Merge branch 'jk/parseopt-string-list'
[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]";
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 git_SHA_CTX ctx;
23 static struct fsck_options fsck_options = FSCK_OPTIONS_STRICT;
24
25 /*
26  * When running under --strict mode, objects whose reachability are
27  * suspect are kept in core without getting written in the object
28  * store.
29  */
30 struct obj_buffer {
31         char *buffer;
32         unsigned long size;
33 };
34
35 static struct decoration obj_decorate;
36
37 static struct obj_buffer *lookup_object_buffer(struct object *base)
38 {
39         return lookup_decoration(&obj_decorate, base);
40 }
41
42 static void add_object_buffer(struct object *object, char *buffer, unsigned long size)
43 {
44         struct obj_buffer *obj;
45         obj = xcalloc(1, sizeof(struct obj_buffer));
46         obj->buffer = buffer;
47         obj->size = size;
48         if (add_decoration(&obj_decorate, object, obj))
49                 die("object %s tried to add buffer twice!", oid_to_hex(&object->oid));
50 }
51
52 /*
53  * Make sure at least "min" bytes are available in the buffer, and
54  * return the pointer to the buffer.
55  */
56 static void *fill(int min)
57 {
58         if (min <= len)
59                 return buffer + offset;
60         if (min > sizeof(buffer))
61                 die("cannot fill %d bytes", min);
62         if (offset) {
63                 git_SHA1_Update(&ctx, buffer, offset);
64                 memmove(buffer, buffer + offset, len);
65                 offset = 0;
66         }
67         do {
68                 ssize_t ret = xread(0, buffer + len, sizeof(buffer) - len);
69                 if (ret <= 0) {
70                         if (!ret)
71                                 die("early EOF");
72                         die_errno("read error on input");
73                 }
74                 len += ret;
75         } while (len < min);
76         return buffer;
77 }
78
79 static void use(int bytes)
80 {
81         if (bytes > len)
82                 die("used more bytes than were available");
83         len -= bytes;
84         offset += bytes;
85
86         /* make sure off_t is sufficiently large not to wrap */
87         if (signed_add_overflows(consumed_bytes, bytes))
88                 die("pack too large for current definition of off_t");
89         consumed_bytes += bytes;
90 }
91
92 static void *get_data(unsigned long size)
93 {
94         git_zstream stream;
95         void *buf = xmallocz(size);
96
97         memset(&stream, 0, sizeof(stream));
98
99         stream.next_out = buf;
100         stream.avail_out = size;
101         stream.next_in = fill(1);
102         stream.avail_in = len;
103         git_inflate_init(&stream);
104
105         for (;;) {
106                 int ret = git_inflate(&stream, 0);
107                 use(len - stream.avail_in);
108                 if (stream.total_out == size && ret == Z_STREAM_END)
109                         break;
110                 if (ret != Z_OK) {
111                         error("inflate returned %d", ret);
112                         free(buf);
113                         buf = NULL;
114                         if (!recover)
115                                 exit(1);
116                         has_errors = 1;
117                         break;
118                 }
119                 stream.next_in = fill(1);
120                 stream.avail_in = len;
121         }
122         git_inflate_end(&stream);
123         return buf;
124 }
125
126 struct delta_info {
127         unsigned char base_sha1[20];
128         unsigned nr;
129         off_t base_offset;
130         unsigned long size;
131         void *delta;
132         struct delta_info *next;
133 };
134
135 static struct delta_info *delta_list;
136
137 static void add_delta_to_list(unsigned nr, unsigned const char *base_sha1,
138                               off_t base_offset,
139                               void *delta, unsigned long size)
140 {
141         struct delta_info *info = xmalloc(sizeof(*info));
142
143         hashcpy(info->base_sha1, base_sha1);
144         info->base_offset = base_offset;
145         info->size = size;
146         info->delta = delta;
147         info->nr = nr;
148         info->next = delta_list;
149         delta_list = info;
150 }
151
152 struct obj_info {
153         off_t offset;
154         unsigned char sha1[20];
155         struct object *obj;
156 };
157
158 #define FLAG_OPEN (1u<<20)
159 #define FLAG_WRITTEN (1u<<21)
160
161 static struct obj_info *obj_list;
162 static unsigned nr_objects;
163
164 /*
165  * Called only from check_object() after it verified this object
166  * is Ok.
167  */
168 static void write_cached_object(struct object *obj, struct obj_buffer *obj_buf)
169 {
170         unsigned char sha1[20];
171
172         if (write_sha1_file(obj_buf->buffer, obj_buf->size, typename(obj->type), sha1) < 0)
173                 die("failed to write object %s", oid_to_hex(&obj->oid));
174         obj->flags |= FLAG_WRITTEN;
175 }
176
177 /*
178  * At the very end of the processing, write_rest() scans the objects
179  * that have reachability requirements and calls this function.
180  * Verify its reachability and validity recursively and write it out.
181  */
182 static int check_object(struct object *obj, int type, void *data, struct fsck_options *options)
183 {
184         struct obj_buffer *obj_buf;
185
186         if (!obj)
187                 return 1;
188
189         if (obj->flags & FLAG_WRITTEN)
190                 return 0;
191
192         if (type != OBJ_ANY && obj->type != type)
193                 die("object type mismatch");
194
195         if (!(obj->flags & FLAG_OPEN)) {
196                 unsigned long size;
197                 int type = sha1_object_info(obj->oid.hash, &size);
198                 if (type != obj->type || type <= 0)
199                         die("object of unexpected type");
200                 obj->flags |= FLAG_WRITTEN;
201                 return 0;
202         }
203
204         obj_buf = lookup_object_buffer(obj);
205         if (!obj_buf)
206                 die("Whoops! Cannot find object '%s'", oid_to_hex(&obj->oid));
207         if (fsck_object(obj, obj_buf->buffer, obj_buf->size, &fsck_options))
208                 die("Error in object");
209         fsck_options.walk = check_object;
210         if (fsck_walk(obj, NULL, &fsck_options))
211                 die("Error on reachable objects of %s", oid_to_hex(&obj->oid));
212         write_cached_object(obj, obj_buf);
213         return 0;
214 }
215
216 static void write_rest(void)
217 {
218         unsigned i;
219         for (i = 0; i < nr_objects; i++) {
220                 if (obj_list[i].obj)
221                         check_object(obj_list[i].obj, OBJ_ANY, NULL, NULL);
222         }
223 }
224
225 static void added_object(unsigned nr, enum object_type type,
226                          void *data, unsigned long size);
227
228 /*
229  * Write out nr-th object from the list, now we know the contents
230  * of it.  Under --strict, this buffers structured objects in-core,
231  * to be checked at the end.
232  */
233 static void write_object(unsigned nr, enum object_type type,
234                          void *buf, unsigned long size)
235 {
236         if (!strict) {
237                 if (write_sha1_file(buf, size, typename(type), obj_list[nr].sha1) < 0)
238                         die("failed to write object");
239                 added_object(nr, type, buf, size);
240                 free(buf);
241                 obj_list[nr].obj = NULL;
242         } else if (type == OBJ_BLOB) {
243                 struct blob *blob;
244                 if (write_sha1_file(buf, size, typename(type), obj_list[nr].sha1) < 0)
245                         die("failed to write object");
246                 added_object(nr, type, buf, size);
247                 free(buf);
248
249                 blob = lookup_blob(obj_list[nr].sha1);
250                 if (blob)
251                         blob->object.flags |= FLAG_WRITTEN;
252                 else
253                         die("invalid blob object");
254                 obj_list[nr].obj = NULL;
255         } else {
256                 struct object *obj;
257                 int eaten;
258                 hash_sha1_file(buf, size, typename(type), obj_list[nr].sha1);
259                 added_object(nr, type, buf, size);
260                 obj = parse_object_buffer(obj_list[nr].sha1, type, size, buf, &eaten);
261                 if (!obj)
262                         die("invalid %s", typename(type));
263                 add_object_buffer(obj, buf, size);
264                 obj->flags |= FLAG_OPEN;
265                 obj_list[nr].obj = obj;
266         }
267 }
268
269 static void resolve_delta(unsigned nr, enum object_type type,
270                           void *base, unsigned long base_size,
271                           void *delta, unsigned long delta_size)
272 {
273         void *result;
274         unsigned long result_size;
275
276         result = patch_delta(base, base_size,
277                              delta, delta_size,
278                              &result_size);
279         if (!result)
280                 die("failed to apply delta");
281         free(delta);
282         write_object(nr, type, result, result_size);
283 }
284
285 /*
286  * We now know the contents of an object (which is nr-th in the pack);
287  * resolve all the deltified objects that are based on it.
288  */
289 static void added_object(unsigned nr, enum object_type type,
290                          void *data, unsigned long size)
291 {
292         struct delta_info **p = &delta_list;
293         struct delta_info *info;
294
295         while ((info = *p) != NULL) {
296                 if (!hashcmp(info->base_sha1, obj_list[nr].sha1) ||
297                     info->base_offset == obj_list[nr].offset) {
298                         *p = info->next;
299                         p = &delta_list;
300                         resolve_delta(info->nr, type, data, size,
301                                       info->delta, info->size);
302                         free(info);
303                         continue;
304                 }
305                 p = &info->next;
306         }
307 }
308
309 static void unpack_non_delta_entry(enum object_type type, unsigned long size,
310                                    unsigned nr)
311 {
312         void *buf = get_data(size);
313
314         if (!dry_run && buf)
315                 write_object(nr, type, buf, size);
316         else
317                 free(buf);
318 }
319
320 static int resolve_against_held(unsigned nr, const unsigned char *base,
321                                 void *delta_data, unsigned long delta_size)
322 {
323         struct object *obj;
324         struct obj_buffer *obj_buffer;
325         obj = lookup_object(base);
326         if (!obj)
327                 return 0;
328         obj_buffer = lookup_object_buffer(obj);
329         if (!obj_buffer)
330                 return 0;
331         resolve_delta(nr, obj->type, obj_buffer->buffer,
332                       obj_buffer->size, delta_data, delta_size);
333         return 1;
334 }
335
336 static void unpack_delta_entry(enum object_type type, unsigned long delta_size,
337                                unsigned nr)
338 {
339         void *delta_data, *base;
340         unsigned long base_size;
341         unsigned char base_sha1[20];
342
343         if (type == OBJ_REF_DELTA) {
344                 hashcpy(base_sha1, fill(20));
345                 use(20);
346                 delta_data = get_data(delta_size);
347                 if (dry_run || !delta_data) {
348                         free(delta_data);
349                         return;
350                 }
351                 if (has_sha1_file(base_sha1))
352                         ; /* Ok we have this one */
353                 else if (resolve_against_held(nr, base_sha1,
354                                               delta_data, delta_size))
355                         return; /* we are done */
356                 else {
357                         /* cannot resolve yet --- queue it */
358                         hashclr(obj_list[nr].sha1);
359                         add_delta_to_list(nr, base_sha1, 0, delta_data, delta_size);
360                         return;
361                 }
362         } else {
363                 unsigned base_found = 0;
364                 unsigned char *pack, c;
365                 off_t base_offset;
366                 unsigned lo, mid, hi;
367
368                 pack = fill(1);
369                 c = *pack;
370                 use(1);
371                 base_offset = c & 127;
372                 while (c & 128) {
373                         base_offset += 1;
374                         if (!base_offset || MSB(base_offset, 7))
375                                 die("offset value overflow for delta base object");
376                         pack = fill(1);
377                         c = *pack;
378                         use(1);
379                         base_offset = (base_offset << 7) + (c & 127);
380                 }
381                 base_offset = obj_list[nr].offset - base_offset;
382                 if (base_offset <= 0 || base_offset >= obj_list[nr].offset)
383                         die("offset value out of bound for delta base object");
384
385                 delta_data = get_data(delta_size);
386                 if (dry_run || !delta_data) {
387                         free(delta_data);
388                         return;
389                 }
390                 lo = 0;
391                 hi = nr;
392                 while (lo < hi) {
393                         mid = (lo + hi)/2;
394                         if (base_offset < obj_list[mid].offset) {
395                                 hi = mid;
396                         } else if (base_offset > obj_list[mid].offset) {
397                                 lo = mid + 1;
398                         } else {
399                                 hashcpy(base_sha1, obj_list[mid].sha1);
400                                 base_found = !is_null_sha1(base_sha1);
401                                 break;
402                         }
403                 }
404                 if (!base_found) {
405                         /*
406                          * The delta base object is itself a delta that
407                          * has not been resolved yet.
408                          */
409                         hashclr(obj_list[nr].sha1);
410                         add_delta_to_list(nr, null_sha1, base_offset, delta_data, delta_size);
411                         return;
412                 }
413         }
414
415         if (resolve_against_held(nr, base_sha1, delta_data, delta_size))
416                 return;
417
418         base = read_sha1_file(base_sha1, &type, &base_size);
419         if (!base) {
420                 error("failed to read delta-pack base object %s",
421                       sha1_to_hex(base_sha1));
422                 if (!recover)
423                         exit(1);
424                 has_errors = 1;
425                 return;
426         }
427         resolve_delta(nr, type, base, base_size, delta_data, delta_size);
428         free(base);
429 }
430
431 static void unpack_one(unsigned nr)
432 {
433         unsigned shift;
434         unsigned char *pack;
435         unsigned long size, c;
436         enum object_type type;
437
438         obj_list[nr].offset = consumed_bytes;
439
440         pack = fill(1);
441         c = *pack;
442         use(1);
443         type = (c >> 4) & 7;
444         size = (c & 15);
445         shift = 4;
446         while (c & 0x80) {
447                 pack = fill(1);
448                 c = *pack;
449                 use(1);
450                 size += (c & 0x7f) << shift;
451                 shift += 7;
452         }
453
454         switch (type) {
455         case OBJ_COMMIT:
456         case OBJ_TREE:
457         case OBJ_BLOB:
458         case OBJ_TAG:
459                 unpack_non_delta_entry(type, size, nr);
460                 return;
461         case OBJ_REF_DELTA:
462         case OBJ_OFS_DELTA:
463                 unpack_delta_entry(type, size, nr);
464                 return;
465         default:
466                 error("bad object type %d", type);
467                 has_errors = 1;
468                 if (recover)
469                         return;
470                 exit(1);
471         }
472 }
473
474 static void unpack_all(void)
475 {
476         int i;
477         struct progress *progress = NULL;
478         struct pack_header *hdr = fill(sizeof(struct pack_header));
479
480         nr_objects = ntohl(hdr->hdr_entries);
481
482         if (ntohl(hdr->hdr_signature) != PACK_SIGNATURE)
483                 die("bad pack file");
484         if (!pack_version_ok(hdr->hdr_version))
485                 die("unknown pack file version %"PRIu32,
486                         ntohl(hdr->hdr_version));
487         use(sizeof(struct pack_header));
488
489         if (!quiet)
490                 progress = start_progress(_("Unpacking objects"), nr_objects);
491         obj_list = xcalloc(nr_objects, sizeof(*obj_list));
492         for (i = 0; i < nr_objects; i++) {
493                 unpack_one(i);
494                 display_progress(progress, i + 1);
495         }
496         stop_progress(&progress);
497
498         if (delta_list)
499                 die("unresolved deltas left after unpacking");
500 }
501
502 int cmd_unpack_objects(int argc, const char **argv, const char *prefix)
503 {
504         int i;
505         unsigned char sha1[20];
506
507         check_replace_refs = 0;
508
509         git_config(git_default_config, NULL);
510
511         quiet = !isatty(2);
512
513         for (i = 1 ; i < argc; i++) {
514                 const char *arg = argv[i];
515
516                 if (*arg == '-') {
517                         if (!strcmp(arg, "-n")) {
518                                 dry_run = 1;
519                                 continue;
520                         }
521                         if (!strcmp(arg, "-q")) {
522                                 quiet = 1;
523                                 continue;
524                         }
525                         if (!strcmp(arg, "-r")) {
526                                 recover = 1;
527                                 continue;
528                         }
529                         if (!strcmp(arg, "--strict")) {
530                                 strict = 1;
531                                 continue;
532                         }
533                         if (skip_prefix(arg, "--strict=", &arg)) {
534                                 strict = 1;
535                                 fsck_set_msg_types(&fsck_options, arg);
536                                 continue;
537                         }
538                         if (starts_with(arg, "--pack_header=")) {
539                                 struct pack_header *hdr;
540                                 char *c;
541
542                                 hdr = (struct pack_header *)buffer;
543                                 hdr->hdr_signature = htonl(PACK_SIGNATURE);
544                                 hdr->hdr_version = htonl(strtoul(arg + 14, &c, 10));
545                                 if (*c != ',')
546                                         die("bad %s", arg);
547                                 hdr->hdr_entries = htonl(strtoul(c + 1, &c, 10));
548                                 if (*c)
549                                         die("bad %s", arg);
550                                 len = sizeof(*hdr);
551                                 continue;
552                         }
553                         usage(unpack_usage);
554                 }
555
556                 /* We don't take any non-flag arguments now.. Maybe some day */
557                 usage(unpack_usage);
558         }
559         git_SHA1_Init(&ctx);
560         unpack_all();
561         git_SHA1_Update(&ctx, buffer, offset);
562         git_SHA1_Final(sha1, &ctx);
563         if (strict)
564                 write_rest();
565         if (hashcmp(fill(20), sha1))
566                 die("final sha1 did not match");
567         use(20);
568
569         /* Write the last part of the buffer to stdout */
570         while (len) {
571                 int ret = xwrite(1, buffer + offset, len);
572                 if (ret <= 0)
573                         break;
574                 len -= ret;
575                 offset += ret;
576         }
577
578         /* All done */
579         return has_errors;
580 }