Code

parse_tag_buffer: don't parse invalid tags
[git.git] / sha1_name.c
1 #include "cache.h"
2 #include "tag.h"
3 #include "commit.h"
4 #include "tree.h"
5 #include "blob.h"
6 #include "tree-walk.h"
7 #include "refs.h"
9 static int find_short_object_filename(int len, const char *name, unsigned char *sha1)
10 {
11         struct alternate_object_database *alt;
12         char hex[40];
13         int found = 0;
14         static struct alternate_object_database *fakeent;
16         if (!fakeent) {
17                 const char *objdir = get_object_directory();
18                 int objdir_len = strlen(objdir);
19                 int entlen = objdir_len + 43;
20                 fakeent = xmalloc(sizeof(*fakeent) + entlen);
21                 memcpy(fakeent->base, objdir, objdir_len);
22                 fakeent->name = fakeent->base + objdir_len + 1;
23                 fakeent->name[-1] = '/';
24         }
25         fakeent->next = alt_odb_list;
27         sprintf(hex, "%.2s", name);
28         for (alt = fakeent; alt && found < 2; alt = alt->next) {
29                 struct dirent *de;
30                 DIR *dir;
31                 sprintf(alt->name, "%.2s/", name);
32                 dir = opendir(alt->base);
33                 if (!dir)
34                         continue;
35                 while ((de = readdir(dir)) != NULL) {
36                         if (strlen(de->d_name) != 38)
37                                 continue;
38                         if (memcmp(de->d_name, name + 2, len - 2))
39                                 continue;
40                         if (!found) {
41                                 memcpy(hex + 2, de->d_name, 38);
42                                 found++;
43                         }
44                         else if (memcmp(hex + 2, de->d_name, 38)) {
45                                 found = 2;
46                                 break;
47                         }
48                 }
49                 closedir(dir);
50         }
51         if (found == 1)
52                 return get_sha1_hex(hex, sha1) == 0;
53         return found;
54 }
56 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
57 {
58         do {
59                 if (*a != *b)
60                         return 0;
61                 a++;
62                 b++;
63                 len -= 2;
64         } while (len > 1);
65         if (len)
66                 if ((*a ^ *b) & 0xf0)
67                         return 0;
68         return 1;
69 }
71 static int find_short_packed_object(int len, const unsigned char *match, unsigned char *sha1)
72 {
73         struct packed_git *p;
74         const unsigned char *found_sha1 = NULL;
75         int found = 0;
77         prepare_packed_git();
78         for (p = packed_git; p && found < 2; p = p->next) {
79                 uint32_t num, last;
80                 uint32_t first = 0;
81                 open_pack_index(p);
82                 num = p->num_objects;
83                 last = num;
84                 while (first < last) {
85                         uint32_t mid = (first + last) / 2;
86                         const unsigned char *now;
87                         int cmp;
89                         now = nth_packed_object_sha1(p, mid);
90                         cmp = hashcmp(match, now);
91                         if (!cmp) {
92                                 first = mid;
93                                 break;
94                         }
95                         if (cmp > 0) {
96                                 first = mid+1;
97                                 continue;
98                         }
99                         last = mid;
100                 }
101                 if (first < num) {
102                         const unsigned char *now, *next;
103                        now = nth_packed_object_sha1(p, first);
104                         if (match_sha(len, match, now)) {
105                                 next = nth_packed_object_sha1(p, first+1);
106                                if (!next|| !match_sha(len, match, next)) {
107                                         /* unique within this pack */
108                                         if (!found) {
109                                                 found_sha1 = now;
110                                                 found++;
111                                         }
112                                         else if (hashcmp(found_sha1, now)) {
113                                                 found = 2;
114                                                 break;
115                                         }
116                                 }
117                                 else {
118                                         /* not even unique within this pack */
119                                         found = 2;
120                                         break;
121                                 }
122                         }
123                 }
124         }
125         if (found == 1)
126                 hashcpy(sha1, found_sha1);
127         return found;
130 #define SHORT_NAME_NOT_FOUND (-1)
131 #define SHORT_NAME_AMBIGUOUS (-2)
133 static int find_unique_short_object(int len, char *canonical,
134                                     unsigned char *res, unsigned char *sha1)
136         int has_unpacked, has_packed;
137         unsigned char unpacked_sha1[20], packed_sha1[20];
139         prepare_alt_odb();
140         has_unpacked = find_short_object_filename(len, canonical, unpacked_sha1);
141         has_packed = find_short_packed_object(len, res, packed_sha1);
142         if (!has_unpacked && !has_packed)
143                 return SHORT_NAME_NOT_FOUND;
144         if (1 < has_unpacked || 1 < has_packed)
145                 return SHORT_NAME_AMBIGUOUS;
146         if (has_unpacked != has_packed) {
147                 hashcpy(sha1, (has_packed ? packed_sha1 : unpacked_sha1));
148                 return 0;
149         }
150         /* Both have unique ones -- do they match? */
151         if (hashcmp(packed_sha1, unpacked_sha1))
152                 return SHORT_NAME_AMBIGUOUS;
153         hashcpy(sha1, packed_sha1);
154         return 0;
157 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
158                           int quietly)
160         int i, status;
161         char canonical[40];
162         unsigned char res[20];
164         if (len < MINIMUM_ABBREV || len > 40)
165                 return -1;
166         hashclr(res);
167         memset(canonical, 'x', 40);
168         for (i = 0; i < len ;i++) {
169                 unsigned char c = name[i];
170                 unsigned char val;
171                 if (c >= '0' && c <= '9')
172                         val = c - '0';
173                 else if (c >= 'a' && c <= 'f')
174                         val = c - 'a' + 10;
175                 else if (c >= 'A' && c <='F') {
176                         val = c - 'A' + 10;
177                         c -= 'A' - 'a';
178                 }
179                 else
180                         return -1;
181                 canonical[i] = c;
182                 if (!(i & 1))
183                         val <<= 4;
184                 res[i >> 1] |= val;
185         }
187         status = find_unique_short_object(i, canonical, res, sha1);
188         if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
189                 return error("short SHA1 %.*s is ambiguous.", len, canonical);
190         return status;
193 const char *find_unique_abbrev(const unsigned char *sha1, int len)
195         int status, is_null;
196         static char hex[41];
198         is_null = is_null_sha1(sha1);
199         memcpy(hex, sha1_to_hex(sha1), 40);
200         if (len == 40 || !len)
201                 return hex;
202         while (len < 40) {
203                 unsigned char sha1_ret[20];
204                 status = get_short_sha1(hex, len, sha1_ret, 1);
205                 if (!status ||
206                     (is_null && status != SHORT_NAME_AMBIGUOUS)) {
207                         hex[len] = 0;
208                         return hex;
209                 }
210                 if (status != SHORT_NAME_AMBIGUOUS)
211                         return NULL;
212                 len++;
213         }
214         return NULL;
217 static int ambiguous_path(const char *path, int len)
219         int slash = 1;
220         int cnt;
222         for (cnt = 0; cnt < len; cnt++) {
223                 switch (*path++) {
224                 case '\0':
225                         break;
226                 case '/':
227                         if (slash)
228                                 break;
229                         slash = 1;
230                         continue;
231                 case '.':
232                         continue;
233                 default:
234                         slash = 0;
235                         continue;
236                 }
237                 break;
238         }
239         return slash;
242 int dwim_ref(const char *str, int len, unsigned char *sha1, char **ref)
244         const char **p, *r;
245         int refs_found = 0;
247         *ref = NULL;
248         for (p = ref_rev_parse_rules; *p; p++) {
249                 unsigned char sha1_from_ref[20];
250                 unsigned char *this_result;
252                 this_result = refs_found ? sha1_from_ref : sha1;
253                 r = resolve_ref(mkpath(*p, len, str), this_result, 1, NULL);
254                 if (r) {
255                         if (!refs_found++)
256                                 *ref = xstrdup(r);
257                         if (!warn_ambiguous_refs)
258                                 break;
259                 }
260         }
261         return refs_found;
264 int dwim_log(const char *str, int len, unsigned char *sha1, char **log)
266         const char **p;
267         int logs_found = 0;
269         *log = NULL;
270         for (p = ref_rev_parse_rules; *p; p++) {
271                 struct stat st;
272                 unsigned char hash[20];
273                 char path[PATH_MAX];
274                 const char *ref, *it;
276                 strcpy(path, mkpath(*p, len, str));
277                 ref = resolve_ref(path, hash, 0, NULL);
278                 if (!ref)
279                         continue;
280                 if (!stat(git_path("logs/%s", path), &st) &&
281                     S_ISREG(st.st_mode))
282                         it = path;
283                 else if (strcmp(ref, path) &&
284                          !stat(git_path("logs/%s", ref), &st) &&
285                          S_ISREG(st.st_mode))
286                         it = ref;
287                 else
288                         continue;
289                 if (!logs_found++) {
290                         *log = xstrdup(it);
291                         hashcpy(sha1, hash);
292                 }
293                 if (!warn_ambiguous_refs)
294                         break;
295         }
296         return logs_found;
299 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
301         static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
302         char *real_ref = NULL;
303         int refs_found = 0;
304         int at, reflog_len;
306         if (len == 40 && !get_sha1_hex(str, sha1))
307                 return 0;
309         /* basic@{time or number} format to query ref-log */
310         reflog_len = at = 0;
311         if (str[len-1] == '}') {
312                 for (at = 0; at < len - 1; at++) {
313                         if (str[at] == '@' && str[at+1] == '{') {
314                                 reflog_len = (len-1) - (at+2);
315                                 len = at;
316                                 break;
317                         }
318                 }
319         }
321         /* Accept only unambiguous ref paths. */
322         if (len && ambiguous_path(str, len))
323                 return -1;
325         if (!len && reflog_len) {
326                 /* allow "@{...}" to mean the current branch reflog */
327                 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
328         } else if (reflog_len)
329                 refs_found = dwim_log(str, len, sha1, &real_ref);
330         else
331                 refs_found = dwim_ref(str, len, sha1, &real_ref);
333         if (!refs_found)
334                 return -1;
336         if (warn_ambiguous_refs && refs_found > 1)
337                 fprintf(stderr, warning, len, str);
339         if (reflog_len) {
340                 int nth, i;
341                 unsigned long at_time;
342                 unsigned long co_time;
343                 int co_tz, co_cnt;
345                 /* Is it asking for N-th entry, or approxidate? */
346                 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
347                         char ch = str[at+2+i];
348                         if ('0' <= ch && ch <= '9')
349                                 nth = nth * 10 + ch - '0';
350                         else
351                                 nth = -1;
352                 }
353                 if (0 <= nth)
354                         at_time = 0;
355                 else
356                         at_time = approxidate(str + at + 2);
357                 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
358                                 &co_time, &co_tz, &co_cnt)) {
359                         if (at_time)
360                                 fprintf(stderr,
361                                         "warning: Log for '%.*s' only goes "
362                                         "back to %s.\n", len, str,
363                                         show_date(co_time, co_tz, DATE_RFC2822));
364                         else
365                                 fprintf(stderr,
366                                         "warning: Log for '%.*s' only has "
367                                         "%d entries.\n", len, str, co_cnt);
368                 }
369         }
371         free(real_ref);
372         return 0;
375 static int get_sha1_1(const char *name, int len, unsigned char *sha1);
377 static int get_parent(const char *name, int len,
378                       unsigned char *result, int idx)
380         unsigned char sha1[20];
381         int ret = get_sha1_1(name, len, sha1);
382         struct commit *commit;
383         struct commit_list *p;
385         if (ret)
386                 return ret;
387         commit = lookup_commit_reference(sha1);
388         if (!commit)
389                 return -1;
390         if (parse_commit(commit))
391                 return -1;
392         if (!idx) {
393                 hashcpy(result, commit->object.sha1);
394                 return 0;
395         }
396         p = commit->parents;
397         while (p) {
398                 if (!--idx) {
399                         hashcpy(result, p->item->object.sha1);
400                         return 0;
401                 }
402                 p = p->next;
403         }
404         return -1;
407 static int get_nth_ancestor(const char *name, int len,
408                             unsigned char *result, int generation)
410         unsigned char sha1[20];
411         int ret = get_sha1_1(name, len, sha1);
412         if (ret)
413                 return ret;
415         while (generation--) {
416                 struct commit *commit = lookup_commit_reference(sha1);
418                 if (!commit || parse_commit(commit) || !commit->parents)
419                         return -1;
420                 hashcpy(sha1, commit->parents->item->object.sha1);
421         }
422         hashcpy(result, sha1);
423         return 0;
426 static int peel_onion(const char *name, int len, unsigned char *sha1)
428         unsigned char outer[20];
429         const char *sp;
430         unsigned int expected_type = 0;
431         struct object *o;
433         /*
434          * "ref^{type}" dereferences ref repeatedly until you cannot
435          * dereference anymore, or you get an object of given type,
436          * whichever comes first.  "ref^{}" means just dereference
437          * tags until you get a non-tag.  "ref^0" is a shorthand for
438          * "ref^{commit}".  "commit^{tree}" could be used to find the
439          * top-level tree of the given commit.
440          */
441         if (len < 4 || name[len-1] != '}')
442                 return -1;
444         for (sp = name + len - 1; name <= sp; sp--) {
445                 int ch = *sp;
446                 if (ch == '{' && name < sp && sp[-1] == '^')
447                         break;
448         }
449         if (sp <= name)
450                 return -1;
452         sp++; /* beginning of type name, or closing brace for empty */
453         if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
454                 expected_type = OBJ_COMMIT;
455         else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
456                 expected_type = OBJ_TREE;
457         else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
458                 expected_type = OBJ_BLOB;
459         else if (sp[0] == '}')
460                 expected_type = OBJ_NONE;
461         else
462                 return -1;
464         if (get_sha1_1(name, sp - name - 2, outer))
465                 return -1;
467         o = parse_object(outer);
468         if (!o)
469                 return -1;
470         if (!expected_type) {
471                 o = deref_tag(o, name, sp - name - 2);
472                 if (!o || (!o->parsed && !parse_object(o->sha1)))
473                         return -1;
474                 hashcpy(sha1, o->sha1);
475         }
476         else {
477                 /* At this point, the syntax look correct, so
478                  * if we do not get the needed object, we should
479                  * barf.
480                  */
482                 while (1) {
483                         if (!o || (!o->parsed && !parse_object(o->sha1)))
484                                 return -1;
485                         if (o->type == expected_type) {
486                                 hashcpy(sha1, o->sha1);
487                                 return 0;
488                         }
489                         if (o->type == OBJ_TAG)
490                                 o = ((struct tag*) o)->tagged;
491                         else if (o->type == OBJ_COMMIT)
492                                 o = &(((struct commit *) o)->tree->object);
493                         else
494                                 return error("%.*s: expected %s type, but the object dereferences to %s type",
495                                              len, name, typename(expected_type),
496                                              typename(o->type));
497                         if (!o->parsed)
498                                 parse_object(o->sha1);
499                 }
500         }
501         return 0;
504 static int get_describe_name(const char *name, int len, unsigned char *sha1)
506         const char *cp;
508         for (cp = name + len - 1; name + 2 <= cp; cp--) {
509                 char ch = *cp;
510                 if (hexval(ch) & ~0377) {
511                         /* We must be looking at g in "SOMETHING-g"
512                          * for it to be describe output.
513                          */
514                         if (ch == 'g' && cp[-1] == '-') {
515                                 cp++;
516                                 len -= cp - name;
517                                 return get_short_sha1(cp, len, sha1, 1);
518                         }
519                 }
520         }
521         return -1;
524 static int get_sha1_1(const char *name, int len, unsigned char *sha1)
526         int ret, has_suffix;
527         const char *cp;
529         /* "name~3" is "name^^^",
530          * "name~" and "name~0" are name -- not "name^0"!
531          * "name^" is not "name^0"; it is "name^1".
532          */
533         has_suffix = 0;
534         for (cp = name + len - 1; name <= cp; cp--) {
535                 int ch = *cp;
536                 if ('0' <= ch && ch <= '9')
537                         continue;
538                 if (ch == '~' || ch == '^')
539                         has_suffix = ch;
540                 break;
541         }
543         if (has_suffix) {
544                 int num = 0;
545                 int len1 = cp - name;
546                 cp++;
547                 while (cp < name + len)
548                         num = num * 10 + *cp++ - '0';
549                 if (has_suffix == '^') {
550                         if (!num && len1 == len - 1)
551                                 num = 1;
552                         return get_parent(name, len1, sha1, num);
553                 }
554                 /* else if (has_suffix == '~') -- goes without saying */
555                 return get_nth_ancestor(name, len1, sha1, num);
556         }
558         ret = peel_onion(name, len, sha1);
559         if (!ret)
560                 return 0;
562         ret = get_sha1_basic(name, len, sha1);
563         if (!ret)
564                 return 0;
566         /* It could be describe output that is "SOMETHING-gXXXX" */
567         ret = get_describe_name(name, len, sha1);
568         if (!ret)
569                 return 0;
571         return get_short_sha1(name, len, sha1, 0);
574 static int handle_one_ref(const char *path,
575                 const unsigned char *sha1, int flag, void *cb_data)
577         struct commit_list **list = cb_data;
578         struct object *object = parse_object(sha1);
579         if (!object)
580                 return 0;
581         if (object->type == OBJ_TAG)
582                 object = deref_tag(object, path, strlen(path));
583         if (object->type != OBJ_COMMIT)
584                 return 0;
585         insert_by_date((struct commit *)object, list);
586         return 0;
589 /*
590  * This interprets names like ':/Initial revision of "git"' by searching
591  * through history and returning the first commit whose message starts
592  * with the given string.
593  *
594  * For future extension, ':/!' is reserved. If you want to match a message
595  * beginning with a '!', you have to repeat the exclamation mark.
596  */
598 #define ONELINE_SEEN (1u<<20)
599 static int get_sha1_oneline(const char *prefix, unsigned char *sha1)
601         struct commit_list *list = NULL, *backup = NULL, *l;
602         int retval = -1;
603         char *temp_commit_buffer = NULL;
605         if (prefix[0] == '!') {
606                 if (prefix[1] != '!')
607                         die ("Invalid search pattern: %s", prefix);
608                 prefix++;
609         }
610         for_each_ref(handle_one_ref, &list);
611         for (l = list; l; l = l->next)
612                 commit_list_insert(l->item, &backup);
613         while (list) {
614                 char *p;
615                 struct commit *commit;
616                 enum object_type type;
617                 unsigned long size;
619                 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
620                 parse_object(commit->object.sha1);
621                 if (temp_commit_buffer)
622                         free(temp_commit_buffer);
623                 if (commit->buffer)
624                         p = commit->buffer;
625                 else {
626                         p = read_sha1_file(commit->object.sha1, &type, &size);
627                         if (!p)
628                                 continue;
629                         temp_commit_buffer = p;
630                 }
631                 if (!(p = strstr(p, "\n\n")))
632                         continue;
633                 if (!prefixcmp(p + 2, prefix)) {
634                         hashcpy(sha1, commit->object.sha1);
635                         retval = 0;
636                         break;
637                 }
638         }
639         if (temp_commit_buffer)
640                 free(temp_commit_buffer);
641         free_commit_list(list);
642         for (l = backup; l; l = l->next)
643                 clear_commit_marks(l->item, ONELINE_SEEN);
644         return retval;
647 /*
648  * This is like "get_sha1_basic()", except it allows "sha1 expressions",
649  * notably "xyz^" for "parent of xyz"
650  */
651 int get_sha1(const char *name, unsigned char *sha1)
653         unsigned unused;
654         return get_sha1_with_mode(name, sha1, &unused);
657 int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
659         int ret, bracket_depth;
660         int namelen = strlen(name);
661         const char *cp;
663         *mode = S_IFINVALID;
664         ret = get_sha1_1(name, namelen, sha1);
665         if (!ret)
666                 return ret;
667         /* sha1:path --> object name of path in ent sha1
668          * :path -> object name of path in index
669          * :[0-3]:path -> object name of path in index at stage
670          */
671         if (name[0] == ':') {
672                 int stage = 0;
673                 struct cache_entry *ce;
674                 int pos;
675                 if (namelen > 2 && name[1] == '/')
676                         return get_sha1_oneline(name + 2, sha1);
677                 if (namelen < 3 ||
678                     name[2] != ':' ||
679                     name[1] < '0' || '3' < name[1])
680                         cp = name + 1;
681                 else {
682                         stage = name[1] - '0';
683                         cp = name + 3;
684                 }
685                 namelen = namelen - (cp - name);
686                 if (!active_cache)
687                         read_cache();
688                 pos = cache_name_pos(cp, namelen);
689                 if (pos < 0)
690                         pos = -pos - 1;
691                 while (pos < active_nr) {
692                         ce = active_cache[pos];
693                         if (ce_namelen(ce) != namelen ||
694                             memcmp(ce->name, cp, namelen))
695                                 break;
696                         if (ce_stage(ce) == stage) {
697                                 hashcpy(sha1, ce->sha1);
698                                 *mode = ntohl(ce->ce_mode);
699                                 return 0;
700                         }
701                         pos++;
702                 }
703                 return -1;
704         }
705         for (cp = name, bracket_depth = 0; *cp; cp++) {
706                 if (*cp == '{')
707                         bracket_depth++;
708                 else if (bracket_depth && *cp == '}')
709                         bracket_depth--;
710                 else if (!bracket_depth && *cp == ':')
711                         break;
712         }
713         if (*cp == ':') {
714                 unsigned char tree_sha1[20];
715                 if (!get_sha1_1(name, cp-name, tree_sha1))
716                         return get_tree_entry(tree_sha1, cp+1, sha1,
717                                               mode);
718         }
719         return ret;