Merge branch 'jk/doc-diff-options'
[git/debian.git] / builtin / fsck.c
blob04846d46f91d6cde14d379da8e0e3dd458c08dc5
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 result |= traverse_one_object(object_array_pop(&pending));
186 display_progress(progress, ++nr);
188 stop_progress(&progress);
189 return !!result;
192 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
194 if (!obj)
195 return 1;
196 obj->flags |= USED;
197 return 0;
201 * Check a single reachable object
203 static void check_reachable_object(struct object *obj)
206 * We obviously want the object to be parsed,
207 * except if it was in a pack-file and we didn't
208 * do a full fsck
210 if (!(obj->flags & HAS_OBJ)) {
211 if (has_sha1_pack(obj->oid.hash))
212 return; /* it is in pack - forget about it */
213 printf("missing %s %s\n", printable_type(obj),
214 describe_object(obj));
215 errors_found |= ERROR_REACHABLE;
216 return;
221 * Check a single unreachable object
223 static void check_unreachable_object(struct object *obj)
226 * Missing unreachable object? Ignore it. It's not like
227 * we miss it (since it can't be reached), nor do we want
228 * to complain about it being unreachable (since it does
229 * not exist).
231 if (!(obj->flags & HAS_OBJ))
232 return;
235 * Unreachable object that exists? Show it if asked to,
236 * since this is something that is prunable.
238 if (show_unreachable) {
239 printf("unreachable %s %s\n", printable_type(obj),
240 describe_object(obj));
241 return;
245 * "!USED" means that nothing at all points to it, including
246 * other unreachable objects. In other words, it's the "tip"
247 * of some set of unreachable objects, usually a commit that
248 * got dropped.
250 * Such starting points are more interesting than some random
251 * set of unreachable objects, so we show them even if the user
252 * hasn't asked for _all_ unreachable objects. If you have
253 * deleted a branch by mistake, this is a prime candidate to
254 * start looking at, for example.
256 if (!(obj->flags & USED)) {
257 if (show_dangling)
258 printf("dangling %s %s\n", printable_type(obj),
259 describe_object(obj));
260 if (write_lost_and_found) {
261 char *filename = git_pathdup("lost-found/%s/%s",
262 obj->type == OBJ_COMMIT ? "commit" : "other",
263 describe_object(obj));
264 FILE *f;
266 if (safe_create_leading_directories_const(filename)) {
267 error("Could not create lost-found");
268 free(filename);
269 return;
271 f = xfopen(filename, "w");
272 if (obj->type == OBJ_BLOB) {
273 if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
274 die_errno("Could not write '%s'", filename);
275 } else
276 fprintf(f, "%s\n", describe_object(obj));
277 if (fclose(f))
278 die_errno("Could not finish '%s'",
279 filename);
280 free(filename);
282 return;
286 * Otherwise? It's there, it's unreachable, and some other unreachable
287 * object points to it. Ignore it - it's not interesting, and we showed
288 * all the interesting cases above.
292 static void check_object(struct object *obj)
294 if (verbose)
295 fprintf(stderr, "Checking %s\n", describe_object(obj));
297 if (obj->flags & REACHABLE)
298 check_reachable_object(obj);
299 else
300 check_unreachable_object(obj);
303 static void check_connectivity(void)
305 int i, max;
307 /* Traverse the pending reachable objects */
308 traverse_reachable();
310 /* Look up all the requirements, warn about missing objects.. */
311 max = get_max_object_index();
312 if (verbose)
313 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
315 for (i = 0; i < max; i++) {
316 struct object *obj = get_indexed_object(i);
318 if (obj)
319 check_object(obj);
323 static int fsck_obj(struct object *obj)
325 int err;
327 if (obj->flags & SEEN)
328 return 0;
329 obj->flags |= SEEN;
331 if (verbose)
332 fprintf(stderr, "Checking %s %s\n",
333 printable_type(obj), describe_object(obj));
335 if (fsck_walk(obj, NULL, &fsck_obj_options))
336 objerror(obj, "broken links");
337 err = fsck_object(obj, NULL, 0, &fsck_obj_options);
338 if (err)
339 goto out;
341 if (obj->type == OBJ_COMMIT) {
342 struct commit *commit = (struct commit *) obj;
344 if (!commit->parents && show_root)
345 printf("root %s\n", describe_object(&commit->object));
348 if (obj->type == OBJ_TAG) {
349 struct tag *tag = (struct tag *) obj;
351 if (show_tags && tag->tagged) {
352 printf("tagged %s %s", printable_type(tag->tagged),
353 describe_object(tag->tagged));
354 printf(" (%s) in %s\n", tag->tag,
355 describe_object(&tag->object));
359 out:
360 if (obj->type == OBJ_TREE)
361 free_tree_buffer((struct tree *)obj);
362 if (obj->type == OBJ_COMMIT)
363 free_commit_buffer((struct commit *)obj);
364 return err;
367 static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
368 unsigned long size, void *buffer, int *eaten)
371 * Note, buffer may be NULL if type is OBJ_BLOB. See
372 * verify_packfile(), data_valid variable for details.
374 struct object *obj;
375 obj = parse_object_buffer(oid, type, size, buffer, eaten);
376 if (!obj) {
377 errors_found |= ERROR_OBJECT;
378 return error("%s: object corrupt or missing", oid_to_hex(oid));
380 obj->flags &= ~(REACHABLE | SEEN);
381 obj->flags |= HAS_OBJ;
382 return fsck_obj(obj);
385 static int default_refs;
387 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
388 timestamp_t timestamp)
390 struct object *obj;
392 if (!is_null_oid(oid)) {
393 obj = lookup_object(oid->hash);
394 if (obj && (obj->flags & HAS_OBJ)) {
395 if (timestamp && name_objects)
396 add_decoration(fsck_walk_options.object_names,
397 obj,
398 xstrfmt("%s@{%"PRItime"}", refname, timestamp));
399 obj->flags |= USED;
400 mark_object_reachable(obj);
401 } else {
402 error("%s: invalid reflog entry %s", refname, oid_to_hex(oid));
403 errors_found |= ERROR_REACHABLE;
408 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
409 const char *email, timestamp_t timestamp, int tz,
410 const char *message, void *cb_data)
412 const char *refname = cb_data;
414 if (verbose)
415 fprintf(stderr, "Checking reflog %s->%s\n",
416 oid_to_hex(ooid), oid_to_hex(noid));
418 fsck_handle_reflog_oid(refname, ooid, 0);
419 fsck_handle_reflog_oid(refname, noid, timestamp);
420 return 0;
423 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
424 int flag, void *cb_data)
426 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
427 return 0;
430 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
431 int flag, void *cb_data)
433 struct object *obj;
435 obj = parse_object(oid);
436 if (!obj) {
437 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
438 errors_found |= ERROR_REACHABLE;
439 /* We'll continue with the rest despite the error.. */
440 return 0;
442 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
443 error("%s: not a commit", refname);
444 errors_found |= ERROR_REFS;
446 default_refs++;
447 obj->flags |= USED;
448 if (name_objects)
449 add_decoration(fsck_walk_options.object_names,
450 obj, xstrdup(refname));
451 mark_object_reachable(obj);
453 return 0;
456 static void get_default_heads(void)
458 if (head_points_at && !is_null_oid(&head_oid))
459 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
460 for_each_rawref(fsck_handle_ref, NULL);
461 if (include_reflogs)
462 for_each_reflog(fsck_handle_reflog, NULL);
465 * Not having any default heads isn't really fatal, but
466 * it does mean that "--unreachable" no longer makes any
467 * sense (since in this case everything will obviously
468 * be unreachable by definition.
470 * Showing dangling objects is valid, though (as those
471 * dangling objects are likely lost heads).
473 * So we just print a warning about it, and clear the
474 * "show_unreachable" flag.
476 if (!default_refs) {
477 fprintf(stderr, "notice: No default references\n");
478 show_unreachable = 0;
482 static struct object *parse_loose_object(const struct object_id *oid,
483 const char *path)
485 struct object *obj;
486 void *contents;
487 enum object_type type;
488 unsigned long size;
489 int eaten;
491 if (read_loose_object(path, oid->hash, &type, &size, &contents) < 0)
492 return NULL;
494 if (!contents && type != OBJ_BLOB)
495 die("BUG: read_loose_object streamed a non-blob");
497 obj = parse_object_buffer(oid, type, size, contents, &eaten);
499 if (!eaten)
500 free(contents);
501 return obj;
504 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
506 struct object *obj = parse_loose_object(oid, path);
508 if (!obj) {
509 errors_found |= ERROR_OBJECT;
510 error("%s: object corrupt or missing: %s",
511 oid_to_hex(oid), path);
512 return 0; /* keep checking other objects */
515 obj->flags &= ~(REACHABLE | SEEN);
516 obj->flags |= HAS_OBJ;
517 if (fsck_obj(obj))
518 errors_found |= ERROR_OBJECT;
519 return 0;
522 static int fsck_cruft(const char *basename, const char *path, void *data)
524 if (!starts_with(basename, "tmp_obj_"))
525 fprintf(stderr, "bad sha1 file: %s\n", path);
526 return 0;
529 static int fsck_subdir(unsigned int nr, const char *path, void *progress)
531 display_progress(progress, nr + 1);
532 return 0;
535 static void fsck_object_dir(const char *path)
537 struct progress *progress = NULL;
539 if (verbose)
540 fprintf(stderr, "Checking object directory\n");
542 if (show_progress)
543 progress = start_progress(_("Checking object directories"), 256);
545 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
546 progress);
547 display_progress(progress, 256);
548 stop_progress(&progress);
551 static int fsck_head_link(void)
553 int null_is_error = 0;
555 if (verbose)
556 fprintf(stderr, "Checking HEAD link\n");
558 head_points_at = resolve_ref_unsafe("HEAD", 0, &head_oid, NULL);
559 if (!head_points_at) {
560 errors_found |= ERROR_REFS;
561 return error("Invalid HEAD");
563 if (!strcmp(head_points_at, "HEAD"))
564 /* detached HEAD */
565 null_is_error = 1;
566 else if (!starts_with(head_points_at, "refs/heads/")) {
567 errors_found |= ERROR_REFS;
568 return error("HEAD points to something strange (%s)",
569 head_points_at);
571 if (is_null_oid(&head_oid)) {
572 if (null_is_error) {
573 errors_found |= ERROR_REFS;
574 return error("HEAD: detached HEAD points at nothing");
576 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
577 head_points_at + 11);
579 return 0;
582 static int fsck_cache_tree(struct cache_tree *it)
584 int i;
585 int err = 0;
587 if (verbose)
588 fprintf(stderr, "Checking cache tree\n");
590 if (0 <= it->entry_count) {
591 struct object *obj = parse_object(&it->oid);
592 if (!obj) {
593 error("%s: invalid sha1 pointer in cache-tree",
594 oid_to_hex(&it->oid));
595 errors_found |= ERROR_REFS;
596 return 1;
598 obj->flags |= USED;
599 if (name_objects)
600 add_decoration(fsck_walk_options.object_names,
601 obj, xstrdup(":"));
602 mark_object_reachable(obj);
603 if (obj->type != OBJ_TREE)
604 err |= objerror(obj, "non-tree in cache-tree");
606 for (i = 0; i < it->subtree_nr; i++)
607 err |= fsck_cache_tree(it->down[i]->cache_tree);
608 return err;
611 static void mark_object_for_connectivity(const struct object_id *oid)
613 struct object *obj = lookup_unknown_object(oid->hash);
614 obj->flags |= HAS_OBJ;
617 static int mark_loose_for_connectivity(const struct object_id *oid,
618 const char *path,
619 void *data)
621 mark_object_for_connectivity(oid);
622 return 0;
625 static int mark_packed_for_connectivity(const struct object_id *oid,
626 struct packed_git *pack,
627 uint32_t pos,
628 void *data)
630 mark_object_for_connectivity(oid);
631 return 0;
634 static char const * const fsck_usage[] = {
635 N_("git fsck [<options>] [<object>...]"),
636 NULL
639 static struct option fsck_opts[] = {
640 OPT__VERBOSE(&verbose, N_("be verbose")),
641 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
642 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
643 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
644 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
645 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
646 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
647 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
648 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
649 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
650 OPT_BOOL(0, "lost-found", &write_lost_and_found,
651 N_("write dangling objects in .git/lost-found")),
652 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
653 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
654 OPT_END(),
657 int cmd_fsck(int argc, const char **argv, const char *prefix)
659 int i;
660 struct alternate_object_database *alt;
662 errors_found = 0;
663 check_replace_refs = 0;
665 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
667 fsck_walk_options.walk = mark_object;
668 fsck_obj_options.walk = mark_used;
669 fsck_obj_options.error_func = fsck_error_func;
670 if (check_strict)
671 fsck_obj_options.strict = 1;
673 if (show_progress == -1)
674 show_progress = isatty(2);
675 if (verbose)
676 show_progress = 0;
678 if (write_lost_and_found) {
679 check_full = 1;
680 include_reflogs = 0;
683 if (name_objects)
684 fsck_walk_options.object_names =
685 xcalloc(1, sizeof(struct decoration));
687 git_config(fsck_config, NULL);
689 fsck_head_link();
690 if (connectivity_only) {
691 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
692 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
693 } else {
694 fsck_object_dir(get_object_directory());
696 prepare_alt_odb();
697 for (alt = alt_odb_list; alt; alt = alt->next)
698 fsck_object_dir(alt->path);
700 if (check_full) {
701 struct packed_git *p;
702 uint32_t total = 0, count = 0;
703 struct progress *progress = NULL;
705 prepare_packed_git();
707 if (show_progress) {
708 for (p = packed_git; p; p = p->next) {
709 if (open_pack_index(p))
710 continue;
711 total += p->num_objects;
714 progress = start_progress(_("Checking objects"), total);
716 for (p = packed_git; p; p = p->next) {
717 /* verify gives error messages itself */
718 if (verify_pack(p, fsck_obj_buffer,
719 progress, count))
720 errors_found |= ERROR_PACK;
721 count += p->num_objects;
723 stop_progress(&progress);
727 for (i = 0; i < argc; i++) {
728 const char *arg = argv[i];
729 struct object_id oid;
730 if (!get_oid(arg, &oid)) {
731 struct object *obj = lookup_object(oid.hash);
733 if (!obj || !(obj->flags & HAS_OBJ)) {
734 error("%s: object missing", oid_to_hex(&oid));
735 errors_found |= ERROR_OBJECT;
736 continue;
739 obj->flags |= USED;
740 if (name_objects)
741 add_decoration(fsck_walk_options.object_names,
742 obj, xstrdup(arg));
743 mark_object_reachable(obj);
744 continue;
746 error("invalid parameter: expected sha1, got '%s'", arg);
747 errors_found |= ERROR_OBJECT;
751 * If we've not been given any explicit head information, do the
752 * default ones from .git/refs. We also consider the index file
753 * in this case (ie this implies --cache).
755 if (!argc) {
756 get_default_heads();
757 keep_cache_objects = 1;
760 if (keep_cache_objects) {
761 verify_index_checksum = 1;
762 verify_ce_order = 1;
763 read_cache();
764 for (i = 0; i < active_nr; i++) {
765 unsigned int mode;
766 struct blob *blob;
767 struct object *obj;
769 mode = active_cache[i]->ce_mode;
770 if (S_ISGITLINK(mode))
771 continue;
772 blob = lookup_blob(&active_cache[i]->oid);
773 if (!blob)
774 continue;
775 obj = &blob->object;
776 obj->flags |= USED;
777 if (name_objects)
778 add_decoration(fsck_walk_options.object_names,
779 obj,
780 xstrfmt(":%s", active_cache[i]->name));
781 mark_object_reachable(obj);
783 if (active_cache_tree)
784 fsck_cache_tree(active_cache_tree);
787 check_connectivity();
788 return errors_found;