]> git.ipfire.org Git - thirdparty/git.git/blob - object-name.c
Merge branch 'gc/bare-repo-discovery'
[thirdparty/git.git] / object-name.c
1 #include "cache.h"
2 #include "config.h"
3 #include "tag.h"
4 #include "commit.h"
5 #include "tree.h"
6 #include "blob.h"
7 #include "tree-walk.h"
8 #include "refs.h"
9 #include "remote.h"
10 #include "dir.h"
11 #include "oid-array.h"
12 #include "packfile.h"
13 #include "object-store.h"
14 #include "repository.h"
15 #include "submodule.h"
16 #include "midx.h"
17 #include "commit-reach.h"
18 #include "date.h"
19
20 static int get_oid_oneline(struct repository *r, const char *, struct object_id *, struct commit_list *);
21
22 typedef int (*disambiguate_hint_fn)(struct repository *, const struct object_id *, void *);
23
24 struct disambiguate_state {
25 int len; /* length of prefix in hex chars */
26 char hex_pfx[GIT_MAX_HEXSZ + 1];
27 struct object_id bin_pfx;
28
29 struct repository *repo;
30 disambiguate_hint_fn fn;
31 void *cb_data;
32 struct object_id candidate;
33 unsigned candidate_exists:1;
34 unsigned candidate_checked:1;
35 unsigned candidate_ok:1;
36 unsigned disambiguate_fn_used:1;
37 unsigned ambiguous:1;
38 unsigned always_call_fn:1;
39 };
40
41 static void update_candidates(struct disambiguate_state *ds, const struct object_id *current)
42 {
43 if (ds->always_call_fn) {
44 ds->ambiguous = ds->fn(ds->repo, current, ds->cb_data) ? 1 : 0;
45 return;
46 }
47 if (!ds->candidate_exists) {
48 /* this is the first candidate */
49 oidcpy(&ds->candidate, current);
50 ds->candidate_exists = 1;
51 return;
52 } else if (oideq(&ds->candidate, current)) {
53 /* the same as what we already have seen */
54 return;
55 }
56
57 if (!ds->fn) {
58 /* cannot disambiguate between ds->candidate and current */
59 ds->ambiguous = 1;
60 return;
61 }
62
63 if (!ds->candidate_checked) {
64 ds->candidate_ok = ds->fn(ds->repo, &ds->candidate, ds->cb_data);
65 ds->disambiguate_fn_used = 1;
66 ds->candidate_checked = 1;
67 }
68
69 if (!ds->candidate_ok) {
70 /* discard the candidate; we know it does not satisfy fn */
71 oidcpy(&ds->candidate, current);
72 ds->candidate_checked = 0;
73 return;
74 }
75
76 /* if we reach this point, we know ds->candidate satisfies fn */
77 if (ds->fn(ds->repo, current, ds->cb_data)) {
78 /*
79 * if both current and candidate satisfy fn, we cannot
80 * disambiguate.
81 */
82 ds->candidate_ok = 0;
83 ds->ambiguous = 1;
84 }
85
86 /* otherwise, current can be discarded and candidate is still good */
87 }
88
89 static int match_hash(unsigned, const unsigned char *, const unsigned char *);
90
91 static enum cb_next match_prefix(const struct object_id *oid, void *arg)
92 {
93 struct disambiguate_state *ds = arg;
94 /* no need to call match_hash, oidtree_each did prefix match */
95 update_candidates(ds, oid);
96 return ds->ambiguous ? CB_BREAK : CB_CONTINUE;
97 }
98
99 static void find_short_object_filename(struct disambiguate_state *ds)
100 {
101 struct object_directory *odb;
102
103 for (odb = ds->repo->objects->odb; odb && !ds->ambiguous; odb = odb->next)
104 oidtree_each(odb_loose_cache(odb, &ds->bin_pfx),
105 &ds->bin_pfx, ds->len, match_prefix, ds);
106 }
107
108 static int match_hash(unsigned len, const unsigned char *a, const unsigned char *b)
109 {
110 do {
111 if (*a != *b)
112 return 0;
113 a++;
114 b++;
115 len -= 2;
116 } while (len > 1);
117 if (len)
118 if ((*a ^ *b) & 0xf0)
119 return 0;
120 return 1;
121 }
122
123 static void unique_in_midx(struct multi_pack_index *m,
124 struct disambiguate_state *ds)
125 {
126 uint32_t num, i, first = 0;
127 const struct object_id *current = NULL;
128 num = m->num_objects;
129
130 if (!num)
131 return;
132
133 bsearch_midx(&ds->bin_pfx, m, &first);
134
135 /*
136 * At this point, "first" is the location of the lowest object
137 * with an object name that could match "bin_pfx". See if we have
138 * 0, 1 or more objects that actually match(es).
139 */
140 for (i = first; i < num && !ds->ambiguous; i++) {
141 struct object_id oid;
142 current = nth_midxed_object_oid(&oid, m, i);
143 if (!match_hash(ds->len, ds->bin_pfx.hash, current->hash))
144 break;
145 update_candidates(ds, current);
146 }
147 }
148
149 static void unique_in_pack(struct packed_git *p,
150 struct disambiguate_state *ds)
151 {
152 uint32_t num, i, first = 0;
153
154 if (p->multi_pack_index)
155 return;
156
157 if (open_pack_index(p) || !p->num_objects)
158 return;
159
160 num = p->num_objects;
161 bsearch_pack(&ds->bin_pfx, p, &first);
162
163 /*
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).
167 */
168 for (i = first; i < num && !ds->ambiguous; i++) {
169 struct object_id oid;
170 nth_packed_object_id(&oid, p, i);
171 if (!match_hash(ds->len, ds->bin_pfx.hash, oid.hash))
172 break;
173 update_candidates(ds, &oid);
174 }
175 }
176
177 static void find_short_packed_object(struct disambiguate_state *ds)
178 {
179 struct multi_pack_index *m;
180 struct packed_git *p;
181
182 for (m = get_multi_pack_index(ds->repo); m && !ds->ambiguous;
183 m = m->next)
184 unique_in_midx(m, ds);
185 for (p = get_packed_git(ds->repo); p && !ds->ambiguous;
186 p = p->next)
187 unique_in_pack(p, ds);
188 }
189
190 static int finish_object_disambiguation(struct disambiguate_state *ds,
191 struct object_id *oid)
192 {
193 if (ds->ambiguous)
194 return SHORT_NAME_AMBIGUOUS;
195
196 if (!ds->candidate_exists)
197 return MISSING_OBJECT;
198
199 if (!ds->candidate_checked)
200 /*
201 * If this is the only candidate, there is no point
202 * calling the disambiguation hint callback.
203 *
204 * On the other hand, if the current candidate
205 * replaced an earlier candidate that did _not_ pass
206 * the disambiguation hint callback, then we do have
207 * more than one objects that match the short name
208 * given, so we should make sure this one matches;
209 * otherwise, if we discovered this one and the one
210 * that we previously discarded in the reverse order,
211 * we would end up showing different results in the
212 * same repository!
213 */
214 ds->candidate_ok = (!ds->disambiguate_fn_used ||
215 ds->fn(ds->repo, &ds->candidate, ds->cb_data));
216
217 if (!ds->candidate_ok)
218 return SHORT_NAME_AMBIGUOUS;
219
220 oidcpy(oid, &ds->candidate);
221 return 0;
222 }
223
224 static int disambiguate_commit_only(struct repository *r,
225 const struct object_id *oid,
226 void *cb_data_unused)
227 {
228 int kind = oid_object_info(r, oid, NULL);
229 return kind == OBJ_COMMIT;
230 }
231
232 static int disambiguate_committish_only(struct repository *r,
233 const struct object_id *oid,
234 void *cb_data_unused)
235 {
236 struct object *obj;
237 int kind;
238
239 kind = oid_object_info(r, oid, NULL);
240 if (kind == OBJ_COMMIT)
241 return 1;
242 if (kind != OBJ_TAG)
243 return 0;
244
245 /* We need to do this the hard way... */
246 obj = deref_tag(r, parse_object(r, oid), NULL, 0);
247 if (obj && obj->type == OBJ_COMMIT)
248 return 1;
249 return 0;
250 }
251
252 static int disambiguate_tree_only(struct repository *r,
253 const struct object_id *oid,
254 void *cb_data_unused)
255 {
256 int kind = oid_object_info(r, oid, NULL);
257 return kind == OBJ_TREE;
258 }
259
260 static int disambiguate_treeish_only(struct repository *r,
261 const struct object_id *oid,
262 void *cb_data_unused)
263 {
264 struct object *obj;
265 int kind;
266
267 kind = oid_object_info(r, oid, NULL);
268 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
269 return 1;
270 if (kind != OBJ_TAG)
271 return 0;
272
273 /* We need to do this the hard way... */
274 obj = deref_tag(r, parse_object(r, oid), NULL, 0);
275 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
276 return 1;
277 return 0;
278 }
279
280 static int disambiguate_blob_only(struct repository *r,
281 const struct object_id *oid,
282 void *cb_data_unused)
283 {
284 int kind = oid_object_info(r, oid, NULL);
285 return kind == OBJ_BLOB;
286 }
287
288 static disambiguate_hint_fn default_disambiguate_hint;
289
290 int set_disambiguate_hint_config(const char *var, const char *value)
291 {
292 static const struct {
293 const char *name;
294 disambiguate_hint_fn fn;
295 } hints[] = {
296 { "none", NULL },
297 { "commit", disambiguate_commit_only },
298 { "committish", disambiguate_committish_only },
299 { "tree", disambiguate_tree_only },
300 { "treeish", disambiguate_treeish_only },
301 { "blob", disambiguate_blob_only }
302 };
303 int i;
304
305 if (!value)
306 return config_error_nonbool(var);
307
308 for (i = 0; i < ARRAY_SIZE(hints); i++) {
309 if (!strcasecmp(value, hints[i].name)) {
310 default_disambiguate_hint = hints[i].fn;
311 return 0;
312 }
313 }
314
315 return error("unknown hint type for '%s': %s", var, value);
316 }
317
318 static int init_object_disambiguation(struct repository *r,
319 const char *name, int len,
320 struct disambiguate_state *ds)
321 {
322 int i;
323
324 if (len < MINIMUM_ABBREV || len > the_hash_algo->hexsz)
325 return -1;
326
327 memset(ds, 0, sizeof(*ds));
328
329 for (i = 0; i < len ;i++) {
330 unsigned char c = name[i];
331 unsigned char val;
332 if (c >= '0' && c <= '9')
333 val = c - '0';
334 else if (c >= 'a' && c <= 'f')
335 val = c - 'a' + 10;
336 else if (c >= 'A' && c <='F') {
337 val = c - 'A' + 10;
338 c -= 'A' - 'a';
339 }
340 else
341 return -1;
342 ds->hex_pfx[i] = c;
343 if (!(i & 1))
344 val <<= 4;
345 ds->bin_pfx.hash[i >> 1] |= val;
346 }
347
348 ds->len = len;
349 ds->hex_pfx[len] = '\0';
350 ds->repo = r;
351 prepare_alt_odb(r);
352 return 0;
353 }
354
355 struct ambiguous_output {
356 const struct disambiguate_state *ds;
357 struct strbuf advice;
358 struct strbuf sb;
359 };
360
361 static int show_ambiguous_object(const struct object_id *oid, void *data)
362 {
363 struct ambiguous_output *state = data;
364 const struct disambiguate_state *ds = state->ds;
365 struct strbuf *advice = &state->advice;
366 struct strbuf *sb = &state->sb;
367 int type;
368 const char *hash;
369
370 if (ds->fn && !ds->fn(ds->repo, oid, ds->cb_data))
371 return 0;
372
373 hash = repo_find_unique_abbrev(ds->repo, oid, DEFAULT_ABBREV);
374 type = oid_object_info(ds->repo, oid, NULL);
375
376 if (type < 0) {
377 /*
378 * TRANSLATORS: This is a line of ambiguous object
379 * output shown when we cannot look up or parse the
380 * object in question. E.g. "deadbeef [bad object]".
381 */
382 strbuf_addf(sb, _("%s [bad object]"), hash);
383 goto out;
384 }
385
386 assert(type == OBJ_TREE || type == OBJ_COMMIT ||
387 type == OBJ_BLOB || type == OBJ_TAG);
388
389 if (type == OBJ_COMMIT) {
390 struct strbuf date = STRBUF_INIT;
391 struct strbuf msg = STRBUF_INIT;
392 struct commit *commit = lookup_commit(ds->repo, oid);
393
394 if (commit) {
395 struct pretty_print_context pp = {0};
396 pp.date_mode.type = DATE_SHORT;
397 format_commit_message(commit, "%ad", &date, &pp);
398 format_commit_message(commit, "%s", &msg, &pp);
399 }
400
401 /*
402 * TRANSLATORS: This is a line of ambiguous commit
403 * object output. E.g.:
404 *
405 * "deadbeef commit 2021-01-01 - Some Commit Message"
406 */
407 strbuf_addf(sb, _("%s commit %s - %s"), hash, date.buf,
408 msg.buf);
409
410 strbuf_release(&date);
411 strbuf_release(&msg);
412 } else if (type == OBJ_TAG) {
413 struct tag *tag = lookup_tag(ds->repo, oid);
414
415 if (!parse_tag(tag) && tag->tag) {
416 /*
417 * TRANSLATORS: This is a line of ambiguous
418 * tag object output. E.g.:
419 *
420 * "deadbeef tag 2022-01-01 - Some Tag Message"
421 *
422 * The second argument is the YYYY-MM-DD found
423 * in the tag.
424 *
425 * The third argument is the "tag" string
426 * from object.c.
427 */
428 strbuf_addf(sb, _("%s tag %s - %s"), hash,
429 show_date(tag->date, 0, DATE_MODE(SHORT)),
430 tag->tag);
431 } else {
432 /*
433 * TRANSLATORS: This is a line of ambiguous
434 * tag object output where we couldn't parse
435 * the tag itself. E.g.:
436 *
437 * "deadbeef [bad tag, could not parse it]"
438 */
439 strbuf_addf(sb, _("%s [bad tag, could not parse it]"),
440 hash);
441 }
442 } else if (type == OBJ_TREE) {
443 /*
444 * TRANSLATORS: This is a line of ambiguous <type>
445 * object output. E.g. "deadbeef tree".
446 */
447 strbuf_addf(sb, _("%s tree"), hash);
448 } else if (type == OBJ_BLOB) {
449 /*
450 * TRANSLATORS: This is a line of ambiguous <type>
451 * object output. E.g. "deadbeef blob".
452 */
453 strbuf_addf(sb, _("%s blob"), hash);
454 }
455
456
457 out:
458 /*
459 * TRANSLATORS: This is line item of ambiguous object output
460 * from describe_ambiguous_object() above. For RTL languages
461 * you'll probably want to swap the "%s" and leading " " space
462 * around.
463 */
464 strbuf_addf(advice, _(" %s\n"), sb->buf);
465
466 strbuf_reset(sb);
467 return 0;
468 }
469
470 static int collect_ambiguous(const struct object_id *oid, void *data)
471 {
472 oid_array_append(data, oid);
473 return 0;
474 }
475
476 static int repo_collect_ambiguous(struct repository *r,
477 const struct object_id *oid,
478 void *data)
479 {
480 return collect_ambiguous(oid, data);
481 }
482
483 static int sort_ambiguous(const void *a, const void *b, void *ctx)
484 {
485 struct repository *sort_ambiguous_repo = ctx;
486 int a_type = oid_object_info(sort_ambiguous_repo, a, NULL);
487 int b_type = oid_object_info(sort_ambiguous_repo, b, NULL);
488 int a_type_sort;
489 int b_type_sort;
490
491 /*
492 * Sorts by hash within the same object type, just as
493 * oid_array_for_each_unique() would do.
494 */
495 if (a_type == b_type)
496 return oidcmp(a, b);
497
498 /*
499 * Between object types show tags, then commits, and finally
500 * trees and blobs.
501 *
502 * The object_type enum is commit, tree, blob, tag, but we
503 * want tag, commit, tree blob. Cleverly (perhaps too
504 * cleverly) do that with modulus, since the enum assigns 1 to
505 * commit, so tag becomes 0.
506 */
507 a_type_sort = a_type % 4;
508 b_type_sort = b_type % 4;
509 return a_type_sort > b_type_sort ? 1 : -1;
510 }
511
512 static void sort_ambiguous_oid_array(struct repository *r, struct oid_array *a)
513 {
514 QSORT_S(a->oid, a->nr, sort_ambiguous, r);
515 }
516
517 static enum get_oid_result get_short_oid(struct repository *r,
518 const char *name, int len,
519 struct object_id *oid,
520 unsigned flags)
521 {
522 int status;
523 struct disambiguate_state ds;
524 int quietly = !!(flags & GET_OID_QUIETLY);
525
526 if (init_object_disambiguation(r, name, len, &ds) < 0)
527 return -1;
528
529 if (HAS_MULTI_BITS(flags & GET_OID_DISAMBIGUATORS))
530 BUG("multiple get_short_oid disambiguator flags");
531
532 if (flags & GET_OID_COMMIT)
533 ds.fn = disambiguate_commit_only;
534 else if (flags & GET_OID_COMMITTISH)
535 ds.fn = disambiguate_committish_only;
536 else if (flags & GET_OID_TREE)
537 ds.fn = disambiguate_tree_only;
538 else if (flags & GET_OID_TREEISH)
539 ds.fn = disambiguate_treeish_only;
540 else if (flags & GET_OID_BLOB)
541 ds.fn = disambiguate_blob_only;
542 else
543 ds.fn = default_disambiguate_hint;
544
545 find_short_object_filename(&ds);
546 find_short_packed_object(&ds);
547 status = finish_object_disambiguation(&ds, oid);
548
549 /*
550 * If we didn't find it, do the usual reprepare() slow-path,
551 * since the object may have recently been added to the repository
552 * or migrated from loose to packed.
553 */
554 if (status == MISSING_OBJECT) {
555 reprepare_packed_git(r);
556 find_short_object_filename(&ds);
557 find_short_packed_object(&ds);
558 status = finish_object_disambiguation(&ds, oid);
559 }
560
561 if (!quietly && (status == SHORT_NAME_AMBIGUOUS)) {
562 struct oid_array collect = OID_ARRAY_INIT;
563 struct ambiguous_output out = {
564 .ds = &ds,
565 .sb = STRBUF_INIT,
566 .advice = STRBUF_INIT,
567 };
568
569 error(_("short object ID %s is ambiguous"), ds.hex_pfx);
570
571 /*
572 * We may still have ambiguity if we simply saw a series of
573 * candidates that did not satisfy our hint function. In
574 * that case, we still want to show them, so disable the hint
575 * function entirely.
576 */
577 if (!ds.ambiguous)
578 ds.fn = NULL;
579
580 repo_for_each_abbrev(r, ds.hex_pfx, collect_ambiguous, &collect);
581 sort_ambiguous_oid_array(r, &collect);
582
583 if (oid_array_for_each(&collect, show_ambiguous_object, &out))
584 BUG("show_ambiguous_object shouldn't return non-zero");
585
586 /*
587 * TRANSLATORS: The argument is the list of ambiguous
588 * objects composed in show_ambiguous_object(). See
589 * its "TRANSLATORS" comments for details.
590 */
591 advise(_("The candidates are:\n%s"), out.advice.buf);
592
593 oid_array_clear(&collect);
594 strbuf_release(&out.advice);
595 strbuf_release(&out.sb);
596 }
597
598 return status;
599 }
600
601 int repo_for_each_abbrev(struct repository *r, const char *prefix,
602 each_abbrev_fn fn, void *cb_data)
603 {
604 struct oid_array collect = OID_ARRAY_INIT;
605 struct disambiguate_state ds;
606 int ret;
607
608 if (init_object_disambiguation(r, prefix, strlen(prefix), &ds) < 0)
609 return -1;
610
611 ds.always_call_fn = 1;
612 ds.fn = repo_collect_ambiguous;
613 ds.cb_data = &collect;
614 find_short_object_filename(&ds);
615 find_short_packed_object(&ds);
616
617 ret = oid_array_for_each_unique(&collect, fn, cb_data);
618 oid_array_clear(&collect);
619 return ret;
620 }
621
622 /*
623 * Return the slot of the most-significant bit set in "val". There are various
624 * ways to do this quickly with fls() or __builtin_clzl(), but speed is
625 * probably not a big deal here.
626 */
627 static unsigned msb(unsigned long val)
628 {
629 unsigned r = 0;
630 while (val >>= 1)
631 r++;
632 return r;
633 }
634
635 struct min_abbrev_data {
636 unsigned int init_len;
637 unsigned int cur_len;
638 char *hex;
639 struct repository *repo;
640 const struct object_id *oid;
641 };
642
643 static inline char get_hex_char_from_oid(const struct object_id *oid,
644 unsigned int pos)
645 {
646 static const char hex[] = "0123456789abcdef";
647
648 if ((pos & 1) == 0)
649 return hex[oid->hash[pos >> 1] >> 4];
650 else
651 return hex[oid->hash[pos >> 1] & 0xf];
652 }
653
654 static int extend_abbrev_len(const struct object_id *oid, void *cb_data)
655 {
656 struct min_abbrev_data *mad = cb_data;
657
658 unsigned int i = mad->init_len;
659 while (mad->hex[i] && mad->hex[i] == get_hex_char_from_oid(oid, i))
660 i++;
661
662 if (i < GIT_MAX_RAWSZ && i >= mad->cur_len)
663 mad->cur_len = i + 1;
664
665 return 0;
666 }
667
668 static int repo_extend_abbrev_len(struct repository *r,
669 const struct object_id *oid,
670 void *cb_data)
671 {
672 return extend_abbrev_len(oid, cb_data);
673 }
674
675 static void find_abbrev_len_for_midx(struct multi_pack_index *m,
676 struct min_abbrev_data *mad)
677 {
678 int match = 0;
679 uint32_t num, first = 0;
680 struct object_id oid;
681 const struct object_id *mad_oid;
682
683 if (!m->num_objects)
684 return;
685
686 num = m->num_objects;
687 mad_oid = mad->oid;
688 match = bsearch_midx(mad_oid, m, &first);
689
690 /*
691 * first is now the position in the packfile where we would insert
692 * mad->hash if it does not exist (or the position of mad->hash if
693 * it does exist). Hence, we consider a maximum of two objects
694 * nearby for the abbreviation length.
695 */
696 mad->init_len = 0;
697 if (!match) {
698 if (nth_midxed_object_oid(&oid, m, first))
699 extend_abbrev_len(&oid, mad);
700 } else if (first < num - 1) {
701 if (nth_midxed_object_oid(&oid, m, first + 1))
702 extend_abbrev_len(&oid, mad);
703 }
704 if (first > 0) {
705 if (nth_midxed_object_oid(&oid, m, first - 1))
706 extend_abbrev_len(&oid, mad);
707 }
708 mad->init_len = mad->cur_len;
709 }
710
711 static void find_abbrev_len_for_pack(struct packed_git *p,
712 struct min_abbrev_data *mad)
713 {
714 int match = 0;
715 uint32_t num, first = 0;
716 struct object_id oid;
717 const struct object_id *mad_oid;
718
719 if (p->multi_pack_index)
720 return;
721
722 if (open_pack_index(p) || !p->num_objects)
723 return;
724
725 num = p->num_objects;
726 mad_oid = mad->oid;
727 match = bsearch_pack(mad_oid, p, &first);
728
729 /*
730 * first is now the position in the packfile where we would insert
731 * mad->hash if it does not exist (or the position of mad->hash if
732 * it does exist). Hence, we consider a maximum of two objects
733 * nearby for the abbreviation length.
734 */
735 mad->init_len = 0;
736 if (!match) {
737 if (!nth_packed_object_id(&oid, p, first))
738 extend_abbrev_len(&oid, mad);
739 } else if (first < num - 1) {
740 if (!nth_packed_object_id(&oid, p, first + 1))
741 extend_abbrev_len(&oid, mad);
742 }
743 if (first > 0) {
744 if (!nth_packed_object_id(&oid, p, first - 1))
745 extend_abbrev_len(&oid, mad);
746 }
747 mad->init_len = mad->cur_len;
748 }
749
750 static void find_abbrev_len_packed(struct min_abbrev_data *mad)
751 {
752 struct multi_pack_index *m;
753 struct packed_git *p;
754
755 for (m = get_multi_pack_index(mad->repo); m; m = m->next)
756 find_abbrev_len_for_midx(m, mad);
757 for (p = get_packed_git(mad->repo); p; p = p->next)
758 find_abbrev_len_for_pack(p, mad);
759 }
760
761 int repo_find_unique_abbrev_r(struct repository *r, char *hex,
762 const struct object_id *oid, int len)
763 {
764 struct disambiguate_state ds;
765 struct min_abbrev_data mad;
766 struct object_id oid_ret;
767 const unsigned hexsz = r->hash_algo->hexsz;
768
769 if (len < 0) {
770 unsigned long count = repo_approximate_object_count(r);
771 /*
772 * Add one because the MSB only tells us the highest bit set,
773 * not including the value of all the _other_ bits (so "15"
774 * is only one off of 2^4, but the MSB is the 3rd bit.
775 */
776 len = msb(count) + 1;
777 /*
778 * We now know we have on the order of 2^len objects, which
779 * expects a collision at 2^(len/2). But we also care about hex
780 * chars, not bits, and there are 4 bits per hex. So all
781 * together we need to divide by 2 and round up.
782 */
783 len = DIV_ROUND_UP(len, 2);
784 /*
785 * For very small repos, we stick with our regular fallback.
786 */
787 if (len < FALLBACK_DEFAULT_ABBREV)
788 len = FALLBACK_DEFAULT_ABBREV;
789 }
790
791 oid_to_hex_r(hex, oid);
792 if (len == hexsz || !len)
793 return hexsz;
794
795 mad.repo = r;
796 mad.init_len = len;
797 mad.cur_len = len;
798 mad.hex = hex;
799 mad.oid = oid;
800
801 find_abbrev_len_packed(&mad);
802
803 if (init_object_disambiguation(r, hex, mad.cur_len, &ds) < 0)
804 return -1;
805
806 ds.fn = repo_extend_abbrev_len;
807 ds.always_call_fn = 1;
808 ds.cb_data = (void *)&mad;
809
810 find_short_object_filename(&ds);
811 (void)finish_object_disambiguation(&ds, &oid_ret);
812
813 hex[mad.cur_len] = 0;
814 return mad.cur_len;
815 }
816
817 const char *repo_find_unique_abbrev(struct repository *r,
818 const struct object_id *oid,
819 int len)
820 {
821 static int bufno;
822 static char hexbuffer[4][GIT_MAX_HEXSZ + 1];
823 char *hex = hexbuffer[bufno];
824 bufno = (bufno + 1) % ARRAY_SIZE(hexbuffer);
825 repo_find_unique_abbrev_r(r, hex, oid, len);
826 return hex;
827 }
828
829 static int ambiguous_path(const char *path, int len)
830 {
831 int slash = 1;
832 int cnt;
833
834 for (cnt = 0; cnt < len; cnt++) {
835 switch (*path++) {
836 case '\0':
837 break;
838 case '/':
839 if (slash)
840 break;
841 slash = 1;
842 continue;
843 case '.':
844 continue;
845 default:
846 slash = 0;
847 continue;
848 }
849 break;
850 }
851 return slash;
852 }
853
854 static inline int at_mark(const char *string, int len,
855 const char **suffix, int nr)
856 {
857 int i;
858
859 for (i = 0; i < nr; i++) {
860 int suffix_len = strlen(suffix[i]);
861 if (suffix_len <= len
862 && !strncasecmp(string, suffix[i], suffix_len))
863 return suffix_len;
864 }
865 return 0;
866 }
867
868 static inline int upstream_mark(const char *string, int len)
869 {
870 const char *suffix[] = { "@{upstream}", "@{u}" };
871 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
872 }
873
874 static inline int push_mark(const char *string, int len)
875 {
876 const char *suffix[] = { "@{push}" };
877 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
878 }
879
880 static enum get_oid_result get_oid_1(struct repository *r, const char *name, int len, struct object_id *oid, unsigned lookup_flags);
881 static int interpret_nth_prior_checkout(struct repository *r, const char *name, int namelen, struct strbuf *buf);
882
883 static int get_oid_basic(struct repository *r, const char *str, int len,
884 struct object_id *oid, unsigned int flags)
885 {
886 static const char *warn_msg = "refname '%.*s' is ambiguous.";
887 static const char *object_name_msg = N_(
888 "Git normally never creates a ref that ends with 40 hex characters\n"
889 "because it will be ignored when you just specify 40-hex. These refs\n"
890 "may be created by mistake. For example,\n"
891 "\n"
892 " git switch -c $br $(git rev-parse ...)\n"
893 "\n"
894 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
895 "examine these refs and maybe delete them. Turn this message off by\n"
896 "running \"git config advice.objectNameWarning false\"");
897 struct object_id tmp_oid;
898 char *real_ref = NULL;
899 int refs_found = 0;
900 int at, reflog_len, nth_prior = 0;
901
902 if (len == r->hash_algo->hexsz && !get_oid_hex(str, oid)) {
903 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
904 refs_found = repo_dwim_ref(r, str, len, &tmp_oid, &real_ref, 0);
905 if (refs_found > 0) {
906 warning(warn_msg, len, str);
907 if (advice_enabled(ADVICE_OBJECT_NAME_WARNING))
908 fprintf(stderr, "%s\n", _(object_name_msg));
909 }
910 free(real_ref);
911 }
912 return 0;
913 }
914
915 /* basic@{time or number or -number} format to query ref-log */
916 reflog_len = at = 0;
917 if (len && str[len-1] == '}') {
918 for (at = len-4; at >= 0; at--) {
919 if (str[at] == '@' && str[at+1] == '{') {
920 if (str[at+2] == '-') {
921 if (at != 0)
922 /* @{-N} not at start */
923 return -1;
924 nth_prior = 1;
925 continue;
926 }
927 if (!upstream_mark(str + at, len - at) &&
928 !push_mark(str + at, len - at)) {
929 reflog_len = (len-1) - (at+2);
930 len = at;
931 }
932 break;
933 }
934 }
935 }
936
937 /* Accept only unambiguous ref paths. */
938 if (len && ambiguous_path(str, len))
939 return -1;
940
941 if (nth_prior) {
942 struct strbuf buf = STRBUF_INIT;
943 int detached;
944
945 if (interpret_nth_prior_checkout(r, str, len, &buf) > 0) {
946 detached = (buf.len == r->hash_algo->hexsz && !get_oid_hex(buf.buf, oid));
947 strbuf_release(&buf);
948 if (detached)
949 return 0;
950 }
951 }
952
953 if (!len && reflog_len)
954 /* allow "@{...}" to mean the current branch reflog */
955 refs_found = repo_dwim_ref(r, "HEAD", 4, oid, &real_ref, 0);
956 else if (reflog_len)
957 refs_found = repo_dwim_log(r, str, len, oid, &real_ref);
958 else
959 refs_found = repo_dwim_ref(r, str, len, oid, &real_ref, 0);
960
961 if (!refs_found)
962 return -1;
963
964 if (warn_ambiguous_refs && !(flags & GET_OID_QUIETLY) &&
965 (refs_found > 1 ||
966 !get_short_oid(r, str, len, &tmp_oid, GET_OID_QUIETLY)))
967 warning(warn_msg, len, str);
968
969 if (reflog_len) {
970 int nth, i;
971 timestamp_t at_time;
972 timestamp_t co_time;
973 int co_tz, co_cnt;
974
975 /* Is it asking for N-th entry, or approxidate? */
976 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
977 char ch = str[at+2+i];
978 if ('0' <= ch && ch <= '9')
979 nth = nth * 10 + ch - '0';
980 else
981 nth = -1;
982 }
983 if (100000000 <= nth) {
984 at_time = nth;
985 nth = -1;
986 } else if (0 <= nth)
987 at_time = 0;
988 else {
989 int errors = 0;
990 char *tmp = xstrndup(str + at + 2, reflog_len);
991 at_time = approxidate_careful(tmp, &errors);
992 free(tmp);
993 if (errors) {
994 free(real_ref);
995 return -1;
996 }
997 }
998 if (read_ref_at(get_main_ref_store(r),
999 real_ref, flags, at_time, nth, oid, NULL,
1000 &co_time, &co_tz, &co_cnt)) {
1001 if (!len) {
1002 if (!skip_prefix(real_ref, "refs/heads/", &str))
1003 str = "HEAD";
1004 len = strlen(str);
1005 }
1006 if (at_time) {
1007 if (!(flags & GET_OID_QUIETLY)) {
1008 warning(_("log for '%.*s' only goes back to %s"),
1009 len, str,
1010 show_date(co_time, co_tz, DATE_MODE(RFC2822)));
1011 }
1012 } else {
1013 if (flags & GET_OID_QUIETLY) {
1014 exit(128);
1015 }
1016 die(_("log for '%.*s' only has %d entries"),
1017 len, str, co_cnt);
1018 }
1019 }
1020 }
1021
1022 free(real_ref);
1023 return 0;
1024 }
1025
1026 static enum get_oid_result get_parent(struct repository *r,
1027 const char *name, int len,
1028 struct object_id *result, int idx)
1029 {
1030 struct object_id oid;
1031 enum get_oid_result ret = get_oid_1(r, name, len, &oid,
1032 GET_OID_COMMITTISH);
1033 struct commit *commit;
1034 struct commit_list *p;
1035
1036 if (ret)
1037 return ret;
1038 commit = lookup_commit_reference(r, &oid);
1039 if (parse_commit(commit))
1040 return MISSING_OBJECT;
1041 if (!idx) {
1042 oidcpy(result, &commit->object.oid);
1043 return FOUND;
1044 }
1045 p = commit->parents;
1046 while (p) {
1047 if (!--idx) {
1048 oidcpy(result, &p->item->object.oid);
1049 return FOUND;
1050 }
1051 p = p->next;
1052 }
1053 return MISSING_OBJECT;
1054 }
1055
1056 static enum get_oid_result get_nth_ancestor(struct repository *r,
1057 const char *name, int len,
1058 struct object_id *result,
1059 int generation)
1060 {
1061 struct object_id oid;
1062 struct commit *commit;
1063 int ret;
1064
1065 ret = get_oid_1(r, name, len, &oid, GET_OID_COMMITTISH);
1066 if (ret)
1067 return ret;
1068 commit = lookup_commit_reference(r, &oid);
1069 if (!commit)
1070 return MISSING_OBJECT;
1071
1072 while (generation--) {
1073 if (parse_commit(commit) || !commit->parents)
1074 return MISSING_OBJECT;
1075 commit = commit->parents->item;
1076 }
1077 oidcpy(result, &commit->object.oid);
1078 return FOUND;
1079 }
1080
1081 struct object *repo_peel_to_type(struct repository *r, const char *name, int namelen,
1082 struct object *o, enum object_type expected_type)
1083 {
1084 if (name && !namelen)
1085 namelen = strlen(name);
1086 while (1) {
1087 if (!o || (!o->parsed && !parse_object(r, &o->oid)))
1088 return NULL;
1089 if (expected_type == OBJ_ANY || o->type == expected_type)
1090 return o;
1091 if (o->type == OBJ_TAG)
1092 o = ((struct tag*) o)->tagged;
1093 else if (o->type == OBJ_COMMIT)
1094 o = &(repo_get_commit_tree(r, ((struct commit *)o))->object);
1095 else {
1096 if (name)
1097 error("%.*s: expected %s type, but the object "
1098 "dereferences to %s type",
1099 namelen, name, type_name(expected_type),
1100 type_name(o->type));
1101 return NULL;
1102 }
1103 }
1104 }
1105
1106 static int peel_onion(struct repository *r, const char *name, int len,
1107 struct object_id *oid, unsigned lookup_flags)
1108 {
1109 struct object_id outer;
1110 const char *sp;
1111 unsigned int expected_type = 0;
1112 struct object *o;
1113
1114 /*
1115 * "ref^{type}" dereferences ref repeatedly until you cannot
1116 * dereference anymore, or you get an object of given type,
1117 * whichever comes first. "ref^{}" means just dereference
1118 * tags until you get a non-tag. "ref^0" is a shorthand for
1119 * "ref^{commit}". "commit^{tree}" could be used to find the
1120 * top-level tree of the given commit.
1121 */
1122 if (len < 4 || name[len-1] != '}')
1123 return -1;
1124
1125 for (sp = name + len - 1; name <= sp; sp--) {
1126 int ch = *sp;
1127 if (ch == '{' && name < sp && sp[-1] == '^')
1128 break;
1129 }
1130 if (sp <= name)
1131 return -1;
1132
1133 sp++; /* beginning of type name, or closing brace for empty */
1134 if (starts_with(sp, "commit}"))
1135 expected_type = OBJ_COMMIT;
1136 else if (starts_with(sp, "tag}"))
1137 expected_type = OBJ_TAG;
1138 else if (starts_with(sp, "tree}"))
1139 expected_type = OBJ_TREE;
1140 else if (starts_with(sp, "blob}"))
1141 expected_type = OBJ_BLOB;
1142 else if (starts_with(sp, "object}"))
1143 expected_type = OBJ_ANY;
1144 else if (sp[0] == '}')
1145 expected_type = OBJ_NONE;
1146 else if (sp[0] == '/')
1147 expected_type = OBJ_COMMIT;
1148 else
1149 return -1;
1150
1151 lookup_flags &= ~GET_OID_DISAMBIGUATORS;
1152 if (expected_type == OBJ_COMMIT)
1153 lookup_flags |= GET_OID_COMMITTISH;
1154 else if (expected_type == OBJ_TREE)
1155 lookup_flags |= GET_OID_TREEISH;
1156
1157 if (get_oid_1(r, name, sp - name - 2, &outer, lookup_flags))
1158 return -1;
1159
1160 o = parse_object(r, &outer);
1161 if (!o)
1162 return -1;
1163 if (!expected_type) {
1164 o = deref_tag(r, o, name, sp - name - 2);
1165 if (!o || (!o->parsed && !parse_object(r, &o->oid)))
1166 return -1;
1167 oidcpy(oid, &o->oid);
1168 return 0;
1169 }
1170
1171 /*
1172 * At this point, the syntax look correct, so
1173 * if we do not get the needed object, we should
1174 * barf.
1175 */
1176 o = repo_peel_to_type(r, name, len, o, expected_type);
1177 if (!o)
1178 return -1;
1179
1180 oidcpy(oid, &o->oid);
1181 if (sp[0] == '/') {
1182 /* "$commit^{/foo}" */
1183 char *prefix;
1184 int ret;
1185 struct commit_list *list = NULL;
1186
1187 /*
1188 * $commit^{/}. Some regex implementation may reject.
1189 * We don't need regex anyway. '' pattern always matches.
1190 */
1191 if (sp[1] == '}')
1192 return 0;
1193
1194 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
1195 commit_list_insert((struct commit *)o, &list);
1196 ret = get_oid_oneline(r, prefix, oid, list);
1197 free(prefix);
1198 return ret;
1199 }
1200 return 0;
1201 }
1202
1203 static int get_describe_name(struct repository *r,
1204 const char *name, int len,
1205 struct object_id *oid)
1206 {
1207 const char *cp;
1208 unsigned flags = GET_OID_QUIETLY | GET_OID_COMMIT;
1209
1210 for (cp = name + len - 1; name + 2 <= cp; cp--) {
1211 char ch = *cp;
1212 if (!isxdigit(ch)) {
1213 /* We must be looking at g in "SOMETHING-g"
1214 * for it to be describe output.
1215 */
1216 if (ch == 'g' && cp[-1] == '-') {
1217 cp++;
1218 len -= cp - name;
1219 return get_short_oid(r,
1220 cp, len, oid, flags);
1221 }
1222 }
1223 }
1224 return -1;
1225 }
1226
1227 static enum get_oid_result get_oid_1(struct repository *r,
1228 const char *name, int len,
1229 struct object_id *oid,
1230 unsigned lookup_flags)
1231 {
1232 int ret, has_suffix;
1233 const char *cp;
1234
1235 /*
1236 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
1237 */
1238 has_suffix = 0;
1239 for (cp = name + len - 1; name <= cp; cp--) {
1240 int ch = *cp;
1241 if ('0' <= ch && ch <= '9')
1242 continue;
1243 if (ch == '~' || ch == '^')
1244 has_suffix = ch;
1245 break;
1246 }
1247
1248 if (has_suffix) {
1249 unsigned int num = 0;
1250 int len1 = cp - name;
1251 cp++;
1252 while (cp < name + len) {
1253 unsigned int digit = *cp++ - '0';
1254 if (unsigned_mult_overflows(num, 10))
1255 return MISSING_OBJECT;
1256 num *= 10;
1257 if (unsigned_add_overflows(num, digit))
1258 return MISSING_OBJECT;
1259 num += digit;
1260 }
1261 if (!num && len1 == len - 1)
1262 num = 1;
1263 else if (num > INT_MAX)
1264 return MISSING_OBJECT;
1265 if (has_suffix == '^')
1266 return get_parent(r, name, len1, oid, num);
1267 /* else if (has_suffix == '~') -- goes without saying */
1268 return get_nth_ancestor(r, name, len1, oid, num);
1269 }
1270
1271 ret = peel_onion(r, name, len, oid, lookup_flags);
1272 if (!ret)
1273 return FOUND;
1274
1275 ret = get_oid_basic(r, name, len, oid, lookup_flags);
1276 if (!ret)
1277 return FOUND;
1278
1279 /* It could be describe output that is "SOMETHING-gXXXX" */
1280 ret = get_describe_name(r, name, len, oid);
1281 if (!ret)
1282 return FOUND;
1283
1284 return get_short_oid(r, name, len, oid, lookup_flags);
1285 }
1286
1287 /*
1288 * This interprets names like ':/Initial revision of "git"' by searching
1289 * through history and returning the first commit whose message starts
1290 * the given regular expression.
1291 *
1292 * For negative-matching, prefix the pattern-part with '!-', like: ':/!-WIP'.
1293 *
1294 * For a literal '!' character at the beginning of a pattern, you have to repeat
1295 * that, like: ':/!!foo'
1296 *
1297 * For future extension, all other sequences beginning with ':/!' are reserved.
1298 */
1299
1300 /* Remember to update object flag allocation in object.h */
1301 #define ONELINE_SEEN (1u<<20)
1302
1303 struct handle_one_ref_cb {
1304 struct repository *repo;
1305 struct commit_list **list;
1306 };
1307
1308 static int handle_one_ref(const char *path, const struct object_id *oid,
1309 int flag UNUSED,
1310 void *cb_data)
1311 {
1312 struct handle_one_ref_cb *cb = cb_data;
1313 struct commit_list **list = cb->list;
1314 struct object *object = parse_object(cb->repo, oid);
1315 if (!object)
1316 return 0;
1317 if (object->type == OBJ_TAG) {
1318 object = deref_tag(cb->repo, object, path,
1319 strlen(path));
1320 if (!object)
1321 return 0;
1322 }
1323 if (object->type != OBJ_COMMIT)
1324 return 0;
1325 commit_list_insert((struct commit *)object, list);
1326 return 0;
1327 }
1328
1329 static int get_oid_oneline(struct repository *r,
1330 const char *prefix, struct object_id *oid,
1331 struct commit_list *list)
1332 {
1333 struct commit_list *backup = NULL, *l;
1334 int found = 0;
1335 int negative = 0;
1336 regex_t regex;
1337
1338 if (prefix[0] == '!') {
1339 prefix++;
1340
1341 if (prefix[0] == '-') {
1342 prefix++;
1343 negative = 1;
1344 } else if (prefix[0] != '!') {
1345 return -1;
1346 }
1347 }
1348
1349 if (regcomp(&regex, prefix, REG_EXTENDED))
1350 return -1;
1351
1352 for (l = list; l; l = l->next) {
1353 l->item->object.flags |= ONELINE_SEEN;
1354 commit_list_insert(l->item, &backup);
1355 }
1356 while (list) {
1357 const char *p, *buf;
1358 struct commit *commit;
1359 int matches;
1360
1361 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
1362 if (!parse_object(r, &commit->object.oid))
1363 continue;
1364 buf = get_commit_buffer(commit, NULL);
1365 p = strstr(buf, "\n\n");
1366 matches = negative ^ (p && !regexec(&regex, p + 2, 0, NULL, 0));
1367 unuse_commit_buffer(commit, buf);
1368
1369 if (matches) {
1370 oidcpy(oid, &commit->object.oid);
1371 found = 1;
1372 break;
1373 }
1374 }
1375 regfree(&regex);
1376 free_commit_list(list);
1377 for (l = backup; l; l = l->next)
1378 clear_commit_marks(l->item, ONELINE_SEEN);
1379 free_commit_list(backup);
1380 return found ? 0 : -1;
1381 }
1382
1383 struct grab_nth_branch_switch_cbdata {
1384 int remaining;
1385 struct strbuf *sb;
1386 };
1387
1388 static int grab_nth_branch_switch(struct object_id *ooid UNUSED,
1389 struct object_id *noid UNUSED,
1390 const char *email UNUSED,
1391 timestamp_t timestamp UNUSED,
1392 int tz UNUSED,
1393 const char *message, void *cb_data)
1394 {
1395 struct grab_nth_branch_switch_cbdata *cb = cb_data;
1396 const char *match = NULL, *target = NULL;
1397 size_t len;
1398
1399 if (skip_prefix(message, "checkout: moving from ", &match))
1400 target = strstr(match, " to ");
1401
1402 if (!match || !target)
1403 return 0;
1404 if (--(cb->remaining) == 0) {
1405 len = target - match;
1406 strbuf_reset(cb->sb);
1407 strbuf_add(cb->sb, match, len);
1408 return 1; /* we are done */
1409 }
1410 return 0;
1411 }
1412
1413 /*
1414 * Parse @{-N} syntax, return the number of characters parsed
1415 * if successful; otherwise signal an error with negative value.
1416 */
1417 static int interpret_nth_prior_checkout(struct repository *r,
1418 const char *name, int namelen,
1419 struct strbuf *buf)
1420 {
1421 long nth;
1422 int retval;
1423 struct grab_nth_branch_switch_cbdata cb;
1424 const char *brace;
1425 char *num_end;
1426
1427 if (namelen < 4)
1428 return -1;
1429 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
1430 return -1;
1431 brace = memchr(name, '}', namelen);
1432 if (!brace)
1433 return -1;
1434 nth = strtol(name + 3, &num_end, 10);
1435 if (num_end != brace)
1436 return -1;
1437 if (nth <= 0)
1438 return -1;
1439 cb.remaining = nth;
1440 cb.sb = buf;
1441
1442 retval = refs_for_each_reflog_ent_reverse(get_main_ref_store(r),
1443 "HEAD", grab_nth_branch_switch, &cb);
1444 if (0 < retval) {
1445 retval = brace - name + 1;
1446 } else
1447 retval = 0;
1448
1449 return retval;
1450 }
1451
1452 int repo_get_oid_mb(struct repository *r,
1453 const char *name,
1454 struct object_id *oid)
1455 {
1456 struct commit *one, *two;
1457 struct commit_list *mbs;
1458 struct object_id oid_tmp;
1459 const char *dots;
1460 int st;
1461
1462 dots = strstr(name, "...");
1463 if (!dots)
1464 return repo_get_oid(r, name, oid);
1465 if (dots == name)
1466 st = repo_get_oid(r, "HEAD", &oid_tmp);
1467 else {
1468 struct strbuf sb;
1469 strbuf_init(&sb, dots - name);
1470 strbuf_add(&sb, name, dots - name);
1471 st = repo_get_oid_committish(r, sb.buf, &oid_tmp);
1472 strbuf_release(&sb);
1473 }
1474 if (st)
1475 return st;
1476 one = lookup_commit_reference_gently(r, &oid_tmp, 0);
1477 if (!one)
1478 return -1;
1479
1480 if (repo_get_oid_committish(r, dots[3] ? (dots + 3) : "HEAD", &oid_tmp))
1481 return -1;
1482 two = lookup_commit_reference_gently(r, &oid_tmp, 0);
1483 if (!two)
1484 return -1;
1485 mbs = repo_get_merge_bases(r, one, two);
1486 if (!mbs || mbs->next)
1487 st = -1;
1488 else {
1489 st = 0;
1490 oidcpy(oid, &mbs->item->object.oid);
1491 }
1492 free_commit_list(mbs);
1493 return st;
1494 }
1495
1496 /* parse @something syntax, when 'something' is not {.*} */
1497 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1498 {
1499 const char *next;
1500
1501 if (len || name[1] == '{')
1502 return -1;
1503
1504 /* make sure it's a single @, or @@{.*}, not @foo */
1505 next = memchr(name + len + 1, '@', namelen - len - 1);
1506 if (next && next[1] != '{')
1507 return -1;
1508 if (!next)
1509 next = name + namelen;
1510 if (next != name + 1)
1511 return -1;
1512
1513 strbuf_reset(buf);
1514 strbuf_add(buf, "HEAD", 4);
1515 return 1;
1516 }
1517
1518 static int reinterpret(struct repository *r,
1519 const char *name, int namelen, int len,
1520 struct strbuf *buf, unsigned allowed)
1521 {
1522 /* we have extra data, which might need further processing */
1523 struct strbuf tmp = STRBUF_INIT;
1524 int used = buf->len;
1525 int ret;
1526 struct interpret_branch_name_options options = {
1527 .allowed = allowed
1528 };
1529
1530 strbuf_add(buf, name + len, namelen - len);
1531 ret = repo_interpret_branch_name(r, buf->buf, buf->len, &tmp, &options);
1532 /* that data was not interpreted, remove our cruft */
1533 if (ret < 0) {
1534 strbuf_setlen(buf, used);
1535 return len;
1536 }
1537 strbuf_reset(buf);
1538 strbuf_addbuf(buf, &tmp);
1539 strbuf_release(&tmp);
1540 /* tweak for size of {-N} versus expanded ref name */
1541 return ret - used + len;
1542 }
1543
1544 static void set_shortened_ref(struct repository *r, struct strbuf *buf, const char *ref)
1545 {
1546 char *s = refs_shorten_unambiguous_ref(get_main_ref_store(r), ref, 0);
1547 strbuf_reset(buf);
1548 strbuf_addstr(buf, s);
1549 free(s);
1550 }
1551
1552 static int branch_interpret_allowed(const char *refname, unsigned allowed)
1553 {
1554 if (!allowed)
1555 return 1;
1556
1557 if ((allowed & INTERPRET_BRANCH_LOCAL) &&
1558 starts_with(refname, "refs/heads/"))
1559 return 1;
1560 if ((allowed & INTERPRET_BRANCH_REMOTE) &&
1561 starts_with(refname, "refs/remotes/"))
1562 return 1;
1563
1564 return 0;
1565 }
1566
1567 static int interpret_branch_mark(struct repository *r,
1568 const char *name, int namelen,
1569 int at, struct strbuf *buf,
1570 int (*get_mark)(const char *, int),
1571 const char *(*get_data)(struct branch *,
1572 struct strbuf *),
1573 const struct interpret_branch_name_options *options)
1574 {
1575 int len;
1576 struct branch *branch;
1577 struct strbuf err = STRBUF_INIT;
1578 const char *value;
1579
1580 len = get_mark(name + at, namelen - at);
1581 if (!len)
1582 return -1;
1583
1584 if (memchr(name, ':', at))
1585 return -1;
1586
1587 if (at) {
1588 char *name_str = xmemdupz(name, at);
1589 branch = branch_get(name_str);
1590 free(name_str);
1591 } else
1592 branch = branch_get(NULL);
1593
1594 value = get_data(branch, &err);
1595 if (!value) {
1596 if (options->nonfatal_dangling_mark) {
1597 strbuf_release(&err);
1598 return -1;
1599 } else {
1600 die("%s", err.buf);
1601 }
1602 }
1603
1604 if (!branch_interpret_allowed(value, options->allowed))
1605 return -1;
1606
1607 set_shortened_ref(r, buf, value);
1608 return len + at;
1609 }
1610
1611 int repo_interpret_branch_name(struct repository *r,
1612 const char *name, int namelen,
1613 struct strbuf *buf,
1614 const struct interpret_branch_name_options *options)
1615 {
1616 char *at;
1617 const char *start;
1618 int len;
1619
1620 if (!namelen)
1621 namelen = strlen(name);
1622
1623 if (!options->allowed || (options->allowed & INTERPRET_BRANCH_LOCAL)) {
1624 len = interpret_nth_prior_checkout(r, name, namelen, buf);
1625 if (!len) {
1626 return len; /* syntax Ok, not enough switches */
1627 } else if (len > 0) {
1628 if (len == namelen)
1629 return len; /* consumed all */
1630 else
1631 return reinterpret(r, name, namelen, len, buf,
1632 options->allowed);
1633 }
1634 }
1635
1636 for (start = name;
1637 (at = memchr(start, '@', namelen - (start - name)));
1638 start = at + 1) {
1639
1640 if (!options->allowed || (options->allowed & INTERPRET_BRANCH_HEAD)) {
1641 len = interpret_empty_at(name, namelen, at - name, buf);
1642 if (len > 0)
1643 return reinterpret(r, name, namelen, len, buf,
1644 options->allowed);
1645 }
1646
1647 len = interpret_branch_mark(r, name, namelen, at - name, buf,
1648 upstream_mark, branch_get_upstream,
1649 options);
1650 if (len > 0)
1651 return len;
1652
1653 len = interpret_branch_mark(r, name, namelen, at - name, buf,
1654 push_mark, branch_get_push,
1655 options);
1656 if (len > 0)
1657 return len;
1658 }
1659
1660 return -1;
1661 }
1662
1663 void strbuf_branchname(struct strbuf *sb, const char *name, unsigned allowed)
1664 {
1665 int len = strlen(name);
1666 struct interpret_branch_name_options options = {
1667 .allowed = allowed
1668 };
1669 int used = interpret_branch_name(name, len, sb, &options);
1670
1671 if (used < 0)
1672 used = 0;
1673 strbuf_add(sb, name + used, len - used);
1674 }
1675
1676 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1677 {
1678 if (startup_info->have_repository)
1679 strbuf_branchname(sb, name, INTERPRET_BRANCH_LOCAL);
1680 else
1681 strbuf_addstr(sb, name);
1682
1683 /*
1684 * This splice must be done even if we end up rejecting the
1685 * name; builtin/branch.c::copy_or_rename_branch() still wants
1686 * to see what the name expanded to so that "branch -m" can be
1687 * used as a tool to correct earlier mistakes.
1688 */
1689 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1690
1691 if (*name == '-' ||
1692 !strcmp(sb->buf, "refs/heads/HEAD"))
1693 return -1;
1694
1695 return check_refname_format(sb->buf, 0);
1696 }
1697
1698 /*
1699 * This is like "get_oid_basic()", except it allows "object ID expressions",
1700 * notably "xyz^" for "parent of xyz"
1701 */
1702 int repo_get_oid(struct repository *r, const char *name, struct object_id *oid)
1703 {
1704 struct object_context unused;
1705 return get_oid_with_context(r, name, 0, oid, &unused);
1706 }
1707
1708 /*
1709 * This returns a non-zero value if the string (built using printf
1710 * format and the given arguments) is not a valid object.
1711 */
1712 int get_oidf(struct object_id *oid, const char *fmt, ...)
1713 {
1714 va_list ap;
1715 int ret;
1716 struct strbuf sb = STRBUF_INIT;
1717
1718 va_start(ap, fmt);
1719 strbuf_vaddf(&sb, fmt, ap);
1720 va_end(ap);
1721
1722 ret = get_oid(sb.buf, oid);
1723 strbuf_release(&sb);
1724
1725 return ret;
1726 }
1727
1728 /*
1729 * Many callers know that the user meant to name a commit-ish by
1730 * syntactical positions where the object name appears. Calling this
1731 * function allows the machinery to disambiguate shorter-than-unique
1732 * abbreviated object names between commit-ish and others.
1733 *
1734 * Note that this does NOT error out when the named object is not a
1735 * commit-ish. It is merely to give a hint to the disambiguation
1736 * machinery.
1737 */
1738 int repo_get_oid_committish(struct repository *r,
1739 const char *name,
1740 struct object_id *oid)
1741 {
1742 struct object_context unused;
1743 return get_oid_with_context(r, name, GET_OID_COMMITTISH,
1744 oid, &unused);
1745 }
1746
1747 int repo_get_oid_treeish(struct repository *r,
1748 const char *name,
1749 struct object_id *oid)
1750 {
1751 struct object_context unused;
1752 return get_oid_with_context(r, name, GET_OID_TREEISH,
1753 oid, &unused);
1754 }
1755
1756 int repo_get_oid_commit(struct repository *r,
1757 const char *name,
1758 struct object_id *oid)
1759 {
1760 struct object_context unused;
1761 return get_oid_with_context(r, name, GET_OID_COMMIT,
1762 oid, &unused);
1763 }
1764
1765 int repo_get_oid_tree(struct repository *r,
1766 const char *name,
1767 struct object_id *oid)
1768 {
1769 struct object_context unused;
1770 return get_oid_with_context(r, name, GET_OID_TREE,
1771 oid, &unused);
1772 }
1773
1774 int repo_get_oid_blob(struct repository *r,
1775 const char *name,
1776 struct object_id *oid)
1777 {
1778 struct object_context unused;
1779 return get_oid_with_context(r, name, GET_OID_BLOB,
1780 oid, &unused);
1781 }
1782
1783 /* Must be called only when object_name:filename doesn't exist. */
1784 static void diagnose_invalid_oid_path(struct repository *r,
1785 const char *prefix,
1786 const char *filename,
1787 const struct object_id *tree_oid,
1788 const char *object_name,
1789 int object_name_len)
1790 {
1791 struct object_id oid;
1792 unsigned short mode;
1793
1794 if (!prefix)
1795 prefix = "";
1796
1797 if (file_exists(filename))
1798 die(_("path '%s' exists on disk, but not in '%.*s'"),
1799 filename, object_name_len, object_name);
1800 if (is_missing_file_error(errno)) {
1801 char *fullname = xstrfmt("%s%s", prefix, filename);
1802
1803 if (!get_tree_entry(r, tree_oid, fullname, &oid, &mode)) {
1804 die(_("path '%s' exists, but not '%s'\n"
1805 "hint: Did you mean '%.*s:%s' aka '%.*s:./%s'?"),
1806 fullname,
1807 filename,
1808 object_name_len, object_name,
1809 fullname,
1810 object_name_len, object_name,
1811 filename);
1812 }
1813 die(_("path '%s' does not exist in '%.*s'"),
1814 filename, object_name_len, object_name);
1815 }
1816 }
1817
1818 /* Must be called only when :stage:filename doesn't exist. */
1819 static void diagnose_invalid_index_path(struct repository *r,
1820 int stage,
1821 const char *prefix,
1822 const char *filename)
1823 {
1824 struct index_state *istate = r->index;
1825 const struct cache_entry *ce;
1826 int pos;
1827 unsigned namelen = strlen(filename);
1828 struct strbuf fullname = STRBUF_INIT;
1829
1830 if (!prefix)
1831 prefix = "";
1832
1833 /* Wrong stage number? */
1834 pos = index_name_pos(istate, filename, namelen);
1835 if (pos < 0)
1836 pos = -pos - 1;
1837 if (pos < istate->cache_nr) {
1838 ce = istate->cache[pos];
1839 if (!S_ISSPARSEDIR(ce->ce_mode) &&
1840 ce_namelen(ce) == namelen &&
1841 !memcmp(ce->name, filename, namelen))
1842 die(_("path '%s' is in the index, but not at stage %d\n"
1843 "hint: Did you mean ':%d:%s'?"),
1844 filename, stage,
1845 ce_stage(ce), filename);
1846 }
1847
1848 /* Confusion between relative and absolute filenames? */
1849 strbuf_addstr(&fullname, prefix);
1850 strbuf_addstr(&fullname, filename);
1851 pos = index_name_pos(istate, fullname.buf, fullname.len);
1852 if (pos < 0)
1853 pos = -pos - 1;
1854 if (pos < istate->cache_nr) {
1855 ce = istate->cache[pos];
1856 if (!S_ISSPARSEDIR(ce->ce_mode) &&
1857 ce_namelen(ce) == fullname.len &&
1858 !memcmp(ce->name, fullname.buf, fullname.len))
1859 die(_("path '%s' is in the index, but not '%s'\n"
1860 "hint: Did you mean ':%d:%s' aka ':%d:./%s'?"),
1861 fullname.buf, filename,
1862 ce_stage(ce), fullname.buf,
1863 ce_stage(ce), filename);
1864 }
1865
1866 if (repo_file_exists(r, filename))
1867 die(_("path '%s' exists on disk, but not in the index"), filename);
1868 if (is_missing_file_error(errno))
1869 die(_("path '%s' does not exist (neither on disk nor in the index)"),
1870 filename);
1871
1872 strbuf_release(&fullname);
1873 }
1874
1875
1876 static char *resolve_relative_path(struct repository *r, const char *rel)
1877 {
1878 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1879 return NULL;
1880
1881 if (r != the_repository || !is_inside_work_tree())
1882 die(_("relative path syntax can't be used outside working tree"));
1883
1884 /* die() inside prefix_path() if resolved path is outside worktree */
1885 return prefix_path(startup_info->prefix,
1886 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1887 rel);
1888 }
1889
1890 static int reject_tree_in_index(struct repository *repo,
1891 int only_to_die,
1892 const struct cache_entry *ce,
1893 int stage,
1894 const char *prefix,
1895 const char *cp)
1896 {
1897 if (!S_ISSPARSEDIR(ce->ce_mode))
1898 return 0;
1899 if (only_to_die)
1900 diagnose_invalid_index_path(repo, stage, prefix, cp);
1901 return -1;
1902 }
1903
1904 static enum get_oid_result get_oid_with_context_1(struct repository *repo,
1905 const char *name,
1906 unsigned flags,
1907 const char *prefix,
1908 struct object_id *oid,
1909 struct object_context *oc)
1910 {
1911 int ret, bracket_depth;
1912 int namelen = strlen(name);
1913 const char *cp;
1914 int only_to_die = flags & GET_OID_ONLY_TO_DIE;
1915
1916 memset(oc, 0, sizeof(*oc));
1917 oc->mode = S_IFINVALID;
1918 strbuf_init(&oc->symlink_path, 0);
1919 ret = get_oid_1(repo, name, namelen, oid, flags);
1920 if (!ret && flags & GET_OID_REQUIRE_PATH)
1921 die(_("<object>:<path> required, only <object> '%s' given"),
1922 name);
1923 if (!ret)
1924 return ret;
1925 /*
1926 * tree:path --> object name of path in tree
1927 * :path -> object name of absolute path in index
1928 * :./path -> object name of path relative to cwd in index
1929 * :[0-3]:path -> object name of path in index at stage
1930 * :/foo -> recent commit matching foo
1931 */
1932 if (name[0] == ':') {
1933 int stage = 0;
1934 const struct cache_entry *ce;
1935 char *new_path = NULL;
1936 int pos;
1937 if (!only_to_die && namelen > 2 && name[1] == '/') {
1938 struct handle_one_ref_cb cb;
1939 struct commit_list *list = NULL;
1940
1941 cb.repo = repo;
1942 cb.list = &list;
1943 refs_for_each_ref(get_main_ref_store(repo), handle_one_ref, &cb);
1944 refs_head_ref(get_main_ref_store(repo), handle_one_ref, &cb);
1945 commit_list_sort_by_date(&list);
1946 return get_oid_oneline(repo, name + 2, oid, list);
1947 }
1948 if (namelen < 3 ||
1949 name[2] != ':' ||
1950 name[1] < '0' || '3' < name[1])
1951 cp = name + 1;
1952 else {
1953 stage = name[1] - '0';
1954 cp = name + 3;
1955 }
1956 new_path = resolve_relative_path(repo, cp);
1957 if (!new_path) {
1958 namelen = namelen - (cp - name);
1959 } else {
1960 cp = new_path;
1961 namelen = strlen(cp);
1962 }
1963
1964 if (flags & GET_OID_RECORD_PATH)
1965 oc->path = xstrdup(cp);
1966
1967 if (!repo->index || !repo->index->cache)
1968 repo_read_index(repo);
1969 pos = index_name_pos(repo->index, cp, namelen);
1970 if (pos < 0)
1971 pos = -pos - 1;
1972 while (pos < repo->index->cache_nr) {
1973 ce = repo->index->cache[pos];
1974 if (ce_namelen(ce) != namelen ||
1975 memcmp(ce->name, cp, namelen))
1976 break;
1977 if (ce_stage(ce) == stage) {
1978 free(new_path);
1979 if (reject_tree_in_index(repo, only_to_die, ce,
1980 stage, prefix, cp))
1981 return -1;
1982 oidcpy(oid, &ce->oid);
1983 oc->mode = ce->ce_mode;
1984 return 0;
1985 }
1986 pos++;
1987 }
1988 if (only_to_die && name[1] && name[1] != '/')
1989 diagnose_invalid_index_path(repo, stage, prefix, cp);
1990 free(new_path);
1991 return -1;
1992 }
1993 for (cp = name, bracket_depth = 0; *cp; cp++) {
1994 if (*cp == '{')
1995 bracket_depth++;
1996 else if (bracket_depth && *cp == '}')
1997 bracket_depth--;
1998 else if (!bracket_depth && *cp == ':')
1999 break;
2000 }
2001 if (*cp == ':') {
2002 struct object_id tree_oid;
2003 int len = cp - name;
2004 unsigned sub_flags = flags;
2005
2006 sub_flags &= ~GET_OID_DISAMBIGUATORS;
2007 sub_flags |= GET_OID_TREEISH;
2008
2009 if (!get_oid_1(repo, name, len, &tree_oid, sub_flags)) {
2010 const char *filename = cp+1;
2011 char *new_filename = NULL;
2012
2013 new_filename = resolve_relative_path(repo, filename);
2014 if (new_filename)
2015 filename = new_filename;
2016 if (flags & GET_OID_FOLLOW_SYMLINKS) {
2017 ret = get_tree_entry_follow_symlinks(repo, &tree_oid,
2018 filename, oid, &oc->symlink_path,
2019 &oc->mode);
2020 } else {
2021 ret = get_tree_entry(repo, &tree_oid, filename, oid,
2022 &oc->mode);
2023 if (ret && only_to_die) {
2024 diagnose_invalid_oid_path(repo, prefix,
2025 filename,
2026 &tree_oid,
2027 name, len);
2028 }
2029 }
2030 if (flags & GET_OID_RECORD_PATH)
2031 oc->path = xstrdup(filename);
2032
2033 free(new_filename);
2034 return ret;
2035 } else {
2036 if (only_to_die)
2037 die(_("invalid object name '%.*s'."), len, name);
2038 }
2039 }
2040 return ret;
2041 }
2042
2043 /*
2044 * Call this function when you know "name" given by the end user must
2045 * name an object but it doesn't; the function _may_ die with a better
2046 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
2047 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
2048 * you have a chance to diagnose the error further.
2049 */
2050 void maybe_die_on_misspelt_object_name(struct repository *r,
2051 const char *name,
2052 const char *prefix)
2053 {
2054 struct object_context oc;
2055 struct object_id oid;
2056 get_oid_with_context_1(r, name, GET_OID_ONLY_TO_DIE | GET_OID_QUIETLY,
2057 prefix, &oid, &oc);
2058 }
2059
2060 enum get_oid_result get_oid_with_context(struct repository *repo,
2061 const char *str,
2062 unsigned flags,
2063 struct object_id *oid,
2064 struct object_context *oc)
2065 {
2066 if (flags & GET_OID_FOLLOW_SYMLINKS && flags & GET_OID_ONLY_TO_DIE)
2067 BUG("incompatible flags for get_oid_with_context");
2068 return get_oid_with_context_1(repo, str, flags, NULL, oid, oc);
2069 }