11 #include "sha1-array.h"
13 #include "object-store.h"
14 #include "repository.h"
16 #include "commit-reach.h"
18 static int get_oid_oneline(const char *, struct object_id *, struct commit_list *);
20 typedef int (*disambiguate_hint_fn)(const struct object_id *, void *);
22 struct disambiguate_state {
23 int len; /* length of prefix in hex chars */
24 char hex_pfx[GIT_MAX_HEXSZ + 1];
25 struct object_id bin_pfx;
27 disambiguate_hint_fn fn;
29 struct object_id candidate;
30 unsigned candidate_exists:1;
31 unsigned candidate_checked:1;
32 unsigned candidate_ok:1;
33 unsigned disambiguate_fn_used:1;
35 unsigned always_call_fn:1;
38 static void update_candidates(struct disambiguate_state *ds, const struct object_id *current)
40 if (ds->always_call_fn) {
41 ds->ambiguous = ds->fn(current, ds->cb_data) ? 1 : 0;
44 if (!ds->candidate_exists) {
45 /* this is the first candidate */
46 oidcpy(&ds->candidate, current);
47 ds->candidate_exists = 1;
49 } else if (oideq(&ds->candidate, current)) {
50 /* the same as what we already have seen */
55 /* cannot disambiguate between ds->candidate and current */
60 if (!ds->candidate_checked) {
61 ds->candidate_ok = ds->fn(&ds->candidate, ds->cb_data);
62 ds->disambiguate_fn_used = 1;
63 ds->candidate_checked = 1;
66 if (!ds->candidate_ok) {
67 /* discard the candidate; we know it does not satisfy fn */
68 oidcpy(&ds->candidate, current);
69 ds->candidate_checked = 0;
73 /* if we reach this point, we know ds->candidate satisfies fn */
74 if (ds->fn(current, ds->cb_data)) {
76 * if both current and candidate satisfy fn, we cannot
83 /* otherwise, current can be discarded and candidate is still good */
86 static int append_loose_object(const struct object_id *oid, const char *path,
89 oid_array_append(data, oid);
93 static int match_sha(unsigned, const unsigned char *, const unsigned char *);
95 static void find_short_object_filename(struct disambiguate_state *ds)
97 int subdir_nr = ds->bin_pfx.hash[0];
98 struct object_directory *odb;
99 struct strbuf buf = STRBUF_INIT;
101 for (odb = the_repository->objects->odb;
102 odb && !ds->ambiguous;
106 if (!odb->loose_objects_subdir_seen[subdir_nr]) {
108 strbuf_addstr(&buf, odb->path);
109 for_each_file_in_obj_subdir(subdir_nr, &buf,
112 &odb->loose_objects_cache);
113 odb->loose_objects_subdir_seen[subdir_nr] = 1;
116 pos = oid_array_lookup(&odb->loose_objects_cache, &ds->bin_pfx);
119 while (!ds->ambiguous && pos < odb->loose_objects_cache.nr) {
120 const struct object_id *oid;
121 oid = odb->loose_objects_cache.oid + pos;
122 if (!match_sha(ds->len, ds->bin_pfx.hash, oid->hash))
124 update_candidates(ds, oid);
129 strbuf_release(&buf);
132 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
142 if ((*a ^ *b) & 0xf0)
147 static void unique_in_midx(struct multi_pack_index *m,
148 struct disambiguate_state *ds)
150 uint32_t num, i, first = 0;
151 const struct object_id *current = NULL;
152 num = m->num_objects;
157 bsearch_midx(&ds->bin_pfx, m, &first);
160 * At this point, "first" is the location of the lowest object
161 * with an object name that could match "bin_pfx". See if we have
162 * 0, 1 or more objects that actually match(es).
164 for (i = first; i < num && !ds->ambiguous; i++) {
165 struct object_id oid;
166 current = nth_midxed_object_oid(&oid, m, i);
167 if (!match_sha(ds->len, ds->bin_pfx.hash, current->hash))
169 update_candidates(ds, current);
173 static void unique_in_pack(struct packed_git *p,
174 struct disambiguate_state *ds)
176 uint32_t num, i, first = 0;
177 const struct object_id *current = NULL;
179 if (open_pack_index(p) || !p->num_objects)
182 num = p->num_objects;
183 bsearch_pack(&ds->bin_pfx, p, &first);
186 * At this point, "first" is the location of the lowest object
187 * with an object name that could match "bin_pfx". See if we have
188 * 0, 1 or more objects that actually match(es).
190 for (i = first; i < num && !ds->ambiguous; i++) {
191 struct object_id oid;
192 current = nth_packed_object_oid(&oid, p, i);
193 if (!match_sha(ds->len, ds->bin_pfx.hash, current->hash))
195 update_candidates(ds, current);
199 static void find_short_packed_object(struct disambiguate_state *ds)
201 struct multi_pack_index *m;
202 struct packed_git *p;
204 for (m = get_multi_pack_index(the_repository); m && !ds->ambiguous;
206 unique_in_midx(m, ds);
207 for (p = get_packed_git(the_repository); p && !ds->ambiguous;
209 unique_in_pack(p, ds);
212 #define SHORT_NAME_NOT_FOUND (-1)
213 #define SHORT_NAME_AMBIGUOUS (-2)
215 static int finish_object_disambiguation(struct disambiguate_state *ds,
216 struct object_id *oid)
219 return SHORT_NAME_AMBIGUOUS;
221 if (!ds->candidate_exists)
222 return SHORT_NAME_NOT_FOUND;
224 if (!ds->candidate_checked)
226 * If this is the only candidate, there is no point
227 * calling the disambiguation hint callback.
229 * On the other hand, if the current candidate
230 * replaced an earlier candidate that did _not_ pass
231 * the disambiguation hint callback, then we do have
232 * more than one objects that match the short name
233 * given, so we should make sure this one matches;
234 * otherwise, if we discovered this one and the one
235 * that we previously discarded in the reverse order,
236 * we would end up showing different results in the
239 ds->candidate_ok = (!ds->disambiguate_fn_used ||
240 ds->fn(&ds->candidate, ds->cb_data));
242 if (!ds->candidate_ok)
243 return SHORT_NAME_AMBIGUOUS;
245 oidcpy(oid, &ds->candidate);
249 static int disambiguate_commit_only(const struct object_id *oid, void *cb_data_unused)
251 int kind = oid_object_info(the_repository, oid, NULL);
252 return kind == OBJ_COMMIT;
255 static int disambiguate_committish_only(const struct object_id *oid, void *cb_data_unused)
260 kind = oid_object_info(the_repository, oid, NULL);
261 if (kind == OBJ_COMMIT)
266 /* We need to do this the hard way... */
267 obj = deref_tag(the_repository, parse_object(the_repository, oid),
269 if (obj && obj->type == OBJ_COMMIT)
274 static int disambiguate_tree_only(const struct object_id *oid, void *cb_data_unused)
276 int kind = oid_object_info(the_repository, oid, NULL);
277 return kind == OBJ_TREE;
280 static int disambiguate_treeish_only(const struct object_id *oid, void *cb_data_unused)
285 kind = oid_object_info(the_repository, oid, NULL);
286 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
291 /* We need to do this the hard way... */
292 obj = deref_tag(the_repository, parse_object(the_repository, oid),
294 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
299 static int disambiguate_blob_only(const struct object_id *oid, void *cb_data_unused)
301 int kind = oid_object_info(the_repository, oid, NULL);
302 return kind == OBJ_BLOB;
305 static disambiguate_hint_fn default_disambiguate_hint;
307 int set_disambiguate_hint_config(const char *var, const char *value)
309 static const struct {
311 disambiguate_hint_fn fn;
314 { "commit", disambiguate_commit_only },
315 { "committish", disambiguate_committish_only },
316 { "tree", disambiguate_tree_only },
317 { "treeish", disambiguate_treeish_only },
318 { "blob", disambiguate_blob_only }
323 return config_error_nonbool(var);
325 for (i = 0; i < ARRAY_SIZE(hints); i++) {
326 if (!strcasecmp(value, hints[i].name)) {
327 default_disambiguate_hint = hints[i].fn;
332 return error("unknown hint type for '%s': %s", var, value);
335 static int init_object_disambiguation(const char *name, int len,
336 struct disambiguate_state *ds)
340 if (len < MINIMUM_ABBREV || len > the_hash_algo->hexsz)
343 memset(ds, 0, sizeof(*ds));
345 for (i = 0; i < len ;i++) {
346 unsigned char c = name[i];
348 if (c >= '0' && c <= '9')
350 else if (c >= 'a' && c <= 'f')
352 else if (c >= 'A' && c <='F') {
361 ds->bin_pfx.hash[i >> 1] |= val;
365 ds->hex_pfx[len] = '\0';
366 prepare_alt_odb(the_repository);
370 static int show_ambiguous_object(const struct object_id *oid, void *data)
372 const struct disambiguate_state *ds = data;
373 struct strbuf desc = STRBUF_INIT;
376 if (ds->fn && !ds->fn(oid, ds->cb_data))
379 type = oid_object_info(the_repository, oid, NULL);
380 if (type == OBJ_COMMIT) {
381 struct commit *commit = lookup_commit(the_repository, oid);
383 struct pretty_print_context pp = {0};
384 pp.date_mode.type = DATE_SHORT;
385 format_commit_message(commit, " %ad - %s", &desc, &pp);
387 } else if (type == OBJ_TAG) {
388 struct tag *tag = lookup_tag(the_repository, oid);
389 if (!parse_tag(tag) && tag->tag)
390 strbuf_addf(&desc, " %s", tag->tag);
394 find_unique_abbrev(oid, DEFAULT_ABBREV),
395 type_name(type) ? type_name(type) : "unknown type",
398 strbuf_release(&desc);
402 static int collect_ambiguous(const struct object_id *oid, void *data)
404 oid_array_append(data, oid);
408 static int sort_ambiguous(const void *a, const void *b)
410 int a_type = oid_object_info(the_repository, a, NULL);
411 int b_type = oid_object_info(the_repository, b, NULL);
416 * Sorts by hash within the same object type, just as
417 * oid_array_for_each_unique() would do.
419 if (a_type == b_type)
423 * Between object types show tags, then commits, and finally
426 * The object_type enum is commit, tree, blob, tag, but we
427 * want tag, commit, tree blob. Cleverly (perhaps too
428 * cleverly) do that with modulus, since the enum assigns 1 to
429 * commit, so tag becomes 0.
431 a_type_sort = a_type % 4;
432 b_type_sort = b_type % 4;
433 return a_type_sort > b_type_sort ? 1 : -1;
436 static int get_short_oid(const char *name, int len, struct object_id *oid,
440 struct disambiguate_state ds;
441 int quietly = !!(flags & GET_OID_QUIETLY);
443 if (init_object_disambiguation(name, len, &ds) < 0)
446 if (HAS_MULTI_BITS(flags & GET_OID_DISAMBIGUATORS))
447 BUG("multiple get_short_oid disambiguator flags");
449 if (flags & GET_OID_COMMIT)
450 ds.fn = disambiguate_commit_only;
451 else if (flags & GET_OID_COMMITTISH)
452 ds.fn = disambiguate_committish_only;
453 else if (flags & GET_OID_TREE)
454 ds.fn = disambiguate_tree_only;
455 else if (flags & GET_OID_TREEISH)
456 ds.fn = disambiguate_treeish_only;
457 else if (flags & GET_OID_BLOB)
458 ds.fn = disambiguate_blob_only;
460 ds.fn = default_disambiguate_hint;
462 find_short_object_filename(&ds);
463 find_short_packed_object(&ds);
464 status = finish_object_disambiguation(&ds, oid);
466 if (!quietly && (status == SHORT_NAME_AMBIGUOUS)) {
467 struct oid_array collect = OID_ARRAY_INIT;
469 error(_("short SHA1 %s is ambiguous"), ds.hex_pfx);
472 * We may still have ambiguity if we simply saw a series of
473 * candidates that did not satisfy our hint function. In
474 * that case, we still want to show them, so disable the hint
480 advise(_("The candidates are:"));
481 for_each_abbrev(ds.hex_pfx, collect_ambiguous, &collect);
482 QSORT(collect.oid, collect.nr, sort_ambiguous);
484 if (oid_array_for_each(&collect, show_ambiguous_object, &ds))
485 BUG("show_ambiguous_object shouldn't return non-zero");
486 oid_array_clear(&collect);
492 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
494 struct oid_array collect = OID_ARRAY_INIT;
495 struct disambiguate_state ds;
498 if (init_object_disambiguation(prefix, strlen(prefix), &ds) < 0)
501 ds.always_call_fn = 1;
502 ds.fn = collect_ambiguous;
503 ds.cb_data = &collect;
504 find_short_object_filename(&ds);
505 find_short_packed_object(&ds);
507 ret = oid_array_for_each_unique(&collect, fn, cb_data);
508 oid_array_clear(&collect);
513 * Return the slot of the most-significant bit set in "val". There are various
514 * ways to do this quickly with fls() or __builtin_clzl(), but speed is
515 * probably not a big deal here.
517 static unsigned msb(unsigned long val)
525 struct min_abbrev_data {
526 unsigned int init_len;
527 unsigned int cur_len;
529 const struct object_id *oid;
532 static inline char get_hex_char_from_oid(const struct object_id *oid,
535 static const char hex[] = "0123456789abcdef";
538 return hex[oid->hash[pos >> 1] >> 4];
540 return hex[oid->hash[pos >> 1] & 0xf];
543 static int extend_abbrev_len(const struct object_id *oid, void *cb_data)
545 struct min_abbrev_data *mad = cb_data;
547 unsigned int i = mad->init_len;
548 while (mad->hex[i] && mad->hex[i] == get_hex_char_from_oid(oid, i))
551 if (i < GIT_MAX_RAWSZ && i >= mad->cur_len)
552 mad->cur_len = i + 1;
557 static void find_abbrev_len_for_midx(struct multi_pack_index *m,
558 struct min_abbrev_data *mad)
561 uint32_t num, first = 0;
562 struct object_id oid;
563 const struct object_id *mad_oid;
568 num = m->num_objects;
570 match = bsearch_midx(mad_oid, m, &first);
573 * first is now the position in the packfile where we would insert
574 * mad->hash if it does not exist (or the position of mad->hash if
575 * it does exist). Hence, we consider a maximum of two objects
576 * nearby for the abbreviation length.
580 if (nth_midxed_object_oid(&oid, m, first))
581 extend_abbrev_len(&oid, mad);
582 } else if (first < num - 1) {
583 if (nth_midxed_object_oid(&oid, m, first + 1))
584 extend_abbrev_len(&oid, mad);
587 if (nth_midxed_object_oid(&oid, m, first - 1))
588 extend_abbrev_len(&oid, mad);
590 mad->init_len = mad->cur_len;
593 static void find_abbrev_len_for_pack(struct packed_git *p,
594 struct min_abbrev_data *mad)
597 uint32_t num, first = 0;
598 struct object_id oid;
599 const struct object_id *mad_oid;
601 if (open_pack_index(p) || !p->num_objects)
604 num = p->num_objects;
606 match = bsearch_pack(mad_oid, p, &first);
609 * first is now the position in the packfile where we would insert
610 * mad->hash if it does not exist (or the position of mad->hash if
611 * it does exist). Hence, we consider a maximum of two objects
612 * nearby for the abbreviation length.
616 if (nth_packed_object_oid(&oid, p, first))
617 extend_abbrev_len(&oid, mad);
618 } else if (first < num - 1) {
619 if (nth_packed_object_oid(&oid, p, first + 1))
620 extend_abbrev_len(&oid, mad);
623 if (nth_packed_object_oid(&oid, p, first - 1))
624 extend_abbrev_len(&oid, mad);
626 mad->init_len = mad->cur_len;
629 static void find_abbrev_len_packed(struct min_abbrev_data *mad)
631 struct multi_pack_index *m;
632 struct packed_git *p;
634 for (m = get_multi_pack_index(the_repository); m; m = m->next)
635 find_abbrev_len_for_midx(m, mad);
636 for (p = get_packed_git(the_repository); p; p = p->next)
637 find_abbrev_len_for_pack(p, mad);
640 int find_unique_abbrev_r(char *hex, const struct object_id *oid, int len)
642 struct disambiguate_state ds;
643 struct min_abbrev_data mad;
644 struct object_id oid_ret;
645 const unsigned hexsz = the_hash_algo->hexsz;
648 unsigned long count = approximate_object_count();
650 * Add one because the MSB only tells us the highest bit set,
651 * not including the value of all the _other_ bits (so "15"
652 * is only one off of 2^4, but the MSB is the 3rd bit.
654 len = msb(count) + 1;
656 * We now know we have on the order of 2^len objects, which
657 * expects a collision at 2^(len/2). But we also care about hex
658 * chars, not bits, and there are 4 bits per hex. So all
659 * together we need to divide by 2 and round up.
661 len = DIV_ROUND_UP(len, 2);
663 * For very small repos, we stick with our regular fallback.
665 if (len < FALLBACK_DEFAULT_ABBREV)
666 len = FALLBACK_DEFAULT_ABBREV;
669 oid_to_hex_r(hex, oid);
670 if (len == hexsz || !len)
678 find_abbrev_len_packed(&mad);
680 if (init_object_disambiguation(hex, mad.cur_len, &ds) < 0)
683 ds.fn = extend_abbrev_len;
684 ds.always_call_fn = 1;
685 ds.cb_data = (void *)&mad;
687 find_short_object_filename(&ds);
688 (void)finish_object_disambiguation(&ds, &oid_ret);
690 hex[mad.cur_len] = 0;
694 const char *find_unique_abbrev(const struct object_id *oid, int len)
697 static char hexbuffer[4][GIT_MAX_HEXSZ + 1];
698 char *hex = hexbuffer[bufno];
699 bufno = (bufno + 1) % ARRAY_SIZE(hexbuffer);
700 find_unique_abbrev_r(hex, oid, len);
704 static int ambiguous_path(const char *path, int len)
709 for (cnt = 0; cnt < len; cnt++) {
729 static inline int at_mark(const char *string, int len,
730 const char **suffix, int nr)
734 for (i = 0; i < nr; i++) {
735 int suffix_len = strlen(suffix[i]);
736 if (suffix_len <= len
737 && !strncasecmp(string, suffix[i], suffix_len))
743 static inline int upstream_mark(const char *string, int len)
745 const char *suffix[] = { "@{upstream}", "@{u}" };
746 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
749 static inline int push_mark(const char *string, int len)
751 const char *suffix[] = { "@{push}" };
752 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
755 static int get_oid_1(const char *name, int len, struct object_id *oid, unsigned lookup_flags);
756 static int interpret_nth_prior_checkout(const char *name, int namelen, struct strbuf *buf);
758 static int get_oid_basic(const char *str, int len, struct object_id *oid,
761 static const char *warn_msg = "refname '%.*s' is ambiguous.";
762 static const char *object_name_msg = N_(
763 "Git normally never creates a ref that ends with 40 hex characters\n"
764 "because it will be ignored when you just specify 40-hex. These refs\n"
765 "may be created by mistake. For example,\n"
767 " git checkout -b $br $(git rev-parse ...)\n"
769 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
770 "examine these refs and maybe delete them. Turn this message off by\n"
771 "running \"git config advice.objectNameWarning false\"");
772 struct object_id tmp_oid;
773 char *real_ref = NULL;
775 int at, reflog_len, nth_prior = 0;
777 if (len == the_hash_algo->hexsz && !get_oid_hex(str, oid)) {
778 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
779 refs_found = dwim_ref(str, len, &tmp_oid, &real_ref);
780 if (refs_found > 0) {
781 warning(warn_msg, len, str);
782 if (advice_object_name_warning)
783 fprintf(stderr, "%s\n", _(object_name_msg));
790 /* basic@{time or number or -number} format to query ref-log */
792 if (len && str[len-1] == '}') {
793 for (at = len-4; at >= 0; at--) {
794 if (str[at] == '@' && str[at+1] == '{') {
795 if (str[at+2] == '-') {
797 /* @{-N} not at start */
802 if (!upstream_mark(str + at, len - at) &&
803 !push_mark(str + at, len - at)) {
804 reflog_len = (len-1) - (at+2);
812 /* Accept only unambiguous ref paths. */
813 if (len && ambiguous_path(str, len))
817 struct strbuf buf = STRBUF_INIT;
820 if (interpret_nth_prior_checkout(str, len, &buf) > 0) {
821 detached = (buf.len == the_hash_algo->hexsz && !get_oid_hex(buf.buf, oid));
822 strbuf_release(&buf);
828 if (!len && reflog_len)
829 /* allow "@{...}" to mean the current branch reflog */
830 refs_found = dwim_ref("HEAD", 4, oid, &real_ref);
832 refs_found = dwim_log(str, len, oid, &real_ref);
834 refs_found = dwim_ref(str, len, oid, &real_ref);
839 if (warn_ambiguous_refs && !(flags & GET_OID_QUIETLY) &&
841 !get_short_oid(str, len, &tmp_oid, GET_OID_QUIETLY)))
842 warning(warn_msg, len, str);
850 /* Is it asking for N-th entry, or approxidate? */
851 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
852 char ch = str[at+2+i];
853 if ('0' <= ch && ch <= '9')
854 nth = nth * 10 + ch - '0';
858 if (100000000 <= nth) {
865 char *tmp = xstrndup(str + at + 2, reflog_len);
866 at_time = approxidate_careful(tmp, &errors);
873 if (read_ref_at(real_ref, flags, at_time, nth, oid, NULL,
874 &co_time, &co_tz, &co_cnt)) {
876 if (starts_with(real_ref, "refs/heads/")) {
878 len = strlen(real_ref + 11);
886 if (!(flags & GET_OID_QUIETLY)) {
887 warning("Log for '%.*s' only goes "
888 "back to %s.", len, str,
889 show_date(co_time, co_tz, DATE_MODE(RFC2822)));
892 if (flags & GET_OID_QUIETLY) {
895 die("Log for '%.*s' only has %d entries.",
905 static int get_parent(const char *name, int len,
906 struct object_id *result, int idx)
908 struct object_id oid;
909 int ret = get_oid_1(name, len, &oid, GET_OID_COMMITTISH);
910 struct commit *commit;
911 struct commit_list *p;
915 commit = lookup_commit_reference(the_repository, &oid);
916 if (parse_commit(commit))
919 oidcpy(result, &commit->object.oid);
925 oidcpy(result, &p->item->object.oid);
933 static int get_nth_ancestor(const char *name, int len,
934 struct object_id *result, int generation)
936 struct object_id oid;
937 struct commit *commit;
940 ret = get_oid_1(name, len, &oid, GET_OID_COMMITTISH);
943 commit = lookup_commit_reference(the_repository, &oid);
947 while (generation--) {
948 if (parse_commit(commit) || !commit->parents)
950 commit = commit->parents->item;
952 oidcpy(result, &commit->object.oid);
956 struct object *peel_to_type(const char *name, int namelen,
957 struct object *o, enum object_type expected_type)
959 if (name && !namelen)
960 namelen = strlen(name);
962 if (!o || (!o->parsed && !parse_object(the_repository, &o->oid)))
964 if (expected_type == OBJ_ANY || o->type == expected_type)
966 if (o->type == OBJ_TAG)
967 o = ((struct tag*) o)->tagged;
968 else if (o->type == OBJ_COMMIT)
969 o = &(get_commit_tree(((struct commit *)o))->object);
972 error("%.*s: expected %s type, but the object "
973 "dereferences to %s type",
974 namelen, name, type_name(expected_type),
981 static int peel_onion(const char *name, int len, struct object_id *oid,
982 unsigned lookup_flags)
984 struct object_id outer;
986 unsigned int expected_type = 0;
990 * "ref^{type}" dereferences ref repeatedly until you cannot
991 * dereference anymore, or you get an object of given type,
992 * whichever comes first. "ref^{}" means just dereference
993 * tags until you get a non-tag. "ref^0" is a shorthand for
994 * "ref^{commit}". "commit^{tree}" could be used to find the
995 * top-level tree of the given commit.
997 if (len < 4 || name[len-1] != '}')
1000 for (sp = name + len - 1; name <= sp; sp--) {
1002 if (ch == '{' && name < sp && sp[-1] == '^')
1008 sp++; /* beginning of type name, or closing brace for empty */
1009 if (starts_with(sp, "commit}"))
1010 expected_type = OBJ_COMMIT;
1011 else if (starts_with(sp, "tag}"))
1012 expected_type = OBJ_TAG;
1013 else if (starts_with(sp, "tree}"))
1014 expected_type = OBJ_TREE;
1015 else if (starts_with(sp, "blob}"))
1016 expected_type = OBJ_BLOB;
1017 else if (starts_with(sp, "object}"))
1018 expected_type = OBJ_ANY;
1019 else if (sp[0] == '}')
1020 expected_type = OBJ_NONE;
1021 else if (sp[0] == '/')
1022 expected_type = OBJ_COMMIT;
1026 lookup_flags &= ~GET_OID_DISAMBIGUATORS;
1027 if (expected_type == OBJ_COMMIT)
1028 lookup_flags |= GET_OID_COMMITTISH;
1029 else if (expected_type == OBJ_TREE)
1030 lookup_flags |= GET_OID_TREEISH;
1032 if (get_oid_1(name, sp - name - 2, &outer, lookup_flags))
1035 o = parse_object(the_repository, &outer);
1038 if (!expected_type) {
1039 o = deref_tag(the_repository, o, name, sp - name - 2);
1040 if (!o || (!o->parsed && !parse_object(the_repository, &o->oid)))
1042 oidcpy(oid, &o->oid);
1047 * At this point, the syntax look correct, so
1048 * if we do not get the needed object, we should
1051 o = peel_to_type(name, len, o, expected_type);
1055 oidcpy(oid, &o->oid);
1057 /* "$commit^{/foo}" */
1060 struct commit_list *list = NULL;
1063 * $commit^{/}. Some regex implementation may reject.
1064 * We don't need regex anyway. '' pattern always matches.
1069 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
1070 commit_list_insert((struct commit *)o, &list);
1071 ret = get_oid_oneline(prefix, oid, list);
1078 static int get_describe_name(const char *name, int len, struct object_id *oid)
1081 unsigned flags = GET_OID_QUIETLY | GET_OID_COMMIT;
1083 for (cp = name + len - 1; name + 2 <= cp; cp--) {
1085 if (!isxdigit(ch)) {
1086 /* We must be looking at g in "SOMETHING-g"
1087 * for it to be describe output.
1089 if (ch == 'g' && cp[-1] == '-') {
1092 return get_short_oid(cp, len, oid, flags);
1099 static int get_oid_1(const char *name, int len, struct object_id *oid, unsigned lookup_flags)
1101 int ret, has_suffix;
1105 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
1108 for (cp = name + len - 1; name <= cp; cp--) {
1110 if ('0' <= ch && ch <= '9')
1112 if (ch == '~' || ch == '^')
1119 int len1 = cp - name;
1121 while (cp < name + len)
1122 num = num * 10 + *cp++ - '0';
1123 if (!num && len1 == len - 1)
1125 if (has_suffix == '^')
1126 return get_parent(name, len1, oid, num);
1127 /* else if (has_suffix == '~') -- goes without saying */
1128 return get_nth_ancestor(name, len1, oid, num);
1131 ret = peel_onion(name, len, oid, lookup_flags);
1135 ret = get_oid_basic(name, len, oid, lookup_flags);
1139 /* It could be describe output that is "SOMETHING-gXXXX" */
1140 ret = get_describe_name(name, len, oid);
1144 return get_short_oid(name, len, oid, lookup_flags);
1148 * This interprets names like ':/Initial revision of "git"' by searching
1149 * through history and returning the first commit whose message starts
1150 * the given regular expression.
1152 * For negative-matching, prefix the pattern-part with '!-', like: ':/!-WIP'.
1154 * For a literal '!' character at the beginning of a pattern, you have to repeat
1155 * that, like: ':/!!foo'
1157 * For future extension, all other sequences beginning with ':/!' are reserved.
1160 /* Remember to update object flag allocation in object.h */
1161 #define ONELINE_SEEN (1u<<20)
1163 static int handle_one_ref(const char *path, const struct object_id *oid,
1164 int flag, void *cb_data)
1166 struct commit_list **list = cb_data;
1167 struct object *object = parse_object(the_repository, oid);
1170 if (object->type == OBJ_TAG) {
1171 object = deref_tag(the_repository, object, path,
1176 if (object->type != OBJ_COMMIT)
1178 commit_list_insert((struct commit *)object, list);
1182 static int get_oid_oneline(const char *prefix, struct object_id *oid,
1183 struct commit_list *list)
1185 struct commit_list *backup = NULL, *l;
1190 if (prefix[0] == '!') {
1193 if (prefix[0] == '-') {
1196 } else if (prefix[0] != '!') {
1201 if (regcomp(®ex, prefix, REG_EXTENDED))
1204 for (l = list; l; l = l->next) {
1205 l->item->object.flags |= ONELINE_SEEN;
1206 commit_list_insert(l->item, &backup);
1209 const char *p, *buf;
1210 struct commit *commit;
1213 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
1214 if (!parse_object(the_repository, &commit->object.oid))
1216 buf = get_commit_buffer(commit, NULL);
1217 p = strstr(buf, "\n\n");
1218 matches = negative ^ (p && !regexec(®ex, p + 2, 0, NULL, 0));
1219 unuse_commit_buffer(commit, buf);
1222 oidcpy(oid, &commit->object.oid);
1228 free_commit_list(list);
1229 for (l = backup; l; l = l->next)
1230 clear_commit_marks(l->item, ONELINE_SEEN);
1231 free_commit_list(backup);
1232 return found ? 0 : -1;
1235 struct grab_nth_branch_switch_cbdata {
1240 static int grab_nth_branch_switch(struct object_id *ooid, struct object_id *noid,
1241 const char *email, timestamp_t timestamp, int tz,
1242 const char *message, void *cb_data)
1244 struct grab_nth_branch_switch_cbdata *cb = cb_data;
1245 const char *match = NULL, *target = NULL;
1248 if (skip_prefix(message, "checkout: moving from ", &match))
1249 target = strstr(match, " to ");
1251 if (!match || !target)
1253 if (--(cb->remaining) == 0) {
1254 len = target - match;
1255 strbuf_reset(&cb->buf);
1256 strbuf_add(&cb->buf, match, len);
1257 return 1; /* we are done */
1263 * Parse @{-N} syntax, return the number of characters parsed
1264 * if successful; otherwise signal an error with negative value.
1266 static int interpret_nth_prior_checkout(const char *name, int namelen,
1271 struct grab_nth_branch_switch_cbdata cb;
1277 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
1279 brace = memchr(name, '}', namelen);
1282 nth = strtol(name + 3, &num_end, 10);
1283 if (num_end != brace)
1288 strbuf_init(&cb.buf, 20);
1291 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
1293 strbuf_addbuf(buf, &cb.buf);
1294 retval = brace - name + 1;
1297 strbuf_release(&cb.buf);
1301 int get_oid_mb(const char *name, struct object_id *oid)
1303 struct commit *one, *two;
1304 struct commit_list *mbs;
1305 struct object_id oid_tmp;
1309 dots = strstr(name, "...");
1311 return get_oid(name, oid);
1313 st = get_oid("HEAD", &oid_tmp);
1316 strbuf_init(&sb, dots - name);
1317 strbuf_add(&sb, name, dots - name);
1318 st = get_oid_committish(sb.buf, &oid_tmp);
1319 strbuf_release(&sb);
1323 one = lookup_commit_reference_gently(the_repository, &oid_tmp, 0);
1327 if (get_oid_committish(dots[3] ? (dots + 3) : "HEAD", &oid_tmp))
1329 two = lookup_commit_reference_gently(the_repository, &oid_tmp, 0);
1332 mbs = get_merge_bases(one, two);
1333 if (!mbs || mbs->next)
1337 oidcpy(oid, &mbs->item->object.oid);
1339 free_commit_list(mbs);
1343 /* parse @something syntax, when 'something' is not {.*} */
1344 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1348 if (len || name[1] == '{')
1351 /* make sure it's a single @, or @@{.*}, not @foo */
1352 next = memchr(name + len + 1, '@', namelen - len - 1);
1353 if (next && next[1] != '{')
1356 next = name + namelen;
1357 if (next != name + 1)
1361 strbuf_add(buf, "HEAD", 4);
1365 static int reinterpret(const char *name, int namelen, int len,
1366 struct strbuf *buf, unsigned allowed)
1368 /* we have extra data, which might need further processing */
1369 struct strbuf tmp = STRBUF_INIT;
1370 int used = buf->len;
1373 strbuf_add(buf, name + len, namelen - len);
1374 ret = interpret_branch_name(buf->buf, buf->len, &tmp, allowed);
1375 /* that data was not interpreted, remove our cruft */
1377 strbuf_setlen(buf, used);
1381 strbuf_addbuf(buf, &tmp);
1382 strbuf_release(&tmp);
1383 /* tweak for size of {-N} versus expanded ref name */
1384 return ret - used + len;
1387 static void set_shortened_ref(struct strbuf *buf, const char *ref)
1389 char *s = shorten_unambiguous_ref(ref, 0);
1391 strbuf_addstr(buf, s);
1395 static int branch_interpret_allowed(const char *refname, unsigned allowed)
1400 if ((allowed & INTERPRET_BRANCH_LOCAL) &&
1401 starts_with(refname, "refs/heads/"))
1403 if ((allowed & INTERPRET_BRANCH_REMOTE) &&
1404 starts_with(refname, "refs/remotes/"))
1410 static int interpret_branch_mark(const char *name, int namelen,
1411 int at, struct strbuf *buf,
1412 int (*get_mark)(const char *, int),
1413 const char *(*get_data)(struct branch *,
1418 struct branch *branch;
1419 struct strbuf err = STRBUF_INIT;
1422 len = get_mark(name + at, namelen - at);
1426 if (memchr(name, ':', at))
1430 char *name_str = xmemdupz(name, at);
1431 branch = branch_get(name_str);
1434 branch = branch_get(NULL);
1436 value = get_data(branch, &err);
1440 if (!branch_interpret_allowed(value, allowed))
1443 set_shortened_ref(buf, value);
1447 int interpret_branch_name(const char *name, int namelen, struct strbuf *buf,
1455 namelen = strlen(name);
1457 if (!allowed || (allowed & INTERPRET_BRANCH_LOCAL)) {
1458 len = interpret_nth_prior_checkout(name, namelen, buf);
1460 return len; /* syntax Ok, not enough switches */
1461 } else if (len > 0) {
1463 return len; /* consumed all */
1465 return reinterpret(name, namelen, len, buf, allowed);
1470 (at = memchr(start, '@', namelen - (start - name)));
1473 if (!allowed || (allowed & INTERPRET_BRANCH_HEAD)) {
1474 len = interpret_empty_at(name, namelen, at - name, buf);
1476 return reinterpret(name, namelen, len, buf,
1480 len = interpret_branch_mark(name, namelen, at - name, buf,
1481 upstream_mark, branch_get_upstream,
1486 len = interpret_branch_mark(name, namelen, at - name, buf,
1487 push_mark, branch_get_push,
1496 void strbuf_branchname(struct strbuf *sb, const char *name, unsigned allowed)
1498 int len = strlen(name);
1499 int used = interpret_branch_name(name, len, sb, allowed);
1503 strbuf_add(sb, name + used, len - used);
1506 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1508 if (startup_info->have_repository)
1509 strbuf_branchname(sb, name, INTERPRET_BRANCH_LOCAL);
1511 strbuf_addstr(sb, name);
1514 * This splice must be done even if we end up rejecting the
1515 * name; builtin/branch.c::copy_or_rename_branch() still wants
1516 * to see what the name expanded to so that "branch -m" can be
1517 * used as a tool to correct earlier mistakes.
1519 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1522 !strcmp(sb->buf, "refs/heads/HEAD"))
1525 return check_refname_format(sb->buf, 0);
1529 * This is like "get_oid_basic()", except it allows "object ID expressions",
1530 * notably "xyz^" for "parent of xyz"
1532 int get_oid(const char *name, struct object_id *oid)
1534 struct object_context unused;
1535 return get_oid_with_context(name, 0, oid, &unused);
1540 * Many callers know that the user meant to name a commit-ish by
1541 * syntactical positions where the object name appears. Calling this
1542 * function allows the machinery to disambiguate shorter-than-unique
1543 * abbreviated object names between commit-ish and others.
1545 * Note that this does NOT error out when the named object is not a
1546 * commit-ish. It is merely to give a hint to the disambiguation
1549 int get_oid_committish(const char *name, struct object_id *oid)
1551 struct object_context unused;
1552 return get_oid_with_context(name, GET_OID_COMMITTISH,
1556 int get_oid_treeish(const char *name, struct object_id *oid)
1558 struct object_context unused;
1559 return get_oid_with_context(name, GET_OID_TREEISH,
1563 int get_oid_commit(const char *name, struct object_id *oid)
1565 struct object_context unused;
1566 return get_oid_with_context(name, GET_OID_COMMIT,
1570 int get_oid_tree(const char *name, struct object_id *oid)
1572 struct object_context unused;
1573 return get_oid_with_context(name, GET_OID_TREE,
1577 int get_oid_blob(const char *name, struct object_id *oid)
1579 struct object_context unused;
1580 return get_oid_with_context(name, GET_OID_BLOB,
1584 /* Must be called only when object_name:filename doesn't exist. */
1585 static void diagnose_invalid_oid_path(const char *prefix,
1586 const char *filename,
1587 const struct object_id *tree_oid,
1588 const char *object_name,
1589 int object_name_len)
1591 struct object_id oid;
1597 if (file_exists(filename))
1598 die("Path '%s' exists on disk, but not in '%.*s'.",
1599 filename, object_name_len, object_name);
1600 if (is_missing_file_error(errno)) {
1601 char *fullname = xstrfmt("%s%s", prefix, filename);
1603 if (!get_tree_entry(tree_oid, fullname, &oid, &mode)) {
1604 die("Path '%s' exists, but not '%s'.\n"
1605 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1608 object_name_len, object_name,
1610 object_name_len, object_name,
1613 die("Path '%s' does not exist in '%.*s'",
1614 filename, object_name_len, object_name);
1618 /* Must be called only when :stage:filename doesn't exist. */
1619 static void diagnose_invalid_index_path(int stage,
1621 const char *filename)
1623 const struct cache_entry *ce;
1625 unsigned namelen = strlen(filename);
1626 struct strbuf fullname = STRBUF_INIT;
1631 /* Wrong stage number? */
1632 pos = cache_name_pos(filename, namelen);
1635 if (pos < active_nr) {
1636 ce = active_cache[pos];
1637 if (ce_namelen(ce) == namelen &&
1638 !memcmp(ce->name, filename, namelen))
1639 die("Path '%s' is in the index, but not at stage %d.\n"
1640 "Did you mean ':%d:%s'?",
1642 ce_stage(ce), filename);
1645 /* Confusion between relative and absolute filenames? */
1646 strbuf_addstr(&fullname, prefix);
1647 strbuf_addstr(&fullname, filename);
1648 pos = cache_name_pos(fullname.buf, fullname.len);
1651 if (pos < active_nr) {
1652 ce = active_cache[pos];
1653 if (ce_namelen(ce) == fullname.len &&
1654 !memcmp(ce->name, fullname.buf, fullname.len))
1655 die("Path '%s' is in the index, but not '%s'.\n"
1656 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1657 fullname.buf, filename,
1658 ce_stage(ce), fullname.buf,
1659 ce_stage(ce), filename);
1662 if (file_exists(filename))
1663 die("Path '%s' exists on disk, but not in the index.", filename);
1664 if (is_missing_file_error(errno))
1665 die("Path '%s' does not exist (neither on disk nor in the index).",
1668 strbuf_release(&fullname);
1672 static char *resolve_relative_path(const char *rel)
1674 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1677 if (!is_inside_work_tree())
1678 die("relative path syntax can't be used outside working tree.");
1680 /* die() inside prefix_path() if resolved path is outside worktree */
1681 return prefix_path(startup_info->prefix,
1682 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1686 static int get_oid_with_context_1(const char *name,
1689 struct object_id *oid,
1690 struct object_context *oc)
1692 int ret, bracket_depth;
1693 int namelen = strlen(name);
1695 int only_to_die = flags & GET_OID_ONLY_TO_DIE;
1698 flags |= GET_OID_QUIETLY;
1700 memset(oc, 0, sizeof(*oc));
1701 oc->mode = S_IFINVALID;
1702 strbuf_init(&oc->symlink_path, 0);
1703 ret = get_oid_1(name, namelen, oid, flags);
1707 * sha1:path --> object name of path in ent sha1
1708 * :path -> object name of absolute path in index
1709 * :./path -> object name of path relative to cwd in index
1710 * :[0-3]:path -> object name of path in index at stage
1711 * :/foo -> recent commit matching foo
1713 if (name[0] == ':') {
1715 const struct cache_entry *ce;
1716 char *new_path = NULL;
1718 if (!only_to_die && namelen > 2 && name[1] == '/') {
1719 struct commit_list *list = NULL;
1721 for_each_ref(handle_one_ref, &list);
1722 head_ref(handle_one_ref, &list);
1723 commit_list_sort_by_date(&list);
1724 return get_oid_oneline(name + 2, oid, list);
1728 name[1] < '0' || '3' < name[1])
1731 stage = name[1] - '0';
1734 new_path = resolve_relative_path(cp);
1736 namelen = namelen - (cp - name);
1739 namelen = strlen(cp);
1742 if (flags & GET_OID_RECORD_PATH)
1743 oc->path = xstrdup(cp);
1747 pos = cache_name_pos(cp, namelen);
1750 while (pos < active_nr) {
1751 ce = active_cache[pos];
1752 if (ce_namelen(ce) != namelen ||
1753 memcmp(ce->name, cp, namelen))
1755 if (ce_stage(ce) == stage) {
1756 oidcpy(oid, &ce->oid);
1757 oc->mode = ce->ce_mode;
1763 if (only_to_die && name[1] && name[1] != '/')
1764 diagnose_invalid_index_path(stage, prefix, cp);
1768 for (cp = name, bracket_depth = 0; *cp; cp++) {
1771 else if (bracket_depth && *cp == '}')
1773 else if (!bracket_depth && *cp == ':')
1777 struct object_id tree_oid;
1778 int len = cp - name;
1779 unsigned sub_flags = flags;
1781 sub_flags &= ~GET_OID_DISAMBIGUATORS;
1782 sub_flags |= GET_OID_TREEISH;
1784 if (!get_oid_1(name, len, &tree_oid, sub_flags)) {
1785 const char *filename = cp+1;
1786 char *new_filename = NULL;
1788 new_filename = resolve_relative_path(filename);
1790 filename = new_filename;
1791 if (flags & GET_OID_FOLLOW_SYMLINKS) {
1792 ret = get_tree_entry_follow_symlinks(&tree_oid,
1793 filename, oid, &oc->symlink_path,
1796 ret = get_tree_entry(&tree_oid, filename, oid,
1798 if (ret && only_to_die) {
1799 diagnose_invalid_oid_path(prefix,
1805 if (flags & GET_OID_RECORD_PATH)
1806 oc->path = xstrdup(filename);
1812 die("Invalid object name '%.*s'.", len, name);
1819 * Call this function when you know "name" given by the end user must
1820 * name an object but it doesn't; the function _may_ die with a better
1821 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1822 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1823 * you have a chance to diagnose the error further.
1825 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1827 struct object_context oc;
1828 struct object_id oid;
1829 get_oid_with_context_1(name, GET_OID_ONLY_TO_DIE, prefix, &oid, &oc);
1832 int get_oid_with_context(const char *str, unsigned flags, struct object_id *oid, struct object_context *oc)
1834 if (flags & GET_OID_FOLLOW_SYMLINKS && flags & GET_OID_ONLY_TO_DIE)
1835 BUG("incompatible flags for get_sha1_with_context");
1836 return get_oid_with_context_1(str, flags, NULL, oid, oc);