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