t5547-push-quarantine: run the path separator test on Windows, too
[git/git-svn.git] / sha1_name.c
blobdefbb3eb05aa3d82c86549337f9cc56e1f6076bd
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"
9 #include "dir.h"
11 static int get_sha1_oneline(const char *, unsigned char *, struct commit_list *);
13 typedef int (*disambiguate_hint_fn)(const unsigned char *, void *);
15 struct disambiguate_state {
16 disambiguate_hint_fn fn;
17 void *cb_data;
18 unsigned char candidate[20];
19 unsigned candidate_exists:1;
20 unsigned candidate_checked:1;
21 unsigned candidate_ok:1;
22 unsigned disambiguate_fn_used:1;
23 unsigned ambiguous:1;
24 unsigned always_call_fn:1;
27 static void update_candidates(struct disambiguate_state *ds, const unsigned char *current)
29 if (ds->always_call_fn) {
30 ds->ambiguous = ds->fn(current, ds->cb_data) ? 1 : 0;
31 return;
33 if (!ds->candidate_exists) {
34 /* this is the first candidate */
35 hashcpy(ds->candidate, current);
36 ds->candidate_exists = 1;
37 return;
38 } else if (!hashcmp(ds->candidate, current)) {
39 /* the same as what we already have seen */
40 return;
43 if (!ds->fn) {
44 /* cannot disambiguate between ds->candidate and current */
45 ds->ambiguous = 1;
46 return;
49 if (!ds->candidate_checked) {
50 ds->candidate_ok = ds->fn(ds->candidate, ds->cb_data);
51 ds->disambiguate_fn_used = 1;
52 ds->candidate_checked = 1;
55 if (!ds->candidate_ok) {
56 /* discard the candidate; we know it does not satisfy fn */
57 hashcpy(ds->candidate, current);
58 ds->candidate_checked = 0;
59 return;
62 /* if we reach this point, we know ds->candidate satisfies fn */
63 if (ds->fn(current, ds->cb_data)) {
65 * if both current and candidate satisfy fn, we cannot
66 * disambiguate.
68 ds->candidate_ok = 0;
69 ds->ambiguous = 1;
72 /* otherwise, current can be discarded and candidate is still good */
75 static void find_short_object_filename(int len, const char *hex_pfx, struct disambiguate_state *ds)
77 struct alternate_object_database *alt;
78 char hex[40];
79 static struct alternate_object_database *fakeent;
81 if (!fakeent) {
83 * Create a "fake" alternate object database that
84 * points to our own object database, to make it
85 * easier to get a temporary working space in
86 * alt->name/alt->base while iterating over the
87 * object databases including our own.
89 fakeent = alloc_alt_odb(get_object_directory());
91 fakeent->next = alt_odb_list;
93 xsnprintf(hex, sizeof(hex), "%.2s", hex_pfx);
94 for (alt = fakeent; alt && !ds->ambiguous; alt = alt->next) {
95 struct strbuf *buf = alt_scratch_buf(alt);
96 struct dirent *de;
97 DIR *dir;
99 strbuf_addf(buf, "%.2s/", hex_pfx);
100 dir = opendir(buf->buf);
101 if (!dir)
102 continue;
104 while (!ds->ambiguous && (de = readdir(dir)) != NULL) {
105 unsigned char sha1[20];
107 if (strlen(de->d_name) != 38)
108 continue;
109 if (memcmp(de->d_name, hex_pfx + 2, len - 2))
110 continue;
111 memcpy(hex + 2, de->d_name, 38);
112 if (!get_sha1_hex(hex, sha1))
113 update_candidates(ds, sha1);
115 closedir(dir);
119 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
121 do {
122 if (*a != *b)
123 return 0;
124 a++;
125 b++;
126 len -= 2;
127 } while (len > 1);
128 if (len)
129 if ((*a ^ *b) & 0xf0)
130 return 0;
131 return 1;
134 static void unique_in_pack(int len,
135 const unsigned char *bin_pfx,
136 struct packed_git *p,
137 struct disambiguate_state *ds)
139 uint32_t num, last, i, first = 0;
140 const unsigned char *current = NULL;
142 open_pack_index(p);
143 num = p->num_objects;
144 last = num;
145 while (first < last) {
146 uint32_t mid = (first + last) / 2;
147 const unsigned char *current;
148 int cmp;
150 current = nth_packed_object_sha1(p, mid);
151 cmp = hashcmp(bin_pfx, current);
152 if (!cmp) {
153 first = mid;
154 break;
156 if (cmp > 0) {
157 first = mid+1;
158 continue;
160 last = mid;
164 * At this point, "first" is the location of the lowest object
165 * with an object name that could match "bin_pfx". See if we have
166 * 0, 1 or more objects that actually match(es).
168 for (i = first; i < num && !ds->ambiguous; i++) {
169 current = nth_packed_object_sha1(p, i);
170 if (!match_sha(len, bin_pfx, current))
171 break;
172 update_candidates(ds, current);
176 static void find_short_packed_object(int len, const unsigned char *bin_pfx,
177 struct disambiguate_state *ds)
179 struct packed_git *p;
181 prepare_packed_git();
182 for (p = packed_git; p && !ds->ambiguous; p = p->next)
183 unique_in_pack(len, bin_pfx, p, ds);
186 #define SHORT_NAME_NOT_FOUND (-1)
187 #define SHORT_NAME_AMBIGUOUS (-2)
189 static int finish_object_disambiguation(struct disambiguate_state *ds,
190 unsigned char *sha1)
192 if (ds->ambiguous)
193 return SHORT_NAME_AMBIGUOUS;
195 if (!ds->candidate_exists)
196 return SHORT_NAME_NOT_FOUND;
198 if (!ds->candidate_checked)
200 * If this is the only candidate, there is no point
201 * calling the disambiguation hint callback.
203 * On the other hand, if the current candidate
204 * replaced an earlier candidate that did _not_ pass
205 * the disambiguation hint callback, then we do have
206 * more than one objects that match the short name
207 * given, so we should make sure this one matches;
208 * otherwise, if we discovered this one and the one
209 * that we previously discarded in the reverse order,
210 * we would end up showing different results in the
211 * same repository!
213 ds->candidate_ok = (!ds->disambiguate_fn_used ||
214 ds->fn(ds->candidate, ds->cb_data));
216 if (!ds->candidate_ok)
217 return SHORT_NAME_AMBIGUOUS;
219 hashcpy(sha1, ds->candidate);
220 return 0;
223 static int disambiguate_commit_only(const unsigned char *sha1, void *cb_data_unused)
225 int kind = sha1_object_info(sha1, NULL);
226 return kind == OBJ_COMMIT;
229 static int disambiguate_committish_only(const unsigned char *sha1, void *cb_data_unused)
231 struct object *obj;
232 int kind;
234 kind = sha1_object_info(sha1, NULL);
235 if (kind == OBJ_COMMIT)
236 return 1;
237 if (kind != OBJ_TAG)
238 return 0;
240 /* We need to do this the hard way... */
241 obj = deref_tag(parse_object(sha1), NULL, 0);
242 if (obj && obj->type == OBJ_COMMIT)
243 return 1;
244 return 0;
247 static int disambiguate_tree_only(const unsigned char *sha1, void *cb_data_unused)
249 int kind = sha1_object_info(sha1, NULL);
250 return kind == OBJ_TREE;
253 static int disambiguate_treeish_only(const unsigned char *sha1, void *cb_data_unused)
255 struct object *obj;
256 int kind;
258 kind = sha1_object_info(sha1, NULL);
259 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
260 return 1;
261 if (kind != OBJ_TAG)
262 return 0;
264 /* We need to do this the hard way... */
265 obj = deref_tag(lookup_object(sha1), NULL, 0);
266 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
267 return 1;
268 return 0;
271 static int disambiguate_blob_only(const unsigned char *sha1, void *cb_data_unused)
273 int kind = sha1_object_info(sha1, NULL);
274 return kind == OBJ_BLOB;
277 static int prepare_prefixes(const char *name, int len,
278 unsigned char *bin_pfx,
279 char *hex_pfx)
281 int i;
283 hashclr(bin_pfx);
284 memset(hex_pfx, 'x', 40);
285 for (i = 0; i < len ;i++) {
286 unsigned char c = name[i];
287 unsigned char val;
288 if (c >= '0' && c <= '9')
289 val = c - '0';
290 else if (c >= 'a' && c <= 'f')
291 val = c - 'a' + 10;
292 else if (c >= 'A' && c <='F') {
293 val = c - 'A' + 10;
294 c -= 'A' - 'a';
296 else
297 return -1;
298 hex_pfx[i] = c;
299 if (!(i & 1))
300 val <<= 4;
301 bin_pfx[i >> 1] |= val;
303 return 0;
306 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
307 unsigned flags)
309 int status;
310 char hex_pfx[40];
311 unsigned char bin_pfx[20];
312 struct disambiguate_state ds;
313 int quietly = !!(flags & GET_SHA1_QUIETLY);
315 if (len < MINIMUM_ABBREV || len > 40)
316 return -1;
317 if (prepare_prefixes(name, len, bin_pfx, hex_pfx) < 0)
318 return -1;
320 prepare_alt_odb();
322 memset(&ds, 0, sizeof(ds));
323 if (flags & GET_SHA1_COMMIT)
324 ds.fn = disambiguate_commit_only;
325 else if (flags & GET_SHA1_COMMITTISH)
326 ds.fn = disambiguate_committish_only;
327 else if (flags & GET_SHA1_TREE)
328 ds.fn = disambiguate_tree_only;
329 else if (flags & GET_SHA1_TREEISH)
330 ds.fn = disambiguate_treeish_only;
331 else if (flags & GET_SHA1_BLOB)
332 ds.fn = disambiguate_blob_only;
334 find_short_object_filename(len, hex_pfx, &ds);
335 find_short_packed_object(len, bin_pfx, &ds);
336 status = finish_object_disambiguation(&ds, sha1);
338 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
339 return error("short SHA1 %.*s is ambiguous.", len, hex_pfx);
340 return status;
343 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
345 char hex_pfx[40];
346 unsigned char bin_pfx[20];
347 struct disambiguate_state ds;
348 int len = strlen(prefix);
350 if (len < MINIMUM_ABBREV || len > 40)
351 return -1;
352 if (prepare_prefixes(prefix, len, bin_pfx, hex_pfx) < 0)
353 return -1;
355 prepare_alt_odb();
357 memset(&ds, 0, sizeof(ds));
358 ds.always_call_fn = 1;
359 ds.cb_data = cb_data;
360 ds.fn = fn;
362 find_short_object_filename(len, hex_pfx, &ds);
363 find_short_packed_object(len, bin_pfx, &ds);
364 return ds.ambiguous;
367 int find_unique_abbrev_r(char *hex, const unsigned char *sha1, int len)
369 int status, exists;
371 sha1_to_hex_r(hex, sha1);
372 if (len == 40 || !len)
373 return 40;
374 exists = has_sha1_file(sha1);
375 while (len < 40) {
376 unsigned char sha1_ret[20];
377 status = get_short_sha1(hex, len, sha1_ret, GET_SHA1_QUIETLY);
378 if (exists
379 ? !status
380 : status == SHORT_NAME_NOT_FOUND) {
381 hex[len] = 0;
382 return len;
384 len++;
386 return len;
389 const char *find_unique_abbrev(const unsigned char *sha1, int len)
391 static char hex[GIT_SHA1_HEXSZ + 1];
392 find_unique_abbrev_r(hex, sha1, len);
393 return hex;
396 static int ambiguous_path(const char *path, int len)
398 int slash = 1;
399 int cnt;
401 for (cnt = 0; cnt < len; cnt++) {
402 switch (*path++) {
403 case '\0':
404 break;
405 case '/':
406 if (slash)
407 break;
408 slash = 1;
409 continue;
410 case '.':
411 continue;
412 default:
413 slash = 0;
414 continue;
416 break;
418 return slash;
421 static inline int at_mark(const char *string, int len,
422 const char **suffix, int nr)
424 int i;
426 for (i = 0; i < nr; i++) {
427 int suffix_len = strlen(suffix[i]);
428 if (suffix_len <= len
429 && !memcmp(string, suffix[i], suffix_len))
430 return suffix_len;
432 return 0;
435 static inline int upstream_mark(const char *string, int len)
437 const char *suffix[] = { "@{upstream}", "@{u}" };
438 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
441 static inline int push_mark(const char *string, int len)
443 const char *suffix[] = { "@{push}" };
444 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
447 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
448 static int interpret_nth_prior_checkout(const char *name, int namelen, struct strbuf *buf);
450 static int get_sha1_basic(const char *str, int len, unsigned char *sha1,
451 unsigned int flags)
453 static const char *warn_msg = "refname '%.*s' is ambiguous.";
454 static const char *object_name_msg = N_(
455 "Git normally never creates a ref that ends with 40 hex characters\n"
456 "because it will be ignored when you just specify 40-hex. These refs\n"
457 "may be created by mistake. For example,\n"
458 "\n"
459 " git checkout -b $br $(git rev-parse ...)\n"
460 "\n"
461 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
462 "examine these refs and maybe delete them. Turn this message off by\n"
463 "running \"git config advice.objectNameWarning false\"");
464 unsigned char tmp_sha1[20];
465 char *real_ref = NULL;
466 int refs_found = 0;
467 int at, reflog_len, nth_prior = 0;
469 if (len == 40 && !get_sha1_hex(str, sha1)) {
470 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
471 refs_found = dwim_ref(str, len, tmp_sha1, &real_ref);
472 if (refs_found > 0) {
473 warning(warn_msg, len, str);
474 if (advice_object_name_warning)
475 fprintf(stderr, "%s\n", _(object_name_msg));
477 free(real_ref);
479 return 0;
482 /* basic@{time or number or -number} format to query ref-log */
483 reflog_len = at = 0;
484 if (len && str[len-1] == '}') {
485 for (at = len-4; at >= 0; at--) {
486 if (str[at] == '@' && str[at+1] == '{') {
487 if (str[at+2] == '-') {
488 if (at != 0)
489 /* @{-N} not at start */
490 return -1;
491 nth_prior = 1;
492 continue;
494 if (!upstream_mark(str + at, len - at) &&
495 !push_mark(str + at, len - at)) {
496 reflog_len = (len-1) - (at+2);
497 len = at;
499 break;
504 /* Accept only unambiguous ref paths. */
505 if (len && ambiguous_path(str, len))
506 return -1;
508 if (nth_prior) {
509 struct strbuf buf = STRBUF_INIT;
510 int detached;
512 if (interpret_nth_prior_checkout(str, len, &buf) > 0) {
513 detached = (buf.len == 40 && !get_sha1_hex(buf.buf, sha1));
514 strbuf_release(&buf);
515 if (detached)
516 return 0;
520 if (!len && reflog_len)
521 /* allow "@{...}" to mean the current branch reflog */
522 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
523 else if (reflog_len)
524 refs_found = dwim_log(str, len, sha1, &real_ref);
525 else
526 refs_found = dwim_ref(str, len, sha1, &real_ref);
528 if (!refs_found)
529 return -1;
531 if (warn_ambiguous_refs && !(flags & GET_SHA1_QUIETLY) &&
532 (refs_found > 1 ||
533 !get_short_sha1(str, len, tmp_sha1, GET_SHA1_QUIETLY)))
534 warning(warn_msg, len, str);
536 if (reflog_len) {
537 int nth, i;
538 unsigned long at_time;
539 unsigned long co_time;
540 int co_tz, co_cnt;
542 /* Is it asking for N-th entry, or approxidate? */
543 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
544 char ch = str[at+2+i];
545 if ('0' <= ch && ch <= '9')
546 nth = nth * 10 + ch - '0';
547 else
548 nth = -1;
550 if (100000000 <= nth) {
551 at_time = nth;
552 nth = -1;
553 } else if (0 <= nth)
554 at_time = 0;
555 else {
556 int errors = 0;
557 char *tmp = xstrndup(str + at + 2, reflog_len);
558 at_time = approxidate_careful(tmp, &errors);
559 free(tmp);
560 if (errors) {
561 free(real_ref);
562 return -1;
565 if (read_ref_at(real_ref, flags, at_time, nth, sha1, NULL,
566 &co_time, &co_tz, &co_cnt)) {
567 if (!len) {
568 if (starts_with(real_ref, "refs/heads/")) {
569 str = real_ref + 11;
570 len = strlen(real_ref + 11);
571 } else {
572 /* detached HEAD */
573 str = "HEAD";
574 len = 4;
577 if (at_time) {
578 if (!(flags & GET_SHA1_QUIETLY)) {
579 warning("Log for '%.*s' only goes "
580 "back to %s.", len, str,
581 show_date(co_time, co_tz, DATE_MODE(RFC2822)));
583 } else {
584 if (flags & GET_SHA1_QUIETLY) {
585 exit(128);
587 die("Log for '%.*s' only has %d entries.",
588 len, str, co_cnt);
593 free(real_ref);
594 return 0;
597 static int get_parent(const char *name, int len,
598 unsigned char *result, int idx)
600 unsigned char sha1[20];
601 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
602 struct commit *commit;
603 struct commit_list *p;
605 if (ret)
606 return ret;
607 commit = lookup_commit_reference(sha1);
608 if (parse_commit(commit))
609 return -1;
610 if (!idx) {
611 hashcpy(result, commit->object.oid.hash);
612 return 0;
614 p = commit->parents;
615 while (p) {
616 if (!--idx) {
617 hashcpy(result, p->item->object.oid.hash);
618 return 0;
620 p = p->next;
622 return -1;
625 static int get_nth_ancestor(const char *name, int len,
626 unsigned char *result, int generation)
628 unsigned char sha1[20];
629 struct commit *commit;
630 int ret;
632 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
633 if (ret)
634 return ret;
635 commit = lookup_commit_reference(sha1);
636 if (!commit)
637 return -1;
639 while (generation--) {
640 if (parse_commit(commit) || !commit->parents)
641 return -1;
642 commit = commit->parents->item;
644 hashcpy(result, commit->object.oid.hash);
645 return 0;
648 struct object *peel_to_type(const char *name, int namelen,
649 struct object *o, enum object_type expected_type)
651 if (name && !namelen)
652 namelen = strlen(name);
653 while (1) {
654 if (!o || (!o->parsed && !parse_object(o->oid.hash)))
655 return NULL;
656 if (expected_type == OBJ_ANY || o->type == expected_type)
657 return o;
658 if (o->type == OBJ_TAG)
659 o = ((struct tag*) o)->tagged;
660 else if (o->type == OBJ_COMMIT)
661 o = &(((struct commit *) o)->tree->object);
662 else {
663 if (name)
664 error("%.*s: expected %s type, but the object "
665 "dereferences to %s type",
666 namelen, name, typename(expected_type),
667 typename(o->type));
668 return NULL;
673 static int peel_onion(const char *name, int len, unsigned char *sha1)
675 unsigned char outer[20];
676 const char *sp;
677 unsigned int expected_type = 0;
678 unsigned lookup_flags = 0;
679 struct object *o;
682 * "ref^{type}" dereferences ref repeatedly until you cannot
683 * dereference anymore, or you get an object of given type,
684 * whichever comes first. "ref^{}" means just dereference
685 * tags until you get a non-tag. "ref^0" is a shorthand for
686 * "ref^{commit}". "commit^{tree}" could be used to find the
687 * top-level tree of the given commit.
689 if (len < 4 || name[len-1] != '}')
690 return -1;
692 for (sp = name + len - 1; name <= sp; sp--) {
693 int ch = *sp;
694 if (ch == '{' && name < sp && sp[-1] == '^')
695 break;
697 if (sp <= name)
698 return -1;
700 sp++; /* beginning of type name, or closing brace for empty */
701 if (starts_with(sp, "commit}"))
702 expected_type = OBJ_COMMIT;
703 else if (starts_with(sp, "tag}"))
704 expected_type = OBJ_TAG;
705 else if (starts_with(sp, "tree}"))
706 expected_type = OBJ_TREE;
707 else if (starts_with(sp, "blob}"))
708 expected_type = OBJ_BLOB;
709 else if (starts_with(sp, "object}"))
710 expected_type = OBJ_ANY;
711 else if (sp[0] == '}')
712 expected_type = OBJ_NONE;
713 else if (sp[0] == '/')
714 expected_type = OBJ_COMMIT;
715 else
716 return -1;
718 if (expected_type == OBJ_COMMIT)
719 lookup_flags = GET_SHA1_COMMITTISH;
720 else if (expected_type == OBJ_TREE)
721 lookup_flags = GET_SHA1_TREEISH;
723 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
724 return -1;
726 o = parse_object(outer);
727 if (!o)
728 return -1;
729 if (!expected_type) {
730 o = deref_tag(o, name, sp - name - 2);
731 if (!o || (!o->parsed && !parse_object(o->oid.hash)))
732 return -1;
733 hashcpy(sha1, o->oid.hash);
734 return 0;
738 * At this point, the syntax look correct, so
739 * if we do not get the needed object, we should
740 * barf.
742 o = peel_to_type(name, len, o, expected_type);
743 if (!o)
744 return -1;
746 hashcpy(sha1, o->oid.hash);
747 if (sp[0] == '/') {
748 /* "$commit^{/foo}" */
749 char *prefix;
750 int ret;
751 struct commit_list *list = NULL;
754 * $commit^{/}. Some regex implementation may reject.
755 * We don't need regex anyway. '' pattern always matches.
757 if (sp[1] == '}')
758 return 0;
760 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
761 commit_list_insert((struct commit *)o, &list);
762 ret = get_sha1_oneline(prefix, sha1, list);
763 free(prefix);
764 return ret;
766 return 0;
769 static int get_describe_name(const char *name, int len, unsigned char *sha1)
771 const char *cp;
772 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
774 for (cp = name + len - 1; name + 2 <= cp; cp--) {
775 char ch = *cp;
776 if (!isxdigit(ch)) {
777 /* We must be looking at g in "SOMETHING-g"
778 * for it to be describe output.
780 if (ch == 'g' && cp[-1] == '-') {
781 cp++;
782 len -= cp - name;
783 return get_short_sha1(cp, len, sha1, flags);
787 return -1;
790 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
792 int ret, has_suffix;
793 const char *cp;
796 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
798 has_suffix = 0;
799 for (cp = name + len - 1; name <= cp; cp--) {
800 int ch = *cp;
801 if ('0' <= ch && ch <= '9')
802 continue;
803 if (ch == '~' || ch == '^')
804 has_suffix = ch;
805 break;
808 if (has_suffix) {
809 int num = 0;
810 int len1 = cp - name;
811 cp++;
812 while (cp < name + len)
813 num = num * 10 + *cp++ - '0';
814 if (!num && len1 == len - 1)
815 num = 1;
816 if (has_suffix == '^')
817 return get_parent(name, len1, sha1, num);
818 /* else if (has_suffix == '~') -- goes without saying */
819 return get_nth_ancestor(name, len1, sha1, num);
822 ret = peel_onion(name, len, sha1);
823 if (!ret)
824 return 0;
826 ret = get_sha1_basic(name, len, sha1, lookup_flags);
827 if (!ret)
828 return 0;
830 /* It could be describe output that is "SOMETHING-gXXXX" */
831 ret = get_describe_name(name, len, sha1);
832 if (!ret)
833 return 0;
835 return get_short_sha1(name, len, sha1, lookup_flags);
839 * This interprets names like ':/Initial revision of "git"' by searching
840 * through history and returning the first commit whose message starts
841 * the given regular expression.
843 * For negative-matching, prefix the pattern-part with '!-', like: ':/!-WIP'.
845 * For a literal '!' character at the beginning of a pattern, you have to repeat
846 * that, like: ':/!!foo'
848 * For future extension, all other sequences beginning with ':/!' are reserved.
851 /* Remember to update object flag allocation in object.h */
852 #define ONELINE_SEEN (1u<<20)
854 static int handle_one_ref(const char *path, const struct object_id *oid,
855 int flag, void *cb_data)
857 struct commit_list **list = cb_data;
858 struct object *object = parse_object(oid->hash);
859 if (!object)
860 return 0;
861 if (object->type == OBJ_TAG) {
862 object = deref_tag(object, path, strlen(path));
863 if (!object)
864 return 0;
866 if (object->type != OBJ_COMMIT)
867 return 0;
868 commit_list_insert((struct commit *)object, list);
869 return 0;
872 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
873 struct commit_list *list)
875 struct commit_list *backup = NULL, *l;
876 int found = 0;
877 int negative = 0;
878 regex_t regex;
880 if (prefix[0] == '!') {
881 prefix++;
883 if (prefix[0] == '-') {
884 prefix++;
885 negative = 1;
886 } else if (prefix[0] != '!') {
887 return -1;
891 if (regcomp(&regex, prefix, REG_EXTENDED))
892 return -1;
894 for (l = list; l; l = l->next) {
895 l->item->object.flags |= ONELINE_SEEN;
896 commit_list_insert(l->item, &backup);
898 while (list) {
899 const char *p, *buf;
900 struct commit *commit;
901 int matches;
903 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
904 if (!parse_object(commit->object.oid.hash))
905 continue;
906 buf = get_commit_buffer(commit, NULL);
907 p = strstr(buf, "\n\n");
908 matches = negative ^ (p && !regexec(&regex, p + 2, 0, NULL, 0));
909 unuse_commit_buffer(commit, buf);
911 if (matches) {
912 hashcpy(sha1, commit->object.oid.hash);
913 found = 1;
914 break;
917 regfree(&regex);
918 free_commit_list(list);
919 for (l = backup; l; l = l->next)
920 clear_commit_marks(l->item, ONELINE_SEEN);
921 free_commit_list(backup);
922 return found ? 0 : -1;
925 struct grab_nth_branch_switch_cbdata {
926 int remaining;
927 struct strbuf buf;
930 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
931 const char *email, unsigned long timestamp, int tz,
932 const char *message, void *cb_data)
934 struct grab_nth_branch_switch_cbdata *cb = cb_data;
935 const char *match = NULL, *target = NULL;
936 size_t len;
938 if (skip_prefix(message, "checkout: moving from ", &match))
939 target = strstr(match, " to ");
941 if (!match || !target)
942 return 0;
943 if (--(cb->remaining) == 0) {
944 len = target - match;
945 strbuf_reset(&cb->buf);
946 strbuf_add(&cb->buf, match, len);
947 return 1; /* we are done */
949 return 0;
953 * Parse @{-N} syntax, return the number of characters parsed
954 * if successful; otherwise signal an error with negative value.
956 static int interpret_nth_prior_checkout(const char *name, int namelen,
957 struct strbuf *buf)
959 long nth;
960 int retval;
961 struct grab_nth_branch_switch_cbdata cb;
962 const char *brace;
963 char *num_end;
965 if (namelen < 4)
966 return -1;
967 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
968 return -1;
969 brace = memchr(name, '}', namelen);
970 if (!brace)
971 return -1;
972 nth = strtol(name + 3, &num_end, 10);
973 if (num_end != brace)
974 return -1;
975 if (nth <= 0)
976 return -1;
977 cb.remaining = nth;
978 strbuf_init(&cb.buf, 20);
980 retval = 0;
981 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
982 strbuf_reset(buf);
983 strbuf_addbuf(buf, &cb.buf);
984 retval = brace - name + 1;
987 strbuf_release(&cb.buf);
988 return retval;
991 int get_oid_mb(const char *name, struct object_id *oid)
993 struct commit *one, *two;
994 struct commit_list *mbs;
995 struct object_id oid_tmp;
996 const char *dots;
997 int st;
999 dots = strstr(name, "...");
1000 if (!dots)
1001 return get_oid(name, oid);
1002 if (dots == name)
1003 st = get_oid("HEAD", &oid_tmp);
1004 else {
1005 struct strbuf sb;
1006 strbuf_init(&sb, dots - name);
1007 strbuf_add(&sb, name, dots - name);
1008 st = get_sha1_committish(sb.buf, oid_tmp.hash);
1009 strbuf_release(&sb);
1011 if (st)
1012 return st;
1013 one = lookup_commit_reference_gently(oid_tmp.hash, 0);
1014 if (!one)
1015 return -1;
1017 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", oid_tmp.hash))
1018 return -1;
1019 two = lookup_commit_reference_gently(oid_tmp.hash, 0);
1020 if (!two)
1021 return -1;
1022 mbs = get_merge_bases(one, two);
1023 if (!mbs || mbs->next)
1024 st = -1;
1025 else {
1026 st = 0;
1027 oidcpy(oid, &mbs->item->object.oid);
1029 free_commit_list(mbs);
1030 return st;
1033 /* parse @something syntax, when 'something' is not {.*} */
1034 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1036 const char *next;
1038 if (len || name[1] == '{')
1039 return -1;
1041 /* make sure it's a single @, or @@{.*}, not @foo */
1042 next = memchr(name + len + 1, '@', namelen - len - 1);
1043 if (next && next[1] != '{')
1044 return -1;
1045 if (!next)
1046 next = name + namelen;
1047 if (next != name + 1)
1048 return -1;
1050 strbuf_reset(buf);
1051 strbuf_add(buf, "HEAD", 4);
1052 return 1;
1055 static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
1057 /* we have extra data, which might need further processing */
1058 struct strbuf tmp = STRBUF_INIT;
1059 int used = buf->len;
1060 int ret;
1062 strbuf_add(buf, name + len, namelen - len);
1063 ret = interpret_branch_name(buf->buf, buf->len, &tmp);
1064 /* that data was not interpreted, remove our cruft */
1065 if (ret < 0) {
1066 strbuf_setlen(buf, used);
1067 return len;
1069 strbuf_reset(buf);
1070 strbuf_addbuf(buf, &tmp);
1071 strbuf_release(&tmp);
1072 /* tweak for size of {-N} versus expanded ref name */
1073 return ret - used + len;
1076 static void set_shortened_ref(struct strbuf *buf, const char *ref)
1078 char *s = shorten_unambiguous_ref(ref, 0);
1079 strbuf_reset(buf);
1080 strbuf_addstr(buf, s);
1081 free(s);
1084 static int interpret_branch_mark(const char *name, int namelen,
1085 int at, struct strbuf *buf,
1086 int (*get_mark)(const char *, int),
1087 const char *(*get_data)(struct branch *,
1088 struct strbuf *))
1090 int len;
1091 struct branch *branch;
1092 struct strbuf err = STRBUF_INIT;
1093 const char *value;
1095 len = get_mark(name + at, namelen - at);
1096 if (!len)
1097 return -1;
1099 if (memchr(name, ':', at))
1100 return -1;
1102 if (at) {
1103 char *name_str = xmemdupz(name, at);
1104 branch = branch_get(name_str);
1105 free(name_str);
1106 } else
1107 branch = branch_get(NULL);
1109 value = get_data(branch, &err);
1110 if (!value)
1111 die("%s", err.buf);
1113 set_shortened_ref(buf, value);
1114 return len + at;
1118 * This reads short-hand syntax that not only evaluates to a commit
1119 * object name, but also can act as if the end user spelled the name
1120 * of the branch from the command line.
1122 * - "@{-N}" finds the name of the Nth previous branch we were on, and
1123 * places the name of the branch in the given buf and returns the
1124 * number of characters parsed if successful.
1126 * - "<branch>@{upstream}" finds the name of the other ref that
1127 * <branch> is configured to merge with (missing <branch> defaults
1128 * to the current branch), and places the name of the branch in the
1129 * given buf and returns the number of characters parsed if
1130 * successful.
1132 * If the input is not of the accepted format, it returns a negative
1133 * number to signal an error.
1135 * If the input was ok but there are not N branch switches in the
1136 * reflog, it returns 0.
1138 int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
1140 char *at;
1141 const char *start;
1142 int len = interpret_nth_prior_checkout(name, namelen, buf);
1144 if (!namelen)
1145 namelen = strlen(name);
1147 if (!len) {
1148 return len; /* syntax Ok, not enough switches */
1149 } else if (len > 0) {
1150 if (len == namelen)
1151 return len; /* consumed all */
1152 else
1153 return reinterpret(name, namelen, len, buf);
1156 for (start = name;
1157 (at = memchr(start, '@', namelen - (start - name)));
1158 start = at + 1) {
1160 len = interpret_empty_at(name, namelen, at - name, buf);
1161 if (len > 0)
1162 return reinterpret(name, namelen, len, buf);
1164 len = interpret_branch_mark(name, namelen, at - name, buf,
1165 upstream_mark, branch_get_upstream);
1166 if (len > 0)
1167 return len;
1169 len = interpret_branch_mark(name, namelen, at - name, buf,
1170 push_mark, branch_get_push);
1171 if (len > 0)
1172 return len;
1175 return -1;
1178 int strbuf_branchname(struct strbuf *sb, const char *name)
1180 int len = strlen(name);
1181 int used = interpret_branch_name(name, len, sb);
1183 if (used == len)
1184 return 0;
1185 if (used < 0)
1186 used = 0;
1187 strbuf_add(sb, name + used, len - used);
1188 return len;
1191 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1193 strbuf_branchname(sb, name);
1194 if (name[0] == '-')
1195 return -1;
1196 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1197 return check_refname_format(sb->buf, 0);
1201 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1202 * notably "xyz^" for "parent of xyz"
1204 int get_sha1(const char *name, unsigned char *sha1)
1206 struct object_context unused;
1207 return get_sha1_with_context(name, 0, sha1, &unused);
1211 * This is like "get_sha1()", but for struct object_id.
1213 int get_oid(const char *name, struct object_id *oid)
1215 return get_sha1(name, oid->hash);
1220 * Many callers know that the user meant to name a commit-ish by
1221 * syntactical positions where the object name appears. Calling this
1222 * function allows the machinery to disambiguate shorter-than-unique
1223 * abbreviated object names between commit-ish and others.
1225 * Note that this does NOT error out when the named object is not a
1226 * commit-ish. It is merely to give a hint to the disambiguation
1227 * machinery.
1229 int get_sha1_committish(const char *name, unsigned char *sha1)
1231 struct object_context unused;
1232 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1233 sha1, &unused);
1236 int get_sha1_treeish(const char *name, unsigned char *sha1)
1238 struct object_context unused;
1239 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1240 sha1, &unused);
1243 int get_sha1_commit(const char *name, unsigned char *sha1)
1245 struct object_context unused;
1246 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1247 sha1, &unused);
1250 int get_sha1_tree(const char *name, unsigned char *sha1)
1252 struct object_context unused;
1253 return get_sha1_with_context(name, GET_SHA1_TREE,
1254 sha1, &unused);
1257 int get_sha1_blob(const char *name, unsigned char *sha1)
1259 struct object_context unused;
1260 return get_sha1_with_context(name, GET_SHA1_BLOB,
1261 sha1, &unused);
1264 /* Must be called only when object_name:filename doesn't exist. */
1265 static void diagnose_invalid_sha1_path(const char *prefix,
1266 const char *filename,
1267 const unsigned char *tree_sha1,
1268 const char *object_name,
1269 int object_name_len)
1271 unsigned char sha1[20];
1272 unsigned mode;
1274 if (!prefix)
1275 prefix = "";
1277 if (file_exists(filename))
1278 die("Path '%s' exists on disk, but not in '%.*s'.",
1279 filename, object_name_len, object_name);
1280 if (errno == ENOENT || errno == ENOTDIR) {
1281 char *fullname = xstrfmt("%s%s", prefix, filename);
1283 if (!get_tree_entry(tree_sha1, fullname,
1284 sha1, &mode)) {
1285 die("Path '%s' exists, but not '%s'.\n"
1286 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1287 fullname,
1288 filename,
1289 object_name_len, object_name,
1290 fullname,
1291 object_name_len, object_name,
1292 filename);
1294 die("Path '%s' does not exist in '%.*s'",
1295 filename, object_name_len, object_name);
1299 /* Must be called only when :stage:filename doesn't exist. */
1300 static void diagnose_invalid_index_path(int stage,
1301 const char *prefix,
1302 const char *filename)
1304 const struct cache_entry *ce;
1305 int pos;
1306 unsigned namelen = strlen(filename);
1307 struct strbuf fullname = STRBUF_INIT;
1309 if (!prefix)
1310 prefix = "";
1312 /* Wrong stage number? */
1313 pos = cache_name_pos(filename, namelen);
1314 if (pos < 0)
1315 pos = -pos - 1;
1316 if (pos < active_nr) {
1317 ce = active_cache[pos];
1318 if (ce_namelen(ce) == namelen &&
1319 !memcmp(ce->name, filename, namelen))
1320 die("Path '%s' is in the index, but not at stage %d.\n"
1321 "Did you mean ':%d:%s'?",
1322 filename, stage,
1323 ce_stage(ce), filename);
1326 /* Confusion between relative and absolute filenames? */
1327 strbuf_addstr(&fullname, prefix);
1328 strbuf_addstr(&fullname, filename);
1329 pos = cache_name_pos(fullname.buf, fullname.len);
1330 if (pos < 0)
1331 pos = -pos - 1;
1332 if (pos < active_nr) {
1333 ce = active_cache[pos];
1334 if (ce_namelen(ce) == fullname.len &&
1335 !memcmp(ce->name, fullname.buf, fullname.len))
1336 die("Path '%s' is in the index, but not '%s'.\n"
1337 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1338 fullname.buf, filename,
1339 ce_stage(ce), fullname.buf,
1340 ce_stage(ce), filename);
1343 if (file_exists(filename))
1344 die("Path '%s' exists on disk, but not in the index.", filename);
1345 if (errno == ENOENT || errno == ENOTDIR)
1346 die("Path '%s' does not exist (neither on disk nor in the index).",
1347 filename);
1349 strbuf_release(&fullname);
1353 static char *resolve_relative_path(const char *rel)
1355 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1356 return NULL;
1358 if (!is_inside_work_tree())
1359 die("relative path syntax can't be used outside working tree.");
1361 /* die() inside prefix_path() if resolved path is outside worktree */
1362 return prefix_path(startup_info->prefix,
1363 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1364 rel);
1367 static int get_sha1_with_context_1(const char *name,
1368 unsigned flags,
1369 const char *prefix,
1370 unsigned char *sha1,
1371 struct object_context *oc)
1373 int ret, bracket_depth;
1374 int namelen = strlen(name);
1375 const char *cp;
1376 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1378 memset(oc, 0, sizeof(*oc));
1379 oc->mode = S_IFINVALID;
1380 ret = get_sha1_1(name, namelen, sha1, flags);
1381 if (!ret)
1382 return ret;
1384 * sha1:path --> object name of path in ent sha1
1385 * :path -> object name of absolute path in index
1386 * :./path -> object name of path relative to cwd in index
1387 * :[0-3]:path -> object name of path in index at stage
1388 * :/foo -> recent commit matching foo
1390 if (name[0] == ':') {
1391 int stage = 0;
1392 const struct cache_entry *ce;
1393 char *new_path = NULL;
1394 int pos;
1395 if (!only_to_die && namelen > 2 && name[1] == '/') {
1396 struct commit_list *list = NULL;
1398 for_each_ref(handle_one_ref, &list);
1399 commit_list_sort_by_date(&list);
1400 return get_sha1_oneline(name + 2, sha1, list);
1402 if (namelen < 3 ||
1403 name[2] != ':' ||
1404 name[1] < '0' || '3' < name[1])
1405 cp = name + 1;
1406 else {
1407 stage = name[1] - '0';
1408 cp = name + 3;
1410 new_path = resolve_relative_path(cp);
1411 if (!new_path) {
1412 namelen = namelen - (cp - name);
1413 } else {
1414 cp = new_path;
1415 namelen = strlen(cp);
1418 strlcpy(oc->path, cp, sizeof(oc->path));
1420 if (!active_cache)
1421 read_cache();
1422 pos = cache_name_pos(cp, namelen);
1423 if (pos < 0)
1424 pos = -pos - 1;
1425 while (pos < active_nr) {
1426 ce = active_cache[pos];
1427 if (ce_namelen(ce) != namelen ||
1428 memcmp(ce->name, cp, namelen))
1429 break;
1430 if (ce_stage(ce) == stage) {
1431 hashcpy(sha1, ce->oid.hash);
1432 oc->mode = ce->ce_mode;
1433 free(new_path);
1434 return 0;
1436 pos++;
1438 if (only_to_die && name[1] && name[1] != '/')
1439 diagnose_invalid_index_path(stage, prefix, cp);
1440 free(new_path);
1441 return -1;
1443 for (cp = name, bracket_depth = 0; *cp; cp++) {
1444 if (*cp == '{')
1445 bracket_depth++;
1446 else if (bracket_depth && *cp == '}')
1447 bracket_depth--;
1448 else if (!bracket_depth && *cp == ':')
1449 break;
1451 if (*cp == ':') {
1452 unsigned char tree_sha1[20];
1453 int len = cp - name;
1454 if (!get_sha1_1(name, len, tree_sha1, GET_SHA1_TREEISH)) {
1455 const char *filename = cp+1;
1456 char *new_filename = NULL;
1458 new_filename = resolve_relative_path(filename);
1459 if (new_filename)
1460 filename = new_filename;
1461 if (flags & GET_SHA1_FOLLOW_SYMLINKS) {
1462 ret = get_tree_entry_follow_symlinks(tree_sha1,
1463 filename, sha1, &oc->symlink_path,
1464 &oc->mode);
1465 } else {
1466 ret = get_tree_entry(tree_sha1, filename,
1467 sha1, &oc->mode);
1468 if (ret && only_to_die) {
1469 diagnose_invalid_sha1_path(prefix,
1470 filename,
1471 tree_sha1,
1472 name, len);
1475 hashcpy(oc->tree, tree_sha1);
1476 strlcpy(oc->path, filename, sizeof(oc->path));
1478 free(new_filename);
1479 return ret;
1480 } else {
1481 if (only_to_die)
1482 die("Invalid object name '%.*s'.", len, name);
1485 return ret;
1489 * Call this function when you know "name" given by the end user must
1490 * name an object but it doesn't; the function _may_ die with a better
1491 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1492 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1493 * you have a chance to diagnose the error further.
1495 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1497 struct object_context oc;
1498 unsigned char sha1[20];
1499 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1502 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1504 if (flags & GET_SHA1_FOLLOW_SYMLINKS && flags & GET_SHA1_ONLY_TO_DIE)
1505 die("BUG: incompatible flags for get_sha1_with_context");
1506 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);