prefer "!=" when checking read_in_full() result
[git.git] / builtin / fsck.c
blob1e4c471b4141e6ec350c4caf58bacafb6ac07b7c
1 #include "builtin.h"
2 #include "cache.h"
3 #include "config.h"
4 #include "commit.h"
5 #include "tree.h"
6 #include "blob.h"
7 #include "tag.h"
8 #include "refs.h"
9 #include "pack.h"
10 #include "cache-tree.h"
11 #include "tree-walk.h"
12 #include "fsck.h"
13 #include "parse-options.h"
14 #include "dir.h"
15 #include "progress.h"
16 #include "streaming.h"
17 #include "decorate.h"
18 #include "packfile.h"
20 #define REACHABLE 0x0001
21 #define SEEN 0x0002
22 #define HAS_OBJ 0x0004
23 /* This flag is set if something points to this object. */
24 #define USED 0x0008
26 static int show_root;
27 static int show_tags;
28 static int show_unreachable;
29 static int include_reflogs = 1;
30 static int check_full = 1;
31 static int connectivity_only;
32 static int check_strict;
33 static int keep_cache_objects;
34 static struct fsck_options fsck_walk_options = FSCK_OPTIONS_DEFAULT;
35 static struct fsck_options fsck_obj_options = FSCK_OPTIONS_DEFAULT;
36 static struct object_id head_oid;
37 static const char *head_points_at;
38 static int errors_found;
39 static int write_lost_and_found;
40 static int verbose;
41 static int show_progress = -1;
42 static int show_dangling = 1;
43 static int name_objects;
44 #define ERROR_OBJECT 01
45 #define ERROR_REACHABLE 02
46 #define ERROR_PACK 04
47 #define ERROR_REFS 010
49 static const char *describe_object(struct object *obj)
51 static struct strbuf buf = STRBUF_INIT;
52 char *name = name_objects ?
53 lookup_decoration(fsck_walk_options.object_names, obj) : NULL;
55 strbuf_reset(&buf);
56 strbuf_addstr(&buf, oid_to_hex(&obj->oid));
57 if (name)
58 strbuf_addf(&buf, " (%s)", name);
60 return buf.buf;
63 static const char *printable_type(struct object *obj)
65 const char *ret;
67 if (obj->type == OBJ_NONE) {
68 enum object_type type = sha1_object_info(obj->oid.hash, NULL);
69 if (type > 0)
70 object_as_type(obj, type, 0);
73 ret = typename(obj->type);
74 if (!ret)
75 ret = "unknown";
77 return ret;
80 static int fsck_config(const char *var, const char *value, void *cb)
82 if (strcmp(var, "fsck.skiplist") == 0) {
83 const char *path;
84 struct strbuf sb = STRBUF_INIT;
86 if (git_config_pathname(&path, var, value))
87 return 1;
88 strbuf_addf(&sb, "skiplist=%s", path);
89 free((char *)path);
90 fsck_set_msg_types(&fsck_obj_options, sb.buf);
91 strbuf_release(&sb);
92 return 0;
95 if (skip_prefix(var, "fsck.", &var)) {
96 fsck_set_msg_type(&fsck_obj_options, var, value);
97 return 0;
100 return git_default_config(var, value, cb);
103 static void objreport(struct object *obj, const char *msg_type,
104 const char *err)
106 fprintf(stderr, "%s in %s %s: %s\n",
107 msg_type, printable_type(obj), describe_object(obj), err);
110 static int objerror(struct object *obj, const char *err)
112 errors_found |= ERROR_OBJECT;
113 objreport(obj, "error", err);
114 return -1;
117 static int fsck_error_func(struct fsck_options *o,
118 struct object *obj, int type, const char *message)
120 objreport(obj, (type == FSCK_WARN) ? "warning" : "error", message);
121 return (type == FSCK_WARN) ? 0 : 1;
124 static struct object_array pending;
126 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
128 struct object *parent = data;
131 * The only case data is NULL or type is OBJ_ANY is when
132 * mark_object_reachable() calls us. All the callers of
133 * that function has non-NULL obj hence ...
135 if (!obj) {
136 /* ... these references to parent->fld are safe here */
137 printf("broken link from %7s %s\n",
138 printable_type(parent), describe_object(parent));
139 printf("broken link from %7s %s\n",
140 (type == OBJ_ANY ? "unknown" : typename(type)), "unknown");
141 errors_found |= ERROR_REACHABLE;
142 return 1;
145 if (type != OBJ_ANY && obj->type != type)
146 /* ... and the reference to parent is safe here */
147 objerror(parent, "wrong object type in link");
149 if (obj->flags & REACHABLE)
150 return 0;
151 obj->flags |= REACHABLE;
152 if (!(obj->flags & HAS_OBJ)) {
153 if (parent && !has_object_file(&obj->oid)) {
154 printf("broken link from %7s %s\n",
155 printable_type(parent), describe_object(parent));
156 printf(" to %7s %s\n",
157 printable_type(obj), describe_object(obj));
158 errors_found |= ERROR_REACHABLE;
160 return 1;
163 add_object_array(obj, NULL, &pending);
164 return 0;
167 static void mark_object_reachable(struct object *obj)
169 mark_object(obj, OBJ_ANY, NULL, NULL);
172 static int traverse_one_object(struct object *obj)
174 return fsck_walk(obj, obj, &fsck_walk_options);
177 static int traverse_reachable(void)
179 struct progress *progress = NULL;
180 unsigned int nr = 0;
181 int result = 0;
182 if (show_progress)
183 progress = start_delayed_progress(_("Checking connectivity"), 0);
184 while (pending.nr) {
185 struct object_array_entry *entry;
186 struct object *obj;
188 entry = pending.objects + --pending.nr;
189 obj = entry->item;
190 result |= traverse_one_object(obj);
191 display_progress(progress, ++nr);
193 stop_progress(&progress);
194 return !!result;
197 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
199 if (!obj)
200 return 1;
201 obj->flags |= USED;
202 return 0;
206 * Check a single reachable object
208 static void check_reachable_object(struct object *obj)
211 * We obviously want the object to be parsed,
212 * except if it was in a pack-file and we didn't
213 * do a full fsck
215 if (!(obj->flags & HAS_OBJ)) {
216 if (has_sha1_pack(obj->oid.hash))
217 return; /* it is in pack - forget about it */
218 printf("missing %s %s\n", printable_type(obj),
219 describe_object(obj));
220 errors_found |= ERROR_REACHABLE;
221 return;
226 * Check a single unreachable object
228 static void check_unreachable_object(struct object *obj)
231 * Missing unreachable object? Ignore it. It's not like
232 * we miss it (since it can't be reached), nor do we want
233 * to complain about it being unreachable (since it does
234 * not exist).
236 if (!(obj->flags & HAS_OBJ))
237 return;
240 * Unreachable object that exists? Show it if asked to,
241 * since this is something that is prunable.
243 if (show_unreachable) {
244 printf("unreachable %s %s\n", printable_type(obj),
245 describe_object(obj));
246 return;
250 * "!USED" means that nothing at all points to it, including
251 * other unreachable objects. In other words, it's the "tip"
252 * of some set of unreachable objects, usually a commit that
253 * got dropped.
255 * Such starting points are more interesting than some random
256 * set of unreachable objects, so we show them even if the user
257 * hasn't asked for _all_ unreachable objects. If you have
258 * deleted a branch by mistake, this is a prime candidate to
259 * start looking at, for example.
261 if (!(obj->flags & USED)) {
262 if (show_dangling)
263 printf("dangling %s %s\n", printable_type(obj),
264 describe_object(obj));
265 if (write_lost_and_found) {
266 char *filename = git_pathdup("lost-found/%s/%s",
267 obj->type == OBJ_COMMIT ? "commit" : "other",
268 describe_object(obj));
269 FILE *f;
271 if (safe_create_leading_directories_const(filename)) {
272 error("Could not create lost-found");
273 free(filename);
274 return;
276 f = xfopen(filename, "w");
277 if (obj->type == OBJ_BLOB) {
278 if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
279 die_errno("Could not write '%s'", filename);
280 } else
281 fprintf(f, "%s\n", describe_object(obj));
282 if (fclose(f))
283 die_errno("Could not finish '%s'",
284 filename);
285 free(filename);
287 return;
291 * Otherwise? It's there, it's unreachable, and some other unreachable
292 * object points to it. Ignore it - it's not interesting, and we showed
293 * all the interesting cases above.
297 static void check_object(struct object *obj)
299 if (verbose)
300 fprintf(stderr, "Checking %s\n", describe_object(obj));
302 if (obj->flags & REACHABLE)
303 check_reachable_object(obj);
304 else
305 check_unreachable_object(obj);
308 static void check_connectivity(void)
310 int i, max;
312 /* Traverse the pending reachable objects */
313 traverse_reachable();
315 /* Look up all the requirements, warn about missing objects.. */
316 max = get_max_object_index();
317 if (verbose)
318 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
320 for (i = 0; i < max; i++) {
321 struct object *obj = get_indexed_object(i);
323 if (obj)
324 check_object(obj);
328 static int fsck_obj(struct object *obj)
330 int err;
332 if (obj->flags & SEEN)
333 return 0;
334 obj->flags |= SEEN;
336 if (verbose)
337 fprintf(stderr, "Checking %s %s\n",
338 printable_type(obj), describe_object(obj));
340 if (fsck_walk(obj, NULL, &fsck_obj_options))
341 objerror(obj, "broken links");
342 err = fsck_object(obj, NULL, 0, &fsck_obj_options);
343 if (err)
344 goto out;
346 if (obj->type == OBJ_COMMIT) {
347 struct commit *commit = (struct commit *) obj;
349 if (!commit->parents && show_root)
350 printf("root %s\n", describe_object(&commit->object));
353 if (obj->type == OBJ_TAG) {
354 struct tag *tag = (struct tag *) obj;
356 if (show_tags && tag->tagged) {
357 printf("tagged %s %s", printable_type(tag->tagged),
358 describe_object(tag->tagged));
359 printf(" (%s) in %s\n", tag->tag,
360 describe_object(&tag->object));
364 out:
365 if (obj->type == OBJ_TREE)
366 free_tree_buffer((struct tree *)obj);
367 if (obj->type == OBJ_COMMIT)
368 free_commit_buffer((struct commit *)obj);
369 return err;
372 static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
373 unsigned long size, void *buffer, int *eaten)
376 * Note, buffer may be NULL if type is OBJ_BLOB. See
377 * verify_packfile(), data_valid variable for details.
379 struct object *obj;
380 obj = parse_object_buffer(oid, type, size, buffer, eaten);
381 if (!obj) {
382 errors_found |= ERROR_OBJECT;
383 return error("%s: object corrupt or missing", oid_to_hex(oid));
385 obj->flags &= ~(REACHABLE | SEEN);
386 obj->flags |= HAS_OBJ;
387 return fsck_obj(obj);
390 static int default_refs;
392 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
393 timestamp_t timestamp)
395 struct object *obj;
397 if (!is_null_oid(oid)) {
398 obj = lookup_object(oid->hash);
399 if (obj && (obj->flags & HAS_OBJ)) {
400 if (timestamp && name_objects)
401 add_decoration(fsck_walk_options.object_names,
402 obj,
403 xstrfmt("%s@{%"PRItime"}", refname, timestamp));
404 obj->flags |= USED;
405 mark_object_reachable(obj);
406 } else {
407 error("%s: invalid reflog entry %s", refname, oid_to_hex(oid));
408 errors_found |= ERROR_REACHABLE;
413 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
414 const char *email, timestamp_t timestamp, int tz,
415 const char *message, void *cb_data)
417 const char *refname = cb_data;
419 if (verbose)
420 fprintf(stderr, "Checking reflog %s->%s\n",
421 oid_to_hex(ooid), oid_to_hex(noid));
423 fsck_handle_reflog_oid(refname, ooid, 0);
424 fsck_handle_reflog_oid(refname, noid, timestamp);
425 return 0;
428 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
429 int flag, void *cb_data)
431 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
432 return 0;
435 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
436 int flag, void *cb_data)
438 struct object *obj;
440 obj = parse_object(oid);
441 if (!obj) {
442 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
443 errors_found |= ERROR_REACHABLE;
444 /* We'll continue with the rest despite the error.. */
445 return 0;
447 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
448 error("%s: not a commit", refname);
449 errors_found |= ERROR_REFS;
451 default_refs++;
452 obj->flags |= USED;
453 if (name_objects)
454 add_decoration(fsck_walk_options.object_names,
455 obj, xstrdup(refname));
456 mark_object_reachable(obj);
458 return 0;
461 static void get_default_heads(void)
463 if (head_points_at && !is_null_oid(&head_oid))
464 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
465 for_each_rawref(fsck_handle_ref, NULL);
466 if (include_reflogs)
467 for_each_reflog(fsck_handle_reflog, NULL);
470 * Not having any default heads isn't really fatal, but
471 * it does mean that "--unreachable" no longer makes any
472 * sense (since in this case everything will obviously
473 * be unreachable by definition.
475 * Showing dangling objects is valid, though (as those
476 * dangling objects are likely lost heads).
478 * So we just print a warning about it, and clear the
479 * "show_unreachable" flag.
481 if (!default_refs) {
482 fprintf(stderr, "notice: No default references\n");
483 show_unreachable = 0;
487 static struct object *parse_loose_object(const struct object_id *oid,
488 const char *path)
490 struct object *obj;
491 void *contents;
492 enum object_type type;
493 unsigned long size;
494 int eaten;
496 if (read_loose_object(path, oid->hash, &type, &size, &contents) < 0)
497 return NULL;
499 if (!contents && type != OBJ_BLOB)
500 die("BUG: read_loose_object streamed a non-blob");
502 obj = parse_object_buffer(oid, type, size, contents, &eaten);
504 if (!eaten)
505 free(contents);
506 return obj;
509 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
511 struct object *obj = parse_loose_object(oid, path);
513 if (!obj) {
514 errors_found |= ERROR_OBJECT;
515 error("%s: object corrupt or missing: %s",
516 oid_to_hex(oid), path);
517 return 0; /* keep checking other objects */
520 obj->flags &= ~(REACHABLE | SEEN);
521 obj->flags |= HAS_OBJ;
522 if (fsck_obj(obj))
523 errors_found |= ERROR_OBJECT;
524 return 0;
527 static int fsck_cruft(const char *basename, const char *path, void *data)
529 if (!starts_with(basename, "tmp_obj_"))
530 fprintf(stderr, "bad sha1 file: %s\n", path);
531 return 0;
534 static int fsck_subdir(unsigned int nr, const char *path, void *progress)
536 display_progress(progress, nr + 1);
537 return 0;
540 static void fsck_object_dir(const char *path)
542 struct progress *progress = NULL;
544 if (verbose)
545 fprintf(stderr, "Checking object directory\n");
547 if (show_progress)
548 progress = start_progress(_("Checking object directories"), 256);
550 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
551 progress);
552 display_progress(progress, 256);
553 stop_progress(&progress);
556 static int fsck_head_link(void)
558 int null_is_error = 0;
560 if (verbose)
561 fprintf(stderr, "Checking HEAD link\n");
563 head_points_at = resolve_ref_unsafe("HEAD", 0, head_oid.hash, NULL);
564 if (!head_points_at) {
565 errors_found |= ERROR_REFS;
566 return error("Invalid HEAD");
568 if (!strcmp(head_points_at, "HEAD"))
569 /* detached HEAD */
570 null_is_error = 1;
571 else if (!starts_with(head_points_at, "refs/heads/")) {
572 errors_found |= ERROR_REFS;
573 return error("HEAD points to something strange (%s)",
574 head_points_at);
576 if (is_null_oid(&head_oid)) {
577 if (null_is_error) {
578 errors_found |= ERROR_REFS;
579 return error("HEAD: detached HEAD points at nothing");
581 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
582 head_points_at + 11);
584 return 0;
587 static int fsck_cache_tree(struct cache_tree *it)
589 int i;
590 int err = 0;
592 if (verbose)
593 fprintf(stderr, "Checking cache tree\n");
595 if (0 <= it->entry_count) {
596 struct object *obj = parse_object(&it->oid);
597 if (!obj) {
598 error("%s: invalid sha1 pointer in cache-tree",
599 oid_to_hex(&it->oid));
600 errors_found |= ERROR_REFS;
601 return 1;
603 obj->flags |= USED;
604 if (name_objects)
605 add_decoration(fsck_walk_options.object_names,
606 obj, xstrdup(":"));
607 mark_object_reachable(obj);
608 if (obj->type != OBJ_TREE)
609 err |= objerror(obj, "non-tree in cache-tree");
611 for (i = 0; i < it->subtree_nr; i++)
612 err |= fsck_cache_tree(it->down[i]->cache_tree);
613 return err;
616 static void mark_object_for_connectivity(const struct object_id *oid)
618 struct object *obj = lookup_unknown_object(oid->hash);
619 obj->flags |= HAS_OBJ;
622 static int mark_loose_for_connectivity(const struct object_id *oid,
623 const char *path,
624 void *data)
626 mark_object_for_connectivity(oid);
627 return 0;
630 static int mark_packed_for_connectivity(const struct object_id *oid,
631 struct packed_git *pack,
632 uint32_t pos,
633 void *data)
635 mark_object_for_connectivity(oid);
636 return 0;
639 static char const * const fsck_usage[] = {
640 N_("git fsck [<options>] [<object>...]"),
641 NULL
644 static struct option fsck_opts[] = {
645 OPT__VERBOSE(&verbose, N_("be verbose")),
646 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
647 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
648 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
649 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
650 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
651 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
652 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
653 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
654 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
655 OPT_BOOL(0, "lost-found", &write_lost_and_found,
656 N_("write dangling objects in .git/lost-found")),
657 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
658 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
659 OPT_END(),
662 int cmd_fsck(int argc, const char **argv, const char *prefix)
664 int i;
665 struct alternate_object_database *alt;
667 errors_found = 0;
668 check_replace_refs = 0;
670 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
672 fsck_walk_options.walk = mark_object;
673 fsck_obj_options.walk = mark_used;
674 fsck_obj_options.error_func = fsck_error_func;
675 if (check_strict)
676 fsck_obj_options.strict = 1;
678 if (show_progress == -1)
679 show_progress = isatty(2);
680 if (verbose)
681 show_progress = 0;
683 if (write_lost_and_found) {
684 check_full = 1;
685 include_reflogs = 0;
688 if (name_objects)
689 fsck_walk_options.object_names =
690 xcalloc(1, sizeof(struct decoration));
692 git_config(fsck_config, NULL);
694 fsck_head_link();
695 if (connectivity_only) {
696 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
697 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
698 } else {
699 fsck_object_dir(get_object_directory());
701 prepare_alt_odb();
702 for (alt = alt_odb_list; alt; alt = alt->next)
703 fsck_object_dir(alt->path);
705 if (check_full) {
706 struct packed_git *p;
707 uint32_t total = 0, count = 0;
708 struct progress *progress = NULL;
710 prepare_packed_git();
712 if (show_progress) {
713 for (p = packed_git; p; p = p->next) {
714 if (open_pack_index(p))
715 continue;
716 total += p->num_objects;
719 progress = start_progress(_("Checking objects"), total);
721 for (p = packed_git; p; p = p->next) {
722 /* verify gives error messages itself */
723 if (verify_pack(p, fsck_obj_buffer,
724 progress, count))
725 errors_found |= ERROR_PACK;
726 count += p->num_objects;
728 stop_progress(&progress);
732 for (i = 0; i < argc; i++) {
733 const char *arg = argv[i];
734 struct object_id oid;
735 if (!get_oid(arg, &oid)) {
736 struct object *obj = lookup_object(oid.hash);
738 if (!obj || !(obj->flags & HAS_OBJ)) {
739 error("%s: object missing", oid_to_hex(&oid));
740 errors_found |= ERROR_OBJECT;
741 continue;
744 obj->flags |= USED;
745 if (name_objects)
746 add_decoration(fsck_walk_options.object_names,
747 obj, xstrdup(arg));
748 mark_object_reachable(obj);
749 continue;
751 error("invalid parameter: expected sha1, got '%s'", arg);
752 errors_found |= ERROR_OBJECT;
756 * If we've not been given any explicit head information, do the
757 * default ones from .git/refs. We also consider the index file
758 * in this case (ie this implies --cache).
760 if (!argc) {
761 get_default_heads();
762 keep_cache_objects = 1;
765 if (keep_cache_objects) {
766 verify_index_checksum = 1;
767 read_cache();
768 for (i = 0; i < active_nr; i++) {
769 unsigned int mode;
770 struct blob *blob;
771 struct object *obj;
773 mode = active_cache[i]->ce_mode;
774 if (S_ISGITLINK(mode))
775 continue;
776 blob = lookup_blob(&active_cache[i]->oid);
777 if (!blob)
778 continue;
779 obj = &blob->object;
780 obj->flags |= USED;
781 if (name_objects)
782 add_decoration(fsck_walk_options.object_names,
783 obj,
784 xstrfmt(":%s", active_cache[i]->name));
785 mark_object_reachable(obj);
787 if (active_cache_tree)
788 fsck_cache_tree(active_cache_tree);
791 check_connectivity();
792 return errors_found;