Merge branch 'jn/gitweb-grep'
[git/git.git] / builtin-unpack-objects.c
1 #include "builtin.h"
2 #include "cache.h"
3 #include "object.h"
4 #include "delta.h"
5 #include "pack.h"
6 #include "blob.h"
7 #include "commit.h"
8 #include "tag.h"
9 #include "tree.h"
10 #include "tree-walk.h"
11 #include "progress.h"
12 #include "decorate.h"
13 #include "fsck.h"
14
15 static int dry_run, quiet, recover, has_errors, strict;
16 static const char unpack_usage[] = "git-unpack-objects [-n] [-q] [-r] [--strict] < pack-file";
17
18 /* We always read in 4kB chunks. */
19 static unsigned char buffer[4096];
20 static unsigned int offset, len;
21 static off_t consumed_bytes;
22 static SHA_CTX ctx;
23
24 struct obj_buffer {
25 char *buffer;
26 unsigned long size;
27 };
28
29 static struct decoration obj_decorate;
30
31 static struct obj_buffer *lookup_object_buffer(struct object *base)
32 {
33 return lookup_decoration(&obj_decorate, base);
34 }
35
36 static void add_object_buffer(struct object *object, char *buffer, unsigned long size)
37 {
38 struct obj_buffer *obj;
39 obj = xcalloc(1, sizeof(struct obj_buffer));
40 obj->buffer = buffer;
41 obj->size = size;
42 if (add_decoration(&obj_decorate, object, obj))
43 die("object %s tried to add buffer twice!", sha1_to_hex(object->sha1));
44 }
45
46 /*
47 * Make sure at least "min" bytes are available in the buffer, and
48 * return the pointer to the buffer.
49 */
50 static void *fill(int min)
51 {
52 if (min <= len)
53 return buffer + offset;
54 if (min > sizeof(buffer))
55 die("cannot fill %d bytes", min);
56 if (offset) {
57 SHA1_Update(&ctx, buffer, offset);
58 memmove(buffer, buffer + offset, len);
59 offset = 0;
60 }
61 do {
62 ssize_t ret = xread(0, buffer + len, sizeof(buffer) - len);
63 if (ret <= 0) {
64 if (!ret)
65 die("early EOF");
66 die("read error on input: %s", strerror(errno));
67 }
68 len += ret;
69 } while (len < min);
70 return buffer;
71 }
72
73 static void use(int bytes)
74 {
75 if (bytes > len)
76 die("used more bytes than were available");
77 len -= bytes;
78 offset += bytes;
79
80 /* make sure off_t is sufficiently large not to wrap */
81 if (consumed_bytes > consumed_bytes + bytes)
82 die("pack too large for current definition of off_t");
83 consumed_bytes += bytes;
84 }
85
86 static void *get_data(unsigned long size)
87 {
88 z_stream stream;
89 void *buf = xmalloc(size);
90
91 memset(&stream, 0, sizeof(stream));
92
93 stream.next_out = buf;
94 stream.avail_out = size;
95 stream.next_in = fill(1);
96 stream.avail_in = len;
97 inflateInit(&stream);
98
99 for (;;) {
100 int ret = inflate(&stream, 0);
101 use(len - stream.avail_in);
102 if (stream.total_out == size && ret == Z_STREAM_END)
103 break;
104 if (ret != Z_OK) {
105 error("inflate returned %d\n", ret);
106 free(buf);
107 buf = NULL;
108 if (!recover)
109 exit(1);
110 has_errors = 1;
111 break;
112 }
113 stream.next_in = fill(1);
114 stream.avail_in = len;
115 }
116 inflateEnd(&stream);
117 return buf;
118 }
119
120 struct delta_info {
121 unsigned char base_sha1[20];
122 unsigned nr;
123 off_t base_offset;
124 unsigned long size;
125 void *delta;
126 struct delta_info *next;
127 };
128
129 static struct delta_info *delta_list;
130
131 static void add_delta_to_list(unsigned nr, unsigned const char *base_sha1,
132 off_t base_offset,
133 void *delta, unsigned long size)
134 {
135 struct delta_info *info = xmalloc(sizeof(*info));
136
137 hashcpy(info->base_sha1, base_sha1);
138 info->base_offset = base_offset;
139 info->size = size;
140 info->delta = delta;
141 info->nr = nr;
142 info->next = delta_list;
143 delta_list = info;
144 }
145
146 struct obj_info {
147 off_t offset;
148 unsigned char sha1[20];
149 struct object *obj;
150 };
151
152 #define FLAG_OPEN (1u<<20)
153 #define FLAG_WRITTEN (1u<<21)
154
155 static struct obj_info *obj_list;
156 unsigned nr_objects;
157
158 static void write_cached_object(struct object *obj)
159 {
160 unsigned char sha1[20];
161 struct obj_buffer *obj_buf = lookup_object_buffer(obj);
162 if (write_sha1_file(obj_buf->buffer, obj_buf->size, typename(obj->type), sha1) < 0)
163 die("failed to write object %s", sha1_to_hex(obj->sha1));
164 obj->flags |= FLAG_WRITTEN;
165 }
166
167 static int check_object(struct object *obj, int type, void *data)
168 {
169 if (!obj)
170 return 0;
171
172 if (obj->flags & FLAG_WRITTEN)
173 return 1;
174
175 if (type != OBJ_ANY && obj->type != type)
176 die("object type mismatch");
177
178 if (!(obj->flags & FLAG_OPEN)) {
179 unsigned long size;
180 int type = sha1_object_info(obj->sha1, &size);
181 if (type != obj->type || type <= 0)
182 die("object of unexpected type");
183 obj->flags |= FLAG_WRITTEN;
184 return 1;
185 }
186
187 if (fsck_object(obj, 1, fsck_error_function))
188 die("Error in object");
189 if (!fsck_walk(obj, check_object, 0))
190 die("Error on reachable objects of %s", sha1_to_hex(obj->sha1));
191 write_cached_object(obj);
192 return 1;
193 }
194
195 static void write_rest(void)
196 {
197 unsigned i;
198 for (i = 0; i < nr_objects; i++)
199 check_object(obj_list[i].obj, OBJ_ANY, 0);
200 }
201
202 static void added_object(unsigned nr, enum object_type type,
203 void *data, unsigned long size);
204
205 static void write_object(unsigned nr, enum object_type type,
206 void *buf, unsigned long size)
207 {
208 added_object(nr, type, buf, size);
209 if (!strict) {
210 if (write_sha1_file(buf, size, typename(type), obj_list[nr].sha1) < 0)
211 die("failed to write object");
212 free(buf);
213 obj_list[nr].obj = 0;
214 } else if (type == OBJ_BLOB) {
215 struct blob *blob;
216 if (write_sha1_file(buf, size, typename(type), obj_list[nr].sha1) < 0)
217 die("failed to write object");
218 free(buf);
219
220 blob = lookup_blob(obj_list[nr].sha1);
221 if (blob)
222 blob->object.flags |= FLAG_WRITTEN;
223 else
224 die("invalid blob object");
225 obj_list[nr].obj = 0;
226 } else {
227 struct object *obj;
228 int eaten;
229 hash_sha1_file(buf, size, typename(type), obj_list[nr].sha1);
230 obj = parse_object_buffer(obj_list[nr].sha1, type, size, buf, &eaten);
231 if (!obj)
232 die("invalid %s", typename(type));
233 /* buf is stored via add_object_buffer and in obj, if its a tree or commit */
234 add_object_buffer(obj, buf, size);
235 obj->flags |= FLAG_OPEN;
236 obj_list[nr].obj = obj;
237 }
238 }
239
240 static void resolve_delta(unsigned nr, enum object_type type,
241 void *base, unsigned long base_size,
242 void *delta, unsigned long delta_size)
243 {
244 void *result;
245 unsigned long result_size;
246
247 result = patch_delta(base, base_size,
248 delta, delta_size,
249 &result_size);
250 if (!result)
251 die("failed to apply delta");
252 free(delta);
253 write_object(nr, type, result, result_size);
254 }
255
256 static void added_object(unsigned nr, enum object_type type,
257 void *data, unsigned long size)
258 {
259 struct delta_info **p = &delta_list;
260 struct delta_info *info;
261
262 while ((info = *p) != NULL) {
263 if (!hashcmp(info->base_sha1, obj_list[nr].sha1) ||
264 info->base_offset == obj_list[nr].offset) {
265 *p = info->next;
266 p = &delta_list;
267 resolve_delta(info->nr, type, data, size,
268 info->delta, info->size);
269 free(info);
270 continue;
271 }
272 p = &info->next;
273 }
274 }
275
276 static void unpack_non_delta_entry(enum object_type type, unsigned long size,
277 unsigned nr)
278 {
279 void *buf = get_data(size);
280
281 if (!dry_run && buf)
282 write_object(nr, type, buf, size);
283 else
284 free(buf);
285 }
286
287 static void unpack_delta_entry(enum object_type type, unsigned long delta_size,
288 unsigned nr)
289 {
290 void *delta_data, *base;
291 unsigned long base_size;
292 unsigned char base_sha1[20];
293 struct object *obj;
294
295 if (type == OBJ_REF_DELTA) {
296 hashcpy(base_sha1, fill(20));
297 use(20);
298 delta_data = get_data(delta_size);
299 if (dry_run || !delta_data) {
300 free(delta_data);
301 return;
302 }
303 if (!has_sha1_file(base_sha1)) {
304 hashcpy(obj_list[nr].sha1, null_sha1);
305 add_delta_to_list(nr, base_sha1, 0, delta_data, delta_size);
306 return;
307 }
308 } else {
309 unsigned base_found = 0;
310 unsigned char *pack, c;
311 off_t base_offset;
312 unsigned lo, mid, hi;
313
314 pack = fill(1);
315 c = *pack;
316 use(1);
317 base_offset = c & 127;
318 while (c & 128) {
319 base_offset += 1;
320 if (!base_offset || MSB(base_offset, 7))
321 die("offset value overflow for delta base object");
322 pack = fill(1);
323 c = *pack;
324 use(1);
325 base_offset = (base_offset << 7) + (c & 127);
326 }
327 base_offset = obj_list[nr].offset - base_offset;
328
329 delta_data = get_data(delta_size);
330 if (dry_run || !delta_data) {
331 free(delta_data);
332 return;
333 }
334 lo = 0;
335 hi = nr;
336 while (lo < hi) {
337 mid = (lo + hi)/2;
338 if (base_offset < obj_list[mid].offset) {
339 hi = mid;
340 } else if (base_offset > obj_list[mid].offset) {
341 lo = mid + 1;
342 } else {
343 hashcpy(base_sha1, obj_list[mid].sha1);
344 base_found = !is_null_sha1(base_sha1);
345 break;
346 }
347 }
348 if (!base_found) {
349 /* The delta base object is itself a delta that
350 has not been resolved yet. */
351 hashcpy(obj_list[nr].sha1, null_sha1);
352 add_delta_to_list(nr, null_sha1, base_offset, delta_data, delta_size);
353 return;
354 }
355 }
356
357 obj = lookup_object(base_sha1);
358 if (obj) {
359 struct obj_buffer *obj_buf = lookup_object_buffer(obj);
360 if (obj_buf) {
361 resolve_delta(nr, obj->type, obj_buf->buffer, obj_buf->size, delta_data, delta_size);
362 return;
363 }
364 }
365
366 base = read_sha1_file(base_sha1, &type, &base_size);
367 if (!base) {
368 error("failed to read delta-pack base object %s",
369 sha1_to_hex(base_sha1));
370 if (!recover)
371 exit(1);
372 has_errors = 1;
373 return;
374 }
375 resolve_delta(nr, type, base, base_size, delta_data, delta_size);
376 free(base);
377 }
378
379 static void unpack_one(unsigned nr)
380 {
381 unsigned shift;
382 unsigned char *pack, c;
383 unsigned long size;
384 enum object_type type;
385
386 obj_list[nr].offset = consumed_bytes;
387
388 pack = fill(1);
389 c = *pack;
390 use(1);
391 type = (c >> 4) & 7;
392 size = (c & 15);
393 shift = 4;
394 while (c & 0x80) {
395 pack = fill(1);
396 c = *pack;
397 use(1);
398 size += (c & 0x7f) << shift;
399 shift += 7;
400 }
401
402 switch (type) {
403 case OBJ_COMMIT:
404 case OBJ_TREE:
405 case OBJ_BLOB:
406 case OBJ_TAG:
407 unpack_non_delta_entry(type, size, nr);
408 return;
409 case OBJ_REF_DELTA:
410 case OBJ_OFS_DELTA:
411 unpack_delta_entry(type, size, nr);
412 return;
413 default:
414 error("bad object type %d", type);
415 has_errors = 1;
416 if (recover)
417 return;
418 exit(1);
419 }
420 }
421
422 static void unpack_all(void)
423 {
424 int i;
425 struct progress *progress = NULL;
426 struct pack_header *hdr = fill(sizeof(struct pack_header));
427
428 nr_objects = ntohl(hdr->hdr_entries);
429
430 if (ntohl(hdr->hdr_signature) != PACK_SIGNATURE)
431 die("bad pack file");
432 if (!pack_version_ok(hdr->hdr_version))
433 die("unknown pack file version %d", ntohl(hdr->hdr_version));
434 use(sizeof(struct pack_header));
435
436 if (!quiet)
437 progress = start_progress("Unpacking objects", nr_objects);
438 obj_list = xmalloc(nr_objects * sizeof(*obj_list));
439 memset(obj_list, 0, nr_objects * sizeof(*obj_list));
440 for (i = 0; i < nr_objects; i++) {
441 unpack_one(i);
442 display_progress(progress, i + 1);
443 }
444 stop_progress(&progress);
445
446 if (delta_list)
447 die("unresolved deltas left after unpacking");
448 }
449
450 int cmd_unpack_objects(int argc, const char **argv, const char *prefix)
451 {
452 int i;
453 unsigned char sha1[20];
454
455 git_config(git_default_config);
456
457 quiet = !isatty(2);
458
459 for (i = 1 ; i < argc; i++) {
460 const char *arg = argv[i];
461
462 if (*arg == '-') {
463 if (!strcmp(arg, "-n")) {
464 dry_run = 1;
465 continue;
466 }
467 if (!strcmp(arg, "-q")) {
468 quiet = 1;
469 continue;
470 }
471 if (!strcmp(arg, "-r")) {
472 recover = 1;
473 continue;
474 }
475 if (!strcmp(arg, "--strict")) {
476 strict = 1;
477 continue;
478 }
479 if (!prefixcmp(arg, "--pack_header=")) {
480 struct pack_header *hdr;
481 char *c;
482
483 hdr = (struct pack_header *)buffer;
484 hdr->hdr_signature = htonl(PACK_SIGNATURE);
485 hdr->hdr_version = htonl(strtoul(arg + 14, &c, 10));
486 if (*c != ',')
487 die("bad %s", arg);
488 hdr->hdr_entries = htonl(strtoul(c + 1, &c, 10));
489 if (*c)
490 die("bad %s", arg);
491 len = sizeof(*hdr);
492 continue;
493 }
494 usage(unpack_usage);
495 }
496
497 /* We don't take any non-flag arguments now.. Maybe some day */
498 usage(unpack_usage);
499 }
500 SHA1_Init(&ctx);
501 unpack_all();
502 SHA1_Update(&ctx, buffer, offset);
503 SHA1_Final(sha1, &ctx);
504 if (strict)
505 write_rest();
506 if (hashcmp(fill(20), sha1))
507 die("final sha1 did not match");
508 use(20);
509
510 /* Write the last part of the buffer to stdout */
511 while (len) {
512 int ret = xwrite(1, buffer + offset, len);
513 if (ret <= 0)
514 break;
515 len -= ret;
516 offset += ret;
517 }
518
519 /* All done */
520 return has_errors;
521 }