]> git.ipfire.org Git - thirdparty/git.git/blob - object-name.c
cache.h: remove this no-longer-used header
[thirdparty/git.git] / object-name.c
1 #include "git-compat-util.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 "read-cache-ll.h"
22 #include "repository.h"
23 #include "setup.h"
24 #include "submodule.h"
25 #include "midx.h"
26 #include "commit-reach.h"
27 #include "date.h"
28
29 static int get_oid_oneline(struct repository *r, const char *, struct object_id *, struct commit_list *);
30
31 typedef int (*disambiguate_hint_fn)(struct repository *, const struct object_id *, void *);
32
33 struct disambiguate_state {
34 int len; /* length of prefix in hex chars */
35 char hex_pfx[GIT_MAX_HEXSZ + 1];
36 struct object_id bin_pfx;
37
38 struct repository *repo;
39 disambiguate_hint_fn fn;
40 void *cb_data;
41 struct object_id candidate;
42 unsigned candidate_exists:1;
43 unsigned candidate_checked:1;
44 unsigned candidate_ok:1;
45 unsigned disambiguate_fn_used:1;
46 unsigned ambiguous:1;
47 unsigned always_call_fn:1;
48 };
49
50 static void update_candidates(struct disambiguate_state *ds, const struct object_id *current)
51 {
52 if (ds->always_call_fn) {
53 ds->ambiguous = ds->fn(ds->repo, current, ds->cb_data) ? 1 : 0;
54 return;
55 }
56 if (!ds->candidate_exists) {
57 /* this is the first candidate */
58 oidcpy(&ds->candidate, current);
59 ds->candidate_exists = 1;
60 return;
61 } else if (oideq(&ds->candidate, current)) {
62 /* the same as what we already have seen */
63 return;
64 }
65
66 if (!ds->fn) {
67 /* cannot disambiguate between ds->candidate and current */
68 ds->ambiguous = 1;
69 return;
70 }
71
72 if (!ds->candidate_checked) {
73 ds->candidate_ok = ds->fn(ds->repo, &ds->candidate, ds->cb_data);
74 ds->disambiguate_fn_used = 1;
75 ds->candidate_checked = 1;
76 }
77
78 if (!ds->candidate_ok) {
79 /* discard the candidate; we know it does not satisfy fn */
80 oidcpy(&ds->candidate, current);
81 ds->candidate_checked = 0;
82 return;
83 }
84
85 /* if we reach this point, we know ds->candidate satisfies fn */
86 if (ds->fn(ds->repo, current, ds->cb_data)) {
87 /*
88 * if both current and candidate satisfy fn, we cannot
89 * disambiguate.
90 */
91 ds->candidate_ok = 0;
92 ds->ambiguous = 1;
93 }
94
95 /* otherwise, current can be discarded and candidate is still good */
96 }
97
98 static int match_hash(unsigned, const unsigned char *, const unsigned char *);
99
100 static enum cb_next match_prefix(const struct object_id *oid, void *arg)
101 {
102 struct disambiguate_state *ds = arg;
103 /* no need to call match_hash, oidtree_each did prefix match */
104 update_candidates(ds, oid);
105 return ds->ambiguous ? CB_BREAK : CB_CONTINUE;
106 }
107
108 static void find_short_object_filename(struct disambiguate_state *ds)
109 {
110 struct object_directory *odb;
111
112 for (odb = ds->repo->objects->odb; odb && !ds->ambiguous; odb = odb->next)
113 oidtree_each(odb_loose_cache(odb, &ds->bin_pfx),
114 &ds->bin_pfx, ds->len, match_prefix, ds);
115 }
116
117 static int match_hash(unsigned len, const unsigned char *a, const unsigned char *b)
118 {
119 do {
120 if (*a != *b)
121 return 0;
122 a++;
123 b++;
124 len -= 2;
125 } while (len > 1);
126 if (len)
127 if ((*a ^ *b) & 0xf0)
128 return 0;
129 return 1;
130 }
131
132 static void unique_in_midx(struct multi_pack_index *m,
133 struct disambiguate_state *ds)
134 {
135 uint32_t num, i, first = 0;
136 const struct object_id *current = NULL;
137 num = m->num_objects;
138
139 if (!num)
140 return;
141
142 bsearch_midx(&ds->bin_pfx, m, &first);
143
144 /*
145 * At this point, "first" is the location of the lowest object
146 * with an object name that could match "bin_pfx". See if we have
147 * 0, 1 or more objects that actually match(es).
148 */
149 for (i = first; i < num && !ds->ambiguous; i++) {
150 struct object_id oid;
151 current = nth_midxed_object_oid(&oid, m, i);
152 if (!match_hash(ds->len, ds->bin_pfx.hash, current->hash))
153 break;
154 update_candidates(ds, current);
155 }
156 }
157
158 static void unique_in_pack(struct packed_git *p,
159 struct disambiguate_state *ds)
160 {
161 uint32_t num, i, first = 0;
162
163 if (p->multi_pack_index)
164 return;
165
166 if (open_pack_index(p) || !p->num_objects)
167 return;
168
169 num = p->num_objects;
170 bsearch_pack(&ds->bin_pfx, p, &first);
171
172 /*
173 * At this point, "first" is the location of the lowest object
174 * with an object name that could match "bin_pfx". See if we have
175 * 0, 1 or more objects that actually match(es).
176 */
177 for (i = first; i < num && !ds->ambiguous; i++) {
178 struct object_id oid;
179 nth_packed_object_id(&oid, p, i);
180 if (!match_hash(ds->len, ds->bin_pfx.hash, oid.hash))
181 break;
182 update_candidates(ds, &oid);
183 }
184 }
185
186 static void find_short_packed_object(struct disambiguate_state *ds)
187 {
188 struct multi_pack_index *m;
189 struct packed_git *p;
190
191 for (m = get_multi_pack_index(ds->repo); m && !ds->ambiguous;
192 m = m->next)
193 unique_in_midx(m, ds);
194 for (p = get_packed_git(ds->repo); p && !ds->ambiguous;
195 p = p->next)
196 unique_in_pack(p, ds);
197 }
198
199 static int finish_object_disambiguation(struct disambiguate_state *ds,
200 struct object_id *oid)
201 {
202 if (ds->ambiguous)
203 return SHORT_NAME_AMBIGUOUS;
204
205 if (!ds->candidate_exists)
206 return MISSING_OBJECT;
207
208 if (!ds->candidate_checked)
209 /*
210 * If this is the only candidate, there is no point
211 * calling the disambiguation hint callback.
212 *
213 * On the other hand, if the current candidate
214 * replaced an earlier candidate that did _not_ pass
215 * the disambiguation hint callback, then we do have
216 * more than one objects that match the short name
217 * given, so we should make sure this one matches;
218 * otherwise, if we discovered this one and the one
219 * that we previously discarded in the reverse order,
220 * we would end up showing different results in the
221 * same repository!
222 */
223 ds->candidate_ok = (!ds->disambiguate_fn_used ||
224 ds->fn(ds->repo, &ds->candidate, ds->cb_data));
225
226 if (!ds->candidate_ok)
227 return SHORT_NAME_AMBIGUOUS;
228
229 oidcpy(oid, &ds->candidate);
230 return 0;
231 }
232
233 static int disambiguate_commit_only(struct repository *r,
234 const struct object_id *oid,
235 void *cb_data UNUSED)
236 {
237 int kind = oid_object_info(r, oid, NULL);
238 return kind == OBJ_COMMIT;
239 }
240
241 static int disambiguate_committish_only(struct repository *r,
242 const struct object_id *oid,
243 void *cb_data UNUSED)
244 {
245 struct object *obj;
246 int kind;
247
248 kind = oid_object_info(r, oid, NULL);
249 if (kind == OBJ_COMMIT)
250 return 1;
251 if (kind != OBJ_TAG)
252 return 0;
253
254 /* We need to do this the hard way... */
255 obj = deref_tag(r, parse_object(r, oid), NULL, 0);
256 if (obj && obj->type == OBJ_COMMIT)
257 return 1;
258 return 0;
259 }
260
261 static int disambiguate_tree_only(struct repository *r,
262 const struct object_id *oid,
263 void *cb_data UNUSED)
264 {
265 int kind = oid_object_info(r, oid, NULL);
266 return kind == OBJ_TREE;
267 }
268
269 static int disambiguate_treeish_only(struct repository *r,
270 const struct object_id *oid,
271 void *cb_data UNUSED)
272 {
273 struct object *obj;
274 int kind;
275
276 kind = oid_object_info(r, oid, NULL);
277 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
278 return 1;
279 if (kind != OBJ_TAG)
280 return 0;
281
282 /* We need to do this the hard way... */
283 obj = deref_tag(r, parse_object(r, oid), NULL, 0);
284 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
285 return 1;
286 return 0;
287 }
288
289 static int disambiguate_blob_only(struct repository *r,
290 const struct object_id *oid,
291 void *cb_data UNUSED)
292 {
293 int kind = oid_object_info(r, oid, NULL);
294 return kind == OBJ_BLOB;
295 }
296
297 static disambiguate_hint_fn default_disambiguate_hint;
298
299 int set_disambiguate_hint_config(const char *var, const char *value)
300 {
301 static const struct {
302 const char *name;
303 disambiguate_hint_fn fn;
304 } hints[] = {
305 { "none", NULL },
306 { "commit", disambiguate_commit_only },
307 { "committish", disambiguate_committish_only },
308 { "tree", disambiguate_tree_only },
309 { "treeish", disambiguate_treeish_only },
310 { "blob", disambiguate_blob_only }
311 };
312 int i;
313
314 if (!value)
315 return config_error_nonbool(var);
316
317 for (i = 0; i < ARRAY_SIZE(hints); i++) {
318 if (!strcasecmp(value, hints[i].name)) {
319 default_disambiguate_hint = hints[i].fn;
320 return 0;
321 }
322 }
323
324 return error("unknown hint type for '%s': %s", var, value);
325 }
326
327 static int init_object_disambiguation(struct repository *r,
328 const char *name, int len,
329 struct disambiguate_state *ds)
330 {
331 int i;
332
333 if (len < MINIMUM_ABBREV || len > the_hash_algo->hexsz)
334 return -1;
335
336 memset(ds, 0, sizeof(*ds));
337
338 for (i = 0; i < len ;i++) {
339 unsigned char c = name[i];
340 unsigned char val;
341 if (c >= '0' && c <= '9')
342 val = c - '0';
343 else if (c >= 'a' && c <= 'f')
344 val = c - 'a' + 10;
345 else if (c >= 'A' && c <='F') {
346 val = c - 'A' + 10;
347 c -= 'A' - 'a';
348 }
349 else
350 return -1;
351 ds->hex_pfx[i] = c;
352 if (!(i & 1))
353 val <<= 4;
354 ds->bin_pfx.hash[i >> 1] |= val;
355 }
356
357 ds->len = len;
358 ds->hex_pfx[len] = '\0';
359 ds->repo = r;
360 prepare_alt_odb(r);
361 return 0;
362 }
363
364 struct ambiguous_output {
365 const struct disambiguate_state *ds;
366 struct strbuf advice;
367 struct strbuf sb;
368 };
369
370 static int show_ambiguous_object(const struct object_id *oid, void *data)
371 {
372 struct ambiguous_output *state = data;
373 const struct disambiguate_state *ds = state->ds;
374 struct strbuf *advice = &state->advice;
375 struct strbuf *sb = &state->sb;
376 int type;
377 const char *hash;
378
379 if (ds->fn && !ds->fn(ds->repo, oid, ds->cb_data))
380 return 0;
381
382 hash = repo_find_unique_abbrev(ds->repo, oid, DEFAULT_ABBREV);
383 type = oid_object_info(ds->repo, oid, NULL);
384
385 if (type < 0) {
386 /*
387 * TRANSLATORS: This is a line of ambiguous object
388 * output shown when we cannot look up or parse the
389 * object in question. E.g. "deadbeef [bad object]".
390 */
391 strbuf_addf(sb, _("%s [bad object]"), hash);
392 goto out;
393 }
394
395 assert(type == OBJ_TREE || type == OBJ_COMMIT ||
396 type == OBJ_BLOB || type == OBJ_TAG);
397
398 if (type == OBJ_COMMIT) {
399 struct strbuf date = STRBUF_INIT;
400 struct strbuf msg = STRBUF_INIT;
401 struct commit *commit = lookup_commit(ds->repo, oid);
402
403 if (commit) {
404 struct pretty_print_context pp = {0};
405 pp.date_mode.type = DATE_SHORT;
406 repo_format_commit_message(the_repository, commit,
407 "%ad", &date, &pp);
408 repo_format_commit_message(the_repository, commit,
409 "%s", &msg, &pp);
410 }
411
412 /*
413 * TRANSLATORS: This is a line of ambiguous commit
414 * object output. E.g.:
415 *
416 * "deadbeef commit 2021-01-01 - Some Commit Message"
417 */
418 strbuf_addf(sb, _("%s commit %s - %s"), hash, date.buf,
419 msg.buf);
420
421 strbuf_release(&date);
422 strbuf_release(&msg);
423 } else if (type == OBJ_TAG) {
424 struct tag *tag = lookup_tag(ds->repo, oid);
425
426 if (!parse_tag(tag) && tag->tag) {
427 /*
428 * TRANSLATORS: This is a line of ambiguous
429 * tag object output. E.g.:
430 *
431 * "deadbeef tag 2022-01-01 - Some Tag Message"
432 *
433 * The second argument is the YYYY-MM-DD found
434 * in the tag.
435 *
436 * The third argument is the "tag" string
437 * from object.c.
438 */
439 strbuf_addf(sb, _("%s tag %s - %s"), hash,
440 show_date(tag->date, 0, DATE_MODE(SHORT)),
441 tag->tag);
442 } else {
443 /*
444 * TRANSLATORS: This is a line of ambiguous
445 * tag object output where we couldn't parse
446 * the tag itself. E.g.:
447 *
448 * "deadbeef [bad tag, could not parse it]"
449 */
450 strbuf_addf(sb, _("%s [bad tag, could not parse it]"),
451 hash);
452 }
453 } else if (type == OBJ_TREE) {
454 /*
455 * TRANSLATORS: This is a line of ambiguous <type>
456 * object output. E.g. "deadbeef tree".
457 */
458 strbuf_addf(sb, _("%s tree"), hash);
459 } else if (type == OBJ_BLOB) {
460 /*
461 * TRANSLATORS: This is a line of ambiguous <type>
462 * object output. E.g. "deadbeef blob".
463 */
464 strbuf_addf(sb, _("%s blob"), hash);
465 }
466
467
468 out:
469 /*
470 * TRANSLATORS: This is line item of ambiguous object output
471 * from describe_ambiguous_object() above. For RTL languages
472 * you'll probably want to swap the "%s" and leading " " space
473 * around.
474 */
475 strbuf_addf(advice, _(" %s\n"), sb->buf);
476
477 strbuf_reset(sb);
478 return 0;
479 }
480
481 static int collect_ambiguous(const struct object_id *oid, void *data)
482 {
483 oid_array_append(data, oid);
484 return 0;
485 }
486
487 static int repo_collect_ambiguous(struct repository *r UNUSED,
488 const struct object_id *oid,
489 void *data)
490 {
491 return collect_ambiguous(oid, data);
492 }
493
494 static int sort_ambiguous(const void *a, const void *b, void *ctx)
495 {
496 struct repository *sort_ambiguous_repo = ctx;
497 int a_type = oid_object_info(sort_ambiguous_repo, a, NULL);
498 int b_type = oid_object_info(sort_ambiguous_repo, b, NULL);
499 int a_type_sort;
500 int b_type_sort;
501
502 /*
503 * Sorts by hash within the same object type, just as
504 * oid_array_for_each_unique() would do.
505 */
506 if (a_type == b_type)
507 return oidcmp(a, b);
508
509 /*
510 * Between object types show tags, then commits, and finally
511 * trees and blobs.
512 *
513 * The object_type enum is commit, tree, blob, tag, but we
514 * want tag, commit, tree blob. Cleverly (perhaps too
515 * cleverly) do that with modulus, since the enum assigns 1 to
516 * commit, so tag becomes 0.
517 */
518 a_type_sort = a_type % 4;
519 b_type_sort = b_type % 4;
520 return a_type_sort > b_type_sort ? 1 : -1;
521 }
522
523 static void sort_ambiguous_oid_array(struct repository *r, struct oid_array *a)
524 {
525 QSORT_S(a->oid, a->nr, sort_ambiguous, r);
526 }
527
528 static enum get_oid_result get_short_oid(struct repository *r,
529 const char *name, int len,
530 struct object_id *oid,
531 unsigned flags)
532 {
533 int status;
534 struct disambiguate_state ds;
535 int quietly = !!(flags & GET_OID_QUIETLY);
536
537 if (init_object_disambiguation(r, name, len, &ds) < 0)
538 return -1;
539
540 if (HAS_MULTI_BITS(flags & GET_OID_DISAMBIGUATORS))
541 BUG("multiple get_short_oid disambiguator flags");
542
543 if (flags & GET_OID_COMMIT)
544 ds.fn = disambiguate_commit_only;
545 else if (flags & GET_OID_COMMITTISH)
546 ds.fn = disambiguate_committish_only;
547 else if (flags & GET_OID_TREE)
548 ds.fn = disambiguate_tree_only;
549 else if (flags & GET_OID_TREEISH)
550 ds.fn = disambiguate_treeish_only;
551 else if (flags & GET_OID_BLOB)
552 ds.fn = disambiguate_blob_only;
553 else
554 ds.fn = default_disambiguate_hint;
555
556 find_short_object_filename(&ds);
557 find_short_packed_object(&ds);
558 status = finish_object_disambiguation(&ds, oid);
559
560 /*
561 * If we didn't find it, do the usual reprepare() slow-path,
562 * since the object may have recently been added to the repository
563 * or migrated from loose to packed.
564 */
565 if (status == MISSING_OBJECT) {
566 reprepare_packed_git(r);
567 find_short_object_filename(&ds);
568 find_short_packed_object(&ds);
569 status = finish_object_disambiguation(&ds, oid);
570 }
571
572 if (!quietly && (status == SHORT_NAME_AMBIGUOUS)) {
573 struct oid_array collect = OID_ARRAY_INIT;
574 struct ambiguous_output out = {
575 .ds = &ds,
576 .sb = STRBUF_INIT,
577 .advice = STRBUF_INIT,
578 };
579
580 error(_("short object ID %s is ambiguous"), ds.hex_pfx);
581
582 /*
583 * We may still have ambiguity if we simply saw a series of
584 * candidates that did not satisfy our hint function. In
585 * that case, we still want to show them, so disable the hint
586 * function entirely.
587 */
588 if (!ds.ambiguous)
589 ds.fn = NULL;
590
591 repo_for_each_abbrev(r, ds.hex_pfx, collect_ambiguous, &collect);
592 sort_ambiguous_oid_array(r, &collect);
593
594 if (oid_array_for_each(&collect, show_ambiguous_object, &out))
595 BUG("show_ambiguous_object shouldn't return non-zero");
596
597 /*
598 * TRANSLATORS: The argument is the list of ambiguous
599 * objects composed in show_ambiguous_object(). See
600 * its "TRANSLATORS" comments for details.
601 */
602 advise(_("The candidates are:\n%s"), out.advice.buf);
603
604 oid_array_clear(&collect);
605 strbuf_release(&out.advice);
606 strbuf_release(&out.sb);
607 }
608
609 return status;
610 }
611
612 int repo_for_each_abbrev(struct repository *r, const char *prefix,
613 each_abbrev_fn fn, void *cb_data)
614 {
615 struct oid_array collect = OID_ARRAY_INIT;
616 struct disambiguate_state ds;
617 int ret;
618
619 if (init_object_disambiguation(r, prefix, strlen(prefix), &ds) < 0)
620 return -1;
621
622 ds.always_call_fn = 1;
623 ds.fn = repo_collect_ambiguous;
624 ds.cb_data = &collect;
625 find_short_object_filename(&ds);
626 find_short_packed_object(&ds);
627
628 ret = oid_array_for_each_unique(&collect, fn, cb_data);
629 oid_array_clear(&collect);
630 return ret;
631 }
632
633 /*
634 * Return the slot of the most-significant bit set in "val". There are various
635 * ways to do this quickly with fls() or __builtin_clzl(), but speed is
636 * probably not a big deal here.
637 */
638 static unsigned msb(unsigned long val)
639 {
640 unsigned r = 0;
641 while (val >>= 1)
642 r++;
643 return r;
644 }
645
646 struct min_abbrev_data {
647 unsigned int init_len;
648 unsigned int cur_len;
649 char *hex;
650 struct repository *repo;
651 const struct object_id *oid;
652 };
653
654 static inline char get_hex_char_from_oid(const struct object_id *oid,
655 unsigned int pos)
656 {
657 static const char hex[] = "0123456789abcdef";
658
659 if ((pos & 1) == 0)
660 return hex[oid->hash[pos >> 1] >> 4];
661 else
662 return hex[oid->hash[pos >> 1] & 0xf];
663 }
664
665 static int extend_abbrev_len(const struct object_id *oid, void *cb_data)
666 {
667 struct min_abbrev_data *mad = cb_data;
668
669 unsigned int i = mad->init_len;
670 while (mad->hex[i] && mad->hex[i] == get_hex_char_from_oid(oid, i))
671 i++;
672
673 if (i < GIT_MAX_RAWSZ && i >= mad->cur_len)
674 mad->cur_len = i + 1;
675
676 return 0;
677 }
678
679 static int repo_extend_abbrev_len(struct repository *r UNUSED,
680 const struct object_id *oid,
681 void *cb_data)
682 {
683 return extend_abbrev_len(oid, cb_data);
684 }
685
686 static void find_abbrev_len_for_midx(struct multi_pack_index *m,
687 struct min_abbrev_data *mad)
688 {
689 int match = 0;
690 uint32_t num, first = 0;
691 struct object_id oid;
692 const struct object_id *mad_oid;
693
694 if (!m->num_objects)
695 return;
696
697 num = m->num_objects;
698 mad_oid = mad->oid;
699 match = bsearch_midx(mad_oid, m, &first);
700
701 /*
702 * first is now the position in the packfile where we would insert
703 * mad->hash if it does not exist (or the position of mad->hash if
704 * it does exist). Hence, we consider a maximum of two objects
705 * nearby for the abbreviation length.
706 */
707 mad->init_len = 0;
708 if (!match) {
709 if (nth_midxed_object_oid(&oid, m, first))
710 extend_abbrev_len(&oid, mad);
711 } else if (first < num - 1) {
712 if (nth_midxed_object_oid(&oid, m, first + 1))
713 extend_abbrev_len(&oid, mad);
714 }
715 if (first > 0) {
716 if (nth_midxed_object_oid(&oid, m, first - 1))
717 extend_abbrev_len(&oid, mad);
718 }
719 mad->init_len = mad->cur_len;
720 }
721
722 static void find_abbrev_len_for_pack(struct packed_git *p,
723 struct min_abbrev_data *mad)
724 {
725 int match = 0;
726 uint32_t num, first = 0;
727 struct object_id oid;
728 const struct object_id *mad_oid;
729
730 if (p->multi_pack_index)
731 return;
732
733 if (open_pack_index(p) || !p->num_objects)
734 return;
735
736 num = p->num_objects;
737 mad_oid = mad->oid;
738 match = bsearch_pack(mad_oid, p, &first);
739
740 /*
741 * first is now the position in the packfile where we would insert
742 * mad->hash if it does not exist (or the position of mad->hash if
743 * it does exist). Hence, we consider a maximum of two objects
744 * nearby for the abbreviation length.
745 */
746 mad->init_len = 0;
747 if (!match) {
748 if (!nth_packed_object_id(&oid, p, first))
749 extend_abbrev_len(&oid, mad);
750 } else if (first < num - 1) {
751 if (!nth_packed_object_id(&oid, p, first + 1))
752 extend_abbrev_len(&oid, mad);
753 }
754 if (first > 0) {
755 if (!nth_packed_object_id(&oid, p, first - 1))
756 extend_abbrev_len(&oid, mad);
757 }
758 mad->init_len = mad->cur_len;
759 }
760
761 static void find_abbrev_len_packed(struct min_abbrev_data *mad)
762 {
763 struct multi_pack_index *m;
764 struct packed_git *p;
765
766 for (m = get_multi_pack_index(mad->repo); m; m = m->next)
767 find_abbrev_len_for_midx(m, mad);
768 for (p = get_packed_git(mad->repo); p; p = p->next)
769 find_abbrev_len_for_pack(p, mad);
770 }
771
772 int repo_find_unique_abbrev_r(struct repository *r, char *hex,
773 const struct object_id *oid, int len)
774 {
775 struct disambiguate_state ds;
776 struct min_abbrev_data mad;
777 struct object_id oid_ret;
778 const unsigned hexsz = r->hash_algo->hexsz;
779
780 if (len < 0) {
781 unsigned long count = repo_approximate_object_count(r);
782 /*
783 * Add one because the MSB only tells us the highest bit set,
784 * not including the value of all the _other_ bits (so "15"
785 * is only one off of 2^4, but the MSB is the 3rd bit.
786 */
787 len = msb(count) + 1;
788 /*
789 * We now know we have on the order of 2^len objects, which
790 * expects a collision at 2^(len/2). But we also care about hex
791 * chars, not bits, and there are 4 bits per hex. So all
792 * together we need to divide by 2 and round up.
793 */
794 len = DIV_ROUND_UP(len, 2);
795 /*
796 * For very small repos, we stick with our regular fallback.
797 */
798 if (len < FALLBACK_DEFAULT_ABBREV)
799 len = FALLBACK_DEFAULT_ABBREV;
800 }
801
802 oid_to_hex_r(hex, oid);
803 if (len == hexsz || !len)
804 return hexsz;
805
806 mad.repo = r;
807 mad.init_len = len;
808 mad.cur_len = len;
809 mad.hex = hex;
810 mad.oid = oid;
811
812 find_abbrev_len_packed(&mad);
813
814 if (init_object_disambiguation(r, hex, mad.cur_len, &ds) < 0)
815 return -1;
816
817 ds.fn = repo_extend_abbrev_len;
818 ds.always_call_fn = 1;
819 ds.cb_data = (void *)&mad;
820
821 find_short_object_filename(&ds);
822 (void)finish_object_disambiguation(&ds, &oid_ret);
823
824 hex[mad.cur_len] = 0;
825 return mad.cur_len;
826 }
827
828 const char *repo_find_unique_abbrev(struct repository *r,
829 const struct object_id *oid,
830 int len)
831 {
832 static int bufno;
833 static char hexbuffer[4][GIT_MAX_HEXSZ + 1];
834 char *hex = hexbuffer[bufno];
835 bufno = (bufno + 1) % ARRAY_SIZE(hexbuffer);
836 repo_find_unique_abbrev_r(r, hex, oid, len);
837 return hex;
838 }
839
840 static int ambiguous_path(const char *path, int len)
841 {
842 int slash = 1;
843 int cnt;
844
845 for (cnt = 0; cnt < len; cnt++) {
846 switch (*path++) {
847 case '\0':
848 break;
849 case '/':
850 if (slash)
851 break;
852 slash = 1;
853 continue;
854 case '.':
855 continue;
856 default:
857 slash = 0;
858 continue;
859 }
860 break;
861 }
862 return slash;
863 }
864
865 static inline int at_mark(const char *string, int len,
866 const char **suffix, int nr)
867 {
868 int i;
869
870 for (i = 0; i < nr; i++) {
871 int suffix_len = strlen(suffix[i]);
872 if (suffix_len <= len
873 && !strncasecmp(string, suffix[i], suffix_len))
874 return suffix_len;
875 }
876 return 0;
877 }
878
879 static inline int upstream_mark(const char *string, int len)
880 {
881 const char *suffix[] = { "@{upstream}", "@{u}" };
882 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
883 }
884
885 static inline int push_mark(const char *string, int len)
886 {
887 const char *suffix[] = { "@{push}" };
888 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
889 }
890
891 static enum get_oid_result get_oid_1(struct repository *r, const char *name, int len, struct object_id *oid, unsigned lookup_flags);
892 static int interpret_nth_prior_checkout(struct repository *r, const char *name, int namelen, struct strbuf *buf);
893
894 static int get_oid_basic(struct repository *r, const char *str, int len,
895 struct object_id *oid, unsigned int flags)
896 {
897 static const char *warn_msg = "refname '%.*s' is ambiguous.";
898 static const char *object_name_msg = N_(
899 "Git normally never creates a ref that ends with 40 hex characters\n"
900 "because it will be ignored when you just specify 40-hex. These refs\n"
901 "may be created by mistake. For example,\n"
902 "\n"
903 " git switch -c $br $(git rev-parse ...)\n"
904 "\n"
905 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
906 "examine these refs and maybe delete them. Turn this message off by\n"
907 "running \"git config advice.objectNameWarning false\"");
908 struct object_id tmp_oid;
909 char *real_ref = NULL;
910 int refs_found = 0;
911 int at, reflog_len, nth_prior = 0;
912 int fatal = !(flags & GET_OID_QUIETLY);
913
914 if (len == r->hash_algo->hexsz && !get_oid_hex(str, oid)) {
915 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
916 refs_found = repo_dwim_ref(r, str, len, &tmp_oid, &real_ref, 0);
917 if (refs_found > 0) {
918 warning(warn_msg, len, str);
919 if (advice_enabled(ADVICE_OBJECT_NAME_WARNING))
920 fprintf(stderr, "%s\n", _(object_name_msg));
921 }
922 free(real_ref);
923 }
924 return 0;
925 }
926
927 /* basic@{time or number or -number} format to query ref-log */
928 reflog_len = at = 0;
929 if (len && str[len-1] == '}') {
930 for (at = len-4; at >= 0; at--) {
931 if (str[at] == '@' && str[at+1] == '{') {
932 if (str[at+2] == '-') {
933 if (at != 0)
934 /* @{-N} not at start */
935 return -1;
936 nth_prior = 1;
937 continue;
938 }
939 if (!upstream_mark(str + at, len - at) &&
940 !push_mark(str + at, len - at)) {
941 reflog_len = (len-1) - (at+2);
942 len = at;
943 }
944 break;
945 }
946 }
947 }
948
949 /* Accept only unambiguous ref paths. */
950 if (len && ambiguous_path(str, len))
951 return -1;
952
953 if (nth_prior) {
954 struct strbuf buf = STRBUF_INIT;
955 int detached;
956
957 if (interpret_nth_prior_checkout(r, str, len, &buf) > 0) {
958 detached = (buf.len == r->hash_algo->hexsz && !get_oid_hex(buf.buf, oid));
959 strbuf_release(&buf);
960 if (detached)
961 return 0;
962 }
963 }
964
965 if (!len && reflog_len)
966 /* allow "@{...}" to mean the current branch reflog */
967 refs_found = repo_dwim_ref(r, "HEAD", 4, oid, &real_ref, !fatal);
968 else if (reflog_len)
969 refs_found = repo_dwim_log(r, str, len, oid, &real_ref);
970 else
971 refs_found = repo_dwim_ref(r, str, len, oid, &real_ref, !fatal);
972
973 if (!refs_found)
974 return -1;
975
976 if (warn_ambiguous_refs && !(flags & GET_OID_QUIETLY) &&
977 (refs_found > 1 ||
978 !get_short_oid(r, str, len, &tmp_oid, GET_OID_QUIETLY)))
979 warning(warn_msg, len, str);
980
981 if (reflog_len) {
982 int nth, i;
983 timestamp_t at_time;
984 timestamp_t co_time;
985 int co_tz, co_cnt;
986
987 /* Is it asking for N-th entry, or approxidate? */
988 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
989 char ch = str[at+2+i];
990 if ('0' <= ch && ch <= '9')
991 nth = nth * 10 + ch - '0';
992 else
993 nth = -1;
994 }
995 if (100000000 <= nth) {
996 at_time = nth;
997 nth = -1;
998 } else if (0 <= nth)
999 at_time = 0;
1000 else {
1001 int errors = 0;
1002 char *tmp = xstrndup(str + at + 2, reflog_len);
1003 at_time = approxidate_careful(tmp, &errors);
1004 free(tmp);
1005 if (errors) {
1006 free(real_ref);
1007 return -1;
1008 }
1009 }
1010 if (read_ref_at(get_main_ref_store(r),
1011 real_ref, flags, at_time, nth, oid, NULL,
1012 &co_time, &co_tz, &co_cnt)) {
1013 if (!len) {
1014 if (!skip_prefix(real_ref, "refs/heads/", &str))
1015 str = "HEAD";
1016 len = strlen(str);
1017 }
1018 if (at_time) {
1019 if (!(flags & GET_OID_QUIETLY)) {
1020 warning(_("log for '%.*s' only goes back to %s"),
1021 len, str,
1022 show_date(co_time, co_tz, DATE_MODE(RFC2822)));
1023 }
1024 } else {
1025 if (flags & GET_OID_QUIETLY) {
1026 exit(128);
1027 }
1028 die(_("log for '%.*s' only has %d entries"),
1029 len, str, co_cnt);
1030 }
1031 }
1032 }
1033
1034 free(real_ref);
1035 return 0;
1036 }
1037
1038 static enum get_oid_result get_parent(struct repository *r,
1039 const char *name, int len,
1040 struct object_id *result, int idx)
1041 {
1042 struct object_id oid;
1043 enum get_oid_result ret = get_oid_1(r, name, len, &oid,
1044 GET_OID_COMMITTISH);
1045 struct commit *commit;
1046 struct commit_list *p;
1047
1048 if (ret)
1049 return ret;
1050 commit = lookup_commit_reference(r, &oid);
1051 if (repo_parse_commit(r, commit))
1052 return MISSING_OBJECT;
1053 if (!idx) {
1054 oidcpy(result, &commit->object.oid);
1055 return FOUND;
1056 }
1057 p = commit->parents;
1058 while (p) {
1059 if (!--idx) {
1060 oidcpy(result, &p->item->object.oid);
1061 return FOUND;
1062 }
1063 p = p->next;
1064 }
1065 return MISSING_OBJECT;
1066 }
1067
1068 static enum get_oid_result get_nth_ancestor(struct repository *r,
1069 const char *name, int len,
1070 struct object_id *result,
1071 int generation)
1072 {
1073 struct object_id oid;
1074 struct commit *commit;
1075 int ret;
1076
1077 ret = get_oid_1(r, name, len, &oid, GET_OID_COMMITTISH);
1078 if (ret)
1079 return ret;
1080 commit = lookup_commit_reference(r, &oid);
1081 if (!commit)
1082 return MISSING_OBJECT;
1083
1084 while (generation--) {
1085 if (repo_parse_commit(r, commit) || !commit->parents)
1086 return MISSING_OBJECT;
1087 commit = commit->parents->item;
1088 }
1089 oidcpy(result, &commit->object.oid);
1090 return FOUND;
1091 }
1092
1093 struct object *repo_peel_to_type(struct repository *r, const char *name, int namelen,
1094 struct object *o, enum object_type expected_type)
1095 {
1096 if (name && !namelen)
1097 namelen = strlen(name);
1098 while (1) {
1099 if (!o || (!o->parsed && !parse_object(r, &o->oid)))
1100 return NULL;
1101 if (expected_type == OBJ_ANY || o->type == expected_type)
1102 return o;
1103 if (o->type == OBJ_TAG)
1104 o = ((struct tag*) o)->tagged;
1105 else if (o->type == OBJ_COMMIT)
1106 o = &(repo_get_commit_tree(r, ((struct commit *)o))->object);
1107 else {
1108 if (name)
1109 error("%.*s: expected %s type, but the object "
1110 "dereferences to %s type",
1111 namelen, name, type_name(expected_type),
1112 type_name(o->type));
1113 return NULL;
1114 }
1115 }
1116 }
1117
1118 static int peel_onion(struct repository *r, const char *name, int len,
1119 struct object_id *oid, unsigned lookup_flags)
1120 {
1121 struct object_id outer;
1122 const char *sp;
1123 unsigned int expected_type = 0;
1124 struct object *o;
1125
1126 /*
1127 * "ref^{type}" dereferences ref repeatedly until you cannot
1128 * dereference anymore, or you get an object of given type,
1129 * whichever comes first. "ref^{}" means just dereference
1130 * tags until you get a non-tag. "ref^0" is a shorthand for
1131 * "ref^{commit}". "commit^{tree}" could be used to find the
1132 * top-level tree of the given commit.
1133 */
1134 if (len < 4 || name[len-1] != '}')
1135 return -1;
1136
1137 for (sp = name + len - 1; name <= sp; sp--) {
1138 int ch = *sp;
1139 if (ch == '{' && name < sp && sp[-1] == '^')
1140 break;
1141 }
1142 if (sp <= name)
1143 return -1;
1144
1145 sp++; /* beginning of type name, or closing brace for empty */
1146 if (starts_with(sp, "commit}"))
1147 expected_type = OBJ_COMMIT;
1148 else if (starts_with(sp, "tag}"))
1149 expected_type = OBJ_TAG;
1150 else if (starts_with(sp, "tree}"))
1151 expected_type = OBJ_TREE;
1152 else if (starts_with(sp, "blob}"))
1153 expected_type = OBJ_BLOB;
1154 else if (starts_with(sp, "object}"))
1155 expected_type = OBJ_ANY;
1156 else if (sp[0] == '}')
1157 expected_type = OBJ_NONE;
1158 else if (sp[0] == '/')
1159 expected_type = OBJ_COMMIT;
1160 else
1161 return -1;
1162
1163 lookup_flags &= ~GET_OID_DISAMBIGUATORS;
1164 if (expected_type == OBJ_COMMIT)
1165 lookup_flags |= GET_OID_COMMITTISH;
1166 else if (expected_type == OBJ_TREE)
1167 lookup_flags |= GET_OID_TREEISH;
1168
1169 if (get_oid_1(r, name, sp - name - 2, &outer, lookup_flags))
1170 return -1;
1171
1172 o = parse_object(r, &outer);
1173 if (!o)
1174 return -1;
1175 if (!expected_type) {
1176 o = deref_tag(r, o, name, sp - name - 2);
1177 if (!o || (!o->parsed && !parse_object(r, &o->oid)))
1178 return -1;
1179 oidcpy(oid, &o->oid);
1180 return 0;
1181 }
1182
1183 /*
1184 * At this point, the syntax look correct, so
1185 * if we do not get the needed object, we should
1186 * barf.
1187 */
1188 o = repo_peel_to_type(r, name, len, o, expected_type);
1189 if (!o)
1190 return -1;
1191
1192 oidcpy(oid, &o->oid);
1193 if (sp[0] == '/') {
1194 /* "$commit^{/foo}" */
1195 char *prefix;
1196 int ret;
1197 struct commit_list *list = NULL;
1198
1199 /*
1200 * $commit^{/}. Some regex implementation may reject.
1201 * We don't need regex anyway. '' pattern always matches.
1202 */
1203 if (sp[1] == '}')
1204 return 0;
1205
1206 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
1207 commit_list_insert((struct commit *)o, &list);
1208 ret = get_oid_oneline(r, prefix, oid, list);
1209 free(prefix);
1210 return ret;
1211 }
1212 return 0;
1213 }
1214
1215 static int get_describe_name(struct repository *r,
1216 const char *name, int len,
1217 struct object_id *oid)
1218 {
1219 const char *cp;
1220 unsigned flags = GET_OID_QUIETLY | GET_OID_COMMIT;
1221
1222 for (cp = name + len - 1; name + 2 <= cp; cp--) {
1223 char ch = *cp;
1224 if (!isxdigit(ch)) {
1225 /* We must be looking at g in "SOMETHING-g"
1226 * for it to be describe output.
1227 */
1228 if (ch == 'g' && cp[-1] == '-') {
1229 cp++;
1230 len -= cp - name;
1231 return get_short_oid(r,
1232 cp, len, oid, flags);
1233 }
1234 }
1235 }
1236 return -1;
1237 }
1238
1239 static enum get_oid_result get_oid_1(struct repository *r,
1240 const char *name, int len,
1241 struct object_id *oid,
1242 unsigned lookup_flags)
1243 {
1244 int ret, has_suffix;
1245 const char *cp;
1246
1247 /*
1248 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
1249 */
1250 has_suffix = 0;
1251 for (cp = name + len - 1; name <= cp; cp--) {
1252 int ch = *cp;
1253 if ('0' <= ch && ch <= '9')
1254 continue;
1255 if (ch == '~' || ch == '^')
1256 has_suffix = ch;
1257 break;
1258 }
1259
1260 if (has_suffix) {
1261 unsigned int num = 0;
1262 int len1 = cp - name;
1263 cp++;
1264 while (cp < name + len) {
1265 unsigned int digit = *cp++ - '0';
1266 if (unsigned_mult_overflows(num, 10))
1267 return MISSING_OBJECT;
1268 num *= 10;
1269 if (unsigned_add_overflows(num, digit))
1270 return MISSING_OBJECT;
1271 num += digit;
1272 }
1273 if (!num && len1 == len - 1)
1274 num = 1;
1275 else if (num > INT_MAX)
1276 return MISSING_OBJECT;
1277 if (has_suffix == '^')
1278 return get_parent(r, name, len1, oid, num);
1279 /* else if (has_suffix == '~') -- goes without saying */
1280 return get_nth_ancestor(r, name, len1, oid, num);
1281 }
1282
1283 ret = peel_onion(r, name, len, oid, lookup_flags);
1284 if (!ret)
1285 return FOUND;
1286
1287 ret = get_oid_basic(r, name, len, oid, lookup_flags);
1288 if (!ret)
1289 return FOUND;
1290
1291 /* It could be describe output that is "SOMETHING-gXXXX" */
1292 ret = get_describe_name(r, name, len, oid);
1293 if (!ret)
1294 return FOUND;
1295
1296 return get_short_oid(r, name, len, oid, lookup_flags);
1297 }
1298
1299 /*
1300 * This interprets names like ':/Initial revision of "git"' by searching
1301 * through history and returning the first commit whose message starts
1302 * the given regular expression.
1303 *
1304 * For negative-matching, prefix the pattern-part with '!-', like: ':/!-WIP'.
1305 *
1306 * For a literal '!' character at the beginning of a pattern, you have to repeat
1307 * that, like: ':/!!foo'
1308 *
1309 * For future extension, all other sequences beginning with ':/!' are reserved.
1310 */
1311
1312 /* Remember to update object flag allocation in object.h */
1313 #define ONELINE_SEEN (1u<<20)
1314
1315 struct handle_one_ref_cb {
1316 struct repository *repo;
1317 struct commit_list **list;
1318 };
1319
1320 static int handle_one_ref(const char *path, const struct object_id *oid,
1321 int flag UNUSED,
1322 void *cb_data)
1323 {
1324 struct handle_one_ref_cb *cb = cb_data;
1325 struct commit_list **list = cb->list;
1326 struct object *object = parse_object(cb->repo, oid);
1327 if (!object)
1328 return 0;
1329 if (object->type == OBJ_TAG) {
1330 object = deref_tag(cb->repo, object, path,
1331 strlen(path));
1332 if (!object)
1333 return 0;
1334 }
1335 if (object->type != OBJ_COMMIT)
1336 return 0;
1337 commit_list_insert((struct commit *)object, list);
1338 return 0;
1339 }
1340
1341 static int get_oid_oneline(struct repository *r,
1342 const char *prefix, struct object_id *oid,
1343 struct commit_list *list)
1344 {
1345 struct commit_list *backup = NULL, *l;
1346 int found = 0;
1347 int negative = 0;
1348 regex_t regex;
1349
1350 if (prefix[0] == '!') {
1351 prefix++;
1352
1353 if (prefix[0] == '-') {
1354 prefix++;
1355 negative = 1;
1356 } else if (prefix[0] != '!') {
1357 return -1;
1358 }
1359 }
1360
1361 if (regcomp(&regex, prefix, REG_EXTENDED))
1362 return -1;
1363
1364 for (l = list; l; l = l->next) {
1365 l->item->object.flags |= ONELINE_SEEN;
1366 commit_list_insert(l->item, &backup);
1367 }
1368 while (list) {
1369 const char *p, *buf;
1370 struct commit *commit;
1371 int matches;
1372
1373 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
1374 if (!parse_object(r, &commit->object.oid))
1375 continue;
1376 buf = repo_get_commit_buffer(r, commit, NULL);
1377 p = strstr(buf, "\n\n");
1378 matches = negative ^ (p && !regexec(&regex, p + 2, 0, NULL, 0));
1379 repo_unuse_commit_buffer(r, commit, buf);
1380
1381 if (matches) {
1382 oidcpy(oid, &commit->object.oid);
1383 found = 1;
1384 break;
1385 }
1386 }
1387 regfree(&regex);
1388 free_commit_list(list);
1389 for (l = backup; l; l = l->next)
1390 clear_commit_marks(l->item, ONELINE_SEEN);
1391 free_commit_list(backup);
1392 return found ? 0 : -1;
1393 }
1394
1395 struct grab_nth_branch_switch_cbdata {
1396 int remaining;
1397 struct strbuf *sb;
1398 };
1399
1400 static int grab_nth_branch_switch(struct object_id *ooid UNUSED,
1401 struct object_id *noid UNUSED,
1402 const char *email UNUSED,
1403 timestamp_t timestamp UNUSED,
1404 int tz UNUSED,
1405 const char *message, void *cb_data)
1406 {
1407 struct grab_nth_branch_switch_cbdata *cb = cb_data;
1408 const char *match = NULL, *target = NULL;
1409 size_t len;
1410
1411 if (skip_prefix(message, "checkout: moving from ", &match))
1412 target = strstr(match, " to ");
1413
1414 if (!match || !target)
1415 return 0;
1416 if (--(cb->remaining) == 0) {
1417 len = target - match;
1418 strbuf_reset(cb->sb);
1419 strbuf_add(cb->sb, match, len);
1420 return 1; /* we are done */
1421 }
1422 return 0;
1423 }
1424
1425 /*
1426 * Parse @{-N} syntax, return the number of characters parsed
1427 * if successful; otherwise signal an error with negative value.
1428 */
1429 static int interpret_nth_prior_checkout(struct repository *r,
1430 const char *name, int namelen,
1431 struct strbuf *buf)
1432 {
1433 long nth;
1434 int retval;
1435 struct grab_nth_branch_switch_cbdata cb;
1436 const char *brace;
1437 char *num_end;
1438
1439 if (namelen < 4)
1440 return -1;
1441 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
1442 return -1;
1443 brace = memchr(name, '}', namelen);
1444 if (!brace)
1445 return -1;
1446 nth = strtol(name + 3, &num_end, 10);
1447 if (num_end != brace)
1448 return -1;
1449 if (nth <= 0)
1450 return -1;
1451 cb.remaining = nth;
1452 cb.sb = buf;
1453
1454 retval = refs_for_each_reflog_ent_reverse(get_main_ref_store(r),
1455 "HEAD", grab_nth_branch_switch, &cb);
1456 if (0 < retval) {
1457 retval = brace - name + 1;
1458 } else
1459 retval = 0;
1460
1461 return retval;
1462 }
1463
1464 int repo_get_oid_mb(struct repository *r,
1465 const char *name,
1466 struct object_id *oid)
1467 {
1468 struct commit *one, *two;
1469 struct commit_list *mbs;
1470 struct object_id oid_tmp;
1471 const char *dots;
1472 int st;
1473
1474 dots = strstr(name, "...");
1475 if (!dots)
1476 return repo_get_oid(r, name, oid);
1477 if (dots == name)
1478 st = repo_get_oid(r, "HEAD", &oid_tmp);
1479 else {
1480 struct strbuf sb;
1481 strbuf_init(&sb, dots - name);
1482 strbuf_add(&sb, name, dots - name);
1483 st = repo_get_oid_committish(r, sb.buf, &oid_tmp);
1484 strbuf_release(&sb);
1485 }
1486 if (st)
1487 return st;
1488 one = lookup_commit_reference_gently(r, &oid_tmp, 0);
1489 if (!one)
1490 return -1;
1491
1492 if (repo_get_oid_committish(r, dots[3] ? (dots + 3) : "HEAD", &oid_tmp))
1493 return -1;
1494 two = lookup_commit_reference_gently(r, &oid_tmp, 0);
1495 if (!two)
1496 return -1;
1497 mbs = repo_get_merge_bases(r, one, two);
1498 if (!mbs || mbs->next)
1499 st = -1;
1500 else {
1501 st = 0;
1502 oidcpy(oid, &mbs->item->object.oid);
1503 }
1504 free_commit_list(mbs);
1505 return st;
1506 }
1507
1508 /* parse @something syntax, when 'something' is not {.*} */
1509 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1510 {
1511 const char *next;
1512
1513 if (len || name[1] == '{')
1514 return -1;
1515
1516 /* make sure it's a single @, or @@{.*}, not @foo */
1517 next = memchr(name + len + 1, '@', namelen - len - 1);
1518 if (next && next[1] != '{')
1519 return -1;
1520 if (!next)
1521 next = name + namelen;
1522 if (next != name + 1)
1523 return -1;
1524
1525 strbuf_reset(buf);
1526 strbuf_add(buf, "HEAD", 4);
1527 return 1;
1528 }
1529
1530 static int reinterpret(struct repository *r,
1531 const char *name, int namelen, int len,
1532 struct strbuf *buf, unsigned allowed)
1533 {
1534 /* we have extra data, which might need further processing */
1535 struct strbuf tmp = STRBUF_INIT;
1536 int used = buf->len;
1537 int ret;
1538 struct interpret_branch_name_options options = {
1539 .allowed = allowed
1540 };
1541
1542 strbuf_add(buf, name + len, namelen - len);
1543 ret = repo_interpret_branch_name(r, buf->buf, buf->len, &tmp, &options);
1544 /* that data was not interpreted, remove our cruft */
1545 if (ret < 0) {
1546 strbuf_setlen(buf, used);
1547 return len;
1548 }
1549 strbuf_reset(buf);
1550 strbuf_addbuf(buf, &tmp);
1551 strbuf_release(&tmp);
1552 /* tweak for size of {-N} versus expanded ref name */
1553 return ret - used + len;
1554 }
1555
1556 static void set_shortened_ref(struct repository *r, struct strbuf *buf, const char *ref)
1557 {
1558 char *s = refs_shorten_unambiguous_ref(get_main_ref_store(r), ref, 0);
1559 strbuf_reset(buf);
1560 strbuf_addstr(buf, s);
1561 free(s);
1562 }
1563
1564 static int branch_interpret_allowed(const char *refname, unsigned allowed)
1565 {
1566 if (!allowed)
1567 return 1;
1568
1569 if ((allowed & INTERPRET_BRANCH_LOCAL) &&
1570 starts_with(refname, "refs/heads/"))
1571 return 1;
1572 if ((allowed & INTERPRET_BRANCH_REMOTE) &&
1573 starts_with(refname, "refs/remotes/"))
1574 return 1;
1575
1576 return 0;
1577 }
1578
1579 static int interpret_branch_mark(struct repository *r,
1580 const char *name, int namelen,
1581 int at, struct strbuf *buf,
1582 int (*get_mark)(const char *, int),
1583 const char *(*get_data)(struct branch *,
1584 struct strbuf *),
1585 const struct interpret_branch_name_options *options)
1586 {
1587 int len;
1588 struct branch *branch;
1589 struct strbuf err = STRBUF_INIT;
1590 const char *value;
1591
1592 len = get_mark(name + at, namelen - at);
1593 if (!len)
1594 return -1;
1595
1596 if (memchr(name, ':', at))
1597 return -1;
1598
1599 if (at) {
1600 char *name_str = xmemdupz(name, at);
1601 branch = branch_get(name_str);
1602 free(name_str);
1603 } else
1604 branch = branch_get(NULL);
1605
1606 value = get_data(branch, &err);
1607 if (!value) {
1608 if (options->nonfatal_dangling_mark) {
1609 strbuf_release(&err);
1610 return -1;
1611 } else {
1612 die("%s", err.buf);
1613 }
1614 }
1615
1616 if (!branch_interpret_allowed(value, options->allowed))
1617 return -1;
1618
1619 set_shortened_ref(r, buf, value);
1620 return len + at;
1621 }
1622
1623 int repo_interpret_branch_name(struct repository *r,
1624 const char *name, int namelen,
1625 struct strbuf *buf,
1626 const struct interpret_branch_name_options *options)
1627 {
1628 char *at;
1629 const char *start;
1630 int len;
1631
1632 if (!namelen)
1633 namelen = strlen(name);
1634
1635 if (!options->allowed || (options->allowed & INTERPRET_BRANCH_LOCAL)) {
1636 len = interpret_nth_prior_checkout(r, name, namelen, buf);
1637 if (!len) {
1638 return len; /* syntax Ok, not enough switches */
1639 } else if (len > 0) {
1640 if (len == namelen)
1641 return len; /* consumed all */
1642 else
1643 return reinterpret(r, name, namelen, len, buf,
1644 options->allowed);
1645 }
1646 }
1647
1648 for (start = name;
1649 (at = memchr(start, '@', namelen - (start - name)));
1650 start = at + 1) {
1651
1652 if (!options->allowed || (options->allowed & INTERPRET_BRANCH_HEAD)) {
1653 len = interpret_empty_at(name, namelen, at - name, buf);
1654 if (len > 0)
1655 return reinterpret(r, name, namelen, len, buf,
1656 options->allowed);
1657 }
1658
1659 len = interpret_branch_mark(r, name, namelen, at - name, buf,
1660 upstream_mark, branch_get_upstream,
1661 options);
1662 if (len > 0)
1663 return len;
1664
1665 len = interpret_branch_mark(r, name, namelen, at - name, buf,
1666 push_mark, branch_get_push,
1667 options);
1668 if (len > 0)
1669 return len;
1670 }
1671
1672 return -1;
1673 }
1674
1675 void strbuf_branchname(struct strbuf *sb, const char *name, unsigned allowed)
1676 {
1677 int len = strlen(name);
1678 struct interpret_branch_name_options options = {
1679 .allowed = allowed
1680 };
1681 int used = repo_interpret_branch_name(the_repository, name, len, sb,
1682 &options);
1683
1684 if (used < 0)
1685 used = 0;
1686 strbuf_add(sb, name + used, len - used);
1687 }
1688
1689 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1690 {
1691 if (startup_info->have_repository)
1692 strbuf_branchname(sb, name, INTERPRET_BRANCH_LOCAL);
1693 else
1694 strbuf_addstr(sb, name);
1695
1696 /*
1697 * This splice must be done even if we end up rejecting the
1698 * name; builtin/branch.c::copy_or_rename_branch() still wants
1699 * to see what the name expanded to so that "branch -m" can be
1700 * used as a tool to correct earlier mistakes.
1701 */
1702 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1703
1704 if (*name == '-' ||
1705 !strcmp(sb->buf, "refs/heads/HEAD"))
1706 return -1;
1707
1708 return check_refname_format(sb->buf, 0);
1709 }
1710
1711 /*
1712 * This is like "get_oid_basic()", except it allows "object ID expressions",
1713 * notably "xyz^" for "parent of xyz"
1714 */
1715 int repo_get_oid(struct repository *r, const char *name, struct object_id *oid)
1716 {
1717 struct object_context unused;
1718 return get_oid_with_context(r, name, 0, oid, &unused);
1719 }
1720
1721 /*
1722 * This returns a non-zero value if the string (built using printf
1723 * format and the given arguments) is not a valid object.
1724 */
1725 int get_oidf(struct object_id *oid, const char *fmt, ...)
1726 {
1727 va_list ap;
1728 int ret;
1729 struct strbuf sb = STRBUF_INIT;
1730
1731 va_start(ap, fmt);
1732 strbuf_vaddf(&sb, fmt, ap);
1733 va_end(ap);
1734
1735 ret = repo_get_oid(the_repository, sb.buf, oid);
1736 strbuf_release(&sb);
1737
1738 return ret;
1739 }
1740
1741 /*
1742 * Many callers know that the user meant to name a commit-ish by
1743 * syntactical positions where the object name appears. Calling this
1744 * function allows the machinery to disambiguate shorter-than-unique
1745 * abbreviated object names between commit-ish and others.
1746 *
1747 * Note that this does NOT error out when the named object is not a
1748 * commit-ish. It is merely to give a hint to the disambiguation
1749 * machinery.
1750 */
1751 int repo_get_oid_committish(struct repository *r,
1752 const char *name,
1753 struct object_id *oid)
1754 {
1755 struct object_context unused;
1756 return get_oid_with_context(r, name, GET_OID_COMMITTISH,
1757 oid, &unused);
1758 }
1759
1760 int repo_get_oid_treeish(struct repository *r,
1761 const char *name,
1762 struct object_id *oid)
1763 {
1764 struct object_context unused;
1765 return get_oid_with_context(r, name, GET_OID_TREEISH,
1766 oid, &unused);
1767 }
1768
1769 int repo_get_oid_commit(struct repository *r,
1770 const char *name,
1771 struct object_id *oid)
1772 {
1773 struct object_context unused;
1774 return get_oid_with_context(r, name, GET_OID_COMMIT,
1775 oid, &unused);
1776 }
1777
1778 int repo_get_oid_tree(struct repository *r,
1779 const char *name,
1780 struct object_id *oid)
1781 {
1782 struct object_context unused;
1783 return get_oid_with_context(r, name, GET_OID_TREE,
1784 oid, &unused);
1785 }
1786
1787 int repo_get_oid_blob(struct repository *r,
1788 const char *name,
1789 struct object_id *oid)
1790 {
1791 struct object_context unused;
1792 return get_oid_with_context(r, name, GET_OID_BLOB,
1793 oid, &unused);
1794 }
1795
1796 /* Must be called only when object_name:filename doesn't exist. */
1797 static void diagnose_invalid_oid_path(struct repository *r,
1798 const char *prefix,
1799 const char *filename,
1800 const struct object_id *tree_oid,
1801 const char *object_name,
1802 int object_name_len)
1803 {
1804 struct object_id oid;
1805 unsigned short mode;
1806
1807 if (!prefix)
1808 prefix = "";
1809
1810 if (file_exists(filename))
1811 die(_("path '%s' exists on disk, but not in '%.*s'"),
1812 filename, object_name_len, object_name);
1813 if (is_missing_file_error(errno)) {
1814 char *fullname = xstrfmt("%s%s", prefix, filename);
1815
1816 if (!get_tree_entry(r, tree_oid, fullname, &oid, &mode)) {
1817 die(_("path '%s' exists, but not '%s'\n"
1818 "hint: Did you mean '%.*s:%s' aka '%.*s:./%s'?"),
1819 fullname,
1820 filename,
1821 object_name_len, object_name,
1822 fullname,
1823 object_name_len, object_name,
1824 filename);
1825 }
1826 die(_("path '%s' does not exist in '%.*s'"),
1827 filename, object_name_len, object_name);
1828 }
1829 }
1830
1831 /* Must be called only when :stage:filename doesn't exist. */
1832 static void diagnose_invalid_index_path(struct repository *r,
1833 int stage,
1834 const char *prefix,
1835 const char *filename)
1836 {
1837 struct index_state *istate = r->index;
1838 const struct cache_entry *ce;
1839 int pos;
1840 unsigned namelen = strlen(filename);
1841 struct strbuf fullname = STRBUF_INIT;
1842
1843 if (!prefix)
1844 prefix = "";
1845
1846 /* Wrong stage number? */
1847 pos = index_name_pos(istate, filename, namelen);
1848 if (pos < 0)
1849 pos = -pos - 1;
1850 if (pos < istate->cache_nr) {
1851 ce = istate->cache[pos];
1852 if (!S_ISSPARSEDIR(ce->ce_mode) &&
1853 ce_namelen(ce) == namelen &&
1854 !memcmp(ce->name, filename, namelen))
1855 die(_("path '%s' is in the index, but not at stage %d\n"
1856 "hint: Did you mean ':%d:%s'?"),
1857 filename, stage,
1858 ce_stage(ce), filename);
1859 }
1860
1861 /* Confusion between relative and absolute filenames? */
1862 strbuf_addstr(&fullname, prefix);
1863 strbuf_addstr(&fullname, filename);
1864 pos = index_name_pos(istate, fullname.buf, fullname.len);
1865 if (pos < 0)
1866 pos = -pos - 1;
1867 if (pos < istate->cache_nr) {
1868 ce = istate->cache[pos];
1869 if (!S_ISSPARSEDIR(ce->ce_mode) &&
1870 ce_namelen(ce) == fullname.len &&
1871 !memcmp(ce->name, fullname.buf, fullname.len))
1872 die(_("path '%s' is in the index, but not '%s'\n"
1873 "hint: Did you mean ':%d:%s' aka ':%d:./%s'?"),
1874 fullname.buf, filename,
1875 ce_stage(ce), fullname.buf,
1876 ce_stage(ce), filename);
1877 }
1878
1879 if (repo_file_exists(r, filename))
1880 die(_("path '%s' exists on disk, but not in the index"), filename);
1881 if (is_missing_file_error(errno))
1882 die(_("path '%s' does not exist (neither on disk nor in the index)"),
1883 filename);
1884
1885 strbuf_release(&fullname);
1886 }
1887
1888
1889 static char *resolve_relative_path(struct repository *r, const char *rel)
1890 {
1891 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1892 return NULL;
1893
1894 if (r != the_repository || !is_inside_work_tree())
1895 die(_("relative path syntax can't be used outside working tree"));
1896
1897 /* die() inside prefix_path() if resolved path is outside worktree */
1898 return prefix_path(startup_info->prefix,
1899 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1900 rel);
1901 }
1902
1903 static int reject_tree_in_index(struct repository *repo,
1904 int only_to_die,
1905 const struct cache_entry *ce,
1906 int stage,
1907 const char *prefix,
1908 const char *cp)
1909 {
1910 if (!S_ISSPARSEDIR(ce->ce_mode))
1911 return 0;
1912 if (only_to_die)
1913 diagnose_invalid_index_path(repo, stage, prefix, cp);
1914 return -1;
1915 }
1916
1917 static enum get_oid_result get_oid_with_context_1(struct repository *repo,
1918 const char *name,
1919 unsigned flags,
1920 const char *prefix,
1921 struct object_id *oid,
1922 struct object_context *oc)
1923 {
1924 int ret, bracket_depth;
1925 int namelen = strlen(name);
1926 const char *cp;
1927 int only_to_die = flags & GET_OID_ONLY_TO_DIE;
1928
1929 memset(oc, 0, sizeof(*oc));
1930 oc->mode = S_IFINVALID;
1931 strbuf_init(&oc->symlink_path, 0);
1932 ret = get_oid_1(repo, name, namelen, oid, flags);
1933 if (!ret && flags & GET_OID_REQUIRE_PATH)
1934 die(_("<object>:<path> required, only <object> '%s' given"),
1935 name);
1936 if (!ret)
1937 return ret;
1938 /*
1939 * tree:path --> object name of path in tree
1940 * :path -> object name of absolute path in index
1941 * :./path -> object name of path relative to cwd in index
1942 * :[0-3]:path -> object name of path in index at stage
1943 * :/foo -> recent commit matching foo
1944 */
1945 if (name[0] == ':') {
1946 int stage = 0;
1947 const struct cache_entry *ce;
1948 char *new_path = NULL;
1949 int pos;
1950 if (!only_to_die && namelen > 2 && name[1] == '/') {
1951 struct handle_one_ref_cb cb;
1952 struct commit_list *list = NULL;
1953
1954 cb.repo = repo;
1955 cb.list = &list;
1956 refs_for_each_ref(get_main_ref_store(repo), handle_one_ref, &cb);
1957 refs_head_ref(get_main_ref_store(repo), handle_one_ref, &cb);
1958 commit_list_sort_by_date(&list);
1959 return get_oid_oneline(repo, name + 2, oid, list);
1960 }
1961 if (namelen < 3 ||
1962 name[2] != ':' ||
1963 name[1] < '0' || '3' < name[1])
1964 cp = name + 1;
1965 else {
1966 stage = name[1] - '0';
1967 cp = name + 3;
1968 }
1969 new_path = resolve_relative_path(repo, cp);
1970 if (!new_path) {
1971 namelen = namelen - (cp - name);
1972 } else {
1973 cp = new_path;
1974 namelen = strlen(cp);
1975 }
1976
1977 if (flags & GET_OID_RECORD_PATH)
1978 oc->path = xstrdup(cp);
1979
1980 if (!repo->index || !repo->index->cache)
1981 repo_read_index(repo);
1982 pos = index_name_pos(repo->index, cp, namelen);
1983 if (pos < 0)
1984 pos = -pos - 1;
1985 while (pos < repo->index->cache_nr) {
1986 ce = repo->index->cache[pos];
1987 if (ce_namelen(ce) != namelen ||
1988 memcmp(ce->name, cp, namelen))
1989 break;
1990 if (ce_stage(ce) == stage) {
1991 free(new_path);
1992 if (reject_tree_in_index(repo, only_to_die, ce,
1993 stage, prefix, cp))
1994 return -1;
1995 oidcpy(oid, &ce->oid);
1996 oc->mode = ce->ce_mode;
1997 return 0;
1998 }
1999 pos++;
2000 }
2001 if (only_to_die && name[1] && name[1] != '/')
2002 diagnose_invalid_index_path(repo, stage, prefix, cp);
2003 free(new_path);
2004 return -1;
2005 }
2006 for (cp = name, bracket_depth = 0; *cp; cp++) {
2007 if (*cp == '{')
2008 bracket_depth++;
2009 else if (bracket_depth && *cp == '}')
2010 bracket_depth--;
2011 else if (!bracket_depth && *cp == ':')
2012 break;
2013 }
2014 if (*cp == ':') {
2015 struct object_id tree_oid;
2016 int len = cp - name;
2017 unsigned sub_flags = flags;
2018
2019 sub_flags &= ~GET_OID_DISAMBIGUATORS;
2020 sub_flags |= GET_OID_TREEISH;
2021
2022 if (!get_oid_1(repo, name, len, &tree_oid, sub_flags)) {
2023 const char *filename = cp+1;
2024 char *new_filename = NULL;
2025
2026 new_filename = resolve_relative_path(repo, filename);
2027 if (new_filename)
2028 filename = new_filename;
2029 if (flags & GET_OID_FOLLOW_SYMLINKS) {
2030 ret = get_tree_entry_follow_symlinks(repo, &tree_oid,
2031 filename, oid, &oc->symlink_path,
2032 &oc->mode);
2033 } else {
2034 ret = get_tree_entry(repo, &tree_oid, filename, oid,
2035 &oc->mode);
2036 if (ret && only_to_die) {
2037 diagnose_invalid_oid_path(repo, prefix,
2038 filename,
2039 &tree_oid,
2040 name, len);
2041 }
2042 }
2043 if (flags & GET_OID_RECORD_PATH)
2044 oc->path = xstrdup(filename);
2045
2046 free(new_filename);
2047 return ret;
2048 } else {
2049 if (only_to_die)
2050 die(_("invalid object name '%.*s'."), len, name);
2051 }
2052 }
2053 return ret;
2054 }
2055
2056 /*
2057 * Call this function when you know "name" given by the end user must
2058 * name an object but it doesn't; the function _may_ die with a better
2059 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
2060 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
2061 * you have a chance to diagnose the error further.
2062 */
2063 void maybe_die_on_misspelt_object_name(struct repository *r,
2064 const char *name,
2065 const char *prefix)
2066 {
2067 struct object_context oc;
2068 struct object_id oid;
2069 get_oid_with_context_1(r, name, GET_OID_ONLY_TO_DIE | GET_OID_QUIETLY,
2070 prefix, &oid, &oc);
2071 }
2072
2073 enum get_oid_result get_oid_with_context(struct repository *repo,
2074 const char *str,
2075 unsigned flags,
2076 struct object_id *oid,
2077 struct object_context *oc)
2078 {
2079 if (flags & GET_OID_FOLLOW_SYMLINKS && flags & GET_OID_ONLY_TO_DIE)
2080 BUG("incompatible flags for get_oid_with_context");
2081 return get_oid_with_context_1(repo, str, flags, NULL, oid, oc);
2082 }