gitweb: use new Git::Repo API, and add optional caching
[git/gitweb-caching.git] / builtin-fsck.c
blobd3f3de9446a9184e9457fe4b743c4e43a9256597
1 #include "builtin.h"
2 #include "cache.h"
3 #include "commit.h"
4 #include "tree.h"
5 #include "blob.h"
6 #include "tag.h"
7 #include "refs.h"
8 #include "pack.h"
9 #include "cache-tree.h"
10 #include "tree-walk.h"
11 #include "fsck.h"
12 #include "parse-options.h"
14 #define REACHABLE 0x0001
15 #define SEEN 0x0002
17 static int show_root;
18 static int show_tags;
19 static int show_unreachable;
20 static int include_reflogs = 1;
21 static int check_full;
22 static int check_strict;
23 static int keep_cache_objects;
24 static unsigned char head_sha1[20];
25 static int errors_found;
26 static int write_lost_and_found;
27 static int verbose;
28 #define ERROR_OBJECT 01
29 #define ERROR_REACHABLE 02
31 #ifdef NO_D_INO_IN_DIRENT
32 #define SORT_DIRENT 0
33 #define DIRENT_SORT_HINT(de) 0
34 #else
35 #define SORT_DIRENT 1
36 #define DIRENT_SORT_HINT(de) ((de)->d_ino)
37 #endif
39 static void objreport(struct object *obj, const char *severity,
40 const char *err, va_list params)
42 fprintf(stderr, "%s in %s %s: ",
43 severity, typename(obj->type), sha1_to_hex(obj->sha1));
44 vfprintf(stderr, err, params);
45 fputs("\n", stderr);
48 static int objerror(struct object *obj, const char *err, ...)
50 va_list params;
51 va_start(params, err);
52 errors_found |= ERROR_OBJECT;
53 objreport(obj, "error", err, params);
54 va_end(params);
55 return -1;
58 static int fsck_error_func(struct object *obj, int type, const char *err, ...)
60 va_list params;
61 va_start(params, err);
62 objreport(obj, (type == FSCK_WARN) ? "warning" : "error", err, params);
63 va_end(params);
64 return (type == FSCK_WARN) ? 0 : 1;
67 static int mark_object(struct object *obj, int type, void *data)
69 struct tree *tree = NULL;
70 struct object *parent = data;
71 int result;
73 if (!obj) {
74 printf("broken link from %7s %s\n",
75 typename(parent->type), sha1_to_hex(parent->sha1));
76 printf("broken link from %7s %s\n",
77 (type == OBJ_ANY ? "unknown" : typename(type)), "unknown");
78 errors_found |= ERROR_REACHABLE;
79 return 1;
82 if (type != OBJ_ANY && obj->type != type)
83 objerror(parent, "wrong object type in link");
85 if (obj->flags & REACHABLE)
86 return 0;
87 obj->flags |= REACHABLE;
88 if (!obj->parsed) {
89 if (parent && !has_sha1_file(obj->sha1)) {
90 printf("broken link from %7s %s\n",
91 typename(parent->type), sha1_to_hex(parent->sha1));
92 printf(" to %7s %s\n",
93 typename(obj->type), sha1_to_hex(obj->sha1));
94 errors_found |= ERROR_REACHABLE;
96 return 1;
99 if (obj->type == OBJ_TREE) {
100 obj->parsed = 0;
101 tree = (struct tree *)obj;
102 if (parse_tree(tree) < 0)
103 return 1; /* error already displayed */
105 result = fsck_walk(obj, mark_object, obj);
106 if (tree) {
107 free(tree->buffer);
108 tree->buffer = NULL;
110 if (result < 0)
111 result = 1;
113 return result;
116 static void mark_object_reachable(struct object *obj)
118 mark_object(obj, OBJ_ANY, 0);
121 static int mark_used(struct object *obj, int type, void *data)
123 if (!obj)
124 return 1;
125 obj->used = 1;
126 return 0;
130 * Check a single reachable object
132 static void check_reachable_object(struct object *obj)
135 * We obviously want the object to be parsed,
136 * except if it was in a pack-file and we didn't
137 * do a full fsck
139 if (!obj->parsed) {
140 if (has_sha1_pack(obj->sha1, NULL))
141 return; /* it is in pack - forget about it */
142 printf("missing %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
143 errors_found |= ERROR_REACHABLE;
144 return;
149 * Check a single unreachable object
151 static void check_unreachable_object(struct object *obj)
154 * Missing unreachable object? Ignore it. It's not like
155 * we miss it (since it can't be reached), nor do we want
156 * to complain about it being unreachable (since it does
157 * not exist).
159 if (!obj->parsed)
160 return;
163 * Unreachable object that exists? Show it if asked to,
164 * since this is something that is prunable.
166 if (show_unreachable) {
167 printf("unreachable %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
168 return;
172 * "!used" means that nothing at all points to it, including
173 * other unreachable objects. In other words, it's the "tip"
174 * of some set of unreachable objects, usually a commit that
175 * got dropped.
177 * Such starting points are more interesting than some random
178 * set of unreachable objects, so we show them even if the user
179 * hasn't asked for _all_ unreachable objects. If you have
180 * deleted a branch by mistake, this is a prime candidate to
181 * start looking at, for example.
183 if (!obj->used) {
184 printf("dangling %s %s\n", typename(obj->type),
185 sha1_to_hex(obj->sha1));
186 if (write_lost_and_found) {
187 char *filename = git_path("lost-found/%s/%s",
188 obj->type == OBJ_COMMIT ? "commit" : "other",
189 sha1_to_hex(obj->sha1));
190 FILE *f;
192 if (safe_create_leading_directories(filename)) {
193 error("Could not create lost-found");
194 return;
196 if (!(f = fopen(filename, "w")))
197 die("Could not open %s", filename);
198 if (obj->type == OBJ_BLOB) {
199 enum object_type type;
200 unsigned long size;
201 char *buf = read_sha1_file(obj->sha1,
202 &type, &size);
203 if (buf) {
204 fwrite(buf, size, 1, f);
205 free(buf);
207 } else
208 fprintf(f, "%s\n", sha1_to_hex(obj->sha1));
209 fclose(f);
211 return;
215 * Otherwise? It's there, it's unreachable, and some other unreachable
216 * object points to it. Ignore it - it's not interesting, and we showed
217 * all the interesting cases above.
221 static void check_object(struct object *obj)
223 if (verbose)
224 fprintf(stderr, "Checking %s\n", sha1_to_hex(obj->sha1));
226 if (obj->flags & REACHABLE)
227 check_reachable_object(obj);
228 else
229 check_unreachable_object(obj);
232 static void check_connectivity(void)
234 int i, max;
236 /* Look up all the requirements, warn about missing objects.. */
237 max = get_max_object_index();
238 if (verbose)
239 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
241 for (i = 0; i < max; i++) {
242 struct object *obj = get_indexed_object(i);
244 if (obj)
245 check_object(obj);
249 static int fsck_sha1(const unsigned char *sha1)
251 struct object *obj = parse_object(sha1);
252 if (!obj) {
253 errors_found |= ERROR_OBJECT;
254 return error("%s: object corrupt or missing",
255 sha1_to_hex(sha1));
257 if (obj->flags & SEEN)
258 return 0;
259 obj->flags |= SEEN;
261 if (verbose)
262 fprintf(stderr, "Checking %s %s\n",
263 typename(obj->type), sha1_to_hex(obj->sha1));
265 if (fsck_walk(obj, mark_used, 0))
266 objerror(obj, "broken links");
267 if (fsck_object(obj, check_strict, fsck_error_func))
268 return -1;
270 if (obj->type == OBJ_TREE) {
271 struct tree *item = (struct tree *) obj;
273 free(item->buffer);
274 item->buffer = NULL;
277 if (obj->type == OBJ_COMMIT) {
278 struct commit *commit = (struct commit *) obj;
280 free(commit->buffer);
281 commit->buffer = NULL;
283 if (!commit->parents && show_root)
284 printf("root %s\n", sha1_to_hex(commit->object.sha1));
287 if (obj->type == OBJ_TAG) {
288 struct tag *tag = (struct tag *) obj;
290 if (show_tags && tag->tagged) {
291 printf("tagged %s %s", typename(tag->tagged->type), sha1_to_hex(tag->tagged->sha1));
292 printf(" (%s) in %s\n", tag->tag, sha1_to_hex(tag->object.sha1));
296 return 0;
300 * This is the sorting chunk size: make it reasonably
301 * big so that we can sort well..
303 #define MAX_SHA1_ENTRIES (1024)
305 struct sha1_entry {
306 unsigned long ino;
307 unsigned char sha1[20];
310 static struct {
311 unsigned long nr;
312 struct sha1_entry *entry[MAX_SHA1_ENTRIES];
313 } sha1_list;
315 static int ino_compare(const void *_a, const void *_b)
317 const struct sha1_entry *a = _a, *b = _b;
318 unsigned long ino1 = a->ino, ino2 = b->ino;
319 return ino1 < ino2 ? -1 : ino1 > ino2 ? 1 : 0;
322 static void fsck_sha1_list(void)
324 int i, nr = sha1_list.nr;
326 if (SORT_DIRENT)
327 qsort(sha1_list.entry, nr,
328 sizeof(struct sha1_entry *), ino_compare);
329 for (i = 0; i < nr; i++) {
330 struct sha1_entry *entry = sha1_list.entry[i];
331 unsigned char *sha1 = entry->sha1;
333 sha1_list.entry[i] = NULL;
334 fsck_sha1(sha1);
335 free(entry);
337 sha1_list.nr = 0;
340 static void add_sha1_list(unsigned char *sha1, unsigned long ino)
342 struct sha1_entry *entry = xmalloc(sizeof(*entry));
343 int nr;
345 entry->ino = ino;
346 hashcpy(entry->sha1, sha1);
347 nr = sha1_list.nr;
348 if (nr == MAX_SHA1_ENTRIES) {
349 fsck_sha1_list();
350 nr = 0;
352 sha1_list.entry[nr] = entry;
353 sha1_list.nr = ++nr;
356 static void fsck_dir(int i, char *path)
358 DIR *dir = opendir(path);
359 struct dirent *de;
361 if (!dir)
362 return;
364 if (verbose)
365 fprintf(stderr, "Checking directory %s\n", path);
367 while ((de = readdir(dir)) != NULL) {
368 char name[100];
369 unsigned char sha1[20];
370 int len = strlen(de->d_name);
372 switch (len) {
373 case 2:
374 if (de->d_name[1] != '.')
375 break;
376 case 1:
377 if (de->d_name[0] != '.')
378 break;
379 continue;
380 case 38:
381 sprintf(name, "%02x", i);
382 memcpy(name+2, de->d_name, len+1);
383 if (get_sha1_hex(name, sha1) < 0)
384 break;
385 add_sha1_list(sha1, DIRENT_SORT_HINT(de));
386 continue;
388 if (!prefixcmp(de->d_name, "tmp_obj_"))
389 continue;
390 fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
392 closedir(dir);
395 static int default_refs;
397 static int fsck_handle_reflog_ent(unsigned char *osha1, unsigned char *nsha1,
398 const char *email, unsigned long timestamp, int tz,
399 const char *message, void *cb_data)
401 struct object *obj;
403 if (verbose)
404 fprintf(stderr, "Checking reflog %s->%s\n",
405 sha1_to_hex(osha1), sha1_to_hex(nsha1));
407 if (!is_null_sha1(osha1)) {
408 obj = lookup_object(osha1);
409 if (obj) {
410 obj->used = 1;
411 mark_object_reachable(obj);
414 obj = lookup_object(nsha1);
415 if (obj) {
416 obj->used = 1;
417 mark_object_reachable(obj);
419 return 0;
422 static int fsck_handle_reflog(const char *logname, const unsigned char *sha1, int flag, void *cb_data)
424 for_each_reflog_ent(logname, fsck_handle_reflog_ent, NULL);
425 return 0;
428 static int is_branch(const char *refname)
430 return !strcmp(refname, "HEAD") || !prefixcmp(refname, "refs/heads/");
433 static int fsck_handle_ref(const char *refname, const unsigned char *sha1, int flag, void *cb_data)
435 struct object *obj;
437 obj = parse_object(sha1);
438 if (!obj) {
439 error("%s: invalid sha1 pointer %s", refname, sha1_to_hex(sha1));
440 /* We'll continue with the rest despite the error.. */
441 return 0;
443 if (obj->type != OBJ_COMMIT && is_branch(refname))
444 error("%s: not a commit", refname);
445 default_refs++;
446 obj->used = 1;
447 mark_object_reachable(obj);
449 return 0;
452 static void get_default_heads(void)
454 for_each_ref(fsck_handle_ref, NULL);
455 if (include_reflogs)
456 for_each_reflog(fsck_handle_reflog, NULL);
459 * Not having any default heads isn't really fatal, but
460 * it does mean that "--unreachable" no longer makes any
461 * sense (since in this case everything will obviously
462 * be unreachable by definition.
464 * Showing dangling objects is valid, though (as those
465 * dangling objects are likely lost heads).
467 * So we just print a warning about it, and clear the
468 * "show_unreachable" flag.
470 if (!default_refs) {
471 fprintf(stderr, "notice: No default references\n");
472 show_unreachable = 0;
476 static void fsck_object_dir(const char *path)
478 int i;
480 if (verbose)
481 fprintf(stderr, "Checking object directory\n");
483 for (i = 0; i < 256; i++) {
484 static char dir[4096];
485 sprintf(dir, "%s/%02x", path, i);
486 fsck_dir(i, dir);
488 fsck_sha1_list();
491 static int fsck_head_link(void)
493 unsigned char sha1[20];
494 int flag;
495 int null_is_error = 0;
496 const char *head_points_at = resolve_ref("HEAD", sha1, 0, &flag);
498 if (verbose)
499 fprintf(stderr, "Checking HEAD link\n");
501 if (!head_points_at)
502 return error("Invalid HEAD");
503 if (!strcmp(head_points_at, "HEAD"))
504 /* detached HEAD */
505 null_is_error = 1;
506 else if (prefixcmp(head_points_at, "refs/heads/"))
507 return error("HEAD points to something strange (%s)",
508 head_points_at);
509 if (is_null_sha1(sha1)) {
510 if (null_is_error)
511 return error("HEAD: detached HEAD points at nothing");
512 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
513 head_points_at + 11);
515 return 0;
518 static int fsck_cache_tree(struct cache_tree *it)
520 int i;
521 int err = 0;
523 if (verbose)
524 fprintf(stderr, "Checking cache tree\n");
526 if (0 <= it->entry_count) {
527 struct object *obj = parse_object(it->sha1);
528 if (!obj) {
529 error("%s: invalid sha1 pointer in cache-tree",
530 sha1_to_hex(it->sha1));
531 return 1;
533 mark_object_reachable(obj);
534 obj->used = 1;
535 if (obj->type != OBJ_TREE)
536 err |= objerror(obj, "non-tree in cache-tree");
538 for (i = 0; i < it->subtree_nr; i++)
539 err |= fsck_cache_tree(it->down[i]->cache_tree);
540 return err;
543 static char const * const fsck_usage[] = {
544 "git fsck [options] [<object>...]",
545 NULL
548 static struct option fsck_opts[] = {
549 OPT__VERBOSE(&verbose),
550 OPT_BOOLEAN(0, "unreachable", &show_unreachable, "show unreachable objects"),
551 OPT_BOOLEAN(0, "tags", &show_tags, "report tags"),
552 OPT_BOOLEAN(0, "root", &show_root, "report root nodes"),
553 OPT_BOOLEAN(0, "cache", &keep_cache_objects, "make index objects head nodes"),
554 OPT_BOOLEAN(0, "reflogs", &include_reflogs, "make reflogs head nodes (default)"),
555 OPT_BOOLEAN(0, "full", &check_full, "also consider alternate objects"),
556 OPT_BOOLEAN(0, "strict", &check_strict, "enable more strict checking"),
557 OPT_BOOLEAN(0, "lost-found", &write_lost_and_found,
558 "write dangling objects in .git/lost-found"),
559 OPT_END(),
562 int cmd_fsck(int argc, const char **argv, const char *prefix)
564 int i, heads;
566 errors_found = 0;
568 argc = parse_options(argc, argv, fsck_opts, fsck_usage, 0);
569 if (write_lost_and_found) {
570 check_full = 1;
571 include_reflogs = 0;
574 fsck_head_link();
575 fsck_object_dir(get_object_directory());
576 if (check_full) {
577 struct alternate_object_database *alt;
578 struct packed_git *p;
579 prepare_alt_odb();
580 for (alt = alt_odb_list; alt; alt = alt->next) {
581 char namebuf[PATH_MAX];
582 int namelen = alt->name - alt->base;
583 memcpy(namebuf, alt->base, namelen);
584 namebuf[namelen - 1] = 0;
585 fsck_object_dir(namebuf);
587 prepare_packed_git();
588 for (p = packed_git; p; p = p->next)
589 /* verify gives error messages itself */
590 verify_pack(p);
592 for (p = packed_git; p; p = p->next) {
593 uint32_t j, num;
594 if (open_pack_index(p))
595 continue;
596 num = p->num_objects;
597 for (j = 0; j < num; j++)
598 fsck_sha1(nth_packed_object_sha1(p, j));
602 heads = 0;
603 for (i = 1; i < argc; i++) {
604 const char *arg = argv[i];
605 if (!get_sha1(arg, head_sha1)) {
606 struct object *obj = lookup_object(head_sha1);
608 /* Error is printed by lookup_object(). */
609 if (!obj)
610 continue;
612 obj->used = 1;
613 mark_object_reachable(obj);
614 heads++;
615 continue;
617 error("invalid parameter: expected sha1, got '%s'", arg);
621 * If we've not been given any explicit head information, do the
622 * default ones from .git/refs. We also consider the index file
623 * in this case (ie this implies --cache).
625 if (!heads) {
626 get_default_heads();
627 keep_cache_objects = 1;
630 if (keep_cache_objects) {
631 read_cache();
632 for (i = 0; i < active_nr; i++) {
633 unsigned int mode;
634 struct blob *blob;
635 struct object *obj;
637 mode = active_cache[i]->ce_mode;
638 if (S_ISGITLINK(mode))
639 continue;
640 blob = lookup_blob(active_cache[i]->sha1);
641 if (!blob)
642 continue;
643 obj = &blob->object;
644 obj->used = 1;
645 mark_object_reachable(obj);
647 if (active_cache_tree)
648 fsck_cache_tree(active_cache_tree);
651 check_connectivity();
652 return errors_found;