pbase_tree_cache); i++) { | |
1462 | if (!pbase_tree_cache[i]) | |
1463 | continue; | |
1464 | free(pbase_tree_cache[i]->tree_data); | |
6a83d902 | 1465 | FREE_AND_NULL(pbase_tree_cache[i]); |
0ef95f72 NP |
1466 | } |
1467 | ||
6a83d902 | 1468 | FREE_AND_NULL(done_pbase_paths); |
0ef95f72 NP |
1469 | done_pbase_paths_num = done_pbase_paths_alloc = 0; |
1470 | } | |
1471 | ||
2fa233a5 JK |
1472 | /* |
1473 | * Return 1 iff the object specified by "delta" can be sent | |
1474 | * literally as a delta against the base in "base_sha1". If | |
1475 | * so, then *base_out will point to the entry in our packing | |
1476 | * list, or NULL if we must use the external-base list. | |
1477 | * | |
1478 | * Depth value does not matter - find_deltas() will | |
1479 | * never consider reused delta as the base object to | |
1480 | * deltify other objects against, in order to avoid | |
1481 | * circular deltas. | |
1482 | */ | |
1483 | static int can_reuse_delta(const unsigned char *base_sha1, | |
1484 | struct object_entry *delta, | |
1485 | struct object_entry **base_out) | |
1486 | { | |
1487 | struct object_entry *base; | |
1488 | ||
1489 | if (!base_sha1) | |
1490 | return 0; | |
1491 | ||
1492 | /* | |
1493 | * First see if we're already sending the base (or it's explicitly in | |
1494 | * our "excluded" list). | |
1495 | */ | |
1496 | base = packlist_find(&to_pack, base_sha1, NULL); | |
1497 | if (base) { | |
1498 | if (!in_same_island(&delta->idx.oid, &base->idx.oid)) | |
1499 | return 0; | |
1500 | *base_out = base; | |
1501 | return 1; | |
1502 | } | |
1503 | ||
1504 | /* | |
1505 | * Otherwise, reachability bitmaps may tell us if the receiver has it, | |
1506 | * even if it was buried too deep in history to make it into the | |
1507 | * packing list. | |
1508 | */ | |
1509 | if (thin && bitmap_has_sha1_in_uninteresting(bitmap_git, base_sha1)) { | |
1510 | if (use_delta_islands) { | |
1511 | struct object_id base_oid; | |
1512 | hashcpy(base_oid.hash, base_sha1); | |
1513 | if (!in_same_island(&delta->idx.oid, &base_oid)) | |
1514 | return 0; | |
1515 | } | |
1516 | *base_out = NULL; | |
1517 | return 1; | |
1518 | } | |
1519 | ||
1520 | return 0; | |
1521 | } | |
1522 | ||
c323ac7d LT |
1523 | static void check_object(struct object_entry *entry) |
1524 | { | |
ac77d0c3 NTND |
1525 | unsigned long canonical_size; |
1526 | ||
43fa44fa NTND |
1527 | if (IN_PACK(entry)) { |
1528 | struct packed_git *p = IN_PACK(entry); | |
03e79c88 | 1529 | struct pack_window *w_curs = NULL; |
5c49c116 NP |
1530 | const unsigned char *base_ref = NULL; |
1531 | struct object_entry *base_entry; | |
1532 | unsigned long used, used_0; | |
ef49a7a0 | 1533 | unsigned long avail; |
5c49c116 NP |
1534 | off_t ofs; |
1535 | unsigned char *buf, c; | |
fd9b1bae | 1536 | enum object_type type; |
27a7d067 | 1537 | unsigned long in_pack_size; |
780e6e73 | 1538 | |
6777a59f | 1539 | buf = use_pack(p, &w_curs, entry->in_pack_offset, &avail); |
ab7cd7bb | 1540 | |
5c49c116 | 1541 | /* |
fa736f72 NP |
1542 | * We want in_pack_type even if we do not reuse delta |
1543 | * since non-delta representations could still be reused. | |
ab7cd7bb | 1544 | */ |
09ded04b | 1545 | used = unpack_object_header_buffer(buf, avail, |
fd9b1bae | 1546 | &type, |
27a7d067 | 1547 | &in_pack_size); |
03d66015 NP |
1548 | if (used == 0) |
1549 | goto give_up; | |
ab7cd7bb | 1550 | |
fd9b1bae NTND |
1551 | if (type < 0) |
1552 | BUG("invalid type %d", type); | |
1553 | entry->in_pack_type = type; | |
1554 | ||
5c49c116 NP |
1555 | /* |
1556 | * Determine if this is a delta and if so whether we can | |
1557 | * reuse it or not. Otherwise let's find out as cheaply as | |
1558 | * possible what the actual type and size for this object is. | |
3f9ac8d2 | 1559 | */ |
5c49c116 NP |
1560 | switch (entry->in_pack_type) { |
1561 | default: | |
1562 | /* Not a delta hence we've already got all we need. */ | |
fd9b1bae | 1563 | oe_set_type(entry, entry->in_pack_type); |
ac77d0c3 | 1564 | SET_SIZE(entry, in_pack_size); |
5c49c116 | 1565 | entry->in_pack_header_size = used; |
fd9b1bae | 1566 | if (oe_type(entry) < OBJ_COMMIT || oe_type(entry) > OBJ_BLOB) |
03d66015 | 1567 | goto give_up; |
5c49c116 NP |
1568 | unuse_pack(&w_curs); |
1569 | return; | |
1570 | case OBJ_REF_DELTA: | |
a7de7130 | 1571 | if (reuse_delta && !entry->preferred_base) |
5c49c116 NP |
1572 | base_ref = use_pack(p, &w_curs, |
1573 | entry->in_pack_offset + used, NULL); | |
41179100 | 1574 | entry->in_pack_header_size = used + the_hash_algo->rawsz; |
5c49c116 NP |
1575 | break; |
1576 | case OBJ_OFS_DELTA: | |
1577 | buf = use_pack(p, &w_curs, | |
1578 | entry->in_pack_offset + used, NULL); | |
1579 | used_0 = 0; | |
1580 | c = buf[used_0++]; | |
1581 | ofs = c & 127; | |
1582 | while (c & 128) { | |
1583 | ofs += 1; | |
03d66015 | 1584 | if (!ofs || MSB(ofs, 7)) { |
f616db6a | 1585 | error(_("delta base offset overflow in pack for %s"), |
e6a492b7 | 1586 | oid_to_hex(&entry->idx.oid)); |
03d66015 NP |
1587 | goto give_up; |
1588 | } | |
5c49c116 NP |
1589 | c = buf[used_0++]; |
1590 | ofs = (ofs << 7) + (c & 127); | |
780e6e73 | 1591 | } |
5c49c116 | 1592 | ofs = entry->in_pack_offset - ofs; |
03d66015 | 1593 | if (ofs <= 0 || ofs >= entry->in_pack_offset) { |
f616db6a | 1594 | error(_("delta base offset out of bound for %s"), |
e6a492b7 | 1595 | oid_to_hex(&entry->idx.oid)); |
03d66015 NP |
1596 | goto give_up; |
1597 | } | |
a7de7130 | 1598 | if (reuse_delta && !entry->preferred_base) { |
3449f8c4 NP |
1599 | struct revindex_entry *revidx; |
1600 | revidx = find_pack_revindex(p, ofs); | |
08698b1e NP |
1601 | if (!revidx) |
1602 | goto give_up; | |
3449f8c4 NP |
1603 | base_ref = nth_packed_object_sha1(p, revidx->nr); |
1604 | } | |
5c49c116 NP |
1605 | entry->in_pack_header_size = used + used_0; |
1606 | break; | |
780e6e73 | 1607 | } |
780e6e73 | 1608 | |
2fa233a5 | 1609 | if (can_reuse_delta(base_ref, entry, &base_entry)) { |
fd9b1bae | 1610 | oe_set_type(entry, entry->in_pack_type); |
ac77d0c3 | 1611 | SET_SIZE(entry, in_pack_size); /* delta size */ |
0aca34e8 | 1612 | SET_DELTA_SIZE(entry, in_pack_size); |
6a1e32d5 JK |
1613 | |
1614 | if (base_entry) { | |
1615 | SET_DELTA(entry, base_entry); | |
1616 | entry->delta_sibling_idx = base_entry->delta_child_idx; | |
1617 | SET_DELTA_CHILD(base_entry, entry); | |
1618 | } else { | |
1619 | SET_DELTA_EXT(entry, base_ref); | |
1620 | } | |
1621 | ||
5c49c116 NP |
1622 | unuse_pack(&w_curs); |
1623 | return; | |
1624 | } | |
ab7cd7bb | 1625 | |
fd9b1bae | 1626 | if (oe_type(entry)) { |
27a7d067 NTND |
1627 | off_t delta_pos; |
1628 | ||
5c49c116 NP |
1629 | /* |
1630 | * This must be a delta and we already know what the | |
1631 | * final object type is. Let's extract the actual | |
1632 | * object size from the delta header. | |
1633 | */ | |
27a7d067 | 1634 | delta_pos = entry->in_pack_offset + entry->in_pack_header_size; |
ac77d0c3 NTND |
1635 | canonical_size = get_size_from_delta(p, &w_curs, delta_pos); |
1636 | if (canonical_size == 0) | |
03d66015 | 1637 | goto give_up; |
ac77d0c3 | 1638 | SET_SIZE(entry, canonical_size); |
5c49c116 | 1639 | unuse_pack(&w_curs); |
3f9ac8d2 JH |
1640 | return; |
1641 | } | |
5c49c116 NP |
1642 | |
1643 | /* | |
1644 | * No choice but to fall back to the recursive delta walk | |
1645 | * with sha1_object_info() to find about the object type | |
1646 | * at this point... | |
1647 | */ | |
03d66015 | 1648 | give_up: |
5c49c116 | 1649 | unuse_pack(&w_curs); |
36e4d74a | 1650 | } |
3f9ac8d2 | 1651 | |
ad635e82 JH |
1652 | oe_set_type(entry, |
1653 | oid_object_info(the_repository, &entry->idx.oid, &canonical_size)); | |
ac77d0c3 NTND |
1654 | if (entry->type_valid) { |
1655 | SET_SIZE(entry, canonical_size); | |
1656 | } else { | |
1657 | /* | |
1658 | * Bad object type is checked in prepare_pack(). This is | |
1659 | * to permit a missing preferred base object to be ignored | |
1660 | * as a preferred base. Doing so can result in a larger | |
1661 | * pack file, but the transfer will still take place. | |
1662 | */ | |
1663 | } | |
3f9ac8d2 JH |
1664 | } |
1665 | ||
5c49c116 NP |
1666 | static int pack_offset_sort(const void *_a, const void *_b) |
1667 | { | |
1668 | const struct object_entry *a = *(struct object_entry **)_a; | |
1669 | const struct object_entry *b = *(struct object_entry **)_b; | |
43fa44fa NTND |
1670 | const struct packed_git *a_in_pack = IN_PACK(a); |
1671 | const struct packed_git *b_in_pack = IN_PACK(b); | |
5c49c116 NP |
1672 | |
1673 | /* avoid filesystem trashing with loose objects */ | |
43fa44fa | 1674 | if (!a_in_pack && !b_in_pack) |
e6a492b7 | 1675 | return oidcmp(&a->idx.oid, &b->idx.oid); |
5c49c116 | 1676 | |
43fa44fa | 1677 | if (a_in_pack < b_in_pack) |
5c49c116 | 1678 | return -1; |
43fa44fa | 1679 | if (a_in_pack > b_in_pack) |
5c49c116 NP |
1680 | return 1; |
1681 | return a->in_pack_offset < b->in_pack_offset ? -1 : | |
1682 | (a->in_pack_offset > b->in_pack_offset); | |
1683 | } | |
1684 | ||
4cf2143e JK |
1685 | /* |
1686 | * Drop an on-disk delta we were planning to reuse. Naively, this would | |
1687 | * just involve blanking out the "delta" field, but we have to deal | |
1688 | * with some extra book-keeping: | |
1689 | * | |
1690 | * 1. Removing ourselves from the delta_sibling linked list. | |
1691 | * | |
1692 | * 2. Updating our size/type to the non-delta representation. These were | |
1693 | * either not recorded initially (size) or overwritten with the delta type | |
1694 | * (type) when check_object() decided to reuse the delta. | |
7dbabbbe JK |
1695 | * |
1696 | * 3. Resetting our delta depth, as we are now a base object. | |
4cf2143e JK |
1697 | */ |
1698 | static void drop_reused_delta(struct object_entry *entry) | |
1699 | { | |
898eba5e | 1700 | unsigned *idx = &to_pack.objects[entry->delta_idx - 1].delta_child_idx; |
4cf2143e | 1701 | struct object_info oi = OBJECT_INFO_INIT; |
fd9b1bae | 1702 | enum object_type type; |
ac77d0c3 | 1703 | unsigned long size; |
4cf2143e | 1704 | |
898eba5e NTND |
1705 | while (*idx) { |
1706 | struct object_entry *oe = &to_pack.objects[*idx - 1]; | |
4cf2143e | 1707 | |
898eba5e NTND |
1708 | if (oe == entry) |
1709 | *idx = oe->delta_sibling_idx; | |
4cf2143e | 1710 | else |
898eba5e | 1711 | idx = &oe->delta_sibling_idx; |
4cf2143e | 1712 | } |
898eba5e | 1713 | SET_DELTA(entry, NULL); |
7dbabbbe | 1714 | entry->depth = 0; |
4cf2143e | 1715 | |
ac77d0c3 | 1716 | oi.sizep = &size; |
fd9b1bae | 1717 | oi.typep = &type; |
ad635e82 | 1718 | if (packed_object_info(the_repository, IN_PACK(entry), entry->in_pack_offset, &oi) < 0) { |
4cf2143e JK |
1719 | /* |
1720 | * We failed to get the info from this pack for some reason; | |
1721 | * fall back to sha1_object_info, which may find another copy. | |
fd9b1bae | 1722 | * And if that fails, the error will be recorded in oe_type(entry) |
4cf2143e JK |
1723 | * and dealt with in prepare_pack(). |
1724 | */ | |
ad635e82 JH |
1725 | oe_set_type(entry, |
1726 | oid_object_info(the_repository, &entry->idx.oid, &size)); | |
fd9b1bae NTND |
1727 | } else { |
1728 | oe_set_type(entry, type); | |
4cf2143e | 1729 | } |
ac77d0c3 | 1730 | SET_SIZE(entry, size); |
4cf2143e JK |
1731 | } |
1732 | ||
1733 | /* | |
1734 | * Follow the chain of deltas from this entry onward, throwing away any links | |
1735 | * that cause us to hit a cycle (as determined by the DFS state flags in | |
1736 | * the entries). | |
7dbabbbe JK |
1737 | * |
1738 | * We also detect too-long reused chains that would violate our --depth | |
1739 | * limit. | |
4cf2143e JK |
1740 | */ |
1741 | static void break_delta_chains(struct object_entry *entry) | |
1742 | { | |
42b766d7 JK |
1743 | /* |
1744 | * The actual depth of each object we will write is stored as an int, | |
1745 | * as it cannot exceed our int "depth" limit. But before we break | |
1746 | * changes based no that limit, we may potentially go as deep as the | |
1747 | * number of objects, which is elsewhere bounded to a uint32_t. | |
1748 | */ | |
1749 | uint32_t total_depth; | |
1750 | struct object_entry *cur, *next; | |
1751 | ||
1752 | for (cur = entry, total_depth = 0; | |
1753 | cur; | |
898eba5e | 1754 | cur = DELTA(cur), total_depth++) { |
42b766d7 JK |
1755 | if (cur->dfs_state == DFS_DONE) { |
1756 | /* | |
1757 | * We've already seen this object and know it isn't | |
1758 | * part of a cycle. We do need to append its depth | |
1759 | * to our count. | |
1760 | */ | |
1761 | total_depth += cur->depth; | |
1762 | break; | |
1763 | } | |
4cf2143e | 1764 | |
4cf2143e | 1765 | /* |
42b766d7 JK |
1766 | * We break cycles before looping, so an ACTIVE state (or any |
1767 | * other cruft which made its way into the state variable) | |
1768 | * is a bug. | |
4cf2143e | 1769 | */ |
42b766d7 | 1770 | if (cur->dfs_state != DFS_NONE) |
033abf97 | 1771 | BUG("confusing delta dfs state in first pass: %d", |
42b766d7 | 1772 | cur->dfs_state); |
4cf2143e | 1773 | |
4cf2143e | 1774 | /* |
42b766d7 JK |
1775 | * Now we know this is the first time we've seen the object. If |
1776 | * it's not a delta, we're done traversing, but we'll mark it | |
1777 | * done to save time on future traversals. | |
4cf2143e | 1778 | */ |
898eba5e | 1779 | if (!DELTA(cur)) { |
42b766d7 JK |
1780 | cur->dfs_state = DFS_DONE; |
1781 | break; | |
1782 | } | |
4cf2143e | 1783 | |
4cf2143e | 1784 | /* |
42b766d7 JK |
1785 | * Mark ourselves as active and see if the next step causes |
1786 | * us to cycle to another active object. It's important to do | |
1787 | * this _before_ we loop, because it impacts where we make the | |
1788 | * cut, and thus how our total_depth counter works. | |
1789 | * E.g., We may see a partial loop like: | |
1790 | * | |
1791 | * A -> B -> C -> D -> B | |
1792 | * | |
1793 | * Cutting B->C breaks the cycle. But now the depth of A is | |
1794 | * only 1, and our total_depth counter is at 3. The size of the | |
1795 | * error is always one less than the size of the cycle we | |
1796 | * broke. Commits C and D were "lost" from A's chain. | |
1797 | * | |
1798 | * If we instead cut D->B, then the depth of A is correct at 3. | |
1799 | * We keep all commits in the chain that we examined. | |
4cf2143e | 1800 | */ |
42b766d7 | 1801 | cur->dfs_state = DFS_ACTIVE; |
898eba5e | 1802 | if (DELTA(cur)->dfs_state == DFS_ACTIVE) { |
42b766d7 JK |
1803 | drop_reused_delta(cur); |
1804 | cur->dfs_state = DFS_DONE; | |
1805 | break; | |
7dbabbbe | 1806 | } |
42b766d7 | 1807 | } |
7dbabbbe | 1808 | |
42b766d7 JK |
1809 | /* |
1810 | * And now that we've gone all the way to the bottom of the chain, we | |
1811 | * need to clear the active flags and set the depth fields as | |
1812 | * appropriate. Unlike the loop above, which can quit when it drops a | |
1813 | * delta, we need to keep going to look for more depth cuts. So we need | |
1814 | * an extra "next" pointer to keep going after we reset cur->delta. | |
1815 | */ | |
1816 | for (cur = entry; cur; cur = next) { | |
898eba5e | 1817 | next = DELTA(cur); |
4cf2143e | 1818 | |
42b766d7 JK |
1819 | /* |
1820 | * We should have a chain of zero or more ACTIVE states down to | |
1821 | * a final DONE. We can quit after the DONE, because either it | |
1822 | * has no bases, or we've already handled them in a previous | |
1823 | * call. | |
1824 | */ | |
1825 | if (cur->dfs_state == DFS_DONE) | |
1826 | break; | |
1827 | else if (cur->dfs_state != DFS_ACTIVE) | |
033abf97 | 1828 | BUG("confusing delta dfs state in second pass: %d", |
42b766d7 | 1829 | cur->dfs_state); |
4cf2143e | 1830 | |
4cf2143e | 1831 | /* |
42b766d7 JK |
1832 | * If the total_depth is more than depth, then we need to snip |
1833 | * the chain into two or more smaller chains that don't exceed | |
1834 | * the maximum depth. Most of the resulting chains will contain | |
1835 | * (depth + 1) entries (i.e., depth deltas plus one base), and | |
1836 | * the last chain (i.e., the one containing entry) will contain | |
1837 | * whatever entries are left over, namely | |
1838 | * (total_depth % (depth + 1)) of them. | |
1839 | * | |
1840 | * Since we are iterating towards decreasing depth, we need to | |
1841 | * decrement total_depth as we go, and we need to write to the | |
1842 | * entry what its final depth will be after all of the | |
1843 | * snipping. Since we're snipping into chains of length (depth | |
1844 | * + 1) entries, the final depth of an entry will be its | |
1845 | * original depth modulo (depth + 1). Any time we encounter an | |
1846 | * entry whose final depth is supposed to be zero, we snip it | |
1847 | * from its delta base, thereby making it so. | |
4cf2143e | 1848 | */ |
42b766d7 JK |
1849 | cur->depth = (total_depth--) % (depth + 1); |
1850 | if (!cur->depth) | |
1851 | drop_reused_delta(cur); | |
1852 | ||
1853 | cur->dfs_state = DFS_DONE; | |
4cf2143e JK |
1854 | } |
1855 | } | |
1856 | ||
c323ac7d LT |
1857 | static void get_object_details(void) |
1858 | { | |
7cadf491 | 1859 | uint32_t i; |
5c49c116 NP |
1860 | struct object_entry **sorted_by_offset; |
1861 | ||
5af05043 NTND |
1862 | if (progress) |
1863 | progress_state = start_progress(_("Counting objects"), | |
1864 | to_pack.nr_objects); | |
1865 | ||
2834bc27 VM |
1866 | sorted_by_offset = xcalloc(to_pack.nr_objects, sizeof(struct object_entry *)); |
1867 | for (i = 0; i < to_pack.nr_objects; i++) | |
1868 | sorted_by_offset[i] = to_pack.objects + i; | |
9ed0d8d6 | 1869 | QSORT(sorted_by_offset, to_pack.nr_objects, pack_offset_sort); |
c323ac7d | 1870 | |
2834bc27 | 1871 | for (i = 0; i < to_pack.nr_objects; i++) { |
15366280 JH |
1872 | struct object_entry *entry = sorted_by_offset[i]; |
1873 | check_object(entry); | |
ac77d0c3 NTND |
1874 | if (entry->type_valid && |
1875 | oe_size_greater_than(&to_pack, entry, big_file_threshold)) | |
15366280 | 1876 | entry->no_try_delta = 1; |
5af05043 | 1877 | display_progress(progress_state, i + 1); |
15366280 | 1878 | } |
5af05043 | 1879 | stop_progress(&progress_state); |
3449f8c4 | 1880 | |
4cf2143e JK |
1881 | /* |
1882 | * This must happen in a second pass, since we rely on the delta | |
1883 | * information for the whole list being completed. | |
1884 | */ | |
1885 | for (i = 0; i < to_pack.nr_objects; i++) | |
1886 | break_delta_chains(&to_pack.objects[i]); | |
1887 | ||
5c49c116 | 1888 | free(sorted_by_offset); |
c323ac7d LT |
1889 | } |
1890 | ||
b904166c NP |
1891 | /* |
1892 | * We search for deltas in a list sorted by type, by filename hash, and then | |
1893 | * by size, so that we see progressively smaller and smaller files. | |
1894 | * That's because we prefer deltas to be from the bigger file | |
1895 | * to the smaller -- deletes are potentially cheaper, but perhaps | |
1896 | * more importantly, the bigger file is likely the more recent | |
1897 | * one. The deepest deltas are therefore the oldest objects which are | |
1898 | * less susceptible to be accessed often. | |
1899 | */ | |
9668cf59 | 1900 | static int type_size_sort(const void *_a, const void *_b) |
c323ac7d | 1901 | { |
9668cf59 NP |
1902 | const struct object_entry *a = *(struct object_entry **)_a; |
1903 | const struct object_entry *b = *(struct object_entry **)_b; | |
fd9b1bae NTND |
1904 | enum object_type a_type = oe_type(a); |
1905 | enum object_type b_type = oe_type(b); | |
ac77d0c3 NTND |
1906 | unsigned long a_size = SIZE(a); |
1907 | unsigned long b_size = SIZE(b); | |
9668cf59 | 1908 | |
fd9b1bae | 1909 | if (a_type > b_type) |
27225f2e | 1910 | return -1; |
fd9b1bae | 1911 | if (a_type < b_type) |
27225f2e | 1912 | return 1; |
b904166c | 1913 | if (a->hash > b->hash) |
7a979d99 | 1914 | return -1; |
b904166c | 1915 | if (a->hash < b->hash) |
7a979d99 | 1916 | return 1; |
b904166c | 1917 | if (a->preferred_base > b->preferred_base) |
c323ac7d | 1918 | return -1; |
b904166c NP |
1919 | if (a->preferred_base < b->preferred_base) |
1920 | return 1; | |
28b8a730 JK |
1921 | if (use_delta_islands) { |
1922 | int island_cmp = island_delta_cmp(&a->idx.oid, &b->idx.oid); | |
1923 | if (island_cmp) | |
1924 | return island_cmp; | |
1925 | } | |
ac77d0c3 | 1926 | if (a_size > b_size) |
b904166c | 1927 | return -1; |
ac77d0c3 | 1928 | if (a_size < b_size) |
c323ac7d | 1929 | return 1; |
b904166c | 1930 | return a < b ? -1 : (a > b); /* newest first */ |
c323ac7d LT |
1931 | } |
1932 | ||
1933 | struct unpacked { | |
1934 | struct object_entry *entry; | |
1935 | void *data; | |
f6c7081a | 1936 | struct delta_index *index; |
5a235b5e | 1937 | unsigned depth; |
c323ac7d LT |
1938 | }; |
1939 | ||
d250626c NP |
1940 | static int delta_cacheable(unsigned long src_size, unsigned long trg_size, |
1941 | unsigned long delta_size) | |
074b2eea MK |
1942 | { |
1943 | if (max_delta_cache_size && delta_cache_size + delta_size > max_delta_cache_size) | |
1944 | return 0; | |
1945 | ||
e3dfddb3 MK |
1946 | if (delta_size < cache_max_small_delta_size) |
1947 | return 1; | |
1948 | ||
074b2eea MK |
1949 | /* cache delta, if objects are large enough compared to delta size */ |
1950 | if ((src_size >> 20) + (trg_size >> 21) > (delta_size >> 10)) | |
1951 | return 1; | |
1952 | ||
1953 | return 0; | |
1954 | } | |
1955 | ||
ffbd51cc | 1956 | /* Protect access to object database */ |
44626dc7 | 1957 | static pthread_mutex_t read_mutex; |
8ecce684 NP |
1958 | #define read_lock() pthread_mutex_lock(&read_mutex) |
1959 | #define read_unlock() pthread_mutex_unlock(&read_mutex) | |
1960 | ||
ffbd51cc | 1961 | /* Protect delta_cache_size */ |
44626dc7 | 1962 | static pthread_mutex_t cache_mutex; |
3c701839 NP |
1963 | #define cache_lock() pthread_mutex_lock(&cache_mutex) |
1964 | #define cache_unlock() pthread_mutex_unlock(&cache_mutex) | |
1965 | ||
ffbd51cc NTND |
1966 | /* |
1967 | * Protect object list partitioning (e.g. struct thread_param) and | |
1968 | * progress_state | |
1969 | */ | |
44626dc7 | 1970 | static pthread_mutex_t progress_mutex; |
8ecce684 NP |
1971 | #define progress_lock() pthread_mutex_lock(&progress_mutex) |
1972 | #define progress_unlock() pthread_mutex_unlock(&progress_mutex) | |
1973 | ||
ffbd51cc NTND |
1974 | /* |
1975 | * Access to struct object_entry is unprotected since each thread owns | |
1976 | * a portion of the main object list. Just don't access object entries | |
1977 | * ahead in the list because they can be stolen and would need | |
1978 | * progress_mutex for protection. | |
1979 | */ | |
8ecce684 | 1980 | |
ac77d0c3 NTND |
1981 | /* |
1982 | * Return the size of the object without doing any delta | |
1983 | * reconstruction (so non-deltas are true object sizes, but deltas | |
1984 | * return the size of the delta data). | |
1985 | */ | |
1986 | unsigned long oe_get_size_slow(struct packing_data *pack, | |
1987 | const struct object_entry *e) | |
1988 | { | |
1989 | struct packed_git *p; | |
1990 | struct pack_window *w_curs; | |
1991 | unsigned char *buf; | |
1992 | enum object_type type; | |
1993 | unsigned long used, avail, size; | |
1994 | ||
1995 | if (e->type_ != OBJ_OFS_DELTA && e->type_ != OBJ_REF_DELTA) { | |
1996 | read_lock(); | |
ad635e82 | 1997 | if (oid_object_info(the_repository, &e->idx.oid, &size) < 0) |
ac77d0c3 NTND |
1998 | die(_("unable to get size of %s"), |
1999 | oid_to_hex(&e->idx.oid)); | |
2000 | read_unlock(); | |
2001 | return size; | |
2002 | } | |
2003 | ||
2004 | p = oe_in_pack(pack, e); | |
2005 | if (!p) | |
2006 | BUG("when e->type is a delta, it must belong to a pack"); | |
2007 | ||
2008 | read_lock(); | |
2009 | w_curs = NULL; | |
2010 | buf = use_pack(p, &w_curs, e->in_pack_offset, &avail); | |
2011 | used = unpack_object_header_buffer(buf, avail, &type, &size); | |
2012 | if (used == 0) | |
2013 | die(_("unable to parse object header of %s"), | |
2014 | oid_to_hex(&e->idx.oid)); | |
2015 | ||
2016 | unuse_pack(&w_curs); | |
2017 | read_unlock(); | |
2018 | return size; | |
2019 | } | |
2020 | ||
f6c7081a | 2021 | static int try_delta(struct unpacked *trg, struct unpacked *src, |
ef0316fc | 2022 | unsigned max_depth, unsigned long *mem_usage) |
c323ac7d | 2023 | { |
f6c7081a NP |
2024 | struct object_entry *trg_entry = trg->entry; |
2025 | struct object_entry *src_entry = src->entry; | |
560b25a8 | 2026 | unsigned long trg_size, src_size, delta_size, sizediff, max_size, sz; |
c83f032e | 2027 | unsigned ref_depth; |
21666f1a | 2028 | enum object_type type; |
c323ac7d LT |
2029 | void *delta_buf; |
2030 | ||
2031 | /* Don't bother doing diffs between different types */ | |
fd9b1bae | 2032 | if (oe_type(trg_entry) != oe_type(src_entry)) |
c323ac7d LT |
2033 | return -1; |
2034 | ||
51d1e83f | 2035 | /* |
15f07e06 JK |
2036 | * We do not bother to try a delta that we discarded on an |
2037 | * earlier try, but only when reusing delta data. Note that | |
2038 | * src_entry that is marked as the preferred_base should always | |
2039 | * be considered, as even if we produce a suboptimal delta against | |
2040 | * it, we will still save the transfer cost, as we already know | |
2041 | * the other side has it and we won't send src_entry at all. | |
51d1e83f | 2042 | */ |
43fa44fa NTND |
2043 | if (reuse_delta && IN_PACK(trg_entry) && |
2044 | IN_PACK(trg_entry) == IN_PACK(src_entry) && | |
15f07e06 | 2045 | !src_entry->preferred_base && |
e9195b58 JH |
2046 | trg_entry->in_pack_type != OBJ_REF_DELTA && |
2047 | trg_entry->in_pack_type != OBJ_OFS_DELTA) | |
51d1e83f LT |
2048 | return 0; |
2049 | ||
898b14ce | 2050 | /* Let's not bust the allowed depth. */ |
5a235b5e | 2051 | if (src->depth >= max_depth) |
d116a45a | 2052 | return 0; |
c323ac7d | 2053 | |
c3b06a69 | 2054 | /* Now some size filtering heuristics. */ |
ac77d0c3 | 2055 | trg_size = SIZE(trg_entry); |
898eba5e | 2056 | if (!DELTA(trg_entry)) { |
41179100 | 2057 | max_size = trg_size/2 - the_hash_algo->rawsz; |
c83f032e NP |
2058 | ref_depth = 1; |
2059 | } else { | |
0aca34e8 | 2060 | max_size = DELTA_SIZE(trg_entry); |
5a235b5e | 2061 | ref_depth = trg->depth; |
c83f032e | 2062 | } |
720fe22d | 2063 | max_size = (uint64_t)max_size * (max_depth - src->depth) / |
c83f032e | 2064 | (max_depth - ref_depth + 1); |
c3b06a69 NP |
2065 | if (max_size == 0) |
2066 | return 0; | |
ac77d0c3 | 2067 | src_size = SIZE(src_entry); |
560b25a8 | 2068 | sizediff = src_size < trg_size ? trg_size - src_size : 0; |
27225f2e | 2069 | if (sizediff >= max_size) |
f527cb8c | 2070 | return 0; |
a1dab41a BD |
2071 | if (trg_size < src_size / 32) |
2072 | return 0; | |
f6c7081a | 2073 | |
28b8a730 JK |
2074 | if (!in_same_island(&trg->entry->idx.oid, &src->entry->idx.oid)) |
2075 | return 0; | |
2076 | ||
560b25a8 NP |
2077 | /* Load data if not already done */ |
2078 | if (!trg->data) { | |
8ecce684 | 2079 | read_lock(); |
b4f5aca4 | 2080 | trg->data = read_object_file(&trg_entry->idx.oid, &type, &sz); |
8ecce684 | 2081 | read_unlock(); |
2e3404c3 | 2082 | if (!trg->data) |
f616db6a | 2083 | die(_("object %s cannot be read"), |
e6a492b7 | 2084 | oid_to_hex(&trg_entry->idx.oid)); |
560b25a8 | 2085 | if (sz != trg_size) |
ca473cef TB |
2086 | die(_("object %s inconsistent object length (%"PRIuMAX" vs %"PRIuMAX")"), |
2087 | oid_to_hex(&trg_entry->idx.oid), (uintmax_t)sz, | |
2088 | (uintmax_t)trg_size); | |
ef0316fc | 2089 | *mem_usage += sz; |
560b25a8 NP |
2090 | } |
2091 | if (!src->data) { | |
8ecce684 | 2092 | read_lock(); |
b4f5aca4 | 2093 | src->data = read_object_file(&src_entry->idx.oid, &type, &sz); |
8ecce684 | 2094 | read_unlock(); |
71064a95 NP |
2095 | if (!src->data) { |
2096 | if (src_entry->preferred_base) { | |
2097 | static int warned = 0; | |
2098 | if (!warned++) | |
f616db6a | 2099 | warning(_("object %s cannot be read"), |
e6a492b7 | 2100 | oid_to_hex(&src_entry->idx.oid)); |
71064a95 NP |
2101 | /* |
2102 | * Those objects are not included in the | |
2103 | * resulting pack. Be resilient and ignore | |
2104 | * them if they can't be read, in case the | |
2105 | * pack could be created nevertheless. | |
2106 | */ | |
2107 | return 0; | |
2108 | } | |
f616db6a | 2109 | die(_("object %s cannot be read"), |
e6a492b7 | 2110 | oid_to_hex(&src_entry->idx.oid)); |
71064a95 | 2111 | } |
560b25a8 | 2112 | if (sz != src_size) |
ca473cef TB |
2113 | die(_("object %s inconsistent object length (%"PRIuMAX" vs %"PRIuMAX")"), |
2114 | oid_to_hex(&src_entry->idx.oid), (uintmax_t)sz, | |
2115 | (uintmax_t)src_size); | |
ef0316fc | 2116 | *mem_usage += sz; |
560b25a8 NP |
2117 | } |
2118 | if (!src->index) { | |
2119 | src->index = create_delta_index(src->data, src_size); | |
a588d88a MK |
2120 | if (!src->index) { |
2121 | static int warned = 0; | |
2122 | if (!warned++) | |
f616db6a | 2123 | warning(_("suboptimal pack - out of memory")); |
a588d88a MK |
2124 | return 0; |
2125 | } | |
ef0316fc | 2126 | *mem_usage += sizeof_delta_index(src->index); |
560b25a8 NP |
2127 | } |
2128 | ||
2129 | delta_buf = create_delta(src->index, trg->data, trg_size, &delta_size, max_size); | |
c323ac7d | 2130 | if (!delta_buf) |
75c42d8c | 2131 | return 0; |
f6c7081a | 2132 | |
898eba5e | 2133 | if (DELTA(trg_entry)) { |
848d732c | 2134 | /* Prefer only shallower same-sized deltas. */ |
0aca34e8 | 2135 | if (delta_size == DELTA_SIZE(trg_entry) && |
5a235b5e | 2136 | src->depth + 1 >= trg->depth) { |
848d732c BD |
2137 | free(delta_buf); |
2138 | return 0; | |
2139 | } | |
074b2eea | 2140 | } |
9e2d57a0 | 2141 | |
3c701839 NP |
2142 | /* |
2143 | * Handle memory allocation outside of the cache | |
2144 | * accounting lock. Compiler will optimize the strangeness | |
7eb151d6 | 2145 | * away when NO_PTHREADS is defined. |
3c701839 | 2146 | */ |
8e0f7003 | 2147 | free(trg_entry->delta_data); |
3c701839 | 2148 | cache_lock(); |
9e2d57a0 | 2149 | if (trg_entry->delta_data) { |
0aca34e8 | 2150 | delta_cache_size -= DELTA_SIZE(trg_entry); |
9e2d57a0 NP |
2151 | trg_entry->delta_data = NULL; |
2152 | } | |
d250626c | 2153 | if (delta_cacheable(src_size, trg_size, delta_size)) { |
b7a28f78 | 2154 | delta_cache_size += delta_size; |
3c701839 NP |
2155 | cache_unlock(); |
2156 | trg_entry->delta_data = xrealloc(delta_buf, delta_size); | |
2157 | } else { | |
2158 | cache_unlock(); | |
074b2eea | 2159 | free(delta_buf); |
3c701839 NP |
2160 | } |
2161 | ||
898eba5e | 2162 | SET_DELTA(trg_entry, src_entry); |
0aca34e8 | 2163 | SET_DELTA_SIZE(trg_entry, delta_size); |
b7a28f78 NP |
2164 | trg->depth = src->depth + 1; |
2165 | ||
f6c7081a | 2166 | return 1; |
c323ac7d LT |
2167 | } |
2168 | ||
898b14ce | 2169 | static unsigned int check_delta_limit(struct object_entry *me, unsigned int n) |
b2504a0d | 2170 | { |
898eba5e | 2171 | struct object_entry *child = DELTA_CHILD(me); |
898b14ce NP |
2172 | unsigned int m = n; |
2173 | while (child) { | |
2174 | unsigned int c = check_delta_limit(child, n + 1); | |
2175 | if (m < c) | |
2176 | m = c; | |
898eba5e | 2177 | child = DELTA_SIBLING(child); |
898b14ce NP |
2178 | } |
2179 | return m; | |
b2504a0d NP |
2180 | } |
2181 | ||
75ad235c | 2182 | static unsigned long free_unpacked(struct unpacked *n) |
a97773ce | 2183 | { |
ef0316fc | 2184 | unsigned long freed_mem = sizeof_delta_index(n->index); |
a97773ce BD |
2185 | free_delta_index(n->index); |
2186 | n->index = NULL; | |
2187 | if (n->data) { | |
ac77d0c3 | 2188 | freed_mem += SIZE(n->entry); |
6a83d902 | 2189 | FREE_AND_NULL(n->data); |
a97773ce BD |
2190 | } |
2191 | n->entry = NULL; | |
7d7baa5e | 2192 | n->depth = 0; |
ef0316fc | 2193 | return freed_mem; |
a97773ce BD |
2194 | } |
2195 | ||
384b32c0 | 2196 | static void find_deltas(struct object_entry **list, unsigned *list_size, |
e334977d | 2197 | int window, int depth, unsigned *processed) |
c323ac7d | 2198 | { |
384b32c0 | 2199 | uint32_t i, idx = 0, count = 0; |
7cadf491 | 2200 | struct unpacked *array; |
ef0316fc | 2201 | unsigned long mem_usage = 0; |
c323ac7d | 2202 | |
19d4b416 | 2203 | array = xcalloc(window, sizeof(struct unpacked)); |
21fcd1bd | 2204 | |
384b32c0 | 2205 | for (;;) { |
421b488a | 2206 | struct object_entry *entry; |
c323ac7d | 2207 | struct unpacked *n = array + idx; |
ef0316fc | 2208 | int j, max_depth, best_base = -1; |
c323ac7d | 2209 | |
384b32c0 NP |
2210 | progress_lock(); |
2211 | if (!*list_size) { | |
2212 | progress_unlock(); | |
2213 | break; | |
2214 | } | |
421b488a | 2215 | entry = *list++; |
384b32c0 NP |
2216 | (*list_size)--; |
2217 | if (!entry->preferred_base) { | |
2218 | (*processed)++; | |
2219 | display_progress(progress_state, *processed); | |
2220 | } | |
2221 | progress_unlock(); | |
2222 | ||
ef0316fc | 2223 | mem_usage -= free_unpacked(n); |
c323ac7d | 2224 | n->entry = entry; |
ab7cd7bb | 2225 | |
a97773ce | 2226 | while (window_memory_limit && |
ef0316fc | 2227 | mem_usage > window_memory_limit && |
a97773ce BD |
2228 | count > 1) { |
2229 | uint32_t tail = (idx + window - count) % window; | |
75ad235c | 2230 | mem_usage -= free_unpacked(array + tail); |
a97773ce BD |
2231 | count--; |
2232 | } | |
2233 | ||
75d39853 NP |
2234 | /* We do not compute delta to *create* objects we are not |
2235 | * going to pack. | |
2236 | */ | |
2237 | if (entry->preferred_base) | |
2238 | goto next; | |
2239 | ||
898b14ce NP |
2240 | /* |
2241 | * If the current object is at pack edge, take the depth the | |
2242 | * objects that depend on the current object into account | |
2243 | * otherwise they would become too deep. | |
2244 | */ | |
2245 | max_depth = depth; | |
898eba5e | 2246 | if (DELTA_CHILD(entry)) { |
898b14ce NP |
2247 | max_depth -= check_delta_limit(entry, 0); |
2248 | if (max_depth <= 0) | |
2249 | goto next; | |
2250 | } | |
2251 | ||
78817c15 LT |
2252 | j = window; |
2253 | while (--j > 0) { | |
77639870 | 2254 | int ret; |
7cadf491 | 2255 | uint32_t other_idx = idx + j; |
c323ac7d | 2256 | struct unpacked *m; |
78817c15 LT |
2257 | if (other_idx >= window) |
2258 | other_idx -= window; | |
c323ac7d LT |
2259 | m = array + other_idx; |
2260 | if (!m->entry) | |
2261 | break; | |
ef0316fc | 2262 | ret = try_delta(n, m, max_depth, &mem_usage); |
77639870 | 2263 | if (ret < 0) |
c323ac7d | 2264 | break; |
77639870 JH |
2265 | else if (ret > 0) |
2266 | best_base = other_idx; | |
c323ac7d | 2267 | } |
898b14ce | 2268 | |
ed4a9031 NP |
2269 | /* |
2270 | * If we decided to cache the delta data, then it is best | |
2271 | * to compress it right away. First because we have to do | |
2272 | * it anyway, and doing it here while we're threaded will | |
2273 | * save a lot of time in the non threaded write phase, | |
2274 | * as well as allow for caching more deltas within | |
2275 | * the same cache size limit. | |
2276 | * ... | |
2277 | * But only if not writing to stdout, since in that case | |
2278 | * the network is most likely throttling writes anyway, | |
2279 | * and therefore it is best to go to the write phase ASAP | |
2280 | * instead, as we can afford spending more time compressing | |
2281 | * between writes at that moment. | |
2282 | */ | |
2283 | if (entry->delta_data && !pack_to_stdout) { | |
0cb3c142 NTND |
2284 | unsigned long size; |
2285 | ||
0aca34e8 | 2286 | size = do_compress(&entry->delta_data, DELTA_SIZE(entry)); |
0cb3c142 NTND |
2287 | if (size < (1U << OE_Z_DELTA_BITS)) { |
2288 | entry->z_delta_size = size; | |
2289 | cache_lock(); | |
0aca34e8 | 2290 | delta_cache_size -= DELTA_SIZE(entry); |
0cb3c142 NTND |
2291 | delta_cache_size += entry->z_delta_size; |
2292 | cache_unlock(); | |
2293 | } else { | |
2294 | FREE_AND_NULL(entry->delta_data); | |
2295 | entry->z_delta_size = 0; | |
2296 | } | |
ed4a9031 NP |
2297 | } |
2298 | ||
70ca1a3f JH |
2299 | /* if we made n a delta, and if n is already at max |
2300 | * depth, leaving it in the window is pointless. we | |
2301 | * should evict it first. | |
70ca1a3f | 2302 | */ |
898eba5e | 2303 | if (DELTA(entry) && max_depth <= n->depth) |
70ca1a3f | 2304 | continue; |
ff45715c | 2305 | |
77639870 JH |
2306 | /* |
2307 | * Move the best delta base up in the window, after the | |
2308 | * currently deltified object, to keep it longer. It will | |
2309 | * be the first base object to be attempted next. | |
2310 | */ | |
898eba5e | 2311 | if (DELTA(entry)) { |
77639870 JH |
2312 | struct unpacked swap = array[best_base]; |
2313 | int dist = (window + idx - best_base) % window; | |
2314 | int dst = best_base; | |
2315 | while (dist--) { | |
2316 | int src = (dst + 1) % window; | |
2317 | array[dst] = array[src]; | |
2318 | dst = src; | |
2319 | } | |
2320 | array[dst] = swap; | |
2321 | } | |
2322 | ||
898b14ce | 2323 | next: |
521a4f4c | 2324 | idx++; |
a97773ce BD |
2325 | if (count + 1 < window) |
2326 | count++; | |
521a4f4c LT |
2327 | if (idx >= window) |
2328 | idx = 0; | |
384b32c0 | 2329 | } |
adee7bdf | 2330 | |
f6c7081a | 2331 | for (i = 0; i < window; ++i) { |
ff45715c | 2332 | free_delta_index(array[i].index); |
adee7bdf | 2333 | free(array[i].data); |
f6c7081a | 2334 | } |
adee7bdf | 2335 | free(array); |
c323ac7d LT |
2336 | } |
2337 | ||
a9a74636 NP |
2338 | static void try_to_free_from_threads(size_t size) |
2339 | { | |
2340 | read_lock(); | |
7c3ecb32 | 2341 | release_pack_memory(size); |
a9a74636 NP |
2342 | read_unlock(); |
2343 | } | |
2344 | ||
bc9b2175 | 2345 | static try_to_free_t old_try_to_free_routine; |
851c34b0 | 2346 | |
50f22ada | 2347 | /* |
ffbd51cc NTND |
2348 | * The main object list is split into smaller lists, each is handed to |
2349 | * one worker. | |
2350 | * | |
50f22ada JS |
2351 | * The main thread waits on the condition that (at least) one of the workers |
2352 | * has stopped working (which is indicated in the .working member of | |
2353 | * struct thread_params). | |
ffbd51cc | 2354 | * |
50f22ada JS |
2355 | * When a work thread has completed its work, it sets .working to 0 and |
2356 | * signals the main thread and waits on the condition that .data_ready | |
2357 | * becomes 1. | |
ffbd51cc NTND |
2358 | * |
2359 | * The main thread steals half of the work from the worker that has | |
2360 | * most work left to hand it to the idle worker. | |
50f22ada JS |
2361 | */ |
2362 | ||
8ecce684 NP |
2363 | struct thread_params { |
2364 | pthread_t thread; | |
2365 | struct object_entry **list; | |
2366 | unsigned list_size; | |
384b32c0 | 2367 | unsigned remaining; |
8ecce684 NP |
2368 | int window; |
2369 | int depth; | |
50f22ada JS |
2370 | int working; |
2371 | int data_ready; | |
2372 | pthread_mutex_t mutex; | |
2373 | pthread_cond_t cond; | |
8ecce684 NP |
2374 | unsigned *processed; |
2375 | }; | |
2376 | ||
44626dc7 AH |
2377 | static pthread_cond_t progress_cond; |
2378 | ||
2379 | /* | |
2380 | * Mutex and conditional variable can't be statically-initialized on Windows. | |
2381 | */ | |
2382 | static void init_threaded_search(void) | |
2383 | { | |
93749194 | 2384 | init_recursive_mutex(&read_mutex); |
44626dc7 AH |
2385 | pthread_mutex_init(&cache_mutex, NULL); |
2386 | pthread_mutex_init(&progress_mutex, NULL); | |
2387 | pthread_cond_init(&progress_cond, NULL); | |
851c34b0 | 2388 | old_try_to_free_routine = set_try_to_free_routine(try_to_free_from_threads); |
44626dc7 AH |
2389 | } |
2390 | ||
2391 | static void cleanup_threaded_search(void) | |
2392 | { | |
851c34b0 | 2393 | set_try_to_free_routine(old_try_to_free_routine); |
44626dc7 AH |
2394 | pthread_cond_destroy(&progress_cond); |
2395 | pthread_mutex_destroy(&read_mutex); | |
2396 | pthread_mutex_destroy(&cache_mutex); | |
2397 | pthread_mutex_destroy(&progress_mutex); | |
2398 | } | |
c2a33679 | 2399 | |
8ecce684 NP |
2400 | static void *threaded_find_deltas(void *arg) |
2401 | { | |
c2a33679 NP |
2402 | struct thread_params *me = arg; |
2403 | ||
0c2ad00b | 2404 | progress_lock(); |
50f22ada | 2405 | while (me->remaining) { |
0c2ad00b MÅ |
2406 | progress_unlock(); |
2407 | ||
384b32c0 | 2408 | find_deltas(me->list, &me->remaining, |
c2a33679 | 2409 | me->window, me->depth, me->processed); |
50f22ada JS |
2410 | |
2411 | progress_lock(); | |
2412 | me->working = 0; | |
2413 | pthread_cond_signal(&progress_cond); | |
2414 | progress_unlock(); | |
2415 | ||
2416 | /* | |
2417 | * We must not set ->data_ready before we wait on the | |
2418 | * condition because the main thread may have set it to 1 | |
2419 | * before we get here. In order to be sure that new | |
2420 | * work is available if we see 1 in ->data_ready, it | |
2421 | * was initialized to 0 before this thread was spawned | |
2422 | * and we reset it to 0 right away. | |
2423 | */ | |
2424 | pthread_mutex_lock(&me->mutex); | |
2425 | while (!me->data_ready) | |
2426 | pthread_cond_wait(&me->cond, &me->mutex); | |
2427 | me->data_ready = 0; | |
2428 | pthread_mutex_unlock(&me->mutex); | |
0c2ad00b MÅ |
2429 | |
2430 | progress_lock(); | |
c2a33679 | 2431 | } |
0c2ad00b | 2432 | progress_unlock(); |
50f22ada JS |
2433 | /* leave ->working 1 so that this doesn't get more work assigned */ |
2434 | return NULL; | |
8ecce684 NP |
2435 | } |
2436 | ||
8ecce684 NP |
2437 | static void ll_find_deltas(struct object_entry **list, unsigned list_size, |
2438 | int window, int depth, unsigned *processed) | |
2439 | { | |
dcda3614 | 2440 | struct thread_params *p; |
384b32c0 | 2441 | int i, ret, active_threads = 0; |
c2a33679 | 2442 | |
44626dc7 AH |
2443 | init_threaded_search(); |
2444 | ||
367f4a43 | 2445 | if (delta_search_threads <= 1) { |
384b32c0 | 2446 | find_deltas(list, &list_size, window, depth, processed); |
44626dc7 | 2447 | cleanup_threaded_search(); |
367f4a43 NP |
2448 | return; |
2449 | } | |
43cc2b42 | 2450 | if (progress > pack_to_stdout) |
f616db6a | 2451 | fprintf_ln(stderr, _("Delta compression using up to %d threads"), |
1a07e59c | 2452 | delta_search_threads); |
dcda3614 | 2453 | p = xcalloc(delta_search_threads, sizeof(*p)); |
367f4a43 | 2454 | |
50f22ada | 2455 | /* Partition the work amongst work threads. */ |
367f4a43 | 2456 | for (i = 0; i < delta_search_threads; i++) { |
50f22ada JS |
2457 | unsigned sub_size = list_size / (delta_search_threads - i); |
2458 | ||
bf874896 NP |
2459 | /* don't use too small segments or no deltas will be found */ |
2460 | if (sub_size < 2*window && i+1 < delta_search_threads) | |
2461 | sub_size = 0; | |
2462 | ||
8ecce684 NP |
2463 | p[i].window = window; |
2464 | p[i].depth = depth; | |
2465 | p[i].processed = processed; | |
50f22ada JS |
2466 | p[i].working = 1; |
2467 | p[i].data_ready = 0; | |
c2a33679 | 2468 | |
59921b4b | 2469 | /* try to split chunks on "path" boundaries */ |
6fc74703 NP |
2470 | while (sub_size && sub_size < list_size && |
2471 | list[sub_size]->hash && | |
384b32c0 NP |
2472 | list[sub_size]->hash == list[sub_size-1]->hash) |
2473 | sub_size++; | |
2474 | ||
50f22ada JS |
2475 | p[i].list = list; |
2476 | p[i].list_size = sub_size; | |
2477 | p[i].remaining = sub_size; | |
59921b4b | 2478 | |
384b32c0 NP |
2479 | list += sub_size; |
2480 | list_size -= sub_size; | |
2481 | } | |
2482 | ||
50f22ada JS |
2483 | /* Start work threads. */ |
2484 | for (i = 0; i < delta_search_threads; i++) { | |
2485 | if (!p[i].list_size) | |
2486 | continue; | |
68e6a4f8 JS |
2487 | pthread_mutex_init(&p[i].mutex, NULL); |
2488 | pthread_cond_init(&p[i].cond, NULL); | |
50f22ada JS |
2489 | ret = pthread_create(&p[i].thread, NULL, |
2490 | threaded_find_deltas, &p[i]); | |
2491 | if (ret) | |
f616db6a | 2492 | die(_("unable to create thread: %s"), strerror(ret)); |
50f22ada JS |
2493 | active_threads++; |
2494 | } | |
2495 | ||
384b32c0 NP |
2496 | /* |
2497 | * Now let's wait for work completion. Each time a thread is done | |
2498 | * with its work, we steal half of the remaining work from the | |
2499 | * thread with the largest number of unprocessed objects and give | |
2500 | * it to that newly idle thread. This ensure good load balancing | |
2501 | * until the remaining object list segments are simply too short | |
2502 | * to be worth splitting anymore. | |
2503 | */ | |
50f22ada JS |
2504 | while (active_threads) { |
2505 | struct thread_params *target = NULL; | |
384b32c0 NP |
2506 | struct thread_params *victim = NULL; |
2507 | unsigned sub_size = 0; | |
384b32c0 NP |
2508 | |
2509 | progress_lock(); | |
50f22ada JS |
2510 | for (;;) { |
2511 | for (i = 0; !target && i < delta_search_threads; i++) | |
2512 | if (!p[i].working) | |
2513 | target = &p[i]; | |
2514 | if (target) | |
2515 | break; | |
2516 | pthread_cond_wait(&progress_cond, &progress_mutex); | |
2517 | } | |
2518 | ||
384b32c0 NP |
2519 | for (i = 0; i < delta_search_threads; i++) |
2520 | if (p[i].remaining > 2*window && | |
2521 | (!victim || victim->remaining < p[i].remaining)) | |
2522 | victim = &p[i]; | |
2523 | if (victim) { | |
2524 | sub_size = victim->remaining / 2; | |
2525 | list = victim->list + victim->list_size - sub_size; | |
2526 | while (sub_size && list[0]->hash && | |
2527 | list[0]->hash == list[-1]->hash) { | |
2528 | list++; | |
2529 | sub_size--; | |
2530 | } | |
eb9688ff NP |
2531 | if (!sub_size) { |
2532 | /* | |
2533 | * It is possible for some "paths" to have | |
2534 | * so many objects that no hash boundary | |
2535 | * might be found. Let's just steal the | |
2536 | * exact half in that case. | |
2537 | */ | |
2538 | sub_size = victim->remaining / 2; | |
2539 | list -= sub_size; | |
2540 | } | |
384b32c0 NP |
2541 | target->list = list; |
2542 | victim->list_size -= sub_size; | |
2543 | victim->remaining -= sub_size; | |
2544 | } | |
384b32c0 NP |
2545 | target->list_size = sub_size; |
2546 | target->remaining = sub_size; | |
50f22ada JS |
2547 | target->working = 1; |
2548 | progress_unlock(); | |
2549 | ||
2550 | pthread_mutex_lock(&target->mutex); | |
2551 | target->data_ready = 1; | |
2552 | pthread_cond_signal(&target->cond); | |
2553 | pthread_mutex_unlock(&target->mutex); | |
c2a33679 | 2554 | |
384b32c0 | 2555 | if (!sub_size) { |
b81d9af7 | 2556 | pthread_join(target->thread, NULL); |
50f22ada JS |
2557 | pthread_cond_destroy(&target->cond); |
2558 | pthread_mutex_destroy(&target->mutex); | |
384b32c0 | 2559 | active_threads--; |
c2a33679 | 2560 | } |
50f22ada | 2561 | } |
44626dc7 | 2562 | cleanup_threaded_search(); |
dcda3614 | 2563 | free(p); |
8ecce684 NP |
2564 | } |
2565 | ||
b773ddea JK |
2566 | static void add_tag_chain(const struct object_id *oid) |
2567 | { | |
2568 | struct tag *tag; | |
2569 | ||
2570 | /* | |
2571 | * We catch duplicates already in add_object_entry(), but we'd | |
2572 | * prefer to do this extra check to avoid having to parse the | |
2573 | * tag at all if we already know that it's being packed (e.g., if | |
2574 | * it was included via bitmaps, we would not have parsed it | |
2575 | * previously). | |
2576 | */ | |
2577 | if (packlist_find(&to_pack, oid->hash, NULL)) | |
2578 | return; | |
2579 | ||
ce71efb7 | 2580 | tag = lookup_tag(the_repository, oid); |
b773ddea JK |
2581 | while (1) { |
2582 | if (!tag || parse_tag(tag) || !tag->tagged) | |
f616db6a | 2583 | die(_("unable to pack objects reachable from tag %s"), |
b773ddea JK |
2584 | oid_to_hex(oid)); |
2585 | ||
188960b4 | 2586 | add_object_entry(&tag->object.oid, OBJ_TAG, NULL, 0); |
b773ddea JK |
2587 | |
2588 | if (tag->tagged->type != OBJ_TAG) | |
2589 | return; | |
2590 | ||
2591 | tag = (struct tag *)tag->tagged; | |
2592 | } | |
2593 | } | |
2594 | ||
d155254c | 2595 | static int add_ref_tag(const char *path, const struct object_id *oid, int flag, void *cb_data) |
f0a24aa5 | 2596 | { |
d155254c | 2597 | struct object_id peeled; |
f0a24aa5 | 2598 | |
59556548 | 2599 | if (starts_with(path, "refs/tags/") && /* is a tag? */ |
b420d909 | 2600 | !peel_ref(path, &peeled) && /* peelable? */ |
d155254c | 2601 | packlist_find(&to_pack, peeled.hash, NULL)) /* object packed? */ |
b773ddea | 2602 | add_tag_chain(oid); |
f0a24aa5 SP |
2603 | return 0; |
2604 | } | |
2605 | ||
f3123c4a JH |
2606 | static void prepare_pack(int window, int depth) |
2607 | { | |
9668cf59 | 2608 | struct object_entry **delta_list; |
6e1c2344 RJ |
2609 | uint32_t i, nr_deltas; |
2610 | unsigned n; | |
9668cf59 | 2611 | |
28b8a730 JK |
2612 | if (use_delta_islands) |
2613 | resolve_tree_islands(progress, &to_pack); | |
2614 | ||
3f9ac8d2 | 2615 | get_object_details(); |
9668cf59 | 2616 | |
0e8189e2 NP |
2617 | /* |
2618 | * If we're locally repacking then we need to be doubly careful | |
2619 | * from now on in order to make sure no stealth corruption gets | |
2620 | * propagated to the new pack. Clients receiving streamed packs | |
2621 | * should validate everything they get anyway so no need to incur | |
2622 | * the additional cost here in that case. | |
2623 | */ | |
2624 | if (!pack_to_stdout) | |
2625 | do_check_packed_object_crc = 1; | |
2626 | ||
2834bc27 | 2627 | if (!to_pack.nr_objects || !window || !depth) |
9668cf59 NP |
2628 | return; |
2629 | ||
b32fa95f | 2630 | ALLOC_ARRAY(delta_list, to_pack.nr_objects); |
75d39853 NP |
2631 | nr_deltas = n = 0; |
2632 | ||
2834bc27 VM |
2633 | for (i = 0; i < to_pack.nr_objects; i++) { |
2634 | struct object_entry *entry = to_pack.objects + i; | |
75d39853 | 2635 | |
898eba5e | 2636 | if (DELTA(entry)) |
75d39853 | 2637 | /* This happens if we decided to reuse existing |
a7de7130 | 2638 | * delta from a pack. "reuse_delta &&" is implied. |
75d39853 NP |
2639 | */ |
2640 | continue; | |
2641 | ||
ac77d0c3 NTND |
2642 | if (!entry->type_valid || |
2643 | oe_size_less_than(&to_pack, entry, 50)) | |
75d39853 NP |
2644 | continue; |
2645 | ||
2646 | if (entry->no_try_delta) | |
2647 | continue; | |
2648 | ||
6d6f9cdd | 2649 | if (!entry->preferred_base) { |
75d39853 | 2650 | nr_deltas++; |
fd9b1bae | 2651 | if (oe_type(entry) < 0) |
f616db6a | 2652 | die(_("unable to get type of object %s"), |
e6a492b7 | 2653 | oid_to_hex(&entry->idx.oid)); |
eede9f42 | 2654 | } else { |
fd9b1bae | 2655 | if (oe_type(entry) < 0) { |
eede9f42 NP |
2656 | /* |
2657 | * This object is not found, but we | |
2658 | * don't have to include it anyway. | |
2659 | */ | |
2660 | continue; | |
2661 | } | |
6d6f9cdd | 2662 | } |
75d39853 NP |
2663 | |
2664 | delta_list[n++] = entry; | |
2665 | } | |
2666 | ||
2f8b8947 | 2667 | if (nr_deltas && n > 1) { |
e334977d NP |
2668 | unsigned nr_done = 0; |
2669 | if (progress) | |
754dbc43 | 2670 | progress_state = start_progress(_("Compressing objects"), |
dc6a0757 | 2671 | nr_deltas); |
9ed0d8d6 | 2672 | QSORT(delta_list, n, type_size_sort); |
8ecce684 | 2673 | ll_find_deltas(delta_list, n, window+1, depth, &nr_done); |
4d4fcc54 | 2674 | stop_progress(&progress_state); |
e334977d | 2675 | if (nr_done != nr_deltas) |
f616db6a | 2676 | die(_("inconsistency with delta count")); |
75d39853 | 2677 | } |
9668cf59 | 2678 | free(delta_list); |
f3123c4a JH |
2679 | } |
2680 | ||
ef90d6d4 | 2681 | static int git_pack_config(const char *k, const char *v, void *cb) |
4812a93a | 2682 | { |
eeefa7c9 | 2683 | if (!strcmp(k, "pack.window")) { |
4812a93a JK |
2684 | window = git_config_int(k, v); |
2685 | return 0; | |
2686 | } | |
a97773ce BD |
2687 | if (!strcmp(k, "pack.windowmemory")) { |
2688 | window_memory_limit = git_config_ulong(k, v); | |
2689 | return 0; | |
2690 | } | |
2691 | if (!strcmp(k, "pack.depth")) { | |
842aaf93 TT |
2692 | depth = git_config_int(k, v); |
2693 | return 0; | |
2694 | } | |
074b2eea MK |
2695 | if (!strcmp(k, "pack.deltacachesize")) { |
2696 | max_delta_cache_size = git_config_int(k, v); | |
2697 | return 0; | |
2698 | } | |
e3dfddb3 MK |
2699 | if (!strcmp(k, "pack.deltacachelimit")) { |
2700 | cache_max_small_delta_size = git_config_int(k, v); | |
2701 | return 0; | |
2702 | } | |
ae4f07fb VM |
2703 | if (!strcmp(k, "pack.writebitmaphashcache")) { |
2704 | if (git_config_bool(k, v)) | |
2705 | write_bitmap_options |= BITMAP_OPT_HASH_CACHE; | |
2706 | else | |
2707 | write_bitmap_options &= ~BITMAP_OPT_HASH_CACHE; | |
2708 | } | |
6b8fda2d | 2709 | if (!strcmp(k, "pack.usebitmaps")) { |
645c432d | 2710 | use_bitmap_index_default = git_config_bool(k, v); |
6b8fda2d VM |
2711 | return 0; |
2712 | } | |
693b86ff NP |
2713 | if (!strcmp(k, "pack.threads")) { |
2714 | delta_search_threads = git_config_int(k, v); | |
833e3df1 | 2715 | if (delta_search_threads < 0) |
f616db6a | 2716 | die(_("invalid number of threads specified (%d)"), |
693b86ff | 2717 | delta_search_threads); |
9c897c5c | 2718 | if (!HAVE_THREADS && delta_search_threads != 1) { |
f616db6a | 2719 | warning(_("no threads support, ignoring %s"), k); |
2e96d815 ÆAB |
2720 | delta_search_threads = 0; |
2721 | } | |
693b86ff NP |
2722 | return 0; |
2723 | } | |
4d00bda2 | 2724 | if (!strcmp(k, "pack.indexversion")) { |
ebcfb379 JH |
2725 | pack_idx_opts.version = git_config_int(k, v); |
2726 | if (pack_idx_opts.version > 2) | |
f616db6a | 2727 | die(_("bad pack.indexversion=%"PRIu32), |
ebcfb379 | 2728 | pack_idx_opts.version); |
4d00bda2 NP |
2729 | return 0; |
2730 | } | |
ef90d6d4 | 2731 | return git_default_config(k, v, cb); |
4812a93a JK |
2732 | } |
2733 | ||
b5d97e6b | 2734 | static void read_object_list_from_stdin(void) |
c323ac7d | 2735 | { |
188960b4 | 2736 | char line[GIT_MAX_HEXSZ + 1 + PATH_MAX + 2]; |
2737 | struct object_id oid; | |
2738 | const char *p; | |
b2504a0d | 2739 | |
da93d12b | 2740 | for (;;) { |
da93d12b LT |
2741 | if (!fgets(line, sizeof(line), stdin)) { |
2742 | if (feof(stdin)) | |
2743 | break; | |
2744 | if (!ferror(stdin)) | |
1a07e59c | 2745 | die("BUG: fgets returned NULL, not EOF, not error!"); |
687dd75c | 2746 | if (errno != EINTR) |
d824cbba | 2747 | die_errno("fgets"); |
687dd75c JH |
2748 | clearerr(stdin); |
2749 | continue; | |
da93d12b | 2750 | } |
7a979d99 | 2751 | if (line[0] == '-') { |
188960b4 | 2752 | if (get_oid_hex(line+1, &oid)) |
f616db6a | 2753 | die(_("expected edge object ID, got garbage:\n %s"), |
b5d97e6b | 2754 | line); |
188960b4 | 2755 | add_preferred_base(&oid); |
7a979d99 | 2756 | continue; |
21fcd1bd | 2757 | } |
188960b4 | 2758 | if (parse_oid_hex(line, &oid, &p)) |
f616db6a | 2759 | die(_("expected object ID, got garbage:\n %s"), line); |
b5d97e6b | 2760 | |
188960b4 | 2761 | add_preferred_base_object(p + 1); |
fd9b1bae | 2762 | add_object_entry(&oid, OBJ_NONE, p + 1, 0); |
c323ac7d | 2763 | } |
b5d97e6b JH |
2764 | } |
2765 | ||
95308d64 | 2766 | /* Remember to update object flag allocation in object.h */ |
08cdfb13 JH |
2767 | #define OBJECT_ADDED (1u<<20) |
2768 | ||
11c211fa | 2769 | static void show_commit(struct commit *commit, void *data) |
b5d97e6b | 2770 | { |
188960b4 | 2771 | add_object_entry(&commit->object.oid, OBJ_COMMIT, NULL, 0); |
08cdfb13 | 2772 | commit->object.flags |= OBJECT_ADDED; |
7cc8f971 VM |
2773 | |
2774 | if (write_bitmap_index) | |
2775 | index_commit_for_bitmap(commit); | |
28b8a730 JK |
2776 | |
2777 | if (use_delta_islands) | |
2778 | propagate_island_marks(commit); | |
b5d97e6b JH |
2779 | } |
2780 | ||
de1e67d0 | 2781 | static void show_object(struct object *obj, const char *name, void *data) |
b5d97e6b | 2782 | { |
8d2dfc49 | 2783 | add_preferred_base_object(name); |
188960b4 | 2784 | add_object_entry(&obj->oid, obj->type, name, 0); |
8d2dfc49 | 2785 | obj->flags |= OBJECT_ADDED; |
28b8a730 JK |
2786 | |
2787 | if (use_delta_islands) { | |
2788 | const char *p; | |
39490536 | 2789 | unsigned depth; |
28b8a730 JK |
2790 | struct object_entry *ent; |
2791 | ||
39490536 JK |
2792 | /* the empty string is a root tree, which is depth 0 */ |
2793 | depth = *name ? 1 : 0; | |
28b8a730 JK |
2794 | for (p = strchr(name, '/'); p; p = strchr(p + 1, '/')) |
2795 | depth++; | |
2796 | ||
2797 | ent = packlist_find(&to_pack, obj->oid.hash, NULL); | |
108f5303 CC |
2798 | if (ent && depth > oe_tree_depth(&to_pack, ent)) |
2799 | oe_set_tree_depth(&to_pack, ent, depth); | |
28b8a730 | 2800 | } |
b5d97e6b JH |
2801 | } |
2802 | ||
9535ce73 JH |
2803 | static void show_object__ma_allow_any(struct object *obj, const char *name, void *data) |
2804 | { | |
2805 | assert(arg_missing_action == MA_ALLOW_ANY); | |
2806 | ||
2807 | /* | |
2808 | * Quietly ignore ALL missing objects. This avoids problems with | |
2809 | * staging them now and getting an odd error later. | |
2810 | */ | |
2811 | if (!has_object_file(&obj->oid)) | |
2812 | return; | |
2813 | ||
2814 | show_object(obj, name, data); | |
2815 | } | |
2816 | ||
0c16cd49 JT |
2817 | static void show_object__ma_allow_promisor(struct object *obj, const char *name, void *data) |
2818 | { | |
2819 | assert(arg_missing_action == MA_ALLOW_PROMISOR); | |
2820 | ||
2821 | /* | |
2822 | * Quietly ignore EXPECTED missing objects. This avoids problems with | |
2823 | * staging them now and getting an odd error later. | |
2824 | */ | |
2825 | if (!has_object_file(&obj->oid) && is_promisor_object(&obj->oid)) | |
2826 | return; | |
2827 | ||
2828 | show_object(obj, name, data); | |
2829 | } | |
2830 | ||
9535ce73 JH |
2831 | static int option_parse_missing_action(const struct option *opt, |
2832 | const char *arg, int unset) | |
2833 | { | |
2834 | assert(arg); | |
2835 | assert(!unset); | |
2836 | ||
2837 | if (!strcmp(arg, "error")) { | |
2838 | arg_missing_action = MA_ERROR; | |
2839 | fn_show_object = show_object; | |
2840 | return 0; | |
2841 | } | |
2842 | ||
2843 | if (!strcmp(arg, "allow-any")) { | |
2844 | arg_missing_action = MA_ALLOW_ANY; | |
0c16cd49 | 2845 | fetch_if_missing = 0; |
9535ce73 JH |
2846 | fn_show_object = show_object__ma_allow_any; |
2847 | return 0; | |
2848 | } | |
2849 | ||
0c16cd49 JT |
2850 | if (!strcmp(arg, "allow-promisor")) { |
2851 | arg_missing_action = MA_ALLOW_PROMISOR; | |
2852 | fetch_if_missing = 0; | |
2853 | fn_show_object = show_object__ma_allow_promisor; | |
2854 | return 0; | |
2855 | } | |
2856 | ||
9535ce73 JH |
2857 | die(_("invalid value for --missing")); |
2858 | return 0; | |
2859 | } | |
2860 | ||
8d1d8f83 JH |
2861 | static void show_edge(struct commit *commit) |
2862 | { | |
188960b4 | 2863 | add_preferred_base(&commit->object.oid); |
8d1d8f83 JH |
2864 | } |
2865 | ||
08cdfb13 JH |
2866 | struct in_pack_object { |
2867 | off_t offset; | |
2868 | struct object *object; | |
2869 | }; | |
2870 | ||
2871 | struct in_pack { | |
071bcaab RJ |
2872 | unsigned int alloc; |
2873 | unsigned int nr; | |
08cdfb13 JH |
2874 | struct in_pack_object *array; |
2875 | }; | |
2876 | ||
2877 | static void mark_in_pack_object(struct object *object, struct packed_git *p, struct in_pack *in_pack) | |
2878 | { | |
ed1c9977 | 2879 | in_pack->array[in_pack->nr].offset = find_pack_entry_one(object->oid.hash, p); |
08cdfb13 JH |
2880 | in_pack->array[in_pack->nr].object = object; |
2881 | in_pack->nr++; | |
2882 | } | |
2883 | ||
2884 | /* | |
2885 | * Compare the objects in the offset order, in order to emulate the | |
f18d244a | 2886 | * "git rev-list --objects" output that produced the pack originally. |
08cdfb13 JH |
2887 | */ |
2888 | static int ofscmp(const void *a_, const void *b_) | |
2889 | { | |
2890 | struct in_pack_object *a = (struct in_pack_object *)a_; | |
2891 | struct in_pack_object *b = (struct in_pack_object *)b_; | |
2892 | ||
2893 | if (a->offset < b->offset) | |
2894 | return -1; | |
2895 | else if (a->offset > b->offset) | |
2896 | return 1; | |
2897 | else | |
f2fd0760 | 2898 | return oidcmp(&a->object->oid, &b->object->oid); |
08cdfb13 JH |
2899 | } |
2900 | ||
2901 | static void add_objects_in_unpacked_packs(struct rev_info *revs) | |
2902 | { | |
2903 | struct packed_git *p; | |
2904 | struct in_pack in_pack; | |
2905 | uint32_t i; | |
2906 | ||
2907 | memset(&in_pack, 0, sizeof(in_pack)); | |
2908 | ||
454ea2e4 | 2909 | for (p = get_all_packs(the_repository); p; p = p->next) { |
188960b4 | 2910 | struct object_id oid; |
08cdfb13 JH |
2911 | struct object *o; |
2912 | ||
ed7e5fc3 | 2913 | if (!p->pack_local || p->pack_keep || p->pack_keep_in_core) |
08cdfb13 JH |
2914 | continue; |
2915 | if (open_pack_index(p)) | |
f616db6a | 2916 | die(_("cannot open pack index")); |
08cdfb13 JH |
2917 | |
2918 | ALLOC_GROW(in_pack.array, | |
2919 | in_pack.nr + p->num_objects, | |
2920 | in_pack.alloc); | |
2921 | ||
2922 | for (i = 0; i < p->num_objects; i++) { | |
188960b4 | 2923 | nth_packed_object_oid(&oid, p, i); |
2924 | o = lookup_unknown_object(oid.hash); | |
08cdfb13 JH |
2925 | if (!(o->flags & OBJECT_ADDED)) |
2926 | mark_in_pack_object(o, p, &in_pack); | |
2927 | o->flags |= OBJECT_ADDED; | |
2928 | } | |
2929 | } | |
2930 | ||
2931 | if (in_pack.nr) { | |
9ed0d8d6 | 2932 | QSORT(in_pack.array, in_pack.nr, ofscmp); |
08cdfb13 JH |
2933 | for (i = 0; i < in_pack.nr; i++) { |
2934 | struct object *o = in_pack.array[i].object; | |
188960b4 | 2935 | add_object_entry(&o->oid, o->type, "", 0); |
08cdfb13 JH |
2936 | } |
2937 | } | |
2938 | free(in_pack.array); | |
2939 | } | |
2940 | ||
76c1d9a0 | 2941 | static int add_loose_object(const struct object_id *oid, const char *path, |
e26a8c47 JK |
2942 | void *data) |
2943 | { | |
0df8e965 | 2944 | enum object_type type = oid_object_info(the_repository, oid, NULL); |
e26a8c47 JK |
2945 | |
2946 | if (type < 0) { | |
f616db6a | 2947 | warning(_("loose object at %s could not be examined"), path); |
e26a8c47 JK |
2948 | return 0; |
2949 | } | |
2950 | ||
188960b4 | 2951 | add_object_entry(oid, type, "", 0); |
e26a8c47 JK |
2952 | return 0; |
2953 | } | |
2954 | ||
2955 | /* | |
2956 | * We actually don't even have to worry about reachability here. | |
2957 | * add_object_entry will weed out duplicates, so we just add every | |
2958 | * loose object we find. | |
2959 | */ | |
2960 | static void add_unreachable_loose_objects(void) | |
2961 | { | |
2962 | for_each_loose_file_in_objdir(get_object_directory(), | |
2963 | add_loose_object, | |
2964 | NULL, NULL, NULL); | |
2965 | } | |
2966 | ||
188960b4 | 2967 | static int has_sha1_pack_kept_or_nonlocal(const struct object_id *oid) |
094085e3 BC |
2968 | { |
2969 | static struct packed_git *last_found = (void *)1; | |
2970 | struct packed_git *p; | |
2971 | ||
a80d72db | 2972 | p = (last_found != (void *)1) ? last_found : |
454ea2e4 | 2973 | get_all_packs(the_repository); |
094085e3 BC |
2974 | |
2975 | while (p) { | |
ed7e5fc3 NTND |
2976 | if ((!p->pack_local || p->pack_keep || |
2977 | p->pack_keep_in_core) && | |
188960b4 | 2978 | find_pack_entry_one(oid->hash, p)) { |
094085e3 BC |
2979 | last_found = p; |
2980 | return 1; | |
2981 | } | |
2982 | if (p == last_found) | |
454ea2e4 | 2983 | p = get_all_packs(the_repository); |
094085e3 BC |
2984 | else |
2985 | p = p->next; | |
2986 | if (p == last_found) | |
2987 | p = p->next; | |
2988 | } | |
2989 | return 0; | |
2990 | } | |
2991 | ||
abcb8655 JK |
2992 | /* |
2993 | * Store a list of sha1s that are should not be discarded | |
2994 | * because they are either written too recently, or are | |
2995 | * reachable from another object that was. | |
2996 | * | |
2997 | * This is filled by get_object_list. | |
2998 | */ | |
910650d2 | 2999 | static struct oid_array recent_objects; |
abcb8655 | 3000 | |
4ce3621a | 3001 | static int loosened_object_can_be_discarded(const struct object_id *oid, |
dddbad72 | 3002 | timestamp_t mtime) |
d0d46abc JK |
3003 | { |
3004 | if (!unpack_unreachable_expiration) | |
3005 | return 0; | |
3006 | if (mtime > unpack_unreachable_expiration) | |
3007 | return 0; | |
910650d2 | 3008 | if (oid_array_lookup(&recent_objects, oid) >= 0) |
abcb8655 | 3009 | return 0; |
d0d46abc JK |
3010 | return 1; |
3011 | } | |
3012 | ||
ca11b212 NP |
3013 | static void loosen_unused_packed_objects(struct rev_info *revs) |
3014 | { | |
3015 | struct packed_git *p; | |
3016 | uint32_t i; | |
4ce3621a | 3017 | struct object_id oid; |
ca11b212 | 3018 | |
454ea2e4 | 3019 | for (p = get_all_packs(the_repository); p; p = p->next) { |
ed7e5fc3 | 3020 | if (!p->pack_local || p->pack_keep || p->pack_keep_in_core) |
ca11b212 NP |
3021 | continue; |
3022 | ||
3023 | if (open_pack_index(p)) | |
f616db6a | 3024 | die(_("cannot open pack index")); |
ca11b212 NP |
3025 | |
3026 | for (i = 0; i < p->num_objects; i++) { | |
4ce3621a | 3027 | nth_packed_object_oid(&oid, p, i); |
3028 | if (!packlist_find(&to_pack, oid.hash, NULL) && | |
188960b4 | 3029 | !has_sha1_pack_kept_or_nonlocal(&oid) && |
4ce3621a | 3030 | !loosened_object_can_be_discarded(&oid, p->mtime)) |
4bdb70a4 | 3031 | if (force_object_loose(&oid, p->mtime)) |
f616db6a | 3032 | die(_("unable to force loose object")); |
ca11b212 NP |
3033 | } |
3034 | } | |
3035 | } | |
3036 | ||
69e4b342 | 3037 | /* |
645c432d KS |
3038 | * This tracks any options which pack-reuse code expects to be on, or which a |
3039 | * reader of the pack might not understand, and which would therefore prevent | |
3040 | * blind reuse of what we have on disk. | |
69e4b342 JK |
3041 | */ |
3042 | static int pack_options_allow_reuse(void) | |
3043 | { | |
9df4a607 JK |
3044 | return pack_to_stdout && |
3045 | allow_ofs_delta && | |
ed7e5fc3 NTND |
3046 | !ignore_packed_keep_on_disk && |
3047 | !ignore_packed_keep_in_core && | |
9df4a607 JK |
3048 | (!local || !have_non_local_packs) && |
3049 | !incremental; | |
69e4b342 JK |
3050 | } |
3051 | ||
6b8fda2d VM |
3052 | static int get_object_list_from_bitmap(struct rev_info *revs) |
3053 | { | |
3ae5fa07 | 3054 | if (!(bitmap_git = prepare_bitmap_walk(revs))) |
6b8fda2d VM |
3055 | return -1; |
3056 | ||
69e4b342 JK |
3057 | if (pack_options_allow_reuse() && |
3058 | !reuse_partial_packfile_from_bitmap( | |
3ae5fa07 | 3059 | bitmap_git, |
6b8fda2d VM |
3060 | &reuse_packfile, |
3061 | &reuse_packfile_objects, | |
3062 | &reuse_packfile_offset)) { | |
3063 | assert(reuse_packfile_objects); | |
3064 | nr_result += reuse_packfile_objects; | |
78d2214e | 3065 | display_progress(progress_state, nr_result); |
6b8fda2d VM |
3066 | } |
3067 | ||
3ae5fa07 | 3068 | traverse_bitmap_commit_list(bitmap_git, &add_object_entry_from_bitmap); |
6b8fda2d VM |
3069 | return 0; |
3070 | } | |
3071 | ||
abcb8655 | 3072 | static void record_recent_object(struct object *obj, |
de1e67d0 | 3073 | const char *name, |
abcb8655 JK |
3074 | void *data) |
3075 | { | |
910650d2 | 3076 | oid_array_append(&recent_objects, &obj->oid); |
abcb8655 JK |
3077 | } |
3078 | ||
3079 | static void record_recent_commit(struct commit *commit, void *data) | |
3080 | { | |
910650d2 | 3081 | oid_array_append(&recent_objects, &commit->object.oid); |
abcb8655 JK |
3082 | } |
3083 | ||
8d1d8f83 | 3084 | static void get_object_list(int ac, const char **av) |
b5d97e6b JH |
3085 | { |
3086 | struct rev_info revs; | |
3087 | char line[1000]; | |
b5d97e6b | 3088 | int flags = 0; |
a4544b31 | 3089 | int save_warning; |
b5d97e6b | 3090 | |
2abf3503 | 3091 | repo_init_revisions(the_repository, &revs, NULL); |
b5d97e6b | 3092 | save_commit_buffer = 0; |
669b1d2a | 3093 | revs.allow_exclude_promisor_objects_opt = 1; |
b5d97e6b JH |
3094 | setup_revisions(ac, av, &revs, NULL); |
3095 | ||
b790e0f6 | 3096 | /* make sure shallows are read */ |
c8813487 | 3097 | is_repository_shallow(the_repository); |
b790e0f6 | 3098 | |
a4544b31 DS |
3099 | save_warning = warn_on_object_refname_ambiguity; |
3100 | warn_on_object_refname_ambiguity = 0; | |
3101 | ||
b5d97e6b JH |
3102 | while (fgets(line, sizeof(line), stdin) != NULL) { |
3103 | int len = strlen(line); | |
872c930d | 3104 | if (len && line[len - 1] == '\n') |
b5d97e6b JH |
3105 | line[--len] = 0; |
3106 | if (!len) | |
3107 | break; | |
3108 | if (*line == '-') { | |
3109 | if (!strcmp(line, "--not")) { | |
3110 | flags ^= UNINTERESTING; | |
7cc8f971 | 3111 | write_bitmap_index = 0; |
b5d97e6b JH |
3112 | continue; |
3113 | } | |
b790e0f6 | 3114 | if (starts_with(line, "--shallow ")) { |
e92b848c | 3115 | struct object_id oid; |
3116 | if (get_oid_hex(line + 10, &oid)) | |
b790e0f6 | 3117 | die("not an SHA-1 '%s'", line + 10); |
19143f13 | 3118 | register_shallow(the_repository, &oid); |
f7f91086 | 3119 | use_bitmap_index = 0; |
b790e0f6 NTND |
3120 | continue; |
3121 | } | |
f616db6a | 3122 | die(_("not a rev '%s'"), line); |
b5d97e6b | 3123 | } |
8e676e8b | 3124 | if (handle_revision_arg(line, &revs, flags, REVARG_CANNOT_BE_FILENAME)) |
f616db6a | 3125 | die(_("bad revision '%s'"), line); |
b5d97e6b JH |
3126 | } |
3127 | ||
a4544b31 DS |
3128 | warn_on_object_refname_ambiguity = save_warning; |
3129 | ||
6b8fda2d VM |
3130 | if (use_bitmap_index && !get_object_list_from_bitmap(&revs)) |
3131 | return; | |
3132 | ||
28b8a730 JK |
3133 | if (use_delta_islands) |
3134 | load_delta_islands(); | |
3135 | ||
3d51e1b5 | 3136 | if (prepare_revision_walk(&revs)) |
f616db6a | 3137 | die(_("revision walk setup failed")); |
e76a5fb4 | 3138 | mark_edges_uninteresting(&revs, show_edge); |
9535ce73 JH |
3139 | |
3140 | if (!fn_show_object) | |
3141 | fn_show_object = show_object; | |
3142 | traverse_commit_list_filtered(&filter_options, &revs, | |
3143 | show_commit, fn_show_object, NULL, | |
3144 | NULL); | |
08cdfb13 | 3145 | |
abcb8655 JK |
3146 | if (unpack_unreachable_expiration) { |
3147 | revs.ignore_missing_links = 1; | |
3148 | if (add_unseen_recent_objects_to_traversal(&revs, | |
3149 | unpack_unreachable_expiration)) | |
f616db6a | 3150 | die(_("unable to add recent objects")); |
abcb8655 | 3151 | if (prepare_revision_walk(&revs)) |
f616db6a | 3152 | die(_("revision walk setup failed")); |
abcb8655 JK |
3153 | traverse_commit_list(&revs, record_recent_commit, |
3154 | record_recent_object, NULL); | |
3155 | } | |
3156 | ||
08cdfb13 JH |
3157 | if (keep_unreachable) |
3158 | add_objects_in_unpacked_packs(&revs); | |
e26a8c47 JK |
3159 | if (pack_loose_unreachable) |
3160 | add_unreachable_loose_objects(); | |
ca11b212 NP |
3161 | if (unpack_unreachable) |
3162 | loosen_unused_packed_objects(&revs); | |
abcb8655 | 3163 | |
910650d2 | 3164 | oid_array_clear(&recent_objects); |
b5d97e6b JH |
3165 | } |
3166 | ||
ed7e5fc3 NTND |
3167 | static void add_extra_kept_packs(const struct string_list *names) |
3168 | { | |
3169 | struct packed_git *p; | |
3170 | ||
3171 | if (!names->nr) | |
3172 | return; | |
3173 | ||
454ea2e4 | 3174 | for (p = get_all_packs(the_repository); p; p = p->next) { |
ed7e5fc3 NTND |
3175 | const char *name = basename(p->pack_name); |
3176 | int i; | |
3177 | ||
3178 | if (!p->pack_local) | |
3179 | continue; | |
3180 | ||
3181 | for (i = 0; i < names->nr; i++) | |
3182 | if (!fspathcmp(name, names->items[i].string)) | |
3183 | break; | |
3184 | ||
3185 | if (i < names->nr) { | |
3186 | p->pack_keep_in_core = 1; | |
3187 | ignore_packed_keep_in_core = 1; | |
3188 | continue; | |
3189 | } | |
3190 | } | |
3191 | } | |
3192 | ||
99fb6e04 NTND |
3193 | static int option_parse_index_version(const struct option *opt, |
3194 | const char *arg, int unset) | |
3195 | { | |
3196 | char *c; | |
3197 | const char *val = arg; | |
517fe807 JK |
3198 | |
3199 | BUG_ON_OPT_NEG(unset); | |
3200 | ||
99fb6e04 NTND |
3201 | pack_idx_opts.version = strtoul(val, &c, 10); |
3202 | if (pack_idx_opts.version > 2) | |
3203 | die(_("unsupported index version %s"), val); | |
3204 | if (*c == ',' && c[1]) | |
3205 | pack_idx_opts.off32_limit = strtoul(c+1, &c, 0); | |
3206 | if (*c || pack_idx_opts.off32_limit & 0x80000000) | |
3207 | die(_("bad index version '%s'"), val); | |
3208 | return 0; | |
3209 | } | |
3210 | ||
7e52f566 JK |
3211 | static int option_parse_unpack_unreachable(const struct option *opt, |
3212 | const char *arg, int unset) | |
3213 | { | |
3214 | if (unset) { | |
3215 | unpack_unreachable = 0; | |
3216 | unpack_unreachable_expiration = 0; | |
3217 | } | |
3218 | else { | |
3219 | unpack_unreachable = 1; | |
3220 | if (arg) | |
3221 | unpack_unreachable_expiration = approxidate(arg); | |
3222 | } | |
3223 | return 0; | |
3224 | } | |
3225 | ||
b5d97e6b JH |
3226 | int cmd_pack_objects(int argc, const char **argv, const char *prefix) |
3227 | { | |
b5d97e6b | 3228 | int use_internal_rev_list = 0; |
2dacf26d | 3229 | int shallow = 0; |
4f366275 | 3230 | int all_progress_implied = 0; |
edfbb2aa | 3231 | struct argv_array rp = ARGV_ARRAY_INIT; |
99fb6e04 | 3232 | int rev_list_unpacked = 0, rev_list_all = 0, rev_list_reflog = 0; |
c90f9e13 | 3233 | int rev_list_index = 0; |
ed7e5fc3 | 3234 | struct string_list keep_pack_list = STRING_LIST_INIT_NODUP; |
99fb6e04 NTND |
3235 | struct option pack_objects_options[] = { |
3236 | OPT_SET_INT('q', "quiet", &progress, | |
4c688120 | 3237 | N_("do not show progress meter"), 0), |
99fb6e04 | 3238 | OPT_SET_INT(0, "progress", &progress, |
4c688120 | 3239 | N_("show progress meter"), 1), |
99fb6e04 | 3240 | OPT_SET_INT(0, "all-progress", &progress, |
4c688120 | 3241 | N_("show progress meter during object writing phase"), 2), |
99fb6e04 NTND |
3242 | OPT_BOOL(0, "all-progress-implied", |
3243 | &all_progress_implied, | |
4c688120 | 3244 | N_("similar to --all-progress when progress meter is shown")), |
cbd23de8 | 3245 | { OPTION_CALLBACK, 0, "index-version", NULL, N_("<version>[,<offset>]"), |
4c688120 | 3246 | N_("write the pack index file in the specified idx format version"), |
f53c163b | 3247 | PARSE_OPT_NONEG, option_parse_index_version }, |
2a514ed8 CB |
3248 | OPT_MAGNITUDE(0, "max-pack-size", &pack_size_limit, |
3249 | N_("maximum size of each output pack file")), | |
99fb6e04 | 3250 | OPT_BOOL(0, "local", &local, |
4c688120 | 3251 | N_("ignore borrowed objects from alternate object store")), |
99fb6e04 | 3252 | OPT_BOOL(0, "incremental", &incremental, |
4c688120 | 3253 | N_("ignore packed objects")), |
99fb6e04 | 3254 | OPT_INTEGER(0, "window", &window, |
4c688120 | 3255 | N_("limit pack window by objects")), |
2a514ed8 CB |
3256 | OPT_MAGNITUDE(0, "window-memory", &window_memory_limit, |
3257 | N_("limit pack window by memory in addition to object limit")), | |
99fb6e04 | 3258 | OPT_INTEGER(0, "depth", &depth, |
4c688120 | 3259 | N_("maximum length of delta chain allowed in the resulting pack")), |
99fb6e04 | 3260 | OPT_BOOL(0, "reuse-delta", &reuse_delta, |
4c688120 | 3261 | N_("reuse existing deltas")), |
99fb6e04 | 3262 | OPT_BOOL(0, "reuse-object", &reuse_object, |
4c688120 | 3263 | N_("reuse existing objects")), |
99fb6e04 | 3264 | OPT_BOOL(0, "delta-base-offset", &allow_ofs_delta, |
4c688120 | 3265 | N_("use OFS_DELTA objects")), |
99fb6e04 | 3266 | OPT_INTEGER(0, "threads", &delta_search_threads, |
4c688120 | 3267 | N_("use threads when searching for best delta matches")), |
99fb6e04 | 3268 | OPT_BOOL(0, "non-empty", &non_empty, |
4c688120 | 3269 | N_("do not create an empty pack output")), |
99fb6e04 | 3270 | OPT_BOOL(0, "revs", &use_internal_rev_list, |
4c688120 | 3271 | N_("read revision arguments from standard input")), |
3e4a67b4 NTND |
3272 | OPT_SET_INT_F(0, "unpacked", &rev_list_unpacked, |
3273 | N_("limit the objects to those that are not yet packed"), | |
3274 | 1, PARSE_OPT_NONEG), | |
3275 | OPT_SET_INT_F(0, "all", &rev_list_all, | |
3276 | N_("include objects reachable from any reference"), | |
3277 | 1, PARSE_OPT_NONEG), | |
3278 | OPT_SET_INT_F(0, "reflog", &rev_list_reflog, | |
3279 | N_("include objects referred by reflog entries"), | |
3280 | 1, PARSE_OPT_NONEG), | |
3281 | OPT_SET_INT_F(0, "indexed-objects", &rev_list_index, | |
3282 | N_("include objects referred to by the index"), | |
3283 | 1, PARSE_OPT_NONEG), | |
99fb6e04 | 3284 | OPT_BOOL(0, "stdout", &pack_to_stdout, |
4c688120 | 3285 | N_("output pack to stdout")), |
99fb6e04 | 3286 | OPT_BOOL(0, "include-tag", &include_tag, |
4c688120 | 3287 | N_("include tag objects that refer to objects to be packed")), |
99fb6e04 | 3288 | OPT_BOOL(0, "keep-unreachable", &keep_unreachable, |
4c688120 | 3289 | N_("keep unreachable objects")), |
e26a8c47 JK |
3290 | OPT_BOOL(0, "pack-loose-unreachable", &pack_loose_unreachable, |
3291 | N_("pack loose unreachable objects")), | |
4c688120 NTND |
3292 | { OPTION_CALLBACK, 0, "unpack-unreachable", NULL, N_("time"), |
3293 | N_("unpack unreachable objects newer than <time>"), | |
7e52f566 | 3294 | PARSE_OPT_OPTARG, option_parse_unpack_unreachable }, |
99fb6e04 | 3295 | OPT_BOOL(0, "thin", &thin, |
4c688120 | 3296 | N_("create thin packs")), |
2dacf26d | 3297 | OPT_BOOL(0, "shallow", &shallow, |
3298 | N_("create packs suitable for shallow fetches")), | |
ed7e5fc3 | 3299 | OPT_BOOL(0, "honor-pack-keep", &ignore_packed_keep_on_disk, |
4c688120 | 3300 | N_("ignore packs that have companion .keep file")), |
ed7e5fc3 NTND |
3301 | OPT_STRING_LIST(0, "keep-pack", &keep_pack_list, N_("name"), |
3302 | N_("ignore this pack")), | |
99fb6e04 | 3303 | OPT_INTEGER(0, "compression", &pack_compression_level, |
4c688120 | 3304 | N_("pack compression level")), |
99fb6e04 | 3305 | OPT_SET_INT(0, "keep-true-parents", &grafts_replace_parents, |
4c688120 | 3306 | N_("do not hide commits by grafts"), 0), |
6b8fda2d VM |
3307 | OPT_BOOL(0, "use-bitmap-index", &use_bitmap_index, |
3308 | N_("use a bitmap index if available to speed up counting objects")), | |
7cc8f971 VM |
3309 | OPT_BOOL(0, "write-bitmap-index", &write_bitmap_index, |
3310 | N_("write a bitmap index together with the pack index")), | |
9535ce73 JH |
3311 | OPT_PARSE_LIST_OBJECTS_FILTER(&filter_options), |
3312 | { OPTION_CALLBACK, 0, "missing", NULL, N_("action"), | |
3313 | N_("handling for missing objects"), PARSE_OPT_NONEG, | |
3314 | option_parse_missing_action }, | |
0c16cd49 JT |
3315 | OPT_BOOL(0, "exclude-promisor-objects", &exclude_promisor_objects, |
3316 | N_("do not pack objects in promisor packfiles")), | |
28b8a730 JK |
3317 | OPT_BOOL(0, "delta-islands", &use_delta_islands, |
3318 | N_("respect islands during delta compression")), | |
99fb6e04 NTND |
3319 | OPT_END(), |
3320 | }; | |
8d1d8f83 | 3321 | |
0c6804ab NTND |
3322 | if (DFS_NUM_STATES > (1 << OE_DFS_STATE_BITS)) |
3323 | BUG("too many dfs states, increase OE_DFS_STATE_BITS"); | |
3324 | ||
6ebd1caf | 3325 | read_replace_refs = 0; |
dae556bd | 3326 | |
ebcfb379 | 3327 | reset_pack_idx_option(&pack_idx_opts); |
ef90d6d4 | 3328 | git_config(git_pack_config, NULL); |
b5d97e6b JH |
3329 | |
3330 | progress = isatty(2); | |
99fb6e04 NTND |
3331 | argc = parse_options(argc, argv, prefix, pack_objects_options, |
3332 | pack_usage, 0); | |
b5d97e6b | 3333 | |
99fb6e04 NTND |
3334 | if (argc) { |
3335 | base_name = argv[0]; | |
3336 | argc--; | |
b5d97e6b | 3337 | } |
99fb6e04 NTND |
3338 | if (pack_to_stdout != !base_name || argc) |
3339 | usage_with_options(pack_usage, pack_objects_options); | |
b5d97e6b | 3340 | |
b5c0cbd8 NTND |
3341 | if (depth >= (1 << OE_DEPTH_BITS)) { |
3342 | warning(_("delta chain depth %d is too deep, forcing %d"), | |
3343 | depth, (1 << OE_DEPTH_BITS) - 1); | |
3344 | depth = (1 << OE_DEPTH_BITS) - 1; | |
3345 | } | |
0cb3c142 NTND |
3346 | if (cache_max_small_delta_size >= (1U << OE_Z_DELTA_BITS)) { |
3347 | warning(_("pack.deltaCacheLimit is too high, forcing %d"), | |
3348 | (1U << OE_Z_DELTA_BITS) - 1); | |
3349 | cache_max_small_delta_size = (1U << OE_Z_DELTA_BITS) - 1; | |
3350 | } | |
b5c0cbd8 | 3351 | |
edfbb2aa | 3352 | argv_array_push(&rp, "pack-objects"); |
99fb6e04 NTND |
3353 | if (thin) { |
3354 | use_internal_rev_list = 1; | |
2dacf26d | 3355 | argv_array_push(&rp, shallow |
3356 | ? "--objects-edge-aggressive" | |
3357 | : "--objects-edge"); | |
99fb6e04 | 3358 | } else |
edfbb2aa | 3359 | argv_array_push(&rp, "--objects"); |
b5d97e6b | 3360 | |
99fb6e04 NTND |
3361 | if (rev_list_all) { |
3362 | use_internal_rev_list = 1; | |
edfbb2aa | 3363 | argv_array_push(&rp, "--all"); |
99fb6e04 NTND |
3364 | } |
3365 | if (rev_list_reflog) { | |
3366 | use_internal_rev_list = 1; | |
edfbb2aa | 3367 | argv_array_push(&rp, "--reflog"); |
99fb6e04 | 3368 | } |
c90f9e13 JK |
3369 | if (rev_list_index) { |
3370 | use_internal_rev_list = 1; | |
3371 | argv_array_push(&rp, "--indexed-objects"); | |
99fb6e04 NTND |
3372 | } |
3373 | if (rev_list_unpacked) { | |
3374 | use_internal_rev_list = 1; | |
edfbb2aa | 3375 | argv_array_push(&rp, "--unpacked"); |
99fb6e04 | 3376 | } |
b5d97e6b | 3377 | |
0c16cd49 JT |
3378 | if (exclude_promisor_objects) { |
3379 | use_internal_rev_list = 1; | |
3380 | fetch_if_missing = 0; | |
3381 | argv_array_push(&rp, "--exclude-promisor-objects"); | |
3382 | } | |
58bd77b6 NTND |
3383 | if (unpack_unreachable || keep_unreachable || pack_loose_unreachable) |
3384 | use_internal_rev_list = 1; | |
0c16cd49 | 3385 | |
99fb6e04 NTND |
3386 | if (!reuse_object) |
3387 | reuse_delta = 0; | |
3388 | if (pack_compression_level == -1) | |
3389 | pack_compression_level = Z_DEFAULT_COMPRESSION; | |
3390 | else if (pack_compression_level < 0 || pack_compression_level > Z_BEST_COMPRESSION) | |
f616db6a | 3391 | die(_("bad pack compression level %d"), pack_compression_level); |
0c45d258 JH |
3392 | |
3393 | if (!delta_search_threads) /* --threads=0 means autodetect */ | |
3394 | delta_search_threads = online_cpus(); | |
3395 | ||
9c897c5c | 3396 | if (!HAVE_THREADS && delta_search_threads != 1) |
f616db6a | 3397 | warning(_("no threads support, ignoring --threads")); |
2b84b5a8 JS |
3398 | if (!pack_to_stdout && !pack_size_limit) |
3399 | pack_size_limit = pack_size_limit_cfg; | |
01c12a23 | 3400 | if (pack_to_stdout && pack_size_limit) |
f616db6a | 3401 | die(_("--max-pack-size cannot be used to build a pack for transfer")); |
07cf0f24 | 3402 | if (pack_size_limit && pack_size_limit < 1024*1024) { |
f616db6a | 3403 | warning(_("minimum pack size limit is 1 MiB")); |
07cf0f24 NP |
3404 | pack_size_limit = 1024*1024; |
3405 | } | |
01c12a23 | 3406 | |
8d1d8f83 | 3407 | if (!pack_to_stdout && thin) |
f616db6a | 3408 | die(_("--thin cannot be used to build an indexable pack")); |
b5d97e6b | 3409 | |
ca11b212 | 3410 | if (keep_unreachable && unpack_unreachable) |
f616db6a | 3411 | die(_("--keep-unreachable and --unpack-unreachable are incompatible")); |
b1e757f3 JK |
3412 | if (!rev_list_all || !rev_list_reflog || !rev_list_index) |
3413 | unpack_unreachable_expiration = 0; | |
ca11b212 | 3414 | |
9535ce73 JH |
3415 | if (filter_options.choice) { |
3416 | if (!pack_to_stdout) | |
f616db6a | 3417 | die(_("cannot use --filter without --stdout")); |
9535ce73 JH |
3418 | use_bitmap_index = 0; |
3419 | } | |
3420 | ||
645c432d KS |
3421 | /* |
3422 | * "soft" reasons not to use bitmaps - for on-disk repack by default we want | |
3423 | * | |
3424 | * - to produce good pack (with bitmap index not-yet-packed objects are | |
3425 | * packed in suboptimal order). | |
3426 | * | |
3427 | * - to use more robust pack-generation codepath (avoiding possible | |
3428 | * bugs in bitmap code and possible bitmap index corruption). | |
3429 | */ | |
3430 | if (!pack_to_stdout) | |
3431 | use_bitmap_index_default = 0; | |
3432 | ||
3433 | if (use_bitmap_index < 0) | |
3434 | use_bitmap_index = use_bitmap_index_default; | |
3435 | ||
3436 | /* "hard" reasons not to use bitmaps; these just won't work at all */ | |
c8813487 | 3437 | if (!use_internal_rev_list || (!pack_to_stdout && write_bitmap_index) || is_repository_shallow(the_repository)) |
6b8fda2d VM |
3438 | use_bitmap_index = 0; |
3439 | ||
7cc8f971 VM |
3440 | if (pack_to_stdout || !rev_list_all) |
3441 | write_bitmap_index = 0; | |
3442 | ||
28b8a730 JK |
3443 | if (use_delta_islands) |
3444 | argv_array_push(&rp, "--topo-order"); | |
3445 | ||
4f366275 NP |
3446 | if (progress && all_progress_implied) |
3447 | progress = 2; | |
3448 | ||
ed7e5fc3 NTND |
3449 | add_extra_kept_packs(&keep_pack_list); |
3450 | if (ignore_packed_keep_on_disk) { | |
56dfeb62 | 3451 | struct packed_git *p; |
454ea2e4 | 3452 | for (p = get_all_packs(the_repository); p; p = p->next) |
56dfeb62 JK |
3453 | if (p->pack_local && p->pack_keep) |
3454 | break; | |
3455 | if (!p) /* no keep-able packs found */ | |
ed7e5fc3 | 3456 | ignore_packed_keep_on_disk = 0; |
56dfeb62 JK |
3457 | } |
3458 | if (local) { | |
3459 | /* | |
ed7e5fc3 NTND |
3460 | * unlike ignore_packed_keep_on_disk above, we do not |
3461 | * want to unset "local" based on looking at packs, as | |
3462 | * it also covers non-local objects | |
56dfeb62 JK |
3463 | */ |
3464 | struct packed_git *p; | |
454ea2e4 | 3465 | for (p = get_all_packs(the_repository); p; p = p->next) { |
56dfeb62 JK |
3466 | if (!p->pack_local) { |
3467 | have_non_local_packs = 1; | |
3468 | break; | |
3469 | } | |
3470 | } | |
3471 | } | |
b5d97e6b | 3472 | |
43fa44fa NTND |
3473 | prepare_packing_data(&to_pack); |
3474 | ||
13aaf148 | 3475 | if (progress) |
5af05043 | 3476 | progress_state = start_progress(_("Enumerating objects"), 0); |
b5d97e6b JH |
3477 | if (!use_internal_rev_list) |
3478 | read_object_list_from_stdin(); | |
8d1d8f83 | 3479 | else { |
edfbb2aa JK |
3480 | get_object_list(rp.argc, rp.argv); |
3481 | argv_array_clear(&rp); | |
8d1d8f83 | 3482 | } |
0ef95f72 | 3483 | cleanup_preferred_base(); |
d155254c MH |
3484 | if (include_tag && nr_result) |
3485 | for_each_ref(add_ref_tag, NULL); | |
4d4fcc54 | 3486 | stop_progress(&progress_state); |
96a02f8f | 3487 | |
f0b0af1b | 3488 | if (non_empty && !nr_result) |
1c4a2912 | 3489 | return 0; |
f7ae6a93 NP |
3490 | if (nr_result) |
3491 | prepare_pack(window, depth); | |
d01fb92f | 3492 | write_pack_file(); |
ab7cd7bb | 3493 | if (progress) |
f616db6a NTND |
3494 | fprintf_ln(stderr, |
3495 | _("Total %"PRIu32" (delta %"PRIu32")," | |
3496 | " reused %"PRIu32" (delta %"PRIu32")"), | |
1a07e59c | 3497 | written, written_delta, reused, reused_delta); |
c323ac7d LT |
3498 | return 0; |
3499 | } |