cat-file: document -p option
[git/dscho.git] / fsck-objects.c
blob59b25904cb0f6e87594e2bad20db3ee2aa726aef
1 #include <sys/types.h>
2 #include <dirent.h>
4 #include "cache.h"
5 #include "commit.h"
6 #include "tree.h"
7 #include "blob.h"
8 #include "tag.h"
9 #include "refs.h"
10 #include "pack.h"
12 #define REACHABLE 0x0001
14 static int show_root = 0;
15 static int show_tags = 0;
16 static int show_unreachable = 0;
17 static int check_full = 0;
18 static int check_strict = 0;
19 static int keep_cache_objects = 0;
20 static unsigned char head_sha1[20];
22 #ifdef NO_D_INO_IN_DIRENT
23 #define SORT_DIRENT 0
24 #define DIRENT_SORT_HINT(de) 0
25 #else
26 #define SORT_DIRENT 1
27 #define DIRENT_SORT_HINT(de) ((de)->d_ino)
28 #endif
30 static void objreport(struct object *obj, const char *severity,
31 const char *err, va_list params)
33 fprintf(stderr, "%s in %s %s: ",
34 severity, obj->type, sha1_to_hex(obj->sha1));
35 vfprintf(stderr, err, params);
36 fputs("\n", stderr);
39 static int objerror(struct object *obj, const char *err, ...)
41 va_list params;
42 va_start(params, err);
43 objreport(obj, "error", err, params);
44 va_end(params);
45 return -1;
48 static int objwarning(struct object *obj, const char *err, ...)
50 va_list params;
51 va_start(params, err);
52 objreport(obj, "warning", err, params);
53 va_end(params);
54 return -1;
58 static void check_connectivity(void)
60 int i;
62 /* Look up all the requirements, warn about missing objects.. */
63 for (i = 0; i < obj_allocs; i++) {
64 struct object *obj = objs[i];
66 if (!obj)
67 continue;
69 if (!obj->parsed) {
70 if (has_sha1_file(obj->sha1))
71 ; /* it is in pack */
72 else
73 printf("missing %s %s\n",
74 obj->type, sha1_to_hex(obj->sha1));
75 continue;
78 if (obj->refs) {
79 const struct object_refs *refs = obj->refs;
80 unsigned j;
81 for (j = 0; j < refs->count; j++) {
82 struct object *ref = refs->ref[j];
83 if (ref->parsed ||
84 (has_sha1_file(ref->sha1)))
85 continue;
86 printf("broken link from %7s %s\n",
87 obj->type, sha1_to_hex(obj->sha1));
88 printf(" to %7s %s\n",
89 ref->type, sha1_to_hex(ref->sha1));
93 if (show_unreachable && !(obj->flags & REACHABLE)) {
94 printf("unreachable %s %s\n",
95 obj->type, sha1_to_hex(obj->sha1));
96 continue;
99 if (!obj->used) {
100 printf("dangling %s %s\n", obj->type,
101 sha1_to_hex(obj->sha1));
107 * The entries in a tree are ordered in the _path_ order,
108 * which means that a directory entry is ordered by adding
109 * a slash to the end of it.
111 * So a directory called "a" is ordered _after_ a file
112 * called "a.c", because "a/" sorts after "a.c".
114 #define TREE_UNORDERED (-1)
115 #define TREE_HAS_DUPS (-2)
117 static int verify_ordered(struct tree_entry_list *a, struct tree_entry_list *b)
119 int len1 = strlen(a->name);
120 int len2 = strlen(b->name);
121 int len = len1 < len2 ? len1 : len2;
122 unsigned char c1, c2;
123 int cmp;
125 cmp = memcmp(a->name, b->name, len);
126 if (cmp < 0)
127 return 0;
128 if (cmp > 0)
129 return TREE_UNORDERED;
132 * Ok, the first <len> characters are the same.
133 * Now we need to order the next one, but turn
134 * a '\0' into a '/' for a directory entry.
136 c1 = a->name[len];
137 c2 = b->name[len];
138 if (!c1 && !c2)
140 * git-write-tree used to write out a nonsense tree that has
141 * entries with the same name, one blob and one tree. Make
142 * sure we do not have duplicate entries.
144 return TREE_HAS_DUPS;
145 if (!c1 && a->directory)
146 c1 = '/';
147 if (!c2 && b->directory)
148 c2 = '/';
149 return c1 < c2 ? 0 : TREE_UNORDERED;
152 static int fsck_tree(struct tree *item)
154 int retval;
155 int has_full_path = 0;
156 int has_zero_pad = 0;
157 int has_bad_modes = 0;
158 int has_dup_entries = 0;
159 int not_properly_sorted = 0;
160 struct tree_entry_list *entry, *last;
162 last = NULL;
163 for (entry = item->entries; entry; entry = entry->next) {
164 if (strchr(entry->name, '/'))
165 has_full_path = 1;
166 has_zero_pad |= entry->zeropad;
168 switch (entry->mode) {
170 * Standard modes..
172 case S_IFREG | 0755:
173 case S_IFREG | 0644:
174 case S_IFLNK:
175 case S_IFDIR:
176 break;
178 * This is nonstandard, but we had a few of these
179 * early on when we honored the full set of mode
180 * bits..
182 case S_IFREG | 0664:
183 if (!check_strict)
184 break;
185 default:
186 has_bad_modes = 1;
189 if (last) {
190 switch (verify_ordered(last, entry)) {
191 case TREE_UNORDERED:
192 not_properly_sorted = 1;
193 break;
194 case TREE_HAS_DUPS:
195 has_dup_entries = 1;
196 break;
197 default:
198 break;
200 free(last->name);
201 free(last);
204 last = entry;
206 if (last) {
207 free(last->name);
208 free(last);
210 item->entries = NULL;
212 retval = 0;
213 if (has_full_path) {
214 objwarning(&item->object, "contains full pathnames");
216 if (has_zero_pad) {
217 objwarning(&item->object, "contains zero-padded file modes");
219 if (has_bad_modes) {
220 objwarning(&item->object, "contains bad file modes");
222 if (has_dup_entries) {
223 retval = objerror(&item->object, "contains duplicate file entries");
225 if (not_properly_sorted) {
226 retval = objerror(&item->object, "not properly sorted");
228 return retval;
231 static int fsck_commit(struct commit *commit)
233 char *buffer = commit->buffer;
234 unsigned char tree_sha1[20], sha1[20];
236 if (memcmp(buffer, "tree ", 5))
237 return objerror(&commit->object, "invalid format - expected 'tree' line");
238 if (get_sha1_hex(buffer+5, tree_sha1) || buffer[45] != '\n')
239 return objerror(&commit->object, "invalid 'tree' line format - bad sha1");
240 buffer += 46;
241 while (!memcmp(buffer, "parent ", 7)) {
242 if (get_sha1_hex(buffer+7, sha1) || buffer[47] != '\n')
243 return objerror(&commit->object, "invalid 'parent' line format - bad sha1");
244 buffer += 48;
246 if (memcmp(buffer, "author ", 7))
247 return objerror(&commit->object, "invalid format - expected 'author' line");
248 free(commit->buffer);
249 commit->buffer = NULL;
250 if (!commit->tree)
251 return objerror(&commit->object, "could not load commit's tree %s", tree_sha1);
252 if (!commit->parents && show_root)
253 printf("root %s\n", sha1_to_hex(commit->object.sha1));
254 if (!commit->date)
255 printf("bad commit date in %s\n",
256 sha1_to_hex(commit->object.sha1));
257 return 0;
260 static int fsck_tag(struct tag *tag)
262 struct object *tagged = tag->tagged;
264 if (!tagged) {
265 return objerror(&tag->object, "could not load tagged object");
267 if (!show_tags)
268 return 0;
270 printf("tagged %s %s", tagged->type, sha1_to_hex(tagged->sha1));
271 printf(" (%s) in %s\n", tag->tag, sha1_to_hex(tag->object.sha1));
272 return 0;
275 static int fsck_sha1(unsigned char *sha1)
277 struct object *obj = parse_object(sha1);
278 if (!obj)
279 return error("%s: object not found", sha1_to_hex(sha1));
280 if (obj->type == blob_type)
281 return 0;
282 if (obj->type == tree_type)
283 return fsck_tree((struct tree *) obj);
284 if (obj->type == commit_type)
285 return fsck_commit((struct commit *) obj);
286 if (obj->type == tag_type)
287 return fsck_tag((struct tag *) obj);
288 /* By now, parse_object() would've returned NULL instead. */
289 return objerror(obj, "unknown type '%s' (internal fsck error)", obj->type);
293 * This is the sorting chunk size: make it reasonably
294 * big so that we can sort well..
296 #define MAX_SHA1_ENTRIES (1024)
298 struct sha1_entry {
299 unsigned long ino;
300 unsigned char sha1[20];
303 static struct {
304 unsigned long nr;
305 struct sha1_entry *entry[MAX_SHA1_ENTRIES];
306 } sha1_list;
308 static int ino_compare(const void *_a, const void *_b)
310 const struct sha1_entry *a = _a, *b = _b;
311 unsigned long ino1 = a->ino, ino2 = b->ino;
312 return ino1 < ino2 ? -1 : ino1 > ino2 ? 1 : 0;
315 static void fsck_sha1_list(void)
317 int i, nr = sha1_list.nr;
319 if (SORT_DIRENT)
320 qsort(sha1_list.entry, nr,
321 sizeof(struct sha1_entry *), ino_compare);
322 for (i = 0; i < nr; i++) {
323 struct sha1_entry *entry = sha1_list.entry[i];
324 unsigned char *sha1 = entry->sha1;
326 sha1_list.entry[i] = NULL;
327 fsck_sha1(sha1);
328 free(entry);
330 sha1_list.nr = 0;
333 static void add_sha1_list(unsigned char *sha1, unsigned long ino)
335 struct sha1_entry *entry = xmalloc(sizeof(*entry));
336 int nr;
338 entry->ino = ino;
339 memcpy(entry->sha1, sha1, 20);
340 nr = sha1_list.nr;
341 if (nr == MAX_SHA1_ENTRIES) {
342 fsck_sha1_list();
343 nr = 0;
345 sha1_list.entry[nr] = entry;
346 sha1_list.nr = ++nr;
349 static int fsck_dir(int i, char *path)
351 DIR *dir = opendir(path);
352 struct dirent *de;
354 if (!dir)
355 return 0;
357 while ((de = readdir(dir)) != NULL) {
358 char name[100];
359 unsigned char sha1[20];
360 int len = strlen(de->d_name);
362 switch (len) {
363 case 2:
364 if (de->d_name[1] != '.')
365 break;
366 case 1:
367 if (de->d_name[0] != '.')
368 break;
369 continue;
370 case 38:
371 sprintf(name, "%02x", i);
372 memcpy(name+2, de->d_name, len+1);
373 if (get_sha1_hex(name, sha1) < 0)
374 break;
375 add_sha1_list(sha1, DIRENT_SORT_HINT(de));
376 continue;
378 fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
380 closedir(dir);
381 return 0;
384 static int default_refs = 0;
386 static int fsck_handle_ref(const char *refname, const unsigned char *sha1)
388 struct object *obj;
390 obj = lookup_object(sha1);
391 if (!obj) {
392 if (has_sha1_file(sha1)) {
393 default_refs++;
394 return 0; /* it is in a pack */
396 error("%s: invalid sha1 pointer %s", refname, sha1_to_hex(sha1));
397 /* We'll continue with the rest despite the error.. */
398 return 0;
400 default_refs++;
401 obj->used = 1;
402 mark_reachable(obj, REACHABLE);
403 return 0;
406 static void get_default_heads(void)
408 for_each_ref(fsck_handle_ref);
409 if (!default_refs)
410 die("No default references");
413 static void fsck_object_dir(const char *path)
415 int i;
416 for (i = 0; i < 256; i++) {
417 static char dir[4096];
418 sprintf(dir, "%s/%02x", path, i);
419 fsck_dir(i, dir);
421 fsck_sha1_list();
424 static int fsck_head_link(void)
426 unsigned char sha1[20];
427 const char *git_HEAD = strdup(git_path("HEAD"));
428 const char *git_refs_heads_master = resolve_ref(git_HEAD, sha1, 1);
429 int pfxlen = strlen(git_HEAD) - 4; /* strip .../.git/ part */
431 if (!git_refs_heads_master)
432 return error("HEAD is not a symbolic ref");
433 if (strncmp(git_refs_heads_master + pfxlen, "refs/heads/", 11))
434 return error("HEAD points to something strange (%s)",
435 git_refs_heads_master + pfxlen);
436 if (!memcmp(null_sha1, sha1, 20))
437 return error("HEAD: not a valid git pointer");
438 return 0;
441 int main(int argc, char **argv)
443 int i, heads;
445 setup_git_directory();
447 for (i = 1; i < argc; i++) {
448 const char *arg = argv[i];
450 if (!strcmp(arg, "--unreachable")) {
451 show_unreachable = 1;
452 continue;
454 if (!strcmp(arg, "--tags")) {
455 show_tags = 1;
456 continue;
458 if (!strcmp(arg, "--root")) {
459 show_root = 1;
460 continue;
462 if (!strcmp(arg, "--cache")) {
463 keep_cache_objects = 1;
464 continue;
466 if (!strcmp(arg, "--full")) {
467 check_full = 1;
468 continue;
470 if (!strcmp(arg, "--strict")) {
471 check_strict = 1;
472 continue;
474 if (*arg == '-')
475 usage("git-fsck-objects [--tags] [--root] [[--unreachable] [--cache] [--full] [--strict] <head-sha1>*]");
478 fsck_head_link();
479 fsck_object_dir(get_object_directory());
480 if (check_full) {
481 struct alternate_object_database *alt;
482 struct packed_git *p;
483 prepare_alt_odb();
484 for (alt = alt_odb_list; alt; alt = alt->next) {
485 char namebuf[PATH_MAX];
486 int namelen = alt->name - alt->base;
487 memcpy(namebuf, alt->base, namelen);
488 namebuf[namelen - 1] = 0;
489 fsck_object_dir(namebuf);
491 prepare_packed_git();
492 for (p = packed_git; p; p = p->next)
493 /* verify gives error messages itself */
494 verify_pack(p, 0);
496 for (p = packed_git; p; p = p->next) {
497 int num = num_packed_objects(p);
498 for (i = 0; i < num; i++) {
499 unsigned char sha1[20];
500 nth_packed_object_sha1(p, i, sha1);
501 fsck_sha1(sha1);
506 heads = 0;
507 for (i = 1; i < argc; i++) {
508 const char *arg = argv[i];
510 if (*arg == '-')
511 continue;
513 if (!get_sha1(arg, head_sha1)) {
514 struct object *obj = lookup_object(head_sha1);
516 /* Error is printed by lookup_object(). */
517 if (!obj)
518 continue;
520 obj->used = 1;
521 mark_reachable(obj, REACHABLE);
522 heads++;
523 continue;
525 error("invalid parameter: expected sha1, got '%s'", arg);
529 * If we've not been given any explicit head information, do the
530 * default ones from .git/refs. We also consider the index file
531 * in this case (ie this implies --cache).
533 if (!heads) {
534 get_default_heads();
535 keep_cache_objects = 1;
538 if (keep_cache_objects) {
539 int i;
540 read_cache();
541 for (i = 0; i < active_nr; i++) {
542 struct blob *blob = lookup_blob(active_cache[i]->sha1);
543 struct object *obj;
544 if (!blob)
545 continue;
546 obj = &blob->object;
547 obj->used = 1;
548 mark_reachable(obj, REACHABLE);
552 check_connectivity();
553 return 0;