pack-objects: match prune logic for discarding objects
[git/debian.git] / sha1_name.c
blob5b004f513b999b31b7968b22dde955c859704853
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"
8 #include "remote.h"
10 static int get_sha1_oneline(const char *, unsigned char *, struct commit_list *);
12 typedef int (*disambiguate_hint_fn)(const unsigned char *, void *);
14 struct disambiguate_state {
15 disambiguate_hint_fn fn;
16 void *cb_data;
17 unsigned char candidate[20];
18 unsigned candidate_exists:1;
19 unsigned candidate_checked:1;
20 unsigned candidate_ok:1;
21 unsigned disambiguate_fn_used:1;
22 unsigned ambiguous:1;
23 unsigned always_call_fn:1;
26 static void update_candidates(struct disambiguate_state *ds, const unsigned char *current)
28 if (ds->always_call_fn) {
29 ds->ambiguous = ds->fn(current, ds->cb_data) ? 1 : 0;
30 return;
32 if (!ds->candidate_exists) {
33 /* this is the first candidate */
34 hashcpy(ds->candidate, current);
35 ds->candidate_exists = 1;
36 return;
37 } else if (!hashcmp(ds->candidate, current)) {
38 /* the same as what we already have seen */
39 return;
42 if (!ds->fn) {
43 /* cannot disambiguate between ds->candidate and current */
44 ds->ambiguous = 1;
45 return;
48 if (!ds->candidate_checked) {
49 ds->candidate_ok = ds->fn(ds->candidate, ds->cb_data);
50 ds->disambiguate_fn_used = 1;
51 ds->candidate_checked = 1;
54 if (!ds->candidate_ok) {
55 /* discard the candidate; we know it does not satisfy fn */
56 hashcpy(ds->candidate, current);
57 ds->candidate_checked = 0;
58 return;
61 /* if we reach this point, we know ds->candidate satisfies fn */
62 if (ds->fn(current, ds->cb_data)) {
64 * if both current and candidate satisfy fn, we cannot
65 * disambiguate.
67 ds->candidate_ok = 0;
68 ds->ambiguous = 1;
71 /* otherwise, current can be discarded and candidate is still good */
74 static void find_short_object_filename(int len, const char *hex_pfx, struct disambiguate_state *ds)
76 struct alternate_object_database *alt;
77 char hex[40];
78 static struct alternate_object_database *fakeent;
80 if (!fakeent) {
82 * Create a "fake" alternate object database that
83 * points to our own object database, to make it
84 * easier to get a temporary working space in
85 * alt->name/alt->base while iterating over the
86 * object databases including our own.
88 const char *objdir = get_object_directory();
89 int objdir_len = strlen(objdir);
90 int entlen = objdir_len + 43;
91 fakeent = xmalloc(sizeof(*fakeent) + entlen);
92 memcpy(fakeent->base, objdir, objdir_len);
93 fakeent->name = fakeent->base + objdir_len + 1;
94 fakeent->name[-1] = '/';
96 fakeent->next = alt_odb_list;
98 sprintf(hex, "%.2s", hex_pfx);
99 for (alt = fakeent; alt && !ds->ambiguous; alt = alt->next) {
100 struct dirent *de;
101 DIR *dir;
102 sprintf(alt->name, "%.2s/", hex_pfx);
103 dir = opendir(alt->base);
104 if (!dir)
105 continue;
107 while (!ds->ambiguous && (de = readdir(dir)) != NULL) {
108 unsigned char sha1[20];
110 if (strlen(de->d_name) != 38)
111 continue;
112 if (memcmp(de->d_name, hex_pfx + 2, len - 2))
113 continue;
114 memcpy(hex + 2, de->d_name, 38);
115 if (!get_sha1_hex(hex, sha1))
116 update_candidates(ds, sha1);
118 closedir(dir);
122 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
124 do {
125 if (*a != *b)
126 return 0;
127 a++;
128 b++;
129 len -= 2;
130 } while (len > 1);
131 if (len)
132 if ((*a ^ *b) & 0xf0)
133 return 0;
134 return 1;
137 static void unique_in_pack(int len,
138 const unsigned char *bin_pfx,
139 struct packed_git *p,
140 struct disambiguate_state *ds)
142 uint32_t num, last, i, first = 0;
143 const unsigned char *current = NULL;
145 open_pack_index(p);
146 num = p->num_objects;
147 last = num;
148 while (first < last) {
149 uint32_t mid = (first + last) / 2;
150 const unsigned char *current;
151 int cmp;
153 current = nth_packed_object_sha1(p, mid);
154 cmp = hashcmp(bin_pfx, current);
155 if (!cmp) {
156 first = mid;
157 break;
159 if (cmp > 0) {
160 first = mid+1;
161 continue;
163 last = mid;
167 * At this point, "first" is the location of the lowest object
168 * with an object name that could match "bin_pfx". See if we have
169 * 0, 1 or more objects that actually match(es).
171 for (i = first; i < num && !ds->ambiguous; i++) {
172 current = nth_packed_object_sha1(p, i);
173 if (!match_sha(len, bin_pfx, current))
174 break;
175 update_candidates(ds, current);
179 static void find_short_packed_object(int len, const unsigned char *bin_pfx,
180 struct disambiguate_state *ds)
182 struct packed_git *p;
184 prepare_packed_git();
185 for (p = packed_git; p && !ds->ambiguous; p = p->next)
186 unique_in_pack(len, bin_pfx, p, ds);
189 #define SHORT_NAME_NOT_FOUND (-1)
190 #define SHORT_NAME_AMBIGUOUS (-2)
192 static int finish_object_disambiguation(struct disambiguate_state *ds,
193 unsigned char *sha1)
195 if (ds->ambiguous)
196 return SHORT_NAME_AMBIGUOUS;
198 if (!ds->candidate_exists)
199 return SHORT_NAME_NOT_FOUND;
201 if (!ds->candidate_checked)
203 * If this is the only candidate, there is no point
204 * calling the disambiguation hint callback.
206 * On the other hand, if the current candidate
207 * replaced an earlier candidate that did _not_ pass
208 * the disambiguation hint callback, then we do have
209 * more than one objects that match the short name
210 * given, so we should make sure this one matches;
211 * otherwise, if we discovered this one and the one
212 * that we previously discarded in the reverse order,
213 * we would end up showing different results in the
214 * same repository!
216 ds->candidate_ok = (!ds->disambiguate_fn_used ||
217 ds->fn(ds->candidate, ds->cb_data));
219 if (!ds->candidate_ok)
220 return SHORT_NAME_AMBIGUOUS;
222 hashcpy(sha1, ds->candidate);
223 return 0;
226 static int disambiguate_commit_only(const unsigned char *sha1, void *cb_data_unused)
228 int kind = sha1_object_info(sha1, NULL);
229 return kind == OBJ_COMMIT;
232 static int disambiguate_committish_only(const unsigned char *sha1, void *cb_data_unused)
234 struct object *obj;
235 int kind;
237 kind = sha1_object_info(sha1, NULL);
238 if (kind == OBJ_COMMIT)
239 return 1;
240 if (kind != OBJ_TAG)
241 return 0;
243 /* We need to do this the hard way... */
244 obj = deref_tag(parse_object(sha1), NULL, 0);
245 if (obj && obj->type == OBJ_COMMIT)
246 return 1;
247 return 0;
250 static int disambiguate_tree_only(const unsigned char *sha1, void *cb_data_unused)
252 int kind = sha1_object_info(sha1, NULL);
253 return kind == OBJ_TREE;
256 static int disambiguate_treeish_only(const unsigned char *sha1, void *cb_data_unused)
258 struct object *obj;
259 int kind;
261 kind = sha1_object_info(sha1, NULL);
262 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
263 return 1;
264 if (kind != OBJ_TAG)
265 return 0;
267 /* We need to do this the hard way... */
268 obj = deref_tag(lookup_object(sha1), NULL, 0);
269 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
270 return 1;
271 return 0;
274 static int disambiguate_blob_only(const unsigned char *sha1, void *cb_data_unused)
276 int kind = sha1_object_info(sha1, NULL);
277 return kind == OBJ_BLOB;
280 static int prepare_prefixes(const char *name, int len,
281 unsigned char *bin_pfx,
282 char *hex_pfx)
284 int i;
286 hashclr(bin_pfx);
287 memset(hex_pfx, 'x', 40);
288 for (i = 0; i < len ;i++) {
289 unsigned char c = name[i];
290 unsigned char val;
291 if (c >= '0' && c <= '9')
292 val = c - '0';
293 else if (c >= 'a' && c <= 'f')
294 val = c - 'a' + 10;
295 else if (c >= 'A' && c <='F') {
296 val = c - 'A' + 10;
297 c -= 'A' - 'a';
299 else
300 return -1;
301 hex_pfx[i] = c;
302 if (!(i & 1))
303 val <<= 4;
304 bin_pfx[i >> 1] |= val;
306 return 0;
309 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
310 unsigned flags)
312 int status;
313 char hex_pfx[40];
314 unsigned char bin_pfx[20];
315 struct disambiguate_state ds;
316 int quietly = !!(flags & GET_SHA1_QUIETLY);
318 if (len < MINIMUM_ABBREV || len > 40)
319 return -1;
320 if (prepare_prefixes(name, len, bin_pfx, hex_pfx) < 0)
321 return -1;
323 prepare_alt_odb();
325 memset(&ds, 0, sizeof(ds));
326 if (flags & GET_SHA1_COMMIT)
327 ds.fn = disambiguate_commit_only;
328 else if (flags & GET_SHA1_COMMITTISH)
329 ds.fn = disambiguate_committish_only;
330 else if (flags & GET_SHA1_TREE)
331 ds.fn = disambiguate_tree_only;
332 else if (flags & GET_SHA1_TREEISH)
333 ds.fn = disambiguate_treeish_only;
334 else if (flags & GET_SHA1_BLOB)
335 ds.fn = disambiguate_blob_only;
337 find_short_object_filename(len, hex_pfx, &ds);
338 find_short_packed_object(len, bin_pfx, &ds);
339 status = finish_object_disambiguation(&ds, sha1);
341 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
342 return error("short SHA1 %.*s is ambiguous.", len, hex_pfx);
343 return status;
346 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
348 char hex_pfx[40];
349 unsigned char bin_pfx[20];
350 struct disambiguate_state ds;
351 int len = strlen(prefix);
353 if (len < MINIMUM_ABBREV || len > 40)
354 return -1;
355 if (prepare_prefixes(prefix, len, bin_pfx, hex_pfx) < 0)
356 return -1;
358 prepare_alt_odb();
360 memset(&ds, 0, sizeof(ds));
361 ds.always_call_fn = 1;
362 ds.cb_data = cb_data;
363 ds.fn = fn;
365 find_short_object_filename(len, hex_pfx, &ds);
366 find_short_packed_object(len, bin_pfx, &ds);
367 return ds.ambiguous;
370 const char *find_unique_abbrev(const unsigned char *sha1, int len)
372 int status, exists;
373 static char hex[41];
375 exists = has_sha1_file(sha1);
376 memcpy(hex, sha1_to_hex(sha1), 40);
377 if (len == 40 || !len)
378 return hex;
379 while (len < 40) {
380 unsigned char sha1_ret[20];
381 status = get_short_sha1(hex, len, sha1_ret, GET_SHA1_QUIETLY);
382 if (exists
383 ? !status
384 : status == SHORT_NAME_NOT_FOUND) {
385 hex[len] = 0;
386 return hex;
388 len++;
390 return hex;
393 static int ambiguous_path(const char *path, int len)
395 int slash = 1;
396 int cnt;
398 for (cnt = 0; cnt < len; cnt++) {
399 switch (*path++) {
400 case '\0':
401 break;
402 case '/':
403 if (slash)
404 break;
405 slash = 1;
406 continue;
407 case '.':
408 continue;
409 default:
410 slash = 0;
411 continue;
413 break;
415 return slash;
418 static inline int upstream_mark(const char *string, int len)
420 const char *suffix[] = { "@{upstream}", "@{u}" };
421 int i;
423 for (i = 0; i < ARRAY_SIZE(suffix); i++) {
424 int suffix_len = strlen(suffix[i]);
425 if (suffix_len <= len
426 && !memcmp(string, suffix[i], suffix_len))
427 return suffix_len;
429 return 0;
432 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
433 static int interpret_nth_prior_checkout(const char *name, int namelen, struct strbuf *buf);
435 static int get_sha1_basic(const char *str, int len, unsigned char *sha1,
436 unsigned int flags)
438 static const char *warn_msg = "refname '%.*s' is ambiguous.";
439 static const char *object_name_msg = N_(
440 "Git normally never creates a ref that ends with 40 hex characters\n"
441 "because it will be ignored when you just specify 40-hex. These refs\n"
442 "may be created by mistake. For example,\n"
443 "\n"
444 " git checkout -b $br $(git rev-parse ...)\n"
445 "\n"
446 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
447 "examine these refs and maybe delete them. Turn this message off by\n"
448 "running \"git config advice.objectNameWarning false\"");
449 unsigned char tmp_sha1[20];
450 char *real_ref = NULL;
451 int refs_found = 0;
452 int at, reflog_len, nth_prior = 0;
454 if (len == 40 && !get_sha1_hex(str, sha1)) {
455 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
456 refs_found = dwim_ref(str, len, tmp_sha1, &real_ref);
457 if (refs_found > 0) {
458 warning(warn_msg, len, str);
459 if (advice_object_name_warning)
460 fprintf(stderr, "%s\n", _(object_name_msg));
462 free(real_ref);
464 return 0;
467 /* basic@{time or number or -number} format to query ref-log */
468 reflog_len = at = 0;
469 if (len && str[len-1] == '}') {
470 for (at = len-4; at >= 0; at--) {
471 if (str[at] == '@' && str[at+1] == '{') {
472 if (str[at+2] == '-') {
473 if (at != 0)
474 /* @{-N} not at start */
475 return -1;
476 nth_prior = 1;
477 continue;
479 if (!upstream_mark(str + at, len - at)) {
480 reflog_len = (len-1) - (at+2);
481 len = at;
483 break;
488 /* Accept only unambiguous ref paths. */
489 if (len && ambiguous_path(str, len))
490 return -1;
492 if (nth_prior) {
493 struct strbuf buf = STRBUF_INIT;
494 int detached;
496 if (interpret_nth_prior_checkout(str, len, &buf) > 0) {
497 detached = (buf.len == 40 && !get_sha1_hex(buf.buf, sha1));
498 strbuf_release(&buf);
499 if (detached)
500 return 0;
504 if (!len && reflog_len)
505 /* allow "@{...}" to mean the current branch reflog */
506 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
507 else if (reflog_len)
508 refs_found = dwim_log(str, len, sha1, &real_ref);
509 else
510 refs_found = dwim_ref(str, len, sha1, &real_ref);
512 if (!refs_found)
513 return -1;
515 if (warn_ambiguous_refs && !(flags & GET_SHA1_QUIETLY) &&
516 (refs_found > 1 ||
517 !get_short_sha1(str, len, tmp_sha1, GET_SHA1_QUIETLY)))
518 warning(warn_msg, len, str);
520 if (reflog_len) {
521 int nth, i;
522 unsigned long at_time;
523 unsigned long co_time;
524 int co_tz, co_cnt;
526 /* Is it asking for N-th entry, or approxidate? */
527 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
528 char ch = str[at+2+i];
529 if ('0' <= ch && ch <= '9')
530 nth = nth * 10 + ch - '0';
531 else
532 nth = -1;
534 if (100000000 <= nth) {
535 at_time = nth;
536 nth = -1;
537 } else if (0 <= nth)
538 at_time = 0;
539 else {
540 int errors = 0;
541 char *tmp = xstrndup(str + at + 2, reflog_len);
542 at_time = approxidate_careful(tmp, &errors);
543 free(tmp);
544 if (errors) {
545 free(real_ref);
546 return -1;
549 if (read_ref_at(real_ref, flags, at_time, nth, sha1, NULL,
550 &co_time, &co_tz, &co_cnt)) {
551 if (!len) {
552 if (starts_with(real_ref, "refs/heads/")) {
553 str = real_ref + 11;
554 len = strlen(real_ref + 11);
555 } else {
556 /* detached HEAD */
557 str = "HEAD";
558 len = 4;
561 if (at_time) {
562 if (!(flags & GET_SHA1_QUIETLY)) {
563 warning("Log for '%.*s' only goes "
564 "back to %s.", len, str,
565 show_date(co_time, co_tz, DATE_RFC2822));
567 } else {
568 if (flags & GET_SHA1_QUIETLY) {
569 exit(128);
571 die("Log for '%.*s' only has %d entries.",
572 len, str, co_cnt);
577 free(real_ref);
578 return 0;
581 static int get_parent(const char *name, int len,
582 unsigned char *result, int idx)
584 unsigned char sha1[20];
585 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
586 struct commit *commit;
587 struct commit_list *p;
589 if (ret)
590 return ret;
591 commit = lookup_commit_reference(sha1);
592 if (parse_commit(commit))
593 return -1;
594 if (!idx) {
595 hashcpy(result, commit->object.sha1);
596 return 0;
598 p = commit->parents;
599 while (p) {
600 if (!--idx) {
601 hashcpy(result, p->item->object.sha1);
602 return 0;
604 p = p->next;
606 return -1;
609 static int get_nth_ancestor(const char *name, int len,
610 unsigned char *result, int generation)
612 unsigned char sha1[20];
613 struct commit *commit;
614 int ret;
616 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
617 if (ret)
618 return ret;
619 commit = lookup_commit_reference(sha1);
620 if (!commit)
621 return -1;
623 while (generation--) {
624 if (parse_commit(commit) || !commit->parents)
625 return -1;
626 commit = commit->parents->item;
628 hashcpy(result, commit->object.sha1);
629 return 0;
632 struct object *peel_to_type(const char *name, int namelen,
633 struct object *o, enum object_type expected_type)
635 if (name && !namelen)
636 namelen = strlen(name);
637 while (1) {
638 if (!o || (!o->parsed && !parse_object(o->sha1)))
639 return NULL;
640 if (expected_type == OBJ_ANY || o->type == expected_type)
641 return o;
642 if (o->type == OBJ_TAG)
643 o = ((struct tag*) o)->tagged;
644 else if (o->type == OBJ_COMMIT)
645 o = &(((struct commit *) o)->tree->object);
646 else {
647 if (name)
648 error("%.*s: expected %s type, but the object "
649 "dereferences to %s type",
650 namelen, name, typename(expected_type),
651 typename(o->type));
652 return NULL;
657 static int peel_onion(const char *name, int len, unsigned char *sha1)
659 unsigned char outer[20];
660 const char *sp;
661 unsigned int expected_type = 0;
662 unsigned lookup_flags = 0;
663 struct object *o;
666 * "ref^{type}" dereferences ref repeatedly until you cannot
667 * dereference anymore, or you get an object of given type,
668 * whichever comes first. "ref^{}" means just dereference
669 * tags until you get a non-tag. "ref^0" is a shorthand for
670 * "ref^{commit}". "commit^{tree}" could be used to find the
671 * top-level tree of the given commit.
673 if (len < 4 || name[len-1] != '}')
674 return -1;
676 for (sp = name + len - 1; name <= sp; sp--) {
677 int ch = *sp;
678 if (ch == '{' && name < sp && sp[-1] == '^')
679 break;
681 if (sp <= name)
682 return -1;
684 sp++; /* beginning of type name, or closing brace for empty */
685 if (starts_with(sp, "commit}"))
686 expected_type = OBJ_COMMIT;
687 else if (starts_with(sp, "tag}"))
688 expected_type = OBJ_TAG;
689 else if (starts_with(sp, "tree}"))
690 expected_type = OBJ_TREE;
691 else if (starts_with(sp, "blob}"))
692 expected_type = OBJ_BLOB;
693 else if (starts_with(sp, "object}"))
694 expected_type = OBJ_ANY;
695 else if (sp[0] == '}')
696 expected_type = OBJ_NONE;
697 else if (sp[0] == '/')
698 expected_type = OBJ_COMMIT;
699 else
700 return -1;
702 if (expected_type == OBJ_COMMIT)
703 lookup_flags = GET_SHA1_COMMITTISH;
704 else if (expected_type == OBJ_TREE)
705 lookup_flags = GET_SHA1_TREEISH;
707 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
708 return -1;
710 o = parse_object(outer);
711 if (!o)
712 return -1;
713 if (!expected_type) {
714 o = deref_tag(o, name, sp - name - 2);
715 if (!o || (!o->parsed && !parse_object(o->sha1)))
716 return -1;
717 hashcpy(sha1, o->sha1);
718 return 0;
722 * At this point, the syntax look correct, so
723 * if we do not get the needed object, we should
724 * barf.
726 o = peel_to_type(name, len, o, expected_type);
727 if (!o)
728 return -1;
730 hashcpy(sha1, o->sha1);
731 if (sp[0] == '/') {
732 /* "$commit^{/foo}" */
733 char *prefix;
734 int ret;
735 struct commit_list *list = NULL;
738 * $commit^{/}. Some regex implementation may reject.
739 * We don't need regex anyway. '' pattern always matches.
741 if (sp[1] == '}')
742 return 0;
744 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
745 commit_list_insert((struct commit *)o, &list);
746 ret = get_sha1_oneline(prefix, sha1, list);
747 free(prefix);
748 return ret;
750 return 0;
753 static int get_describe_name(const char *name, int len, unsigned char *sha1)
755 const char *cp;
756 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
758 for (cp = name + len - 1; name + 2 <= cp; cp--) {
759 char ch = *cp;
760 if (hexval(ch) & ~0377) {
761 /* We must be looking at g in "SOMETHING-g"
762 * for it to be describe output.
764 if (ch == 'g' && cp[-1] == '-') {
765 cp++;
766 len -= cp - name;
767 return get_short_sha1(cp, len, sha1, flags);
771 return -1;
774 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
776 int ret, has_suffix;
777 const char *cp;
780 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
782 has_suffix = 0;
783 for (cp = name + len - 1; name <= cp; cp--) {
784 int ch = *cp;
785 if ('0' <= ch && ch <= '9')
786 continue;
787 if (ch == '~' || ch == '^')
788 has_suffix = ch;
789 break;
792 if (has_suffix) {
793 int num = 0;
794 int len1 = cp - name;
795 cp++;
796 while (cp < name + len)
797 num = num * 10 + *cp++ - '0';
798 if (!num && len1 == len - 1)
799 num = 1;
800 if (has_suffix == '^')
801 return get_parent(name, len1, sha1, num);
802 /* else if (has_suffix == '~') -- goes without saying */
803 return get_nth_ancestor(name, len1, sha1, num);
806 ret = peel_onion(name, len, sha1);
807 if (!ret)
808 return 0;
810 ret = get_sha1_basic(name, len, sha1, lookup_flags);
811 if (!ret)
812 return 0;
814 /* It could be describe output that is "SOMETHING-gXXXX" */
815 ret = get_describe_name(name, len, sha1);
816 if (!ret)
817 return 0;
819 return get_short_sha1(name, len, sha1, lookup_flags);
823 * This interprets names like ':/Initial revision of "git"' by searching
824 * through history and returning the first commit whose message starts
825 * the given regular expression.
827 * For future extension, ':/!' is reserved. If you want to match a message
828 * beginning with a '!', you have to repeat the exclamation mark.
831 /* Remember to update object flag allocation in object.h */
832 #define ONELINE_SEEN (1u<<20)
834 static int handle_one_ref(const char *path,
835 const unsigned char *sha1, int flag, void *cb_data)
837 struct commit_list **list = cb_data;
838 struct object *object = parse_object(sha1);
839 if (!object)
840 return 0;
841 if (object->type == OBJ_TAG) {
842 object = deref_tag(object, path, strlen(path));
843 if (!object)
844 return 0;
846 if (object->type != OBJ_COMMIT)
847 return 0;
848 commit_list_insert((struct commit *)object, list);
849 return 0;
852 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
853 struct commit_list *list)
855 struct commit_list *backup = NULL, *l;
856 int found = 0;
857 regex_t regex;
859 if (prefix[0] == '!') {
860 if (prefix[1] != '!')
861 die ("Invalid search pattern: %s", prefix);
862 prefix++;
865 if (regcomp(&regex, prefix, REG_EXTENDED))
866 die("Invalid search pattern: %s", prefix);
868 for (l = list; l; l = l->next) {
869 l->item->object.flags |= ONELINE_SEEN;
870 commit_list_insert(l->item, &backup);
872 while (list) {
873 const char *p, *buf;
874 struct commit *commit;
875 int matches;
877 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
878 if (!parse_object(commit->object.sha1))
879 continue;
880 buf = get_commit_buffer(commit, NULL);
881 p = strstr(buf, "\n\n");
882 matches = p && !regexec(&regex, p + 2, 0, NULL, 0);
883 unuse_commit_buffer(commit, buf);
885 if (matches) {
886 hashcpy(sha1, commit->object.sha1);
887 found = 1;
888 break;
891 regfree(&regex);
892 free_commit_list(list);
893 for (l = backup; l; l = l->next)
894 clear_commit_marks(l->item, ONELINE_SEEN);
895 free_commit_list(backup);
896 return found ? 0 : -1;
899 struct grab_nth_branch_switch_cbdata {
900 int remaining;
901 struct strbuf buf;
904 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
905 const char *email, unsigned long timestamp, int tz,
906 const char *message, void *cb_data)
908 struct grab_nth_branch_switch_cbdata *cb = cb_data;
909 const char *match = NULL, *target = NULL;
910 size_t len;
912 if (skip_prefix(message, "checkout: moving from ", &match))
913 target = strstr(match, " to ");
915 if (!match || !target)
916 return 0;
917 if (--(cb->remaining) == 0) {
918 len = target - match;
919 strbuf_reset(&cb->buf);
920 strbuf_add(&cb->buf, match, len);
921 return 1; /* we are done */
923 return 0;
927 * Parse @{-N} syntax, return the number of characters parsed
928 * if successful; otherwise signal an error with negative value.
930 static int interpret_nth_prior_checkout(const char *name, int namelen,
931 struct strbuf *buf)
933 long nth;
934 int retval;
935 struct grab_nth_branch_switch_cbdata cb;
936 const char *brace;
937 char *num_end;
939 if (namelen < 4)
940 return -1;
941 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
942 return -1;
943 brace = memchr(name, '}', namelen);
944 if (!brace)
945 return -1;
946 nth = strtol(name + 3, &num_end, 10);
947 if (num_end != brace)
948 return -1;
949 if (nth <= 0)
950 return -1;
951 cb.remaining = nth;
952 strbuf_init(&cb.buf, 20);
954 retval = 0;
955 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
956 strbuf_reset(buf);
957 strbuf_addbuf(buf, &cb.buf);
958 retval = brace - name + 1;
961 strbuf_release(&cb.buf);
962 return retval;
965 int get_sha1_mb(const char *name, unsigned char *sha1)
967 struct commit *one, *two;
968 struct commit_list *mbs;
969 unsigned char sha1_tmp[20];
970 const char *dots;
971 int st;
973 dots = strstr(name, "...");
974 if (!dots)
975 return get_sha1(name, sha1);
976 if (dots == name)
977 st = get_sha1("HEAD", sha1_tmp);
978 else {
979 struct strbuf sb;
980 strbuf_init(&sb, dots - name);
981 strbuf_add(&sb, name, dots - name);
982 st = get_sha1_committish(sb.buf, sha1_tmp);
983 strbuf_release(&sb);
985 if (st)
986 return st;
987 one = lookup_commit_reference_gently(sha1_tmp, 0);
988 if (!one)
989 return -1;
991 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", sha1_tmp))
992 return -1;
993 two = lookup_commit_reference_gently(sha1_tmp, 0);
994 if (!two)
995 return -1;
996 mbs = get_merge_bases(one, two, 1);
997 if (!mbs || mbs->next)
998 st = -1;
999 else {
1000 st = 0;
1001 hashcpy(sha1, mbs->item->object.sha1);
1003 free_commit_list(mbs);
1004 return st;
1007 /* parse @something syntax, when 'something' is not {.*} */
1008 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1010 const char *next;
1012 if (len || name[1] == '{')
1013 return -1;
1015 /* make sure it's a single @, or @@{.*}, not @foo */
1016 next = memchr(name + len + 1, '@', namelen - len - 1);
1017 if (next && next[1] != '{')
1018 return -1;
1019 if (!next)
1020 next = name + namelen;
1021 if (next != name + 1)
1022 return -1;
1024 strbuf_reset(buf);
1025 strbuf_add(buf, "HEAD", 4);
1026 return 1;
1029 static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
1031 /* we have extra data, which might need further processing */
1032 struct strbuf tmp = STRBUF_INIT;
1033 int used = buf->len;
1034 int ret;
1036 strbuf_add(buf, name + len, namelen - len);
1037 ret = interpret_branch_name(buf->buf, buf->len, &tmp);
1038 /* that data was not interpreted, remove our cruft */
1039 if (ret < 0) {
1040 strbuf_setlen(buf, used);
1041 return len;
1043 strbuf_reset(buf);
1044 strbuf_addbuf(buf, &tmp);
1045 strbuf_release(&tmp);
1046 /* tweak for size of {-N} versus expanded ref name */
1047 return ret - used + len;
1050 static void set_shortened_ref(struct strbuf *buf, const char *ref)
1052 char *s = shorten_unambiguous_ref(ref, 0);
1053 strbuf_reset(buf);
1054 strbuf_addstr(buf, s);
1055 free(s);
1058 static const char *get_upstream_branch(const char *branch_buf, int len)
1060 char *branch = xstrndup(branch_buf, len);
1061 struct branch *upstream = branch_get(*branch ? branch : NULL);
1064 * Upstream can be NULL only if branch refers to HEAD and HEAD
1065 * points to something different than a branch.
1067 if (!upstream)
1068 die(_("HEAD does not point to a branch"));
1069 if (!upstream->merge || !upstream->merge[0]->dst) {
1070 if (!ref_exists(upstream->refname))
1071 die(_("No such branch: '%s'"), branch);
1072 if (!upstream->merge) {
1073 die(_("No upstream configured for branch '%s'"),
1074 upstream->name);
1076 die(
1077 _("Upstream branch '%s' not stored as a remote-tracking branch"),
1078 upstream->merge[0]->src);
1080 free(branch);
1082 return upstream->merge[0]->dst;
1085 static int interpret_upstream_mark(const char *name, int namelen,
1086 int at, struct strbuf *buf)
1088 int len;
1090 len = upstream_mark(name + at, namelen - at);
1091 if (!len)
1092 return -1;
1094 if (memchr(name, ':', at))
1095 return -1;
1097 set_shortened_ref(buf, get_upstream_branch(name, at));
1098 return len + at;
1102 * This reads short-hand syntax that not only evaluates to a commit
1103 * object name, but also can act as if the end user spelled the name
1104 * of the branch from the command line.
1106 * - "@{-N}" finds the name of the Nth previous branch we were on, and
1107 * places the name of the branch in the given buf and returns the
1108 * number of characters parsed if successful.
1110 * - "<branch>@{upstream}" finds the name of the other ref that
1111 * <branch> is configured to merge with (missing <branch> defaults
1112 * to the current branch), and places the name of the branch in the
1113 * given buf and returns the number of characters parsed if
1114 * successful.
1116 * If the input is not of the accepted format, it returns a negative
1117 * number to signal an error.
1119 * If the input was ok but there are not N branch switches in the
1120 * reflog, it returns 0.
1122 int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
1124 char *at;
1125 const char *start;
1126 int len = interpret_nth_prior_checkout(name, namelen, buf);
1128 if (!namelen)
1129 namelen = strlen(name);
1131 if (!len) {
1132 return len; /* syntax Ok, not enough switches */
1133 } else if (len > 0) {
1134 if (len == namelen)
1135 return len; /* consumed all */
1136 else
1137 return reinterpret(name, namelen, len, buf);
1140 for (start = name;
1141 (at = memchr(start, '@', namelen - (start - name)));
1142 start = at + 1) {
1144 len = interpret_empty_at(name, namelen, at - name, buf);
1145 if (len > 0)
1146 return reinterpret(name, namelen, len, buf);
1148 len = interpret_upstream_mark(name, namelen, at - name, buf);
1149 if (len > 0)
1150 return len;
1153 return -1;
1156 int strbuf_branchname(struct strbuf *sb, const char *name)
1158 int len = strlen(name);
1159 int used = interpret_branch_name(name, len, sb);
1161 if (used == len)
1162 return 0;
1163 if (used < 0)
1164 used = 0;
1165 strbuf_add(sb, name + used, len - used);
1166 return len;
1169 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1171 strbuf_branchname(sb, name);
1172 if (name[0] == '-')
1173 return -1;
1174 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1175 return check_refname_format(sb->buf, 0);
1179 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1180 * notably "xyz^" for "parent of xyz"
1182 int get_sha1(const char *name, unsigned char *sha1)
1184 struct object_context unused;
1185 return get_sha1_with_context(name, 0, sha1, &unused);
1189 * Many callers know that the user meant to name a commit-ish by
1190 * syntactical positions where the object name appears. Calling this
1191 * function allows the machinery to disambiguate shorter-than-unique
1192 * abbreviated object names between commit-ish and others.
1194 * Note that this does NOT error out when the named object is not a
1195 * commit-ish. It is merely to give a hint to the disambiguation
1196 * machinery.
1198 int get_sha1_committish(const char *name, unsigned char *sha1)
1200 struct object_context unused;
1201 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1202 sha1, &unused);
1205 int get_sha1_treeish(const char *name, unsigned char *sha1)
1207 struct object_context unused;
1208 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1209 sha1, &unused);
1212 int get_sha1_commit(const char *name, unsigned char *sha1)
1214 struct object_context unused;
1215 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1216 sha1, &unused);
1219 int get_sha1_tree(const char *name, unsigned char *sha1)
1221 struct object_context unused;
1222 return get_sha1_with_context(name, GET_SHA1_TREE,
1223 sha1, &unused);
1226 int get_sha1_blob(const char *name, unsigned char *sha1)
1228 struct object_context unused;
1229 return get_sha1_with_context(name, GET_SHA1_BLOB,
1230 sha1, &unused);
1233 /* Must be called only when object_name:filename doesn't exist. */
1234 static void diagnose_invalid_sha1_path(const char *prefix,
1235 const char *filename,
1236 const unsigned char *tree_sha1,
1237 const char *object_name,
1238 int object_name_len)
1240 struct stat st;
1241 unsigned char sha1[20];
1242 unsigned mode;
1244 if (!prefix)
1245 prefix = "";
1247 if (!lstat(filename, &st))
1248 die("Path '%s' exists on disk, but not in '%.*s'.",
1249 filename, object_name_len, object_name);
1250 if (errno == ENOENT || errno == ENOTDIR) {
1251 char *fullname = xstrfmt("%s%s", prefix, filename);
1253 if (!get_tree_entry(tree_sha1, fullname,
1254 sha1, &mode)) {
1255 die("Path '%s' exists, but not '%s'.\n"
1256 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1257 fullname,
1258 filename,
1259 object_name_len, object_name,
1260 fullname,
1261 object_name_len, object_name,
1262 filename);
1264 die("Path '%s' does not exist in '%.*s'",
1265 filename, object_name_len, object_name);
1269 /* Must be called only when :stage:filename doesn't exist. */
1270 static void diagnose_invalid_index_path(int stage,
1271 const char *prefix,
1272 const char *filename)
1274 struct stat st;
1275 const struct cache_entry *ce;
1276 int pos;
1277 unsigned namelen = strlen(filename);
1278 unsigned fullnamelen;
1279 char *fullname;
1281 if (!prefix)
1282 prefix = "";
1284 /* Wrong stage number? */
1285 pos = cache_name_pos(filename, namelen);
1286 if (pos < 0)
1287 pos = -pos - 1;
1288 if (pos < active_nr) {
1289 ce = active_cache[pos];
1290 if (ce_namelen(ce) == namelen &&
1291 !memcmp(ce->name, filename, namelen))
1292 die("Path '%s' is in the index, but not at stage %d.\n"
1293 "Did you mean ':%d:%s'?",
1294 filename, stage,
1295 ce_stage(ce), filename);
1298 /* Confusion between relative and absolute filenames? */
1299 fullnamelen = namelen + strlen(prefix);
1300 fullname = xmalloc(fullnamelen + 1);
1301 strcpy(fullname, prefix);
1302 strcat(fullname, filename);
1303 pos = cache_name_pos(fullname, fullnamelen);
1304 if (pos < 0)
1305 pos = -pos - 1;
1306 if (pos < active_nr) {
1307 ce = active_cache[pos];
1308 if (ce_namelen(ce) == fullnamelen &&
1309 !memcmp(ce->name, fullname, fullnamelen))
1310 die("Path '%s' is in the index, but not '%s'.\n"
1311 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1312 fullname, filename,
1313 ce_stage(ce), fullname,
1314 ce_stage(ce), filename);
1317 if (!lstat(filename, &st))
1318 die("Path '%s' exists on disk, but not in the index.", filename);
1319 if (errno == ENOENT || errno == ENOTDIR)
1320 die("Path '%s' does not exist (neither on disk nor in the index).",
1321 filename);
1323 free(fullname);
1327 static char *resolve_relative_path(const char *rel)
1329 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1330 return NULL;
1332 if (!startup_info)
1333 die("BUG: startup_info struct is not initialized.");
1335 if (!is_inside_work_tree())
1336 die("relative path syntax can't be used outside working tree.");
1338 /* die() inside prefix_path() if resolved path is outside worktree */
1339 return prefix_path(startup_info->prefix,
1340 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1341 rel);
1344 static int get_sha1_with_context_1(const char *name,
1345 unsigned flags,
1346 const char *prefix,
1347 unsigned char *sha1,
1348 struct object_context *oc)
1350 int ret, bracket_depth;
1351 int namelen = strlen(name);
1352 const char *cp;
1353 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1355 memset(oc, 0, sizeof(*oc));
1356 oc->mode = S_IFINVALID;
1357 ret = get_sha1_1(name, namelen, sha1, flags);
1358 if (!ret)
1359 return ret;
1361 * sha1:path --> object name of path in ent sha1
1362 * :path -> object name of absolute path in index
1363 * :./path -> object name of path relative to cwd in index
1364 * :[0-3]:path -> object name of path in index at stage
1365 * :/foo -> recent commit matching foo
1367 if (name[0] == ':') {
1368 int stage = 0;
1369 const struct cache_entry *ce;
1370 char *new_path = NULL;
1371 int pos;
1372 if (!only_to_die && namelen > 2 && name[1] == '/') {
1373 struct commit_list *list = NULL;
1374 for_each_ref(handle_one_ref, &list);
1375 commit_list_sort_by_date(&list);
1376 return get_sha1_oneline(name + 2, sha1, list);
1378 if (namelen < 3 ||
1379 name[2] != ':' ||
1380 name[1] < '0' || '3' < name[1])
1381 cp = name + 1;
1382 else {
1383 stage = name[1] - '0';
1384 cp = name + 3;
1386 new_path = resolve_relative_path(cp);
1387 if (!new_path) {
1388 namelen = namelen - (cp - name);
1389 } else {
1390 cp = new_path;
1391 namelen = strlen(cp);
1394 strncpy(oc->path, cp,
1395 sizeof(oc->path));
1396 oc->path[sizeof(oc->path)-1] = '\0';
1398 if (!active_cache)
1399 read_cache();
1400 pos = cache_name_pos(cp, namelen);
1401 if (pos < 0)
1402 pos = -pos - 1;
1403 while (pos < active_nr) {
1404 ce = active_cache[pos];
1405 if (ce_namelen(ce) != namelen ||
1406 memcmp(ce->name, cp, namelen))
1407 break;
1408 if (ce_stage(ce) == stage) {
1409 hashcpy(sha1, ce->sha1);
1410 oc->mode = ce->ce_mode;
1411 free(new_path);
1412 return 0;
1414 pos++;
1416 if (only_to_die && name[1] && name[1] != '/')
1417 diagnose_invalid_index_path(stage, prefix, cp);
1418 free(new_path);
1419 return -1;
1421 for (cp = name, bracket_depth = 0; *cp; cp++) {
1422 if (*cp == '{')
1423 bracket_depth++;
1424 else if (bracket_depth && *cp == '}')
1425 bracket_depth--;
1426 else if (!bracket_depth && *cp == ':')
1427 break;
1429 if (*cp == ':') {
1430 unsigned char tree_sha1[20];
1431 int len = cp - name;
1432 if (!get_sha1_1(name, len, tree_sha1, GET_SHA1_TREEISH)) {
1433 const char *filename = cp+1;
1434 char *new_filename = NULL;
1436 new_filename = resolve_relative_path(filename);
1437 if (new_filename)
1438 filename = new_filename;
1439 ret = get_tree_entry(tree_sha1, filename, sha1, &oc->mode);
1440 if (ret && only_to_die) {
1441 diagnose_invalid_sha1_path(prefix, filename,
1442 tree_sha1,
1443 name, len);
1445 hashcpy(oc->tree, tree_sha1);
1446 strncpy(oc->path, filename,
1447 sizeof(oc->path));
1448 oc->path[sizeof(oc->path)-1] = '\0';
1450 free(new_filename);
1451 return ret;
1452 } else {
1453 if (only_to_die)
1454 die("Invalid object name '%.*s'.", len, name);
1457 return ret;
1461 * Call this function when you know "name" given by the end user must
1462 * name an object but it doesn't; the function _may_ die with a better
1463 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1464 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1465 * you have a chance to diagnose the error further.
1467 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1469 struct object_context oc;
1470 unsigned char sha1[20];
1471 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1474 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1476 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);