bulk-checkin: convert index_bulk_checkin to struct object_id
[git.git] / builtin / fsck.c
blobef78c6c00cbf4401ed672d6ad954bbbc68c9c115
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 = type_name(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" : type_name(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;
153 if (is_promisor_object(&obj->oid))
155 * Further recursion does not need to be performed on this
156 * object since it is a promisor object (so it does not need to
157 * be added to "pending").
159 return 0;
161 if (!(obj->flags & HAS_OBJ)) {
162 if (parent && !has_object_file(&obj->oid)) {
163 printf("broken link from %7s %s\n",
164 printable_type(parent), describe_object(parent));
165 printf(" to %7s %s\n",
166 printable_type(obj), describe_object(obj));
167 errors_found |= ERROR_REACHABLE;
169 return 1;
172 add_object_array(obj, NULL, &pending);
173 return 0;
176 static void mark_object_reachable(struct object *obj)
178 mark_object(obj, OBJ_ANY, NULL, NULL);
181 static int traverse_one_object(struct object *obj)
183 int result = fsck_walk(obj, obj, &fsck_walk_options);
185 if (obj->type == OBJ_TREE) {
186 struct tree *tree = (struct tree *)obj;
187 free_tree_buffer(tree);
189 return result;
192 static int traverse_reachable(void)
194 struct progress *progress = NULL;
195 unsigned int nr = 0;
196 int result = 0;
197 if (show_progress)
198 progress = start_delayed_progress(_("Checking connectivity"), 0);
199 while (pending.nr) {
200 result |= traverse_one_object(object_array_pop(&pending));
201 display_progress(progress, ++nr);
203 stop_progress(&progress);
204 return !!result;
207 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
209 if (!obj)
210 return 1;
211 obj->flags |= USED;
212 return 0;
216 * Check a single reachable object
218 static void check_reachable_object(struct object *obj)
221 * We obviously want the object to be parsed,
222 * except if it was in a pack-file and we didn't
223 * do a full fsck
225 if (!(obj->flags & HAS_OBJ)) {
226 if (is_promisor_object(&obj->oid))
227 return;
228 if (has_sha1_pack(obj->oid.hash))
229 return; /* it is in pack - forget about it */
230 printf("missing %s %s\n", printable_type(obj),
231 describe_object(obj));
232 errors_found |= ERROR_REACHABLE;
233 return;
238 * Check a single unreachable object
240 static void check_unreachable_object(struct object *obj)
243 * Missing unreachable object? Ignore it. It's not like
244 * we miss it (since it can't be reached), nor do we want
245 * to complain about it being unreachable (since it does
246 * not exist).
248 if (!(obj->flags & HAS_OBJ))
249 return;
252 * Unreachable object that exists? Show it if asked to,
253 * since this is something that is prunable.
255 if (show_unreachable) {
256 printf("unreachable %s %s\n", printable_type(obj),
257 describe_object(obj));
258 return;
262 * "!USED" means that nothing at all points to it, including
263 * other unreachable objects. In other words, it's the "tip"
264 * of some set of unreachable objects, usually a commit that
265 * got dropped.
267 * Such starting points are more interesting than some random
268 * set of unreachable objects, so we show them even if the user
269 * hasn't asked for _all_ unreachable objects. If you have
270 * deleted a branch by mistake, this is a prime candidate to
271 * start looking at, for example.
273 if (!(obj->flags & USED)) {
274 if (show_dangling)
275 printf("dangling %s %s\n", printable_type(obj),
276 describe_object(obj));
277 if (write_lost_and_found) {
278 char *filename = git_pathdup("lost-found/%s/%s",
279 obj->type == OBJ_COMMIT ? "commit" : "other",
280 describe_object(obj));
281 FILE *f;
283 if (safe_create_leading_directories_const(filename)) {
284 error("Could not create lost-found");
285 free(filename);
286 return;
288 f = xfopen(filename, "w");
289 if (obj->type == OBJ_BLOB) {
290 if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
291 die_errno("Could not write '%s'", filename);
292 } else
293 fprintf(f, "%s\n", describe_object(obj));
294 if (fclose(f))
295 die_errno("Could not finish '%s'",
296 filename);
297 free(filename);
299 return;
303 * Otherwise? It's there, it's unreachable, and some other unreachable
304 * object points to it. Ignore it - it's not interesting, and we showed
305 * all the interesting cases above.
309 static void check_object(struct object *obj)
311 if (verbose)
312 fprintf(stderr, "Checking %s\n", describe_object(obj));
314 if (obj->flags & REACHABLE)
315 check_reachable_object(obj);
316 else
317 check_unreachable_object(obj);
320 static void check_connectivity(void)
322 int i, max;
324 /* Traverse the pending reachable objects */
325 traverse_reachable();
327 /* Look up all the requirements, warn about missing objects.. */
328 max = get_max_object_index();
329 if (verbose)
330 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
332 for (i = 0; i < max; i++) {
333 struct object *obj = get_indexed_object(i);
335 if (obj)
336 check_object(obj);
340 static int fsck_obj(struct object *obj)
342 int err;
344 if (obj->flags & SEEN)
345 return 0;
346 obj->flags |= SEEN;
348 if (verbose)
349 fprintf(stderr, "Checking %s %s\n",
350 printable_type(obj), describe_object(obj));
352 if (fsck_walk(obj, NULL, &fsck_obj_options))
353 objerror(obj, "broken links");
354 err = fsck_object(obj, NULL, 0, &fsck_obj_options);
355 if (err)
356 goto out;
358 if (obj->type == OBJ_COMMIT) {
359 struct commit *commit = (struct commit *) obj;
361 if (!commit->parents && show_root)
362 printf("root %s\n", describe_object(&commit->object));
365 if (obj->type == OBJ_TAG) {
366 struct tag *tag = (struct tag *) obj;
368 if (show_tags && tag->tagged) {
369 printf("tagged %s %s", printable_type(tag->tagged),
370 describe_object(tag->tagged));
371 printf(" (%s) in %s\n", tag->tag,
372 describe_object(&tag->object));
376 out:
377 if (obj->type == OBJ_TREE)
378 free_tree_buffer((struct tree *)obj);
379 if (obj->type == OBJ_COMMIT)
380 free_commit_buffer((struct commit *)obj);
381 return err;
384 static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
385 unsigned long size, void *buffer, int *eaten)
388 * Note, buffer may be NULL if type is OBJ_BLOB. See
389 * verify_packfile(), data_valid variable for details.
391 struct object *obj;
392 obj = parse_object_buffer(oid, type, size, buffer, eaten);
393 if (!obj) {
394 errors_found |= ERROR_OBJECT;
395 return error("%s: object corrupt or missing", oid_to_hex(oid));
397 obj->flags &= ~(REACHABLE | SEEN);
398 obj->flags |= HAS_OBJ;
399 return fsck_obj(obj);
402 static int default_refs;
404 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
405 timestamp_t timestamp)
407 struct object *obj;
409 if (!is_null_oid(oid)) {
410 obj = lookup_object(oid->hash);
411 if (obj && (obj->flags & HAS_OBJ)) {
412 if (timestamp && name_objects)
413 add_decoration(fsck_walk_options.object_names,
414 obj,
415 xstrfmt("%s@{%"PRItime"}", refname, timestamp));
416 obj->flags |= USED;
417 mark_object_reachable(obj);
418 } else if (!is_promisor_object(oid)) {
419 error("%s: invalid reflog entry %s", refname, oid_to_hex(oid));
420 errors_found |= ERROR_REACHABLE;
425 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
426 const char *email, timestamp_t timestamp, int tz,
427 const char *message, void *cb_data)
429 const char *refname = cb_data;
431 if (verbose)
432 fprintf(stderr, "Checking reflog %s->%s\n",
433 oid_to_hex(ooid), oid_to_hex(noid));
435 fsck_handle_reflog_oid(refname, ooid, 0);
436 fsck_handle_reflog_oid(refname, noid, timestamp);
437 return 0;
440 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
441 int flag, void *cb_data)
443 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
444 return 0;
447 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
448 int flag, void *cb_data)
450 struct object *obj;
452 obj = parse_object(oid);
453 if (!obj) {
454 if (is_promisor_object(oid)) {
456 * Increment default_refs anyway, because this is a
457 * valid ref.
459 default_refs++;
460 return 0;
462 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
463 errors_found |= ERROR_REACHABLE;
464 /* We'll continue with the rest despite the error.. */
465 return 0;
467 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
468 error("%s: not a commit", refname);
469 errors_found |= ERROR_REFS;
471 default_refs++;
472 obj->flags |= USED;
473 if (name_objects)
474 add_decoration(fsck_walk_options.object_names,
475 obj, xstrdup(refname));
476 mark_object_reachable(obj);
478 return 0;
481 static void get_default_heads(void)
483 if (head_points_at && !is_null_oid(&head_oid))
484 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
485 for_each_rawref(fsck_handle_ref, NULL);
486 if (include_reflogs)
487 for_each_reflog(fsck_handle_reflog, NULL);
490 * Not having any default heads isn't really fatal, but
491 * it does mean that "--unreachable" no longer makes any
492 * sense (since in this case everything will obviously
493 * be unreachable by definition.
495 * Showing dangling objects is valid, though (as those
496 * dangling objects are likely lost heads).
498 * So we just print a warning about it, and clear the
499 * "show_unreachable" flag.
501 if (!default_refs) {
502 fprintf(stderr, "notice: No default references\n");
503 show_unreachable = 0;
507 static struct object *parse_loose_object(const struct object_id *oid,
508 const char *path)
510 struct object *obj;
511 void *contents;
512 enum object_type type;
513 unsigned long size;
514 int eaten;
516 if (read_loose_object(path, oid->hash, &type, &size, &contents) < 0)
517 return NULL;
519 if (!contents && type != OBJ_BLOB)
520 die("BUG: read_loose_object streamed a non-blob");
522 obj = parse_object_buffer(oid, type, size, contents, &eaten);
524 if (!eaten)
525 free(contents);
526 return obj;
529 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
531 struct object *obj = parse_loose_object(oid, path);
533 if (!obj) {
534 errors_found |= ERROR_OBJECT;
535 error("%s: object corrupt or missing: %s",
536 oid_to_hex(oid), path);
537 return 0; /* keep checking other objects */
540 obj->flags &= ~(REACHABLE | SEEN);
541 obj->flags |= HAS_OBJ;
542 if (fsck_obj(obj))
543 errors_found |= ERROR_OBJECT;
544 return 0;
547 static int fsck_cruft(const char *basename, const char *path, void *data)
549 if (!starts_with(basename, "tmp_obj_"))
550 fprintf(stderr, "bad sha1 file: %s\n", path);
551 return 0;
554 static int fsck_subdir(unsigned int nr, const char *path, void *progress)
556 display_progress(progress, nr + 1);
557 return 0;
560 static void fsck_object_dir(const char *path)
562 struct progress *progress = NULL;
564 if (verbose)
565 fprintf(stderr, "Checking object directory\n");
567 if (show_progress)
568 progress = start_progress(_("Checking object directories"), 256);
570 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
571 progress);
572 display_progress(progress, 256);
573 stop_progress(&progress);
576 static int fsck_head_link(void)
578 int null_is_error = 0;
580 if (verbose)
581 fprintf(stderr, "Checking HEAD link\n");
583 head_points_at = resolve_ref_unsafe("HEAD", 0, &head_oid, NULL);
584 if (!head_points_at) {
585 errors_found |= ERROR_REFS;
586 return error("Invalid HEAD");
588 if (!strcmp(head_points_at, "HEAD"))
589 /* detached HEAD */
590 null_is_error = 1;
591 else if (!starts_with(head_points_at, "refs/heads/")) {
592 errors_found |= ERROR_REFS;
593 return error("HEAD points to something strange (%s)",
594 head_points_at);
596 if (is_null_oid(&head_oid)) {
597 if (null_is_error) {
598 errors_found |= ERROR_REFS;
599 return error("HEAD: detached HEAD points at nothing");
601 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
602 head_points_at + 11);
604 return 0;
607 static int fsck_cache_tree(struct cache_tree *it)
609 int i;
610 int err = 0;
612 if (verbose)
613 fprintf(stderr, "Checking cache tree\n");
615 if (0 <= it->entry_count) {
616 struct object *obj = parse_object(&it->oid);
617 if (!obj) {
618 error("%s: invalid sha1 pointer in cache-tree",
619 oid_to_hex(&it->oid));
620 errors_found |= ERROR_REFS;
621 return 1;
623 obj->flags |= USED;
624 if (name_objects)
625 add_decoration(fsck_walk_options.object_names,
626 obj, xstrdup(":"));
627 mark_object_reachable(obj);
628 if (obj->type != OBJ_TREE)
629 err |= objerror(obj, "non-tree in cache-tree");
631 for (i = 0; i < it->subtree_nr; i++)
632 err |= fsck_cache_tree(it->down[i]->cache_tree);
633 return err;
636 static void mark_object_for_connectivity(const struct object_id *oid)
638 struct object *obj = lookup_unknown_object(oid->hash);
639 obj->flags |= HAS_OBJ;
642 static int mark_loose_for_connectivity(const struct object_id *oid,
643 const char *path,
644 void *data)
646 mark_object_for_connectivity(oid);
647 return 0;
650 static int mark_packed_for_connectivity(const struct object_id *oid,
651 struct packed_git *pack,
652 uint32_t pos,
653 void *data)
655 mark_object_for_connectivity(oid);
656 return 0;
659 static char const * const fsck_usage[] = {
660 N_("git fsck [<options>] [<object>...]"),
661 NULL
664 static struct option fsck_opts[] = {
665 OPT__VERBOSE(&verbose, N_("be verbose")),
666 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
667 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
668 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
669 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
670 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
671 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
672 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
673 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
674 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
675 OPT_BOOL(0, "lost-found", &write_lost_and_found,
676 N_("write dangling objects in .git/lost-found")),
677 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
678 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
679 OPT_END(),
682 int cmd_fsck(int argc, const char **argv, const char *prefix)
684 int i;
685 struct alternate_object_database *alt;
687 /* fsck knows how to handle missing promisor objects */
688 fetch_if_missing = 0;
690 errors_found = 0;
691 check_replace_refs = 0;
693 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
695 fsck_walk_options.walk = mark_object;
696 fsck_obj_options.walk = mark_used;
697 fsck_obj_options.error_func = fsck_error_func;
698 if (check_strict)
699 fsck_obj_options.strict = 1;
701 if (show_progress == -1)
702 show_progress = isatty(2);
703 if (verbose)
704 show_progress = 0;
706 if (write_lost_and_found) {
707 check_full = 1;
708 include_reflogs = 0;
711 if (name_objects)
712 fsck_walk_options.object_names =
713 xcalloc(1, sizeof(struct decoration));
715 git_config(fsck_config, NULL);
717 fsck_head_link();
718 if (connectivity_only) {
719 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
720 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
721 } else {
722 fsck_object_dir(get_object_directory());
724 prepare_alt_odb();
725 for (alt = alt_odb_list; alt; alt = alt->next)
726 fsck_object_dir(alt->path);
728 if (check_full) {
729 struct packed_git *p;
730 uint32_t total = 0, count = 0;
731 struct progress *progress = NULL;
733 prepare_packed_git();
735 if (show_progress) {
736 for (p = packed_git; p; p = p->next) {
737 if (open_pack_index(p))
738 continue;
739 total += p->num_objects;
742 progress = start_progress(_("Checking objects"), total);
744 for (p = packed_git; p; p = p->next) {
745 /* verify gives error messages itself */
746 if (verify_pack(p, fsck_obj_buffer,
747 progress, count))
748 errors_found |= ERROR_PACK;
749 count += p->num_objects;
751 stop_progress(&progress);
755 for (i = 0; i < argc; i++) {
756 const char *arg = argv[i];
757 struct object_id oid;
758 if (!get_oid(arg, &oid)) {
759 struct object *obj = lookup_object(oid.hash);
761 if (!obj || !(obj->flags & HAS_OBJ)) {
762 if (is_promisor_object(&oid))
763 continue;
764 error("%s: object missing", oid_to_hex(&oid));
765 errors_found |= ERROR_OBJECT;
766 continue;
769 obj->flags |= USED;
770 if (name_objects)
771 add_decoration(fsck_walk_options.object_names,
772 obj, xstrdup(arg));
773 mark_object_reachable(obj);
774 continue;
776 error("invalid parameter: expected sha1, got '%s'", arg);
777 errors_found |= ERROR_OBJECT;
781 * If we've not been given any explicit head information, do the
782 * default ones from .git/refs. We also consider the index file
783 * in this case (ie this implies --cache).
785 if (!argc) {
786 get_default_heads();
787 keep_cache_objects = 1;
790 if (keep_cache_objects) {
791 verify_index_checksum = 1;
792 verify_ce_order = 1;
793 read_cache();
794 for (i = 0; i < active_nr; i++) {
795 unsigned int mode;
796 struct blob *blob;
797 struct object *obj;
799 mode = active_cache[i]->ce_mode;
800 if (S_ISGITLINK(mode))
801 continue;
802 blob = lookup_blob(&active_cache[i]->oid);
803 if (!blob)
804 continue;
805 obj = &blob->object;
806 obj->flags |= USED;
807 if (name_objects)
808 add_decoration(fsck_walk_options.object_names,
809 obj,
810 xstrfmt(":%s", active_cache[i]->name));
811 mark_object_reachable(obj);
813 if (active_cache_tree)
814 fsck_cache_tree(active_cache_tree);
817 check_connectivity();
818 return errors_found;