sha1_name: refactor upstream_mark
[alt-git.git] / sha1_name.c
blob1005f45d75927a05a7f4dfabea6ccba82ef367a4
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 memcpy(hex, sha1_to_hex(sha1), 40);
376 if (len == 40 || !len)
377 return hex;
378 exists = has_sha1_file(sha1);
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 at_mark(const char *string, int len,
419 const char **suffix, int nr)
421 int i;
423 for (i = 0; i < nr; 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 inline int upstream_mark(const char *string, int len)
434 const char *suffix[] = { "@{upstream}", "@{u}" };
435 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
438 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
439 static int interpret_nth_prior_checkout(const char *name, int namelen, struct strbuf *buf);
441 static int get_sha1_basic(const char *str, int len, unsigned char *sha1,
442 unsigned int flags)
444 static const char *warn_msg = "refname '%.*s' is ambiguous.";
445 static const char *object_name_msg = N_(
446 "Git normally never creates a ref that ends with 40 hex characters\n"
447 "because it will be ignored when you just specify 40-hex. These refs\n"
448 "may be created by mistake. For example,\n"
449 "\n"
450 " git checkout -b $br $(git rev-parse ...)\n"
451 "\n"
452 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
453 "examine these refs and maybe delete them. Turn this message off by\n"
454 "running \"git config advice.objectNameWarning false\"");
455 unsigned char tmp_sha1[20];
456 char *real_ref = NULL;
457 int refs_found = 0;
458 int at, reflog_len, nth_prior = 0;
460 if (len == 40 && !get_sha1_hex(str, sha1)) {
461 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
462 refs_found = dwim_ref(str, len, tmp_sha1, &real_ref);
463 if (refs_found > 0) {
464 warning(warn_msg, len, str);
465 if (advice_object_name_warning)
466 fprintf(stderr, "%s\n", _(object_name_msg));
468 free(real_ref);
470 return 0;
473 /* basic@{time or number or -number} format to query ref-log */
474 reflog_len = at = 0;
475 if (len && str[len-1] == '}') {
476 for (at = len-4; at >= 0; at--) {
477 if (str[at] == '@' && str[at+1] == '{') {
478 if (str[at+2] == '-') {
479 if (at != 0)
480 /* @{-N} not at start */
481 return -1;
482 nth_prior = 1;
483 continue;
485 if (!upstream_mark(str + at, len - at)) {
486 reflog_len = (len-1) - (at+2);
487 len = at;
489 break;
494 /* Accept only unambiguous ref paths. */
495 if (len && ambiguous_path(str, len))
496 return -1;
498 if (nth_prior) {
499 struct strbuf buf = STRBUF_INIT;
500 int detached;
502 if (interpret_nth_prior_checkout(str, len, &buf) > 0) {
503 detached = (buf.len == 40 && !get_sha1_hex(buf.buf, sha1));
504 strbuf_release(&buf);
505 if (detached)
506 return 0;
510 if (!len && reflog_len)
511 /* allow "@{...}" to mean the current branch reflog */
512 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
513 else if (reflog_len)
514 refs_found = dwim_log(str, len, sha1, &real_ref);
515 else
516 refs_found = dwim_ref(str, len, sha1, &real_ref);
518 if (!refs_found)
519 return -1;
521 if (warn_ambiguous_refs && !(flags & GET_SHA1_QUIETLY) &&
522 (refs_found > 1 ||
523 !get_short_sha1(str, len, tmp_sha1, GET_SHA1_QUIETLY)))
524 warning(warn_msg, len, str);
526 if (reflog_len) {
527 int nth, i;
528 unsigned long at_time;
529 unsigned long co_time;
530 int co_tz, co_cnt;
532 /* Is it asking for N-th entry, or approxidate? */
533 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
534 char ch = str[at+2+i];
535 if ('0' <= ch && ch <= '9')
536 nth = nth * 10 + ch - '0';
537 else
538 nth = -1;
540 if (100000000 <= nth) {
541 at_time = nth;
542 nth = -1;
543 } else if (0 <= nth)
544 at_time = 0;
545 else {
546 int errors = 0;
547 char *tmp = xstrndup(str + at + 2, reflog_len);
548 at_time = approxidate_careful(tmp, &errors);
549 free(tmp);
550 if (errors) {
551 free(real_ref);
552 return -1;
555 if (read_ref_at(real_ref, flags, at_time, nth, sha1, NULL,
556 &co_time, &co_tz, &co_cnt)) {
557 if (!len) {
558 if (starts_with(real_ref, "refs/heads/")) {
559 str = real_ref + 11;
560 len = strlen(real_ref + 11);
561 } else {
562 /* detached HEAD */
563 str = "HEAD";
564 len = 4;
567 if (at_time) {
568 if (!(flags & GET_SHA1_QUIETLY)) {
569 warning("Log for '%.*s' only goes "
570 "back to %s.", len, str,
571 show_date(co_time, co_tz, DATE_RFC2822));
573 } else {
574 if (flags & GET_SHA1_QUIETLY) {
575 exit(128);
577 die("Log for '%.*s' only has %d entries.",
578 len, str, co_cnt);
583 free(real_ref);
584 return 0;
587 static int get_parent(const char *name, int len,
588 unsigned char *result, int idx)
590 unsigned char sha1[20];
591 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
592 struct commit *commit;
593 struct commit_list *p;
595 if (ret)
596 return ret;
597 commit = lookup_commit_reference(sha1);
598 if (parse_commit(commit))
599 return -1;
600 if (!idx) {
601 hashcpy(result, commit->object.sha1);
602 return 0;
604 p = commit->parents;
605 while (p) {
606 if (!--idx) {
607 hashcpy(result, p->item->object.sha1);
608 return 0;
610 p = p->next;
612 return -1;
615 static int get_nth_ancestor(const char *name, int len,
616 unsigned char *result, int generation)
618 unsigned char sha1[20];
619 struct commit *commit;
620 int ret;
622 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
623 if (ret)
624 return ret;
625 commit = lookup_commit_reference(sha1);
626 if (!commit)
627 return -1;
629 while (generation--) {
630 if (parse_commit(commit) || !commit->parents)
631 return -1;
632 commit = commit->parents->item;
634 hashcpy(result, commit->object.sha1);
635 return 0;
638 struct object *peel_to_type(const char *name, int namelen,
639 struct object *o, enum object_type expected_type)
641 if (name && !namelen)
642 namelen = strlen(name);
643 while (1) {
644 if (!o || (!o->parsed && !parse_object(o->sha1)))
645 return NULL;
646 if (expected_type == OBJ_ANY || o->type == expected_type)
647 return o;
648 if (o->type == OBJ_TAG)
649 o = ((struct tag*) o)->tagged;
650 else if (o->type == OBJ_COMMIT)
651 o = &(((struct commit *) o)->tree->object);
652 else {
653 if (name)
654 error("%.*s: expected %s type, but the object "
655 "dereferences to %s type",
656 namelen, name, typename(expected_type),
657 typename(o->type));
658 return NULL;
663 static int peel_onion(const char *name, int len, unsigned char *sha1)
665 unsigned char outer[20];
666 const char *sp;
667 unsigned int expected_type = 0;
668 unsigned lookup_flags = 0;
669 struct object *o;
672 * "ref^{type}" dereferences ref repeatedly until you cannot
673 * dereference anymore, or you get an object of given type,
674 * whichever comes first. "ref^{}" means just dereference
675 * tags until you get a non-tag. "ref^0" is a shorthand for
676 * "ref^{commit}". "commit^{tree}" could be used to find the
677 * top-level tree of the given commit.
679 if (len < 4 || name[len-1] != '}')
680 return -1;
682 for (sp = name + len - 1; name <= sp; sp--) {
683 int ch = *sp;
684 if (ch == '{' && name < sp && sp[-1] == '^')
685 break;
687 if (sp <= name)
688 return -1;
690 sp++; /* beginning of type name, or closing brace for empty */
691 if (starts_with(sp, "commit}"))
692 expected_type = OBJ_COMMIT;
693 else if (starts_with(sp, "tag}"))
694 expected_type = OBJ_TAG;
695 else if (starts_with(sp, "tree}"))
696 expected_type = OBJ_TREE;
697 else if (starts_with(sp, "blob}"))
698 expected_type = OBJ_BLOB;
699 else if (starts_with(sp, "object}"))
700 expected_type = OBJ_ANY;
701 else if (sp[0] == '}')
702 expected_type = OBJ_NONE;
703 else if (sp[0] == '/')
704 expected_type = OBJ_COMMIT;
705 else
706 return -1;
708 if (expected_type == OBJ_COMMIT)
709 lookup_flags = GET_SHA1_COMMITTISH;
710 else if (expected_type == OBJ_TREE)
711 lookup_flags = GET_SHA1_TREEISH;
713 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
714 return -1;
716 o = parse_object(outer);
717 if (!o)
718 return -1;
719 if (!expected_type) {
720 o = deref_tag(o, name, sp - name - 2);
721 if (!o || (!o->parsed && !parse_object(o->sha1)))
722 return -1;
723 hashcpy(sha1, o->sha1);
724 return 0;
728 * At this point, the syntax look correct, so
729 * if we do not get the needed object, we should
730 * barf.
732 o = peel_to_type(name, len, o, expected_type);
733 if (!o)
734 return -1;
736 hashcpy(sha1, o->sha1);
737 if (sp[0] == '/') {
738 /* "$commit^{/foo}" */
739 char *prefix;
740 int ret;
741 struct commit_list *list = NULL;
744 * $commit^{/}. Some regex implementation may reject.
745 * We don't need regex anyway. '' pattern always matches.
747 if (sp[1] == '}')
748 return 0;
750 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
751 commit_list_insert((struct commit *)o, &list);
752 ret = get_sha1_oneline(prefix, sha1, list);
753 free(prefix);
754 return ret;
756 return 0;
759 static int get_describe_name(const char *name, int len, unsigned char *sha1)
761 const char *cp;
762 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
764 for (cp = name + len - 1; name + 2 <= cp; cp--) {
765 char ch = *cp;
766 if (!isxdigit(ch)) {
767 /* We must be looking at g in "SOMETHING-g"
768 * for it to be describe output.
770 if (ch == 'g' && cp[-1] == '-') {
771 cp++;
772 len -= cp - name;
773 return get_short_sha1(cp, len, sha1, flags);
777 return -1;
780 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
782 int ret, has_suffix;
783 const char *cp;
786 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
788 has_suffix = 0;
789 for (cp = name + len - 1; name <= cp; cp--) {
790 int ch = *cp;
791 if ('0' <= ch && ch <= '9')
792 continue;
793 if (ch == '~' || ch == '^')
794 has_suffix = ch;
795 break;
798 if (has_suffix) {
799 int num = 0;
800 int len1 = cp - name;
801 cp++;
802 while (cp < name + len)
803 num = num * 10 + *cp++ - '0';
804 if (!num && len1 == len - 1)
805 num = 1;
806 if (has_suffix == '^')
807 return get_parent(name, len1, sha1, num);
808 /* else if (has_suffix == '~') -- goes without saying */
809 return get_nth_ancestor(name, len1, sha1, num);
812 ret = peel_onion(name, len, sha1);
813 if (!ret)
814 return 0;
816 ret = get_sha1_basic(name, len, sha1, lookup_flags);
817 if (!ret)
818 return 0;
820 /* It could be describe output that is "SOMETHING-gXXXX" */
821 ret = get_describe_name(name, len, sha1);
822 if (!ret)
823 return 0;
825 return get_short_sha1(name, len, sha1, lookup_flags);
829 * This interprets names like ':/Initial revision of "git"' by searching
830 * through history and returning the first commit whose message starts
831 * the given regular expression.
833 * For future extension, ':/!' is reserved. If you want to match a message
834 * beginning with a '!', you have to repeat the exclamation mark.
837 /* Remember to update object flag allocation in object.h */
838 #define ONELINE_SEEN (1u<<20)
840 static int handle_one_ref(const char *path,
841 const unsigned char *sha1, int flag, void *cb_data)
843 struct commit_list **list = cb_data;
844 struct object *object = parse_object(sha1);
845 if (!object)
846 return 0;
847 if (object->type == OBJ_TAG) {
848 object = deref_tag(object, path, strlen(path));
849 if (!object)
850 return 0;
852 if (object->type != OBJ_COMMIT)
853 return 0;
854 commit_list_insert((struct commit *)object, list);
855 return 0;
858 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
859 struct commit_list *list)
861 struct commit_list *backup = NULL, *l;
862 int found = 0;
863 regex_t regex;
865 if (prefix[0] == '!') {
866 if (prefix[1] != '!')
867 die ("Invalid search pattern: %s", prefix);
868 prefix++;
871 if (regcomp(&regex, prefix, REG_EXTENDED))
872 die("Invalid search pattern: %s", prefix);
874 for (l = list; l; l = l->next) {
875 l->item->object.flags |= ONELINE_SEEN;
876 commit_list_insert(l->item, &backup);
878 while (list) {
879 const char *p, *buf;
880 struct commit *commit;
881 int matches;
883 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
884 if (!parse_object(commit->object.sha1))
885 continue;
886 buf = get_commit_buffer(commit, NULL);
887 p = strstr(buf, "\n\n");
888 matches = p && !regexec(&regex, p + 2, 0, NULL, 0);
889 unuse_commit_buffer(commit, buf);
891 if (matches) {
892 hashcpy(sha1, commit->object.sha1);
893 found = 1;
894 break;
897 regfree(&regex);
898 free_commit_list(list);
899 for (l = backup; l; l = l->next)
900 clear_commit_marks(l->item, ONELINE_SEEN);
901 free_commit_list(backup);
902 return found ? 0 : -1;
905 struct grab_nth_branch_switch_cbdata {
906 int remaining;
907 struct strbuf buf;
910 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
911 const char *email, unsigned long timestamp, int tz,
912 const char *message, void *cb_data)
914 struct grab_nth_branch_switch_cbdata *cb = cb_data;
915 const char *match = NULL, *target = NULL;
916 size_t len;
918 if (skip_prefix(message, "checkout: moving from ", &match))
919 target = strstr(match, " to ");
921 if (!match || !target)
922 return 0;
923 if (--(cb->remaining) == 0) {
924 len = target - match;
925 strbuf_reset(&cb->buf);
926 strbuf_add(&cb->buf, match, len);
927 return 1; /* we are done */
929 return 0;
933 * Parse @{-N} syntax, return the number of characters parsed
934 * if successful; otherwise signal an error with negative value.
936 static int interpret_nth_prior_checkout(const char *name, int namelen,
937 struct strbuf *buf)
939 long nth;
940 int retval;
941 struct grab_nth_branch_switch_cbdata cb;
942 const char *brace;
943 char *num_end;
945 if (namelen < 4)
946 return -1;
947 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
948 return -1;
949 brace = memchr(name, '}', namelen);
950 if (!brace)
951 return -1;
952 nth = strtol(name + 3, &num_end, 10);
953 if (num_end != brace)
954 return -1;
955 if (nth <= 0)
956 return -1;
957 cb.remaining = nth;
958 strbuf_init(&cb.buf, 20);
960 retval = 0;
961 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
962 strbuf_reset(buf);
963 strbuf_addbuf(buf, &cb.buf);
964 retval = brace - name + 1;
967 strbuf_release(&cb.buf);
968 return retval;
971 int get_sha1_mb(const char *name, unsigned char *sha1)
973 struct commit *one, *two;
974 struct commit_list *mbs;
975 unsigned char sha1_tmp[20];
976 const char *dots;
977 int st;
979 dots = strstr(name, "...");
980 if (!dots)
981 return get_sha1(name, sha1);
982 if (dots == name)
983 st = get_sha1("HEAD", sha1_tmp);
984 else {
985 struct strbuf sb;
986 strbuf_init(&sb, dots - name);
987 strbuf_add(&sb, name, dots - name);
988 st = get_sha1_committish(sb.buf, sha1_tmp);
989 strbuf_release(&sb);
991 if (st)
992 return st;
993 one = lookup_commit_reference_gently(sha1_tmp, 0);
994 if (!one)
995 return -1;
997 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", sha1_tmp))
998 return -1;
999 two = lookup_commit_reference_gently(sha1_tmp, 0);
1000 if (!two)
1001 return -1;
1002 mbs = get_merge_bases(one, two);
1003 if (!mbs || mbs->next)
1004 st = -1;
1005 else {
1006 st = 0;
1007 hashcpy(sha1, mbs->item->object.sha1);
1009 free_commit_list(mbs);
1010 return st;
1013 /* parse @something syntax, when 'something' is not {.*} */
1014 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1016 const char *next;
1018 if (len || name[1] == '{')
1019 return -1;
1021 /* make sure it's a single @, or @@{.*}, not @foo */
1022 next = memchr(name + len + 1, '@', namelen - len - 1);
1023 if (next && next[1] != '{')
1024 return -1;
1025 if (!next)
1026 next = name + namelen;
1027 if (next != name + 1)
1028 return -1;
1030 strbuf_reset(buf);
1031 strbuf_add(buf, "HEAD", 4);
1032 return 1;
1035 static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
1037 /* we have extra data, which might need further processing */
1038 struct strbuf tmp = STRBUF_INIT;
1039 int used = buf->len;
1040 int ret;
1042 strbuf_add(buf, name + len, namelen - len);
1043 ret = interpret_branch_name(buf->buf, buf->len, &tmp);
1044 /* that data was not interpreted, remove our cruft */
1045 if (ret < 0) {
1046 strbuf_setlen(buf, used);
1047 return len;
1049 strbuf_reset(buf);
1050 strbuf_addbuf(buf, &tmp);
1051 strbuf_release(&tmp);
1052 /* tweak for size of {-N} versus expanded ref name */
1053 return ret - used + len;
1056 static void set_shortened_ref(struct strbuf *buf, const char *ref)
1058 char *s = shorten_unambiguous_ref(ref, 0);
1059 strbuf_reset(buf);
1060 strbuf_addstr(buf, s);
1061 free(s);
1064 static const char *get_upstream_branch(const char *branch_buf, int len)
1066 char *branch = xstrndup(branch_buf, len);
1067 struct branch *upstream = branch_get(*branch ? branch : NULL);
1068 struct strbuf err = STRBUF_INIT;
1069 const char *ret;
1071 free(branch);
1073 ret = branch_get_upstream(upstream, &err);
1074 if (!ret)
1075 die("%s", err.buf);
1077 return ret;
1080 static int interpret_upstream_mark(const char *name, int namelen,
1081 int at, struct strbuf *buf)
1083 int len;
1085 len = upstream_mark(name + at, namelen - at);
1086 if (!len)
1087 return -1;
1089 if (memchr(name, ':', at))
1090 return -1;
1092 set_shortened_ref(buf, get_upstream_branch(name, at));
1093 return len + at;
1097 * This reads short-hand syntax that not only evaluates to a commit
1098 * object name, but also can act as if the end user spelled the name
1099 * of the branch from the command line.
1101 * - "@{-N}" finds the name of the Nth previous branch we were on, and
1102 * places the name of the branch in the given buf and returns the
1103 * number of characters parsed if successful.
1105 * - "<branch>@{upstream}" finds the name of the other ref that
1106 * <branch> is configured to merge with (missing <branch> defaults
1107 * to the current branch), and places the name of the branch in the
1108 * given buf and returns the number of characters parsed if
1109 * successful.
1111 * If the input is not of the accepted format, it returns a negative
1112 * number to signal an error.
1114 * If the input was ok but there are not N branch switches in the
1115 * reflog, it returns 0.
1117 int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
1119 char *at;
1120 const char *start;
1121 int len = interpret_nth_prior_checkout(name, namelen, buf);
1123 if (!namelen)
1124 namelen = strlen(name);
1126 if (!len) {
1127 return len; /* syntax Ok, not enough switches */
1128 } else if (len > 0) {
1129 if (len == namelen)
1130 return len; /* consumed all */
1131 else
1132 return reinterpret(name, namelen, len, buf);
1135 for (start = name;
1136 (at = memchr(start, '@', namelen - (start - name)));
1137 start = at + 1) {
1139 len = interpret_empty_at(name, namelen, at - name, buf);
1140 if (len > 0)
1141 return reinterpret(name, namelen, len, buf);
1143 len = interpret_upstream_mark(name, namelen, at - name, buf);
1144 if (len > 0)
1145 return len;
1148 return -1;
1151 int strbuf_branchname(struct strbuf *sb, const char *name)
1153 int len = strlen(name);
1154 int used = interpret_branch_name(name, len, sb);
1156 if (used == len)
1157 return 0;
1158 if (used < 0)
1159 used = 0;
1160 strbuf_add(sb, name + used, len - used);
1161 return len;
1164 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1166 strbuf_branchname(sb, name);
1167 if (name[0] == '-')
1168 return -1;
1169 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1170 return check_refname_format(sb->buf, 0);
1174 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1175 * notably "xyz^" for "parent of xyz"
1177 int get_sha1(const char *name, unsigned char *sha1)
1179 struct object_context unused;
1180 return get_sha1_with_context(name, 0, sha1, &unused);
1184 * Many callers know that the user meant to name a commit-ish by
1185 * syntactical positions where the object name appears. Calling this
1186 * function allows the machinery to disambiguate shorter-than-unique
1187 * abbreviated object names between commit-ish and others.
1189 * Note that this does NOT error out when the named object is not a
1190 * commit-ish. It is merely to give a hint to the disambiguation
1191 * machinery.
1193 int get_sha1_committish(const char *name, unsigned char *sha1)
1195 struct object_context unused;
1196 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1197 sha1, &unused);
1200 int get_sha1_treeish(const char *name, unsigned char *sha1)
1202 struct object_context unused;
1203 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1204 sha1, &unused);
1207 int get_sha1_commit(const char *name, unsigned char *sha1)
1209 struct object_context unused;
1210 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1211 sha1, &unused);
1214 int get_sha1_tree(const char *name, unsigned char *sha1)
1216 struct object_context unused;
1217 return get_sha1_with_context(name, GET_SHA1_TREE,
1218 sha1, &unused);
1221 int get_sha1_blob(const char *name, unsigned char *sha1)
1223 struct object_context unused;
1224 return get_sha1_with_context(name, GET_SHA1_BLOB,
1225 sha1, &unused);
1228 /* Must be called only when object_name:filename doesn't exist. */
1229 static void diagnose_invalid_sha1_path(const char *prefix,
1230 const char *filename,
1231 const unsigned char *tree_sha1,
1232 const char *object_name,
1233 int object_name_len)
1235 struct stat st;
1236 unsigned char sha1[20];
1237 unsigned mode;
1239 if (!prefix)
1240 prefix = "";
1242 if (!lstat(filename, &st))
1243 die("Path '%s' exists on disk, but not in '%.*s'.",
1244 filename, object_name_len, object_name);
1245 if (errno == ENOENT || errno == ENOTDIR) {
1246 char *fullname = xstrfmt("%s%s", prefix, filename);
1248 if (!get_tree_entry(tree_sha1, fullname,
1249 sha1, &mode)) {
1250 die("Path '%s' exists, but not '%s'.\n"
1251 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1252 fullname,
1253 filename,
1254 object_name_len, object_name,
1255 fullname,
1256 object_name_len, object_name,
1257 filename);
1259 die("Path '%s' does not exist in '%.*s'",
1260 filename, object_name_len, object_name);
1264 /* Must be called only when :stage:filename doesn't exist. */
1265 static void diagnose_invalid_index_path(int stage,
1266 const char *prefix,
1267 const char *filename)
1269 struct stat st;
1270 const struct cache_entry *ce;
1271 int pos;
1272 unsigned namelen = strlen(filename);
1273 unsigned fullnamelen;
1274 char *fullname;
1276 if (!prefix)
1277 prefix = "";
1279 /* Wrong stage number? */
1280 pos = cache_name_pos(filename, namelen);
1281 if (pos < 0)
1282 pos = -pos - 1;
1283 if (pos < active_nr) {
1284 ce = active_cache[pos];
1285 if (ce_namelen(ce) == namelen &&
1286 !memcmp(ce->name, filename, namelen))
1287 die("Path '%s' is in the index, but not at stage %d.\n"
1288 "Did you mean ':%d:%s'?",
1289 filename, stage,
1290 ce_stage(ce), filename);
1293 /* Confusion between relative and absolute filenames? */
1294 fullnamelen = namelen + strlen(prefix);
1295 fullname = xmalloc(fullnamelen + 1);
1296 strcpy(fullname, prefix);
1297 strcat(fullname, filename);
1298 pos = cache_name_pos(fullname, fullnamelen);
1299 if (pos < 0)
1300 pos = -pos - 1;
1301 if (pos < active_nr) {
1302 ce = active_cache[pos];
1303 if (ce_namelen(ce) == fullnamelen &&
1304 !memcmp(ce->name, fullname, fullnamelen))
1305 die("Path '%s' is in the index, but not '%s'.\n"
1306 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1307 fullname, filename,
1308 ce_stage(ce), fullname,
1309 ce_stage(ce), filename);
1312 if (!lstat(filename, &st))
1313 die("Path '%s' exists on disk, but not in the index.", filename);
1314 if (errno == ENOENT || errno == ENOTDIR)
1315 die("Path '%s' does not exist (neither on disk nor in the index).",
1316 filename);
1318 free(fullname);
1322 static char *resolve_relative_path(const char *rel)
1324 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1325 return NULL;
1327 if (!startup_info)
1328 die("BUG: startup_info struct is not initialized.");
1330 if (!is_inside_work_tree())
1331 die("relative path syntax can't be used outside working tree.");
1333 /* die() inside prefix_path() if resolved path is outside worktree */
1334 return prefix_path(startup_info->prefix,
1335 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1336 rel);
1339 static int get_sha1_with_context_1(const char *name,
1340 unsigned flags,
1341 const char *prefix,
1342 unsigned char *sha1,
1343 struct object_context *oc)
1345 int ret, bracket_depth;
1346 int namelen = strlen(name);
1347 const char *cp;
1348 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1350 memset(oc, 0, sizeof(*oc));
1351 oc->mode = S_IFINVALID;
1352 ret = get_sha1_1(name, namelen, sha1, flags);
1353 if (!ret)
1354 return ret;
1356 * sha1:path --> object name of path in ent sha1
1357 * :path -> object name of absolute path in index
1358 * :./path -> object name of path relative to cwd in index
1359 * :[0-3]:path -> object name of path in index at stage
1360 * :/foo -> recent commit matching foo
1362 if (name[0] == ':') {
1363 int stage = 0;
1364 const struct cache_entry *ce;
1365 char *new_path = NULL;
1366 int pos;
1367 if (!only_to_die && namelen > 2 && name[1] == '/') {
1368 struct commit_list *list = NULL;
1369 for_each_ref(handle_one_ref, &list);
1370 commit_list_sort_by_date(&list);
1371 return get_sha1_oneline(name + 2, sha1, list);
1373 if (namelen < 3 ||
1374 name[2] != ':' ||
1375 name[1] < '0' || '3' < name[1])
1376 cp = name + 1;
1377 else {
1378 stage = name[1] - '0';
1379 cp = name + 3;
1381 new_path = resolve_relative_path(cp);
1382 if (!new_path) {
1383 namelen = namelen - (cp - name);
1384 } else {
1385 cp = new_path;
1386 namelen = strlen(cp);
1389 strlcpy(oc->path, cp, sizeof(oc->path));
1391 if (!active_cache)
1392 read_cache();
1393 pos = cache_name_pos(cp, namelen);
1394 if (pos < 0)
1395 pos = -pos - 1;
1396 while (pos < active_nr) {
1397 ce = active_cache[pos];
1398 if (ce_namelen(ce) != namelen ||
1399 memcmp(ce->name, cp, namelen))
1400 break;
1401 if (ce_stage(ce) == stage) {
1402 hashcpy(sha1, ce->sha1);
1403 oc->mode = ce->ce_mode;
1404 free(new_path);
1405 return 0;
1407 pos++;
1409 if (only_to_die && name[1] && name[1] != '/')
1410 diagnose_invalid_index_path(stage, prefix, cp);
1411 free(new_path);
1412 return -1;
1414 for (cp = name, bracket_depth = 0; *cp; cp++) {
1415 if (*cp == '{')
1416 bracket_depth++;
1417 else if (bracket_depth && *cp == '}')
1418 bracket_depth--;
1419 else if (!bracket_depth && *cp == ':')
1420 break;
1422 if (*cp == ':') {
1423 unsigned char tree_sha1[20];
1424 int len = cp - name;
1425 if (!get_sha1_1(name, len, tree_sha1, GET_SHA1_TREEISH)) {
1426 const char *filename = cp+1;
1427 char *new_filename = NULL;
1429 new_filename = resolve_relative_path(filename);
1430 if (new_filename)
1431 filename = new_filename;
1432 ret = get_tree_entry(tree_sha1, filename, sha1, &oc->mode);
1433 if (ret && only_to_die) {
1434 diagnose_invalid_sha1_path(prefix, filename,
1435 tree_sha1,
1436 name, len);
1438 hashcpy(oc->tree, tree_sha1);
1439 strlcpy(oc->path, filename, sizeof(oc->path));
1441 free(new_filename);
1442 return ret;
1443 } else {
1444 if (only_to_die)
1445 die("Invalid object name '%.*s'.", len, name);
1448 return ret;
1452 * Call this function when you know "name" given by the end user must
1453 * name an object but it doesn't; the function _may_ die with a better
1454 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1455 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1456 * you have a chance to diagnose the error further.
1458 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1460 struct object_context oc;
1461 unsigned char sha1[20];
1462 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1465 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1467 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);