Merge branch 'md/list-objects-filter-memfix'
[git] / object.c
1 #include "cache.h"
2 #include "object.h"
3 #include "replace-object.h"
4 #include "object-store.h"
5 #include "blob.h"
6 #include "tree.h"
7 #include "commit.h"
8 #include "tag.h"
9 #include "alloc.h"
10 #include "object-store.h"
11 #include "packfile.h"
12 #include "commit-graph.h"
13
14 unsigned int get_max_object_index(void)
15 {
16         return the_repository->parsed_objects->obj_hash_size;
17 }
18
19 struct object *get_indexed_object(unsigned int idx)
20 {
21         return the_repository->parsed_objects->obj_hash[idx];
22 }
23
24 static const char *object_type_strings[] = {
25         NULL,           /* OBJ_NONE = 0 */
26         "commit",       /* OBJ_COMMIT = 1 */
27         "tree",         /* OBJ_TREE = 2 */
28         "blob",         /* OBJ_BLOB = 3 */
29         "tag",          /* OBJ_TAG = 4 */
30 };
31
32 const char *type_name(unsigned int type)
33 {
34         if (type >= ARRAY_SIZE(object_type_strings))
35                 return NULL;
36         return object_type_strings[type];
37 }
38
39 int type_from_string_gently(const char *str, ssize_t len, int gentle)
40 {
41         int i;
42
43         if (len < 0)
44                 len = strlen(str);
45
46         for (i = 1; i < ARRAY_SIZE(object_type_strings); i++)
47                 if (!strncmp(str, object_type_strings[i], len) &&
48                     object_type_strings[i][len] == '\0')
49                         return i;
50
51         if (gentle)
52                 return -1;
53
54         die(_("invalid object type \"%s\""), str);
55 }
56
57 /*
58  * Return a numerical hash value between 0 and n-1 for the object with
59  * the specified sha1.  n must be a power of 2.  Please note that the
60  * return value is *not* consistent across computer architectures.
61  */
62 static unsigned int hash_obj(const unsigned char *sha1, unsigned int n)
63 {
64         return sha1hash(sha1) & (n - 1);
65 }
66
67 /*
68  * Insert obj into the hash table hash, which has length size (which
69  * must be a power of 2).  On collisions, simply overflow to the next
70  * empty bucket.
71  */
72 static void insert_obj_hash(struct object *obj, struct object **hash, unsigned int size)
73 {
74         unsigned int j = hash_obj(obj->oid.hash, size);
75
76         while (hash[j]) {
77                 j++;
78                 if (j >= size)
79                         j = 0;
80         }
81         hash[j] = obj;
82 }
83
84 /*
85  * Look up the record for the given sha1 in the hash map stored in
86  * obj_hash.  Return NULL if it was not found.
87  */
88 struct object *lookup_object(struct repository *r, const unsigned char *sha1)
89 {
90         unsigned int i, first;
91         struct object *obj;
92
93         if (!r->parsed_objects->obj_hash)
94                 return NULL;
95
96         first = i = hash_obj(sha1, r->parsed_objects->obj_hash_size);
97         while ((obj = r->parsed_objects->obj_hash[i]) != NULL) {
98                 if (hasheq(sha1, obj->oid.hash))
99                         break;
100                 i++;
101                 if (i == r->parsed_objects->obj_hash_size)
102                         i = 0;
103         }
104         if (obj && i != first) {
105                 /*
106                  * Move object to where we started to look for it so
107                  * that we do not need to walk the hash table the next
108                  * time we look for it.
109                  */
110                 SWAP(r->parsed_objects->obj_hash[i],
111                      r->parsed_objects->obj_hash[first]);
112         }
113         return obj;
114 }
115
116 /*
117  * Increase the size of the hash map stored in obj_hash to the next
118  * power of 2 (but at least 32).  Copy the existing values to the new
119  * hash map.
120  */
121 static void grow_object_hash(struct repository *r)
122 {
123         int i;
124         /*
125          * Note that this size must always be power-of-2 to match hash_obj
126          * above.
127          */
128         int new_hash_size = r->parsed_objects->obj_hash_size < 32 ? 32 : 2 * r->parsed_objects->obj_hash_size;
129         struct object **new_hash;
130
131         new_hash = xcalloc(new_hash_size, sizeof(struct object *));
132         for (i = 0; i < r->parsed_objects->obj_hash_size; i++) {
133                 struct object *obj = r->parsed_objects->obj_hash[i];
134
135                 if (!obj)
136                         continue;
137                 insert_obj_hash(obj, new_hash, new_hash_size);
138         }
139         free(r->parsed_objects->obj_hash);
140         r->parsed_objects->obj_hash = new_hash;
141         r->parsed_objects->obj_hash_size = new_hash_size;
142 }
143
144 void *create_object(struct repository *r, const unsigned char *sha1, void *o)
145 {
146         struct object *obj = o;
147
148         obj->parsed = 0;
149         obj->flags = 0;
150         hashcpy(obj->oid.hash, sha1);
151
152         if (r->parsed_objects->obj_hash_size - 1 <= r->parsed_objects->nr_objs * 2)
153                 grow_object_hash(r);
154
155         insert_obj_hash(obj, r->parsed_objects->obj_hash,
156                         r->parsed_objects->obj_hash_size);
157         r->parsed_objects->nr_objs++;
158         return obj;
159 }
160
161 void *object_as_type(struct repository *r, struct object *obj, enum object_type type, int quiet)
162 {
163         if (obj->type == type)
164                 return obj;
165         else if (obj->type == OBJ_NONE) {
166                 if (type == OBJ_COMMIT)
167                         init_commit_node(r, (struct commit *) obj);
168                 else
169                         obj->type = type;
170                 return obj;
171         }
172         else {
173                 if (!quiet)
174                         error(_("object %s is a %s, not a %s"),
175                               oid_to_hex(&obj->oid),
176                               type_name(obj->type), type_name(type));
177                 return NULL;
178         }
179 }
180
181 struct object *lookup_unknown_object(const unsigned char *sha1)
182 {
183         struct object *obj = lookup_object(the_repository, sha1);
184         if (!obj)
185                 obj = create_object(the_repository, sha1,
186                                     alloc_object_node(the_repository));
187         return obj;
188 }
189
190 struct object *parse_object_buffer(struct repository *r, const struct object_id *oid, enum object_type type, unsigned long size, void *buffer, int *eaten_p)
191 {
192         struct object *obj;
193         *eaten_p = 0;
194
195         obj = NULL;
196         if (type == OBJ_BLOB) {
197                 struct blob *blob = lookup_blob(r, oid);
198                 if (blob) {
199                         if (parse_blob_buffer(blob, buffer, size))
200                                 return NULL;
201                         obj = &blob->object;
202                 }
203         } else if (type == OBJ_TREE) {
204                 struct tree *tree = lookup_tree(r, oid);
205                 if (tree) {
206                         obj = &tree->object;
207                         if (!tree->buffer)
208                                 tree->object.parsed = 0;
209                         if (!tree->object.parsed) {
210                                 if (parse_tree_buffer(tree, buffer, size))
211                                         return NULL;
212                                 *eaten_p = 1;
213                         }
214                 }
215         } else if (type == OBJ_COMMIT) {
216                 struct commit *commit = lookup_commit(r, oid);
217                 if (commit) {
218                         if (parse_commit_buffer(r, commit, buffer, size, 1))
219                                 return NULL;
220                         if (!get_cached_commit_buffer(r, commit, NULL)) {
221                                 set_commit_buffer(r, commit, buffer, size);
222                                 *eaten_p = 1;
223                         }
224                         obj = &commit->object;
225                 }
226         } else if (type == OBJ_TAG) {
227                 struct tag *tag = lookup_tag(r, oid);
228                 if (tag) {
229                         if (parse_tag_buffer(r, tag, buffer, size))
230                                return NULL;
231                         obj = &tag->object;
232                 }
233         } else {
234                 warning(_("object %s has unknown type id %d"), oid_to_hex(oid), type);
235                 obj = NULL;
236         }
237         return obj;
238 }
239
240 struct object *parse_object_or_die(const struct object_id *oid,
241                                    const char *name)
242 {
243         struct object *o = parse_object(the_repository, oid);
244         if (o)
245                 return o;
246
247         die(_("unable to parse object: %s"), name ? name : oid_to_hex(oid));
248 }
249
250 struct object *parse_object(struct repository *r, const struct object_id *oid)
251 {
252         unsigned long size;
253         enum object_type type;
254         int eaten;
255         const struct object_id *repl = lookup_replace_object(r, oid);
256         void *buffer;
257         struct object *obj;
258
259         obj = lookup_object(r, oid->hash);
260         if (obj && obj->parsed)
261                 return obj;
262
263         if ((obj && obj->type == OBJ_BLOB && repo_has_object_file(r, oid)) ||
264             (!obj && repo_has_object_file(r, oid) &&
265              oid_object_info(r, oid, NULL) == OBJ_BLOB)) {
266                 if (check_object_signature(repl, NULL, 0, NULL) < 0) {
267                         error(_("hash mismatch %s"), oid_to_hex(oid));
268                         return NULL;
269                 }
270                 parse_blob_buffer(lookup_blob(r, oid), NULL, 0);
271                 return lookup_object(r, oid->hash);
272         }
273
274         buffer = repo_read_object_file(r, oid, &type, &size);
275         if (buffer) {
276                 if (check_object_signature(repl, buffer, size, type_name(type)) < 0) {
277                         free(buffer);
278                         error(_("hash mismatch %s"), oid_to_hex(repl));
279                         return NULL;
280                 }
281
282                 obj = parse_object_buffer(r, oid, type, size,
283                                           buffer, &eaten);
284                 if (!eaten)
285                         free(buffer);
286                 return obj;
287         }
288         return NULL;
289 }
290
291 struct object_list *object_list_insert(struct object *item,
292                                        struct object_list **list_p)
293 {
294         struct object_list *new_list = xmalloc(sizeof(struct object_list));
295         new_list->item = item;
296         new_list->next = *list_p;
297         *list_p = new_list;
298         return new_list;
299 }
300
301 int object_list_contains(struct object_list *list, struct object *obj)
302 {
303         while (list) {
304                 if (list->item == obj)
305                         return 1;
306                 list = list->next;
307         }
308         return 0;
309 }
310
311 /*
312  * A zero-length string to which object_array_entry::name can be
313  * initialized without requiring a malloc/free.
314  */
315 static char object_array_slopbuf[1];
316
317 void add_object_array_with_path(struct object *obj, const char *name,
318                                 struct object_array *array,
319                                 unsigned mode, const char *path)
320 {
321         unsigned nr = array->nr;
322         unsigned alloc = array->alloc;
323         struct object_array_entry *objects = array->objects;
324         struct object_array_entry *entry;
325
326         if (nr >= alloc) {
327                 alloc = (alloc + 32) * 2;
328                 REALLOC_ARRAY(objects, alloc);
329                 array->alloc = alloc;
330                 array->objects = objects;
331         }
332         entry = &objects[nr];
333         entry->item = obj;
334         if (!name)
335                 entry->name = NULL;
336         else if (!*name)
337                 /* Use our own empty string instead of allocating one: */
338                 entry->name = object_array_slopbuf;
339         else
340                 entry->name = xstrdup(name);
341         entry->mode = mode;
342         if (path)
343                 entry->path = xstrdup(path);
344         else
345                 entry->path = NULL;
346         array->nr = ++nr;
347 }
348
349 void add_object_array(struct object *obj, const char *name, struct object_array *array)
350 {
351         add_object_array_with_path(obj, name, array, S_IFINVALID, NULL);
352 }
353
354 /*
355  * Free all memory associated with an entry; the result is
356  * in an unspecified state and should not be examined.
357  */
358 static void object_array_release_entry(struct object_array_entry *ent)
359 {
360         if (ent->name != object_array_slopbuf)
361                 free(ent->name);
362         free(ent->path);
363 }
364
365 struct object *object_array_pop(struct object_array *array)
366 {
367         struct object *ret;
368
369         if (!array->nr)
370                 return NULL;
371
372         ret = array->objects[array->nr - 1].item;
373         object_array_release_entry(&array->objects[array->nr - 1]);
374         array->nr--;
375         return ret;
376 }
377
378 void object_array_filter(struct object_array *array,
379                          object_array_each_func_t want, void *cb_data)
380 {
381         unsigned nr = array->nr, src, dst;
382         struct object_array_entry *objects = array->objects;
383
384         for (src = dst = 0; src < nr; src++) {
385                 if (want(&objects[src], cb_data)) {
386                         if (src != dst)
387                                 objects[dst] = objects[src];
388                         dst++;
389                 } else {
390                         object_array_release_entry(&objects[src]);
391                 }
392         }
393         array->nr = dst;
394 }
395
396 void object_array_clear(struct object_array *array)
397 {
398         int i;
399         for (i = 0; i < array->nr; i++)
400                 object_array_release_entry(&array->objects[i]);
401         FREE_AND_NULL(array->objects);
402         array->nr = array->alloc = 0;
403 }
404
405 /*
406  * Return true iff array already contains an entry with name.
407  */
408 static int contains_name(struct object_array *array, const char *name)
409 {
410         unsigned nr = array->nr, i;
411         struct object_array_entry *object = array->objects;
412
413         for (i = 0; i < nr; i++, object++)
414                 if (!strcmp(object->name, name))
415                         return 1;
416         return 0;
417 }
418
419 void object_array_remove_duplicates(struct object_array *array)
420 {
421         unsigned nr = array->nr, src;
422         struct object_array_entry *objects = array->objects;
423
424         array->nr = 0;
425         for (src = 0; src < nr; src++) {
426                 if (!contains_name(array, objects[src].name)) {
427                         if (src != array->nr)
428                                 objects[array->nr] = objects[src];
429                         array->nr++;
430                 } else {
431                         object_array_release_entry(&objects[src]);
432                 }
433         }
434 }
435
436 void clear_object_flags(unsigned flags)
437 {
438         int i;
439
440         for (i=0; i < the_repository->parsed_objects->obj_hash_size; i++) {
441                 struct object *obj = the_repository->parsed_objects->obj_hash[i];
442                 if (obj)
443                         obj->flags &= ~flags;
444         }
445 }
446
447 void clear_commit_marks_all(unsigned int flags)
448 {
449         int i;
450
451         for (i = 0; i < the_repository->parsed_objects->obj_hash_size; i++) {
452                 struct object *obj = the_repository->parsed_objects->obj_hash[i];
453                 if (obj && obj->type == OBJ_COMMIT)
454                         obj->flags &= ~flags;
455         }
456 }
457
458 struct parsed_object_pool *parsed_object_pool_new(void)
459 {
460         struct parsed_object_pool *o = xmalloc(sizeof(*o));
461         memset(o, 0, sizeof(*o));
462
463         o->blob_state = allocate_alloc_state();
464         o->tree_state = allocate_alloc_state();
465         o->commit_state = allocate_alloc_state();
466         o->tag_state = allocate_alloc_state();
467         o->object_state = allocate_alloc_state();
468
469         o->is_shallow = -1;
470         o->shallow_stat = xcalloc(1, sizeof(*o->shallow_stat));
471
472         o->buffer_slab = allocate_commit_buffer_slab();
473
474         return o;
475 }
476
477 struct raw_object_store *raw_object_store_new(void)
478 {
479         struct raw_object_store *o = xmalloc(sizeof(*o));
480
481         memset(o, 0, sizeof(*o));
482         INIT_LIST_HEAD(&o->packed_git_mru);
483         return o;
484 }
485
486 static void free_object_directory(struct object_directory *odb)
487 {
488         free(odb->path);
489         odb_clear_loose_cache(odb);
490         free(odb);
491 }
492
493 static void free_object_directories(struct raw_object_store *o)
494 {
495         while (o->odb) {
496                 struct object_directory *next;
497
498                 next = o->odb->next;
499                 free_object_directory(o->odb);
500                 o->odb = next;
501         }
502 }
503
504 void raw_object_store_clear(struct raw_object_store *o)
505 {
506         FREE_AND_NULL(o->alternate_db);
507
508         oidmap_free(o->replace_map, 1);
509         FREE_AND_NULL(o->replace_map);
510
511         free_commit_graph(o->commit_graph);
512         o->commit_graph = NULL;
513         o->commit_graph_attempted = 0;
514
515         free_object_directories(o);
516         o->odb_tail = NULL;
517         o->loaded_alternates = 0;
518
519         INIT_LIST_HEAD(&o->packed_git_mru);
520         close_all_packs(o);
521         o->packed_git = NULL;
522 }
523
524 void parsed_object_pool_clear(struct parsed_object_pool *o)
525 {
526         /*
527          * As objects are allocated in slabs (see alloc.c), we do
528          * not need to free each object, but each slab instead.
529          *
530          * Before doing so, we need to free any additional memory
531          * the objects may hold.
532          */
533         unsigned i;
534
535         for (i = 0; i < o->obj_hash_size; i++) {
536                 struct object *obj = o->obj_hash[i];
537
538                 if (!obj)
539                         continue;
540
541                 if (obj->type == OBJ_TREE)
542                         free_tree_buffer((struct tree*)obj);
543                 else if (obj->type == OBJ_COMMIT)
544                         release_commit_memory(o, (struct commit*)obj);
545                 else if (obj->type == OBJ_TAG)
546                         release_tag_memory((struct tag*)obj);
547         }
548
549         FREE_AND_NULL(o->obj_hash);
550         o->obj_hash_size = 0;
551
552         free_commit_buffer_slab(o->buffer_slab);
553         o->buffer_slab = NULL;
554
555         clear_alloc_state(o->blob_state);
556         clear_alloc_state(o->tree_state);
557         clear_alloc_state(o->commit_state);
558         clear_alloc_state(o->tag_state);
559         clear_alloc_state(o->object_state);
560         stat_validity_clear(o->shallow_stat);
561         FREE_AND_NULL(o->blob_state);
562         FREE_AND_NULL(o->tree_state);
563         FREE_AND_NULL(o->commit_state);
564         FREE_AND_NULL(o->tag_state);
565         FREE_AND_NULL(o->object_state);
566         FREE_AND_NULL(o->shallow_stat);
567 }