Merge branch 'jk/submodule-fsck-loose' into maint
[git] / fsck.c
1 #include "cache.h"
2 #include "object.h"
3 #include "blob.h"
4 #include "tree.h"
5 #include "tree-walk.h"
6 #include "commit.h"
7 #include "tag.h"
8 #include "fsck.h"
9 #include "refs.h"
10 #include "utf8.h"
11 #include "sha1-array.h"
12 #include "decorate.h"
13 #include "oidset.h"
14 #include "packfile.h"
15 #include "submodule-config.h"
16 #include "config.h"
17
18 static struct oidset gitmodules_found = OIDSET_INIT;
19 static struct oidset gitmodules_done = OIDSET_INIT;
20
21 #define FSCK_FATAL -1
22 #define FSCK_INFO -2
23
24 #define FOREACH_MSG_ID(FUNC) \
25         /* fatal errors */ \
26         FUNC(NUL_IN_HEADER, FATAL) \
27         FUNC(UNTERMINATED_HEADER, FATAL) \
28         /* errors */ \
29         FUNC(BAD_DATE, ERROR) \
30         FUNC(BAD_DATE_OVERFLOW, ERROR) \
31         FUNC(BAD_EMAIL, ERROR) \
32         FUNC(BAD_NAME, ERROR) \
33         FUNC(BAD_OBJECT_SHA1, ERROR) \
34         FUNC(BAD_PARENT_SHA1, ERROR) \
35         FUNC(BAD_TAG_OBJECT, ERROR) \
36         FUNC(BAD_TIMEZONE, ERROR) \
37         FUNC(BAD_TREE, ERROR) \
38         FUNC(BAD_TREE_SHA1, ERROR) \
39         FUNC(BAD_TYPE, ERROR) \
40         FUNC(DUPLICATE_ENTRIES, ERROR) \
41         FUNC(MISSING_AUTHOR, ERROR) \
42         FUNC(MISSING_COMMITTER, ERROR) \
43         FUNC(MISSING_EMAIL, ERROR) \
44         FUNC(MISSING_GRAFT, ERROR) \
45         FUNC(MISSING_NAME_BEFORE_EMAIL, ERROR) \
46         FUNC(MISSING_OBJECT, ERROR) \
47         FUNC(MISSING_PARENT, ERROR) \
48         FUNC(MISSING_SPACE_BEFORE_DATE, ERROR) \
49         FUNC(MISSING_SPACE_BEFORE_EMAIL, ERROR) \
50         FUNC(MISSING_TAG, ERROR) \
51         FUNC(MISSING_TAG_ENTRY, ERROR) \
52         FUNC(MISSING_TAG_OBJECT, ERROR) \
53         FUNC(MISSING_TREE, ERROR) \
54         FUNC(MISSING_TREE_OBJECT, ERROR) \
55         FUNC(MISSING_TYPE, ERROR) \
56         FUNC(MISSING_TYPE_ENTRY, ERROR) \
57         FUNC(MULTIPLE_AUTHORS, ERROR) \
58         FUNC(TAG_OBJECT_NOT_TAG, ERROR) \
59         FUNC(TREE_NOT_SORTED, ERROR) \
60         FUNC(UNKNOWN_TYPE, ERROR) \
61         FUNC(ZERO_PADDED_DATE, ERROR) \
62         FUNC(GITMODULES_MISSING, ERROR) \
63         FUNC(GITMODULES_BLOB, ERROR) \
64         FUNC(GITMODULES_PARSE, ERROR) \
65         FUNC(GITMODULES_NAME, ERROR) \
66         FUNC(GITMODULES_SYMLINK, ERROR) \
67         /* warnings */ \
68         FUNC(BAD_FILEMODE, WARN) \
69         FUNC(EMPTY_NAME, WARN) \
70         FUNC(FULL_PATHNAME, WARN) \
71         FUNC(HAS_DOT, WARN) \
72         FUNC(HAS_DOTDOT, WARN) \
73         FUNC(HAS_DOTGIT, WARN) \
74         FUNC(NULL_SHA1, WARN) \
75         FUNC(ZERO_PADDED_FILEMODE, WARN) \
76         FUNC(NUL_IN_COMMIT, WARN) \
77         /* infos (reported as warnings, but ignored by default) */ \
78         FUNC(BAD_TAG_NAME, INFO) \
79         FUNC(MISSING_TAGGER_ENTRY, INFO)
80
81 #define MSG_ID(id, msg_type) FSCK_MSG_##id,
82 enum fsck_msg_id {
83         FOREACH_MSG_ID(MSG_ID)
84         FSCK_MSG_MAX
85 };
86 #undef MSG_ID
87
88 #define STR(x) #x
89 #define MSG_ID(id, msg_type) { STR(id), NULL, FSCK_##msg_type },
90 static struct {
91         const char *id_string;
92         const char *downcased;
93         int msg_type;
94 } msg_id_info[FSCK_MSG_MAX + 1] = {
95         FOREACH_MSG_ID(MSG_ID)
96         { NULL, NULL, -1 }
97 };
98 #undef MSG_ID
99
100 static int parse_msg_id(const char *text)
101 {
102         int i;
103
104         if (!msg_id_info[0].downcased) {
105                 /* convert id_string to lower case, without underscores. */
106                 for (i = 0; i < FSCK_MSG_MAX; i++) {
107                         const char *p = msg_id_info[i].id_string;
108                         int len = strlen(p);
109                         char *q = xmalloc(len);
110
111                         msg_id_info[i].downcased = q;
112                         while (*p)
113                                 if (*p == '_')
114                                         p++;
115                                 else
116                                         *(q)++ = tolower(*(p)++);
117                         *q = '\0';
118                 }
119         }
120
121         for (i = 0; i < FSCK_MSG_MAX; i++)
122                 if (!strcmp(text, msg_id_info[i].downcased))
123                         return i;
124
125         return -1;
126 }
127
128 static int fsck_msg_type(enum fsck_msg_id msg_id,
129         struct fsck_options *options)
130 {
131         int msg_type;
132
133         assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
134
135         if (options->msg_type)
136                 msg_type = options->msg_type[msg_id];
137         else {
138                 msg_type = msg_id_info[msg_id].msg_type;
139                 if (options->strict && msg_type == FSCK_WARN)
140                         msg_type = FSCK_ERROR;
141         }
142
143         return msg_type;
144 }
145
146 static void init_skiplist(struct fsck_options *options, const char *path)
147 {
148         static struct oid_array skiplist = OID_ARRAY_INIT;
149         int sorted, fd;
150         char buffer[GIT_MAX_HEXSZ + 1];
151         struct object_id oid;
152
153         if (options->skiplist)
154                 sorted = options->skiplist->sorted;
155         else {
156                 sorted = 1;
157                 options->skiplist = &skiplist;
158         }
159
160         fd = open(path, O_RDONLY);
161         if (fd < 0)
162                 die("Could not open skip list: %s", path);
163         for (;;) {
164                 const char *p;
165                 int result = read_in_full(fd, buffer, sizeof(buffer));
166                 if (result < 0)
167                         die_errno("Could not read '%s'", path);
168                 if (!result)
169                         break;
170                 if (parse_oid_hex(buffer, &oid, &p) || *p != '\n')
171                         die("Invalid SHA-1: %s", buffer);
172                 oid_array_append(&skiplist, &oid);
173                 if (sorted && skiplist.nr > 1 &&
174                                 oidcmp(&skiplist.oid[skiplist.nr - 2],
175                                        &oid) > 0)
176                         sorted = 0;
177         }
178         close(fd);
179
180         if (sorted)
181                 skiplist.sorted = 1;
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                         init_skiplist(options, 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 __attribute__((format (printf, 4, 5)))
285 static int report(struct fsck_options *options, struct object *object,
286         enum fsck_msg_id id, const char *fmt, ...)
287 {
288         va_list ap;
289         struct strbuf sb = STRBUF_INIT;
290         int msg_type = fsck_msg_type(id, options), result;
291
292         if (msg_type == FSCK_IGNORE)
293                 return 0;
294
295         if (options->skiplist && object &&
296                         oid_array_lookup(options->skiplist, &object->oid) >= 0)
297                 return 0;
298
299         if (msg_type == FSCK_FATAL)
300                 msg_type = FSCK_ERROR;
301         else if (msg_type == FSCK_INFO)
302                 msg_type = FSCK_WARN;
303
304         append_msg_id(&sb, msg_id_info[id].id_string);
305
306         va_start(ap, fmt);
307         strbuf_vaddf(&sb, fmt, ap);
308         result = options->error_func(options, object, msg_type, sb.buf);
309         strbuf_release(&sb);
310         va_end(ap);
311
312         return result;
313 }
314
315 static char *get_object_name(struct fsck_options *options, struct object *obj)
316 {
317         if (!options->object_names)
318                 return NULL;
319         return lookup_decoration(options->object_names, obj);
320 }
321
322 static void put_object_name(struct fsck_options *options, struct object *obj,
323         const char *fmt, ...)
324 {
325         va_list ap;
326         struct strbuf buf = STRBUF_INIT;
327         char *existing;
328
329         if (!options->object_names)
330                 return;
331         existing = lookup_decoration(options->object_names, obj);
332         if (existing)
333                 return;
334         va_start(ap, fmt);
335         strbuf_vaddf(&buf, fmt, ap);
336         add_decoration(options->object_names, obj, strbuf_detach(&buf, NULL));
337         va_end(ap);
338 }
339
340 static const char *describe_object(struct fsck_options *o, struct object *obj)
341 {
342         static struct strbuf buf = STRBUF_INIT;
343         char *name;
344
345         strbuf_reset(&buf);
346         strbuf_addstr(&buf, oid_to_hex(&obj->oid));
347         if (o->object_names && (name = lookup_decoration(o->object_names, obj)))
348                 strbuf_addf(&buf, " (%s)", name);
349
350         return buf.buf;
351 }
352
353 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
354 {
355         struct tree_desc desc;
356         struct name_entry entry;
357         int res = 0;
358         const char *name;
359
360         if (parse_tree(tree))
361                 return -1;
362
363         name = get_object_name(options, &tree->object);
364         if (init_tree_desc_gently(&desc, tree->buffer, tree->size))
365                 return -1;
366         while (tree_entry_gently(&desc, &entry)) {
367                 struct object *obj;
368                 int result;
369
370                 if (S_ISGITLINK(entry.mode))
371                         continue;
372
373                 if (S_ISDIR(entry.mode)) {
374                         obj = (struct object *)lookup_tree(entry.oid);
375                         if (name && obj)
376                                 put_object_name(options, obj, "%s%s/", name,
377                                         entry.path);
378                         result = options->walk(obj, OBJ_TREE, data, options);
379                 }
380                 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
381                         obj = (struct object *)lookup_blob(entry.oid);
382                         if (name && obj)
383                                 put_object_name(options, obj, "%s%s", name,
384                                         entry.path);
385                         result = options->walk(obj, OBJ_BLOB, data, options);
386                 }
387                 else {
388                         result = error("in tree %s: entry %s has bad mode %.6o",
389                                         describe_object(options, &tree->object), entry.path, entry.mode);
390                 }
391                 if (result < 0)
392                         return result;
393                 if (!res)
394                         res = result;
395         }
396         return res;
397 }
398
399 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
400 {
401         int counter = 0, generation = 0, name_prefix_len = 0;
402         struct commit_list *parents;
403         int res;
404         int result;
405         const char *name;
406
407         if (parse_commit(commit))
408                 return -1;
409
410         name = get_object_name(options, &commit->object);
411         if (name)
412                 put_object_name(options, &commit->tree->object, "%s:", name);
413
414         result = options->walk((struct object *)commit->tree, OBJ_TREE, data, options);
415         if (result < 0)
416                 return result;
417         res = result;
418
419         parents = commit->parents;
420         if (name && parents) {
421                 int len = strlen(name), power;
422
423                 if (len && name[len - 1] == '^') {
424                         generation = 1;
425                         name_prefix_len = len - 1;
426                 }
427                 else { /* parse ~<generation> suffix */
428                         for (generation = 0, power = 1;
429                              len && isdigit(name[len - 1]);
430                              power *= 10)
431                                 generation += power * (name[--len] - '0');
432                         if (power > 1 && len && name[len - 1] == '~')
433                                 name_prefix_len = len - 1;
434                 }
435         }
436
437         while (parents) {
438                 if (name) {
439                         struct object *obj = &parents->item->object;
440
441                         if (++counter > 1)
442                                 put_object_name(options, obj, "%s^%d",
443                                         name, counter);
444                         else if (generation > 0)
445                                 put_object_name(options, obj, "%.*s~%d",
446                                         name_prefix_len, name, generation + 1);
447                         else
448                                 put_object_name(options, obj, "%s^", name);
449                 }
450                 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
451                 if (result < 0)
452                         return result;
453                 if (!res)
454                         res = result;
455                 parents = parents->next;
456         }
457         return res;
458 }
459
460 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
461 {
462         char *name = get_object_name(options, &tag->object);
463
464         if (parse_tag(tag))
465                 return -1;
466         if (name)
467                 put_object_name(options, tag->tagged, "%s", name);
468         return options->walk(tag->tagged, OBJ_ANY, data, options);
469 }
470
471 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
472 {
473         if (!obj)
474                 return -1;
475
476         if (obj->type == OBJ_NONE)
477                 parse_object(&obj->oid);
478
479         switch (obj->type) {
480         case OBJ_BLOB:
481                 return 0;
482         case OBJ_TREE:
483                 return fsck_walk_tree((struct tree *)obj, data, options);
484         case OBJ_COMMIT:
485                 return fsck_walk_commit((struct commit *)obj, data, options);
486         case OBJ_TAG:
487                 return fsck_walk_tag((struct tag *)obj, data, options);
488         default:
489                 error("Unknown object type for %s", describe_object(options, obj));
490                 return -1;
491         }
492 }
493
494 /*
495  * The entries in a tree are ordered in the _path_ order,
496  * which means that a directory entry is ordered by adding
497  * a slash to the end of it.
498  *
499  * So a directory called "a" is ordered _after_ a file
500  * called "a.c", because "a/" sorts after "a.c".
501  */
502 #define TREE_UNORDERED (-1)
503 #define TREE_HAS_DUPS  (-2)
504
505 static int verify_ordered(unsigned mode1, const char *name1, unsigned mode2, const char *name2)
506 {
507         int len1 = strlen(name1);
508         int len2 = strlen(name2);
509         int len = len1 < len2 ? len1 : len2;
510         unsigned char c1, c2;
511         int cmp;
512
513         cmp = memcmp(name1, name2, len);
514         if (cmp < 0)
515                 return 0;
516         if (cmp > 0)
517                 return TREE_UNORDERED;
518
519         /*
520          * Ok, the first <len> characters are the same.
521          * Now we need to order the next one, but turn
522          * a '\0' into a '/' for a directory entry.
523          */
524         c1 = name1[len];
525         c2 = name2[len];
526         if (!c1 && !c2)
527                 /*
528                  * git-write-tree used to write out a nonsense tree that has
529                  * entries with the same name, one blob and one tree.  Make
530                  * sure we do not have duplicate entries.
531                  */
532                 return TREE_HAS_DUPS;
533         if (!c1 && S_ISDIR(mode1))
534                 c1 = '/';
535         if (!c2 && S_ISDIR(mode2))
536                 c2 = '/';
537         return c1 < c2 ? 0 : TREE_UNORDERED;
538 }
539
540 static int fsck_tree(struct tree *item, struct fsck_options *options)
541 {
542         int retval = 0;
543         int has_null_sha1 = 0;
544         int has_full_path = 0;
545         int has_empty_name = 0;
546         int has_dot = 0;
547         int has_dotdot = 0;
548         int has_dotgit = 0;
549         int has_zero_pad = 0;
550         int has_bad_modes = 0;
551         int has_dup_entries = 0;
552         int not_properly_sorted = 0;
553         struct tree_desc desc;
554         unsigned o_mode;
555         const char *o_name;
556
557         if (init_tree_desc_gently(&desc, item->buffer, item->size)) {
558                 retval += report(options, &item->object, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
559                 return retval;
560         }
561
562         o_mode = 0;
563         o_name = NULL;
564
565         while (desc.size) {
566                 unsigned mode;
567                 const char *name;
568                 const struct object_id *oid;
569
570                 oid = tree_entry_extract(&desc, &name, &mode);
571
572                 has_null_sha1 |= is_null_oid(oid);
573                 has_full_path |= !!strchr(name, '/');
574                 has_empty_name |= !*name;
575                 has_dot |= !strcmp(name, ".");
576                 has_dotdot |= !strcmp(name, "..");
577                 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
578                 has_zero_pad |= *(char *)desc.buffer == '0';
579
580                 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
581                         if (!S_ISLNK(mode))
582                                 oidset_insert(&gitmodules_found, oid);
583                         else
584                                 retval += report(options, &item->object,
585                                                  FSCK_MSG_GITMODULES_SYMLINK,
586                                                  ".gitmodules is a symbolic link");
587                 }
588
589                 if (update_tree_entry_gently(&desc)) {
590                         retval += report(options, &item->object, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
591                         break;
592                 }
593
594                 switch (mode) {
595                 /*
596                  * Standard modes..
597                  */
598                 case S_IFREG | 0755:
599                 case S_IFREG | 0644:
600                 case S_IFLNK:
601                 case S_IFDIR:
602                 case S_IFGITLINK:
603                         break;
604                 /*
605                  * This is nonstandard, but we had a few of these
606                  * early on when we honored the full set of mode
607                  * bits..
608                  */
609                 case S_IFREG | 0664:
610                         if (!options->strict)
611                                 break;
612                         /* fallthrough */
613                 default:
614                         has_bad_modes = 1;
615                 }
616
617                 if (o_name) {
618                         switch (verify_ordered(o_mode, o_name, mode, name)) {
619                         case TREE_UNORDERED:
620                                 not_properly_sorted = 1;
621                                 break;
622                         case TREE_HAS_DUPS:
623                                 has_dup_entries = 1;
624                                 break;
625                         default:
626                                 break;
627                         }
628                 }
629
630                 o_mode = mode;
631                 o_name = name;
632         }
633
634         if (has_null_sha1)
635                 retval += report(options, &item->object, FSCK_MSG_NULL_SHA1, "contains entries pointing to null sha1");
636         if (has_full_path)
637                 retval += report(options, &item->object, FSCK_MSG_FULL_PATHNAME, "contains full pathnames");
638         if (has_empty_name)
639                 retval += report(options, &item->object, FSCK_MSG_EMPTY_NAME, "contains empty pathname");
640         if (has_dot)
641                 retval += report(options, &item->object, FSCK_MSG_HAS_DOT, "contains '.'");
642         if (has_dotdot)
643                 retval += report(options, &item->object, FSCK_MSG_HAS_DOTDOT, "contains '..'");
644         if (has_dotgit)
645                 retval += report(options, &item->object, FSCK_MSG_HAS_DOTGIT, "contains '.git'");
646         if (has_zero_pad)
647                 retval += report(options, &item->object, FSCK_MSG_ZERO_PADDED_FILEMODE, "contains zero-padded file modes");
648         if (has_bad_modes)
649                 retval += report(options, &item->object, FSCK_MSG_BAD_FILEMODE, "contains bad file modes");
650         if (has_dup_entries)
651                 retval += report(options, &item->object, FSCK_MSG_DUPLICATE_ENTRIES, "contains duplicate file entries");
652         if (not_properly_sorted)
653                 retval += report(options, &item->object, FSCK_MSG_TREE_NOT_SORTED, "not properly sorted");
654         return retval;
655 }
656
657 static int verify_headers(const void *data, unsigned long size,
658                           struct object *obj, struct fsck_options *options)
659 {
660         const char *buffer = (const char *)data;
661         unsigned long i;
662
663         for (i = 0; i < size; i++) {
664                 switch (buffer[i]) {
665                 case '\0':
666                         return report(options, obj,
667                                 FSCK_MSG_NUL_IN_HEADER,
668                                 "unterminated header: NUL at offset %ld", i);
669                 case '\n':
670                         if (i + 1 < size && buffer[i + 1] == '\n')
671                                 return 0;
672                 }
673         }
674
675         /*
676          * We did not find double-LF that separates the header
677          * and the body.  Not having a body is not a crime but
678          * we do want to see the terminating LF for the last header
679          * line.
680          */
681         if (size && buffer[size - 1] == '\n')
682                 return 0;
683
684         return report(options, obj,
685                 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
686 }
687
688 static int fsck_ident(const char **ident, struct object *obj, struct fsck_options *options)
689 {
690         const char *p = *ident;
691         char *end;
692
693         *ident = strchrnul(*ident, '\n');
694         if (**ident == '\n')
695                 (*ident)++;
696
697         if (*p == '<')
698                 return report(options, obj, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
699         p += strcspn(p, "<>\n");
700         if (*p == '>')
701                 return report(options, obj, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
702         if (*p != '<')
703                 return report(options, obj, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
704         if (p[-1] != ' ')
705                 return report(options, obj, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
706         p++;
707         p += strcspn(p, "<>\n");
708         if (*p != '>')
709                 return report(options, obj, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
710         p++;
711         if (*p != ' ')
712                 return report(options, obj, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
713         p++;
714         if (*p == '0' && p[1] != ' ')
715                 return report(options, obj, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
716         if (date_overflows(parse_timestamp(p, &end, 10)))
717                 return report(options, obj, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
718         if ((end == p || *end != ' '))
719                 return report(options, obj, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
720         p = end + 1;
721         if ((*p != '+' && *p != '-') ||
722             !isdigit(p[1]) ||
723             !isdigit(p[2]) ||
724             !isdigit(p[3]) ||
725             !isdigit(p[4]) ||
726             (p[5] != '\n'))
727                 return report(options, obj, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
728         p += 6;
729         return 0;
730 }
731
732 static int fsck_commit_buffer(struct commit *commit, const char *buffer,
733         unsigned long size, struct fsck_options *options)
734 {
735         unsigned char tree_sha1[20], sha1[20];
736         struct commit_graft *graft;
737         unsigned parent_count, parent_line_count = 0, author_count;
738         int err;
739         const char *buffer_begin = buffer;
740
741         if (verify_headers(buffer, size, &commit->object, options))
742                 return -1;
743
744         if (!skip_prefix(buffer, "tree ", &buffer))
745                 return report(options, &commit->object, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
746         if (get_sha1_hex(buffer, tree_sha1) || buffer[40] != '\n') {
747                 err = report(options, &commit->object, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
748                 if (err)
749                         return err;
750         }
751         buffer += 41;
752         while (skip_prefix(buffer, "parent ", &buffer)) {
753                 if (get_sha1_hex(buffer, sha1) || buffer[40] != '\n') {
754                         err = report(options, &commit->object, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
755                         if (err)
756                                 return err;
757                 }
758                 buffer += 41;
759                 parent_line_count++;
760         }
761         graft = lookup_commit_graft(&commit->object.oid);
762         parent_count = commit_list_count(commit->parents);
763         if (graft) {
764                 if (graft->nr_parent == -1 && !parent_count)
765                         ; /* shallow commit */
766                 else if (graft->nr_parent != parent_count) {
767                         err = report(options, &commit->object, FSCK_MSG_MISSING_GRAFT, "graft objects missing");
768                         if (err)
769                                 return err;
770                 }
771         } else {
772                 if (parent_count != parent_line_count) {
773                         err = report(options, &commit->object, FSCK_MSG_MISSING_PARENT, "parent objects missing");
774                         if (err)
775                                 return err;
776                 }
777         }
778         author_count = 0;
779         while (skip_prefix(buffer, "author ", &buffer)) {
780                 author_count++;
781                 err = fsck_ident(&buffer, &commit->object, options);
782                 if (err)
783                         return err;
784         }
785         if (author_count < 1)
786                 err = report(options, &commit->object, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
787         else if (author_count > 1)
788                 err = report(options, &commit->object, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
789         if (err)
790                 return err;
791         if (!skip_prefix(buffer, "committer ", &buffer))
792                 return report(options, &commit->object, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
793         err = fsck_ident(&buffer, &commit->object, options);
794         if (err)
795                 return err;
796         if (!commit->tree) {
797                 err = report(options, &commit->object, FSCK_MSG_BAD_TREE, "could not load commit's tree %s", sha1_to_hex(tree_sha1));
798                 if (err)
799                         return err;
800         }
801         if (memchr(buffer_begin, '\0', size)) {
802                 err = report(options, &commit->object, FSCK_MSG_NUL_IN_COMMIT,
803                              "NUL byte in the commit object body");
804                 if (err)
805                         return err;
806         }
807         return 0;
808 }
809
810 static int fsck_commit(struct commit *commit, const char *data,
811         unsigned long size, struct fsck_options *options)
812 {
813         const char *buffer = data ?  data : get_commit_buffer(commit, &size);
814         int ret = fsck_commit_buffer(commit, buffer, size, options);
815         if (!data)
816                 unuse_commit_buffer(commit, buffer);
817         return ret;
818 }
819
820 static int fsck_tag_buffer(struct tag *tag, const char *data,
821         unsigned long size, struct fsck_options *options)
822 {
823         unsigned char sha1[20];
824         int ret = 0;
825         const char *buffer;
826         char *to_free = NULL, *eol;
827         struct strbuf sb = STRBUF_INIT;
828
829         if (data)
830                 buffer = data;
831         else {
832                 enum object_type type;
833
834                 buffer = to_free =
835                         read_sha1_file(tag->object.oid.hash, &type, &size);
836                 if (!buffer)
837                         return report(options, &tag->object,
838                                 FSCK_MSG_MISSING_TAG_OBJECT,
839                                 "cannot read tag object");
840
841                 if (type != OBJ_TAG) {
842                         ret = report(options, &tag->object,
843                                 FSCK_MSG_TAG_OBJECT_NOT_TAG,
844                                 "expected tag got %s",
845                             type_name(type));
846                         goto done;
847                 }
848         }
849
850         ret = verify_headers(buffer, size, &tag->object, options);
851         if (ret)
852                 goto done;
853
854         if (!skip_prefix(buffer, "object ", &buffer)) {
855                 ret = report(options, &tag->object, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
856                 goto done;
857         }
858         if (get_sha1_hex(buffer, sha1) || buffer[40] != '\n') {
859                 ret = report(options, &tag->object, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
860                 if (ret)
861                         goto done;
862         }
863         buffer += 41;
864
865         if (!skip_prefix(buffer, "type ", &buffer)) {
866                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
867                 goto done;
868         }
869         eol = strchr(buffer, '\n');
870         if (!eol) {
871                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
872                 goto done;
873         }
874         if (type_from_string_gently(buffer, eol - buffer, 1) < 0)
875                 ret = report(options, &tag->object, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
876         if (ret)
877                 goto done;
878         buffer = eol + 1;
879
880         if (!skip_prefix(buffer, "tag ", &buffer)) {
881                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
882                 goto done;
883         }
884         eol = strchr(buffer, '\n');
885         if (!eol) {
886                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
887                 goto done;
888         }
889         strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
890         if (check_refname_format(sb.buf, 0)) {
891                 ret = report(options, &tag->object, FSCK_MSG_BAD_TAG_NAME,
892                            "invalid 'tag' name: %.*s",
893                            (int)(eol - buffer), buffer);
894                 if (ret)
895                         goto done;
896         }
897         buffer = eol + 1;
898
899         if (!skip_prefix(buffer, "tagger ", &buffer)) {
900                 /* early tags do not contain 'tagger' lines; warn only */
901                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
902                 if (ret)
903                         goto done;
904         }
905         else
906                 ret = fsck_ident(&buffer, &tag->object, options);
907
908 done:
909         strbuf_release(&sb);
910         free(to_free);
911         return ret;
912 }
913
914 static int fsck_tag(struct tag *tag, const char *data,
915         unsigned long size, struct fsck_options *options)
916 {
917         struct object *tagged = tag->tagged;
918
919         if (!tagged)
920                 return report(options, &tag->object, FSCK_MSG_BAD_TAG_OBJECT, "could not load tagged object");
921
922         return fsck_tag_buffer(tag, data, size, options);
923 }
924
925 struct fsck_gitmodules_data {
926         struct object *obj;
927         struct fsck_options *options;
928         int ret;
929 };
930
931 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
932 {
933         struct fsck_gitmodules_data *data = vdata;
934         const char *subsection, *key;
935         int subsection_len;
936         char *name;
937
938         if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
939             !subsection)
940                 return 0;
941
942         name = xmemdupz(subsection, subsection_len);
943         if (check_submodule_name(name) < 0)
944                 data->ret |= report(data->options, data->obj,
945                                     FSCK_MSG_GITMODULES_NAME,
946                                     "disallowed submodule name: %s",
947                                     name);
948         free(name);
949
950         return 0;
951 }
952
953 static int fsck_blob(struct blob *blob, const char *buf,
954                      unsigned long size, struct fsck_options *options)
955 {
956         struct fsck_gitmodules_data data;
957
958         if (!oidset_contains(&gitmodules_found, &blob->object.oid))
959                 return 0;
960         oidset_insert(&gitmodules_done, &blob->object.oid);
961
962         if (!buf) {
963                 /*
964                  * A missing buffer here is a sign that the caller found the
965                  * blob too gigantic to load into memory. Let's just consider
966                  * that an error.
967                  */
968                 return report(options, &blob->object,
969                               FSCK_MSG_GITMODULES_PARSE,
970                               ".gitmodules too large to parse");
971         }
972
973         data.obj = &blob->object;
974         data.options = options;
975         data.ret = 0;
976         if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
977                                 ".gitmodules", buf, size, &data))
978                 data.ret |= report(options, &blob->object,
979                                    FSCK_MSG_GITMODULES_PARSE,
980                                    "could not parse gitmodules blob");
981
982         return data.ret;
983 }
984
985 int fsck_object(struct object *obj, void *data, unsigned long size,
986         struct fsck_options *options)
987 {
988         if (!obj)
989                 return report(options, obj, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
990
991         if (obj->type == OBJ_BLOB)
992                 return fsck_blob((struct blob *)obj, data, size, options);
993         if (obj->type == OBJ_TREE)
994                 return fsck_tree((struct tree *) obj, options);
995         if (obj->type == OBJ_COMMIT)
996                 return fsck_commit((struct commit *) obj, (const char *) data,
997                         size, options);
998         if (obj->type == OBJ_TAG)
999                 return fsck_tag((struct tag *) obj, (const char *) data,
1000                         size, options);
1001
1002         return report(options, obj, FSCK_MSG_UNKNOWN_TYPE, "unknown type '%d' (internal fsck error)",
1003                           obj->type);
1004 }
1005
1006 int fsck_error_function(struct fsck_options *o,
1007         struct object *obj, int msg_type, const char *message)
1008 {
1009         if (msg_type == FSCK_WARN) {
1010                 warning("object %s: %s", describe_object(o, obj), message);
1011                 return 0;
1012         }
1013         error("object %s: %s", describe_object(o, obj), message);
1014         return 1;
1015 }
1016
1017 int fsck_finish(struct fsck_options *options)
1018 {
1019         int ret = 0;
1020         struct oidset_iter iter;
1021         const struct object_id *oid;
1022
1023         oidset_iter_init(&gitmodules_found, &iter);
1024         while ((oid = oidset_iter_next(&iter))) {
1025                 struct blob *blob;
1026                 enum object_type type;
1027                 unsigned long size;
1028                 char *buf;
1029
1030                 if (oidset_contains(&gitmodules_done, oid))
1031                         continue;
1032
1033                 blob = lookup_blob(oid);
1034                 if (!blob) {
1035                         ret |= report(options, &blob->object,
1036                                       FSCK_MSG_GITMODULES_BLOB,
1037                                       "non-blob found at .gitmodules");
1038                         continue;
1039                 }
1040
1041                 buf = read_sha1_file(oid->hash, &type, &size);
1042                 if (!buf) {
1043                         if (is_promisor_object(&blob->object.oid))
1044                                 continue;
1045                         ret |= report(options, &blob->object,
1046                                       FSCK_MSG_GITMODULES_MISSING,
1047                                       "unable to read .gitmodules blob");
1048                         continue;
1049                 }
1050
1051                 if (type == OBJ_BLOB)
1052                         ret |= fsck_blob(blob, buf, size, options);
1053                 else
1054                         ret |= report(options, &blob->object,
1055                                       FSCK_MSG_GITMODULES_BLOB,
1056                                       "non-blob found at .gitmodules");
1057                 free(buf);
1058         }
1059
1060
1061         oidset_clear(&gitmodules_found);
1062         oidset_clear(&gitmodules_done);
1063         return ret;
1064 }