26a5811c846e5bff3afa226625e7b13909b5752f
[git/git.git] / sha1_name.c
1 #include "cache.h"
2 #include "tag.h"
3 #include "commit.h"
4 #include "tree.h"
5 #include "blob.h"
6 #include "tree-walk.h"
7 #include "refs.h"
8 #include "remote.h"
9
10 static int get_sha1_oneline(const char *, unsigned char *, struct commit_list *);
11
12 typedef int (*disambiguate_hint_fn)(const unsigned char *, void *);
13
14 struct disambiguate_state {
15 disambiguate_hint_fn fn;
16 void *cb_data;
17 unsigned char candidate[20];
18 unsigned candidate_exists:1;
19 unsigned candidate_checked:1;
20 unsigned candidate_ok:1;
21 unsigned disambiguate_fn_used:1;
22 unsigned ambiguous:1;
23 unsigned always_call_fn:1;
24 };
25
26 static void update_candidates(struct disambiguate_state *ds, const unsigned char *current)
27 {
28 if (ds->always_call_fn) {
29 ds->ambiguous = ds->fn(current, ds->cb_data) ? 1 : 0;
30 return;
31 }
32 if (!ds->candidate_exists) {
33 /* this is the first candidate */
34 hashcpy(ds->candidate, current);
35 ds->candidate_exists = 1;
36 return;
37 } else if (!hashcmp(ds->candidate, current)) {
38 /* the same as what we already have seen */
39 return;
40 }
41
42 if (!ds->fn) {
43 /* cannot disambiguate between ds->candidate and current */
44 ds->ambiguous = 1;
45 return;
46 }
47
48 if (!ds->candidate_checked) {
49 ds->candidate_ok = ds->fn(ds->candidate, ds->cb_data);
50 ds->disambiguate_fn_used = 1;
51 ds->candidate_checked = 1;
52 }
53
54 if (!ds->candidate_ok) {
55 /* discard the candidate; we know it does not satisfy fn */
56 hashcpy(ds->candidate, current);
57 ds->candidate_checked = 0;
58 return;
59 }
60
61 /* if we reach this point, we know ds->candidate satisfies fn */
62 if (ds->fn(current, ds->cb_data)) {
63 /*
64 * if both current and candidate satisfy fn, we cannot
65 * disambiguate.
66 */
67 ds->candidate_ok = 0;
68 ds->ambiguous = 1;
69 }
70
71 /* otherwise, current can be discarded and candidate is still good */
72 }
73
74 static void find_short_object_filename(int len, const char *hex_pfx, struct disambiguate_state *ds)
75 {
76 struct alternate_object_database *alt;
77 char hex[40];
78 static struct alternate_object_database *fakeent;
79
80 if (!fakeent) {
81 /*
82 * Create a "fake" alternate object database that
83 * points to our own object database, to make it
84 * easier to get a temporary working space in
85 * alt->name/alt->base while iterating over the
86 * object databases including our own.
87 */
88 const char *objdir = get_object_directory();
89 int objdir_len = strlen(objdir);
90 int entlen = objdir_len + 43;
91 fakeent = xmalloc(sizeof(*fakeent) + entlen);
92 memcpy(fakeent->base, objdir, objdir_len);
93 fakeent->name = fakeent->base + objdir_len + 1;
94 fakeent->name[-1] = '/';
95 }
96 fakeent->next = alt_odb_list;
97
98 sprintf(hex, "%.2s", hex_pfx);
99 for (alt = fakeent; alt && !ds->ambiguous; alt = alt->next) {
100 struct dirent *de;
101 DIR *dir;
102 sprintf(alt->name, "%.2s/", hex_pfx);
103 dir = opendir(alt->base);
104 if (!dir)
105 continue;
106
107 while (!ds->ambiguous && (de = readdir(dir)) != NULL) {
108 unsigned char sha1[20];
109
110 if (strlen(de->d_name) != 38)
111 continue;
112 if (memcmp(de->d_name, hex_pfx + 2, len - 2))
113 continue;
114 memcpy(hex + 2, de->d_name, 38);
115 if (!get_sha1_hex(hex, sha1))
116 update_candidates(ds, sha1);
117 }
118 closedir(dir);
119 }
120 }
121
122 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
123 {
124 do {
125 if (*a != *b)
126 return 0;
127 a++;
128 b++;
129 len -= 2;
130 } while (len > 1);
131 if (len)
132 if ((*a ^ *b) & 0xf0)
133 return 0;
134 return 1;
135 }
136
137 static void unique_in_pack(int len,
138 const unsigned char *bin_pfx,
139 struct packed_git *p,
140 struct disambiguate_state *ds)
141 {
142 uint32_t num, last, i, first = 0;
143 const unsigned char *current = NULL;
144
145 open_pack_index(p);
146 num = p->num_objects;
147 last = num;
148 while (first < last) {
149 uint32_t mid = (first + last) / 2;
150 const unsigned char *current;
151 int cmp;
152
153 current = nth_packed_object_sha1(p, mid);
154 cmp = hashcmp(bin_pfx, current);
155 if (!cmp) {
156 first = mid;
157 break;
158 }
159 if (cmp > 0) {
160 first = mid+1;
161 continue;
162 }
163 last = mid;
164 }
165
166 /*
167 * At this point, "first" is the location of the lowest object
168 * with an object name that could match "bin_pfx". See if we have
169 * 0, 1 or more objects that actually match(es).
170 */
171 for (i = first; i < num && !ds->ambiguous; i++) {
172 current = nth_packed_object_sha1(p, i);
173 if (!match_sha(len, bin_pfx, current))
174 break;
175 update_candidates(ds, current);
176 }
177 }
178
179 static void find_short_packed_object(int len, const unsigned char *bin_pfx,
180 struct disambiguate_state *ds)
181 {
182 struct packed_git *p;
183
184 prepare_packed_git();
185 for (p = packed_git; p && !ds->ambiguous; p = p->next)
186 unique_in_pack(len, bin_pfx, p, ds);
187 }
188
189 #define SHORT_NAME_NOT_FOUND (-1)
190 #define SHORT_NAME_AMBIGUOUS (-2)
191
192 static int finish_object_disambiguation(struct disambiguate_state *ds,
193 unsigned char *sha1)
194 {
195 if (ds->ambiguous)
196 return SHORT_NAME_AMBIGUOUS;
197
198 if (!ds->candidate_exists)
199 return SHORT_NAME_NOT_FOUND;
200
201 if (!ds->candidate_checked)
202 /*
203 * If this is the only candidate, there is no point
204 * calling the disambiguation hint callback.
205 *
206 * On the other hand, if the current candidate
207 * replaced an earlier candidate that did _not_ pass
208 * the disambiguation hint callback, then we do have
209 * more than one objects that match the short name
210 * given, so we should make sure this one matches;
211 * otherwise, if we discovered this one and the one
212 * that we previously discarded in the reverse order,
213 * we would end up showing different results in the
214 * same repository!
215 */
216 ds->candidate_ok = (!ds->disambiguate_fn_used ||
217 ds->fn(ds->candidate, ds->cb_data));
218
219 if (!ds->candidate_ok)
220 return SHORT_NAME_AMBIGUOUS;
221
222 hashcpy(sha1, ds->candidate);
223 return 0;
224 }
225
226 static int disambiguate_commit_only(const unsigned char *sha1, void *cb_data_unused)
227 {
228 int kind = sha1_object_info(sha1, NULL);
229 return kind == OBJ_COMMIT;
230 }
231
232 static int disambiguate_committish_only(const unsigned char *sha1, void *cb_data_unused)
233 {
234 struct object *obj;
235 int kind;
236
237 kind = sha1_object_info(sha1, NULL);
238 if (kind == OBJ_COMMIT)
239 return 1;
240 if (kind != OBJ_TAG)
241 return 0;
242
243 /* We need to do this the hard way... */
244 obj = deref_tag(parse_object(sha1), NULL, 0);
245 if (obj && obj->type == OBJ_COMMIT)
246 return 1;
247 return 0;
248 }
249
250 static int disambiguate_tree_only(const unsigned char *sha1, void *cb_data_unused)
251 {
252 int kind = sha1_object_info(sha1, NULL);
253 return kind == OBJ_TREE;
254 }
255
256 static int disambiguate_treeish_only(const unsigned char *sha1, void *cb_data_unused)
257 {
258 struct object *obj;
259 int kind;
260
261 kind = sha1_object_info(sha1, NULL);
262 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
263 return 1;
264 if (kind != OBJ_TAG)
265 return 0;
266
267 /* We need to do this the hard way... */
268 obj = deref_tag(lookup_object(sha1), NULL, 0);
269 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
270 return 1;
271 return 0;
272 }
273
274 static int disambiguate_blob_only(const unsigned char *sha1, void *cb_data_unused)
275 {
276 int kind = sha1_object_info(sha1, NULL);
277 return kind == OBJ_BLOB;
278 }
279
280 static int prepare_prefixes(const char *name, int len,
281 unsigned char *bin_pfx,
282 char *hex_pfx)
283 {
284 int i;
285
286 hashclr(bin_pfx);
287 memset(hex_pfx, 'x', 40);
288 for (i = 0; i < len ;i++) {
289 unsigned char c = name[i];
290 unsigned char val;
291 if (c >= '0' && c <= '9')
292 val = c - '0';
293 else if (c >= 'a' && c <= 'f')
294 val = c - 'a' + 10;
295 else if (c >= 'A' && c <='F') {
296 val = c - 'A' + 10;
297 c -= 'A' - 'a';
298 }
299 else
300 return -1;
301 hex_pfx[i] = c;
302 if (!(i & 1))
303 val <<= 4;
304 bin_pfx[i >> 1] |= val;
305 }
306 return 0;
307 }
308
309 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
310 unsigned flags)
311 {
312 int status;
313 char hex_pfx[40];
314 unsigned char bin_pfx[20];
315 struct disambiguate_state ds;
316 int quietly = !!(flags & GET_SHA1_QUIETLY);
317
318 if (len < MINIMUM_ABBREV || len > 40)
319 return -1;
320 if (prepare_prefixes(name, len, bin_pfx, hex_pfx) < 0)
321 return -1;
322
323 prepare_alt_odb();
324
325 memset(&ds, 0, sizeof(ds));
326 if (flags & GET_SHA1_COMMIT)
327 ds.fn = disambiguate_commit_only;
328 else if (flags & GET_SHA1_COMMITTISH)
329 ds.fn = disambiguate_committish_only;
330 else if (flags & GET_SHA1_TREE)
331 ds.fn = disambiguate_tree_only;
332 else if (flags & GET_SHA1_TREEISH)
333 ds.fn = disambiguate_treeish_only;
334 else if (flags & GET_SHA1_BLOB)
335 ds.fn = disambiguate_blob_only;
336
337 find_short_object_filename(len, hex_pfx, &ds);
338 find_short_packed_object(len, bin_pfx, &ds);
339 status = finish_object_disambiguation(&ds, sha1);
340
341 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
342 return error("short SHA1 %.*s is ambiguous.", len, hex_pfx);
343 return status;
344 }
345
346 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
347 {
348 char hex_pfx[40];
349 unsigned char bin_pfx[20];
350 struct disambiguate_state ds;
351 int len = strlen(prefix);
352
353 if (len < MINIMUM_ABBREV || len > 40)
354 return -1;
355 if (prepare_prefixes(prefix, len, bin_pfx, hex_pfx) < 0)
356 return -1;
357
358 prepare_alt_odb();
359
360 memset(&ds, 0, sizeof(ds));
361 ds.always_call_fn = 1;
362 ds.cb_data = cb_data;
363 ds.fn = fn;
364
365 find_short_object_filename(len, hex_pfx, &ds);
366 find_short_packed_object(len, bin_pfx, &ds);
367 return ds.ambiguous;
368 }
369
370 const char *find_unique_abbrev(const unsigned char *sha1, int len)
371 {
372 int status, exists;
373 static char hex[41];
374
375 exists = has_sha1_file(sha1);
376 memcpy(hex, sha1_to_hex(sha1), 40);
377 if (len == 40 || !len)
378 return hex;
379 while (len < 40) {
380 unsigned char sha1_ret[20];
381 status = get_short_sha1(hex, len, sha1_ret, GET_SHA1_QUIETLY);
382 if (exists
383 ? !status
384 : status == SHORT_NAME_NOT_FOUND) {
385 hex[len] = 0;
386 return hex;
387 }
388 len++;
389 }
390 return hex;
391 }
392
393 static int ambiguous_path(const char *path, int len)
394 {
395 int slash = 1;
396 int cnt;
397
398 for (cnt = 0; cnt < len; cnt++) {
399 switch (*path++) {
400 case '\0':
401 break;
402 case '/':
403 if (slash)
404 break;
405 slash = 1;
406 continue;
407 case '.':
408 continue;
409 default:
410 slash = 0;
411 continue;
412 }
413 break;
414 }
415 return slash;
416 }
417
418 static inline int upstream_mark(const char *string, int len)
419 {
420 const char *suffix[] = { "@{upstream}", "@{u}" };
421 int i;
422
423 for (i = 0; i < ARRAY_SIZE(suffix); i++) {
424 int suffix_len = strlen(suffix[i]);
425 if (suffix_len <= len
426 && !memcmp(string, suffix[i], suffix_len))
427 return suffix_len;
428 }
429 return 0;
430 }
431
432 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
433 static int interpret_nth_prior_checkout(const char *name, int namelen, struct strbuf *buf);
434
435 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
436 {
437 static const char *warn_msg = "refname '%.*s' is ambiguous.";
438 static const char *object_name_msg = N_(
439 "Git normally never creates a ref that ends with 40 hex characters\n"
440 "because it will be ignored when you just specify 40-hex. These refs\n"
441 "may be created by mistake. For example,\n"
442 "\n"
443 " git checkout -b $br $(git rev-parse ...)\n"
444 "\n"
445 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
446 "examine these refs and maybe delete them. Turn this message off by\n"
447 "running \"git config advice.objectNameWarning false\"");
448 unsigned char tmp_sha1[20];
449 char *real_ref = NULL;
450 int refs_found = 0;
451 int at, reflog_len, nth_prior = 0;
452
453 if (len == 40 && !get_sha1_hex(str, sha1)) {
454 if (warn_on_object_refname_ambiguity) {
455 refs_found = dwim_ref(str, len, tmp_sha1, &real_ref);
456 if (refs_found > 0 && warn_ambiguous_refs) {
457 warning(warn_msg, len, str);
458 if (advice_object_name_warning)
459 fprintf(stderr, "%s\n", _(object_name_msg));
460 }
461 free(real_ref);
462 }
463 return 0;
464 }
465
466 /* basic@{time or number or -number} format to query ref-log */
467 reflog_len = at = 0;
468 if (len && str[len-1] == '}') {
469 for (at = len-4; at >= 0; at--) {
470 if (str[at] == '@' && str[at+1] == '{') {
471 if (str[at+2] == '-') {
472 if (at != 0)
473 /* @{-N} not at start */
474 return -1;
475 nth_prior = 1;
476 continue;
477 }
478 if (!upstream_mark(str + at, len - at)) {
479 reflog_len = (len-1) - (at+2);
480 len = at;
481 }
482 break;
483 }
484 }
485 }
486
487 /* Accept only unambiguous ref paths. */
488 if (len && ambiguous_path(str, len))
489 return -1;
490
491 if (nth_prior) {
492 struct strbuf buf = STRBUF_INIT;
493 int detached;
494
495 if (interpret_nth_prior_checkout(str, len, &buf) > 0) {
496 detached = (buf.len == 40 && !get_sha1_hex(buf.buf, sha1));
497 strbuf_release(&buf);
498 if (detached)
499 return 0;
500 }
501 }
502
503 if (!len && reflog_len)
504 /* allow "@{...}" to mean the current branch reflog */
505 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
506 else if (reflog_len)
507 refs_found = dwim_log(str, len, sha1, &real_ref);
508 else
509 refs_found = dwim_ref(str, len, sha1, &real_ref);
510
511 if (!refs_found)
512 return -1;
513
514 if (warn_ambiguous_refs &&
515 (refs_found > 1 ||
516 !get_short_sha1(str, len, tmp_sha1, GET_SHA1_QUIETLY)))
517 warning(warn_msg, len, str);
518
519 if (reflog_len) {
520 int nth, i;
521 unsigned long at_time;
522 unsigned long co_time;
523 int co_tz, co_cnt;
524
525 /* Is it asking for N-th entry, or approxidate? */
526 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
527 char ch = str[at+2+i];
528 if ('0' <= ch && ch <= '9')
529 nth = nth * 10 + ch - '0';
530 else
531 nth = -1;
532 }
533 if (100000000 <= nth) {
534 at_time = nth;
535 nth = -1;
536 } else if (0 <= nth)
537 at_time = 0;
538 else {
539 int errors = 0;
540 char *tmp = xstrndup(str + at + 2, reflog_len);
541 at_time = approxidate_careful(tmp, &errors);
542 free(tmp);
543 if (errors)
544 return -1;
545 }
546 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
547 &co_time, &co_tz, &co_cnt)) {
548 if (!len) {
549 if (!prefixcmp(real_ref, "refs/heads/")) {
550 str = real_ref + 11;
551 len = strlen(real_ref + 11);
552 } else {
553 /* detached HEAD */
554 str = "HEAD";
555 len = 4;
556 }
557 }
558 if (at_time)
559 warning("Log for '%.*s' only goes "
560 "back to %s.", len, str,
561 show_date(co_time, co_tz, DATE_RFC2822));
562 else {
563 die("Log for '%.*s' only has %d entries.",
564 len, str, co_cnt);
565 }
566 }
567 }
568
569 free(real_ref);
570 return 0;
571 }
572
573 static int get_parent(const char *name, int len,
574 unsigned char *result, int idx)
575 {
576 unsigned char sha1[20];
577 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
578 struct commit *commit;
579 struct commit_list *p;
580
581 if (ret)
582 return ret;
583 commit = lookup_commit_reference(sha1);
584 if (!commit)
585 return -1;
586 if (parse_commit(commit))
587 return -1;
588 if (!idx) {
589 hashcpy(result, commit->object.sha1);
590 return 0;
591 }
592 p = commit->parents;
593 while (p) {
594 if (!--idx) {
595 hashcpy(result, p->item->object.sha1);
596 return 0;
597 }
598 p = p->next;
599 }
600 return -1;
601 }
602
603 static int get_nth_ancestor(const char *name, int len,
604 unsigned char *result, int generation)
605 {
606 unsigned char sha1[20];
607 struct commit *commit;
608 int ret;
609
610 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
611 if (ret)
612 return ret;
613 commit = lookup_commit_reference(sha1);
614 if (!commit)
615 return -1;
616
617 while (generation--) {
618 if (parse_commit(commit) || !commit->parents)
619 return -1;
620 commit = commit->parents->item;
621 }
622 hashcpy(result, commit->object.sha1);
623 return 0;
624 }
625
626 struct object *peel_to_type(const char *name, int namelen,
627 struct object *o, enum object_type expected_type)
628 {
629 if (name && !namelen)
630 namelen = strlen(name);
631 while (1) {
632 if (!o || (!o->parsed && !parse_object(o->sha1)))
633 return NULL;
634 if (expected_type == OBJ_ANY || o->type == expected_type)
635 return o;
636 if (o->type == OBJ_TAG)
637 o = ((struct tag*) o)->tagged;
638 else if (o->type == OBJ_COMMIT)
639 o = &(((struct commit *) o)->tree->object);
640 else {
641 if (name)
642 error("%.*s: expected %s type, but the object "
643 "dereferences to %s type",
644 namelen, name, typename(expected_type),
645 typename(o->type));
646 return NULL;
647 }
648 }
649 }
650
651 static int peel_onion(const char *name, int len, unsigned char *sha1)
652 {
653 unsigned char outer[20];
654 const char *sp;
655 unsigned int expected_type = 0;
656 unsigned lookup_flags = 0;
657 struct object *o;
658
659 /*
660 * "ref^{type}" dereferences ref repeatedly until you cannot
661 * dereference anymore, or you get an object of given type,
662 * whichever comes first. "ref^{}" means just dereference
663 * tags until you get a non-tag. "ref^0" is a shorthand for
664 * "ref^{commit}". "commit^{tree}" could be used to find the
665 * top-level tree of the given commit.
666 */
667 if (len < 4 || name[len-1] != '}')
668 return -1;
669
670 for (sp = name + len - 1; name <= sp; sp--) {
671 int ch = *sp;
672 if (ch == '{' && name < sp && sp[-1] == '^')
673 break;
674 }
675 if (sp <= name)
676 return -1;
677
678 sp++; /* beginning of type name, or closing brace for empty */
679 if (!prefixcmp(sp, "commit}"))
680 expected_type = OBJ_COMMIT;
681 else if (!prefixcmp(sp, "tag}"))
682 expected_type = OBJ_TAG;
683 else if (!prefixcmp(sp, "tree}"))
684 expected_type = OBJ_TREE;
685 else if (!prefixcmp(sp, "blob}"))
686 expected_type = OBJ_BLOB;
687 else if (!prefixcmp(sp, "object}"))
688 expected_type = OBJ_ANY;
689 else if (sp[0] == '}')
690 expected_type = OBJ_NONE;
691 else if (sp[0] == '/')
692 expected_type = OBJ_COMMIT;
693 else
694 return -1;
695
696 if (expected_type == OBJ_COMMIT)
697 lookup_flags = GET_SHA1_COMMITTISH;
698 else if (expected_type == OBJ_TREE)
699 lookup_flags = GET_SHA1_TREEISH;
700
701 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
702 return -1;
703
704 o = parse_object(outer);
705 if (!o)
706 return -1;
707 if (!expected_type) {
708 o = deref_tag(o, name, sp - name - 2);
709 if (!o || (!o->parsed && !parse_object(o->sha1)))
710 return -1;
711 hashcpy(sha1, o->sha1);
712 return 0;
713 }
714
715 /*
716 * At this point, the syntax look correct, so
717 * if we do not get the needed object, we should
718 * barf.
719 */
720 o = peel_to_type(name, len, o, expected_type);
721 if (!o)
722 return -1;
723
724 hashcpy(sha1, o->sha1);
725 if (sp[0] == '/') {
726 /* "$commit^{/foo}" */
727 char *prefix;
728 int ret;
729 struct commit_list *list = NULL;
730
731 /*
732 * $commit^{/}. Some regex implementation may reject.
733 * We don't need regex anyway. '' pattern always matches.
734 */
735 if (sp[1] == '}')
736 return 0;
737
738 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
739 commit_list_insert((struct commit *)o, &list);
740 ret = get_sha1_oneline(prefix, sha1, list);
741 free(prefix);
742 return ret;
743 }
744 return 0;
745 }
746
747 static int get_describe_name(const char *name, int len, unsigned char *sha1)
748 {
749 const char *cp;
750 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
751
752 for (cp = name + len - 1; name + 2 <= cp; cp--) {
753 char ch = *cp;
754 if (hexval(ch) & ~0377) {
755 /* We must be looking at g in "SOMETHING-g"
756 * for it to be describe output.
757 */
758 if (ch == 'g' && cp[-1] == '-') {
759 cp++;
760 len -= cp - name;
761 return get_short_sha1(cp, len, sha1, flags);
762 }
763 }
764 }
765 return -1;
766 }
767
768 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
769 {
770 int ret, has_suffix;
771 const char *cp;
772
773 /*
774 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
775 */
776 has_suffix = 0;
777 for (cp = name + len - 1; name <= cp; cp--) {
778 int ch = *cp;
779 if ('0' <= ch && ch <= '9')
780 continue;
781 if (ch == '~' || ch == '^')
782 has_suffix = ch;
783 break;
784 }
785
786 if (has_suffix) {
787 int num = 0;
788 int len1 = cp - name;
789 cp++;
790 while (cp < name + len)
791 num = num * 10 + *cp++ - '0';
792 if (!num && len1 == len - 1)
793 num = 1;
794 if (has_suffix == '^')
795 return get_parent(name, len1, sha1, num);
796 /* else if (has_suffix == '~') -- goes without saying */
797 return get_nth_ancestor(name, len1, sha1, num);
798 }
799
800 ret = peel_onion(name, len, sha1);
801 if (!ret)
802 return 0;
803
804 ret = get_sha1_basic(name, len, sha1);
805 if (!ret)
806 return 0;
807
808 /* It could be describe output that is "SOMETHING-gXXXX" */
809 ret = get_describe_name(name, len, sha1);
810 if (!ret)
811 return 0;
812
813 return get_short_sha1(name, len, sha1, lookup_flags);
814 }
815
816 /*
817 * This interprets names like ':/Initial revision of "git"' by searching
818 * through history and returning the first commit whose message starts
819 * the given regular expression.
820 *
821 * For future extension, ':/!' is reserved. If you want to match a message
822 * beginning with a '!', you have to repeat the exclamation mark.
823 */
824 #define ONELINE_SEEN (1u<<20)
825
826 static int handle_one_ref(const char *path,
827 const unsigned char *sha1, int flag, void *cb_data)
828 {
829 struct commit_list **list = cb_data;
830 struct object *object = parse_object(sha1);
831 if (!object)
832 return 0;
833 if (object->type == OBJ_TAG) {
834 object = deref_tag(object, path, strlen(path));
835 if (!object)
836 return 0;
837 }
838 if (object->type != OBJ_COMMIT)
839 return 0;
840 commit_list_insert_by_date((struct commit *)object, list);
841 return 0;
842 }
843
844 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
845 struct commit_list *list)
846 {
847 struct commit_list *backup = NULL, *l;
848 int found = 0;
849 regex_t regex;
850
851 if (prefix[0] == '!') {
852 if (prefix[1] != '!')
853 die ("Invalid search pattern: %s", prefix);
854 prefix++;
855 }
856
857 if (regcomp(&regex, prefix, REG_EXTENDED))
858 die("Invalid search pattern: %s", prefix);
859
860 for (l = list; l; l = l->next) {
861 l->item->object.flags |= ONELINE_SEEN;
862 commit_list_insert(l->item, &backup);
863 }
864 while (list) {
865 char *p, *to_free = NULL;
866 struct commit *commit;
867 enum object_type type;
868 unsigned long size;
869 int matches;
870
871 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
872 if (!parse_object(commit->object.sha1))
873 continue;
874 if (commit->buffer)
875 p = commit->buffer;
876 else {
877 p = read_sha1_file(commit->object.sha1, &type, &size);
878 if (!p)
879 continue;
880 to_free = p;
881 }
882
883 p = strstr(p, "\n\n");
884 matches = p && !regexec(&regex, p + 2, 0, NULL, 0);
885 free(to_free);
886
887 if (matches) {
888 hashcpy(sha1, commit->object.sha1);
889 found = 1;
890 break;
891 }
892 }
893 regfree(&regex);
894 free_commit_list(list);
895 for (l = backup; l; l = l->next)
896 clear_commit_marks(l->item, ONELINE_SEEN);
897 free_commit_list(backup);
898 return found ? 0 : -1;
899 }
900
901 struct grab_nth_branch_switch_cbdata {
902 int remaining;
903 struct strbuf buf;
904 };
905
906 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
907 const char *email, unsigned long timestamp, int tz,
908 const char *message, void *cb_data)
909 {
910 struct grab_nth_branch_switch_cbdata *cb = cb_data;
911 const char *match = NULL, *target = NULL;
912 size_t len;
913
914 if (!prefixcmp(message, "checkout: moving from ")) {
915 match = message + strlen("checkout: moving from ");
916 target = strstr(match, " to ");
917 }
918
919 if (!match || !target)
920 return 0;
921 if (--(cb->remaining) == 0) {
922 len = target - match;
923 strbuf_reset(&cb->buf);
924 strbuf_add(&cb->buf, match, len);
925 return 1; /* we are done */
926 }
927 return 0;
928 }
929
930 /*
931 * Parse @{-N} syntax, return the number of characters parsed
932 * if successful; otherwise signal an error with negative value.
933 */
934 static int interpret_nth_prior_checkout(const char *name, int namelen,
935 struct strbuf *buf)
936 {
937 long nth;
938 int retval;
939 struct grab_nth_branch_switch_cbdata cb;
940 const char *brace;
941 char *num_end;
942
943 if (namelen < 4)
944 return -1;
945 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
946 return -1;
947 brace = memchr(name, '}', namelen);
948 if (!brace)
949 return -1;
950 nth = strtol(name + 3, &num_end, 10);
951 if (num_end != brace)
952 return -1;
953 if (nth <= 0)
954 return -1;
955 cb.remaining = nth;
956 strbuf_init(&cb.buf, 20);
957
958 retval = 0;
959 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
960 strbuf_reset(buf);
961 strbuf_add(buf, cb.buf.buf, cb.buf.len);
962 retval = brace - name + 1;
963 }
964
965 strbuf_release(&cb.buf);
966 return retval;
967 }
968
969 int get_sha1_mb(const char *name, unsigned char *sha1)
970 {
971 struct commit *one, *two;
972 struct commit_list *mbs;
973 unsigned char sha1_tmp[20];
974 const char *dots;
975 int st;
976
977 dots = strstr(name, "...");
978 if (!dots)
979 return get_sha1(name, sha1);
980 if (dots == name)
981 st = get_sha1("HEAD", sha1_tmp);
982 else {
983 struct strbuf sb;
984 strbuf_init(&sb, dots - name);
985 strbuf_add(&sb, name, dots - name);
986 st = get_sha1_committish(sb.buf, sha1_tmp);
987 strbuf_release(&sb);
988 }
989 if (st)
990 return st;
991 one = lookup_commit_reference_gently(sha1_tmp, 0);
992 if (!one)
993 return -1;
994
995 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", sha1_tmp))
996 return -1;
997 two = lookup_commit_reference_gently(sha1_tmp, 0);
998 if (!two)
999 return -1;
1000 mbs = get_merge_bases(one, two, 1);
1001 if (!mbs || mbs->next)
1002 st = -1;
1003 else {
1004 st = 0;
1005 hashcpy(sha1, mbs->item->object.sha1);
1006 }
1007 free_commit_list(mbs);
1008 return st;
1009 }
1010
1011 /* parse @something syntax, when 'something' is not {.*} */
1012 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1013 {
1014 const char *next;
1015
1016 if (len || name[1] == '{')
1017 return -1;
1018
1019 /* make sure it's a single @, or @@{.*}, not @foo */
1020 next = memchr(name + len + 1, '@', namelen - len - 1);
1021 if (next && next[1] != '{')
1022 return -1;
1023 if (!next)
1024 next = name + namelen;
1025 if (next != name + 1)
1026 return -1;
1027
1028 strbuf_reset(buf);
1029 strbuf_add(buf, "HEAD", 4);
1030 return 1;
1031 }
1032
1033 static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
1034 {
1035 /* we have extra data, which might need further processing */
1036 struct strbuf tmp = STRBUF_INIT;
1037 int used = buf->len;
1038 int ret;
1039
1040 strbuf_add(buf, name + len, namelen - len);
1041 ret = interpret_branch_name(buf->buf, buf->len, &tmp);
1042 /* that data was not interpreted, remove our cruft */
1043 if (ret < 0) {
1044 strbuf_setlen(buf, used);
1045 return len;
1046 }
1047 strbuf_reset(buf);
1048 strbuf_addbuf(buf, &tmp);
1049 strbuf_release(&tmp);
1050 /* tweak for size of {-N} versus expanded ref name */
1051 return ret - used + len;
1052 }
1053
1054 static void set_shortened_ref(struct strbuf *buf, const char *ref)
1055 {
1056 char *s = shorten_unambiguous_ref(ref, 0);
1057 strbuf_reset(buf);
1058 strbuf_addstr(buf, s);
1059 free(s);
1060 }
1061
1062 static const char *get_upstream_branch(const char *branch_buf, int len)
1063 {
1064 char *branch = xstrndup(branch_buf, len);
1065 struct branch *upstream = branch_get(*branch ? branch : NULL);
1066
1067 /*
1068 * Upstream can be NULL only if branch refers to HEAD and HEAD
1069 * points to something different than a branch.
1070 */
1071 if (!upstream)
1072 die(_("HEAD does not point to a branch"));
1073 if (!upstream->merge || !upstream->merge[0]->dst) {
1074 if (!ref_exists(upstream->refname))
1075 die(_("No such branch: '%s'"), branch);
1076 if (!upstream->merge) {
1077 die(_("No upstream configured for branch '%s'"),
1078 upstream->name);
1079 }
1080 die(
1081 _("Upstream branch '%s' not stored as a remote-tracking branch"),
1082 upstream->merge[0]->src);
1083 }
1084 free(branch);
1085
1086 return upstream->merge[0]->dst;
1087 }
1088
1089 static int interpret_upstream_mark(const char *name, int namelen,
1090 int at, struct strbuf *buf)
1091 {
1092 int len;
1093
1094 len = upstream_mark(name + at, namelen - at);
1095 if (!len)
1096 return -1;
1097
1098 if (memchr(name, ':', at))
1099 return -1;
1100
1101 set_shortened_ref(buf, get_upstream_branch(name, at));
1102 return len + at;
1103 }
1104
1105 /*
1106 * This reads short-hand syntax that not only evaluates to a commit
1107 * object name, but also can act as if the end user spelled the name
1108 * of the branch from the command line.
1109 *
1110 * - "@{-N}" finds the name of the Nth previous branch we were on, and
1111 * places the name of the branch in the given buf and returns the
1112 * number of characters parsed if successful.
1113 *
1114 * - "<branch>@{upstream}" finds the name of the other ref that
1115 * <branch> is configured to merge with (missing <branch> defaults
1116 * to the current branch), and places the name of the branch in the
1117 * given buf and returns the number of characters parsed if
1118 * successful.
1119 *
1120 * If the input is not of the accepted format, it returns a negative
1121 * number to signal an error.
1122 *
1123 * If the input was ok but there are not N branch switches in the
1124 * reflog, it returns 0.
1125 */
1126 int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
1127 {
1128 char *at;
1129 int len = interpret_nth_prior_checkout(name, namelen, buf);
1130
1131 if (!namelen)
1132 namelen = strlen(name);
1133
1134 if (!len) {
1135 return len; /* syntax Ok, not enough switches */
1136 } else if (len > 0) {
1137 if (len == namelen)
1138 return len; /* consumed all */
1139 else
1140 return reinterpret(name, namelen, len, buf);
1141 }
1142
1143 at = memchr(name, '@', namelen);
1144 if (!at)
1145 return -1;
1146
1147 len = interpret_empty_at(name, namelen, at - name, buf);
1148 if (len > 0)
1149 return reinterpret(name, namelen, len, buf);
1150
1151 len = interpret_upstream_mark(name, namelen, at - name, buf);
1152 if (len > 0)
1153 return len;
1154
1155 return -1;
1156 }
1157
1158 int strbuf_branchname(struct strbuf *sb, const char *name)
1159 {
1160 int len = strlen(name);
1161 int used = interpret_branch_name(name, len, sb);
1162
1163 if (used == len)
1164 return 0;
1165 if (used < 0)
1166 used = 0;
1167 strbuf_add(sb, name + used, len - used);
1168 return len;
1169 }
1170
1171 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1172 {
1173 strbuf_branchname(sb, name);
1174 if (name[0] == '-')
1175 return -1;
1176 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1177 return check_refname_format(sb->buf, 0);
1178 }
1179
1180 /*
1181 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1182 * notably "xyz^" for "parent of xyz"
1183 */
1184 int get_sha1(const char *name, unsigned char *sha1)
1185 {
1186 struct object_context unused;
1187 return get_sha1_with_context(name, 0, sha1, &unused);
1188 }
1189
1190 /*
1191 * Many callers know that the user meant to name a commit-ish by
1192 * syntactical positions where the object name appears. Calling this
1193 * function allows the machinery to disambiguate shorter-than-unique
1194 * abbreviated object names between commit-ish and others.
1195 *
1196 * Note that this does NOT error out when the named object is not a
1197 * commit-ish. It is merely to give a hint to the disambiguation
1198 * machinery.
1199 */
1200 int get_sha1_committish(const char *name, unsigned char *sha1)
1201 {
1202 struct object_context unused;
1203 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1204 sha1, &unused);
1205 }
1206
1207 int get_sha1_treeish(const char *name, unsigned char *sha1)
1208 {
1209 struct object_context unused;
1210 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1211 sha1, &unused);
1212 }
1213
1214 int get_sha1_commit(const char *name, unsigned char *sha1)
1215 {
1216 struct object_context unused;
1217 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1218 sha1, &unused);
1219 }
1220
1221 int get_sha1_tree(const char *name, unsigned char *sha1)
1222 {
1223 struct object_context unused;
1224 return get_sha1_with_context(name, GET_SHA1_TREE,
1225 sha1, &unused);
1226 }
1227
1228 int get_sha1_blob(const char *name, unsigned char *sha1)
1229 {
1230 struct object_context unused;
1231 return get_sha1_with_context(name, GET_SHA1_BLOB,
1232 sha1, &unused);
1233 }
1234
1235 /* Must be called only when object_name:filename doesn't exist. */
1236 static void diagnose_invalid_sha1_path(const char *prefix,
1237 const char *filename,
1238 const unsigned char *tree_sha1,
1239 const char *object_name,
1240 int object_name_len)
1241 {
1242 struct stat st;
1243 unsigned char sha1[20];
1244 unsigned mode;
1245
1246 if (!prefix)
1247 prefix = "";
1248
1249 if (!lstat(filename, &st))
1250 die("Path '%s' exists on disk, but not in '%.*s'.",
1251 filename, object_name_len, object_name);
1252 if (errno == ENOENT || errno == ENOTDIR) {
1253 char *fullname = xmalloc(strlen(filename)
1254 + strlen(prefix) + 1);
1255 strcpy(fullname, prefix);
1256 strcat(fullname, filename);
1257
1258 if (!get_tree_entry(tree_sha1, fullname,
1259 sha1, &mode)) {
1260 die("Path '%s' exists, but not '%s'.\n"
1261 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1262 fullname,
1263 filename,
1264 object_name_len, object_name,
1265 fullname,
1266 object_name_len, object_name,
1267 filename);
1268 }
1269 die("Path '%s' does not exist in '%.*s'",
1270 filename, object_name_len, object_name);
1271 }
1272 }
1273
1274 /* Must be called only when :stage:filename doesn't exist. */
1275 static void diagnose_invalid_index_path(int stage,
1276 const char *prefix,
1277 const char *filename)
1278 {
1279 struct stat st;
1280 const struct cache_entry *ce;
1281 int pos;
1282 unsigned namelen = strlen(filename);
1283 unsigned fullnamelen;
1284 char *fullname;
1285
1286 if (!prefix)
1287 prefix = "";
1288
1289 /* Wrong stage number? */
1290 pos = cache_name_pos(filename, namelen);
1291 if (pos < 0)
1292 pos = -pos - 1;
1293 if (pos < active_nr) {
1294 ce = active_cache[pos];
1295 if (ce_namelen(ce) == namelen &&
1296 !memcmp(ce->name, filename, namelen))
1297 die("Path '%s' is in the index, but not at stage %d.\n"
1298 "Did you mean ':%d:%s'?",
1299 filename, stage,
1300 ce_stage(ce), filename);
1301 }
1302
1303 /* Confusion between relative and absolute filenames? */
1304 fullnamelen = namelen + strlen(prefix);
1305 fullname = xmalloc(fullnamelen + 1);
1306 strcpy(fullname, prefix);
1307 strcat(fullname, filename);
1308 pos = cache_name_pos(fullname, fullnamelen);
1309 if (pos < 0)
1310 pos = -pos - 1;
1311 if (pos < active_nr) {
1312 ce = active_cache[pos];
1313 if (ce_namelen(ce) == fullnamelen &&
1314 !memcmp(ce->name, fullname, fullnamelen))
1315 die("Path '%s' is in the index, but not '%s'.\n"
1316 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1317 fullname, filename,
1318 ce_stage(ce), fullname,
1319 ce_stage(ce), filename);
1320 }
1321
1322 if (!lstat(filename, &st))
1323 die("Path '%s' exists on disk, but not in the index.", filename);
1324 if (errno == ENOENT || errno == ENOTDIR)
1325 die("Path '%s' does not exist (neither on disk nor in the index).",
1326 filename);
1327
1328 free(fullname);
1329 }
1330
1331
1332 static char *resolve_relative_path(const char *rel)
1333 {
1334 if (prefixcmp(rel, "./") && prefixcmp(rel, "../"))
1335 return NULL;
1336
1337 if (!startup_info)
1338 die("BUG: startup_info struct is not initialized.");
1339
1340 if (!is_inside_work_tree())
1341 die("relative path syntax can't be used outside working tree.");
1342
1343 /* die() inside prefix_path() if resolved path is outside worktree */
1344 return prefix_path(startup_info->prefix,
1345 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1346 rel);
1347 }
1348
1349 static int get_sha1_with_context_1(const char *name,
1350 unsigned flags,
1351 const char *prefix,
1352 unsigned char *sha1,
1353 struct object_context *oc)
1354 {
1355 int ret, bracket_depth;
1356 int namelen = strlen(name);
1357 const char *cp;
1358 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1359
1360 memset(oc, 0, sizeof(*oc));
1361 oc->mode = S_IFINVALID;
1362 ret = get_sha1_1(name, namelen, sha1, flags);
1363 if (!ret)
1364 return ret;
1365 /*
1366 * sha1:path --> object name of path in ent sha1
1367 * :path -> object name of absolute path in index
1368 * :./path -> object name of path relative to cwd in index
1369 * :[0-3]:path -> object name of path in index at stage
1370 * :/foo -> recent commit matching foo
1371 */
1372 if (name[0] == ':') {
1373 int stage = 0;
1374 const struct cache_entry *ce;
1375 char *new_path = NULL;
1376 int pos;
1377 if (!only_to_die && namelen > 2 && name[1] == '/') {
1378 struct commit_list *list = NULL;
1379 for_each_ref(handle_one_ref, &list);
1380 return get_sha1_oneline(name + 2, sha1, list);
1381 }
1382 if (namelen < 3 ||
1383 name[2] != ':' ||
1384 name[1] < '0' || '3' < name[1])
1385 cp = name + 1;
1386 else {
1387 stage = name[1] - '0';
1388 cp = name + 3;
1389 }
1390 new_path = resolve_relative_path(cp);
1391 if (!new_path) {
1392 namelen = namelen - (cp - name);
1393 } else {
1394 cp = new_path;
1395 namelen = strlen(cp);
1396 }
1397
1398 strncpy(oc->path, cp,
1399 sizeof(oc->path));
1400 oc->path[sizeof(oc->path)-1] = '\0';
1401
1402 if (!active_cache)
1403 read_cache();
1404 pos = cache_name_pos(cp, namelen);
1405 if (pos < 0)
1406 pos = -pos - 1;
1407 while (pos < active_nr) {
1408 ce = active_cache[pos];
1409 if (ce_namelen(ce) != namelen ||
1410 memcmp(ce->name, cp, namelen))
1411 break;
1412 if (ce_stage(ce) == stage) {
1413 hashcpy(sha1, ce->sha1);
1414 oc->mode = ce->ce_mode;
1415 free(new_path);
1416 return 0;
1417 }
1418 pos++;
1419 }
1420 if (only_to_die && name[1] && name[1] != '/')
1421 diagnose_invalid_index_path(stage, prefix, cp);
1422 free(new_path);
1423 return -1;
1424 }
1425 for (cp = name, bracket_depth = 0; *cp; cp++) {
1426 if (*cp == '{')
1427 bracket_depth++;
1428 else if (bracket_depth && *cp == '}')
1429 bracket_depth--;
1430 else if (!bracket_depth && *cp == ':')
1431 break;
1432 }
1433 if (*cp == ':') {
1434 unsigned char tree_sha1[20];
1435 int len = cp - name;
1436 if (!get_sha1_1(name, len, tree_sha1, GET_SHA1_TREEISH)) {
1437 const char *filename = cp+1;
1438 char *new_filename = NULL;
1439
1440 new_filename = resolve_relative_path(filename);
1441 if (new_filename)
1442 filename = new_filename;
1443 ret = get_tree_entry(tree_sha1, filename, sha1, &oc->mode);
1444 if (ret && only_to_die) {
1445 diagnose_invalid_sha1_path(prefix, filename,
1446 tree_sha1,
1447 name, len);
1448 }
1449 hashcpy(oc->tree, tree_sha1);
1450 strncpy(oc->path, filename,
1451 sizeof(oc->path));
1452 oc->path[sizeof(oc->path)-1] = '\0';
1453
1454 free(new_filename);
1455 return ret;
1456 } else {
1457 if (only_to_die)
1458 die("Invalid object name '%.*s'.", len, name);
1459 }
1460 }
1461 return ret;
1462 }
1463
1464 /*
1465 * Call this function when you know "name" given by the end user must
1466 * name an object but it doesn't; the function _may_ die with a better
1467 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1468 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1469 * you have a chance to diagnose the error further.
1470 */
1471 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1472 {
1473 struct object_context oc;
1474 unsigned char sha1[20];
1475 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1476 }
1477
1478 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1479 {
1480 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);
1481 }