]>
Commit | Line | Data |
---|---|---|
5d4a6003 | 1 | #include "builtin.h" |
c323ac7d | 2 | #include "cache.h" |
a74db82e | 3 | #include "attr.h" |
c323ac7d | 4 | #include "object.h" |
8e440259 PE |
5 | #include "blob.h" |
6 | #include "commit.h" | |
7 | #include "tag.h" | |
8 | #include "tree.h" | |
c323ac7d | 9 | #include "delta.h" |
a733cb60 | 10 | #include "pack.h" |
3449f8c4 | 11 | #include "pack-revindex.h" |
c38138cd | 12 | #include "csum-file.h" |
1b0c7174 | 13 | #include "tree-walk.h" |
b5d97e6b JH |
14 | #include "diff.h" |
15 | #include "revision.h" | |
16 | #include "list-objects.h" | |
2834bc27 | 17 | #include "pack-objects.h" |
96a02f8f | 18 | #include "progress.h" |
f0a24aa5 | 19 | #include "refs.h" |
cf2ba13a | 20 | #include "streaming.h" |
93749194 | 21 | #include "thread-utils.h" |
6b8fda2d | 22 | #include "pack-bitmap.h" |
abcb8655 JK |
23 | #include "reachable.h" |
24 | #include "sha1-array.h" | |
edfbb2aa | 25 | #include "argv-array.h" |
c9af708b | 26 | #include "mru.h" |
8ecce684 | 27 | |
99fb6e04 | 28 | static const char *pack_usage[] = { |
b8c1d275 AH |
29 | N_("git pack-objects --stdout [<options>...] [< <ref-list> | < <object-list>]"), |
30 | N_("git pack-objects [<options>...] <base-name> [< <ref-list> | < <object-list>]"), | |
99fb6e04 NTND |
31 | NULL |
32 | }; | |
c323ac7d | 33 | |
3f9ac8d2 | 34 | /* |
2834bc27 VM |
35 | * Objects we are going to pack are collected in the `to_pack` structure. |
36 | * It contains an array (dynamically expanded) of the object data, and a map | |
37 | * that can resolve SHA1s to their position in the array. | |
3f9ac8d2 | 38 | */ |
2834bc27 VM |
39 | static struct packing_data to_pack; |
40 | ||
79814f42 | 41 | static struct pack_idx_entry **written_list; |
2834bc27 | 42 | static uint32_t nr_result, nr_written; |
3f9ac8d2 | 43 | |
96f1e58f | 44 | static int non_empty; |
a7de7130 | 45 | static int reuse_delta = 1, reuse_object = 1; |
e5e9714a | 46 | static int keep_unreachable, unpack_unreachable, include_tag; |
dddbad72 | 47 | static timestamp_t unpack_unreachable_expiration; |
e26a8c47 | 48 | static int pack_loose_unreachable; |
96f1e58f | 49 | static int local; |
56dfeb62 | 50 | static int have_non_local_packs; |
96f1e58f | 51 | static int incremental; |
e96fb9b8 | 52 | static int ignore_packed_keep; |
be6b1914 | 53 | static int allow_ofs_delta; |
ebcfb379 | 54 | static struct pack_idx_option pack_idx_opts; |
d01fb92f | 55 | static const char *base_name; |
024701f1 | 56 | static int progress = 1; |
4812a93a | 57 | static int window = 10; |
568508e7 | 58 | static unsigned long pack_size_limit; |
618e613a | 59 | static int depth = 50; |
43cc2b42 | 60 | static int delta_search_threads; |
df6d6101 | 61 | static int pack_to_stdout; |
8d1d8f83 | 62 | static int num_preferred_base; |
dc6a0757 | 63 | static struct progress *progress_state; |
c323ac7d | 64 | |
6b8fda2d VM |
65 | static struct packed_git *reuse_packfile; |
66 | static uint32_t reuse_packfile_objects; | |
67 | static off_t reuse_packfile_offset; | |
68 | ||
645c432d KS |
69 | static int use_bitmap_index_default = 1; |
70 | static int use_bitmap_index = -1; | |
7cc8f971 | 71 | static int write_bitmap_index; |
ae4f07fb | 72 | static uint16_t write_bitmap_options; |
6b8fda2d | 73 | |
074b2eea | 74 | static unsigned long delta_cache_size = 0; |
5749b0b2 | 75 | static unsigned long max_delta_cache_size = 256 * 1024 * 1024; |
e3dfddb3 | 76 | static unsigned long cache_max_small_delta_size = 1000; |
074b2eea | 77 | |
a97773ce BD |
78 | static unsigned long window_memory_limit = 0; |
79 | ||
3f9ac8d2 JH |
80 | /* |
81 | * stats | |
82 | */ | |
7cadf491 SP |
83 | static uint32_t written, written_delta; |
84 | static uint32_t reused, reused_delta; | |
3f9ac8d2 | 85 | |
7cc8f971 VM |
86 | /* |
87 | * Indexed commits | |
88 | */ | |
89 | static struct commit **indexed_commits; | |
90 | static unsigned int indexed_commits_nr; | |
91 | static unsigned int indexed_commits_alloc; | |
92 | ||
93 | static void index_commit_for_bitmap(struct commit *commit) | |
94 | { | |
95 | if (indexed_commits_nr >= indexed_commits_alloc) { | |
96 | indexed_commits_alloc = (indexed_commits_alloc + 32) * 2; | |
2756ca43 | 97 | REALLOC_ARRAY(indexed_commits, indexed_commits_alloc); |
7cc8f971 VM |
98 | } |
99 | ||
100 | indexed_commits[indexed_commits_nr++] = commit; | |
101 | } | |
780e6e73 | 102 | |
3613f9b4 | 103 | static void *get_delta(struct object_entry *entry) |
c323ac7d | 104 | { |
3613f9b4 NP |
105 | unsigned long size, base_size, delta_size; |
106 | void *buf, *base_buf, *delta_buf; | |
21666f1a | 107 | enum object_type type; |
c323ac7d | 108 | |
e6a492b7 | 109 | buf = read_sha1_file(entry->idx.oid.hash, &type, &size); |
3613f9b4 | 110 | if (!buf) |
e6a492b7 | 111 | die("unable to read %s", oid_to_hex(&entry->idx.oid)); |
112 | base_buf = read_sha1_file(entry->delta->idx.oid.hash, &type, | |
113 | &base_size); | |
3613f9b4 | 114 | if (!base_buf) |
e6a492b7 | 115 | die("unable to read %s", |
116 | oid_to_hex(&entry->delta->idx.oid)); | |
3613f9b4 | 117 | delta_buf = diff_delta(base_buf, base_size, |
dcde55bc | 118 | buf, size, &delta_size, 0); |
3613f9b4 | 119 | if (!delta_buf || delta_size != entry->delta_size) |
a6080a0a | 120 | die("delta size changed"); |
3613f9b4 NP |
121 | free(buf); |
122 | free(base_buf); | |
c323ac7d LT |
123 | return delta_buf; |
124 | } | |
125 | ||
30ebb40a NP |
126 | static unsigned long do_compress(void **pptr, unsigned long size) |
127 | { | |
ef49a7a0 | 128 | git_zstream stream; |
30ebb40a NP |
129 | void *in, *out; |
130 | unsigned long maxsize; | |
131 | ||
55bb5c91 | 132 | git_deflate_init(&stream, pack_compression_level); |
225a6f10 | 133 | maxsize = git_deflate_bound(&stream, size); |
30ebb40a NP |
134 | |
135 | in = *pptr; | |
136 | out = xmalloc(maxsize); | |
137 | *pptr = out; | |
138 | ||
139 | stream.next_in = in; | |
140 | stream.avail_in = size; | |
141 | stream.next_out = out; | |
142 | stream.avail_out = maxsize; | |
55bb5c91 | 143 | while (git_deflate(&stream, Z_FINISH) == Z_OK) |
30ebb40a | 144 | ; /* nothing */ |
55bb5c91 | 145 | git_deflate_end(&stream); |
30ebb40a NP |
146 | |
147 | free(in); | |
148 | return stream.total_out; | |
149 | } | |
150 | ||
cf2ba13a NTND |
151 | static unsigned long write_large_blob_data(struct git_istream *st, struct sha1file *f, |
152 | const unsigned char *sha1) | |
153 | { | |
154 | git_zstream stream; | |
155 | unsigned char ibuf[1024 * 16]; | |
156 | unsigned char obuf[1024 * 16]; | |
157 | unsigned long olen = 0; | |
158 | ||
cf2ba13a NTND |
159 | git_deflate_init(&stream, pack_compression_level); |
160 | ||
161 | for (;;) { | |
162 | ssize_t readlen; | |
163 | int zret = Z_OK; | |
164 | readlen = read_istream(st, ibuf, sizeof(ibuf)); | |
165 | if (readlen == -1) | |
166 | die(_("unable to read %s"), sha1_to_hex(sha1)); | |
167 | ||
168 | stream.next_in = ibuf; | |
169 | stream.avail_in = readlen; | |
170 | while ((stream.avail_in || readlen == 0) && | |
171 | (zret == Z_OK || zret == Z_BUF_ERROR)) { | |
172 | stream.next_out = obuf; | |
173 | stream.avail_out = sizeof(obuf); | |
174 | zret = git_deflate(&stream, readlen ? 0 : Z_FINISH); | |
175 | sha1write(f, obuf, stream.next_out - obuf); | |
176 | olen += stream.next_out - obuf; | |
177 | } | |
178 | if (stream.avail_in) | |
179 | die(_("deflate error (%d)"), zret); | |
180 | if (readlen == 0) { | |
181 | if (zret != Z_STREAM_END) | |
182 | die(_("deflate error (%d)"), zret); | |
183 | break; | |
184 | } | |
185 | } | |
186 | git_deflate_end(&stream); | |
187 | return olen; | |
188 | } | |
189 | ||
780e6e73 NP |
190 | /* |
191 | * we are going to reuse the existing object data as is. make | |
192 | * sure it is not corrupt. | |
193 | */ | |
079afb18 SP |
194 | static int check_pack_inflate(struct packed_git *p, |
195 | struct pack_window **w_curs, | |
6777a59f SP |
196 | off_t offset, |
197 | off_t len, | |
079afb18 SP |
198 | unsigned long expect) |
199 | { | |
ef49a7a0 | 200 | git_zstream stream; |
079afb18 SP |
201 | unsigned char fakebuf[4096], *in; |
202 | int st; | |
203 | ||
204 | memset(&stream, 0, sizeof(stream)); | |
39c68542 | 205 | git_inflate_init(&stream); |
079afb18 SP |
206 | do { |
207 | in = use_pack(p, w_curs, offset, &stream.avail_in); | |
208 | stream.next_in = in; | |
209 | stream.next_out = fakebuf; | |
210 | stream.avail_out = sizeof(fakebuf); | |
39c68542 | 211 | st = git_inflate(&stream, Z_FINISH); |
079afb18 SP |
212 | offset += stream.next_in - in; |
213 | } while (st == Z_OK || st == Z_BUF_ERROR); | |
39c68542 | 214 | git_inflate_end(&stream); |
079afb18 SP |
215 | return (st == Z_STREAM_END && |
216 | stream.total_out == expect && | |
217 | stream.total_in == len) ? 0 : -1; | |
218 | } | |
219 | ||
220 | static void copy_pack_data(struct sha1file *f, | |
221 | struct packed_git *p, | |
222 | struct pack_window **w_curs, | |
6777a59f SP |
223 | off_t offset, |
224 | off_t len) | |
079afb18 SP |
225 | { |
226 | unsigned char *in; | |
ef49a7a0 | 227 | unsigned long avail; |
079afb18 SP |
228 | |
229 | while (len) { | |
230 | in = use_pack(p, w_curs, offset, &avail); | |
231 | if (avail > len) | |
ef49a7a0 | 232 | avail = (unsigned long)len; |
079afb18 SP |
233 | sha1write(f, in, avail); |
234 | offset += avail; | |
235 | len -= avail; | |
236 | } | |
237 | } | |
238 | ||
1b4bb16b | 239 | /* Return 0 if we will bust the pack-size limit */ |
c9018b03 NTND |
240 | static unsigned long write_no_reuse_object(struct sha1file *f, struct object_entry *entry, |
241 | unsigned long limit, int usable_delta) | |
c323ac7d | 242 | { |
c9018b03 | 243 | unsigned long size, datalen; |
2c5e2865 JK |
244 | unsigned char header[MAX_PACK_OBJECT_HEADER], |
245 | dheader[MAX_PACK_OBJECT_HEADER]; | |
6777a59f | 246 | unsigned hdrlen; |
2c5ef824 | 247 | enum object_type type; |
c9018b03 | 248 | void *buf; |
cf2ba13a | 249 | struct git_istream *st = NULL; |
c9018b03 NTND |
250 | |
251 | if (!usable_delta) { | |
cf2ba13a NTND |
252 | if (entry->type == OBJ_BLOB && |
253 | entry->size > big_file_threshold && | |
e6a492b7 | 254 | (st = open_istream(entry->idx.oid.hash, &type, &size, NULL)) != NULL) |
cf2ba13a NTND |
255 | buf = NULL; |
256 | else { | |
e6a492b7 | 257 | buf = read_sha1_file(entry->idx.oid.hash, &type, |
258 | &size); | |
cf2ba13a | 259 | if (!buf) |
e6a492b7 | 260 | die(_("unable to read %s"), |
261 | oid_to_hex(&entry->idx.oid)); | |
cf2ba13a | 262 | } |
c9018b03 NTND |
263 | /* |
264 | * make sure no cached delta data remains from a | |
265 | * previous attempt before a pack split occurred. | |
266 | */ | |
267 | free(entry->delta_data); | |
268 | entry->delta_data = NULL; | |
269 | entry->z_delta_size = 0; | |
270 | } else if (entry->delta_data) { | |
271 | size = entry->delta_size; | |
272 | buf = entry->delta_data; | |
273 | entry->delta_data = NULL; | |
274 | type = (allow_ofs_delta && entry->delta->idx.offset) ? | |
275 | OBJ_OFS_DELTA : OBJ_REF_DELTA; | |
276 | } else { | |
277 | buf = get_delta(entry); | |
278 | size = entry->delta_size; | |
279 | type = (allow_ofs_delta && entry->delta->idx.offset) ? | |
280 | OBJ_OFS_DELTA : OBJ_REF_DELTA; | |
281 | } | |
282 | ||
cf2ba13a NTND |
283 | if (st) /* large blob case, just assume we don't compress well */ |
284 | datalen = size; | |
285 | else if (entry->z_delta_size) | |
c9018b03 NTND |
286 | datalen = entry->z_delta_size; |
287 | else | |
288 | datalen = do_compress(&buf, size); | |
289 | ||
290 | /* | |
291 | * The object header is a byte of 'type' followed by zero or | |
292 | * more bytes of length. | |
293 | */ | |
7202a6fa JK |
294 | hdrlen = encode_in_pack_object_header(header, sizeof(header), |
295 | type, size); | |
c9018b03 NTND |
296 | |
297 | if (type == OBJ_OFS_DELTA) { | |
298 | /* | |
299 | * Deltas with relative base contain an additional | |
300 | * encoding of the relative offset for the delta | |
301 | * base from this object's position in the pack. | |
302 | */ | |
303 | off_t ofs = entry->idx.offset - entry->delta->idx.offset; | |
304 | unsigned pos = sizeof(dheader) - 1; | |
305 | dheader[pos] = ofs & 127; | |
306 | while (ofs >>= 7) | |
307 | dheader[--pos] = 128 | (--ofs & 127); | |
308 | if (limit && hdrlen + sizeof(dheader) - pos + datalen + 20 >= limit) { | |
cf2ba13a NTND |
309 | if (st) |
310 | close_istream(st); | |
c9018b03 NTND |
311 | free(buf); |
312 | return 0; | |
313 | } | |
314 | sha1write(f, header, hdrlen); | |
315 | sha1write(f, dheader + pos, sizeof(dheader) - pos); | |
316 | hdrlen += sizeof(dheader) - pos; | |
317 | } else if (type == OBJ_REF_DELTA) { | |
318 | /* | |
319 | * Deltas with a base reference contain | |
320 | * an additional 20 bytes for the base sha1. | |
321 | */ | |
322 | if (limit && hdrlen + 20 + datalen + 20 >= limit) { | |
cf2ba13a NTND |
323 | if (st) |
324 | close_istream(st); | |
c9018b03 NTND |
325 | free(buf); |
326 | return 0; | |
327 | } | |
328 | sha1write(f, header, hdrlen); | |
e6a492b7 | 329 | sha1write(f, entry->delta->idx.oid.hash, 20); |
c9018b03 NTND |
330 | hdrlen += 20; |
331 | } else { | |
332 | if (limit && hdrlen + datalen + 20 >= limit) { | |
cf2ba13a NTND |
333 | if (st) |
334 | close_istream(st); | |
c9018b03 NTND |
335 | free(buf); |
336 | return 0; | |
337 | } | |
338 | sha1write(f, header, hdrlen); | |
339 | } | |
cf2ba13a | 340 | if (st) { |
e6a492b7 | 341 | datalen = write_large_blob_data(st, f, entry->idx.oid.hash); |
cf2ba13a NTND |
342 | close_istream(st); |
343 | } else { | |
344 | sha1write(f, buf, datalen); | |
345 | free(buf); | |
346 | } | |
c9018b03 NTND |
347 | |
348 | return hdrlen + datalen; | |
349 | } | |
350 | ||
351 | /* Return 0 if we will bust the pack-size limit */ | |
af92a645 NTND |
352 | static off_t write_reuse_object(struct sha1file *f, struct object_entry *entry, |
353 | unsigned long limit, int usable_delta) | |
c9018b03 NTND |
354 | { |
355 | struct packed_git *p = entry->in_pack; | |
356 | struct pack_window *w_curs = NULL; | |
357 | struct revindex_entry *revidx; | |
358 | off_t offset; | |
359 | enum object_type type = entry->type; | |
211c61c6 | 360 | off_t datalen; |
2c5e2865 JK |
361 | unsigned char header[MAX_PACK_OBJECT_HEADER], |
362 | dheader[MAX_PACK_OBJECT_HEADER]; | |
c9018b03 NTND |
363 | unsigned hdrlen; |
364 | ||
365 | if (entry->delta) | |
366 | type = (allow_ofs_delta && entry->delta->idx.offset) ? | |
367 | OBJ_OFS_DELTA : OBJ_REF_DELTA; | |
7202a6fa JK |
368 | hdrlen = encode_in_pack_object_header(header, sizeof(header), |
369 | type, entry->size); | |
c9018b03 NTND |
370 | |
371 | offset = entry->in_pack_offset; | |
372 | revidx = find_pack_revindex(p, offset); | |
373 | datalen = revidx[1].offset - offset; | |
374 | if (!pack_to_stdout && p->index_version > 1 && | |
375 | check_pack_crc(p, &w_curs, offset, datalen, revidx->nr)) { | |
e6a492b7 | 376 | error("bad packed object CRC for %s", |
377 | oid_to_hex(&entry->idx.oid)); | |
c9018b03 NTND |
378 | unuse_pack(&w_curs); |
379 | return write_no_reuse_object(f, entry, limit, usable_delta); | |
380 | } | |
381 | ||
382 | offset += entry->in_pack_header_size; | |
383 | datalen -= entry->in_pack_header_size; | |
384 | ||
385 | if (!pack_to_stdout && p->index_version == 1 && | |
386 | check_pack_inflate(p, &w_curs, offset, datalen, entry->size)) { | |
e6a492b7 | 387 | error("corrupt packed object for %s", |
388 | oid_to_hex(&entry->idx.oid)); | |
c9018b03 NTND |
389 | unuse_pack(&w_curs); |
390 | return write_no_reuse_object(f, entry, limit, usable_delta); | |
391 | } | |
392 | ||
393 | if (type == OBJ_OFS_DELTA) { | |
394 | off_t ofs = entry->idx.offset - entry->delta->idx.offset; | |
395 | unsigned pos = sizeof(dheader) - 1; | |
396 | dheader[pos] = ofs & 127; | |
397 | while (ofs >>= 7) | |
398 | dheader[--pos] = 128 | (--ofs & 127); | |
399 | if (limit && hdrlen + sizeof(dheader) - pos + datalen + 20 >= limit) { | |
400 | unuse_pack(&w_curs); | |
401 | return 0; | |
402 | } | |
403 | sha1write(f, header, hdrlen); | |
404 | sha1write(f, dheader + pos, sizeof(dheader) - pos); | |
405 | hdrlen += sizeof(dheader) - pos; | |
406 | reused_delta++; | |
407 | } else if (type == OBJ_REF_DELTA) { | |
408 | if (limit && hdrlen + 20 + datalen + 20 >= limit) { | |
409 | unuse_pack(&w_curs); | |
410 | return 0; | |
411 | } | |
412 | sha1write(f, header, hdrlen); | |
e6a492b7 | 413 | sha1write(f, entry->delta->idx.oid.hash, 20); |
c9018b03 NTND |
414 | hdrlen += 20; |
415 | reused_delta++; | |
416 | } else { | |
417 | if (limit && hdrlen + datalen + 20 >= limit) { | |
418 | unuse_pack(&w_curs); | |
419 | return 0; | |
420 | } | |
421 | sha1write(f, header, hdrlen); | |
422 | } | |
423 | copy_pack_data(f, p, &w_curs, offset, datalen); | |
424 | unuse_pack(&w_curs); | |
425 | reused++; | |
426 | return hdrlen + datalen; | |
427 | } | |
428 | ||
429 | /* Return 0 if we will bust the pack-size limit */ | |
af92a645 NTND |
430 | static off_t write_object(struct sha1file *f, |
431 | struct object_entry *entry, | |
432 | off_t write_offset) | |
c9018b03 | 433 | { |
af92a645 NTND |
434 | unsigned long limit; |
435 | off_t len; | |
2c5ef824 | 436 | int usable_delta, to_reuse; |
c323ac7d | 437 | |
78d1e84f NP |
438 | if (!pack_to_stdout) |
439 | crc32_begin(f); | |
440 | ||
a2430dde | 441 | /* apply size limit if limited packsize and not first object */ |
a1e4760f NP |
442 | if (!pack_size_limit || !nr_written) |
443 | limit = 0; | |
444 | else if (pack_size_limit <= write_offset) | |
445 | /* | |
446 | * the earlier object did not fit the limit; avoid | |
447 | * mistaking this with unlimited (i.e. limit = 0). | |
448 | */ | |
449 | limit = 1; | |
450 | else | |
451 | limit = pack_size_limit - write_offset; | |
2c5ef824 NP |
452 | |
453 | if (!entry->delta) | |
454 | usable_delta = 0; /* no delta */ | |
455 | else if (!pack_size_limit) | |
456 | usable_delta = 1; /* unlimited packfile */ | |
457 | else if (entry->delta->idx.offset == (off_t)-1) | |
458 | usable_delta = 0; /* base was written to another pack */ | |
459 | else if (entry->delta->idx.offset) | |
460 | usable_delta = 1; /* base already exists in this pack */ | |
461 | else | |
462 | usable_delta = 0; /* base could end up in another pack */ | |
463 | ||
a7de7130 | 464 | if (!reuse_object) |
fa736f72 NP |
465 | to_reuse = 0; /* explicit */ |
466 | else if (!entry->in_pack) | |
ab7cd7bb | 467 | to_reuse = 0; /* can't reuse what we don't have */ |
c9018b03 | 468 | else if (entry->type == OBJ_REF_DELTA || entry->type == OBJ_OFS_DELTA) |
17b08f2c DH |
469 | /* check_object() decided it for us ... */ |
470 | to_reuse = usable_delta; | |
471 | /* ... but pack split may override that */ | |
c9018b03 | 472 | else if (entry->type != entry->in_pack_type) |
ab7cd7bb JH |
473 | to_reuse = 0; /* pack has delta which is unusable */ |
474 | else if (entry->delta) | |
475 | to_reuse = 0; /* we want to pack afresh */ | |
476 | else | |
477 | to_reuse = 1; /* we have it in-pack undeltified, | |
478 | * and we do not need to deltify it. | |
479 | */ | |
480 | ||
c9018b03 NTND |
481 | if (!to_reuse) |
482 | len = write_no_reuse_object(f, entry, limit, usable_delta); | |
483 | else | |
484 | len = write_reuse_object(f, entry, limit, usable_delta); | |
485 | if (!len) | |
486 | return 0; | |
64bd76b1 | 487 | |
17b08f2c | 488 | if (usable_delta) |
ab7cd7bb | 489 | written_delta++; |
3f9ac8d2 | 490 | written++; |
78d1e84f | 491 | if (!pack_to_stdout) |
aa7e44bf | 492 | entry->idx.crc32 = crc32_end(f); |
c9018b03 | 493 | return len; |
c323ac7d LT |
494 | } |
495 | ||
f63c79db JH |
496 | enum write_one_status { |
497 | WRITE_ONE_SKIP = -1, /* already written */ | |
498 | WRITE_ONE_BREAK = 0, /* writing this will bust the limit; not written */ | |
499 | WRITE_ONE_WRITTEN = 1, /* normal */ | |
500 | WRITE_ONE_RECURSIVE = 2 /* already scheduled to be written */ | |
501 | }; | |
502 | ||
503 | static enum write_one_status write_one(struct sha1file *f, | |
504 | struct object_entry *e, | |
505 | off_t *offset) | |
9d5ab962 | 506 | { |
af92a645 | 507 | off_t size; |
f63c79db | 508 | int recursing; |
d7dd0223 | 509 | |
f63c79db JH |
510 | /* |
511 | * we set offset to 1 (which is an impossible value) to mark | |
512 | * the fact that this object is involved in "write its base | |
513 | * first before writing a deltified object" recursion. | |
514 | */ | |
515 | recursing = (e->idx.offset == 1); | |
516 | if (recursing) { | |
517 | warning("recursive delta detected for object %s", | |
e6a492b7 | 518 | oid_to_hex(&e->idx.oid)); |
f63c79db JH |
519 | return WRITE_ONE_RECURSIVE; |
520 | } else if (e->idx.offset || e->preferred_base) { | |
521 | /* offset is non zero if object is written already. */ | |
522 | return WRITE_ONE_SKIP; | |
523 | } | |
d7dd0223 | 524 | |
720c9f7b | 525 | /* if we are deltified, write out base object first. */ |
f63c79db JH |
526 | if (e->delta) { |
527 | e->idx.offset = 1; /* now recurse */ | |
528 | switch (write_one(f, e->delta, offset)) { | |
529 | case WRITE_ONE_RECURSIVE: | |
530 | /* we cannot depend on this one */ | |
531 | e->delta = NULL; | |
532 | break; | |
533 | default: | |
534 | break; | |
535 | case WRITE_ONE_BREAK: | |
536 | e->idx.offset = recursing; | |
537 | return WRITE_ONE_BREAK; | |
538 | } | |
539 | } | |
d7dd0223 | 540 | |
6ed7f25e NP |
541 | e->idx.offset = *offset; |
542 | size = write_object(f, e, *offset); | |
17b08f2c | 543 | if (!size) { |
f63c79db JH |
544 | e->idx.offset = recursing; |
545 | return WRITE_ONE_BREAK; | |
17b08f2c | 546 | } |
79814f42 | 547 | written_list[nr_written++] = &e->idx; |
d7dd0223 NP |
548 | |
549 | /* make sure off_t is sufficiently large not to wrap */ | |
c03c8315 | 550 | if (signed_add_overflows(*offset, size)) |
d7dd0223 | 551 | die("pack too large for current definition of off_t"); |
6ed7f25e | 552 | *offset += size; |
f63c79db | 553 | return WRITE_ONE_WRITTEN; |
9d5ab962 JH |
554 | } |
555 | ||
d155254c | 556 | static int mark_tagged(const char *path, const struct object_id *oid, int flag, |
1b4bb16b JH |
557 | void *cb_data) |
558 | { | |
559 | unsigned char peeled[20]; | |
d155254c | 560 | struct object_entry *entry = packlist_find(&to_pack, oid->hash, NULL); |
1b4bb16b JH |
561 | |
562 | if (entry) | |
563 | entry->tagged = 1; | |
564 | if (!peel_ref(path, peeled)) { | |
2834bc27 | 565 | entry = packlist_find(&to_pack, peeled, NULL); |
1b4bb16b JH |
566 | if (entry) |
567 | entry->tagged = 1; | |
568 | } | |
569 | return 0; | |
570 | } | |
571 | ||
be126818 | 572 | static inline void add_to_write_order(struct object_entry **wo, |
92bef1a1 | 573 | unsigned int *endp, |
1b4bb16b JH |
574 | struct object_entry *e) |
575 | { | |
576 | if (e->filled) | |
577 | return; | |
578 | wo[(*endp)++] = e; | |
579 | e->filled = 1; | |
580 | } | |
581 | ||
582 | static void add_descendants_to_write_order(struct object_entry **wo, | |
92bef1a1 | 583 | unsigned int *endp, |
1b4bb16b JH |
584 | struct object_entry *e) |
585 | { | |
f380872f DM |
586 | int add_to_order = 1; |
587 | while (e) { | |
588 | if (add_to_order) { | |
589 | struct object_entry *s; | |
590 | /* add this node... */ | |
591 | add_to_write_order(wo, endp, e); | |
592 | /* all its siblings... */ | |
593 | for (s = e->delta_sibling; s; s = s->delta_sibling) { | |
594 | add_to_write_order(wo, endp, s); | |
595 | } | |
596 | } | |
597 | /* drop down a level to add left subtree nodes if possible */ | |
598 | if (e->delta_child) { | |
599 | add_to_order = 1; | |
600 | e = e->delta_child; | |
601 | } else { | |
602 | add_to_order = 0; | |
603 | /* our sibling might have some children, it is next */ | |
604 | if (e->delta_sibling) { | |
605 | e = e->delta_sibling; | |
606 | continue; | |
607 | } | |
608 | /* go back to our parent node */ | |
609 | e = e->delta; | |
610 | while (e && !e->delta_sibling) { | |
611 | /* we're on the right side of a subtree, keep | |
612 | * going up until we can go right again */ | |
613 | e = e->delta; | |
614 | } | |
615 | if (!e) { | |
616 | /* done- we hit our original root node */ | |
617 | return; | |
618 | } | |
619 | /* pass it off to sibling at this level */ | |
620 | e = e->delta_sibling; | |
621 | } | |
622 | }; | |
1b4bb16b JH |
623 | } |
624 | ||
625 | static void add_family_to_write_order(struct object_entry **wo, | |
92bef1a1 | 626 | unsigned int *endp, |
1b4bb16b JH |
627 | struct object_entry *e) |
628 | { | |
629 | struct object_entry *root; | |
630 | ||
631 | for (root = e; root->delta; root = root->delta) | |
632 | ; /* nothing */ | |
1b4bb16b JH |
633 | add_descendants_to_write_order(wo, endp, root); |
634 | } | |
635 | ||
636 | static struct object_entry **compute_write_order(void) | |
637 | { | |
38d4debb | 638 | unsigned int i, wo_end, last_untagged; |
1b4bb16b | 639 | |
b32fa95f | 640 | struct object_entry **wo; |
2834bc27 | 641 | struct object_entry *objects = to_pack.objects; |
1b4bb16b | 642 | |
2834bc27 | 643 | for (i = 0; i < to_pack.nr_objects; i++) { |
1b4bb16b JH |
644 | objects[i].tagged = 0; |
645 | objects[i].filled = 0; | |
646 | objects[i].delta_child = NULL; | |
647 | objects[i].delta_sibling = NULL; | |
648 | } | |
649 | ||
650 | /* | |
651 | * Fully connect delta_child/delta_sibling network. | |
652 | * Make sure delta_sibling is sorted in the original | |
653 | * recency order. | |
654 | */ | |
2834bc27 | 655 | for (i = to_pack.nr_objects; i > 0;) { |
92bef1a1 | 656 | struct object_entry *e = &objects[--i]; |
1b4bb16b JH |
657 | if (!e->delta) |
658 | continue; | |
659 | /* Mark me as the first child */ | |
660 | e->delta_sibling = e->delta->delta_child; | |
661 | e->delta->delta_child = e; | |
662 | } | |
663 | ||
664 | /* | |
665 | * Mark objects that are at the tip of tags. | |
666 | */ | |
d155254c | 667 | for_each_tag_ref(mark_tagged, NULL); |
1b4bb16b JH |
668 | |
669 | /* | |
38d4debb | 670 | * Give the objects in the original recency order until |
1b4bb16b JH |
671 | * we see a tagged tip. |
672 | */ | |
b32fa95f | 673 | ALLOC_ARRAY(wo, to_pack.nr_objects); |
2834bc27 | 674 | for (i = wo_end = 0; i < to_pack.nr_objects; i++) { |
1b4bb16b JH |
675 | if (objects[i].tagged) |
676 | break; | |
677 | add_to_write_order(wo, &wo_end, &objects[i]); | |
678 | } | |
38d4debb | 679 | last_untagged = i; |
1b4bb16b JH |
680 | |
681 | /* | |
682 | * Then fill all the tagged tips. | |
683 | */ | |
2834bc27 | 684 | for (; i < to_pack.nr_objects; i++) { |
1b4bb16b JH |
685 | if (objects[i].tagged) |
686 | add_to_write_order(wo, &wo_end, &objects[i]); | |
687 | } | |
688 | ||
689 | /* | |
690 | * And then all remaining commits and tags. | |
691 | */ | |
2834bc27 | 692 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
1b4bb16b JH |
693 | if (objects[i].type != OBJ_COMMIT && |
694 | objects[i].type != OBJ_TAG) | |
695 | continue; | |
696 | add_to_write_order(wo, &wo_end, &objects[i]); | |
697 | } | |
698 | ||
699 | /* | |
700 | * And then all the trees. | |
701 | */ | |
2834bc27 | 702 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
1b4bb16b JH |
703 | if (objects[i].type != OBJ_TREE) |
704 | continue; | |
705 | add_to_write_order(wo, &wo_end, &objects[i]); | |
706 | } | |
707 | ||
708 | /* | |
709 | * Finally all the rest in really tight order | |
710 | */ | |
2834bc27 | 711 | for (i = last_untagged; i < to_pack.nr_objects; i++) { |
38d4debb DM |
712 | if (!objects[i].filled) |
713 | add_family_to_write_order(wo, &wo_end, &objects[i]); | |
714 | } | |
715 | ||
2834bc27 VM |
716 | if (wo_end != to_pack.nr_objects) |
717 | die("ordered %u objects, expected %"PRIu32, wo_end, to_pack.nr_objects); | |
1b4bb16b JH |
718 | |
719 | return wo; | |
720 | } | |
721 | ||
6b8fda2d VM |
722 | static off_t write_reused_pack(struct sha1file *f) |
723 | { | |
724 | unsigned char buffer[8192]; | |
657673f1 | 725 | off_t to_write, total; |
6b8fda2d VM |
726 | int fd; |
727 | ||
728 | if (!is_pack_valid(reuse_packfile)) | |
729 | die("packfile is invalid: %s", reuse_packfile->pack_name); | |
730 | ||
a5436b57 | 731 | fd = git_open(reuse_packfile->pack_name); |
6b8fda2d VM |
732 | if (fd < 0) |
733 | die_errno("unable to open packfile for reuse: %s", | |
734 | reuse_packfile->pack_name); | |
735 | ||
736 | if (lseek(fd, sizeof(struct pack_header), SEEK_SET) == -1) | |
737 | die_errno("unable to seek in reused packfile"); | |
738 | ||
739 | if (reuse_packfile_offset < 0) | |
740 | reuse_packfile_offset = reuse_packfile->pack_size - 20; | |
741 | ||
657673f1 | 742 | total = to_write = reuse_packfile_offset - sizeof(struct pack_header); |
6b8fda2d VM |
743 | |
744 | while (to_write) { | |
745 | int read_pack = xread(fd, buffer, sizeof(buffer)); | |
746 | ||
747 | if (read_pack <= 0) | |
748 | die_errno("unable to read from reused packfile"); | |
749 | ||
750 | if (read_pack > to_write) | |
751 | read_pack = to_write; | |
752 | ||
753 | sha1write(f, buffer, read_pack); | |
754 | to_write -= read_pack; | |
657673f1 JK |
755 | |
756 | /* | |
757 | * We don't know the actual number of objects written, | |
758 | * only how many bytes written, how many bytes total, and | |
759 | * how many objects total. So we can fake it by pretending all | |
760 | * objects we are writing are the same size. This gives us a | |
761 | * smooth progress meter, and at the end it matches the true | |
762 | * answer. | |
763 | */ | |
764 | written = reuse_packfile_objects * | |
765 | (((double)(total - to_write)) / total); | |
766 | display_progress(progress_state, written); | |
6b8fda2d VM |
767 | } |
768 | ||
769 | close(fd); | |
657673f1 JK |
770 | written = reuse_packfile_objects; |
771 | display_progress(progress_state, written); | |
6b8fda2d VM |
772 | return reuse_packfile_offset - sizeof(struct pack_header); |
773 | } | |
774 | ||
9cea46cd EW |
775 | static const char no_split_warning[] = N_( |
776 | "disabling bitmap writing, packs are split due to pack.packSizeLimit" | |
777 | ); | |
778 | ||
d01fb92f | 779 | static void write_pack_file(void) |
c323ac7d | 780 | { |
ebe27b13 | 781 | uint32_t i = 0, j; |
d22b9290 | 782 | struct sha1file *f; |
6ed7f25e | 783 | off_t offset; |
ebe27b13 | 784 | uint32_t nr_remaining = nr_result; |
f746bae8 | 785 | time_t last_mtime = 0; |
1b4bb16b | 786 | struct object_entry **write_order; |
81a216a5 | 787 | |
bcd7954e | 788 | if (progress > pack_to_stdout) |
754dbc43 | 789 | progress_state = start_progress(_("Writing objects"), nr_result); |
b32fa95f | 790 | ALLOC_ARRAY(written_list, to_pack.nr_objects); |
1b4bb16b | 791 | write_order = compute_write_order(); |
183bdb2c | 792 | |
ebe27b13 | 793 | do { |
aa7e44bf | 794 | unsigned char sha1[20]; |
7ba502c4 | 795 | char *pack_tmp_name = NULL; |
aa7e44bf | 796 | |
cdf9db3c | 797 | if (pack_to_stdout) |
2a128d63 | 798 | f = sha1fd_throughput(1, "<stdout>", progress_state); |
cdf9db3c JH |
799 | else |
800 | f = create_tmp_packfile(&pack_tmp_name); | |
ebe27b13 | 801 | |
c0ad4657 | 802 | offset = write_pack_header(f, nr_remaining); |
6b8fda2d VM |
803 | |
804 | if (reuse_packfile) { | |
805 | off_t packfile_size; | |
806 | assert(pack_to_stdout); | |
807 | ||
808 | packfile_size = write_reused_pack(f); | |
809 | offset += packfile_size; | |
810 | } | |
811 | ||
ebe27b13 | 812 | nr_written = 0; |
2834bc27 | 813 | for (; i < to_pack.nr_objects; i++) { |
1b4bb16b | 814 | struct object_entry *e = write_order[i]; |
cddec4f8 | 815 | if (write_one(f, e, &offset) == WRITE_ONE_BREAK) |
720c9f7b NP |
816 | break; |
817 | display_progress(progress_state, written); | |
818 | } | |
c553ca25 | 819 | |
ebe27b13 DH |
820 | /* |
821 | * Did we write the wrong # entries in the header? | |
822 | * If so, rewrite it like in fast-import | |
823 | */ | |
54352bb2 LT |
824 | if (pack_to_stdout) { |
825 | sha1close(f, sha1, CSUM_CLOSE); | |
826 | } else if (nr_written == nr_remaining) { | |
827 | sha1close(f, sha1, CSUM_FSYNC); | |
ebe27b13 | 828 | } else { |
ac0463ed | 829 | int fd = sha1close(f, sha1, 0); |
abeb40e5 | 830 | fixup_pack_header_footer(fd, sha1, pack_tmp_name, |
ac0463ed | 831 | nr_written, sha1, offset); |
7ba502c4 | 832 | close(fd); |
9cea46cd EW |
833 | if (write_bitmap_index) { |
834 | warning(_(no_split_warning)); | |
835 | write_bitmap_index = 0; | |
836 | } | |
ebe27b13 DH |
837 | } |
838 | ||
839 | if (!pack_to_stdout) { | |
f746bae8 | 840 | struct stat st; |
58892711 | 841 | struct strbuf tmpname = STRBUF_INIT; |
d01fb92f | 842 | |
f746bae8 NP |
843 | /* |
844 | * Packs are runtime accessed in their mtime | |
845 | * order since newer packs are more likely to contain | |
846 | * younger objects. So if we are creating multiple | |
847 | * packs then we should modify the mtime of later ones | |
848 | * to preserve this property. | |
849 | */ | |
0e990530 | 850 | if (stat(pack_tmp_name, &st) < 0) { |
54d47394 | 851 | warning_errno("failed to stat %s", pack_tmp_name); |
f746bae8 NP |
852 | } else if (!last_mtime) { |
853 | last_mtime = st.st_mtime; | |
854 | } else { | |
855 | struct utimbuf utb; | |
856 | utb.actime = st.st_atime; | |
857 | utb.modtime = --last_mtime; | |
0e990530 | 858 | if (utime(pack_tmp_name, &utb) < 0) |
54d47394 | 859 | warning_errno("failed utime() on %s", pack_tmp_name); |
f746bae8 NP |
860 | } |
861 | ||
58892711 | 862 | strbuf_addf(&tmpname, "%s-", base_name); |
7cc8f971 VM |
863 | |
864 | if (write_bitmap_index) { | |
865 | bitmap_writer_set_checksum(sha1); | |
866 | bitmap_writer_build_type_index(written_list, nr_written); | |
867 | } | |
868 | ||
58892711 | 869 | finish_tmp_packfile(&tmpname, pack_tmp_name, |
0e990530 JH |
870 | written_list, nr_written, |
871 | &pack_idx_opts, sha1); | |
7cc8f971 VM |
872 | |
873 | if (write_bitmap_index) { | |
58892711 | 874 | strbuf_addf(&tmpname, "%s.bitmap", sha1_to_hex(sha1)); |
7cc8f971 VM |
875 | |
876 | stop_progress(&progress_state); | |
877 | ||
878 | bitmap_writer_show_progress(progress); | |
879 | bitmap_writer_reuse_bitmaps(&to_pack); | |
880 | bitmap_writer_select_commits(indexed_commits, indexed_commits_nr, -1); | |
881 | bitmap_writer_build(&to_pack); | |
ae4f07fb | 882 | bitmap_writer_finish(written_list, nr_written, |
58892711 | 883 | tmpname.buf, write_bitmap_options); |
7cc8f971 VM |
884 | write_bitmap_index = 0; |
885 | } | |
886 | ||
58892711 | 887 | strbuf_release(&tmpname); |
7ba502c4 | 888 | free(pack_tmp_name); |
aa7e44bf | 889 | puts(sha1_to_hex(sha1)); |
ebe27b13 DH |
890 | } |
891 | ||
892 | /* mark written objects as written to previous pack */ | |
893 | for (j = 0; j < nr_written; j++) { | |
79814f42 | 894 | written_list[j]->offset = (off_t)-1; |
ebe27b13 DH |
895 | } |
896 | nr_remaining -= nr_written; | |
2834bc27 | 897 | } while (nr_remaining && i < to_pack.nr_objects); |
ebe27b13 DH |
898 | |
899 | free(written_list); | |
1b4bb16b | 900 | free(write_order); |
4d4fcc54 | 901 | stop_progress(&progress_state); |
67c08ce1 | 902 | if (written != nr_result) |
6e1c2344 RJ |
903 | die("wrote %"PRIu32" objects while expecting %"PRIu32, |
904 | written, nr_result); | |
c323ac7d LT |
905 | } |
906 | ||
a74db82e JH |
907 | static int no_try_delta(const char *path) |
908 | { | |
2aef63d3 | 909 | static struct attr_check *check; |
a74db82e | 910 | |
2aef63d3 JH |
911 | if (!check) |
912 | check = attr_check_initl("delta", NULL); | |
913 | if (git_check_attr(path, check)) | |
a74db82e | 914 | return 0; |
2aef63d3 | 915 | if (ATTR_FALSE(check->items[0].value)) |
a74db82e JH |
916 | return 1; |
917 | return 0; | |
918 | } | |
919 | ||
ce2bc424 JK |
920 | /* |
921 | * When adding an object, check whether we have already added it | |
922 | * to our packing list. If so, we can skip. However, if we are | |
923 | * being asked to excludei t, but the previous mention was to include | |
924 | * it, make sure to adjust its flags and tweak our numbers accordingly. | |
925 | * | |
926 | * As an optimization, we pass out the index position where we would have | |
927 | * found the item, since that saves us from having to look it up again a | |
928 | * few lines later when we want to add the new entry. | |
929 | */ | |
930 | static int have_duplicate_entry(const unsigned char *sha1, | |
931 | int exclude, | |
932 | uint32_t *index_pos) | |
c323ac7d | 933 | { |
c323ac7d | 934 | struct object_entry *entry; |
29b734e4 | 935 | |
ce2bc424 JK |
936 | entry = packlist_find(&to_pack, sha1, index_pos); |
937 | if (!entry) | |
29b734e4 | 938 | return 0; |
ce2bc424 JK |
939 | |
940 | if (exclude) { | |
941 | if (!entry->preferred_base) | |
942 | nr_result--; | |
943 | entry->preferred_base = 1; | |
29b734e4 | 944 | } |
c323ac7d | 945 | |
ce2bc424 JK |
946 | return 1; |
947 | } | |
948 | ||
702d1b95 KS |
949 | static int want_found_object(int exclude, struct packed_git *p) |
950 | { | |
951 | if (exclude) | |
952 | return 1; | |
953 | if (incremental) | |
954 | return 0; | |
955 | ||
956 | /* | |
957 | * When asked to do --local (do not include an object that appears in a | |
958 | * pack we borrow from elsewhere) or --honor-pack-keep (do not include | |
959 | * an object that appears in a pack marked with .keep), finding a pack | |
960 | * that matches the criteria is sufficient for us to decide to omit it. | |
961 | * However, even if this pack does not satisfy the criteria, we need to | |
962 | * make sure no copy of this object appears in _any_ pack that makes us | |
963 | * to omit the object, so we need to check all the packs. | |
964 | * | |
965 | * We can however first check whether these options can possible matter; | |
966 | * if they do not matter we know we want the object in generated pack. | |
967 | * Otherwise, we signal "-1" at the end to tell the caller that we do | |
968 | * not know either way, and it needs to check more packs. | |
969 | */ | |
970 | if (!ignore_packed_keep && | |
971 | (!local || !have_non_local_packs)) | |
972 | return 1; | |
973 | ||
974 | if (local && !p->pack_local) | |
975 | return 0; | |
976 | if (ignore_packed_keep && p->pack_local && p->pack_keep) | |
977 | return 0; | |
978 | ||
979 | /* we don't know yet; keep looking for more packs */ | |
980 | return -1; | |
981 | } | |
982 | ||
ce2bc424 JK |
983 | /* |
984 | * Check whether we want the object in the pack (e.g., we do not want | |
985 | * objects found in non-local stores if the "--local" option was used). | |
986 | * | |
702d1b95 KS |
987 | * If the caller already knows an existing pack it wants to take the object |
988 | * from, that is passed in *found_pack and *found_offset; otherwise this | |
989 | * function finds if there is any pack that has the object and returns the pack | |
990 | * and its offset in these variables. | |
ce2bc424 JK |
991 | */ |
992 | static int want_object_in_pack(const unsigned char *sha1, | |
993 | int exclude, | |
994 | struct packed_git **found_pack, | |
995 | off_t *found_offset) | |
996 | { | |
c9af708b | 997 | struct mru_entry *entry; |
702d1b95 | 998 | int want; |
ce2bc424 | 999 | |
daae0625 BC |
1000 | if (!exclude && local && has_loose_object_nonlocal(sha1)) |
1001 | return 0; | |
1002 | ||
702d1b95 KS |
1003 | /* |
1004 | * If we already know the pack object lives in, start checks from that | |
1005 | * pack - in the usual case when neither --local was given nor .keep files | |
1006 | * are present we will determine the answer right now. | |
1007 | */ | |
1008 | if (*found_pack) { | |
1009 | want = want_found_object(exclude, *found_pack); | |
1010 | if (want != -1) | |
1011 | return want; | |
1012 | } | |
ce2bc424 | 1013 | |
c9af708b JK |
1014 | for (entry = packed_git_mru->head; entry; entry = entry->next) { |
1015 | struct packed_git *p = entry->item; | |
702d1b95 KS |
1016 | off_t offset; |
1017 | ||
1018 | if (p == *found_pack) | |
1019 | offset = *found_offset; | |
1020 | else | |
1021 | offset = find_pack_entry_one(sha1, p); | |
1022 | ||
5c49c116 | 1023 | if (offset) { |
ce2bc424 | 1024 | if (!*found_pack) { |
319b678a | 1025 | if (!is_pack_valid(p)) |
4c080182 | 1026 | continue; |
ce2bc424 JK |
1027 | *found_offset = offset; |
1028 | *found_pack = p; | |
64560374 | 1029 | } |
702d1b95 | 1030 | want = want_found_object(exclude, p); |
e6e24c94 | 1031 | if (!exclude && want > 0) |
c9af708b | 1032 | mru_mark(packed_git_mru, entry); |
702d1b95 KS |
1033 | if (want != -1) |
1034 | return want; | |
64560374 LT |
1035 | } |
1036 | } | |
eb019375 | 1037 | |
ce2bc424 JK |
1038 | return 1; |
1039 | } | |
1040 | ||
1041 | static void create_object_entry(const unsigned char *sha1, | |
1042 | enum object_type type, | |
1043 | uint32_t hash, | |
1044 | int exclude, | |
1045 | int no_try_delta, | |
1046 | uint32_t index_pos, | |
1047 | struct packed_git *found_pack, | |
1048 | off_t found_offset) | |
1049 | { | |
1050 | struct object_entry *entry; | |
29b734e4 | 1051 | |
2834bc27 | 1052 | entry = packlist_alloc(&to_pack, sha1, index_pos); |
27225f2e | 1053 | entry->hash = hash; |
5c49c116 NP |
1054 | if (type) |
1055 | entry->type = type; | |
29b734e4 NP |
1056 | if (exclude) |
1057 | entry->preferred_base = 1; | |
81a216a5 NP |
1058 | else |
1059 | nr_result++; | |
29b734e4 NP |
1060 | if (found_pack) { |
1061 | entry->in_pack = found_pack; | |
1062 | entry->in_pack_offset = found_offset; | |
1063 | } | |
7a979d99 | 1064 | |
ce2bc424 JK |
1065 | entry->no_try_delta = no_try_delta; |
1066 | } | |
29b734e4 | 1067 | |
373c67da JK |
1068 | static const char no_closure_warning[] = N_( |
1069 | "disabling bitmap writing, as some objects are not being packed" | |
1070 | ); | |
1071 | ||
ce2bc424 JK |
1072 | static int add_object_entry(const unsigned char *sha1, enum object_type type, |
1073 | const char *name, int exclude) | |
1074 | { | |
702d1b95 KS |
1075 | struct packed_git *found_pack = NULL; |
1076 | off_t found_offset = 0; | |
ce2bc424 | 1077 | uint32_t index_pos; |
7a979d99 | 1078 | |
ce2bc424 JK |
1079 | if (have_duplicate_entry(sha1, exclude, &index_pos)) |
1080 | return 0; | |
a74db82e | 1081 | |
373c67da JK |
1082 | if (!want_object_in_pack(sha1, exclude, &found_pack, &found_offset)) { |
1083 | /* The pack is missing an object, so it will not have closure */ | |
1084 | if (write_bitmap_index) { | |
1085 | warning(_(no_closure_warning)); | |
1086 | write_bitmap_index = 0; | |
1087 | } | |
ce2bc424 | 1088 | return 0; |
373c67da | 1089 | } |
29b734e4 | 1090 | |
ce2bc424 JK |
1091 | create_object_entry(sha1, type, pack_name_hash(name), |
1092 | exclude, name && no_try_delta(name), | |
1093 | index_pos, found_pack, found_offset); | |
a74db82e | 1094 | |
78d2214e | 1095 | display_progress(progress_state, nr_result); |
29b734e4 | 1096 | return 1; |
c323ac7d LT |
1097 | } |
1098 | ||
6b8fda2d VM |
1099 | static int add_object_entry_from_bitmap(const unsigned char *sha1, |
1100 | enum object_type type, | |
1101 | int flags, uint32_t name_hash, | |
1102 | struct packed_git *pack, off_t offset) | |
1103 | { | |
1104 | uint32_t index_pos; | |
1105 | ||
1106 | if (have_duplicate_entry(sha1, 0, &index_pos)) | |
1107 | return 0; | |
1108 | ||
702d1b95 KS |
1109 | if (!want_object_in_pack(sha1, 0, &pack, &offset)) |
1110 | return 0; | |
1111 | ||
6b8fda2d VM |
1112 | create_object_entry(sha1, type, name_hash, 0, 0, index_pos, pack, offset); |
1113 | ||
78d2214e | 1114 | display_progress(progress_state, nr_result); |
29b734e4 | 1115 | return 1; |
c323ac7d LT |
1116 | } |
1117 | ||
5379a5c5 JH |
1118 | struct pbase_tree_cache { |
1119 | unsigned char sha1[20]; | |
1120 | int ref; | |
1121 | int temporary; | |
1122 | void *tree_data; | |
1123 | unsigned long tree_size; | |
1124 | }; | |
1125 | ||
1126 | static struct pbase_tree_cache *(pbase_tree_cache[256]); | |
1127 | static int pbase_tree_cache_ix(const unsigned char *sha1) | |
1128 | { | |
1129 | return sha1[0] % ARRAY_SIZE(pbase_tree_cache); | |
1130 | } | |
1131 | static int pbase_tree_cache_ix_incr(int ix) | |
1132 | { | |
1133 | return (ix+1) % ARRAY_SIZE(pbase_tree_cache); | |
1134 | } | |
1135 | ||
1136 | static struct pbase_tree { | |
1137 | struct pbase_tree *next; | |
1138 | /* This is a phony "cache" entry; we are not | |
01689909 | 1139 | * going to evict it or find it through _get() |
5379a5c5 JH |
1140 | * mechanism -- this is for the toplevel node that |
1141 | * would almost always change with any commit. | |
1142 | */ | |
1143 | struct pbase_tree_cache pcache; | |
1144 | } *pbase_tree; | |
1145 | ||
1146 | static struct pbase_tree_cache *pbase_tree_get(const unsigned char *sha1) | |
1147 | { | |
1148 | struct pbase_tree_cache *ent, *nent; | |
1149 | void *data; | |
1150 | unsigned long size; | |
21666f1a | 1151 | enum object_type type; |
5379a5c5 JH |
1152 | int neigh; |
1153 | int my_ix = pbase_tree_cache_ix(sha1); | |
1154 | int available_ix = -1; | |
1155 | ||
1156 | /* pbase-tree-cache acts as a limited hashtable. | |
1157 | * your object will be found at your index or within a few | |
1158 | * slots after that slot if it is cached. | |
1159 | */ | |
1160 | for (neigh = 0; neigh < 8; neigh++) { | |
1161 | ent = pbase_tree_cache[my_ix]; | |
a89fccd2 | 1162 | if (ent && !hashcmp(ent->sha1, sha1)) { |
5379a5c5 JH |
1163 | ent->ref++; |
1164 | return ent; | |
1165 | } | |
1166 | else if (((available_ix < 0) && (!ent || !ent->ref)) || | |
1167 | ((0 <= available_ix) && | |
1168 | (!ent && pbase_tree_cache[available_ix]))) | |
1169 | available_ix = my_ix; | |
1170 | if (!ent) | |
1171 | break; | |
1172 | my_ix = pbase_tree_cache_ix_incr(my_ix); | |
1173 | } | |
1174 | ||
1175 | /* Did not find one. Either we got a bogus request or | |
1176 | * we need to read and perhaps cache. | |
1177 | */ | |
21666f1a | 1178 | data = read_sha1_file(sha1, &type, &size); |
5379a5c5 JH |
1179 | if (!data) |
1180 | return NULL; | |
21666f1a | 1181 | if (type != OBJ_TREE) { |
5379a5c5 JH |
1182 | free(data); |
1183 | return NULL; | |
1184 | } | |
1185 | ||
1186 | /* We need to either cache or return a throwaway copy */ | |
1187 | ||
1188 | if (available_ix < 0) | |
1189 | ent = NULL; | |
1190 | else { | |
1191 | ent = pbase_tree_cache[available_ix]; | |
1192 | my_ix = available_ix; | |
1193 | } | |
1194 | ||
1195 | if (!ent) { | |
1196 | nent = xmalloc(sizeof(*nent)); | |
1197 | nent->temporary = (available_ix < 0); | |
1198 | } | |
1199 | else { | |
1200 | /* evict and reuse */ | |
1201 | free(ent->tree_data); | |
1202 | nent = ent; | |
1203 | } | |
e702496e | 1204 | hashcpy(nent->sha1, sha1); |
5379a5c5 JH |
1205 | nent->tree_data = data; |
1206 | nent->tree_size = size; | |
1207 | nent->ref = 1; | |
1208 | if (!nent->temporary) | |
1209 | pbase_tree_cache[my_ix] = nent; | |
1210 | return nent; | |
1211 | } | |
1212 | ||
1213 | static void pbase_tree_put(struct pbase_tree_cache *cache) | |
1214 | { | |
1215 | if (!cache->temporary) { | |
1216 | cache->ref--; | |
1217 | return; | |
1218 | } | |
1219 | free(cache->tree_data); | |
1220 | free(cache); | |
1221 | } | |
1222 | ||
1223 | static int name_cmp_len(const char *name) | |
1224 | { | |
1225 | int i; | |
1226 | for (i = 0; name[i] && name[i] != '\n' && name[i] != '/'; i++) | |
1227 | ; | |
1228 | return i; | |
1229 | } | |
1230 | ||
1231 | static void add_pbase_object(struct tree_desc *tree, | |
5379a5c5 | 1232 | const char *name, |
ce0bd642 LT |
1233 | int cmplen, |
1234 | const char *fullname) | |
3f9ac8d2 | 1235 | { |
4c068a98 | 1236 | struct name_entry entry; |
8a5a8d6c | 1237 | int cmp; |
4c068a98 LT |
1238 | |
1239 | while (tree_entry(tree,&entry)) { | |
1211be6b LT |
1240 | if (S_ISGITLINK(entry.mode)) |
1241 | continue; | |
0de16337 | 1242 | cmp = tree_entry_len(&entry) != cmplen ? 1 : |
8a5a8d6c NP |
1243 | memcmp(name, entry.path, cmplen); |
1244 | if (cmp > 0) | |
7a979d99 | 1245 | continue; |
8a5a8d6c NP |
1246 | if (cmp < 0) |
1247 | return; | |
5379a5c5 | 1248 | if (name[cmplen] != '/') { |
7d924c91 | 1249 | add_object_entry(entry.oid->hash, |
4d1012c3 | 1250 | object_type(entry.mode), |
bc32fed5 | 1251 | fullname, 1); |
5379a5c5 JH |
1252 | return; |
1253 | } | |
8a5a8d6c | 1254 | if (S_ISDIR(entry.mode)) { |
7a979d99 | 1255 | struct tree_desc sub; |
5379a5c5 JH |
1256 | struct pbase_tree_cache *tree; |
1257 | const char *down = name+cmplen+1; | |
1258 | int downlen = name_cmp_len(down); | |
1259 | ||
7d924c91 | 1260 | tree = pbase_tree_get(entry.oid->hash); |
5379a5c5 JH |
1261 | if (!tree) |
1262 | return; | |
6fda5e51 | 1263 | init_tree_desc(&sub, tree->tree_data, tree->tree_size); |
5379a5c5 | 1264 | |
ce0bd642 | 1265 | add_pbase_object(&sub, down, downlen, fullname); |
5379a5c5 JH |
1266 | pbase_tree_put(tree); |
1267 | } | |
1268 | } | |
1269 | } | |
1d6b38cc | 1270 | |
5379a5c5 JH |
1271 | static unsigned *done_pbase_paths; |
1272 | static int done_pbase_paths_num; | |
1273 | static int done_pbase_paths_alloc; | |
1274 | static int done_pbase_path_pos(unsigned hash) | |
1275 | { | |
1276 | int lo = 0; | |
1277 | int hi = done_pbase_paths_num; | |
1278 | while (lo < hi) { | |
1279 | int mi = (hi + lo) / 2; | |
1280 | if (done_pbase_paths[mi] == hash) | |
1281 | return mi; | |
1282 | if (done_pbase_paths[mi] < hash) | |
1283 | hi = mi; | |
1284 | else | |
1285 | lo = mi + 1; | |
1286 | } | |
1287 | return -lo-1; | |
1288 | } | |
1289 | ||
1290 | static int check_pbase_path(unsigned hash) | |
1291 | { | |
1292 | int pos = (!done_pbase_paths) ? -1 : done_pbase_path_pos(hash); | |
1293 | if (0 <= pos) | |
1294 | return 1; | |
1295 | pos = -pos - 1; | |
25e19407 DD |
1296 | ALLOC_GROW(done_pbase_paths, |
1297 | done_pbase_paths_num + 1, | |
1298 | done_pbase_paths_alloc); | |
5379a5c5 JH |
1299 | done_pbase_paths_num++; |
1300 | if (pos < done_pbase_paths_num) | |
1301 | memmove(done_pbase_paths + pos + 1, | |
1302 | done_pbase_paths + pos, | |
1303 | (done_pbase_paths_num - pos - 1) * sizeof(unsigned)); | |
1304 | done_pbase_paths[pos] = hash; | |
1305 | return 0; | |
1306 | } | |
1307 | ||
bc32fed5 | 1308 | static void add_preferred_base_object(const char *name) |
5379a5c5 JH |
1309 | { |
1310 | struct pbase_tree *it; | |
8a5a8d6c | 1311 | int cmplen; |
68fb36eb | 1312 | unsigned hash = pack_name_hash(name); |
5379a5c5 | 1313 | |
8a5a8d6c | 1314 | if (!num_preferred_base || check_pbase_path(hash)) |
5379a5c5 JH |
1315 | return; |
1316 | ||
8a5a8d6c | 1317 | cmplen = name_cmp_len(name); |
5379a5c5 JH |
1318 | for (it = pbase_tree; it; it = it->next) { |
1319 | if (cmplen == 0) { | |
bc32fed5 | 1320 | add_object_entry(it->pcache.sha1, OBJ_TREE, NULL, 1); |
5379a5c5 JH |
1321 | } |
1322 | else { | |
1323 | struct tree_desc tree; | |
6fda5e51 | 1324 | init_tree_desc(&tree, it->pcache.tree_data, it->pcache.tree_size); |
ce0bd642 | 1325 | add_pbase_object(&tree, name, cmplen, name); |
7a979d99 | 1326 | } |
3f9ac8d2 | 1327 | } |
3f9ac8d2 JH |
1328 | } |
1329 | ||
7a979d99 | 1330 | static void add_preferred_base(unsigned char *sha1) |
3f9ac8d2 | 1331 | { |
5379a5c5 JH |
1332 | struct pbase_tree *it; |
1333 | void *data; | |
1334 | unsigned long size; | |
1335 | unsigned char tree_sha1[20]; | |
1d6b38cc | 1336 | |
8d1d8f83 JH |
1337 | if (window <= num_preferred_base++) |
1338 | return; | |
1339 | ||
5379a5c5 JH |
1340 | data = read_object_with_reference(sha1, tree_type, &size, tree_sha1); |
1341 | if (!data) | |
7a979d99 | 1342 | return; |
5379a5c5 JH |
1343 | |
1344 | for (it = pbase_tree; it; it = it->next) { | |
a89fccd2 | 1345 | if (!hashcmp(it->pcache.sha1, tree_sha1)) { |
5379a5c5 JH |
1346 | free(data); |
1347 | return; | |
1348 | } | |
1349 | } | |
1350 | ||
1351 | it = xcalloc(1, sizeof(*it)); | |
1352 | it->next = pbase_tree; | |
1353 | pbase_tree = it; | |
1354 | ||
e702496e | 1355 | hashcpy(it->pcache.sha1, tree_sha1); |
5379a5c5 JH |
1356 | it->pcache.tree_data = data; |
1357 | it->pcache.tree_size = size; | |
3f9ac8d2 JH |
1358 | } |
1359 | ||
0ef95f72 NP |
1360 | static void cleanup_preferred_base(void) |
1361 | { | |
1362 | struct pbase_tree *it; | |
1363 | unsigned i; | |
1364 | ||
1365 | it = pbase_tree; | |
1366 | pbase_tree = NULL; | |
1367 | while (it) { | |
1368 | struct pbase_tree *this = it; | |
1369 | it = this->next; | |
1370 | free(this->pcache.tree_data); | |
1371 | free(this); | |
1372 | } | |
1373 | ||
1374 | for (i = 0; i < ARRAY_SIZE(pbase_tree_cache); i++) { | |
1375 | if (!pbase_tree_cache[i]) | |
1376 | continue; | |
1377 | free(pbase_tree_cache[i]->tree_data); | |
1378 | free(pbase_tree_cache[i]); | |
1379 | pbase_tree_cache[i] = NULL; | |
1380 | } | |
1381 | ||
1382 | free(done_pbase_paths); | |
1383 | done_pbase_paths = NULL; | |
1384 | done_pbase_paths_num = done_pbase_paths_alloc = 0; | |
1385 | } | |
1386 | ||
c323ac7d LT |
1387 | static void check_object(struct object_entry *entry) |
1388 | { | |
5c49c116 | 1389 | if (entry->in_pack) { |
780e6e73 | 1390 | struct packed_git *p = entry->in_pack; |
03e79c88 | 1391 | struct pack_window *w_curs = NULL; |
5c49c116 NP |
1392 | const unsigned char *base_ref = NULL; |
1393 | struct object_entry *base_entry; | |
1394 | unsigned long used, used_0; | |
ef49a7a0 | 1395 | unsigned long avail; |
5c49c116 NP |
1396 | off_t ofs; |
1397 | unsigned char *buf, c; | |
780e6e73 | 1398 | |
6777a59f | 1399 | buf = use_pack(p, &w_curs, entry->in_pack_offset, &avail); |
ab7cd7bb | 1400 | |
5c49c116 | 1401 | /* |
fa736f72 NP |
1402 | * We want in_pack_type even if we do not reuse delta |
1403 | * since non-delta representations could still be reused. | |
ab7cd7bb | 1404 | */ |
09ded04b | 1405 | used = unpack_object_header_buffer(buf, avail, |
5c49c116 NP |
1406 | &entry->in_pack_type, |
1407 | &entry->size); | |
03d66015 NP |
1408 | if (used == 0) |
1409 | goto give_up; | |
ab7cd7bb | 1410 | |
5c49c116 NP |
1411 | /* |
1412 | * Determine if this is a delta and if so whether we can | |
1413 | * reuse it or not. Otherwise let's find out as cheaply as | |
1414 | * possible what the actual type and size for this object is. | |
3f9ac8d2 | 1415 | */ |
5c49c116 NP |
1416 | switch (entry->in_pack_type) { |
1417 | default: | |
1418 | /* Not a delta hence we've already got all we need. */ | |
1419 | entry->type = entry->in_pack_type; | |
1420 | entry->in_pack_header_size = used; | |
03d66015 NP |
1421 | if (entry->type < OBJ_COMMIT || entry->type > OBJ_BLOB) |
1422 | goto give_up; | |
5c49c116 NP |
1423 | unuse_pack(&w_curs); |
1424 | return; | |
1425 | case OBJ_REF_DELTA: | |
a7de7130 | 1426 | if (reuse_delta && !entry->preferred_base) |
5c49c116 NP |
1427 | base_ref = use_pack(p, &w_curs, |
1428 | entry->in_pack_offset + used, NULL); | |
1429 | entry->in_pack_header_size = used + 20; | |
1430 | break; | |
1431 | case OBJ_OFS_DELTA: | |
1432 | buf = use_pack(p, &w_curs, | |
1433 | entry->in_pack_offset + used, NULL); | |
1434 | used_0 = 0; | |
1435 | c = buf[used_0++]; | |
1436 | ofs = c & 127; | |
1437 | while (c & 128) { | |
1438 | ofs += 1; | |
03d66015 NP |
1439 | if (!ofs || MSB(ofs, 7)) { |
1440 | error("delta base offset overflow in pack for %s", | |
e6a492b7 | 1441 | oid_to_hex(&entry->idx.oid)); |
03d66015 NP |
1442 | goto give_up; |
1443 | } | |
5c49c116 NP |
1444 | c = buf[used_0++]; |
1445 | ofs = (ofs << 7) + (c & 127); | |
780e6e73 | 1446 | } |
5c49c116 | 1447 | ofs = entry->in_pack_offset - ofs; |
03d66015 NP |
1448 | if (ofs <= 0 || ofs >= entry->in_pack_offset) { |
1449 | error("delta base offset out of bound for %s", | |
e6a492b7 | 1450 | oid_to_hex(&entry->idx.oid)); |
03d66015 NP |
1451 | goto give_up; |
1452 | } | |
a7de7130 | 1453 | if (reuse_delta && !entry->preferred_base) { |
3449f8c4 NP |
1454 | struct revindex_entry *revidx; |
1455 | revidx = find_pack_revindex(p, ofs); | |
08698b1e NP |
1456 | if (!revidx) |
1457 | goto give_up; | |
3449f8c4 NP |
1458 | base_ref = nth_packed_object_sha1(p, revidx->nr); |
1459 | } | |
5c49c116 NP |
1460 | entry->in_pack_header_size = used + used_0; |
1461 | break; | |
780e6e73 | 1462 | } |
780e6e73 | 1463 | |
2834bc27 | 1464 | if (base_ref && (base_entry = packlist_find(&to_pack, base_ref, NULL))) { |
5c49c116 NP |
1465 | /* |
1466 | * If base_ref was set above that means we wish to | |
1467 | * reuse delta data, and we even found that base | |
1468 | * in the list of objects we want to pack. Goodie! | |
1469 | * | |
1470 | * Depth value does not matter - find_deltas() will | |
1471 | * never consider reused delta as the base object to | |
1472 | * deltify other objects against, in order to avoid | |
1473 | * circular deltas. | |
3f9ac8d2 | 1474 | */ |
780e6e73 | 1475 | entry->type = entry->in_pack_type; |
5c49c116 | 1476 | entry->delta = base_entry; |
64bd76b1 | 1477 | entry->delta_size = entry->size; |
15b4d577 JH |
1478 | entry->delta_sibling = base_entry->delta_child; |
1479 | base_entry->delta_child = entry; | |
5c49c116 NP |
1480 | unuse_pack(&w_curs); |
1481 | return; | |
1482 | } | |
ab7cd7bb | 1483 | |
5c49c116 NP |
1484 | if (entry->type) { |
1485 | /* | |
1486 | * This must be a delta and we already know what the | |
1487 | * final object type is. Let's extract the actual | |
1488 | * object size from the delta header. | |
1489 | */ | |
1490 | entry->size = get_size_from_delta(p, &w_curs, | |
1491 | entry->in_pack_offset + entry->in_pack_header_size); | |
03d66015 NP |
1492 | if (entry->size == 0) |
1493 | goto give_up; | |
5c49c116 | 1494 | unuse_pack(&w_curs); |
3f9ac8d2 JH |
1495 | return; |
1496 | } | |
5c49c116 NP |
1497 | |
1498 | /* | |
1499 | * No choice but to fall back to the recursive delta walk | |
1500 | * with sha1_object_info() to find about the object type | |
1501 | * at this point... | |
1502 | */ | |
03d66015 | 1503 | give_up: |
5c49c116 | 1504 | unuse_pack(&w_curs); |
36e4d74a | 1505 | } |
3f9ac8d2 | 1506 | |
e6a492b7 | 1507 | entry->type = sha1_object_info(entry->idx.oid.hash, &entry->size); |
6d6f9cdd SP |
1508 | /* |
1509 | * The error condition is checked in prepare_pack(). This is | |
1510 | * to permit a missing preferred base object to be ignored | |
1511 | * as a preferred base. Doing so can result in a larger | |
1512 | * pack file, but the transfer will still take place. | |
1513 | */ | |
3f9ac8d2 JH |
1514 | } |
1515 | ||
5c49c116 NP |
1516 | static int pack_offset_sort(const void *_a, const void *_b) |
1517 | { | |
1518 | const struct object_entry *a = *(struct object_entry **)_a; | |
1519 | const struct object_entry *b = *(struct object_entry **)_b; | |
1520 | ||
1521 | /* avoid filesystem trashing with loose objects */ | |
1522 | if (!a->in_pack && !b->in_pack) | |
e6a492b7 | 1523 | return oidcmp(&a->idx.oid, &b->idx.oid); |
5c49c116 NP |
1524 | |
1525 | if (a->in_pack < b->in_pack) | |
1526 | return -1; | |
1527 | if (a->in_pack > b->in_pack) | |
1528 | return 1; | |
1529 | return a->in_pack_offset < b->in_pack_offset ? -1 : | |
1530 | (a->in_pack_offset > b->in_pack_offset); | |
1531 | } | |
1532 | ||
4cf2143e JK |
1533 | /* |
1534 | * Drop an on-disk delta we were planning to reuse. Naively, this would | |
1535 | * just involve blanking out the "delta" field, but we have to deal | |
1536 | * with some extra book-keeping: | |
1537 | * | |
1538 | * 1. Removing ourselves from the delta_sibling linked list. | |
1539 | * | |
1540 | * 2. Updating our size/type to the non-delta representation. These were | |
1541 | * either not recorded initially (size) or overwritten with the delta type | |
1542 | * (type) when check_object() decided to reuse the delta. | |
7dbabbbe JK |
1543 | * |
1544 | * 3. Resetting our delta depth, as we are now a base object. | |
4cf2143e JK |
1545 | */ |
1546 | static void drop_reused_delta(struct object_entry *entry) | |
1547 | { | |
1548 | struct object_entry **p = &entry->delta->delta_child; | |
1549 | struct object_info oi = OBJECT_INFO_INIT; | |
1550 | ||
1551 | while (*p) { | |
1552 | if (*p == entry) | |
1553 | *p = (*p)->delta_sibling; | |
1554 | else | |
1555 | p = &(*p)->delta_sibling; | |
1556 | } | |
1557 | entry->delta = NULL; | |
7dbabbbe | 1558 | entry->depth = 0; |
4cf2143e JK |
1559 | |
1560 | oi.sizep = &entry->size; | |
1561 | oi.typep = &entry->type; | |
1562 | if (packed_object_info(entry->in_pack, entry->in_pack_offset, &oi) < 0) { | |
1563 | /* | |
1564 | * We failed to get the info from this pack for some reason; | |
1565 | * fall back to sha1_object_info, which may find another copy. | |
1566 | * And if that fails, the error will be recorded in entry->type | |
1567 | * and dealt with in prepare_pack(). | |
1568 | */ | |
e6a492b7 | 1569 | entry->type = sha1_object_info(entry->idx.oid.hash, |
1570 | &entry->size); | |
4cf2143e JK |
1571 | } |
1572 | } | |
1573 | ||
1574 | /* | |
1575 | * Follow the chain of deltas from this entry onward, throwing away any links | |
1576 | * that cause us to hit a cycle (as determined by the DFS state flags in | |
1577 | * the entries). | |
7dbabbbe JK |
1578 | * |
1579 | * We also detect too-long reused chains that would violate our --depth | |
1580 | * limit. | |
4cf2143e JK |
1581 | */ |
1582 | static void break_delta_chains(struct object_entry *entry) | |
1583 | { | |
42b766d7 JK |
1584 | /* |
1585 | * The actual depth of each object we will write is stored as an int, | |
1586 | * as it cannot exceed our int "depth" limit. But before we break | |
1587 | * changes based no that limit, we may potentially go as deep as the | |
1588 | * number of objects, which is elsewhere bounded to a uint32_t. | |
1589 | */ | |
1590 | uint32_t total_depth; | |
1591 | struct object_entry *cur, *next; | |
1592 | ||
1593 | for (cur = entry, total_depth = 0; | |
1594 | cur; | |
1595 | cur = cur->delta, total_depth++) { | |
1596 | if (cur->dfs_state == DFS_DONE) { | |
1597 | /* | |
1598 | * We've already seen this object and know it isn't | |
1599 | * part of a cycle. We do need to append its depth | |
1600 | * to our count. | |
1601 | */ | |
1602 | total_depth += cur->depth; | |
1603 | break; | |
1604 | } | |
4cf2143e | 1605 | |
4cf2143e | 1606 | /* |
42b766d7 JK |
1607 | * We break cycles before looping, so an ACTIVE state (or any |
1608 | * other cruft which made its way into the state variable) | |
1609 | * is a bug. | |
4cf2143e | 1610 | */ |
42b766d7 JK |
1611 | if (cur->dfs_state != DFS_NONE) |
1612 | die("BUG: confusing delta dfs state in first pass: %d", | |
1613 | cur->dfs_state); | |
4cf2143e | 1614 | |
4cf2143e | 1615 | /* |
42b766d7 JK |
1616 | * Now we know this is the first time we've seen the object. If |
1617 | * it's not a delta, we're done traversing, but we'll mark it | |
1618 | * done to save time on future traversals. | |
4cf2143e | 1619 | */ |
42b766d7 JK |
1620 | if (!cur->delta) { |
1621 | cur->dfs_state = DFS_DONE; | |
1622 | break; | |
1623 | } | |
4cf2143e | 1624 | |
4cf2143e | 1625 | /* |
42b766d7 JK |
1626 | * Mark ourselves as active and see if the next step causes |
1627 | * us to cycle to another active object. It's important to do | |
1628 | * this _before_ we loop, because it impacts where we make the | |
1629 | * cut, and thus how our total_depth counter works. | |
1630 | * E.g., We may see a partial loop like: | |
1631 | * | |
1632 | * A -> B -> C -> D -> B | |
1633 | * | |
1634 | * Cutting B->C breaks the cycle. But now the depth of A is | |
1635 | * only 1, and our total_depth counter is at 3. The size of the | |
1636 | * error is always one less than the size of the cycle we | |
1637 | * broke. Commits C and D were "lost" from A's chain. | |
1638 | * | |
1639 | * If we instead cut D->B, then the depth of A is correct at 3. | |
1640 | * We keep all commits in the chain that we examined. | |
4cf2143e | 1641 | */ |
42b766d7 JK |
1642 | cur->dfs_state = DFS_ACTIVE; |
1643 | if (cur->delta->dfs_state == DFS_ACTIVE) { | |
1644 | drop_reused_delta(cur); | |
1645 | cur->dfs_state = DFS_DONE; | |
1646 | break; | |
7dbabbbe | 1647 | } |
42b766d7 | 1648 | } |
7dbabbbe | 1649 | |
42b766d7 JK |
1650 | /* |
1651 | * And now that we've gone all the way to the bottom of the chain, we | |
1652 | * need to clear the active flags and set the depth fields as | |
1653 | * appropriate. Unlike the loop above, which can quit when it drops a | |
1654 | * delta, we need to keep going to look for more depth cuts. So we need | |
1655 | * an extra "next" pointer to keep going after we reset cur->delta. | |
1656 | */ | |
1657 | for (cur = entry; cur; cur = next) { | |
1658 | next = cur->delta; | |
4cf2143e | 1659 | |
42b766d7 JK |
1660 | /* |
1661 | * We should have a chain of zero or more ACTIVE states down to | |
1662 | * a final DONE. We can quit after the DONE, because either it | |
1663 | * has no bases, or we've already handled them in a previous | |
1664 | * call. | |
1665 | */ | |
1666 | if (cur->dfs_state == DFS_DONE) | |
1667 | break; | |
1668 | else if (cur->dfs_state != DFS_ACTIVE) | |
1669 | die("BUG: confusing delta dfs state in second pass: %d", | |
1670 | cur->dfs_state); | |
4cf2143e | 1671 | |
4cf2143e | 1672 | /* |
42b766d7 JK |
1673 | * If the total_depth is more than depth, then we need to snip |
1674 | * the chain into two or more smaller chains that don't exceed | |
1675 | * the maximum depth. Most of the resulting chains will contain | |
1676 | * (depth + 1) entries (i.e., depth deltas plus one base), and | |
1677 | * the last chain (i.e., the one containing entry) will contain | |
1678 | * whatever entries are left over, namely | |
1679 | * (total_depth % (depth + 1)) of them. | |
1680 | * | |
1681 | * Since we are iterating towards decreasing depth, we need to | |
1682 | * decrement total_depth as we go, and we need to write to the | |
1683 | * entry what its final depth will be after all of the | |
1684 | * snipping. Since we're snipping into chains of length (depth | |
1685 | * + 1) entries, the final depth of an entry will be its | |
1686 | * original depth modulo (depth + 1). Any time we encounter an | |
1687 | * entry whose final depth is supposed to be zero, we snip it | |
1688 | * from its delta base, thereby making it so. | |
4cf2143e | 1689 | */ |
42b766d7 JK |
1690 | cur->depth = (total_depth--) % (depth + 1); |
1691 | if (!cur->depth) | |
1692 | drop_reused_delta(cur); | |
1693 | ||
1694 | cur->dfs_state = DFS_DONE; | |
4cf2143e JK |
1695 | } |
1696 | } | |
1697 | ||
c323ac7d LT |
1698 | static void get_object_details(void) |
1699 | { | |
7cadf491 | 1700 | uint32_t i; |
5c49c116 NP |
1701 | struct object_entry **sorted_by_offset; |
1702 | ||
2834bc27 VM |
1703 | sorted_by_offset = xcalloc(to_pack.nr_objects, sizeof(struct object_entry *)); |
1704 | for (i = 0; i < to_pack.nr_objects; i++) | |
1705 | sorted_by_offset[i] = to_pack.objects + i; | |
9ed0d8d6 | 1706 | QSORT(sorted_by_offset, to_pack.nr_objects, pack_offset_sort); |
c323ac7d | 1707 | |
2834bc27 | 1708 | for (i = 0; i < to_pack.nr_objects; i++) { |
15366280 JH |
1709 | struct object_entry *entry = sorted_by_offset[i]; |
1710 | check_object(entry); | |
754980d0 | 1711 | if (big_file_threshold < entry->size) |
15366280 JH |
1712 | entry->no_try_delta = 1; |
1713 | } | |
3449f8c4 | 1714 | |
4cf2143e JK |
1715 | /* |
1716 | * This must happen in a second pass, since we rely on the delta | |
1717 | * information for the whole list being completed. | |
1718 | */ | |
1719 | for (i = 0; i < to_pack.nr_objects; i++) | |
1720 | break_delta_chains(&to_pack.objects[i]); | |
1721 | ||
5c49c116 | 1722 | free(sorted_by_offset); |
c323ac7d LT |
1723 | } |
1724 | ||
b904166c NP |
1725 | /* |
1726 | * We search for deltas in a list sorted by type, by filename hash, and then | |
1727 | * by size, so that we see progressively smaller and smaller files. | |
1728 | * That's because we prefer deltas to be from the bigger file | |
1729 | * to the smaller -- deletes are potentially cheaper, but perhaps | |
1730 | * more importantly, the bigger file is likely the more recent | |
1731 | * one. The deepest deltas are therefore the oldest objects which are | |
1732 | * less susceptible to be accessed often. | |
1733 | */ | |
9668cf59 | 1734 | static int type_size_sort(const void *_a, const void *_b) |
c323ac7d | 1735 | { |
9668cf59 NP |
1736 | const struct object_entry *a = *(struct object_entry **)_a; |
1737 | const struct object_entry *b = *(struct object_entry **)_b; | |
1738 | ||
c323ac7d | 1739 | if (a->type > b->type) |
27225f2e | 1740 | return -1; |
b904166c | 1741 | if (a->type < b->type) |
27225f2e | 1742 | return 1; |
b904166c | 1743 | if (a->hash > b->hash) |
7a979d99 | 1744 | return -1; |
b904166c | 1745 | if (a->hash < b->hash) |
7a979d99 | 1746 | return 1; |
b904166c | 1747 | if (a->preferred_base > b->preferred_base) |
c323ac7d | 1748 | return -1; |
b904166c NP |
1749 | if (a->preferred_base < b->preferred_base) |
1750 | return 1; | |
c323ac7d | 1751 | if (a->size > b->size) |
b904166c NP |
1752 | return -1; |
1753 | if (a->size < b->size) | |
c323ac7d | 1754 | return 1; |
b904166c | 1755 | return a < b ? -1 : (a > b); /* newest first */ |
c323ac7d LT |
1756 | } |
1757 | ||
1758 | struct unpacked { | |
1759 | struct object_entry *entry; | |
1760 | void *data; | |
f6c7081a | 1761 | struct delta_index *index; |
5a235b5e | 1762 | unsigned depth; |
c323ac7d LT |
1763 | }; |
1764 | ||
d250626c NP |
1765 | static int delta_cacheable(unsigned long src_size, unsigned long trg_size, |
1766 | unsigned long delta_size) | |
074b2eea MK |
1767 | { |
1768 | if (max_delta_cache_size && delta_cache_size + delta_size > max_delta_cache_size) | |
1769 | return 0; | |
1770 | ||
e3dfddb3 MK |
1771 | if (delta_size < cache_max_small_delta_size) |
1772 | return 1; | |
1773 | ||
074b2eea MK |
1774 | /* cache delta, if objects are large enough compared to delta size */ |
1775 | if ((src_size >> 20) + (trg_size >> 21) > (delta_size >> 10)) | |
1776 | return 1; | |
1777 | ||
1778 | return 0; | |
1779 | } | |
1780 | ||
7eb151d6 | 1781 | #ifndef NO_PTHREADS |
8ecce684 | 1782 | |
44626dc7 | 1783 | static pthread_mutex_t read_mutex; |
8ecce684 NP |
1784 | #define read_lock() pthread_mutex_lock(&read_mutex) |
1785 | #define read_unlock() pthread_mutex_unlock(&read_mutex) | |
1786 | ||
44626dc7 | 1787 | static pthread_mutex_t cache_mutex; |
3c701839 NP |
1788 | #define cache_lock() pthread_mutex_lock(&cache_mutex) |
1789 | #define cache_unlock() pthread_mutex_unlock(&cache_mutex) | |
1790 | ||
44626dc7 | 1791 | static pthread_mutex_t progress_mutex; |
8ecce684 NP |
1792 | #define progress_lock() pthread_mutex_lock(&progress_mutex) |
1793 | #define progress_unlock() pthread_mutex_unlock(&progress_mutex) | |
1794 | ||
1795 | #else | |
1796 | ||
e1ef8673 JH |
1797 | #define read_lock() (void)0 |
1798 | #define read_unlock() (void)0 | |
1799 | #define cache_lock() (void)0 | |
1800 | #define cache_unlock() (void)0 | |
1801 | #define progress_lock() (void)0 | |
1802 | #define progress_unlock() (void)0 | |
8ecce684 NP |
1803 | |
1804 | #endif | |
1805 | ||
f6c7081a | 1806 | static int try_delta(struct unpacked *trg, struct unpacked *src, |
ef0316fc | 1807 | unsigned max_depth, unsigned long *mem_usage) |
c323ac7d | 1808 | { |
f6c7081a NP |
1809 | struct object_entry *trg_entry = trg->entry; |
1810 | struct object_entry *src_entry = src->entry; | |
560b25a8 | 1811 | unsigned long trg_size, src_size, delta_size, sizediff, max_size, sz; |
c83f032e | 1812 | unsigned ref_depth; |
21666f1a | 1813 | enum object_type type; |
c323ac7d LT |
1814 | void *delta_buf; |
1815 | ||
1816 | /* Don't bother doing diffs between different types */ | |
f6c7081a | 1817 | if (trg_entry->type != src_entry->type) |
c323ac7d LT |
1818 | return -1; |
1819 | ||
51d1e83f | 1820 | /* |
15f07e06 JK |
1821 | * We do not bother to try a delta that we discarded on an |
1822 | * earlier try, but only when reusing delta data. Note that | |
1823 | * src_entry that is marked as the preferred_base should always | |
1824 | * be considered, as even if we produce a suboptimal delta against | |
1825 | * it, we will still save the transfer cost, as we already know | |
1826 | * the other side has it and we won't send src_entry at all. | |
51d1e83f | 1827 | */ |
a7de7130 | 1828 | if (reuse_delta && trg_entry->in_pack && |
e9195b58 | 1829 | trg_entry->in_pack == src_entry->in_pack && |
15f07e06 | 1830 | !src_entry->preferred_base && |
e9195b58 JH |
1831 | trg_entry->in_pack_type != OBJ_REF_DELTA && |
1832 | trg_entry->in_pack_type != OBJ_OFS_DELTA) | |
51d1e83f LT |
1833 | return 0; |
1834 | ||
898b14ce | 1835 | /* Let's not bust the allowed depth. */ |
5a235b5e | 1836 | if (src->depth >= max_depth) |
d116a45a | 1837 | return 0; |
c323ac7d | 1838 | |
c3b06a69 | 1839 | /* Now some size filtering heuristics. */ |
560b25a8 | 1840 | trg_size = trg_entry->size; |
c83f032e NP |
1841 | if (!trg_entry->delta) { |
1842 | max_size = trg_size/2 - 20; | |
1843 | ref_depth = 1; | |
1844 | } else { | |
1845 | max_size = trg_entry->delta_size; | |
5a235b5e | 1846 | ref_depth = trg->depth; |
c83f032e | 1847 | } |
720fe22d | 1848 | max_size = (uint64_t)max_size * (max_depth - src->depth) / |
c83f032e | 1849 | (max_depth - ref_depth + 1); |
c3b06a69 NP |
1850 | if (max_size == 0) |
1851 | return 0; | |
f6c7081a | 1852 | src_size = src_entry->size; |
560b25a8 | 1853 | sizediff = src_size < trg_size ? trg_size - src_size : 0; |
27225f2e | 1854 | if (sizediff >= max_size) |
f527cb8c | 1855 | return 0; |
a1dab41a BD |
1856 | if (trg_size < src_size / 32) |
1857 | return 0; | |
f6c7081a | 1858 | |
560b25a8 NP |
1859 | /* Load data if not already done */ |
1860 | if (!trg->data) { | |
8ecce684 | 1861 | read_lock(); |
e6a492b7 | 1862 | trg->data = read_sha1_file(trg_entry->idx.oid.hash, &type, |
1863 | &sz); | |
8ecce684 | 1864 | read_unlock(); |
2e3404c3 JH |
1865 | if (!trg->data) |
1866 | die("object %s cannot be read", | |
e6a492b7 | 1867 | oid_to_hex(&trg_entry->idx.oid)); |
560b25a8 NP |
1868 | if (sz != trg_size) |
1869 | die("object %s inconsistent object length (%lu vs %lu)", | |
e6a492b7 | 1870 | oid_to_hex(&trg_entry->idx.oid), sz, |
1871 | trg_size); | |
ef0316fc | 1872 | *mem_usage += sz; |
560b25a8 NP |
1873 | } |
1874 | if (!src->data) { | |
8ecce684 | 1875 | read_lock(); |
e6a492b7 | 1876 | src->data = read_sha1_file(src_entry->idx.oid.hash, &type, |
1877 | &sz); | |
8ecce684 | 1878 | read_unlock(); |
71064a95 NP |
1879 | if (!src->data) { |
1880 | if (src_entry->preferred_base) { | |
1881 | static int warned = 0; | |
1882 | if (!warned++) | |
1883 | warning("object %s cannot be read", | |
e6a492b7 | 1884 | oid_to_hex(&src_entry->idx.oid)); |
71064a95 NP |
1885 | /* |
1886 | * Those objects are not included in the | |
1887 | * resulting pack. Be resilient and ignore | |
1888 | * them if they can't be read, in case the | |
1889 | * pack could be created nevertheless. | |
1890 | */ | |
1891 | return 0; | |
1892 | } | |
2e3404c3 | 1893 | die("object %s cannot be read", |
e6a492b7 | 1894 | oid_to_hex(&src_entry->idx.oid)); |
71064a95 | 1895 | } |
560b25a8 NP |
1896 | if (sz != src_size) |
1897 | die("object %s inconsistent object length (%lu vs %lu)", | |
e6a492b7 | 1898 | oid_to_hex(&src_entry->idx.oid), sz, |
1899 | src_size); | |
ef0316fc | 1900 | *mem_usage += sz; |
560b25a8 NP |
1901 | } |
1902 | if (!src->index) { | |
1903 | src->index = create_delta_index(src->data, src_size); | |
a588d88a MK |
1904 | if (!src->index) { |
1905 | static int warned = 0; | |
1906 | if (!warned++) | |
1907 | warning("suboptimal pack - out of memory"); | |
1908 | return 0; | |
1909 | } | |
ef0316fc | 1910 | *mem_usage += sizeof_delta_index(src->index); |
560b25a8 NP |
1911 | } |
1912 | ||
1913 | delta_buf = create_delta(src->index, trg->data, trg_size, &delta_size, max_size); | |
c323ac7d | 1914 | if (!delta_buf) |
75c42d8c | 1915 | return 0; |
f6c7081a | 1916 | |
9e2d57a0 | 1917 | if (trg_entry->delta) { |
848d732c BD |
1918 | /* Prefer only shallower same-sized deltas. */ |
1919 | if (delta_size == trg_entry->delta_size && | |
5a235b5e | 1920 | src->depth + 1 >= trg->depth) { |
848d732c BD |
1921 | free(delta_buf); |
1922 | return 0; | |
1923 | } | |
074b2eea | 1924 | } |
9e2d57a0 | 1925 | |
3c701839 NP |
1926 | /* |
1927 | * Handle memory allocation outside of the cache | |
1928 | * accounting lock. Compiler will optimize the strangeness | |
7eb151d6 | 1929 | * away when NO_PTHREADS is defined. |
3c701839 | 1930 | */ |
8e0f7003 | 1931 | free(trg_entry->delta_data); |
3c701839 | 1932 | cache_lock(); |
9e2d57a0 NP |
1933 | if (trg_entry->delta_data) { |
1934 | delta_cache_size -= trg_entry->delta_size; | |
9e2d57a0 NP |
1935 | trg_entry->delta_data = NULL; |
1936 | } | |
d250626c | 1937 | if (delta_cacheable(src_size, trg_size, delta_size)) { |
b7a28f78 | 1938 | delta_cache_size += delta_size; |
3c701839 NP |
1939 | cache_unlock(); |
1940 | trg_entry->delta_data = xrealloc(delta_buf, delta_size); | |
1941 | } else { | |
1942 | cache_unlock(); | |
074b2eea | 1943 | free(delta_buf); |
3c701839 NP |
1944 | } |
1945 | ||
b7a28f78 NP |
1946 | trg_entry->delta = src_entry; |
1947 | trg_entry->delta_size = delta_size; | |
1948 | trg->depth = src->depth + 1; | |
1949 | ||
f6c7081a | 1950 | return 1; |
c323ac7d LT |
1951 | } |
1952 | ||
898b14ce | 1953 | static unsigned int check_delta_limit(struct object_entry *me, unsigned int n) |
b2504a0d | 1954 | { |
898b14ce NP |
1955 | struct object_entry *child = me->delta_child; |
1956 | unsigned int m = n; | |
1957 | while (child) { | |
1958 | unsigned int c = check_delta_limit(child, n + 1); | |
1959 | if (m < c) | |
1960 | m = c; | |
1961 | child = child->delta_sibling; | |
1962 | } | |
1963 | return m; | |
b2504a0d NP |
1964 | } |
1965 | ||
75ad235c | 1966 | static unsigned long free_unpacked(struct unpacked *n) |
a97773ce | 1967 | { |
ef0316fc | 1968 | unsigned long freed_mem = sizeof_delta_index(n->index); |
a97773ce BD |
1969 | free_delta_index(n->index); |
1970 | n->index = NULL; | |
1971 | if (n->data) { | |
ef0316fc | 1972 | freed_mem += n->entry->size; |
a97773ce BD |
1973 | free(n->data); |
1974 | n->data = NULL; | |
a97773ce BD |
1975 | } |
1976 | n->entry = NULL; | |
7d7baa5e | 1977 | n->depth = 0; |
ef0316fc | 1978 | return freed_mem; |
a97773ce BD |
1979 | } |
1980 | ||
384b32c0 | 1981 | static void find_deltas(struct object_entry **list, unsigned *list_size, |
e334977d | 1982 | int window, int depth, unsigned *processed) |
c323ac7d | 1983 | { |
384b32c0 | 1984 | uint32_t i, idx = 0, count = 0; |
7cadf491 | 1985 | struct unpacked *array; |
ef0316fc | 1986 | unsigned long mem_usage = 0; |
c323ac7d | 1987 | |
19d4b416 | 1988 | array = xcalloc(window, sizeof(struct unpacked)); |
21fcd1bd | 1989 | |
384b32c0 | 1990 | for (;;) { |
421b488a | 1991 | struct object_entry *entry; |
c323ac7d | 1992 | struct unpacked *n = array + idx; |
ef0316fc | 1993 | int j, max_depth, best_base = -1; |
c323ac7d | 1994 | |
384b32c0 NP |
1995 | progress_lock(); |
1996 | if (!*list_size) { | |
1997 | progress_unlock(); | |
1998 | break; | |
1999 | } | |
421b488a | 2000 | entry = *list++; |
384b32c0 NP |
2001 | (*list_size)--; |
2002 | if (!entry->preferred_base) { | |
2003 | (*processed)++; | |
2004 | display_progress(progress_state, *processed); | |
2005 | } | |
2006 | progress_unlock(); | |
2007 | ||
ef0316fc | 2008 | mem_usage -= free_unpacked(n); |
c323ac7d | 2009 | n->entry = entry; |
ab7cd7bb | 2010 | |
a97773ce | 2011 | while (window_memory_limit && |
ef0316fc | 2012 | mem_usage > window_memory_limit && |
a97773ce BD |
2013 | count > 1) { |
2014 | uint32_t tail = (idx + window - count) % window; | |
75ad235c | 2015 | mem_usage -= free_unpacked(array + tail); |
a97773ce BD |
2016 | count--; |
2017 | } | |
2018 | ||
75d39853 NP |
2019 | /* We do not compute delta to *create* objects we are not |
2020 | * going to pack. | |
2021 | */ | |
2022 | if (entry->preferred_base) | |
2023 | goto next; | |
2024 | ||
898b14ce NP |
2025 | /* |
2026 | * If the current object is at pack edge, take the depth the | |
2027 | * objects that depend on the current object into account | |
2028 | * otherwise they would become too deep. | |
2029 | */ | |
2030 | max_depth = depth; | |
2031 | if (entry->delta_child) { | |
2032 | max_depth -= check_delta_limit(entry, 0); | |
2033 | if (max_depth <= 0) | |
2034 | goto next; | |
2035 | } | |
2036 | ||
78817c15 LT |
2037 | j = window; |
2038 | while (--j > 0) { | |
77639870 | 2039 | int ret; |
7cadf491 | 2040 | uint32_t other_idx = idx + j; |
c323ac7d | 2041 | struct unpacked *m; |
78817c15 LT |
2042 | if (other_idx >= window) |
2043 | other_idx -= window; | |
c323ac7d LT |
2044 | m = array + other_idx; |
2045 | if (!m->entry) | |
2046 | break; | |
ef0316fc | 2047 | ret = try_delta(n, m, max_depth, &mem_usage); |
77639870 | 2048 | if (ret < 0) |
c323ac7d | 2049 | break; |
77639870 JH |
2050 | else if (ret > 0) |
2051 | best_base = other_idx; | |
c323ac7d | 2052 | } |
898b14ce | 2053 | |
ed4a9031 NP |
2054 | /* |
2055 | * If we decided to cache the delta data, then it is best | |
2056 | * to compress it right away. First because we have to do | |
2057 | * it anyway, and doing it here while we're threaded will | |
2058 | * save a lot of time in the non threaded write phase, | |
2059 | * as well as allow for caching more deltas within | |
2060 | * the same cache size limit. | |
2061 | * ... | |
2062 | * But only if not writing to stdout, since in that case | |
2063 | * the network is most likely throttling writes anyway, | |
2064 | * and therefore it is best to go to the write phase ASAP | |
2065 | * instead, as we can afford spending more time compressing | |
2066 | * between writes at that moment. | |
2067 | */ | |
2068 | if (entry->delta_data && !pack_to_stdout) { | |
2069 | entry->z_delta_size = do_compress(&entry->delta_data, | |
2070 | entry->delta_size); | |
2071 | cache_lock(); | |
2072 | delta_cache_size -= entry->delta_size; | |
2073 | delta_cache_size += entry->z_delta_size; | |
2074 | cache_unlock(); | |
2075 | } | |
2076 | ||
70ca1a3f JH |
2077 | /* if we made n a delta, and if n is already at max |
2078 | * depth, leaving it in the window is pointless. we | |
2079 | * should evict it first. | |
70ca1a3f | 2080 | */ |
70baf5d4 | 2081 | if (entry->delta && max_depth <= n->depth) |
70ca1a3f | 2082 | continue; |
ff45715c | 2083 | |
77639870 JH |
2084 | /* |
2085 | * Move the best delta base up in the window, after the | |
2086 | * currently deltified object, to keep it longer. It will | |
2087 | * be the first base object to be attempted next. | |
2088 | */ | |
2089 | if (entry->delta) { | |
2090 | struct unpacked swap = array[best_base]; | |
2091 | int dist = (window + idx - best_base) % window; | |
2092 | int dst = best_base; | |
2093 | while (dist--) { | |
2094 | int src = (dst + 1) % window; | |
2095 | array[dst] = array[src]; | |
2096 | dst = src; | |
2097 | } | |
2098 | array[dst] = swap; | |
2099 | } | |
2100 | ||
898b14ce | 2101 | next: |
521a4f4c | 2102 | idx++; |
a97773ce BD |
2103 | if (count + 1 < window) |
2104 | count++; | |
521a4f4c LT |
2105 | if (idx >= window) |
2106 | idx = 0; | |
384b32c0 | 2107 | } |
adee7bdf | 2108 | |
f6c7081a | 2109 | for (i = 0; i < window; ++i) { |
ff45715c | 2110 | free_delta_index(array[i].index); |
adee7bdf | 2111 | free(array[i].data); |
f6c7081a | 2112 | } |
adee7bdf | 2113 | free(array); |
c323ac7d LT |
2114 | } |
2115 | ||
7eb151d6 | 2116 | #ifndef NO_PTHREADS |
8ecce684 | 2117 | |
a9a74636 NP |
2118 | static void try_to_free_from_threads(size_t size) |
2119 | { | |
2120 | read_lock(); | |
7c3ecb32 | 2121 | release_pack_memory(size); |
a9a74636 NP |
2122 | read_unlock(); |
2123 | } | |
2124 | ||
bc9b2175 | 2125 | static try_to_free_t old_try_to_free_routine; |
851c34b0 | 2126 | |
50f22ada JS |
2127 | /* |
2128 | * The main thread waits on the condition that (at least) one of the workers | |
2129 | * has stopped working (which is indicated in the .working member of | |
2130 | * struct thread_params). | |
2131 | * When a work thread has completed its work, it sets .working to 0 and | |
2132 | * signals the main thread and waits on the condition that .data_ready | |
2133 | * becomes 1. | |
2134 | */ | |
2135 | ||
8ecce684 NP |
2136 | struct thread_params { |
2137 | pthread_t thread; | |
2138 | struct object_entry **list; | |
2139 | unsigned list_size; | |
384b32c0 | 2140 | unsigned remaining; |
8ecce684 NP |
2141 | int window; |
2142 | int depth; | |
50f22ada JS |
2143 | int working; |
2144 | int data_ready; | |
2145 | pthread_mutex_t mutex; | |
2146 | pthread_cond_t cond; | |
8ecce684 NP |
2147 | unsigned *processed; |
2148 | }; | |
2149 | ||
44626dc7 AH |
2150 | static pthread_cond_t progress_cond; |
2151 | ||
2152 | /* | |
2153 | * Mutex and conditional variable can't be statically-initialized on Windows. | |
2154 | */ | |
2155 | static void init_threaded_search(void) | |
2156 | { | |
93749194 | 2157 | init_recursive_mutex(&read_mutex); |
44626dc7 AH |
2158 | pthread_mutex_init(&cache_mutex, NULL); |
2159 | pthread_mutex_init(&progress_mutex, NULL); | |
2160 | pthread_cond_init(&progress_cond, NULL); | |
851c34b0 | 2161 | old_try_to_free_routine = set_try_to_free_routine(try_to_free_from_threads); |
44626dc7 AH |
2162 | } |
2163 | ||
2164 | static void cleanup_threaded_search(void) | |
2165 | { | |
851c34b0 | 2166 | set_try_to_free_routine(old_try_to_free_routine); |
44626dc7 AH |
2167 | pthread_cond_destroy(&progress_cond); |
2168 | pthread_mutex_destroy(&read_mutex); | |
2169 | pthread_mutex_destroy(&cache_mutex); | |
2170 | pthread_mutex_destroy(&progress_mutex); | |
2171 | } | |
c2a33679 | 2172 | |
8ecce684 NP |
2173 | static void *threaded_find_deltas(void *arg) |
2174 | { | |
c2a33679 NP |
2175 | struct thread_params *me = arg; |
2176 | ||
50f22ada | 2177 | while (me->remaining) { |
384b32c0 | 2178 | find_deltas(me->list, &me->remaining, |
c2a33679 | 2179 | me->window, me->depth, me->processed); |
50f22ada JS |
2180 | |
2181 | progress_lock(); | |
2182 | me->working = 0; | |
2183 | pthread_cond_signal(&progress_cond); | |
2184 | progress_unlock(); | |
2185 | ||
2186 | /* | |
2187 | * We must not set ->data_ready before we wait on the | |
2188 | * condition because the main thread may have set it to 1 | |
2189 | * before we get here. In order to be sure that new | |
2190 | * work is available if we see 1 in ->data_ready, it | |
2191 | * was initialized to 0 before this thread was spawned | |
2192 | * and we reset it to 0 right away. | |
2193 | */ | |
2194 | pthread_mutex_lock(&me->mutex); | |
2195 | while (!me->data_ready) | |
2196 | pthread_cond_wait(&me->cond, &me->mutex); | |
2197 | me->data_ready = 0; | |
2198 | pthread_mutex_unlock(&me->mutex); | |
c2a33679 | 2199 | } |
50f22ada JS |
2200 | /* leave ->working 1 so that this doesn't get more work assigned */ |
2201 | return NULL; | |
8ecce684 NP |
2202 | } |
2203 | ||
8ecce684 NP |
2204 | static void ll_find_deltas(struct object_entry **list, unsigned list_size, |
2205 | int window, int depth, unsigned *processed) | |
2206 | { | |
dcda3614 | 2207 | struct thread_params *p; |
384b32c0 | 2208 | int i, ret, active_threads = 0; |
c2a33679 | 2209 | |
44626dc7 AH |
2210 | init_threaded_search(); |
2211 | ||
367f4a43 | 2212 | if (delta_search_threads <= 1) { |
384b32c0 | 2213 | find_deltas(list, &list_size, window, depth, processed); |
44626dc7 | 2214 | cleanup_threaded_search(); |
367f4a43 NP |
2215 | return; |
2216 | } | |
43cc2b42 | 2217 | if (progress > pack_to_stdout) |
b6c29915 | 2218 | fprintf(stderr, "Delta compression using up to %d threads.\n", |
43cc2b42 | 2219 | delta_search_threads); |
dcda3614 | 2220 | p = xcalloc(delta_search_threads, sizeof(*p)); |
367f4a43 | 2221 | |
50f22ada | 2222 | /* Partition the work amongst work threads. */ |
367f4a43 | 2223 | for (i = 0; i < delta_search_threads; i++) { |
50f22ada JS |
2224 | unsigned sub_size = list_size / (delta_search_threads - i); |
2225 | ||
bf874896 NP |
2226 | /* don't use too small segments or no deltas will be found */ |
2227 | if (sub_size < 2*window && i+1 < delta_search_threads) | |
2228 | sub_size = 0; | |
2229 | ||
8ecce684 NP |
2230 | p[i].window = window; |
2231 | p[i].depth = depth; | |
2232 | p[i].processed = processed; | |
50f22ada JS |
2233 | p[i].working = 1; |
2234 | p[i].data_ready = 0; | |
c2a33679 | 2235 | |
59921b4b | 2236 | /* try to split chunks on "path" boundaries */ |
6fc74703 NP |
2237 | while (sub_size && sub_size < list_size && |
2238 | list[sub_size]->hash && | |
384b32c0 NP |
2239 | list[sub_size]->hash == list[sub_size-1]->hash) |
2240 | sub_size++; | |
2241 | ||
50f22ada JS |
2242 | p[i].list = list; |
2243 | p[i].list_size = sub_size; | |
2244 | p[i].remaining = sub_size; | |
59921b4b | 2245 | |
384b32c0 NP |
2246 | list += sub_size; |
2247 | list_size -= sub_size; | |
2248 | } | |
2249 | ||
50f22ada JS |
2250 | /* Start work threads. */ |
2251 | for (i = 0; i < delta_search_threads; i++) { | |
2252 | if (!p[i].list_size) | |
2253 | continue; | |
68e6a4f8 JS |
2254 | pthread_mutex_init(&p[i].mutex, NULL); |
2255 | pthread_cond_init(&p[i].cond, NULL); | |
50f22ada JS |
2256 | ret = pthread_create(&p[i].thread, NULL, |
2257 | threaded_find_deltas, &p[i]); | |
2258 | if (ret) | |
2259 | die("unable to create thread: %s", strerror(ret)); | |
2260 | active_threads++; | |
2261 | } | |
2262 | ||
384b32c0 NP |
2263 | /* |
2264 | * Now let's wait for work completion. Each time a thread is done | |
2265 | * with its work, we steal half of the remaining work from the | |
2266 | * thread with the largest number of unprocessed objects and give | |
2267 | * it to that newly idle thread. This ensure good load balancing | |
2268 | * until the remaining object list segments are simply too short | |
2269 | * to be worth splitting anymore. | |
2270 | */ | |
50f22ada JS |
2271 | while (active_threads) { |
2272 | struct thread_params *target = NULL; | |
384b32c0 NP |
2273 | struct thread_params *victim = NULL; |
2274 | unsigned sub_size = 0; | |
384b32c0 NP |
2275 | |
2276 | progress_lock(); | |
50f22ada JS |
2277 | for (;;) { |
2278 | for (i = 0; !target && i < delta_search_threads; i++) | |
2279 | if (!p[i].working) | |
2280 | target = &p[i]; | |
2281 | if (target) | |
2282 | break; | |
2283 | pthread_cond_wait(&progress_cond, &progress_mutex); | |
2284 | } | |
2285 | ||
384b32c0 NP |
2286 | for (i = 0; i < delta_search_threads; i++) |
2287 | if (p[i].remaining > 2*window && | |
2288 | (!victim || victim->remaining < p[i].remaining)) | |
2289 | victim = &p[i]; | |
2290 | if (victim) { | |
2291 | sub_size = victim->remaining / 2; | |
2292 | list = victim->list + victim->list_size - sub_size; | |
2293 | while (sub_size && list[0]->hash && | |
2294 | list[0]->hash == list[-1]->hash) { | |
2295 | list++; | |
2296 | sub_size--; | |
2297 | } | |
eb9688ff NP |
2298 | if (!sub_size) { |
2299 | /* | |
2300 | * It is possible for some "paths" to have | |
2301 | * so many objects that no hash boundary | |
2302 | * might be found. Let's just steal the | |
2303 | * exact half in that case. | |
2304 | */ | |
2305 | sub_size = victim->remaining / 2; | |
2306 | list -= sub_size; | |
2307 | } | |
384b32c0 NP |
2308 | target->list = list; |
2309 | victim->list_size -= sub_size; | |
2310 | victim->remaining -= sub_size; | |
2311 | } | |
384b32c0 NP |
2312 | target->list_size = sub_size; |
2313 | target->remaining = sub_size; | |
50f22ada JS |
2314 | target->working = 1; |
2315 | progress_unlock(); | |
2316 | ||
2317 | pthread_mutex_lock(&target->mutex); | |
2318 | target->data_ready = 1; | |
2319 | pthread_cond_signal(&target->cond); | |
2320 | pthread_mutex_unlock(&target->mutex); | |
c2a33679 | 2321 | |
384b32c0 | 2322 | if (!sub_size) { |
b81d9af7 | 2323 | pthread_join(target->thread, NULL); |
50f22ada JS |
2324 | pthread_cond_destroy(&target->cond); |
2325 | pthread_mutex_destroy(&target->mutex); | |
384b32c0 | 2326 | active_threads--; |
c2a33679 | 2327 | } |
50f22ada | 2328 | } |
44626dc7 | 2329 | cleanup_threaded_search(); |
dcda3614 | 2330 | free(p); |
8ecce684 NP |
2331 | } |
2332 | ||
2333 | #else | |
384b32c0 | 2334 | #define ll_find_deltas(l, s, w, d, p) find_deltas(l, &s, w, d, p) |
8ecce684 NP |
2335 | #endif |
2336 | ||
b773ddea JK |
2337 | static void add_tag_chain(const struct object_id *oid) |
2338 | { | |
2339 | struct tag *tag; | |
2340 | ||
2341 | /* | |
2342 | * We catch duplicates already in add_object_entry(), but we'd | |
2343 | * prefer to do this extra check to avoid having to parse the | |
2344 | * tag at all if we already know that it's being packed (e.g., if | |
2345 | * it was included via bitmaps, we would not have parsed it | |
2346 | * previously). | |
2347 | */ | |
2348 | if (packlist_find(&to_pack, oid->hash, NULL)) | |
2349 | return; | |
2350 | ||
d3101b53 | 2351 | tag = lookup_tag(oid); |
b773ddea JK |
2352 | while (1) { |
2353 | if (!tag || parse_tag(tag) || !tag->tagged) | |
2354 | die("unable to pack objects reachable from tag %s", | |
2355 | oid_to_hex(oid)); | |
2356 | ||
2357 | add_object_entry(tag->object.oid.hash, OBJ_TAG, NULL, 0); | |
2358 | ||
2359 | if (tag->tagged->type != OBJ_TAG) | |
2360 | return; | |
2361 | ||
2362 | tag = (struct tag *)tag->tagged; | |
2363 | } | |
2364 | } | |
2365 | ||
d155254c | 2366 | static int add_ref_tag(const char *path, const struct object_id *oid, int flag, void *cb_data) |
f0a24aa5 | 2367 | { |
d155254c | 2368 | struct object_id peeled; |
f0a24aa5 | 2369 | |
59556548 | 2370 | if (starts_with(path, "refs/tags/") && /* is a tag? */ |
d155254c MH |
2371 | !peel_ref(path, peeled.hash) && /* peelable? */ |
2372 | packlist_find(&to_pack, peeled.hash, NULL)) /* object packed? */ | |
b773ddea | 2373 | add_tag_chain(oid); |
f0a24aa5 SP |
2374 | return 0; |
2375 | } | |
2376 | ||
f3123c4a JH |
2377 | static void prepare_pack(int window, int depth) |
2378 | { | |
9668cf59 | 2379 | struct object_entry **delta_list; |
6e1c2344 RJ |
2380 | uint32_t i, nr_deltas; |
2381 | unsigned n; | |
9668cf59 | 2382 | |
3f9ac8d2 | 2383 | get_object_details(); |
9668cf59 | 2384 | |
0e8189e2 NP |
2385 | /* |
2386 | * If we're locally repacking then we need to be doubly careful | |
2387 | * from now on in order to make sure no stealth corruption gets | |
2388 | * propagated to the new pack. Clients receiving streamed packs | |
2389 | * should validate everything they get anyway so no need to incur | |
2390 | * the additional cost here in that case. | |
2391 | */ | |
2392 | if (!pack_to_stdout) | |
2393 | do_check_packed_object_crc = 1; | |
2394 | ||
2834bc27 | 2395 | if (!to_pack.nr_objects || !window || !depth) |
9668cf59 NP |
2396 | return; |
2397 | ||
b32fa95f | 2398 | ALLOC_ARRAY(delta_list, to_pack.nr_objects); |
75d39853 NP |
2399 | nr_deltas = n = 0; |
2400 | ||
2834bc27 VM |
2401 | for (i = 0; i < to_pack.nr_objects; i++) { |
2402 | struct object_entry *entry = to_pack.objects + i; | |
75d39853 NP |
2403 | |
2404 | if (entry->delta) | |
2405 | /* This happens if we decided to reuse existing | |
a7de7130 | 2406 | * delta from a pack. "reuse_delta &&" is implied. |
75d39853 NP |
2407 | */ |
2408 | continue; | |
2409 | ||
2410 | if (entry->size < 50) | |
2411 | continue; | |
2412 | ||
2413 | if (entry->no_try_delta) | |
2414 | continue; | |
2415 | ||
6d6f9cdd | 2416 | if (!entry->preferred_base) { |
75d39853 | 2417 | nr_deltas++; |
6d6f9cdd SP |
2418 | if (entry->type < 0) |
2419 | die("unable to get type of object %s", | |
e6a492b7 | 2420 | oid_to_hex(&entry->idx.oid)); |
eede9f42 NP |
2421 | } else { |
2422 | if (entry->type < 0) { | |
2423 | /* | |
2424 | * This object is not found, but we | |
2425 | * don't have to include it anyway. | |
2426 | */ | |
2427 | continue; | |
2428 | } | |
6d6f9cdd | 2429 | } |
75d39853 NP |
2430 | |
2431 | delta_list[n++] = entry; | |
2432 | } | |
2433 | ||
2f8b8947 | 2434 | if (nr_deltas && n > 1) { |
e334977d NP |
2435 | unsigned nr_done = 0; |
2436 | if (progress) | |
754dbc43 | 2437 | progress_state = start_progress(_("Compressing objects"), |
dc6a0757 | 2438 | nr_deltas); |
9ed0d8d6 | 2439 | QSORT(delta_list, n, type_size_sort); |
8ecce684 | 2440 | ll_find_deltas(delta_list, n, window+1, depth, &nr_done); |
4d4fcc54 | 2441 | stop_progress(&progress_state); |
e334977d NP |
2442 | if (nr_done != nr_deltas) |
2443 | die("inconsistency with delta count"); | |
75d39853 | 2444 | } |
9668cf59 | 2445 | free(delta_list); |
f3123c4a JH |
2446 | } |
2447 | ||
ef90d6d4 | 2448 | static int git_pack_config(const char *k, const char *v, void *cb) |
4812a93a | 2449 | { |
eeefa7c9 | 2450 | if (!strcmp(k, "pack.window")) { |
4812a93a JK |
2451 | window = git_config_int(k, v); |
2452 | return 0; | |
2453 | } | |
a97773ce BD |
2454 | if (!strcmp(k, "pack.windowmemory")) { |
2455 | window_memory_limit = git_config_ulong(k, v); | |
2456 | return 0; | |
2457 | } | |
2458 | if (!strcmp(k, "pack.depth")) { | |
842aaf93 TT |
2459 | depth = git_config_int(k, v); |
2460 | return 0; | |
2461 | } | |
074b2eea MK |
2462 | if (!strcmp(k, "pack.deltacachesize")) { |
2463 | max_delta_cache_size = git_config_int(k, v); | |
2464 | return 0; | |
2465 | } | |
e3dfddb3 MK |
2466 | if (!strcmp(k, "pack.deltacachelimit")) { |
2467 | cache_max_small_delta_size = git_config_int(k, v); | |
2468 | return 0; | |
2469 | } | |
ae4f07fb VM |
2470 | if (!strcmp(k, "pack.writebitmaphashcache")) { |
2471 | if (git_config_bool(k, v)) | |
2472 | write_bitmap_options |= BITMAP_OPT_HASH_CACHE; | |
2473 | else | |
2474 | write_bitmap_options &= ~BITMAP_OPT_HASH_CACHE; | |
2475 | } | |
6b8fda2d | 2476 | if (!strcmp(k, "pack.usebitmaps")) { |
645c432d | 2477 | use_bitmap_index_default = git_config_bool(k, v); |
6b8fda2d VM |
2478 | return 0; |
2479 | } | |
693b86ff NP |
2480 | if (!strcmp(k, "pack.threads")) { |
2481 | delta_search_threads = git_config_int(k, v); | |
833e3df1 | 2482 | if (delta_search_threads < 0) |
693b86ff NP |
2483 | die("invalid number of threads specified (%d)", |
2484 | delta_search_threads); | |
7eb151d6 | 2485 | #ifdef NO_PTHREADS |
2e96d815 | 2486 | if (delta_search_threads != 1) { |
693b86ff | 2487 | warning("no threads support, ignoring %s", k); |
2e96d815 ÆAB |
2488 | delta_search_threads = 0; |
2489 | } | |
693b86ff NP |
2490 | #endif |
2491 | return 0; | |
2492 | } | |
4d00bda2 | 2493 | if (!strcmp(k, "pack.indexversion")) { |
ebcfb379 JH |
2494 | pack_idx_opts.version = git_config_int(k, v); |
2495 | if (pack_idx_opts.version > 2) | |
6e1c2344 | 2496 | die("bad pack.indexversion=%"PRIu32, |
ebcfb379 | 2497 | pack_idx_opts.version); |
4d00bda2 NP |
2498 | return 0; |
2499 | } | |
ef90d6d4 | 2500 | return git_default_config(k, v, cb); |
4812a93a JK |
2501 | } |
2502 | ||
b5d97e6b | 2503 | static void read_object_list_from_stdin(void) |
c323ac7d | 2504 | { |
b5d97e6b JH |
2505 | char line[40 + 1 + PATH_MAX + 2]; |
2506 | unsigned char sha1[20]; | |
b2504a0d | 2507 | |
da93d12b | 2508 | for (;;) { |
da93d12b LT |
2509 | if (!fgets(line, sizeof(line), stdin)) { |
2510 | if (feof(stdin)) | |
2511 | break; | |
2512 | if (!ferror(stdin)) | |
2513 | die("fgets returned NULL, not EOF, not error!"); | |
687dd75c | 2514 | if (errno != EINTR) |
d824cbba | 2515 | die_errno("fgets"); |
687dd75c JH |
2516 | clearerr(stdin); |
2517 | continue; | |
da93d12b | 2518 | } |
7a979d99 JH |
2519 | if (line[0] == '-') { |
2520 | if (get_sha1_hex(line+1, sha1)) | |
2521 | die("expected edge sha1, got garbage:\n %s", | |
b5d97e6b | 2522 | line); |
8d1d8f83 | 2523 | add_preferred_base(sha1); |
7a979d99 | 2524 | continue; |
21fcd1bd | 2525 | } |
c323ac7d | 2526 | if (get_sha1_hex(line, sha1)) |
ef07618f | 2527 | die("expected sha1, got garbage:\n %s", line); |
b5d97e6b | 2528 | |
bc32fed5 JH |
2529 | add_preferred_base_object(line+41); |
2530 | add_object_entry(sha1, 0, line+41, 0); | |
c323ac7d | 2531 | } |
b5d97e6b JH |
2532 | } |
2533 | ||
08cdfb13 JH |
2534 | #define OBJECT_ADDED (1u<<20) |
2535 | ||
11c211fa | 2536 | static void show_commit(struct commit *commit, void *data) |
b5d97e6b | 2537 | { |
ed1c9977 | 2538 | add_object_entry(commit->object.oid.hash, OBJ_COMMIT, NULL, 0); |
08cdfb13 | 2539 | commit->object.flags |= OBJECT_ADDED; |
7cc8f971 VM |
2540 | |
2541 | if (write_bitmap_index) | |
2542 | index_commit_for_bitmap(commit); | |
b5d97e6b JH |
2543 | } |
2544 | ||
de1e67d0 | 2545 | static void show_object(struct object *obj, const char *name, void *data) |
b5d97e6b | 2546 | { |
8d2dfc49 | 2547 | add_preferred_base_object(name); |
ed1c9977 | 2548 | add_object_entry(obj->oid.hash, obj->type, name, 0); |
8d2dfc49 | 2549 | obj->flags |= OBJECT_ADDED; |
b5d97e6b JH |
2550 | } |
2551 | ||
8d1d8f83 JH |
2552 | static void show_edge(struct commit *commit) |
2553 | { | |
ed1c9977 | 2554 | add_preferred_base(commit->object.oid.hash); |
8d1d8f83 JH |
2555 | } |
2556 | ||
08cdfb13 JH |
2557 | struct in_pack_object { |
2558 | off_t offset; | |
2559 | struct object *object; | |
2560 | }; | |
2561 | ||
2562 | struct in_pack { | |
2563 | int alloc; | |
2564 | int nr; | |
2565 | struct in_pack_object *array; | |
2566 | }; | |
2567 | ||
2568 | static void mark_in_pack_object(struct object *object, struct packed_git *p, struct in_pack *in_pack) | |
2569 | { | |
ed1c9977 | 2570 | in_pack->array[in_pack->nr].offset = find_pack_entry_one(object->oid.hash, p); |
08cdfb13 JH |
2571 | in_pack->array[in_pack->nr].object = object; |
2572 | in_pack->nr++; | |
2573 | } | |
2574 | ||
2575 | /* | |
2576 | * Compare the objects in the offset order, in order to emulate the | |
f18d244a | 2577 | * "git rev-list --objects" output that produced the pack originally. |
08cdfb13 JH |
2578 | */ |
2579 | static int ofscmp(const void *a_, const void *b_) | |
2580 | { | |
2581 | struct in_pack_object *a = (struct in_pack_object *)a_; | |
2582 | struct in_pack_object *b = (struct in_pack_object *)b_; | |
2583 | ||
2584 | if (a->offset < b->offset) | |
2585 | return -1; | |
2586 | else if (a->offset > b->offset) | |
2587 | return 1; | |
2588 | else | |
f2fd0760 | 2589 | return oidcmp(&a->object->oid, &b->object->oid); |
08cdfb13 JH |
2590 | } |
2591 | ||
2592 | static void add_objects_in_unpacked_packs(struct rev_info *revs) | |
2593 | { | |
2594 | struct packed_git *p; | |
2595 | struct in_pack in_pack; | |
2596 | uint32_t i; | |
2597 | ||
2598 | memset(&in_pack, 0, sizeof(in_pack)); | |
2599 | ||
2600 | for (p = packed_git; p; p = p->next) { | |
2601 | const unsigned char *sha1; | |
2602 | struct object *o; | |
2603 | ||
79bc4c71 | 2604 | if (!p->pack_local || p->pack_keep) |
08cdfb13 JH |
2605 | continue; |
2606 | if (open_pack_index(p)) | |
2607 | die("cannot open pack index"); | |
2608 | ||
2609 | ALLOC_GROW(in_pack.array, | |
2610 | in_pack.nr + p->num_objects, | |
2611 | in_pack.alloc); | |
2612 | ||
2613 | for (i = 0; i < p->num_objects; i++) { | |
2614 | sha1 = nth_packed_object_sha1(p, i); | |
2615 | o = lookup_unknown_object(sha1); | |
2616 | if (!(o->flags & OBJECT_ADDED)) | |
2617 | mark_in_pack_object(o, p, &in_pack); | |
2618 | o->flags |= OBJECT_ADDED; | |
2619 | } | |
2620 | } | |
2621 | ||
2622 | if (in_pack.nr) { | |
9ed0d8d6 | 2623 | QSORT(in_pack.array, in_pack.nr, ofscmp); |
08cdfb13 JH |
2624 | for (i = 0; i < in_pack.nr; i++) { |
2625 | struct object *o = in_pack.array[i].object; | |
ed1c9977 | 2626 | add_object_entry(o->oid.hash, o->type, "", 0); |
08cdfb13 JH |
2627 | } |
2628 | } | |
2629 | free(in_pack.array); | |
2630 | } | |
2631 | ||
76c1d9a0 | 2632 | static int add_loose_object(const struct object_id *oid, const char *path, |
e26a8c47 JK |
2633 | void *data) |
2634 | { | |
76c1d9a0 | 2635 | enum object_type type = sha1_object_info(oid->hash, NULL); |
e26a8c47 JK |
2636 | |
2637 | if (type < 0) { | |
2638 | warning("loose object at %s could not be examined", path); | |
2639 | return 0; | |
2640 | } | |
2641 | ||
76c1d9a0 | 2642 | add_object_entry(oid->hash, type, "", 0); |
e26a8c47 JK |
2643 | return 0; |
2644 | } | |
2645 | ||
2646 | /* | |
2647 | * We actually don't even have to worry about reachability here. | |
2648 | * add_object_entry will weed out duplicates, so we just add every | |
2649 | * loose object we find. | |
2650 | */ | |
2651 | static void add_unreachable_loose_objects(void) | |
2652 | { | |
2653 | for_each_loose_file_in_objdir(get_object_directory(), | |
2654 | add_loose_object, | |
2655 | NULL, NULL, NULL); | |
2656 | } | |
2657 | ||
094085e3 BC |
2658 | static int has_sha1_pack_kept_or_nonlocal(const unsigned char *sha1) |
2659 | { | |
2660 | static struct packed_git *last_found = (void *)1; | |
2661 | struct packed_git *p; | |
2662 | ||
2663 | p = (last_found != (void *)1) ? last_found : packed_git; | |
2664 | ||
2665 | while (p) { | |
2666 | if ((!p->pack_local || p->pack_keep) && | |
2667 | find_pack_entry_one(sha1, p)) { | |
2668 | last_found = p; | |
2669 | return 1; | |
2670 | } | |
2671 | if (p == last_found) | |
2672 | p = packed_git; | |
2673 | else | |
2674 | p = p->next; | |
2675 | if (p == last_found) | |
2676 | p = p->next; | |
2677 | } | |
2678 | return 0; | |
2679 | } | |
2680 | ||
abcb8655 JK |
2681 | /* |
2682 | * Store a list of sha1s that are should not be discarded | |
2683 | * because they are either written too recently, or are | |
2684 | * reachable from another object that was. | |
2685 | * | |
2686 | * This is filled by get_object_list. | |
2687 | */ | |
910650d2 | 2688 | static struct oid_array recent_objects; |
abcb8655 | 2689 | |
4ce3621a | 2690 | static int loosened_object_can_be_discarded(const struct object_id *oid, |
dddbad72 | 2691 | timestamp_t mtime) |
d0d46abc JK |
2692 | { |
2693 | if (!unpack_unreachable_expiration) | |
2694 | return 0; | |
2695 | if (mtime > unpack_unreachable_expiration) | |
2696 | return 0; | |
910650d2 | 2697 | if (oid_array_lookup(&recent_objects, oid) >= 0) |
abcb8655 | 2698 | return 0; |
d0d46abc JK |
2699 | return 1; |
2700 | } | |
2701 | ||
ca11b212 NP |
2702 | static void loosen_unused_packed_objects(struct rev_info *revs) |
2703 | { | |
2704 | struct packed_git *p; | |
2705 | uint32_t i; | |
4ce3621a | 2706 | struct object_id oid; |
ca11b212 NP |
2707 | |
2708 | for (p = packed_git; p; p = p->next) { | |
79bc4c71 | 2709 | if (!p->pack_local || p->pack_keep) |
ca11b212 NP |
2710 | continue; |
2711 | ||
2712 | if (open_pack_index(p)) | |
2713 | die("cannot open pack index"); | |
2714 | ||
2715 | for (i = 0; i < p->num_objects; i++) { | |
4ce3621a | 2716 | nth_packed_object_oid(&oid, p, i); |
2717 | if (!packlist_find(&to_pack, oid.hash, NULL) && | |
2718 | !has_sha1_pack_kept_or_nonlocal(oid.hash) && | |
2719 | !loosened_object_can_be_discarded(&oid, p->mtime)) | |
2720 | if (force_object_loose(oid.hash, p->mtime)) | |
ca11b212 NP |
2721 | die("unable to force loose object"); |
2722 | } | |
2723 | } | |
2724 | } | |
2725 | ||
69e4b342 | 2726 | /* |
645c432d KS |
2727 | * This tracks any options which pack-reuse code expects to be on, or which a |
2728 | * reader of the pack might not understand, and which would therefore prevent | |
2729 | * blind reuse of what we have on disk. | |
69e4b342 JK |
2730 | */ |
2731 | static int pack_options_allow_reuse(void) | |
2732 | { | |
9df4a607 JK |
2733 | return pack_to_stdout && |
2734 | allow_ofs_delta && | |
2735 | !ignore_packed_keep && | |
2736 | (!local || !have_non_local_packs) && | |
2737 | !incremental; | |
69e4b342 JK |
2738 | } |
2739 | ||
6b8fda2d VM |
2740 | static int get_object_list_from_bitmap(struct rev_info *revs) |
2741 | { | |
2742 | if (prepare_bitmap_walk(revs) < 0) | |
2743 | return -1; | |
2744 | ||
69e4b342 JK |
2745 | if (pack_options_allow_reuse() && |
2746 | !reuse_partial_packfile_from_bitmap( | |
6b8fda2d VM |
2747 | &reuse_packfile, |
2748 | &reuse_packfile_objects, | |
2749 | &reuse_packfile_offset)) { | |
2750 | assert(reuse_packfile_objects); | |
2751 | nr_result += reuse_packfile_objects; | |
78d2214e | 2752 | display_progress(progress_state, nr_result); |
6b8fda2d VM |
2753 | } |
2754 | ||
2755 | traverse_bitmap_commit_list(&add_object_entry_from_bitmap); | |
2756 | return 0; | |
2757 | } | |
2758 | ||
abcb8655 | 2759 | static void record_recent_object(struct object *obj, |
de1e67d0 | 2760 | const char *name, |
abcb8655 JK |
2761 | void *data) |
2762 | { | |
910650d2 | 2763 | oid_array_append(&recent_objects, &obj->oid); |
abcb8655 JK |
2764 | } |
2765 | ||
2766 | static void record_recent_commit(struct commit *commit, void *data) | |
2767 | { | |
910650d2 | 2768 | oid_array_append(&recent_objects, &commit->object.oid); |
abcb8655 JK |
2769 | } |
2770 | ||
8d1d8f83 | 2771 | static void get_object_list(int ac, const char **av) |
b5d97e6b JH |
2772 | { |
2773 | struct rev_info revs; | |
2774 | char line[1000]; | |
b5d97e6b JH |
2775 | int flags = 0; |
2776 | ||
b5d97e6b JH |
2777 | init_revisions(&revs, NULL); |
2778 | save_commit_buffer = 0; | |
b5d97e6b JH |
2779 | setup_revisions(ac, av, &revs, NULL); |
2780 | ||
b790e0f6 NTND |
2781 | /* make sure shallows are read */ |
2782 | is_repository_shallow(); | |
2783 | ||
b5d97e6b JH |
2784 | while (fgets(line, sizeof(line), stdin) != NULL) { |
2785 | int len = strlen(line); | |
872c930d | 2786 | if (len && line[len - 1] == '\n') |
b5d97e6b JH |
2787 | line[--len] = 0; |
2788 | if (!len) | |
2789 | break; | |
2790 | if (*line == '-') { | |
2791 | if (!strcmp(line, "--not")) { | |
2792 | flags ^= UNINTERESTING; | |
7cc8f971 | 2793 | write_bitmap_index = 0; |
b5d97e6b JH |
2794 | continue; |
2795 | } | |
b790e0f6 | 2796 | if (starts_with(line, "--shallow ")) { |
e92b848c | 2797 | struct object_id oid; |
2798 | if (get_oid_hex(line + 10, &oid)) | |
b790e0f6 | 2799 | die("not an SHA-1 '%s'", line + 10); |
e92b848c | 2800 | register_shallow(&oid); |
f7f91086 | 2801 | use_bitmap_index = 0; |
b790e0f6 NTND |
2802 | continue; |
2803 | } | |
b5d97e6b JH |
2804 | die("not a rev '%s'", line); |
2805 | } | |
8e676e8b | 2806 | if (handle_revision_arg(line, &revs, flags, REVARG_CANNOT_BE_FILENAME)) |
b5d97e6b JH |
2807 | die("bad revision '%s'", line); |
2808 | } | |
2809 | ||
6b8fda2d VM |
2810 | if (use_bitmap_index && !get_object_list_from_bitmap(&revs)) |
2811 | return; | |
2812 | ||
3d51e1b5 MK |
2813 | if (prepare_revision_walk(&revs)) |
2814 | die("revision walk setup failed"); | |
e76a5fb4 | 2815 | mark_edges_uninteresting(&revs, show_edge); |
11c211fa | 2816 | traverse_commit_list(&revs, show_commit, show_object, NULL); |
08cdfb13 | 2817 | |
abcb8655 JK |
2818 | if (unpack_unreachable_expiration) { |
2819 | revs.ignore_missing_links = 1; | |
2820 | if (add_unseen_recent_objects_to_traversal(&revs, | |
2821 | unpack_unreachable_expiration)) | |
2822 | die("unable to add recent objects"); | |
2823 | if (prepare_revision_walk(&revs)) | |
2824 | die("revision walk setup failed"); | |
2825 | traverse_commit_list(&revs, record_recent_commit, | |
2826 | record_recent_object, NULL); | |
2827 | } | |
2828 | ||
08cdfb13 JH |
2829 | if (keep_unreachable) |
2830 | add_objects_in_unpacked_packs(&revs); | |
e26a8c47 JK |
2831 | if (pack_loose_unreachable) |
2832 | add_unreachable_loose_objects(); | |
ca11b212 NP |
2833 | if (unpack_unreachable) |
2834 | loosen_unused_packed_objects(&revs); | |
abcb8655 | 2835 | |
910650d2 | 2836 | oid_array_clear(&recent_objects); |
b5d97e6b JH |
2837 | } |
2838 | ||
99fb6e04 NTND |
2839 | static int option_parse_index_version(const struct option *opt, |
2840 | const char *arg, int unset) | |
2841 | { | |
2842 | char *c; | |
2843 | const char *val = arg; | |
2844 | pack_idx_opts.version = strtoul(val, &c, 10); | |
2845 | if (pack_idx_opts.version > 2) | |
2846 | die(_("unsupported index version %s"), val); | |
2847 | if (*c == ',' && c[1]) | |
2848 | pack_idx_opts.off32_limit = strtoul(c+1, &c, 0); | |
2849 | if (*c || pack_idx_opts.off32_limit & 0x80000000) | |
2850 | die(_("bad index version '%s'"), val); | |
2851 | return 0; | |
2852 | } | |
2853 | ||
7e52f566 JK |
2854 | static int option_parse_unpack_unreachable(const struct option *opt, |
2855 | const char *arg, int unset) | |
2856 | { | |
2857 | if (unset) { | |
2858 | unpack_unreachable = 0; | |
2859 | unpack_unreachable_expiration = 0; | |
2860 | } | |
2861 | else { | |
2862 | unpack_unreachable = 1; | |
2863 | if (arg) | |
2864 | unpack_unreachable_expiration = approxidate(arg); | |
2865 | } | |
2866 | return 0; | |
2867 | } | |
2868 | ||
b5d97e6b JH |
2869 | int cmd_pack_objects(int argc, const char **argv, const char *prefix) |
2870 | { | |
b5d97e6b | 2871 | int use_internal_rev_list = 0; |
8d1d8f83 | 2872 | int thin = 0; |
2dacf26d | 2873 | int shallow = 0; |
4f366275 | 2874 | int all_progress_implied = 0; |
edfbb2aa | 2875 | struct argv_array rp = ARGV_ARRAY_INIT; |
99fb6e04 | 2876 | int rev_list_unpacked = 0, rev_list_all = 0, rev_list_reflog = 0; |
c90f9e13 | 2877 | int rev_list_index = 0; |
99fb6e04 NTND |
2878 | struct option pack_objects_options[] = { |
2879 | OPT_SET_INT('q', "quiet", &progress, | |
4c688120 | 2880 | N_("do not show progress meter"), 0), |
99fb6e04 | 2881 | OPT_SET_INT(0, "progress", &progress, |
4c688120 | 2882 | N_("show progress meter"), 1), |
99fb6e04 | 2883 | OPT_SET_INT(0, "all-progress", &progress, |
4c688120 | 2884 | N_("show progress meter during object writing phase"), 2), |
99fb6e04 NTND |
2885 | OPT_BOOL(0, "all-progress-implied", |
2886 | &all_progress_implied, | |
4c688120 NTND |
2887 | N_("similar to --all-progress when progress meter is shown")), |
2888 | { OPTION_CALLBACK, 0, "index-version", NULL, N_("version[,offset]"), | |
2889 | N_("write the pack index file in the specified idx format version"), | |
99fb6e04 | 2890 | 0, option_parse_index_version }, |
2a514ed8 CB |
2891 | OPT_MAGNITUDE(0, "max-pack-size", &pack_size_limit, |
2892 | N_("maximum size of each output pack file")), | |
99fb6e04 | 2893 | OPT_BOOL(0, "local", &local, |
4c688120 | 2894 | N_("ignore borrowed objects from alternate object store")), |
99fb6e04 | 2895 | OPT_BOOL(0, "incremental", &incremental, |
4c688120 | 2896 | N_("ignore packed objects")), |
99fb6e04 | 2897 | OPT_INTEGER(0, "window", &window, |
4c688120 | 2898 | N_("limit pack window by objects")), |
2a514ed8 CB |
2899 | OPT_MAGNITUDE(0, "window-memory", &window_memory_limit, |
2900 | N_("limit pack window by memory in addition to object limit")), | |
99fb6e04 | 2901 | OPT_INTEGER(0, "depth", &depth, |
4c688120 | 2902 | N_("maximum length of delta chain allowed in the resulting pack")), |
99fb6e04 | 2903 | OPT_BOOL(0, "reuse-delta", &reuse_delta, |
4c688120 | 2904 | N_("reuse existing deltas")), |
99fb6e04 | 2905 | OPT_BOOL(0, "reuse-object", &reuse_object, |
4c688120 | 2906 | N_("reuse existing objects")), |
99fb6e04 | 2907 | OPT_BOOL(0, "delta-base-offset", &allow_ofs_delta, |
4c688120 | 2908 | N_("use OFS_DELTA objects")), |
99fb6e04 | 2909 | OPT_INTEGER(0, "threads", &delta_search_threads, |
4c688120 | 2910 | N_("use threads when searching for best delta matches")), |
99fb6e04 | 2911 | OPT_BOOL(0, "non-empty", &non_empty, |
4c688120 | 2912 | N_("do not create an empty pack output")), |
99fb6e04 | 2913 | OPT_BOOL(0, "revs", &use_internal_rev_list, |
4c688120 | 2914 | N_("read revision arguments from standard input")), |
99fb6e04 | 2915 | { OPTION_SET_INT, 0, "unpacked", &rev_list_unpacked, NULL, |
4c688120 | 2916 | N_("limit the objects to those that are not yet packed"), |
99fb6e04 NTND |
2917 | PARSE_OPT_NOARG | PARSE_OPT_NONEG, NULL, 1 }, |
2918 | { OPTION_SET_INT, 0, "all", &rev_list_all, NULL, | |
4c688120 | 2919 | N_("include objects reachable from any reference"), |
99fb6e04 NTND |
2920 | PARSE_OPT_NOARG | PARSE_OPT_NONEG, NULL, 1 }, |
2921 | { OPTION_SET_INT, 0, "reflog", &rev_list_reflog, NULL, | |
4c688120 | 2922 | N_("include objects referred by reflog entries"), |
99fb6e04 | 2923 | PARSE_OPT_NOARG | PARSE_OPT_NONEG, NULL, 1 }, |
c90f9e13 JK |
2924 | { OPTION_SET_INT, 0, "indexed-objects", &rev_list_index, NULL, |
2925 | N_("include objects referred to by the index"), | |
2926 | PARSE_OPT_NOARG | PARSE_OPT_NONEG, NULL, 1 }, | |
99fb6e04 | 2927 | OPT_BOOL(0, "stdout", &pack_to_stdout, |
4c688120 | 2928 | N_("output pack to stdout")), |
99fb6e04 | 2929 | OPT_BOOL(0, "include-tag", &include_tag, |
4c688120 | 2930 | N_("include tag objects that refer to objects to be packed")), |
99fb6e04 | 2931 | OPT_BOOL(0, "keep-unreachable", &keep_unreachable, |
4c688120 | 2932 | N_("keep unreachable objects")), |
e26a8c47 JK |
2933 | OPT_BOOL(0, "pack-loose-unreachable", &pack_loose_unreachable, |
2934 | N_("pack loose unreachable objects")), | |
4c688120 NTND |
2935 | { OPTION_CALLBACK, 0, "unpack-unreachable", NULL, N_("time"), |
2936 | N_("unpack unreachable objects newer than <time>"), | |
7e52f566 | 2937 | PARSE_OPT_OPTARG, option_parse_unpack_unreachable }, |
99fb6e04 | 2938 | OPT_BOOL(0, "thin", &thin, |
4c688120 | 2939 | N_("create thin packs")), |
2dacf26d | 2940 | OPT_BOOL(0, "shallow", &shallow, |
2941 | N_("create packs suitable for shallow fetches")), | |
99fb6e04 | 2942 | OPT_BOOL(0, "honor-pack-keep", &ignore_packed_keep, |
4c688120 | 2943 | N_("ignore packs that have companion .keep file")), |
99fb6e04 | 2944 | OPT_INTEGER(0, "compression", &pack_compression_level, |
4c688120 | 2945 | N_("pack compression level")), |
99fb6e04 | 2946 | OPT_SET_INT(0, "keep-true-parents", &grafts_replace_parents, |
4c688120 | 2947 | N_("do not hide commits by grafts"), 0), |
6b8fda2d VM |
2948 | OPT_BOOL(0, "use-bitmap-index", &use_bitmap_index, |
2949 | N_("use a bitmap index if available to speed up counting objects")), | |
7cc8f971 VM |
2950 | OPT_BOOL(0, "write-bitmap-index", &write_bitmap_index, |
2951 | N_("write a bitmap index together with the pack index")), | |
99fb6e04 NTND |
2952 | OPT_END(), |
2953 | }; | |
8d1d8f83 | 2954 | |
afc711b8 | 2955 | check_replace_refs = 0; |
dae556bd | 2956 | |
ebcfb379 | 2957 | reset_pack_idx_option(&pack_idx_opts); |
ef90d6d4 | 2958 | git_config(git_pack_config, NULL); |
b5d97e6b JH |
2959 | |
2960 | progress = isatty(2); | |
99fb6e04 NTND |
2961 | argc = parse_options(argc, argv, prefix, pack_objects_options, |
2962 | pack_usage, 0); | |
b5d97e6b | 2963 | |
99fb6e04 NTND |
2964 | if (argc) { |
2965 | base_name = argv[0]; | |
2966 | argc--; | |
b5d97e6b | 2967 | } |
99fb6e04 NTND |
2968 | if (pack_to_stdout != !base_name || argc) |
2969 | usage_with_options(pack_usage, pack_objects_options); | |
b5d97e6b | 2970 | |
edfbb2aa | 2971 | argv_array_push(&rp, "pack-objects"); |
99fb6e04 NTND |
2972 | if (thin) { |
2973 | use_internal_rev_list = 1; | |
2dacf26d | 2974 | argv_array_push(&rp, shallow |
2975 | ? "--objects-edge-aggressive" | |
2976 | : "--objects-edge"); | |
99fb6e04 | 2977 | } else |
edfbb2aa | 2978 | argv_array_push(&rp, "--objects"); |
b5d97e6b | 2979 | |
99fb6e04 NTND |
2980 | if (rev_list_all) { |
2981 | use_internal_rev_list = 1; | |
edfbb2aa | 2982 | argv_array_push(&rp, "--all"); |
99fb6e04 NTND |
2983 | } |
2984 | if (rev_list_reflog) { | |
2985 | use_internal_rev_list = 1; | |
edfbb2aa | 2986 | argv_array_push(&rp, "--reflog"); |
99fb6e04 | 2987 | } |
c90f9e13 JK |
2988 | if (rev_list_index) { |
2989 | use_internal_rev_list = 1; | |
2990 | argv_array_push(&rp, "--indexed-objects"); | |
99fb6e04 NTND |
2991 | } |
2992 | if (rev_list_unpacked) { | |
2993 | use_internal_rev_list = 1; | |
edfbb2aa | 2994 | argv_array_push(&rp, "--unpacked"); |
99fb6e04 | 2995 | } |
b5d97e6b | 2996 | |
99fb6e04 NTND |
2997 | if (!reuse_object) |
2998 | reuse_delta = 0; | |
2999 | if (pack_compression_level == -1) | |
3000 | pack_compression_level = Z_DEFAULT_COMPRESSION; | |
3001 | else if (pack_compression_level < 0 || pack_compression_level > Z_BEST_COMPRESSION) | |
3002 | die("bad pack compression level %d", pack_compression_level); | |
0c45d258 JH |
3003 | |
3004 | if (!delta_search_threads) /* --threads=0 means autodetect */ | |
3005 | delta_search_threads = online_cpus(); | |
3006 | ||
99fb6e04 NTND |
3007 | #ifdef NO_PTHREADS |
3008 | if (delta_search_threads != 1) | |
2b34e486 | 3009 | warning("no threads support, ignoring --threads"); |
99fb6e04 | 3010 | #endif |
2b84b5a8 JS |
3011 | if (!pack_to_stdout && !pack_size_limit) |
3012 | pack_size_limit = pack_size_limit_cfg; | |
01c12a23 DH |
3013 | if (pack_to_stdout && pack_size_limit) |
3014 | die("--max-pack-size cannot be used to build a pack for transfer."); | |
07cf0f24 NP |
3015 | if (pack_size_limit && pack_size_limit < 1024*1024) { |
3016 | warning("minimum pack size limit is 1 MiB"); | |
3017 | pack_size_limit = 1024*1024; | |
3018 | } | |
01c12a23 | 3019 | |
8d1d8f83 JH |
3020 | if (!pack_to_stdout && thin) |
3021 | die("--thin cannot be used to build an indexable pack."); | |
b5d97e6b | 3022 | |
ca11b212 NP |
3023 | if (keep_unreachable && unpack_unreachable) |
3024 | die("--keep-unreachable and --unpack-unreachable are incompatible."); | |
b1e757f3 JK |
3025 | if (!rev_list_all || !rev_list_reflog || !rev_list_index) |
3026 | unpack_unreachable_expiration = 0; | |
ca11b212 | 3027 | |
645c432d KS |
3028 | /* |
3029 | * "soft" reasons not to use bitmaps - for on-disk repack by default we want | |
3030 | * | |
3031 | * - to produce good pack (with bitmap index not-yet-packed objects are | |
3032 | * packed in suboptimal order). | |
3033 | * | |
3034 | * - to use more robust pack-generation codepath (avoiding possible | |
3035 | * bugs in bitmap code and possible bitmap index corruption). | |
3036 | */ | |
3037 | if (!pack_to_stdout) | |
3038 | use_bitmap_index_default = 0; | |
3039 | ||
3040 | if (use_bitmap_index < 0) | |
3041 | use_bitmap_index = use_bitmap_index_default; | |
3042 | ||
3043 | /* "hard" reasons not to use bitmaps; these just won't work at all */ | |
3044 | if (!use_internal_rev_list || (!pack_to_stdout && write_bitmap_index) || is_repository_shallow()) | |
6b8fda2d VM |
3045 | use_bitmap_index = 0; |
3046 | ||
7cc8f971 VM |
3047 | if (pack_to_stdout || !rev_list_all) |
3048 | write_bitmap_index = 0; | |
3049 | ||
4f366275 NP |
3050 | if (progress && all_progress_implied) |
3051 | progress = 2; | |
3052 | ||
b5d97e6b | 3053 | prepare_packed_git(); |
56dfeb62 JK |
3054 | if (ignore_packed_keep) { |
3055 | struct packed_git *p; | |
3056 | for (p = packed_git; p; p = p->next) | |
3057 | if (p->pack_local && p->pack_keep) | |
3058 | break; | |
3059 | if (!p) /* no keep-able packs found */ | |
3060 | ignore_packed_keep = 0; | |
3061 | } | |
3062 | if (local) { | |
3063 | /* | |
3064 | * unlike ignore_packed_keep above, we do not want to | |
3065 | * unset "local" based on looking at packs, as it | |
3066 | * also covers non-local objects | |
3067 | */ | |
3068 | struct packed_git *p; | |
3069 | for (p = packed_git; p; p = p->next) { | |
3070 | if (!p->pack_local) { | |
3071 | have_non_local_packs = 1; | |
3072 | break; | |
3073 | } | |
3074 | } | |
3075 | } | |
b5d97e6b | 3076 | |
13aaf148 | 3077 | if (progress) |
754dbc43 | 3078 | progress_state = start_progress(_("Counting objects"), 0); |
b5d97e6b JH |
3079 | if (!use_internal_rev_list) |
3080 | read_object_list_from_stdin(); | |
8d1d8f83 | 3081 | else { |
edfbb2aa JK |
3082 | get_object_list(rp.argc, rp.argv); |
3083 | argv_array_clear(&rp); | |
8d1d8f83 | 3084 | } |
0ef95f72 | 3085 | cleanup_preferred_base(); |
d155254c MH |
3086 | if (include_tag && nr_result) |
3087 | for_each_ref(add_ref_tag, NULL); | |
4d4fcc54 | 3088 | stop_progress(&progress_state); |
96a02f8f | 3089 | |
f0b0af1b | 3090 | if (non_empty && !nr_result) |
1c4a2912 | 3091 | return 0; |
f7ae6a93 NP |
3092 | if (nr_result) |
3093 | prepare_pack(window, depth); | |
d01fb92f | 3094 | write_pack_file(); |
ab7cd7bb | 3095 | if (progress) |
6e1c2344 RJ |
3096 | fprintf(stderr, "Total %"PRIu32" (delta %"PRIu32")," |
3097 | " reused %"PRIu32" (delta %"PRIu32")\n", | |
67c08ce1 | 3098 | written, written_delta, reused, reused_delta); |
c323ac7d LT |
3099 | return 0; |
3100 | } |