Merge branch 'js/bundle-verify-require-object-store' into maint
[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 void init_skiplist(struct fsck_options *options, const char *path)
185 {
186         FILE *fp;
187         struct strbuf sb = STRBUF_INIT;
188         struct object_id oid;
189
190         fp = fopen(path, "r");
191         if (!fp)
192                 die("Could not open skip list: %s", path);
193         while (!strbuf_getline(&sb, fp)) {
194                 const char *p;
195                 const char *hash;
196
197                 /*
198                  * Allow trailing comments, leading whitespace
199                  * (including before commits), and empty or whitespace
200                  * only lines.
201                  */
202                 hash = strchr(sb.buf, '#');
203                 if (hash)
204                         strbuf_setlen(&sb, hash - sb.buf);
205                 strbuf_trim(&sb);
206                 if (!sb.len)
207                         continue;
208
209                 if (parse_oid_hex(sb.buf, &oid, &p) || *p != '\0')
210                         die("Invalid SHA-1: %s", sb.buf);
211                 oidset_insert(&options->skiplist, &oid);
212         }
213         if (ferror(fp))
214                 die_errno("Could not read '%s'", path);
215         fclose(fp);
216         strbuf_release(&sb);
217 }
218
219 static int parse_msg_type(const char *str)
220 {
221         if (!strcmp(str, "error"))
222                 return FSCK_ERROR;
223         else if (!strcmp(str, "warn"))
224                 return FSCK_WARN;
225         else if (!strcmp(str, "ignore"))
226                 return FSCK_IGNORE;
227         else
228                 die("Unknown fsck message type: '%s'", str);
229 }
230
231 int is_valid_msg_type(const char *msg_id, const char *msg_type)
232 {
233         if (parse_msg_id(msg_id) < 0)
234                 return 0;
235         parse_msg_type(msg_type);
236         return 1;
237 }
238
239 void fsck_set_msg_type(struct fsck_options *options,
240                 const char *msg_id, const char *msg_type)
241 {
242         int id = parse_msg_id(msg_id), type;
243
244         if (id < 0)
245                 die("Unhandled message id: %s", msg_id);
246         type = parse_msg_type(msg_type);
247
248         if (type != FSCK_ERROR && msg_id_info[id].msg_type == FSCK_FATAL)
249                 die("Cannot demote %s to %s", msg_id, msg_type);
250
251         if (!options->msg_type) {
252                 int i;
253                 int *msg_type;
254                 ALLOC_ARRAY(msg_type, FSCK_MSG_MAX);
255                 for (i = 0; i < FSCK_MSG_MAX; i++)
256                         msg_type[i] = fsck_msg_type(i, options);
257                 options->msg_type = msg_type;
258         }
259
260         options->msg_type[id] = type;
261 }
262
263 void fsck_set_msg_types(struct fsck_options *options, const char *values)
264 {
265         char *buf = xstrdup(values), *to_free = buf;
266         int done = 0;
267
268         while (!done) {
269                 int len = strcspn(buf, " ,|"), equal;
270
271                 done = !buf[len];
272                 if (!len) {
273                         buf++;
274                         continue;
275                 }
276                 buf[len] = '\0';
277
278                 for (equal = 0;
279                      equal < len && buf[equal] != '=' && buf[equal] != ':';
280                      equal++)
281                         buf[equal] = tolower(buf[equal]);
282                 buf[equal] = '\0';
283
284                 if (!strcmp(buf, "skiplist")) {
285                         if (equal == len)
286                                 die("skiplist requires a path");
287                         init_skiplist(options, buf + equal + 1);
288                         buf += len + 1;
289                         continue;
290                 }
291
292                 if (equal == len)
293                         die("Missing '=': '%s'", buf);
294
295                 fsck_set_msg_type(options, buf, buf + equal + 1);
296                 buf += len + 1;
297         }
298         free(to_free);
299 }
300
301 static void append_msg_id(struct strbuf *sb, const char *msg_id)
302 {
303         for (;;) {
304                 char c = *(msg_id)++;
305
306                 if (!c)
307                         break;
308                 if (c != '_')
309                         strbuf_addch(sb, tolower(c));
310                 else {
311                         assert(*msg_id);
312                         strbuf_addch(sb, *(msg_id)++);
313                 }
314         }
315
316         strbuf_addstr(sb, ": ");
317 }
318
319 static int object_on_skiplist(struct fsck_options *opts, struct object *obj)
320 {
321         return opts && obj && oidset_contains(&opts->skiplist, &obj->oid);
322 }
323
324 __attribute__((format (printf, 4, 5)))
325 static int report(struct fsck_options *options, struct object *object,
326         enum fsck_msg_id id, const char *fmt, ...)
327 {
328         va_list ap;
329         struct strbuf sb = STRBUF_INIT;
330         int msg_type = fsck_msg_type(id, options), result;
331
332         if (msg_type == FSCK_IGNORE)
333                 return 0;
334
335         if (object_on_skiplist(options, object))
336                 return 0;
337
338         if (msg_type == FSCK_FATAL)
339                 msg_type = FSCK_ERROR;
340         else if (msg_type == FSCK_INFO)
341                 msg_type = FSCK_WARN;
342
343         append_msg_id(&sb, msg_id_info[id].id_string);
344
345         va_start(ap, fmt);
346         strbuf_vaddf(&sb, fmt, ap);
347         result = options->error_func(options, object, msg_type, sb.buf);
348         strbuf_release(&sb);
349         va_end(ap);
350
351         return result;
352 }
353
354 static char *get_object_name(struct fsck_options *options, struct object *obj)
355 {
356         if (!options->object_names)
357                 return NULL;
358         return lookup_decoration(options->object_names, obj);
359 }
360
361 static void put_object_name(struct fsck_options *options, struct object *obj,
362         const char *fmt, ...)
363 {
364         va_list ap;
365         struct strbuf buf = STRBUF_INIT;
366         char *existing;
367
368         if (!options->object_names)
369                 return;
370         existing = lookup_decoration(options->object_names, obj);
371         if (existing)
372                 return;
373         va_start(ap, fmt);
374         strbuf_vaddf(&buf, fmt, ap);
375         add_decoration(options->object_names, obj, strbuf_detach(&buf, NULL));
376         va_end(ap);
377 }
378
379 static const char *describe_object(struct fsck_options *o, struct object *obj)
380 {
381         static struct strbuf buf = STRBUF_INIT;
382         char *name;
383
384         strbuf_reset(&buf);
385         strbuf_addstr(&buf, oid_to_hex(&obj->oid));
386         if (o->object_names && (name = lookup_decoration(o->object_names, obj)))
387                 strbuf_addf(&buf, " (%s)", name);
388
389         return buf.buf;
390 }
391
392 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
393 {
394         struct tree_desc desc;
395         struct name_entry entry;
396         int res = 0;
397         const char *name;
398
399         if (parse_tree(tree))
400                 return -1;
401
402         name = get_object_name(options, &tree->object);
403         if (init_tree_desc_gently(&desc, tree->buffer, tree->size))
404                 return -1;
405         while (tree_entry_gently(&desc, &entry)) {
406                 struct object *obj;
407                 int result;
408
409                 if (S_ISGITLINK(entry.mode))
410                         continue;
411
412                 if (S_ISDIR(entry.mode)) {
413                         obj = (struct object *)lookup_tree(the_repository, &entry.oid);
414                         if (name && obj)
415                                 put_object_name(options, obj, "%s%s/", name,
416                                         entry.path);
417                         result = options->walk(obj, OBJ_TREE, data, options);
418                 }
419                 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
420                         obj = (struct object *)lookup_blob(the_repository, &entry.oid);
421                         if (name && obj)
422                                 put_object_name(options, obj, "%s%s", name,
423                                         entry.path);
424                         result = options->walk(obj, OBJ_BLOB, data, options);
425                 }
426                 else {
427                         result = error("in tree %s: entry %s has bad mode %.6o",
428                                         describe_object(options, &tree->object), entry.path, entry.mode);
429                 }
430                 if (result < 0)
431                         return result;
432                 if (!res)
433                         res = result;
434         }
435         return res;
436 }
437
438 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
439 {
440         int counter = 0, generation = 0, name_prefix_len = 0;
441         struct commit_list *parents;
442         int res;
443         int result;
444         const char *name;
445
446         if (parse_commit(commit))
447                 return -1;
448
449         name = get_object_name(options, &commit->object);
450         if (name)
451                 put_object_name(options, &get_commit_tree(commit)->object,
452                                 "%s:", name);
453
454         result = options->walk((struct object *)get_commit_tree(commit),
455                                OBJ_TREE, data, options);
456         if (result < 0)
457                 return result;
458         res = result;
459
460         parents = commit->parents;
461         if (name && parents) {
462                 int len = strlen(name), power;
463
464                 if (len && name[len - 1] == '^') {
465                         generation = 1;
466                         name_prefix_len = len - 1;
467                 }
468                 else { /* parse ~<generation> suffix */
469                         for (generation = 0, power = 1;
470                              len && isdigit(name[len - 1]);
471                              power *= 10)
472                                 generation += power * (name[--len] - '0');
473                         if (power > 1 && len && name[len - 1] == '~')
474                                 name_prefix_len = len - 1;
475                 }
476         }
477
478         while (parents) {
479                 if (name) {
480                         struct object *obj = &parents->item->object;
481
482                         if (counter++)
483                                 put_object_name(options, obj, "%s^%d",
484                                         name, counter);
485                         else if (generation > 0)
486                                 put_object_name(options, obj, "%.*s~%d",
487                                         name_prefix_len, name, generation + 1);
488                         else
489                                 put_object_name(options, obj, "%s^", name);
490                 }
491                 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
492                 if (result < 0)
493                         return result;
494                 if (!res)
495                         res = result;
496                 parents = parents->next;
497         }
498         return res;
499 }
500
501 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
502 {
503         char *name = get_object_name(options, &tag->object);
504
505         if (parse_tag(tag))
506                 return -1;
507         if (name)
508                 put_object_name(options, tag->tagged, "%s", name);
509         return options->walk(tag->tagged, OBJ_ANY, data, options);
510 }
511
512 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
513 {
514         if (!obj)
515                 return -1;
516
517         if (obj->type == OBJ_NONE)
518                 parse_object(the_repository, &obj->oid);
519
520         switch (obj->type) {
521         case OBJ_BLOB:
522                 return 0;
523         case OBJ_TREE:
524                 return fsck_walk_tree((struct tree *)obj, data, options);
525         case OBJ_COMMIT:
526                 return fsck_walk_commit((struct commit *)obj, data, options);
527         case OBJ_TAG:
528                 return fsck_walk_tag((struct tag *)obj, data, options);
529         default:
530                 error("Unknown object type for %s", describe_object(options, obj));
531                 return -1;
532         }
533 }
534
535 /*
536  * The entries in a tree are ordered in the _path_ order,
537  * which means that a directory entry is ordered by adding
538  * a slash to the end of it.
539  *
540  * So a directory called "a" is ordered _after_ a file
541  * called "a.c", because "a/" sorts after "a.c".
542  */
543 #define TREE_UNORDERED (-1)
544 #define TREE_HAS_DUPS  (-2)
545
546 static int verify_ordered(unsigned mode1, const char *name1, unsigned mode2, const char *name2)
547 {
548         int len1 = strlen(name1);
549         int len2 = strlen(name2);
550         int len = len1 < len2 ? len1 : len2;
551         unsigned char c1, c2;
552         int cmp;
553
554         cmp = memcmp(name1, name2, len);
555         if (cmp < 0)
556                 return 0;
557         if (cmp > 0)
558                 return TREE_UNORDERED;
559
560         /*
561          * Ok, the first <len> characters are the same.
562          * Now we need to order the next one, but turn
563          * a '\0' into a '/' for a directory entry.
564          */
565         c1 = name1[len];
566         c2 = name2[len];
567         if (!c1 && !c2)
568                 /*
569                  * git-write-tree used to write out a nonsense tree that has
570                  * entries with the same name, one blob and one tree.  Make
571                  * sure we do not have duplicate entries.
572                  */
573                 return TREE_HAS_DUPS;
574         if (!c1 && S_ISDIR(mode1))
575                 c1 = '/';
576         if (!c2 && S_ISDIR(mode2))
577                 c2 = '/';
578         return c1 < c2 ? 0 : TREE_UNORDERED;
579 }
580
581 static int fsck_tree(struct tree *item, struct fsck_options *options)
582 {
583         int retval = 0;
584         int has_null_sha1 = 0;
585         int has_full_path = 0;
586         int has_empty_name = 0;
587         int has_dot = 0;
588         int has_dotdot = 0;
589         int has_dotgit = 0;
590         int has_zero_pad = 0;
591         int has_bad_modes = 0;
592         int has_dup_entries = 0;
593         int not_properly_sorted = 0;
594         struct tree_desc desc;
595         unsigned o_mode;
596         const char *o_name;
597
598         if (init_tree_desc_gently(&desc, item->buffer, item->size)) {
599                 retval += report(options, &item->object, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
600                 return retval;
601         }
602
603         o_mode = 0;
604         o_name = NULL;
605
606         while (desc.size) {
607                 unsigned short mode;
608                 const char *name;
609                 const struct object_id *oid;
610
611                 oid = tree_entry_extract(&desc, &name, &mode);
612
613                 has_null_sha1 |= is_null_oid(oid);
614                 has_full_path |= !!strchr(name, '/');
615                 has_empty_name |= !*name;
616                 has_dot |= !strcmp(name, ".");
617                 has_dotdot |= !strcmp(name, "..");
618                 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
619                 has_zero_pad |= *(char *)desc.buffer == '0';
620
621                 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
622                         if (!S_ISLNK(mode))
623                                 oidset_insert(&gitmodules_found, oid);
624                         else
625                                 retval += report(options, &item->object,
626                                                  FSCK_MSG_GITMODULES_SYMLINK,
627                                                  ".gitmodules is a symbolic link");
628                 }
629
630                 if (update_tree_entry_gently(&desc)) {
631                         retval += report(options, &item->object, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
632                         break;
633                 }
634
635                 switch (mode) {
636                 /*
637                  * Standard modes..
638                  */
639                 case S_IFREG | 0755:
640                 case S_IFREG | 0644:
641                 case S_IFLNK:
642                 case S_IFDIR:
643                 case S_IFGITLINK:
644                         break;
645                 /*
646                  * This is nonstandard, but we had a few of these
647                  * early on when we honored the full set of mode
648                  * bits..
649                  */
650                 case S_IFREG | 0664:
651                         if (!options->strict)
652                                 break;
653                         /* fallthrough */
654                 default:
655                         has_bad_modes = 1;
656                 }
657
658                 if (o_name) {
659                         switch (verify_ordered(o_mode, o_name, mode, name)) {
660                         case TREE_UNORDERED:
661                                 not_properly_sorted = 1;
662                                 break;
663                         case TREE_HAS_DUPS:
664                                 has_dup_entries = 1;
665                                 break;
666                         default:
667                                 break;
668                         }
669                 }
670
671                 o_mode = mode;
672                 o_name = name;
673         }
674
675         if (has_null_sha1)
676                 retval += report(options, &item->object, FSCK_MSG_NULL_SHA1, "contains entries pointing to null sha1");
677         if (has_full_path)
678                 retval += report(options, &item->object, FSCK_MSG_FULL_PATHNAME, "contains full pathnames");
679         if (has_empty_name)
680                 retval += report(options, &item->object, FSCK_MSG_EMPTY_NAME, "contains empty pathname");
681         if (has_dot)
682                 retval += report(options, &item->object, FSCK_MSG_HAS_DOT, "contains '.'");
683         if (has_dotdot)
684                 retval += report(options, &item->object, FSCK_MSG_HAS_DOTDOT, "contains '..'");
685         if (has_dotgit)
686                 retval += report(options, &item->object, FSCK_MSG_HAS_DOTGIT, "contains '.git'");
687         if (has_zero_pad)
688                 retval += report(options, &item->object, FSCK_MSG_ZERO_PADDED_FILEMODE, "contains zero-padded file modes");
689         if (has_bad_modes)
690                 retval += report(options, &item->object, FSCK_MSG_BAD_FILEMODE, "contains bad file modes");
691         if (has_dup_entries)
692                 retval += report(options, &item->object, FSCK_MSG_DUPLICATE_ENTRIES, "contains duplicate file entries");
693         if (not_properly_sorted)
694                 retval += report(options, &item->object, FSCK_MSG_TREE_NOT_SORTED, "not properly sorted");
695         return retval;
696 }
697
698 static int verify_headers(const void *data, unsigned long size,
699                           struct object *obj, struct fsck_options *options)
700 {
701         const char *buffer = (const char *)data;
702         unsigned long i;
703
704         for (i = 0; i < size; i++) {
705                 switch (buffer[i]) {
706                 case '\0':
707                         return report(options, obj,
708                                 FSCK_MSG_NUL_IN_HEADER,
709                                 "unterminated header: NUL at offset %ld", i);
710                 case '\n':
711                         if (i + 1 < size && buffer[i + 1] == '\n')
712                                 return 0;
713                 }
714         }
715
716         /*
717          * We did not find double-LF that separates the header
718          * and the body.  Not having a body is not a crime but
719          * we do want to see the terminating LF for the last header
720          * line.
721          */
722         if (size && buffer[size - 1] == '\n')
723                 return 0;
724
725         return report(options, obj,
726                 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
727 }
728
729 static int fsck_ident(const char **ident, struct object *obj, struct fsck_options *options)
730 {
731         const char *p = *ident;
732         char *end;
733
734         *ident = strchrnul(*ident, '\n');
735         if (**ident == '\n')
736                 (*ident)++;
737
738         if (*p == '<')
739                 return report(options, obj, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
740         p += strcspn(p, "<>\n");
741         if (*p == '>')
742                 return report(options, obj, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
743         if (*p != '<')
744                 return report(options, obj, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
745         if (p[-1] != ' ')
746                 return report(options, obj, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
747         p++;
748         p += strcspn(p, "<>\n");
749         if (*p != '>')
750                 return report(options, obj, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
751         p++;
752         if (*p != ' ')
753                 return report(options, obj, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
754         p++;
755         if (*p == '0' && p[1] != ' ')
756                 return report(options, obj, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
757         if (date_overflows(parse_timestamp(p, &end, 10)))
758                 return report(options, obj, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
759         if ((end == p || *end != ' '))
760                 return report(options, obj, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
761         p = end + 1;
762         if ((*p != '+' && *p != '-') ||
763             !isdigit(p[1]) ||
764             !isdigit(p[2]) ||
765             !isdigit(p[3]) ||
766             !isdigit(p[4]) ||
767             (p[5] != '\n'))
768                 return report(options, obj, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
769         p += 6;
770         return 0;
771 }
772
773 static int fsck_commit_buffer(struct commit *commit, const char *buffer,
774         unsigned long size, struct fsck_options *options)
775 {
776         struct object_id tree_oid, oid;
777         struct commit_graft *graft;
778         unsigned parent_count, parent_line_count = 0, author_count;
779         int err;
780         const char *buffer_begin = buffer;
781         const char *p;
782
783         if (verify_headers(buffer, size, &commit->object, options))
784                 return -1;
785
786         if (!skip_prefix(buffer, "tree ", &buffer))
787                 return report(options, &commit->object, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
788         if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
789                 err = report(options, &commit->object, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
790                 if (err)
791                         return err;
792         }
793         buffer = p + 1;
794         while (skip_prefix(buffer, "parent ", &buffer)) {
795                 if (parse_oid_hex(buffer, &oid, &p) || *p != '\n') {
796                         err = report(options, &commit->object, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
797                         if (err)
798                                 return err;
799                 }
800                 buffer = p + 1;
801                 parent_line_count++;
802         }
803         graft = lookup_commit_graft(the_repository, &commit->object.oid);
804         parent_count = commit_list_count(commit->parents);
805         if (graft) {
806                 if (graft->nr_parent == -1 && !parent_count)
807                         ; /* shallow commit */
808                 else if (graft->nr_parent != parent_count) {
809                         err = report(options, &commit->object, FSCK_MSG_MISSING_GRAFT, "graft objects missing");
810                         if (err)
811                                 return err;
812                 }
813         } else {
814                 if (parent_count != parent_line_count) {
815                         err = report(options, &commit->object, FSCK_MSG_MISSING_PARENT, "parent objects missing");
816                         if (err)
817                                 return err;
818                 }
819         }
820         author_count = 0;
821         while (skip_prefix(buffer, "author ", &buffer)) {
822                 author_count++;
823                 err = fsck_ident(&buffer, &commit->object, options);
824                 if (err)
825                         return err;
826         }
827         if (author_count < 1)
828                 err = report(options, &commit->object, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
829         else if (author_count > 1)
830                 err = report(options, &commit->object, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
831         if (err)
832                 return err;
833         if (!skip_prefix(buffer, "committer ", &buffer))
834                 return report(options, &commit->object, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
835         err = fsck_ident(&buffer, &commit->object, options);
836         if (err)
837                 return err;
838         if (!get_commit_tree(commit)) {
839                 err = report(options, &commit->object, FSCK_MSG_BAD_TREE, "could not load commit's tree %s", oid_to_hex(&tree_oid));
840                 if (err)
841                         return err;
842         }
843         if (memchr(buffer_begin, '\0', size)) {
844                 err = report(options, &commit->object, FSCK_MSG_NUL_IN_COMMIT,
845                              "NUL byte in the commit object body");
846                 if (err)
847                         return err;
848         }
849         return 0;
850 }
851
852 static int fsck_commit(struct commit *commit, const char *data,
853         unsigned long size, struct fsck_options *options)
854 {
855         const char *buffer = data ?  data : get_commit_buffer(commit, &size);
856         int ret = fsck_commit_buffer(commit, buffer, size, options);
857         if (!data)
858                 unuse_commit_buffer(commit, buffer);
859         return ret;
860 }
861
862 static int fsck_tag_buffer(struct tag *tag, const char *data,
863         unsigned long size, struct fsck_options *options)
864 {
865         struct object_id oid;
866         int ret = 0;
867         const char *buffer;
868         char *to_free = NULL, *eol;
869         struct strbuf sb = STRBUF_INIT;
870         const char *p;
871
872         if (data)
873                 buffer = data;
874         else {
875                 enum object_type type;
876
877                 buffer = to_free =
878                         read_object_file(&tag->object.oid, &type, &size);
879                 if (!buffer)
880                         return report(options, &tag->object,
881                                 FSCK_MSG_MISSING_TAG_OBJECT,
882                                 "cannot read tag object");
883
884                 if (type != OBJ_TAG) {
885                         ret = report(options, &tag->object,
886                                 FSCK_MSG_TAG_OBJECT_NOT_TAG,
887                                 "expected tag got %s",
888                             type_name(type));
889                         goto done;
890                 }
891         }
892
893         ret = verify_headers(buffer, size, &tag->object, options);
894         if (ret)
895                 goto done;
896
897         if (!skip_prefix(buffer, "object ", &buffer)) {
898                 ret = report(options, &tag->object, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
899                 goto done;
900         }
901         if (parse_oid_hex(buffer, &oid, &p) || *p != '\n') {
902                 ret = report(options, &tag->object, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
903                 if (ret)
904                         goto done;
905         }
906         buffer = p + 1;
907
908         if (!skip_prefix(buffer, "type ", &buffer)) {
909                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
910                 goto done;
911         }
912         eol = strchr(buffer, '\n');
913         if (!eol) {
914                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
915                 goto done;
916         }
917         if (type_from_string_gently(buffer, eol - buffer, 1) < 0)
918                 ret = report(options, &tag->object, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
919         if (ret)
920                 goto done;
921         buffer = eol + 1;
922
923         if (!skip_prefix(buffer, "tag ", &buffer)) {
924                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
925                 goto done;
926         }
927         eol = strchr(buffer, '\n');
928         if (!eol) {
929                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
930                 goto done;
931         }
932         strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
933         if (check_refname_format(sb.buf, 0)) {
934                 ret = report(options, &tag->object, FSCK_MSG_BAD_TAG_NAME,
935                            "invalid 'tag' name: %.*s",
936                            (int)(eol - buffer), buffer);
937                 if (ret)
938                         goto done;
939         }
940         buffer = eol + 1;
941
942         if (!skip_prefix(buffer, "tagger ", &buffer)) {
943                 /* early tags do not contain 'tagger' lines; warn only */
944                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
945                 if (ret)
946                         goto done;
947         }
948         else
949                 ret = fsck_ident(&buffer, &tag->object, options);
950
951 done:
952         strbuf_release(&sb);
953         free(to_free);
954         return ret;
955 }
956
957 static int fsck_tag(struct tag *tag, const char *data,
958         unsigned long size, struct fsck_options *options)
959 {
960         struct object *tagged = tag->tagged;
961
962         if (!tagged)
963                 return report(options, &tag->object, FSCK_MSG_BAD_TAG_OBJECT, "could not load tagged object");
964
965         return fsck_tag_buffer(tag, data, size, options);
966 }
967
968 struct fsck_gitmodules_data {
969         struct object *obj;
970         struct fsck_options *options;
971         int ret;
972 };
973
974 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
975 {
976         struct fsck_gitmodules_data *data = vdata;
977         const char *subsection, *key;
978         int subsection_len;
979         char *name;
980
981         if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
982             !subsection)
983                 return 0;
984
985         name = xmemdupz(subsection, subsection_len);
986         if (check_submodule_name(name) < 0)
987                 data->ret |= report(data->options, data->obj,
988                                     FSCK_MSG_GITMODULES_NAME,
989                                     "disallowed submodule name: %s",
990                                     name);
991         if (!strcmp(key, "url") && value &&
992             looks_like_command_line_option(value))
993                 data->ret |= report(data->options, data->obj,
994                                     FSCK_MSG_GITMODULES_URL,
995                                     "disallowed submodule url: %s",
996                                     value);
997         if (!strcmp(key, "path") && value &&
998             looks_like_command_line_option(value))
999                 data->ret |= report(data->options, data->obj,
1000                                     FSCK_MSG_GITMODULES_PATH,
1001                                     "disallowed submodule path: %s",
1002                                     value);
1003         free(name);
1004
1005         return 0;
1006 }
1007
1008 static int fsck_blob(struct blob *blob, const char *buf,
1009                      unsigned long size, struct fsck_options *options)
1010 {
1011         struct fsck_gitmodules_data data;
1012         struct config_options config_opts = { 0 };
1013
1014         if (!oidset_contains(&gitmodules_found, &blob->object.oid))
1015                 return 0;
1016         oidset_insert(&gitmodules_done, &blob->object.oid);
1017
1018         if (object_on_skiplist(options, &blob->object))
1019                 return 0;
1020
1021         if (!buf) {
1022                 /*
1023                  * A missing buffer here is a sign that the caller found the
1024                  * blob too gigantic to load into memory. Let's just consider
1025                  * that an error.
1026                  */
1027                 return report(options, &blob->object,
1028                               FSCK_MSG_GITMODULES_LARGE,
1029                               ".gitmodules too large to parse");
1030         }
1031
1032         data.obj = &blob->object;
1033         data.options = options;
1034         data.ret = 0;
1035         config_opts.error_action = CONFIG_ERROR_SILENT;
1036         if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1037                                 ".gitmodules", buf, size, &data, &config_opts))
1038                 data.ret |= report(options, &blob->object,
1039                                    FSCK_MSG_GITMODULES_PARSE,
1040                                    "could not parse gitmodules blob");
1041
1042         return data.ret;
1043 }
1044
1045 int fsck_object(struct object *obj, void *data, unsigned long size,
1046         struct fsck_options *options)
1047 {
1048         if (!obj)
1049                 return report(options, obj, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1050
1051         if (obj->type == OBJ_BLOB)
1052                 return fsck_blob((struct blob *)obj, data, size, options);
1053         if (obj->type == OBJ_TREE)
1054                 return fsck_tree((struct tree *) obj, options);
1055         if (obj->type == OBJ_COMMIT)
1056                 return fsck_commit((struct commit *) obj, (const char *) data,
1057                         size, options);
1058         if (obj->type == OBJ_TAG)
1059                 return fsck_tag((struct tag *) obj, (const char *) data,
1060                         size, options);
1061
1062         return report(options, obj, FSCK_MSG_UNKNOWN_TYPE, "unknown type '%d' (internal fsck error)",
1063                           obj->type);
1064 }
1065
1066 int fsck_error_function(struct fsck_options *o,
1067         struct object *obj, int msg_type, const char *message)
1068 {
1069         if (msg_type == FSCK_WARN) {
1070                 warning("object %s: %s", describe_object(o, obj), message);
1071                 return 0;
1072         }
1073         error("object %s: %s", describe_object(o, obj), message);
1074         return 1;
1075 }
1076
1077 int fsck_finish(struct fsck_options *options)
1078 {
1079         int ret = 0;
1080         struct oidset_iter iter;
1081         const struct object_id *oid;
1082
1083         oidset_iter_init(&gitmodules_found, &iter);
1084         while ((oid = oidset_iter_next(&iter))) {
1085                 struct blob *blob;
1086                 enum object_type type;
1087                 unsigned long size;
1088                 char *buf;
1089
1090                 if (oidset_contains(&gitmodules_done, oid))
1091                         continue;
1092
1093                 blob = lookup_blob(the_repository, oid);
1094                 if (!blob) {
1095                         struct object *obj = lookup_unknown_object(oid->hash);
1096                         ret |= report(options, obj,
1097                                       FSCK_MSG_GITMODULES_BLOB,
1098                                       "non-blob found at .gitmodules");
1099                         continue;
1100                 }
1101
1102                 buf = read_object_file(oid, &type, &size);
1103                 if (!buf) {
1104                         if (is_promisor_object(&blob->object.oid))
1105                                 continue;
1106                         ret |= report(options, &blob->object,
1107                                       FSCK_MSG_GITMODULES_MISSING,
1108                                       "unable to read .gitmodules blob");
1109                         continue;
1110                 }
1111
1112                 if (type == OBJ_BLOB)
1113                         ret |= fsck_blob(blob, buf, size, options);
1114                 else
1115                         ret |= report(options, &blob->object,
1116                                       FSCK_MSG_GITMODULES_BLOB,
1117                                       "non-blob found at .gitmodules");
1118                 free(buf);
1119         }
1120
1121
1122         oidset_clear(&gitmodules_found);
1123         oidset_clear(&gitmodules_done);
1124         return ret;
1125 }