sparse-checkout: update working directory in-process for 'init'
[git] / fsck.c
1 #include "cache.h"
2 #include "object-store.h"
3 #include "repository.h"
4 #include "object.h"
5 #include "blob.h"
6 #include "tree.h"
7 #include "tree-walk.h"
8 #include "commit.h"
9 #include "tag.h"
10 #include "fsck.h"
11 #include "refs.h"
12 #include "utf8.h"
13 #include "decorate.h"
14 #include "oidset.h"
15 #include "packfile.h"
16 #include "submodule-config.h"
17 #include "config.h"
18 #include "help.h"
19
20 static struct oidset gitmodules_found = OIDSET_INIT;
21 static struct oidset gitmodules_done = OIDSET_INIT;
22
23 #define FSCK_FATAL -1
24 #define FSCK_INFO -2
25
26 #define FOREACH_MSG_ID(FUNC) \
27         /* fatal errors */ \
28         FUNC(NUL_IN_HEADER, FATAL) \
29         FUNC(UNTERMINATED_HEADER, FATAL) \
30         /* errors */ \
31         FUNC(BAD_DATE, ERROR) \
32         FUNC(BAD_DATE_OVERFLOW, ERROR) \
33         FUNC(BAD_EMAIL, ERROR) \
34         FUNC(BAD_NAME, ERROR) \
35         FUNC(BAD_OBJECT_SHA1, ERROR) \
36         FUNC(BAD_PARENT_SHA1, ERROR) \
37         FUNC(BAD_TAG_OBJECT, ERROR) \
38         FUNC(BAD_TIMEZONE, ERROR) \
39         FUNC(BAD_TREE, ERROR) \
40         FUNC(BAD_TREE_SHA1, ERROR) \
41         FUNC(BAD_TYPE, ERROR) \
42         FUNC(DUPLICATE_ENTRIES, ERROR) \
43         FUNC(MISSING_AUTHOR, ERROR) \
44         FUNC(MISSING_COMMITTER, ERROR) \
45         FUNC(MISSING_EMAIL, ERROR) \
46         FUNC(MISSING_GRAFT, ERROR) \
47         FUNC(MISSING_NAME_BEFORE_EMAIL, ERROR) \
48         FUNC(MISSING_OBJECT, ERROR) \
49         FUNC(MISSING_PARENT, ERROR) \
50         FUNC(MISSING_SPACE_BEFORE_DATE, ERROR) \
51         FUNC(MISSING_SPACE_BEFORE_EMAIL, ERROR) \
52         FUNC(MISSING_TAG, ERROR) \
53         FUNC(MISSING_TAG_ENTRY, ERROR) \
54         FUNC(MISSING_TAG_OBJECT, ERROR) \
55         FUNC(MISSING_TREE, ERROR) \
56         FUNC(MISSING_TREE_OBJECT, ERROR) \
57         FUNC(MISSING_TYPE, ERROR) \
58         FUNC(MISSING_TYPE_ENTRY, ERROR) \
59         FUNC(MULTIPLE_AUTHORS, ERROR) \
60         FUNC(TAG_OBJECT_NOT_TAG, ERROR) \
61         FUNC(TREE_NOT_SORTED, ERROR) \
62         FUNC(UNKNOWN_TYPE, ERROR) \
63         FUNC(ZERO_PADDED_DATE, ERROR) \
64         FUNC(GITMODULES_MISSING, ERROR) \
65         FUNC(GITMODULES_BLOB, ERROR) \
66         FUNC(GITMODULES_LARGE, ERROR) \
67         FUNC(GITMODULES_NAME, ERROR) \
68         FUNC(GITMODULES_SYMLINK, ERROR) \
69         FUNC(GITMODULES_URL, ERROR) \
70         FUNC(GITMODULES_PATH, ERROR) \
71         /* warnings */ \
72         FUNC(BAD_FILEMODE, WARN) \
73         FUNC(EMPTY_NAME, WARN) \
74         FUNC(FULL_PATHNAME, WARN) \
75         FUNC(HAS_DOT, WARN) \
76         FUNC(HAS_DOTDOT, WARN) \
77         FUNC(HAS_DOTGIT, WARN) \
78         FUNC(NULL_SHA1, WARN) \
79         FUNC(ZERO_PADDED_FILEMODE, WARN) \
80         FUNC(NUL_IN_COMMIT, WARN) \
81         /* infos (reported as warnings, but ignored by default) */ \
82         FUNC(GITMODULES_PARSE, INFO) \
83         FUNC(BAD_TAG_NAME, INFO) \
84         FUNC(MISSING_TAGGER_ENTRY, INFO)
85
86 #define MSG_ID(id, msg_type) FSCK_MSG_##id,
87 enum fsck_msg_id {
88         FOREACH_MSG_ID(MSG_ID)
89         FSCK_MSG_MAX
90 };
91 #undef MSG_ID
92
93 #define STR(x) #x
94 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
95 static struct {
96         const char *id_string;
97         const char *downcased;
98         const char *camelcased;
99         int msg_type;
100 } msg_id_info[FSCK_MSG_MAX + 1] = {
101         FOREACH_MSG_ID(MSG_ID)
102         { NULL, NULL, NULL, -1 }
103 };
104 #undef MSG_ID
105
106 static void prepare_msg_ids(void)
107 {
108         int i;
109
110         if (msg_id_info[0].downcased)
111                 return;
112
113         /* convert id_string to lower case, without underscores. */
114         for (i = 0; i < FSCK_MSG_MAX; i++) {
115                 const char *p = msg_id_info[i].id_string;
116                 int len = strlen(p);
117                 char *q = xmalloc(len);
118
119                 msg_id_info[i].downcased = q;
120                 while (*p)
121                         if (*p == '_')
122                                 p++;
123                         else
124                                 *(q)++ = tolower(*(p)++);
125                 *q = '\0';
126
127                 p = msg_id_info[i].id_string;
128                 q = xmalloc(len);
129                 msg_id_info[i].camelcased = q;
130                 while (*p) {
131                         if (*p == '_') {
132                                 p++;
133                                 if (*p)
134                                         *q++ = *p++;
135                         } else {
136                                 *q++ = tolower(*p++);
137                         }
138                 }
139                 *q = '\0';
140         }
141 }
142
143 static int parse_msg_id(const char *text)
144 {
145         int i;
146
147         prepare_msg_ids();
148
149         for (i = 0; i < FSCK_MSG_MAX; i++)
150                 if (!strcmp(text, msg_id_info[i].downcased))
151                         return i;
152
153         return -1;
154 }
155
156 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
157 {
158         int i;
159
160         prepare_msg_ids();
161
162         for (i = 0; i < FSCK_MSG_MAX; i++)
163                 list_config_item(list, prefix, msg_id_info[i].camelcased);
164 }
165
166 static int fsck_msg_type(enum fsck_msg_id msg_id,
167         struct fsck_options *options)
168 {
169         int msg_type;
170
171         assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
172
173         if (options->msg_type)
174                 msg_type = options->msg_type[msg_id];
175         else {
176                 msg_type = msg_id_info[msg_id].msg_type;
177                 if (options->strict && msg_type == FSCK_WARN)
178                         msg_type = FSCK_ERROR;
179         }
180
181         return msg_type;
182 }
183
184 static int parse_msg_type(const char *str)
185 {
186         if (!strcmp(str, "error"))
187                 return FSCK_ERROR;
188         else if (!strcmp(str, "warn"))
189                 return FSCK_WARN;
190         else if (!strcmp(str, "ignore"))
191                 return FSCK_IGNORE;
192         else
193                 die("Unknown fsck message type: '%s'", str);
194 }
195
196 int is_valid_msg_type(const char *msg_id, const char *msg_type)
197 {
198         if (parse_msg_id(msg_id) < 0)
199                 return 0;
200         parse_msg_type(msg_type);
201         return 1;
202 }
203
204 void fsck_set_msg_type(struct fsck_options *options,
205                 const char *msg_id, const char *msg_type)
206 {
207         int id = parse_msg_id(msg_id), type;
208
209         if (id < 0)
210                 die("Unhandled message id: %s", msg_id);
211         type = parse_msg_type(msg_type);
212
213         if (type != FSCK_ERROR && msg_id_info[id].msg_type == FSCK_FATAL)
214                 die("Cannot demote %s to %s", msg_id, msg_type);
215
216         if (!options->msg_type) {
217                 int i;
218                 int *msg_type;
219                 ALLOC_ARRAY(msg_type, FSCK_MSG_MAX);
220                 for (i = 0; i < FSCK_MSG_MAX; i++)
221                         msg_type[i] = fsck_msg_type(i, options);
222                 options->msg_type = msg_type;
223         }
224
225         options->msg_type[id] = type;
226 }
227
228 void fsck_set_msg_types(struct fsck_options *options, const char *values)
229 {
230         char *buf = xstrdup(values), *to_free = buf;
231         int done = 0;
232
233         while (!done) {
234                 int len = strcspn(buf, " ,|"), equal;
235
236                 done = !buf[len];
237                 if (!len) {
238                         buf++;
239                         continue;
240                 }
241                 buf[len] = '\0';
242
243                 for (equal = 0;
244                      equal < len && buf[equal] != '=' && buf[equal] != ':';
245                      equal++)
246                         buf[equal] = tolower(buf[equal]);
247                 buf[equal] = '\0';
248
249                 if (!strcmp(buf, "skiplist")) {
250                         if (equal == len)
251                                 die("skiplist requires a path");
252                         oidset_parse_file(&options->skiplist, buf + equal + 1);
253                         buf += len + 1;
254                         continue;
255                 }
256
257                 if (equal == len)
258                         die("Missing '=': '%s'", buf);
259
260                 fsck_set_msg_type(options, buf, buf + equal + 1);
261                 buf += len + 1;
262         }
263         free(to_free);
264 }
265
266 static void append_msg_id(struct strbuf *sb, const char *msg_id)
267 {
268         for (;;) {
269                 char c = *(msg_id)++;
270
271                 if (!c)
272                         break;
273                 if (c != '_')
274                         strbuf_addch(sb, tolower(c));
275                 else {
276                         assert(*msg_id);
277                         strbuf_addch(sb, *(msg_id)++);
278                 }
279         }
280
281         strbuf_addstr(sb, ": ");
282 }
283
284 static int object_on_skiplist(struct fsck_options *opts, struct object *obj)
285 {
286         return opts && obj && oidset_contains(&opts->skiplist, &obj->oid);
287 }
288
289 __attribute__((format (printf, 4, 5)))
290 static int report(struct fsck_options *options, struct object *object,
291         enum fsck_msg_id id, const char *fmt, ...)
292 {
293         va_list ap;
294         struct strbuf sb = STRBUF_INIT;
295         int msg_type = fsck_msg_type(id, options), result;
296
297         if (msg_type == FSCK_IGNORE)
298                 return 0;
299
300         if (object_on_skiplist(options, object))
301                 return 0;
302
303         if (msg_type == FSCK_FATAL)
304                 msg_type = FSCK_ERROR;
305         else if (msg_type == FSCK_INFO)
306                 msg_type = FSCK_WARN;
307
308         append_msg_id(&sb, msg_id_info[id].id_string);
309
310         va_start(ap, fmt);
311         strbuf_vaddf(&sb, fmt, ap);
312         result = options->error_func(options, object, msg_type, sb.buf);
313         strbuf_release(&sb);
314         va_end(ap);
315
316         return result;
317 }
318
319 static char *get_object_name(struct fsck_options *options, struct object *obj)
320 {
321         if (!options->object_names)
322                 return NULL;
323         return lookup_decoration(options->object_names, obj);
324 }
325
326 static void put_object_name(struct fsck_options *options, struct object *obj,
327         const char *fmt, ...)
328 {
329         va_list ap;
330         struct strbuf buf = STRBUF_INIT;
331         char *existing;
332
333         if (!options->object_names)
334                 return;
335         existing = lookup_decoration(options->object_names, obj);
336         if (existing)
337                 return;
338         va_start(ap, fmt);
339         strbuf_vaddf(&buf, fmt, ap);
340         add_decoration(options->object_names, obj, strbuf_detach(&buf, NULL));
341         va_end(ap);
342 }
343
344 static const char *describe_object(struct fsck_options *o, struct object *obj)
345 {
346         static struct strbuf buf = STRBUF_INIT;
347         char *name;
348
349         strbuf_reset(&buf);
350         strbuf_addstr(&buf, oid_to_hex(&obj->oid));
351         if (o->object_names && (name = lookup_decoration(o->object_names, obj)))
352                 strbuf_addf(&buf, " (%s)", name);
353
354         return buf.buf;
355 }
356
357 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
358 {
359         struct tree_desc desc;
360         struct name_entry entry;
361         int res = 0;
362         const char *name;
363
364         if (parse_tree(tree))
365                 return -1;
366
367         name = get_object_name(options, &tree->object);
368         if (init_tree_desc_gently(&desc, tree->buffer, tree->size))
369                 return -1;
370         while (tree_entry_gently(&desc, &entry)) {
371                 struct object *obj;
372                 int result;
373
374                 if (S_ISGITLINK(entry.mode))
375                         continue;
376
377                 if (S_ISDIR(entry.mode)) {
378                         obj = (struct object *)lookup_tree(the_repository, &entry.oid);
379                         if (name && obj)
380                                 put_object_name(options, obj, "%s%s/", name,
381                                         entry.path);
382                         result = options->walk(obj, OBJ_TREE, data, options);
383                 }
384                 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
385                         obj = (struct object *)lookup_blob(the_repository, &entry.oid);
386                         if (name && obj)
387                                 put_object_name(options, obj, "%s%s", name,
388                                         entry.path);
389                         result = options->walk(obj, OBJ_BLOB, data, options);
390                 }
391                 else {
392                         result = error("in tree %s: entry %s has bad mode %.6o",
393                                         describe_object(options, &tree->object), entry.path, entry.mode);
394                 }
395                 if (result < 0)
396                         return result;
397                 if (!res)
398                         res = result;
399         }
400         return res;
401 }
402
403 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
404 {
405         int counter = 0, generation = 0, name_prefix_len = 0;
406         struct commit_list *parents;
407         int res;
408         int result;
409         const char *name;
410
411         if (parse_commit(commit))
412                 return -1;
413
414         name = get_object_name(options, &commit->object);
415         if (name)
416                 put_object_name(options, &get_commit_tree(commit)->object,
417                                 "%s:", name);
418
419         result = options->walk((struct object *)get_commit_tree(commit),
420                                OBJ_TREE, data, options);
421         if (result < 0)
422                 return result;
423         res = result;
424
425         parents = commit->parents;
426         if (name && parents) {
427                 int len = strlen(name), power;
428
429                 if (len && name[len - 1] == '^') {
430                         generation = 1;
431                         name_prefix_len = len - 1;
432                 }
433                 else { /* parse ~<generation> suffix */
434                         for (generation = 0, power = 1;
435                              len && isdigit(name[len - 1]);
436                              power *= 10)
437                                 generation += power * (name[--len] - '0');
438                         if (power > 1 && len && name[len - 1] == '~')
439                                 name_prefix_len = len - 1;
440                 }
441         }
442
443         while (parents) {
444                 if (name) {
445                         struct object *obj = &parents->item->object;
446
447                         if (counter++)
448                                 put_object_name(options, obj, "%s^%d",
449                                         name, counter);
450                         else if (generation > 0)
451                                 put_object_name(options, obj, "%.*s~%d",
452                                         name_prefix_len, name, generation + 1);
453                         else
454                                 put_object_name(options, obj, "%s^", name);
455                 }
456                 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
457                 if (result < 0)
458                         return result;
459                 if (!res)
460                         res = result;
461                 parents = parents->next;
462         }
463         return res;
464 }
465
466 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
467 {
468         char *name = get_object_name(options, &tag->object);
469
470         if (parse_tag(tag))
471                 return -1;
472         if (name)
473                 put_object_name(options, tag->tagged, "%s", name);
474         return options->walk(tag->tagged, OBJ_ANY, data, options);
475 }
476
477 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
478 {
479         if (!obj)
480                 return -1;
481
482         if (obj->type == OBJ_NONE)
483                 parse_object(the_repository, &obj->oid);
484
485         switch (obj->type) {
486         case OBJ_BLOB:
487                 return 0;
488         case OBJ_TREE:
489                 return fsck_walk_tree((struct tree *)obj, data, options);
490         case OBJ_COMMIT:
491                 return fsck_walk_commit((struct commit *)obj, data, options);
492         case OBJ_TAG:
493                 return fsck_walk_tag((struct tag *)obj, data, options);
494         default:
495                 error("Unknown object type for %s", describe_object(options, obj));
496                 return -1;
497         }
498 }
499
500 /*
501  * The entries in a tree are ordered in the _path_ order,
502  * which means that a directory entry is ordered by adding
503  * a slash to the end of it.
504  *
505  * So a directory called "a" is ordered _after_ a file
506  * called "a.c", because "a/" sorts after "a.c".
507  */
508 #define TREE_UNORDERED (-1)
509 #define TREE_HAS_DUPS  (-2)
510
511 static int verify_ordered(unsigned mode1, const char *name1, unsigned mode2, const char *name2)
512 {
513         int len1 = strlen(name1);
514         int len2 = strlen(name2);
515         int len = len1 < len2 ? len1 : len2;
516         unsigned char c1, c2;
517         int cmp;
518
519         cmp = memcmp(name1, name2, len);
520         if (cmp < 0)
521                 return 0;
522         if (cmp > 0)
523                 return TREE_UNORDERED;
524
525         /*
526          * Ok, the first <len> characters are the same.
527          * Now we need to order the next one, but turn
528          * a '\0' into a '/' for a directory entry.
529          */
530         c1 = name1[len];
531         c2 = name2[len];
532         if (!c1 && !c2)
533                 /*
534                  * git-write-tree used to write out a nonsense tree that has
535                  * entries with the same name, one blob and one tree.  Make
536                  * sure we do not have duplicate entries.
537                  */
538                 return TREE_HAS_DUPS;
539         if (!c1 && S_ISDIR(mode1))
540                 c1 = '/';
541         if (!c2 && S_ISDIR(mode2))
542                 c2 = '/';
543         return c1 < c2 ? 0 : TREE_UNORDERED;
544 }
545
546 static int fsck_tree(struct tree *item, struct fsck_options *options)
547 {
548         int retval = 0;
549         int has_null_sha1 = 0;
550         int has_full_path = 0;
551         int has_empty_name = 0;
552         int has_dot = 0;
553         int has_dotdot = 0;
554         int has_dotgit = 0;
555         int has_zero_pad = 0;
556         int has_bad_modes = 0;
557         int has_dup_entries = 0;
558         int not_properly_sorted = 0;
559         struct tree_desc desc;
560         unsigned o_mode;
561         const char *o_name;
562
563         if (init_tree_desc_gently(&desc, item->buffer, item->size)) {
564                 retval += report(options, &item->object, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
565                 return retval;
566         }
567
568         o_mode = 0;
569         o_name = NULL;
570
571         while (desc.size) {
572                 unsigned short mode;
573                 const char *name;
574                 const struct object_id *oid;
575
576                 oid = tree_entry_extract(&desc, &name, &mode);
577
578                 has_null_sha1 |= is_null_oid(oid);
579                 has_full_path |= !!strchr(name, '/');
580                 has_empty_name |= !*name;
581                 has_dot |= !strcmp(name, ".");
582                 has_dotdot |= !strcmp(name, "..");
583                 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
584                 has_zero_pad |= *(char *)desc.buffer == '0';
585
586                 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
587                         if (!S_ISLNK(mode))
588                                 oidset_insert(&gitmodules_found, oid);
589                         else
590                                 retval += report(options, &item->object,
591                                                  FSCK_MSG_GITMODULES_SYMLINK,
592                                                  ".gitmodules is a symbolic link");
593                 }
594
595                 if (update_tree_entry_gently(&desc)) {
596                         retval += report(options, &item->object, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
597                         break;
598                 }
599
600                 switch (mode) {
601                 /*
602                  * Standard modes..
603                  */
604                 case S_IFREG | 0755:
605                 case S_IFREG | 0644:
606                 case S_IFLNK:
607                 case S_IFDIR:
608                 case S_IFGITLINK:
609                         break;
610                 /*
611                  * This is nonstandard, but we had a few of these
612                  * early on when we honored the full set of mode
613                  * bits..
614                  */
615                 case S_IFREG | 0664:
616                         if (!options->strict)
617                                 break;
618                         /* fallthrough */
619                 default:
620                         has_bad_modes = 1;
621                 }
622
623                 if (o_name) {
624                         switch (verify_ordered(o_mode, o_name, mode, name)) {
625                         case TREE_UNORDERED:
626                                 not_properly_sorted = 1;
627                                 break;
628                         case TREE_HAS_DUPS:
629                                 has_dup_entries = 1;
630                                 break;
631                         default:
632                                 break;
633                         }
634                 }
635
636                 o_mode = mode;
637                 o_name = name;
638         }
639
640         if (has_null_sha1)
641                 retval += report(options, &item->object, FSCK_MSG_NULL_SHA1, "contains entries pointing to null sha1");
642         if (has_full_path)
643                 retval += report(options, &item->object, FSCK_MSG_FULL_PATHNAME, "contains full pathnames");
644         if (has_empty_name)
645                 retval += report(options, &item->object, FSCK_MSG_EMPTY_NAME, "contains empty pathname");
646         if (has_dot)
647                 retval += report(options, &item->object, FSCK_MSG_HAS_DOT, "contains '.'");
648         if (has_dotdot)
649                 retval += report(options, &item->object, FSCK_MSG_HAS_DOTDOT, "contains '..'");
650         if (has_dotgit)
651                 retval += report(options, &item->object, FSCK_MSG_HAS_DOTGIT, "contains '.git'");
652         if (has_zero_pad)
653                 retval += report(options, &item->object, FSCK_MSG_ZERO_PADDED_FILEMODE, "contains zero-padded file modes");
654         if (has_bad_modes)
655                 retval += report(options, &item->object, FSCK_MSG_BAD_FILEMODE, "contains bad file modes");
656         if (has_dup_entries)
657                 retval += report(options, &item->object, FSCK_MSG_DUPLICATE_ENTRIES, "contains duplicate file entries");
658         if (not_properly_sorted)
659                 retval += report(options, &item->object, FSCK_MSG_TREE_NOT_SORTED, "not properly sorted");
660         return retval;
661 }
662
663 static int verify_headers(const void *data, unsigned long size,
664                           struct object *obj, struct fsck_options *options)
665 {
666         const char *buffer = (const char *)data;
667         unsigned long i;
668
669         for (i = 0; i < size; i++) {
670                 switch (buffer[i]) {
671                 case '\0':
672                         return report(options, obj,
673                                 FSCK_MSG_NUL_IN_HEADER,
674                                 "unterminated header: NUL at offset %ld", i);
675                 case '\n':
676                         if (i + 1 < size && buffer[i + 1] == '\n')
677                                 return 0;
678                 }
679         }
680
681         /*
682          * We did not find double-LF that separates the header
683          * and the body.  Not having a body is not a crime but
684          * we do want to see the terminating LF for the last header
685          * line.
686          */
687         if (size && buffer[size - 1] == '\n')
688                 return 0;
689
690         return report(options, obj,
691                 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
692 }
693
694 static int fsck_ident(const char **ident, struct object *obj, struct fsck_options *options)
695 {
696         const char *p = *ident;
697         char *end;
698
699         *ident = strchrnul(*ident, '\n');
700         if (**ident == '\n')
701                 (*ident)++;
702
703         if (*p == '<')
704                 return report(options, obj, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
705         p += strcspn(p, "<>\n");
706         if (*p == '>')
707                 return report(options, obj, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
708         if (*p != '<')
709                 return report(options, obj, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
710         if (p[-1] != ' ')
711                 return report(options, obj, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
712         p++;
713         p += strcspn(p, "<>\n");
714         if (*p != '>')
715                 return report(options, obj, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
716         p++;
717         if (*p != ' ')
718                 return report(options, obj, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
719         p++;
720         if (*p == '0' && p[1] != ' ')
721                 return report(options, obj, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
722         if (date_overflows(parse_timestamp(p, &end, 10)))
723                 return report(options, obj, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
724         if ((end == p || *end != ' '))
725                 return report(options, obj, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
726         p = end + 1;
727         if ((*p != '+' && *p != '-') ||
728             !isdigit(p[1]) ||
729             !isdigit(p[2]) ||
730             !isdigit(p[3]) ||
731             !isdigit(p[4]) ||
732             (p[5] != '\n'))
733                 return report(options, obj, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
734         p += 6;
735         return 0;
736 }
737
738 static int fsck_commit_buffer(struct commit *commit, const char *buffer,
739         unsigned long size, struct fsck_options *options)
740 {
741         struct object_id tree_oid, oid;
742         struct commit_graft *graft;
743         unsigned parent_count, parent_line_count = 0, author_count;
744         int err;
745         const char *buffer_begin = buffer;
746         const char *p;
747
748         if (verify_headers(buffer, size, &commit->object, options))
749                 return -1;
750
751         if (!skip_prefix(buffer, "tree ", &buffer))
752                 return report(options, &commit->object, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
753         if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
754                 err = report(options, &commit->object, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
755                 if (err)
756                         return err;
757         }
758         buffer = p + 1;
759         while (skip_prefix(buffer, "parent ", &buffer)) {
760                 if (parse_oid_hex(buffer, &oid, &p) || *p != '\n') {
761                         err = report(options, &commit->object, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
762                         if (err)
763                                 return err;
764                 }
765                 buffer = p + 1;
766                 parent_line_count++;
767         }
768         graft = lookup_commit_graft(the_repository, &commit->object.oid);
769         parent_count = commit_list_count(commit->parents);
770         if (graft) {
771                 if (graft->nr_parent == -1 && !parent_count)
772                         ; /* shallow commit */
773                 else if (graft->nr_parent != parent_count) {
774                         err = report(options, &commit->object, FSCK_MSG_MISSING_GRAFT, "graft objects missing");
775                         if (err)
776                                 return err;
777                 }
778         } else {
779                 if (parent_count != parent_line_count) {
780                         err = report(options, &commit->object, FSCK_MSG_MISSING_PARENT, "parent objects missing");
781                         if (err)
782                                 return err;
783                 }
784         }
785         author_count = 0;
786         while (skip_prefix(buffer, "author ", &buffer)) {
787                 author_count++;
788                 err = fsck_ident(&buffer, &commit->object, options);
789                 if (err)
790                         return err;
791         }
792         if (author_count < 1)
793                 err = report(options, &commit->object, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
794         else if (author_count > 1)
795                 err = report(options, &commit->object, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
796         if (err)
797                 return err;
798         if (!skip_prefix(buffer, "committer ", &buffer))
799                 return report(options, &commit->object, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
800         err = fsck_ident(&buffer, &commit->object, options);
801         if (err)
802                 return err;
803         if (!get_commit_tree(commit)) {
804                 err = report(options, &commit->object, FSCK_MSG_BAD_TREE, "could not load commit's tree %s", oid_to_hex(&tree_oid));
805                 if (err)
806                         return err;
807         }
808         if (memchr(buffer_begin, '\0', size)) {
809                 err = report(options, &commit->object, FSCK_MSG_NUL_IN_COMMIT,
810                              "NUL byte in the commit object body");
811                 if (err)
812                         return err;
813         }
814         return 0;
815 }
816
817 static int fsck_commit(struct commit *commit, const char *data,
818         unsigned long size, struct fsck_options *options)
819 {
820         const char *buffer = data ?  data : get_commit_buffer(commit, &size);
821         int ret = fsck_commit_buffer(commit, buffer, size, options);
822         if (!data)
823                 unuse_commit_buffer(commit, buffer);
824         return ret;
825 }
826
827 static int fsck_tag_buffer(struct tag *tag, const char *data,
828         unsigned long size, struct fsck_options *options)
829 {
830         struct object_id oid;
831         int ret = 0;
832         const char *buffer;
833         char *to_free = NULL, *eol;
834         struct strbuf sb = STRBUF_INIT;
835         const char *p;
836
837         if (data)
838                 buffer = data;
839         else {
840                 enum object_type type;
841
842                 buffer = to_free =
843                         read_object_file(&tag->object.oid, &type, &size);
844                 if (!buffer)
845                         return report(options, &tag->object,
846                                 FSCK_MSG_MISSING_TAG_OBJECT,
847                                 "cannot read tag object");
848
849                 if (type != OBJ_TAG) {
850                         ret = report(options, &tag->object,
851                                 FSCK_MSG_TAG_OBJECT_NOT_TAG,
852                                 "expected tag got %s",
853                             type_name(type));
854                         goto done;
855                 }
856         }
857
858         ret = verify_headers(buffer, size, &tag->object, options);
859         if (ret)
860                 goto done;
861
862         if (!skip_prefix(buffer, "object ", &buffer)) {
863                 ret = report(options, &tag->object, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
864                 goto done;
865         }
866         if (parse_oid_hex(buffer, &oid, &p) || *p != '\n') {
867                 ret = report(options, &tag->object, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
868                 if (ret)
869                         goto done;
870         }
871         buffer = p + 1;
872
873         if (!skip_prefix(buffer, "type ", &buffer)) {
874                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
875                 goto done;
876         }
877         eol = strchr(buffer, '\n');
878         if (!eol) {
879                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
880                 goto done;
881         }
882         if (type_from_string_gently(buffer, eol - buffer, 1) < 0)
883                 ret = report(options, &tag->object, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
884         if (ret)
885                 goto done;
886         buffer = eol + 1;
887
888         if (!skip_prefix(buffer, "tag ", &buffer)) {
889                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
890                 goto done;
891         }
892         eol = strchr(buffer, '\n');
893         if (!eol) {
894                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
895                 goto done;
896         }
897         strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
898         if (check_refname_format(sb.buf, 0)) {
899                 ret = report(options, &tag->object, FSCK_MSG_BAD_TAG_NAME,
900                            "invalid 'tag' name: %.*s",
901                            (int)(eol - buffer), buffer);
902                 if (ret)
903                         goto done;
904         }
905         buffer = eol + 1;
906
907         if (!skip_prefix(buffer, "tagger ", &buffer)) {
908                 /* early tags do not contain 'tagger' lines; warn only */
909                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
910                 if (ret)
911                         goto done;
912         }
913         else
914                 ret = fsck_ident(&buffer, &tag->object, options);
915
916 done:
917         strbuf_release(&sb);
918         free(to_free);
919         return ret;
920 }
921
922 static int fsck_tag(struct tag *tag, const char *data,
923         unsigned long size, struct fsck_options *options)
924 {
925         struct object *tagged = tag->tagged;
926
927         if (!tagged)
928                 return report(options, &tag->object, FSCK_MSG_BAD_TAG_OBJECT, "could not load tagged object");
929
930         return fsck_tag_buffer(tag, data, size, options);
931 }
932
933 struct fsck_gitmodules_data {
934         struct object *obj;
935         struct fsck_options *options;
936         int ret;
937 };
938
939 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
940 {
941         struct fsck_gitmodules_data *data = vdata;
942         const char *subsection, *key;
943         int subsection_len;
944         char *name;
945
946         if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
947             !subsection)
948                 return 0;
949
950         name = xmemdupz(subsection, subsection_len);
951         if (check_submodule_name(name) < 0)
952                 data->ret |= report(data->options, data->obj,
953                                     FSCK_MSG_GITMODULES_NAME,
954                                     "disallowed submodule name: %s",
955                                     name);
956         if (!strcmp(key, "url") && value &&
957             looks_like_command_line_option(value))
958                 data->ret |= report(data->options, data->obj,
959                                     FSCK_MSG_GITMODULES_URL,
960                                     "disallowed submodule url: %s",
961                                     value);
962         if (!strcmp(key, "path") && value &&
963             looks_like_command_line_option(value))
964                 data->ret |= report(data->options, data->obj,
965                                     FSCK_MSG_GITMODULES_PATH,
966                                     "disallowed submodule path: %s",
967                                     value);
968         free(name);
969
970         return 0;
971 }
972
973 static int fsck_blob(struct blob *blob, const char *buf,
974                      unsigned long size, struct fsck_options *options)
975 {
976         struct fsck_gitmodules_data data;
977         struct config_options config_opts = { 0 };
978
979         if (!oidset_contains(&gitmodules_found, &blob->object.oid))
980                 return 0;
981         oidset_insert(&gitmodules_done, &blob->object.oid);
982
983         if (object_on_skiplist(options, &blob->object))
984                 return 0;
985
986         if (!buf) {
987                 /*
988                  * A missing buffer here is a sign that the caller found the
989                  * blob too gigantic to load into memory. Let's just consider
990                  * that an error.
991                  */
992                 return report(options, &blob->object,
993                               FSCK_MSG_GITMODULES_LARGE,
994                               ".gitmodules too large to parse");
995         }
996
997         data.obj = &blob->object;
998         data.options = options;
999         data.ret = 0;
1000         config_opts.error_action = CONFIG_ERROR_SILENT;
1001         if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1002                                 ".gitmodules", buf, size, &data, &config_opts))
1003                 data.ret |= report(options, &blob->object,
1004                                    FSCK_MSG_GITMODULES_PARSE,
1005                                    "could not parse gitmodules blob");
1006
1007         return data.ret;
1008 }
1009
1010 int fsck_object(struct object *obj, void *data, unsigned long size,
1011         struct fsck_options *options)
1012 {
1013         if (!obj)
1014                 return report(options, obj, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1015
1016         if (obj->type == OBJ_BLOB)
1017                 return fsck_blob((struct blob *)obj, data, size, options);
1018         if (obj->type == OBJ_TREE)
1019                 return fsck_tree((struct tree *) obj, options);
1020         if (obj->type == OBJ_COMMIT)
1021                 return fsck_commit((struct commit *) obj, (const char *) data,
1022                         size, options);
1023         if (obj->type == OBJ_TAG)
1024                 return fsck_tag((struct tag *) obj, (const char *) data,
1025                         size, options);
1026
1027         return report(options, obj, FSCK_MSG_UNKNOWN_TYPE, "unknown type '%d' (internal fsck error)",
1028                           obj->type);
1029 }
1030
1031 int fsck_error_function(struct fsck_options *o,
1032         struct object *obj, int msg_type, const char *message)
1033 {
1034         if (msg_type == FSCK_WARN) {
1035                 warning("object %s: %s", describe_object(o, obj), message);
1036                 return 0;
1037         }
1038         error("object %s: %s", describe_object(o, obj), message);
1039         return 1;
1040 }
1041
1042 int fsck_finish(struct fsck_options *options)
1043 {
1044         int ret = 0;
1045         struct oidset_iter iter;
1046         const struct object_id *oid;
1047
1048         oidset_iter_init(&gitmodules_found, &iter);
1049         while ((oid = oidset_iter_next(&iter))) {
1050                 struct blob *blob;
1051                 enum object_type type;
1052                 unsigned long size;
1053                 char *buf;
1054
1055                 if (oidset_contains(&gitmodules_done, oid))
1056                         continue;
1057
1058                 blob = lookup_blob(the_repository, oid);
1059                 if (!blob) {
1060                         struct object *obj = lookup_unknown_object(oid);
1061                         ret |= report(options, obj,
1062                                       FSCK_MSG_GITMODULES_BLOB,
1063                                       "non-blob found at .gitmodules");
1064                         continue;
1065                 }
1066
1067                 buf = read_object_file(oid, &type, &size);
1068                 if (!buf) {
1069                         if (is_promisor_object(&blob->object.oid))
1070                                 continue;
1071                         ret |= report(options, &blob->object,
1072                                       FSCK_MSG_GITMODULES_MISSING,
1073                                       "unable to read .gitmodules blob");
1074                         continue;
1075                 }
1076
1077                 if (type == OBJ_BLOB)
1078                         ret |= fsck_blob(blob, buf, size, options);
1079                 else
1080                         ret |= report(options, &blob->object,
1081                                       FSCK_MSG_GITMODULES_BLOB,
1082                                       "non-blob found at .gitmodules");
1083                 free(buf);
1084         }
1085
1086
1087         oidset_clear(&gitmodules_found);
1088         oidset_clear(&gitmodules_done);
1089         return ret;
1090 }