]>
Commit | Line | Data |
---|---|---|
1 | #include "cache.h" | |
2 | #include "delta.h" | |
3 | #include "pack.h" | |
4 | #include "csum-file.h" | |
5 | #include "blob.h" | |
6 | #include "commit.h" | |
7 | #include "tag.h" | |
8 | #include "tree.h" | |
9 | #include "progress.h" | |
10 | #include "fsck.h" | |
11 | ||
12 | static const char index_pack_usage[] = | |
13 | "git index-pack [-v] [-o <index-file>] [{ ---keep | --keep=<msg> }] [--strict] { <pack-file> | --stdin [--fix-thin] [<pack-file>] }"; | |
14 | ||
15 | struct object_entry | |
16 | { | |
17 | struct pack_idx_entry idx; | |
18 | unsigned long size; | |
19 | unsigned int hdr_size; | |
20 | enum object_type type; | |
21 | enum object_type real_type; | |
22 | }; | |
23 | ||
24 | union delta_base { | |
25 | unsigned char sha1[20]; | |
26 | off_t offset; | |
27 | }; | |
28 | ||
29 | /* | |
30 | * Even if sizeof(union delta_base) == 24 on 64-bit archs, we really want | |
31 | * to memcmp() only the first 20 bytes. | |
32 | */ | |
33 | #define UNION_BASE_SZ 20 | |
34 | ||
35 | #define FLAG_LINK (1u<<20) | |
36 | #define FLAG_CHECKED (1u<<21) | |
37 | ||
38 | struct delta_entry | |
39 | { | |
40 | union delta_base base; | |
41 | int obj_no; | |
42 | }; | |
43 | ||
44 | static struct object_entry *objects; | |
45 | static struct delta_entry *deltas; | |
46 | static int nr_objects; | |
47 | static int nr_deltas; | |
48 | static int nr_resolved_deltas; | |
49 | ||
50 | static int from_stdin; | |
51 | static int strict; | |
52 | static int verbose; | |
53 | ||
54 | static struct progress *progress; | |
55 | ||
56 | /* We always read in 4kB chunks. */ | |
57 | static unsigned char input_buffer[4096]; | |
58 | static unsigned int input_offset, input_len; | |
59 | static off_t consumed_bytes; | |
60 | static SHA_CTX input_ctx; | |
61 | static uint32_t input_crc32; | |
62 | static int input_fd, output_fd, pack_fd; | |
63 | ||
64 | static int mark_link(struct object *obj, int type, void *data) | |
65 | { | |
66 | if (!obj) | |
67 | return -1; | |
68 | ||
69 | if (type != OBJ_ANY && obj->type != type) | |
70 | die("object type mismatch at %s", sha1_to_hex(obj->sha1)); | |
71 | ||
72 | obj->flags |= FLAG_LINK; | |
73 | return 0; | |
74 | } | |
75 | ||
76 | /* The content of each linked object must have been checked | |
77 | or it must be already present in the object database */ | |
78 | static void check_object(struct object *obj) | |
79 | { | |
80 | if (!obj) | |
81 | return; | |
82 | ||
83 | if (!(obj->flags & FLAG_LINK)) | |
84 | return; | |
85 | ||
86 | if (!(obj->flags & FLAG_CHECKED)) { | |
87 | unsigned long size; | |
88 | int type = sha1_object_info(obj->sha1, &size); | |
89 | if (type != obj->type || type <= 0) | |
90 | die("object of unexpected type"); | |
91 | obj->flags |= FLAG_CHECKED; | |
92 | return; | |
93 | } | |
94 | } | |
95 | ||
96 | static void check_objects(void) | |
97 | { | |
98 | unsigned i, max; | |
99 | ||
100 | max = get_max_object_index(); | |
101 | for (i = 0; i < max; i++) | |
102 | check_object(get_indexed_object(i)); | |
103 | } | |
104 | ||
105 | ||
106 | /* Discard current buffer used content. */ | |
107 | static void flush(void) | |
108 | { | |
109 | if (input_offset) { | |
110 | if (output_fd >= 0) | |
111 | write_or_die(output_fd, input_buffer, input_offset); | |
112 | SHA1_Update(&input_ctx, input_buffer, input_offset); | |
113 | memmove(input_buffer, input_buffer + input_offset, input_len); | |
114 | input_offset = 0; | |
115 | } | |
116 | } | |
117 | ||
118 | /* | |
119 | * Make sure at least "min" bytes are available in the buffer, and | |
120 | * return the pointer to the buffer. | |
121 | */ | |
122 | static void *fill(int min) | |
123 | { | |
124 | if (min <= input_len) | |
125 | return input_buffer + input_offset; | |
126 | if (min > sizeof(input_buffer)) | |
127 | die("cannot fill %d bytes", min); | |
128 | flush(); | |
129 | do { | |
130 | ssize_t ret = xread(input_fd, input_buffer + input_len, | |
131 | sizeof(input_buffer) - input_len); | |
132 | if (ret <= 0) { | |
133 | if (!ret) | |
134 | die("early EOF"); | |
135 | die("read error on input: %s", strerror(errno)); | |
136 | } | |
137 | input_len += ret; | |
138 | if (from_stdin) | |
139 | display_throughput(progress, consumed_bytes + input_len); | |
140 | } while (input_len < min); | |
141 | return input_buffer; | |
142 | } | |
143 | ||
144 | static void use(int bytes) | |
145 | { | |
146 | if (bytes > input_len) | |
147 | die("used more bytes than were available"); | |
148 | input_crc32 = crc32(input_crc32, input_buffer + input_offset, bytes); | |
149 | input_len -= bytes; | |
150 | input_offset += bytes; | |
151 | ||
152 | /* make sure off_t is sufficiently large not to wrap */ | |
153 | if (consumed_bytes > consumed_bytes + bytes) | |
154 | die("pack too large for current definition of off_t"); | |
155 | consumed_bytes += bytes; | |
156 | } | |
157 | ||
158 | static char *open_pack_file(char *pack_name) | |
159 | { | |
160 | if (from_stdin) { | |
161 | input_fd = 0; | |
162 | if (!pack_name) { | |
163 | static char tmpfile[PATH_MAX]; | |
164 | snprintf(tmpfile, sizeof(tmpfile), | |
165 | "%s/tmp_pack_XXXXXX", get_object_directory()); | |
166 | output_fd = xmkstemp(tmpfile); | |
167 | pack_name = xstrdup(tmpfile); | |
168 | } else | |
169 | output_fd = open(pack_name, O_CREAT|O_EXCL|O_RDWR, 0600); | |
170 | if (output_fd < 0) | |
171 | die("unable to create %s: %s\n", pack_name, strerror(errno)); | |
172 | pack_fd = output_fd; | |
173 | } else { | |
174 | input_fd = open(pack_name, O_RDONLY); | |
175 | if (input_fd < 0) | |
176 | die("cannot open packfile '%s': %s", | |
177 | pack_name, strerror(errno)); | |
178 | output_fd = -1; | |
179 | pack_fd = input_fd; | |
180 | } | |
181 | SHA1_Init(&input_ctx); | |
182 | return pack_name; | |
183 | } | |
184 | ||
185 | static void parse_pack_header(void) | |
186 | { | |
187 | struct pack_header *hdr = fill(sizeof(struct pack_header)); | |
188 | ||
189 | /* Header consistency check */ | |
190 | if (hdr->hdr_signature != htonl(PACK_SIGNATURE)) | |
191 | die("pack signature mismatch"); | |
192 | if (!pack_version_ok(hdr->hdr_version)) | |
193 | die("pack version %"PRIu32" unsupported", | |
194 | ntohl(hdr->hdr_version)); | |
195 | ||
196 | nr_objects = ntohl(hdr->hdr_entries); | |
197 | use(sizeof(struct pack_header)); | |
198 | } | |
199 | ||
200 | static void bad_object(unsigned long offset, const char *format, | |
201 | ...) NORETURN __attribute__((format (printf, 2, 3))); | |
202 | ||
203 | static void bad_object(unsigned long offset, const char *format, ...) | |
204 | { | |
205 | va_list params; | |
206 | char buf[1024]; | |
207 | ||
208 | va_start(params, format); | |
209 | vsnprintf(buf, sizeof(buf), format, params); | |
210 | va_end(params); | |
211 | die("pack has bad object at offset %lu: %s", offset, buf); | |
212 | } | |
213 | ||
214 | static void *unpack_entry_data(unsigned long offset, unsigned long size) | |
215 | { | |
216 | z_stream stream; | |
217 | void *buf = xmalloc(size); | |
218 | ||
219 | memset(&stream, 0, sizeof(stream)); | |
220 | stream.next_out = buf; | |
221 | stream.avail_out = size; | |
222 | stream.next_in = fill(1); | |
223 | stream.avail_in = input_len; | |
224 | inflateInit(&stream); | |
225 | ||
226 | for (;;) { | |
227 | int ret = inflate(&stream, 0); | |
228 | use(input_len - stream.avail_in); | |
229 | if (stream.total_out == size && ret == Z_STREAM_END) | |
230 | break; | |
231 | if (ret != Z_OK) | |
232 | bad_object(offset, "inflate returned %d", ret); | |
233 | stream.next_in = fill(1); | |
234 | stream.avail_in = input_len; | |
235 | } | |
236 | inflateEnd(&stream); | |
237 | return buf; | |
238 | } | |
239 | ||
240 | static void *unpack_raw_entry(struct object_entry *obj, union delta_base *delta_base) | |
241 | { | |
242 | unsigned char *p, c; | |
243 | unsigned long size; | |
244 | off_t base_offset; | |
245 | unsigned shift; | |
246 | void *data; | |
247 | ||
248 | obj->idx.offset = consumed_bytes; | |
249 | input_crc32 = crc32(0, Z_NULL, 0); | |
250 | ||
251 | p = fill(1); | |
252 | c = *p; | |
253 | use(1); | |
254 | obj->type = (c >> 4) & 7; | |
255 | size = (c & 15); | |
256 | shift = 4; | |
257 | while (c & 0x80) { | |
258 | p = fill(1); | |
259 | c = *p; | |
260 | use(1); | |
261 | size += (c & 0x7fUL) << shift; | |
262 | shift += 7; | |
263 | } | |
264 | obj->size = size; | |
265 | ||
266 | switch (obj->type) { | |
267 | case OBJ_REF_DELTA: | |
268 | hashcpy(delta_base->sha1, fill(20)); | |
269 | use(20); | |
270 | break; | |
271 | case OBJ_OFS_DELTA: | |
272 | memset(delta_base, 0, sizeof(*delta_base)); | |
273 | p = fill(1); | |
274 | c = *p; | |
275 | use(1); | |
276 | base_offset = c & 127; | |
277 | while (c & 128) { | |
278 | base_offset += 1; | |
279 | if (!base_offset || MSB(base_offset, 7)) | |
280 | bad_object(obj->idx.offset, "offset value overflow for delta base object"); | |
281 | p = fill(1); | |
282 | c = *p; | |
283 | use(1); | |
284 | base_offset = (base_offset << 7) + (c & 127); | |
285 | } | |
286 | delta_base->offset = obj->idx.offset - base_offset; | |
287 | if (delta_base->offset >= obj->idx.offset) | |
288 | bad_object(obj->idx.offset, "delta base offset is out of bound"); | |
289 | break; | |
290 | case OBJ_COMMIT: | |
291 | case OBJ_TREE: | |
292 | case OBJ_BLOB: | |
293 | case OBJ_TAG: | |
294 | break; | |
295 | default: | |
296 | bad_object(obj->idx.offset, "unknown object type %d", obj->type); | |
297 | } | |
298 | obj->hdr_size = consumed_bytes - obj->idx.offset; | |
299 | ||
300 | data = unpack_entry_data(obj->idx.offset, obj->size); | |
301 | obj->idx.crc32 = input_crc32; | |
302 | return data; | |
303 | } | |
304 | ||
305 | static void *get_data_from_pack(struct object_entry *obj) | |
306 | { | |
307 | off_t from = obj[0].idx.offset + obj[0].hdr_size; | |
308 | unsigned long len = obj[1].idx.offset - from; | |
309 | unsigned long rdy = 0; | |
310 | unsigned char *src, *data; | |
311 | z_stream stream; | |
312 | int st; | |
313 | ||
314 | src = xmalloc(len); | |
315 | data = src; | |
316 | do { | |
317 | ssize_t n = pread(pack_fd, data + rdy, len - rdy, from + rdy); | |
318 | if (n <= 0) | |
319 | die("cannot pread pack file: %s", strerror(errno)); | |
320 | rdy += n; | |
321 | } while (rdy < len); | |
322 | data = xmalloc(obj->size); | |
323 | memset(&stream, 0, sizeof(stream)); | |
324 | stream.next_out = data; | |
325 | stream.avail_out = obj->size; | |
326 | stream.next_in = src; | |
327 | stream.avail_in = len; | |
328 | inflateInit(&stream); | |
329 | while ((st = inflate(&stream, Z_FINISH)) == Z_OK); | |
330 | inflateEnd(&stream); | |
331 | if (st != Z_STREAM_END || stream.total_out != obj->size) | |
332 | die("serious inflate inconsistency"); | |
333 | free(src); | |
334 | return data; | |
335 | } | |
336 | ||
337 | static int find_delta(const union delta_base *base) | |
338 | { | |
339 | int first = 0, last = nr_deltas; | |
340 | ||
341 | while (first < last) { | |
342 | int next = (first + last) / 2; | |
343 | struct delta_entry *delta = &deltas[next]; | |
344 | int cmp; | |
345 | ||
346 | cmp = memcmp(base, &delta->base, UNION_BASE_SZ); | |
347 | if (!cmp) | |
348 | return next; | |
349 | if (cmp < 0) { | |
350 | last = next; | |
351 | continue; | |
352 | } | |
353 | first = next+1; | |
354 | } | |
355 | return -first-1; | |
356 | } | |
357 | ||
358 | static int find_delta_children(const union delta_base *base, | |
359 | int *first_index, int *last_index) | |
360 | { | |
361 | int first = find_delta(base); | |
362 | int last = first; | |
363 | int end = nr_deltas - 1; | |
364 | ||
365 | if (first < 0) | |
366 | return -1; | |
367 | while (first > 0 && !memcmp(&deltas[first - 1].base, base, UNION_BASE_SZ)) | |
368 | --first; | |
369 | while (last < end && !memcmp(&deltas[last + 1].base, base, UNION_BASE_SZ)) | |
370 | ++last; | |
371 | *first_index = first; | |
372 | *last_index = last; | |
373 | return 0; | |
374 | } | |
375 | ||
376 | static void sha1_object(const void *data, unsigned long size, | |
377 | enum object_type type, unsigned char *sha1) | |
378 | { | |
379 | hash_sha1_file(data, size, typename(type), sha1); | |
380 | if (has_sha1_file(sha1)) { | |
381 | void *has_data; | |
382 | enum object_type has_type; | |
383 | unsigned long has_size; | |
384 | has_data = read_sha1_file(sha1, &has_type, &has_size); | |
385 | if (!has_data) | |
386 | die("cannot read existing object %s", sha1_to_hex(sha1)); | |
387 | if (size != has_size || type != has_type || | |
388 | memcmp(data, has_data, size) != 0) | |
389 | die("SHA1 COLLISION FOUND WITH %s !", sha1_to_hex(sha1)); | |
390 | free(has_data); | |
391 | } | |
392 | if (strict) { | |
393 | if (type == OBJ_BLOB) { | |
394 | struct blob *blob = lookup_blob(sha1); | |
395 | if (blob) | |
396 | blob->object.flags |= FLAG_CHECKED; | |
397 | else | |
398 | die("invalid blob object %s", sha1_to_hex(sha1)); | |
399 | } else { | |
400 | struct object *obj; | |
401 | int eaten; | |
402 | void *buf = (void *) data; | |
403 | ||
404 | /* | |
405 | * we do not need to free the memory here, as the | |
406 | * buf is deleted by the caller. | |
407 | */ | |
408 | obj = parse_object_buffer(sha1, type, size, buf, &eaten); | |
409 | if (!obj) | |
410 | die("invalid %s", typename(type)); | |
411 | if (fsck_object(obj, 1, fsck_error_function)) | |
412 | die("Error in object"); | |
413 | if (fsck_walk(obj, mark_link, 0)) | |
414 | die("Not all child objects of %s are reachable", sha1_to_hex(obj->sha1)); | |
415 | ||
416 | if (obj->type == OBJ_TREE) { | |
417 | struct tree *item = (struct tree *) obj; | |
418 | item->buffer = NULL; | |
419 | } | |
420 | if (obj->type == OBJ_COMMIT) { | |
421 | struct commit *commit = (struct commit *) obj; | |
422 | commit->buffer = NULL; | |
423 | } | |
424 | obj->flags |= FLAG_CHECKED; | |
425 | } | |
426 | } | |
427 | } | |
428 | ||
429 | static void resolve_delta(struct object_entry *delta_obj, void *base_data, | |
430 | unsigned long base_size, enum object_type type) | |
431 | { | |
432 | void *delta_data; | |
433 | unsigned long delta_size; | |
434 | void *result; | |
435 | unsigned long result_size; | |
436 | union delta_base delta_base; | |
437 | int j, first, last; | |
438 | ||
439 | delta_obj->real_type = type; | |
440 | delta_data = get_data_from_pack(delta_obj); | |
441 | delta_size = delta_obj->size; | |
442 | result = patch_delta(base_data, base_size, delta_data, delta_size, | |
443 | &result_size); | |
444 | free(delta_data); | |
445 | if (!result) | |
446 | bad_object(delta_obj->idx.offset, "failed to apply delta"); | |
447 | sha1_object(result, result_size, type, delta_obj->idx.sha1); | |
448 | nr_resolved_deltas++; | |
449 | ||
450 | hashcpy(delta_base.sha1, delta_obj->idx.sha1); | |
451 | if (!find_delta_children(&delta_base, &first, &last)) { | |
452 | for (j = first; j <= last; j++) { | |
453 | struct object_entry *child = objects + deltas[j].obj_no; | |
454 | if (child->real_type == OBJ_REF_DELTA) | |
455 | resolve_delta(child, result, result_size, type); | |
456 | } | |
457 | } | |
458 | ||
459 | memset(&delta_base, 0, sizeof(delta_base)); | |
460 | delta_base.offset = delta_obj->idx.offset; | |
461 | if (!find_delta_children(&delta_base, &first, &last)) { | |
462 | for (j = first; j <= last; j++) { | |
463 | struct object_entry *child = objects + deltas[j].obj_no; | |
464 | if (child->real_type == OBJ_OFS_DELTA) | |
465 | resolve_delta(child, result, result_size, type); | |
466 | } | |
467 | } | |
468 | ||
469 | free(result); | |
470 | } | |
471 | ||
472 | static int compare_delta_entry(const void *a, const void *b) | |
473 | { | |
474 | const struct delta_entry *delta_a = a; | |
475 | const struct delta_entry *delta_b = b; | |
476 | return memcmp(&delta_a->base, &delta_b->base, UNION_BASE_SZ); | |
477 | } | |
478 | ||
479 | /* Parse all objects and return the pack content SHA1 hash */ | |
480 | static void parse_pack_objects(unsigned char *sha1) | |
481 | { | |
482 | int i; | |
483 | struct delta_entry *delta = deltas; | |
484 | void *data; | |
485 | struct stat st; | |
486 | ||
487 | /* | |
488 | * First pass: | |
489 | * - find locations of all objects; | |
490 | * - calculate SHA1 of all non-delta objects; | |
491 | * - remember base (SHA1 or offset) for all deltas. | |
492 | */ | |
493 | if (verbose) | |
494 | progress = start_progress( | |
495 | from_stdin ? "Receiving objects" : "Indexing objects", | |
496 | nr_objects); | |
497 | for (i = 0; i < nr_objects; i++) { | |
498 | struct object_entry *obj = &objects[i]; | |
499 | data = unpack_raw_entry(obj, &delta->base); | |
500 | obj->real_type = obj->type; | |
501 | if (obj->type == OBJ_REF_DELTA || obj->type == OBJ_OFS_DELTA) { | |
502 | nr_deltas++; | |
503 | delta->obj_no = i; | |
504 | delta++; | |
505 | } else | |
506 | sha1_object(data, obj->size, obj->type, obj->idx.sha1); | |
507 | free(data); | |
508 | display_progress(progress, i+1); | |
509 | } | |
510 | objects[i].idx.offset = consumed_bytes; | |
511 | stop_progress(&progress); | |
512 | ||
513 | /* Check pack integrity */ | |
514 | flush(); | |
515 | SHA1_Final(sha1, &input_ctx); | |
516 | if (hashcmp(fill(20), sha1)) | |
517 | die("pack is corrupted (SHA1 mismatch)"); | |
518 | use(20); | |
519 | ||
520 | /* If input_fd is a file, we should have reached its end now. */ | |
521 | if (fstat(input_fd, &st)) | |
522 | die("cannot fstat packfile: %s", strerror(errno)); | |
523 | if (S_ISREG(st.st_mode) && | |
524 | lseek(input_fd, 0, SEEK_CUR) - input_len != st.st_size) | |
525 | die("pack has junk at the end"); | |
526 | ||
527 | if (!nr_deltas) | |
528 | return; | |
529 | ||
530 | /* Sort deltas by base SHA1/offset for fast searching */ | |
531 | qsort(deltas, nr_deltas, sizeof(struct delta_entry), | |
532 | compare_delta_entry); | |
533 | ||
534 | /* | |
535 | * Second pass: | |
536 | * - for all non-delta objects, look if it is used as a base for | |
537 | * deltas; | |
538 | * - if used as a base, uncompress the object and apply all deltas, | |
539 | * recursively checking if the resulting object is used as a base | |
540 | * for some more deltas. | |
541 | */ | |
542 | if (verbose) | |
543 | progress = start_progress("Resolving deltas", nr_deltas); | |
544 | for (i = 0; i < nr_objects; i++) { | |
545 | struct object_entry *obj = &objects[i]; | |
546 | union delta_base base; | |
547 | int j, ref, ref_first, ref_last, ofs, ofs_first, ofs_last; | |
548 | ||
549 | if (obj->type == OBJ_REF_DELTA || obj->type == OBJ_OFS_DELTA) | |
550 | continue; | |
551 | hashcpy(base.sha1, obj->idx.sha1); | |
552 | ref = !find_delta_children(&base, &ref_first, &ref_last); | |
553 | memset(&base, 0, sizeof(base)); | |
554 | base.offset = obj->idx.offset; | |
555 | ofs = !find_delta_children(&base, &ofs_first, &ofs_last); | |
556 | if (!ref && !ofs) | |
557 | continue; | |
558 | data = get_data_from_pack(obj); | |
559 | if (ref) | |
560 | for (j = ref_first; j <= ref_last; j++) { | |
561 | struct object_entry *child = objects + deltas[j].obj_no; | |
562 | if (child->real_type == OBJ_REF_DELTA) | |
563 | resolve_delta(child, data, | |
564 | obj->size, obj->type); | |
565 | } | |
566 | if (ofs) | |
567 | for (j = ofs_first; j <= ofs_last; j++) { | |
568 | struct object_entry *child = objects + deltas[j].obj_no; | |
569 | if (child->real_type == OBJ_OFS_DELTA) | |
570 | resolve_delta(child, data, | |
571 | obj->size, obj->type); | |
572 | } | |
573 | free(data); | |
574 | display_progress(progress, nr_resolved_deltas); | |
575 | } | |
576 | } | |
577 | ||
578 | static int write_compressed(int fd, void *in, unsigned int size, uint32_t *obj_crc) | |
579 | { | |
580 | z_stream stream; | |
581 | unsigned long maxsize; | |
582 | void *out; | |
583 | ||
584 | memset(&stream, 0, sizeof(stream)); | |
585 | deflateInit(&stream, zlib_compression_level); | |
586 | maxsize = deflateBound(&stream, size); | |
587 | out = xmalloc(maxsize); | |
588 | ||
589 | /* Compress it */ | |
590 | stream.next_in = in; | |
591 | stream.avail_in = size; | |
592 | stream.next_out = out; | |
593 | stream.avail_out = maxsize; | |
594 | while (deflate(&stream, Z_FINISH) == Z_OK); | |
595 | deflateEnd(&stream); | |
596 | ||
597 | size = stream.total_out; | |
598 | write_or_die(fd, out, size); | |
599 | *obj_crc = crc32(*obj_crc, out, size); | |
600 | free(out); | |
601 | return size; | |
602 | } | |
603 | ||
604 | static void append_obj_to_pack(const unsigned char *sha1, void *buf, | |
605 | unsigned long size, enum object_type type) | |
606 | { | |
607 | struct object_entry *obj = &objects[nr_objects++]; | |
608 | unsigned char header[10]; | |
609 | unsigned long s = size; | |
610 | int n = 0; | |
611 | unsigned char c = (type << 4) | (s & 15); | |
612 | s >>= 4; | |
613 | while (s) { | |
614 | header[n++] = c | 0x80; | |
615 | c = s & 0x7f; | |
616 | s >>= 7; | |
617 | } | |
618 | header[n++] = c; | |
619 | write_or_die(output_fd, header, n); | |
620 | obj[0].idx.crc32 = crc32(0, Z_NULL, 0); | |
621 | obj[0].idx.crc32 = crc32(obj[0].idx.crc32, header, n); | |
622 | obj[1].idx.offset = obj[0].idx.offset + n; | |
623 | obj[1].idx.offset += write_compressed(output_fd, buf, size, &obj[0].idx.crc32); | |
624 | hashcpy(obj->idx.sha1, sha1); | |
625 | } | |
626 | ||
627 | static int delta_pos_compare(const void *_a, const void *_b) | |
628 | { | |
629 | struct delta_entry *a = *(struct delta_entry **)_a; | |
630 | struct delta_entry *b = *(struct delta_entry **)_b; | |
631 | return a->obj_no - b->obj_no; | |
632 | } | |
633 | ||
634 | static void fix_unresolved_deltas(int nr_unresolved) | |
635 | { | |
636 | struct delta_entry **sorted_by_pos; | |
637 | int i, n = 0; | |
638 | ||
639 | /* | |
640 | * Since many unresolved deltas may well be themselves base objects | |
641 | * for more unresolved deltas, we really want to include the | |
642 | * smallest number of base objects that would cover as much delta | |
643 | * as possible by picking the | |
644 | * trunc deltas first, allowing for other deltas to resolve without | |
645 | * additional base objects. Since most base objects are to be found | |
646 | * before deltas depending on them, a good heuristic is to start | |
647 | * resolving deltas in the same order as their position in the pack. | |
648 | */ | |
649 | sorted_by_pos = xmalloc(nr_unresolved * sizeof(*sorted_by_pos)); | |
650 | for (i = 0; i < nr_deltas; i++) { | |
651 | if (objects[deltas[i].obj_no].real_type != OBJ_REF_DELTA) | |
652 | continue; | |
653 | sorted_by_pos[n++] = &deltas[i]; | |
654 | } | |
655 | qsort(sorted_by_pos, n, sizeof(*sorted_by_pos), delta_pos_compare); | |
656 | ||
657 | for (i = 0; i < n; i++) { | |
658 | struct delta_entry *d = sorted_by_pos[i]; | |
659 | void *data; | |
660 | unsigned long size; | |
661 | enum object_type type; | |
662 | int j, first, last; | |
663 | ||
664 | if (objects[d->obj_no].real_type != OBJ_REF_DELTA) | |
665 | continue; | |
666 | data = read_sha1_file(d->base.sha1, &type, &size); | |
667 | if (!data) | |
668 | continue; | |
669 | ||
670 | find_delta_children(&d->base, &first, &last); | |
671 | for (j = first; j <= last; j++) { | |
672 | struct object_entry *child = objects + deltas[j].obj_no; | |
673 | if (child->real_type == OBJ_REF_DELTA) | |
674 | resolve_delta(child, data, size, type); | |
675 | } | |
676 | ||
677 | if (check_sha1_signature(d->base.sha1, data, size, typename(type))) | |
678 | die("local object %s is corrupt", sha1_to_hex(d->base.sha1)); | |
679 | append_obj_to_pack(d->base.sha1, data, size, type); | |
680 | free(data); | |
681 | display_progress(progress, nr_resolved_deltas); | |
682 | } | |
683 | free(sorted_by_pos); | |
684 | } | |
685 | ||
686 | static void final(const char *final_pack_name, const char *curr_pack_name, | |
687 | const char *final_index_name, const char *curr_index_name, | |
688 | const char *keep_name, const char *keep_msg, | |
689 | unsigned char *sha1) | |
690 | { | |
691 | const char *report = "pack"; | |
692 | char name[PATH_MAX]; | |
693 | int err; | |
694 | ||
695 | if (!from_stdin) { | |
696 | close(input_fd); | |
697 | } else { | |
698 | fsync_or_die(output_fd, curr_pack_name); | |
699 | err = close(output_fd); | |
700 | if (err) | |
701 | die("error while closing pack file: %s", strerror(errno)); | |
702 | chmod(curr_pack_name, 0444); | |
703 | } | |
704 | ||
705 | if (keep_msg) { | |
706 | int keep_fd, keep_msg_len = strlen(keep_msg); | |
707 | if (!keep_name) { | |
708 | snprintf(name, sizeof(name), "%s/pack/pack-%s.keep", | |
709 | get_object_directory(), sha1_to_hex(sha1)); | |
710 | keep_name = name; | |
711 | } | |
712 | keep_fd = open(keep_name, O_RDWR|O_CREAT|O_EXCL, 0600); | |
713 | if (keep_fd < 0) { | |
714 | if (errno != EEXIST) | |
715 | die("cannot write keep file"); | |
716 | } else { | |
717 | if (keep_msg_len > 0) { | |
718 | write_or_die(keep_fd, keep_msg, keep_msg_len); | |
719 | write_or_die(keep_fd, "\n", 1); | |
720 | } | |
721 | if (close(keep_fd) != 0) | |
722 | die("cannot write keep file"); | |
723 | report = "keep"; | |
724 | } | |
725 | } | |
726 | ||
727 | if (final_pack_name != curr_pack_name) { | |
728 | if (!final_pack_name) { | |
729 | snprintf(name, sizeof(name), "%s/pack/pack-%s.pack", | |
730 | get_object_directory(), sha1_to_hex(sha1)); | |
731 | final_pack_name = name; | |
732 | } | |
733 | if (move_temp_to_file(curr_pack_name, final_pack_name)) | |
734 | die("cannot store pack file"); | |
735 | } | |
736 | ||
737 | chmod(curr_index_name, 0444); | |
738 | if (final_index_name != curr_index_name) { | |
739 | if (!final_index_name) { | |
740 | snprintf(name, sizeof(name), "%s/pack/pack-%s.idx", | |
741 | get_object_directory(), sha1_to_hex(sha1)); | |
742 | final_index_name = name; | |
743 | } | |
744 | if (move_temp_to_file(curr_index_name, final_index_name)) | |
745 | die("cannot store index file"); | |
746 | } | |
747 | ||
748 | if (!from_stdin) { | |
749 | printf("%s\n", sha1_to_hex(sha1)); | |
750 | } else { | |
751 | char buf[48]; | |
752 | int len = snprintf(buf, sizeof(buf), "%s\t%s\n", | |
753 | report, sha1_to_hex(sha1)); | |
754 | write_or_die(1, buf, len); | |
755 | ||
756 | /* | |
757 | * Let's just mimic git-unpack-objects here and write | |
758 | * the last part of the input buffer to stdout. | |
759 | */ | |
760 | while (input_len) { | |
761 | err = xwrite(1, input_buffer + input_offset, input_len); | |
762 | if (err <= 0) | |
763 | break; | |
764 | input_len -= err; | |
765 | input_offset += err; | |
766 | } | |
767 | } | |
768 | } | |
769 | ||
770 | static int git_index_pack_config(const char *k, const char *v, void *cb) | |
771 | { | |
772 | if (!strcmp(k, "pack.indexversion")) { | |
773 | pack_idx_default_version = git_config_int(k, v); | |
774 | if (pack_idx_default_version > 2) | |
775 | die("bad pack.indexversion=%"PRIu32, | |
776 | pack_idx_default_version); | |
777 | return 0; | |
778 | } | |
779 | return git_default_config(k, v, cb); | |
780 | } | |
781 | ||
782 | int main(int argc, char **argv) | |
783 | { | |
784 | int i, fix_thin_pack = 0; | |
785 | char *curr_pack, *pack_name = NULL; | |
786 | char *curr_index, *index_name = NULL; | |
787 | const char *keep_name = NULL, *keep_msg = NULL; | |
788 | char *index_name_buf = NULL, *keep_name_buf = NULL; | |
789 | struct pack_idx_entry **idx_objects; | |
790 | unsigned char sha1[20]; | |
791 | ||
792 | git_config(git_index_pack_config, NULL); | |
793 | ||
794 | for (i = 1; i < argc; i++) { | |
795 | char *arg = argv[i]; | |
796 | ||
797 | if (*arg == '-') { | |
798 | if (!strcmp(arg, "--stdin")) { | |
799 | from_stdin = 1; | |
800 | } else if (!strcmp(arg, "--fix-thin")) { | |
801 | fix_thin_pack = 1; | |
802 | } else if (!strcmp(arg, "--strict")) { | |
803 | strict = 1; | |
804 | } else if (!strcmp(arg, "--keep")) { | |
805 | keep_msg = ""; | |
806 | } else if (!prefixcmp(arg, "--keep=")) { | |
807 | keep_msg = arg + 7; | |
808 | } else if (!prefixcmp(arg, "--pack_header=")) { | |
809 | struct pack_header *hdr; | |
810 | char *c; | |
811 | ||
812 | hdr = (struct pack_header *)input_buffer; | |
813 | hdr->hdr_signature = htonl(PACK_SIGNATURE); | |
814 | hdr->hdr_version = htonl(strtoul(arg + 14, &c, 10)); | |
815 | if (*c != ',') | |
816 | die("bad %s", arg); | |
817 | hdr->hdr_entries = htonl(strtoul(c + 1, &c, 10)); | |
818 | if (*c) | |
819 | die("bad %s", arg); | |
820 | input_len = sizeof(*hdr); | |
821 | } else if (!strcmp(arg, "-v")) { | |
822 | verbose = 1; | |
823 | } else if (!strcmp(arg, "-o")) { | |
824 | if (index_name || (i+1) >= argc) | |
825 | usage(index_pack_usage); | |
826 | index_name = argv[++i]; | |
827 | } else if (!prefixcmp(arg, "--index-version=")) { | |
828 | char *c; | |
829 | pack_idx_default_version = strtoul(arg + 16, &c, 10); | |
830 | if (pack_idx_default_version > 2) | |
831 | die("bad %s", arg); | |
832 | if (*c == ',') | |
833 | pack_idx_off32_limit = strtoul(c+1, &c, 0); | |
834 | if (*c || pack_idx_off32_limit & 0x80000000) | |
835 | die("bad %s", arg); | |
836 | } else | |
837 | usage(index_pack_usage); | |
838 | continue; | |
839 | } | |
840 | ||
841 | if (pack_name) | |
842 | usage(index_pack_usage); | |
843 | pack_name = arg; | |
844 | } | |
845 | ||
846 | if (!pack_name && !from_stdin) | |
847 | usage(index_pack_usage); | |
848 | if (fix_thin_pack && !from_stdin) | |
849 | die("--fix-thin cannot be used without --stdin"); | |
850 | if (!index_name && pack_name) { | |
851 | int len = strlen(pack_name); | |
852 | if (!has_extension(pack_name, ".pack")) | |
853 | die("packfile name '%s' does not end with '.pack'", | |
854 | pack_name); | |
855 | index_name_buf = xmalloc(len); | |
856 | memcpy(index_name_buf, pack_name, len - 5); | |
857 | strcpy(index_name_buf + len - 5, ".idx"); | |
858 | index_name = index_name_buf; | |
859 | } | |
860 | if (keep_msg && !keep_name && pack_name) { | |
861 | int len = strlen(pack_name); | |
862 | if (!has_extension(pack_name, ".pack")) | |
863 | die("packfile name '%s' does not end with '.pack'", | |
864 | pack_name); | |
865 | keep_name_buf = xmalloc(len); | |
866 | memcpy(keep_name_buf, pack_name, len - 5); | |
867 | strcpy(keep_name_buf + len - 5, ".keep"); | |
868 | keep_name = keep_name_buf; | |
869 | } | |
870 | ||
871 | curr_pack = open_pack_file(pack_name); | |
872 | parse_pack_header(); | |
873 | objects = xmalloc((nr_objects + 1) * sizeof(struct object_entry)); | |
874 | deltas = xmalloc(nr_objects * sizeof(struct delta_entry)); | |
875 | parse_pack_objects(sha1); | |
876 | if (nr_deltas == nr_resolved_deltas) { | |
877 | stop_progress(&progress); | |
878 | /* Flush remaining pack final 20-byte SHA1. */ | |
879 | flush(); | |
880 | } else { | |
881 | if (fix_thin_pack) { | |
882 | char msg[48]; | |
883 | int nr_unresolved = nr_deltas - nr_resolved_deltas; | |
884 | int nr_objects_initial = nr_objects; | |
885 | if (nr_unresolved <= 0) | |
886 | die("confusion beyond insanity"); | |
887 | objects = xrealloc(objects, | |
888 | (nr_objects + nr_unresolved + 1) | |
889 | * sizeof(*objects)); | |
890 | fix_unresolved_deltas(nr_unresolved); | |
891 | sprintf(msg, "completed with %d local objects", | |
892 | nr_objects - nr_objects_initial); | |
893 | stop_progress_msg(&progress, msg); | |
894 | fixup_pack_header_footer(output_fd, sha1, | |
895 | curr_pack, nr_objects); | |
896 | } | |
897 | if (nr_deltas != nr_resolved_deltas) | |
898 | die("pack has %d unresolved deltas", | |
899 | nr_deltas - nr_resolved_deltas); | |
900 | } | |
901 | free(deltas); | |
902 | if (strict) | |
903 | check_objects(); | |
904 | ||
905 | idx_objects = xmalloc((nr_objects) * sizeof(struct pack_idx_entry *)); | |
906 | for (i = 0; i < nr_objects; i++) | |
907 | idx_objects[i] = &objects[i].idx; | |
908 | curr_index = write_idx_file(index_name, idx_objects, nr_objects, sha1); | |
909 | free(idx_objects); | |
910 | ||
911 | final(pack_name, curr_pack, | |
912 | index_name, curr_index, | |
913 | keep_name, keep_msg, | |
914 | sha1); | |
915 | free(objects); | |
916 | free(index_name_buf); | |
917 | free(keep_name_buf); | |
918 | if (pack_name == NULL) | |
919 | free(curr_pack); | |
920 | if (index_name == NULL) | |
921 | free(curr_index); | |
922 | ||
923 | return 0; | |
924 | } |