list-objects: check if filter is NULL before using
[git/gitster.git] / builtin / fsck.c
blob3b76c0ef0f4beddd2af5500aa435ea67ebf573e7
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;
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 return fsck_walk(obj, obj, &fsck_walk_options);
186 static int traverse_reachable(void)
188 struct progress *progress = NULL;
189 unsigned int nr = 0;
190 int result = 0;
191 if (show_progress)
192 progress = start_delayed_progress(_("Checking connectivity"), 0);
193 while (pending.nr) {
194 result |= traverse_one_object(object_array_pop(&pending));
195 display_progress(progress, ++nr);
197 stop_progress(&progress);
198 return !!result;
201 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
203 if (!obj)
204 return 1;
205 obj->flags |= USED;
206 return 0;
210 * Check a single reachable object
212 static void check_reachable_object(struct object *obj)
215 * We obviously want the object to be parsed,
216 * except if it was in a pack-file and we didn't
217 * do a full fsck
219 if (!(obj->flags & HAS_OBJ)) {
220 if (is_promisor_object(&obj->oid))
221 return;
222 if (has_sha1_pack(obj->oid.hash))
223 return; /* it is in pack - forget about it */
224 printf("missing %s %s\n", printable_type(obj),
225 describe_object(obj));
226 errors_found |= ERROR_REACHABLE;
227 return;
232 * Check a single unreachable object
234 static void check_unreachable_object(struct object *obj)
237 * Missing unreachable object? Ignore it. It's not like
238 * we miss it (since it can't be reached), nor do we want
239 * to complain about it being unreachable (since it does
240 * not exist).
242 if (!(obj->flags & HAS_OBJ))
243 return;
246 * Unreachable object that exists? Show it if asked to,
247 * since this is something that is prunable.
249 if (show_unreachable) {
250 printf("unreachable %s %s\n", printable_type(obj),
251 describe_object(obj));
252 return;
256 * "!USED" means that nothing at all points to it, including
257 * other unreachable objects. In other words, it's the "tip"
258 * of some set of unreachable objects, usually a commit that
259 * got dropped.
261 * Such starting points are more interesting than some random
262 * set of unreachable objects, so we show them even if the user
263 * hasn't asked for _all_ unreachable objects. If you have
264 * deleted a branch by mistake, this is a prime candidate to
265 * start looking at, for example.
267 if (!(obj->flags & USED)) {
268 if (show_dangling)
269 printf("dangling %s %s\n", printable_type(obj),
270 describe_object(obj));
271 if (write_lost_and_found) {
272 char *filename = git_pathdup("lost-found/%s/%s",
273 obj->type == OBJ_COMMIT ? "commit" : "other",
274 describe_object(obj));
275 FILE *f;
277 if (safe_create_leading_directories_const(filename)) {
278 error("Could not create lost-found");
279 free(filename);
280 return;
282 f = xfopen(filename, "w");
283 if (obj->type == OBJ_BLOB) {
284 if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
285 die_errno("Could not write '%s'", filename);
286 } else
287 fprintf(f, "%s\n", describe_object(obj));
288 if (fclose(f))
289 die_errno("Could not finish '%s'",
290 filename);
291 free(filename);
293 return;
297 * Otherwise? It's there, it's unreachable, and some other unreachable
298 * object points to it. Ignore it - it's not interesting, and we showed
299 * all the interesting cases above.
303 static void check_object(struct object *obj)
305 if (verbose)
306 fprintf(stderr, "Checking %s\n", describe_object(obj));
308 if (obj->flags & REACHABLE)
309 check_reachable_object(obj);
310 else
311 check_unreachable_object(obj);
314 static void check_connectivity(void)
316 int i, max;
318 /* Traverse the pending reachable objects */
319 traverse_reachable();
321 /* Look up all the requirements, warn about missing objects.. */
322 max = get_max_object_index();
323 if (verbose)
324 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
326 for (i = 0; i < max; i++) {
327 struct object *obj = get_indexed_object(i);
329 if (obj)
330 check_object(obj);
334 static int fsck_obj(struct object *obj)
336 int err;
338 if (obj->flags & SEEN)
339 return 0;
340 obj->flags |= SEEN;
342 if (verbose)
343 fprintf(stderr, "Checking %s %s\n",
344 printable_type(obj), describe_object(obj));
346 if (fsck_walk(obj, NULL, &fsck_obj_options))
347 objerror(obj, "broken links");
348 err = fsck_object(obj, NULL, 0, &fsck_obj_options);
349 if (err)
350 goto out;
352 if (obj->type == OBJ_COMMIT) {
353 struct commit *commit = (struct commit *) obj;
355 if (!commit->parents && show_root)
356 printf("root %s\n", describe_object(&commit->object));
359 if (obj->type == OBJ_TAG) {
360 struct tag *tag = (struct tag *) obj;
362 if (show_tags && tag->tagged) {
363 printf("tagged %s %s", printable_type(tag->tagged),
364 describe_object(tag->tagged));
365 printf(" (%s) in %s\n", tag->tag,
366 describe_object(&tag->object));
370 out:
371 if (obj->type == OBJ_TREE)
372 free_tree_buffer((struct tree *)obj);
373 if (obj->type == OBJ_COMMIT)
374 free_commit_buffer((struct commit *)obj);
375 return err;
378 static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
379 unsigned long size, void *buffer, int *eaten)
382 * Note, buffer may be NULL if type is OBJ_BLOB. See
383 * verify_packfile(), data_valid variable for details.
385 struct object *obj;
386 obj = parse_object_buffer(oid, type, size, buffer, eaten);
387 if (!obj) {
388 errors_found |= ERROR_OBJECT;
389 return error("%s: object corrupt or missing", oid_to_hex(oid));
391 obj->flags &= ~(REACHABLE | SEEN);
392 obj->flags |= HAS_OBJ;
393 return fsck_obj(obj);
396 static int default_refs;
398 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
399 timestamp_t timestamp)
401 struct object *obj;
403 if (!is_null_oid(oid)) {
404 obj = lookup_object(oid->hash);
405 if (obj && (obj->flags & HAS_OBJ)) {
406 if (timestamp && name_objects)
407 add_decoration(fsck_walk_options.object_names,
408 obj,
409 xstrfmt("%s@{%"PRItime"}", refname, timestamp));
410 obj->flags |= USED;
411 mark_object_reachable(obj);
412 } else if (!is_promisor_object(oid)) {
413 error("%s: invalid reflog entry %s", refname, oid_to_hex(oid));
414 errors_found |= ERROR_REACHABLE;
419 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
420 const char *email, timestamp_t timestamp, int tz,
421 const char *message, void *cb_data)
423 const char *refname = cb_data;
425 if (verbose)
426 fprintf(stderr, "Checking reflog %s->%s\n",
427 oid_to_hex(ooid), oid_to_hex(noid));
429 fsck_handle_reflog_oid(refname, ooid, 0);
430 fsck_handle_reflog_oid(refname, noid, timestamp);
431 return 0;
434 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
435 int flag, void *cb_data)
437 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
438 return 0;
441 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
442 int flag, void *cb_data)
444 struct object *obj;
446 obj = parse_object(oid);
447 if (!obj) {
448 if (is_promisor_object(oid)) {
450 * Increment default_refs anyway, because this is a
451 * valid ref.
453 default_refs++;
454 return 0;
456 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
457 errors_found |= ERROR_REACHABLE;
458 /* We'll continue with the rest despite the error.. */
459 return 0;
461 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
462 error("%s: not a commit", refname);
463 errors_found |= ERROR_REFS;
465 default_refs++;
466 obj->flags |= USED;
467 if (name_objects)
468 add_decoration(fsck_walk_options.object_names,
469 obj, xstrdup(refname));
470 mark_object_reachable(obj);
472 return 0;
475 static void get_default_heads(void)
477 if (head_points_at && !is_null_oid(&head_oid))
478 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
479 for_each_rawref(fsck_handle_ref, NULL);
480 if (include_reflogs)
481 for_each_reflog(fsck_handle_reflog, NULL);
484 * Not having any default heads isn't really fatal, but
485 * it does mean that "--unreachable" no longer makes any
486 * sense (since in this case everything will obviously
487 * be unreachable by definition.
489 * Showing dangling objects is valid, though (as those
490 * dangling objects are likely lost heads).
492 * So we just print a warning about it, and clear the
493 * "show_unreachable" flag.
495 if (!default_refs) {
496 fprintf(stderr, "notice: No default references\n");
497 show_unreachable = 0;
501 static struct object *parse_loose_object(const struct object_id *oid,
502 const char *path)
504 struct object *obj;
505 void *contents;
506 enum object_type type;
507 unsigned long size;
508 int eaten;
510 if (read_loose_object(path, oid->hash, &type, &size, &contents) < 0)
511 return NULL;
513 if (!contents && type != OBJ_BLOB)
514 die("BUG: read_loose_object streamed a non-blob");
516 obj = parse_object_buffer(oid, type, size, contents, &eaten);
518 if (!eaten)
519 free(contents);
520 return obj;
523 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
525 struct object *obj = parse_loose_object(oid, path);
527 if (!obj) {
528 errors_found |= ERROR_OBJECT;
529 error("%s: object corrupt or missing: %s",
530 oid_to_hex(oid), path);
531 return 0; /* keep checking other objects */
534 obj->flags &= ~(REACHABLE | SEEN);
535 obj->flags |= HAS_OBJ;
536 if (fsck_obj(obj))
537 errors_found |= ERROR_OBJECT;
538 return 0;
541 static int fsck_cruft(const char *basename, const char *path, void *data)
543 if (!starts_with(basename, "tmp_obj_"))
544 fprintf(stderr, "bad sha1 file: %s\n", path);
545 return 0;
548 static int fsck_subdir(unsigned int nr, const char *path, void *progress)
550 display_progress(progress, nr + 1);
551 return 0;
554 static void fsck_object_dir(const char *path)
556 struct progress *progress = NULL;
558 if (verbose)
559 fprintf(stderr, "Checking object directory\n");
561 if (show_progress)
562 progress = start_progress(_("Checking object directories"), 256);
564 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
565 progress);
566 display_progress(progress, 256);
567 stop_progress(&progress);
570 static int fsck_head_link(void)
572 int null_is_error = 0;
574 if (verbose)
575 fprintf(stderr, "Checking HEAD link\n");
577 head_points_at = resolve_ref_unsafe("HEAD", 0, head_oid.hash, NULL);
578 if (!head_points_at) {
579 errors_found |= ERROR_REFS;
580 return error("Invalid HEAD");
582 if (!strcmp(head_points_at, "HEAD"))
583 /* detached HEAD */
584 null_is_error = 1;
585 else if (!starts_with(head_points_at, "refs/heads/")) {
586 errors_found |= ERROR_REFS;
587 return error("HEAD points to something strange (%s)",
588 head_points_at);
590 if (is_null_oid(&head_oid)) {
591 if (null_is_error) {
592 errors_found |= ERROR_REFS;
593 return error("HEAD: detached HEAD points at nothing");
595 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
596 head_points_at + 11);
598 return 0;
601 static int fsck_cache_tree(struct cache_tree *it)
603 int i;
604 int err = 0;
606 if (verbose)
607 fprintf(stderr, "Checking cache tree\n");
609 if (0 <= it->entry_count) {
610 struct object *obj = parse_object(&it->oid);
611 if (!obj) {
612 error("%s: invalid sha1 pointer in cache-tree",
613 oid_to_hex(&it->oid));
614 errors_found |= ERROR_REFS;
615 return 1;
617 obj->flags |= USED;
618 if (name_objects)
619 add_decoration(fsck_walk_options.object_names,
620 obj, xstrdup(":"));
621 mark_object_reachable(obj);
622 if (obj->type != OBJ_TREE)
623 err |= objerror(obj, "non-tree in cache-tree");
625 for (i = 0; i < it->subtree_nr; i++)
626 err |= fsck_cache_tree(it->down[i]->cache_tree);
627 return err;
630 static void mark_object_for_connectivity(const struct object_id *oid)
632 struct object *obj = lookup_unknown_object(oid->hash);
633 obj->flags |= HAS_OBJ;
636 static int mark_loose_for_connectivity(const struct object_id *oid,
637 const char *path,
638 void *data)
640 mark_object_for_connectivity(oid);
641 return 0;
644 static int mark_packed_for_connectivity(const struct object_id *oid,
645 struct packed_git *pack,
646 uint32_t pos,
647 void *data)
649 mark_object_for_connectivity(oid);
650 return 0;
653 static char const * const fsck_usage[] = {
654 N_("git fsck [<options>] [<object>...]"),
655 NULL
658 static struct option fsck_opts[] = {
659 OPT__VERBOSE(&verbose, N_("be verbose")),
660 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
661 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
662 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
663 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
664 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
665 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
666 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
667 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
668 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
669 OPT_BOOL(0, "lost-found", &write_lost_and_found,
670 N_("write dangling objects in .git/lost-found")),
671 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
672 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
673 OPT_END(),
676 int cmd_fsck(int argc, const char **argv, const char *prefix)
678 int i;
679 struct alternate_object_database *alt;
681 /* fsck knows how to handle missing promisor objects */
682 fetch_if_missing = 0;
684 errors_found = 0;
685 check_replace_refs = 0;
687 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
689 fsck_walk_options.walk = mark_object;
690 fsck_obj_options.walk = mark_used;
691 fsck_obj_options.error_func = fsck_error_func;
692 if (check_strict)
693 fsck_obj_options.strict = 1;
695 if (show_progress == -1)
696 show_progress = isatty(2);
697 if (verbose)
698 show_progress = 0;
700 if (write_lost_and_found) {
701 check_full = 1;
702 include_reflogs = 0;
705 if (name_objects)
706 fsck_walk_options.object_names =
707 xcalloc(1, sizeof(struct decoration));
709 git_config(fsck_config, NULL);
711 fsck_head_link();
712 if (connectivity_only) {
713 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
714 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
715 } else {
716 fsck_object_dir(get_object_directory());
718 prepare_alt_odb();
719 for (alt = alt_odb_list; alt; alt = alt->next)
720 fsck_object_dir(alt->path);
722 if (check_full) {
723 struct packed_git *p;
724 uint32_t total = 0, count = 0;
725 struct progress *progress = NULL;
727 prepare_packed_git();
729 if (show_progress) {
730 for (p = packed_git; p; p = p->next) {
731 if (open_pack_index(p))
732 continue;
733 total += p->num_objects;
736 progress = start_progress(_("Checking objects"), total);
738 for (p = packed_git; p; p = p->next) {
739 /* verify gives error messages itself */
740 if (verify_pack(p, fsck_obj_buffer,
741 progress, count))
742 errors_found |= ERROR_PACK;
743 count += p->num_objects;
745 stop_progress(&progress);
749 for (i = 0; i < argc; i++) {
750 const char *arg = argv[i];
751 struct object_id oid;
752 if (!get_oid(arg, &oid)) {
753 struct object *obj = lookup_object(oid.hash);
755 if (!obj || !(obj->flags & HAS_OBJ)) {
756 if (is_promisor_object(&oid))
757 continue;
758 error("%s: object missing", oid_to_hex(&oid));
759 errors_found |= ERROR_OBJECT;
760 continue;
763 obj->flags |= USED;
764 if (name_objects)
765 add_decoration(fsck_walk_options.object_names,
766 obj, xstrdup(arg));
767 mark_object_reachable(obj);
768 continue;
770 error("invalid parameter: expected sha1, got '%s'", arg);
771 errors_found |= ERROR_OBJECT;
775 * If we've not been given any explicit head information, do the
776 * default ones from .git/refs. We also consider the index file
777 * in this case (ie this implies --cache).
779 if (!argc) {
780 get_default_heads();
781 keep_cache_objects = 1;
784 if (keep_cache_objects) {
785 verify_index_checksum = 1;
786 read_cache();
787 for (i = 0; i < active_nr; i++) {
788 unsigned int mode;
789 struct blob *blob;
790 struct object *obj;
792 mode = active_cache[i]->ce_mode;
793 if (S_ISGITLINK(mode))
794 continue;
795 blob = lookup_blob(&active_cache[i]->oid);
796 if (!blob)
797 continue;
798 obj = &blob->object;
799 obj->flags |= USED;
800 if (name_objects)
801 add_decoration(fsck_walk_options.object_names,
802 obj,
803 xstrfmt(":%s", active_cache[i]->name));
804 mark_object_reachable(obj);
806 if (active_cache_tree)
807 fsck_cache_tree(active_cache_tree);
810 check_connectivity();
811 return errors_found;