]>
Commit | Line | Data |
---|---|---|
1 | #include "builtin.h" | |
2 | #include "config.h" | |
3 | #include "delta.h" | |
4 | #include "pack.h" | |
5 | #include "csum-file.h" | |
6 | #include "blob.h" | |
7 | #include "commit.h" | |
8 | #include "tag.h" | |
9 | #include "tree.h" | |
10 | #include "progress.h" | |
11 | #include "fsck.h" | |
12 | #include "exec-cmd.h" | |
13 | #include "streaming.h" | |
14 | #include "thread-utils.h" | |
15 | #include "packfile.h" | |
16 | #include "object-store.h" | |
17 | #include "promisor-remote.h" | |
18 | ||
19 | static const char index_pack_usage[] = | |
20 | "git index-pack [-v] [-o <index-file>] [--keep | --keep=<msg>] [--verify] [--strict] (<pack-file> | --stdin [--fix-thin] [<pack-file>])"; | |
21 | ||
22 | struct object_entry { | |
23 | struct pack_idx_entry idx; | |
24 | unsigned long size; | |
25 | unsigned char hdr_size; | |
26 | signed char type; | |
27 | signed char real_type; | |
28 | }; | |
29 | ||
30 | struct object_stat { | |
31 | unsigned delta_depth; | |
32 | int base_object_no; | |
33 | }; | |
34 | ||
35 | struct base_data { | |
36 | /* Initialized by make_base(). */ | |
37 | struct base_data *base; | |
38 | struct object_entry *obj; | |
39 | int ref_first, ref_last; | |
40 | int ofs_first, ofs_last; | |
41 | ||
42 | /* Not initialized by make_base(). */ | |
43 | void *data; | |
44 | unsigned long size; | |
45 | }; | |
46 | ||
47 | struct thread_local { | |
48 | pthread_t thread; | |
49 | size_t base_cache_used; | |
50 | int pack_fd; | |
51 | }; | |
52 | ||
53 | /* Remember to update object flag allocation in object.h */ | |
54 | #define FLAG_LINK (1u<<20) | |
55 | #define FLAG_CHECKED (1u<<21) | |
56 | ||
57 | struct ofs_delta_entry { | |
58 | off_t offset; | |
59 | int obj_no; | |
60 | }; | |
61 | ||
62 | struct ref_delta_entry { | |
63 | struct object_id oid; | |
64 | int obj_no; | |
65 | }; | |
66 | ||
67 | static struct object_entry *objects; | |
68 | static struct object_stat *obj_stat; | |
69 | static struct ofs_delta_entry *ofs_deltas; | |
70 | static struct ref_delta_entry *ref_deltas; | |
71 | static struct thread_local nothread_data; | |
72 | static int nr_objects; | |
73 | static int nr_ofs_deltas; | |
74 | static int nr_ref_deltas; | |
75 | static int ref_deltas_alloc; | |
76 | static int nr_resolved_deltas; | |
77 | static int nr_threads; | |
78 | ||
79 | static int from_stdin; | |
80 | static int strict; | |
81 | static int do_fsck_object; | |
82 | static struct fsck_options fsck_options = FSCK_OPTIONS_STRICT; | |
83 | static int verbose; | |
84 | static int show_resolving_progress; | |
85 | static int show_stat; | |
86 | static int check_self_contained_and_connected; | |
87 | ||
88 | static struct progress *progress; | |
89 | ||
90 | /* We always read in 4kB chunks. */ | |
91 | static unsigned char input_buffer[4096]; | |
92 | static unsigned int input_offset, input_len; | |
93 | static off_t consumed_bytes; | |
94 | static off_t max_input_size; | |
95 | static unsigned deepest_delta; | |
96 | static git_hash_ctx input_ctx; | |
97 | static uint32_t input_crc32; | |
98 | static int input_fd, output_fd; | |
99 | static const char *curr_pack; | |
100 | ||
101 | static struct thread_local *thread_data; | |
102 | static int nr_dispatched; | |
103 | static int threads_active; | |
104 | ||
105 | static pthread_mutex_t read_mutex; | |
106 | #define read_lock() lock_mutex(&read_mutex) | |
107 | #define read_unlock() unlock_mutex(&read_mutex) | |
108 | ||
109 | static pthread_mutex_t counter_mutex; | |
110 | #define counter_lock() lock_mutex(&counter_mutex) | |
111 | #define counter_unlock() unlock_mutex(&counter_mutex) | |
112 | ||
113 | static pthread_mutex_t work_mutex; | |
114 | #define work_lock() lock_mutex(&work_mutex) | |
115 | #define work_unlock() unlock_mutex(&work_mutex) | |
116 | ||
117 | static pthread_mutex_t deepest_delta_mutex; | |
118 | #define deepest_delta_lock() lock_mutex(&deepest_delta_mutex) | |
119 | #define deepest_delta_unlock() unlock_mutex(&deepest_delta_mutex) | |
120 | ||
121 | static pthread_mutex_t type_cas_mutex; | |
122 | #define type_cas_lock() lock_mutex(&type_cas_mutex) | |
123 | #define type_cas_unlock() unlock_mutex(&type_cas_mutex) | |
124 | ||
125 | static pthread_key_t key; | |
126 | ||
127 | static inline void lock_mutex(pthread_mutex_t *mutex) | |
128 | { | |
129 | if (threads_active) | |
130 | pthread_mutex_lock(mutex); | |
131 | } | |
132 | ||
133 | static inline void unlock_mutex(pthread_mutex_t *mutex) | |
134 | { | |
135 | if (threads_active) | |
136 | pthread_mutex_unlock(mutex); | |
137 | } | |
138 | ||
139 | /* | |
140 | * Mutex and conditional variable can't be statically-initialized on Windows. | |
141 | */ | |
142 | static void init_thread(void) | |
143 | { | |
144 | int i; | |
145 | init_recursive_mutex(&read_mutex); | |
146 | pthread_mutex_init(&counter_mutex, NULL); | |
147 | pthread_mutex_init(&work_mutex, NULL); | |
148 | pthread_mutex_init(&type_cas_mutex, NULL); | |
149 | if (show_stat) | |
150 | pthread_mutex_init(&deepest_delta_mutex, NULL); | |
151 | pthread_key_create(&key, NULL); | |
152 | thread_data = xcalloc(nr_threads, sizeof(*thread_data)); | |
153 | for (i = 0; i < nr_threads; i++) { | |
154 | thread_data[i].pack_fd = open(curr_pack, O_RDONLY); | |
155 | if (thread_data[i].pack_fd == -1) | |
156 | die_errno(_("unable to open %s"), curr_pack); | |
157 | } | |
158 | ||
159 | threads_active = 1; | |
160 | } | |
161 | ||
162 | static void cleanup_thread(void) | |
163 | { | |
164 | int i; | |
165 | if (!threads_active) | |
166 | return; | |
167 | threads_active = 0; | |
168 | pthread_mutex_destroy(&read_mutex); | |
169 | pthread_mutex_destroy(&counter_mutex); | |
170 | pthread_mutex_destroy(&work_mutex); | |
171 | pthread_mutex_destroy(&type_cas_mutex); | |
172 | if (show_stat) | |
173 | pthread_mutex_destroy(&deepest_delta_mutex); | |
174 | for (i = 0; i < nr_threads; i++) | |
175 | close(thread_data[i].pack_fd); | |
176 | pthread_key_delete(key); | |
177 | free(thread_data); | |
178 | } | |
179 | ||
180 | static int mark_link(struct object *obj, int type, void *data, struct fsck_options *options) | |
181 | { | |
182 | if (!obj) | |
183 | return -1; | |
184 | ||
185 | if (type != OBJ_ANY && obj->type != type) | |
186 | die(_("object type mismatch at %s"), oid_to_hex(&obj->oid)); | |
187 | ||
188 | obj->flags |= FLAG_LINK; | |
189 | return 0; | |
190 | } | |
191 | ||
192 | /* The content of each linked object must have been checked | |
193 | or it must be already present in the object database */ | |
194 | static unsigned check_object(struct object *obj) | |
195 | { | |
196 | if (!obj) | |
197 | return 0; | |
198 | ||
199 | if (!(obj->flags & FLAG_LINK)) | |
200 | return 0; | |
201 | ||
202 | if (!(obj->flags & FLAG_CHECKED)) { | |
203 | unsigned long size; | |
204 | int type = oid_object_info(the_repository, &obj->oid, &size); | |
205 | if (type <= 0) | |
206 | die(_("did not receive expected object %s"), | |
207 | oid_to_hex(&obj->oid)); | |
208 | if (type != obj->type) | |
209 | die(_("object %s: expected type %s, found %s"), | |
210 | oid_to_hex(&obj->oid), | |
211 | type_name(obj->type), type_name(type)); | |
212 | obj->flags |= FLAG_CHECKED; | |
213 | return 1; | |
214 | } | |
215 | ||
216 | return 0; | |
217 | } | |
218 | ||
219 | static unsigned check_objects(void) | |
220 | { | |
221 | unsigned i, max, foreign_nr = 0; | |
222 | ||
223 | max = get_max_object_index(); | |
224 | ||
225 | if (verbose) | |
226 | progress = start_delayed_progress(_("Checking objects"), max); | |
227 | ||
228 | for (i = 0; i < max; i++) { | |
229 | foreign_nr += check_object(get_indexed_object(i)); | |
230 | display_progress(progress, i + 1); | |
231 | } | |
232 | ||
233 | stop_progress(&progress); | |
234 | return foreign_nr; | |
235 | } | |
236 | ||
237 | ||
238 | /* Discard current buffer used content. */ | |
239 | static void flush(void) | |
240 | { | |
241 | if (input_offset) { | |
242 | if (output_fd >= 0) | |
243 | write_or_die(output_fd, input_buffer, input_offset); | |
244 | the_hash_algo->update_fn(&input_ctx, input_buffer, input_offset); | |
245 | memmove(input_buffer, input_buffer + input_offset, input_len); | |
246 | input_offset = 0; | |
247 | } | |
248 | } | |
249 | ||
250 | /* | |
251 | * Make sure at least "min" bytes are available in the buffer, and | |
252 | * return the pointer to the buffer. | |
253 | */ | |
254 | static void *fill(int min) | |
255 | { | |
256 | if (min <= input_len) | |
257 | return input_buffer + input_offset; | |
258 | if (min > sizeof(input_buffer)) | |
259 | die(Q_("cannot fill %d byte", | |
260 | "cannot fill %d bytes", | |
261 | min), | |
262 | min); | |
263 | flush(); | |
264 | do { | |
265 | ssize_t ret = xread(input_fd, input_buffer + input_len, | |
266 | sizeof(input_buffer) - input_len); | |
267 | if (ret <= 0) { | |
268 | if (!ret) | |
269 | die(_("early EOF")); | |
270 | die_errno(_("read error on input")); | |
271 | } | |
272 | input_len += ret; | |
273 | if (from_stdin) | |
274 | display_throughput(progress, consumed_bytes + input_len); | |
275 | } while (input_len < min); | |
276 | return input_buffer; | |
277 | } | |
278 | ||
279 | static void use(int bytes) | |
280 | { | |
281 | if (bytes > input_len) | |
282 | die(_("used more bytes than were available")); | |
283 | input_crc32 = crc32(input_crc32, input_buffer + input_offset, bytes); | |
284 | input_len -= bytes; | |
285 | input_offset += bytes; | |
286 | ||
287 | /* make sure off_t is sufficiently large not to wrap */ | |
288 | if (signed_add_overflows(consumed_bytes, bytes)) | |
289 | die(_("pack too large for current definition of off_t")); | |
290 | consumed_bytes += bytes; | |
291 | if (max_input_size && consumed_bytes > max_input_size) | |
292 | die(_("pack exceeds maximum allowed size")); | |
293 | } | |
294 | ||
295 | static const char *open_pack_file(const char *pack_name) | |
296 | { | |
297 | if (from_stdin) { | |
298 | input_fd = 0; | |
299 | if (!pack_name) { | |
300 | struct strbuf tmp_file = STRBUF_INIT; | |
301 | output_fd = odb_mkstemp(&tmp_file, | |
302 | "pack/tmp_pack_XXXXXX"); | |
303 | pack_name = strbuf_detach(&tmp_file, NULL); | |
304 | } else { | |
305 | output_fd = open(pack_name, O_CREAT|O_EXCL|O_RDWR, 0600); | |
306 | if (output_fd < 0) | |
307 | die_errno(_("unable to create '%s'"), pack_name); | |
308 | } | |
309 | nothread_data.pack_fd = output_fd; | |
310 | } else { | |
311 | input_fd = open(pack_name, O_RDONLY); | |
312 | if (input_fd < 0) | |
313 | die_errno(_("cannot open packfile '%s'"), pack_name); | |
314 | output_fd = -1; | |
315 | nothread_data.pack_fd = input_fd; | |
316 | } | |
317 | the_hash_algo->init_fn(&input_ctx); | |
318 | return pack_name; | |
319 | } | |
320 | ||
321 | static void parse_pack_header(void) | |
322 | { | |
323 | struct pack_header *hdr = fill(sizeof(struct pack_header)); | |
324 | ||
325 | /* Header consistency check */ | |
326 | if (hdr->hdr_signature != htonl(PACK_SIGNATURE)) | |
327 | die(_("pack signature mismatch")); | |
328 | if (!pack_version_ok(hdr->hdr_version)) | |
329 | die(_("pack version %"PRIu32" unsupported"), | |
330 | ntohl(hdr->hdr_version)); | |
331 | ||
332 | nr_objects = ntohl(hdr->hdr_entries); | |
333 | use(sizeof(struct pack_header)); | |
334 | } | |
335 | ||
336 | static NORETURN void bad_object(off_t offset, const char *format, | |
337 | ...) __attribute__((format (printf, 2, 3))); | |
338 | ||
339 | static NORETURN void bad_object(off_t offset, const char *format, ...) | |
340 | { | |
341 | va_list params; | |
342 | char buf[1024]; | |
343 | ||
344 | va_start(params, format); | |
345 | vsnprintf(buf, sizeof(buf), format, params); | |
346 | va_end(params); | |
347 | die(_("pack has bad object at offset %"PRIuMAX": %s"), | |
348 | (uintmax_t)offset, buf); | |
349 | } | |
350 | ||
351 | static inline struct thread_local *get_thread_data(void) | |
352 | { | |
353 | if (HAVE_THREADS) { | |
354 | if (threads_active) | |
355 | return pthread_getspecific(key); | |
356 | assert(!threads_active && | |
357 | "This should only be reached when all threads are gone"); | |
358 | } | |
359 | return ¬hread_data; | |
360 | } | |
361 | ||
362 | static void set_thread_data(struct thread_local *data) | |
363 | { | |
364 | if (threads_active) | |
365 | pthread_setspecific(key, data); | |
366 | } | |
367 | ||
368 | static void free_base_data(struct base_data *c) | |
369 | { | |
370 | if (c->data) { | |
371 | FREE_AND_NULL(c->data); | |
372 | get_thread_data()->base_cache_used -= c->size; | |
373 | } | |
374 | } | |
375 | ||
376 | static void prune_base_data(struct base_data *youngest_child) | |
377 | { | |
378 | struct base_data *b; | |
379 | struct thread_local *data = get_thread_data(); | |
380 | struct base_data **ancestry = NULL; | |
381 | size_t nr = 0, alloc = 0; | |
382 | ssize_t i; | |
383 | ||
384 | if (data->base_cache_used <= delta_base_cache_limit) | |
385 | return; | |
386 | ||
387 | /* | |
388 | * Free all ancestors of youngest_child until we have enough space, | |
389 | * starting with the oldest. (We cannot free youngest_child itself.) | |
390 | */ | |
391 | for (b = youngest_child->base; b != NULL; b = b->base) { | |
392 | ALLOC_GROW(ancestry, nr + 1, alloc); | |
393 | ancestry[nr++] = b; | |
394 | } | |
395 | for (i = nr - 1; | |
396 | i >= 0 && data->base_cache_used > delta_base_cache_limit; | |
397 | i--) { | |
398 | if (ancestry[i]->data) | |
399 | free_base_data(ancestry[i]); | |
400 | } | |
401 | free(ancestry); | |
402 | } | |
403 | ||
404 | static int is_delta_type(enum object_type type) | |
405 | { | |
406 | return (type == OBJ_REF_DELTA || type == OBJ_OFS_DELTA); | |
407 | } | |
408 | ||
409 | static void *unpack_entry_data(off_t offset, unsigned long size, | |
410 | enum object_type type, struct object_id *oid) | |
411 | { | |
412 | static char fixed_buf[8192]; | |
413 | int status; | |
414 | git_zstream stream; | |
415 | void *buf; | |
416 | git_hash_ctx c; | |
417 | char hdr[32]; | |
418 | int hdrlen; | |
419 | ||
420 | if (!is_delta_type(type)) { | |
421 | hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %"PRIuMAX, | |
422 | type_name(type),(uintmax_t)size) + 1; | |
423 | the_hash_algo->init_fn(&c); | |
424 | the_hash_algo->update_fn(&c, hdr, hdrlen); | |
425 | } else | |
426 | oid = NULL; | |
427 | if (type == OBJ_BLOB && size > big_file_threshold) | |
428 | buf = fixed_buf; | |
429 | else | |
430 | buf = xmallocz(size); | |
431 | ||
432 | memset(&stream, 0, sizeof(stream)); | |
433 | git_inflate_init(&stream); | |
434 | stream.next_out = buf; | |
435 | stream.avail_out = buf == fixed_buf ? sizeof(fixed_buf) : size; | |
436 | ||
437 | do { | |
438 | unsigned char *last_out = stream.next_out; | |
439 | stream.next_in = fill(1); | |
440 | stream.avail_in = input_len; | |
441 | status = git_inflate(&stream, 0); | |
442 | use(input_len - stream.avail_in); | |
443 | if (oid) | |
444 | the_hash_algo->update_fn(&c, last_out, stream.next_out - last_out); | |
445 | if (buf == fixed_buf) { | |
446 | stream.next_out = buf; | |
447 | stream.avail_out = sizeof(fixed_buf); | |
448 | } | |
449 | } while (status == Z_OK); | |
450 | if (stream.total_out != size || status != Z_STREAM_END) | |
451 | bad_object(offset, _("inflate returned %d"), status); | |
452 | git_inflate_end(&stream); | |
453 | if (oid) | |
454 | the_hash_algo->final_fn(oid->hash, &c); | |
455 | return buf == fixed_buf ? NULL : buf; | |
456 | } | |
457 | ||
458 | static void *unpack_raw_entry(struct object_entry *obj, | |
459 | off_t *ofs_offset, | |
460 | struct object_id *ref_oid, | |
461 | struct object_id *oid) | |
462 | { | |
463 | unsigned char *p; | |
464 | unsigned long size, c; | |
465 | off_t base_offset; | |
466 | unsigned shift; | |
467 | void *data; | |
468 | ||
469 | obj->idx.offset = consumed_bytes; | |
470 | input_crc32 = crc32(0, NULL, 0); | |
471 | ||
472 | p = fill(1); | |
473 | c = *p; | |
474 | use(1); | |
475 | obj->type = (c >> 4) & 7; | |
476 | size = (c & 15); | |
477 | shift = 4; | |
478 | while (c & 0x80) { | |
479 | p = fill(1); | |
480 | c = *p; | |
481 | use(1); | |
482 | size += (c & 0x7f) << shift; | |
483 | shift += 7; | |
484 | } | |
485 | obj->size = size; | |
486 | ||
487 | switch (obj->type) { | |
488 | case OBJ_REF_DELTA: | |
489 | hashcpy(ref_oid->hash, fill(the_hash_algo->rawsz)); | |
490 | use(the_hash_algo->rawsz); | |
491 | break; | |
492 | case OBJ_OFS_DELTA: | |
493 | p = fill(1); | |
494 | c = *p; | |
495 | use(1); | |
496 | base_offset = c & 127; | |
497 | while (c & 128) { | |
498 | base_offset += 1; | |
499 | if (!base_offset || MSB(base_offset, 7)) | |
500 | bad_object(obj->idx.offset, _("offset value overflow for delta base object")); | |
501 | p = fill(1); | |
502 | c = *p; | |
503 | use(1); | |
504 | base_offset = (base_offset << 7) + (c & 127); | |
505 | } | |
506 | *ofs_offset = obj->idx.offset - base_offset; | |
507 | if (*ofs_offset <= 0 || *ofs_offset >= obj->idx.offset) | |
508 | bad_object(obj->idx.offset, _("delta base offset is out of bound")); | |
509 | break; | |
510 | case OBJ_COMMIT: | |
511 | case OBJ_TREE: | |
512 | case OBJ_BLOB: | |
513 | case OBJ_TAG: | |
514 | break; | |
515 | default: | |
516 | bad_object(obj->idx.offset, _("unknown object type %d"), obj->type); | |
517 | } | |
518 | obj->hdr_size = consumed_bytes - obj->idx.offset; | |
519 | ||
520 | data = unpack_entry_data(obj->idx.offset, obj->size, obj->type, oid); | |
521 | obj->idx.crc32 = input_crc32; | |
522 | return data; | |
523 | } | |
524 | ||
525 | static void *unpack_data(struct object_entry *obj, | |
526 | int (*consume)(const unsigned char *, unsigned long, void *), | |
527 | void *cb_data) | |
528 | { | |
529 | off_t from = obj[0].idx.offset + obj[0].hdr_size; | |
530 | off_t len = obj[1].idx.offset - from; | |
531 | unsigned char *data, *inbuf; | |
532 | git_zstream stream; | |
533 | int status; | |
534 | ||
535 | data = xmallocz(consume ? 64*1024 : obj->size); | |
536 | inbuf = xmalloc((len < 64*1024) ? (int)len : 64*1024); | |
537 | ||
538 | memset(&stream, 0, sizeof(stream)); | |
539 | git_inflate_init(&stream); | |
540 | stream.next_out = data; | |
541 | stream.avail_out = consume ? 64*1024 : obj->size; | |
542 | ||
543 | do { | |
544 | ssize_t n = (len < 64*1024) ? (ssize_t)len : 64*1024; | |
545 | n = xpread(get_thread_data()->pack_fd, inbuf, n, from); | |
546 | if (n < 0) | |
547 | die_errno(_("cannot pread pack file")); | |
548 | if (!n) | |
549 | die(Q_("premature end of pack file, %"PRIuMAX" byte missing", | |
550 | "premature end of pack file, %"PRIuMAX" bytes missing", | |
551 | (unsigned int)len), | |
552 | (uintmax_t)len); | |
553 | from += n; | |
554 | len -= n; | |
555 | stream.next_in = inbuf; | |
556 | stream.avail_in = n; | |
557 | if (!consume) | |
558 | status = git_inflate(&stream, 0); | |
559 | else { | |
560 | do { | |
561 | status = git_inflate(&stream, 0); | |
562 | if (consume(data, stream.next_out - data, cb_data)) { | |
563 | free(inbuf); | |
564 | free(data); | |
565 | return NULL; | |
566 | } | |
567 | stream.next_out = data; | |
568 | stream.avail_out = 64*1024; | |
569 | } while (status == Z_OK && stream.avail_in); | |
570 | } | |
571 | } while (len && status == Z_OK && !stream.avail_in); | |
572 | ||
573 | /* This has been inflated OK when first encountered, so... */ | |
574 | if (status != Z_STREAM_END || stream.total_out != obj->size) | |
575 | die(_("serious inflate inconsistency")); | |
576 | ||
577 | git_inflate_end(&stream); | |
578 | free(inbuf); | |
579 | if (consume) { | |
580 | FREE_AND_NULL(data); | |
581 | } | |
582 | return data; | |
583 | } | |
584 | ||
585 | static void *get_data_from_pack(struct object_entry *obj) | |
586 | { | |
587 | return unpack_data(obj, NULL, NULL); | |
588 | } | |
589 | ||
590 | static int compare_ofs_delta_bases(off_t offset1, off_t offset2, | |
591 | enum object_type type1, | |
592 | enum object_type type2) | |
593 | { | |
594 | int cmp = type1 - type2; | |
595 | if (cmp) | |
596 | return cmp; | |
597 | return offset1 < offset2 ? -1 : | |
598 | offset1 > offset2 ? 1 : | |
599 | 0; | |
600 | } | |
601 | ||
602 | static int find_ofs_delta(const off_t offset) | |
603 | { | |
604 | int first = 0, last = nr_ofs_deltas; | |
605 | ||
606 | while (first < last) { | |
607 | int next = first + (last - first) / 2; | |
608 | struct ofs_delta_entry *delta = &ofs_deltas[next]; | |
609 | int cmp; | |
610 | ||
611 | cmp = compare_ofs_delta_bases(offset, delta->offset, | |
612 | OBJ_OFS_DELTA, | |
613 | objects[delta->obj_no].type); | |
614 | if (!cmp) | |
615 | return next; | |
616 | if (cmp < 0) { | |
617 | last = next; | |
618 | continue; | |
619 | } | |
620 | first = next+1; | |
621 | } | |
622 | return -first-1; | |
623 | } | |
624 | ||
625 | static void find_ofs_delta_children(off_t offset, | |
626 | int *first_index, int *last_index) | |
627 | { | |
628 | int first = find_ofs_delta(offset); | |
629 | int last = first; | |
630 | int end = nr_ofs_deltas - 1; | |
631 | ||
632 | if (first < 0) { | |
633 | *first_index = 0; | |
634 | *last_index = -1; | |
635 | return; | |
636 | } | |
637 | while (first > 0 && ofs_deltas[first - 1].offset == offset) | |
638 | --first; | |
639 | while (last < end && ofs_deltas[last + 1].offset == offset) | |
640 | ++last; | |
641 | *first_index = first; | |
642 | *last_index = last; | |
643 | } | |
644 | ||
645 | static int compare_ref_delta_bases(const struct object_id *oid1, | |
646 | const struct object_id *oid2, | |
647 | enum object_type type1, | |
648 | enum object_type type2) | |
649 | { | |
650 | int cmp = type1 - type2; | |
651 | if (cmp) | |
652 | return cmp; | |
653 | return oidcmp(oid1, oid2); | |
654 | } | |
655 | ||
656 | static int find_ref_delta(const struct object_id *oid) | |
657 | { | |
658 | int first = 0, last = nr_ref_deltas; | |
659 | ||
660 | while (first < last) { | |
661 | int next = first + (last - first) / 2; | |
662 | struct ref_delta_entry *delta = &ref_deltas[next]; | |
663 | int cmp; | |
664 | ||
665 | cmp = compare_ref_delta_bases(oid, &delta->oid, | |
666 | OBJ_REF_DELTA, | |
667 | objects[delta->obj_no].type); | |
668 | if (!cmp) | |
669 | return next; | |
670 | if (cmp < 0) { | |
671 | last = next; | |
672 | continue; | |
673 | } | |
674 | first = next+1; | |
675 | } | |
676 | return -first-1; | |
677 | } | |
678 | ||
679 | static void find_ref_delta_children(const struct object_id *oid, | |
680 | int *first_index, int *last_index) | |
681 | { | |
682 | int first = find_ref_delta(oid); | |
683 | int last = first; | |
684 | int end = nr_ref_deltas - 1; | |
685 | ||
686 | if (first < 0) { | |
687 | *first_index = 0; | |
688 | *last_index = -1; | |
689 | return; | |
690 | } | |
691 | while (first > 0 && oideq(&ref_deltas[first - 1].oid, oid)) | |
692 | --first; | |
693 | while (last < end && oideq(&ref_deltas[last + 1].oid, oid)) | |
694 | ++last; | |
695 | *first_index = first; | |
696 | *last_index = last; | |
697 | } | |
698 | ||
699 | struct compare_data { | |
700 | struct object_entry *entry; | |
701 | struct git_istream *st; | |
702 | unsigned char *buf; | |
703 | unsigned long buf_size; | |
704 | }; | |
705 | ||
706 | static int compare_objects(const unsigned char *buf, unsigned long size, | |
707 | void *cb_data) | |
708 | { | |
709 | struct compare_data *data = cb_data; | |
710 | ||
711 | if (data->buf_size < size) { | |
712 | free(data->buf); | |
713 | data->buf = xmalloc(size); | |
714 | data->buf_size = size; | |
715 | } | |
716 | ||
717 | while (size) { | |
718 | ssize_t len = read_istream(data->st, data->buf, size); | |
719 | if (len == 0) | |
720 | die(_("SHA1 COLLISION FOUND WITH %s !"), | |
721 | oid_to_hex(&data->entry->idx.oid)); | |
722 | if (len < 0) | |
723 | die(_("unable to read %s"), | |
724 | oid_to_hex(&data->entry->idx.oid)); | |
725 | if (memcmp(buf, data->buf, len)) | |
726 | die(_("SHA1 COLLISION FOUND WITH %s !"), | |
727 | oid_to_hex(&data->entry->idx.oid)); | |
728 | size -= len; | |
729 | buf += len; | |
730 | } | |
731 | return 0; | |
732 | } | |
733 | ||
734 | static int check_collison(struct object_entry *entry) | |
735 | { | |
736 | struct compare_data data; | |
737 | enum object_type type; | |
738 | unsigned long size; | |
739 | ||
740 | if (entry->size <= big_file_threshold || entry->type != OBJ_BLOB) | |
741 | return -1; | |
742 | ||
743 | memset(&data, 0, sizeof(data)); | |
744 | data.entry = entry; | |
745 | data.st = open_istream(the_repository, &entry->idx.oid, &type, &size, | |
746 | NULL); | |
747 | if (!data.st) | |
748 | return -1; | |
749 | if (size != entry->size || type != entry->type) | |
750 | die(_("SHA1 COLLISION FOUND WITH %s !"), | |
751 | oid_to_hex(&entry->idx.oid)); | |
752 | unpack_data(entry, compare_objects, &data); | |
753 | close_istream(data.st); | |
754 | free(data.buf); | |
755 | return 0; | |
756 | } | |
757 | ||
758 | static void sha1_object(const void *data, struct object_entry *obj_entry, | |
759 | unsigned long size, enum object_type type, | |
760 | const struct object_id *oid) | |
761 | { | |
762 | void *new_data = NULL; | |
763 | int collision_test_needed = 0; | |
764 | ||
765 | assert(data || obj_entry); | |
766 | ||
767 | if (startup_info->have_repository) { | |
768 | read_lock(); | |
769 | collision_test_needed = | |
770 | has_object_file_with_flags(oid, OBJECT_INFO_QUICK); | |
771 | read_unlock(); | |
772 | } | |
773 | ||
774 | if (collision_test_needed && !data) { | |
775 | read_lock(); | |
776 | if (!check_collison(obj_entry)) | |
777 | collision_test_needed = 0; | |
778 | read_unlock(); | |
779 | } | |
780 | if (collision_test_needed) { | |
781 | void *has_data; | |
782 | enum object_type has_type; | |
783 | unsigned long has_size; | |
784 | read_lock(); | |
785 | has_type = oid_object_info(the_repository, oid, &has_size); | |
786 | if (has_type < 0) | |
787 | die(_("cannot read existing object info %s"), oid_to_hex(oid)); | |
788 | if (has_type != type || has_size != size) | |
789 | die(_("SHA1 COLLISION FOUND WITH %s !"), oid_to_hex(oid)); | |
790 | has_data = read_object_file(oid, &has_type, &has_size); | |
791 | read_unlock(); | |
792 | if (!data) | |
793 | data = new_data = get_data_from_pack(obj_entry); | |
794 | if (!has_data) | |
795 | die(_("cannot read existing object %s"), oid_to_hex(oid)); | |
796 | if (size != has_size || type != has_type || | |
797 | memcmp(data, has_data, size) != 0) | |
798 | die(_("SHA1 COLLISION FOUND WITH %s !"), oid_to_hex(oid)); | |
799 | free(has_data); | |
800 | } | |
801 | ||
802 | if (strict || do_fsck_object) { | |
803 | read_lock(); | |
804 | if (type == OBJ_BLOB) { | |
805 | struct blob *blob = lookup_blob(the_repository, oid); | |
806 | if (blob) | |
807 | blob->object.flags |= FLAG_CHECKED; | |
808 | else | |
809 | die(_("invalid blob object %s"), oid_to_hex(oid)); | |
810 | if (do_fsck_object && | |
811 | fsck_object(&blob->object, (void *)data, size, &fsck_options)) | |
812 | die(_("fsck error in packed object")); | |
813 | } else { | |
814 | struct object *obj; | |
815 | int eaten; | |
816 | void *buf = (void *) data; | |
817 | ||
818 | assert(data && "data can only be NULL for large _blobs_"); | |
819 | ||
820 | /* | |
821 | * we do not need to free the memory here, as the | |
822 | * buf is deleted by the caller. | |
823 | */ | |
824 | obj = parse_object_buffer(the_repository, oid, type, | |
825 | size, buf, | |
826 | &eaten); | |
827 | if (!obj) | |
828 | die(_("invalid %s"), type_name(type)); | |
829 | if (do_fsck_object && | |
830 | fsck_object(obj, buf, size, &fsck_options)) | |
831 | die(_("fsck error in packed object")); | |
832 | if (strict && fsck_walk(obj, NULL, &fsck_options)) | |
833 | die(_("Not all child objects of %s are reachable"), oid_to_hex(&obj->oid)); | |
834 | ||
835 | if (obj->type == OBJ_TREE) { | |
836 | struct tree *item = (struct tree *) obj; | |
837 | item->buffer = NULL; | |
838 | obj->parsed = 0; | |
839 | } | |
840 | if (obj->type == OBJ_COMMIT) { | |
841 | struct commit *commit = (struct commit *) obj; | |
842 | if (detach_commit_buffer(commit, NULL) != data) | |
843 | BUG("parse_object_buffer transmogrified our buffer"); | |
844 | } | |
845 | obj->flags |= FLAG_CHECKED; | |
846 | } | |
847 | read_unlock(); | |
848 | } | |
849 | ||
850 | free(new_data); | |
851 | } | |
852 | ||
853 | /* | |
854 | * This function is part of find_unresolved_deltas(). There are two | |
855 | * walkers going in the opposite ways. | |
856 | * | |
857 | * The first one in find_unresolved_deltas() traverses down from | |
858 | * parent node to children, deflating nodes along the way. However, | |
859 | * memory for deflated nodes is limited by delta_base_cache_limit, so | |
860 | * at some point parent node's deflated content may be freed. | |
861 | * | |
862 | * The second walker is this function, which goes from current node up | |
863 | * to top parent if necessary to deflate the node. In normal | |
864 | * situation, its parent node would be already deflated, so it just | |
865 | * needs to apply delta. | |
866 | * | |
867 | * In the worst case scenario, parent node is no longer deflated because | |
868 | * we're running out of delta_base_cache_limit; we need to re-deflate | |
869 | * parents, possibly up to the top base. | |
870 | * | |
871 | * All deflated objects here are subject to be freed if we exceed | |
872 | * delta_base_cache_limit, just like in find_unresolved_deltas(), we | |
873 | * just need to make sure the last node is not freed. | |
874 | */ | |
875 | static void *get_base_data(struct base_data *c) | |
876 | { | |
877 | if (!c->data) { | |
878 | struct object_entry *obj = c->obj; | |
879 | struct base_data **delta = NULL; | |
880 | int delta_nr = 0, delta_alloc = 0; | |
881 | ||
882 | while (is_delta_type(c->obj->type) && !c->data) { | |
883 | ALLOC_GROW(delta, delta_nr + 1, delta_alloc); | |
884 | delta[delta_nr++] = c; | |
885 | c = c->base; | |
886 | } | |
887 | if (!delta_nr) { | |
888 | c->data = get_data_from_pack(obj); | |
889 | c->size = obj->size; | |
890 | get_thread_data()->base_cache_used += c->size; | |
891 | prune_base_data(c); | |
892 | } | |
893 | for (; delta_nr > 0; delta_nr--) { | |
894 | void *base, *raw; | |
895 | c = delta[delta_nr - 1]; | |
896 | obj = c->obj; | |
897 | base = get_base_data(c->base); | |
898 | raw = get_data_from_pack(obj); | |
899 | c->data = patch_delta( | |
900 | base, c->base->size, | |
901 | raw, obj->size, | |
902 | &c->size); | |
903 | free(raw); | |
904 | if (!c->data) | |
905 | bad_object(obj->idx.offset, _("failed to apply delta")); | |
906 | get_thread_data()->base_cache_used += c->size; | |
907 | prune_base_data(c); | |
908 | } | |
909 | free(delta); | |
910 | } | |
911 | return c->data; | |
912 | } | |
913 | ||
914 | static struct base_data *make_base(struct object_entry *obj, | |
915 | struct base_data *parent) | |
916 | { | |
917 | struct base_data *base = xcalloc(1, sizeof(struct base_data)); | |
918 | base->base = parent; | |
919 | base->obj = obj; | |
920 | find_ref_delta_children(&obj->idx.oid, | |
921 | &base->ref_first, &base->ref_last); | |
922 | find_ofs_delta_children(obj->idx.offset, | |
923 | &base->ofs_first, &base->ofs_last); | |
924 | return base; | |
925 | } | |
926 | ||
927 | static struct base_data *resolve_delta(struct object_entry *delta_obj, | |
928 | struct base_data *base) | |
929 | { | |
930 | void *base_data, *delta_data, *result_data; | |
931 | struct base_data *result; | |
932 | unsigned long result_size; | |
933 | ||
934 | if (show_stat) { | |
935 | int i = delta_obj - objects; | |
936 | int j = base->obj - objects; | |
937 | obj_stat[i].delta_depth = obj_stat[j].delta_depth + 1; | |
938 | deepest_delta_lock(); | |
939 | if (deepest_delta < obj_stat[i].delta_depth) | |
940 | deepest_delta = obj_stat[i].delta_depth; | |
941 | deepest_delta_unlock(); | |
942 | obj_stat[i].base_object_no = j; | |
943 | } | |
944 | delta_data = get_data_from_pack(delta_obj); | |
945 | base_data = get_base_data(base); | |
946 | result_data = patch_delta(base_data, base->size, | |
947 | delta_data, delta_obj->size, &result_size); | |
948 | free(delta_data); | |
949 | if (!result_data) | |
950 | bad_object(delta_obj->idx.offset, _("failed to apply delta")); | |
951 | hash_object_file(the_hash_algo, result_data, result_size, | |
952 | type_name(delta_obj->real_type), &delta_obj->idx.oid); | |
953 | sha1_object(result_data, NULL, result_size, delta_obj->real_type, | |
954 | &delta_obj->idx.oid); | |
955 | ||
956 | result = make_base(delta_obj, base); | |
957 | if (result->ref_last == -1 && result->ofs_last == -1) { | |
958 | free(result_data); | |
959 | } else { | |
960 | result->data = result_data; | |
961 | result->size = result_size; | |
962 | get_thread_data()->base_cache_used += result->size; | |
963 | prune_base_data(result); | |
964 | } | |
965 | ||
966 | counter_lock(); | |
967 | nr_resolved_deltas++; | |
968 | counter_unlock(); | |
969 | ||
970 | return result; | |
971 | } | |
972 | ||
973 | /* | |
974 | * Standard boolean compare-and-swap: atomically check whether "*type" is | |
975 | * "want"; if so, swap in "set" and return true. Otherwise, leave it untouched | |
976 | * and return false. | |
977 | */ | |
978 | static int compare_and_swap_type(signed char *type, | |
979 | enum object_type want, | |
980 | enum object_type set) | |
981 | { | |
982 | enum object_type old; | |
983 | ||
984 | type_cas_lock(); | |
985 | old = *type; | |
986 | if (old == want) | |
987 | *type = set; | |
988 | type_cas_unlock(); | |
989 | ||
990 | return old == want; | |
991 | } | |
992 | ||
993 | static struct base_data *find_unresolved_deltas_1(struct base_data *base, | |
994 | struct base_data *prev_base) | |
995 | { | |
996 | if (base->ref_first <= base->ref_last) { | |
997 | struct object_entry *child = objects + ref_deltas[base->ref_first].obj_no; | |
998 | struct base_data *result; | |
999 | ||
1000 | if (!compare_and_swap_type(&child->real_type, OBJ_REF_DELTA, | |
1001 | base->obj->real_type)) | |
1002 | die("REF_DELTA at offset %"PRIuMAX" already resolved (duplicate base %s?)", | |
1003 | (uintmax_t)child->idx.offset, | |
1004 | oid_to_hex(&base->obj->idx.oid)); | |
1005 | ||
1006 | result = resolve_delta(child, base); | |
1007 | if (base->ref_first == base->ref_last && base->ofs_last == -1) | |
1008 | free_base_data(base); | |
1009 | ||
1010 | base->ref_first++; | |
1011 | return result; | |
1012 | } | |
1013 | ||
1014 | if (base->ofs_first <= base->ofs_last) { | |
1015 | struct object_entry *child = objects + ofs_deltas[base->ofs_first].obj_no; | |
1016 | struct base_data *result; | |
1017 | ||
1018 | assert(child->real_type == OBJ_OFS_DELTA); | |
1019 | child->real_type = base->obj->real_type; | |
1020 | result = resolve_delta(child, base); | |
1021 | if (base->ofs_first == base->ofs_last) | |
1022 | free_base_data(base); | |
1023 | ||
1024 | base->ofs_first++; | |
1025 | return result; | |
1026 | } | |
1027 | ||
1028 | free_base_data(base); | |
1029 | return NULL; | |
1030 | } | |
1031 | ||
1032 | static void find_unresolved_deltas(struct base_data *base) | |
1033 | { | |
1034 | struct base_data *new_base, *prev_base = NULL; | |
1035 | for (;;) { | |
1036 | new_base = find_unresolved_deltas_1(base, prev_base); | |
1037 | ||
1038 | if (new_base) { | |
1039 | prev_base = base; | |
1040 | base = new_base; | |
1041 | } else { | |
1042 | free(base); | |
1043 | base = prev_base; | |
1044 | if (!base) | |
1045 | return; | |
1046 | prev_base = base->base; | |
1047 | } | |
1048 | } | |
1049 | } | |
1050 | ||
1051 | static int compare_ofs_delta_entry(const void *a, const void *b) | |
1052 | { | |
1053 | const struct ofs_delta_entry *delta_a = a; | |
1054 | const struct ofs_delta_entry *delta_b = b; | |
1055 | ||
1056 | return delta_a->offset < delta_b->offset ? -1 : | |
1057 | delta_a->offset > delta_b->offset ? 1 : | |
1058 | 0; | |
1059 | } | |
1060 | ||
1061 | static int compare_ref_delta_entry(const void *a, const void *b) | |
1062 | { | |
1063 | const struct ref_delta_entry *delta_a = a; | |
1064 | const struct ref_delta_entry *delta_b = b; | |
1065 | ||
1066 | return oidcmp(&delta_a->oid, &delta_b->oid); | |
1067 | } | |
1068 | ||
1069 | static void resolve_base(struct object_entry *obj) | |
1070 | { | |
1071 | struct base_data *base_obj = make_base(obj, NULL); | |
1072 | ||
1073 | find_unresolved_deltas(base_obj); | |
1074 | } | |
1075 | ||
1076 | static void *threaded_second_pass(void *data) | |
1077 | { | |
1078 | set_thread_data(data); | |
1079 | for (;;) { | |
1080 | int i; | |
1081 | counter_lock(); | |
1082 | display_progress(progress, nr_resolved_deltas); | |
1083 | counter_unlock(); | |
1084 | work_lock(); | |
1085 | while (nr_dispatched < nr_objects && | |
1086 | is_delta_type(objects[nr_dispatched].type)) | |
1087 | nr_dispatched++; | |
1088 | if (nr_dispatched >= nr_objects) { | |
1089 | work_unlock(); | |
1090 | break; | |
1091 | } | |
1092 | i = nr_dispatched++; | |
1093 | work_unlock(); | |
1094 | ||
1095 | resolve_base(&objects[i]); | |
1096 | } | |
1097 | return NULL; | |
1098 | } | |
1099 | ||
1100 | /* | |
1101 | * First pass: | |
1102 | * - find locations of all objects; | |
1103 | * - calculate SHA1 of all non-delta objects; | |
1104 | * - remember base (SHA1 or offset) for all deltas. | |
1105 | */ | |
1106 | static void parse_pack_objects(unsigned char *hash) | |
1107 | { | |
1108 | int i, nr_delays = 0; | |
1109 | struct ofs_delta_entry *ofs_delta = ofs_deltas; | |
1110 | struct object_id ref_delta_oid; | |
1111 | struct stat st; | |
1112 | ||
1113 | if (verbose) | |
1114 | progress = start_progress( | |
1115 | from_stdin ? _("Receiving objects") : _("Indexing objects"), | |
1116 | nr_objects); | |
1117 | for (i = 0; i < nr_objects; i++) { | |
1118 | struct object_entry *obj = &objects[i]; | |
1119 | void *data = unpack_raw_entry(obj, &ofs_delta->offset, | |
1120 | &ref_delta_oid, | |
1121 | &obj->idx.oid); | |
1122 | obj->real_type = obj->type; | |
1123 | if (obj->type == OBJ_OFS_DELTA) { | |
1124 | nr_ofs_deltas++; | |
1125 | ofs_delta->obj_no = i; | |
1126 | ofs_delta++; | |
1127 | } else if (obj->type == OBJ_REF_DELTA) { | |
1128 | ALLOC_GROW(ref_deltas, nr_ref_deltas + 1, ref_deltas_alloc); | |
1129 | oidcpy(&ref_deltas[nr_ref_deltas].oid, &ref_delta_oid); | |
1130 | ref_deltas[nr_ref_deltas].obj_no = i; | |
1131 | nr_ref_deltas++; | |
1132 | } else if (!data) { | |
1133 | /* large blobs, check later */ | |
1134 | obj->real_type = OBJ_BAD; | |
1135 | nr_delays++; | |
1136 | } else | |
1137 | sha1_object(data, NULL, obj->size, obj->type, | |
1138 | &obj->idx.oid); | |
1139 | free(data); | |
1140 | display_progress(progress, i+1); | |
1141 | } | |
1142 | objects[i].idx.offset = consumed_bytes; | |
1143 | stop_progress(&progress); | |
1144 | ||
1145 | /* Check pack integrity */ | |
1146 | flush(); | |
1147 | the_hash_algo->final_fn(hash, &input_ctx); | |
1148 | if (!hasheq(fill(the_hash_algo->rawsz), hash)) | |
1149 | die(_("pack is corrupted (SHA1 mismatch)")); | |
1150 | use(the_hash_algo->rawsz); | |
1151 | ||
1152 | /* If input_fd is a file, we should have reached its end now. */ | |
1153 | if (fstat(input_fd, &st)) | |
1154 | die_errno(_("cannot fstat packfile")); | |
1155 | if (S_ISREG(st.st_mode) && | |
1156 | lseek(input_fd, 0, SEEK_CUR) - input_len != st.st_size) | |
1157 | die(_("pack has junk at the end")); | |
1158 | ||
1159 | for (i = 0; i < nr_objects; i++) { | |
1160 | struct object_entry *obj = &objects[i]; | |
1161 | if (obj->real_type != OBJ_BAD) | |
1162 | continue; | |
1163 | obj->real_type = obj->type; | |
1164 | sha1_object(NULL, obj, obj->size, obj->type, | |
1165 | &obj->idx.oid); | |
1166 | nr_delays--; | |
1167 | } | |
1168 | if (nr_delays) | |
1169 | die(_("confusion beyond insanity in parse_pack_objects()")); | |
1170 | } | |
1171 | ||
1172 | /* | |
1173 | * Second pass: | |
1174 | * - for all non-delta objects, look if it is used as a base for | |
1175 | * deltas; | |
1176 | * - if used as a base, uncompress the object and apply all deltas, | |
1177 | * recursively checking if the resulting object is used as a base | |
1178 | * for some more deltas. | |
1179 | */ | |
1180 | static void resolve_deltas(void) | |
1181 | { | |
1182 | int i; | |
1183 | ||
1184 | if (!nr_ofs_deltas && !nr_ref_deltas) | |
1185 | return; | |
1186 | ||
1187 | /* Sort deltas by base SHA1/offset for fast searching */ | |
1188 | QSORT(ofs_deltas, nr_ofs_deltas, compare_ofs_delta_entry); | |
1189 | QSORT(ref_deltas, nr_ref_deltas, compare_ref_delta_entry); | |
1190 | ||
1191 | if (verbose || show_resolving_progress) | |
1192 | progress = start_progress(_("Resolving deltas"), | |
1193 | nr_ref_deltas + nr_ofs_deltas); | |
1194 | ||
1195 | nr_dispatched = 0; | |
1196 | if (nr_threads > 1 || getenv("GIT_FORCE_THREADS")) { | |
1197 | init_thread(); | |
1198 | for (i = 0; i < nr_threads; i++) { | |
1199 | int ret = pthread_create(&thread_data[i].thread, NULL, | |
1200 | threaded_second_pass, thread_data + i); | |
1201 | if (ret) | |
1202 | die(_("unable to create thread: %s"), | |
1203 | strerror(ret)); | |
1204 | } | |
1205 | for (i = 0; i < nr_threads; i++) | |
1206 | pthread_join(thread_data[i].thread, NULL); | |
1207 | cleanup_thread(); | |
1208 | return; | |
1209 | } | |
1210 | threaded_second_pass(¬hread_data); | |
1211 | } | |
1212 | ||
1213 | /* | |
1214 | * Third pass: | |
1215 | * - append objects to convert thin pack to full pack if required | |
1216 | * - write the final pack hash | |
1217 | */ | |
1218 | static void fix_unresolved_deltas(struct hashfile *f); | |
1219 | static void conclude_pack(int fix_thin_pack, const char *curr_pack, unsigned char *pack_hash) | |
1220 | { | |
1221 | if (nr_ref_deltas + nr_ofs_deltas == nr_resolved_deltas) { | |
1222 | stop_progress(&progress); | |
1223 | /* Flush remaining pack final hash. */ | |
1224 | flush(); | |
1225 | return; | |
1226 | } | |
1227 | ||
1228 | if (fix_thin_pack) { | |
1229 | struct hashfile *f; | |
1230 | unsigned char read_hash[GIT_MAX_RAWSZ], tail_hash[GIT_MAX_RAWSZ]; | |
1231 | struct strbuf msg = STRBUF_INIT; | |
1232 | int nr_unresolved = nr_ofs_deltas + nr_ref_deltas - nr_resolved_deltas; | |
1233 | int nr_objects_initial = nr_objects; | |
1234 | if (nr_unresolved <= 0) | |
1235 | die(_("confusion beyond insanity")); | |
1236 | REALLOC_ARRAY(objects, nr_objects + nr_unresolved + 1); | |
1237 | memset(objects + nr_objects + 1, 0, | |
1238 | nr_unresolved * sizeof(*objects)); | |
1239 | f = hashfd(output_fd, curr_pack); | |
1240 | fix_unresolved_deltas(f); | |
1241 | strbuf_addf(&msg, Q_("completed with %d local object", | |
1242 | "completed with %d local objects", | |
1243 | nr_objects - nr_objects_initial), | |
1244 | nr_objects - nr_objects_initial); | |
1245 | stop_progress_msg(&progress, msg.buf); | |
1246 | strbuf_release(&msg); | |
1247 | finalize_hashfile(f, tail_hash, 0); | |
1248 | hashcpy(read_hash, pack_hash); | |
1249 | fixup_pack_header_footer(output_fd, pack_hash, | |
1250 | curr_pack, nr_objects, | |
1251 | read_hash, consumed_bytes-the_hash_algo->rawsz); | |
1252 | if (!hasheq(read_hash, tail_hash)) | |
1253 | die(_("Unexpected tail checksum for %s " | |
1254 | "(disk corruption?)"), curr_pack); | |
1255 | } | |
1256 | if (nr_ofs_deltas + nr_ref_deltas != nr_resolved_deltas) | |
1257 | die(Q_("pack has %d unresolved delta", | |
1258 | "pack has %d unresolved deltas", | |
1259 | nr_ofs_deltas + nr_ref_deltas - nr_resolved_deltas), | |
1260 | nr_ofs_deltas + nr_ref_deltas - nr_resolved_deltas); | |
1261 | } | |
1262 | ||
1263 | static int write_compressed(struct hashfile *f, void *in, unsigned int size) | |
1264 | { | |
1265 | git_zstream stream; | |
1266 | int status; | |
1267 | unsigned char outbuf[4096]; | |
1268 | ||
1269 | git_deflate_init(&stream, zlib_compression_level); | |
1270 | stream.next_in = in; | |
1271 | stream.avail_in = size; | |
1272 | ||
1273 | do { | |
1274 | stream.next_out = outbuf; | |
1275 | stream.avail_out = sizeof(outbuf); | |
1276 | status = git_deflate(&stream, Z_FINISH); | |
1277 | hashwrite(f, outbuf, sizeof(outbuf) - stream.avail_out); | |
1278 | } while (status == Z_OK); | |
1279 | ||
1280 | if (status != Z_STREAM_END) | |
1281 | die(_("unable to deflate appended object (%d)"), status); | |
1282 | size = stream.total_out; | |
1283 | git_deflate_end(&stream); | |
1284 | return size; | |
1285 | } | |
1286 | ||
1287 | static struct object_entry *append_obj_to_pack(struct hashfile *f, | |
1288 | const unsigned char *sha1, void *buf, | |
1289 | unsigned long size, enum object_type type) | |
1290 | { | |
1291 | struct object_entry *obj = &objects[nr_objects++]; | |
1292 | unsigned char header[10]; | |
1293 | unsigned long s = size; | |
1294 | int n = 0; | |
1295 | unsigned char c = (type << 4) | (s & 15); | |
1296 | s >>= 4; | |
1297 | while (s) { | |
1298 | header[n++] = c | 0x80; | |
1299 | c = s & 0x7f; | |
1300 | s >>= 7; | |
1301 | } | |
1302 | header[n++] = c; | |
1303 | crc32_begin(f); | |
1304 | hashwrite(f, header, n); | |
1305 | obj[0].size = size; | |
1306 | obj[0].hdr_size = n; | |
1307 | obj[0].type = type; | |
1308 | obj[0].real_type = type; | |
1309 | obj[1].idx.offset = obj[0].idx.offset + n; | |
1310 | obj[1].idx.offset += write_compressed(f, buf, size); | |
1311 | obj[0].idx.crc32 = crc32_end(f); | |
1312 | hashflush(f); | |
1313 | hashcpy(obj->idx.oid.hash, sha1); | |
1314 | return obj; | |
1315 | } | |
1316 | ||
1317 | static int delta_pos_compare(const void *_a, const void *_b) | |
1318 | { | |
1319 | struct ref_delta_entry *a = *(struct ref_delta_entry **)_a; | |
1320 | struct ref_delta_entry *b = *(struct ref_delta_entry **)_b; | |
1321 | return a->obj_no - b->obj_no; | |
1322 | } | |
1323 | ||
1324 | static void fix_unresolved_deltas(struct hashfile *f) | |
1325 | { | |
1326 | struct ref_delta_entry **sorted_by_pos; | |
1327 | int i; | |
1328 | ||
1329 | /* | |
1330 | * Since many unresolved deltas may well be themselves base objects | |
1331 | * for more unresolved deltas, we really want to include the | |
1332 | * smallest number of base objects that would cover as much delta | |
1333 | * as possible by picking the | |
1334 | * trunc deltas first, allowing for other deltas to resolve without | |
1335 | * additional base objects. Since most base objects are to be found | |
1336 | * before deltas depending on them, a good heuristic is to start | |
1337 | * resolving deltas in the same order as their position in the pack. | |
1338 | */ | |
1339 | ALLOC_ARRAY(sorted_by_pos, nr_ref_deltas); | |
1340 | for (i = 0; i < nr_ref_deltas; i++) | |
1341 | sorted_by_pos[i] = &ref_deltas[i]; | |
1342 | QSORT(sorted_by_pos, nr_ref_deltas, delta_pos_compare); | |
1343 | ||
1344 | if (has_promisor_remote()) { | |
1345 | /* | |
1346 | * Prefetch the delta bases. | |
1347 | */ | |
1348 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
1349 | for (i = 0; i < nr_ref_deltas; i++) { | |
1350 | struct ref_delta_entry *d = sorted_by_pos[i]; | |
1351 | if (!oid_object_info_extended(the_repository, &d->oid, | |
1352 | NULL, | |
1353 | OBJECT_INFO_FOR_PREFETCH)) | |
1354 | continue; | |
1355 | oid_array_append(&to_fetch, &d->oid); | |
1356 | } | |
1357 | promisor_remote_get_direct(the_repository, | |
1358 | to_fetch.oid, to_fetch.nr); | |
1359 | oid_array_clear(&to_fetch); | |
1360 | } | |
1361 | ||
1362 | for (i = 0; i < nr_ref_deltas; i++) { | |
1363 | struct ref_delta_entry *d = sorted_by_pos[i]; | |
1364 | enum object_type type; | |
1365 | struct base_data *base; | |
1366 | void *data; | |
1367 | unsigned long size; | |
1368 | struct object_entry *obj; | |
1369 | ||
1370 | if (objects[d->obj_no].real_type != OBJ_REF_DELTA) | |
1371 | continue; | |
1372 | data = read_object_file(&d->oid, &type, &size); | |
1373 | if (!data) | |
1374 | continue; | |
1375 | ||
1376 | if (check_object_signature(the_repository, &d->oid, | |
1377 | data, size, | |
1378 | type_name(type))) | |
1379 | die(_("local object %s is corrupt"), oid_to_hex(&d->oid)); | |
1380 | obj = append_obj_to_pack(f, d->oid.hash, data, size, type); | |
1381 | base = make_base(obj, NULL); | |
1382 | base->data = data; | |
1383 | base->size = size; | |
1384 | find_unresolved_deltas(base); | |
1385 | display_progress(progress, nr_resolved_deltas); | |
1386 | } | |
1387 | free(sorted_by_pos); | |
1388 | } | |
1389 | ||
1390 | static const char *derive_filename(const char *pack_name, const char *suffix, | |
1391 | struct strbuf *buf) | |
1392 | { | |
1393 | size_t len; | |
1394 | if (!strip_suffix(pack_name, ".pack", &len)) | |
1395 | die(_("packfile name '%s' does not end with '.pack'"), | |
1396 | pack_name); | |
1397 | strbuf_add(buf, pack_name, len); | |
1398 | strbuf_addch(buf, '.'); | |
1399 | strbuf_addstr(buf, suffix); | |
1400 | return buf->buf; | |
1401 | } | |
1402 | ||
1403 | static void write_special_file(const char *suffix, const char *msg, | |
1404 | const char *pack_name, const unsigned char *hash, | |
1405 | const char **report) | |
1406 | { | |
1407 | struct strbuf name_buf = STRBUF_INIT; | |
1408 | const char *filename; | |
1409 | int fd; | |
1410 | int msg_len = strlen(msg); | |
1411 | ||
1412 | if (pack_name) | |
1413 | filename = derive_filename(pack_name, suffix, &name_buf); | |
1414 | else | |
1415 | filename = odb_pack_name(&name_buf, hash, suffix); | |
1416 | ||
1417 | fd = odb_pack_keep(filename); | |
1418 | if (fd < 0) { | |
1419 | if (errno != EEXIST) | |
1420 | die_errno(_("cannot write %s file '%s'"), | |
1421 | suffix, filename); | |
1422 | } else { | |
1423 | if (msg_len > 0) { | |
1424 | write_or_die(fd, msg, msg_len); | |
1425 | write_or_die(fd, "\n", 1); | |
1426 | } | |
1427 | if (close(fd) != 0) | |
1428 | die_errno(_("cannot close written %s file '%s'"), | |
1429 | suffix, filename); | |
1430 | if (report) | |
1431 | *report = suffix; | |
1432 | } | |
1433 | strbuf_release(&name_buf); | |
1434 | } | |
1435 | ||
1436 | static void final(const char *final_pack_name, const char *curr_pack_name, | |
1437 | const char *final_index_name, const char *curr_index_name, | |
1438 | const char *keep_msg, const char *promisor_msg, | |
1439 | unsigned char *hash) | |
1440 | { | |
1441 | const char *report = "pack"; | |
1442 | struct strbuf pack_name = STRBUF_INIT; | |
1443 | struct strbuf index_name = STRBUF_INIT; | |
1444 | int err; | |
1445 | ||
1446 | if (!from_stdin) { | |
1447 | close(input_fd); | |
1448 | } else { | |
1449 | fsync_or_die(output_fd, curr_pack_name); | |
1450 | err = close(output_fd); | |
1451 | if (err) | |
1452 | die_errno(_("error while closing pack file")); | |
1453 | } | |
1454 | ||
1455 | if (keep_msg) | |
1456 | write_special_file("keep", keep_msg, final_pack_name, hash, | |
1457 | &report); | |
1458 | if (promisor_msg) | |
1459 | write_special_file("promisor", promisor_msg, final_pack_name, | |
1460 | hash, NULL); | |
1461 | ||
1462 | if (final_pack_name != curr_pack_name) { | |
1463 | if (!final_pack_name) | |
1464 | final_pack_name = odb_pack_name(&pack_name, hash, "pack"); | |
1465 | if (finalize_object_file(curr_pack_name, final_pack_name)) | |
1466 | die(_("cannot store pack file")); | |
1467 | } else if (from_stdin) | |
1468 | chmod(final_pack_name, 0444); | |
1469 | ||
1470 | if (final_index_name != curr_index_name) { | |
1471 | if (!final_index_name) | |
1472 | final_index_name = odb_pack_name(&index_name, hash, "idx"); | |
1473 | if (finalize_object_file(curr_index_name, final_index_name)) | |
1474 | die(_("cannot store index file")); | |
1475 | } else | |
1476 | chmod(final_index_name, 0444); | |
1477 | ||
1478 | if (do_fsck_object) { | |
1479 | struct packed_git *p; | |
1480 | p = add_packed_git(final_index_name, strlen(final_index_name), 0); | |
1481 | if (p) | |
1482 | install_packed_git(the_repository, p); | |
1483 | } | |
1484 | ||
1485 | if (!from_stdin) { | |
1486 | printf("%s\n", hash_to_hex(hash)); | |
1487 | } else { | |
1488 | struct strbuf buf = STRBUF_INIT; | |
1489 | ||
1490 | strbuf_addf(&buf, "%s\t%s\n", report, hash_to_hex(hash)); | |
1491 | write_or_die(1, buf.buf, buf.len); | |
1492 | strbuf_release(&buf); | |
1493 | ||
1494 | /* | |
1495 | * Let's just mimic git-unpack-objects here and write | |
1496 | * the last part of the input buffer to stdout. | |
1497 | */ | |
1498 | while (input_len) { | |
1499 | err = xwrite(1, input_buffer + input_offset, input_len); | |
1500 | if (err <= 0) | |
1501 | break; | |
1502 | input_len -= err; | |
1503 | input_offset += err; | |
1504 | } | |
1505 | } | |
1506 | ||
1507 | strbuf_release(&index_name); | |
1508 | strbuf_release(&pack_name); | |
1509 | } | |
1510 | ||
1511 | static int git_index_pack_config(const char *k, const char *v, void *cb) | |
1512 | { | |
1513 | struct pack_idx_option *opts = cb; | |
1514 | ||
1515 | if (!strcmp(k, "pack.indexversion")) { | |
1516 | opts->version = git_config_int(k, v); | |
1517 | if (opts->version > 2) | |
1518 | die(_("bad pack.indexversion=%"PRIu32), opts->version); | |
1519 | return 0; | |
1520 | } | |
1521 | if (!strcmp(k, "pack.threads")) { | |
1522 | nr_threads = git_config_int(k, v); | |
1523 | if (nr_threads < 0) | |
1524 | die(_("invalid number of threads specified (%d)"), | |
1525 | nr_threads); | |
1526 | if (!HAVE_THREADS && nr_threads != 1) { | |
1527 | warning(_("no threads support, ignoring %s"), k); | |
1528 | nr_threads = 1; | |
1529 | } | |
1530 | return 0; | |
1531 | } | |
1532 | return git_default_config(k, v, cb); | |
1533 | } | |
1534 | ||
1535 | static int cmp_uint32(const void *a_, const void *b_) | |
1536 | { | |
1537 | uint32_t a = *((uint32_t *)a_); | |
1538 | uint32_t b = *((uint32_t *)b_); | |
1539 | ||
1540 | return (a < b) ? -1 : (a != b); | |
1541 | } | |
1542 | ||
1543 | static void read_v2_anomalous_offsets(struct packed_git *p, | |
1544 | struct pack_idx_option *opts) | |
1545 | { | |
1546 | const uint32_t *idx1, *idx2; | |
1547 | uint32_t i; | |
1548 | ||
1549 | /* The address of the 4-byte offset table */ | |
1550 | idx1 = (((const uint32_t *)((const uint8_t *)p->index_data + p->crc_offset)) | |
1551 | + p->num_objects /* CRC32 table */ | |
1552 | ); | |
1553 | ||
1554 | /* The address of the 8-byte offset table */ | |
1555 | idx2 = idx1 + p->num_objects; | |
1556 | ||
1557 | for (i = 0; i < p->num_objects; i++) { | |
1558 | uint32_t off = ntohl(idx1[i]); | |
1559 | if (!(off & 0x80000000)) | |
1560 | continue; | |
1561 | off = off & 0x7fffffff; | |
1562 | check_pack_index_ptr(p, &idx2[off * 2]); | |
1563 | if (idx2[off * 2]) | |
1564 | continue; | |
1565 | /* | |
1566 | * The real offset is ntohl(idx2[off * 2]) in high 4 | |
1567 | * octets, and ntohl(idx2[off * 2 + 1]) in low 4 | |
1568 | * octets. But idx2[off * 2] is Zero!!! | |
1569 | */ | |
1570 | ALLOC_GROW(opts->anomaly, opts->anomaly_nr + 1, opts->anomaly_alloc); | |
1571 | opts->anomaly[opts->anomaly_nr++] = ntohl(idx2[off * 2 + 1]); | |
1572 | } | |
1573 | ||
1574 | QSORT(opts->anomaly, opts->anomaly_nr, cmp_uint32); | |
1575 | } | |
1576 | ||
1577 | static void read_idx_option(struct pack_idx_option *opts, const char *pack_name) | |
1578 | { | |
1579 | struct packed_git *p = add_packed_git(pack_name, strlen(pack_name), 1); | |
1580 | ||
1581 | if (!p) | |
1582 | die(_("Cannot open existing pack file '%s'"), pack_name); | |
1583 | if (open_pack_index(p)) | |
1584 | die(_("Cannot open existing pack idx file for '%s'"), pack_name); | |
1585 | ||
1586 | /* Read the attributes from the existing idx file */ | |
1587 | opts->version = p->index_version; | |
1588 | ||
1589 | if (opts->version == 2) | |
1590 | read_v2_anomalous_offsets(p, opts); | |
1591 | ||
1592 | /* | |
1593 | * Get rid of the idx file as we do not need it anymore. | |
1594 | * NEEDSWORK: extract this bit from free_pack_by_name() in | |
1595 | * sha1-file.c, perhaps? It shouldn't matter very much as we | |
1596 | * know we haven't installed this pack (hence we never have | |
1597 | * read anything from it). | |
1598 | */ | |
1599 | close_pack_index(p); | |
1600 | free(p); | |
1601 | } | |
1602 | ||
1603 | static void show_pack_info(int stat_only) | |
1604 | { | |
1605 | int i, baseobjects = nr_objects - nr_ref_deltas - nr_ofs_deltas; | |
1606 | unsigned long *chain_histogram = NULL; | |
1607 | ||
1608 | if (deepest_delta) | |
1609 | chain_histogram = xcalloc(deepest_delta, sizeof(unsigned long)); | |
1610 | ||
1611 | for (i = 0; i < nr_objects; i++) { | |
1612 | struct object_entry *obj = &objects[i]; | |
1613 | ||
1614 | if (is_delta_type(obj->type)) | |
1615 | chain_histogram[obj_stat[i].delta_depth - 1]++; | |
1616 | if (stat_only) | |
1617 | continue; | |
1618 | printf("%s %-6s %"PRIuMAX" %"PRIuMAX" %"PRIuMAX, | |
1619 | oid_to_hex(&obj->idx.oid), | |
1620 | type_name(obj->real_type), (uintmax_t)obj->size, | |
1621 | (uintmax_t)(obj[1].idx.offset - obj->idx.offset), | |
1622 | (uintmax_t)obj->idx.offset); | |
1623 | if (is_delta_type(obj->type)) { | |
1624 | struct object_entry *bobj = &objects[obj_stat[i].base_object_no]; | |
1625 | printf(" %u %s", obj_stat[i].delta_depth, | |
1626 | oid_to_hex(&bobj->idx.oid)); | |
1627 | } | |
1628 | putchar('\n'); | |
1629 | } | |
1630 | ||
1631 | if (baseobjects) | |
1632 | printf_ln(Q_("non delta: %d object", | |
1633 | "non delta: %d objects", | |
1634 | baseobjects), | |
1635 | baseobjects); | |
1636 | for (i = 0; i < deepest_delta; i++) { | |
1637 | if (!chain_histogram[i]) | |
1638 | continue; | |
1639 | printf_ln(Q_("chain length = %d: %lu object", | |
1640 | "chain length = %d: %lu objects", | |
1641 | chain_histogram[i]), | |
1642 | i + 1, | |
1643 | chain_histogram[i]); | |
1644 | } | |
1645 | } | |
1646 | ||
1647 | int cmd_index_pack(int argc, const char **argv, const char *prefix) | |
1648 | { | |
1649 | int i, fix_thin_pack = 0, verify = 0, stat_only = 0; | |
1650 | const char *curr_index; | |
1651 | const char *index_name = NULL, *pack_name = NULL; | |
1652 | const char *keep_msg = NULL; | |
1653 | const char *promisor_msg = NULL; | |
1654 | struct strbuf index_name_buf = STRBUF_INIT; | |
1655 | struct pack_idx_entry **idx_objects; | |
1656 | struct pack_idx_option opts; | |
1657 | unsigned char pack_hash[GIT_MAX_RAWSZ]; | |
1658 | unsigned foreign_nr = 1; /* zero is a "good" value, assume bad */ | |
1659 | int report_end_of_input = 0; | |
1660 | int hash_algo = 0; | |
1661 | ||
1662 | /* | |
1663 | * index-pack never needs to fetch missing objects except when | |
1664 | * REF_DELTA bases are missing (which are explicitly handled). It only | |
1665 | * accesses the repo to do hash collision checks and to check which | |
1666 | * REF_DELTA bases need to be fetched. | |
1667 | */ | |
1668 | fetch_if_missing = 0; | |
1669 | ||
1670 | if (argc == 2 && !strcmp(argv[1], "-h")) | |
1671 | usage(index_pack_usage); | |
1672 | ||
1673 | read_replace_refs = 0; | |
1674 | fsck_options.walk = mark_link; | |
1675 | ||
1676 | reset_pack_idx_option(&opts); | |
1677 | git_config(git_index_pack_config, &opts); | |
1678 | if (prefix && chdir(prefix)) | |
1679 | die(_("Cannot come back to cwd")); | |
1680 | ||
1681 | for (i = 1; i < argc; i++) { | |
1682 | const char *arg = argv[i]; | |
1683 | ||
1684 | if (*arg == '-') { | |
1685 | if (!strcmp(arg, "--stdin")) { | |
1686 | from_stdin = 1; | |
1687 | } else if (!strcmp(arg, "--fix-thin")) { | |
1688 | fix_thin_pack = 1; | |
1689 | } else if (skip_to_optional_arg(arg, "--strict", &arg)) { | |
1690 | strict = 1; | |
1691 | do_fsck_object = 1; | |
1692 | fsck_set_msg_types(&fsck_options, arg); | |
1693 | } else if (!strcmp(arg, "--check-self-contained-and-connected")) { | |
1694 | strict = 1; | |
1695 | check_self_contained_and_connected = 1; | |
1696 | } else if (!strcmp(arg, "--fsck-objects")) { | |
1697 | do_fsck_object = 1; | |
1698 | } else if (!strcmp(arg, "--verify")) { | |
1699 | verify = 1; | |
1700 | } else if (!strcmp(arg, "--verify-stat")) { | |
1701 | verify = 1; | |
1702 | show_stat = 1; | |
1703 | } else if (!strcmp(arg, "--verify-stat-only")) { | |
1704 | verify = 1; | |
1705 | show_stat = 1; | |
1706 | stat_only = 1; | |
1707 | } else if (skip_to_optional_arg(arg, "--keep", &keep_msg)) { | |
1708 | ; /* nothing to do */ | |
1709 | } else if (skip_to_optional_arg(arg, "--promisor", &promisor_msg)) { | |
1710 | ; /* already parsed */ | |
1711 | } else if (starts_with(arg, "--threads=")) { | |
1712 | char *end; | |
1713 | nr_threads = strtoul(arg+10, &end, 0); | |
1714 | if (!arg[10] || *end || nr_threads < 0) | |
1715 | usage(index_pack_usage); | |
1716 | if (!HAVE_THREADS && nr_threads != 1) { | |
1717 | warning(_("no threads support, ignoring %s"), arg); | |
1718 | nr_threads = 1; | |
1719 | } | |
1720 | } else if (starts_with(arg, "--pack_header=")) { | |
1721 | struct pack_header *hdr; | |
1722 | char *c; | |
1723 | ||
1724 | hdr = (struct pack_header *)input_buffer; | |
1725 | hdr->hdr_signature = htonl(PACK_SIGNATURE); | |
1726 | hdr->hdr_version = htonl(strtoul(arg + 14, &c, 10)); | |
1727 | if (*c != ',') | |
1728 | die(_("bad %s"), arg); | |
1729 | hdr->hdr_entries = htonl(strtoul(c + 1, &c, 10)); | |
1730 | if (*c) | |
1731 | die(_("bad %s"), arg); | |
1732 | input_len = sizeof(*hdr); | |
1733 | } else if (!strcmp(arg, "-v")) { | |
1734 | verbose = 1; | |
1735 | } else if (!strcmp(arg, "--show-resolving-progress")) { | |
1736 | show_resolving_progress = 1; | |
1737 | } else if (!strcmp(arg, "--report-end-of-input")) { | |
1738 | report_end_of_input = 1; | |
1739 | } else if (!strcmp(arg, "-o")) { | |
1740 | if (index_name || (i+1) >= argc) | |
1741 | usage(index_pack_usage); | |
1742 | index_name = argv[++i]; | |
1743 | } else if (starts_with(arg, "--index-version=")) { | |
1744 | char *c; | |
1745 | opts.version = strtoul(arg + 16, &c, 10); | |
1746 | if (opts.version > 2) | |
1747 | die(_("bad %s"), arg); | |
1748 | if (*c == ',') | |
1749 | opts.off32_limit = strtoul(c+1, &c, 0); | |
1750 | if (*c || opts.off32_limit & 0x80000000) | |
1751 | die(_("bad %s"), arg); | |
1752 | } else if (skip_prefix(arg, "--max-input-size=", &arg)) { | |
1753 | max_input_size = strtoumax(arg, NULL, 10); | |
1754 | } else if (skip_prefix(arg, "--object-format=", &arg)) { | |
1755 | hash_algo = hash_algo_by_name(arg); | |
1756 | if (hash_algo == GIT_HASH_UNKNOWN) | |
1757 | die(_("unknown hash algorithm '%s'"), arg); | |
1758 | repo_set_hash_algo(the_repository, hash_algo); | |
1759 | } else | |
1760 | usage(index_pack_usage); | |
1761 | continue; | |
1762 | } | |
1763 | ||
1764 | if (pack_name) | |
1765 | usage(index_pack_usage); | |
1766 | pack_name = arg; | |
1767 | } | |
1768 | ||
1769 | if (!pack_name && !from_stdin) | |
1770 | usage(index_pack_usage); | |
1771 | if (fix_thin_pack && !from_stdin) | |
1772 | die(_("--fix-thin cannot be used without --stdin")); | |
1773 | if (from_stdin && !startup_info->have_repository) | |
1774 | die(_("--stdin requires a git repository")); | |
1775 | if (from_stdin && hash_algo) | |
1776 | die(_("--object-format cannot be used with --stdin")); | |
1777 | if (!index_name && pack_name) | |
1778 | index_name = derive_filename(pack_name, "idx", &index_name_buf); | |
1779 | ||
1780 | if (verify) { | |
1781 | if (!index_name) | |
1782 | die(_("--verify with no packfile name given")); | |
1783 | read_idx_option(&opts, index_name); | |
1784 | opts.flags |= WRITE_IDX_VERIFY | WRITE_IDX_STRICT; | |
1785 | } | |
1786 | if (strict) | |
1787 | opts.flags |= WRITE_IDX_STRICT; | |
1788 | ||
1789 | if (HAVE_THREADS && !nr_threads) { | |
1790 | nr_threads = online_cpus(); | |
1791 | /* An experiment showed that more threads does not mean faster */ | |
1792 | if (nr_threads > 3) | |
1793 | nr_threads = 3; | |
1794 | } | |
1795 | ||
1796 | curr_pack = open_pack_file(pack_name); | |
1797 | parse_pack_header(); | |
1798 | objects = xcalloc(st_add(nr_objects, 1), sizeof(struct object_entry)); | |
1799 | if (show_stat) | |
1800 | obj_stat = xcalloc(st_add(nr_objects, 1), sizeof(struct object_stat)); | |
1801 | ofs_deltas = xcalloc(nr_objects, sizeof(struct ofs_delta_entry)); | |
1802 | parse_pack_objects(pack_hash); | |
1803 | if (report_end_of_input) | |
1804 | write_in_full(2, "\0", 1); | |
1805 | resolve_deltas(); | |
1806 | conclude_pack(fix_thin_pack, curr_pack, pack_hash); | |
1807 | free(ofs_deltas); | |
1808 | free(ref_deltas); | |
1809 | if (strict) | |
1810 | foreign_nr = check_objects(); | |
1811 | ||
1812 | if (show_stat) | |
1813 | show_pack_info(stat_only); | |
1814 | ||
1815 | ALLOC_ARRAY(idx_objects, nr_objects); | |
1816 | for (i = 0; i < nr_objects; i++) | |
1817 | idx_objects[i] = &objects[i].idx; | |
1818 | curr_index = write_idx_file(index_name, idx_objects, nr_objects, &opts, pack_hash); | |
1819 | free(idx_objects); | |
1820 | ||
1821 | if (!verify) | |
1822 | final(pack_name, curr_pack, | |
1823 | index_name, curr_index, | |
1824 | keep_msg, promisor_msg, | |
1825 | pack_hash); | |
1826 | else | |
1827 | close(input_fd); | |
1828 | ||
1829 | if (do_fsck_object && fsck_finish(&fsck_options)) | |
1830 | die(_("fsck error in pack objects")); | |
1831 | ||
1832 | free(objects); | |
1833 | strbuf_release(&index_name_buf); | |
1834 | if (pack_name == NULL) | |
1835 | free((void *) curr_pack); | |
1836 | if (index_name == NULL) | |
1837 | free((void *) curr_index); | |
1838 | ||
1839 | /* | |
1840 | * Let the caller know this pack is not self contained | |
1841 | */ | |
1842 | if (check_self_contained_and_connected && foreign_nr) | |
1843 | return 1; | |
1844 | ||
1845 | return 0; | |
1846 | } |