]> git.ipfire.org Git - thirdparty/git.git/blob - object-name.c
Merge branch 'add-uk-initial-l10n' of github.com:arkid15r/git-ukrainian-l10n
[thirdparty/git.git] / object-name.c
1 #include "cache.h"
2 #include "object-name.h"
3 #include "advice.h"
4 #include "config.h"
5 #include "environment.h"
6 #include "gettext.h"
7 #include "hex.h"
8 #include "tag.h"
9 #include "commit.h"
10 #include "tree.h"
11 #include "blob.h"
12 #include "tree-walk.h"
13 #include "refs.h"
14 #include "remote.h"
15 #include "dir.h"
16 #include "oid-array.h"
17 #include "oidtree.h"
18 #include "packfile.h"
19 #include "pretty.h"
20 #include "object-store.h"
21 #include "repository.h"
22 #include "setup.h"
23 #include "submodule.h"
24 #include "midx.h"
25 #include "commit-reach.h"
26 #include "date.h"
27
28 static int get_oid_oneline(struct repository *r, const char *, struct object_id *, struct commit_list *);
29
30 typedef int (*disambiguate_hint_fn)(struct repository *, const struct object_id *, void *);
31
32 struct disambiguate_state {
33 int len; /* length of prefix in hex chars */
34 char hex_pfx[GIT_MAX_HEXSZ + 1];
35 struct object_id bin_pfx;
36
37 struct repository *repo;
38 disambiguate_hint_fn fn;
39 void *cb_data;
40 struct object_id candidate;
41 unsigned candidate_exists:1;
42 unsigned candidate_checked:1;
43 unsigned candidate_ok:1;
44 unsigned disambiguate_fn_used:1;
45 unsigned ambiguous:1;
46 unsigned always_call_fn:1;
47 };
48
49 static void update_candidates(struct disambiguate_state *ds, const struct object_id *current)
50 {
51 if (ds->always_call_fn) {
52 ds->ambiguous = ds->fn(ds->repo, current, ds->cb_data) ? 1 : 0;
53 return;
54 }
55 if (!ds->candidate_exists) {
56 /* this is the first candidate */
57 oidcpy(&ds->candidate, current);
58 ds->candidate_exists = 1;
59 return;
60 } else if (oideq(&ds->candidate, current)) {
61 /* the same as what we already have seen */
62 return;
63 }
64
65 if (!ds->fn) {
66 /* cannot disambiguate between ds->candidate and current */
67 ds->ambiguous = 1;
68 return;
69 }
70
71 if (!ds->candidate_checked) {
72 ds->candidate_ok = ds->fn(ds->repo, &ds->candidate, ds->cb_data);
73 ds->disambiguate_fn_used = 1;
74 ds->candidate_checked = 1;
75 }
76
77 if (!ds->candidate_ok) {
78 /* discard the candidate; we know it does not satisfy fn */
79 oidcpy(&ds->candidate, current);
80 ds->candidate_checked = 0;
81 return;
82 }
83
84 /* if we reach this point, we know ds->candidate satisfies fn */
85 if (ds->fn(ds->repo, current, ds->cb_data)) {
86 /*
87 * if both current and candidate satisfy fn, we cannot
88 * disambiguate.
89 */
90 ds->candidate_ok = 0;
91 ds->ambiguous = 1;
92 }
93
94 /* otherwise, current can be discarded and candidate is still good */
95 }
96
97 static int match_hash(unsigned, const unsigned char *, const unsigned char *);
98
99 static enum cb_next match_prefix(const struct object_id *oid, void *arg)
100 {
101 struct disambiguate_state *ds = arg;
102 /* no need to call match_hash, oidtree_each did prefix match */
103 update_candidates(ds, oid);
104 return ds->ambiguous ? CB_BREAK : CB_CONTINUE;
105 }
106
107 static void find_short_object_filename(struct disambiguate_state *ds)
108 {
109 struct object_directory *odb;
110
111 for (odb = ds->repo->objects->odb; odb && !ds->ambiguous; odb = odb->next)
112 oidtree_each(odb_loose_cache(odb, &ds->bin_pfx),
113 &ds->bin_pfx, ds->len, match_prefix, ds);
114 }
115
116 static int match_hash(unsigned len, const unsigned char *a, const unsigned char *b)
117 {
118 do {
119 if (*a != *b)
120 return 0;
121 a++;
122 b++;
123 len -= 2;
124 } while (len > 1);
125 if (len)
126 if ((*a ^ *b) & 0xf0)
127 return 0;
128 return 1;
129 }
130
131 static void unique_in_midx(struct multi_pack_index *m,
132 struct disambiguate_state *ds)
133 {
134 uint32_t num, i, first = 0;
135 const struct object_id *current = NULL;
136 num = m->num_objects;
137
138 if (!num)
139 return;
140
141 bsearch_midx(&ds->bin_pfx, m, &first);
142
143 /*
144 * At this point, "first" is the location of the lowest object
145 * with an object name that could match "bin_pfx". See if we have
146 * 0, 1 or more objects that actually match(es).
147 */
148 for (i = first; i < num && !ds->ambiguous; i++) {
149 struct object_id oid;
150 current = nth_midxed_object_oid(&oid, m, i);
151 if (!match_hash(ds->len, ds->bin_pfx.hash, current->hash))
152 break;
153 update_candidates(ds, current);
154 }
155 }
156
157 static void unique_in_pack(struct packed_git *p,
158 struct disambiguate_state *ds)
159 {
160 uint32_t num, i, first = 0;
161
162 if (p->multi_pack_index)
163 return;
164
165 if (open_pack_index(p) || !p->num_objects)
166 return;
167
168 num = p->num_objects;
169 bsearch_pack(&ds->bin_pfx, p, &first);
170
171 /*
172 * At this point, "first" is the location of the lowest object
173 * with an object name that could match "bin_pfx". See if we have
174 * 0, 1 or more objects that actually match(es).
175 */
176 for (i = first; i < num && !ds->ambiguous; i++) {
177 struct object_id oid;
178 nth_packed_object_id(&oid, p, i);
179 if (!match_hash(ds->len, ds->bin_pfx.hash, oid.hash))
180 break;
181 update_candidates(ds, &oid);
182 }
183 }
184
185 static void find_short_packed_object(struct disambiguate_state *ds)
186 {
187 struct multi_pack_index *m;
188 struct packed_git *p;
189
190 for (m = get_multi_pack_index(ds->repo); m && !ds->ambiguous;
191 m = m->next)
192 unique_in_midx(m, ds);
193 for (p = get_packed_git(ds->repo); p && !ds->ambiguous;
194 p = p->next)
195 unique_in_pack(p, ds);
196 }
197
198 static int finish_object_disambiguation(struct disambiguate_state *ds,
199 struct object_id *oid)
200 {
201 if (ds->ambiguous)
202 return SHORT_NAME_AMBIGUOUS;
203
204 if (!ds->candidate_exists)
205 return MISSING_OBJECT;
206
207 if (!ds->candidate_checked)
208 /*
209 * If this is the only candidate, there is no point
210 * calling the disambiguation hint callback.
211 *
212 * On the other hand, if the current candidate
213 * replaced an earlier candidate that did _not_ pass
214 * the disambiguation hint callback, then we do have
215 * more than one objects that match the short name
216 * given, so we should make sure this one matches;
217 * otherwise, if we discovered this one and the one
218 * that we previously discarded in the reverse order,
219 * we would end up showing different results in the
220 * same repository!
221 */
222 ds->candidate_ok = (!ds->disambiguate_fn_used ||
223 ds->fn(ds->repo, &ds->candidate, ds->cb_data));
224
225 if (!ds->candidate_ok)
226 return SHORT_NAME_AMBIGUOUS;
227
228 oidcpy(oid, &ds->candidate);
229 return 0;
230 }
231
232 static int disambiguate_commit_only(struct repository *r,
233 const struct object_id *oid,
234 void *cb_data UNUSED)
235 {
236 int kind = oid_object_info(r, oid, NULL);
237 return kind == OBJ_COMMIT;
238 }
239
240 static int disambiguate_committish_only(struct repository *r,
241 const struct object_id *oid,
242 void *cb_data UNUSED)
243 {
244 struct object *obj;
245 int kind;
246
247 kind = oid_object_info(r, oid, NULL);
248 if (kind == OBJ_COMMIT)
249 return 1;
250 if (kind != OBJ_TAG)
251 return 0;
252
253 /* We need to do this the hard way... */
254 obj = deref_tag(r, parse_object(r, oid), NULL, 0);
255 if (obj && obj->type == OBJ_COMMIT)
256 return 1;
257 return 0;
258 }
259
260 static int disambiguate_tree_only(struct repository *r,
261 const struct object_id *oid,
262 void *cb_data UNUSED)
263 {
264 int kind = oid_object_info(r, oid, NULL);
265 return kind == OBJ_TREE;
266 }
267
268 static int disambiguate_treeish_only(struct repository *r,
269 const struct object_id *oid,
270 void *cb_data UNUSED)
271 {
272 struct object *obj;
273 int kind;
274
275 kind = oid_object_info(r, oid, NULL);
276 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
277 return 1;
278 if (kind != OBJ_TAG)
279 return 0;
280
281 /* We need to do this the hard way... */
282 obj = deref_tag(r, parse_object(r, oid), NULL, 0);
283 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
284 return 1;
285 return 0;
286 }
287
288 static int disambiguate_blob_only(struct repository *r,
289 const struct object_id *oid,
290 void *cb_data UNUSED)
291 {
292 int kind = oid_object_info(r, oid, NULL);
293 return kind == OBJ_BLOB;
294 }
295
296 static disambiguate_hint_fn default_disambiguate_hint;
297
298 int set_disambiguate_hint_config(const char *var, const char *value)
299 {
300 static const struct {
301 const char *name;
302 disambiguate_hint_fn fn;
303 } hints[] = {
304 { "none", NULL },
305 { "commit", disambiguate_commit_only },
306 { "committish", disambiguate_committish_only },
307 { "tree", disambiguate_tree_only },
308 { "treeish", disambiguate_treeish_only },
309 { "blob", disambiguate_blob_only }
310 };
311 int i;
312
313 if (!value)
314 return config_error_nonbool(var);
315
316 for (i = 0; i < ARRAY_SIZE(hints); i++) {
317 if (!strcasecmp(value, hints[i].name)) {
318 default_disambiguate_hint = hints[i].fn;
319 return 0;
320 }
321 }
322
323 return error("unknown hint type for '%s': %s", var, value);
324 }
325
326 static int init_object_disambiguation(struct repository *r,
327 const char *name, int len,
328 struct disambiguate_state *ds)
329 {
330 int i;
331
332 if (len < MINIMUM_ABBREV || len > the_hash_algo->hexsz)
333 return -1;
334
335 memset(ds, 0, sizeof(*ds));
336
337 for (i = 0; i < len ;i++) {
338 unsigned char c = name[i];
339 unsigned char val;
340 if (c >= '0' && c <= '9')
341 val = c - '0';
342 else if (c >= 'a' && c <= 'f')
343 val = c - 'a' + 10;
344 else if (c >= 'A' && c <='F') {
345 val = c - 'A' + 10;
346 c -= 'A' - 'a';
347 }
348 else
349 return -1;
350 ds->hex_pfx[i] = c;
351 if (!(i & 1))
352 val <<= 4;
353 ds->bin_pfx.hash[i >> 1] |= val;
354 }
355
356 ds->len = len;
357 ds->hex_pfx[len] = '\0';
358 ds->repo = r;
359 prepare_alt_odb(r);
360 return 0;
361 }
362
363 struct ambiguous_output {
364 const struct disambiguate_state *ds;
365 struct strbuf advice;
366 struct strbuf sb;
367 };
368
369 static int show_ambiguous_object(const struct object_id *oid, void *data)
370 {
371 struct ambiguous_output *state = data;
372 const struct disambiguate_state *ds = state->ds;
373 struct strbuf *advice = &state->advice;
374 struct strbuf *sb = &state->sb;
375 int type;
376 const char *hash;
377
378 if (ds->fn && !ds->fn(ds->repo, oid, ds->cb_data))
379 return 0;
380
381 hash = repo_find_unique_abbrev(ds->repo, oid, DEFAULT_ABBREV);
382 type = oid_object_info(ds->repo, oid, NULL);
383
384 if (type < 0) {
385 /*
386 * TRANSLATORS: This is a line of ambiguous object
387 * output shown when we cannot look up or parse the
388 * object in question. E.g. "deadbeef [bad object]".
389 */
390 strbuf_addf(sb, _("%s [bad object]"), hash);
391 goto out;
392 }
393
394 assert(type == OBJ_TREE || type == OBJ_COMMIT ||
395 type == OBJ_BLOB || type == OBJ_TAG);
396
397 if (type == OBJ_COMMIT) {
398 struct strbuf date = STRBUF_INIT;
399 struct strbuf msg = STRBUF_INIT;
400 struct commit *commit = lookup_commit(ds->repo, oid);
401
402 if (commit) {
403 struct pretty_print_context pp = {0};
404 pp.date_mode.type = DATE_SHORT;
405 repo_format_commit_message(the_repository, commit,
406 "%ad", &date, &pp);
407 repo_format_commit_message(the_repository, commit,
408 "%s", &msg, &pp);
409 }
410
411 /*
412 * TRANSLATORS: This is a line of ambiguous commit
413 * object output. E.g.:
414 *
415 * "deadbeef commit 2021-01-01 - Some Commit Message"
416 */
417 strbuf_addf(sb, _("%s commit %s - %s"), hash, date.buf,
418 msg.buf);
419
420 strbuf_release(&date);
421 strbuf_release(&msg);
422 } else if (type == OBJ_TAG) {
423 struct tag *tag = lookup_tag(ds->repo, oid);
424
425 if (!parse_tag(tag) && tag->tag) {
426 /*
427 * TRANSLATORS: This is a line of ambiguous
428 * tag object output. E.g.:
429 *
430 * "deadbeef tag 2022-01-01 - Some Tag Message"
431 *
432 * The second argument is the YYYY-MM-DD found
433 * in the tag.
434 *
435 * The third argument is the "tag" string
436 * from object.c.
437 */
438 strbuf_addf(sb, _("%s tag %s - %s"), hash,
439 show_date(tag->date, 0, DATE_MODE(SHORT)),
440 tag->tag);
441 } else {
442 /*
443 * TRANSLATORS: This is a line of ambiguous
444 * tag object output where we couldn't parse
445 * the tag itself. E.g.:
446 *
447 * "deadbeef [bad tag, could not parse it]"
448 */
449 strbuf_addf(sb, _("%s [bad tag, could not parse it]"),
450 hash);
451 }
452 } else if (type == OBJ_TREE) {
453 /*
454 * TRANSLATORS: This is a line of ambiguous <type>
455 * object output. E.g. "deadbeef tree".
456 */
457 strbuf_addf(sb, _("%s tree"), hash);
458 } else if (type == OBJ_BLOB) {
459 /*
460 * TRANSLATORS: This is a line of ambiguous <type>
461 * object output. E.g. "deadbeef blob".
462 */
463 strbuf_addf(sb, _("%s blob"), hash);
464 }
465
466
467 out:
468 /*
469 * TRANSLATORS: This is line item of ambiguous object output
470 * from describe_ambiguous_object() above. For RTL languages
471 * you'll probably want to swap the "%s" and leading " " space
472 * around.
473 */
474 strbuf_addf(advice, _(" %s\n"), sb->buf);
475
476 strbuf_reset(sb);
477 return 0;
478 }
479
480 static int collect_ambiguous(const struct object_id *oid, void *data)
481 {
482 oid_array_append(data, oid);
483 return 0;
484 }
485
486 static int repo_collect_ambiguous(struct repository *r UNUSED,
487 const struct object_id *oid,
488 void *data)
489 {
490 return collect_ambiguous(oid, data);
491 }
492
493 static int sort_ambiguous(const void *a, const void *b, void *ctx)
494 {
495 struct repository *sort_ambiguous_repo = ctx;
496 int a_type = oid_object_info(sort_ambiguous_repo, a, NULL);
497 int b_type = oid_object_info(sort_ambiguous_repo, b, NULL);
498 int a_type_sort;
499 int b_type_sort;
500
501 /*
502 * Sorts by hash within the same object type, just as
503 * oid_array_for_each_unique() would do.
504 */
505 if (a_type == b_type)
506 return oidcmp(a, b);
507
508 /*
509 * Between object types show tags, then commits, and finally
510 * trees and blobs.
511 *
512 * The object_type enum is commit, tree, blob, tag, but we
513 * want tag, commit, tree blob. Cleverly (perhaps too
514 * cleverly) do that with modulus, since the enum assigns 1 to
515 * commit, so tag becomes 0.
516 */
517 a_type_sort = a_type % 4;
518 b_type_sort = b_type % 4;
519 return a_type_sort > b_type_sort ? 1 : -1;
520 }
521
522 static void sort_ambiguous_oid_array(struct repository *r, struct oid_array *a)
523 {
524 QSORT_S(a->oid, a->nr, sort_ambiguous, r);
525 }
526
527 static enum get_oid_result get_short_oid(struct repository *r,
528 const char *name, int len,
529 struct object_id *oid,
530 unsigned flags)
531 {
532 int status;
533 struct disambiguate_state ds;
534 int quietly = !!(flags & GET_OID_QUIETLY);
535
536 if (init_object_disambiguation(r, name, len, &ds) < 0)
537 return -1;
538
539 if (HAS_MULTI_BITS(flags & GET_OID_DISAMBIGUATORS))
540 BUG("multiple get_short_oid disambiguator flags");
541
542 if (flags & GET_OID_COMMIT)
543 ds.fn = disambiguate_commit_only;
544 else if (flags & GET_OID_COMMITTISH)
545 ds.fn = disambiguate_committish_only;
546 else if (flags & GET_OID_TREE)
547 ds.fn = disambiguate_tree_only;
548 else if (flags & GET_OID_TREEISH)
549 ds.fn = disambiguate_treeish_only;
550 else if (flags & GET_OID_BLOB)
551 ds.fn = disambiguate_blob_only;
552 else
553 ds.fn = default_disambiguate_hint;
554
555 find_short_object_filename(&ds);
556 find_short_packed_object(&ds);
557 status = finish_object_disambiguation(&ds, oid);
558
559 /*
560 * If we didn't find it, do the usual reprepare() slow-path,
561 * since the object may have recently been added to the repository
562 * or migrated from loose to packed.
563 */
564 if (status == MISSING_OBJECT) {
565 reprepare_packed_git(r);
566 find_short_object_filename(&ds);
567 find_short_packed_object(&ds);
568 status = finish_object_disambiguation(&ds, oid);
569 }
570
571 if (!quietly && (status == SHORT_NAME_AMBIGUOUS)) {
572 struct oid_array collect = OID_ARRAY_INIT;
573 struct ambiguous_output out = {
574 .ds = &ds,
575 .sb = STRBUF_INIT,
576 .advice = STRBUF_INIT,
577 };
578
579 error(_("short object ID %s is ambiguous"), ds.hex_pfx);
580
581 /*
582 * We may still have ambiguity if we simply saw a series of
583 * candidates that did not satisfy our hint function. In
584 * that case, we still want to show them, so disable the hint
585 * function entirely.
586 */
587 if (!ds.ambiguous)
588 ds.fn = NULL;
589
590 repo_for_each_abbrev(r, ds.hex_pfx, collect_ambiguous, &collect);
591 sort_ambiguous_oid_array(r, &collect);
592
593 if (oid_array_for_each(&collect, show_ambiguous_object, &out))
594 BUG("show_ambiguous_object shouldn't return non-zero");
595
596 /*
597 * TRANSLATORS: The argument is the list of ambiguous
598 * objects composed in show_ambiguous_object(). See
599 * its "TRANSLATORS" comments for details.
600 */
601 advise(_("The candidates are:\n%s"), out.advice.buf);
602
603 oid_array_clear(&collect);
604 strbuf_release(&out.advice);
605 strbuf_release(&out.sb);
606 }
607
608 return status;
609 }
610
611 int repo_for_each_abbrev(struct repository *r, const char *prefix,
612 each_abbrev_fn fn, void *cb_data)
613 {
614 struct oid_array collect = OID_ARRAY_INIT;
615 struct disambiguate_state ds;
616 int ret;
617
618 if (init_object_disambiguation(r, prefix, strlen(prefix), &ds) < 0)
619 return -1;
620
621 ds.always_call_fn = 1;
622 ds.fn = repo_collect_ambiguous;
623 ds.cb_data = &collect;
624 find_short_object_filename(&ds);
625 find_short_packed_object(&ds);
626
627 ret = oid_array_for_each_unique(&collect, fn, cb_data);
628 oid_array_clear(&collect);
629 return ret;
630 }
631
632 /*
633 * Return the slot of the most-significant bit set in "val". There are various
634 * ways to do this quickly with fls() or __builtin_clzl(), but speed is
635 * probably not a big deal here.
636 */
637 static unsigned msb(unsigned long val)
638 {
639 unsigned r = 0;
640 while (val >>= 1)
641 r++;
642 return r;
643 }
644
645 struct min_abbrev_data {
646 unsigned int init_len;
647 unsigned int cur_len;
648 char *hex;
649 struct repository *repo;
650 const struct object_id *oid;
651 };
652
653 static inline char get_hex_char_from_oid(const struct object_id *oid,
654 unsigned int pos)
655 {
656 static const char hex[] = "0123456789abcdef";
657
658 if ((pos & 1) == 0)
659 return hex[oid->hash[pos >> 1] >> 4];
660 else
661 return hex[oid->hash[pos >> 1] & 0xf];
662 }
663
664 static int extend_abbrev_len(const struct object_id *oid, void *cb_data)
665 {
666 struct min_abbrev_data *mad = cb_data;
667
668 unsigned int i = mad->init_len;
669 while (mad->hex[i] && mad->hex[i] == get_hex_char_from_oid(oid, i))
670 i++;
671
672 if (i < GIT_MAX_RAWSZ && i >= mad->cur_len)
673 mad->cur_len = i + 1;
674
675 return 0;
676 }
677
678 static int repo_extend_abbrev_len(struct repository *r UNUSED,
679 const struct object_id *oid,
680 void *cb_data)
681 {
682 return extend_abbrev_len(oid, cb_data);
683 }
684
685 static void find_abbrev_len_for_midx(struct multi_pack_index *m,
686 struct min_abbrev_data *mad)
687 {
688 int match = 0;
689 uint32_t num, first = 0;
690 struct object_id oid;
691 const struct object_id *mad_oid;
692
693 if (!m->num_objects)
694 return;
695
696 num = m->num_objects;
697 mad_oid = mad->oid;
698 match = bsearch_midx(mad_oid, m, &first);
699
700 /*
701 * first is now the position in the packfile where we would insert
702 * mad->hash if it does not exist (or the position of mad->hash if
703 * it does exist). Hence, we consider a maximum of two objects
704 * nearby for the abbreviation length.
705 */
706 mad->init_len = 0;
707 if (!match) {
708 if (nth_midxed_object_oid(&oid, m, first))
709 extend_abbrev_len(&oid, mad);
710 } else if (first < num - 1) {
711 if (nth_midxed_object_oid(&oid, m, first + 1))
712 extend_abbrev_len(&oid, mad);
713 }
714 if (first > 0) {
715 if (nth_midxed_object_oid(&oid, m, first - 1))
716 extend_abbrev_len(&oid, mad);
717 }
718 mad->init_len = mad->cur_len;
719 }
720
721 static void find_abbrev_len_for_pack(struct packed_git *p,
722 struct min_abbrev_data *mad)
723 {
724 int match = 0;
725 uint32_t num, first = 0;
726 struct object_id oid;
727 const struct object_id *mad_oid;
728
729 if (p->multi_pack_index)
730 return;
731
732 if (open_pack_index(p) || !p->num_objects)
733 return;
734
735 num = p->num_objects;
736 mad_oid = mad->oid;
737 match = bsearch_pack(mad_oid, p, &first);
738
739 /*
740 * first is now the position in the packfile where we would insert
741 * mad->hash if it does not exist (or the position of mad->hash if
742 * it does exist). Hence, we consider a maximum of two objects
743 * nearby for the abbreviation length.
744 */
745 mad->init_len = 0;
746 if (!match) {
747 if (!nth_packed_object_id(&oid, p, first))
748 extend_abbrev_len(&oid, mad);
749 } else if (first < num - 1) {
750 if (!nth_packed_object_id(&oid, p, first + 1))
751 extend_abbrev_len(&oid, mad);
752 }
753 if (first > 0) {
754 if (!nth_packed_object_id(&oid, p, first - 1))
755 extend_abbrev_len(&oid, mad);
756 }
757 mad->init_len = mad->cur_len;
758 }
759
760 static void find_abbrev_len_packed(struct min_abbrev_data *mad)
761 {
762 struct multi_pack_index *m;
763 struct packed_git *p;
764
765 for (m = get_multi_pack_index(mad->repo); m; m = m->next)
766 find_abbrev_len_for_midx(m, mad);
767 for (p = get_packed_git(mad->repo); p; p = p->next)
768 find_abbrev_len_for_pack(p, mad);
769 }
770
771 int repo_find_unique_abbrev_r(struct repository *r, char *hex,
772 const struct object_id *oid, int len)
773 {
774 struct disambiguate_state ds;
775 struct min_abbrev_data mad;
776 struct object_id oid_ret;
777 const unsigned hexsz = r->hash_algo->hexsz;
778
779 if (len < 0) {
780 unsigned long count = repo_approximate_object_count(r);
781 /*
782 * Add one because the MSB only tells us the highest bit set,
783 * not including the value of all the _other_ bits (so "15"
784 * is only one off of 2^4, but the MSB is the 3rd bit.
785 */
786 len = msb(count) + 1;
787 /*
788 * We now know we have on the order of 2^len objects, which
789 * expects a collision at 2^(len/2). But we also care about hex
790 * chars, not bits, and there are 4 bits per hex. So all
791 * together we need to divide by 2 and round up.
792 */
793 len = DIV_ROUND_UP(len, 2);
794 /*
795 * For very small repos, we stick with our regular fallback.
796 */
797 if (len < FALLBACK_DEFAULT_ABBREV)
798 len = FALLBACK_DEFAULT_ABBREV;
799 }
800
801 oid_to_hex_r(hex, oid);
802 if (len == hexsz || !len)
803 return hexsz;
804
805 mad.repo = r;
806 mad.init_len = len;
807 mad.cur_len = len;
808 mad.hex = hex;
809 mad.oid = oid;
810
811 find_abbrev_len_packed(&mad);
812
813 if (init_object_disambiguation(r, hex, mad.cur_len, &ds) < 0)
814 return -1;
815
816 ds.fn = repo_extend_abbrev_len;
817 ds.always_call_fn = 1;
818 ds.cb_data = (void *)&mad;
819
820 find_short_object_filename(&ds);
821 (void)finish_object_disambiguation(&ds, &oid_ret);
822
823 hex[mad.cur_len] = 0;
824 return mad.cur_len;
825 }
826
827 const char *repo_find_unique_abbrev(struct repository *r,
828 const struct object_id *oid,
829 int len)
830 {
831 static int bufno;
832 static char hexbuffer[4][GIT_MAX_HEXSZ + 1];
833 char *hex = hexbuffer[bufno];
834 bufno = (bufno + 1) % ARRAY_SIZE(hexbuffer);
835 repo_find_unique_abbrev_r(r, hex, oid, len);
836 return hex;
837 }
838
839 static int ambiguous_path(const char *path, int len)
840 {
841 int slash = 1;
842 int cnt;
843
844 for (cnt = 0; cnt < len; cnt++) {
845 switch (*path++) {
846 case '\0':
847 break;
848 case '/':
849 if (slash)
850 break;
851 slash = 1;
852 continue;
853 case '.':
854 continue;
855 default:
856 slash = 0;
857 continue;
858 }
859 break;
860 }
861 return slash;
862 }
863
864 static inline int at_mark(const char *string, int len,
865 const char **suffix, int nr)
866 {
867 int i;
868
869 for (i = 0; i < nr; i++) {
870 int suffix_len = strlen(suffix[i]);
871 if (suffix_len <= len
872 && !strncasecmp(string, suffix[i], suffix_len))
873 return suffix_len;
874 }
875 return 0;
876 }
877
878 static inline int upstream_mark(const char *string, int len)
879 {
880 const char *suffix[] = { "@{upstream}", "@{u}" };
881 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
882 }
883
884 static inline int push_mark(const char *string, int len)
885 {
886 const char *suffix[] = { "@{push}" };
887 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
888 }
889
890 static enum get_oid_result get_oid_1(struct repository *r, const char *name, int len, struct object_id *oid, unsigned lookup_flags);
891 static int interpret_nth_prior_checkout(struct repository *r, const char *name, int namelen, struct strbuf *buf);
892
893 static int get_oid_basic(struct repository *r, const char *str, int len,
894 struct object_id *oid, unsigned int flags)
895 {
896 static const char *warn_msg = "refname '%.*s' is ambiguous.";
897 static const char *object_name_msg = N_(
898 "Git normally never creates a ref that ends with 40 hex characters\n"
899 "because it will be ignored when you just specify 40-hex. These refs\n"
900 "may be created by mistake. For example,\n"
901 "\n"
902 " git switch -c $br $(git rev-parse ...)\n"
903 "\n"
904 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
905 "examine these refs and maybe delete them. Turn this message off by\n"
906 "running \"git config advice.objectNameWarning false\"");
907 struct object_id tmp_oid;
908 char *real_ref = NULL;
909 int refs_found = 0;
910 int at, reflog_len, nth_prior = 0;
911 int fatal = !(flags & GET_OID_QUIETLY);
912
913 if (len == r->hash_algo->hexsz && !get_oid_hex(str, oid)) {
914 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
915 refs_found = repo_dwim_ref(r, str, len, &tmp_oid, &real_ref, 0);
916 if (refs_found > 0) {
917 warning(warn_msg, len, str);
918 if (advice_enabled(ADVICE_OBJECT_NAME_WARNING))
919 fprintf(stderr, "%s\n", _(object_name_msg));
920 }
921 free(real_ref);
922 }
923 return 0;
924 }
925
926 /* basic@{time or number or -number} format to query ref-log */
927 reflog_len = at = 0;
928 if (len && str[len-1] == '}') {
929 for (at = len-4; at >= 0; at--) {
930 if (str[at] == '@' && str[at+1] == '{') {
931 if (str[at+2] == '-') {
932 if (at != 0)
933 /* @{-N} not at start */
934 return -1;
935 nth_prior = 1;
936 continue;
937 }
938 if (!upstream_mark(str + at, len - at) &&
939 !push_mark(str + at, len - at)) {
940 reflog_len = (len-1) - (at+2);
941 len = at;
942 }
943 break;
944 }
945 }
946 }
947
948 /* Accept only unambiguous ref paths. */
949 if (len && ambiguous_path(str, len))
950 return -1;
951
952 if (nth_prior) {
953 struct strbuf buf = STRBUF_INIT;
954 int detached;
955
956 if (interpret_nth_prior_checkout(r, str, len, &buf) > 0) {
957 detached = (buf.len == r->hash_algo->hexsz && !get_oid_hex(buf.buf, oid));
958 strbuf_release(&buf);
959 if (detached)
960 return 0;
961 }
962 }
963
964 if (!len && reflog_len)
965 /* allow "@{...}" to mean the current branch reflog */
966 refs_found = repo_dwim_ref(r, "HEAD", 4, oid, &real_ref, !fatal);
967 else if (reflog_len)
968 refs_found = repo_dwim_log(r, str, len, oid, &real_ref);
969 else
970 refs_found = repo_dwim_ref(r, str, len, oid, &real_ref, !fatal);
971
972 if (!refs_found)
973 return -1;
974
975 if (warn_ambiguous_refs && !(flags & GET_OID_QUIETLY) &&
976 (refs_found > 1 ||
977 !get_short_oid(r, str, len, &tmp_oid, GET_OID_QUIETLY)))
978 warning(warn_msg, len, str);
979
980 if (reflog_len) {
981 int nth, i;
982 timestamp_t at_time;
983 timestamp_t co_time;
984 int co_tz, co_cnt;
985
986 /* Is it asking for N-th entry, or approxidate? */
987 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
988 char ch = str[at+2+i];
989 if ('0' <= ch && ch <= '9')
990 nth = nth * 10 + ch - '0';
991 else
992 nth = -1;
993 }
994 if (100000000 <= nth) {
995 at_time = nth;
996 nth = -1;
997 } else if (0 <= nth)
998 at_time = 0;
999 else {
1000 int errors = 0;
1001 char *tmp = xstrndup(str + at + 2, reflog_len);
1002 at_time = approxidate_careful(tmp, &errors);
1003 free(tmp);
1004 if (errors) {
1005 free(real_ref);
1006 return -1;
1007 }
1008 }
1009 if (read_ref_at(get_main_ref_store(r),
1010 real_ref, flags, at_time, nth, oid, NULL,
1011 &co_time, &co_tz, &co_cnt)) {
1012 if (!len) {
1013 if (!skip_prefix(real_ref, "refs/heads/", &str))
1014 str = "HEAD";
1015 len = strlen(str);
1016 }
1017 if (at_time) {
1018 if (!(flags & GET_OID_QUIETLY)) {
1019 warning(_("log for '%.*s' only goes back to %s"),
1020 len, str,
1021 show_date(co_time, co_tz, DATE_MODE(RFC2822)));
1022 }
1023 } else {
1024 if (flags & GET_OID_QUIETLY) {
1025 exit(128);
1026 }
1027 die(_("log for '%.*s' only has %d entries"),
1028 len, str, co_cnt);
1029 }
1030 }
1031 }
1032
1033 free(real_ref);
1034 return 0;
1035 }
1036
1037 static enum get_oid_result get_parent(struct repository *r,
1038 const char *name, int len,
1039 struct object_id *result, int idx)
1040 {
1041 struct object_id oid;
1042 enum get_oid_result ret = get_oid_1(r, name, len, &oid,
1043 GET_OID_COMMITTISH);
1044 struct commit *commit;
1045 struct commit_list *p;
1046
1047 if (ret)
1048 return ret;
1049 commit = lookup_commit_reference(r, &oid);
1050 if (repo_parse_commit(r, commit))
1051 return MISSING_OBJECT;
1052 if (!idx) {
1053 oidcpy(result, &commit->object.oid);
1054 return FOUND;
1055 }
1056 p = commit->parents;
1057 while (p) {
1058 if (!--idx) {
1059 oidcpy(result, &p->item->object.oid);
1060 return FOUND;
1061 }
1062 p = p->next;
1063 }
1064 return MISSING_OBJECT;
1065 }
1066
1067 static enum get_oid_result get_nth_ancestor(struct repository *r,
1068 const char *name, int len,
1069 struct object_id *result,
1070 int generation)
1071 {
1072 struct object_id oid;
1073 struct commit *commit;
1074 int ret;
1075
1076 ret = get_oid_1(r, name, len, &oid, GET_OID_COMMITTISH);
1077 if (ret)
1078 return ret;
1079 commit = lookup_commit_reference(r, &oid);
1080 if (!commit)
1081 return MISSING_OBJECT;
1082
1083 while (generation--) {
1084 if (repo_parse_commit(r, commit) || !commit->parents)
1085 return MISSING_OBJECT;
1086 commit = commit->parents->item;
1087 }
1088 oidcpy(result, &commit->object.oid);
1089 return FOUND;
1090 }
1091
1092 struct object *repo_peel_to_type(struct repository *r, const char *name, int namelen,
1093 struct object *o, enum object_type expected_type)
1094 {
1095 if (name && !namelen)
1096 namelen = strlen(name);
1097 while (1) {
1098 if (!o || (!o->parsed && !parse_object(r, &o->oid)))
1099 return NULL;
1100 if (expected_type == OBJ_ANY || o->type == expected_type)
1101 return o;
1102 if (o->type == OBJ_TAG)
1103 o = ((struct tag*) o)->tagged;
1104 else if (o->type == OBJ_COMMIT)
1105 o = &(repo_get_commit_tree(r, ((struct commit *)o))->object);
1106 else {
1107 if (name)
1108 error("%.*s: expected %s type, but the object "
1109 "dereferences to %s type",
1110 namelen, name, type_name(expected_type),
1111 type_name(o->type));
1112 return NULL;
1113 }
1114 }
1115 }
1116
1117 static int peel_onion(struct repository *r, const char *name, int len,
1118 struct object_id *oid, unsigned lookup_flags)
1119 {
1120 struct object_id outer;
1121 const char *sp;
1122 unsigned int expected_type = 0;
1123 struct object *o;
1124
1125 /*
1126 * "ref^{type}" dereferences ref repeatedly until you cannot
1127 * dereference anymore, or you get an object of given type,
1128 * whichever comes first. "ref^{}" means just dereference
1129 * tags until you get a non-tag. "ref^0" is a shorthand for
1130 * "ref^{commit}". "commit^{tree}" could be used to find the
1131 * top-level tree of the given commit.
1132 */
1133 if (len < 4 || name[len-1] != '}')
1134 return -1;
1135
1136 for (sp = name + len - 1; name <= sp; sp--) {
1137 int ch = *sp;
1138 if (ch == '{' && name < sp && sp[-1] == '^')
1139 break;
1140 }
1141 if (sp <= name)
1142 return -1;
1143
1144 sp++; /* beginning of type name, or closing brace for empty */
1145 if (starts_with(sp, "commit}"))
1146 expected_type = OBJ_COMMIT;
1147 else if (starts_with(sp, "tag}"))
1148 expected_type = OBJ_TAG;
1149 else if (starts_with(sp, "tree}"))
1150 expected_type = OBJ_TREE;
1151 else if (starts_with(sp, "blob}"))
1152 expected_type = OBJ_BLOB;
1153 else if (starts_with(sp, "object}"))
1154 expected_type = OBJ_ANY;
1155 else if (sp[0] == '}')
1156 expected_type = OBJ_NONE;
1157 else if (sp[0] == '/')
1158 expected_type = OBJ_COMMIT;
1159 else
1160 return -1;
1161
1162 lookup_flags &= ~GET_OID_DISAMBIGUATORS;
1163 if (expected_type == OBJ_COMMIT)
1164 lookup_flags |= GET_OID_COMMITTISH;
1165 else if (expected_type == OBJ_TREE)
1166 lookup_flags |= GET_OID_TREEISH;
1167
1168 if (get_oid_1(r, name, sp - name - 2, &outer, lookup_flags))
1169 return -1;
1170
1171 o = parse_object(r, &outer);
1172 if (!o)
1173 return -1;
1174 if (!expected_type) {
1175 o = deref_tag(r, o, name, sp - name - 2);
1176 if (!o || (!o->parsed && !parse_object(r, &o->oid)))
1177 return -1;
1178 oidcpy(oid, &o->oid);
1179 return 0;
1180 }
1181
1182 /*
1183 * At this point, the syntax look correct, so
1184 * if we do not get the needed object, we should
1185 * barf.
1186 */
1187 o = repo_peel_to_type(r, name, len, o, expected_type);
1188 if (!o)
1189 return -1;
1190
1191 oidcpy(oid, &o->oid);
1192 if (sp[0] == '/') {
1193 /* "$commit^{/foo}" */
1194 char *prefix;
1195 int ret;
1196 struct commit_list *list = NULL;
1197
1198 /*
1199 * $commit^{/}. Some regex implementation may reject.
1200 * We don't need regex anyway. '' pattern always matches.
1201 */
1202 if (sp[1] == '}')
1203 return 0;
1204
1205 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
1206 commit_list_insert((struct commit *)o, &list);
1207 ret = get_oid_oneline(r, prefix, oid, list);
1208 free(prefix);
1209 return ret;
1210 }
1211 return 0;
1212 }
1213
1214 static int get_describe_name(struct repository *r,
1215 const char *name, int len,
1216 struct object_id *oid)
1217 {
1218 const char *cp;
1219 unsigned flags = GET_OID_QUIETLY | GET_OID_COMMIT;
1220
1221 for (cp = name + len - 1; name + 2 <= cp; cp--) {
1222 char ch = *cp;
1223 if (!isxdigit(ch)) {
1224 /* We must be looking at g in "SOMETHING-g"
1225 * for it to be describe output.
1226 */
1227 if (ch == 'g' && cp[-1] == '-') {
1228 cp++;
1229 len -= cp - name;
1230 return get_short_oid(r,
1231 cp, len, oid, flags);
1232 }
1233 }
1234 }
1235 return -1;
1236 }
1237
1238 static enum get_oid_result get_oid_1(struct repository *r,
1239 const char *name, int len,
1240 struct object_id *oid,
1241 unsigned lookup_flags)
1242 {
1243 int ret, has_suffix;
1244 const char *cp;
1245
1246 /*
1247 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
1248 */
1249 has_suffix = 0;
1250 for (cp = name + len - 1; name <= cp; cp--) {
1251 int ch = *cp;
1252 if ('0' <= ch && ch <= '9')
1253 continue;
1254 if (ch == '~' || ch == '^')
1255 has_suffix = ch;
1256 break;
1257 }
1258
1259 if (has_suffix) {
1260 unsigned int num = 0;
1261 int len1 = cp - name;
1262 cp++;
1263 while (cp < name + len) {
1264 unsigned int digit = *cp++ - '0';
1265 if (unsigned_mult_overflows(num, 10))
1266 return MISSING_OBJECT;
1267 num *= 10;
1268 if (unsigned_add_overflows(num, digit))
1269 return MISSING_OBJECT;
1270 num += digit;
1271 }
1272 if (!num && len1 == len - 1)
1273 num = 1;
1274 else if (num > INT_MAX)
1275 return MISSING_OBJECT;
1276 if (has_suffix == '^')
1277 return get_parent(r, name, len1, oid, num);
1278 /* else if (has_suffix == '~') -- goes without saying */
1279 return get_nth_ancestor(r, name, len1, oid, num);
1280 }
1281
1282 ret = peel_onion(r, name, len, oid, lookup_flags);
1283 if (!ret)
1284 return FOUND;
1285
1286 ret = get_oid_basic(r, name, len, oid, lookup_flags);
1287 if (!ret)
1288 return FOUND;
1289
1290 /* It could be describe output that is "SOMETHING-gXXXX" */
1291 ret = get_describe_name(r, name, len, oid);
1292 if (!ret)
1293 return FOUND;
1294
1295 return get_short_oid(r, name, len, oid, lookup_flags);
1296 }
1297
1298 /*
1299 * This interprets names like ':/Initial revision of "git"' by searching
1300 * through history and returning the first commit whose message starts
1301 * the given regular expression.
1302 *
1303 * For negative-matching, prefix the pattern-part with '!-', like: ':/!-WIP'.
1304 *
1305 * For a literal '!' character at the beginning of a pattern, you have to repeat
1306 * that, like: ':/!!foo'
1307 *
1308 * For future extension, all other sequences beginning with ':/!' are reserved.
1309 */
1310
1311 /* Remember to update object flag allocation in object.h */
1312 #define ONELINE_SEEN (1u<<20)
1313
1314 struct handle_one_ref_cb {
1315 struct repository *repo;
1316 struct commit_list **list;
1317 };
1318
1319 static int handle_one_ref(const char *path, const struct object_id *oid,
1320 int flag UNUSED,
1321 void *cb_data)
1322 {
1323 struct handle_one_ref_cb *cb = cb_data;
1324 struct commit_list **list = cb->list;
1325 struct object *object = parse_object(cb->repo, oid);
1326 if (!object)
1327 return 0;
1328 if (object->type == OBJ_TAG) {
1329 object = deref_tag(cb->repo, object, path,
1330 strlen(path));
1331 if (!object)
1332 return 0;
1333 }
1334 if (object->type != OBJ_COMMIT)
1335 return 0;
1336 commit_list_insert((struct commit *)object, list);
1337 return 0;
1338 }
1339
1340 static int get_oid_oneline(struct repository *r,
1341 const char *prefix, struct object_id *oid,
1342 struct commit_list *list)
1343 {
1344 struct commit_list *backup = NULL, *l;
1345 int found = 0;
1346 int negative = 0;
1347 regex_t regex;
1348
1349 if (prefix[0] == '!') {
1350 prefix++;
1351
1352 if (prefix[0] == '-') {
1353 prefix++;
1354 negative = 1;
1355 } else if (prefix[0] != '!') {
1356 return -1;
1357 }
1358 }
1359
1360 if (regcomp(&regex, prefix, REG_EXTENDED))
1361 return -1;
1362
1363 for (l = list; l; l = l->next) {
1364 l->item->object.flags |= ONELINE_SEEN;
1365 commit_list_insert(l->item, &backup);
1366 }
1367 while (list) {
1368 const char *p, *buf;
1369 struct commit *commit;
1370 int matches;
1371
1372 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
1373 if (!parse_object(r, &commit->object.oid))
1374 continue;
1375 buf = repo_get_commit_buffer(r, commit, NULL);
1376 p = strstr(buf, "\n\n");
1377 matches = negative ^ (p && !regexec(&regex, p + 2, 0, NULL, 0));
1378 repo_unuse_commit_buffer(r, commit, buf);
1379
1380 if (matches) {
1381 oidcpy(oid, &commit->object.oid);
1382 found = 1;
1383 break;
1384 }
1385 }
1386 regfree(&regex);
1387 free_commit_list(list);
1388 for (l = backup; l; l = l->next)
1389 clear_commit_marks(l->item, ONELINE_SEEN);
1390 free_commit_list(backup);
1391 return found ? 0 : -1;
1392 }
1393
1394 struct grab_nth_branch_switch_cbdata {
1395 int remaining;
1396 struct strbuf *sb;
1397 };
1398
1399 static int grab_nth_branch_switch(struct object_id *ooid UNUSED,
1400 struct object_id *noid UNUSED,
1401 const char *email UNUSED,
1402 timestamp_t timestamp UNUSED,
1403 int tz UNUSED,
1404 const char *message, void *cb_data)
1405 {
1406 struct grab_nth_branch_switch_cbdata *cb = cb_data;
1407 const char *match = NULL, *target = NULL;
1408 size_t len;
1409
1410 if (skip_prefix(message, "checkout: moving from ", &match))
1411 target = strstr(match, " to ");
1412
1413 if (!match || !target)
1414 return 0;
1415 if (--(cb->remaining) == 0) {
1416 len = target - match;
1417 strbuf_reset(cb->sb);
1418 strbuf_add(cb->sb, match, len);
1419 return 1; /* we are done */
1420 }
1421 return 0;
1422 }
1423
1424 /*
1425 * Parse @{-N} syntax, return the number of characters parsed
1426 * if successful; otherwise signal an error with negative value.
1427 */
1428 static int interpret_nth_prior_checkout(struct repository *r,
1429 const char *name, int namelen,
1430 struct strbuf *buf)
1431 {
1432 long nth;
1433 int retval;
1434 struct grab_nth_branch_switch_cbdata cb;
1435 const char *brace;
1436 char *num_end;
1437
1438 if (namelen < 4)
1439 return -1;
1440 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
1441 return -1;
1442 brace = memchr(name, '}', namelen);
1443 if (!brace)
1444 return -1;
1445 nth = strtol(name + 3, &num_end, 10);
1446 if (num_end != brace)
1447 return -1;
1448 if (nth <= 0)
1449 return -1;
1450 cb.remaining = nth;
1451 cb.sb = buf;
1452
1453 retval = refs_for_each_reflog_ent_reverse(get_main_ref_store(r),
1454 "HEAD", grab_nth_branch_switch, &cb);
1455 if (0 < retval) {
1456 retval = brace - name + 1;
1457 } else
1458 retval = 0;
1459
1460 return retval;
1461 }
1462
1463 int repo_get_oid_mb(struct repository *r,
1464 const char *name,
1465 struct object_id *oid)
1466 {
1467 struct commit *one, *two;
1468 struct commit_list *mbs;
1469 struct object_id oid_tmp;
1470 const char *dots;
1471 int st;
1472
1473 dots = strstr(name, "...");
1474 if (!dots)
1475 return repo_get_oid(r, name, oid);
1476 if (dots == name)
1477 st = repo_get_oid(r, "HEAD", &oid_tmp);
1478 else {
1479 struct strbuf sb;
1480 strbuf_init(&sb, dots - name);
1481 strbuf_add(&sb, name, dots - name);
1482 st = repo_get_oid_committish(r, sb.buf, &oid_tmp);
1483 strbuf_release(&sb);
1484 }
1485 if (st)
1486 return st;
1487 one = lookup_commit_reference_gently(r, &oid_tmp, 0);
1488 if (!one)
1489 return -1;
1490
1491 if (repo_get_oid_committish(r, dots[3] ? (dots + 3) : "HEAD", &oid_tmp))
1492 return -1;
1493 two = lookup_commit_reference_gently(r, &oid_tmp, 0);
1494 if (!two)
1495 return -1;
1496 mbs = repo_get_merge_bases(r, one, two);
1497 if (!mbs || mbs->next)
1498 st = -1;
1499 else {
1500 st = 0;
1501 oidcpy(oid, &mbs->item->object.oid);
1502 }
1503 free_commit_list(mbs);
1504 return st;
1505 }
1506
1507 /* parse @something syntax, when 'something' is not {.*} */
1508 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1509 {
1510 const char *next;
1511
1512 if (len || name[1] == '{')
1513 return -1;
1514
1515 /* make sure it's a single @, or @@{.*}, not @foo */
1516 next = memchr(name + len + 1, '@', namelen - len - 1);
1517 if (next && next[1] != '{')
1518 return -1;
1519 if (!next)
1520 next = name + namelen;
1521 if (next != name + 1)
1522 return -1;
1523
1524 strbuf_reset(buf);
1525 strbuf_add(buf, "HEAD", 4);
1526 return 1;
1527 }
1528
1529 static int reinterpret(struct repository *r,
1530 const char *name, int namelen, int len,
1531 struct strbuf *buf, unsigned allowed)
1532 {
1533 /* we have extra data, which might need further processing */
1534 struct strbuf tmp = STRBUF_INIT;
1535 int used = buf->len;
1536 int ret;
1537 struct interpret_branch_name_options options = {
1538 .allowed = allowed
1539 };
1540
1541 strbuf_add(buf, name + len, namelen - len);
1542 ret = repo_interpret_branch_name(r, buf->buf, buf->len, &tmp, &options);
1543 /* that data was not interpreted, remove our cruft */
1544 if (ret < 0) {
1545 strbuf_setlen(buf, used);
1546 return len;
1547 }
1548 strbuf_reset(buf);
1549 strbuf_addbuf(buf, &tmp);
1550 strbuf_release(&tmp);
1551 /* tweak for size of {-N} versus expanded ref name */
1552 return ret - used + len;
1553 }
1554
1555 static void set_shortened_ref(struct repository *r, struct strbuf *buf, const char *ref)
1556 {
1557 char *s = refs_shorten_unambiguous_ref(get_main_ref_store(r), ref, 0);
1558 strbuf_reset(buf);
1559 strbuf_addstr(buf, s);
1560 free(s);
1561 }
1562
1563 static int branch_interpret_allowed(const char *refname, unsigned allowed)
1564 {
1565 if (!allowed)
1566 return 1;
1567
1568 if ((allowed & INTERPRET_BRANCH_LOCAL) &&
1569 starts_with(refname, "refs/heads/"))
1570 return 1;
1571 if ((allowed & INTERPRET_BRANCH_REMOTE) &&
1572 starts_with(refname, "refs/remotes/"))
1573 return 1;
1574
1575 return 0;
1576 }
1577
1578 static int interpret_branch_mark(struct repository *r,
1579 const char *name, int namelen,
1580 int at, struct strbuf *buf,
1581 int (*get_mark)(const char *, int),
1582 const char *(*get_data)(struct branch *,
1583 struct strbuf *),
1584 const struct interpret_branch_name_options *options)
1585 {
1586 int len;
1587 struct branch *branch;
1588 struct strbuf err = STRBUF_INIT;
1589 const char *value;
1590
1591 len = get_mark(name + at, namelen - at);
1592 if (!len)
1593 return -1;
1594
1595 if (memchr(name, ':', at))
1596 return -1;
1597
1598 if (at) {
1599 char *name_str = xmemdupz(name, at);
1600 branch = branch_get(name_str);
1601 free(name_str);
1602 } else
1603 branch = branch_get(NULL);
1604
1605 value = get_data(branch, &err);
1606 if (!value) {
1607 if (options->nonfatal_dangling_mark) {
1608 strbuf_release(&err);
1609 return -1;
1610 } else {
1611 die("%s", err.buf);
1612 }
1613 }
1614
1615 if (!branch_interpret_allowed(value, options->allowed))
1616 return -1;
1617
1618 set_shortened_ref(r, buf, value);
1619 return len + at;
1620 }
1621
1622 int repo_interpret_branch_name(struct repository *r,
1623 const char *name, int namelen,
1624 struct strbuf *buf,
1625 const struct interpret_branch_name_options *options)
1626 {
1627 char *at;
1628 const char *start;
1629 int len;
1630
1631 if (!namelen)
1632 namelen = strlen(name);
1633
1634 if (!options->allowed || (options->allowed & INTERPRET_BRANCH_LOCAL)) {
1635 len = interpret_nth_prior_checkout(r, name, namelen, buf);
1636 if (!len) {
1637 return len; /* syntax Ok, not enough switches */
1638 } else if (len > 0) {
1639 if (len == namelen)
1640 return len; /* consumed all */
1641 else
1642 return reinterpret(r, name, namelen, len, buf,
1643 options->allowed);
1644 }
1645 }
1646
1647 for (start = name;
1648 (at = memchr(start, '@', namelen - (start - name)));
1649 start = at + 1) {
1650
1651 if (!options->allowed || (options->allowed & INTERPRET_BRANCH_HEAD)) {
1652 len = interpret_empty_at(name, namelen, at - name, buf);
1653 if (len > 0)
1654 return reinterpret(r, name, namelen, len, buf,
1655 options->allowed);
1656 }
1657
1658 len = interpret_branch_mark(r, name, namelen, at - name, buf,
1659 upstream_mark, branch_get_upstream,
1660 options);
1661 if (len > 0)
1662 return len;
1663
1664 len = interpret_branch_mark(r, name, namelen, at - name, buf,
1665 push_mark, branch_get_push,
1666 options);
1667 if (len > 0)
1668 return len;
1669 }
1670
1671 return -1;
1672 }
1673
1674 void strbuf_branchname(struct strbuf *sb, const char *name, unsigned allowed)
1675 {
1676 int len = strlen(name);
1677 struct interpret_branch_name_options options = {
1678 .allowed = allowed
1679 };
1680 int used = repo_interpret_branch_name(the_repository, name, len, sb,
1681 &options);
1682
1683 if (used < 0)
1684 used = 0;
1685 strbuf_add(sb, name + used, len - used);
1686 }
1687
1688 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1689 {
1690 if (startup_info->have_repository)
1691 strbuf_branchname(sb, name, INTERPRET_BRANCH_LOCAL);
1692 else
1693 strbuf_addstr(sb, name);
1694
1695 /*
1696 * This splice must be done even if we end up rejecting the
1697 * name; builtin/branch.c::copy_or_rename_branch() still wants
1698 * to see what the name expanded to so that "branch -m" can be
1699 * used as a tool to correct earlier mistakes.
1700 */
1701 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1702
1703 if (*name == '-' ||
1704 !strcmp(sb->buf, "refs/heads/HEAD"))
1705 return -1;
1706
1707 return check_refname_format(sb->buf, 0);
1708 }
1709
1710 /*
1711 * This is like "get_oid_basic()", except it allows "object ID expressions",
1712 * notably "xyz^" for "parent of xyz"
1713 */
1714 int repo_get_oid(struct repository *r, const char *name, struct object_id *oid)
1715 {
1716 struct object_context unused;
1717 return get_oid_with_context(r, name, 0, oid, &unused);
1718 }
1719
1720 /*
1721 * This returns a non-zero value if the string (built using printf
1722 * format and the given arguments) is not a valid object.
1723 */
1724 int get_oidf(struct object_id *oid, const char *fmt, ...)
1725 {
1726 va_list ap;
1727 int ret;
1728 struct strbuf sb = STRBUF_INIT;
1729
1730 va_start(ap, fmt);
1731 strbuf_vaddf(&sb, fmt, ap);
1732 va_end(ap);
1733
1734 ret = repo_get_oid(the_repository, sb.buf, oid);
1735 strbuf_release(&sb);
1736
1737 return ret;
1738 }
1739
1740 /*
1741 * Many callers know that the user meant to name a commit-ish by
1742 * syntactical positions where the object name appears. Calling this
1743 * function allows the machinery to disambiguate shorter-than-unique
1744 * abbreviated object names between commit-ish and others.
1745 *
1746 * Note that this does NOT error out when the named object is not a
1747 * commit-ish. It is merely to give a hint to the disambiguation
1748 * machinery.
1749 */
1750 int repo_get_oid_committish(struct repository *r,
1751 const char *name,
1752 struct object_id *oid)
1753 {
1754 struct object_context unused;
1755 return get_oid_with_context(r, name, GET_OID_COMMITTISH,
1756 oid, &unused);
1757 }
1758
1759 int repo_get_oid_treeish(struct repository *r,
1760 const char *name,
1761 struct object_id *oid)
1762 {
1763 struct object_context unused;
1764 return get_oid_with_context(r, name, GET_OID_TREEISH,
1765 oid, &unused);
1766 }
1767
1768 int repo_get_oid_commit(struct repository *r,
1769 const char *name,
1770 struct object_id *oid)
1771 {
1772 struct object_context unused;
1773 return get_oid_with_context(r, name, GET_OID_COMMIT,
1774 oid, &unused);
1775 }
1776
1777 int repo_get_oid_tree(struct repository *r,
1778 const char *name,
1779 struct object_id *oid)
1780 {
1781 struct object_context unused;
1782 return get_oid_with_context(r, name, GET_OID_TREE,
1783 oid, &unused);
1784 }
1785
1786 int repo_get_oid_blob(struct repository *r,
1787 const char *name,
1788 struct object_id *oid)
1789 {
1790 struct object_context unused;
1791 return get_oid_with_context(r, name, GET_OID_BLOB,
1792 oid, &unused);
1793 }
1794
1795 /* Must be called only when object_name:filename doesn't exist. */
1796 static void diagnose_invalid_oid_path(struct repository *r,
1797 const char *prefix,
1798 const char *filename,
1799 const struct object_id *tree_oid,
1800 const char *object_name,
1801 int object_name_len)
1802 {
1803 struct object_id oid;
1804 unsigned short mode;
1805
1806 if (!prefix)
1807 prefix = "";
1808
1809 if (file_exists(filename))
1810 die(_("path '%s' exists on disk, but not in '%.*s'"),
1811 filename, object_name_len, object_name);
1812 if (is_missing_file_error(errno)) {
1813 char *fullname = xstrfmt("%s%s", prefix, filename);
1814
1815 if (!get_tree_entry(r, tree_oid, fullname, &oid, &mode)) {
1816 die(_("path '%s' exists, but not '%s'\n"
1817 "hint: Did you mean '%.*s:%s' aka '%.*s:./%s'?"),
1818 fullname,
1819 filename,
1820 object_name_len, object_name,
1821 fullname,
1822 object_name_len, object_name,
1823 filename);
1824 }
1825 die(_("path '%s' does not exist in '%.*s'"),
1826 filename, object_name_len, object_name);
1827 }
1828 }
1829
1830 /* Must be called only when :stage:filename doesn't exist. */
1831 static void diagnose_invalid_index_path(struct repository *r,
1832 int stage,
1833 const char *prefix,
1834 const char *filename)
1835 {
1836 struct index_state *istate = r->index;
1837 const struct cache_entry *ce;
1838 int pos;
1839 unsigned namelen = strlen(filename);
1840 struct strbuf fullname = STRBUF_INIT;
1841
1842 if (!prefix)
1843 prefix = "";
1844
1845 /* Wrong stage number? */
1846 pos = index_name_pos(istate, filename, namelen);
1847 if (pos < 0)
1848 pos = -pos - 1;
1849 if (pos < istate->cache_nr) {
1850 ce = istate->cache[pos];
1851 if (!S_ISSPARSEDIR(ce->ce_mode) &&
1852 ce_namelen(ce) == namelen &&
1853 !memcmp(ce->name, filename, namelen))
1854 die(_("path '%s' is in the index, but not at stage %d\n"
1855 "hint: Did you mean ':%d:%s'?"),
1856 filename, stage,
1857 ce_stage(ce), filename);
1858 }
1859
1860 /* Confusion between relative and absolute filenames? */
1861 strbuf_addstr(&fullname, prefix);
1862 strbuf_addstr(&fullname, filename);
1863 pos = index_name_pos(istate, fullname.buf, fullname.len);
1864 if (pos < 0)
1865 pos = -pos - 1;
1866 if (pos < istate->cache_nr) {
1867 ce = istate->cache[pos];
1868 if (!S_ISSPARSEDIR(ce->ce_mode) &&
1869 ce_namelen(ce) == fullname.len &&
1870 !memcmp(ce->name, fullname.buf, fullname.len))
1871 die(_("path '%s' is in the index, but not '%s'\n"
1872 "hint: Did you mean ':%d:%s' aka ':%d:./%s'?"),
1873 fullname.buf, filename,
1874 ce_stage(ce), fullname.buf,
1875 ce_stage(ce), filename);
1876 }
1877
1878 if (repo_file_exists(r, filename))
1879 die(_("path '%s' exists on disk, but not in the index"), filename);
1880 if (is_missing_file_error(errno))
1881 die(_("path '%s' does not exist (neither on disk nor in the index)"),
1882 filename);
1883
1884 strbuf_release(&fullname);
1885 }
1886
1887
1888 static char *resolve_relative_path(struct repository *r, const char *rel)
1889 {
1890 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1891 return NULL;
1892
1893 if (r != the_repository || !is_inside_work_tree())
1894 die(_("relative path syntax can't be used outside working tree"));
1895
1896 /* die() inside prefix_path() if resolved path is outside worktree */
1897 return prefix_path(startup_info->prefix,
1898 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1899 rel);
1900 }
1901
1902 static int reject_tree_in_index(struct repository *repo,
1903 int only_to_die,
1904 const struct cache_entry *ce,
1905 int stage,
1906 const char *prefix,
1907 const char *cp)
1908 {
1909 if (!S_ISSPARSEDIR(ce->ce_mode))
1910 return 0;
1911 if (only_to_die)
1912 diagnose_invalid_index_path(repo, stage, prefix, cp);
1913 return -1;
1914 }
1915
1916 static enum get_oid_result get_oid_with_context_1(struct repository *repo,
1917 const char *name,
1918 unsigned flags,
1919 const char *prefix,
1920 struct object_id *oid,
1921 struct object_context *oc)
1922 {
1923 int ret, bracket_depth;
1924 int namelen = strlen(name);
1925 const char *cp;
1926 int only_to_die = flags & GET_OID_ONLY_TO_DIE;
1927
1928 memset(oc, 0, sizeof(*oc));
1929 oc->mode = S_IFINVALID;
1930 strbuf_init(&oc->symlink_path, 0);
1931 ret = get_oid_1(repo, name, namelen, oid, flags);
1932 if (!ret && flags & GET_OID_REQUIRE_PATH)
1933 die(_("<object>:<path> required, only <object> '%s' given"),
1934 name);
1935 if (!ret)
1936 return ret;
1937 /*
1938 * tree:path --> object name of path in tree
1939 * :path -> object name of absolute path in index
1940 * :./path -> object name of path relative to cwd in index
1941 * :[0-3]:path -> object name of path in index at stage
1942 * :/foo -> recent commit matching foo
1943 */
1944 if (name[0] == ':') {
1945 int stage = 0;
1946 const struct cache_entry *ce;
1947 char *new_path = NULL;
1948 int pos;
1949 if (!only_to_die && namelen > 2 && name[1] == '/') {
1950 struct handle_one_ref_cb cb;
1951 struct commit_list *list = NULL;
1952
1953 cb.repo = repo;
1954 cb.list = &list;
1955 refs_for_each_ref(get_main_ref_store(repo), handle_one_ref, &cb);
1956 refs_head_ref(get_main_ref_store(repo), handle_one_ref, &cb);
1957 commit_list_sort_by_date(&list);
1958 return get_oid_oneline(repo, name + 2, oid, list);
1959 }
1960 if (namelen < 3 ||
1961 name[2] != ':' ||
1962 name[1] < '0' || '3' < name[1])
1963 cp = name + 1;
1964 else {
1965 stage = name[1] - '0';
1966 cp = name + 3;
1967 }
1968 new_path = resolve_relative_path(repo, cp);
1969 if (!new_path) {
1970 namelen = namelen - (cp - name);
1971 } else {
1972 cp = new_path;
1973 namelen = strlen(cp);
1974 }
1975
1976 if (flags & GET_OID_RECORD_PATH)
1977 oc->path = xstrdup(cp);
1978
1979 if (!repo->index || !repo->index->cache)
1980 repo_read_index(repo);
1981 pos = index_name_pos(repo->index, cp, namelen);
1982 if (pos < 0)
1983 pos = -pos - 1;
1984 while (pos < repo->index->cache_nr) {
1985 ce = repo->index->cache[pos];
1986 if (ce_namelen(ce) != namelen ||
1987 memcmp(ce->name, cp, namelen))
1988 break;
1989 if (ce_stage(ce) == stage) {
1990 free(new_path);
1991 if (reject_tree_in_index(repo, only_to_die, ce,
1992 stage, prefix, cp))
1993 return -1;
1994 oidcpy(oid, &ce->oid);
1995 oc->mode = ce->ce_mode;
1996 return 0;
1997 }
1998 pos++;
1999 }
2000 if (only_to_die && name[1] && name[1] != '/')
2001 diagnose_invalid_index_path(repo, stage, prefix, cp);
2002 free(new_path);
2003 return -1;
2004 }
2005 for (cp = name, bracket_depth = 0; *cp; cp++) {
2006 if (*cp == '{')
2007 bracket_depth++;
2008 else if (bracket_depth && *cp == '}')
2009 bracket_depth--;
2010 else if (!bracket_depth && *cp == ':')
2011 break;
2012 }
2013 if (*cp == ':') {
2014 struct object_id tree_oid;
2015 int len = cp - name;
2016 unsigned sub_flags = flags;
2017
2018 sub_flags &= ~GET_OID_DISAMBIGUATORS;
2019 sub_flags |= GET_OID_TREEISH;
2020
2021 if (!get_oid_1(repo, name, len, &tree_oid, sub_flags)) {
2022 const char *filename = cp+1;
2023 char *new_filename = NULL;
2024
2025 new_filename = resolve_relative_path(repo, filename);
2026 if (new_filename)
2027 filename = new_filename;
2028 if (flags & GET_OID_FOLLOW_SYMLINKS) {
2029 ret = get_tree_entry_follow_symlinks(repo, &tree_oid,
2030 filename, oid, &oc->symlink_path,
2031 &oc->mode);
2032 } else {
2033 ret = get_tree_entry(repo, &tree_oid, filename, oid,
2034 &oc->mode);
2035 if (ret && only_to_die) {
2036 diagnose_invalid_oid_path(repo, prefix,
2037 filename,
2038 &tree_oid,
2039 name, len);
2040 }
2041 }
2042 if (flags & GET_OID_RECORD_PATH)
2043 oc->path = xstrdup(filename);
2044
2045 free(new_filename);
2046 return ret;
2047 } else {
2048 if (only_to_die)
2049 die(_("invalid object name '%.*s'."), len, name);
2050 }
2051 }
2052 return ret;
2053 }
2054
2055 /*
2056 * Call this function when you know "name" given by the end user must
2057 * name an object but it doesn't; the function _may_ die with a better
2058 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
2059 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
2060 * you have a chance to diagnose the error further.
2061 */
2062 void maybe_die_on_misspelt_object_name(struct repository *r,
2063 const char *name,
2064 const char *prefix)
2065 {
2066 struct object_context oc;
2067 struct object_id oid;
2068 get_oid_with_context_1(r, name, GET_OID_ONLY_TO_DIE | GET_OID_QUIETLY,
2069 prefix, &oid, &oc);
2070 }
2071
2072 enum get_oid_result get_oid_with_context(struct repository *repo,
2073 const char *str,
2074 unsigned flags,
2075 struct object_id *oid,
2076 struct object_context *oc)
2077 {
2078 if (flags & GET_OID_FOLLOW_SYMLINKS && flags & GET_OID_ONLY_TO_DIE)
2079 BUG("incompatible flags for get_oid_with_context");
2080 return get_oid_with_context_1(repo, str, flags, NULL, oid, oc);
2081 }