]>
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 | 28 | #include "reachable.h" |
fe299ec5 | 29 | #include "oid-array.h" |
dbbcd44f | 30 | #include "strvec.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" |
ae417807 | 36 | #include "trace2.h" |
120ad2b0 | 37 | #include "shallow.h" |
e00549aa | 38 | #include "promisor-remote.h" |
8ecce684 | 39 | |
7d089fb9 ÆAB |
40 | /* |
41 | * Objects we are going to pack are collected in the `to_pack` structure. | |
42 | * It contains an array (dynamically expanded) of the object data, and a map | |
43 | * that can resolve SHA1s to their position in the array. | |
44 | */ | |
45 | static struct packing_data to_pack; | |
46 | ||
47 | static inline struct object_entry *oe_delta( | |
48 | const struct packing_data *pack, | |
49 | const struct object_entry *e) | |
50 | { | |
51 | if (!e->delta_idx) | |
52 | return NULL; | |
53 | if (e->ext_base) | |
54 | return &pack->ext_bases[e->delta_idx - 1]; | |
55 | else | |
56 | return &pack->objects[e->delta_idx - 1]; | |
57 | } | |
58 | ||
59 | static inline unsigned long oe_delta_size(struct packing_data *pack, | |
60 | const struct object_entry *e) | |
61 | { | |
62 | if (e->delta_size_valid) | |
63 | return e->delta_size_; | |
64 | ||
65 | /* | |
66 | * pack->delta_size[] can't be NULL because oe_set_delta_size() | |
67 | * must have been called when a new delta is saved with | |
68 | * oe_set_delta(). | |
69 | * If oe_delta() returns NULL (i.e. default state, which means | |
70 | * delta_size_valid is also false), then the caller must never | |
71 | * call oe_delta_size(). | |
72 | */ | |
73 | return pack->delta_size[e - pack->objects]; | |
74 | } | |
75 | ||
76 | unsigned long oe_get_size_slow(struct packing_data *pack, | |
77 | const struct object_entry *e); | |
78 | ||
79 | static inline unsigned long oe_size(struct packing_data *pack, | |
80 | const struct object_entry *e) | |
81 | { | |
82 | if (e->size_valid) | |
83 | return e->size_; | |
84 | ||
85 | return oe_get_size_slow(pack, e); | |
86 | } | |
87 | ||
88 | static inline void oe_set_delta(struct packing_data *pack, | |
89 | struct object_entry *e, | |
90 | struct object_entry *delta) | |
91 | { | |
92 | if (delta) | |
93 | e->delta_idx = (delta - pack->objects) + 1; | |
94 | else | |
95 | e->delta_idx = 0; | |
96 | } | |
97 | ||
98 | static inline struct object_entry *oe_delta_sibling( | |
99 | const struct packing_data *pack, | |
100 | const struct object_entry *e) | |
101 | { | |
102 | if (e->delta_sibling_idx) | |
103 | return &pack->objects[e->delta_sibling_idx - 1]; | |
104 | return NULL; | |
105 | } | |
106 | ||
107 | static inline struct object_entry *oe_delta_child( | |
108 | const struct packing_data *pack, | |
109 | const struct object_entry *e) | |
110 | { | |
111 | if (e->delta_child_idx) | |
112 | return &pack->objects[e->delta_child_idx - 1]; | |
113 | return NULL; | |
114 | } | |
115 | ||
116 | static inline void oe_set_delta_child(struct packing_data *pack, | |
117 | struct object_entry *e, | |
118 | struct object_entry *delta) | |
119 | { | |
120 | if (delta) | |
121 | e->delta_child_idx = (delta - pack->objects) + 1; | |
122 | else | |
123 | e->delta_child_idx = 0; | |
124 | } | |
125 | ||
126 | static inline void oe_set_delta_sibling(struct packing_data *pack, | |
127 | struct object_entry *e, | |
128 | struct object_entry *delta) | |
129 | { | |
130 | if (delta) | |
131 | e->delta_sibling_idx = (delta - pack->objects) + 1; | |
132 | else | |
133 | e->delta_sibling_idx = 0; | |
134 | } | |
135 | ||
136 | static inline void oe_set_size(struct packing_data *pack, | |
137 | struct object_entry *e, | |
138 | unsigned long size) | |
139 | { | |
140 | if (size < pack->oe_size_limit) { | |
141 | e->size_ = size; | |
142 | e->size_valid = 1; | |
143 | } else { | |
144 | e->size_valid = 0; | |
145 | if (oe_get_size_slow(pack, e) != size) | |
146 | BUG("'size' is supposed to be the object size!"); | |
147 | } | |
148 | } | |
149 | ||
150 | static inline void oe_set_delta_size(struct packing_data *pack, | |
151 | struct object_entry *e, | |
152 | unsigned long size) | |
153 | { | |
154 | if (size < pack->oe_delta_size_limit) { | |
155 | e->delta_size_ = size; | |
156 | e->delta_size_valid = 1; | |
157 | } else { | |
158 | packing_data_lock(pack); | |
159 | if (!pack->delta_size) | |
160 | ALLOC_ARRAY(pack->delta_size, pack->nr_alloc); | |
161 | packing_data_unlock(pack); | |
162 | ||
163 | pack->delta_size[e - pack->objects] = size; | |
164 | e->delta_size_valid = 0; | |
165 | } | |
166 | } | |
167 | ||
43fa44fa | 168 | #define IN_PACK(obj) oe_in_pack(&to_pack, obj) |
ac77d0c3 NTND |
169 | #define SIZE(obj) oe_size(&to_pack, obj) |
170 | #define SET_SIZE(obj,size) oe_set_size(&to_pack, obj, size) | |
0aca34e8 | 171 | #define DELTA_SIZE(obj) oe_delta_size(&to_pack, obj) |
898eba5e NTND |
172 | #define DELTA(obj) oe_delta(&to_pack, obj) |
173 | #define DELTA_CHILD(obj) oe_delta_child(&to_pack, obj) | |
174 | #define DELTA_SIBLING(obj) oe_delta_sibling(&to_pack, obj) | |
175 | #define SET_DELTA(obj, val) oe_set_delta(&to_pack, obj, val) | |
6a1e32d5 | 176 | #define SET_DELTA_EXT(obj, oid) oe_set_delta_ext(&to_pack, obj, oid) |
0aca34e8 | 177 | #define SET_DELTA_SIZE(obj, val) oe_set_delta_size(&to_pack, obj, val) |
898eba5e NTND |
178 | #define SET_DELTA_CHILD(obj, val) oe_set_delta_child(&to_pack, obj, val) |
179 | #define SET_DELTA_SIBLING(obj, val) oe_set_delta_sibling(&to_pack, obj, val) | |
43fa44fa | 180 | |
99fb6e04 | 181 | static const char *pack_usage[] = { |
b8c1d275 AH |
182 | N_("git pack-objects --stdout [<options>...] [< <ref-list> | < <object-list>]"), |
183 | N_("git pack-objects [<options>...] <base-name> [< <ref-list> | < <object-list>]"), | |
99fb6e04 NTND |
184 | NULL |
185 | }; | |
c323ac7d | 186 | |
79814f42 | 187 | static struct pack_idx_entry **written_list; |
5af05043 | 188 | static uint32_t nr_result, nr_written, nr_seen; |
6a1e32d5 | 189 | static struct bitmap_index *bitmap_git; |
28b8a730 | 190 | static uint32_t write_layer; |
3f9ac8d2 | 191 | |
96f1e58f | 192 | static int non_empty; |
a7de7130 | 193 | static int reuse_delta = 1, reuse_object = 1; |
e5e9714a | 194 | static int keep_unreachable, unpack_unreachable, include_tag; |
dddbad72 | 195 | static timestamp_t unpack_unreachable_expiration; |
e26a8c47 | 196 | static int pack_loose_unreachable; |
96f1e58f | 197 | static int local; |
56dfeb62 | 198 | static int have_non_local_packs; |
96f1e58f | 199 | static int incremental; |
ed7e5fc3 NTND |
200 | static int ignore_packed_keep_on_disk; |
201 | static int ignore_packed_keep_in_core; | |
be6b1914 | 202 | static int allow_ofs_delta; |
ebcfb379 | 203 | static struct pack_idx_option pack_idx_opts; |
d01fb92f | 204 | static const char *base_name; |
024701f1 | 205 | static int progress = 1; |
4812a93a | 206 | static int window = 10; |
568508e7 | 207 | static unsigned long pack_size_limit; |
618e613a | 208 | static int depth = 50; |
43cc2b42 | 209 | static int delta_search_threads; |
df6d6101 | 210 | static int pack_to_stdout; |
4f6d26b1 | 211 | static int sparse; |
6a1e32d5 | 212 | static int thin; |
8d1d8f83 | 213 | static int num_preferred_base; |
dc6a0757 | 214 | static struct progress *progress_state; |
c323ac7d | 215 | |
6b8fda2d VM |
216 | static struct packed_git *reuse_packfile; |
217 | static uint32_t reuse_packfile_objects; | |
bb514de3 | 218 | static struct bitmap *reuse_packfile_bitmap; |
6b8fda2d | 219 | |
645c432d KS |
220 | static int use_bitmap_index_default = 1; |
221 | static int use_bitmap_index = -1; | |
e704fc79 | 222 | static int allow_pack_reuse = 1; |
25575015 JK |
223 | static enum { |
224 | WRITE_BITMAP_FALSE = 0, | |
225 | WRITE_BITMAP_QUIET, | |
226 | WRITE_BITMAP_TRUE, | |
227 | } write_bitmap_index; | |
d4316604 | 228 | static uint16_t write_bitmap_options = BITMAP_OPT_HASH_CACHE; |
6b8fda2d | 229 | |
0c16cd49 JT |
230 | static int exclude_promisor_objects; |
231 | ||
28b8a730 JK |
232 | static int use_delta_islands; |
233 | ||
074b2eea | 234 | static unsigned long delta_cache_size = 0; |
9806f5a7 | 235 | static unsigned long max_delta_cache_size = DEFAULT_DELTA_CACHE_SIZE; |
e3dfddb3 | 236 | static unsigned long cache_max_small_delta_size = 1000; |
074b2eea | 237 | |
a97773ce BD |
238 | static unsigned long window_memory_limit = 0; |
239 | ||
9535ce73 JH |
240 | static struct list_objects_filter_options filter_options; |
241 | ||
dd4b732d JT |
242 | static struct string_list uri_protocols = STRING_LIST_INIT_NODUP; |
243 | ||
9535ce73 | 244 | enum missing_action { |
0c16cd49 JT |
245 | MA_ERROR = 0, /* fail if any missing objects are encountered */ |
246 | MA_ALLOW_ANY, /* silently allow ALL missing objects */ | |
247 | MA_ALLOW_PROMISOR, /* silently allow all missing PROMISOR objects */ | |
9535ce73 JH |
248 | }; |
249 | static enum missing_action arg_missing_action; | |
250 | static show_object_fn fn_show_object; | |
251 | ||
dd4b732d JT |
252 | struct configured_exclusion { |
253 | struct oidmap_entry e; | |
254 | char *pack_hash_hex; | |
255 | char *uri; | |
256 | }; | |
257 | static struct oidmap configured_exclusions; | |
258 | ||
259 | static struct oidset excluded_by_config; | |
260 | ||
3f9ac8d2 JH |
261 | /* |
262 | * stats | |
263 | */ | |
7cadf491 SP |
264 | static uint32_t written, written_delta; |
265 | static uint32_t reused, reused_delta; | |
3f9ac8d2 | 266 | |
7cc8f971 VM |
267 | /* |
268 | * Indexed commits | |
269 | */ | |
270 | static struct commit **indexed_commits; | |
271 | static unsigned int indexed_commits_nr; | |
272 | static unsigned int indexed_commits_alloc; | |
273 | ||
274 | static void index_commit_for_bitmap(struct commit *commit) | |
275 | { | |
276 | if (indexed_commits_nr >= indexed_commits_alloc) { | |
277 | indexed_commits_alloc = (indexed_commits_alloc + 32) * 2; | |
2756ca43 | 278 | REALLOC_ARRAY(indexed_commits, indexed_commits_alloc); |
7cc8f971 VM |
279 | } |
280 | ||
281 | indexed_commits[indexed_commits_nr++] = commit; | |
282 | } | |
780e6e73 | 283 | |
3613f9b4 | 284 | static void *get_delta(struct object_entry *entry) |
c323ac7d | 285 | { |
3613f9b4 NP |
286 | unsigned long size, base_size, delta_size; |
287 | void *buf, *base_buf, *delta_buf; | |
21666f1a | 288 | enum object_type type; |
c323ac7d | 289 | |
b4f5aca4 | 290 | buf = read_object_file(&entry->idx.oid, &type, &size); |
3613f9b4 | 291 | if (!buf) |
f616db6a | 292 | die(_("unable to read %s"), oid_to_hex(&entry->idx.oid)); |
898eba5e NTND |
293 | base_buf = read_object_file(&DELTA(entry)->idx.oid, &type, |
294 | &base_size); | |
3613f9b4 | 295 | if (!base_buf) |
e6a492b7 | 296 | die("unable to read %s", |
898eba5e | 297 | oid_to_hex(&DELTA(entry)->idx.oid)); |
3613f9b4 | 298 | delta_buf = diff_delta(base_buf, base_size, |
dcde55bc | 299 | buf, size, &delta_size, 0); |
1a07e59c | 300 | /* |
15beaaa3 | 301 | * We successfully computed this delta once but dropped it for |
1a07e59c NTND |
302 | * memory reasons. Something is very wrong if this time we |
303 | * recompute and create a different delta. | |
304 | */ | |
0aca34e8 | 305 | if (!delta_buf || delta_size != DELTA_SIZE(entry)) |
1a07e59c | 306 | BUG("delta size changed"); |
3613f9b4 NP |
307 | free(buf); |
308 | free(base_buf); | |
c323ac7d LT |
309 | return delta_buf; |
310 | } | |
311 | ||
30ebb40a NP |
312 | static unsigned long do_compress(void **pptr, unsigned long size) |
313 | { | |
ef49a7a0 | 314 | git_zstream stream; |
30ebb40a NP |
315 | void *in, *out; |
316 | unsigned long maxsize; | |
317 | ||
55bb5c91 | 318 | git_deflate_init(&stream, pack_compression_level); |
225a6f10 | 319 | maxsize = git_deflate_bound(&stream, size); |
30ebb40a NP |
320 | |
321 | in = *pptr; | |
322 | out = xmalloc(maxsize); | |
323 | *pptr = out; | |
324 | ||
325 | stream.next_in = in; | |
326 | stream.avail_in = size; | |
327 | stream.next_out = out; | |
328 | stream.avail_out = maxsize; | |
55bb5c91 | 329 | while (git_deflate(&stream, Z_FINISH) == Z_OK) |
30ebb40a | 330 | ; /* nothing */ |
55bb5c91 | 331 | git_deflate_end(&stream); |
30ebb40a NP |
332 | |
333 | free(in); | |
334 | return stream.total_out; | |
335 | } | |
336 | ||
98a3beab | 337 | static unsigned long write_large_blob_data(struct git_istream *st, struct hashfile *f, |
188960b4 | 338 | const struct object_id *oid) |
cf2ba13a NTND |
339 | { |
340 | git_zstream stream; | |
341 | unsigned char ibuf[1024 * 16]; | |
342 | unsigned char obuf[1024 * 16]; | |
343 | unsigned long olen = 0; | |
344 | ||
cf2ba13a NTND |
345 | git_deflate_init(&stream, pack_compression_level); |
346 | ||
347 | for (;;) { | |
348 | ssize_t readlen; | |
349 | int zret = Z_OK; | |
350 | readlen = read_istream(st, ibuf, sizeof(ibuf)); | |
351 | if (readlen == -1) | |
188960b4 | 352 | die(_("unable to read %s"), oid_to_hex(oid)); |
cf2ba13a NTND |
353 | |
354 | stream.next_in = ibuf; | |
355 | stream.avail_in = readlen; | |
356 | while ((stream.avail_in || readlen == 0) && | |
357 | (zret == Z_OK || zret == Z_BUF_ERROR)) { | |
358 | stream.next_out = obuf; | |
359 | stream.avail_out = sizeof(obuf); | |
360 | zret = git_deflate(&stream, readlen ? 0 : Z_FINISH); | |
98a3beab | 361 | hashwrite(f, obuf, stream.next_out - obuf); |
cf2ba13a NTND |
362 | olen += stream.next_out - obuf; |
363 | } | |
364 | if (stream.avail_in) | |
365 | die(_("deflate error (%d)"), zret); | |
366 | if (readlen == 0) { | |
367 | if (zret != Z_STREAM_END) | |
368 | die(_("deflate error (%d)"), zret); | |
369 | break; | |
370 | } | |
371 | } | |
372 | git_deflate_end(&stream); | |
373 | return olen; | |
374 | } | |
375 | ||
780e6e73 NP |
376 | /* |
377 | * we are going to reuse the existing object data as is. make | |
378 | * sure it is not corrupt. | |
379 | */ | |
079afb18 SP |
380 | static int check_pack_inflate(struct packed_git *p, |
381 | struct pack_window **w_curs, | |
6777a59f SP |
382 | off_t offset, |
383 | off_t len, | |
079afb18 SP |
384 | unsigned long expect) |
385 | { | |
ef49a7a0 | 386 | git_zstream stream; |
079afb18 SP |
387 | unsigned char fakebuf[4096], *in; |
388 | int st; | |
389 | ||
390 | memset(&stream, 0, sizeof(stream)); | |
39c68542 | 391 | git_inflate_init(&stream); |
079afb18 SP |
392 | do { |
393 | in = use_pack(p, w_curs, offset, &stream.avail_in); | |
394 | stream.next_in = in; | |
395 | stream.next_out = fakebuf; | |
396 | stream.avail_out = sizeof(fakebuf); | |
39c68542 | 397 | st = git_inflate(&stream, Z_FINISH); |
079afb18 SP |
398 | offset += stream.next_in - in; |
399 | } while (st == Z_OK || st == Z_BUF_ERROR); | |
39c68542 | 400 | git_inflate_end(&stream); |
079afb18 SP |
401 | return (st == Z_STREAM_END && |
402 | stream.total_out == expect && | |
403 | stream.total_in == len) ? 0 : -1; | |
404 | } | |
405 | ||
98a3beab | 406 | static void copy_pack_data(struct hashfile *f, |
079afb18 SP |
407 | struct packed_git *p, |
408 | struct pack_window **w_curs, | |
6777a59f SP |
409 | off_t offset, |
410 | off_t len) | |
079afb18 SP |
411 | { |
412 | unsigned char *in; | |
ef49a7a0 | 413 | unsigned long avail; |
079afb18 SP |
414 | |
415 | while (len) { | |
416 | in = use_pack(p, w_curs, offset, &avail); | |
417 | if (avail > len) | |
ef49a7a0 | 418 | avail = (unsigned long)len; |
98a3beab | 419 | hashwrite(f, in, avail); |
079afb18 SP |
420 | offset += avail; |
421 | len -= avail; | |
422 | } | |
423 | } | |
424 | ||
7d089fb9 ÆAB |
425 | static inline int oe_size_greater_than(struct packing_data *pack, |
426 | const struct object_entry *lhs, | |
427 | unsigned long rhs) | |
428 | { | |
429 | if (lhs->size_valid) | |
430 | return lhs->size_ > rhs; | |
431 | if (rhs < pack->oe_size_limit) /* rhs < 2^x <= lhs ? */ | |
432 | return 1; | |
433 | return oe_get_size_slow(pack, lhs) > rhs; | |
434 | } | |
435 | ||
1b4bb16b | 436 | /* Return 0 if we will bust the pack-size limit */ |
98a3beab | 437 | static unsigned long write_no_reuse_object(struct hashfile *f, struct object_entry *entry, |
c9018b03 | 438 | unsigned long limit, int usable_delta) |
c323ac7d | 439 | { |
c9018b03 | 440 | unsigned long size, datalen; |
2c5e2865 JK |
441 | unsigned char header[MAX_PACK_OBJECT_HEADER], |
442 | dheader[MAX_PACK_OBJECT_HEADER]; | |
6777a59f | 443 | unsigned hdrlen; |
2c5ef824 | 444 | enum object_type type; |
c9018b03 | 445 | void *buf; |
cf2ba13a | 446 | struct git_istream *st = NULL; |
41179100 | 447 | const unsigned hashsz = the_hash_algo->rawsz; |
c9018b03 NTND |
448 | |
449 | if (!usable_delta) { | |
fd9b1bae | 450 | if (oe_type(entry) == OBJ_BLOB && |
ac77d0c3 | 451 | oe_size_greater_than(&to_pack, entry, big_file_threshold) && |
c8123e72 MT |
452 | (st = open_istream(the_repository, &entry->idx.oid, &type, |
453 | &size, NULL)) != NULL) | |
cf2ba13a NTND |
454 | buf = NULL; |
455 | else { | |
b4f5aca4 | 456 | buf = read_object_file(&entry->idx.oid, &type, &size); |
cf2ba13a | 457 | if (!buf) |
e6a492b7 | 458 | die(_("unable to read %s"), |
459 | oid_to_hex(&entry->idx.oid)); | |
cf2ba13a | 460 | } |
c9018b03 NTND |
461 | /* |
462 | * make sure no cached delta data remains from a | |
463 | * previous attempt before a pack split occurred. | |
464 | */ | |
6a83d902 | 465 | FREE_AND_NULL(entry->delta_data); |
c9018b03 NTND |
466 | entry->z_delta_size = 0; |
467 | } else if (entry->delta_data) { | |
0aca34e8 | 468 | size = DELTA_SIZE(entry); |
c9018b03 NTND |
469 | buf = entry->delta_data; |
470 | entry->delta_data = NULL; | |
898eba5e | 471 | type = (allow_ofs_delta && DELTA(entry)->idx.offset) ? |
c9018b03 NTND |
472 | OBJ_OFS_DELTA : OBJ_REF_DELTA; |
473 | } else { | |
474 | buf = get_delta(entry); | |
0aca34e8 | 475 | size = DELTA_SIZE(entry); |
898eba5e | 476 | type = (allow_ofs_delta && DELTA(entry)->idx.offset) ? |
c9018b03 NTND |
477 | OBJ_OFS_DELTA : OBJ_REF_DELTA; |
478 | } | |
479 | ||
cf2ba13a NTND |
480 | if (st) /* large blob case, just assume we don't compress well */ |
481 | datalen = size; | |
482 | else if (entry->z_delta_size) | |
c9018b03 NTND |
483 | datalen = entry->z_delta_size; |
484 | else | |
485 | datalen = do_compress(&buf, size); | |
486 | ||
487 | /* | |
488 | * The object header is a byte of 'type' followed by zero or | |
489 | * more bytes of length. | |
490 | */ | |
7202a6fa JK |
491 | hdrlen = encode_in_pack_object_header(header, sizeof(header), |
492 | type, size); | |
c9018b03 NTND |
493 | |
494 | if (type == OBJ_OFS_DELTA) { | |
495 | /* | |
496 | * Deltas with relative base contain an additional | |
497 | * encoding of the relative offset for the delta | |
498 | * base from this object's position in the pack. | |
499 | */ | |
898eba5e | 500 | off_t ofs = entry->idx.offset - DELTA(entry)->idx.offset; |
c9018b03 NTND |
501 | unsigned pos = sizeof(dheader) - 1; |
502 | dheader[pos] = ofs & 127; | |
503 | while (ofs >>= 7) | |
504 | dheader[--pos] = 128 | (--ofs & 127); | |
41179100 | 505 | if (limit && hdrlen + sizeof(dheader) - pos + datalen + hashsz >= limit) { |
cf2ba13a NTND |
506 | if (st) |
507 | close_istream(st); | |
c9018b03 NTND |
508 | free(buf); |
509 | return 0; | |
510 | } | |
98a3beab | 511 | hashwrite(f, header, hdrlen); |
512 | hashwrite(f, dheader + pos, sizeof(dheader) - pos); | |
c9018b03 NTND |
513 | hdrlen += sizeof(dheader) - pos; |
514 | } else if (type == OBJ_REF_DELTA) { | |
515 | /* | |
516 | * Deltas with a base reference contain | |
41179100 | 517 | * additional bytes for the base object ID. |
c9018b03 | 518 | */ |
41179100 | 519 | if (limit && hdrlen + hashsz + datalen + hashsz >= limit) { |
cf2ba13a NTND |
520 | if (st) |
521 | close_istream(st); | |
c9018b03 NTND |
522 | free(buf); |
523 | return 0; | |
524 | } | |
98a3beab | 525 | hashwrite(f, header, hdrlen); |
42c8ce1c | 526 | hashwrite(f, DELTA(entry)->idx.oid.hash, hashsz); |
41179100 | 527 | hdrlen += hashsz; |
c9018b03 | 528 | } else { |
41179100 | 529 | if (limit && hdrlen + datalen + hashsz >= limit) { |
cf2ba13a NTND |
530 | if (st) |
531 | close_istream(st); | |
c9018b03 NTND |
532 | free(buf); |
533 | return 0; | |
534 | } | |
98a3beab | 535 | hashwrite(f, header, hdrlen); |
c9018b03 | 536 | } |
cf2ba13a | 537 | if (st) { |
188960b4 | 538 | datalen = write_large_blob_data(st, f, &entry->idx.oid); |
cf2ba13a NTND |
539 | close_istream(st); |
540 | } else { | |
98a3beab | 541 | hashwrite(f, buf, datalen); |
cf2ba13a NTND |
542 | free(buf); |
543 | } | |
c9018b03 NTND |
544 | |
545 | return hdrlen + datalen; | |
546 | } | |
547 | ||
548 | /* Return 0 if we will bust the pack-size limit */ | |
98a3beab | 549 | static off_t write_reuse_object(struct hashfile *f, struct object_entry *entry, |
af92a645 | 550 | unsigned long limit, int usable_delta) |
c9018b03 | 551 | { |
43fa44fa | 552 | struct packed_git *p = IN_PACK(entry); |
c9018b03 | 553 | struct pack_window *w_curs = NULL; |
952fc687 | 554 | uint32_t pos; |
c9018b03 | 555 | off_t offset; |
fd9b1bae | 556 | enum object_type type = oe_type(entry); |
211c61c6 | 557 | off_t datalen; |
2c5e2865 JK |
558 | unsigned char header[MAX_PACK_OBJECT_HEADER], |
559 | dheader[MAX_PACK_OBJECT_HEADER]; | |
c9018b03 | 560 | unsigned hdrlen; |
41179100 | 561 | const unsigned hashsz = the_hash_algo->rawsz; |
ac77d0c3 | 562 | unsigned long entry_size = SIZE(entry); |
c9018b03 | 563 | |
898eba5e NTND |
564 | if (DELTA(entry)) |
565 | type = (allow_ofs_delta && DELTA(entry)->idx.offset) ? | |
c9018b03 | 566 | OBJ_OFS_DELTA : OBJ_REF_DELTA; |
7202a6fa | 567 | hdrlen = encode_in_pack_object_header(header, sizeof(header), |
ac77d0c3 | 568 | type, entry_size); |
c9018b03 NTND |
569 | |
570 | offset = entry->in_pack_offset; | |
952fc687 TB |
571 | if (offset_to_pack_pos(p, offset, &pos) < 0) |
572 | die(_("write_reuse_object: could not locate %s, expected at " | |
573 | "offset %"PRIuMAX" in pack %s"), | |
574 | oid_to_hex(&entry->idx.oid), (uintmax_t)offset, | |
575 | p->pack_name); | |
576 | datalen = pack_pos_to_offset(p, pos + 1) - offset; | |
c9018b03 | 577 | if (!pack_to_stdout && p->index_version > 1 && |
952fc687 TB |
578 | check_pack_crc(p, &w_curs, offset, datalen, |
579 | pack_pos_to_index(p, pos))) { | |
f616db6a | 580 | error(_("bad packed object CRC for %s"), |
e6a492b7 | 581 | oid_to_hex(&entry->idx.oid)); |
c9018b03 NTND |
582 | unuse_pack(&w_curs); |
583 | return write_no_reuse_object(f, entry, limit, usable_delta); | |
584 | } | |
585 | ||
586 | offset += entry->in_pack_header_size; | |
587 | datalen -= entry->in_pack_header_size; | |
588 | ||
589 | if (!pack_to_stdout && p->index_version == 1 && | |
ac77d0c3 | 590 | check_pack_inflate(p, &w_curs, offset, datalen, entry_size)) { |
f616db6a | 591 | error(_("corrupt packed object for %s"), |
e6a492b7 | 592 | oid_to_hex(&entry->idx.oid)); |
c9018b03 NTND |
593 | unuse_pack(&w_curs); |
594 | return write_no_reuse_object(f, entry, limit, usable_delta); | |
595 | } | |
596 | ||
597 | if (type == OBJ_OFS_DELTA) { | |
898eba5e | 598 | off_t ofs = entry->idx.offset - DELTA(entry)->idx.offset; |
c9018b03 NTND |
599 | unsigned pos = sizeof(dheader) - 1; |
600 | dheader[pos] = ofs & 127; | |
601 | while (ofs >>= 7) | |
602 | dheader[--pos] = 128 | (--ofs & 127); | |
41179100 | 603 | if (limit && hdrlen + sizeof(dheader) - pos + datalen + hashsz >= limit) { |
c9018b03 NTND |
604 | unuse_pack(&w_curs); |
605 | return 0; | |
606 | } | |
98a3beab | 607 | hashwrite(f, header, hdrlen); |
608 | hashwrite(f, dheader + pos, sizeof(dheader) - pos); | |
c9018b03 NTND |
609 | hdrlen += sizeof(dheader) - pos; |
610 | reused_delta++; | |
611 | } else if (type == OBJ_REF_DELTA) { | |
41179100 | 612 | if (limit && hdrlen + hashsz + datalen + hashsz >= limit) { |
c9018b03 NTND |
613 | unuse_pack(&w_curs); |
614 | return 0; | |
615 | } | |
98a3beab | 616 | hashwrite(f, header, hdrlen); |
42c8ce1c | 617 | hashwrite(f, DELTA(entry)->idx.oid.hash, hashsz); |
41179100 | 618 | hdrlen += hashsz; |
c9018b03 NTND |
619 | reused_delta++; |
620 | } else { | |
41179100 | 621 | if (limit && hdrlen + datalen + hashsz >= limit) { |
c9018b03 NTND |
622 | unuse_pack(&w_curs); |
623 | return 0; | |
624 | } | |
98a3beab | 625 | hashwrite(f, header, hdrlen); |
c9018b03 NTND |
626 | } |
627 | copy_pack_data(f, p, &w_curs, offset, datalen); | |
628 | unuse_pack(&w_curs); | |
629 | reused++; | |
630 | return hdrlen + datalen; | |
631 | } | |
632 | ||
633 | /* Return 0 if we will bust the pack-size limit */ | |
98a3beab | 634 | static off_t write_object(struct hashfile *f, |
af92a645 NTND |
635 | struct object_entry *entry, |
636 | off_t write_offset) | |
c9018b03 | 637 | { |
af92a645 NTND |
638 | unsigned long limit; |
639 | off_t len; | |
2c5ef824 | 640 | int usable_delta, to_reuse; |
c323ac7d | 641 | |
78d1e84f NP |
642 | if (!pack_to_stdout) |
643 | crc32_begin(f); | |
644 | ||
a2430dde | 645 | /* apply size limit if limited packsize and not first object */ |
a1e4760f NP |
646 | if (!pack_size_limit || !nr_written) |
647 | limit = 0; | |
648 | else if (pack_size_limit <= write_offset) | |
649 | /* | |
650 | * the earlier object did not fit the limit; avoid | |
651 | * mistaking this with unlimited (i.e. limit = 0). | |
652 | */ | |
653 | limit = 1; | |
654 | else | |
655 | limit = pack_size_limit - write_offset; | |
2c5ef824 | 656 | |
898eba5e | 657 | if (!DELTA(entry)) |
2c5ef824 NP |
658 | usable_delta = 0; /* no delta */ |
659 | else if (!pack_size_limit) | |
660 | usable_delta = 1; /* unlimited packfile */ | |
898eba5e | 661 | else if (DELTA(entry)->idx.offset == (off_t)-1) |
2c5ef824 | 662 | usable_delta = 0; /* base was written to another pack */ |
898eba5e | 663 | else if (DELTA(entry)->idx.offset) |
2c5ef824 NP |
664 | usable_delta = 1; /* base already exists in this pack */ |
665 | else | |
666 | usable_delta = 0; /* base could end up in another pack */ | |
667 | ||
a7de7130 | 668 | if (!reuse_object) |
fa736f72 | 669 | to_reuse = 0; /* explicit */ |
43fa44fa | 670 | else if (!IN_PACK(entry)) |
ab7cd7bb | 671 | to_reuse = 0; /* can't reuse what we don't have */ |
fd9b1bae NTND |
672 | else if (oe_type(entry) == OBJ_REF_DELTA || |
673 | oe_type(entry) == OBJ_OFS_DELTA) | |
17b08f2c DH |
674 | /* check_object() decided it for us ... */ |
675 | to_reuse = usable_delta; | |
676 | /* ... but pack split may override that */ | |
fd9b1bae | 677 | else if (oe_type(entry) != entry->in_pack_type) |
ab7cd7bb | 678 | to_reuse = 0; /* pack has delta which is unusable */ |
898eba5e | 679 | else if (DELTA(entry)) |
ab7cd7bb JH |
680 | to_reuse = 0; /* we want to pack afresh */ |
681 | else | |
682 | to_reuse = 1; /* we have it in-pack undeltified, | |
683 | * and we do not need to deltify it. | |
684 | */ | |
685 | ||
c9018b03 NTND |
686 | if (!to_reuse) |
687 | len = write_no_reuse_object(f, entry, limit, usable_delta); | |
688 | else | |
689 | len = write_reuse_object(f, entry, limit, usable_delta); | |
690 | if (!len) | |
691 | return 0; | |
64bd76b1 | 692 | |
17b08f2c | 693 | if (usable_delta) |
ab7cd7bb | 694 | written_delta++; |
3f9ac8d2 | 695 | written++; |
78d1e84f | 696 | if (!pack_to_stdout) |
aa7e44bf | 697 | entry->idx.crc32 = crc32_end(f); |
c9018b03 | 698 | return len; |
c323ac7d LT |
699 | } |
700 | ||
f63c79db JH |
701 | enum write_one_status { |
702 | WRITE_ONE_SKIP = -1, /* already written */ | |
703 | WRITE_ONE_BREAK = 0, /* writing this will bust the limit; not written */ | |
704 | WRITE_ONE_WRITTEN = 1, /* normal */ | |
705 | WRITE_ONE_RECURSIVE = 2 /* already scheduled to be written */ | |
706 | }; | |
707 | ||
98a3beab | 708 | static enum write_one_status write_one(struct hashfile *f, |
f63c79db JH |
709 | struct object_entry *e, |
710 | off_t *offset) | |
9d5ab962 | 711 | { |
af92a645 | 712 | off_t size; |
f63c79db | 713 | int recursing; |
d7dd0223 | 714 | |
f63c79db JH |
715 | /* |
716 | * we set offset to 1 (which is an impossible value) to mark | |
717 | * the fact that this object is involved in "write its base | |
718 | * first before writing a deltified object" recursion. | |
719 | */ | |
720 | recursing = (e->idx.offset == 1); | |
721 | if (recursing) { | |
f616db6a | 722 | warning(_("recursive delta detected for object %s"), |
e6a492b7 | 723 | oid_to_hex(&e->idx.oid)); |
f63c79db JH |
724 | return WRITE_ONE_RECURSIVE; |
725 | } else if (e->idx.offset || e->preferred_base) { | |
726 | /* offset is non zero if object is written already. */ | |
727 | return WRITE_ONE_SKIP; | |
728 | } | |
d7dd0223 | 729 | |
720c9f7b | 730 | /* if we are deltified, write out base object first. */ |
898eba5e | 731 | if (DELTA(e)) { |
f63c79db | 732 | e->idx.offset = 1; /* now recurse */ |
898eba5e | 733 | switch (write_one(f, DELTA(e), offset)) { |
f63c79db JH |
734 | case WRITE_ONE_RECURSIVE: |
735 | /* we cannot depend on this one */ | |
898eba5e | 736 | SET_DELTA(e, NULL); |
f63c79db JH |
737 | break; |
738 | default: | |
739 | break; | |
740 | case WRITE_ONE_BREAK: | |
741 | e->idx.offset = recursing; | |
742 | return WRITE_ONE_BREAK; | |
743 | } | |
744 | } | |
d7dd0223 | 745 | |
6ed7f25e NP |
746 | e->idx.offset = *offset; |
747 | size = write_object(f, e, *offset); | |
17b08f2c | 748 | if (!size) { |
f63c79db JH |
749 | e->idx.offset = recursing; |
750 | return WRITE_ONE_BREAK; | |
17b08f2c | 751 | } |
79814f42 | 752 | written_list[nr_written++] = &e->idx; |
d7dd0223 NP |
753 | |
754 | /* make sure off_t is sufficiently large not to wrap */ | |
c03c8315 | 755 | if (signed_add_overflows(*offset, size)) |
f616db6a | 756 | die(_("pack too large for current definition of off_t")); |
6ed7f25e | 757 | *offset += size; |
f63c79db | 758 | return WRITE_ONE_WRITTEN; |
9d5ab962 JH |
759 | } |
760 | ||
d155254c | 761 | static int mark_tagged(const char *path, const struct object_id *oid, int flag, |
1b4bb16b JH |
762 | void *cb_data) |
763 | { | |
188960b4 | 764 | struct object_id peeled; |
3a37876b | 765 | struct object_entry *entry = packlist_find(&to_pack, oid); |
1b4bb16b JH |
766 | |
767 | if (entry) | |
768 | entry->tagged = 1; | |
36a31792 | 769 | if (!peel_iterated_oid(oid, &peeled)) { |
3a37876b | 770 | entry = packlist_find(&to_pack, &peeled); |
1b4bb16b JH |
771 | if (entry) |
772 | entry->tagged = 1; | |
773 | } | |
774 | return 0; | |
775 | } | |
776 | ||
7d089fb9 ÆAB |
777 | static inline unsigned char oe_layer(struct packing_data *pack, |
778 | struct object_entry *e) | |
779 | { | |
780 | if (!pack->layer) | |
781 | return 0; | |
782 | return pack->layer[e - pack->objects]; | |
783 | } | |
784 | ||
be126818 | 785 | static inline void add_to_write_order(struct object_entry **wo, |
92bef1a1 | 786 | unsigned int *endp, |
1b4bb16b JH |
787 | struct object_entry *e) |
788 | { | |
fe0ac2fb | 789 | if (e->filled || oe_layer(&to_pack, e) != write_layer) |
1b4bb16b JH |
790 | return; |
791 | wo[(*endp)++] = e; | |
792 | e->filled = 1; | |
793 | } | |
794 | ||
795 | static void add_descendants_to_write_order(struct object_entry **wo, | |
92bef1a1 | 796 | unsigned int *endp, |
1b4bb16b JH |
797 | struct object_entry *e) |
798 | { | |
f380872f DM |
799 | int add_to_order = 1; |
800 | while (e) { | |
801 | if (add_to_order) { | |
802 | struct object_entry *s; | |
803 | /* add this node... */ | |
804 | add_to_write_order(wo, endp, e); | |
805 | /* all its siblings... */ | |
898eba5e | 806 | for (s = DELTA_SIBLING(e); s; s = DELTA_SIBLING(s)) { |
f380872f DM |
807 | add_to_write_order(wo, endp, s); |
808 | } | |
809 | } | |
810 | /* drop down a level to add left subtree nodes if possible */ | |
898eba5e | 811 | if (DELTA_CHILD(e)) { |
f380872f | 812 | add_to_order = 1; |
898eba5e | 813 | e = DELTA_CHILD(e); |
f380872f DM |
814 | } else { |
815 | add_to_order = 0; | |
816 | /* our sibling might have some children, it is next */ | |
898eba5e NTND |
817 | if (DELTA_SIBLING(e)) { |
818 | e = DELTA_SIBLING(e); | |
f380872f DM |
819 | continue; |
820 | } | |
821 | /* go back to our parent node */ | |
898eba5e NTND |
822 | e = DELTA(e); |
823 | while (e && !DELTA_SIBLING(e)) { | |
f380872f DM |
824 | /* we're on the right side of a subtree, keep |
825 | * going up until we can go right again */ | |
898eba5e | 826 | e = DELTA(e); |
f380872f DM |
827 | } |
828 | if (!e) { | |
829 | /* done- we hit our original root node */ | |
830 | return; | |
831 | } | |
832 | /* pass it off to sibling at this level */ | |
898eba5e | 833 | e = DELTA_SIBLING(e); |
f380872f DM |
834 | } |
835 | }; | |
1b4bb16b JH |
836 | } |
837 | ||
838 | static void add_family_to_write_order(struct object_entry **wo, | |
92bef1a1 | 839 | unsigned int *endp, |
1b4bb16b JH |
840 | struct object_entry *e) |
841 | { | |
842 | struct object_entry *root; | |
843 | ||
898eba5e | 844 | for (root = e; DELTA(root); root = DELTA(root)) |
1b4bb16b | 845 | ; /* nothing */ |
1b4bb16b JH |
846 | add_descendants_to_write_order(wo, endp, root); |
847 | } | |
848 | ||
f64ba53e | 849 | static void compute_layer_order(struct object_entry **wo, unsigned int *wo_end) |
1b4bb16b | 850 | { |
f64ba53e | 851 | unsigned int i, last_untagged; |
2834bc27 | 852 | struct object_entry *objects = to_pack.objects; |
1b4bb16b | 853 | |
2834bc27 | 854 | for (i = 0; i < to_pack.nr_objects; i++) { |
1b4bb16b JH |
855 | if (objects[i].tagged) |
856 | break; | |
f64ba53e | 857 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b | 858 | } |
38d4debb | 859 | last_untagged = i; |
1b4bb16b JH |
860 | |
861 | /* | |
862 | * Then fill all the tagged tips. | |
863 | */ | |
2834bc27 | 864 | for (; i < to_pack.nr_objects; i++) { |
1b4bb16b | 865 | if (objects[i].tagged) |
f64ba53e | 866 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b JH |
867 | } |
868 | ||
869 | /* | |
870 | * And then all remaining commits and tags. | |
871 | */ | |
2834bc27 | 872 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
fd9b1bae NTND |
873 | if (oe_type(&objects[i]) != OBJ_COMMIT && |
874 | oe_type(&objects[i]) != OBJ_TAG) | |
1b4bb16b | 875 | continue; |
f64ba53e | 876 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b JH |
877 | } |
878 | ||
879 | /* | |
880 | * And then all the trees. | |
881 | */ | |
2834bc27 | 882 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
fd9b1bae | 883 | if (oe_type(&objects[i]) != OBJ_TREE) |
1b4bb16b | 884 | continue; |
f64ba53e | 885 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b JH |
886 | } |
887 | ||
888 | /* | |
889 | * Finally all the rest in really tight order | |
890 | */ | |
2834bc27 | 891 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
fe0ac2fb | 892 | if (!objects[i].filled && oe_layer(&to_pack, &objects[i]) == write_layer) |
f64ba53e | 893 | add_family_to_write_order(wo, wo_end, &objects[i]); |
38d4debb | 894 | } |
f64ba53e CC |
895 | } |
896 | ||
897 | static struct object_entry **compute_write_order(void) | |
898 | { | |
28b8a730 | 899 | uint32_t max_layers = 1; |
f64ba53e CC |
900 | unsigned int i, wo_end; |
901 | ||
902 | struct object_entry **wo; | |
903 | struct object_entry *objects = to_pack.objects; | |
904 | ||
905 | for (i = 0; i < to_pack.nr_objects; i++) { | |
906 | objects[i].tagged = 0; | |
907 | objects[i].filled = 0; | |
908 | SET_DELTA_CHILD(&objects[i], NULL); | |
909 | SET_DELTA_SIBLING(&objects[i], NULL); | |
910 | } | |
911 | ||
912 | /* | |
913 | * Fully connect delta_child/delta_sibling network. | |
914 | * Make sure delta_sibling is sorted in the original | |
915 | * recency order. | |
916 | */ | |
917 | for (i = to_pack.nr_objects; i > 0;) { | |
918 | struct object_entry *e = &objects[--i]; | |
919 | if (!DELTA(e)) | |
920 | continue; | |
921 | /* Mark me as the first child */ | |
922 | e->delta_sibling_idx = DELTA(e)->delta_child_idx; | |
923 | SET_DELTA_CHILD(DELTA(e), e); | |
38d4debb DM |
924 | } |
925 | ||
f64ba53e CC |
926 | /* |
927 | * Mark objects that are at the tip of tags. | |
928 | */ | |
929 | for_each_tag_ref(mark_tagged, NULL); | |
930 | ||
28b8a730 JK |
931 | if (use_delta_islands) |
932 | max_layers = compute_pack_layers(&to_pack); | |
933 | ||
f64ba53e CC |
934 | ALLOC_ARRAY(wo, to_pack.nr_objects); |
935 | wo_end = 0; | |
936 | ||
28b8a730 JK |
937 | for (; write_layer < max_layers; ++write_layer) |
938 | compute_layer_order(wo, &wo_end); | |
38d4debb | 939 | |
2834bc27 | 940 | if (wo_end != to_pack.nr_objects) |
f616db6a NTND |
941 | die(_("ordered %u objects, expected %"PRIu32), |
942 | wo_end, to_pack.nr_objects); | |
1b4bb16b JH |
943 | |
944 | return wo; | |
945 | } | |
946 | ||
bb514de3 JK |
947 | |
948 | /* | |
949 | * A reused set of objects. All objects in a chunk have the same | |
950 | * relative position in the original packfile and the generated | |
951 | * packfile. | |
952 | */ | |
953 | ||
954 | static struct reused_chunk { | |
955 | /* The offset of the first object of this chunk in the original | |
956 | * packfile. */ | |
957 | off_t original; | |
bf12013f HX |
958 | /* The difference for "original" minus the offset of the first object of |
959 | * this chunk in the generated packfile. */ | |
bb514de3 JK |
960 | off_t difference; |
961 | } *reused_chunks; | |
962 | static int reused_chunks_nr; | |
963 | static int reused_chunks_alloc; | |
964 | ||
965 | static void record_reused_object(off_t where, off_t offset) | |
966 | { | |
967 | if (reused_chunks_nr && reused_chunks[reused_chunks_nr-1].difference == offset) | |
968 | return; | |
969 | ||
970 | ALLOC_GROW(reused_chunks, reused_chunks_nr + 1, | |
971 | reused_chunks_alloc); | |
972 | reused_chunks[reused_chunks_nr].original = where; | |
973 | reused_chunks[reused_chunks_nr].difference = offset; | |
974 | reused_chunks_nr++; | |
975 | } | |
976 | ||
977 | /* | |
978 | * Binary search to find the chunk that "where" is in. Note | |
979 | * that we're not looking for an exact match, just the first | |
980 | * chunk that contains it (which implicitly ends at the start | |
981 | * of the next chunk. | |
982 | */ | |
983 | static off_t find_reused_offset(off_t where) | |
6b8fda2d | 984 | { |
bb514de3 JK |
985 | int lo = 0, hi = reused_chunks_nr; |
986 | while (lo < hi) { | |
987 | int mi = lo + ((hi - lo) / 2); | |
988 | if (where == reused_chunks[mi].original) | |
989 | return reused_chunks[mi].difference; | |
990 | if (where < reused_chunks[mi].original) | |
991 | hi = mi; | |
992 | else | |
993 | lo = mi + 1; | |
994 | } | |
6b8fda2d | 995 | |
bb514de3 JK |
996 | /* |
997 | * The first chunk starts at zero, so we can't have gone below | |
998 | * there. | |
999 | */ | |
1000 | assert(lo); | |
1001 | return reused_chunks[lo-1].difference; | |
1002 | } | |
6b8fda2d | 1003 | |
bb514de3 JK |
1004 | static void write_reused_pack_one(size_t pos, struct hashfile *out, |
1005 | struct pack_window **w_curs) | |
6b8fda2d | 1006 | { |
bb514de3 JK |
1007 | off_t offset, next, cur; |
1008 | enum object_type type; | |
1009 | unsigned long size; | |
6b8fda2d | 1010 | |
66cbd3e2 TB |
1011 | offset = pack_pos_to_offset(reuse_packfile, pos); |
1012 | next = pack_pos_to_offset(reuse_packfile, pos + 1); | |
6b8fda2d | 1013 | |
bb514de3 | 1014 | record_reused_object(offset, offset - hashfile_total(out)); |
6b8fda2d | 1015 | |
bb514de3 JK |
1016 | cur = offset; |
1017 | type = unpack_object_header(reuse_packfile, w_curs, &cur, &size); | |
1018 | assert(type >= 0); | |
6b8fda2d | 1019 | |
bb514de3 JK |
1020 | if (type == OBJ_OFS_DELTA) { |
1021 | off_t base_offset; | |
1022 | off_t fixup; | |
1023 | ||
1024 | unsigned char header[MAX_PACK_OBJECT_HEADER]; | |
1025 | unsigned len; | |
1026 | ||
1027 | base_offset = get_delta_base(reuse_packfile, w_curs, &cur, type, offset); | |
1028 | assert(base_offset != 0); | |
1029 | ||
1030 | /* Convert to REF_DELTA if we must... */ | |
1031 | if (!allow_ofs_delta) { | |
66cbd3e2 | 1032 | uint32_t base_pos; |
f66d4e02 JK |
1033 | struct object_id base_oid; |
1034 | ||
66cbd3e2 TB |
1035 | if (offset_to_pack_pos(reuse_packfile, base_offset, &base_pos) < 0) |
1036 | die(_("expected object at offset %"PRIuMAX" " | |
1037 | "in pack %s"), | |
1038 | (uintmax_t)base_offset, | |
1039 | reuse_packfile->pack_name); | |
1040 | ||
f66d4e02 | 1041 | nth_packed_object_id(&base_oid, reuse_packfile, |
66cbd3e2 | 1042 | pack_pos_to_index(reuse_packfile, base_pos)); |
bb514de3 JK |
1043 | |
1044 | len = encode_in_pack_object_header(header, sizeof(header), | |
1045 | OBJ_REF_DELTA, size); | |
1046 | hashwrite(out, header, len); | |
f8cb64e3 | 1047 | hashwrite(out, base_oid.hash, the_hash_algo->rawsz); |
bb514de3 JK |
1048 | copy_pack_data(out, reuse_packfile, w_curs, cur, next - cur); |
1049 | return; | |
1050 | } | |
6b8fda2d | 1051 | |
bb514de3 JK |
1052 | /* Otherwise see if we need to rewrite the offset... */ |
1053 | fixup = find_reused_offset(offset) - | |
1054 | find_reused_offset(base_offset); | |
1055 | if (fixup) { | |
1056 | unsigned char ofs_header[10]; | |
1057 | unsigned i, ofs_len; | |
1058 | off_t ofs = offset - base_offset - fixup; | |
6b8fda2d | 1059 | |
bb514de3 JK |
1060 | len = encode_in_pack_object_header(header, sizeof(header), |
1061 | OBJ_OFS_DELTA, size); | |
6b8fda2d | 1062 | |
bb514de3 JK |
1063 | i = sizeof(ofs_header) - 1; |
1064 | ofs_header[i] = ofs & 127; | |
1065 | while (ofs >>= 7) | |
1066 | ofs_header[--i] = 128 | (--ofs & 127); | |
6b8fda2d | 1067 | |
bb514de3 | 1068 | ofs_len = sizeof(ofs_header) - i; |
6b8fda2d | 1069 | |
bb514de3 JK |
1070 | hashwrite(out, header, len); |
1071 | hashwrite(out, ofs_header + sizeof(ofs_header) - ofs_len, ofs_len); | |
1072 | copy_pack_data(out, reuse_packfile, w_curs, cur, next - cur); | |
1073 | return; | |
1074 | } | |
1075 | ||
1076 | /* ...otherwise we have no fixup, and can write it verbatim */ | |
1077 | } | |
1078 | ||
1079 | copy_pack_data(out, reuse_packfile, w_curs, offset, next - offset); | |
1080 | } | |
1081 | ||
1082 | static size_t write_reused_pack_verbatim(struct hashfile *out, | |
1083 | struct pack_window **w_curs) | |
1084 | { | |
1085 | size_t pos = 0; | |
1086 | ||
1087 | while (pos < reuse_packfile_bitmap->word_alloc && | |
1088 | reuse_packfile_bitmap->words[pos] == (eword_t)~0) | |
1089 | pos++; | |
1090 | ||
1091 | if (pos) { | |
1092 | off_t to_write; | |
1093 | ||
1094 | written = (pos * BITS_IN_EWORD); | |
6a5c10c4 | 1095 | to_write = pack_pos_to_offset(reuse_packfile, written) |
bb514de3 JK |
1096 | - sizeof(struct pack_header); |
1097 | ||
1098 | /* We're recording one chunk, not one object. */ | |
1099 | record_reused_object(sizeof(struct pack_header), 0); | |
1100 | hashflush(out); | |
1101 | copy_pack_data(out, reuse_packfile, w_curs, | |
1102 | sizeof(struct pack_header), to_write); | |
657673f1 | 1103 | |
657673f1 | 1104 | display_progress(progress_state, written); |
6b8fda2d | 1105 | } |
bb514de3 JK |
1106 | return pos; |
1107 | } | |
1108 | ||
1109 | static void write_reused_pack(struct hashfile *f) | |
1110 | { | |
1111 | size_t i = 0; | |
1112 | uint32_t offset; | |
1113 | struct pack_window *w_curs = NULL; | |
1114 | ||
1115 | if (allow_ofs_delta) | |
1116 | i = write_reused_pack_verbatim(f, &w_curs); | |
1117 | ||
1118 | for (; i < reuse_packfile_bitmap->word_alloc; ++i) { | |
1119 | eword_t word = reuse_packfile_bitmap->words[i]; | |
1120 | size_t pos = (i * BITS_IN_EWORD); | |
6b8fda2d | 1121 | |
bb514de3 JK |
1122 | for (offset = 0; offset < BITS_IN_EWORD; ++offset) { |
1123 | if ((word >> offset) == 0) | |
1124 | break; | |
1125 | ||
1126 | offset += ewah_bit_ctz64(word >> offset); | |
1127 | write_reused_pack_one(pos + offset, f, &w_curs); | |
1128 | display_progress(progress_state, ++written); | |
1129 | } | |
1130 | } | |
6b8fda2d | 1131 | |
bb514de3 | 1132 | unuse_pack(&w_curs); |
6b8fda2d VM |
1133 | } |
1134 | ||
dd4b732d JT |
1135 | static void write_excluded_by_configs(void) |
1136 | { | |
1137 | struct oidset_iter iter; | |
1138 | const struct object_id *oid; | |
1139 | ||
1140 | oidset_iter_init(&excluded_by_config, &iter); | |
1141 | while ((oid = oidset_iter_next(&iter))) { | |
1142 | struct configured_exclusion *ex = | |
1143 | oidmap_get(&configured_exclusions, oid); | |
1144 | ||
1145 | if (!ex) | |
1146 | BUG("configured exclusion wasn't configured"); | |
1147 | write_in_full(1, ex->pack_hash_hex, strlen(ex->pack_hash_hex)); | |
1148 | write_in_full(1, " ", 1); | |
1149 | write_in_full(1, ex->uri, strlen(ex->uri)); | |
1150 | write_in_full(1, "\n", 1); | |
1151 | } | |
1152 | } | |
1153 | ||
9cea46cd EW |
1154 | static const char no_split_warning[] = N_( |
1155 | "disabling bitmap writing, packs are split due to pack.packSizeLimit" | |
1156 | ); | |
1157 | ||
d01fb92f | 1158 | static void write_pack_file(void) |
c323ac7d | 1159 | { |
ebe27b13 | 1160 | uint32_t i = 0, j; |
98a3beab | 1161 | struct hashfile *f; |
6ed7f25e | 1162 | off_t offset; |
ebe27b13 | 1163 | uint32_t nr_remaining = nr_result; |
f746bae8 | 1164 | time_t last_mtime = 0; |
1b4bb16b | 1165 | struct object_entry **write_order; |
81a216a5 | 1166 | |
bcd7954e | 1167 | if (progress > pack_to_stdout) |
754dbc43 | 1168 | progress_state = start_progress(_("Writing objects"), nr_result); |
b32fa95f | 1169 | ALLOC_ARRAY(written_list, to_pack.nr_objects); |
1b4bb16b | 1170 | write_order = compute_write_order(); |
183bdb2c | 1171 | |
ebe27b13 | 1172 | do { |
71b7672b | 1173 | unsigned char hash[GIT_MAX_RAWSZ]; |
7ba502c4 | 1174 | char *pack_tmp_name = NULL; |
aa7e44bf | 1175 | |
cdf9db3c | 1176 | if (pack_to_stdout) |
98a3beab | 1177 | f = hashfd_throughput(1, "<stdout>", progress_state); |
cdf9db3c JH |
1178 | else |
1179 | f = create_tmp_packfile(&pack_tmp_name); | |
ebe27b13 | 1180 | |
c0ad4657 | 1181 | offset = write_pack_header(f, nr_remaining); |
6b8fda2d VM |
1182 | |
1183 | if (reuse_packfile) { | |
6b8fda2d | 1184 | assert(pack_to_stdout); |
bb514de3 JK |
1185 | write_reused_pack(f); |
1186 | offset = hashfile_total(f); | |
6b8fda2d VM |
1187 | } |
1188 | ||
ebe27b13 | 1189 | nr_written = 0; |
2834bc27 | 1190 | for (; i < to_pack.nr_objects; i++) { |
1b4bb16b | 1191 | struct object_entry *e = write_order[i]; |
cddec4f8 | 1192 | if (write_one(f, e, &offset) == WRITE_ONE_BREAK) |
720c9f7b NP |
1193 | break; |
1194 | display_progress(progress_state, written); | |
1195 | } | |
c553ca25 | 1196 | |
ebe27b13 DH |
1197 | /* |
1198 | * Did we write the wrong # entries in the header? | |
1199 | * If so, rewrite it like in fast-import | |
1200 | */ | |
54352bb2 | 1201 | if (pack_to_stdout) { |
71b7672b | 1202 | finalize_hashfile(f, hash, CSUM_HASH_IN_STREAM | CSUM_CLOSE); |
54352bb2 | 1203 | } else if (nr_written == nr_remaining) { |
71b7672b | 1204 | finalize_hashfile(f, hash, CSUM_HASH_IN_STREAM | CSUM_FSYNC | CSUM_CLOSE); |
ebe27b13 | 1205 | } else { |
71b7672b | 1206 | int fd = finalize_hashfile(f, hash, 0); |
1207 | fixup_pack_header_footer(fd, hash, pack_tmp_name, | |
1208 | nr_written, hash, offset); | |
7ba502c4 | 1209 | close(fd); |
9cea46cd | 1210 | if (write_bitmap_index) { |
25575015 JK |
1211 | if (write_bitmap_index != WRITE_BITMAP_QUIET) |
1212 | warning(_(no_split_warning)); | |
9cea46cd EW |
1213 | write_bitmap_index = 0; |
1214 | } | |
ebe27b13 DH |
1215 | } |
1216 | ||
1217 | if (!pack_to_stdout) { | |
f746bae8 | 1218 | struct stat st; |
58892711 | 1219 | struct strbuf tmpname = STRBUF_INIT; |
d01fb92f | 1220 | |
f746bae8 NP |
1221 | /* |
1222 | * Packs are runtime accessed in their mtime | |
1223 | * order since newer packs are more likely to contain | |
1224 | * younger objects. So if we are creating multiple | |
1225 | * packs then we should modify the mtime of later ones | |
1226 | * to preserve this property. | |
1227 | */ | |
0e990530 | 1228 | if (stat(pack_tmp_name, &st) < 0) { |
f616db6a | 1229 | warning_errno(_("failed to stat %s"), pack_tmp_name); |
f746bae8 NP |
1230 | } else if (!last_mtime) { |
1231 | last_mtime = st.st_mtime; | |
1232 | } else { | |
1233 | struct utimbuf utb; | |
1234 | utb.actime = st.st_atime; | |
1235 | utb.modtime = --last_mtime; | |
0e990530 | 1236 | if (utime(pack_tmp_name, &utb) < 0) |
f616db6a | 1237 | warning_errno(_("failed utime() on %s"), pack_tmp_name); |
f746bae8 NP |
1238 | } |
1239 | ||
58892711 | 1240 | strbuf_addf(&tmpname, "%s-", base_name); |
7cc8f971 VM |
1241 | |
1242 | if (write_bitmap_index) { | |
71b7672b | 1243 | bitmap_writer_set_checksum(hash); |
06af3bba NTND |
1244 | bitmap_writer_build_type_index( |
1245 | &to_pack, written_list, nr_written); | |
7cc8f971 VM |
1246 | } |
1247 | ||
58892711 | 1248 | finish_tmp_packfile(&tmpname, pack_tmp_name, |
0e990530 | 1249 | written_list, nr_written, |
71b7672b | 1250 | &pack_idx_opts, hash); |
7cc8f971 VM |
1251 | |
1252 | if (write_bitmap_index) { | |
71b7672b | 1253 | strbuf_addf(&tmpname, "%s.bitmap", hash_to_hex(hash)); |
7cc8f971 VM |
1254 | |
1255 | stop_progress(&progress_state); | |
1256 | ||
1257 | bitmap_writer_show_progress(progress); | |
7cc8f971 VM |
1258 | bitmap_writer_select_commits(indexed_commits, indexed_commits_nr, -1); |
1259 | bitmap_writer_build(&to_pack); | |
ae4f07fb | 1260 | bitmap_writer_finish(written_list, nr_written, |
58892711 | 1261 | tmpname.buf, write_bitmap_options); |
7cc8f971 VM |
1262 | write_bitmap_index = 0; |
1263 | } | |
1264 | ||
58892711 | 1265 | strbuf_release(&tmpname); |
7ba502c4 | 1266 | free(pack_tmp_name); |
71b7672b | 1267 | puts(hash_to_hex(hash)); |
ebe27b13 DH |
1268 | } |
1269 | ||
1270 | /* mark written objects as written to previous pack */ | |
1271 | for (j = 0; j < nr_written; j++) { | |
79814f42 | 1272 | written_list[j]->offset = (off_t)-1; |
ebe27b13 DH |
1273 | } |
1274 | nr_remaining -= nr_written; | |
2834bc27 | 1275 | } while (nr_remaining && i < to_pack.nr_objects); |
ebe27b13 DH |
1276 | |
1277 | free(written_list); | |
1b4bb16b | 1278 | free(write_order); |
4d4fcc54 | 1279 | stop_progress(&progress_state); |
67c08ce1 | 1280 | if (written != nr_result) |
f616db6a NTND |
1281 | die(_("wrote %"PRIu32" objects while expecting %"PRIu32), |
1282 | written, nr_result); | |
9ed87902 JT |
1283 | trace2_data_intmax("pack-objects", the_repository, |
1284 | "write_pack_file/wrote", nr_result); | |
c323ac7d LT |
1285 | } |
1286 | ||
a74db82e JH |
1287 | static int no_try_delta(const char *path) |
1288 | { | |
2aef63d3 | 1289 | static struct attr_check *check; |
a74db82e | 1290 | |
2aef63d3 JH |
1291 | if (!check) |
1292 | check = attr_check_initl("delta", NULL); | |
f8adbec9 | 1293 | git_check_attr(the_repository->index, path, check); |
2aef63d3 | 1294 | if (ATTR_FALSE(check->items[0].value)) |
a74db82e JH |
1295 | return 1; |
1296 | return 0; | |
1297 | } | |
1298 | ||
ce2bc424 JK |
1299 | /* |
1300 | * When adding an object, check whether we have already added it | |
1301 | * to our packing list. If so, we can skip. However, if we are | |
1302 | * being asked to excludei t, but the previous mention was to include | |
1303 | * it, make sure to adjust its flags and tweak our numbers accordingly. | |
1304 | * | |
1305 | * As an optimization, we pass out the index position where we would have | |
1306 | * found the item, since that saves us from having to look it up again a | |
1307 | * few lines later when we want to add the new entry. | |
1308 | */ | |
188960b4 | 1309 | static int have_duplicate_entry(const struct object_id *oid, |
3a37876b | 1310 | int exclude) |
c323ac7d | 1311 | { |
c323ac7d | 1312 | struct object_entry *entry; |
29b734e4 | 1313 | |
92fb0db9 JK |
1314 | if (reuse_packfile_bitmap && |
1315 | bitmap_walk_contains(bitmap_git, reuse_packfile_bitmap, oid)) | |
1316 | return 1; | |
1317 | ||
3a37876b | 1318 | entry = packlist_find(&to_pack, oid); |
ce2bc424 | 1319 | if (!entry) |
29b734e4 | 1320 | return 0; |
ce2bc424 JK |
1321 | |
1322 | if (exclude) { | |
1323 | if (!entry->preferred_base) | |
1324 | nr_result--; | |
1325 | entry->preferred_base = 1; | |
29b734e4 | 1326 | } |
c323ac7d | 1327 | |
ce2bc424 JK |
1328 | return 1; |
1329 | } | |
1330 | ||
6325da14 JK |
1331 | static int want_found_object(const struct object_id *oid, int exclude, |
1332 | struct packed_git *p) | |
702d1b95 KS |
1333 | { |
1334 | if (exclude) | |
1335 | return 1; | |
1336 | if (incremental) | |
1337 | return 0; | |
1338 | ||
1339 | /* | |
1340 | * When asked to do --local (do not include an object that appears in a | |
1341 | * pack we borrow from elsewhere) or --honor-pack-keep (do not include | |
1342 | * an object that appears in a pack marked with .keep), finding a pack | |
1343 | * that matches the criteria is sufficient for us to decide to omit it. | |
1344 | * However, even if this pack does not satisfy the criteria, we need to | |
1345 | * make sure no copy of this object appears in _any_ pack that makes us | |
1346 | * to omit the object, so we need to check all the packs. | |
1347 | * | |
6325da14 | 1348 | * We can however first check whether these options can possibly matter; |
702d1b95 KS |
1349 | * if they do not matter we know we want the object in generated pack. |
1350 | * Otherwise, we signal "-1" at the end to tell the caller that we do | |
1351 | * not know either way, and it needs to check more packs. | |
1352 | */ | |
702d1b95 | 1353 | |
6325da14 JK |
1354 | /* |
1355 | * Objects in packs borrowed from elsewhere are discarded regardless of | |
1356 | * if they appear in other packs that weren't borrowed. | |
1357 | */ | |
702d1b95 KS |
1358 | if (local && !p->pack_local) |
1359 | return 0; | |
6325da14 JK |
1360 | |
1361 | /* | |
1362 | * Then handle .keep first, as we have a fast(er) path there. | |
1363 | */ | |
1364 | if (ignore_packed_keep_on_disk || ignore_packed_keep_in_core) { | |
1365 | /* | |
1366 | * Set the flags for the kept-pack cache to be the ones we want | |
1367 | * to ignore. | |
1368 | * | |
1369 | * That is, if we are ignoring objects in on-disk keep packs, | |
1370 | * then we want to search through the on-disk keep and ignore | |
1371 | * the in-core ones. | |
1372 | */ | |
1373 | unsigned flags = 0; | |
1374 | if (ignore_packed_keep_on_disk) | |
1375 | flags |= ON_DISK_KEEP_PACKS; | |
1376 | if (ignore_packed_keep_in_core) | |
1377 | flags |= IN_CORE_KEEP_PACKS; | |
1378 | ||
1379 | if (ignore_packed_keep_on_disk && p->pack_keep) | |
1380 | return 0; | |
1381 | if (ignore_packed_keep_in_core && p->pack_keep_in_core) | |
1382 | return 0; | |
1383 | if (has_object_kept_pack(oid, flags)) | |
1384 | return 0; | |
1385 | } | |
1386 | ||
1387 | /* | |
1388 | * At this point we know definitively that either we don't care about | |
1389 | * keep-packs, or the object is not in one. Keep checking other | |
1390 | * conditions... | |
1391 | */ | |
1392 | if (!local || !have_non_local_packs) | |
1393 | return 1; | |
702d1b95 KS |
1394 | |
1395 | /* we don't know yet; keep looking for more packs */ | |
1396 | return -1; | |
1397 | } | |
1398 | ||
6325da14 JK |
1399 | static int want_object_in_pack_one(struct packed_git *p, |
1400 | const struct object_id *oid, | |
1401 | int exclude, | |
1402 | struct packed_git **found_pack, | |
1403 | off_t *found_offset) | |
1404 | { | |
1405 | off_t offset; | |
1406 | ||
1407 | if (p == *found_pack) | |
1408 | offset = *found_offset; | |
1409 | else | |
1410 | offset = find_pack_entry_one(oid->hash, p); | |
1411 | ||
1412 | if (offset) { | |
1413 | if (!*found_pack) { | |
1414 | if (!is_pack_valid(p)) | |
1415 | return -1; | |
1416 | *found_offset = offset; | |
1417 | *found_pack = p; | |
1418 | } | |
1419 | return want_found_object(oid, exclude, p); | |
1420 | } | |
1421 | return -1; | |
1422 | } | |
1423 | ||
ce2bc424 JK |
1424 | /* |
1425 | * Check whether we want the object in the pack (e.g., we do not want | |
1426 | * objects found in non-local stores if the "--local" option was used). | |
1427 | * | |
702d1b95 KS |
1428 | * If the caller already knows an existing pack it wants to take the object |
1429 | * from, that is passed in *found_pack and *found_offset; otherwise this | |
1430 | * function finds if there is any pack that has the object and returns the pack | |
1431 | * and its offset in these variables. | |
ce2bc424 | 1432 | */ |
188960b4 | 1433 | static int want_object_in_pack(const struct object_id *oid, |
ce2bc424 JK |
1434 | int exclude, |
1435 | struct packed_git **found_pack, | |
1436 | off_t *found_offset) | |
1437 | { | |
702d1b95 | 1438 | int want; |
8865859d | 1439 | struct list_head *pos; |
6a22d521 | 1440 | struct multi_pack_index *m; |
ce2bc424 | 1441 | |
6862ebbf | 1442 | if (!exclude && local && has_loose_object_nonlocal(oid)) |
daae0625 BC |
1443 | return 0; |
1444 | ||
702d1b95 KS |
1445 | /* |
1446 | * If we already know the pack object lives in, start checks from that | |
1447 | * pack - in the usual case when neither --local was given nor .keep files | |
1448 | * are present we will determine the answer right now. | |
1449 | */ | |
1450 | if (*found_pack) { | |
6325da14 | 1451 | want = want_found_object(oid, exclude, *found_pack); |
702d1b95 KS |
1452 | if (want != -1) |
1453 | return want; | |
1454 | } | |
6a22d521 DS |
1455 | |
1456 | for (m = get_multi_pack_index(the_repository); m; m = m->next) { | |
1457 | struct pack_entry e; | |
64404a24 | 1458 | if (fill_midx_entry(the_repository, oid, &e, m)) { |
6325da14 JK |
1459 | want = want_object_in_pack_one(e.p, oid, exclude, found_pack, found_offset); |
1460 | if (want != -1) | |
1461 | return want; | |
6a22d521 DS |
1462 | } |
1463 | } | |
1464 | ||
a80d72db | 1465 | list_for_each(pos, get_packed_git_mru(the_repository)) { |
ec2dd32c | 1466 | struct packed_git *p = list_entry(pos, struct packed_git, mru); |
6325da14 JK |
1467 | want = want_object_in_pack_one(p, oid, exclude, found_pack, found_offset); |
1468 | if (!exclude && want > 0) | |
1469 | list_move(&p->mru, | |
1470 | get_packed_git_mru(the_repository)); | |
1471 | if (want != -1) | |
1472 | return want; | |
64560374 | 1473 | } |
eb019375 | 1474 | |
dd4b732d JT |
1475 | if (uri_protocols.nr) { |
1476 | struct configured_exclusion *ex = | |
1477 | oidmap_get(&configured_exclusions, oid); | |
1478 | int i; | |
1479 | const char *p; | |
1480 | ||
1481 | if (ex) { | |
1482 | for (i = 0; i < uri_protocols.nr; i++) { | |
1483 | if (skip_prefix(ex->uri, | |
1484 | uri_protocols.items[i].string, | |
1485 | &p) && | |
1486 | *p == ':') { | |
1487 | oidset_insert(&excluded_by_config, oid); | |
1488 | return 0; | |
1489 | } | |
1490 | } | |
1491 | } | |
1492 | } | |
1493 | ||
ce2bc424 JK |
1494 | return 1; |
1495 | } | |
1496 | ||
188960b4 | 1497 | static void create_object_entry(const struct object_id *oid, |
ce2bc424 JK |
1498 | enum object_type type, |
1499 | uint32_t hash, | |
1500 | int exclude, | |
1501 | int no_try_delta, | |
ce2bc424 JK |
1502 | struct packed_git *found_pack, |
1503 | off_t found_offset) | |
1504 | { | |
1505 | struct object_entry *entry; | |
29b734e4 | 1506 | |
3a37876b | 1507 | entry = packlist_alloc(&to_pack, oid); |
27225f2e | 1508 | entry->hash = hash; |
fd9b1bae | 1509 | oe_set_type(entry, type); |
29b734e4 NP |
1510 | if (exclude) |
1511 | entry->preferred_base = 1; | |
81a216a5 NP |
1512 | else |
1513 | nr_result++; | |
29b734e4 | 1514 | if (found_pack) { |
43fa44fa | 1515 | oe_set_in_pack(&to_pack, entry, found_pack); |
29b734e4 NP |
1516 | entry->in_pack_offset = found_offset; |
1517 | } | |
7a979d99 | 1518 | |
ce2bc424 JK |
1519 | entry->no_try_delta = no_try_delta; |
1520 | } | |
29b734e4 | 1521 | |
373c67da JK |
1522 | static const char no_closure_warning[] = N_( |
1523 | "disabling bitmap writing, as some objects are not being packed" | |
1524 | ); | |
1525 | ||
188960b4 | 1526 | static int add_object_entry(const struct object_id *oid, enum object_type type, |
ce2bc424 JK |
1527 | const char *name, int exclude) |
1528 | { | |
702d1b95 KS |
1529 | struct packed_git *found_pack = NULL; |
1530 | off_t found_offset = 0; | |
7a979d99 | 1531 | |
5af05043 NTND |
1532 | display_progress(progress_state, ++nr_seen); |
1533 | ||
3a37876b | 1534 | if (have_duplicate_entry(oid, exclude)) |
ce2bc424 | 1535 | return 0; |
a74db82e | 1536 | |
188960b4 | 1537 | if (!want_object_in_pack(oid, exclude, &found_pack, &found_offset)) { |
373c67da JK |
1538 | /* The pack is missing an object, so it will not have closure */ |
1539 | if (write_bitmap_index) { | |
25575015 JK |
1540 | if (write_bitmap_index != WRITE_BITMAP_QUIET) |
1541 | warning(_(no_closure_warning)); | |
373c67da JK |
1542 | write_bitmap_index = 0; |
1543 | } | |
ce2bc424 | 1544 | return 0; |
373c67da | 1545 | } |
29b734e4 | 1546 | |
188960b4 | 1547 | create_object_entry(oid, type, pack_name_hash(name), |
ce2bc424 | 1548 | exclude, name && no_try_delta(name), |
3a37876b | 1549 | found_pack, found_offset); |
29b734e4 | 1550 | return 1; |
c323ac7d LT |
1551 | } |
1552 | ||
20664967 | 1553 | static int add_object_entry_from_bitmap(const struct object_id *oid, |
6b8fda2d VM |
1554 | enum object_type type, |
1555 | int flags, uint32_t name_hash, | |
1556 | struct packed_git *pack, off_t offset) | |
1557 | { | |
5af05043 NTND |
1558 | display_progress(progress_state, ++nr_seen); |
1559 | ||
3a37876b | 1560 | if (have_duplicate_entry(oid, 0)) |
6b8fda2d VM |
1561 | return 0; |
1562 | ||
188960b4 | 1563 | if (!want_object_in_pack(oid, 0, &pack, &offset)) |
702d1b95 KS |
1564 | return 0; |
1565 | ||
3a37876b | 1566 | create_object_entry(oid, type, name_hash, 0, 0, pack, offset); |
29b734e4 | 1567 | return 1; |
c323ac7d LT |
1568 | } |
1569 | ||
5379a5c5 | 1570 | struct pbase_tree_cache { |
188960b4 | 1571 | struct object_id oid; |
5379a5c5 JH |
1572 | int ref; |
1573 | int temporary; | |
1574 | void *tree_data; | |
1575 | unsigned long tree_size; | |
1576 | }; | |
1577 | ||
1578 | static struct pbase_tree_cache *(pbase_tree_cache[256]); | |
188960b4 | 1579 | static int pbase_tree_cache_ix(const struct object_id *oid) |
5379a5c5 | 1580 | { |
188960b4 | 1581 | return oid->hash[0] % ARRAY_SIZE(pbase_tree_cache); |
5379a5c5 JH |
1582 | } |
1583 | static int pbase_tree_cache_ix_incr(int ix) | |
1584 | { | |
1585 | return (ix+1) % ARRAY_SIZE(pbase_tree_cache); | |
1586 | } | |
1587 | ||
1588 | static struct pbase_tree { | |
1589 | struct pbase_tree *next; | |
1590 | /* This is a phony "cache" entry; we are not | |
01689909 | 1591 | * going to evict it or find it through _get() |
5379a5c5 JH |
1592 | * mechanism -- this is for the toplevel node that |
1593 | * would almost always change with any commit. | |
1594 | */ | |
1595 | struct pbase_tree_cache pcache; | |
1596 | } *pbase_tree; | |
1597 | ||
188960b4 | 1598 | static struct pbase_tree_cache *pbase_tree_get(const struct object_id *oid) |
5379a5c5 JH |
1599 | { |
1600 | struct pbase_tree_cache *ent, *nent; | |
1601 | void *data; | |
1602 | unsigned long size; | |
21666f1a | 1603 | enum object_type type; |
5379a5c5 | 1604 | int neigh; |
188960b4 | 1605 | int my_ix = pbase_tree_cache_ix(oid); |
5379a5c5 JH |
1606 | int available_ix = -1; |
1607 | ||
1608 | /* pbase-tree-cache acts as a limited hashtable. | |
1609 | * your object will be found at your index or within a few | |
1610 | * slots after that slot if it is cached. | |
1611 | */ | |
1612 | for (neigh = 0; neigh < 8; neigh++) { | |
1613 | ent = pbase_tree_cache[my_ix]; | |
4a7e27e9 | 1614 | if (ent && oideq(&ent->oid, oid)) { |
5379a5c5 JH |
1615 | ent->ref++; |
1616 | return ent; | |
1617 | } | |
1618 | else if (((available_ix < 0) && (!ent || !ent->ref)) || | |
1619 | ((0 <= available_ix) && | |
1620 | (!ent && pbase_tree_cache[available_ix]))) | |
1621 | available_ix = my_ix; | |
1622 | if (!ent) | |
1623 | break; | |
1624 | my_ix = pbase_tree_cache_ix_incr(my_ix); | |
1625 | } | |
1626 | ||
1627 | /* Did not find one. Either we got a bogus request or | |
1628 | * we need to read and perhaps cache. | |
1629 | */ | |
b4f5aca4 | 1630 | data = read_object_file(oid, &type, &size); |
5379a5c5 JH |
1631 | if (!data) |
1632 | return NULL; | |
21666f1a | 1633 | if (type != OBJ_TREE) { |
5379a5c5 JH |
1634 | free(data); |
1635 | return NULL; | |
1636 | } | |
1637 | ||
1638 | /* We need to either cache or return a throwaway copy */ | |
1639 | ||
1640 | if (available_ix < 0) | |
1641 | ent = NULL; | |
1642 | else { | |
1643 | ent = pbase_tree_cache[available_ix]; | |
1644 | my_ix = available_ix; | |
1645 | } | |
1646 | ||
1647 | if (!ent) { | |
1648 | nent = xmalloc(sizeof(*nent)); | |
1649 | nent->temporary = (available_ix < 0); | |
1650 | } | |
1651 | else { | |
1652 | /* evict and reuse */ | |
1653 | free(ent->tree_data); | |
1654 | nent = ent; | |
1655 | } | |
188960b4 | 1656 | oidcpy(&nent->oid, oid); |
5379a5c5 JH |
1657 | nent->tree_data = data; |
1658 | nent->tree_size = size; | |
1659 | nent->ref = 1; | |
1660 | if (!nent->temporary) | |
1661 | pbase_tree_cache[my_ix] = nent; | |
1662 | return nent; | |
1663 | } | |
1664 | ||
1665 | static void pbase_tree_put(struct pbase_tree_cache *cache) | |
1666 | { | |
1667 | if (!cache->temporary) { | |
1668 | cache->ref--; | |
1669 | return; | |
1670 | } | |
1671 | free(cache->tree_data); | |
1672 | free(cache); | |
1673 | } | |
1674 | ||
1675 | static int name_cmp_len(const char *name) | |
1676 | { | |
1677 | int i; | |
1678 | for (i = 0; name[i] && name[i] != '\n' && name[i] != '/'; i++) | |
1679 | ; | |
1680 | return i; | |
1681 | } | |
1682 | ||
1683 | static void add_pbase_object(struct tree_desc *tree, | |
5379a5c5 | 1684 | const char *name, |
ce0bd642 LT |
1685 | int cmplen, |
1686 | const char *fullname) | |
3f9ac8d2 | 1687 | { |
4c068a98 | 1688 | struct name_entry entry; |
8a5a8d6c | 1689 | int cmp; |
4c068a98 LT |
1690 | |
1691 | while (tree_entry(tree,&entry)) { | |
1211be6b LT |
1692 | if (S_ISGITLINK(entry.mode)) |
1693 | continue; | |
0de16337 | 1694 | cmp = tree_entry_len(&entry) != cmplen ? 1 : |
8a5a8d6c NP |
1695 | memcmp(name, entry.path, cmplen); |
1696 | if (cmp > 0) | |
7a979d99 | 1697 | continue; |
8a5a8d6c NP |
1698 | if (cmp < 0) |
1699 | return; | |
5379a5c5 | 1700 | if (name[cmplen] != '/') { |
ea82b2a0 | 1701 | add_object_entry(&entry.oid, |
4d1012c3 | 1702 | object_type(entry.mode), |
bc32fed5 | 1703 | fullname, 1); |
5379a5c5 JH |
1704 | return; |
1705 | } | |
8a5a8d6c | 1706 | if (S_ISDIR(entry.mode)) { |
7a979d99 | 1707 | struct tree_desc sub; |
5379a5c5 JH |
1708 | struct pbase_tree_cache *tree; |
1709 | const char *down = name+cmplen+1; | |
1710 | int downlen = name_cmp_len(down); | |
1711 | ||
ea82b2a0 | 1712 | tree = pbase_tree_get(&entry.oid); |
5379a5c5 JH |
1713 | if (!tree) |
1714 | return; | |
6fda5e51 | 1715 | init_tree_desc(&sub, tree->tree_data, tree->tree_size); |
5379a5c5 | 1716 | |
ce0bd642 | 1717 | add_pbase_object(&sub, down, downlen, fullname); |
5379a5c5 JH |
1718 | pbase_tree_put(tree); |
1719 | } | |
1720 | } | |
1721 | } | |
1d6b38cc | 1722 | |
5379a5c5 JH |
1723 | static unsigned *done_pbase_paths; |
1724 | static int done_pbase_paths_num; | |
1725 | static int done_pbase_paths_alloc; | |
1726 | static int done_pbase_path_pos(unsigned hash) | |
1727 | { | |
1728 | int lo = 0; | |
1729 | int hi = done_pbase_paths_num; | |
1730 | while (lo < hi) { | |
19716b21 | 1731 | int mi = lo + (hi - lo) / 2; |
5379a5c5 JH |
1732 | if (done_pbase_paths[mi] == hash) |
1733 | return mi; | |
1734 | if (done_pbase_paths[mi] < hash) | |
1735 | hi = mi; | |
1736 | else | |
1737 | lo = mi + 1; | |
1738 | } | |
1739 | return -lo-1; | |
1740 | } | |
1741 | ||
1742 | static int check_pbase_path(unsigned hash) | |
1743 | { | |
c7b07805 | 1744 | int pos = done_pbase_path_pos(hash); |
5379a5c5 JH |
1745 | if (0 <= pos) |
1746 | return 1; | |
1747 | pos = -pos - 1; | |
25e19407 DD |
1748 | ALLOC_GROW(done_pbase_paths, |
1749 | done_pbase_paths_num + 1, | |
1750 | done_pbase_paths_alloc); | |
5379a5c5 JH |
1751 | done_pbase_paths_num++; |
1752 | if (pos < done_pbase_paths_num) | |
f331ab9d RS |
1753 | MOVE_ARRAY(done_pbase_paths + pos + 1, done_pbase_paths + pos, |
1754 | done_pbase_paths_num - pos - 1); | |
5379a5c5 JH |
1755 | done_pbase_paths[pos] = hash; |
1756 | return 0; | |
1757 | } | |
1758 | ||
bc32fed5 | 1759 | static void add_preferred_base_object(const char *name) |
5379a5c5 JH |
1760 | { |
1761 | struct pbase_tree *it; | |
8a5a8d6c | 1762 | int cmplen; |
68fb36eb | 1763 | unsigned hash = pack_name_hash(name); |
5379a5c5 | 1764 | |
8a5a8d6c | 1765 | if (!num_preferred_base || check_pbase_path(hash)) |
5379a5c5 JH |
1766 | return; |
1767 | ||
8a5a8d6c | 1768 | cmplen = name_cmp_len(name); |
5379a5c5 JH |
1769 | for (it = pbase_tree; it; it = it->next) { |
1770 | if (cmplen == 0) { | |
188960b4 | 1771 | add_object_entry(&it->pcache.oid, OBJ_TREE, NULL, 1); |
5379a5c5 JH |
1772 | } |
1773 | else { | |
1774 | struct tree_desc tree; | |
6fda5e51 | 1775 | init_tree_desc(&tree, it->pcache.tree_data, it->pcache.tree_size); |
ce0bd642 | 1776 | add_pbase_object(&tree, name, cmplen, name); |
7a979d99 | 1777 | } |
3f9ac8d2 | 1778 | } |
3f9ac8d2 JH |
1779 | } |
1780 | ||
188960b4 | 1781 | static void add_preferred_base(struct object_id *oid) |
3f9ac8d2 | 1782 | { |
5379a5c5 JH |
1783 | struct pbase_tree *it; |
1784 | void *data; | |
1785 | unsigned long size; | |
188960b4 | 1786 | struct object_id tree_oid; |
1d6b38cc | 1787 | |
8d1d8f83 JH |
1788 | if (window <= num_preferred_base++) |
1789 | return; | |
1790 | ||
d3b4705a NTND |
1791 | data = read_object_with_reference(the_repository, oid, |
1792 | tree_type, &size, &tree_oid); | |
5379a5c5 | 1793 | if (!data) |
7a979d99 | 1794 | return; |
5379a5c5 JH |
1795 | |
1796 | for (it = pbase_tree; it; it = it->next) { | |
4a7e27e9 | 1797 | if (oideq(&it->pcache.oid, &tree_oid)) { |
5379a5c5 JH |
1798 | free(data); |
1799 | return; | |
1800 | } | |
1801 | } | |
1802 | ||
ca56dadb | 1803 | CALLOC_ARRAY(it, 1); |
5379a5c5 JH |
1804 | it->next = pbase_tree; |
1805 | pbase_tree = it; | |
1806 | ||
188960b4 | 1807 | oidcpy(&it->pcache.oid, &tree_oid); |
5379a5c5 JH |
1808 | it->pcache.tree_data = data; |
1809 | it->pcache.tree_size = size; | |
3f9ac8d2 JH |
1810 | } |
1811 | ||
0ef95f72 NP |
1812 | static void cleanup_preferred_base(void) |
1813 | { | |
1814 | struct pbase_tree *it; | |
1815 | unsigned i; | |
1816 | ||
1817 | it = pbase_tree; | |
1818 | pbase_tree = NULL; | |
1819 | while (it) { | |
095b3b2c BW |
1820 | struct pbase_tree *tmp = it; |
1821 | it = tmp->next; | |
1822 | free(tmp->pcache.tree_data); | |
1823 | free(tmp); | |
0ef95f72 NP |
1824 | } |
1825 | ||
1826 | for (i = 0; i < ARRAY_SIZE(pbase_tree_cache); i++) { | |
1827 | if (!pbase_tree_cache[i]) | |
1828 | continue; | |
1829 | free(pbase_tree_cache[i]->tree_data); | |
6a83d902 | 1830 | FREE_AND_NULL(pbase_tree_cache[i]); |
0ef95f72 NP |
1831 | } |
1832 | ||
6a83d902 | 1833 | FREE_AND_NULL(done_pbase_paths); |
0ef95f72 NP |
1834 | done_pbase_paths_num = done_pbase_paths_alloc = 0; |
1835 | } | |
1836 | ||
2fa233a5 JK |
1837 | /* |
1838 | * Return 1 iff the object specified by "delta" can be sent | |
1839 | * literally as a delta against the base in "base_sha1". If | |
1840 | * so, then *base_out will point to the entry in our packing | |
1841 | * list, or NULL if we must use the external-base list. | |
1842 | * | |
1843 | * Depth value does not matter - find_deltas() will | |
1844 | * never consider reused delta as the base object to | |
1845 | * deltify other objects against, in order to avoid | |
1846 | * circular deltas. | |
1847 | */ | |
3f83fd5e | 1848 | static int can_reuse_delta(const struct object_id *base_oid, |
2fa233a5 JK |
1849 | struct object_entry *delta, |
1850 | struct object_entry **base_out) | |
1851 | { | |
1852 | struct object_entry *base; | |
3df28cae | 1853 | |
2fa233a5 JK |
1854 | /* |
1855 | * First see if we're already sending the base (or it's explicitly in | |
1856 | * our "excluded" list). | |
1857 | */ | |
3f83fd5e | 1858 | base = packlist_find(&to_pack, base_oid); |
2fa233a5 JK |
1859 | if (base) { |
1860 | if (!in_same_island(&delta->idx.oid, &base->idx.oid)) | |
1861 | return 0; | |
1862 | *base_out = base; | |
1863 | return 1; | |
1864 | } | |
1865 | ||
1866 | /* | |
1867 | * Otherwise, reachability bitmaps may tell us if the receiver has it, | |
1868 | * even if it was buried too deep in history to make it into the | |
1869 | * packing list. | |
1870 | */ | |
3f83fd5e | 1871 | if (thin && bitmap_has_oid_in_uninteresting(bitmap_git, base_oid)) { |
2fa233a5 | 1872 | if (use_delta_islands) { |
3f83fd5e | 1873 | if (!in_same_island(&delta->idx.oid, base_oid)) |
2fa233a5 JK |
1874 | return 0; |
1875 | } | |
1876 | *base_out = NULL; | |
1877 | return 1; | |
1878 | } | |
1879 | ||
1880 | return 0; | |
1881 | } | |
1882 | ||
e00549aa JT |
1883 | static void prefetch_to_pack(uint32_t object_index_start) { |
1884 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
1885 | uint32_t i; | |
1886 | ||
1887 | for (i = object_index_start; i < to_pack.nr_objects; i++) { | |
1888 | struct object_entry *entry = to_pack.objects + i; | |
1889 | ||
1890 | if (!oid_object_info_extended(the_repository, | |
1891 | &entry->idx.oid, | |
1892 | NULL, | |
1893 | OBJECT_INFO_FOR_PREFETCH)) | |
1894 | continue; | |
1895 | oid_array_append(&to_fetch, &entry->idx.oid); | |
1896 | } | |
1897 | promisor_remote_get_direct(the_repository, | |
1898 | to_fetch.oid, to_fetch.nr); | |
1899 | oid_array_clear(&to_fetch); | |
1900 | } | |
1901 | ||
1902 | static void check_object(struct object_entry *entry, uint32_t object_index) | |
c323ac7d | 1903 | { |
ac77d0c3 | 1904 | unsigned long canonical_size; |
8d5cf957 JT |
1905 | enum object_type type; |
1906 | struct object_info oi = {.typep = &type, .sizep = &canonical_size}; | |
ac77d0c3 | 1907 | |
43fa44fa NTND |
1908 | if (IN_PACK(entry)) { |
1909 | struct packed_git *p = IN_PACK(entry); | |
03e79c88 | 1910 | struct pack_window *w_curs = NULL; |
3f83fd5e JK |
1911 | int have_base = 0; |
1912 | struct object_id base_ref; | |
5c49c116 NP |
1913 | struct object_entry *base_entry; |
1914 | unsigned long used, used_0; | |
ef49a7a0 | 1915 | unsigned long avail; |
5c49c116 NP |
1916 | off_t ofs; |
1917 | unsigned char *buf, c; | |
fd9b1bae | 1918 | enum object_type type; |
27a7d067 | 1919 | unsigned long in_pack_size; |
780e6e73 | 1920 | |
6777a59f | 1921 | buf = use_pack(p, &w_curs, entry->in_pack_offset, &avail); |
ab7cd7bb | 1922 | |
5c49c116 | 1923 | /* |
fa736f72 NP |
1924 | * We want in_pack_type even if we do not reuse delta |
1925 | * since non-delta representations could still be reused. | |
ab7cd7bb | 1926 | */ |
09ded04b | 1927 | used = unpack_object_header_buffer(buf, avail, |
fd9b1bae | 1928 | &type, |
27a7d067 | 1929 | &in_pack_size); |
03d66015 NP |
1930 | if (used == 0) |
1931 | goto give_up; | |
ab7cd7bb | 1932 | |
fd9b1bae NTND |
1933 | if (type < 0) |
1934 | BUG("invalid type %d", type); | |
1935 | entry->in_pack_type = type; | |
1936 | ||
5c49c116 NP |
1937 | /* |
1938 | * Determine if this is a delta and if so whether we can | |
1939 | * reuse it or not. Otherwise let's find out as cheaply as | |
1940 | * possible what the actual type and size for this object is. | |
3f9ac8d2 | 1941 | */ |
5c49c116 NP |
1942 | switch (entry->in_pack_type) { |
1943 | default: | |
1944 | /* Not a delta hence we've already got all we need. */ | |
fd9b1bae | 1945 | oe_set_type(entry, entry->in_pack_type); |
ac77d0c3 | 1946 | SET_SIZE(entry, in_pack_size); |
5c49c116 | 1947 | entry->in_pack_header_size = used; |
fd9b1bae | 1948 | if (oe_type(entry) < OBJ_COMMIT || oe_type(entry) > OBJ_BLOB) |
03d66015 | 1949 | goto give_up; |
5c49c116 NP |
1950 | unuse_pack(&w_curs); |
1951 | return; | |
1952 | case OBJ_REF_DELTA: | |
3f83fd5e JK |
1953 | if (reuse_delta && !entry->preferred_base) { |
1954 | oidread(&base_ref, | |
1955 | use_pack(p, &w_curs, | |
1956 | entry->in_pack_offset + used, | |
1957 | NULL)); | |
1958 | have_base = 1; | |
1959 | } | |
41179100 | 1960 | entry->in_pack_header_size = used + the_hash_algo->rawsz; |
5c49c116 NP |
1961 | break; |
1962 | case OBJ_OFS_DELTA: | |
1963 | buf = use_pack(p, &w_curs, | |
1964 | entry->in_pack_offset + used, NULL); | |
1965 | used_0 = 0; | |
1966 | c = buf[used_0++]; | |
1967 | ofs = c & 127; | |
1968 | while (c & 128) { | |
1969 | ofs += 1; | |
03d66015 | 1970 | if (!ofs || MSB(ofs, 7)) { |
f616db6a | 1971 | error(_("delta base offset overflow in pack for %s"), |
e6a492b7 | 1972 | oid_to_hex(&entry->idx.oid)); |
03d66015 NP |
1973 | goto give_up; |
1974 | } | |
5c49c116 NP |
1975 | c = buf[used_0++]; |
1976 | ofs = (ofs << 7) + (c & 127); | |
780e6e73 | 1977 | } |
5c49c116 | 1978 | ofs = entry->in_pack_offset - ofs; |
03d66015 | 1979 | if (ofs <= 0 || ofs >= entry->in_pack_offset) { |
f616db6a | 1980 | error(_("delta base offset out of bound for %s"), |
e6a492b7 | 1981 | oid_to_hex(&entry->idx.oid)); |
03d66015 NP |
1982 | goto give_up; |
1983 | } | |
a7de7130 | 1984 | if (reuse_delta && !entry->preferred_base) { |
eb3fd99e TB |
1985 | uint32_t pos; |
1986 | if (offset_to_pack_pos(p, ofs, &pos) < 0) | |
08698b1e | 1987 | goto give_up; |
eb3fd99e TB |
1988 | if (!nth_packed_object_id(&base_ref, p, |
1989 | pack_pos_to_index(p, pos))) | |
3f83fd5e | 1990 | have_base = 1; |
3449f8c4 | 1991 | } |
5c49c116 NP |
1992 | entry->in_pack_header_size = used + used_0; |
1993 | break; | |
780e6e73 | 1994 | } |
780e6e73 | 1995 | |
3f83fd5e JK |
1996 | if (have_base && |
1997 | can_reuse_delta(&base_ref, entry, &base_entry)) { | |
fd9b1bae | 1998 | oe_set_type(entry, entry->in_pack_type); |
ac77d0c3 | 1999 | SET_SIZE(entry, in_pack_size); /* delta size */ |
0aca34e8 | 2000 | SET_DELTA_SIZE(entry, in_pack_size); |
6a1e32d5 JK |
2001 | |
2002 | if (base_entry) { | |
2003 | SET_DELTA(entry, base_entry); | |
2004 | entry->delta_sibling_idx = base_entry->delta_child_idx; | |
2005 | SET_DELTA_CHILD(base_entry, entry); | |
2006 | } else { | |
a93c141d | 2007 | SET_DELTA_EXT(entry, &base_ref); |
6a1e32d5 JK |
2008 | } |
2009 | ||
5c49c116 NP |
2010 | unuse_pack(&w_curs); |
2011 | return; | |
2012 | } | |
ab7cd7bb | 2013 | |
fd9b1bae | 2014 | if (oe_type(entry)) { |
27a7d067 NTND |
2015 | off_t delta_pos; |
2016 | ||
5c49c116 NP |
2017 | /* |
2018 | * This must be a delta and we already know what the | |
2019 | * final object type is. Let's extract the actual | |
2020 | * object size from the delta header. | |
2021 | */ | |
27a7d067 | 2022 | delta_pos = entry->in_pack_offset + entry->in_pack_header_size; |
ac77d0c3 NTND |
2023 | canonical_size = get_size_from_delta(p, &w_curs, delta_pos); |
2024 | if (canonical_size == 0) | |
03d66015 | 2025 | goto give_up; |
ac77d0c3 | 2026 | SET_SIZE(entry, canonical_size); |
5c49c116 | 2027 | unuse_pack(&w_curs); |
3f9ac8d2 JH |
2028 | return; |
2029 | } | |
5c49c116 NP |
2030 | |
2031 | /* | |
2032 | * No choice but to fall back to the recursive delta walk | |
c93206b4 | 2033 | * with oid_object_info() to find about the object type |
5c49c116 NP |
2034 | * at this point... |
2035 | */ | |
03d66015 | 2036 | give_up: |
5c49c116 | 2037 | unuse_pack(&w_curs); |
36e4d74a | 2038 | } |
3f9ac8d2 | 2039 | |
8d5cf957 | 2040 | if (oid_object_info_extended(the_repository, &entry->idx.oid, &oi, |
e00549aa JT |
2041 | OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_LOOKUP_REPLACE) < 0) { |
2042 | if (has_promisor_remote()) { | |
2043 | prefetch_to_pack(object_index); | |
2044 | if (oid_object_info_extended(the_repository, &entry->idx.oid, &oi, | |
2045 | OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_LOOKUP_REPLACE) < 0) | |
2046 | type = -1; | |
2047 | } else { | |
2048 | type = -1; | |
2049 | } | |
2050 | } | |
8d5cf957 | 2051 | oe_set_type(entry, type); |
ac77d0c3 NTND |
2052 | if (entry->type_valid) { |
2053 | SET_SIZE(entry, canonical_size); | |
2054 | } else { | |
2055 | /* | |
2056 | * Bad object type is checked in prepare_pack(). This is | |
2057 | * to permit a missing preferred base object to be ignored | |
2058 | * as a preferred base. Doing so can result in a larger | |
2059 | * pack file, but the transfer will still take place. | |
2060 | */ | |
2061 | } | |
3f9ac8d2 JH |
2062 | } |
2063 | ||
5c49c116 NP |
2064 | static int pack_offset_sort(const void *_a, const void *_b) |
2065 | { | |
2066 | const struct object_entry *a = *(struct object_entry **)_a; | |
2067 | const struct object_entry *b = *(struct object_entry **)_b; | |
43fa44fa NTND |
2068 | const struct packed_git *a_in_pack = IN_PACK(a); |
2069 | const struct packed_git *b_in_pack = IN_PACK(b); | |
5c49c116 NP |
2070 | |
2071 | /* avoid filesystem trashing with loose objects */ | |
43fa44fa | 2072 | if (!a_in_pack && !b_in_pack) |
e6a492b7 | 2073 | return oidcmp(&a->idx.oid, &b->idx.oid); |
5c49c116 | 2074 | |
43fa44fa | 2075 | if (a_in_pack < b_in_pack) |
5c49c116 | 2076 | return -1; |
43fa44fa | 2077 | if (a_in_pack > b_in_pack) |
5c49c116 NP |
2078 | return 1; |
2079 | return a->in_pack_offset < b->in_pack_offset ? -1 : | |
2080 | (a->in_pack_offset > b->in_pack_offset); | |
2081 | } | |
2082 | ||
4cf2143e JK |
2083 | /* |
2084 | * Drop an on-disk delta we were planning to reuse. Naively, this would | |
2085 | * just involve blanking out the "delta" field, but we have to deal | |
2086 | * with some extra book-keeping: | |
2087 | * | |
2088 | * 1. Removing ourselves from the delta_sibling linked list. | |
2089 | * | |
2090 | * 2. Updating our size/type to the non-delta representation. These were | |
2091 | * either not recorded initially (size) or overwritten with the delta type | |
2092 | * (type) when check_object() decided to reuse the delta. | |
7dbabbbe JK |
2093 | * |
2094 | * 3. Resetting our delta depth, as we are now a base object. | |
4cf2143e JK |
2095 | */ |
2096 | static void drop_reused_delta(struct object_entry *entry) | |
2097 | { | |
898eba5e | 2098 | unsigned *idx = &to_pack.objects[entry->delta_idx - 1].delta_child_idx; |
4cf2143e | 2099 | struct object_info oi = OBJECT_INFO_INIT; |
fd9b1bae | 2100 | enum object_type type; |
ac77d0c3 | 2101 | unsigned long size; |
4cf2143e | 2102 | |
898eba5e NTND |
2103 | while (*idx) { |
2104 | struct object_entry *oe = &to_pack.objects[*idx - 1]; | |
4cf2143e | 2105 | |
898eba5e NTND |
2106 | if (oe == entry) |
2107 | *idx = oe->delta_sibling_idx; | |
4cf2143e | 2108 | else |
898eba5e | 2109 | idx = &oe->delta_sibling_idx; |
4cf2143e | 2110 | } |
898eba5e | 2111 | SET_DELTA(entry, NULL); |
7dbabbbe | 2112 | entry->depth = 0; |
4cf2143e | 2113 | |
ac77d0c3 | 2114 | oi.sizep = &size; |
fd9b1bae | 2115 | oi.typep = &type; |
ad635e82 | 2116 | if (packed_object_info(the_repository, IN_PACK(entry), entry->in_pack_offset, &oi) < 0) { |
4cf2143e JK |
2117 | /* |
2118 | * We failed to get the info from this pack for some reason; | |
c93206b4 | 2119 | * fall back to oid_object_info, which may find another copy. |
fd9b1bae | 2120 | * And if that fails, the error will be recorded in oe_type(entry) |
4cf2143e JK |
2121 | * and dealt with in prepare_pack(). |
2122 | */ | |
ad635e82 JH |
2123 | oe_set_type(entry, |
2124 | oid_object_info(the_repository, &entry->idx.oid, &size)); | |
fd9b1bae NTND |
2125 | } else { |
2126 | oe_set_type(entry, type); | |
4cf2143e | 2127 | } |
ac77d0c3 | 2128 | SET_SIZE(entry, size); |
4cf2143e JK |
2129 | } |
2130 | ||
2131 | /* | |
2132 | * Follow the chain of deltas from this entry onward, throwing away any links | |
2133 | * that cause us to hit a cycle (as determined by the DFS state flags in | |
2134 | * the entries). | |
7dbabbbe JK |
2135 | * |
2136 | * We also detect too-long reused chains that would violate our --depth | |
2137 | * limit. | |
4cf2143e JK |
2138 | */ |
2139 | static void break_delta_chains(struct object_entry *entry) | |
2140 | { | |
42b766d7 JK |
2141 | /* |
2142 | * The actual depth of each object we will write is stored as an int, | |
2143 | * as it cannot exceed our int "depth" limit. But before we break | |
2144 | * changes based no that limit, we may potentially go as deep as the | |
2145 | * number of objects, which is elsewhere bounded to a uint32_t. | |
2146 | */ | |
2147 | uint32_t total_depth; | |
2148 | struct object_entry *cur, *next; | |
2149 | ||
2150 | for (cur = entry, total_depth = 0; | |
2151 | cur; | |
898eba5e | 2152 | cur = DELTA(cur), total_depth++) { |
42b766d7 JK |
2153 | if (cur->dfs_state == DFS_DONE) { |
2154 | /* | |
2155 | * We've already seen this object and know it isn't | |
2156 | * part of a cycle. We do need to append its depth | |
2157 | * to our count. | |
2158 | */ | |
2159 | total_depth += cur->depth; | |
2160 | break; | |
2161 | } | |
4cf2143e | 2162 | |
4cf2143e | 2163 | /* |
42b766d7 JK |
2164 | * We break cycles before looping, so an ACTIVE state (or any |
2165 | * other cruft which made its way into the state variable) | |
2166 | * is a bug. | |
4cf2143e | 2167 | */ |
42b766d7 | 2168 | if (cur->dfs_state != DFS_NONE) |
033abf97 | 2169 | BUG("confusing delta dfs state in first pass: %d", |
42b766d7 | 2170 | cur->dfs_state); |
4cf2143e | 2171 | |
4cf2143e | 2172 | /* |
42b766d7 JK |
2173 | * Now we know this is the first time we've seen the object. If |
2174 | * it's not a delta, we're done traversing, but we'll mark it | |
2175 | * done to save time on future traversals. | |
4cf2143e | 2176 | */ |
898eba5e | 2177 | if (!DELTA(cur)) { |
42b766d7 JK |
2178 | cur->dfs_state = DFS_DONE; |
2179 | break; | |
2180 | } | |
4cf2143e | 2181 | |
4cf2143e | 2182 | /* |
42b766d7 JK |
2183 | * Mark ourselves as active and see if the next step causes |
2184 | * us to cycle to another active object. It's important to do | |
2185 | * this _before_ we loop, because it impacts where we make the | |
2186 | * cut, and thus how our total_depth counter works. | |
2187 | * E.g., We may see a partial loop like: | |
2188 | * | |
2189 | * A -> B -> C -> D -> B | |
2190 | * | |
2191 | * Cutting B->C breaks the cycle. But now the depth of A is | |
2192 | * only 1, and our total_depth counter is at 3. The size of the | |
2193 | * error is always one less than the size of the cycle we | |
2194 | * broke. Commits C and D were "lost" from A's chain. | |
2195 | * | |
2196 | * If we instead cut D->B, then the depth of A is correct at 3. | |
2197 | * We keep all commits in the chain that we examined. | |
4cf2143e | 2198 | */ |
42b766d7 | 2199 | cur->dfs_state = DFS_ACTIVE; |
898eba5e | 2200 | if (DELTA(cur)->dfs_state == DFS_ACTIVE) { |
42b766d7 JK |
2201 | drop_reused_delta(cur); |
2202 | cur->dfs_state = DFS_DONE; | |
2203 | break; | |
7dbabbbe | 2204 | } |
42b766d7 | 2205 | } |
7dbabbbe | 2206 | |
42b766d7 JK |
2207 | /* |
2208 | * And now that we've gone all the way to the bottom of the chain, we | |
2209 | * need to clear the active flags and set the depth fields as | |
2210 | * appropriate. Unlike the loop above, which can quit when it drops a | |
2211 | * delta, we need to keep going to look for more depth cuts. So we need | |
2212 | * an extra "next" pointer to keep going after we reset cur->delta. | |
2213 | */ | |
2214 | for (cur = entry; cur; cur = next) { | |
898eba5e | 2215 | next = DELTA(cur); |
4cf2143e | 2216 | |
42b766d7 JK |
2217 | /* |
2218 | * We should have a chain of zero or more ACTIVE states down to | |
2219 | * a final DONE. We can quit after the DONE, because either it | |
2220 | * has no bases, or we've already handled them in a previous | |
2221 | * call. | |
2222 | */ | |
2223 | if (cur->dfs_state == DFS_DONE) | |
2224 | break; | |
2225 | else if (cur->dfs_state != DFS_ACTIVE) | |
033abf97 | 2226 | BUG("confusing delta dfs state in second pass: %d", |
42b766d7 | 2227 | cur->dfs_state); |
4cf2143e | 2228 | |
4cf2143e | 2229 | /* |
42b766d7 JK |
2230 | * If the total_depth is more than depth, then we need to snip |
2231 | * the chain into two or more smaller chains that don't exceed | |
2232 | * the maximum depth. Most of the resulting chains will contain | |
2233 | * (depth + 1) entries (i.e., depth deltas plus one base), and | |
2234 | * the last chain (i.e., the one containing entry) will contain | |
2235 | * whatever entries are left over, namely | |
2236 | * (total_depth % (depth + 1)) of them. | |
2237 | * | |
2238 | * Since we are iterating towards decreasing depth, we need to | |
2239 | * decrement total_depth as we go, and we need to write to the | |
2240 | * entry what its final depth will be after all of the | |
2241 | * snipping. Since we're snipping into chains of length (depth | |
2242 | * + 1) entries, the final depth of an entry will be its | |
2243 | * original depth modulo (depth + 1). Any time we encounter an | |
2244 | * entry whose final depth is supposed to be zero, we snip it | |
2245 | * from its delta base, thereby making it so. | |
4cf2143e | 2246 | */ |
42b766d7 JK |
2247 | cur->depth = (total_depth--) % (depth + 1); |
2248 | if (!cur->depth) | |
2249 | drop_reused_delta(cur); | |
2250 | ||
2251 | cur->dfs_state = DFS_DONE; | |
4cf2143e JK |
2252 | } |
2253 | } | |
2254 | ||
c323ac7d LT |
2255 | static void get_object_details(void) |
2256 | { | |
7cadf491 | 2257 | uint32_t i; |
5c49c116 NP |
2258 | struct object_entry **sorted_by_offset; |
2259 | ||
5af05043 NTND |
2260 | if (progress) |
2261 | progress_state = start_progress(_("Counting objects"), | |
2262 | to_pack.nr_objects); | |
2263 | ||
ca56dadb | 2264 | CALLOC_ARRAY(sorted_by_offset, to_pack.nr_objects); |
2834bc27 VM |
2265 | for (i = 0; i < to_pack.nr_objects; i++) |
2266 | sorted_by_offset[i] = to_pack.objects + i; | |
9ed0d8d6 | 2267 | QSORT(sorted_by_offset, to_pack.nr_objects, pack_offset_sort); |
c323ac7d | 2268 | |
2834bc27 | 2269 | for (i = 0; i < to_pack.nr_objects; i++) { |
15366280 | 2270 | struct object_entry *entry = sorted_by_offset[i]; |
e00549aa | 2271 | check_object(entry, i); |
ac77d0c3 NTND |
2272 | if (entry->type_valid && |
2273 | oe_size_greater_than(&to_pack, entry, big_file_threshold)) | |
15366280 | 2274 | entry->no_try_delta = 1; |
5af05043 | 2275 | display_progress(progress_state, i + 1); |
15366280 | 2276 | } |
5af05043 | 2277 | stop_progress(&progress_state); |
3449f8c4 | 2278 | |
4cf2143e JK |
2279 | /* |
2280 | * This must happen in a second pass, since we rely on the delta | |
2281 | * information for the whole list being completed. | |
2282 | */ | |
2283 | for (i = 0; i < to_pack.nr_objects; i++) | |
2284 | break_delta_chains(&to_pack.objects[i]); | |
2285 | ||
5c49c116 | 2286 | free(sorted_by_offset); |
c323ac7d LT |
2287 | } |
2288 | ||
b904166c NP |
2289 | /* |
2290 | * We search for deltas in a list sorted by type, by filename hash, and then | |
2291 | * by size, so that we see progressively smaller and smaller files. | |
2292 | * That's because we prefer deltas to be from the bigger file | |
2293 | * to the smaller -- deletes are potentially cheaper, but perhaps | |
2294 | * more importantly, the bigger file is likely the more recent | |
2295 | * one. The deepest deltas are therefore the oldest objects which are | |
2296 | * less susceptible to be accessed often. | |
2297 | */ | |
9668cf59 | 2298 | static int type_size_sort(const void *_a, const void *_b) |
c323ac7d | 2299 | { |
9668cf59 NP |
2300 | const struct object_entry *a = *(struct object_entry **)_a; |
2301 | const struct object_entry *b = *(struct object_entry **)_b; | |
33de80b1 SL |
2302 | const enum object_type a_type = oe_type(a); |
2303 | const enum object_type b_type = oe_type(b); | |
2304 | const unsigned long a_size = SIZE(a); | |
2305 | const unsigned long b_size = SIZE(b); | |
9668cf59 | 2306 | |
fd9b1bae | 2307 | if (a_type > b_type) |
27225f2e | 2308 | return -1; |
fd9b1bae | 2309 | if (a_type < b_type) |
27225f2e | 2310 | return 1; |
b904166c | 2311 | if (a->hash > b->hash) |
7a979d99 | 2312 | return -1; |
b904166c | 2313 | if (a->hash < b->hash) |
7a979d99 | 2314 | return 1; |
b904166c | 2315 | if (a->preferred_base > b->preferred_base) |
c323ac7d | 2316 | return -1; |
b904166c NP |
2317 | if (a->preferred_base < b->preferred_base) |
2318 | return 1; | |
28b8a730 | 2319 | if (use_delta_islands) { |
33de80b1 | 2320 | const int island_cmp = island_delta_cmp(&a->idx.oid, &b->idx.oid); |
28b8a730 JK |
2321 | if (island_cmp) |
2322 | return island_cmp; | |
2323 | } | |
ac77d0c3 | 2324 | if (a_size > b_size) |
b904166c | 2325 | return -1; |
ac77d0c3 | 2326 | if (a_size < b_size) |
c323ac7d | 2327 | return 1; |
b904166c | 2328 | return a < b ? -1 : (a > b); /* newest first */ |
c323ac7d LT |
2329 | } |
2330 | ||
2331 | struct unpacked { | |
2332 | struct object_entry *entry; | |
2333 | void *data; | |
f6c7081a | 2334 | struct delta_index *index; |
5a235b5e | 2335 | unsigned depth; |
c323ac7d LT |
2336 | }; |
2337 | ||
d250626c NP |
2338 | static int delta_cacheable(unsigned long src_size, unsigned long trg_size, |
2339 | unsigned long delta_size) | |
074b2eea MK |
2340 | { |
2341 | if (max_delta_cache_size && delta_cache_size + delta_size > max_delta_cache_size) | |
2342 | return 0; | |
2343 | ||
e3dfddb3 MK |
2344 | if (delta_size < cache_max_small_delta_size) |
2345 | return 1; | |
2346 | ||
074b2eea MK |
2347 | /* cache delta, if objects are large enough compared to delta size */ |
2348 | if ((src_size >> 20) + (trg_size >> 21) > (delta_size >> 10)) | |
2349 | return 1; | |
2350 | ||
2351 | return 0; | |
2352 | } | |
2353 | ||
ffbd51cc | 2354 | /* Protect delta_cache_size */ |
44626dc7 | 2355 | static pthread_mutex_t cache_mutex; |
3c701839 NP |
2356 | #define cache_lock() pthread_mutex_lock(&cache_mutex) |
2357 | #define cache_unlock() pthread_mutex_unlock(&cache_mutex) | |
2358 | ||
ffbd51cc NTND |
2359 | /* |
2360 | * Protect object list partitioning (e.g. struct thread_param) and | |
2361 | * progress_state | |
2362 | */ | |
44626dc7 | 2363 | static pthread_mutex_t progress_mutex; |
8ecce684 NP |
2364 | #define progress_lock() pthread_mutex_lock(&progress_mutex) |
2365 | #define progress_unlock() pthread_mutex_unlock(&progress_mutex) | |
2366 | ||
ffbd51cc NTND |
2367 | /* |
2368 | * Access to struct object_entry is unprotected since each thread owns | |
2369 | * a portion of the main object list. Just don't access object entries | |
2370 | * ahead in the list because they can be stolen and would need | |
2371 | * progress_mutex for protection. | |
2372 | */ | |
8ecce684 | 2373 | |
7d089fb9 ÆAB |
2374 | static inline int oe_size_less_than(struct packing_data *pack, |
2375 | const struct object_entry *lhs, | |
2376 | unsigned long rhs) | |
2377 | { | |
2378 | if (lhs->size_valid) | |
2379 | return lhs->size_ < rhs; | |
2380 | if (rhs < pack->oe_size_limit) /* rhs < 2^x <= lhs ? */ | |
2381 | return 0; | |
2382 | return oe_get_size_slow(pack, lhs) < rhs; | |
2383 | } | |
2384 | ||
2385 | static inline void oe_set_tree_depth(struct packing_data *pack, | |
2386 | struct object_entry *e, | |
2387 | unsigned int tree_depth) | |
2388 | { | |
2389 | if (!pack->tree_depth) | |
2390 | CALLOC_ARRAY(pack->tree_depth, pack->nr_alloc); | |
2391 | pack->tree_depth[e - pack->objects] = tree_depth; | |
2392 | } | |
2393 | ||
ac77d0c3 NTND |
2394 | /* |
2395 | * Return the size of the object without doing any delta | |
2396 | * reconstruction (so non-deltas are true object sizes, but deltas | |
2397 | * return the size of the delta data). | |
2398 | */ | |
2399 | unsigned long oe_get_size_slow(struct packing_data *pack, | |
2400 | const struct object_entry *e) | |
2401 | { | |
2402 | struct packed_git *p; | |
2403 | struct pack_window *w_curs; | |
2404 | unsigned char *buf; | |
2405 | enum object_type type; | |
2406 | unsigned long used, avail, size; | |
2407 | ||
2408 | if (e->type_ != OBJ_OFS_DELTA && e->type_ != OBJ_REF_DELTA) { | |
edb673cf | 2409 | packing_data_lock(&to_pack); |
ad635e82 | 2410 | if (oid_object_info(the_repository, &e->idx.oid, &size) < 0) |
ac77d0c3 NTND |
2411 | die(_("unable to get size of %s"), |
2412 | oid_to_hex(&e->idx.oid)); | |
edb673cf | 2413 | packing_data_unlock(&to_pack); |
ac77d0c3 NTND |
2414 | return size; |
2415 | } | |
2416 | ||
2417 | p = oe_in_pack(pack, e); | |
2418 | if (!p) | |
2419 | BUG("when e->type is a delta, it must belong to a pack"); | |
2420 | ||
edb673cf | 2421 | packing_data_lock(&to_pack); |
ac77d0c3 NTND |
2422 | w_curs = NULL; |
2423 | buf = use_pack(p, &w_curs, e->in_pack_offset, &avail); | |
2424 | used = unpack_object_header_buffer(buf, avail, &type, &size); | |
2425 | if (used == 0) | |
2426 | die(_("unable to parse object header of %s"), | |
2427 | oid_to_hex(&e->idx.oid)); | |
2428 | ||
2429 | unuse_pack(&w_curs); | |
edb673cf | 2430 | packing_data_unlock(&to_pack); |
ac77d0c3 NTND |
2431 | return size; |
2432 | } | |
2433 | ||
f6c7081a | 2434 | static int try_delta(struct unpacked *trg, struct unpacked *src, |
ef0316fc | 2435 | unsigned max_depth, unsigned long *mem_usage) |
c323ac7d | 2436 | { |
f6c7081a NP |
2437 | struct object_entry *trg_entry = trg->entry; |
2438 | struct object_entry *src_entry = src->entry; | |
560b25a8 | 2439 | unsigned long trg_size, src_size, delta_size, sizediff, max_size, sz; |
c83f032e | 2440 | unsigned ref_depth; |
21666f1a | 2441 | enum object_type type; |
c323ac7d LT |
2442 | void *delta_buf; |
2443 | ||
2444 | /* Don't bother doing diffs between different types */ | |
fd9b1bae | 2445 | if (oe_type(trg_entry) != oe_type(src_entry)) |
c323ac7d LT |
2446 | return -1; |
2447 | ||
51d1e83f | 2448 | /* |
15f07e06 JK |
2449 | * We do not bother to try a delta that we discarded on an |
2450 | * earlier try, but only when reusing delta data. Note that | |
2451 | * src_entry that is marked as the preferred_base should always | |
2452 | * be considered, as even if we produce a suboptimal delta against | |
2453 | * it, we will still save the transfer cost, as we already know | |
2454 | * the other side has it and we won't send src_entry at all. | |
51d1e83f | 2455 | */ |
43fa44fa NTND |
2456 | if (reuse_delta && IN_PACK(trg_entry) && |
2457 | IN_PACK(trg_entry) == IN_PACK(src_entry) && | |
15f07e06 | 2458 | !src_entry->preferred_base && |
e9195b58 JH |
2459 | trg_entry->in_pack_type != OBJ_REF_DELTA && |
2460 | trg_entry->in_pack_type != OBJ_OFS_DELTA) | |
51d1e83f LT |
2461 | return 0; |
2462 | ||
898b14ce | 2463 | /* Let's not bust the allowed depth. */ |
5a235b5e | 2464 | if (src->depth >= max_depth) |
d116a45a | 2465 | return 0; |
c323ac7d | 2466 | |
c3b06a69 | 2467 | /* Now some size filtering heuristics. */ |
ac77d0c3 | 2468 | trg_size = SIZE(trg_entry); |
898eba5e | 2469 | if (!DELTA(trg_entry)) { |
41179100 | 2470 | max_size = trg_size/2 - the_hash_algo->rawsz; |
c83f032e NP |
2471 | ref_depth = 1; |
2472 | } else { | |
0aca34e8 | 2473 | max_size = DELTA_SIZE(trg_entry); |
5a235b5e | 2474 | ref_depth = trg->depth; |
c83f032e | 2475 | } |
720fe22d | 2476 | max_size = (uint64_t)max_size * (max_depth - src->depth) / |
c83f032e | 2477 | (max_depth - ref_depth + 1); |
c3b06a69 NP |
2478 | if (max_size == 0) |
2479 | return 0; | |
ac77d0c3 | 2480 | src_size = SIZE(src_entry); |
560b25a8 | 2481 | sizediff = src_size < trg_size ? trg_size - src_size : 0; |
27225f2e | 2482 | if (sizediff >= max_size) |
f527cb8c | 2483 | return 0; |
a1dab41a BD |
2484 | if (trg_size < src_size / 32) |
2485 | return 0; | |
f6c7081a | 2486 | |
28b8a730 JK |
2487 | if (!in_same_island(&trg->entry->idx.oid, &src->entry->idx.oid)) |
2488 | return 0; | |
2489 | ||
560b25a8 NP |
2490 | /* Load data if not already done */ |
2491 | if (!trg->data) { | |
edb673cf | 2492 | packing_data_lock(&to_pack); |
b4f5aca4 | 2493 | trg->data = read_object_file(&trg_entry->idx.oid, &type, &sz); |
edb673cf | 2494 | packing_data_unlock(&to_pack); |
2e3404c3 | 2495 | if (!trg->data) |
f616db6a | 2496 | die(_("object %s cannot be read"), |
e6a492b7 | 2497 | oid_to_hex(&trg_entry->idx.oid)); |
560b25a8 | 2498 | if (sz != trg_size) |
ca473cef TB |
2499 | die(_("object %s inconsistent object length (%"PRIuMAX" vs %"PRIuMAX")"), |
2500 | oid_to_hex(&trg_entry->idx.oid), (uintmax_t)sz, | |
2501 | (uintmax_t)trg_size); | |
ef0316fc | 2502 | *mem_usage += sz; |
560b25a8 NP |
2503 | } |
2504 | if (!src->data) { | |
edb673cf | 2505 | packing_data_lock(&to_pack); |
b4f5aca4 | 2506 | src->data = read_object_file(&src_entry->idx.oid, &type, &sz); |
edb673cf | 2507 | packing_data_unlock(&to_pack); |
71064a95 NP |
2508 | if (!src->data) { |
2509 | if (src_entry->preferred_base) { | |
2510 | static int warned = 0; | |
2511 | if (!warned++) | |
f616db6a | 2512 | warning(_("object %s cannot be read"), |
e6a492b7 | 2513 | oid_to_hex(&src_entry->idx.oid)); |
71064a95 NP |
2514 | /* |
2515 | * Those objects are not included in the | |
2516 | * resulting pack. Be resilient and ignore | |
2517 | * them if they can't be read, in case the | |
2518 | * pack could be created nevertheless. | |
2519 | */ | |
2520 | return 0; | |
2521 | } | |
f616db6a | 2522 | die(_("object %s cannot be read"), |
e6a492b7 | 2523 | oid_to_hex(&src_entry->idx.oid)); |
71064a95 | 2524 | } |
560b25a8 | 2525 | if (sz != src_size) |
ca473cef TB |
2526 | die(_("object %s inconsistent object length (%"PRIuMAX" vs %"PRIuMAX")"), |
2527 | oid_to_hex(&src_entry->idx.oid), (uintmax_t)sz, | |
2528 | (uintmax_t)src_size); | |
ef0316fc | 2529 | *mem_usage += sz; |
560b25a8 NP |
2530 | } |
2531 | if (!src->index) { | |
2532 | src->index = create_delta_index(src->data, src_size); | |
a588d88a MK |
2533 | if (!src->index) { |
2534 | static int warned = 0; | |
2535 | if (!warned++) | |
f616db6a | 2536 | warning(_("suboptimal pack - out of memory")); |
a588d88a MK |
2537 | return 0; |
2538 | } | |
ef0316fc | 2539 | *mem_usage += sizeof_delta_index(src->index); |
560b25a8 NP |
2540 | } |
2541 | ||
2542 | delta_buf = create_delta(src->index, trg->data, trg_size, &delta_size, max_size); | |
c323ac7d | 2543 | if (!delta_buf) |
75c42d8c | 2544 | return 0; |
f6c7081a | 2545 | |
898eba5e | 2546 | if (DELTA(trg_entry)) { |
848d732c | 2547 | /* Prefer only shallower same-sized deltas. */ |
0aca34e8 | 2548 | if (delta_size == DELTA_SIZE(trg_entry) && |
5a235b5e | 2549 | src->depth + 1 >= trg->depth) { |
848d732c BD |
2550 | free(delta_buf); |
2551 | return 0; | |
2552 | } | |
074b2eea | 2553 | } |
9e2d57a0 | 2554 | |
3c701839 NP |
2555 | /* |
2556 | * Handle memory allocation outside of the cache | |
2557 | * accounting lock. Compiler will optimize the strangeness | |
7eb151d6 | 2558 | * away when NO_PTHREADS is defined. |
3c701839 | 2559 | */ |
8e0f7003 | 2560 | free(trg_entry->delta_data); |
3c701839 | 2561 | cache_lock(); |
9e2d57a0 | 2562 | if (trg_entry->delta_data) { |
0aca34e8 | 2563 | delta_cache_size -= DELTA_SIZE(trg_entry); |
9e2d57a0 NP |
2564 | trg_entry->delta_data = NULL; |
2565 | } | |
d250626c | 2566 | if (delta_cacheable(src_size, trg_size, delta_size)) { |
b7a28f78 | 2567 | delta_cache_size += delta_size; |
3c701839 NP |
2568 | cache_unlock(); |
2569 | trg_entry->delta_data = xrealloc(delta_buf, delta_size); | |
2570 | } else { | |
2571 | cache_unlock(); | |
074b2eea | 2572 | free(delta_buf); |
3c701839 NP |
2573 | } |
2574 | ||
898eba5e | 2575 | SET_DELTA(trg_entry, src_entry); |
0aca34e8 | 2576 | SET_DELTA_SIZE(trg_entry, delta_size); |
b7a28f78 NP |
2577 | trg->depth = src->depth + 1; |
2578 | ||
f6c7081a | 2579 | return 1; |
c323ac7d LT |
2580 | } |
2581 | ||
898b14ce | 2582 | static unsigned int check_delta_limit(struct object_entry *me, unsigned int n) |
b2504a0d | 2583 | { |
898eba5e | 2584 | struct object_entry *child = DELTA_CHILD(me); |
898b14ce NP |
2585 | unsigned int m = n; |
2586 | while (child) { | |
33de80b1 | 2587 | const unsigned int c = check_delta_limit(child, n + 1); |
898b14ce NP |
2588 | if (m < c) |
2589 | m = c; | |
898eba5e | 2590 | child = DELTA_SIBLING(child); |
898b14ce NP |
2591 | } |
2592 | return m; | |
b2504a0d NP |
2593 | } |
2594 | ||
75ad235c | 2595 | static unsigned long free_unpacked(struct unpacked *n) |
a97773ce | 2596 | { |
ef0316fc | 2597 | unsigned long freed_mem = sizeof_delta_index(n->index); |
a97773ce BD |
2598 | free_delta_index(n->index); |
2599 | n->index = NULL; | |
2600 | if (n->data) { | |
ac77d0c3 | 2601 | freed_mem += SIZE(n->entry); |
6a83d902 | 2602 | FREE_AND_NULL(n->data); |
a97773ce BD |
2603 | } |
2604 | n->entry = NULL; | |
7d7baa5e | 2605 | n->depth = 0; |
ef0316fc | 2606 | return freed_mem; |
a97773ce BD |
2607 | } |
2608 | ||
384b32c0 | 2609 | static void find_deltas(struct object_entry **list, unsigned *list_size, |
e334977d | 2610 | int window, int depth, unsigned *processed) |
c323ac7d | 2611 | { |
384b32c0 | 2612 | uint32_t i, idx = 0, count = 0; |
7cadf491 | 2613 | struct unpacked *array; |
ef0316fc | 2614 | unsigned long mem_usage = 0; |
c323ac7d | 2615 | |
ca56dadb | 2616 | CALLOC_ARRAY(array, window); |
21fcd1bd | 2617 | |
384b32c0 | 2618 | for (;;) { |
421b488a | 2619 | struct object_entry *entry; |
c323ac7d | 2620 | struct unpacked *n = array + idx; |
ef0316fc | 2621 | int j, max_depth, best_base = -1; |
c323ac7d | 2622 | |
384b32c0 NP |
2623 | progress_lock(); |
2624 | if (!*list_size) { | |
2625 | progress_unlock(); | |
2626 | break; | |
2627 | } | |
421b488a | 2628 | entry = *list++; |
384b32c0 NP |
2629 | (*list_size)--; |
2630 | if (!entry->preferred_base) { | |
2631 | (*processed)++; | |
2632 | display_progress(progress_state, *processed); | |
2633 | } | |
2634 | progress_unlock(); | |
2635 | ||
ef0316fc | 2636 | mem_usage -= free_unpacked(n); |
c323ac7d | 2637 | n->entry = entry; |
ab7cd7bb | 2638 | |
a97773ce | 2639 | while (window_memory_limit && |
ef0316fc | 2640 | mem_usage > window_memory_limit && |
a97773ce | 2641 | count > 1) { |
33de80b1 | 2642 | const uint32_t tail = (idx + window - count) % window; |
75ad235c | 2643 | mem_usage -= free_unpacked(array + tail); |
a97773ce BD |
2644 | count--; |
2645 | } | |
2646 | ||
75d39853 NP |
2647 | /* We do not compute delta to *create* objects we are not |
2648 | * going to pack. | |
2649 | */ | |
2650 | if (entry->preferred_base) | |
2651 | goto next; | |
2652 | ||
898b14ce NP |
2653 | /* |
2654 | * If the current object is at pack edge, take the depth the | |
2655 | * objects that depend on the current object into account | |
2656 | * otherwise they would become too deep. | |
2657 | */ | |
2658 | max_depth = depth; | |
898eba5e | 2659 | if (DELTA_CHILD(entry)) { |
898b14ce NP |
2660 | max_depth -= check_delta_limit(entry, 0); |
2661 | if (max_depth <= 0) | |
2662 | goto next; | |
2663 | } | |
2664 | ||
78817c15 LT |
2665 | j = window; |
2666 | while (--j > 0) { | |
77639870 | 2667 | int ret; |
7cadf491 | 2668 | uint32_t other_idx = idx + j; |
c323ac7d | 2669 | struct unpacked *m; |
78817c15 LT |
2670 | if (other_idx >= window) |
2671 | other_idx -= window; | |
c323ac7d LT |
2672 | m = array + other_idx; |
2673 | if (!m->entry) | |
2674 | break; | |
ef0316fc | 2675 | ret = try_delta(n, m, max_depth, &mem_usage); |
77639870 | 2676 | if (ret < 0) |
c323ac7d | 2677 | break; |
77639870 JH |
2678 | else if (ret > 0) |
2679 | best_base = other_idx; | |
c323ac7d | 2680 | } |
898b14ce | 2681 | |
ed4a9031 NP |
2682 | /* |
2683 | * If we decided to cache the delta data, then it is best | |
2684 | * to compress it right away. First because we have to do | |
2685 | * it anyway, and doing it here while we're threaded will | |
2686 | * save a lot of time in the non threaded write phase, | |
2687 | * as well as allow for caching more deltas within | |
2688 | * the same cache size limit. | |
2689 | * ... | |
2690 | * But only if not writing to stdout, since in that case | |
2691 | * the network is most likely throttling writes anyway, | |
2692 | * and therefore it is best to go to the write phase ASAP | |
2693 | * instead, as we can afford spending more time compressing | |
2694 | * between writes at that moment. | |
2695 | */ | |
2696 | if (entry->delta_data && !pack_to_stdout) { | |
0cb3c142 NTND |
2697 | unsigned long size; |
2698 | ||
0aca34e8 | 2699 | size = do_compress(&entry->delta_data, DELTA_SIZE(entry)); |
0cb3c142 NTND |
2700 | if (size < (1U << OE_Z_DELTA_BITS)) { |
2701 | entry->z_delta_size = size; | |
2702 | cache_lock(); | |
0aca34e8 | 2703 | delta_cache_size -= DELTA_SIZE(entry); |
0cb3c142 NTND |
2704 | delta_cache_size += entry->z_delta_size; |
2705 | cache_unlock(); | |
2706 | } else { | |
2707 | FREE_AND_NULL(entry->delta_data); | |
2708 | entry->z_delta_size = 0; | |
2709 | } | |
ed4a9031 NP |
2710 | } |
2711 | ||
70ca1a3f JH |
2712 | /* if we made n a delta, and if n is already at max |
2713 | * depth, leaving it in the window is pointless. we | |
2714 | * should evict it first. | |
70ca1a3f | 2715 | */ |
898eba5e | 2716 | if (DELTA(entry) && max_depth <= n->depth) |
70ca1a3f | 2717 | continue; |
ff45715c | 2718 | |
77639870 JH |
2719 | /* |
2720 | * Move the best delta base up in the window, after the | |
2721 | * currently deltified object, to keep it longer. It will | |
2722 | * be the first base object to be attempted next. | |
2723 | */ | |
898eba5e | 2724 | if (DELTA(entry)) { |
77639870 JH |
2725 | struct unpacked swap = array[best_base]; |
2726 | int dist = (window + idx - best_base) % window; | |
2727 | int dst = best_base; | |
2728 | while (dist--) { | |
2729 | int src = (dst + 1) % window; | |
2730 | array[dst] = array[src]; | |
2731 | dst = src; | |
2732 | } | |
2733 | array[dst] = swap; | |
2734 | } | |
2735 | ||
898b14ce | 2736 | next: |
521a4f4c | 2737 | idx++; |
a97773ce BD |
2738 | if (count + 1 < window) |
2739 | count++; | |
521a4f4c LT |
2740 | if (idx >= window) |
2741 | idx = 0; | |
384b32c0 | 2742 | } |
adee7bdf | 2743 | |
f6c7081a | 2744 | for (i = 0; i < window; ++i) { |
ff45715c | 2745 | free_delta_index(array[i].index); |
adee7bdf | 2746 | free(array[i].data); |
f6c7081a | 2747 | } |
adee7bdf | 2748 | free(array); |
c323ac7d LT |
2749 | } |
2750 | ||
50f22ada | 2751 | /* |
ffbd51cc NTND |
2752 | * The main object list is split into smaller lists, each is handed to |
2753 | * one worker. | |
2754 | * | |
50f22ada JS |
2755 | * The main thread waits on the condition that (at least) one of the workers |
2756 | * has stopped working (which is indicated in the .working member of | |
2757 | * struct thread_params). | |
ffbd51cc | 2758 | * |
50f22ada JS |
2759 | * When a work thread has completed its work, it sets .working to 0 and |
2760 | * signals the main thread and waits on the condition that .data_ready | |
2761 | * becomes 1. | |
ffbd51cc NTND |
2762 | * |
2763 | * The main thread steals half of the work from the worker that has | |
2764 | * most work left to hand it to the idle worker. | |
50f22ada JS |
2765 | */ |
2766 | ||
8ecce684 NP |
2767 | struct thread_params { |
2768 | pthread_t thread; | |
2769 | struct object_entry **list; | |
2770 | unsigned list_size; | |
384b32c0 | 2771 | unsigned remaining; |
8ecce684 NP |
2772 | int window; |
2773 | int depth; | |
50f22ada JS |
2774 | int working; |
2775 | int data_ready; | |
2776 | pthread_mutex_t mutex; | |
2777 | pthread_cond_t cond; | |
8ecce684 NP |
2778 | unsigned *processed; |
2779 | }; | |
2780 | ||
44626dc7 AH |
2781 | static pthread_cond_t progress_cond; |
2782 | ||
2783 | /* | |
2784 | * Mutex and conditional variable can't be statically-initialized on Windows. | |
2785 | */ | |
2786 | static void init_threaded_search(void) | |
2787 | { | |
44626dc7 AH |
2788 | pthread_mutex_init(&cache_mutex, NULL); |
2789 | pthread_mutex_init(&progress_mutex, NULL); | |
2790 | pthread_cond_init(&progress_cond, NULL); | |
2791 | } | |
2792 | ||
2793 | static void cleanup_threaded_search(void) | |
2794 | { | |
2795 | pthread_cond_destroy(&progress_cond); | |
44626dc7 AH |
2796 | pthread_mutex_destroy(&cache_mutex); |
2797 | pthread_mutex_destroy(&progress_mutex); | |
2798 | } | |
c2a33679 | 2799 | |
8ecce684 NP |
2800 | static void *threaded_find_deltas(void *arg) |
2801 | { | |
c2a33679 NP |
2802 | struct thread_params *me = arg; |
2803 | ||
0c2ad00b | 2804 | progress_lock(); |
50f22ada | 2805 | while (me->remaining) { |
0c2ad00b MÅ |
2806 | progress_unlock(); |
2807 | ||
384b32c0 | 2808 | find_deltas(me->list, &me->remaining, |
c2a33679 | 2809 | me->window, me->depth, me->processed); |
50f22ada JS |
2810 | |
2811 | progress_lock(); | |
2812 | me->working = 0; | |
2813 | pthread_cond_signal(&progress_cond); | |
2814 | progress_unlock(); | |
2815 | ||
2816 | /* | |
2817 | * We must not set ->data_ready before we wait on the | |
2818 | * condition because the main thread may have set it to 1 | |
2819 | * before we get here. In order to be sure that new | |
2820 | * work is available if we see 1 in ->data_ready, it | |
2821 | * was initialized to 0 before this thread was spawned | |
2822 | * and we reset it to 0 right away. | |
2823 | */ | |
2824 | pthread_mutex_lock(&me->mutex); | |
2825 | while (!me->data_ready) | |
2826 | pthread_cond_wait(&me->cond, &me->mutex); | |
2827 | me->data_ready = 0; | |
2828 | pthread_mutex_unlock(&me->mutex); | |
0c2ad00b MÅ |
2829 | |
2830 | progress_lock(); | |
c2a33679 | 2831 | } |
0c2ad00b | 2832 | progress_unlock(); |
50f22ada JS |
2833 | /* leave ->working 1 so that this doesn't get more work assigned */ |
2834 | return NULL; | |
8ecce684 NP |
2835 | } |
2836 | ||
8ecce684 NP |
2837 | static void ll_find_deltas(struct object_entry **list, unsigned list_size, |
2838 | int window, int depth, unsigned *processed) | |
2839 | { | |
dcda3614 | 2840 | struct thread_params *p; |
384b32c0 | 2841 | int i, ret, active_threads = 0; |
c2a33679 | 2842 | |
44626dc7 AH |
2843 | init_threaded_search(); |
2844 | ||
367f4a43 | 2845 | if (delta_search_threads <= 1) { |
384b32c0 | 2846 | find_deltas(list, &list_size, window, depth, processed); |
44626dc7 | 2847 | cleanup_threaded_search(); |
367f4a43 NP |
2848 | return; |
2849 | } | |
43cc2b42 | 2850 | if (progress > pack_to_stdout) |
f616db6a | 2851 | fprintf_ln(stderr, _("Delta compression using up to %d threads"), |
1a07e59c | 2852 | delta_search_threads); |
ca56dadb | 2853 | CALLOC_ARRAY(p, delta_search_threads); |
367f4a43 | 2854 | |
50f22ada | 2855 | /* Partition the work amongst work threads. */ |
367f4a43 | 2856 | for (i = 0; i < delta_search_threads; i++) { |
50f22ada JS |
2857 | unsigned sub_size = list_size / (delta_search_threads - i); |
2858 | ||
bf874896 NP |
2859 | /* don't use too small segments or no deltas will be found */ |
2860 | if (sub_size < 2*window && i+1 < delta_search_threads) | |
2861 | sub_size = 0; | |
2862 | ||
8ecce684 NP |
2863 | p[i].window = window; |
2864 | p[i].depth = depth; | |
2865 | p[i].processed = processed; | |
50f22ada JS |
2866 | p[i].working = 1; |
2867 | p[i].data_ready = 0; | |
c2a33679 | 2868 | |
59921b4b | 2869 | /* try to split chunks on "path" boundaries */ |
6fc74703 NP |
2870 | while (sub_size && sub_size < list_size && |
2871 | list[sub_size]->hash && | |
384b32c0 NP |
2872 | list[sub_size]->hash == list[sub_size-1]->hash) |
2873 | sub_size++; | |
2874 | ||
50f22ada JS |
2875 | p[i].list = list; |
2876 | p[i].list_size = sub_size; | |
2877 | p[i].remaining = sub_size; | |
59921b4b | 2878 | |
384b32c0 NP |
2879 | list += sub_size; |
2880 | list_size -= sub_size; | |
2881 | } | |
2882 | ||
50f22ada JS |
2883 | /* Start work threads. */ |
2884 | for (i = 0; i < delta_search_threads; i++) { | |
2885 | if (!p[i].list_size) | |
2886 | continue; | |
68e6a4f8 JS |
2887 | pthread_mutex_init(&p[i].mutex, NULL); |
2888 | pthread_cond_init(&p[i].cond, NULL); | |
50f22ada JS |
2889 | ret = pthread_create(&p[i].thread, NULL, |
2890 | threaded_find_deltas, &p[i]); | |
2891 | if (ret) | |
f616db6a | 2892 | die(_("unable to create thread: %s"), strerror(ret)); |
50f22ada JS |
2893 | active_threads++; |
2894 | } | |
2895 | ||
384b32c0 NP |
2896 | /* |
2897 | * Now let's wait for work completion. Each time a thread is done | |
2898 | * with its work, we steal half of the remaining work from the | |
2899 | * thread with the largest number of unprocessed objects and give | |
2900 | * it to that newly idle thread. This ensure good load balancing | |
2901 | * until the remaining object list segments are simply too short | |
2902 | * to be worth splitting anymore. | |
2903 | */ | |
50f22ada JS |
2904 | while (active_threads) { |
2905 | struct thread_params *target = NULL; | |
384b32c0 NP |
2906 | struct thread_params *victim = NULL; |
2907 | unsigned sub_size = 0; | |
384b32c0 NP |
2908 | |
2909 | progress_lock(); | |
50f22ada JS |
2910 | for (;;) { |
2911 | for (i = 0; !target && i < delta_search_threads; i++) | |
2912 | if (!p[i].working) | |
2913 | target = &p[i]; | |
2914 | if (target) | |
2915 | break; | |
2916 | pthread_cond_wait(&progress_cond, &progress_mutex); | |
2917 | } | |
2918 | ||
384b32c0 NP |
2919 | for (i = 0; i < delta_search_threads; i++) |
2920 | if (p[i].remaining > 2*window && | |
2921 | (!victim || victim->remaining < p[i].remaining)) | |
2922 | victim = &p[i]; | |
2923 | if (victim) { | |
2924 | sub_size = victim->remaining / 2; | |
2925 | list = victim->list + victim->list_size - sub_size; | |
2926 | while (sub_size && list[0]->hash && | |
2927 | list[0]->hash == list[-1]->hash) { | |
2928 | list++; | |
2929 | sub_size--; | |
2930 | } | |
eb9688ff NP |
2931 | if (!sub_size) { |
2932 | /* | |
2933 | * It is possible for some "paths" to have | |
2934 | * so many objects that no hash boundary | |
2935 | * might be found. Let's just steal the | |
2936 | * exact half in that case. | |
2937 | */ | |
2938 | sub_size = victim->remaining / 2; | |
2939 | list -= sub_size; | |
2940 | } | |
384b32c0 NP |
2941 | target->list = list; |
2942 | victim->list_size -= sub_size; | |
2943 | victim->remaining -= sub_size; | |
2944 | } | |
384b32c0 NP |
2945 | target->list_size = sub_size; |
2946 | target->remaining = sub_size; | |
50f22ada JS |
2947 | target->working = 1; |
2948 | progress_unlock(); | |
2949 | ||
2950 | pthread_mutex_lock(&target->mutex); | |
2951 | target->data_ready = 1; | |
2952 | pthread_cond_signal(&target->cond); | |
2953 | pthread_mutex_unlock(&target->mutex); | |
c2a33679 | 2954 | |
384b32c0 | 2955 | if (!sub_size) { |
b81d9af7 | 2956 | pthread_join(target->thread, NULL); |
50f22ada JS |
2957 | pthread_cond_destroy(&target->cond); |
2958 | pthread_mutex_destroy(&target->mutex); | |
384b32c0 | 2959 | active_threads--; |
c2a33679 | 2960 | } |
50f22ada | 2961 | } |
44626dc7 | 2962 | cleanup_threaded_search(); |
dcda3614 | 2963 | free(p); |
8ecce684 NP |
2964 | } |
2965 | ||
ff483026 JK |
2966 | static int obj_is_packed(const struct object_id *oid) |
2967 | { | |
a14aebea | 2968 | return packlist_find(&to_pack, oid) || |
92fb0db9 JK |
2969 | (reuse_packfile_bitmap && |
2970 | bitmap_walk_contains(bitmap_git, reuse_packfile_bitmap, oid)); | |
ff483026 JK |
2971 | } |
2972 | ||
b773ddea JK |
2973 | static void add_tag_chain(const struct object_id *oid) |
2974 | { | |
2975 | struct tag *tag; | |
2976 | ||
2977 | /* | |
2978 | * We catch duplicates already in add_object_entry(), but we'd | |
2979 | * prefer to do this extra check to avoid having to parse the | |
2980 | * tag at all if we already know that it's being packed (e.g., if | |
2981 | * it was included via bitmaps, we would not have parsed it | |
2982 | * previously). | |
2983 | */ | |
ff483026 | 2984 | if (obj_is_packed(oid)) |
b773ddea JK |
2985 | return; |
2986 | ||
ce71efb7 | 2987 | tag = lookup_tag(the_repository, oid); |
b773ddea JK |
2988 | while (1) { |
2989 | if (!tag || parse_tag(tag) || !tag->tagged) | |
f616db6a | 2990 | die(_("unable to pack objects reachable from tag %s"), |
b773ddea JK |
2991 | oid_to_hex(oid)); |
2992 | ||
188960b4 | 2993 | add_object_entry(&tag->object.oid, OBJ_TAG, NULL, 0); |
b773ddea JK |
2994 | |
2995 | if (tag->tagged->type != OBJ_TAG) | |
2996 | return; | |
2997 | ||
2998 | tag = (struct tag *)tag->tagged; | |
2999 | } | |
3000 | } | |
3001 | ||
be18153b | 3002 | static int add_ref_tag(const char *tag, const struct object_id *oid, int flag, void *cb_data) |
f0a24aa5 | 3003 | { |
d155254c | 3004 | struct object_id peeled; |
f0a24aa5 | 3005 | |
77db59c2 | 3006 | if (!peel_iterated_oid(oid, &peeled) && obj_is_packed(&peeled)) |
b773ddea | 3007 | add_tag_chain(oid); |
f0a24aa5 SP |
3008 | return 0; |
3009 | } | |
3010 | ||
f3123c4a JH |
3011 | static void prepare_pack(int window, int depth) |
3012 | { | |
9668cf59 | 3013 | struct object_entry **delta_list; |
6e1c2344 RJ |
3014 | uint32_t i, nr_deltas; |
3015 | unsigned n; | |
9668cf59 | 3016 | |
28b8a730 | 3017 | if (use_delta_islands) |
385cb64f | 3018 | resolve_tree_islands(the_repository, progress, &to_pack); |
28b8a730 | 3019 | |
3f9ac8d2 | 3020 | get_object_details(); |
9668cf59 | 3021 | |
0e8189e2 NP |
3022 | /* |
3023 | * If we're locally repacking then we need to be doubly careful | |
3024 | * from now on in order to make sure no stealth corruption gets | |
3025 | * propagated to the new pack. Clients receiving streamed packs | |
3026 | * should validate everything they get anyway so no need to incur | |
3027 | * the additional cost here in that case. | |
3028 | */ | |
3029 | if (!pack_to_stdout) | |
3030 | do_check_packed_object_crc = 1; | |
3031 | ||
2834bc27 | 3032 | if (!to_pack.nr_objects || !window || !depth) |
9668cf59 NP |
3033 | return; |
3034 | ||
b32fa95f | 3035 | ALLOC_ARRAY(delta_list, to_pack.nr_objects); |
75d39853 NP |
3036 | nr_deltas = n = 0; |
3037 | ||
2834bc27 VM |
3038 | for (i = 0; i < to_pack.nr_objects; i++) { |
3039 | struct object_entry *entry = to_pack.objects + i; | |
75d39853 | 3040 | |
898eba5e | 3041 | if (DELTA(entry)) |
75d39853 | 3042 | /* This happens if we decided to reuse existing |
a7de7130 | 3043 | * delta from a pack. "reuse_delta &&" is implied. |
75d39853 NP |
3044 | */ |
3045 | continue; | |
3046 | ||
ac77d0c3 NTND |
3047 | if (!entry->type_valid || |
3048 | oe_size_less_than(&to_pack, entry, 50)) | |
75d39853 NP |
3049 | continue; |
3050 | ||
3051 | if (entry->no_try_delta) | |
3052 | continue; | |
3053 | ||
6d6f9cdd | 3054 | if (!entry->preferred_base) { |
75d39853 | 3055 | nr_deltas++; |
fd9b1bae | 3056 | if (oe_type(entry) < 0) |
f616db6a | 3057 | die(_("unable to get type of object %s"), |
e6a492b7 | 3058 | oid_to_hex(&entry->idx.oid)); |
eede9f42 | 3059 | } else { |
fd9b1bae | 3060 | if (oe_type(entry) < 0) { |
eede9f42 NP |
3061 | /* |
3062 | * This object is not found, but we | |
3063 | * don't have to include it anyway. | |
3064 | */ | |
3065 | continue; | |
3066 | } | |
6d6f9cdd | 3067 | } |
75d39853 NP |
3068 | |
3069 | delta_list[n++] = entry; | |
3070 | } | |
3071 | ||
2f8b8947 | 3072 | if (nr_deltas && n > 1) { |
e334977d | 3073 | unsigned nr_done = 0; |
bb514de3 | 3074 | |
e334977d | 3075 | if (progress) |
754dbc43 | 3076 | progress_state = start_progress(_("Compressing objects"), |
dc6a0757 | 3077 | nr_deltas); |
9ed0d8d6 | 3078 | QSORT(delta_list, n, type_size_sort); |
8ecce684 | 3079 | ll_find_deltas(delta_list, n, window+1, depth, &nr_done); |
4d4fcc54 | 3080 | stop_progress(&progress_state); |
e334977d | 3081 | if (nr_done != nr_deltas) |
f616db6a | 3082 | die(_("inconsistency with delta count")); |
75d39853 | 3083 | } |
9668cf59 | 3084 | free(delta_list); |
f3123c4a JH |
3085 | } |
3086 | ||
ef90d6d4 | 3087 | static int git_pack_config(const char *k, const char *v, void *cb) |
4812a93a | 3088 | { |
eeefa7c9 | 3089 | if (!strcmp(k, "pack.window")) { |
4812a93a JK |
3090 | window = git_config_int(k, v); |
3091 | return 0; | |
3092 | } | |
a97773ce BD |
3093 | if (!strcmp(k, "pack.windowmemory")) { |
3094 | window_memory_limit = git_config_ulong(k, v); | |
3095 | return 0; | |
3096 | } | |
3097 | if (!strcmp(k, "pack.depth")) { | |
842aaf93 TT |
3098 | depth = git_config_int(k, v); |
3099 | return 0; | |
3100 | } | |
074b2eea MK |
3101 | if (!strcmp(k, "pack.deltacachesize")) { |
3102 | max_delta_cache_size = git_config_int(k, v); | |
3103 | return 0; | |
3104 | } | |
e3dfddb3 MK |
3105 | if (!strcmp(k, "pack.deltacachelimit")) { |
3106 | cache_max_small_delta_size = git_config_int(k, v); | |
3107 | return 0; | |
3108 | } | |
ae4f07fb VM |
3109 | if (!strcmp(k, "pack.writebitmaphashcache")) { |
3110 | if (git_config_bool(k, v)) | |
3111 | write_bitmap_options |= BITMAP_OPT_HASH_CACHE; | |
3112 | else | |
3113 | write_bitmap_options &= ~BITMAP_OPT_HASH_CACHE; | |
3114 | } | |
6b8fda2d | 3115 | if (!strcmp(k, "pack.usebitmaps")) { |
645c432d | 3116 | use_bitmap_index_default = git_config_bool(k, v); |
6b8fda2d VM |
3117 | return 0; |
3118 | } | |
e704fc79 JK |
3119 | if (!strcmp(k, "pack.allowpackreuse")) { |
3120 | allow_pack_reuse = git_config_bool(k, v); | |
3121 | return 0; | |
3122 | } | |
693b86ff NP |
3123 | if (!strcmp(k, "pack.threads")) { |
3124 | delta_search_threads = git_config_int(k, v); | |
833e3df1 | 3125 | if (delta_search_threads < 0) |
f616db6a | 3126 | die(_("invalid number of threads specified (%d)"), |
693b86ff | 3127 | delta_search_threads); |
9c897c5c | 3128 | if (!HAVE_THREADS && delta_search_threads != 1) { |
f616db6a | 3129 | warning(_("no threads support, ignoring %s"), k); |
2e96d815 ÆAB |
3130 | delta_search_threads = 0; |
3131 | } | |
693b86ff NP |
3132 | return 0; |
3133 | } | |
4d00bda2 | 3134 | if (!strcmp(k, "pack.indexversion")) { |
ebcfb379 JH |
3135 | pack_idx_opts.version = git_config_int(k, v); |
3136 | if (pack_idx_opts.version > 2) | |
f616db6a | 3137 | die(_("bad pack.indexversion=%"PRIu32), |
ebcfb379 | 3138 | pack_idx_opts.version); |
4d00bda2 NP |
3139 | return 0; |
3140 | } | |
c9773343 TB |
3141 | if (!strcmp(k, "pack.writereverseindex")) { |
3142 | if (git_config_bool(k, v)) | |
3143 | pack_idx_opts.flags |= WRITE_REV; | |
3144 | else | |
3145 | pack_idx_opts.flags &= ~WRITE_REV; | |
3146 | return 0; | |
3147 | } | |
dd4b732d JT |
3148 | if (!strcmp(k, "uploadpack.blobpackfileuri")) { |
3149 | struct configured_exclusion *ex = xmalloc(sizeof(*ex)); | |
3150 | const char *oid_end, *pack_end; | |
3151 | /* | |
3152 | * Stores the pack hash. This is not a true object ID, but is | |
3153 | * of the same form. | |
3154 | */ | |
3155 | struct object_id pack_hash; | |
3156 | ||
3157 | if (parse_oid_hex(v, &ex->e.oid, &oid_end) || | |
3158 | *oid_end != ' ' || | |
3159 | parse_oid_hex(oid_end + 1, &pack_hash, &pack_end) || | |
3160 | *pack_end != ' ') | |
3161 | die(_("value of uploadpack.blobpackfileuri must be " | |
3162 | "of the form '<object-hash> <pack-hash> <uri>' (got '%s')"), v); | |
3163 | if (oidmap_get(&configured_exclusions, &ex->e.oid)) | |
3164 | die(_("object already configured in another " | |
3165 | "uploadpack.blobpackfileuri (got '%s')"), v); | |
3166 | ex->pack_hash_hex = xcalloc(1, pack_end - oid_end); | |
3167 | memcpy(ex->pack_hash_hex, oid_end + 1, pack_end - oid_end - 1); | |
3168 | ex->uri = xstrdup(pack_end + 1); | |
3169 | oidmap_put(&configured_exclusions, ex); | |
3170 | } | |
ef90d6d4 | 3171 | return git_default_config(k, v, cb); |
4812a93a JK |
3172 | } |
3173 | ||
339bce27 TB |
3174 | /* Counters for trace2 output when in --stdin-packs mode. */ |
3175 | static int stdin_packs_found_nr; | |
3176 | static int stdin_packs_hints_nr; | |
3177 | ||
3178 | static int add_object_entry_from_pack(const struct object_id *oid, | |
3179 | struct packed_git *p, | |
3180 | uint32_t pos, | |
3181 | void *_data) | |
3182 | { | |
3183 | struct rev_info *revs = _data; | |
3184 | struct object_info oi = OBJECT_INFO_INIT; | |
3185 | off_t ofs; | |
3186 | enum object_type type; | |
3187 | ||
3188 | display_progress(progress_state, ++nr_seen); | |
3189 | ||
3190 | if (have_duplicate_entry(oid, 0)) | |
3191 | return 0; | |
3192 | ||
3193 | ofs = nth_packed_object_offset(p, pos); | |
3194 | if (!want_object_in_pack(oid, 0, &p, &ofs)) | |
3195 | return 0; | |
3196 | ||
3197 | oi.typep = &type; | |
3198 | if (packed_object_info(the_repository, p, ofs, &oi) < 0) | |
3199 | die(_("could not get type of object %s in pack %s"), | |
3200 | oid_to_hex(oid), p->pack_name); | |
3201 | else if (type == OBJ_COMMIT) { | |
3202 | /* | |
3203 | * commits in included packs are used as starting points for the | |
3204 | * subsequent revision walk | |
3205 | */ | |
3206 | add_pending_oid(revs, NULL, oid, 0); | |
3207 | } | |
3208 | ||
3209 | stdin_packs_found_nr++; | |
3210 | ||
3211 | create_object_entry(oid, type, 0, 0, 0, p, ofs); | |
3212 | ||
3213 | return 0; | |
3214 | } | |
3215 | ||
3216 | static void show_commit_pack_hint(struct commit *commit, void *_data) | |
3217 | { | |
3218 | /* nothing to do; commits don't have a namehash */ | |
3219 | } | |
3220 | ||
3221 | static void show_object_pack_hint(struct object *object, const char *name, | |
3222 | void *_data) | |
3223 | { | |
3224 | struct object_entry *oe = packlist_find(&to_pack, &object->oid); | |
3225 | if (!oe) | |
3226 | return; | |
3227 | ||
3228 | /* | |
3229 | * Our 'to_pack' list was constructed by iterating all objects packed in | |
3230 | * included packs, and so doesn't have a non-zero hash field that you | |
3231 | * would typically pick up during a reachability traversal. | |
3232 | * | |
3233 | * Make a best-effort attempt to fill in the ->hash and ->no_try_delta | |
3234 | * here using a now in order to perhaps improve the delta selection | |
3235 | * process. | |
3236 | */ | |
3237 | oe->hash = pack_name_hash(name); | |
3238 | oe->no_try_delta = name && no_try_delta(name); | |
3239 | ||
3240 | stdin_packs_hints_nr++; | |
3241 | } | |
3242 | ||
3243 | static int pack_mtime_cmp(const void *_a, const void *_b) | |
3244 | { | |
3245 | struct packed_git *a = ((const struct string_list_item*)_a)->util; | |
3246 | struct packed_git *b = ((const struct string_list_item*)_b)->util; | |
3247 | ||
3248 | /* | |
3249 | * order packs by descending mtime so that objects are laid out | |
3250 | * roughly as newest-to-oldest | |
3251 | */ | |
3252 | if (a->mtime < b->mtime) | |
3253 | return 1; | |
3254 | else if (b->mtime < a->mtime) | |
3255 | return -1; | |
3256 | else | |
3257 | return 0; | |
3258 | } | |
3259 | ||
3260 | static void read_packs_list_from_stdin(void) | |
3261 | { | |
3262 | struct strbuf buf = STRBUF_INIT; | |
3263 | struct string_list include_packs = STRING_LIST_INIT_DUP; | |
3264 | struct string_list exclude_packs = STRING_LIST_INIT_DUP; | |
3265 | struct string_list_item *item = NULL; | |
3266 | ||
3267 | struct packed_git *p; | |
3268 | struct rev_info revs; | |
3269 | ||
3270 | repo_init_revisions(the_repository, &revs, NULL); | |
3271 | /* | |
3272 | * Use a revision walk to fill in the namehash of objects in the include | |
3273 | * packs. To save time, we'll avoid traversing through objects that are | |
3274 | * in excluded packs. | |
3275 | * | |
3276 | * That may cause us to avoid populating all of the namehash fields of | |
3277 | * all included objects, but our goal is best-effort, since this is only | |
3278 | * an optimization during delta selection. | |
3279 | */ | |
3280 | revs.no_kept_objects = 1; | |
3281 | revs.keep_pack_cache_flags |= IN_CORE_KEEP_PACKS; | |
3282 | revs.blob_objects = 1; | |
3283 | revs.tree_objects = 1; | |
3284 | revs.tag_objects = 1; | |
14e7b834 | 3285 | revs.ignore_missing_links = 1; |
339bce27 TB |
3286 | |
3287 | while (strbuf_getline(&buf, stdin) != EOF) { | |
3288 | if (!buf.len) | |
3289 | continue; | |
3290 | ||
3291 | if (*buf.buf == '^') | |
3292 | string_list_append(&exclude_packs, buf.buf + 1); | |
3293 | else | |
3294 | string_list_append(&include_packs, buf.buf); | |
3295 | ||
3296 | strbuf_reset(&buf); | |
3297 | } | |
3298 | ||
3299 | string_list_sort(&include_packs); | |
3300 | string_list_sort(&exclude_packs); | |
3301 | ||
3302 | for (p = get_all_packs(the_repository); p; p = p->next) { | |
3303 | const char *pack_name = pack_basename(p); | |
3304 | ||
3305 | item = string_list_lookup(&include_packs, pack_name); | |
3306 | if (!item) | |
3307 | item = string_list_lookup(&exclude_packs, pack_name); | |
3308 | ||
3309 | if (item) | |
3310 | item->util = p; | |
3311 | } | |
3312 | ||
3313 | /* | |
3314 | * First handle all of the excluded packs, marking them as kept in-core | |
3315 | * so that later calls to add_object_entry() discards any objects that | |
3316 | * are also found in excluded packs. | |
3317 | */ | |
3318 | for_each_string_list_item(item, &exclude_packs) { | |
3319 | struct packed_git *p = item->util; | |
3320 | if (!p) | |
3321 | die(_("could not find pack '%s'"), item->string); | |
3322 | p->pack_keep_in_core = 1; | |
3323 | } | |
3324 | ||
3325 | /* | |
3326 | * Order packs by ascending mtime; use QSORT directly to access the | |
3327 | * string_list_item's ->util pointer, which string_list_sort() does not | |
3328 | * provide. | |
3329 | */ | |
3330 | QSORT(include_packs.items, include_packs.nr, pack_mtime_cmp); | |
3331 | ||
3332 | for_each_string_list_item(item, &include_packs) { | |
3333 | struct packed_git *p = item->util; | |
3334 | if (!p) | |
3335 | die(_("could not find pack '%s'"), item->string); | |
3336 | for_each_object_in_pack(p, | |
3337 | add_object_entry_from_pack, | |
3338 | &revs, | |
3339 | FOR_EACH_OBJECT_PACK_ORDER); | |
3340 | } | |
3341 | ||
3342 | if (prepare_revision_walk(&revs)) | |
3343 | die(_("revision walk setup failed")); | |
3344 | traverse_commit_list(&revs, | |
3345 | show_commit_pack_hint, | |
3346 | show_object_pack_hint, | |
3347 | NULL); | |
3348 | ||
3349 | trace2_data_intmax("pack-objects", the_repository, "stdin_packs_found", | |
3350 | stdin_packs_found_nr); | |
3351 | trace2_data_intmax("pack-objects", the_repository, "stdin_packs_hints", | |
3352 | stdin_packs_hints_nr); | |
3353 | ||
3354 | strbuf_release(&buf); | |
3355 | string_list_clear(&include_packs, 0); | |
3356 | string_list_clear(&exclude_packs, 0); | |
3357 | } | |
3358 | ||
b5d97e6b | 3359 | static void read_object_list_from_stdin(void) |
c323ac7d | 3360 | { |
188960b4 | 3361 | char line[GIT_MAX_HEXSZ + 1 + PATH_MAX + 2]; |
3362 | struct object_id oid; | |
3363 | const char *p; | |
b2504a0d | 3364 | |
da93d12b | 3365 | for (;;) { |
da93d12b LT |
3366 | if (!fgets(line, sizeof(line), stdin)) { |
3367 | if (feof(stdin)) | |
3368 | break; | |
3369 | if (!ferror(stdin)) | |
1a07e59c | 3370 | die("BUG: fgets returned NULL, not EOF, not error!"); |
687dd75c | 3371 | if (errno != EINTR) |
d824cbba | 3372 | die_errno("fgets"); |
687dd75c JH |
3373 | clearerr(stdin); |
3374 | continue; | |
da93d12b | 3375 | } |
7a979d99 | 3376 | if (line[0] == '-') { |
188960b4 | 3377 | if (get_oid_hex(line+1, &oid)) |
f616db6a | 3378 | die(_("expected edge object ID, got garbage:\n %s"), |
b5d97e6b | 3379 | line); |
188960b4 | 3380 | add_preferred_base(&oid); |
7a979d99 | 3381 | continue; |
21fcd1bd | 3382 | } |
188960b4 | 3383 | if (parse_oid_hex(line, &oid, &p)) |
f616db6a | 3384 | die(_("expected object ID, got garbage:\n %s"), line); |
b5d97e6b | 3385 | |
188960b4 | 3386 | add_preferred_base_object(p + 1); |
fd9b1bae | 3387 | add_object_entry(&oid, OBJ_NONE, p + 1, 0); |
c323ac7d | 3388 | } |
b5d97e6b JH |
3389 | } |
3390 | ||
95308d64 | 3391 | /* Remember to update object flag allocation in object.h */ |
08cdfb13 JH |
3392 | #define OBJECT_ADDED (1u<<20) |
3393 | ||
11c211fa | 3394 | static void show_commit(struct commit *commit, void *data) |
b5d97e6b | 3395 | { |
188960b4 | 3396 | add_object_entry(&commit->object.oid, OBJ_COMMIT, NULL, 0); |
08cdfb13 | 3397 | commit->object.flags |= OBJECT_ADDED; |
7cc8f971 VM |
3398 | |
3399 | if (write_bitmap_index) | |
3400 | index_commit_for_bitmap(commit); | |
28b8a730 JK |
3401 | |
3402 | if (use_delta_islands) | |
3403 | propagate_island_marks(commit); | |
b5d97e6b JH |
3404 | } |
3405 | ||
de1e67d0 | 3406 | static void show_object(struct object *obj, const char *name, void *data) |
b5d97e6b | 3407 | { |
8d2dfc49 | 3408 | add_preferred_base_object(name); |
188960b4 | 3409 | add_object_entry(&obj->oid, obj->type, name, 0); |
8d2dfc49 | 3410 | obj->flags |= OBJECT_ADDED; |
28b8a730 JK |
3411 | |
3412 | if (use_delta_islands) { | |
3413 | const char *p; | |
39490536 | 3414 | unsigned depth; |
28b8a730 JK |
3415 | struct object_entry *ent; |
3416 | ||
39490536 JK |
3417 | /* the empty string is a root tree, which is depth 0 */ |
3418 | depth = *name ? 1 : 0; | |
28b8a730 JK |
3419 | for (p = strchr(name, '/'); p; p = strchr(p + 1, '/')) |
3420 | depth++; | |
3421 | ||
3a37876b | 3422 | ent = packlist_find(&to_pack, &obj->oid); |
108f5303 CC |
3423 | if (ent && depth > oe_tree_depth(&to_pack, ent)) |
3424 | oe_set_tree_depth(&to_pack, ent, depth); | |
28b8a730 | 3425 | } |
b5d97e6b JH |
3426 | } |
3427 | ||
9535ce73 JH |
3428 | static void show_object__ma_allow_any(struct object *obj, const char *name, void *data) |
3429 | { | |
3430 | assert(arg_missing_action == MA_ALLOW_ANY); | |
3431 | ||
3432 | /* | |
3433 | * Quietly ignore ALL missing objects. This avoids problems with | |
3434 | * staging them now and getting an odd error later. | |
3435 | */ | |
ee47243d | 3436 | if (!has_object(the_repository, &obj->oid, 0)) |
9535ce73 JH |
3437 | return; |
3438 | ||
3439 | show_object(obj, name, data); | |
3440 | } | |
3441 | ||
0c16cd49 JT |
3442 | static void show_object__ma_allow_promisor(struct object *obj, const char *name, void *data) |
3443 | { | |
3444 | assert(arg_missing_action == MA_ALLOW_PROMISOR); | |
3445 | ||
3446 | /* | |
3447 | * Quietly ignore EXPECTED missing objects. This avoids problems with | |
3448 | * staging them now and getting an odd error later. | |
3449 | */ | |
ee47243d | 3450 | if (!has_object(the_repository, &obj->oid, 0) && is_promisor_object(&obj->oid)) |
0c16cd49 JT |
3451 | return; |
3452 | ||
3453 | show_object(obj, name, data); | |
3454 | } | |
3455 | ||
9535ce73 JH |
3456 | static int option_parse_missing_action(const struct option *opt, |
3457 | const char *arg, int unset) | |
3458 | { | |
3459 | assert(arg); | |
3460 | assert(!unset); | |
3461 | ||
3462 | if (!strcmp(arg, "error")) { | |
3463 | arg_missing_action = MA_ERROR; | |
3464 | fn_show_object = show_object; | |
3465 | return 0; | |
3466 | } | |
3467 | ||
3468 | if (!strcmp(arg, "allow-any")) { | |
3469 | arg_missing_action = MA_ALLOW_ANY; | |
0c16cd49 | 3470 | fetch_if_missing = 0; |
9535ce73 JH |
3471 | fn_show_object = show_object__ma_allow_any; |
3472 | return 0; | |
3473 | } | |
3474 | ||
0c16cd49 JT |
3475 | if (!strcmp(arg, "allow-promisor")) { |
3476 | arg_missing_action = MA_ALLOW_PROMISOR; | |
3477 | fetch_if_missing = 0; | |
3478 | fn_show_object = show_object__ma_allow_promisor; | |
3479 | return 0; | |
3480 | } | |
3481 | ||
9535ce73 JH |
3482 | die(_("invalid value for --missing")); |
3483 | return 0; | |
3484 | } | |
3485 | ||
8d1d8f83 JH |
3486 | static void show_edge(struct commit *commit) |
3487 | { | |
188960b4 | 3488 | add_preferred_base(&commit->object.oid); |
8d1d8f83 JH |
3489 | } |
3490 | ||
08cdfb13 JH |
3491 | struct in_pack_object { |
3492 | off_t offset; | |
3493 | struct object *object; | |
3494 | }; | |
3495 | ||
3496 | struct in_pack { | |
071bcaab RJ |
3497 | unsigned int alloc; |
3498 | unsigned int nr; | |
08cdfb13 JH |
3499 | struct in_pack_object *array; |
3500 | }; | |
3501 | ||
3502 | static void mark_in_pack_object(struct object *object, struct packed_git *p, struct in_pack *in_pack) | |
3503 | { | |
ed1c9977 | 3504 | in_pack->array[in_pack->nr].offset = find_pack_entry_one(object->oid.hash, p); |
08cdfb13 JH |
3505 | in_pack->array[in_pack->nr].object = object; |
3506 | in_pack->nr++; | |
3507 | } | |
3508 | ||
3509 | /* | |
3510 | * Compare the objects in the offset order, in order to emulate the | |
f18d244a | 3511 | * "git rev-list --objects" output that produced the pack originally. |
08cdfb13 JH |
3512 | */ |
3513 | static int ofscmp(const void *a_, const void *b_) | |
3514 | { | |
3515 | struct in_pack_object *a = (struct in_pack_object *)a_; | |
3516 | struct in_pack_object *b = (struct in_pack_object *)b_; | |
3517 | ||
3518 | if (a->offset < b->offset) | |
3519 | return -1; | |
3520 | else if (a->offset > b->offset) | |
3521 | return 1; | |
3522 | else | |
f2fd0760 | 3523 | return oidcmp(&a->object->oid, &b->object->oid); |
08cdfb13 JH |
3524 | } |
3525 | ||
7a2a7216 | 3526 | static void add_objects_in_unpacked_packs(void) |
08cdfb13 JH |
3527 | { |
3528 | struct packed_git *p; | |
3529 | struct in_pack in_pack; | |
3530 | uint32_t i; | |
3531 | ||
3532 | memset(&in_pack, 0, sizeof(in_pack)); | |
3533 | ||
454ea2e4 | 3534 | for (p = get_all_packs(the_repository); p; p = p->next) { |
188960b4 | 3535 | struct object_id oid; |
08cdfb13 JH |
3536 | struct object *o; |
3537 | ||
ed7e5fc3 | 3538 | if (!p->pack_local || p->pack_keep || p->pack_keep_in_core) |
08cdfb13 JH |
3539 | continue; |
3540 | if (open_pack_index(p)) | |
f616db6a | 3541 | die(_("cannot open pack index")); |
08cdfb13 JH |
3542 | |
3543 | ALLOC_GROW(in_pack.array, | |
3544 | in_pack.nr + p->num_objects, | |
3545 | in_pack.alloc); | |
3546 | ||
3547 | for (i = 0; i < p->num_objects; i++) { | |
0763671b | 3548 | nth_packed_object_id(&oid, p, i); |
45a187cc | 3549 | o = lookup_unknown_object(the_repository, &oid); |
08cdfb13 JH |
3550 | if (!(o->flags & OBJECT_ADDED)) |
3551 | mark_in_pack_object(o, p, &in_pack); | |
3552 | o->flags |= OBJECT_ADDED; | |
3553 | } | |
3554 | } | |
3555 | ||
3556 | if (in_pack.nr) { | |
9ed0d8d6 | 3557 | QSORT(in_pack.array, in_pack.nr, ofscmp); |
08cdfb13 JH |
3558 | for (i = 0; i < in_pack.nr; i++) { |
3559 | struct object *o = in_pack.array[i].object; | |
188960b4 | 3560 | add_object_entry(&o->oid, o->type, "", 0); |
08cdfb13 JH |
3561 | } |
3562 | } | |
3563 | free(in_pack.array); | |
3564 | } | |
3565 | ||
76c1d9a0 | 3566 | static int add_loose_object(const struct object_id *oid, const char *path, |
e26a8c47 JK |
3567 | void *data) |
3568 | { | |
0df8e965 | 3569 | enum object_type type = oid_object_info(the_repository, oid, NULL); |
e26a8c47 JK |
3570 | |
3571 | if (type < 0) { | |
f616db6a | 3572 | warning(_("loose object at %s could not be examined"), path); |
e26a8c47 JK |
3573 | return 0; |
3574 | } | |
3575 | ||
188960b4 | 3576 | add_object_entry(oid, type, "", 0); |
e26a8c47 JK |
3577 | return 0; |
3578 | } | |
3579 | ||
3580 | /* | |
3581 | * We actually don't even have to worry about reachability here. | |
3582 | * add_object_entry will weed out duplicates, so we just add every | |
3583 | * loose object we find. | |
3584 | */ | |
3585 | static void add_unreachable_loose_objects(void) | |
3586 | { | |
3587 | for_each_loose_file_in_objdir(get_object_directory(), | |
3588 | add_loose_object, | |
3589 | NULL, NULL, NULL); | |
3590 | } | |
3591 | ||
188960b4 | 3592 | static int has_sha1_pack_kept_or_nonlocal(const struct object_id *oid) |
094085e3 BC |
3593 | { |
3594 | static struct packed_git *last_found = (void *)1; | |
3595 | struct packed_git *p; | |
3596 | ||
a80d72db | 3597 | p = (last_found != (void *)1) ? last_found : |
454ea2e4 | 3598 | get_all_packs(the_repository); |
094085e3 BC |
3599 | |
3600 | while (p) { | |
ed7e5fc3 NTND |
3601 | if ((!p->pack_local || p->pack_keep || |
3602 | p->pack_keep_in_core) && | |
188960b4 | 3603 | find_pack_entry_one(oid->hash, p)) { |
094085e3 BC |
3604 | last_found = p; |
3605 | return 1; | |
3606 | } | |
3607 | if (p == last_found) | |
454ea2e4 | 3608 | p = get_all_packs(the_repository); |
094085e3 BC |
3609 | else |
3610 | p = p->next; | |
3611 | if (p == last_found) | |
3612 | p = p->next; | |
3613 | } | |
3614 | return 0; | |
3615 | } | |
3616 | ||
abcb8655 JK |
3617 | /* |
3618 | * Store a list of sha1s that are should not be discarded | |
3619 | * because they are either written too recently, or are | |
3620 | * reachable from another object that was. | |
3621 | * | |
3622 | * This is filled by get_object_list. | |
3623 | */ | |
910650d2 | 3624 | static struct oid_array recent_objects; |
abcb8655 | 3625 | |
4ce3621a | 3626 | static int loosened_object_can_be_discarded(const struct object_id *oid, |
dddbad72 | 3627 | timestamp_t mtime) |
d0d46abc JK |
3628 | { |
3629 | if (!unpack_unreachable_expiration) | |
3630 | return 0; | |
3631 | if (mtime > unpack_unreachable_expiration) | |
3632 | return 0; | |
910650d2 | 3633 | if (oid_array_lookup(&recent_objects, oid) >= 0) |
abcb8655 | 3634 | return 0; |
d0d46abc JK |
3635 | return 1; |
3636 | } | |
3637 | ||
7a2a7216 | 3638 | static void loosen_unused_packed_objects(void) |
ca11b212 NP |
3639 | { |
3640 | struct packed_git *p; | |
3641 | uint32_t i; | |
a643157d | 3642 | uint32_t loosened_objects_nr = 0; |
4ce3621a | 3643 | struct object_id oid; |
ca11b212 | 3644 | |
454ea2e4 | 3645 | for (p = get_all_packs(the_repository); p; p = p->next) { |
ed7e5fc3 | 3646 | if (!p->pack_local || p->pack_keep || p->pack_keep_in_core) |
ca11b212 NP |
3647 | continue; |
3648 | ||
3649 | if (open_pack_index(p)) | |
f616db6a | 3650 | die(_("cannot open pack index")); |
ca11b212 NP |
3651 | |
3652 | for (i = 0; i < p->num_objects; i++) { | |
0763671b | 3653 | nth_packed_object_id(&oid, p, i); |
3a37876b | 3654 | if (!packlist_find(&to_pack, &oid) && |
188960b4 | 3655 | !has_sha1_pack_kept_or_nonlocal(&oid) && |
a643157d | 3656 | !loosened_object_can_be_discarded(&oid, p->mtime)) { |
4bdb70a4 | 3657 | if (force_object_loose(&oid, p->mtime)) |
f616db6a | 3658 | die(_("unable to force loose object")); |
a643157d RS |
3659 | loosened_objects_nr++; |
3660 | } | |
ca11b212 NP |
3661 | } |
3662 | } | |
a643157d RS |
3663 | |
3664 | trace2_data_intmax("pack-objects", the_repository, | |
3665 | "loosen_unused_packed_objects/loosened", loosened_objects_nr); | |
ca11b212 NP |
3666 | } |
3667 | ||
69e4b342 | 3668 | /* |
645c432d KS |
3669 | * This tracks any options which pack-reuse code expects to be on, or which a |
3670 | * reader of the pack might not understand, and which would therefore prevent | |
3671 | * blind reuse of what we have on disk. | |
69e4b342 JK |
3672 | */ |
3673 | static int pack_options_allow_reuse(void) | |
3674 | { | |
e704fc79 JK |
3675 | return allow_pack_reuse && |
3676 | pack_to_stdout && | |
ed7e5fc3 NTND |
3677 | !ignore_packed_keep_on_disk && |
3678 | !ignore_packed_keep_in_core && | |
9df4a607 JK |
3679 | (!local || !have_non_local_packs) && |
3680 | !incremental; | |
69e4b342 JK |
3681 | } |
3682 | ||
6b8fda2d VM |
3683 | static int get_object_list_from_bitmap(struct rev_info *revs) |
3684 | { | |
9cf68b27 | 3685 | if (!(bitmap_git = prepare_bitmap_walk(revs, &filter_options, 0))) |
6b8fda2d VM |
3686 | return -1; |
3687 | ||
69e4b342 JK |
3688 | if (pack_options_allow_reuse() && |
3689 | !reuse_partial_packfile_from_bitmap( | |
3ae5fa07 | 3690 | bitmap_git, |
6b8fda2d VM |
3691 | &reuse_packfile, |
3692 | &reuse_packfile_objects, | |
bb514de3 | 3693 | &reuse_packfile_bitmap)) { |
6b8fda2d VM |
3694 | assert(reuse_packfile_objects); |
3695 | nr_result += reuse_packfile_objects; | |
8e118e84 JK |
3696 | nr_seen += reuse_packfile_objects; |
3697 | display_progress(progress_state, nr_seen); | |
6b8fda2d VM |
3698 | } |
3699 | ||
4eb707eb JK |
3700 | traverse_bitmap_commit_list(bitmap_git, revs, |
3701 | &add_object_entry_from_bitmap); | |
6b8fda2d VM |
3702 | return 0; |
3703 | } | |
3704 | ||
abcb8655 | 3705 | static void record_recent_object(struct object *obj, |
de1e67d0 | 3706 | const char *name, |
abcb8655 JK |
3707 | void *data) |
3708 | { | |
910650d2 | 3709 | oid_array_append(&recent_objects, &obj->oid); |
abcb8655 JK |
3710 | } |
3711 | ||
3712 | static void record_recent_commit(struct commit *commit, void *data) | |
3713 | { | |
910650d2 | 3714 | oid_array_append(&recent_objects, &commit->object.oid); |
abcb8655 JK |
3715 | } |
3716 | ||
3f267a11 TB |
3717 | static int mark_bitmap_preferred_tip(const char *refname, |
3718 | const struct object_id *oid, int flags, | |
3719 | void *_data) | |
3720 | { | |
3721 | struct object_id peeled; | |
3722 | struct object *object; | |
3723 | ||
3724 | if (!peel_iterated_oid(oid, &peeled)) | |
3725 | oid = &peeled; | |
3726 | ||
3727 | object = parse_object_or_die(oid, refname); | |
3728 | if (object->type == OBJ_COMMIT) | |
3729 | object->flags |= NEEDS_BITMAP; | |
3730 | ||
3731 | return 0; | |
3732 | } | |
3733 | ||
3734 | static void mark_bitmap_preferred_tips(void) | |
3735 | { | |
3736 | struct string_list_item *item; | |
3737 | const struct string_list *preferred_tips; | |
3738 | ||
3739 | preferred_tips = bitmap_preferred_tips(the_repository); | |
3740 | if (!preferred_tips) | |
3741 | return; | |
3742 | ||
3743 | for_each_string_list_item(item, preferred_tips) { | |
3744 | for_each_ref_in(item->string, mark_bitmap_preferred_tip, NULL); | |
3745 | } | |
3746 | } | |
3747 | ||
8d1d8f83 | 3748 | static void get_object_list(int ac, const char **av) |
b5d97e6b JH |
3749 | { |
3750 | struct rev_info revs; | |
bbcde41a MD |
3751 | struct setup_revision_opt s_r_opt = { |
3752 | .allow_exclude_promisor_objects = 1, | |
3753 | }; | |
b5d97e6b | 3754 | char line[1000]; |
b5d97e6b | 3755 | int flags = 0; |
a4544b31 | 3756 | int save_warning; |
b5d97e6b | 3757 | |
2abf3503 | 3758 | repo_init_revisions(the_repository, &revs, NULL); |
b5d97e6b | 3759 | save_commit_buffer = 0; |
bbcde41a | 3760 | setup_revisions(ac, av, &revs, &s_r_opt); |
b5d97e6b | 3761 | |
b790e0f6 | 3762 | /* make sure shallows are read */ |
c8813487 | 3763 | is_repository_shallow(the_repository); |
b790e0f6 | 3764 | |
a4544b31 DS |
3765 | save_warning = warn_on_object_refname_ambiguity; |
3766 | warn_on_object_refname_ambiguity = 0; | |
3767 | ||
b5d97e6b JH |
3768 | while (fgets(line, sizeof(line), stdin) != NULL) { |
3769 | int len = strlen(line); | |
872c930d | 3770 | if (len && line[len - 1] == '\n') |
b5d97e6b JH |
3771 | line[--len] = 0; |
3772 | if (!len) | |
3773 | break; | |
3774 | if (*line == '-') { | |
3775 | if (!strcmp(line, "--not")) { | |
3776 | flags ^= UNINTERESTING; | |
7cc8f971 | 3777 | write_bitmap_index = 0; |
b5d97e6b JH |
3778 | continue; |
3779 | } | |
b790e0f6 | 3780 | if (starts_with(line, "--shallow ")) { |
e92b848c | 3781 | struct object_id oid; |
3782 | if (get_oid_hex(line + 10, &oid)) | |
4279000d | 3783 | die("not an object name '%s'", line + 10); |
19143f13 | 3784 | register_shallow(the_repository, &oid); |
f7f91086 | 3785 | use_bitmap_index = 0; |
b790e0f6 NTND |
3786 | continue; |
3787 | } | |
f616db6a | 3788 | die(_("not a rev '%s'"), line); |
b5d97e6b | 3789 | } |
8e676e8b | 3790 | if (handle_revision_arg(line, &revs, flags, REVARG_CANNOT_BE_FILENAME)) |
f616db6a | 3791 | die(_("bad revision '%s'"), line); |
b5d97e6b JH |
3792 | } |
3793 | ||
a4544b31 DS |
3794 | warn_on_object_refname_ambiguity = save_warning; |
3795 | ||
6b8fda2d VM |
3796 | if (use_bitmap_index && !get_object_list_from_bitmap(&revs)) |
3797 | return; | |
3798 | ||
28b8a730 | 3799 | if (use_delta_islands) |
bdbdf42f | 3800 | load_delta_islands(the_repository, progress); |
28b8a730 | 3801 | |
3f267a11 TB |
3802 | if (write_bitmap_index) |
3803 | mark_bitmap_preferred_tips(); | |
3804 | ||
3d51e1b5 | 3805 | if (prepare_revision_walk(&revs)) |
f616db6a | 3806 | die(_("revision walk setup failed")); |
4f6d26b1 | 3807 | mark_edges_uninteresting(&revs, show_edge, sparse); |
9535ce73 JH |
3808 | |
3809 | if (!fn_show_object) | |
3810 | fn_show_object = show_object; | |
3811 | traverse_commit_list_filtered(&filter_options, &revs, | |
3812 | show_commit, fn_show_object, NULL, | |
3813 | NULL); | |
08cdfb13 | 3814 | |
abcb8655 JK |
3815 | if (unpack_unreachable_expiration) { |
3816 | revs.ignore_missing_links = 1; | |
3817 | if (add_unseen_recent_objects_to_traversal(&revs, | |
3818 | unpack_unreachable_expiration)) | |
f616db6a | 3819 | die(_("unable to add recent objects")); |
abcb8655 | 3820 | if (prepare_revision_walk(&revs)) |
f616db6a | 3821 | die(_("revision walk setup failed")); |
abcb8655 JK |
3822 | traverse_commit_list(&revs, record_recent_commit, |
3823 | record_recent_object, NULL); | |
3824 | } | |
3825 | ||
08cdfb13 | 3826 | if (keep_unreachable) |
7a2a7216 | 3827 | add_objects_in_unpacked_packs(); |
e26a8c47 JK |
3828 | if (pack_loose_unreachable) |
3829 | add_unreachable_loose_objects(); | |
ca11b212 | 3830 | if (unpack_unreachable) |
7a2a7216 | 3831 | loosen_unused_packed_objects(); |
abcb8655 | 3832 | |
910650d2 | 3833 | oid_array_clear(&recent_objects); |
b5d97e6b JH |
3834 | } |
3835 | ||
ed7e5fc3 NTND |
3836 | static void add_extra_kept_packs(const struct string_list *names) |
3837 | { | |
3838 | struct packed_git *p; | |
3839 | ||
3840 | if (!names->nr) | |
3841 | return; | |
3842 | ||
454ea2e4 | 3843 | for (p = get_all_packs(the_repository); p; p = p->next) { |
ed7e5fc3 NTND |
3844 | const char *name = basename(p->pack_name); |
3845 | int i; | |
3846 | ||
3847 | if (!p->pack_local) | |
3848 | continue; | |
3849 | ||
3850 | for (i = 0; i < names->nr; i++) | |
3851 | if (!fspathcmp(name, names->items[i].string)) | |
3852 | break; | |
3853 | ||
3854 | if (i < names->nr) { | |
3855 | p->pack_keep_in_core = 1; | |
3856 | ignore_packed_keep_in_core = 1; | |
3857 | continue; | |
3858 | } | |
3859 | } | |
3860 | } | |
3861 | ||
99fb6e04 NTND |
3862 | static int option_parse_index_version(const struct option *opt, |
3863 | const char *arg, int unset) | |
3864 | { | |
3865 | char *c; | |
3866 | const char *val = arg; | |
517fe807 JK |
3867 | |
3868 | BUG_ON_OPT_NEG(unset); | |
3869 | ||
99fb6e04 NTND |
3870 | pack_idx_opts.version = strtoul(val, &c, 10); |
3871 | if (pack_idx_opts.version > 2) | |
3872 | die(_("unsupported index version %s"), val); | |
3873 | if (*c == ',' && c[1]) | |
3874 | pack_idx_opts.off32_limit = strtoul(c+1, &c, 0); | |
3875 | if (*c || pack_idx_opts.off32_limit & 0x80000000) | |
3876 | die(_("bad index version '%s'"), val); | |
3877 | return 0; | |
3878 | } | |
3879 | ||
7e52f566 JK |
3880 | static int option_parse_unpack_unreachable(const struct option *opt, |
3881 | const char *arg, int unset) | |
3882 | { | |
3883 | if (unset) { | |
3884 | unpack_unreachable = 0; | |
3885 | unpack_unreachable_expiration = 0; | |
3886 | } | |
3887 | else { | |
3888 | unpack_unreachable = 1; | |
3889 | if (arg) | |
3890 | unpack_unreachable_expiration = approxidate(arg); | |
3891 | } | |
3892 | return 0; | |
3893 | } | |
3894 | ||
b5d97e6b JH |
3895 | int cmd_pack_objects(int argc, const char **argv, const char *prefix) |
3896 | { | |
b5d97e6b | 3897 | int use_internal_rev_list = 0; |
2dacf26d | 3898 | int shallow = 0; |
4f366275 | 3899 | int all_progress_implied = 0; |
22f9b7f3 | 3900 | struct strvec rp = STRVEC_INIT; |
99fb6e04 | 3901 | int rev_list_unpacked = 0, rev_list_all = 0, rev_list_reflog = 0; |
c90f9e13 | 3902 | int rev_list_index = 0; |
339bce27 | 3903 | int stdin_packs = 0; |
ed7e5fc3 | 3904 | struct string_list keep_pack_list = STRING_LIST_INIT_NODUP; |
99fb6e04 NTND |
3905 | struct option pack_objects_options[] = { |
3906 | OPT_SET_INT('q', "quiet", &progress, | |
4c688120 | 3907 | N_("do not show progress meter"), 0), |
99fb6e04 | 3908 | OPT_SET_INT(0, "progress", &progress, |
4c688120 | 3909 | N_("show progress meter"), 1), |
99fb6e04 | 3910 | OPT_SET_INT(0, "all-progress", &progress, |
4c688120 | 3911 | N_("show progress meter during object writing phase"), 2), |
99fb6e04 NTND |
3912 | OPT_BOOL(0, "all-progress-implied", |
3913 | &all_progress_implied, | |
4c688120 | 3914 | N_("similar to --all-progress when progress meter is shown")), |
203c8533 | 3915 | OPT_CALLBACK_F(0, "index-version", NULL, N_("<version>[,<offset>]"), |
4c688120 | 3916 | N_("write the pack index file in the specified idx format version"), |
203c8533 | 3917 | PARSE_OPT_NONEG, option_parse_index_version), |
2a514ed8 CB |
3918 | OPT_MAGNITUDE(0, "max-pack-size", &pack_size_limit, |
3919 | N_("maximum size of each output pack file")), | |
99fb6e04 | 3920 | OPT_BOOL(0, "local", &local, |
4c688120 | 3921 | N_("ignore borrowed objects from alternate object store")), |
99fb6e04 | 3922 | OPT_BOOL(0, "incremental", &incremental, |
4c688120 | 3923 | N_("ignore packed objects")), |
99fb6e04 | 3924 | OPT_INTEGER(0, "window", &window, |
4c688120 | 3925 | N_("limit pack window by objects")), |
2a514ed8 CB |
3926 | OPT_MAGNITUDE(0, "window-memory", &window_memory_limit, |
3927 | N_("limit pack window by memory in addition to object limit")), | |
99fb6e04 | 3928 | OPT_INTEGER(0, "depth", &depth, |
4c688120 | 3929 | N_("maximum length of delta chain allowed in the resulting pack")), |
99fb6e04 | 3930 | OPT_BOOL(0, "reuse-delta", &reuse_delta, |
4c688120 | 3931 | N_("reuse existing deltas")), |
99fb6e04 | 3932 | OPT_BOOL(0, "reuse-object", &reuse_object, |
4c688120 | 3933 | N_("reuse existing objects")), |
99fb6e04 | 3934 | OPT_BOOL(0, "delta-base-offset", &allow_ofs_delta, |
4c688120 | 3935 | N_("use OFS_DELTA objects")), |
99fb6e04 | 3936 | OPT_INTEGER(0, "threads", &delta_search_threads, |
4c688120 | 3937 | N_("use threads when searching for best delta matches")), |
99fb6e04 | 3938 | OPT_BOOL(0, "non-empty", &non_empty, |
4c688120 | 3939 | N_("do not create an empty pack output")), |
99fb6e04 | 3940 | OPT_BOOL(0, "revs", &use_internal_rev_list, |
4c688120 | 3941 | N_("read revision arguments from standard input")), |
3e4a67b4 NTND |
3942 | OPT_SET_INT_F(0, "unpacked", &rev_list_unpacked, |
3943 | N_("limit the objects to those that are not yet packed"), | |
3944 | 1, PARSE_OPT_NONEG), | |
3945 | OPT_SET_INT_F(0, "all", &rev_list_all, | |
3946 | N_("include objects reachable from any reference"), | |
3947 | 1, PARSE_OPT_NONEG), | |
3948 | OPT_SET_INT_F(0, "reflog", &rev_list_reflog, | |
3949 | N_("include objects referred by reflog entries"), | |
3950 | 1, PARSE_OPT_NONEG), | |
3951 | OPT_SET_INT_F(0, "indexed-objects", &rev_list_index, | |
3952 | N_("include objects referred to by the index"), | |
3953 | 1, PARSE_OPT_NONEG), | |
339bce27 TB |
3954 | OPT_BOOL(0, "stdin-packs", &stdin_packs, |
3955 | N_("read packs from stdin")), | |
99fb6e04 | 3956 | OPT_BOOL(0, "stdout", &pack_to_stdout, |
4c688120 | 3957 | N_("output pack to stdout")), |
99fb6e04 | 3958 | OPT_BOOL(0, "include-tag", &include_tag, |
4c688120 | 3959 | N_("include tag objects that refer to objects to be packed")), |
99fb6e04 | 3960 | OPT_BOOL(0, "keep-unreachable", &keep_unreachable, |
4c688120 | 3961 | N_("keep unreachable objects")), |
e26a8c47 JK |
3962 | OPT_BOOL(0, "pack-loose-unreachable", &pack_loose_unreachable, |
3963 | N_("pack loose unreachable objects")), | |
203c8533 | 3964 | OPT_CALLBACK_F(0, "unpack-unreachable", NULL, N_("time"), |
4c688120 | 3965 | N_("unpack unreachable objects newer than <time>"), |
203c8533 | 3966 | PARSE_OPT_OPTARG, option_parse_unpack_unreachable), |
4f6d26b1 DS |
3967 | OPT_BOOL(0, "sparse", &sparse, |
3968 | N_("use the sparse reachability algorithm")), | |
99fb6e04 | 3969 | OPT_BOOL(0, "thin", &thin, |
4c688120 | 3970 | N_("create thin packs")), |
2dacf26d | 3971 | OPT_BOOL(0, "shallow", &shallow, |
3972 | N_("create packs suitable for shallow fetches")), | |
ed7e5fc3 | 3973 | OPT_BOOL(0, "honor-pack-keep", &ignore_packed_keep_on_disk, |
4c688120 | 3974 | N_("ignore packs that have companion .keep file")), |
ed7e5fc3 NTND |
3975 | OPT_STRING_LIST(0, "keep-pack", &keep_pack_list, N_("name"), |
3976 | N_("ignore this pack")), | |
99fb6e04 | 3977 | OPT_INTEGER(0, "compression", &pack_compression_level, |
4c688120 | 3978 | N_("pack compression level")), |
99fb6e04 | 3979 | OPT_SET_INT(0, "keep-true-parents", &grafts_replace_parents, |
4c688120 | 3980 | N_("do not hide commits by grafts"), 0), |
6b8fda2d VM |
3981 | OPT_BOOL(0, "use-bitmap-index", &use_bitmap_index, |
3982 | N_("use a bitmap index if available to speed up counting objects")), | |
25575015 JK |
3983 | OPT_SET_INT(0, "write-bitmap-index", &write_bitmap_index, |
3984 | N_("write a bitmap index together with the pack index"), | |
3985 | WRITE_BITMAP_TRUE), | |
3986 | OPT_SET_INT_F(0, "write-bitmap-index-quiet", | |
3987 | &write_bitmap_index, | |
3988 | N_("write a bitmap index if possible"), | |
3989 | WRITE_BITMAP_QUIET, PARSE_OPT_HIDDEN), | |
9535ce73 | 3990 | OPT_PARSE_LIST_OBJECTS_FILTER(&filter_options), |
203c8533 | 3991 | OPT_CALLBACK_F(0, "missing", NULL, N_("action"), |
9535ce73 | 3992 | N_("handling for missing objects"), PARSE_OPT_NONEG, |
203c8533 | 3993 | option_parse_missing_action), |
0c16cd49 JT |
3994 | OPT_BOOL(0, "exclude-promisor-objects", &exclude_promisor_objects, |
3995 | N_("do not pack objects in promisor packfiles")), | |
28b8a730 JK |
3996 | OPT_BOOL(0, "delta-islands", &use_delta_islands, |
3997 | N_("respect islands during delta compression")), | |
dd4b732d JT |
3998 | OPT_STRING_LIST(0, "uri-protocol", &uri_protocols, |
3999 | N_("protocol"), | |
4000 | N_("exclude any configured uploadpack.blobpackfileuri with this protocol")), | |
99fb6e04 NTND |
4001 | OPT_END(), |
4002 | }; | |
8d1d8f83 | 4003 | |
0c6804ab NTND |
4004 | if (DFS_NUM_STATES > (1 << OE_DFS_STATE_BITS)) |
4005 | BUG("too many dfs states, increase OE_DFS_STATE_BITS"); | |
4006 | ||
6ebd1caf | 4007 | read_replace_refs = 0; |
dae556bd | 4008 | |
2d657ab9 | 4009 | sparse = git_env_bool("GIT_TEST_PACK_SPARSE", -1); |
7211b9e7 | 4010 | prepare_repo_settings(the_repository); |
2d657ab9 | 4011 | if (sparse < 0) |
7211b9e7 DS |
4012 | sparse = the_repository->settings.pack_use_sparse; |
4013 | ||
ebcfb379 | 4014 | reset_pack_idx_option(&pack_idx_opts); |
ef90d6d4 | 4015 | git_config(git_pack_config, NULL); |
e8c58f89 TB |
4016 | if (git_env_bool(GIT_TEST_WRITE_REV_INDEX, 0)) |
4017 | pack_idx_opts.flags |= WRITE_REV; | |
b5d97e6b JH |
4018 | |
4019 | progress = isatty(2); | |
99fb6e04 NTND |
4020 | argc = parse_options(argc, argv, prefix, pack_objects_options, |
4021 | pack_usage, 0); | |
b5d97e6b | 4022 | |
99fb6e04 NTND |
4023 | if (argc) { |
4024 | base_name = argv[0]; | |
4025 | argc--; | |
b5d97e6b | 4026 | } |
99fb6e04 NTND |
4027 | if (pack_to_stdout != !base_name || argc) |
4028 | usage_with_options(pack_usage, pack_objects_options); | |
b5d97e6b | 4029 | |
6d52b6a5 JK |
4030 | if (depth < 0) |
4031 | depth = 0; | |
b5c0cbd8 NTND |
4032 | if (depth >= (1 << OE_DEPTH_BITS)) { |
4033 | warning(_("delta chain depth %d is too deep, forcing %d"), | |
4034 | depth, (1 << OE_DEPTH_BITS) - 1); | |
4035 | depth = (1 << OE_DEPTH_BITS) - 1; | |
4036 | } | |
0cb3c142 NTND |
4037 | if (cache_max_small_delta_size >= (1U << OE_Z_DELTA_BITS)) { |
4038 | warning(_("pack.deltaCacheLimit is too high, forcing %d"), | |
4039 | (1U << OE_Z_DELTA_BITS) - 1); | |
4040 | cache_max_small_delta_size = (1U << OE_Z_DELTA_BITS) - 1; | |
4041 | } | |
953aa54e JK |
4042 | if (window < 0) |
4043 | window = 0; | |
b5c0cbd8 | 4044 | |
22f9b7f3 | 4045 | strvec_push(&rp, "pack-objects"); |
99fb6e04 NTND |
4046 | if (thin) { |
4047 | use_internal_rev_list = 1; | |
22f9b7f3 | 4048 | strvec_push(&rp, shallow |
2dacf26d | 4049 | ? "--objects-edge-aggressive" |
4050 | : "--objects-edge"); | |
99fb6e04 | 4051 | } else |
22f9b7f3 | 4052 | strvec_push(&rp, "--objects"); |
b5d97e6b | 4053 | |
99fb6e04 NTND |
4054 | if (rev_list_all) { |
4055 | use_internal_rev_list = 1; | |
22f9b7f3 | 4056 | strvec_push(&rp, "--all"); |
99fb6e04 NTND |
4057 | } |
4058 | if (rev_list_reflog) { | |
4059 | use_internal_rev_list = 1; | |
22f9b7f3 | 4060 | strvec_push(&rp, "--reflog"); |
99fb6e04 | 4061 | } |
c90f9e13 JK |
4062 | if (rev_list_index) { |
4063 | use_internal_rev_list = 1; | |
22f9b7f3 | 4064 | strvec_push(&rp, "--indexed-objects"); |
99fb6e04 | 4065 | } |
339bce27 | 4066 | if (rev_list_unpacked && !stdin_packs) { |
99fb6e04 | 4067 | use_internal_rev_list = 1; |
22f9b7f3 | 4068 | strvec_push(&rp, "--unpacked"); |
99fb6e04 | 4069 | } |
b5d97e6b | 4070 | |
0c16cd49 JT |
4071 | if (exclude_promisor_objects) { |
4072 | use_internal_rev_list = 1; | |
4073 | fetch_if_missing = 0; | |
22f9b7f3 | 4074 | strvec_push(&rp, "--exclude-promisor-objects"); |
0c16cd49 | 4075 | } |
58bd77b6 NTND |
4076 | if (unpack_unreachable || keep_unreachable || pack_loose_unreachable) |
4077 | use_internal_rev_list = 1; | |
0c16cd49 | 4078 | |
99fb6e04 NTND |
4079 | if (!reuse_object) |
4080 | reuse_delta = 0; | |
4081 | if (pack_compression_level == -1) | |
4082 | pack_compression_level = Z_DEFAULT_COMPRESSION; | |
4083 | else if (pack_compression_level < 0 || pack_compression_level > Z_BEST_COMPRESSION) | |
f616db6a | 4084 | die(_("bad pack compression level %d"), pack_compression_level); |
0c45d258 JH |
4085 | |
4086 | if (!delta_search_threads) /* --threads=0 means autodetect */ | |
4087 | delta_search_threads = online_cpus(); | |
4088 | ||
9c897c5c | 4089 | if (!HAVE_THREADS && delta_search_threads != 1) |
f616db6a | 4090 | warning(_("no threads support, ignoring --threads")); |
2b84b5a8 JS |
4091 | if (!pack_to_stdout && !pack_size_limit) |
4092 | pack_size_limit = pack_size_limit_cfg; | |
01c12a23 | 4093 | if (pack_to_stdout && pack_size_limit) |
f616db6a | 4094 | die(_("--max-pack-size cannot be used to build a pack for transfer")); |
07cf0f24 | 4095 | if (pack_size_limit && pack_size_limit < 1024*1024) { |
f616db6a | 4096 | warning(_("minimum pack size limit is 1 MiB")); |
07cf0f24 NP |
4097 | pack_size_limit = 1024*1024; |
4098 | } | |
01c12a23 | 4099 | |
8d1d8f83 | 4100 | if (!pack_to_stdout && thin) |
f616db6a | 4101 | die(_("--thin cannot be used to build an indexable pack")); |
b5d97e6b | 4102 | |
ca11b212 | 4103 | if (keep_unreachable && unpack_unreachable) |
f616db6a | 4104 | die(_("--keep-unreachable and --unpack-unreachable are incompatible")); |
b1e757f3 JK |
4105 | if (!rev_list_all || !rev_list_reflog || !rev_list_index) |
4106 | unpack_unreachable_expiration = 0; | |
ca11b212 | 4107 | |
9535ce73 JH |
4108 | if (filter_options.choice) { |
4109 | if (!pack_to_stdout) | |
f616db6a | 4110 | die(_("cannot use --filter without --stdout")); |
339bce27 TB |
4111 | if (stdin_packs) |
4112 | die(_("cannot use --filter with --stdin-packs")); | |
9535ce73 JH |
4113 | } |
4114 | ||
339bce27 TB |
4115 | if (stdin_packs && use_internal_rev_list) |
4116 | die(_("cannot use internal rev list with --stdin-packs")); | |
4117 | ||
645c432d KS |
4118 | /* |
4119 | * "soft" reasons not to use bitmaps - for on-disk repack by default we want | |
4120 | * | |
4121 | * - to produce good pack (with bitmap index not-yet-packed objects are | |
4122 | * packed in suboptimal order). | |
4123 | * | |
4124 | * - to use more robust pack-generation codepath (avoiding possible | |
4125 | * bugs in bitmap code and possible bitmap index corruption). | |
4126 | */ | |
4127 | if (!pack_to_stdout) | |
4128 | use_bitmap_index_default = 0; | |
4129 | ||
4130 | if (use_bitmap_index < 0) | |
4131 | use_bitmap_index = use_bitmap_index_default; | |
4132 | ||
4133 | /* "hard" reasons not to use bitmaps; these just won't work at all */ | |
c8813487 | 4134 | if (!use_internal_rev_list || (!pack_to_stdout && write_bitmap_index) || is_repository_shallow(the_repository)) |
6b8fda2d VM |
4135 | use_bitmap_index = 0; |
4136 | ||
7cc8f971 VM |
4137 | if (pack_to_stdout || !rev_list_all) |
4138 | write_bitmap_index = 0; | |
4139 | ||
28b8a730 | 4140 | if (use_delta_islands) |
22f9b7f3 | 4141 | strvec_push(&rp, "--topo-order"); |
28b8a730 | 4142 | |
4f366275 NP |
4143 | if (progress && all_progress_implied) |
4144 | progress = 2; | |
4145 | ||
ed7e5fc3 NTND |
4146 | add_extra_kept_packs(&keep_pack_list); |
4147 | if (ignore_packed_keep_on_disk) { | |
56dfeb62 | 4148 | struct packed_git *p; |
454ea2e4 | 4149 | for (p = get_all_packs(the_repository); p; p = p->next) |
56dfeb62 JK |
4150 | if (p->pack_local && p->pack_keep) |
4151 | break; | |
4152 | if (!p) /* no keep-able packs found */ | |
ed7e5fc3 | 4153 | ignore_packed_keep_on_disk = 0; |
56dfeb62 JK |
4154 | } |
4155 | if (local) { | |
4156 | /* | |
ed7e5fc3 NTND |
4157 | * unlike ignore_packed_keep_on_disk above, we do not |
4158 | * want to unset "local" based on looking at packs, as | |
4159 | * it also covers non-local objects | |
56dfeb62 JK |
4160 | */ |
4161 | struct packed_git *p; | |
454ea2e4 | 4162 | for (p = get_all_packs(the_repository); p; p = p->next) { |
56dfeb62 JK |
4163 | if (!p->pack_local) { |
4164 | have_non_local_packs = 1; | |
4165 | break; | |
4166 | } | |
4167 | } | |
4168 | } | |
b5d97e6b | 4169 | |
ae417807 DS |
4170 | trace2_region_enter("pack-objects", "enumerate-objects", |
4171 | the_repository); | |
7c141127 | 4172 | prepare_packing_data(the_repository, &to_pack); |
43fa44fa | 4173 | |
13aaf148 | 4174 | if (progress) |
5af05043 | 4175 | progress_state = start_progress(_("Enumerating objects"), 0); |
339bce27 TB |
4176 | if (stdin_packs) { |
4177 | /* avoids adding objects in excluded packs */ | |
4178 | ignore_packed_keep_in_core = 1; | |
4179 | read_packs_list_from_stdin(); | |
4180 | if (rev_list_unpacked) | |
4181 | add_unreachable_loose_objects(); | |
4182 | } else if (!use_internal_rev_list) | |
b5d97e6b | 4183 | read_object_list_from_stdin(); |
8d1d8f83 | 4184 | else { |
d70a9eb6 | 4185 | get_object_list(rp.nr, rp.v); |
22f9b7f3 | 4186 | strvec_clear(&rp); |
8d1d8f83 | 4187 | } |
0ef95f72 | 4188 | cleanup_preferred_base(); |
d155254c | 4189 | if (include_tag && nr_result) |
be18153b | 4190 | for_each_tag_ref(add_ref_tag, NULL); |
4d4fcc54 | 4191 | stop_progress(&progress_state); |
ae417807 DS |
4192 | trace2_region_leave("pack-objects", "enumerate-objects", |
4193 | the_repository); | |
96a02f8f | 4194 | |
f0b0af1b | 4195 | if (non_empty && !nr_result) |
1c4a2912 | 4196 | return 0; |
ae417807 DS |
4197 | if (nr_result) { |
4198 | trace2_region_enter("pack-objects", "prepare-pack", | |
4199 | the_repository); | |
f7ae6a93 | 4200 | prepare_pack(window, depth); |
ae417807 DS |
4201 | trace2_region_leave("pack-objects", "prepare-pack", |
4202 | the_repository); | |
4203 | } | |
4204 | ||
4205 | trace2_region_enter("pack-objects", "write-pack-file", the_repository); | |
dd4b732d | 4206 | write_excluded_by_configs(); |
d01fb92f | 4207 | write_pack_file(); |
ae417807 DS |
4208 | trace2_region_leave("pack-objects", "write-pack-file", the_repository); |
4209 | ||
ab7cd7bb | 4210 | if (progress) |
f616db6a NTND |
4211 | fprintf_ln(stderr, |
4212 | _("Total %"PRIu32" (delta %"PRIu32")," | |
bab28d9f JK |
4213 | " reused %"PRIu32" (delta %"PRIu32")," |
4214 | " pack-reused %"PRIu32), | |
4215 | written, written_delta, reused, reused_delta, | |
4216 | reuse_packfile_objects); | |
c323ac7d LT |
4217 | return 0; |
4218 | } |