git-fast-import possible memory corruption problem
[git/msvc.git] / sha1_name.c
blob159c2ab84fa2cdde0e540024a1ca22e0bbb43af8
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)
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] = '/';
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++;
44 else if (memcmp(hex + 2, de->d_name, 38)) {
45 found = 2;
46 break;
49 closedir(dir);
51 if (found == 1)
52 return get_sha1_hex(hex, sha1) == 0;
53 return found;
56 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
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;
71 static int find_short_packed_object(int len, const unsigned char *match, unsigned char *sha1)
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;
95 if (cmp > 0) {
96 first = mid+1;
97 continue;
99 last = mid;
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++;
112 else if (hashcmp(found_sha1, now)) {
113 found = 2;
114 break;
117 else {
118 /* not even unique within this pack */
119 found = 2;
120 break;
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;
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';
179 else
180 return -1;
181 canonical[i] = c;
182 if (!(i & 1))
183 val <<= 4;
184 res[i >> 1] |= val;
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, exists;
196 static char hex[41];
198 exists = has_sha1_file(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 (exists
206 ? !status
207 : status == SHORT_NAME_NOT_FOUND) {
208 hex[len] = 0;
209 return hex;
211 len++;
213 return hex;
216 static int ambiguous_path(const char *path, int len)
218 int slash = 1;
219 int cnt;
221 for (cnt = 0; cnt < len; cnt++) {
222 switch (*path++) {
223 case '\0':
224 break;
225 case '/':
226 if (slash)
227 break;
228 slash = 1;
229 continue;
230 case '.':
231 continue;
232 default:
233 slash = 0;
234 continue;
236 break;
238 return slash;
241 int dwim_ref(const char *str, int len, unsigned char *sha1, char **ref)
243 const char **p, *r;
244 int refs_found = 0;
246 *ref = NULL;
247 for (p = ref_rev_parse_rules; *p; p++) {
248 char fullref[PATH_MAX];
249 unsigned char sha1_from_ref[20];
250 unsigned char *this_result;
252 this_result = refs_found ? sha1_from_ref : sha1;
253 mksnpath(fullref, sizeof(fullref), *p, len, str);
254 r = resolve_ref(fullref, this_result, 1, NULL);
255 if (r) {
256 if (!refs_found++)
257 *ref = xstrdup(r);
258 if (!warn_ambiguous_refs)
259 break;
262 return refs_found;
265 int dwim_log(const char *str, int len, unsigned char *sha1, char **log)
267 const char **p;
268 int logs_found = 0;
270 *log = NULL;
271 for (p = ref_rev_parse_rules; *p; p++) {
272 struct stat st;
273 unsigned char hash[20];
274 char path[PATH_MAX];
275 const char *ref, *it;
277 mksnpath(path, sizeof(path), *p, len, str);
278 ref = resolve_ref(path, hash, 1, NULL);
279 if (!ref)
280 continue;
281 if (!stat(git_path("logs/%s", path), &st) &&
282 S_ISREG(st.st_mode))
283 it = path;
284 else if (strcmp(ref, path) &&
285 !stat(git_path("logs/%s", ref), &st) &&
286 S_ISREG(st.st_mode))
287 it = ref;
288 else
289 continue;
290 if (!logs_found++) {
291 *log = xstrdup(it);
292 hashcpy(sha1, hash);
294 if (!warn_ambiguous_refs)
295 break;
297 return logs_found;
300 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
302 static const char *warning = "warning: refname '%.*s' is ambiguous.\n";
303 char *real_ref = NULL;
304 int refs_found = 0;
305 int at, reflog_len;
307 if (len == 40 && !get_sha1_hex(str, sha1))
308 return 0;
310 /* basic@{time or number} format to query ref-log */
311 reflog_len = at = 0;
312 if (str[len-1] == '}') {
313 for (at = 0; at < len - 1; at++) {
314 if (str[at] == '@' && str[at+1] == '{') {
315 reflog_len = (len-1) - (at+2);
316 len = at;
317 break;
322 /* Accept only unambiguous ref paths. */
323 if (len && ambiguous_path(str, len))
324 return -1;
326 if (!len && reflog_len) {
327 /* allow "@{...}" to mean the current branch reflog */
328 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
329 } else if (reflog_len)
330 refs_found = dwim_log(str, len, sha1, &real_ref);
331 else
332 refs_found = dwim_ref(str, len, sha1, &real_ref);
334 if (!refs_found)
335 return -1;
337 if (warn_ambiguous_refs && refs_found > 1)
338 fprintf(stderr, warning, len, str);
340 if (reflog_len) {
341 int nth, i;
342 unsigned long at_time;
343 unsigned long co_time;
344 int co_tz, co_cnt;
346 /* Is it asking for N-th entry, or approxidate? */
347 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
348 char ch = str[at+2+i];
349 if ('0' <= ch && ch <= '9')
350 nth = nth * 10 + ch - '0';
351 else
352 nth = -1;
354 if (100000000 <= nth) {
355 at_time = nth;
356 nth = -1;
357 } else if (0 <= nth)
358 at_time = 0;
359 else {
360 char *tmp = xstrndup(str + at + 2, reflog_len);
361 at_time = approxidate(tmp);
362 free(tmp);
364 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
365 &co_time, &co_tz, &co_cnt)) {
366 if (at_time)
367 fprintf(stderr,
368 "warning: Log for '%.*s' only goes "
369 "back to %s.\n", len, str,
370 show_date(co_time, co_tz, DATE_RFC2822));
371 else
372 fprintf(stderr,
373 "warning: Log for '%.*s' only has "
374 "%d entries.\n", len, str, co_cnt);
378 free(real_ref);
379 return 0;
382 static int get_sha1_1(const char *name, int len, unsigned char *sha1);
384 static int get_parent(const char *name, int len,
385 unsigned char *result, int idx)
387 unsigned char sha1[20];
388 int ret = get_sha1_1(name, len, sha1);
389 struct commit *commit;
390 struct commit_list *p;
392 if (ret)
393 return ret;
394 commit = lookup_commit_reference(sha1);
395 if (!commit)
396 return -1;
397 if (parse_commit(commit))
398 return -1;
399 if (!idx) {
400 hashcpy(result, commit->object.sha1);
401 return 0;
403 p = commit->parents;
404 while (p) {
405 if (!--idx) {
406 hashcpy(result, p->item->object.sha1);
407 return 0;
409 p = p->next;
411 return -1;
414 static int get_nth_ancestor(const char *name, int len,
415 unsigned char *result, int generation)
417 unsigned char sha1[20];
418 struct commit *commit;
419 int ret;
421 ret = get_sha1_1(name, len, sha1);
422 if (ret)
423 return ret;
424 commit = lookup_commit_reference(sha1);
425 if (!commit)
426 return -1;
428 while (generation--) {
429 if (parse_commit(commit) || !commit->parents)
430 return -1;
431 commit = commit->parents->item;
433 hashcpy(result, commit->object.sha1);
434 return 0;
437 struct object *peel_to_type(const char *name, int namelen,
438 struct object *o, enum object_type expected_type)
440 if (name && !namelen)
441 namelen = strlen(name);
442 if (!o) {
443 unsigned char sha1[20];
444 if (get_sha1_1(name, namelen, sha1))
445 return NULL;
446 o = parse_object(sha1);
448 while (1) {
449 if (!o || (!o->parsed && !parse_object(o->sha1)))
450 return NULL;
451 if (o->type == expected_type)
452 return o;
453 if (o->type == OBJ_TAG)
454 o = ((struct tag*) o)->tagged;
455 else if (o->type == OBJ_COMMIT)
456 o = &(((struct commit *) o)->tree->object);
457 else {
458 if (name)
459 error("%.*s: expected %s type, but the object "
460 "dereferences to %s type",
461 namelen, name, typename(expected_type),
462 typename(o->type));
463 return NULL;
468 static int peel_onion(const char *name, int len, unsigned char *sha1)
470 unsigned char outer[20];
471 const char *sp;
472 unsigned int expected_type = 0;
473 struct object *o;
476 * "ref^{type}" dereferences ref repeatedly until you cannot
477 * dereference anymore, or you get an object of given type,
478 * whichever comes first. "ref^{}" means just dereference
479 * tags until you get a non-tag. "ref^0" is a shorthand for
480 * "ref^{commit}". "commit^{tree}" could be used to find the
481 * top-level tree of the given commit.
483 if (len < 4 || name[len-1] != '}')
484 return -1;
486 for (sp = name + len - 1; name <= sp; sp--) {
487 int ch = *sp;
488 if (ch == '{' && name < sp && sp[-1] == '^')
489 break;
491 if (sp <= name)
492 return -1;
494 sp++; /* beginning of type name, or closing brace for empty */
495 if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
496 expected_type = OBJ_COMMIT;
497 else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
498 expected_type = OBJ_TREE;
499 else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
500 expected_type = OBJ_BLOB;
501 else if (sp[0] == '}')
502 expected_type = OBJ_NONE;
503 else
504 return -1;
506 if (get_sha1_1(name, sp - name - 2, outer))
507 return -1;
509 o = parse_object(outer);
510 if (!o)
511 return -1;
512 if (!expected_type) {
513 o = deref_tag(o, name, sp - name - 2);
514 if (!o || (!o->parsed && !parse_object(o->sha1)))
515 return -1;
516 hashcpy(sha1, o->sha1);
518 else {
520 * At this point, the syntax look correct, so
521 * if we do not get the needed object, we should
522 * barf.
524 o = peel_to_type(name, len, o, expected_type);
525 if (o) {
526 hashcpy(sha1, o->sha1);
527 return 0;
529 return -1;
531 return 0;
534 static int get_describe_name(const char *name, int len, unsigned char *sha1)
536 const char *cp;
538 for (cp = name + len - 1; name + 2 <= cp; cp--) {
539 char ch = *cp;
540 if (hexval(ch) & ~0377) {
541 /* We must be looking at g in "SOMETHING-g"
542 * for it to be describe output.
544 if (ch == 'g' && cp[-1] == '-') {
545 cp++;
546 len -= cp - name;
547 return get_short_sha1(cp, len, sha1, 1);
551 return -1;
554 static int get_sha1_1(const char *name, int len, unsigned char *sha1)
556 int ret, has_suffix;
557 const char *cp;
560 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
562 has_suffix = 0;
563 for (cp = name + len - 1; name <= cp; cp--) {
564 int ch = *cp;
565 if ('0' <= ch && ch <= '9')
566 continue;
567 if (ch == '~' || ch == '^')
568 has_suffix = ch;
569 break;
572 if (has_suffix) {
573 int num = 0;
574 int len1 = cp - name;
575 cp++;
576 while (cp < name + len)
577 num = num * 10 + *cp++ - '0';
578 if (!num && len1 == len - 1)
579 num = 1;
580 if (has_suffix == '^')
581 return get_parent(name, len1, sha1, num);
582 /* else if (has_suffix == '~') -- goes without saying */
583 return get_nth_ancestor(name, len1, sha1, num);
586 ret = peel_onion(name, len, sha1);
587 if (!ret)
588 return 0;
590 ret = get_sha1_basic(name, len, sha1);
591 if (!ret)
592 return 0;
594 /* It could be describe output that is "SOMETHING-gXXXX" */
595 ret = get_describe_name(name, len, sha1);
596 if (!ret)
597 return 0;
599 return get_short_sha1(name, len, sha1, 0);
602 static int handle_one_ref(const char *path,
603 const unsigned char *sha1, int flag, void *cb_data)
605 struct commit_list **list = cb_data;
606 struct object *object = parse_object(sha1);
607 if (!object)
608 return 0;
609 if (object->type == OBJ_TAG) {
610 object = deref_tag(object, path, strlen(path));
611 if (!object)
612 return 0;
614 if (object->type != OBJ_COMMIT)
615 return 0;
616 insert_by_date((struct commit *)object, list);
617 return 0;
621 * This interprets names like ':/Initial revision of "git"' by searching
622 * through history and returning the first commit whose message starts
623 * with the given string.
625 * For future extension, ':/!' is reserved. If you want to match a message
626 * beginning with a '!', you have to repeat the exclamation mark.
629 #define ONELINE_SEEN (1u<<20)
630 static int get_sha1_oneline(const char *prefix, unsigned char *sha1)
632 struct commit_list *list = NULL, *backup = NULL, *l;
633 int retval = -1;
634 char *temp_commit_buffer = NULL;
636 if (prefix[0] == '!') {
637 if (prefix[1] != '!')
638 die ("Invalid search pattern: %s", prefix);
639 prefix++;
641 for_each_ref(handle_one_ref, &list);
642 for (l = list; l; l = l->next)
643 commit_list_insert(l->item, &backup);
644 while (list) {
645 char *p;
646 struct commit *commit;
647 enum object_type type;
648 unsigned long size;
650 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
651 if (!parse_object(commit->object.sha1))
652 continue;
653 free(temp_commit_buffer);
654 if (commit->buffer)
655 p = commit->buffer;
656 else {
657 p = read_sha1_file(commit->object.sha1, &type, &size);
658 if (!p)
659 continue;
660 temp_commit_buffer = p;
662 if (!(p = strstr(p, "\n\n")))
663 continue;
664 if (!prefixcmp(p + 2, prefix)) {
665 hashcpy(sha1, commit->object.sha1);
666 retval = 0;
667 break;
670 free(temp_commit_buffer);
671 free_commit_list(list);
672 for (l = backup; l; l = l->next)
673 clear_commit_marks(l->item, ONELINE_SEEN);
674 return retval;
678 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
679 * notably "xyz^" for "parent of xyz"
681 int get_sha1(const char *name, unsigned char *sha1)
683 unsigned unused;
684 return get_sha1_with_mode(name, sha1, &unused);
687 int get_sha1_with_mode(const char *name, unsigned char *sha1, unsigned *mode)
689 int ret, bracket_depth;
690 int namelen = strlen(name);
691 const char *cp;
693 *mode = S_IFINVALID;
694 ret = get_sha1_1(name, namelen, sha1);
695 if (!ret)
696 return ret;
697 /* sha1:path --> object name of path in ent sha1
698 * :path -> object name of path in index
699 * :[0-3]:path -> object name of path in index at stage
701 if (name[0] == ':') {
702 int stage = 0;
703 struct cache_entry *ce;
704 int pos;
705 if (namelen > 2 && name[1] == '/')
706 return get_sha1_oneline(name + 2, sha1);
707 if (namelen < 3 ||
708 name[2] != ':' ||
709 name[1] < '0' || '3' < name[1])
710 cp = name + 1;
711 else {
712 stage = name[1] - '0';
713 cp = name + 3;
715 namelen = namelen - (cp - name);
716 if (!active_cache)
717 read_cache();
718 pos = cache_name_pos(cp, namelen);
719 if (pos < 0)
720 pos = -pos - 1;
721 while (pos < active_nr) {
722 ce = active_cache[pos];
723 if (ce_namelen(ce) != namelen ||
724 memcmp(ce->name, cp, namelen))
725 break;
726 if (ce_stage(ce) == stage) {
727 hashcpy(sha1, ce->sha1);
728 *mode = ce->ce_mode;
729 return 0;
731 pos++;
733 return -1;
735 for (cp = name, bracket_depth = 0; *cp; cp++) {
736 if (*cp == '{')
737 bracket_depth++;
738 else if (bracket_depth && *cp == '}')
739 bracket_depth--;
740 else if (!bracket_depth && *cp == ':')
741 break;
743 if (*cp == ':') {
744 unsigned char tree_sha1[20];
745 if (!get_sha1_1(name, cp-name, tree_sha1))
746 return get_tree_entry(tree_sha1, cp+1, sha1,
747 mode);
749 return ret;