]>
Commit | Line | Data |
---|---|---|
5d4a6003 | 1 | #include "builtin.h" |
36bf1958 | 2 | #include "alloc.h" |
32a8f510 | 3 | #include "environment.h" |
f394e093 | 4 | #include "gettext.h" |
41771fa4 | 5 | #include "hex.h" |
a80d72db | 6 | #include "repository.h" |
b2141fc1 | 7 | #include "config.h" |
a74db82e | 8 | #include "attr.h" |
c323ac7d | 9 | #include "object.h" |
8e440259 PE |
10 | #include "blob.h" |
11 | #include "commit.h" | |
12 | #include "tag.h" | |
13 | #include "tree.h" | |
c323ac7d | 14 | #include "delta.h" |
a733cb60 | 15 | #include "pack.h" |
3449f8c4 | 16 | #include "pack-revindex.h" |
c38138cd | 17 | #include "csum-file.h" |
1b0c7174 | 18 | #include "tree-walk.h" |
b5d97e6b JH |
19 | #include "diff.h" |
20 | #include "revision.h" | |
21 | #include "list-objects.h" | |
9535ce73 JH |
22 | #include "list-objects-filter.h" |
23 | #include "list-objects-filter-options.h" | |
2834bc27 | 24 | #include "pack-objects.h" |
96a02f8f | 25 | #include "progress.h" |
f0a24aa5 | 26 | #include "refs.h" |
cf2ba13a | 27 | #include "streaming.h" |
93749194 | 28 | #include "thread-utils.h" |
6b8fda2d | 29 | #include "pack-bitmap.h" |
28b8a730 | 30 | #include "delta-islands.h" |
abcb8655 | 31 | #include "reachable.h" |
fe299ec5 | 32 | #include "oid-array.h" |
dbbcd44f | 33 | #include "strvec.h" |
ec2dd32c | 34 | #include "list.h" |
0317f455 | 35 | #include "packfile.h" |
a80d72db | 36 | #include "object-store.h" |
cbeab747 | 37 | #include "replace-object.h" |
ed7e5fc3 | 38 | #include "dir.h" |
6a22d521 | 39 | #include "midx.h" |
ae417807 | 40 | #include "trace2.h" |
120ad2b0 | 41 | #include "shallow.h" |
e00549aa | 42 | #include "promisor-remote.h" |
b7573536 | 43 | #include "pack-mtimes.h" |
d5ebb50d | 44 | #include "wrapper.h" |
8ecce684 | 45 | |
7d089fb9 ÆAB |
46 | /* |
47 | * Objects we are going to pack are collected in the `to_pack` structure. | |
48 | * It contains an array (dynamically expanded) of the object data, and a map | |
49 | * that can resolve SHA1s to their position in the array. | |
50 | */ | |
51 | static struct packing_data to_pack; | |
52 | ||
53 | static inline struct object_entry *oe_delta( | |
54 | const struct packing_data *pack, | |
55 | const struct object_entry *e) | |
56 | { | |
57 | if (!e->delta_idx) | |
58 | return NULL; | |
59 | if (e->ext_base) | |
60 | return &pack->ext_bases[e->delta_idx - 1]; | |
61 | else | |
62 | return &pack->objects[e->delta_idx - 1]; | |
63 | } | |
64 | ||
65 | static inline unsigned long oe_delta_size(struct packing_data *pack, | |
66 | const struct object_entry *e) | |
67 | { | |
68 | if (e->delta_size_valid) | |
69 | return e->delta_size_; | |
70 | ||
71 | /* | |
72 | * pack->delta_size[] can't be NULL because oe_set_delta_size() | |
73 | * must have been called when a new delta is saved with | |
74 | * oe_set_delta(). | |
75 | * If oe_delta() returns NULL (i.e. default state, which means | |
76 | * delta_size_valid is also false), then the caller must never | |
77 | * call oe_delta_size(). | |
78 | */ | |
79 | return pack->delta_size[e - pack->objects]; | |
80 | } | |
81 | ||
82 | unsigned long oe_get_size_slow(struct packing_data *pack, | |
83 | const struct object_entry *e); | |
84 | ||
85 | static inline unsigned long oe_size(struct packing_data *pack, | |
86 | const struct object_entry *e) | |
87 | { | |
88 | if (e->size_valid) | |
89 | return e->size_; | |
90 | ||
91 | return oe_get_size_slow(pack, e); | |
92 | } | |
93 | ||
94 | static inline void oe_set_delta(struct packing_data *pack, | |
95 | struct object_entry *e, | |
96 | struct object_entry *delta) | |
97 | { | |
98 | if (delta) | |
99 | e->delta_idx = (delta - pack->objects) + 1; | |
100 | else | |
101 | e->delta_idx = 0; | |
102 | } | |
103 | ||
104 | static inline struct object_entry *oe_delta_sibling( | |
105 | const struct packing_data *pack, | |
106 | const struct object_entry *e) | |
107 | { | |
108 | if (e->delta_sibling_idx) | |
109 | return &pack->objects[e->delta_sibling_idx - 1]; | |
110 | return NULL; | |
111 | } | |
112 | ||
113 | static inline struct object_entry *oe_delta_child( | |
114 | const struct packing_data *pack, | |
115 | const struct object_entry *e) | |
116 | { | |
117 | if (e->delta_child_idx) | |
118 | return &pack->objects[e->delta_child_idx - 1]; | |
119 | return NULL; | |
120 | } | |
121 | ||
122 | static inline void oe_set_delta_child(struct packing_data *pack, | |
123 | struct object_entry *e, | |
124 | struct object_entry *delta) | |
125 | { | |
126 | if (delta) | |
127 | e->delta_child_idx = (delta - pack->objects) + 1; | |
128 | else | |
129 | e->delta_child_idx = 0; | |
130 | } | |
131 | ||
132 | static inline void oe_set_delta_sibling(struct packing_data *pack, | |
133 | struct object_entry *e, | |
134 | struct object_entry *delta) | |
135 | { | |
136 | if (delta) | |
137 | e->delta_sibling_idx = (delta - pack->objects) + 1; | |
138 | else | |
139 | e->delta_sibling_idx = 0; | |
140 | } | |
141 | ||
142 | static inline void oe_set_size(struct packing_data *pack, | |
143 | struct object_entry *e, | |
144 | unsigned long size) | |
145 | { | |
146 | if (size < pack->oe_size_limit) { | |
147 | e->size_ = size; | |
148 | e->size_valid = 1; | |
149 | } else { | |
150 | e->size_valid = 0; | |
151 | if (oe_get_size_slow(pack, e) != size) | |
152 | BUG("'size' is supposed to be the object size!"); | |
153 | } | |
154 | } | |
155 | ||
156 | static inline void oe_set_delta_size(struct packing_data *pack, | |
157 | struct object_entry *e, | |
158 | unsigned long size) | |
159 | { | |
160 | if (size < pack->oe_delta_size_limit) { | |
161 | e->delta_size_ = size; | |
162 | e->delta_size_valid = 1; | |
163 | } else { | |
164 | packing_data_lock(pack); | |
165 | if (!pack->delta_size) | |
166 | ALLOC_ARRAY(pack->delta_size, pack->nr_alloc); | |
167 | packing_data_unlock(pack); | |
168 | ||
169 | pack->delta_size[e - pack->objects] = size; | |
170 | e->delta_size_valid = 0; | |
171 | } | |
172 | } | |
173 | ||
43fa44fa | 174 | #define IN_PACK(obj) oe_in_pack(&to_pack, obj) |
ac77d0c3 NTND |
175 | #define SIZE(obj) oe_size(&to_pack, obj) |
176 | #define SET_SIZE(obj,size) oe_set_size(&to_pack, obj, size) | |
0aca34e8 | 177 | #define DELTA_SIZE(obj) oe_delta_size(&to_pack, obj) |
898eba5e NTND |
178 | #define DELTA(obj) oe_delta(&to_pack, obj) |
179 | #define DELTA_CHILD(obj) oe_delta_child(&to_pack, obj) | |
180 | #define DELTA_SIBLING(obj) oe_delta_sibling(&to_pack, obj) | |
181 | #define SET_DELTA(obj, val) oe_set_delta(&to_pack, obj, val) | |
6a1e32d5 | 182 | #define SET_DELTA_EXT(obj, oid) oe_set_delta_ext(&to_pack, obj, oid) |
0aca34e8 | 183 | #define SET_DELTA_SIZE(obj, val) oe_set_delta_size(&to_pack, obj, val) |
898eba5e NTND |
184 | #define SET_DELTA_CHILD(obj, val) oe_set_delta_child(&to_pack, obj, val) |
185 | #define SET_DELTA_SIBLING(obj, val) oe_set_delta_sibling(&to_pack, obj, val) | |
43fa44fa | 186 | |
99fb6e04 | 187 | static const char *pack_usage[] = { |
23a9235d ÆAB |
188 | N_("git pack-objects --stdout [<options>] [< <ref-list> | < <object-list>]"), |
189 | N_("git pack-objects [<options>] <base-name> [< <ref-list> | < <object-list>]"), | |
99fb6e04 NTND |
190 | NULL |
191 | }; | |
c323ac7d | 192 | |
79814f42 | 193 | static struct pack_idx_entry **written_list; |
5af05043 | 194 | static uint32_t nr_result, nr_written, nr_seen; |
6a1e32d5 | 195 | static struct bitmap_index *bitmap_git; |
28b8a730 | 196 | static uint32_t write_layer; |
3f9ac8d2 | 197 | |
96f1e58f | 198 | static int non_empty; |
a7de7130 | 199 | static int reuse_delta = 1, reuse_object = 1; |
e5e9714a | 200 | static int keep_unreachable, unpack_unreachable, include_tag; |
dddbad72 | 201 | static timestamp_t unpack_unreachable_expiration; |
e26a8c47 | 202 | static int pack_loose_unreachable; |
b7573536 TB |
203 | static int cruft; |
204 | static timestamp_t cruft_expiration; | |
96f1e58f | 205 | static int local; |
56dfeb62 | 206 | static int have_non_local_packs; |
96f1e58f | 207 | static int incremental; |
ed7e5fc3 NTND |
208 | static int ignore_packed_keep_on_disk; |
209 | static int ignore_packed_keep_in_core; | |
be6b1914 | 210 | static int allow_ofs_delta; |
ebcfb379 | 211 | static struct pack_idx_option pack_idx_opts; |
d01fb92f | 212 | static const char *base_name; |
024701f1 | 213 | static int progress = 1; |
4812a93a | 214 | static int window = 10; |
568508e7 | 215 | static unsigned long pack_size_limit; |
618e613a | 216 | static int depth = 50; |
43cc2b42 | 217 | static int delta_search_threads; |
df6d6101 | 218 | static int pack_to_stdout; |
4f6d26b1 | 219 | static int sparse; |
6a1e32d5 | 220 | static int thin; |
8d1d8f83 | 221 | static int num_preferred_base; |
dc6a0757 | 222 | static struct progress *progress_state; |
c323ac7d | 223 | |
6b8fda2d VM |
224 | static struct packed_git *reuse_packfile; |
225 | static uint32_t reuse_packfile_objects; | |
bb514de3 | 226 | static struct bitmap *reuse_packfile_bitmap; |
6b8fda2d | 227 | |
645c432d KS |
228 | static int use_bitmap_index_default = 1; |
229 | static int use_bitmap_index = -1; | |
e704fc79 | 230 | static int allow_pack_reuse = 1; |
25575015 JK |
231 | static enum { |
232 | WRITE_BITMAP_FALSE = 0, | |
233 | WRITE_BITMAP_QUIET, | |
234 | WRITE_BITMAP_TRUE, | |
235 | } write_bitmap_index; | |
d4316604 | 236 | static uint16_t write_bitmap_options = BITMAP_OPT_HASH_CACHE; |
6b8fda2d | 237 | |
0c16cd49 JT |
238 | static int exclude_promisor_objects; |
239 | ||
28b8a730 JK |
240 | static int use_delta_islands; |
241 | ||
074b2eea | 242 | static unsigned long delta_cache_size = 0; |
9806f5a7 | 243 | static unsigned long max_delta_cache_size = DEFAULT_DELTA_CACHE_SIZE; |
e3dfddb3 | 244 | static unsigned long cache_max_small_delta_size = 1000; |
074b2eea | 245 | |
a97773ce BD |
246 | static unsigned long window_memory_limit = 0; |
247 | ||
dd4b732d JT |
248 | static struct string_list uri_protocols = STRING_LIST_INIT_NODUP; |
249 | ||
9535ce73 | 250 | enum missing_action { |
0c16cd49 JT |
251 | MA_ERROR = 0, /* fail if any missing objects are encountered */ |
252 | MA_ALLOW_ANY, /* silently allow ALL missing objects */ | |
253 | MA_ALLOW_PROMISOR, /* silently allow all missing PROMISOR objects */ | |
9535ce73 JH |
254 | }; |
255 | static enum missing_action arg_missing_action; | |
256 | static show_object_fn fn_show_object; | |
257 | ||
dd4b732d JT |
258 | struct configured_exclusion { |
259 | struct oidmap_entry e; | |
260 | char *pack_hash_hex; | |
261 | char *uri; | |
262 | }; | |
263 | static struct oidmap configured_exclusions; | |
264 | ||
265 | static struct oidset excluded_by_config; | |
266 | ||
3f9ac8d2 JH |
267 | /* |
268 | * stats | |
269 | */ | |
7cadf491 SP |
270 | static uint32_t written, written_delta; |
271 | static uint32_t reused, reused_delta; | |
3f9ac8d2 | 272 | |
7cc8f971 VM |
273 | /* |
274 | * Indexed commits | |
275 | */ | |
276 | static struct commit **indexed_commits; | |
277 | static unsigned int indexed_commits_nr; | |
278 | static unsigned int indexed_commits_alloc; | |
279 | ||
280 | static void index_commit_for_bitmap(struct commit *commit) | |
281 | { | |
282 | if (indexed_commits_nr >= indexed_commits_alloc) { | |
283 | indexed_commits_alloc = (indexed_commits_alloc + 32) * 2; | |
2756ca43 | 284 | REALLOC_ARRAY(indexed_commits, indexed_commits_alloc); |
7cc8f971 VM |
285 | } |
286 | ||
287 | indexed_commits[indexed_commits_nr++] = commit; | |
288 | } | |
780e6e73 | 289 | |
3613f9b4 | 290 | static void *get_delta(struct object_entry *entry) |
c323ac7d | 291 | { |
3613f9b4 NP |
292 | unsigned long size, base_size, delta_size; |
293 | void *buf, *base_buf, *delta_buf; | |
21666f1a | 294 | enum object_type type; |
c323ac7d | 295 | |
b4f5aca4 | 296 | buf = read_object_file(&entry->idx.oid, &type, &size); |
3613f9b4 | 297 | if (!buf) |
f616db6a | 298 | die(_("unable to read %s"), oid_to_hex(&entry->idx.oid)); |
898eba5e NTND |
299 | base_buf = read_object_file(&DELTA(entry)->idx.oid, &type, |
300 | &base_size); | |
3613f9b4 | 301 | if (!base_buf) |
e6a492b7 | 302 | die("unable to read %s", |
898eba5e | 303 | oid_to_hex(&DELTA(entry)->idx.oid)); |
3613f9b4 | 304 | delta_buf = diff_delta(base_buf, base_size, |
dcde55bc | 305 | buf, size, &delta_size, 0); |
1a07e59c | 306 | /* |
15beaaa3 | 307 | * We successfully computed this delta once but dropped it for |
1a07e59c NTND |
308 | * memory reasons. Something is very wrong if this time we |
309 | * recompute and create a different delta. | |
310 | */ | |
0aca34e8 | 311 | if (!delta_buf || delta_size != DELTA_SIZE(entry)) |
1a07e59c | 312 | BUG("delta size changed"); |
3613f9b4 NP |
313 | free(buf); |
314 | free(base_buf); | |
c323ac7d LT |
315 | return delta_buf; |
316 | } | |
317 | ||
30ebb40a NP |
318 | static unsigned long do_compress(void **pptr, unsigned long size) |
319 | { | |
ef49a7a0 | 320 | git_zstream stream; |
30ebb40a NP |
321 | void *in, *out; |
322 | unsigned long maxsize; | |
323 | ||
55bb5c91 | 324 | git_deflate_init(&stream, pack_compression_level); |
225a6f10 | 325 | maxsize = git_deflate_bound(&stream, size); |
30ebb40a NP |
326 | |
327 | in = *pptr; | |
328 | out = xmalloc(maxsize); | |
329 | *pptr = out; | |
330 | ||
331 | stream.next_in = in; | |
332 | stream.avail_in = size; | |
333 | stream.next_out = out; | |
334 | stream.avail_out = maxsize; | |
55bb5c91 | 335 | while (git_deflate(&stream, Z_FINISH) == Z_OK) |
30ebb40a | 336 | ; /* nothing */ |
55bb5c91 | 337 | git_deflate_end(&stream); |
30ebb40a NP |
338 | |
339 | free(in); | |
340 | return stream.total_out; | |
341 | } | |
342 | ||
98a3beab | 343 | static unsigned long write_large_blob_data(struct git_istream *st, struct hashfile *f, |
188960b4 | 344 | const struct object_id *oid) |
cf2ba13a NTND |
345 | { |
346 | git_zstream stream; | |
347 | unsigned char ibuf[1024 * 16]; | |
348 | unsigned char obuf[1024 * 16]; | |
349 | unsigned long olen = 0; | |
350 | ||
cf2ba13a NTND |
351 | git_deflate_init(&stream, pack_compression_level); |
352 | ||
353 | for (;;) { | |
354 | ssize_t readlen; | |
355 | int zret = Z_OK; | |
356 | readlen = read_istream(st, ibuf, sizeof(ibuf)); | |
357 | if (readlen == -1) | |
188960b4 | 358 | die(_("unable to read %s"), oid_to_hex(oid)); |
cf2ba13a NTND |
359 | |
360 | stream.next_in = ibuf; | |
361 | stream.avail_in = readlen; | |
362 | while ((stream.avail_in || readlen == 0) && | |
363 | (zret == Z_OK || zret == Z_BUF_ERROR)) { | |
364 | stream.next_out = obuf; | |
365 | stream.avail_out = sizeof(obuf); | |
366 | zret = git_deflate(&stream, readlen ? 0 : Z_FINISH); | |
98a3beab | 367 | hashwrite(f, obuf, stream.next_out - obuf); |
cf2ba13a NTND |
368 | olen += stream.next_out - obuf; |
369 | } | |
370 | if (stream.avail_in) | |
371 | die(_("deflate error (%d)"), zret); | |
372 | if (readlen == 0) { | |
373 | if (zret != Z_STREAM_END) | |
374 | die(_("deflate error (%d)"), zret); | |
375 | break; | |
376 | } | |
377 | } | |
378 | git_deflate_end(&stream); | |
379 | return olen; | |
380 | } | |
381 | ||
780e6e73 NP |
382 | /* |
383 | * we are going to reuse the existing object data as is. make | |
384 | * sure it is not corrupt. | |
385 | */ | |
079afb18 SP |
386 | static int check_pack_inflate(struct packed_git *p, |
387 | struct pack_window **w_curs, | |
6777a59f SP |
388 | off_t offset, |
389 | off_t len, | |
079afb18 SP |
390 | unsigned long expect) |
391 | { | |
ef49a7a0 | 392 | git_zstream stream; |
079afb18 SP |
393 | unsigned char fakebuf[4096], *in; |
394 | int st; | |
395 | ||
396 | memset(&stream, 0, sizeof(stream)); | |
39c68542 | 397 | git_inflate_init(&stream); |
079afb18 SP |
398 | do { |
399 | in = use_pack(p, w_curs, offset, &stream.avail_in); | |
400 | stream.next_in = in; | |
401 | stream.next_out = fakebuf; | |
402 | stream.avail_out = sizeof(fakebuf); | |
39c68542 | 403 | st = git_inflate(&stream, Z_FINISH); |
079afb18 SP |
404 | offset += stream.next_in - in; |
405 | } while (st == Z_OK || st == Z_BUF_ERROR); | |
39c68542 | 406 | git_inflate_end(&stream); |
079afb18 SP |
407 | return (st == Z_STREAM_END && |
408 | stream.total_out == expect && | |
409 | stream.total_in == len) ? 0 : -1; | |
410 | } | |
411 | ||
98a3beab | 412 | static void copy_pack_data(struct hashfile *f, |
079afb18 SP |
413 | struct packed_git *p, |
414 | struct pack_window **w_curs, | |
6777a59f SP |
415 | off_t offset, |
416 | off_t len) | |
079afb18 SP |
417 | { |
418 | unsigned char *in; | |
ef49a7a0 | 419 | unsigned long avail; |
079afb18 SP |
420 | |
421 | while (len) { | |
422 | in = use_pack(p, w_curs, offset, &avail); | |
423 | if (avail > len) | |
ef49a7a0 | 424 | avail = (unsigned long)len; |
98a3beab | 425 | hashwrite(f, in, avail); |
079afb18 SP |
426 | offset += avail; |
427 | len -= avail; | |
428 | } | |
429 | } | |
430 | ||
7d089fb9 ÆAB |
431 | static inline int oe_size_greater_than(struct packing_data *pack, |
432 | const struct object_entry *lhs, | |
433 | unsigned long rhs) | |
434 | { | |
435 | if (lhs->size_valid) | |
436 | return lhs->size_ > rhs; | |
437 | if (rhs < pack->oe_size_limit) /* rhs < 2^x <= lhs ? */ | |
438 | return 1; | |
439 | return oe_get_size_slow(pack, lhs) > rhs; | |
440 | } | |
441 | ||
1b4bb16b | 442 | /* Return 0 if we will bust the pack-size limit */ |
98a3beab | 443 | static unsigned long write_no_reuse_object(struct hashfile *f, struct object_entry *entry, |
c9018b03 | 444 | unsigned long limit, int usable_delta) |
c323ac7d | 445 | { |
c9018b03 | 446 | unsigned long size, datalen; |
2c5e2865 JK |
447 | unsigned char header[MAX_PACK_OBJECT_HEADER], |
448 | dheader[MAX_PACK_OBJECT_HEADER]; | |
6777a59f | 449 | unsigned hdrlen; |
2c5ef824 | 450 | enum object_type type; |
c9018b03 | 451 | void *buf; |
cf2ba13a | 452 | struct git_istream *st = NULL; |
41179100 | 453 | const unsigned hashsz = the_hash_algo->rawsz; |
c9018b03 NTND |
454 | |
455 | if (!usable_delta) { | |
fd9b1bae | 456 | if (oe_type(entry) == OBJ_BLOB && |
ac77d0c3 | 457 | oe_size_greater_than(&to_pack, entry, big_file_threshold) && |
c8123e72 MT |
458 | (st = open_istream(the_repository, &entry->idx.oid, &type, |
459 | &size, NULL)) != NULL) | |
cf2ba13a NTND |
460 | buf = NULL; |
461 | else { | |
b4f5aca4 | 462 | buf = read_object_file(&entry->idx.oid, &type, &size); |
cf2ba13a | 463 | if (!buf) |
e6a492b7 | 464 | die(_("unable to read %s"), |
465 | oid_to_hex(&entry->idx.oid)); | |
cf2ba13a | 466 | } |
c9018b03 NTND |
467 | /* |
468 | * make sure no cached delta data remains from a | |
469 | * previous attempt before a pack split occurred. | |
470 | */ | |
6a83d902 | 471 | FREE_AND_NULL(entry->delta_data); |
c9018b03 NTND |
472 | entry->z_delta_size = 0; |
473 | } else if (entry->delta_data) { | |
0aca34e8 | 474 | size = DELTA_SIZE(entry); |
c9018b03 NTND |
475 | buf = entry->delta_data; |
476 | entry->delta_data = NULL; | |
898eba5e | 477 | type = (allow_ofs_delta && DELTA(entry)->idx.offset) ? |
c9018b03 NTND |
478 | OBJ_OFS_DELTA : OBJ_REF_DELTA; |
479 | } else { | |
480 | buf = get_delta(entry); | |
0aca34e8 | 481 | size = DELTA_SIZE(entry); |
898eba5e | 482 | type = (allow_ofs_delta && DELTA(entry)->idx.offset) ? |
c9018b03 NTND |
483 | OBJ_OFS_DELTA : OBJ_REF_DELTA; |
484 | } | |
485 | ||
cf2ba13a NTND |
486 | if (st) /* large blob case, just assume we don't compress well */ |
487 | datalen = size; | |
488 | else if (entry->z_delta_size) | |
c9018b03 NTND |
489 | datalen = entry->z_delta_size; |
490 | else | |
491 | datalen = do_compress(&buf, size); | |
492 | ||
493 | /* | |
494 | * The object header is a byte of 'type' followed by zero or | |
495 | * more bytes of length. | |
496 | */ | |
7202a6fa JK |
497 | hdrlen = encode_in_pack_object_header(header, sizeof(header), |
498 | type, size); | |
c9018b03 NTND |
499 | |
500 | if (type == OBJ_OFS_DELTA) { | |
501 | /* | |
502 | * Deltas with relative base contain an additional | |
503 | * encoding of the relative offset for the delta | |
504 | * base from this object's position in the pack. | |
505 | */ | |
898eba5e | 506 | off_t ofs = entry->idx.offset - DELTA(entry)->idx.offset; |
c9018b03 NTND |
507 | unsigned pos = sizeof(dheader) - 1; |
508 | dheader[pos] = ofs & 127; | |
509 | while (ofs >>= 7) | |
510 | dheader[--pos] = 128 | (--ofs & 127); | |
41179100 | 511 | if (limit && hdrlen + sizeof(dheader) - pos + datalen + hashsz >= limit) { |
cf2ba13a NTND |
512 | if (st) |
513 | close_istream(st); | |
c9018b03 NTND |
514 | free(buf); |
515 | return 0; | |
516 | } | |
98a3beab | 517 | hashwrite(f, header, hdrlen); |
518 | hashwrite(f, dheader + pos, sizeof(dheader) - pos); | |
c9018b03 NTND |
519 | hdrlen += sizeof(dheader) - pos; |
520 | } else if (type == OBJ_REF_DELTA) { | |
521 | /* | |
522 | * Deltas with a base reference contain | |
41179100 | 523 | * additional bytes for the base object ID. |
c9018b03 | 524 | */ |
41179100 | 525 | if (limit && hdrlen + hashsz + datalen + hashsz >= limit) { |
cf2ba13a NTND |
526 | if (st) |
527 | close_istream(st); | |
c9018b03 NTND |
528 | free(buf); |
529 | return 0; | |
530 | } | |
98a3beab | 531 | hashwrite(f, header, hdrlen); |
42c8ce1c | 532 | hashwrite(f, DELTA(entry)->idx.oid.hash, hashsz); |
41179100 | 533 | hdrlen += hashsz; |
c9018b03 | 534 | } else { |
41179100 | 535 | if (limit && hdrlen + datalen + hashsz >= limit) { |
cf2ba13a NTND |
536 | if (st) |
537 | close_istream(st); | |
c9018b03 NTND |
538 | free(buf); |
539 | return 0; | |
540 | } | |
98a3beab | 541 | hashwrite(f, header, hdrlen); |
c9018b03 | 542 | } |
cf2ba13a | 543 | if (st) { |
188960b4 | 544 | datalen = write_large_blob_data(st, f, &entry->idx.oid); |
cf2ba13a NTND |
545 | close_istream(st); |
546 | } else { | |
98a3beab | 547 | hashwrite(f, buf, datalen); |
cf2ba13a NTND |
548 | free(buf); |
549 | } | |
c9018b03 NTND |
550 | |
551 | return hdrlen + datalen; | |
552 | } | |
553 | ||
554 | /* Return 0 if we will bust the pack-size limit */ | |
98a3beab | 555 | static off_t write_reuse_object(struct hashfile *f, struct object_entry *entry, |
af92a645 | 556 | unsigned long limit, int usable_delta) |
c9018b03 | 557 | { |
43fa44fa | 558 | struct packed_git *p = IN_PACK(entry); |
c9018b03 | 559 | struct pack_window *w_curs = NULL; |
952fc687 | 560 | uint32_t pos; |
c9018b03 | 561 | off_t offset; |
fd9b1bae | 562 | enum object_type type = oe_type(entry); |
211c61c6 | 563 | off_t datalen; |
2c5e2865 JK |
564 | unsigned char header[MAX_PACK_OBJECT_HEADER], |
565 | dheader[MAX_PACK_OBJECT_HEADER]; | |
c9018b03 | 566 | unsigned hdrlen; |
41179100 | 567 | const unsigned hashsz = the_hash_algo->rawsz; |
ac77d0c3 | 568 | unsigned long entry_size = SIZE(entry); |
c9018b03 | 569 | |
898eba5e NTND |
570 | if (DELTA(entry)) |
571 | type = (allow_ofs_delta && DELTA(entry)->idx.offset) ? | |
c9018b03 | 572 | OBJ_OFS_DELTA : OBJ_REF_DELTA; |
7202a6fa | 573 | hdrlen = encode_in_pack_object_header(header, sizeof(header), |
ac77d0c3 | 574 | type, entry_size); |
c9018b03 NTND |
575 | |
576 | offset = entry->in_pack_offset; | |
952fc687 TB |
577 | if (offset_to_pack_pos(p, offset, &pos) < 0) |
578 | die(_("write_reuse_object: could not locate %s, expected at " | |
579 | "offset %"PRIuMAX" in pack %s"), | |
580 | oid_to_hex(&entry->idx.oid), (uintmax_t)offset, | |
581 | p->pack_name); | |
582 | datalen = pack_pos_to_offset(p, pos + 1) - offset; | |
c9018b03 | 583 | if (!pack_to_stdout && p->index_version > 1 && |
952fc687 TB |
584 | check_pack_crc(p, &w_curs, offset, datalen, |
585 | pack_pos_to_index(p, pos))) { | |
f616db6a | 586 | error(_("bad packed object CRC for %s"), |
e6a492b7 | 587 | oid_to_hex(&entry->idx.oid)); |
c9018b03 NTND |
588 | unuse_pack(&w_curs); |
589 | return write_no_reuse_object(f, entry, limit, usable_delta); | |
590 | } | |
591 | ||
592 | offset += entry->in_pack_header_size; | |
593 | datalen -= entry->in_pack_header_size; | |
594 | ||
595 | if (!pack_to_stdout && p->index_version == 1 && | |
ac77d0c3 | 596 | check_pack_inflate(p, &w_curs, offset, datalen, entry_size)) { |
f616db6a | 597 | error(_("corrupt packed object for %s"), |
e6a492b7 | 598 | oid_to_hex(&entry->idx.oid)); |
c9018b03 NTND |
599 | unuse_pack(&w_curs); |
600 | return write_no_reuse_object(f, entry, limit, usable_delta); | |
601 | } | |
602 | ||
603 | if (type == OBJ_OFS_DELTA) { | |
898eba5e | 604 | off_t ofs = entry->idx.offset - DELTA(entry)->idx.offset; |
c9018b03 NTND |
605 | unsigned pos = sizeof(dheader) - 1; |
606 | dheader[pos] = ofs & 127; | |
607 | while (ofs >>= 7) | |
608 | dheader[--pos] = 128 | (--ofs & 127); | |
41179100 | 609 | if (limit && hdrlen + sizeof(dheader) - pos + datalen + hashsz >= limit) { |
c9018b03 NTND |
610 | unuse_pack(&w_curs); |
611 | return 0; | |
612 | } | |
98a3beab | 613 | hashwrite(f, header, hdrlen); |
614 | hashwrite(f, dheader + pos, sizeof(dheader) - pos); | |
c9018b03 NTND |
615 | hdrlen += sizeof(dheader) - pos; |
616 | reused_delta++; | |
617 | } else if (type == OBJ_REF_DELTA) { | |
41179100 | 618 | if (limit && hdrlen + hashsz + datalen + hashsz >= limit) { |
c9018b03 NTND |
619 | unuse_pack(&w_curs); |
620 | return 0; | |
621 | } | |
98a3beab | 622 | hashwrite(f, header, hdrlen); |
42c8ce1c | 623 | hashwrite(f, DELTA(entry)->idx.oid.hash, hashsz); |
41179100 | 624 | hdrlen += hashsz; |
c9018b03 NTND |
625 | reused_delta++; |
626 | } else { | |
41179100 | 627 | if (limit && hdrlen + datalen + hashsz >= limit) { |
c9018b03 NTND |
628 | unuse_pack(&w_curs); |
629 | return 0; | |
630 | } | |
98a3beab | 631 | hashwrite(f, header, hdrlen); |
c9018b03 NTND |
632 | } |
633 | copy_pack_data(f, p, &w_curs, offset, datalen); | |
634 | unuse_pack(&w_curs); | |
635 | reused++; | |
636 | return hdrlen + datalen; | |
637 | } | |
638 | ||
639 | /* Return 0 if we will bust the pack-size limit */ | |
98a3beab | 640 | static off_t write_object(struct hashfile *f, |
af92a645 NTND |
641 | struct object_entry *entry, |
642 | off_t write_offset) | |
c9018b03 | 643 | { |
af92a645 NTND |
644 | unsigned long limit; |
645 | off_t len; | |
2c5ef824 | 646 | int usable_delta, to_reuse; |
c323ac7d | 647 | |
78d1e84f NP |
648 | if (!pack_to_stdout) |
649 | crc32_begin(f); | |
650 | ||
a2430dde | 651 | /* apply size limit if limited packsize and not first object */ |
a1e4760f NP |
652 | if (!pack_size_limit || !nr_written) |
653 | limit = 0; | |
654 | else if (pack_size_limit <= write_offset) | |
655 | /* | |
656 | * the earlier object did not fit the limit; avoid | |
657 | * mistaking this with unlimited (i.e. limit = 0). | |
658 | */ | |
659 | limit = 1; | |
660 | else | |
661 | limit = pack_size_limit - write_offset; | |
2c5ef824 | 662 | |
898eba5e | 663 | if (!DELTA(entry)) |
2c5ef824 NP |
664 | usable_delta = 0; /* no delta */ |
665 | else if (!pack_size_limit) | |
666 | usable_delta = 1; /* unlimited packfile */ | |
898eba5e | 667 | else if (DELTA(entry)->idx.offset == (off_t)-1) |
2c5ef824 | 668 | usable_delta = 0; /* base was written to another pack */ |
898eba5e | 669 | else if (DELTA(entry)->idx.offset) |
2c5ef824 NP |
670 | usable_delta = 1; /* base already exists in this pack */ |
671 | else | |
672 | usable_delta = 0; /* base could end up in another pack */ | |
673 | ||
a7de7130 | 674 | if (!reuse_object) |
fa736f72 | 675 | to_reuse = 0; /* explicit */ |
43fa44fa | 676 | else if (!IN_PACK(entry)) |
ab7cd7bb | 677 | to_reuse = 0; /* can't reuse what we don't have */ |
fd9b1bae NTND |
678 | else if (oe_type(entry) == OBJ_REF_DELTA || |
679 | oe_type(entry) == OBJ_OFS_DELTA) | |
17b08f2c DH |
680 | /* check_object() decided it for us ... */ |
681 | to_reuse = usable_delta; | |
682 | /* ... but pack split may override that */ | |
fd9b1bae | 683 | else if (oe_type(entry) != entry->in_pack_type) |
ab7cd7bb | 684 | to_reuse = 0; /* pack has delta which is unusable */ |
898eba5e | 685 | else if (DELTA(entry)) |
ab7cd7bb JH |
686 | to_reuse = 0; /* we want to pack afresh */ |
687 | else | |
688 | to_reuse = 1; /* we have it in-pack undeltified, | |
689 | * and we do not need to deltify it. | |
690 | */ | |
691 | ||
c9018b03 NTND |
692 | if (!to_reuse) |
693 | len = write_no_reuse_object(f, entry, limit, usable_delta); | |
694 | else | |
695 | len = write_reuse_object(f, entry, limit, usable_delta); | |
696 | if (!len) | |
697 | return 0; | |
64bd76b1 | 698 | |
17b08f2c | 699 | if (usable_delta) |
ab7cd7bb | 700 | written_delta++; |
3f9ac8d2 | 701 | written++; |
78d1e84f | 702 | if (!pack_to_stdout) |
aa7e44bf | 703 | entry->idx.crc32 = crc32_end(f); |
c9018b03 | 704 | return len; |
c323ac7d LT |
705 | } |
706 | ||
f63c79db JH |
707 | enum write_one_status { |
708 | WRITE_ONE_SKIP = -1, /* already written */ | |
709 | WRITE_ONE_BREAK = 0, /* writing this will bust the limit; not written */ | |
710 | WRITE_ONE_WRITTEN = 1, /* normal */ | |
711 | WRITE_ONE_RECURSIVE = 2 /* already scheduled to be written */ | |
712 | }; | |
713 | ||
98a3beab | 714 | static enum write_one_status write_one(struct hashfile *f, |
f63c79db JH |
715 | struct object_entry *e, |
716 | off_t *offset) | |
9d5ab962 | 717 | { |
af92a645 | 718 | off_t size; |
f63c79db | 719 | int recursing; |
d7dd0223 | 720 | |
f63c79db JH |
721 | /* |
722 | * we set offset to 1 (which is an impossible value) to mark | |
723 | * the fact that this object is involved in "write its base | |
724 | * first before writing a deltified object" recursion. | |
725 | */ | |
726 | recursing = (e->idx.offset == 1); | |
727 | if (recursing) { | |
f616db6a | 728 | warning(_("recursive delta detected for object %s"), |
e6a492b7 | 729 | oid_to_hex(&e->idx.oid)); |
f63c79db JH |
730 | return WRITE_ONE_RECURSIVE; |
731 | } else if (e->idx.offset || e->preferred_base) { | |
732 | /* offset is non zero if object is written already. */ | |
733 | return WRITE_ONE_SKIP; | |
734 | } | |
d7dd0223 | 735 | |
720c9f7b | 736 | /* if we are deltified, write out base object first. */ |
898eba5e | 737 | if (DELTA(e)) { |
f63c79db | 738 | e->idx.offset = 1; /* now recurse */ |
898eba5e | 739 | switch (write_one(f, DELTA(e), offset)) { |
f63c79db JH |
740 | case WRITE_ONE_RECURSIVE: |
741 | /* we cannot depend on this one */ | |
898eba5e | 742 | SET_DELTA(e, NULL); |
f63c79db JH |
743 | break; |
744 | default: | |
745 | break; | |
746 | case WRITE_ONE_BREAK: | |
747 | e->idx.offset = recursing; | |
748 | return WRITE_ONE_BREAK; | |
749 | } | |
750 | } | |
d7dd0223 | 751 | |
6ed7f25e NP |
752 | e->idx.offset = *offset; |
753 | size = write_object(f, e, *offset); | |
17b08f2c | 754 | if (!size) { |
f63c79db JH |
755 | e->idx.offset = recursing; |
756 | return WRITE_ONE_BREAK; | |
17b08f2c | 757 | } |
79814f42 | 758 | written_list[nr_written++] = &e->idx; |
d7dd0223 NP |
759 | |
760 | /* make sure off_t is sufficiently large not to wrap */ | |
c03c8315 | 761 | if (signed_add_overflows(*offset, size)) |
f616db6a | 762 | die(_("pack too large for current definition of off_t")); |
6ed7f25e | 763 | *offset += size; |
f63c79db | 764 | return WRITE_ONE_WRITTEN; |
9d5ab962 JH |
765 | } |
766 | ||
5cf88fd8 ÆAB |
767 | static int mark_tagged(const char *path UNUSED, const struct object_id *oid, |
768 | int flag UNUSED, void *cb_data UNUSED) | |
1b4bb16b | 769 | { |
188960b4 | 770 | struct object_id peeled; |
3a37876b | 771 | struct object_entry *entry = packlist_find(&to_pack, oid); |
1b4bb16b JH |
772 | |
773 | if (entry) | |
774 | entry->tagged = 1; | |
36a31792 | 775 | if (!peel_iterated_oid(oid, &peeled)) { |
3a37876b | 776 | entry = packlist_find(&to_pack, &peeled); |
1b4bb16b JH |
777 | if (entry) |
778 | entry->tagged = 1; | |
779 | } | |
780 | return 0; | |
781 | } | |
782 | ||
7d089fb9 ÆAB |
783 | static inline unsigned char oe_layer(struct packing_data *pack, |
784 | struct object_entry *e) | |
785 | { | |
786 | if (!pack->layer) | |
787 | return 0; | |
788 | return pack->layer[e - pack->objects]; | |
789 | } | |
790 | ||
be126818 | 791 | static inline void add_to_write_order(struct object_entry **wo, |
92bef1a1 | 792 | unsigned int *endp, |
1b4bb16b JH |
793 | struct object_entry *e) |
794 | { | |
fe0ac2fb | 795 | if (e->filled || oe_layer(&to_pack, e) != write_layer) |
1b4bb16b JH |
796 | return; |
797 | wo[(*endp)++] = e; | |
798 | e->filled = 1; | |
799 | } | |
800 | ||
801 | static void add_descendants_to_write_order(struct object_entry **wo, | |
92bef1a1 | 802 | unsigned int *endp, |
1b4bb16b JH |
803 | struct object_entry *e) |
804 | { | |
f380872f DM |
805 | int add_to_order = 1; |
806 | while (e) { | |
807 | if (add_to_order) { | |
808 | struct object_entry *s; | |
809 | /* add this node... */ | |
810 | add_to_write_order(wo, endp, e); | |
811 | /* all its siblings... */ | |
898eba5e | 812 | for (s = DELTA_SIBLING(e); s; s = DELTA_SIBLING(s)) { |
f380872f DM |
813 | add_to_write_order(wo, endp, s); |
814 | } | |
815 | } | |
816 | /* drop down a level to add left subtree nodes if possible */ | |
898eba5e | 817 | if (DELTA_CHILD(e)) { |
f380872f | 818 | add_to_order = 1; |
898eba5e | 819 | e = DELTA_CHILD(e); |
f380872f DM |
820 | } else { |
821 | add_to_order = 0; | |
822 | /* our sibling might have some children, it is next */ | |
898eba5e NTND |
823 | if (DELTA_SIBLING(e)) { |
824 | e = DELTA_SIBLING(e); | |
f380872f DM |
825 | continue; |
826 | } | |
827 | /* go back to our parent node */ | |
898eba5e NTND |
828 | e = DELTA(e); |
829 | while (e && !DELTA_SIBLING(e)) { | |
f380872f DM |
830 | /* we're on the right side of a subtree, keep |
831 | * going up until we can go right again */ | |
898eba5e | 832 | e = DELTA(e); |
f380872f DM |
833 | } |
834 | if (!e) { | |
835 | /* done- we hit our original root node */ | |
836 | return; | |
837 | } | |
838 | /* pass it off to sibling at this level */ | |
898eba5e | 839 | e = DELTA_SIBLING(e); |
f380872f DM |
840 | } |
841 | }; | |
1b4bb16b JH |
842 | } |
843 | ||
844 | static void add_family_to_write_order(struct object_entry **wo, | |
92bef1a1 | 845 | unsigned int *endp, |
1b4bb16b JH |
846 | struct object_entry *e) |
847 | { | |
848 | struct object_entry *root; | |
849 | ||
898eba5e | 850 | for (root = e; DELTA(root); root = DELTA(root)) |
1b4bb16b | 851 | ; /* nothing */ |
1b4bb16b JH |
852 | add_descendants_to_write_order(wo, endp, root); |
853 | } | |
854 | ||
f64ba53e | 855 | static void compute_layer_order(struct object_entry **wo, unsigned int *wo_end) |
1b4bb16b | 856 | { |
f64ba53e | 857 | unsigned int i, last_untagged; |
2834bc27 | 858 | struct object_entry *objects = to_pack.objects; |
1b4bb16b | 859 | |
2834bc27 | 860 | for (i = 0; i < to_pack.nr_objects; i++) { |
1b4bb16b JH |
861 | if (objects[i].tagged) |
862 | break; | |
f64ba53e | 863 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b | 864 | } |
38d4debb | 865 | last_untagged = i; |
1b4bb16b JH |
866 | |
867 | /* | |
868 | * Then fill all the tagged tips. | |
869 | */ | |
2834bc27 | 870 | for (; i < to_pack.nr_objects; i++) { |
1b4bb16b | 871 | if (objects[i].tagged) |
f64ba53e | 872 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b JH |
873 | } |
874 | ||
875 | /* | |
876 | * And then all remaining commits and tags. | |
877 | */ | |
2834bc27 | 878 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
fd9b1bae NTND |
879 | if (oe_type(&objects[i]) != OBJ_COMMIT && |
880 | oe_type(&objects[i]) != OBJ_TAG) | |
1b4bb16b | 881 | continue; |
f64ba53e | 882 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b JH |
883 | } |
884 | ||
885 | /* | |
886 | * And then all the trees. | |
887 | */ | |
2834bc27 | 888 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
fd9b1bae | 889 | if (oe_type(&objects[i]) != OBJ_TREE) |
1b4bb16b | 890 | continue; |
f64ba53e | 891 | add_to_write_order(wo, wo_end, &objects[i]); |
1b4bb16b JH |
892 | } |
893 | ||
894 | /* | |
895 | * Finally all the rest in really tight order | |
896 | */ | |
2834bc27 | 897 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
fe0ac2fb | 898 | if (!objects[i].filled && oe_layer(&to_pack, &objects[i]) == write_layer) |
f64ba53e | 899 | add_family_to_write_order(wo, wo_end, &objects[i]); |
38d4debb | 900 | } |
f64ba53e CC |
901 | } |
902 | ||
903 | static struct object_entry **compute_write_order(void) | |
904 | { | |
28b8a730 | 905 | uint32_t max_layers = 1; |
f64ba53e CC |
906 | unsigned int i, wo_end; |
907 | ||
908 | struct object_entry **wo; | |
909 | struct object_entry *objects = to_pack.objects; | |
910 | ||
911 | for (i = 0; i < to_pack.nr_objects; i++) { | |
912 | objects[i].tagged = 0; | |
913 | objects[i].filled = 0; | |
914 | SET_DELTA_CHILD(&objects[i], NULL); | |
915 | SET_DELTA_SIBLING(&objects[i], NULL); | |
916 | } | |
917 | ||
918 | /* | |
919 | * Fully connect delta_child/delta_sibling network. | |
920 | * Make sure delta_sibling is sorted in the original | |
921 | * recency order. | |
922 | */ | |
923 | for (i = to_pack.nr_objects; i > 0;) { | |
924 | struct object_entry *e = &objects[--i]; | |
925 | if (!DELTA(e)) | |
926 | continue; | |
927 | /* Mark me as the first child */ | |
928 | e->delta_sibling_idx = DELTA(e)->delta_child_idx; | |
929 | SET_DELTA_CHILD(DELTA(e), e); | |
38d4debb DM |
930 | } |
931 | ||
f64ba53e CC |
932 | /* |
933 | * Mark objects that are at the tip of tags. | |
934 | */ | |
935 | for_each_tag_ref(mark_tagged, NULL); | |
936 | ||
647982bb | 937 | if (use_delta_islands) { |
28b8a730 | 938 | max_layers = compute_pack_layers(&to_pack); |
647982bb EW |
939 | free_island_marks(); |
940 | } | |
28b8a730 | 941 | |
f64ba53e CC |
942 | ALLOC_ARRAY(wo, to_pack.nr_objects); |
943 | wo_end = 0; | |
944 | ||
28b8a730 JK |
945 | for (; write_layer < max_layers; ++write_layer) |
946 | compute_layer_order(wo, &wo_end); | |
38d4debb | 947 | |
2834bc27 | 948 | if (wo_end != to_pack.nr_objects) |
f616db6a NTND |
949 | die(_("ordered %u objects, expected %"PRIu32), |
950 | wo_end, to_pack.nr_objects); | |
1b4bb16b JH |
951 | |
952 | return wo; | |
953 | } | |
954 | ||
bb514de3 JK |
955 | |
956 | /* | |
957 | * A reused set of objects. All objects in a chunk have the same | |
958 | * relative position in the original packfile and the generated | |
959 | * packfile. | |
960 | */ | |
961 | ||
962 | static struct reused_chunk { | |
963 | /* The offset of the first object of this chunk in the original | |
964 | * packfile. */ | |
965 | off_t original; | |
bf12013f HX |
966 | /* The difference for "original" minus the offset of the first object of |
967 | * this chunk in the generated packfile. */ | |
bb514de3 JK |
968 | off_t difference; |
969 | } *reused_chunks; | |
970 | static int reused_chunks_nr; | |
971 | static int reused_chunks_alloc; | |
972 | ||
973 | static void record_reused_object(off_t where, off_t offset) | |
974 | { | |
975 | if (reused_chunks_nr && reused_chunks[reused_chunks_nr-1].difference == offset) | |
976 | return; | |
977 | ||
978 | ALLOC_GROW(reused_chunks, reused_chunks_nr + 1, | |
979 | reused_chunks_alloc); | |
980 | reused_chunks[reused_chunks_nr].original = where; | |
981 | reused_chunks[reused_chunks_nr].difference = offset; | |
982 | reused_chunks_nr++; | |
983 | } | |
984 | ||
985 | /* | |
986 | * Binary search to find the chunk that "where" is in. Note | |
987 | * that we're not looking for an exact match, just the first | |
988 | * chunk that contains it (which implicitly ends at the start | |
989 | * of the next chunk. | |
990 | */ | |
991 | static off_t find_reused_offset(off_t where) | |
6b8fda2d | 992 | { |
bb514de3 JK |
993 | int lo = 0, hi = reused_chunks_nr; |
994 | while (lo < hi) { | |
995 | int mi = lo + ((hi - lo) / 2); | |
996 | if (where == reused_chunks[mi].original) | |
997 | return reused_chunks[mi].difference; | |
998 | if (where < reused_chunks[mi].original) | |
999 | hi = mi; | |
1000 | else | |
1001 | lo = mi + 1; | |
1002 | } | |
6b8fda2d | 1003 | |
bb514de3 JK |
1004 | /* |
1005 | * The first chunk starts at zero, so we can't have gone below | |
1006 | * there. | |
1007 | */ | |
1008 | assert(lo); | |
1009 | return reused_chunks[lo-1].difference; | |
1010 | } | |
6b8fda2d | 1011 | |
bb514de3 JK |
1012 | static void write_reused_pack_one(size_t pos, struct hashfile *out, |
1013 | struct pack_window **w_curs) | |
6b8fda2d | 1014 | { |
bb514de3 JK |
1015 | off_t offset, next, cur; |
1016 | enum object_type type; | |
1017 | unsigned long size; | |
6b8fda2d | 1018 | |
66cbd3e2 TB |
1019 | offset = pack_pos_to_offset(reuse_packfile, pos); |
1020 | next = pack_pos_to_offset(reuse_packfile, pos + 1); | |
6b8fda2d | 1021 | |
bb514de3 | 1022 | record_reused_object(offset, offset - hashfile_total(out)); |
6b8fda2d | 1023 | |
bb514de3 JK |
1024 | cur = offset; |
1025 | type = unpack_object_header(reuse_packfile, w_curs, &cur, &size); | |
1026 | assert(type >= 0); | |
6b8fda2d | 1027 | |
bb514de3 JK |
1028 | if (type == OBJ_OFS_DELTA) { |
1029 | off_t base_offset; | |
1030 | off_t fixup; | |
1031 | ||
1032 | unsigned char header[MAX_PACK_OBJECT_HEADER]; | |
1033 | unsigned len; | |
1034 | ||
1035 | base_offset = get_delta_base(reuse_packfile, w_curs, &cur, type, offset); | |
1036 | assert(base_offset != 0); | |
1037 | ||
1038 | /* Convert to REF_DELTA if we must... */ | |
1039 | if (!allow_ofs_delta) { | |
66cbd3e2 | 1040 | uint32_t base_pos; |
f66d4e02 JK |
1041 | struct object_id base_oid; |
1042 | ||
66cbd3e2 TB |
1043 | if (offset_to_pack_pos(reuse_packfile, base_offset, &base_pos) < 0) |
1044 | die(_("expected object at offset %"PRIuMAX" " | |
1045 | "in pack %s"), | |
1046 | (uintmax_t)base_offset, | |
1047 | reuse_packfile->pack_name); | |
1048 | ||
f66d4e02 | 1049 | nth_packed_object_id(&base_oid, reuse_packfile, |
66cbd3e2 | 1050 | pack_pos_to_index(reuse_packfile, base_pos)); |
bb514de3 JK |
1051 | |
1052 | len = encode_in_pack_object_header(header, sizeof(header), | |
1053 | OBJ_REF_DELTA, size); | |
1054 | hashwrite(out, header, len); | |
f8cb64e3 | 1055 | hashwrite(out, base_oid.hash, the_hash_algo->rawsz); |
bb514de3 JK |
1056 | copy_pack_data(out, reuse_packfile, w_curs, cur, next - cur); |
1057 | return; | |
1058 | } | |
6b8fda2d | 1059 | |
bb514de3 JK |
1060 | /* Otherwise see if we need to rewrite the offset... */ |
1061 | fixup = find_reused_offset(offset) - | |
1062 | find_reused_offset(base_offset); | |
1063 | if (fixup) { | |
1064 | unsigned char ofs_header[10]; | |
1065 | unsigned i, ofs_len; | |
1066 | off_t ofs = offset - base_offset - fixup; | |
6b8fda2d | 1067 | |
bb514de3 JK |
1068 | len = encode_in_pack_object_header(header, sizeof(header), |
1069 | OBJ_OFS_DELTA, size); | |
6b8fda2d | 1070 | |
bb514de3 JK |
1071 | i = sizeof(ofs_header) - 1; |
1072 | ofs_header[i] = ofs & 127; | |
1073 | while (ofs >>= 7) | |
1074 | ofs_header[--i] = 128 | (--ofs & 127); | |
6b8fda2d | 1075 | |
bb514de3 | 1076 | ofs_len = sizeof(ofs_header) - i; |
6b8fda2d | 1077 | |
bb514de3 JK |
1078 | hashwrite(out, header, len); |
1079 | hashwrite(out, ofs_header + sizeof(ofs_header) - ofs_len, ofs_len); | |
1080 | copy_pack_data(out, reuse_packfile, w_curs, cur, next - cur); | |
1081 | return; | |
1082 | } | |
1083 | ||
1084 | /* ...otherwise we have no fixup, and can write it verbatim */ | |
1085 | } | |
1086 | ||
1087 | copy_pack_data(out, reuse_packfile, w_curs, offset, next - offset); | |
1088 | } | |
1089 | ||
1090 | static size_t write_reused_pack_verbatim(struct hashfile *out, | |
1091 | struct pack_window **w_curs) | |
1092 | { | |
1093 | size_t pos = 0; | |
1094 | ||
1095 | while (pos < reuse_packfile_bitmap->word_alloc && | |
1096 | reuse_packfile_bitmap->words[pos] == (eword_t)~0) | |
1097 | pos++; | |
1098 | ||
1099 | if (pos) { | |
1100 | off_t to_write; | |
1101 | ||
1102 | written = (pos * BITS_IN_EWORD); | |
6a5c10c4 | 1103 | to_write = pack_pos_to_offset(reuse_packfile, written) |
bb514de3 JK |
1104 | - sizeof(struct pack_header); |
1105 | ||
1106 | /* We're recording one chunk, not one object. */ | |
1107 | record_reused_object(sizeof(struct pack_header), 0); | |
1108 | hashflush(out); | |
1109 | copy_pack_data(out, reuse_packfile, w_curs, | |
1110 | sizeof(struct pack_header), to_write); | |
657673f1 | 1111 | |
657673f1 | 1112 | display_progress(progress_state, written); |
6b8fda2d | 1113 | } |
bb514de3 JK |
1114 | return pos; |
1115 | } | |
1116 | ||
1117 | static void write_reused_pack(struct hashfile *f) | |
1118 | { | |
1119 | size_t i = 0; | |
1120 | uint32_t offset; | |
1121 | struct pack_window *w_curs = NULL; | |
1122 | ||
1123 | if (allow_ofs_delta) | |
1124 | i = write_reused_pack_verbatim(f, &w_curs); | |
1125 | ||
1126 | for (; i < reuse_packfile_bitmap->word_alloc; ++i) { | |
1127 | eword_t word = reuse_packfile_bitmap->words[i]; | |
1128 | size_t pos = (i * BITS_IN_EWORD); | |
6b8fda2d | 1129 | |
bb514de3 JK |
1130 | for (offset = 0; offset < BITS_IN_EWORD; ++offset) { |
1131 | if ((word >> offset) == 0) | |
1132 | break; | |
1133 | ||
1134 | offset += ewah_bit_ctz64(word >> offset); | |
0f533c72 TB |
1135 | /* |
1136 | * Can use bit positions directly, even for MIDX | |
1137 | * bitmaps. See comment in try_partial_reuse() | |
1138 | * for why. | |
1139 | */ | |
bb514de3 JK |
1140 | write_reused_pack_one(pos + offset, f, &w_curs); |
1141 | display_progress(progress_state, ++written); | |
1142 | } | |
1143 | } | |
6b8fda2d | 1144 | |
bb514de3 | 1145 | unuse_pack(&w_curs); |
6b8fda2d VM |
1146 | } |
1147 | ||
dd4b732d JT |
1148 | static void write_excluded_by_configs(void) |
1149 | { | |
1150 | struct oidset_iter iter; | |
1151 | const struct object_id *oid; | |
1152 | ||
1153 | oidset_iter_init(&excluded_by_config, &iter); | |
1154 | while ((oid = oidset_iter_next(&iter))) { | |
1155 | struct configured_exclusion *ex = | |
1156 | oidmap_get(&configured_exclusions, oid); | |
1157 | ||
1158 | if (!ex) | |
1159 | BUG("configured exclusion wasn't configured"); | |
1160 | write_in_full(1, ex->pack_hash_hex, strlen(ex->pack_hash_hex)); | |
1161 | write_in_full(1, " ", 1); | |
1162 | write_in_full(1, ex->uri, strlen(ex->uri)); | |
1163 | write_in_full(1, "\n", 1); | |
1164 | } | |
1165 | } | |
1166 | ||
9cea46cd EW |
1167 | static const char no_split_warning[] = N_( |
1168 | "disabling bitmap writing, packs are split due to pack.packSizeLimit" | |
1169 | ); | |
1170 | ||
d01fb92f | 1171 | static void write_pack_file(void) |
c323ac7d | 1172 | { |
ebe27b13 | 1173 | uint32_t i = 0, j; |
98a3beab | 1174 | struct hashfile *f; |
6ed7f25e | 1175 | off_t offset; |
ebe27b13 | 1176 | uint32_t nr_remaining = nr_result; |
f746bae8 | 1177 | time_t last_mtime = 0; |
1b4bb16b | 1178 | struct object_entry **write_order; |
81a216a5 | 1179 | |
bcd7954e | 1180 | if (progress > pack_to_stdout) |
754dbc43 | 1181 | progress_state = start_progress(_("Writing objects"), nr_result); |
b32fa95f | 1182 | ALLOC_ARRAY(written_list, to_pack.nr_objects); |
1b4bb16b | 1183 | write_order = compute_write_order(); |
183bdb2c | 1184 | |
ebe27b13 | 1185 | do { |
71b7672b | 1186 | unsigned char hash[GIT_MAX_RAWSZ]; |
7ba502c4 | 1187 | char *pack_tmp_name = NULL; |
aa7e44bf | 1188 | |
cdf9db3c | 1189 | if (pack_to_stdout) |
98a3beab | 1190 | f = hashfd_throughput(1, "<stdout>", progress_state); |
cdf9db3c JH |
1191 | else |
1192 | f = create_tmp_packfile(&pack_tmp_name); | |
ebe27b13 | 1193 | |
c0ad4657 | 1194 | offset = write_pack_header(f, nr_remaining); |
6b8fda2d VM |
1195 | |
1196 | if (reuse_packfile) { | |
6b8fda2d | 1197 | assert(pack_to_stdout); |
bb514de3 JK |
1198 | write_reused_pack(f); |
1199 | offset = hashfile_total(f); | |
6b8fda2d VM |
1200 | } |
1201 | ||
ebe27b13 | 1202 | nr_written = 0; |
2834bc27 | 1203 | for (; i < to_pack.nr_objects; i++) { |
1b4bb16b | 1204 | struct object_entry *e = write_order[i]; |
cddec4f8 | 1205 | if (write_one(f, e, &offset) == WRITE_ONE_BREAK) |
720c9f7b NP |
1206 | break; |
1207 | display_progress(progress_state, written); | |
1208 | } | |
c553ca25 | 1209 | |
54352bb2 | 1210 | if (pack_to_stdout) { |
020406ea NS |
1211 | /* |
1212 | * We never fsync when writing to stdout since we may | |
1213 | * not be writing to an actual pack file. For instance, | |
1214 | * the upload-pack code passes a pipe here. Calling | |
1215 | * fsync on a pipe results in unnecessary | |
1216 | * synchronization with the reader on some platforms. | |
1217 | */ | |
1218 | finalize_hashfile(f, hash, FSYNC_COMPONENT_NONE, | |
1219 | CSUM_HASH_IN_STREAM | CSUM_CLOSE); | |
54352bb2 | 1220 | } else if (nr_written == nr_remaining) { |
020406ea NS |
1221 | finalize_hashfile(f, hash, FSYNC_COMPONENT_PACK, |
1222 | CSUM_HASH_IN_STREAM | CSUM_FSYNC | CSUM_CLOSE); | |
ebe27b13 | 1223 | } else { |
020406ea NS |
1224 | /* |
1225 | * If we wrote the wrong number of entries in the | |
1226 | * header, rewrite it like in fast-import. | |
1227 | */ | |
1228 | ||
1229 | int fd = finalize_hashfile(f, hash, FSYNC_COMPONENT_PACK, 0); | |
71b7672b | 1230 | fixup_pack_header_footer(fd, hash, pack_tmp_name, |
1231 | nr_written, hash, offset); | |
7ba502c4 | 1232 | close(fd); |
9cea46cd | 1233 | if (write_bitmap_index) { |
25575015 JK |
1234 | if (write_bitmap_index != WRITE_BITMAP_QUIET) |
1235 | warning(_(no_split_warning)); | |
9cea46cd EW |
1236 | write_bitmap_index = 0; |
1237 | } | |
ebe27b13 DH |
1238 | } |
1239 | ||
1240 | if (!pack_to_stdout) { | |
f746bae8 | 1241 | struct stat st; |
58892711 | 1242 | struct strbuf tmpname = STRBUF_INIT; |
2ec02dd5 | 1243 | char *idx_tmp_name = NULL; |
d01fb92f | 1244 | |
f746bae8 NP |
1245 | /* |
1246 | * Packs are runtime accessed in their mtime | |
1247 | * order since newer packs are more likely to contain | |
1248 | * younger objects. So if we are creating multiple | |
1249 | * packs then we should modify the mtime of later ones | |
1250 | * to preserve this property. | |
1251 | */ | |
0e990530 | 1252 | if (stat(pack_tmp_name, &st) < 0) { |
f616db6a | 1253 | warning_errno(_("failed to stat %s"), pack_tmp_name); |
f746bae8 NP |
1254 | } else if (!last_mtime) { |
1255 | last_mtime = st.st_mtime; | |
1256 | } else { | |
1257 | struct utimbuf utb; | |
1258 | utb.actime = st.st_atime; | |
1259 | utb.modtime = --last_mtime; | |
0e990530 | 1260 | if (utime(pack_tmp_name, &utb) < 0) |
f616db6a | 1261 | warning_errno(_("failed utime() on %s"), pack_tmp_name); |
f746bae8 NP |
1262 | } |
1263 | ||
66833f0e ÆAB |
1264 | strbuf_addf(&tmpname, "%s-%s.", base_name, |
1265 | hash_to_hex(hash)); | |
7cc8f971 VM |
1266 | |
1267 | if (write_bitmap_index) { | |
71b7672b | 1268 | bitmap_writer_set_checksum(hash); |
06af3bba NTND |
1269 | bitmap_writer_build_type_index( |
1270 | &to_pack, written_list, nr_written); | |
7cc8f971 VM |
1271 | } |
1272 | ||
b7573536 TB |
1273 | if (cruft) |
1274 | pack_idx_opts.flags |= WRITE_MTIMES; | |
1275 | ||
2ec02dd5 | 1276 | stage_tmp_packfiles(&tmpname, pack_tmp_name, |
0e990530 | 1277 | written_list, nr_written, |
1c573cdd TB |
1278 | &to_pack, &pack_idx_opts, hash, |
1279 | &idx_tmp_name); | |
7cc8f971 VM |
1280 | |
1281 | if (write_bitmap_index) { | |
66833f0e | 1282 | size_t tmpname_len = tmpname.len; |
7cc8f971 | 1283 | |
66833f0e | 1284 | strbuf_addstr(&tmpname, "bitmap"); |
7cc8f971 VM |
1285 | stop_progress(&progress_state); |
1286 | ||
1287 | bitmap_writer_show_progress(progress); | |
7cc8f971 | 1288 | bitmap_writer_select_commits(indexed_commits, indexed_commits_nr, -1); |
3ba3d062 TB |
1289 | if (bitmap_writer_build(&to_pack) < 0) |
1290 | die(_("failed to write bitmap index")); | |
ae4f07fb | 1291 | bitmap_writer_finish(written_list, nr_written, |
58892711 | 1292 | tmpname.buf, write_bitmap_options); |
7cc8f971 | 1293 | write_bitmap_index = 0; |
66833f0e | 1294 | strbuf_setlen(&tmpname, tmpname_len); |
7cc8f971 VM |
1295 | } |
1296 | ||
4bc1fd6e ÆAB |
1297 | rename_tmp_packfile_idx(&tmpname, &idx_tmp_name); |
1298 | ||
2ec02dd5 | 1299 | free(idx_tmp_name); |
58892711 | 1300 | strbuf_release(&tmpname); |
7ba502c4 | 1301 | free(pack_tmp_name); |
71b7672b | 1302 | puts(hash_to_hex(hash)); |
ebe27b13 DH |
1303 | } |
1304 | ||
1305 | /* mark written objects as written to previous pack */ | |
1306 | for (j = 0; j < nr_written; j++) { | |
79814f42 | 1307 | written_list[j]->offset = (off_t)-1; |
ebe27b13 DH |
1308 | } |
1309 | nr_remaining -= nr_written; | |
2834bc27 | 1310 | } while (nr_remaining && i < to_pack.nr_objects); |
ebe27b13 DH |
1311 | |
1312 | free(written_list); | |
1b4bb16b | 1313 | free(write_order); |
4d4fcc54 | 1314 | stop_progress(&progress_state); |
67c08ce1 | 1315 | if (written != nr_result) |
f616db6a NTND |
1316 | die(_("wrote %"PRIu32" objects while expecting %"PRIu32), |
1317 | written, nr_result); | |
9ed87902 JT |
1318 | trace2_data_intmax("pack-objects", the_repository, |
1319 | "write_pack_file/wrote", nr_result); | |
c323ac7d LT |
1320 | } |
1321 | ||
a74db82e JH |
1322 | static int no_try_delta(const char *path) |
1323 | { | |
2aef63d3 | 1324 | static struct attr_check *check; |
a74db82e | 1325 | |
2aef63d3 JH |
1326 | if (!check) |
1327 | check = attr_check_initl("delta", NULL); | |
47cfc9bd | 1328 | git_check_attr(the_repository->index, NULL, path, check); |
2aef63d3 | 1329 | if (ATTR_FALSE(check->items[0].value)) |
a74db82e JH |
1330 | return 1; |
1331 | return 0; | |
1332 | } | |
1333 | ||
ce2bc424 JK |
1334 | /* |
1335 | * When adding an object, check whether we have already added it | |
1336 | * to our packing list. If so, we can skip. However, if we are | |
1337 | * being asked to excludei t, but the previous mention was to include | |
1338 | * it, make sure to adjust its flags and tweak our numbers accordingly. | |
1339 | * | |
1340 | * As an optimization, we pass out the index position where we would have | |
1341 | * found the item, since that saves us from having to look it up again a | |
1342 | * few lines later when we want to add the new entry. | |
1343 | */ | |
188960b4 | 1344 | static int have_duplicate_entry(const struct object_id *oid, |
3a37876b | 1345 | int exclude) |
c323ac7d | 1346 | { |
c323ac7d | 1347 | struct object_entry *entry; |
29b734e4 | 1348 | |
92fb0db9 JK |
1349 | if (reuse_packfile_bitmap && |
1350 | bitmap_walk_contains(bitmap_git, reuse_packfile_bitmap, oid)) | |
1351 | return 1; | |
1352 | ||
3a37876b | 1353 | entry = packlist_find(&to_pack, oid); |
ce2bc424 | 1354 | if (!entry) |
29b734e4 | 1355 | return 0; |
ce2bc424 JK |
1356 | |
1357 | if (exclude) { | |
1358 | if (!entry->preferred_base) | |
1359 | nr_result--; | |
1360 | entry->preferred_base = 1; | |
29b734e4 | 1361 | } |
c323ac7d | 1362 | |
ce2bc424 JK |
1363 | return 1; |
1364 | } | |
1365 | ||
6325da14 JK |
1366 | static int want_found_object(const struct object_id *oid, int exclude, |
1367 | struct packed_git *p) | |
702d1b95 KS |
1368 | { |
1369 | if (exclude) | |
1370 | return 1; | |
1371 | if (incremental) | |
1372 | return 0; | |
1373 | ||
4090511e TB |
1374 | if (!is_pack_valid(p)) |
1375 | return -1; | |
1376 | ||
702d1b95 KS |
1377 | /* |
1378 | * When asked to do --local (do not include an object that appears in a | |
1379 | * pack we borrow from elsewhere) or --honor-pack-keep (do not include | |
1380 | * an object that appears in a pack marked with .keep), finding a pack | |
1381 | * that matches the criteria is sufficient for us to decide to omit it. | |
1382 | * However, even if this pack does not satisfy the criteria, we need to | |
1383 | * make sure no copy of this object appears in _any_ pack that makes us | |
1384 | * to omit the object, so we need to check all the packs. | |
1385 | * | |
6325da14 | 1386 | * We can however first check whether these options can possibly matter; |
702d1b95 KS |
1387 | * if they do not matter we know we want the object in generated pack. |
1388 | * Otherwise, we signal "-1" at the end to tell the caller that we do | |
1389 | * not know either way, and it needs to check more packs. | |
1390 | */ | |
702d1b95 | 1391 | |
6325da14 JK |
1392 | /* |
1393 | * Objects in packs borrowed from elsewhere are discarded regardless of | |
1394 | * if they appear in other packs that weren't borrowed. | |
1395 | */ | |
702d1b95 KS |
1396 | if (local && !p->pack_local) |
1397 | return 0; | |
6325da14 JK |
1398 | |
1399 | /* | |
1400 | * Then handle .keep first, as we have a fast(er) path there. | |
1401 | */ | |
1402 | if (ignore_packed_keep_on_disk || ignore_packed_keep_in_core) { | |
1403 | /* | |
1404 | * Set the flags for the kept-pack cache to be the ones we want | |
1405 | * to ignore. | |
1406 | * | |
1407 | * That is, if we are ignoring objects in on-disk keep packs, | |
1408 | * then we want to search through the on-disk keep and ignore | |
1409 | * the in-core ones. | |
1410 | */ | |
1411 | unsigned flags = 0; | |
1412 | if (ignore_packed_keep_on_disk) | |
1413 | flags |= ON_DISK_KEEP_PACKS; | |
1414 | if (ignore_packed_keep_in_core) | |
1415 | flags |= IN_CORE_KEEP_PACKS; | |
1416 | ||
1417 | if (ignore_packed_keep_on_disk && p->pack_keep) | |
1418 | return 0; | |
1419 | if (ignore_packed_keep_in_core && p->pack_keep_in_core) | |
1420 | return 0; | |
1421 | if (has_object_kept_pack(oid, flags)) | |
1422 | return 0; | |
1423 | } | |
1424 | ||
1425 | /* | |
1426 | * At this point we know definitively that either we don't care about | |
1427 | * keep-packs, or the object is not in one. Keep checking other | |
1428 | * conditions... | |
1429 | */ | |
1430 | if (!local || !have_non_local_packs) | |
1431 | return 1; | |
702d1b95 KS |
1432 | |
1433 | /* we don't know yet; keep looking for more packs */ | |
1434 | return -1; | |
1435 | } | |
1436 | ||
6325da14 JK |
1437 | static int want_object_in_pack_one(struct packed_git *p, |
1438 | const struct object_id *oid, | |
1439 | int exclude, | |
1440 | struct packed_git **found_pack, | |
1441 | off_t *found_offset) | |
1442 | { | |
1443 | off_t offset; | |
1444 | ||
1445 | if (p == *found_pack) | |
1446 | offset = *found_offset; | |
1447 | else | |
1448 | offset = find_pack_entry_one(oid->hash, p); | |
1449 | ||
1450 | if (offset) { | |
1451 | if (!*found_pack) { | |
1452 | if (!is_pack_valid(p)) | |
1453 | return -1; | |
1454 | *found_offset = offset; | |
1455 | *found_pack = p; | |
1456 | } | |
1457 | return want_found_object(oid, exclude, p); | |
1458 | } | |
1459 | return -1; | |
1460 | } | |
1461 | ||
ce2bc424 JK |
1462 | /* |
1463 | * Check whether we want the object in the pack (e.g., we do not want | |
1464 | * objects found in non-local stores if the "--local" option was used). | |
1465 | * | |
702d1b95 KS |
1466 | * If the caller already knows an existing pack it wants to take the object |
1467 | * from, that is passed in *found_pack and *found_offset; otherwise this | |
1468 | * function finds if there is any pack that has the object and returns the pack | |
1469 | * and its offset in these variables. | |
ce2bc424 | 1470 | */ |
188960b4 | 1471 | static int want_object_in_pack(const struct object_id *oid, |
ce2bc424 JK |
1472 | int exclude, |
1473 | struct packed_git **found_pack, | |
1474 | off_t *found_offset) | |
1475 | { | |
702d1b95 | 1476 | int want; |
8865859d | 1477 | struct list_head *pos; |
6a22d521 | 1478 | struct multi_pack_index *m; |
ce2bc424 | 1479 | |
6862ebbf | 1480 | if (!exclude && local && has_loose_object_nonlocal(oid)) |
daae0625 BC |
1481 | return 0; |
1482 | ||
702d1b95 KS |
1483 | /* |
1484 | * If we already know the pack object lives in, start checks from that | |
1485 | * pack - in the usual case when neither --local was given nor .keep files | |
1486 | * are present we will determine the answer right now. | |
1487 | */ | |
1488 | if (*found_pack) { | |
6325da14 | 1489 | want = want_found_object(oid, exclude, *found_pack); |
702d1b95 KS |
1490 | if (want != -1) |
1491 | return want; | |
4090511e TB |
1492 | |
1493 | *found_pack = NULL; | |
1494 | *found_offset = 0; | |
702d1b95 | 1495 | } |
6a22d521 DS |
1496 | |
1497 | for (m = get_multi_pack_index(the_repository); m; m = m->next) { | |
1498 | struct pack_entry e; | |
64404a24 | 1499 | if (fill_midx_entry(the_repository, oid, &e, m)) { |
6325da14 JK |
1500 | want = want_object_in_pack_one(e.p, oid, exclude, found_pack, found_offset); |
1501 | if (want != -1) | |
1502 | return want; | |
6a22d521 DS |
1503 | } |
1504 | } | |
1505 | ||
a80d72db | 1506 | list_for_each(pos, get_packed_git_mru(the_repository)) { |
ec2dd32c | 1507 | struct packed_git *p = list_entry(pos, struct packed_git, mru); |
6325da14 JK |
1508 | want = want_object_in_pack_one(p, oid, exclude, found_pack, found_offset); |
1509 | if (!exclude && want > 0) | |
1510 | list_move(&p->mru, | |
1511 | get_packed_git_mru(the_repository)); | |
1512 | if (want != -1) | |
1513 | return want; | |
64560374 | 1514 | } |
eb019375 | 1515 | |
dd4b732d JT |
1516 | if (uri_protocols.nr) { |
1517 | struct configured_exclusion *ex = | |
1518 | oidmap_get(&configured_exclusions, oid); | |
1519 | int i; | |
1520 | const char *p; | |
1521 | ||
1522 | if (ex) { | |
1523 | for (i = 0; i < uri_protocols.nr; i++) { | |
1524 | if (skip_prefix(ex->uri, | |
1525 | uri_protocols.items[i].string, | |
1526 | &p) && | |
1527 | *p == ':') { | |
1528 | oidset_insert(&excluded_by_config, oid); | |
1529 | return 0; | |
1530 | } | |
1531 | } | |
1532 | } | |
1533 | } | |
1534 | ||
ce2bc424 JK |
1535 | return 1; |
1536 | } | |
1537 | ||
fa23090b TB |
1538 | static struct object_entry *create_object_entry(const struct object_id *oid, |
1539 | enum object_type type, | |
1540 | uint32_t hash, | |
1541 | int exclude, | |
1542 | int no_try_delta, | |
1543 | struct packed_git *found_pack, | |
1544 | off_t found_offset) | |
ce2bc424 JK |
1545 | { |
1546 | struct object_entry *entry; | |
29b734e4 | 1547 | |
3a37876b | 1548 | entry = packlist_alloc(&to_pack, oid); |
27225f2e | 1549 | entry->hash = hash; |
fd9b1bae | 1550 | oe_set_type(entry, type); |
29b734e4 NP |
1551 | if (exclude) |
1552 | entry->preferred_base = 1; | |
81a216a5 NP |
1553 | else |
1554 | nr_result++; | |
29b734e4 | 1555 | if (found_pack) { |
43fa44fa | 1556 | oe_set_in_pack(&to_pack, entry, found_pack); |
29b734e4 NP |
1557 | entry->in_pack_offset = found_offset; |
1558 | } | |
7a979d99 | 1559 | |
ce2bc424 | 1560 | entry->no_try_delta = no_try_delta; |
fa23090b TB |
1561 | |
1562 | return entry; | |
ce2bc424 | 1563 | } |
29b734e4 | 1564 | |
373c67da JK |
1565 | static const char no_closure_warning[] = N_( |
1566 | "disabling bitmap writing, as some objects are not being packed" | |
1567 | ); | |
1568 | ||
188960b4 | 1569 | static int add_object_entry(const struct object_id *oid, enum object_type type, |
ce2bc424 JK |
1570 | const char *name, int exclude) |
1571 | { | |
702d1b95 KS |
1572 | struct packed_git *found_pack = NULL; |
1573 | off_t found_offset = 0; | |
7a979d99 | 1574 | |
5af05043 NTND |
1575 | display_progress(progress_state, ++nr_seen); |
1576 | ||
3a37876b | 1577 | if (have_duplicate_entry(oid, exclude)) |
ce2bc424 | 1578 | return 0; |
a74db82e | 1579 | |
188960b4 | 1580 | if (!want_object_in_pack(oid, exclude, &found_pack, &found_offset)) { |
373c67da JK |
1581 | /* The pack is missing an object, so it will not have closure */ |
1582 | if (write_bitmap_index) { | |
25575015 JK |
1583 | if (write_bitmap_index != WRITE_BITMAP_QUIET) |
1584 | warning(_(no_closure_warning)); | |
373c67da JK |
1585 | write_bitmap_index = 0; |
1586 | } | |
ce2bc424 | 1587 | return 0; |
373c67da | 1588 | } |
29b734e4 | 1589 | |
188960b4 | 1590 | create_object_entry(oid, type, pack_name_hash(name), |
ce2bc424 | 1591 | exclude, name && no_try_delta(name), |
3a37876b | 1592 | found_pack, found_offset); |
29b734e4 | 1593 | return 1; |
c323ac7d LT |
1594 | } |
1595 | ||
20664967 | 1596 | static int add_object_entry_from_bitmap(const struct object_id *oid, |
6b8fda2d | 1597 | enum object_type type, |
c50dca2a | 1598 | int flags UNUSED, uint32_t name_hash, |
6b8fda2d VM |
1599 | struct packed_git *pack, off_t offset) |
1600 | { | |
5af05043 NTND |
1601 | display_progress(progress_state, ++nr_seen); |
1602 | ||
3a37876b | 1603 | if (have_duplicate_entry(oid, 0)) |
6b8fda2d VM |
1604 | return 0; |
1605 | ||
188960b4 | 1606 | if (!want_object_in_pack(oid, 0, &pack, &offset)) |
702d1b95 KS |
1607 | return 0; |
1608 | ||
3a37876b | 1609 | create_object_entry(oid, type, name_hash, 0, 0, pack, offset); |
29b734e4 | 1610 | return 1; |
c323ac7d LT |
1611 | } |
1612 | ||
5379a5c5 | 1613 | struct pbase_tree_cache { |
188960b4 | 1614 | struct object_id oid; |
5379a5c5 JH |
1615 | int ref; |
1616 | int temporary; | |
1617 | void *tree_data; | |
1618 | unsigned long tree_size; | |
1619 | }; | |
1620 | ||
1621 | static struct pbase_tree_cache *(pbase_tree_cache[256]); | |
188960b4 | 1622 | static int pbase_tree_cache_ix(const struct object_id *oid) |
5379a5c5 | 1623 | { |
188960b4 | 1624 | return oid->hash[0] % ARRAY_SIZE(pbase_tree_cache); |
5379a5c5 JH |
1625 | } |
1626 | static int pbase_tree_cache_ix_incr(int ix) | |
1627 | { | |
1628 | return (ix+1) % ARRAY_SIZE(pbase_tree_cache); | |
1629 | } | |
1630 | ||
1631 | static struct pbase_tree { | |
1632 | struct pbase_tree *next; | |
1633 | /* This is a phony "cache" entry; we are not | |
01689909 | 1634 | * going to evict it or find it through _get() |
5379a5c5 JH |
1635 | * mechanism -- this is for the toplevel node that |
1636 | * would almost always change with any commit. | |
1637 | */ | |
1638 | struct pbase_tree_cache pcache; | |
1639 | } *pbase_tree; | |
1640 | ||
188960b4 | 1641 | static struct pbase_tree_cache *pbase_tree_get(const struct object_id *oid) |
5379a5c5 JH |
1642 | { |
1643 | struct pbase_tree_cache *ent, *nent; | |
1644 | void *data; | |
1645 | unsigned long size; | |
21666f1a | 1646 | enum object_type type; |
5379a5c5 | 1647 | int neigh; |
188960b4 | 1648 | int my_ix = pbase_tree_cache_ix(oid); |
5379a5c5 JH |
1649 | int available_ix = -1; |
1650 | ||
1651 | /* pbase-tree-cache acts as a limited hashtable. | |
1652 | * your object will be found at your index or within a few | |
1653 | * slots after that slot if it is cached. | |
1654 | */ | |
1655 | for (neigh = 0; neigh < 8; neigh++) { | |
1656 | ent = pbase_tree_cache[my_ix]; | |
4a7e27e9 | 1657 | if (ent && oideq(&ent->oid, oid)) { |
5379a5c5 JH |
1658 | ent->ref++; |
1659 | return ent; | |
1660 | } | |
1661 | else if (((available_ix < 0) && (!ent || !ent->ref)) || | |
1662 | ((0 <= available_ix) && | |
1663 | (!ent && pbase_tree_cache[available_ix]))) | |
1664 | available_ix = my_ix; | |
1665 | if (!ent) | |
1666 | break; | |
1667 | my_ix = pbase_tree_cache_ix_incr(my_ix); | |
1668 | } | |
1669 | ||
1670 | /* Did not find one. Either we got a bogus request or | |
1671 | * we need to read and perhaps cache. | |
1672 | */ | |
b4f5aca4 | 1673 | data = read_object_file(oid, &type, &size); |
5379a5c5 JH |
1674 | if (!data) |
1675 | return NULL; | |
21666f1a | 1676 | if (type != OBJ_TREE) { |
5379a5c5 JH |
1677 | free(data); |
1678 | return NULL; | |
1679 | } | |
1680 | ||
1681 | /* We need to either cache or return a throwaway copy */ | |
1682 | ||
1683 | if (available_ix < 0) | |
1684 | ent = NULL; | |
1685 | else { | |
1686 | ent = pbase_tree_cache[available_ix]; | |
1687 | my_ix = available_ix; | |
1688 | } | |
1689 | ||
1690 | if (!ent) { | |
1691 | nent = xmalloc(sizeof(*nent)); | |
1692 | nent->temporary = (available_ix < 0); | |
1693 | } | |
1694 | else { | |
1695 | /* evict and reuse */ | |
1696 | free(ent->tree_data); | |
1697 | nent = ent; | |
1698 | } | |
188960b4 | 1699 | oidcpy(&nent->oid, oid); |
5379a5c5 JH |
1700 | nent->tree_data = data; |
1701 | nent->tree_size = size; | |
1702 | nent->ref = 1; | |
1703 | if (!nent->temporary) | |
1704 | pbase_tree_cache[my_ix] = nent; | |
1705 | return nent; | |
1706 | } | |
1707 | ||
1708 | static void pbase_tree_put(struct pbase_tree_cache *cache) | |
1709 | { | |
1710 | if (!cache->temporary) { | |
1711 | cache->ref--; | |
1712 | return; | |
1713 | } | |
1714 | free(cache->tree_data); | |
1715 | free(cache); | |
1716 | } | |
1717 | ||
e65b868d | 1718 | static size_t name_cmp_len(const char *name) |
5379a5c5 | 1719 | { |
e65b868d | 1720 | return strcspn(name, "\n/"); |
5379a5c5 JH |
1721 | } |
1722 | ||
1723 | static void add_pbase_object(struct tree_desc *tree, | |
5379a5c5 | 1724 | const char *name, |
e65b868d | 1725 | size_t cmplen, |
ce0bd642 | 1726 | const char *fullname) |
3f9ac8d2 | 1727 | { |
4c068a98 | 1728 | struct name_entry entry; |
8a5a8d6c | 1729 | int cmp; |
4c068a98 LT |
1730 | |
1731 | while (tree_entry(tree,&entry)) { | |
1211be6b LT |
1732 | if (S_ISGITLINK(entry.mode)) |
1733 | continue; | |
0de16337 | 1734 | cmp = tree_entry_len(&entry) != cmplen ? 1 : |
8a5a8d6c NP |
1735 | memcmp(name, entry.path, cmplen); |
1736 | if (cmp > 0) | |
7a979d99 | 1737 | continue; |
8a5a8d6c NP |
1738 | if (cmp < 0) |
1739 | return; | |
5379a5c5 | 1740 | if (name[cmplen] != '/') { |
ea82b2a0 | 1741 | add_object_entry(&entry.oid, |
4d1012c3 | 1742 | object_type(entry.mode), |
bc32fed5 | 1743 | fullname, 1); |
5379a5c5 JH |
1744 | return; |
1745 | } | |
8a5a8d6c | 1746 | if (S_ISDIR(entry.mode)) { |
7a979d99 | 1747 | struct tree_desc sub; |
5379a5c5 JH |
1748 | struct pbase_tree_cache *tree; |
1749 | const char *down = name+cmplen+1; | |
e65b868d | 1750 | size_t downlen = name_cmp_len(down); |
5379a5c5 | 1751 | |
ea82b2a0 | 1752 | tree = pbase_tree_get(&entry.oid); |
5379a5c5 JH |
1753 | if (!tree) |
1754 | return; | |
6fda5e51 | 1755 | init_tree_desc(&sub, tree->tree_data, tree->tree_size); |
5379a5c5 | 1756 | |
ce0bd642 | 1757 | add_pbase_object(&sub, down, downlen, fullname); |
5379a5c5 JH |
1758 | pbase_tree_put(tree); |
1759 | } | |
1760 | } | |
1761 | } | |
1d6b38cc | 1762 | |
5379a5c5 JH |
1763 | static unsigned *done_pbase_paths; |
1764 | static int done_pbase_paths_num; | |
1765 | static int done_pbase_paths_alloc; | |
1766 | static int done_pbase_path_pos(unsigned hash) | |
1767 | { | |
1768 | int lo = 0; | |
1769 | int hi = done_pbase_paths_num; | |
1770 | while (lo < hi) { | |
19716b21 | 1771 | int mi = lo + (hi - lo) / 2; |
5379a5c5 JH |
1772 | if (done_pbase_paths[mi] == hash) |
1773 | return mi; | |
1774 | if (done_pbase_paths[mi] < hash) | |
1775 | hi = mi; | |
1776 | else | |
1777 | lo = mi + 1; | |
1778 | } | |
1779 | return -lo-1; | |
1780 | } | |
1781 | ||
1782 | static int check_pbase_path(unsigned hash) | |
1783 | { | |
c7b07805 | 1784 | int pos = done_pbase_path_pos(hash); |
5379a5c5 JH |
1785 | if (0 <= pos) |
1786 | return 1; | |
1787 | pos = -pos - 1; | |
25e19407 DD |
1788 | ALLOC_GROW(done_pbase_paths, |
1789 | done_pbase_paths_num + 1, | |
1790 | done_pbase_paths_alloc); | |
5379a5c5 JH |
1791 | done_pbase_paths_num++; |
1792 | if (pos < done_pbase_paths_num) | |
f331ab9d RS |
1793 | MOVE_ARRAY(done_pbase_paths + pos + 1, done_pbase_paths + pos, |
1794 | done_pbase_paths_num - pos - 1); | |
5379a5c5 JH |
1795 | done_pbase_paths[pos] = hash; |
1796 | return 0; | |
1797 | } | |
1798 | ||
bc32fed5 | 1799 | static void add_preferred_base_object(const char *name) |
5379a5c5 JH |
1800 | { |
1801 | struct pbase_tree *it; | |
e65b868d | 1802 | size_t cmplen; |
68fb36eb | 1803 | unsigned hash = pack_name_hash(name); |
5379a5c5 | 1804 | |
8a5a8d6c | 1805 | if (!num_preferred_base || check_pbase_path(hash)) |
5379a5c5 JH |
1806 | return; |
1807 | ||
8a5a8d6c | 1808 | cmplen = name_cmp_len(name); |
5379a5c5 JH |
1809 | for (it = pbase_tree; it; it = it->next) { |
1810 | if (cmplen == 0) { | |
188960b4 | 1811 | add_object_entry(&it->pcache.oid, OBJ_TREE, NULL, 1); |
5379a5c5 JH |
1812 | } |
1813 | else { | |
1814 | struct tree_desc tree; | |
6fda5e51 | 1815 | init_tree_desc(&tree, it->pcache.tree_data, it->pcache.tree_size); |
ce0bd642 | 1816 | add_pbase_object(&tree, name, cmplen, name); |
7a979d99 | 1817 | } |
3f9ac8d2 | 1818 | } |
3f9ac8d2 JH |
1819 | } |
1820 | ||
188960b4 | 1821 | static void add_preferred_base(struct object_id *oid) |
3f9ac8d2 | 1822 | { |
5379a5c5 JH |
1823 | struct pbase_tree *it; |
1824 | void *data; | |
1825 | unsigned long size; | |
188960b4 | 1826 | struct object_id tree_oid; |
1d6b38cc | 1827 | |
8d1d8f83 JH |
1828 | if (window <= num_preferred_base++) |
1829 | return; | |
1830 | ||
d3b4705a | 1831 | data = read_object_with_reference(the_repository, oid, |
6aea6bae | 1832 | OBJ_TREE, &size, &tree_oid); |
5379a5c5 | 1833 | if (!data) |
7a979d99 | 1834 | return; |
5379a5c5 JH |
1835 | |
1836 | for (it = pbase_tree; it; it = it->next) { | |
4a7e27e9 | 1837 | if (oideq(&it->pcache.oid, &tree_oid)) { |
5379a5c5 JH |
1838 | free(data); |
1839 | return; | |
1840 | } | |
1841 | } | |
1842 | ||
ca56dadb | 1843 | CALLOC_ARRAY(it, 1); |
5379a5c5 JH |
1844 | it->next = pbase_tree; |
1845 | pbase_tree = it; | |
1846 | ||
188960b4 | 1847 | oidcpy(&it->pcache.oid, &tree_oid); |
5379a5c5 JH |
1848 | it->pcache.tree_data = data; |
1849 | it->pcache.tree_size = size; | |
3f9ac8d2 JH |
1850 | } |
1851 | ||
0ef95f72 NP |
1852 | static void cleanup_preferred_base(void) |
1853 | { | |
1854 | struct pbase_tree *it; | |
1855 | unsigned i; | |
1856 | ||
1857 | it = pbase_tree; | |
1858 | pbase_tree = NULL; | |
1859 | while (it) { | |
095b3b2c BW |
1860 | struct pbase_tree *tmp = it; |
1861 | it = tmp->next; | |
1862 | free(tmp->pcache.tree_data); | |
1863 | free(tmp); | |
0ef95f72 NP |
1864 | } |
1865 | ||
1866 | for (i = 0; i < ARRAY_SIZE(pbase_tree_cache); i++) { | |
1867 | if (!pbase_tree_cache[i]) | |
1868 | continue; | |
1869 | free(pbase_tree_cache[i]->tree_data); | |
6a83d902 | 1870 | FREE_AND_NULL(pbase_tree_cache[i]); |
0ef95f72 NP |
1871 | } |
1872 | ||
6a83d902 | 1873 | FREE_AND_NULL(done_pbase_paths); |
0ef95f72 NP |
1874 | done_pbase_paths_num = done_pbase_paths_alloc = 0; |
1875 | } | |
1876 | ||
2fa233a5 JK |
1877 | /* |
1878 | * Return 1 iff the object specified by "delta" can be sent | |
1879 | * literally as a delta against the base in "base_sha1". If | |
1880 | * so, then *base_out will point to the entry in our packing | |
1881 | * list, or NULL if we must use the external-base list. | |
1882 | * | |
1883 | * Depth value does not matter - find_deltas() will | |
1884 | * never consider reused delta as the base object to | |
1885 | * deltify other objects against, in order to avoid | |
1886 | * circular deltas. | |
1887 | */ | |
3f83fd5e | 1888 | static int can_reuse_delta(const struct object_id *base_oid, |
2fa233a5 JK |
1889 | struct object_entry *delta, |
1890 | struct object_entry **base_out) | |
1891 | { | |
1892 | struct object_entry *base; | |
3df28cae | 1893 | |
2fa233a5 JK |
1894 | /* |
1895 | * First see if we're already sending the base (or it's explicitly in | |
1896 | * our "excluded" list). | |
1897 | */ | |
3f83fd5e | 1898 | base = packlist_find(&to_pack, base_oid); |
2fa233a5 JK |
1899 | if (base) { |
1900 | if (!in_same_island(&delta->idx.oid, &base->idx.oid)) | |
1901 | return 0; | |
1902 | *base_out = base; | |
1903 | return 1; | |
1904 | } | |
1905 | ||
1906 | /* | |
1907 | * Otherwise, reachability bitmaps may tell us if the receiver has it, | |
1908 | * even if it was buried too deep in history to make it into the | |
1909 | * packing list. | |
1910 | */ | |
3f83fd5e | 1911 | if (thin && bitmap_has_oid_in_uninteresting(bitmap_git, base_oid)) { |
2fa233a5 | 1912 | if (use_delta_islands) { |
3f83fd5e | 1913 | if (!in_same_island(&delta->idx.oid, base_oid)) |
2fa233a5 JK |
1914 | return 0; |
1915 | } | |
1916 | *base_out = NULL; | |
1917 | return 1; | |
1918 | } | |
1919 | ||
1920 | return 0; | |
1921 | } | |
1922 | ||
e00549aa JT |
1923 | static void prefetch_to_pack(uint32_t object_index_start) { |
1924 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
1925 | uint32_t i; | |
1926 | ||
1927 | for (i = object_index_start; i < to_pack.nr_objects; i++) { | |
1928 | struct object_entry *entry = to_pack.objects + i; | |
1929 | ||
1930 | if (!oid_object_info_extended(the_repository, | |
1931 | &entry->idx.oid, | |
1932 | NULL, | |
1933 | OBJECT_INFO_FOR_PREFETCH)) | |
1934 | continue; | |
1935 | oid_array_append(&to_fetch, &entry->idx.oid); | |
1936 | } | |
1937 | promisor_remote_get_direct(the_repository, | |
1938 | to_fetch.oid, to_fetch.nr); | |
1939 | oid_array_clear(&to_fetch); | |
1940 | } | |
1941 | ||
1942 | static void check_object(struct object_entry *entry, uint32_t object_index) | |
c323ac7d | 1943 | { |
ac77d0c3 | 1944 | unsigned long canonical_size; |
8d5cf957 JT |
1945 | enum object_type type; |
1946 | struct object_info oi = {.typep = &type, .sizep = &canonical_size}; | |
ac77d0c3 | 1947 | |
43fa44fa NTND |
1948 | if (IN_PACK(entry)) { |
1949 | struct packed_git *p = IN_PACK(entry); | |
03e79c88 | 1950 | struct pack_window *w_curs = NULL; |
3f83fd5e JK |
1951 | int have_base = 0; |
1952 | struct object_id base_ref; | |
5c49c116 NP |
1953 | struct object_entry *base_entry; |
1954 | unsigned long used, used_0; | |
ef49a7a0 | 1955 | unsigned long avail; |
5c49c116 NP |
1956 | off_t ofs; |
1957 | unsigned char *buf, c; | |
fd9b1bae | 1958 | enum object_type type; |
27a7d067 | 1959 | unsigned long in_pack_size; |
780e6e73 | 1960 | |
6777a59f | 1961 | buf = use_pack(p, &w_curs, entry->in_pack_offset, &avail); |
ab7cd7bb | 1962 | |
5c49c116 | 1963 | /* |
fa736f72 NP |
1964 | * We want in_pack_type even if we do not reuse delta |
1965 | * since non-delta representations could still be reused. | |
ab7cd7bb | 1966 | */ |
09ded04b | 1967 | used = unpack_object_header_buffer(buf, avail, |
fd9b1bae | 1968 | &type, |
27a7d067 | 1969 | &in_pack_size); |
03d66015 NP |
1970 | if (used == 0) |
1971 | goto give_up; | |
ab7cd7bb | 1972 | |
fd9b1bae NTND |
1973 | if (type < 0) |
1974 | BUG("invalid type %d", type); | |
1975 | entry->in_pack_type = type; | |
1976 | ||
5c49c116 NP |
1977 | /* |
1978 | * Determine if this is a delta and if so whether we can | |
1979 | * reuse it or not. Otherwise let's find out as cheaply as | |
1980 | * possible what the actual type and size for this object is. | |
3f9ac8d2 | 1981 | */ |
5c49c116 NP |
1982 | switch (entry->in_pack_type) { |
1983 | default: | |
1984 | /* Not a delta hence we've already got all we need. */ | |
fd9b1bae | 1985 | oe_set_type(entry, entry->in_pack_type); |
ac77d0c3 | 1986 | SET_SIZE(entry, in_pack_size); |
5c49c116 | 1987 | entry->in_pack_header_size = used; |
fd9b1bae | 1988 | if (oe_type(entry) < OBJ_COMMIT || oe_type(entry) > OBJ_BLOB) |
03d66015 | 1989 | goto give_up; |
5c49c116 NP |
1990 | unuse_pack(&w_curs); |
1991 | return; | |
1992 | case OBJ_REF_DELTA: | |
3f83fd5e JK |
1993 | if (reuse_delta && !entry->preferred_base) { |
1994 | oidread(&base_ref, | |
1995 | use_pack(p, &w_curs, | |
1996 | entry->in_pack_offset + used, | |
1997 | NULL)); | |
1998 | have_base = 1; | |
1999 | } | |
41179100 | 2000 | entry->in_pack_header_size = used + the_hash_algo->rawsz; |
5c49c116 NP |
2001 | break; |
2002 | case OBJ_OFS_DELTA: | |
2003 | buf = use_pack(p, &w_curs, | |
2004 | entry->in_pack_offset + used, NULL); | |
2005 | used_0 = 0; | |
2006 | c = buf[used_0++]; | |
2007 | ofs = c & 127; | |
2008 | while (c & 128) { | |
2009 | ofs += 1; | |
03d66015 | 2010 | if (!ofs || MSB(ofs, 7)) { |
f616db6a | 2011 | error(_("delta base offset overflow in pack for %s"), |
e6a492b7 | 2012 | oid_to_hex(&entry->idx.oid)); |
03d66015 NP |
2013 | goto give_up; |
2014 | } | |
5c49c116 NP |
2015 | c = buf[used_0++]; |
2016 | ofs = (ofs << 7) + (c & 127); | |
780e6e73 | 2017 | } |
5c49c116 | 2018 | ofs = entry->in_pack_offset - ofs; |
03d66015 | 2019 | if (ofs <= 0 || ofs >= entry->in_pack_offset) { |
f616db6a | 2020 | error(_("delta base offset out of bound for %s"), |
e6a492b7 | 2021 | oid_to_hex(&entry->idx.oid)); |
03d66015 NP |
2022 | goto give_up; |
2023 | } | |
a7de7130 | 2024 | if (reuse_delta && !entry->preferred_base) { |
eb3fd99e TB |
2025 | uint32_t pos; |
2026 | if (offset_to_pack_pos(p, ofs, &pos) < 0) | |
08698b1e | 2027 | goto give_up; |
eb3fd99e TB |
2028 | if (!nth_packed_object_id(&base_ref, p, |
2029 | pack_pos_to_index(p, pos))) | |
3f83fd5e | 2030 | have_base = 1; |
3449f8c4 | 2031 | } |
5c49c116 NP |
2032 | entry->in_pack_header_size = used + used_0; |
2033 | break; | |
780e6e73 | 2034 | } |
780e6e73 | 2035 | |
3f83fd5e JK |
2036 | if (have_base && |
2037 | can_reuse_delta(&base_ref, entry, &base_entry)) { | |
fd9b1bae | 2038 | oe_set_type(entry, entry->in_pack_type); |
ac77d0c3 | 2039 | SET_SIZE(entry, in_pack_size); /* delta size */ |
0aca34e8 | 2040 | SET_DELTA_SIZE(entry, in_pack_size); |
6a1e32d5 JK |
2041 | |
2042 | if (base_entry) { | |
2043 | SET_DELTA(entry, base_entry); | |
2044 | entry->delta_sibling_idx = base_entry->delta_child_idx; | |
2045 | SET_DELTA_CHILD(base_entry, entry); | |
2046 | } else { | |
a93c141d | 2047 | SET_DELTA_EXT(entry, &base_ref); |
6a1e32d5 JK |
2048 | } |
2049 | ||
5c49c116 NP |
2050 | unuse_pack(&w_curs); |
2051 | return; | |
2052 | } | |
ab7cd7bb | 2053 | |
fd9b1bae | 2054 | if (oe_type(entry)) { |
27a7d067 NTND |
2055 | off_t delta_pos; |
2056 | ||
5c49c116 NP |
2057 | /* |
2058 | * This must be a delta and we already know what the | |
2059 | * final object type is. Let's extract the actual | |
2060 | * object size from the delta header. | |
2061 | */ | |
27a7d067 | 2062 | delta_pos = entry->in_pack_offset + entry->in_pack_header_size; |
ac77d0c3 NTND |
2063 | canonical_size = get_size_from_delta(p, &w_curs, delta_pos); |
2064 | if (canonical_size == 0) | |
03d66015 | 2065 | goto give_up; |
ac77d0c3 | 2066 | SET_SIZE(entry, canonical_size); |
5c49c116 | 2067 | unuse_pack(&w_curs); |
3f9ac8d2 JH |
2068 | return; |
2069 | } | |
5c49c116 NP |
2070 | |
2071 | /* | |
2072 | * No choice but to fall back to the recursive delta walk | |
c93206b4 | 2073 | * with oid_object_info() to find about the object type |
5c49c116 NP |
2074 | * at this point... |
2075 | */ | |
03d66015 | 2076 | give_up: |
5c49c116 | 2077 | unuse_pack(&w_curs); |
36e4d74a | 2078 | } |
3f9ac8d2 | 2079 | |
8d5cf957 | 2080 | if (oid_object_info_extended(the_repository, &entry->idx.oid, &oi, |
e00549aa JT |
2081 | OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_LOOKUP_REPLACE) < 0) { |
2082 | if (has_promisor_remote()) { | |
2083 | prefetch_to_pack(object_index); | |
2084 | if (oid_object_info_extended(the_repository, &entry->idx.oid, &oi, | |
2085 | OBJECT_INFO_SKIP_FETCH_OBJECT | OBJECT_INFO_LOOKUP_REPLACE) < 0) | |
2086 | type = -1; | |
2087 | } else { | |
2088 | type = -1; | |
2089 | } | |
2090 | } | |
8d5cf957 | 2091 | oe_set_type(entry, type); |
ac77d0c3 NTND |
2092 | if (entry->type_valid) { |
2093 | SET_SIZE(entry, canonical_size); | |
2094 | } else { | |
2095 | /* | |
2096 | * Bad object type is checked in prepare_pack(). This is | |
2097 | * to permit a missing preferred base object to be ignored | |
2098 | * as a preferred base. Doing so can result in a larger | |
2099 | * pack file, but the transfer will still take place. | |
2100 | */ | |
2101 | } | |
3f9ac8d2 JH |
2102 | } |
2103 | ||
5c49c116 NP |
2104 | static int pack_offset_sort(const void *_a, const void *_b) |
2105 | { | |
2106 | const struct object_entry *a = *(struct object_entry **)_a; | |
2107 | const struct object_entry *b = *(struct object_entry **)_b; | |
43fa44fa NTND |
2108 | const struct packed_git *a_in_pack = IN_PACK(a); |
2109 | const struct packed_git *b_in_pack = IN_PACK(b); | |
5c49c116 NP |
2110 | |
2111 | /* avoid filesystem trashing with loose objects */ | |
43fa44fa | 2112 | if (!a_in_pack && !b_in_pack) |
e6a492b7 | 2113 | return oidcmp(&a->idx.oid, &b->idx.oid); |
5c49c116 | 2114 | |
43fa44fa | 2115 | if (a_in_pack < b_in_pack) |
5c49c116 | 2116 | return -1; |
43fa44fa | 2117 | if (a_in_pack > b_in_pack) |
5c49c116 NP |
2118 | return 1; |
2119 | return a->in_pack_offset < b->in_pack_offset ? -1 : | |
2120 | (a->in_pack_offset > b->in_pack_offset); | |
2121 | } | |
2122 | ||
4cf2143e JK |
2123 | /* |
2124 | * Drop an on-disk delta we were planning to reuse. Naively, this would | |
2125 | * just involve blanking out the "delta" field, but we have to deal | |
2126 | * with some extra book-keeping: | |
2127 | * | |
2128 | * 1. Removing ourselves from the delta_sibling linked list. | |
2129 | * | |
2130 | * 2. Updating our size/type to the non-delta representation. These were | |
2131 | * either not recorded initially (size) or overwritten with the delta type | |
2132 | * (type) when check_object() decided to reuse the delta. | |
7dbabbbe JK |
2133 | * |
2134 | * 3. Resetting our delta depth, as we are now a base object. | |
4cf2143e JK |
2135 | */ |
2136 | static void drop_reused_delta(struct object_entry *entry) | |
2137 | { | |
898eba5e | 2138 | unsigned *idx = &to_pack.objects[entry->delta_idx - 1].delta_child_idx; |
4cf2143e | 2139 | struct object_info oi = OBJECT_INFO_INIT; |
fd9b1bae | 2140 | enum object_type type; |
ac77d0c3 | 2141 | unsigned long size; |
4cf2143e | 2142 | |
898eba5e NTND |
2143 | while (*idx) { |
2144 | struct object_entry *oe = &to_pack.objects[*idx - 1]; | |
4cf2143e | 2145 | |
898eba5e NTND |
2146 | if (oe == entry) |
2147 | *idx = oe->delta_sibling_idx; | |
4cf2143e | 2148 | else |
898eba5e | 2149 | idx = &oe->delta_sibling_idx; |
4cf2143e | 2150 | } |
898eba5e | 2151 | SET_DELTA(entry, NULL); |
7dbabbbe | 2152 | entry->depth = 0; |
4cf2143e | 2153 | |
ac77d0c3 | 2154 | oi.sizep = &size; |
fd9b1bae | 2155 | oi.typep = &type; |
ad635e82 | 2156 | if (packed_object_info(the_repository, IN_PACK(entry), entry->in_pack_offset, &oi) < 0) { |
4cf2143e JK |
2157 | /* |
2158 | * We failed to get the info from this pack for some reason; | |
c93206b4 | 2159 | * fall back to oid_object_info, which may find another copy. |
fd9b1bae | 2160 | * And if that fails, the error will be recorded in oe_type(entry) |
4cf2143e JK |
2161 | * and dealt with in prepare_pack(). |
2162 | */ | |
ad635e82 JH |
2163 | oe_set_type(entry, |
2164 | oid_object_info(the_repository, &entry->idx.oid, &size)); | |
fd9b1bae NTND |
2165 | } else { |
2166 | oe_set_type(entry, type); | |
4cf2143e | 2167 | } |
ac77d0c3 | 2168 | SET_SIZE(entry, size); |
4cf2143e JK |
2169 | } |
2170 | ||
2171 | /* | |
2172 | * Follow the chain of deltas from this entry onward, throwing away any links | |
2173 | * that cause us to hit a cycle (as determined by the DFS state flags in | |
2174 | * the entries). | |
7dbabbbe JK |
2175 | * |
2176 | * We also detect too-long reused chains that would violate our --depth | |
2177 | * limit. | |
4cf2143e JK |
2178 | */ |
2179 | static void break_delta_chains(struct object_entry *entry) | |
2180 | { | |
42b766d7 JK |
2181 | /* |
2182 | * The actual depth of each object we will write is stored as an int, | |
2183 | * as it cannot exceed our int "depth" limit. But before we break | |
2184 | * changes based no that limit, we may potentially go as deep as the | |
2185 | * number of objects, which is elsewhere bounded to a uint32_t. | |
2186 | */ | |
2187 | uint32_t total_depth; | |
2188 | struct object_entry *cur, *next; | |
2189 | ||
2190 | for (cur = entry, total_depth = 0; | |
2191 | cur; | |
898eba5e | 2192 | cur = DELTA(cur), total_depth++) { |
42b766d7 JK |
2193 | if (cur->dfs_state == DFS_DONE) { |
2194 | /* | |
2195 | * We've already seen this object and know it isn't | |
2196 | * part of a cycle. We do need to append its depth | |
2197 | * to our count. | |
2198 | */ | |
2199 | total_depth += cur->depth; | |
2200 | break; | |
2201 | } | |
4cf2143e | 2202 | |
4cf2143e | 2203 | /* |
42b766d7 JK |
2204 | * We break cycles before looping, so an ACTIVE state (or any |
2205 | * other cruft which made its way into the state variable) | |
2206 | * is a bug. | |
4cf2143e | 2207 | */ |
42b766d7 | 2208 | if (cur->dfs_state != DFS_NONE) |
033abf97 | 2209 | BUG("confusing delta dfs state in first pass: %d", |
42b766d7 | 2210 | cur->dfs_state); |
4cf2143e | 2211 | |
4cf2143e | 2212 | /* |
42b766d7 JK |
2213 | * Now we know this is the first time we've seen the object. If |
2214 | * it's not a delta, we're done traversing, but we'll mark it | |
2215 | * done to save time on future traversals. | |
4cf2143e | 2216 | */ |
898eba5e | 2217 | if (!DELTA(cur)) { |
42b766d7 JK |
2218 | cur->dfs_state = DFS_DONE; |
2219 | break; | |
2220 | } | |
4cf2143e | 2221 | |
4cf2143e | 2222 | /* |
42b766d7 JK |
2223 | * Mark ourselves as active and see if the next step causes |
2224 | * us to cycle to another active object. It's important to do | |
2225 | * this _before_ we loop, because it impacts where we make the | |
2226 | * cut, and thus how our total_depth counter works. | |
2227 | * E.g., We may see a partial loop like: | |
2228 | * | |
2229 | * A -> B -> C -> D -> B | |
2230 | * | |
2231 | * Cutting B->C breaks the cycle. But now the depth of A is | |
2232 | * only 1, and our total_depth counter is at 3. The size of the | |
2233 | * error is always one less than the size of the cycle we | |
2234 | * broke. Commits C and D were "lost" from A's chain. | |
2235 | * | |
2236 | * If we instead cut D->B, then the depth of A is correct at 3. | |
2237 | * We keep all commits in the chain that we examined. | |
4cf2143e | 2238 | */ |
42b766d7 | 2239 | cur->dfs_state = DFS_ACTIVE; |
898eba5e | 2240 | if (DELTA(cur)->dfs_state == DFS_ACTIVE) { |
42b766d7 JK |
2241 | drop_reused_delta(cur); |
2242 | cur->dfs_state = DFS_DONE; | |
2243 | break; | |
7dbabbbe | 2244 | } |
42b766d7 | 2245 | } |
7dbabbbe | 2246 | |
42b766d7 JK |
2247 | /* |
2248 | * And now that we've gone all the way to the bottom of the chain, we | |
2249 | * need to clear the active flags and set the depth fields as | |
2250 | * appropriate. Unlike the loop above, which can quit when it drops a | |
2251 | * delta, we need to keep going to look for more depth cuts. So we need | |
2252 | * an extra "next" pointer to keep going after we reset cur->delta. | |
2253 | */ | |
2254 | for (cur = entry; cur; cur = next) { | |
898eba5e | 2255 | next = DELTA(cur); |
4cf2143e | 2256 | |
42b766d7 JK |
2257 | /* |
2258 | * We should have a chain of zero or more ACTIVE states down to | |
2259 | * a final DONE. We can quit after the DONE, because either it | |
2260 | * has no bases, or we've already handled them in a previous | |
2261 | * call. | |
2262 | */ | |
2263 | if (cur->dfs_state == DFS_DONE) | |
2264 | break; | |
2265 | else if (cur->dfs_state != DFS_ACTIVE) | |
033abf97 | 2266 | BUG("confusing delta dfs state in second pass: %d", |
42b766d7 | 2267 | cur->dfs_state); |
4cf2143e | 2268 | |
4cf2143e | 2269 | /* |
42b766d7 JK |
2270 | * If the total_depth is more than depth, then we need to snip |
2271 | * the chain into two or more smaller chains that don't exceed | |
2272 | * the maximum depth. Most of the resulting chains will contain | |
2273 | * (depth + 1) entries (i.e., depth deltas plus one base), and | |
2274 | * the last chain (i.e., the one containing entry) will contain | |
2275 | * whatever entries are left over, namely | |
2276 | * (total_depth % (depth + 1)) of them. | |
2277 | * | |
2278 | * Since we are iterating towards decreasing depth, we need to | |
2279 | * decrement total_depth as we go, and we need to write to the | |
2280 | * entry what its final depth will be after all of the | |
2281 | * snipping. Since we're snipping into chains of length (depth | |
2282 | * + 1) entries, the final depth of an entry will be its | |
2283 | * original depth modulo (depth + 1). Any time we encounter an | |
2284 | * entry whose final depth is supposed to be zero, we snip it | |
2285 | * from its delta base, thereby making it so. | |
4cf2143e | 2286 | */ |
42b766d7 JK |
2287 | cur->depth = (total_depth--) % (depth + 1); |
2288 | if (!cur->depth) | |
2289 | drop_reused_delta(cur); | |
2290 | ||
2291 | cur->dfs_state = DFS_DONE; | |
4cf2143e JK |
2292 | } |
2293 | } | |
2294 | ||
c323ac7d LT |
2295 | static void get_object_details(void) |
2296 | { | |
7cadf491 | 2297 | uint32_t i; |
5c49c116 NP |
2298 | struct object_entry **sorted_by_offset; |
2299 | ||
5af05043 NTND |
2300 | if (progress) |
2301 | progress_state = start_progress(_("Counting objects"), | |
2302 | to_pack.nr_objects); | |
2303 | ||
ca56dadb | 2304 | CALLOC_ARRAY(sorted_by_offset, to_pack.nr_objects); |
2834bc27 VM |
2305 | for (i = 0; i < to_pack.nr_objects; i++) |
2306 | sorted_by_offset[i] = to_pack.objects + i; | |
9ed0d8d6 | 2307 | QSORT(sorted_by_offset, to_pack.nr_objects, pack_offset_sort); |
c323ac7d | 2308 | |
2834bc27 | 2309 | for (i = 0; i < to_pack.nr_objects; i++) { |
15366280 | 2310 | struct object_entry *entry = sorted_by_offset[i]; |
e00549aa | 2311 | check_object(entry, i); |
ac77d0c3 NTND |
2312 | if (entry->type_valid && |
2313 | oe_size_greater_than(&to_pack, entry, big_file_threshold)) | |
15366280 | 2314 | entry->no_try_delta = 1; |
5af05043 | 2315 | display_progress(progress_state, i + 1); |
15366280 | 2316 | } |
5af05043 | 2317 | stop_progress(&progress_state); |
3449f8c4 | 2318 | |
4cf2143e JK |
2319 | /* |
2320 | * This must happen in a second pass, since we rely on the delta | |
2321 | * information for the whole list being completed. | |
2322 | */ | |
2323 | for (i = 0; i < to_pack.nr_objects; i++) | |
2324 | break_delta_chains(&to_pack.objects[i]); | |
2325 | ||
5c49c116 | 2326 | free(sorted_by_offset); |
c323ac7d LT |
2327 | } |
2328 | ||
b904166c NP |
2329 | /* |
2330 | * We search for deltas in a list sorted by type, by filename hash, and then | |
2331 | * by size, so that we see progressively smaller and smaller files. | |
2332 | * That's because we prefer deltas to be from the bigger file | |
2333 | * to the smaller -- deletes are potentially cheaper, but perhaps | |
2334 | * more importantly, the bigger file is likely the more recent | |
2335 | * one. The deepest deltas are therefore the oldest objects which are | |
2336 | * less susceptible to be accessed often. | |
2337 | */ | |
9668cf59 | 2338 | static int type_size_sort(const void *_a, const void *_b) |
c323ac7d | 2339 | { |
9668cf59 NP |
2340 | const struct object_entry *a = *(struct object_entry **)_a; |
2341 | const struct object_entry *b = *(struct object_entry **)_b; | |
33de80b1 SL |
2342 | const enum object_type a_type = oe_type(a); |
2343 | const enum object_type b_type = oe_type(b); | |
2344 | const unsigned long a_size = SIZE(a); | |
2345 | const unsigned long b_size = SIZE(b); | |
9668cf59 | 2346 | |
fd9b1bae | 2347 | if (a_type > b_type) |
27225f2e | 2348 | return -1; |
fd9b1bae | 2349 | if (a_type < b_type) |
27225f2e | 2350 | return 1; |
b904166c | 2351 | if (a->hash > b->hash) |
7a979d99 | 2352 | return -1; |
b904166c | 2353 | if (a->hash < b->hash) |
7a979d99 | 2354 | return 1; |
b904166c | 2355 | if (a->preferred_base > b->preferred_base) |
c323ac7d | 2356 | return -1; |
b904166c NP |
2357 | if (a->preferred_base < b->preferred_base) |
2358 | return 1; | |
28b8a730 | 2359 | if (use_delta_islands) { |
33de80b1 | 2360 | const int island_cmp = island_delta_cmp(&a->idx.oid, &b->idx.oid); |
28b8a730 JK |
2361 | if (island_cmp) |
2362 | return island_cmp; | |
2363 | } | |
ac77d0c3 | 2364 | if (a_size > b_size) |
b904166c | 2365 | return -1; |
ac77d0c3 | 2366 | if (a_size < b_size) |
c323ac7d | 2367 | return 1; |
b904166c | 2368 | return a < b ? -1 : (a > b); /* newest first */ |
c323ac7d LT |
2369 | } |
2370 | ||
2371 | struct unpacked { | |
2372 | struct object_entry *entry; | |
2373 | void *data; | |
f6c7081a | 2374 | struct delta_index *index; |
5a235b5e | 2375 | unsigned depth; |
c323ac7d LT |
2376 | }; |
2377 | ||
d250626c NP |
2378 | static int delta_cacheable(unsigned long src_size, unsigned long trg_size, |
2379 | unsigned long delta_size) | |
074b2eea MK |
2380 | { |
2381 | if (max_delta_cache_size && delta_cache_size + delta_size > max_delta_cache_size) | |
2382 | return 0; | |
2383 | ||
e3dfddb3 MK |
2384 | if (delta_size < cache_max_small_delta_size) |
2385 | return 1; | |
2386 | ||
074b2eea MK |
2387 | /* cache delta, if objects are large enough compared to delta size */ |
2388 | if ((src_size >> 20) + (trg_size >> 21) > (delta_size >> 10)) | |
2389 | return 1; | |
2390 | ||
2391 | return 0; | |
2392 | } | |
2393 | ||
ffbd51cc | 2394 | /* Protect delta_cache_size */ |
44626dc7 | 2395 | static pthread_mutex_t cache_mutex; |
3c701839 NP |
2396 | #define cache_lock() pthread_mutex_lock(&cache_mutex) |
2397 | #define cache_unlock() pthread_mutex_unlock(&cache_mutex) | |
2398 | ||
ffbd51cc NTND |
2399 | /* |
2400 | * Protect object list partitioning (e.g. struct thread_param) and | |
2401 | * progress_state | |
2402 | */ | |
44626dc7 | 2403 | static pthread_mutex_t progress_mutex; |
8ecce684 NP |
2404 | #define progress_lock() pthread_mutex_lock(&progress_mutex) |
2405 | #define progress_unlock() pthread_mutex_unlock(&progress_mutex) | |
2406 | ||
ffbd51cc NTND |
2407 | /* |
2408 | * Access to struct object_entry is unprotected since each thread owns | |
2409 | * a portion of the main object list. Just don't access object entries | |
2410 | * ahead in the list because they can be stolen and would need | |
2411 | * progress_mutex for protection. | |
2412 | */ | |
8ecce684 | 2413 | |
7d089fb9 ÆAB |
2414 | static inline int oe_size_less_than(struct packing_data *pack, |
2415 | const struct object_entry *lhs, | |
2416 | unsigned long rhs) | |
2417 | { | |
2418 | if (lhs->size_valid) | |
2419 | return lhs->size_ < rhs; | |
2420 | if (rhs < pack->oe_size_limit) /* rhs < 2^x <= lhs ? */ | |
2421 | return 0; | |
2422 | return oe_get_size_slow(pack, lhs) < rhs; | |
2423 | } | |
2424 | ||
2425 | static inline void oe_set_tree_depth(struct packing_data *pack, | |
2426 | struct object_entry *e, | |
2427 | unsigned int tree_depth) | |
2428 | { | |
2429 | if (!pack->tree_depth) | |
2430 | CALLOC_ARRAY(pack->tree_depth, pack->nr_alloc); | |
2431 | pack->tree_depth[e - pack->objects] = tree_depth; | |
2432 | } | |
2433 | ||
ac77d0c3 NTND |
2434 | /* |
2435 | * Return the size of the object without doing any delta | |
2436 | * reconstruction (so non-deltas are true object sizes, but deltas | |
2437 | * return the size of the delta data). | |
2438 | */ | |
2439 | unsigned long oe_get_size_slow(struct packing_data *pack, | |
2440 | const struct object_entry *e) | |
2441 | { | |
2442 | struct packed_git *p; | |
2443 | struct pack_window *w_curs; | |
2444 | unsigned char *buf; | |
2445 | enum object_type type; | |
2446 | unsigned long used, avail, size; | |
2447 | ||
2448 | if (e->type_ != OBJ_OFS_DELTA && e->type_ != OBJ_REF_DELTA) { | |
edb673cf | 2449 | packing_data_lock(&to_pack); |
ad635e82 | 2450 | if (oid_object_info(the_repository, &e->idx.oid, &size) < 0) |
ac77d0c3 NTND |
2451 | die(_("unable to get size of %s"), |
2452 | oid_to_hex(&e->idx.oid)); | |
edb673cf | 2453 | packing_data_unlock(&to_pack); |
ac77d0c3 NTND |
2454 | return size; |
2455 | } | |
2456 | ||
2457 | p = oe_in_pack(pack, e); | |
2458 | if (!p) | |
2459 | BUG("when e->type is a delta, it must belong to a pack"); | |
2460 | ||
edb673cf | 2461 | packing_data_lock(&to_pack); |
ac77d0c3 NTND |
2462 | w_curs = NULL; |
2463 | buf = use_pack(p, &w_curs, e->in_pack_offset, &avail); | |
2464 | used = unpack_object_header_buffer(buf, avail, &type, &size); | |
2465 | if (used == 0) | |
2466 | die(_("unable to parse object header of %s"), | |
2467 | oid_to_hex(&e->idx.oid)); | |
2468 | ||
2469 | unuse_pack(&w_curs); | |
edb673cf | 2470 | packing_data_unlock(&to_pack); |
ac77d0c3 NTND |
2471 | return size; |
2472 | } | |
2473 | ||
f6c7081a | 2474 | static int try_delta(struct unpacked *trg, struct unpacked *src, |
ef0316fc | 2475 | unsigned max_depth, unsigned long *mem_usage) |
c323ac7d | 2476 | { |
f6c7081a NP |
2477 | struct object_entry *trg_entry = trg->entry; |
2478 | struct object_entry *src_entry = src->entry; | |
560b25a8 | 2479 | unsigned long trg_size, src_size, delta_size, sizediff, max_size, sz; |
c83f032e | 2480 | unsigned ref_depth; |
21666f1a | 2481 | enum object_type type; |
c323ac7d LT |
2482 | void *delta_buf; |
2483 | ||
2484 | /* Don't bother doing diffs between different types */ | |
fd9b1bae | 2485 | if (oe_type(trg_entry) != oe_type(src_entry)) |
c323ac7d LT |
2486 | return -1; |
2487 | ||
51d1e83f | 2488 | /* |
15f07e06 JK |
2489 | * We do not bother to try a delta that we discarded on an |
2490 | * earlier try, but only when reusing delta data. Note that | |
2491 | * src_entry that is marked as the preferred_base should always | |
2492 | * be considered, as even if we produce a suboptimal delta against | |
2493 | * it, we will still save the transfer cost, as we already know | |
2494 | * the other side has it and we won't send src_entry at all. | |
51d1e83f | 2495 | */ |
43fa44fa NTND |
2496 | if (reuse_delta && IN_PACK(trg_entry) && |
2497 | IN_PACK(trg_entry) == IN_PACK(src_entry) && | |
15f07e06 | 2498 | !src_entry->preferred_base && |
e9195b58 JH |
2499 | trg_entry->in_pack_type != OBJ_REF_DELTA && |
2500 | trg_entry->in_pack_type != OBJ_OFS_DELTA) | |
51d1e83f LT |
2501 | return 0; |
2502 | ||
898b14ce | 2503 | /* Let's not bust the allowed depth. */ |
5a235b5e | 2504 | if (src->depth >= max_depth) |
d116a45a | 2505 | return 0; |
c323ac7d | 2506 | |
c3b06a69 | 2507 | /* Now some size filtering heuristics. */ |
ac77d0c3 | 2508 | trg_size = SIZE(trg_entry); |
898eba5e | 2509 | if (!DELTA(trg_entry)) { |
41179100 | 2510 | max_size = trg_size/2 - the_hash_algo->rawsz; |
c83f032e NP |
2511 | ref_depth = 1; |
2512 | } else { | |
0aca34e8 | 2513 | max_size = DELTA_SIZE(trg_entry); |
5a235b5e | 2514 | ref_depth = trg->depth; |
c83f032e | 2515 | } |
720fe22d | 2516 | max_size = (uint64_t)max_size * (max_depth - src->depth) / |
c83f032e | 2517 | (max_depth - ref_depth + 1); |
c3b06a69 NP |
2518 | if (max_size == 0) |
2519 | return 0; | |
ac77d0c3 | 2520 | src_size = SIZE(src_entry); |
560b25a8 | 2521 | sizediff = src_size < trg_size ? trg_size - src_size : 0; |
27225f2e | 2522 | if (sizediff >= max_size) |
f527cb8c | 2523 | return 0; |
a1dab41a BD |
2524 | if (trg_size < src_size / 32) |
2525 | return 0; | |
f6c7081a | 2526 | |
28b8a730 JK |
2527 | if (!in_same_island(&trg->entry->idx.oid, &src->entry->idx.oid)) |
2528 | return 0; | |
2529 | ||
560b25a8 NP |
2530 | /* Load data if not already done */ |
2531 | if (!trg->data) { | |
edb673cf | 2532 | packing_data_lock(&to_pack); |
b4f5aca4 | 2533 | trg->data = read_object_file(&trg_entry->idx.oid, &type, &sz); |
edb673cf | 2534 | packing_data_unlock(&to_pack); |
2e3404c3 | 2535 | if (!trg->data) |
f616db6a | 2536 | die(_("object %s cannot be read"), |
e6a492b7 | 2537 | oid_to_hex(&trg_entry->idx.oid)); |
560b25a8 | 2538 | if (sz != trg_size) |
ca473cef TB |
2539 | die(_("object %s inconsistent object length (%"PRIuMAX" vs %"PRIuMAX")"), |
2540 | oid_to_hex(&trg_entry->idx.oid), (uintmax_t)sz, | |
2541 | (uintmax_t)trg_size); | |
ef0316fc | 2542 | *mem_usage += sz; |
560b25a8 NP |
2543 | } |
2544 | if (!src->data) { | |
edb673cf | 2545 | packing_data_lock(&to_pack); |
b4f5aca4 | 2546 | src->data = read_object_file(&src_entry->idx.oid, &type, &sz); |
edb673cf | 2547 | packing_data_unlock(&to_pack); |
71064a95 NP |
2548 | if (!src->data) { |
2549 | if (src_entry->preferred_base) { | |
2550 | static int warned = 0; | |
2551 | if (!warned++) | |
f616db6a | 2552 | warning(_("object %s cannot be read"), |
e6a492b7 | 2553 | oid_to_hex(&src_entry->idx.oid)); |
71064a95 NP |
2554 | /* |
2555 | * Those objects are not included in the | |
2556 | * resulting pack. Be resilient and ignore | |
2557 | * them if they can't be read, in case the | |
2558 | * pack could be created nevertheless. | |
2559 | */ | |
2560 | return 0; | |
2561 | } | |
f616db6a | 2562 | die(_("object %s cannot be read"), |
e6a492b7 | 2563 | oid_to_hex(&src_entry->idx.oid)); |
71064a95 | 2564 | } |
560b25a8 | 2565 | if (sz != src_size) |
ca473cef TB |
2566 | die(_("object %s inconsistent object length (%"PRIuMAX" vs %"PRIuMAX")"), |
2567 | oid_to_hex(&src_entry->idx.oid), (uintmax_t)sz, | |
2568 | (uintmax_t)src_size); | |
ef0316fc | 2569 | *mem_usage += sz; |
560b25a8 NP |
2570 | } |
2571 | if (!src->index) { | |
2572 | src->index = create_delta_index(src->data, src_size); | |
a588d88a MK |
2573 | if (!src->index) { |
2574 | static int warned = 0; | |
2575 | if (!warned++) | |
f616db6a | 2576 | warning(_("suboptimal pack - out of memory")); |
a588d88a MK |
2577 | return 0; |
2578 | } | |
ef0316fc | 2579 | *mem_usage += sizeof_delta_index(src->index); |
560b25a8 NP |
2580 | } |
2581 | ||
2582 | delta_buf = create_delta(src->index, trg->data, trg_size, &delta_size, max_size); | |
c323ac7d | 2583 | if (!delta_buf) |
75c42d8c | 2584 | return 0; |
f6c7081a | 2585 | |
898eba5e | 2586 | if (DELTA(trg_entry)) { |
848d732c | 2587 | /* Prefer only shallower same-sized deltas. */ |
0aca34e8 | 2588 | if (delta_size == DELTA_SIZE(trg_entry) && |
5a235b5e | 2589 | src->depth + 1 >= trg->depth) { |
848d732c BD |
2590 | free(delta_buf); |
2591 | return 0; | |
2592 | } | |
074b2eea | 2593 | } |
9e2d57a0 | 2594 | |
3c701839 NP |
2595 | /* |
2596 | * Handle memory allocation outside of the cache | |
2597 | * accounting lock. Compiler will optimize the strangeness | |
7eb151d6 | 2598 | * away when NO_PTHREADS is defined. |
3c701839 | 2599 | */ |
8e0f7003 | 2600 | free(trg_entry->delta_data); |
3c701839 | 2601 | cache_lock(); |
9e2d57a0 | 2602 | if (trg_entry->delta_data) { |
0aca34e8 | 2603 | delta_cache_size -= DELTA_SIZE(trg_entry); |
9e2d57a0 NP |
2604 | trg_entry->delta_data = NULL; |
2605 | } | |
d250626c | 2606 | if (delta_cacheable(src_size, trg_size, delta_size)) { |
b7a28f78 | 2607 | delta_cache_size += delta_size; |
3c701839 NP |
2608 | cache_unlock(); |
2609 | trg_entry->delta_data = xrealloc(delta_buf, delta_size); | |
2610 | } else { | |
2611 | cache_unlock(); | |
074b2eea | 2612 | free(delta_buf); |
3c701839 NP |
2613 | } |
2614 | ||
898eba5e | 2615 | SET_DELTA(trg_entry, src_entry); |
0aca34e8 | 2616 | SET_DELTA_SIZE(trg_entry, delta_size); |
b7a28f78 NP |
2617 | trg->depth = src->depth + 1; |
2618 | ||
f6c7081a | 2619 | return 1; |
c323ac7d LT |
2620 | } |
2621 | ||
898b14ce | 2622 | static unsigned int check_delta_limit(struct object_entry *me, unsigned int n) |
b2504a0d | 2623 | { |
898eba5e | 2624 | struct object_entry *child = DELTA_CHILD(me); |
898b14ce NP |
2625 | unsigned int m = n; |
2626 | while (child) { | |
33de80b1 | 2627 | const unsigned int c = check_delta_limit(child, n + 1); |
898b14ce NP |
2628 | if (m < c) |
2629 | m = c; | |
898eba5e | 2630 | child = DELTA_SIBLING(child); |
898b14ce NP |
2631 | } |
2632 | return m; | |
b2504a0d NP |
2633 | } |
2634 | ||
75ad235c | 2635 | static unsigned long free_unpacked(struct unpacked *n) |
a97773ce | 2636 | { |
ef0316fc | 2637 | unsigned long freed_mem = sizeof_delta_index(n->index); |
a97773ce BD |
2638 | free_delta_index(n->index); |
2639 | n->index = NULL; | |
2640 | if (n->data) { | |
ac77d0c3 | 2641 | freed_mem += SIZE(n->entry); |
6a83d902 | 2642 | FREE_AND_NULL(n->data); |
a97773ce BD |
2643 | } |
2644 | n->entry = NULL; | |
7d7baa5e | 2645 | n->depth = 0; |
ef0316fc | 2646 | return freed_mem; |
a97773ce BD |
2647 | } |
2648 | ||
384b32c0 | 2649 | static void find_deltas(struct object_entry **list, unsigned *list_size, |
e334977d | 2650 | int window, int depth, unsigned *processed) |
c323ac7d | 2651 | { |
384b32c0 | 2652 | uint32_t i, idx = 0, count = 0; |
7cadf491 | 2653 | struct unpacked *array; |
ef0316fc | 2654 | unsigned long mem_usage = 0; |
c323ac7d | 2655 | |
ca56dadb | 2656 | CALLOC_ARRAY(array, window); |
21fcd1bd | 2657 | |
384b32c0 | 2658 | for (;;) { |
421b488a | 2659 | struct object_entry *entry; |
c323ac7d | 2660 | struct unpacked *n = array + idx; |
ef0316fc | 2661 | int j, max_depth, best_base = -1; |
c323ac7d | 2662 | |
384b32c0 NP |
2663 | progress_lock(); |
2664 | if (!*list_size) { | |
2665 | progress_unlock(); | |
2666 | break; | |
2667 | } | |
421b488a | 2668 | entry = *list++; |
384b32c0 NP |
2669 | (*list_size)--; |
2670 | if (!entry->preferred_base) { | |
2671 | (*processed)++; | |
2672 | display_progress(progress_state, *processed); | |
2673 | } | |
2674 | progress_unlock(); | |
2675 | ||
ef0316fc | 2676 | mem_usage -= free_unpacked(n); |
c323ac7d | 2677 | n->entry = entry; |
ab7cd7bb | 2678 | |
a97773ce | 2679 | while (window_memory_limit && |
ef0316fc | 2680 | mem_usage > window_memory_limit && |
a97773ce | 2681 | count > 1) { |
33de80b1 | 2682 | const uint32_t tail = (idx + window - count) % window; |
75ad235c | 2683 | mem_usage -= free_unpacked(array + tail); |
a97773ce BD |
2684 | count--; |
2685 | } | |
2686 | ||
75d39853 NP |
2687 | /* We do not compute delta to *create* objects we are not |
2688 | * going to pack. | |
2689 | */ | |
2690 | if (entry->preferred_base) | |
2691 | goto next; | |
2692 | ||
898b14ce NP |
2693 | /* |
2694 | * If the current object is at pack edge, take the depth the | |
2695 | * objects that depend on the current object into account | |
2696 | * otherwise they would become too deep. | |
2697 | */ | |
2698 | max_depth = depth; | |
898eba5e | 2699 | if (DELTA_CHILD(entry)) { |
898b14ce NP |
2700 | max_depth -= check_delta_limit(entry, 0); |
2701 | if (max_depth <= 0) | |
2702 | goto next; | |
2703 | } | |
2704 | ||
78817c15 LT |
2705 | j = window; |
2706 | while (--j > 0) { | |
77639870 | 2707 | int ret; |
7cadf491 | 2708 | uint32_t other_idx = idx + j; |
c323ac7d | 2709 | struct unpacked *m; |
78817c15 LT |
2710 | if (other_idx >= window) |
2711 | other_idx -= window; | |
c323ac7d LT |
2712 | m = array + other_idx; |
2713 | if (!m->entry) | |
2714 | break; | |
ef0316fc | 2715 | ret = try_delta(n, m, max_depth, &mem_usage); |
77639870 | 2716 | if (ret < 0) |
c323ac7d | 2717 | break; |
77639870 JH |
2718 | else if (ret > 0) |
2719 | best_base = other_idx; | |
c323ac7d | 2720 | } |
898b14ce | 2721 | |
ed4a9031 NP |
2722 | /* |
2723 | * If we decided to cache the delta data, then it is best | |
2724 | * to compress it right away. First because we have to do | |
2725 | * it anyway, and doing it here while we're threaded will | |
2726 | * save a lot of time in the non threaded write phase, | |
2727 | * as well as allow for caching more deltas within | |
2728 | * the same cache size limit. | |
2729 | * ... | |
2730 | * But only if not writing to stdout, since in that case | |
2731 | * the network is most likely throttling writes anyway, | |
2732 | * and therefore it is best to go to the write phase ASAP | |
2733 | * instead, as we can afford spending more time compressing | |
2734 | * between writes at that moment. | |
2735 | */ | |
2736 | if (entry->delta_data && !pack_to_stdout) { | |
0cb3c142 NTND |
2737 | unsigned long size; |
2738 | ||
0aca34e8 | 2739 | size = do_compress(&entry->delta_data, DELTA_SIZE(entry)); |
0cb3c142 NTND |
2740 | if (size < (1U << OE_Z_DELTA_BITS)) { |
2741 | entry->z_delta_size = size; | |
2742 | cache_lock(); | |
0aca34e8 | 2743 | delta_cache_size -= DELTA_SIZE(entry); |
0cb3c142 NTND |
2744 | delta_cache_size += entry->z_delta_size; |
2745 | cache_unlock(); | |
2746 | } else { | |
2747 | FREE_AND_NULL(entry->delta_data); | |
2748 | entry->z_delta_size = 0; | |
2749 | } | |
ed4a9031 NP |
2750 | } |
2751 | ||
70ca1a3f JH |
2752 | /* if we made n a delta, and if n is already at max |
2753 | * depth, leaving it in the window is pointless. we | |
2754 | * should evict it first. | |
70ca1a3f | 2755 | */ |
898eba5e | 2756 | if (DELTA(entry) && max_depth <= n->depth) |
70ca1a3f | 2757 | continue; |
ff45715c | 2758 | |
77639870 JH |
2759 | /* |
2760 | * Move the best delta base up in the window, after the | |
2761 | * currently deltified object, to keep it longer. It will | |
2762 | * be the first base object to be attempted next. | |
2763 | */ | |
898eba5e | 2764 | if (DELTA(entry)) { |
77639870 JH |
2765 | struct unpacked swap = array[best_base]; |
2766 | int dist = (window + idx - best_base) % window; | |
2767 | int dst = best_base; | |
2768 | while (dist--) { | |
2769 | int src = (dst + 1) % window; | |
2770 | array[dst] = array[src]; | |
2771 | dst = src; | |
2772 | } | |
2773 | array[dst] = swap; | |
2774 | } | |
2775 | ||
898b14ce | 2776 | next: |
521a4f4c | 2777 | idx++; |
a97773ce BD |
2778 | if (count + 1 < window) |
2779 | count++; | |
521a4f4c LT |
2780 | if (idx >= window) |
2781 | idx = 0; | |
384b32c0 | 2782 | } |
adee7bdf | 2783 | |
f6c7081a | 2784 | for (i = 0; i < window; ++i) { |
ff45715c | 2785 | free_delta_index(array[i].index); |
adee7bdf | 2786 | free(array[i].data); |
f6c7081a | 2787 | } |
adee7bdf | 2788 | free(array); |
c323ac7d LT |
2789 | } |
2790 | ||
50f22ada | 2791 | /* |
ffbd51cc NTND |
2792 | * The main object list is split into smaller lists, each is handed to |
2793 | * one worker. | |
2794 | * | |
50f22ada JS |
2795 | * The main thread waits on the condition that (at least) one of the workers |
2796 | * has stopped working (which is indicated in the .working member of | |
2797 | * struct thread_params). | |
ffbd51cc | 2798 | * |
50f22ada JS |
2799 | * When a work thread has completed its work, it sets .working to 0 and |
2800 | * signals the main thread and waits on the condition that .data_ready | |
2801 | * becomes 1. | |
ffbd51cc NTND |
2802 | * |
2803 | * The main thread steals half of the work from the worker that has | |
2804 | * most work left to hand it to the idle worker. | |
50f22ada JS |
2805 | */ |
2806 | ||
8ecce684 NP |
2807 | struct thread_params { |
2808 | pthread_t thread; | |
2809 | struct object_entry **list; | |
2810 | unsigned list_size; | |
384b32c0 | 2811 | unsigned remaining; |
8ecce684 NP |
2812 | int window; |
2813 | int depth; | |
50f22ada JS |
2814 | int working; |
2815 | int data_ready; | |
2816 | pthread_mutex_t mutex; | |
2817 | pthread_cond_t cond; | |
8ecce684 NP |
2818 | unsigned *processed; |
2819 | }; | |
2820 | ||
44626dc7 AH |
2821 | static pthread_cond_t progress_cond; |
2822 | ||
2823 | /* | |
2824 | * Mutex and conditional variable can't be statically-initialized on Windows. | |
2825 | */ | |
2826 | static void init_threaded_search(void) | |
2827 | { | |
44626dc7 AH |
2828 | pthread_mutex_init(&cache_mutex, NULL); |
2829 | pthread_mutex_init(&progress_mutex, NULL); | |
2830 | pthread_cond_init(&progress_cond, NULL); | |
2831 | } | |
2832 | ||
2833 | static void cleanup_threaded_search(void) | |
2834 | { | |
2835 | pthread_cond_destroy(&progress_cond); | |
44626dc7 AH |
2836 | pthread_mutex_destroy(&cache_mutex); |
2837 | pthread_mutex_destroy(&progress_mutex); | |
2838 | } | |
c2a33679 | 2839 | |
8ecce684 NP |
2840 | static void *threaded_find_deltas(void *arg) |
2841 | { | |
c2a33679 NP |
2842 | struct thread_params *me = arg; |
2843 | ||
0c2ad00b | 2844 | progress_lock(); |
50f22ada | 2845 | while (me->remaining) { |
0c2ad00b MÅ |
2846 | progress_unlock(); |
2847 | ||
384b32c0 | 2848 | find_deltas(me->list, &me->remaining, |
c2a33679 | 2849 | me->window, me->depth, me->processed); |
50f22ada JS |
2850 | |
2851 | progress_lock(); | |
2852 | me->working = 0; | |
2853 | pthread_cond_signal(&progress_cond); | |
2854 | progress_unlock(); | |
2855 | ||
2856 | /* | |
2857 | * We must not set ->data_ready before we wait on the | |
2858 | * condition because the main thread may have set it to 1 | |
2859 | * before we get here. In order to be sure that new | |
2860 | * work is available if we see 1 in ->data_ready, it | |
2861 | * was initialized to 0 before this thread was spawned | |
2862 | * and we reset it to 0 right away. | |
2863 | */ | |
2864 | pthread_mutex_lock(&me->mutex); | |
2865 | while (!me->data_ready) | |
2866 | pthread_cond_wait(&me->cond, &me->mutex); | |
2867 | me->data_ready = 0; | |
2868 | pthread_mutex_unlock(&me->mutex); | |
0c2ad00b MÅ |
2869 | |
2870 | progress_lock(); | |
c2a33679 | 2871 | } |
0c2ad00b | 2872 | progress_unlock(); |
50f22ada JS |
2873 | /* leave ->working 1 so that this doesn't get more work assigned */ |
2874 | return NULL; | |
8ecce684 NP |
2875 | } |
2876 | ||
8ecce684 NP |
2877 | static void ll_find_deltas(struct object_entry **list, unsigned list_size, |
2878 | int window, int depth, unsigned *processed) | |
2879 | { | |
dcda3614 | 2880 | struct thread_params *p; |
384b32c0 | 2881 | int i, ret, active_threads = 0; |
c2a33679 | 2882 | |
44626dc7 AH |
2883 | init_threaded_search(); |
2884 | ||
367f4a43 | 2885 | if (delta_search_threads <= 1) { |
384b32c0 | 2886 | find_deltas(list, &list_size, window, depth, processed); |
44626dc7 | 2887 | cleanup_threaded_search(); |
367f4a43 NP |
2888 | return; |
2889 | } | |
43cc2b42 | 2890 | if (progress > pack_to_stdout) |
f616db6a | 2891 | fprintf_ln(stderr, _("Delta compression using up to %d threads"), |
1a07e59c | 2892 | delta_search_threads); |
ca56dadb | 2893 | CALLOC_ARRAY(p, delta_search_threads); |
367f4a43 | 2894 | |
50f22ada | 2895 | /* Partition the work amongst work threads. */ |
367f4a43 | 2896 | for (i = 0; i < delta_search_threads; i++) { |
50f22ada JS |
2897 | unsigned sub_size = list_size / (delta_search_threads - i); |
2898 | ||
bf874896 NP |
2899 | /* don't use too small segments or no deltas will be found */ |
2900 | if (sub_size < 2*window && i+1 < delta_search_threads) | |
2901 | sub_size = 0; | |
2902 | ||
8ecce684 NP |
2903 | p[i].window = window; |
2904 | p[i].depth = depth; | |
2905 | p[i].processed = processed; | |
50f22ada JS |
2906 | p[i].working = 1; |
2907 | p[i].data_ready = 0; | |
c2a33679 | 2908 | |
59921b4b | 2909 | /* try to split chunks on "path" boundaries */ |
6fc74703 NP |
2910 | while (sub_size && sub_size < list_size && |
2911 | list[sub_size]->hash && | |
384b32c0 NP |
2912 | list[sub_size]->hash == list[sub_size-1]->hash) |
2913 | sub_size++; | |
2914 | ||
50f22ada JS |
2915 | p[i].list = list; |
2916 | p[i].list_size = sub_size; | |
2917 | p[i].remaining = sub_size; | |
59921b4b | 2918 | |
384b32c0 NP |
2919 | list += sub_size; |
2920 | list_size -= sub_size; | |
2921 | } | |
2922 | ||
50f22ada JS |
2923 | /* Start work threads. */ |
2924 | for (i = 0; i < delta_search_threads; i++) { | |
2925 | if (!p[i].list_size) | |
2926 | continue; | |
68e6a4f8 JS |
2927 | pthread_mutex_init(&p[i].mutex, NULL); |
2928 | pthread_cond_init(&p[i].cond, NULL); | |
50f22ada JS |
2929 | ret = pthread_create(&p[i].thread, NULL, |
2930 | threaded_find_deltas, &p[i]); | |
2931 | if (ret) | |
f616db6a | 2932 | die(_("unable to create thread: %s"), strerror(ret)); |
50f22ada JS |
2933 | active_threads++; |
2934 | } | |
2935 | ||
384b32c0 NP |
2936 | /* |
2937 | * Now let's wait for work completion. Each time a thread is done | |
2938 | * with its work, we steal half of the remaining work from the | |
2939 | * thread with the largest number of unprocessed objects and give | |
2940 | * it to that newly idle thread. This ensure good load balancing | |
2941 | * until the remaining object list segments are simply too short | |
2942 | * to be worth splitting anymore. | |
2943 | */ | |
50f22ada JS |
2944 | while (active_threads) { |
2945 | struct thread_params *target = NULL; | |
384b32c0 NP |
2946 | struct thread_params *victim = NULL; |
2947 | unsigned sub_size = 0; | |
384b32c0 NP |
2948 | |
2949 | progress_lock(); | |
50f22ada JS |
2950 | for (;;) { |
2951 | for (i = 0; !target && i < delta_search_threads; i++) | |
2952 | if (!p[i].working) | |
2953 | target = &p[i]; | |
2954 | if (target) | |
2955 | break; | |
2956 | pthread_cond_wait(&progress_cond, &progress_mutex); | |
2957 | } | |
2958 | ||
384b32c0 NP |
2959 | for (i = 0; i < delta_search_threads; i++) |
2960 | if (p[i].remaining > 2*window && | |
2961 | (!victim || victim->remaining < p[i].remaining)) | |
2962 | victim = &p[i]; | |
2963 | if (victim) { | |
2964 | sub_size = victim->remaining / 2; | |
2965 | list = victim->list + victim->list_size - sub_size; | |
2966 | while (sub_size && list[0]->hash && | |
2967 | list[0]->hash == list[-1]->hash) { | |
2968 | list++; | |
2969 | sub_size--; | |
2970 | } | |
eb9688ff NP |
2971 | if (!sub_size) { |
2972 | /* | |
2973 | * It is possible for some "paths" to have | |
2974 | * so many objects that no hash boundary | |
2975 | * might be found. Let's just steal the | |
2976 | * exact half in that case. | |
2977 | */ | |
2978 | sub_size = victim->remaining / 2; | |
2979 | list -= sub_size; | |
2980 | } | |
384b32c0 NP |
2981 | target->list = list; |
2982 | victim->list_size -= sub_size; | |
2983 | victim->remaining -= sub_size; | |
2984 | } | |
384b32c0 NP |
2985 | target->list_size = sub_size; |
2986 | target->remaining = sub_size; | |
50f22ada JS |
2987 | target->working = 1; |
2988 | progress_unlock(); | |
2989 | ||
2990 | pthread_mutex_lock(&target->mutex); | |
2991 | target->data_ready = 1; | |
2992 | pthread_cond_signal(&target->cond); | |
2993 | pthread_mutex_unlock(&target->mutex); | |
c2a33679 | 2994 | |
384b32c0 | 2995 | if (!sub_size) { |
b81d9af7 | 2996 | pthread_join(target->thread, NULL); |
50f22ada JS |
2997 | pthread_cond_destroy(&target->cond); |
2998 | pthread_mutex_destroy(&target->mutex); | |
384b32c0 | 2999 | active_threads--; |
c2a33679 | 3000 | } |
50f22ada | 3001 | } |
44626dc7 | 3002 | cleanup_threaded_search(); |
dcda3614 | 3003 | free(p); |
8ecce684 NP |
3004 | } |
3005 | ||
ff483026 JK |
3006 | static int obj_is_packed(const struct object_id *oid) |
3007 | { | |
a14aebea | 3008 | return packlist_find(&to_pack, oid) || |
92fb0db9 JK |
3009 | (reuse_packfile_bitmap && |
3010 | bitmap_walk_contains(bitmap_git, reuse_packfile_bitmap, oid)); | |
ff483026 JK |
3011 | } |
3012 | ||
b773ddea JK |
3013 | static void add_tag_chain(const struct object_id *oid) |
3014 | { | |
3015 | struct tag *tag; | |
3016 | ||
3017 | /* | |
3018 | * We catch duplicates already in add_object_entry(), but we'd | |
3019 | * prefer to do this extra check to avoid having to parse the | |
3020 | * tag at all if we already know that it's being packed (e.g., if | |
3021 | * it was included via bitmaps, we would not have parsed it | |
3022 | * previously). | |
3023 | */ | |
ff483026 | 3024 | if (obj_is_packed(oid)) |
b773ddea JK |
3025 | return; |
3026 | ||
ce71efb7 | 3027 | tag = lookup_tag(the_repository, oid); |
b773ddea JK |
3028 | while (1) { |
3029 | if (!tag || parse_tag(tag) || !tag->tagged) | |
f616db6a | 3030 | die(_("unable to pack objects reachable from tag %s"), |
b773ddea JK |
3031 | oid_to_hex(oid)); |
3032 | ||
188960b4 | 3033 | add_object_entry(&tag->object.oid, OBJ_TAG, NULL, 0); |
b773ddea JK |
3034 | |
3035 | if (tag->tagged->type != OBJ_TAG) | |
3036 | return; | |
3037 | ||
3038 | tag = (struct tag *)tag->tagged; | |
3039 | } | |
3040 | } | |
3041 | ||
5cf88fd8 ÆAB |
3042 | static int add_ref_tag(const char *tag UNUSED, const struct object_id *oid, |
3043 | int flag UNUSED, void *cb_data UNUSED) | |
f0a24aa5 | 3044 | { |
d155254c | 3045 | struct object_id peeled; |
f0a24aa5 | 3046 | |
77db59c2 | 3047 | if (!peel_iterated_oid(oid, &peeled) && obj_is_packed(&peeled)) |
b773ddea | 3048 | add_tag_chain(oid); |
f0a24aa5 SP |
3049 | return 0; |
3050 | } | |
3051 | ||
f3123c4a JH |
3052 | static void prepare_pack(int window, int depth) |
3053 | { | |
9668cf59 | 3054 | struct object_entry **delta_list; |
6e1c2344 RJ |
3055 | uint32_t i, nr_deltas; |
3056 | unsigned n; | |
9668cf59 | 3057 | |
28b8a730 | 3058 | if (use_delta_islands) |
385cb64f | 3059 | resolve_tree_islands(the_repository, progress, &to_pack); |
28b8a730 | 3060 | |
3f9ac8d2 | 3061 | get_object_details(); |
9668cf59 | 3062 | |
0e8189e2 NP |
3063 | /* |
3064 | * If we're locally repacking then we need to be doubly careful | |
3065 | * from now on in order to make sure no stealth corruption gets | |
3066 | * propagated to the new pack. Clients receiving streamed packs | |
3067 | * should validate everything they get anyway so no need to incur | |
3068 | * the additional cost here in that case. | |
3069 | */ | |
3070 | if (!pack_to_stdout) | |
3071 | do_check_packed_object_crc = 1; | |
3072 | ||
2834bc27 | 3073 | if (!to_pack.nr_objects || !window || !depth) |
9668cf59 NP |
3074 | return; |
3075 | ||
b32fa95f | 3076 | ALLOC_ARRAY(delta_list, to_pack.nr_objects); |
75d39853 NP |
3077 | nr_deltas = n = 0; |
3078 | ||
2834bc27 VM |
3079 | for (i = 0; i < to_pack.nr_objects; i++) { |
3080 | struct object_entry *entry = to_pack.objects + i; | |
75d39853 | 3081 | |
898eba5e | 3082 | if (DELTA(entry)) |
75d39853 | 3083 | /* This happens if we decided to reuse existing |
a7de7130 | 3084 | * delta from a pack. "reuse_delta &&" is implied. |
75d39853 NP |
3085 | */ |
3086 | continue; | |
3087 | ||
ac77d0c3 NTND |
3088 | if (!entry->type_valid || |
3089 | oe_size_less_than(&to_pack, entry, 50)) | |
75d39853 NP |
3090 | continue; |
3091 | ||
3092 | if (entry->no_try_delta) | |
3093 | continue; | |
3094 | ||
6d6f9cdd | 3095 | if (!entry->preferred_base) { |
75d39853 | 3096 | nr_deltas++; |
fd9b1bae | 3097 | if (oe_type(entry) < 0) |
f616db6a | 3098 | die(_("unable to get type of object %s"), |
e6a492b7 | 3099 | oid_to_hex(&entry->idx.oid)); |
eede9f42 | 3100 | } else { |
fd9b1bae | 3101 | if (oe_type(entry) < 0) { |
eede9f42 NP |
3102 | /* |
3103 | * This object is not found, but we | |
3104 | * don't have to include it anyway. | |
3105 | */ | |
3106 | continue; | |
3107 | } | |
6d6f9cdd | 3108 | } |
75d39853 NP |
3109 | |
3110 | delta_list[n++] = entry; | |
3111 | } | |
3112 | ||
2f8b8947 | 3113 | if (nr_deltas && n > 1) { |
e334977d | 3114 | unsigned nr_done = 0; |
bb514de3 | 3115 | |
e334977d | 3116 | if (progress) |
754dbc43 | 3117 | progress_state = start_progress(_("Compressing objects"), |
dc6a0757 | 3118 | nr_deltas); |
9ed0d8d6 | 3119 | QSORT(delta_list, n, type_size_sort); |
8ecce684 | 3120 | ll_find_deltas(delta_list, n, window+1, depth, &nr_done); |
4d4fcc54 | 3121 | stop_progress(&progress_state); |
e334977d | 3122 | if (nr_done != nr_deltas) |
f616db6a | 3123 | die(_("inconsistency with delta count")); |
75d39853 | 3124 | } |
9668cf59 | 3125 | free(delta_list); |
f3123c4a JH |
3126 | } |
3127 | ||
ef90d6d4 | 3128 | static int git_pack_config(const char *k, const char *v, void *cb) |
4812a93a | 3129 | { |
eeefa7c9 | 3130 | if (!strcmp(k, "pack.window")) { |
4812a93a JK |
3131 | window = git_config_int(k, v); |
3132 | return 0; | |
3133 | } | |
a97773ce BD |
3134 | if (!strcmp(k, "pack.windowmemory")) { |
3135 | window_memory_limit = git_config_ulong(k, v); | |
3136 | return 0; | |
3137 | } | |
3138 | if (!strcmp(k, "pack.depth")) { | |
842aaf93 TT |
3139 | depth = git_config_int(k, v); |
3140 | return 0; | |
3141 | } | |
074b2eea MK |
3142 | if (!strcmp(k, "pack.deltacachesize")) { |
3143 | max_delta_cache_size = git_config_int(k, v); | |
3144 | return 0; | |
3145 | } | |
e3dfddb3 MK |
3146 | if (!strcmp(k, "pack.deltacachelimit")) { |
3147 | cache_max_small_delta_size = git_config_int(k, v); | |
3148 | return 0; | |
3149 | } | |
ae4f07fb VM |
3150 | if (!strcmp(k, "pack.writebitmaphashcache")) { |
3151 | if (git_config_bool(k, v)) | |
3152 | write_bitmap_options |= BITMAP_OPT_HASH_CACHE; | |
3153 | else | |
3154 | write_bitmap_options &= ~BITMAP_OPT_HASH_CACHE; | |
3155 | } | |
76f14b77 AC |
3156 | |
3157 | if (!strcmp(k, "pack.writebitmaplookuptable")) { | |
3158 | if (git_config_bool(k, v)) | |
3159 | write_bitmap_options |= BITMAP_OPT_LOOKUP_TABLE; | |
3160 | else | |
3161 | write_bitmap_options &= ~BITMAP_OPT_LOOKUP_TABLE; | |
3162 | } | |
3163 | ||
6b8fda2d | 3164 | if (!strcmp(k, "pack.usebitmaps")) { |
645c432d | 3165 | use_bitmap_index_default = git_config_bool(k, v); |
6b8fda2d VM |
3166 | return 0; |
3167 | } | |
e704fc79 JK |
3168 | if (!strcmp(k, "pack.allowpackreuse")) { |
3169 | allow_pack_reuse = git_config_bool(k, v); | |
3170 | return 0; | |
3171 | } | |
693b86ff NP |
3172 | if (!strcmp(k, "pack.threads")) { |
3173 | delta_search_threads = git_config_int(k, v); | |
833e3df1 | 3174 | if (delta_search_threads < 0) |
f616db6a | 3175 | die(_("invalid number of threads specified (%d)"), |
693b86ff | 3176 | delta_search_threads); |
9c897c5c | 3177 | if (!HAVE_THREADS && delta_search_threads != 1) { |
f616db6a | 3178 | warning(_("no threads support, ignoring %s"), k); |
2e96d815 ÆAB |
3179 | delta_search_threads = 0; |
3180 | } | |
693b86ff NP |
3181 | return 0; |
3182 | } | |
4d00bda2 | 3183 | if (!strcmp(k, "pack.indexversion")) { |
ebcfb379 JH |
3184 | pack_idx_opts.version = git_config_int(k, v); |
3185 | if (pack_idx_opts.version > 2) | |
b4eda05d | 3186 | die(_("bad pack.indexVersion=%"PRIu32), |
ebcfb379 | 3187 | pack_idx_opts.version); |
4d00bda2 NP |
3188 | return 0; |
3189 | } | |
c9773343 TB |
3190 | if (!strcmp(k, "pack.writereverseindex")) { |
3191 | if (git_config_bool(k, v)) | |
3192 | pack_idx_opts.flags |= WRITE_REV; | |
3193 | else | |
3194 | pack_idx_opts.flags &= ~WRITE_REV; | |
3195 | return 0; | |
3196 | } | |
dd4b732d JT |
3197 | if (!strcmp(k, "uploadpack.blobpackfileuri")) { |
3198 | struct configured_exclusion *ex = xmalloc(sizeof(*ex)); | |
3199 | const char *oid_end, *pack_end; | |
3200 | /* | |
3201 | * Stores the pack hash. This is not a true object ID, but is | |
3202 | * of the same form. | |
3203 | */ | |
3204 | struct object_id pack_hash; | |
3205 | ||
3206 | if (parse_oid_hex(v, &ex->e.oid, &oid_end) || | |
3207 | *oid_end != ' ' || | |
3208 | parse_oid_hex(oid_end + 1, &pack_hash, &pack_end) || | |
3209 | *pack_end != ' ') | |
3210 | die(_("value of uploadpack.blobpackfileuri must be " | |
3211 | "of the form '<object-hash> <pack-hash> <uri>' (got '%s')"), v); | |
3212 | if (oidmap_get(&configured_exclusions, &ex->e.oid)) | |
3213 | die(_("object already configured in another " | |
3214 | "uploadpack.blobpackfileuri (got '%s')"), v); | |
3215 | ex->pack_hash_hex = xcalloc(1, pack_end - oid_end); | |
3216 | memcpy(ex->pack_hash_hex, oid_end + 1, pack_end - oid_end - 1); | |
3217 | ex->uri = xstrdup(pack_end + 1); | |
3218 | oidmap_put(&configured_exclusions, ex); | |
3219 | } | |
ef90d6d4 | 3220 | return git_default_config(k, v, cb); |
4812a93a JK |
3221 | } |
3222 | ||
339bce27 TB |
3223 | /* Counters for trace2 output when in --stdin-packs mode. */ |
3224 | static int stdin_packs_found_nr; | |
3225 | static int stdin_packs_hints_nr; | |
3226 | ||
3227 | static int add_object_entry_from_pack(const struct object_id *oid, | |
3228 | struct packed_git *p, | |
3229 | uint32_t pos, | |
3230 | void *_data) | |
3231 | { | |
339bce27 | 3232 | off_t ofs; |
5045759d | 3233 | enum object_type type = OBJ_NONE; |
339bce27 TB |
3234 | |
3235 | display_progress(progress_state, ++nr_seen); | |
3236 | ||
3237 | if (have_duplicate_entry(oid, 0)) | |
3238 | return 0; | |
3239 | ||
3240 | ofs = nth_packed_object_offset(p, pos); | |
3241 | if (!want_object_in_pack(oid, 0, &p, &ofs)) | |
3242 | return 0; | |
3243 | ||
5045759d TB |
3244 | if (p) { |
3245 | struct rev_info *revs = _data; | |
3246 | struct object_info oi = OBJECT_INFO_INIT; | |
339bce27 | 3247 | |
5045759d TB |
3248 | oi.typep = &type; |
3249 | if (packed_object_info(the_repository, p, ofs, &oi) < 0) { | |
3250 | die(_("could not get type of object %s in pack %s"), | |
3251 | oid_to_hex(oid), p->pack_name); | |
3252 | } else if (type == OBJ_COMMIT) { | |
3253 | /* | |
3254 | * commits in included packs are used as starting points for the | |
3255 | * subsequent revision walk | |
3256 | */ | |
3257 | add_pending_oid(revs, NULL, oid, 0); | |
3258 | } | |
339bce27 | 3259 | |
5045759d TB |
3260 | stdin_packs_found_nr++; |
3261 | } | |
339bce27 TB |
3262 | |
3263 | create_object_entry(oid, type, 0, 0, 0, p, ofs); | |
3264 | ||
3265 | return 0; | |
3266 | } | |
3267 | ||
be252d33 JK |
3268 | static void show_commit_pack_hint(struct commit *commit UNUSED, |
3269 | void *data UNUSED) | |
339bce27 TB |
3270 | { |
3271 | /* nothing to do; commits don't have a namehash */ | |
3272 | } | |
3273 | ||
3274 | static void show_object_pack_hint(struct object *object, const char *name, | |
be252d33 | 3275 | void *data UNUSED) |
339bce27 TB |
3276 | { |
3277 | struct object_entry *oe = packlist_find(&to_pack, &object->oid); | |
3278 | if (!oe) | |
3279 | return; | |
3280 | ||
3281 | /* | |
3282 | * Our 'to_pack' list was constructed by iterating all objects packed in | |
3283 | * included packs, and so doesn't have a non-zero hash field that you | |
3284 | * would typically pick up during a reachability traversal. | |
3285 | * | |
3286 | * Make a best-effort attempt to fill in the ->hash and ->no_try_delta | |
3287 | * here using a now in order to perhaps improve the delta selection | |
3288 | * process. | |
3289 | */ | |
3290 | oe->hash = pack_name_hash(name); | |
3291 | oe->no_try_delta = name && no_try_delta(name); | |
3292 | ||
3293 | stdin_packs_hints_nr++; | |
3294 | } | |
3295 | ||
3296 | static int pack_mtime_cmp(const void *_a, const void *_b) | |
3297 | { | |
3298 | struct packed_git *a = ((const struct string_list_item*)_a)->util; | |
3299 | struct packed_git *b = ((const struct string_list_item*)_b)->util; | |
3300 | ||
3301 | /* | |
3302 | * order packs by descending mtime so that objects are laid out | |
3303 | * roughly as newest-to-oldest | |
3304 | */ | |
3305 | if (a->mtime < b->mtime) | |
3306 | return 1; | |
3307 | else if (b->mtime < a->mtime) | |
3308 | return -1; | |
3309 | else | |
3310 | return 0; | |
3311 | } | |
3312 | ||
3313 | static void read_packs_list_from_stdin(void) | |
3314 | { | |
3315 | struct strbuf buf = STRBUF_INIT; | |
3316 | struct string_list include_packs = STRING_LIST_INIT_DUP; | |
3317 | struct string_list exclude_packs = STRING_LIST_INIT_DUP; | |
3318 | struct string_list_item *item = NULL; | |
3319 | ||
3320 | struct packed_git *p; | |
3321 | struct rev_info revs; | |
3322 | ||
3323 | repo_init_revisions(the_repository, &revs, NULL); | |
3324 | /* | |
3325 | * Use a revision walk to fill in the namehash of objects in the include | |
3326 | * packs. To save time, we'll avoid traversing through objects that are | |
3327 | * in excluded packs. | |
3328 | * | |
3329 | * That may cause us to avoid populating all of the namehash fields of | |
3330 | * all included objects, but our goal is best-effort, since this is only | |
3331 | * an optimization during delta selection. | |
3332 | */ | |
3333 | revs.no_kept_objects = 1; | |
3334 | revs.keep_pack_cache_flags |= IN_CORE_KEEP_PACKS; | |
3335 | revs.blob_objects = 1; | |
3336 | revs.tree_objects = 1; | |
3337 | revs.tag_objects = 1; | |
14e7b834 | 3338 | revs.ignore_missing_links = 1; |
339bce27 TB |
3339 | |
3340 | while (strbuf_getline(&buf, stdin) != EOF) { | |
3341 | if (!buf.len) | |
3342 | continue; | |
3343 | ||
3344 | if (*buf.buf == '^') | |
3345 | string_list_append(&exclude_packs, buf.buf + 1); | |
3346 | else | |
3347 | string_list_append(&include_packs, buf.buf); | |
3348 | ||
3349 | strbuf_reset(&buf); | |
3350 | } | |
3351 | ||
3352 | string_list_sort(&include_packs); | |
3353 | string_list_sort(&exclude_packs); | |
3354 | ||
3355 | for (p = get_all_packs(the_repository); p; p = p->next) { | |
3356 | const char *pack_name = pack_basename(p); | |
3357 | ||
3358 | item = string_list_lookup(&include_packs, pack_name); | |
3359 | if (!item) | |
3360 | item = string_list_lookup(&exclude_packs, pack_name); | |
3361 | ||
3362 | if (item) | |
3363 | item->util = p; | |
3364 | } | |
3365 | ||
3366 | /* | |
561fa035 ÆAB |
3367 | * Arguments we got on stdin may not even be packs. First |
3368 | * check that to avoid segfaulting later on in | |
3369 | * e.g. pack_mtime_cmp(), excluded packs are handled below. | |
3370 | * | |
3371 | * Since we first parsed our STDIN and then sorted the input | |
3372 | * lines the pack we error on will be whatever line happens to | |
3373 | * sort first. This is lazy, it's enough that we report one | |
3374 | * bad case here, we don't need to report the first/last one, | |
3375 | * or all of them. | |
3376 | */ | |
3377 | for_each_string_list_item(item, &include_packs) { | |
3378 | struct packed_git *p = item->util; | |
3379 | if (!p) | |
3380 | die(_("could not find pack '%s'"), item->string); | |
5045759d TB |
3381 | if (!is_pack_valid(p)) |
3382 | die(_("packfile %s cannot be accessed"), p->pack_name); | |
561fa035 ÆAB |
3383 | } |
3384 | ||
3385 | /* | |
3386 | * Then, handle all of the excluded packs, marking them as | |
3387 | * kept in-core so that later calls to add_object_entry() | |
3388 | * discards any objects that are also found in excluded packs. | |
339bce27 TB |
3389 | */ |
3390 | for_each_string_list_item(item, &exclude_packs) { | |
3391 | struct packed_git *p = item->util; | |
3392 | if (!p) | |
3393 | die(_("could not find pack '%s'"), item->string); | |
3394 | p->pack_keep_in_core = 1; | |
3395 | } | |
3396 | ||
3397 | /* | |
3398 | * Order packs by ascending mtime; use QSORT directly to access the | |
3399 | * string_list_item's ->util pointer, which string_list_sort() does not | |
3400 | * provide. | |
3401 | */ | |
3402 | QSORT(include_packs.items, include_packs.nr, pack_mtime_cmp); | |
3403 | ||
3404 | for_each_string_list_item(item, &include_packs) { | |
3405 | struct packed_git *p = item->util; | |
339bce27 TB |
3406 | for_each_object_in_pack(p, |
3407 | add_object_entry_from_pack, | |
3408 | &revs, | |
3409 | FOR_EACH_OBJECT_PACK_ORDER); | |
3410 | } | |
3411 | ||
3412 | if (prepare_revision_walk(&revs)) | |
3413 | die(_("revision walk setup failed")); | |
3414 | traverse_commit_list(&revs, | |
3415 | show_commit_pack_hint, | |
3416 | show_object_pack_hint, | |
3417 | NULL); | |
3418 | ||
3419 | trace2_data_intmax("pack-objects", the_repository, "stdin_packs_found", | |
3420 | stdin_packs_found_nr); | |
3421 | trace2_data_intmax("pack-objects", the_repository, "stdin_packs_hints", | |
3422 | stdin_packs_hints_nr); | |
3423 | ||
3424 | strbuf_release(&buf); | |
3425 | string_list_clear(&include_packs, 0); | |
3426 | string_list_clear(&exclude_packs, 0); | |
3427 | } | |
3428 | ||
b7573536 TB |
3429 | static void add_cruft_object_entry(const struct object_id *oid, enum object_type type, |
3430 | struct packed_git *pack, off_t offset, | |
3431 | const char *name, uint32_t mtime) | |
3432 | { | |
3433 | struct object_entry *entry; | |
3434 | ||
3435 | display_progress(progress_state, ++nr_seen); | |
3436 | ||
3437 | entry = packlist_find(&to_pack, oid); | |
3438 | if (entry) { | |
3439 | if (name) { | |
3440 | entry->hash = pack_name_hash(name); | |
3441 | entry->no_try_delta = no_try_delta(name); | |
3442 | } | |
3443 | } else { | |
3444 | if (!want_object_in_pack(oid, 0, &pack, &offset)) | |
3445 | return; | |
3446 | if (!pack && type == OBJ_BLOB && !has_loose_object(oid)) { | |
3447 | /* | |
3448 | * If a traversed tree has a missing blob then we want | |
3449 | * to avoid adding that missing object to our pack. | |
3450 | * | |
3451 | * This only applies to missing blobs, not trees, | |
3452 | * because the traversal needs to parse sub-trees but | |
3453 | * not blobs. | |
3454 | * | |
3455 | * Note we only perform this check when we couldn't | |
3456 | * already find the object in a pack, so we're really | |
3457 | * limited to "ensure non-tip blobs which don't exist in | |
3458 | * packs do exist via loose objects". Confused? | |
3459 | */ | |
3460 | return; | |
3461 | } | |
3462 | ||
3463 | entry = create_object_entry(oid, type, pack_name_hash(name), | |
3464 | 0, name && no_try_delta(name), | |
3465 | pack, offset); | |
3466 | } | |
3467 | ||
3468 | if (mtime > oe_cruft_mtime(&to_pack, entry)) | |
3469 | oe_set_cruft_mtime(&to_pack, entry, mtime); | |
3470 | return; | |
3471 | } | |
3472 | ||
c50dca2a | 3473 | static void show_cruft_object(struct object *obj, const char *name, void *data UNUSED) |
a7d49383 TB |
3474 | { |
3475 | /* | |
3476 | * if we did not record it earlier, it's at least as old as our | |
3477 | * expiration value. Rather than find it exactly, just use that | |
3478 | * value. This may bump it forward from its real mtime, but it | |
3479 | * will still be "too old" next time we run with the same | |
3480 | * expiration. | |
3481 | * | |
3482 | * if obj does appear in the packing list, this call is a noop (or may | |
3483 | * set the namehash). | |
3484 | */ | |
3485 | add_cruft_object_entry(&obj->oid, obj->type, NULL, 0, name, cruft_expiration); | |
3486 | } | |
3487 | ||
3488 | static void show_cruft_commit(struct commit *commit, void *data) | |
3489 | { | |
3490 | show_cruft_object((struct object*)commit, NULL, data); | |
3491 | } | |
3492 | ||
c50dca2a | 3493 | static int cruft_include_check_obj(struct object *obj, void *data UNUSED) |
a7d49383 TB |
3494 | { |
3495 | return !has_object_kept_pack(&obj->oid, IN_CORE_KEEP_PACKS); | |
3496 | } | |
3497 | ||
3498 | static int cruft_include_check(struct commit *commit, void *data) | |
3499 | { | |
3500 | return cruft_include_check_obj((struct object*)commit, data); | |
3501 | } | |
3502 | ||
3503 | static void set_cruft_mtime(const struct object *object, | |
3504 | struct packed_git *pack, | |
3505 | off_t offset, time_t mtime) | |
3506 | { | |
3507 | add_cruft_object_entry(&object->oid, object->type, pack, offset, NULL, | |
3508 | mtime); | |
3509 | } | |
3510 | ||
b7573536 TB |
3511 | static void mark_pack_kept_in_core(struct string_list *packs, unsigned keep) |
3512 | { | |
3513 | struct string_list_item *item = NULL; | |
3514 | for_each_string_list_item(item, packs) { | |
3515 | struct packed_git *p = item->util; | |
3516 | if (!p) | |
3517 | die(_("could not find pack '%s'"), item->string); | |
3518 | p->pack_keep_in_core = keep; | |
3519 | } | |
3520 | } | |
3521 | ||
3522 | static void add_unreachable_loose_objects(void); | |
3523 | static void add_objects_in_unpacked_packs(void); | |
3524 | ||
3525 | static void enumerate_cruft_objects(void) | |
3526 | { | |
3527 | if (progress) | |
3528 | progress_state = start_progress(_("Enumerating cruft objects"), 0); | |
3529 | ||
3530 | add_objects_in_unpacked_packs(); | |
3531 | add_unreachable_loose_objects(); | |
3532 | ||
3533 | stop_progress(&progress_state); | |
3534 | } | |
3535 | ||
a7d49383 TB |
3536 | static void enumerate_and_traverse_cruft_objects(struct string_list *fresh_packs) |
3537 | { | |
3538 | struct packed_git *p; | |
3539 | struct rev_info revs; | |
3540 | int ret; | |
3541 | ||
3542 | repo_init_revisions(the_repository, &revs, NULL); | |
3543 | ||
3544 | revs.tag_objects = 1; | |
3545 | revs.tree_objects = 1; | |
3546 | revs.blob_objects = 1; | |
3547 | ||
3548 | revs.include_check = cruft_include_check; | |
3549 | revs.include_check_obj = cruft_include_check_obj; | |
3550 | ||
3551 | revs.ignore_missing_links = 1; | |
3552 | ||
3553 | if (progress) | |
3554 | progress_state = start_progress(_("Enumerating cruft objects"), 0); | |
3555 | ret = add_unseen_recent_objects_to_traversal(&revs, cruft_expiration, | |
3556 | set_cruft_mtime, 1); | |
3557 | stop_progress(&progress_state); | |
3558 | ||
3559 | if (ret) | |
3560 | die(_("unable to add cruft objects")); | |
3561 | ||
3562 | /* | |
3563 | * Re-mark only the fresh packs as kept so that objects in | |
3564 | * unknown packs do not halt the reachability traversal early. | |
3565 | */ | |
3566 | for (p = get_all_packs(the_repository); p; p = p->next) | |
3567 | p->pack_keep_in_core = 0; | |
3568 | mark_pack_kept_in_core(fresh_packs, 1); | |
3569 | ||
3570 | if (prepare_revision_walk(&revs)) | |
3571 | die(_("revision walk setup failed")); | |
3572 | if (progress) | |
3573 | progress_state = start_progress(_("Traversing cruft objects"), 0); | |
3574 | nr_seen = 0; | |
3575 | traverse_commit_list(&revs, show_cruft_commit, show_cruft_object, NULL); | |
3576 | ||
3577 | stop_progress(&progress_state); | |
3578 | } | |
3579 | ||
b7573536 TB |
3580 | static void read_cruft_objects(void) |
3581 | { | |
3582 | struct strbuf buf = STRBUF_INIT; | |
3583 | struct string_list discard_packs = STRING_LIST_INIT_DUP; | |
3584 | struct string_list fresh_packs = STRING_LIST_INIT_DUP; | |
3585 | struct packed_git *p; | |
3586 | ||
3587 | ignore_packed_keep_in_core = 1; | |
3588 | ||
3589 | while (strbuf_getline(&buf, stdin) != EOF) { | |
3590 | if (!buf.len) | |
3591 | continue; | |
3592 | ||
3593 | if (*buf.buf == '-') | |
3594 | string_list_append(&discard_packs, buf.buf + 1); | |
3595 | else | |
3596 | string_list_append(&fresh_packs, buf.buf); | |
3597 | strbuf_reset(&buf); | |
3598 | } | |
3599 | ||
3600 | string_list_sort(&discard_packs); | |
3601 | string_list_sort(&fresh_packs); | |
3602 | ||
3603 | for (p = get_all_packs(the_repository); p; p = p->next) { | |
3604 | const char *pack_name = pack_basename(p); | |
3605 | struct string_list_item *item; | |
3606 | ||
3607 | item = string_list_lookup(&fresh_packs, pack_name); | |
3608 | if (!item) | |
3609 | item = string_list_lookup(&discard_packs, pack_name); | |
3610 | ||
3611 | if (item) { | |
3612 | item->util = p; | |
3613 | } else { | |
3614 | /* | |
3615 | * This pack wasn't mentioned in either the "fresh" or | |
3616 | * "discard" list, so the caller didn't know about it. | |
3617 | * | |
3618 | * Mark it as kept so that its objects are ignored by | |
3619 | * add_unseen_recent_objects_to_traversal(). We'll | |
3620 | * unmark it before starting the traversal so it doesn't | |
3621 | * halt the traversal early. | |
3622 | */ | |
3623 | p->pack_keep_in_core = 1; | |
3624 | } | |
3625 | } | |
3626 | ||
3627 | mark_pack_kept_in_core(&fresh_packs, 1); | |
3628 | mark_pack_kept_in_core(&discard_packs, 0); | |
3629 | ||
3630 | if (cruft_expiration) | |
a7d49383 | 3631 | enumerate_and_traverse_cruft_objects(&fresh_packs); |
b7573536 TB |
3632 | else |
3633 | enumerate_cruft_objects(); | |
3634 | ||
3635 | strbuf_release(&buf); | |
3636 | string_list_clear(&discard_packs, 0); | |
3637 | string_list_clear(&fresh_packs, 0); | |
3638 | } | |
3639 | ||
b5d97e6b | 3640 | static void read_object_list_from_stdin(void) |
c323ac7d | 3641 | { |
188960b4 | 3642 | char line[GIT_MAX_HEXSZ + 1 + PATH_MAX + 2]; |
3643 | struct object_id oid; | |
3644 | const char *p; | |
b2504a0d | 3645 | |
da93d12b | 3646 | for (;;) { |
da93d12b LT |
3647 | if (!fgets(line, sizeof(line), stdin)) { |
3648 | if (feof(stdin)) | |
3649 | break; | |
3650 | if (!ferror(stdin)) | |
5867757d | 3651 | BUG("fgets returned NULL, not EOF, not error!"); |
687dd75c | 3652 | if (errno != EINTR) |
d824cbba | 3653 | die_errno("fgets"); |
687dd75c JH |
3654 | clearerr(stdin); |
3655 | continue; | |
da93d12b | 3656 | } |
7a979d99 | 3657 | if (line[0] == '-') { |
188960b4 | 3658 | if (get_oid_hex(line+1, &oid)) |
f616db6a | 3659 | die(_("expected edge object ID, got garbage:\n %s"), |
b5d97e6b | 3660 | line); |
188960b4 | 3661 | add_preferred_base(&oid); |
7a979d99 | 3662 | continue; |
21fcd1bd | 3663 | } |
188960b4 | 3664 | if (parse_oid_hex(line, &oid, &p)) |
f616db6a | 3665 | die(_("expected object ID, got garbage:\n %s"), line); |
b5d97e6b | 3666 | |
188960b4 | 3667 | add_preferred_base_object(p + 1); |
fd9b1bae | 3668 | add_object_entry(&oid, OBJ_NONE, p + 1, 0); |
c323ac7d | 3669 | } |
b5d97e6b JH |
3670 | } |
3671 | ||
c50dca2a | 3672 | static void show_commit(struct commit *commit, void *data UNUSED) |
b5d97e6b | 3673 | { |
188960b4 | 3674 | add_object_entry(&commit->object.oid, OBJ_COMMIT, NULL, 0); |
7cc8f971 VM |
3675 | |
3676 | if (write_bitmap_index) | |
3677 | index_commit_for_bitmap(commit); | |
28b8a730 JK |
3678 | |
3679 | if (use_delta_islands) | |
3680 | propagate_island_marks(commit); | |
b5d97e6b JH |
3681 | } |
3682 | ||
c50dca2a JK |
3683 | static void show_object(struct object *obj, const char *name, |
3684 | void *data UNUSED) | |
b5d97e6b | 3685 | { |
8d2dfc49 | 3686 | add_preferred_base_object(name); |
188960b4 | 3687 | add_object_entry(&obj->oid, obj->type, name, 0); |
28b8a730 JK |
3688 | |
3689 | if (use_delta_islands) { | |
3690 | const char *p; | |
39490536 | 3691 | unsigned depth; |
28b8a730 JK |
3692 | struct object_entry *ent; |
3693 | ||
39490536 JK |
3694 | /* the empty string is a root tree, which is depth 0 */ |
3695 | depth = *name ? 1 : 0; | |
28b8a730 JK |
3696 | for (p = strchr(name, '/'); p; p = strchr(p + 1, '/')) |
3697 | depth++; | |
3698 | ||
3a37876b | 3699 | ent = packlist_find(&to_pack, &obj->oid); |
108f5303 CC |
3700 | if (ent && depth > oe_tree_depth(&to_pack, ent)) |
3701 | oe_set_tree_depth(&to_pack, ent, depth); | |
28b8a730 | 3702 | } |
b5d97e6b JH |
3703 | } |
3704 | ||
9535ce73 JH |
3705 | static void show_object__ma_allow_any(struct object *obj, const char *name, void *data) |
3706 | { | |
3707 | assert(arg_missing_action == MA_ALLOW_ANY); | |
3708 | ||
3709 | /* | |
3710 | * Quietly ignore ALL missing objects. This avoids problems with | |
3711 | * staging them now and getting an odd error later. | |
3712 | */ | |
ee47243d | 3713 | if (!has_object(the_repository, &obj->oid, 0)) |
9535ce73 JH |
3714 | return; |
3715 | ||
3716 | show_object(obj, name, data); | |
3717 | } | |
3718 | ||
0c16cd49 JT |
3719 | static void show_object__ma_allow_promisor(struct object *obj, const char *name, void *data) |
3720 | { | |
3721 | assert(arg_missing_action == MA_ALLOW_PROMISOR); | |
3722 | ||
3723 | /* | |
3724 | * Quietly ignore EXPECTED missing objects. This avoids problems with | |
3725 | * staging them now and getting an odd error later. | |
3726 | */ | |
ee47243d | 3727 | if (!has_object(the_repository, &obj->oid, 0) && is_promisor_object(&obj->oid)) |
0c16cd49 JT |
3728 | return; |
3729 | ||
3730 | show_object(obj, name, data); | |
3731 | } | |
3732 | ||
9535ce73 JH |
3733 | static int option_parse_missing_action(const struct option *opt, |
3734 | const char *arg, int unset) | |
3735 | { | |
3736 | assert(arg); | |
3737 | assert(!unset); | |
3738 | ||
3739 | if (!strcmp(arg, "error")) { | |
3740 | arg_missing_action = MA_ERROR; | |
3741 | fn_show_object = show_object; | |
3742 | return 0; | |
3743 | } | |
3744 | ||
3745 | if (!strcmp(arg, "allow-any")) { | |
3746 | arg_missing_action = MA_ALLOW_ANY; | |
0c16cd49 | 3747 | fetch_if_missing = 0; |
9535ce73 JH |
3748 | fn_show_object = show_object__ma_allow_any; |
3749 | return 0; | |
3750 | } | |
3751 | ||
0c16cd49 JT |
3752 | if (!strcmp(arg, "allow-promisor")) { |
3753 | arg_missing_action = MA_ALLOW_PROMISOR; | |
3754 | fetch_if_missing = 0; | |
3755 | fn_show_object = show_object__ma_allow_promisor; | |
3756 | return 0; | |
3757 | } | |
3758 | ||
1a8aea85 | 3759 | die(_("invalid value for '%s': '%s'"), "--missing", arg); |
9535ce73 JH |
3760 | return 0; |
3761 | } | |
3762 | ||
8d1d8f83 JH |
3763 | static void show_edge(struct commit *commit) |
3764 | { | |
188960b4 | 3765 | add_preferred_base(&commit->object.oid); |
8d1d8f83 JH |
3766 | } |
3767 | ||
a9fd2f20 TB |
3768 | static int add_object_in_unpacked_pack(const struct object_id *oid, |
3769 | struct packed_git *pack, | |
3770 | uint32_t pos, | |
be252d33 | 3771 | void *data UNUSED) |
08cdfb13 | 3772 | { |
b7573536 TB |
3773 | if (cruft) { |
3774 | off_t offset; | |
3775 | time_t mtime; | |
3776 | ||
3777 | if (pack->is_cruft) { | |
3778 | if (load_pack_mtimes(pack) < 0) | |
3779 | die(_("could not load cruft pack .mtimes")); | |
3780 | mtime = nth_packed_mtime(pack, pos); | |
3781 | } else { | |
3782 | mtime = pack->mtime; | |
3783 | } | |
3784 | offset = nth_packed_object_offset(pack, pos); | |
3785 | ||
3786 | add_cruft_object_entry(oid, OBJ_NONE, pack, offset, | |
3787 | NULL, mtime); | |
3788 | } else { | |
3789 | add_object_entry(oid, OBJ_NONE, "", 0); | |
3790 | } | |
a9fd2f20 | 3791 | return 0; |
08cdfb13 JH |
3792 | } |
3793 | ||
7a2a7216 | 3794 | static void add_objects_in_unpacked_packs(void) |
08cdfb13 | 3795 | { |
a9fd2f20 TB |
3796 | if (for_each_packed_object(add_object_in_unpacked_pack, NULL, |
3797 | FOR_EACH_OBJECT_PACK_ORDER | | |
3798 | FOR_EACH_OBJECT_LOCAL_ONLY | | |
3799 | FOR_EACH_OBJECT_SKIP_IN_CORE_KEPT_PACKS | | |
3800 | FOR_EACH_OBJECT_SKIP_ON_DISK_KEPT_PACKS)) | |
3801 | die(_("cannot open pack index")); | |
08cdfb13 JH |
3802 | } |
3803 | ||
76c1d9a0 | 3804 | static int add_loose_object(const struct object_id *oid, const char *path, |
be252d33 | 3805 | void *data UNUSED) |
e26a8c47 | 3806 | { |
0df8e965 | 3807 | enum object_type type = oid_object_info(the_repository, oid, NULL); |
e26a8c47 JK |
3808 | |
3809 | if (type < 0) { | |
f616db6a | 3810 | warning(_("loose object at %s could not be examined"), path); |
e26a8c47 JK |
3811 | return 0; |
3812 | } | |
3813 | ||
b7573536 TB |
3814 | if (cruft) { |
3815 | struct stat st; | |
3816 | if (stat(path, &st) < 0) { | |
3817 | if (errno == ENOENT) | |
3818 | return 0; | |
3819 | return error_errno("unable to stat %s", oid_to_hex(oid)); | |
3820 | } | |
3821 | ||
3822 | add_cruft_object_entry(oid, type, NULL, 0, NULL, | |
3823 | st.st_mtime); | |
3824 | } else { | |
3825 | add_object_entry(oid, type, "", 0); | |
3826 | } | |
e26a8c47 JK |
3827 | return 0; |
3828 | } | |
3829 | ||
3830 | /* | |
3831 | * We actually don't even have to worry about reachability here. | |
3832 | * add_object_entry will weed out duplicates, so we just add every | |
3833 | * loose object we find. | |
3834 | */ | |
3835 | static void add_unreachable_loose_objects(void) | |
3836 | { | |
3837 | for_each_loose_file_in_objdir(get_object_directory(), | |
3838 | add_loose_object, | |
3839 | NULL, NULL, NULL); | |
3840 | } | |
3841 | ||
188960b4 | 3842 | static int has_sha1_pack_kept_or_nonlocal(const struct object_id *oid) |
094085e3 BC |
3843 | { |
3844 | static struct packed_git *last_found = (void *)1; | |
3845 | struct packed_git *p; | |
3846 | ||
a80d72db | 3847 | p = (last_found != (void *)1) ? last_found : |
454ea2e4 | 3848 | get_all_packs(the_repository); |
094085e3 BC |
3849 | |
3850 | while (p) { | |
ed7e5fc3 NTND |
3851 | if ((!p->pack_local || p->pack_keep || |
3852 | p->pack_keep_in_core) && | |
188960b4 | 3853 | find_pack_entry_one(oid->hash, p)) { |
094085e3 BC |
3854 | last_found = p; |
3855 | return 1; | |
3856 | } | |
3857 | if (p == last_found) | |
454ea2e4 | 3858 | p = get_all_packs(the_repository); |
094085e3 BC |
3859 | else |
3860 | p = p->next; | |
3861 | if (p == last_found) | |
3862 | p = p->next; | |
3863 | } | |
3864 | return 0; | |
3865 | } | |
3866 | ||
abcb8655 JK |
3867 | /* |
3868 | * Store a list of sha1s that are should not be discarded | |
3869 | * because they are either written too recently, or are | |
3870 | * reachable from another object that was. | |
3871 | * | |
3872 | * This is filled by get_object_list. | |
3873 | */ | |
910650d2 | 3874 | static struct oid_array recent_objects; |
abcb8655 | 3875 | |
4ce3621a | 3876 | static int loosened_object_can_be_discarded(const struct object_id *oid, |
dddbad72 | 3877 | timestamp_t mtime) |
d0d46abc JK |
3878 | { |
3879 | if (!unpack_unreachable_expiration) | |
3880 | return 0; | |
3881 | if (mtime > unpack_unreachable_expiration) | |
3882 | return 0; | |
910650d2 | 3883 | if (oid_array_lookup(&recent_objects, oid) >= 0) |
abcb8655 | 3884 | return 0; |
d0d46abc JK |
3885 | return 1; |
3886 | } | |
3887 | ||
7a2a7216 | 3888 | static void loosen_unused_packed_objects(void) |
ca11b212 NP |
3889 | { |
3890 | struct packed_git *p; | |
3891 | uint32_t i; | |
a643157d | 3892 | uint32_t loosened_objects_nr = 0; |
4ce3621a | 3893 | struct object_id oid; |
ca11b212 | 3894 | |
454ea2e4 | 3895 | for (p = get_all_packs(the_repository); p; p = p->next) { |
ed7e5fc3 | 3896 | if (!p->pack_local || p->pack_keep || p->pack_keep_in_core) |
ca11b212 NP |
3897 | continue; |
3898 | ||
3899 | if (open_pack_index(p)) | |
f616db6a | 3900 | die(_("cannot open pack index")); |
ca11b212 NP |
3901 | |
3902 | for (i = 0; i < p->num_objects; i++) { | |
0763671b | 3903 | nth_packed_object_id(&oid, p, i); |
3a37876b | 3904 | if (!packlist_find(&to_pack, &oid) && |
188960b4 | 3905 | !has_sha1_pack_kept_or_nonlocal(&oid) && |
a643157d | 3906 | !loosened_object_can_be_discarded(&oid, p->mtime)) { |
4bdb70a4 | 3907 | if (force_object_loose(&oid, p->mtime)) |
f616db6a | 3908 | die(_("unable to force loose object")); |
a643157d RS |
3909 | loosened_objects_nr++; |
3910 | } | |
ca11b212 NP |
3911 | } |
3912 | } | |
a643157d RS |
3913 | |
3914 | trace2_data_intmax("pack-objects", the_repository, | |
3915 | "loosen_unused_packed_objects/loosened", loosened_objects_nr); | |
ca11b212 NP |
3916 | } |
3917 | ||
69e4b342 | 3918 | /* |
645c432d KS |
3919 | * This tracks any options which pack-reuse code expects to be on, or which a |
3920 | * reader of the pack might not understand, and which would therefore prevent | |
3921 | * blind reuse of what we have on disk. | |
69e4b342 JK |
3922 | */ |
3923 | static int pack_options_allow_reuse(void) | |
3924 | { | |
e704fc79 JK |
3925 | return allow_pack_reuse && |
3926 | pack_to_stdout && | |
ed7e5fc3 NTND |
3927 | !ignore_packed_keep_on_disk && |
3928 | !ignore_packed_keep_in_core && | |
9df4a607 JK |
3929 | (!local || !have_non_local_packs) && |
3930 | !incremental; | |
69e4b342 JK |
3931 | } |
3932 | ||
6b8fda2d VM |
3933 | static int get_object_list_from_bitmap(struct rev_info *revs) |
3934 | { | |
09d4a79e | 3935 | if (!(bitmap_git = prepare_bitmap_walk(revs, 0))) |
6b8fda2d VM |
3936 | return -1; |
3937 | ||
69e4b342 JK |
3938 | if (pack_options_allow_reuse() && |
3939 | !reuse_partial_packfile_from_bitmap( | |
3ae5fa07 | 3940 | bitmap_git, |
6b8fda2d VM |
3941 | &reuse_packfile, |
3942 | &reuse_packfile_objects, | |
bb514de3 | 3943 | &reuse_packfile_bitmap)) { |
6b8fda2d VM |
3944 | assert(reuse_packfile_objects); |
3945 | nr_result += reuse_packfile_objects; | |
8e118e84 JK |
3946 | nr_seen += reuse_packfile_objects; |
3947 | display_progress(progress_state, nr_seen); | |
6b8fda2d VM |
3948 | } |
3949 | ||
4eb707eb JK |
3950 | traverse_bitmap_commit_list(bitmap_git, revs, |
3951 | &add_object_entry_from_bitmap); | |
6b8fda2d VM |
3952 | return 0; |
3953 | } | |
3954 | ||
abcb8655 | 3955 | static void record_recent_object(struct object *obj, |
be252d33 JK |
3956 | const char *name UNUSED, |
3957 | void *data UNUSED) | |
abcb8655 | 3958 | { |
910650d2 | 3959 | oid_array_append(&recent_objects, &obj->oid); |
abcb8655 JK |
3960 | } |
3961 | ||
be252d33 | 3962 | static void record_recent_commit(struct commit *commit, void *data UNUSED) |
abcb8655 | 3963 | { |
910650d2 | 3964 | oid_array_append(&recent_objects, &commit->object.oid); |
abcb8655 JK |
3965 | } |
3966 | ||
3f267a11 | 3967 | static int mark_bitmap_preferred_tip(const char *refname, |
63e14ee2 | 3968 | const struct object_id *oid, |
5cf88fd8 ÆAB |
3969 | int flags UNUSED, |
3970 | void *data UNUSED) | |
3f267a11 TB |
3971 | { |
3972 | struct object_id peeled; | |
3973 | struct object *object; | |
3974 | ||
3975 | if (!peel_iterated_oid(oid, &peeled)) | |
3976 | oid = &peeled; | |
3977 | ||
3978 | object = parse_object_or_die(oid, refname); | |
3979 | if (object->type == OBJ_COMMIT) | |
3980 | object->flags |= NEEDS_BITMAP; | |
3981 | ||
3982 | return 0; | |
3983 | } | |
3984 | ||
3985 | static void mark_bitmap_preferred_tips(void) | |
3986 | { | |
3987 | struct string_list_item *item; | |
3988 | const struct string_list *preferred_tips; | |
3989 | ||
3990 | preferred_tips = bitmap_preferred_tips(the_repository); | |
3991 | if (!preferred_tips) | |
3992 | return; | |
3993 | ||
3994 | for_each_string_list_item(item, preferred_tips) { | |
3995 | for_each_ref_in(item->string, mark_bitmap_preferred_tip, NULL); | |
3996 | } | |
3997 | } | |
3998 | ||
80f6de4f | 3999 | static void get_object_list(struct rev_info *revs, int ac, const char **av) |
b5d97e6b | 4000 | { |
bbcde41a MD |
4001 | struct setup_revision_opt s_r_opt = { |
4002 | .allow_exclude_promisor_objects = 1, | |
4003 | }; | |
b5d97e6b | 4004 | char line[1000]; |
b5d97e6b | 4005 | int flags = 0; |
a4544b31 | 4006 | int save_warning; |
b5d97e6b | 4007 | |
b5d97e6b | 4008 | save_commit_buffer = 0; |
80f6de4f | 4009 | setup_revisions(ac, av, revs, &s_r_opt); |
b5d97e6b | 4010 | |
b790e0f6 | 4011 | /* make sure shallows are read */ |
c8813487 | 4012 | is_repository_shallow(the_repository); |
b790e0f6 | 4013 | |
a4544b31 DS |
4014 | save_warning = warn_on_object_refname_ambiguity; |
4015 | warn_on_object_refname_ambiguity = 0; | |
4016 | ||
b5d97e6b JH |
4017 | while (fgets(line, sizeof(line), stdin) != NULL) { |
4018 | int len = strlen(line); | |
872c930d | 4019 | if (len && line[len - 1] == '\n') |
b5d97e6b JH |
4020 | line[--len] = 0; |
4021 | if (!len) | |
4022 | break; | |
4023 | if (*line == '-') { | |
4024 | if (!strcmp(line, "--not")) { | |
4025 | flags ^= UNINTERESTING; | |
7cc8f971 | 4026 | write_bitmap_index = 0; |
b5d97e6b JH |
4027 | continue; |
4028 | } | |
b790e0f6 | 4029 | if (starts_with(line, "--shallow ")) { |
e92b848c | 4030 | struct object_id oid; |
4031 | if (get_oid_hex(line + 10, &oid)) | |
4279000d | 4032 | die("not an object name '%s'", line + 10); |
19143f13 | 4033 | register_shallow(the_repository, &oid); |
f7f91086 | 4034 | use_bitmap_index = 0; |
b790e0f6 NTND |
4035 | continue; |
4036 | } | |
f616db6a | 4037 | die(_("not a rev '%s'"), line); |
b5d97e6b | 4038 | } |
80f6de4f | 4039 | if (handle_revision_arg(line, revs, flags, REVARG_CANNOT_BE_FILENAME)) |
f616db6a | 4040 | die(_("bad revision '%s'"), line); |
b5d97e6b JH |
4041 | } |
4042 | ||
a4544b31 DS |
4043 | warn_on_object_refname_ambiguity = save_warning; |
4044 | ||
80f6de4f | 4045 | if (use_bitmap_index && !get_object_list_from_bitmap(revs)) |
6b8fda2d VM |
4046 | return; |
4047 | ||
28b8a730 | 4048 | if (use_delta_islands) |
bdbdf42f | 4049 | load_delta_islands(the_repository, progress); |
28b8a730 | 4050 | |
3f267a11 TB |
4051 | if (write_bitmap_index) |
4052 | mark_bitmap_preferred_tips(); | |
4053 | ||
80f6de4f | 4054 | if (prepare_revision_walk(revs)) |
f616db6a | 4055 | die(_("revision walk setup failed")); |
80f6de4f | 4056 | mark_edges_uninteresting(revs, show_edge, sparse); |
9535ce73 JH |
4057 | |
4058 | if (!fn_show_object) | |
4059 | fn_show_object = show_object; | |
80f6de4f | 4060 | traverse_commit_list(revs, |
3e0370a8 DS |
4061 | show_commit, fn_show_object, |
4062 | NULL); | |
08cdfb13 | 4063 | |
abcb8655 | 4064 | if (unpack_unreachable_expiration) { |
80f6de4f DS |
4065 | revs->ignore_missing_links = 1; |
4066 | if (add_unseen_recent_objects_to_traversal(revs, | |
2fb90409 | 4067 | unpack_unreachable_expiration, NULL, 0)) |
f616db6a | 4068 | die(_("unable to add recent objects")); |
80f6de4f | 4069 | if (prepare_revision_walk(revs)) |
f616db6a | 4070 | die(_("revision walk setup failed")); |
80f6de4f | 4071 | traverse_commit_list(revs, record_recent_commit, |
abcb8655 JK |
4072 | record_recent_object, NULL); |
4073 | } | |
4074 | ||
08cdfb13 | 4075 | if (keep_unreachable) |
7a2a7216 | 4076 | add_objects_in_unpacked_packs(); |
e26a8c47 JK |
4077 | if (pack_loose_unreachable) |
4078 | add_unreachable_loose_objects(); | |
ca11b212 | 4079 | if (unpack_unreachable) |
7a2a7216 | 4080 | loosen_unused_packed_objects(); |
abcb8655 | 4081 | |
910650d2 | 4082 | oid_array_clear(&recent_objects); |
b5d97e6b JH |
4083 | } |
4084 | ||
ed7e5fc3 NTND |
4085 | static void add_extra_kept_packs(const struct string_list *names) |
4086 | { | |
4087 | struct packed_git *p; | |
4088 | ||
4089 | if (!names->nr) | |
4090 | return; | |
4091 | ||
454ea2e4 | 4092 | for (p = get_all_packs(the_repository); p; p = p->next) { |
ed7e5fc3 NTND |
4093 | const char *name = basename(p->pack_name); |
4094 | int i; | |
4095 | ||
4096 | if (!p->pack_local) | |
4097 | continue; | |
4098 | ||
4099 | for (i = 0; i < names->nr; i++) | |
4100 | if (!fspathcmp(name, names->items[i].string)) | |
4101 | break; | |
4102 | ||
4103 | if (i < names->nr) { | |
4104 | p->pack_keep_in_core = 1; | |
4105 | ignore_packed_keep_in_core = 1; | |
4106 | continue; | |
4107 | } | |
4108 | } | |
4109 | } | |
4110 | ||
99fb6e04 NTND |
4111 | static int option_parse_index_version(const struct option *opt, |
4112 | const char *arg, int unset) | |
4113 | { | |
4114 | char *c; | |
4115 | const char *val = arg; | |
517fe807 JK |
4116 | |
4117 | BUG_ON_OPT_NEG(unset); | |
4118 | ||
99fb6e04 NTND |
4119 | pack_idx_opts.version = strtoul(val, &c, 10); |
4120 | if (pack_idx_opts.version > 2) | |
4121 | die(_("unsupported index version %s"), val); | |
4122 | if (*c == ',' && c[1]) | |
4123 | pack_idx_opts.off32_limit = strtoul(c+1, &c, 0); | |
4124 | if (*c || pack_idx_opts.off32_limit & 0x80000000) | |
4125 | die(_("bad index version '%s'"), val); | |
4126 | return 0; | |
4127 | } | |
4128 | ||
7e52f566 JK |
4129 | static int option_parse_unpack_unreachable(const struct option *opt, |
4130 | const char *arg, int unset) | |
4131 | { | |
4132 | if (unset) { | |
4133 | unpack_unreachable = 0; | |
4134 | unpack_unreachable_expiration = 0; | |
4135 | } | |
4136 | else { | |
4137 | unpack_unreachable = 1; | |
4138 | if (arg) | |
4139 | unpack_unreachable_expiration = approxidate(arg); | |
4140 | } | |
4141 | return 0; | |
4142 | } | |
4143 | ||
b7573536 TB |
4144 | static int option_parse_cruft_expiration(const struct option *opt, |
4145 | const char *arg, int unset) | |
4146 | { | |
4147 | if (unset) { | |
4148 | cruft = 0; | |
4149 | cruft_expiration = 0; | |
4150 | } else { | |
4151 | cruft = 1; | |
4152 | if (arg) | |
4153 | cruft_expiration = approxidate(arg); | |
4154 | } | |
4155 | return 0; | |
4156 | } | |
4157 | ||
b5d97e6b JH |
4158 | int cmd_pack_objects(int argc, const char **argv, const char *prefix) |
4159 | { | |
b5d97e6b | 4160 | int use_internal_rev_list = 0; |
2dacf26d | 4161 | int shallow = 0; |
4f366275 | 4162 | int all_progress_implied = 0; |
22f9b7f3 | 4163 | struct strvec rp = STRVEC_INIT; |
99fb6e04 | 4164 | int rev_list_unpacked = 0, rev_list_all = 0, rev_list_reflog = 0; |
c90f9e13 | 4165 | int rev_list_index = 0; |
339bce27 | 4166 | int stdin_packs = 0; |
ed7e5fc3 | 4167 | struct string_list keep_pack_list = STRING_LIST_INIT_NODUP; |
0d5448a5 RS |
4168 | struct list_objects_filter_options filter_options = |
4169 | LIST_OBJECTS_FILTER_INIT; | |
80f6de4f | 4170 | |
99fb6e04 NTND |
4171 | struct option pack_objects_options[] = { |
4172 | OPT_SET_INT('q', "quiet", &progress, | |
4c688120 | 4173 | N_("do not show progress meter"), 0), |
99fb6e04 | 4174 | OPT_SET_INT(0, "progress", &progress, |
4c688120 | 4175 | N_("show progress meter"), 1), |
99fb6e04 | 4176 | OPT_SET_INT(0, "all-progress", &progress, |
4c688120 | 4177 | N_("show progress meter during object writing phase"), 2), |
99fb6e04 NTND |
4178 | OPT_BOOL(0, "all-progress-implied", |
4179 | &all_progress_implied, | |
4c688120 | 4180 | N_("similar to --all-progress when progress meter is shown")), |
203c8533 | 4181 | OPT_CALLBACK_F(0, "index-version", NULL, N_("<version>[,<offset>]"), |
4c688120 | 4182 | N_("write the pack index file in the specified idx format version"), |
203c8533 | 4183 | PARSE_OPT_NONEG, option_parse_index_version), |
2a514ed8 CB |
4184 | OPT_MAGNITUDE(0, "max-pack-size", &pack_size_limit, |
4185 | N_("maximum size of each output pack file")), | |
99fb6e04 | 4186 | OPT_BOOL(0, "local", &local, |
4c688120 | 4187 | N_("ignore borrowed objects from alternate object store")), |
99fb6e04 | 4188 | OPT_BOOL(0, "incremental", &incremental, |
4c688120 | 4189 | N_("ignore packed objects")), |
99fb6e04 | 4190 | OPT_INTEGER(0, "window", &window, |
4c688120 | 4191 | N_("limit pack window by objects")), |
2a514ed8 CB |
4192 | OPT_MAGNITUDE(0, "window-memory", &window_memory_limit, |
4193 | N_("limit pack window by memory in addition to object limit")), | |
99fb6e04 | 4194 | OPT_INTEGER(0, "depth", &depth, |
4c688120 | 4195 | N_("maximum length of delta chain allowed in the resulting pack")), |
99fb6e04 | 4196 | OPT_BOOL(0, "reuse-delta", &reuse_delta, |
4c688120 | 4197 | N_("reuse existing deltas")), |
99fb6e04 | 4198 | OPT_BOOL(0, "reuse-object", &reuse_object, |
4c688120 | 4199 | N_("reuse existing objects")), |
99fb6e04 | 4200 | OPT_BOOL(0, "delta-base-offset", &allow_ofs_delta, |
4c688120 | 4201 | N_("use OFS_DELTA objects")), |
99fb6e04 | 4202 | OPT_INTEGER(0, "threads", &delta_search_threads, |
4c688120 | 4203 | N_("use threads when searching for best delta matches")), |
99fb6e04 | 4204 | OPT_BOOL(0, "non-empty", &non_empty, |
4c688120 | 4205 | N_("do not create an empty pack output")), |
99fb6e04 | 4206 | OPT_BOOL(0, "revs", &use_internal_rev_list, |
4c688120 | 4207 | N_("read revision arguments from standard input")), |
3e4a67b4 NTND |
4208 | OPT_SET_INT_F(0, "unpacked", &rev_list_unpacked, |
4209 | N_("limit the objects to those that are not yet packed"), | |
4210 | 1, PARSE_OPT_NONEG), | |
4211 | OPT_SET_INT_F(0, "all", &rev_list_all, | |
4212 | N_("include objects reachable from any reference"), | |
4213 | 1, PARSE_OPT_NONEG), | |
4214 | OPT_SET_INT_F(0, "reflog", &rev_list_reflog, | |
4215 | N_("include objects referred by reflog entries"), | |
4216 | 1, PARSE_OPT_NONEG), | |
4217 | OPT_SET_INT_F(0, "indexed-objects", &rev_list_index, | |
4218 | N_("include objects referred to by the index"), | |
4219 | 1, PARSE_OPT_NONEG), | |
339bce27 TB |
4220 | OPT_BOOL(0, "stdin-packs", &stdin_packs, |
4221 | N_("read packs from stdin")), | |
99fb6e04 | 4222 | OPT_BOOL(0, "stdout", &pack_to_stdout, |
4c688120 | 4223 | N_("output pack to stdout")), |
99fb6e04 | 4224 | OPT_BOOL(0, "include-tag", &include_tag, |
4c688120 | 4225 | N_("include tag objects that refer to objects to be packed")), |
99fb6e04 | 4226 | OPT_BOOL(0, "keep-unreachable", &keep_unreachable, |
4c688120 | 4227 | N_("keep unreachable objects")), |
e26a8c47 JK |
4228 | OPT_BOOL(0, "pack-loose-unreachable", &pack_loose_unreachable, |
4229 | N_("pack loose unreachable objects")), | |
203c8533 | 4230 | OPT_CALLBACK_F(0, "unpack-unreachable", NULL, N_("time"), |
4c688120 | 4231 | N_("unpack unreachable objects newer than <time>"), |
203c8533 | 4232 | PARSE_OPT_OPTARG, option_parse_unpack_unreachable), |
b7573536 TB |
4233 | OPT_BOOL(0, "cruft", &cruft, N_("create a cruft pack")), |
4234 | OPT_CALLBACK_F(0, "cruft-expiration", NULL, N_("time"), | |
4235 | N_("expire cruft objects older than <time>"), | |
4236 | PARSE_OPT_OPTARG, option_parse_cruft_expiration), | |
4f6d26b1 DS |
4237 | OPT_BOOL(0, "sparse", &sparse, |
4238 | N_("use the sparse reachability algorithm")), | |
99fb6e04 | 4239 | OPT_BOOL(0, "thin", &thin, |
4c688120 | 4240 | N_("create thin packs")), |
2dacf26d | 4241 | OPT_BOOL(0, "shallow", &shallow, |
4242 | N_("create packs suitable for shallow fetches")), | |
ed7e5fc3 | 4243 | OPT_BOOL(0, "honor-pack-keep", &ignore_packed_keep_on_disk, |
4c688120 | 4244 | N_("ignore packs that have companion .keep file")), |
ed7e5fc3 NTND |
4245 | OPT_STRING_LIST(0, "keep-pack", &keep_pack_list, N_("name"), |
4246 | N_("ignore this pack")), | |
99fb6e04 | 4247 | OPT_INTEGER(0, "compression", &pack_compression_level, |
4c688120 | 4248 | N_("pack compression level")), |
99fb6e04 | 4249 | OPT_SET_INT(0, "keep-true-parents", &grafts_replace_parents, |
4c688120 | 4250 | N_("do not hide commits by grafts"), 0), |
6b8fda2d VM |
4251 | OPT_BOOL(0, "use-bitmap-index", &use_bitmap_index, |
4252 | N_("use a bitmap index if available to speed up counting objects")), | |
25575015 JK |
4253 | OPT_SET_INT(0, "write-bitmap-index", &write_bitmap_index, |
4254 | N_("write a bitmap index together with the pack index"), | |
4255 | WRITE_BITMAP_TRUE), | |
4256 | OPT_SET_INT_F(0, "write-bitmap-index-quiet", | |
4257 | &write_bitmap_index, | |
4258 | N_("write a bitmap index if possible"), | |
4259 | WRITE_BITMAP_QUIET, PARSE_OPT_HIDDEN), | |
0d5448a5 | 4260 | OPT_PARSE_LIST_OBJECTS_FILTER(&filter_options), |
203c8533 | 4261 | OPT_CALLBACK_F(0, "missing", NULL, N_("action"), |
9535ce73 | 4262 | N_("handling for missing objects"), PARSE_OPT_NONEG, |
203c8533 | 4263 | option_parse_missing_action), |
0c16cd49 JT |
4264 | OPT_BOOL(0, "exclude-promisor-objects", &exclude_promisor_objects, |
4265 | N_("do not pack objects in promisor packfiles")), | |
28b8a730 JK |
4266 | OPT_BOOL(0, "delta-islands", &use_delta_islands, |
4267 | N_("respect islands during delta compression")), | |
dd4b732d JT |
4268 | OPT_STRING_LIST(0, "uri-protocol", &uri_protocols, |
4269 | N_("protocol"), | |
4270 | N_("exclude any configured uploadpack.blobpackfileuri with this protocol")), | |
99fb6e04 NTND |
4271 | OPT_END(), |
4272 | }; | |
8d1d8f83 | 4273 | |
0c6804ab NTND |
4274 | if (DFS_NUM_STATES > (1 << OE_DFS_STATE_BITS)) |
4275 | BUG("too many dfs states, increase OE_DFS_STATE_BITS"); | |
4276 | ||
6ebd1caf | 4277 | read_replace_refs = 0; |
dae556bd | 4278 | |
2d657ab9 | 4279 | sparse = git_env_bool("GIT_TEST_PACK_SPARSE", -1); |
059fda19 JS |
4280 | if (the_repository->gitdir) { |
4281 | prepare_repo_settings(the_repository); | |
4282 | if (sparse < 0) | |
4283 | sparse = the_repository->settings.pack_use_sparse; | |
4284 | } | |
7211b9e7 | 4285 | |
ebcfb379 | 4286 | reset_pack_idx_option(&pack_idx_opts); |
ef90d6d4 | 4287 | git_config(git_pack_config, NULL); |
e8c58f89 TB |
4288 | if (git_env_bool(GIT_TEST_WRITE_REV_INDEX, 0)) |
4289 | pack_idx_opts.flags |= WRITE_REV; | |
b5d97e6b JH |
4290 | |
4291 | progress = isatty(2); | |
99fb6e04 NTND |
4292 | argc = parse_options(argc, argv, prefix, pack_objects_options, |
4293 | pack_usage, 0); | |
b5d97e6b | 4294 | |
99fb6e04 NTND |
4295 | if (argc) { |
4296 | base_name = argv[0]; | |
4297 | argc--; | |
b5d97e6b | 4298 | } |
99fb6e04 NTND |
4299 | if (pack_to_stdout != !base_name || argc) |
4300 | usage_with_options(pack_usage, pack_objects_options); | |
b5d97e6b | 4301 | |
6d52b6a5 JK |
4302 | if (depth < 0) |
4303 | depth = 0; | |
b5c0cbd8 NTND |
4304 | if (depth >= (1 << OE_DEPTH_BITS)) { |
4305 | warning(_("delta chain depth %d is too deep, forcing %d"), | |
4306 | depth, (1 << OE_DEPTH_BITS) - 1); | |
4307 | depth = (1 << OE_DEPTH_BITS) - 1; | |
4308 | } | |
0cb3c142 NTND |
4309 | if (cache_max_small_delta_size >= (1U << OE_Z_DELTA_BITS)) { |
4310 | warning(_("pack.deltaCacheLimit is too high, forcing %d"), | |
4311 | (1U << OE_Z_DELTA_BITS) - 1); | |
4312 | cache_max_small_delta_size = (1U << OE_Z_DELTA_BITS) - 1; | |
4313 | } | |
953aa54e JK |
4314 | if (window < 0) |
4315 | window = 0; | |
b5c0cbd8 | 4316 | |
22f9b7f3 | 4317 | strvec_push(&rp, "pack-objects"); |
99fb6e04 NTND |
4318 | if (thin) { |
4319 | use_internal_rev_list = 1; | |
22f9b7f3 | 4320 | strvec_push(&rp, shallow |
2dacf26d | 4321 | ? "--objects-edge-aggressive" |
4322 | : "--objects-edge"); | |
99fb6e04 | 4323 | } else |
22f9b7f3 | 4324 | strvec_push(&rp, "--objects"); |
b5d97e6b | 4325 | |
99fb6e04 NTND |
4326 | if (rev_list_all) { |
4327 | use_internal_rev_list = 1; | |
22f9b7f3 | 4328 | strvec_push(&rp, "--all"); |
99fb6e04 NTND |
4329 | } |
4330 | if (rev_list_reflog) { | |
4331 | use_internal_rev_list = 1; | |
22f9b7f3 | 4332 | strvec_push(&rp, "--reflog"); |
99fb6e04 | 4333 | } |
c90f9e13 JK |
4334 | if (rev_list_index) { |
4335 | use_internal_rev_list = 1; | |
22f9b7f3 | 4336 | strvec_push(&rp, "--indexed-objects"); |
99fb6e04 | 4337 | } |
339bce27 | 4338 | if (rev_list_unpacked && !stdin_packs) { |
99fb6e04 | 4339 | use_internal_rev_list = 1; |
22f9b7f3 | 4340 | strvec_push(&rp, "--unpacked"); |
99fb6e04 | 4341 | } |
b5d97e6b | 4342 | |
0c16cd49 JT |
4343 | if (exclude_promisor_objects) { |
4344 | use_internal_rev_list = 1; | |
4345 | fetch_if_missing = 0; | |
22f9b7f3 | 4346 | strvec_push(&rp, "--exclude-promisor-objects"); |
0c16cd49 | 4347 | } |
58bd77b6 NTND |
4348 | if (unpack_unreachable || keep_unreachable || pack_loose_unreachable) |
4349 | use_internal_rev_list = 1; | |
0c16cd49 | 4350 | |
99fb6e04 NTND |
4351 | if (!reuse_object) |
4352 | reuse_delta = 0; | |
4353 | if (pack_compression_level == -1) | |
4354 | pack_compression_level = Z_DEFAULT_COMPRESSION; | |
4355 | else if (pack_compression_level < 0 || pack_compression_level > Z_BEST_COMPRESSION) | |
f616db6a | 4356 | die(_("bad pack compression level %d"), pack_compression_level); |
0c45d258 JH |
4357 | |
4358 | if (!delta_search_threads) /* --threads=0 means autodetect */ | |
4359 | delta_search_threads = online_cpus(); | |
4360 | ||
9c897c5c | 4361 | if (!HAVE_THREADS && delta_search_threads != 1) |
f616db6a | 4362 | warning(_("no threads support, ignoring --threads")); |
b7573536 | 4363 | if (!pack_to_stdout && !pack_size_limit && !cruft) |
2b84b5a8 | 4364 | pack_size_limit = pack_size_limit_cfg; |
01c12a23 | 4365 | if (pack_to_stdout && pack_size_limit) |
f616db6a | 4366 | die(_("--max-pack-size cannot be used to build a pack for transfer")); |
07cf0f24 | 4367 | if (pack_size_limit && pack_size_limit < 1024*1024) { |
f616db6a | 4368 | warning(_("minimum pack size limit is 1 MiB")); |
07cf0f24 NP |
4369 | pack_size_limit = 1024*1024; |
4370 | } | |
01c12a23 | 4371 | |
8d1d8f83 | 4372 | if (!pack_to_stdout && thin) |
f616db6a | 4373 | die(_("--thin cannot be used to build an indexable pack")); |
b5d97e6b | 4374 | |
ca11b212 | 4375 | if (keep_unreachable && unpack_unreachable) |
12909b6b | 4376 | die(_("options '%s' and '%s' cannot be used together"), "--keep-unreachable", "--unpack-unreachable"); |
b1e757f3 JK |
4377 | if (!rev_list_all || !rev_list_reflog || !rev_list_index) |
4378 | unpack_unreachable_expiration = 0; | |
ca11b212 | 4379 | |
0d5448a5 | 4380 | if (filter_options.choice) { |
9535ce73 | 4381 | if (!pack_to_stdout) |
f616db6a | 4382 | die(_("cannot use --filter without --stdout")); |
339bce27 TB |
4383 | if (stdin_packs) |
4384 | die(_("cannot use --filter with --stdin-packs")); | |
9535ce73 JH |
4385 | } |
4386 | ||
339bce27 TB |
4387 | if (stdin_packs && use_internal_rev_list) |
4388 | die(_("cannot use internal rev list with --stdin-packs")); | |
4389 | ||
b7573536 TB |
4390 | if (cruft) { |
4391 | if (use_internal_rev_list) | |
4392 | die(_("cannot use internal rev list with --cruft")); | |
4393 | if (stdin_packs) | |
4394 | die(_("cannot use --stdin-packs with --cruft")); | |
4395 | if (pack_size_limit) | |
4396 | die(_("cannot use --max-pack-size with --cruft")); | |
4397 | } | |
4398 | ||
645c432d KS |
4399 | /* |
4400 | * "soft" reasons not to use bitmaps - for on-disk repack by default we want | |
4401 | * | |
4402 | * - to produce good pack (with bitmap index not-yet-packed objects are | |
4403 | * packed in suboptimal order). | |
4404 | * | |
4405 | * - to use more robust pack-generation codepath (avoiding possible | |
4406 | * bugs in bitmap code and possible bitmap index corruption). | |
4407 | */ | |
4408 | if (!pack_to_stdout) | |
4409 | use_bitmap_index_default = 0; | |
4410 | ||
4411 | if (use_bitmap_index < 0) | |
4412 | use_bitmap_index = use_bitmap_index_default; | |
4413 | ||
4414 | /* "hard" reasons not to use bitmaps; these just won't work at all */ | |
c8813487 | 4415 | if (!use_internal_rev_list || (!pack_to_stdout && write_bitmap_index) || is_repository_shallow(the_repository)) |
6b8fda2d VM |
4416 | use_bitmap_index = 0; |
4417 | ||
7cc8f971 VM |
4418 | if (pack_to_stdout || !rev_list_all) |
4419 | write_bitmap_index = 0; | |
4420 | ||
28b8a730 | 4421 | if (use_delta_islands) |
22f9b7f3 | 4422 | strvec_push(&rp, "--topo-order"); |
28b8a730 | 4423 | |
4f366275 NP |
4424 | if (progress && all_progress_implied) |
4425 | progress = 2; | |
4426 | ||
ed7e5fc3 NTND |
4427 | add_extra_kept_packs(&keep_pack_list); |
4428 | if (ignore_packed_keep_on_disk) { | |
56dfeb62 | 4429 | struct packed_git *p; |
454ea2e4 | 4430 | for (p = get_all_packs(the_repository); p; p = p->next) |
56dfeb62 JK |
4431 | if (p->pack_local && p->pack_keep) |
4432 | break; | |
4433 | if (!p) /* no keep-able packs found */ | |
ed7e5fc3 | 4434 | ignore_packed_keep_on_disk = 0; |
56dfeb62 JK |
4435 | } |
4436 | if (local) { | |
4437 | /* | |
ed7e5fc3 NTND |
4438 | * unlike ignore_packed_keep_on_disk above, we do not |
4439 | * want to unset "local" based on looking at packs, as | |
4440 | * it also covers non-local objects | |
56dfeb62 JK |
4441 | */ |
4442 | struct packed_git *p; | |
454ea2e4 | 4443 | for (p = get_all_packs(the_repository); p; p = p->next) { |
56dfeb62 JK |
4444 | if (!p->pack_local) { |
4445 | have_non_local_packs = 1; | |
4446 | break; | |
4447 | } | |
4448 | } | |
4449 | } | |
b5d97e6b | 4450 | |
ae417807 DS |
4451 | trace2_region_enter("pack-objects", "enumerate-objects", |
4452 | the_repository); | |
7c141127 | 4453 | prepare_packing_data(the_repository, &to_pack); |
43fa44fa | 4454 | |
b7573536 | 4455 | if (progress && !cruft) |
5af05043 | 4456 | progress_state = start_progress(_("Enumerating objects"), 0); |
339bce27 TB |
4457 | if (stdin_packs) { |
4458 | /* avoids adding objects in excluded packs */ | |
4459 | ignore_packed_keep_in_core = 1; | |
4460 | read_packs_list_from_stdin(); | |
4461 | if (rev_list_unpacked) | |
4462 | add_unreachable_loose_objects(); | |
b7573536 TB |
4463 | } else if (cruft) { |
4464 | read_cruft_objects(); | |
9e39acc9 | 4465 | } else if (!use_internal_rev_list) { |
b5d97e6b | 4466 | read_object_list_from_stdin(); |
9e39acc9 | 4467 | } else { |
5cb28270 ÆAB |
4468 | struct rev_info revs; |
4469 | ||
4470 | repo_init_revisions(the_repository, &revs, NULL); | |
0d5448a5 | 4471 | list_objects_filter_copy(&revs.filter, &filter_options); |
80f6de4f | 4472 | get_object_list(&revs, rp.nr, rp.v); |
2108fe4a | 4473 | release_revisions(&revs); |
8d1d8f83 | 4474 | } |
0ef95f72 | 4475 | cleanup_preferred_base(); |
d155254c | 4476 | if (include_tag && nr_result) |
be18153b | 4477 | for_each_tag_ref(add_ref_tag, NULL); |
4d4fcc54 | 4478 | stop_progress(&progress_state); |
ae417807 DS |
4479 | trace2_region_leave("pack-objects", "enumerate-objects", |
4480 | the_repository); | |
96a02f8f | 4481 | |
f0b0af1b | 4482 | if (non_empty && !nr_result) |
9e39acc9 | 4483 | goto cleanup; |
ae417807 DS |
4484 | if (nr_result) { |
4485 | trace2_region_enter("pack-objects", "prepare-pack", | |
4486 | the_repository); | |
f7ae6a93 | 4487 | prepare_pack(window, depth); |
ae417807 DS |
4488 | trace2_region_leave("pack-objects", "prepare-pack", |
4489 | the_repository); | |
4490 | } | |
4491 | ||
4492 | trace2_region_enter("pack-objects", "write-pack-file", the_repository); | |
dd4b732d | 4493 | write_excluded_by_configs(); |
d01fb92f | 4494 | write_pack_file(); |
ae417807 DS |
4495 | trace2_region_leave("pack-objects", "write-pack-file", the_repository); |
4496 | ||
ab7cd7bb | 4497 | if (progress) |
f616db6a NTND |
4498 | fprintf_ln(stderr, |
4499 | _("Total %"PRIu32" (delta %"PRIu32")," | |
bab28d9f JK |
4500 | " reused %"PRIu32" (delta %"PRIu32")," |
4501 | " pack-reused %"PRIu32), | |
4502 | written, written_delta, reused, reused_delta, | |
4503 | reuse_packfile_objects); | |
9e39acc9 TB |
4504 | |
4505 | cleanup: | |
0d5448a5 | 4506 | list_objects_filter_release(&filter_options); |
9e39acc9 TB |
4507 | strvec_clear(&rp); |
4508 | ||
c323ac7d LT |
4509 | return 0; |
4510 | } |