assume parse_commit checks for NULL commit
[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
347 int for_each_abbrev(const char *prefix, each_abbrev_fn fn, void *cb_data)
348 {
349 char hex_pfx[40];
350 unsigned char bin_pfx[20];
351 struct disambiguate_state ds;
352 int len = strlen(prefix);
353
354 if (len < MINIMUM_ABBREV || len > 40)
355 return -1;
356 if (prepare_prefixes(prefix, len, bin_pfx, hex_pfx) < 0)
357 return -1;
358
359 prepare_alt_odb();
360
361 memset(&ds, 0, sizeof(ds));
362 ds.always_call_fn = 1;
363 ds.cb_data = cb_data;
364 ds.fn = fn;
365
366 find_short_object_filename(len, hex_pfx, &ds);
367 find_short_packed_object(len, bin_pfx, &ds);
368 return ds.ambiguous;
369 }
370
371 const char *find_unique_abbrev(const unsigned char *sha1, int len)
372 {
373 int status, exists;
374 static char hex[41];
375
376 exists = has_sha1_file(sha1);
377 memcpy(hex, sha1_to_hex(sha1), 40);
378 if (len == 40 || !len)
379 return hex;
380 while (len < 40) {
381 unsigned char sha1_ret[20];
382 status = get_short_sha1(hex, len, sha1_ret, GET_SHA1_QUIETLY);
383 if (exists
384 ? !status
385 : status == SHORT_NAME_NOT_FOUND) {
386 hex[len] = 0;
387 return hex;
388 }
389 len++;
390 }
391 return hex;
392 }
393
394 static int ambiguous_path(const char *path, int len)
395 {
396 int slash = 1;
397 int cnt;
398
399 for (cnt = 0; cnt < len; cnt++) {
400 switch (*path++) {
401 case '\0':
402 break;
403 case '/':
404 if (slash)
405 break;
406 slash = 1;
407 continue;
408 case '.':
409 continue;
410 default:
411 slash = 0;
412 continue;
413 }
414 break;
415 }
416 return slash;
417 }
418
419 static inline int upstream_mark(const char *string, int len)
420 {
421 const char *suffix[] = { "@{upstream}", "@{u}" };
422 int i;
423
424 for (i = 0; i < ARRAY_SIZE(suffix); i++) {
425 int suffix_len = strlen(suffix[i]);
426 if (suffix_len <= len
427 && !memcmp(string, suffix[i], suffix_len))
428 return suffix_len;
429 }
430 return 0;
431 }
432
433 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
434 static int interpret_nth_prior_checkout(const char *name, struct strbuf *buf);
435
436 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
437 {
438 static const char *warn_msg = "refname '%.*s' is ambiguous.";
439 static const char *object_name_msg = N_(
440 "Git normally never creates a ref that ends with 40 hex characters\n"
441 "because it will be ignored when you just specify 40-hex. These refs\n"
442 "may be created by mistake. For example,\n"
443 "\n"
444 " git checkout -b $br $(git rev-parse ...)\n"
445 "\n"
446 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
447 "examine these refs and maybe delete them. Turn this message off by\n"
448 "running \"git config advice.objectNameWarning false\"");
449 unsigned char tmp_sha1[20];
450 char *real_ref = NULL;
451 int refs_found = 0;
452 int at, reflog_len, nth_prior = 0;
453
454 if (len == 40 && !get_sha1_hex(str, sha1)) {
455 if (warn_on_object_refname_ambiguity) {
456 refs_found = dwim_ref(str, len, tmp_sha1, &real_ref);
457 if (refs_found > 0 && warn_ambiguous_refs) {
458 warning(warn_msg, len, str);
459 if (advice_object_name_warning)
460 fprintf(stderr, "%s\n", _(object_name_msg));
461 }
462 free(real_ref);
463 }
464 return 0;
465 }
466
467 /* basic@{time or number or -number} format to query ref-log */
468 reflog_len = at = 0;
469 if (len && str[len-1] == '}') {
470 for (at = len-4; at >= 0; at--) {
471 if (str[at] == '@' && str[at+1] == '{') {
472 if (str[at+2] == '-') {
473 if (at != 0)
474 /* @{-N} not at start */
475 return -1;
476 nth_prior = 1;
477 continue;
478 }
479 if (!upstream_mark(str + at, len - at)) {
480 reflog_len = (len-1) - (at+2);
481 len = at;
482 }
483 break;
484 }
485 }
486 }
487
488 /* Accept only unambiguous ref paths. */
489 if (len && ambiguous_path(str, len))
490 return -1;
491
492 if (nth_prior) {
493 struct strbuf buf = STRBUF_INIT;
494 int detached;
495
496 if (interpret_nth_prior_checkout(str, &buf) > 0) {
497 detached = (buf.len == 40 && !get_sha1_hex(buf.buf, sha1));
498 strbuf_release(&buf);
499 if (detached)
500 return 0;
501 }
502 }
503
504 if (!len && reflog_len)
505 /* allow "@{...}" to mean the current branch reflog */
506 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
507 else if (reflog_len)
508 refs_found = dwim_log(str, len, sha1, &real_ref);
509 else
510 refs_found = dwim_ref(str, len, sha1, &real_ref);
511
512 if (!refs_found)
513 return -1;
514
515 if (warn_ambiguous_refs &&
516 (refs_found > 1 ||
517 !get_short_sha1(str, len, tmp_sha1, GET_SHA1_QUIETLY)))
518 warning(warn_msg, len, str);
519
520 if (reflog_len) {
521 int nth, i;
522 unsigned long at_time;
523 unsigned long co_time;
524 int co_tz, co_cnt;
525
526 /* Is it asking for N-th entry, or approxidate? */
527 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
528 char ch = str[at+2+i];
529 if ('0' <= ch && ch <= '9')
530 nth = nth * 10 + ch - '0';
531 else
532 nth = -1;
533 }
534 if (100000000 <= nth) {
535 at_time = nth;
536 nth = -1;
537 } else if (0 <= nth)
538 at_time = 0;
539 else {
540 int errors = 0;
541 char *tmp = xstrndup(str + at + 2, reflog_len);
542 at_time = approxidate_careful(tmp, &errors);
543 free(tmp);
544 if (errors)
545 return -1;
546 }
547 if (read_ref_at(real_ref, at_time, nth, sha1, NULL,
548 &co_time, &co_tz, &co_cnt)) {
549 if (!len) {
550 if (!prefixcmp(real_ref, "refs/heads/")) {
551 str = real_ref + 11;
552 len = strlen(real_ref + 11);
553 } else {
554 /* detached HEAD */
555 str = "HEAD";
556 len = 4;
557 }
558 }
559 if (at_time)
560 warning("Log for '%.*s' only goes "
561 "back to %s.", len, str,
562 show_date(co_time, co_tz, DATE_RFC2822));
563 else {
564 die("Log for '%.*s' only has %d entries.",
565 len, str, co_cnt);
566 }
567 }
568 }
569
570 free(real_ref);
571 return 0;
572 }
573
574 static int get_parent(const char *name, int len,
575 unsigned char *result, int idx)
576 {
577 unsigned char sha1[20];
578 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
579 struct commit *commit;
580 struct commit_list *p;
581
582 if (ret)
583 return ret;
584 commit = lookup_commit_reference(sha1);
585 if (parse_commit(commit))
586 return -1;
587 if (!idx) {
588 hashcpy(result, commit->object.sha1);
589 return 0;
590 }
591 p = commit->parents;
592 while (p) {
593 if (!--idx) {
594 hashcpy(result, p->item->object.sha1);
595 return 0;
596 }
597 p = p->next;
598 }
599 return -1;
600 }
601
602 static int get_nth_ancestor(const char *name, int len,
603 unsigned char *result, int generation)
604 {
605 unsigned char sha1[20];
606 struct commit *commit;
607 int ret;
608
609 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
610 if (ret)
611 return ret;
612 commit = lookup_commit_reference(sha1);
613 if (!commit)
614 return -1;
615
616 while (generation--) {
617 if (parse_commit(commit) || !commit->parents)
618 return -1;
619 commit = commit->parents->item;
620 }
621 hashcpy(result, commit->object.sha1);
622 return 0;
623 }
624
625 struct object *peel_to_type(const char *name, int namelen,
626 struct object *o, enum object_type expected_type)
627 {
628 if (name && !namelen)
629 namelen = strlen(name);
630 while (1) {
631 if (!o || (!o->parsed && !parse_object(o->sha1)))
632 return NULL;
633 if (expected_type == OBJ_ANY || o->type == expected_type)
634 return o;
635 if (o->type == OBJ_TAG)
636 o = ((struct tag*) o)->tagged;
637 else if (o->type == OBJ_COMMIT)
638 o = &(((struct commit *) o)->tree->object);
639 else {
640 if (name)
641 error("%.*s: expected %s type, but the object "
642 "dereferences to %s type",
643 namelen, name, typename(expected_type),
644 typename(o->type));
645 return NULL;
646 }
647 }
648 }
649
650 static int peel_onion(const char *name, int len, unsigned char *sha1)
651 {
652 unsigned char outer[20];
653 const char *sp;
654 unsigned int expected_type = 0;
655 unsigned lookup_flags = 0;
656 struct object *o;
657
658 /*
659 * "ref^{type}" dereferences ref repeatedly until you cannot
660 * dereference anymore, or you get an object of given type,
661 * whichever comes first. "ref^{}" means just dereference
662 * tags until you get a non-tag. "ref^0" is a shorthand for
663 * "ref^{commit}". "commit^{tree}" could be used to find the
664 * top-level tree of the given commit.
665 */
666 if (len < 4 || name[len-1] != '}')
667 return -1;
668
669 for (sp = name + len - 1; name <= sp; sp--) {
670 int ch = *sp;
671 if (ch == '{' && name < sp && sp[-1] == '^')
672 break;
673 }
674 if (sp <= name)
675 return -1;
676
677 sp++; /* beginning of type name, or closing brace for empty */
678 if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
679 expected_type = OBJ_COMMIT;
680 else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
681 expected_type = OBJ_TREE;
682 else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
683 expected_type = OBJ_BLOB;
684 else if (!prefixcmp(sp, "object}"))
685 expected_type = OBJ_ANY;
686 else if (sp[0] == '}')
687 expected_type = OBJ_NONE;
688 else if (sp[0] == '/')
689 expected_type = OBJ_COMMIT;
690 else
691 return -1;
692
693 if (expected_type == OBJ_COMMIT)
694 lookup_flags = GET_SHA1_COMMITTISH;
695 else if (expected_type == OBJ_TREE)
696 lookup_flags = GET_SHA1_TREEISH;
697
698 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
699 return -1;
700
701 o = parse_object(outer);
702 if (!o)
703 return -1;
704 if (!expected_type) {
705 o = deref_tag(o, name, sp - name - 2);
706 if (!o || (!o->parsed && !parse_object(o->sha1)))
707 return -1;
708 hashcpy(sha1, o->sha1);
709 return 0;
710 }
711
712 /*
713 * At this point, the syntax look correct, so
714 * if we do not get the needed object, we should
715 * barf.
716 */
717 o = peel_to_type(name, len, o, expected_type);
718 if (!o)
719 return -1;
720
721 hashcpy(sha1, o->sha1);
722 if (sp[0] == '/') {
723 /* "$commit^{/foo}" */
724 char *prefix;
725 int ret;
726 struct commit_list *list = NULL;
727
728 /*
729 * $commit^{/}. Some regex implementation may reject.
730 * We don't need regex anyway. '' pattern always matches.
731 */
732 if (sp[1] == '}')
733 return 0;
734
735 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
736 commit_list_insert((struct commit *)o, &list);
737 ret = get_sha1_oneline(prefix, sha1, list);
738 free(prefix);
739 return ret;
740 }
741 return 0;
742 }
743
744 static int get_describe_name(const char *name, int len, unsigned char *sha1)
745 {
746 const char *cp;
747 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
748
749 for (cp = name + len - 1; name + 2 <= cp; cp--) {
750 char ch = *cp;
751 if (hexval(ch) & ~0377) {
752 /* We must be looking at g in "SOMETHING-g"
753 * for it to be describe output.
754 */
755 if (ch == 'g' && cp[-1] == '-') {
756 cp++;
757 len -= cp - name;
758 return get_short_sha1(cp, len, sha1, flags);
759 }
760 }
761 }
762 return -1;
763 }
764
765 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
766 {
767 int ret, has_suffix;
768 const char *cp;
769
770 /*
771 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
772 */
773 has_suffix = 0;
774 for (cp = name + len - 1; name <= cp; cp--) {
775 int ch = *cp;
776 if ('0' <= ch && ch <= '9')
777 continue;
778 if (ch == '~' || ch == '^')
779 has_suffix = ch;
780 break;
781 }
782
783 if (has_suffix) {
784 int num = 0;
785 int len1 = cp - name;
786 cp++;
787 while (cp < name + len)
788 num = num * 10 + *cp++ - '0';
789 if (!num && len1 == len - 1)
790 num = 1;
791 if (has_suffix == '^')
792 return get_parent(name, len1, sha1, num);
793 /* else if (has_suffix == '~') -- goes without saying */
794 return get_nth_ancestor(name, len1, sha1, num);
795 }
796
797 ret = peel_onion(name, len, sha1);
798 if (!ret)
799 return 0;
800
801 ret = get_sha1_basic(name, len, sha1);
802 if (!ret)
803 return 0;
804
805 /* It could be describe output that is "SOMETHING-gXXXX" */
806 ret = get_describe_name(name, len, sha1);
807 if (!ret)
808 return 0;
809
810 return get_short_sha1(name, len, sha1, lookup_flags);
811 }
812
813 /*
814 * This interprets names like ':/Initial revision of "git"' by searching
815 * through history and returning the first commit whose message starts
816 * the given regular expression.
817 *
818 * For future extension, ':/!' is reserved. If you want to match a message
819 * beginning with a '!', you have to repeat the exclamation mark.
820 */
821 #define ONELINE_SEEN (1u<<20)
822
823 static int handle_one_ref(const char *path,
824 const unsigned char *sha1, int flag, void *cb_data)
825 {
826 struct commit_list **list = cb_data;
827 struct object *object = parse_object(sha1);
828 if (!object)
829 return 0;
830 if (object->type == OBJ_TAG) {
831 object = deref_tag(object, path, strlen(path));
832 if (!object)
833 return 0;
834 }
835 if (object->type != OBJ_COMMIT)
836 return 0;
837 commit_list_insert_by_date((struct commit *)object, list);
838 return 0;
839 }
840
841 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
842 struct commit_list *list)
843 {
844 struct commit_list *backup = NULL, *l;
845 int found = 0;
846 regex_t regex;
847
848 if (prefix[0] == '!') {
849 if (prefix[1] != '!')
850 die ("Invalid search pattern: %s", prefix);
851 prefix++;
852 }
853
854 if (regcomp(&regex, prefix, REG_EXTENDED))
855 die("Invalid search pattern: %s", prefix);
856
857 for (l = list; l; l = l->next) {
858 l->item->object.flags |= ONELINE_SEEN;
859 commit_list_insert(l->item, &backup);
860 }
861 while (list) {
862 char *p, *to_free = NULL;
863 struct commit *commit;
864 enum object_type type;
865 unsigned long size;
866 int matches;
867
868 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
869 if (!parse_object(commit->object.sha1))
870 continue;
871 if (commit->buffer)
872 p = commit->buffer;
873 else {
874 p = read_sha1_file(commit->object.sha1, &type, &size);
875 if (!p)
876 continue;
877 to_free = p;
878 }
879
880 p = strstr(p, "\n\n");
881 matches = p && !regexec(&regex, p + 2, 0, NULL, 0);
882 free(to_free);
883
884 if (matches) {
885 hashcpy(sha1, commit->object.sha1);
886 found = 1;
887 break;
888 }
889 }
890 regfree(&regex);
891 free_commit_list(list);
892 for (l = backup; l; l = l->next)
893 clear_commit_marks(l->item, ONELINE_SEEN);
894 free_commit_list(backup);
895 return found ? 0 : -1;
896 }
897
898 struct grab_nth_branch_switch_cbdata {
899 int remaining;
900 struct strbuf buf;
901 };
902
903 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
904 const char *email, unsigned long timestamp, int tz,
905 const char *message, void *cb_data)
906 {
907 struct grab_nth_branch_switch_cbdata *cb = cb_data;
908 const char *match = NULL, *target = NULL;
909 size_t len;
910
911 if (!prefixcmp(message, "checkout: moving from ")) {
912 match = message + strlen("checkout: moving from ");
913 target = strstr(match, " to ");
914 }
915
916 if (!match || !target)
917 return 0;
918 if (--(cb->remaining) == 0) {
919 len = target - match;
920 strbuf_reset(&cb->buf);
921 strbuf_add(&cb->buf, match, len);
922 return 1; /* we are done */
923 }
924 return 0;
925 }
926
927 /*
928 * Parse @{-N} syntax, return the number of characters parsed
929 * if successful; otherwise signal an error with negative value.
930 */
931 static int interpret_nth_prior_checkout(const char *name, struct strbuf *buf)
932 {
933 long nth;
934 int retval;
935 struct grab_nth_branch_switch_cbdata cb;
936 const char *brace;
937 char *num_end;
938
939 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
940 return -1;
941 brace = strchr(name, '}');
942 if (!brace)
943 return -1;
944 nth = strtol(name + 3, &num_end, 10);
945 if (num_end != brace)
946 return -1;
947 if (nth <= 0)
948 return -1;
949 cb.remaining = nth;
950 strbuf_init(&cb.buf, 20);
951
952 retval = 0;
953 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
954 strbuf_reset(buf);
955 strbuf_add(buf, cb.buf.buf, cb.buf.len);
956 retval = brace - name + 1;
957 }
958
959 strbuf_release(&cb.buf);
960 return retval;
961 }
962
963 int get_sha1_mb(const char *name, unsigned char *sha1)
964 {
965 struct commit *one, *two;
966 struct commit_list *mbs;
967 unsigned char sha1_tmp[20];
968 const char *dots;
969 int st;
970
971 dots = strstr(name, "...");
972 if (!dots)
973 return get_sha1(name, sha1);
974 if (dots == name)
975 st = get_sha1("HEAD", sha1_tmp);
976 else {
977 struct strbuf sb;
978 strbuf_init(&sb, dots - name);
979 strbuf_add(&sb, name, dots - name);
980 st = get_sha1_committish(sb.buf, sha1_tmp);
981 strbuf_release(&sb);
982 }
983 if (st)
984 return st;
985 one = lookup_commit_reference_gently(sha1_tmp, 0);
986 if (!one)
987 return -1;
988
989 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", sha1_tmp))
990 return -1;
991 two = lookup_commit_reference_gently(sha1_tmp, 0);
992 if (!two)
993 return -1;
994 mbs = get_merge_bases(one, two, 1);
995 if (!mbs || mbs->next)
996 st = -1;
997 else {
998 st = 0;
999 hashcpy(sha1, mbs->item->object.sha1);
1000 }
1001 free_commit_list(mbs);
1002 return st;
1003 }
1004
1005 static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
1006 {
1007 /* we have extra data, which might need further processing */
1008 struct strbuf tmp = STRBUF_INIT;
1009 int used = buf->len;
1010 int ret;
1011
1012 strbuf_add(buf, name + len, namelen - len);
1013 ret = interpret_branch_name(buf->buf, &tmp);
1014 /* that data was not interpreted, remove our cruft */
1015 if (ret < 0) {
1016 strbuf_setlen(buf, used);
1017 return len;
1018 }
1019 strbuf_reset(buf);
1020 strbuf_addbuf(buf, &tmp);
1021 strbuf_release(&tmp);
1022 /* tweak for size of {-N} versus expanded ref name */
1023 return ret - used + len;
1024 }
1025
1026 /*
1027 * This reads short-hand syntax that not only evaluates to a commit
1028 * object name, but also can act as if the end user spelled the name
1029 * of the branch from the command line.
1030 *
1031 * - "@{-N}" finds the name of the Nth previous branch we were on, and
1032 * places the name of the branch in the given buf and returns the
1033 * number of characters parsed if successful.
1034 *
1035 * - "<branch>@{upstream}" finds the name of the other ref that
1036 * <branch> is configured to merge with (missing <branch> defaults
1037 * to the current branch), and places the name of the branch in the
1038 * given buf and returns the number of characters parsed if
1039 * successful.
1040 *
1041 * If the input is not of the accepted format, it returns a negative
1042 * number to signal an error.
1043 *
1044 * If the input was ok but there are not N branch switches in the
1045 * reflog, it returns 0.
1046 */
1047 int interpret_branch_name(const char *name, struct strbuf *buf)
1048 {
1049 char *cp;
1050 struct branch *upstream;
1051 int namelen = strlen(name);
1052 int len = interpret_nth_prior_checkout(name, buf);
1053 int tmp_len;
1054
1055 if (!len) {
1056 return len; /* syntax Ok, not enough switches */
1057 } else if (len > 0) {
1058 if (len == namelen)
1059 return len; /* consumed all */
1060 else
1061 return reinterpret(name, namelen, len, buf);
1062 }
1063
1064 cp = strchr(name, '@');
1065 if (!cp)
1066 return -1;
1067 tmp_len = upstream_mark(cp, namelen - (cp - name));
1068 if (!tmp_len)
1069 return -1;
1070 len = cp + tmp_len - name;
1071 cp = xstrndup(name, cp - name);
1072 upstream = branch_get(*cp ? cp : NULL);
1073 /*
1074 * Upstream can be NULL only if cp refers to HEAD and HEAD
1075 * points to something different than a branch.
1076 */
1077 if (!upstream)
1078 die(_("HEAD does not point to a branch"));
1079 if (!upstream->merge || !upstream->merge[0]->dst) {
1080 if (!ref_exists(upstream->refname))
1081 die(_("No such branch: '%s'"), cp);
1082 if (!upstream->merge) {
1083 die(_("No upstream configured for branch '%s'"),
1084 upstream->name);
1085 }
1086 die(
1087 _("Upstream branch '%s' not stored as a remote-tracking branch"),
1088 upstream->merge[0]->src);
1089 }
1090 free(cp);
1091 cp = shorten_unambiguous_ref(upstream->merge[0]->dst, 0);
1092 strbuf_reset(buf);
1093 strbuf_addstr(buf, cp);
1094 free(cp);
1095 return len;
1096 }
1097
1098 int strbuf_branchname(struct strbuf *sb, const char *name)
1099 {
1100 int len = strlen(name);
1101 int used = interpret_branch_name(name, sb);
1102
1103 if (used == len)
1104 return 0;
1105 if (used < 0)
1106 used = 0;
1107 strbuf_add(sb, name + used, len - used);
1108 return len;
1109 }
1110
1111 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1112 {
1113 strbuf_branchname(sb, name);
1114 if (name[0] == '-')
1115 return -1;
1116 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1117 return check_refname_format(sb->buf, 0);
1118 }
1119
1120 /*
1121 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1122 * notably "xyz^" for "parent of xyz"
1123 */
1124 int get_sha1(const char *name, unsigned char *sha1)
1125 {
1126 struct object_context unused;
1127 return get_sha1_with_context(name, 0, sha1, &unused);
1128 }
1129
1130 /*
1131 * Many callers know that the user meant to name a committish by
1132 * syntactical positions where the object name appears. Calling this
1133 * function allows the machinery to disambiguate shorter-than-unique
1134 * abbreviated object names between committish and others.
1135 *
1136 * Note that this does NOT error out when the named object is not a
1137 * committish. It is merely to give a hint to the disambiguation
1138 * machinery.
1139 */
1140 int get_sha1_committish(const char *name, unsigned char *sha1)
1141 {
1142 struct object_context unused;
1143 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1144 sha1, &unused);
1145 }
1146
1147 int get_sha1_treeish(const char *name, unsigned char *sha1)
1148 {
1149 struct object_context unused;
1150 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1151 sha1, &unused);
1152 }
1153
1154 int get_sha1_commit(const char *name, unsigned char *sha1)
1155 {
1156 struct object_context unused;
1157 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1158 sha1, &unused);
1159 }
1160
1161 int get_sha1_tree(const char *name, unsigned char *sha1)
1162 {
1163 struct object_context unused;
1164 return get_sha1_with_context(name, GET_SHA1_TREE,
1165 sha1, &unused);
1166 }
1167
1168 int get_sha1_blob(const char *name, unsigned char *sha1)
1169 {
1170 struct object_context unused;
1171 return get_sha1_with_context(name, GET_SHA1_BLOB,
1172 sha1, &unused);
1173 }
1174
1175 /* Must be called only when object_name:filename doesn't exist. */
1176 static void diagnose_invalid_sha1_path(const char *prefix,
1177 const char *filename,
1178 const unsigned char *tree_sha1,
1179 const char *object_name,
1180 int object_name_len)
1181 {
1182 struct stat st;
1183 unsigned char sha1[20];
1184 unsigned mode;
1185
1186 if (!prefix)
1187 prefix = "";
1188
1189 if (!lstat(filename, &st))
1190 die("Path '%s' exists on disk, but not in '%.*s'.",
1191 filename, object_name_len, object_name);
1192 if (errno == ENOENT || errno == ENOTDIR) {
1193 char *fullname = xmalloc(strlen(filename)
1194 + strlen(prefix) + 1);
1195 strcpy(fullname, prefix);
1196 strcat(fullname, filename);
1197
1198 if (!get_tree_entry(tree_sha1, fullname,
1199 sha1, &mode)) {
1200 die("Path '%s' exists, but not '%s'.\n"
1201 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1202 fullname,
1203 filename,
1204 object_name_len, object_name,
1205 fullname,
1206 object_name_len, object_name,
1207 filename);
1208 }
1209 die("Path '%s' does not exist in '%.*s'",
1210 filename, object_name_len, object_name);
1211 }
1212 }
1213
1214 /* Must be called only when :stage:filename doesn't exist. */
1215 static void diagnose_invalid_index_path(int stage,
1216 const char *prefix,
1217 const char *filename)
1218 {
1219 struct stat st;
1220 const struct cache_entry *ce;
1221 int pos;
1222 unsigned namelen = strlen(filename);
1223 unsigned fullnamelen;
1224 char *fullname;
1225
1226 if (!prefix)
1227 prefix = "";
1228
1229 /* Wrong stage number? */
1230 pos = cache_name_pos(filename, namelen);
1231 if (pos < 0)
1232 pos = -pos - 1;
1233 if (pos < active_nr) {
1234 ce = active_cache[pos];
1235 if (ce_namelen(ce) == namelen &&
1236 !memcmp(ce->name, filename, namelen))
1237 die("Path '%s' is in the index, but not at stage %d.\n"
1238 "Did you mean ':%d:%s'?",
1239 filename, stage,
1240 ce_stage(ce), filename);
1241 }
1242
1243 /* Confusion between relative and absolute filenames? */
1244 fullnamelen = namelen + strlen(prefix);
1245 fullname = xmalloc(fullnamelen + 1);
1246 strcpy(fullname, prefix);
1247 strcat(fullname, filename);
1248 pos = cache_name_pos(fullname, fullnamelen);
1249 if (pos < 0)
1250 pos = -pos - 1;
1251 if (pos < active_nr) {
1252 ce = active_cache[pos];
1253 if (ce_namelen(ce) == fullnamelen &&
1254 !memcmp(ce->name, fullname, fullnamelen))
1255 die("Path '%s' is in the index, but not '%s'.\n"
1256 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1257 fullname, filename,
1258 ce_stage(ce), fullname,
1259 ce_stage(ce), filename);
1260 }
1261
1262 if (!lstat(filename, &st))
1263 die("Path '%s' exists on disk, but not in the index.", filename);
1264 if (errno == ENOENT || errno == ENOTDIR)
1265 die("Path '%s' does not exist (neither on disk nor in the index).",
1266 filename);
1267
1268 free(fullname);
1269 }
1270
1271
1272 static char *resolve_relative_path(const char *rel)
1273 {
1274 if (prefixcmp(rel, "./") && prefixcmp(rel, "../"))
1275 return NULL;
1276
1277 if (!startup_info)
1278 die("BUG: startup_info struct is not initialized.");
1279
1280 if (!is_inside_work_tree())
1281 die("relative path syntax can't be used outside working tree.");
1282
1283 /* die() inside prefix_path() if resolved path is outside worktree */
1284 return prefix_path(startup_info->prefix,
1285 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1286 rel);
1287 }
1288
1289 static int get_sha1_with_context_1(const char *name,
1290 unsigned flags,
1291 const char *prefix,
1292 unsigned char *sha1,
1293 struct object_context *oc)
1294 {
1295 int ret, bracket_depth;
1296 int namelen = strlen(name);
1297 const char *cp;
1298 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1299
1300 memset(oc, 0, sizeof(*oc));
1301 oc->mode = S_IFINVALID;
1302 ret = get_sha1_1(name, namelen, sha1, flags);
1303 if (!ret)
1304 return ret;
1305 /*
1306 * sha1:path --> object name of path in ent sha1
1307 * :path -> object name of absolute path in index
1308 * :./path -> object name of path relative to cwd in index
1309 * :[0-3]:path -> object name of path in index at stage
1310 * :/foo -> recent commit matching foo
1311 */
1312 if (name[0] == ':') {
1313 int stage = 0;
1314 const struct cache_entry *ce;
1315 char *new_path = NULL;
1316 int pos;
1317 if (!only_to_die && namelen > 2 && name[1] == '/') {
1318 struct commit_list *list = NULL;
1319 for_each_ref(handle_one_ref, &list);
1320 return get_sha1_oneline(name + 2, sha1, list);
1321 }
1322 if (namelen < 3 ||
1323 name[2] != ':' ||
1324 name[1] < '0' || '3' < name[1])
1325 cp = name + 1;
1326 else {
1327 stage = name[1] - '0';
1328 cp = name + 3;
1329 }
1330 new_path = resolve_relative_path(cp);
1331 if (!new_path) {
1332 namelen = namelen - (cp - name);
1333 } else {
1334 cp = new_path;
1335 namelen = strlen(cp);
1336 }
1337
1338 strncpy(oc->path, cp,
1339 sizeof(oc->path));
1340 oc->path[sizeof(oc->path)-1] = '\0';
1341
1342 if (!active_cache)
1343 read_cache();
1344 pos = cache_name_pos(cp, namelen);
1345 if (pos < 0)
1346 pos = -pos - 1;
1347 while (pos < active_nr) {
1348 ce = active_cache[pos];
1349 if (ce_namelen(ce) != namelen ||
1350 memcmp(ce->name, cp, namelen))
1351 break;
1352 if (ce_stage(ce) == stage) {
1353 hashcpy(sha1, ce->sha1);
1354 oc->mode = ce->ce_mode;
1355 free(new_path);
1356 return 0;
1357 }
1358 pos++;
1359 }
1360 if (only_to_die && name[1] && name[1] != '/')
1361 diagnose_invalid_index_path(stage, prefix, cp);
1362 free(new_path);
1363 return -1;
1364 }
1365 for (cp = name, bracket_depth = 0; *cp; cp++) {
1366 if (*cp == '{')
1367 bracket_depth++;
1368 else if (bracket_depth && *cp == '}')
1369 bracket_depth--;
1370 else if (!bracket_depth && *cp == ':')
1371 break;
1372 }
1373 if (*cp == ':') {
1374 unsigned char tree_sha1[20];
1375 int len = cp - name;
1376 if (!get_sha1_1(name, len, tree_sha1, GET_SHA1_TREEISH)) {
1377 const char *filename = cp+1;
1378 char *new_filename = NULL;
1379
1380 new_filename = resolve_relative_path(filename);
1381 if (new_filename)
1382 filename = new_filename;
1383 ret = get_tree_entry(tree_sha1, filename, sha1, &oc->mode);
1384 if (ret && only_to_die) {
1385 diagnose_invalid_sha1_path(prefix, filename,
1386 tree_sha1,
1387 name, len);
1388 }
1389 hashcpy(oc->tree, tree_sha1);
1390 strncpy(oc->path, filename,
1391 sizeof(oc->path));
1392 oc->path[sizeof(oc->path)-1] = '\0';
1393
1394 free(new_filename);
1395 return ret;
1396 } else {
1397 if (only_to_die)
1398 die("Invalid object name '%.*s'.", len, name);
1399 }
1400 }
1401 return ret;
1402 }
1403
1404 /*
1405 * Call this function when you know "name" given by the end user must
1406 * name an object but it doesn't; the function _may_ die with a better
1407 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1408 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1409 * you have a chance to diagnose the error further.
1410 */
1411 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1412 {
1413 struct object_context oc;
1414 unsigned char sha1[20];
1415 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1416 }
1417
1418 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1419 {
1420 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);
1421 }