use xsnprintf for generating git object headers
[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 #include "dir.h"
10
11 static int get_sha1_oneline(const char *, unsigned char *, struct commit_list *);
12
13 typedef int (*disambiguate_hint_fn)(const unsigned char *, void *);
14
15 struct disambiguate_state {
16 disambiguate_hint_fn fn;
17 void *cb_data;
18 unsigned char candidate[20];
19 unsigned candidate_exists:1;
20 unsigned candidate_checked:1;
21 unsigned candidate_ok:1;
22 unsigned disambiguate_fn_used:1;
23 unsigned ambiguous:1;
24 unsigned always_call_fn:1;
25 };
26
27 static void update_candidates(struct disambiguate_state *ds, const unsigned char *current)
28 {
29 if (ds->always_call_fn) {
30 ds->ambiguous = ds->fn(current, ds->cb_data) ? 1 : 0;
31 return;
32 }
33 if (!ds->candidate_exists) {
34 /* this is the first candidate */
35 hashcpy(ds->candidate, current);
36 ds->candidate_exists = 1;
37 return;
38 } else if (!hashcmp(ds->candidate, current)) {
39 /* the same as what we already have seen */
40 return;
41 }
42
43 if (!ds->fn) {
44 /* cannot disambiguate between ds->candidate and current */
45 ds->ambiguous = 1;
46 return;
47 }
48
49 if (!ds->candidate_checked) {
50 ds->candidate_ok = ds->fn(ds->candidate, ds->cb_data);
51 ds->disambiguate_fn_used = 1;
52 ds->candidate_checked = 1;
53 }
54
55 if (!ds->candidate_ok) {
56 /* discard the candidate; we know it does not satisfy fn */
57 hashcpy(ds->candidate, current);
58 ds->candidate_checked = 0;
59 return;
60 }
61
62 /* if we reach this point, we know ds->candidate satisfies fn */
63 if (ds->fn(current, ds->cb_data)) {
64 /*
65 * if both current and candidate satisfy fn, we cannot
66 * disambiguate.
67 */
68 ds->candidate_ok = 0;
69 ds->ambiguous = 1;
70 }
71
72 /* otherwise, current can be discarded and candidate is still good */
73 }
74
75 static void find_short_object_filename(int len, const char *hex_pfx, struct disambiguate_state *ds)
76 {
77 struct alternate_object_database *alt;
78 char hex[40];
79 static struct alternate_object_database *fakeent;
80
81 if (!fakeent) {
82 /*
83 * Create a "fake" alternate object database that
84 * points to our own object database, to make it
85 * easier to get a temporary working space in
86 * alt->name/alt->base while iterating over the
87 * object databases including our own.
88 */
89 const char *objdir = get_object_directory();
90 int objdir_len = strlen(objdir);
91 int entlen = objdir_len + 43;
92 fakeent = xmalloc(sizeof(*fakeent) + entlen);
93 memcpy(fakeent->base, objdir, objdir_len);
94 fakeent->name = fakeent->base + objdir_len + 1;
95 fakeent->name[-1] = '/';
96 }
97 fakeent->next = alt_odb_list;
98
99 sprintf(hex, "%.2s", hex_pfx);
100 for (alt = fakeent; alt && !ds->ambiguous; alt = alt->next) {
101 struct dirent *de;
102 DIR *dir;
103 sprintf(alt->name, "%.2s/", hex_pfx);
104 dir = opendir(alt->base);
105 if (!dir)
106 continue;
107
108 while (!ds->ambiguous && (de = readdir(dir)) != NULL) {
109 unsigned char sha1[20];
110
111 if (strlen(de->d_name) != 38)
112 continue;
113 if (memcmp(de->d_name, hex_pfx + 2, len - 2))
114 continue;
115 memcpy(hex + 2, de->d_name, 38);
116 if (!get_sha1_hex(hex, sha1))
117 update_candidates(ds, sha1);
118 }
119 closedir(dir);
120 }
121 }
122
123 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
124 {
125 do {
126 if (*a != *b)
127 return 0;
128 a++;
129 b++;
130 len -= 2;
131 } while (len > 1);
132 if (len)
133 if ((*a ^ *b) & 0xf0)
134 return 0;
135 return 1;
136 }
137
138 static void unique_in_pack(int len,
139 const unsigned char *bin_pfx,
140 struct packed_git *p,
141 struct disambiguate_state *ds)
142 {
143 uint32_t num, last, i, first = 0;
144 const unsigned char *current = NULL;
145
146 open_pack_index(p);
147 num = p->num_objects;
148 last = num;
149 while (first < last) {
150 uint32_t mid = (first + last) / 2;
151 const unsigned char *current;
152 int cmp;
153
154 current = nth_packed_object_sha1(p, mid);
155 cmp = hashcmp(bin_pfx, current);
156 if (!cmp) {
157 first = mid;
158 break;
159 }
160 if (cmp > 0) {
161 first = mid+1;
162 continue;
163 }
164 last = mid;
165 }
166
167 /*
168 * At this point, "first" is the location of the lowest object
169 * with an object name that could match "bin_pfx". See if we have
170 * 0, 1 or more objects that actually match(es).
171 */
172 for (i = first; i < num && !ds->ambiguous; i++) {
173 current = nth_packed_object_sha1(p, i);
174 if (!match_sha(len, bin_pfx, current))
175 break;
176 update_candidates(ds, current);
177 }
178 }
179
180 static void find_short_packed_object(int len, const unsigned char *bin_pfx,
181 struct disambiguate_state *ds)
182 {
183 struct packed_git *p;
184
185 prepare_packed_git();
186 for (p = packed_git; p && !ds->ambiguous; p = p->next)
187 unique_in_pack(len, bin_pfx, p, ds);
188 }
189
190 #define SHORT_NAME_NOT_FOUND (-1)
191 #define SHORT_NAME_AMBIGUOUS (-2)
192
193 static int finish_object_disambiguation(struct disambiguate_state *ds,
194 unsigned char *sha1)
195 {
196 if (ds->ambiguous)
197 return SHORT_NAME_AMBIGUOUS;
198
199 if (!ds->candidate_exists)
200 return SHORT_NAME_NOT_FOUND;
201
202 if (!ds->candidate_checked)
203 /*
204 * If this is the only candidate, there is no point
205 * calling the disambiguation hint callback.
206 *
207 * On the other hand, if the current candidate
208 * replaced an earlier candidate that did _not_ pass
209 * the disambiguation hint callback, then we do have
210 * more than one objects that match the short name
211 * given, so we should make sure this one matches;
212 * otherwise, if we discovered this one and the one
213 * that we previously discarded in the reverse order,
214 * we would end up showing different results in the
215 * same repository!
216 */
217 ds->candidate_ok = (!ds->disambiguate_fn_used ||
218 ds->fn(ds->candidate, ds->cb_data));
219
220 if (!ds->candidate_ok)
221 return SHORT_NAME_AMBIGUOUS;
222
223 hashcpy(sha1, ds->candidate);
224 return 0;
225 }
226
227 static int disambiguate_commit_only(const unsigned char *sha1, void *cb_data_unused)
228 {
229 int kind = sha1_object_info(sha1, NULL);
230 return kind == OBJ_COMMIT;
231 }
232
233 static int disambiguate_committish_only(const unsigned char *sha1, void *cb_data_unused)
234 {
235 struct object *obj;
236 int kind;
237
238 kind = sha1_object_info(sha1, NULL);
239 if (kind == OBJ_COMMIT)
240 return 1;
241 if (kind != OBJ_TAG)
242 return 0;
243
244 /* We need to do this the hard way... */
245 obj = deref_tag(parse_object(sha1), NULL, 0);
246 if (obj && obj->type == OBJ_COMMIT)
247 return 1;
248 return 0;
249 }
250
251 static int disambiguate_tree_only(const unsigned char *sha1, void *cb_data_unused)
252 {
253 int kind = sha1_object_info(sha1, NULL);
254 return kind == OBJ_TREE;
255 }
256
257 static int disambiguate_treeish_only(const unsigned char *sha1, void *cb_data_unused)
258 {
259 struct object *obj;
260 int kind;
261
262 kind = sha1_object_info(sha1, NULL);
263 if (kind == OBJ_TREE || kind == OBJ_COMMIT)
264 return 1;
265 if (kind != OBJ_TAG)
266 return 0;
267
268 /* We need to do this the hard way... */
269 obj = deref_tag(lookup_object(sha1), NULL, 0);
270 if (obj && (obj->type == OBJ_TREE || obj->type == OBJ_COMMIT))
271 return 1;
272 return 0;
273 }
274
275 static int disambiguate_blob_only(const unsigned char *sha1, void *cb_data_unused)
276 {
277 int kind = sha1_object_info(sha1, NULL);
278 return kind == OBJ_BLOB;
279 }
280
281 static int prepare_prefixes(const char *name, int len,
282 unsigned char *bin_pfx,
283 char *hex_pfx)
284 {
285 int i;
286
287 hashclr(bin_pfx);
288 memset(hex_pfx, 'x', 40);
289 for (i = 0; i < len ;i++) {
290 unsigned char c = name[i];
291 unsigned char val;
292 if (c >= '0' && c <= '9')
293 val = c - '0';
294 else if (c >= 'a' && c <= 'f')
295 val = c - 'a' + 10;
296 else if (c >= 'A' && c <='F') {
297 val = c - 'A' + 10;
298 c -= 'A' - 'a';
299 }
300 else
301 return -1;
302 hex_pfx[i] = c;
303 if (!(i & 1))
304 val <<= 4;
305 bin_pfx[i >> 1] |= val;
306 }
307 return 0;
308 }
309
310 static int get_short_sha1(const char *name, int len, unsigned char *sha1,
311 unsigned flags)
312 {
313 int status;
314 char hex_pfx[40];
315 unsigned char bin_pfx[20];
316 struct disambiguate_state ds;
317 int quietly = !!(flags & GET_SHA1_QUIETLY);
318
319 if (len < MINIMUM_ABBREV || len > 40)
320 return -1;
321 if (prepare_prefixes(name, len, bin_pfx, hex_pfx) < 0)
322 return -1;
323
324 prepare_alt_odb();
325
326 memset(&ds, 0, sizeof(ds));
327 if (flags & GET_SHA1_COMMIT)
328 ds.fn = disambiguate_commit_only;
329 else if (flags & GET_SHA1_COMMITTISH)
330 ds.fn = disambiguate_committish_only;
331 else if (flags & GET_SHA1_TREE)
332 ds.fn = disambiguate_tree_only;
333 else if (flags & GET_SHA1_TREEISH)
334 ds.fn = disambiguate_treeish_only;
335 else if (flags & GET_SHA1_BLOB)
336 ds.fn = disambiguate_blob_only;
337
338 find_short_object_filename(len, hex_pfx, &ds);
339 find_short_packed_object(len, bin_pfx, &ds);
340 status = finish_object_disambiguation(&ds, sha1);
341
342 if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
343 return error("short SHA1 %.*s is ambiguous.", len, hex_pfx);
344 return status;
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 int find_unique_abbrev_r(char *hex, const unsigned char *sha1, int len)
372 {
373 int status, exists;
374
375 sha1_to_hex_r(hex, sha1);
376 if (len == 40 || !len)
377 return 40;
378 exists = has_sha1_file(sha1);
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 len;
387 }
388 len++;
389 }
390 return len;
391 }
392
393 const char *find_unique_abbrev(const unsigned char *sha1, int len)
394 {
395 static char hex[GIT_SHA1_HEXSZ + 1];
396 find_unique_abbrev_r(hex, sha1, len);
397 return hex;
398 }
399
400 static int ambiguous_path(const char *path, int len)
401 {
402 int slash = 1;
403 int cnt;
404
405 for (cnt = 0; cnt < len; cnt++) {
406 switch (*path++) {
407 case '\0':
408 break;
409 case '/':
410 if (slash)
411 break;
412 slash = 1;
413 continue;
414 case '.':
415 continue;
416 default:
417 slash = 0;
418 continue;
419 }
420 break;
421 }
422 return slash;
423 }
424
425 static inline int at_mark(const char *string, int len,
426 const char **suffix, int nr)
427 {
428 int i;
429
430 for (i = 0; i < nr; i++) {
431 int suffix_len = strlen(suffix[i]);
432 if (suffix_len <= len
433 && !memcmp(string, suffix[i], suffix_len))
434 return suffix_len;
435 }
436 return 0;
437 }
438
439 static inline int upstream_mark(const char *string, int len)
440 {
441 const char *suffix[] = { "@{upstream}", "@{u}" };
442 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
443 }
444
445 static inline int push_mark(const char *string, int len)
446 {
447 const char *suffix[] = { "@{push}" };
448 return at_mark(string, len, suffix, ARRAY_SIZE(suffix));
449 }
450
451 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags);
452 static int interpret_nth_prior_checkout(const char *name, int namelen, struct strbuf *buf);
453
454 static int get_sha1_basic(const char *str, int len, unsigned char *sha1,
455 unsigned int flags)
456 {
457 static const char *warn_msg = "refname '%.*s' is ambiguous.";
458 static const char *object_name_msg = N_(
459 "Git normally never creates a ref that ends with 40 hex characters\n"
460 "because it will be ignored when you just specify 40-hex. These refs\n"
461 "may be created by mistake. For example,\n"
462 "\n"
463 " git checkout -b $br $(git rev-parse ...)\n"
464 "\n"
465 "where \"$br\" is somehow empty and a 40-hex ref is created. Please\n"
466 "examine these refs and maybe delete them. Turn this message off by\n"
467 "running \"git config advice.objectNameWarning false\"");
468 unsigned char tmp_sha1[20];
469 char *real_ref = NULL;
470 int refs_found = 0;
471 int at, reflog_len, nth_prior = 0;
472
473 if (len == 40 && !get_sha1_hex(str, sha1)) {
474 if (warn_ambiguous_refs && warn_on_object_refname_ambiguity) {
475 refs_found = dwim_ref(str, len, tmp_sha1, &real_ref);
476 if (refs_found > 0) {
477 warning(warn_msg, len, str);
478 if (advice_object_name_warning)
479 fprintf(stderr, "%s\n", _(object_name_msg));
480 }
481 free(real_ref);
482 }
483 return 0;
484 }
485
486 /* basic@{time or number or -number} format to query ref-log */
487 reflog_len = at = 0;
488 if (len && str[len-1] == '}') {
489 for (at = len-4; at >= 0; at--) {
490 if (str[at] == '@' && str[at+1] == '{') {
491 if (str[at+2] == '-') {
492 if (at != 0)
493 /* @{-N} not at start */
494 return -1;
495 nth_prior = 1;
496 continue;
497 }
498 if (!upstream_mark(str + at, len - at) &&
499 !push_mark(str + at, len - at)) {
500 reflog_len = (len-1) - (at+2);
501 len = at;
502 }
503 break;
504 }
505 }
506 }
507
508 /* Accept only unambiguous ref paths. */
509 if (len && ambiguous_path(str, len))
510 return -1;
511
512 if (nth_prior) {
513 struct strbuf buf = STRBUF_INIT;
514 int detached;
515
516 if (interpret_nth_prior_checkout(str, len, &buf) > 0) {
517 detached = (buf.len == 40 && !get_sha1_hex(buf.buf, sha1));
518 strbuf_release(&buf);
519 if (detached)
520 return 0;
521 }
522 }
523
524 if (!len && reflog_len)
525 /* allow "@{...}" to mean the current branch reflog */
526 refs_found = dwim_ref("HEAD", 4, sha1, &real_ref);
527 else if (reflog_len)
528 refs_found = dwim_log(str, len, sha1, &real_ref);
529 else
530 refs_found = dwim_ref(str, len, sha1, &real_ref);
531
532 if (!refs_found)
533 return -1;
534
535 if (warn_ambiguous_refs && !(flags & GET_SHA1_QUIETLY) &&
536 (refs_found > 1 ||
537 !get_short_sha1(str, len, tmp_sha1, GET_SHA1_QUIETLY)))
538 warning(warn_msg, len, str);
539
540 if (reflog_len) {
541 int nth, i;
542 unsigned long at_time;
543 unsigned long co_time;
544 int co_tz, co_cnt;
545
546 /* Is it asking for N-th entry, or approxidate? */
547 for (i = nth = 0; 0 <= nth && i < reflog_len; i++) {
548 char ch = str[at+2+i];
549 if ('0' <= ch && ch <= '9')
550 nth = nth * 10 + ch - '0';
551 else
552 nth = -1;
553 }
554 if (100000000 <= nth) {
555 at_time = nth;
556 nth = -1;
557 } else if (0 <= nth)
558 at_time = 0;
559 else {
560 int errors = 0;
561 char *tmp = xstrndup(str + at + 2, reflog_len);
562 at_time = approxidate_careful(tmp, &errors);
563 free(tmp);
564 if (errors) {
565 free(real_ref);
566 return -1;
567 }
568 }
569 if (read_ref_at(real_ref, flags, at_time, nth, sha1, NULL,
570 &co_time, &co_tz, &co_cnt)) {
571 if (!len) {
572 if (starts_with(real_ref, "refs/heads/")) {
573 str = real_ref + 11;
574 len = strlen(real_ref + 11);
575 } else {
576 /* detached HEAD */
577 str = "HEAD";
578 len = 4;
579 }
580 }
581 if (at_time) {
582 if (!(flags & GET_SHA1_QUIETLY)) {
583 warning("Log for '%.*s' only goes "
584 "back to %s.", len, str,
585 show_date(co_time, co_tz, DATE_MODE(RFC2822)));
586 }
587 } else {
588 if (flags & GET_SHA1_QUIETLY) {
589 exit(128);
590 }
591 die("Log for '%.*s' only has %d entries.",
592 len, str, co_cnt);
593 }
594 }
595 }
596
597 free(real_ref);
598 return 0;
599 }
600
601 static int get_parent(const char *name, int len,
602 unsigned char *result, int idx)
603 {
604 unsigned char sha1[20];
605 int ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
606 struct commit *commit;
607 struct commit_list *p;
608
609 if (ret)
610 return ret;
611 commit = lookup_commit_reference(sha1);
612 if (parse_commit(commit))
613 return -1;
614 if (!idx) {
615 hashcpy(result, commit->object.sha1);
616 return 0;
617 }
618 p = commit->parents;
619 while (p) {
620 if (!--idx) {
621 hashcpy(result, p->item->object.sha1);
622 return 0;
623 }
624 p = p->next;
625 }
626 return -1;
627 }
628
629 static int get_nth_ancestor(const char *name, int len,
630 unsigned char *result, int generation)
631 {
632 unsigned char sha1[20];
633 struct commit *commit;
634 int ret;
635
636 ret = get_sha1_1(name, len, sha1, GET_SHA1_COMMITTISH);
637 if (ret)
638 return ret;
639 commit = lookup_commit_reference(sha1);
640 if (!commit)
641 return -1;
642
643 while (generation--) {
644 if (parse_commit(commit) || !commit->parents)
645 return -1;
646 commit = commit->parents->item;
647 }
648 hashcpy(result, commit->object.sha1);
649 return 0;
650 }
651
652 struct object *peel_to_type(const char *name, int namelen,
653 struct object *o, enum object_type expected_type)
654 {
655 if (name && !namelen)
656 namelen = strlen(name);
657 while (1) {
658 if (!o || (!o->parsed && !parse_object(o->sha1)))
659 return NULL;
660 if (expected_type == OBJ_ANY || o->type == expected_type)
661 return o;
662 if (o->type == OBJ_TAG)
663 o = ((struct tag*) o)->tagged;
664 else if (o->type == OBJ_COMMIT)
665 o = &(((struct commit *) o)->tree->object);
666 else {
667 if (name)
668 error("%.*s: expected %s type, but the object "
669 "dereferences to %s type",
670 namelen, name, typename(expected_type),
671 typename(o->type));
672 return NULL;
673 }
674 }
675 }
676
677 static int peel_onion(const char *name, int len, unsigned char *sha1)
678 {
679 unsigned char outer[20];
680 const char *sp;
681 unsigned int expected_type = 0;
682 unsigned lookup_flags = 0;
683 struct object *o;
684
685 /*
686 * "ref^{type}" dereferences ref repeatedly until you cannot
687 * dereference anymore, or you get an object of given type,
688 * whichever comes first. "ref^{}" means just dereference
689 * tags until you get a non-tag. "ref^0" is a shorthand for
690 * "ref^{commit}". "commit^{tree}" could be used to find the
691 * top-level tree of the given commit.
692 */
693 if (len < 4 || name[len-1] != '}')
694 return -1;
695
696 for (sp = name + len - 1; name <= sp; sp--) {
697 int ch = *sp;
698 if (ch == '{' && name < sp && sp[-1] == '^')
699 break;
700 }
701 if (sp <= name)
702 return -1;
703
704 sp++; /* beginning of type name, or closing brace for empty */
705 if (starts_with(sp, "commit}"))
706 expected_type = OBJ_COMMIT;
707 else if (starts_with(sp, "tag}"))
708 expected_type = OBJ_TAG;
709 else if (starts_with(sp, "tree}"))
710 expected_type = OBJ_TREE;
711 else if (starts_with(sp, "blob}"))
712 expected_type = OBJ_BLOB;
713 else if (starts_with(sp, "object}"))
714 expected_type = OBJ_ANY;
715 else if (sp[0] == '}')
716 expected_type = OBJ_NONE;
717 else if (sp[0] == '/')
718 expected_type = OBJ_COMMIT;
719 else
720 return -1;
721
722 if (expected_type == OBJ_COMMIT)
723 lookup_flags = GET_SHA1_COMMITTISH;
724 else if (expected_type == OBJ_TREE)
725 lookup_flags = GET_SHA1_TREEISH;
726
727 if (get_sha1_1(name, sp - name - 2, outer, lookup_flags))
728 return -1;
729
730 o = parse_object(outer);
731 if (!o)
732 return -1;
733 if (!expected_type) {
734 o = deref_tag(o, name, sp - name - 2);
735 if (!o || (!o->parsed && !parse_object(o->sha1)))
736 return -1;
737 hashcpy(sha1, o->sha1);
738 return 0;
739 }
740
741 /*
742 * At this point, the syntax look correct, so
743 * if we do not get the needed object, we should
744 * barf.
745 */
746 o = peel_to_type(name, len, o, expected_type);
747 if (!o)
748 return -1;
749
750 hashcpy(sha1, o->sha1);
751 if (sp[0] == '/') {
752 /* "$commit^{/foo}" */
753 char *prefix;
754 int ret;
755 struct commit_list *list = NULL;
756
757 /*
758 * $commit^{/}. Some regex implementation may reject.
759 * We don't need regex anyway. '' pattern always matches.
760 */
761 if (sp[1] == '}')
762 return 0;
763
764 prefix = xstrndup(sp + 1, name + len - 1 - (sp + 1));
765 commit_list_insert((struct commit *)o, &list);
766 ret = get_sha1_oneline(prefix, sha1, list);
767 free(prefix);
768 return ret;
769 }
770 return 0;
771 }
772
773 static int get_describe_name(const char *name, int len, unsigned char *sha1)
774 {
775 const char *cp;
776 unsigned flags = GET_SHA1_QUIETLY | GET_SHA1_COMMIT;
777
778 for (cp = name + len - 1; name + 2 <= cp; cp--) {
779 char ch = *cp;
780 if (!isxdigit(ch)) {
781 /* We must be looking at g in "SOMETHING-g"
782 * for it to be describe output.
783 */
784 if (ch == 'g' && cp[-1] == '-') {
785 cp++;
786 len -= cp - name;
787 return get_short_sha1(cp, len, sha1, flags);
788 }
789 }
790 }
791 return -1;
792 }
793
794 static int get_sha1_1(const char *name, int len, unsigned char *sha1, unsigned lookup_flags)
795 {
796 int ret, has_suffix;
797 const char *cp;
798
799 /*
800 * "name~3" is "name^^^", "name~" is "name~1", and "name^" is "name^1".
801 */
802 has_suffix = 0;
803 for (cp = name + len - 1; name <= cp; cp--) {
804 int ch = *cp;
805 if ('0' <= ch && ch <= '9')
806 continue;
807 if (ch == '~' || ch == '^')
808 has_suffix = ch;
809 break;
810 }
811
812 if (has_suffix) {
813 int num = 0;
814 int len1 = cp - name;
815 cp++;
816 while (cp < name + len)
817 num = num * 10 + *cp++ - '0';
818 if (!num && len1 == len - 1)
819 num = 1;
820 if (has_suffix == '^')
821 return get_parent(name, len1, sha1, num);
822 /* else if (has_suffix == '~') -- goes without saying */
823 return get_nth_ancestor(name, len1, sha1, num);
824 }
825
826 ret = peel_onion(name, len, sha1);
827 if (!ret)
828 return 0;
829
830 ret = get_sha1_basic(name, len, sha1, lookup_flags);
831 if (!ret)
832 return 0;
833
834 /* It could be describe output that is "SOMETHING-gXXXX" */
835 ret = get_describe_name(name, len, sha1);
836 if (!ret)
837 return 0;
838
839 return get_short_sha1(name, len, sha1, lookup_flags);
840 }
841
842 /*
843 * This interprets names like ':/Initial revision of "git"' by searching
844 * through history and returning the first commit whose message starts
845 * the given regular expression.
846 *
847 * For future extension, ':/!' is reserved. If you want to match a message
848 * beginning with a '!', you have to repeat the exclamation mark.
849 */
850
851 /* Remember to update object flag allocation in object.h */
852 #define ONELINE_SEEN (1u<<20)
853
854 static int handle_one_ref(const char *path, const struct object_id *oid,
855 int flag, void *cb_data)
856 {
857 struct commit_list **list = cb_data;
858 struct object *object = parse_object(oid->hash);
859 if (!object)
860 return 0;
861 if (object->type == OBJ_TAG) {
862 object = deref_tag(object, path, strlen(path));
863 if (!object)
864 return 0;
865 }
866 if (object->type != OBJ_COMMIT)
867 return 0;
868 commit_list_insert((struct commit *)object, list);
869 return 0;
870 }
871
872 static int get_sha1_oneline(const char *prefix, unsigned char *sha1,
873 struct commit_list *list)
874 {
875 struct commit_list *backup = NULL, *l;
876 int found = 0;
877 regex_t regex;
878
879 if (prefix[0] == '!') {
880 if (prefix[1] != '!')
881 die ("Invalid search pattern: %s", prefix);
882 prefix++;
883 }
884
885 if (regcomp(&regex, prefix, REG_EXTENDED))
886 die("Invalid search pattern: %s", prefix);
887
888 for (l = list; l; l = l->next) {
889 l->item->object.flags |= ONELINE_SEEN;
890 commit_list_insert(l->item, &backup);
891 }
892 while (list) {
893 const char *p, *buf;
894 struct commit *commit;
895 int matches;
896
897 commit = pop_most_recent_commit(&list, ONELINE_SEEN);
898 if (!parse_object(commit->object.sha1))
899 continue;
900 buf = get_commit_buffer(commit, NULL);
901 p = strstr(buf, "\n\n");
902 matches = p && !regexec(&regex, p + 2, 0, NULL, 0);
903 unuse_commit_buffer(commit, buf);
904
905 if (matches) {
906 hashcpy(sha1, commit->object.sha1);
907 found = 1;
908 break;
909 }
910 }
911 regfree(&regex);
912 free_commit_list(list);
913 for (l = backup; l; l = l->next)
914 clear_commit_marks(l->item, ONELINE_SEEN);
915 free_commit_list(backup);
916 return found ? 0 : -1;
917 }
918
919 struct grab_nth_branch_switch_cbdata {
920 int remaining;
921 struct strbuf buf;
922 };
923
924 static int grab_nth_branch_switch(unsigned char *osha1, unsigned char *nsha1,
925 const char *email, unsigned long timestamp, int tz,
926 const char *message, void *cb_data)
927 {
928 struct grab_nth_branch_switch_cbdata *cb = cb_data;
929 const char *match = NULL, *target = NULL;
930 size_t len;
931
932 if (skip_prefix(message, "checkout: moving from ", &match))
933 target = strstr(match, " to ");
934
935 if (!match || !target)
936 return 0;
937 if (--(cb->remaining) == 0) {
938 len = target - match;
939 strbuf_reset(&cb->buf);
940 strbuf_add(&cb->buf, match, len);
941 return 1; /* we are done */
942 }
943 return 0;
944 }
945
946 /*
947 * Parse @{-N} syntax, return the number of characters parsed
948 * if successful; otherwise signal an error with negative value.
949 */
950 static int interpret_nth_prior_checkout(const char *name, int namelen,
951 struct strbuf *buf)
952 {
953 long nth;
954 int retval;
955 struct grab_nth_branch_switch_cbdata cb;
956 const char *brace;
957 char *num_end;
958
959 if (namelen < 4)
960 return -1;
961 if (name[0] != '@' || name[1] != '{' || name[2] != '-')
962 return -1;
963 brace = memchr(name, '}', namelen);
964 if (!brace)
965 return -1;
966 nth = strtol(name + 3, &num_end, 10);
967 if (num_end != brace)
968 return -1;
969 if (nth <= 0)
970 return -1;
971 cb.remaining = nth;
972 strbuf_init(&cb.buf, 20);
973
974 retval = 0;
975 if (0 < for_each_reflog_ent_reverse("HEAD", grab_nth_branch_switch, &cb)) {
976 strbuf_reset(buf);
977 strbuf_addbuf(buf, &cb.buf);
978 retval = brace - name + 1;
979 }
980
981 strbuf_release(&cb.buf);
982 return retval;
983 }
984
985 int get_sha1_mb(const char *name, unsigned char *sha1)
986 {
987 struct commit *one, *two;
988 struct commit_list *mbs;
989 unsigned char sha1_tmp[20];
990 const char *dots;
991 int st;
992
993 dots = strstr(name, "...");
994 if (!dots)
995 return get_sha1(name, sha1);
996 if (dots == name)
997 st = get_sha1("HEAD", sha1_tmp);
998 else {
999 struct strbuf sb;
1000 strbuf_init(&sb, dots - name);
1001 strbuf_add(&sb, name, dots - name);
1002 st = get_sha1_committish(sb.buf, sha1_tmp);
1003 strbuf_release(&sb);
1004 }
1005 if (st)
1006 return st;
1007 one = lookup_commit_reference_gently(sha1_tmp, 0);
1008 if (!one)
1009 return -1;
1010
1011 if (get_sha1_committish(dots[3] ? (dots + 3) : "HEAD", sha1_tmp))
1012 return -1;
1013 two = lookup_commit_reference_gently(sha1_tmp, 0);
1014 if (!two)
1015 return -1;
1016 mbs = get_merge_bases(one, two);
1017 if (!mbs || mbs->next)
1018 st = -1;
1019 else {
1020 st = 0;
1021 hashcpy(sha1, mbs->item->object.sha1);
1022 }
1023 free_commit_list(mbs);
1024 return st;
1025 }
1026
1027 /* parse @something syntax, when 'something' is not {.*} */
1028 static int interpret_empty_at(const char *name, int namelen, int len, struct strbuf *buf)
1029 {
1030 const char *next;
1031
1032 if (len || name[1] == '{')
1033 return -1;
1034
1035 /* make sure it's a single @, or @@{.*}, not @foo */
1036 next = memchr(name + len + 1, '@', namelen - len - 1);
1037 if (next && next[1] != '{')
1038 return -1;
1039 if (!next)
1040 next = name + namelen;
1041 if (next != name + 1)
1042 return -1;
1043
1044 strbuf_reset(buf);
1045 strbuf_add(buf, "HEAD", 4);
1046 return 1;
1047 }
1048
1049 static int reinterpret(const char *name, int namelen, int len, struct strbuf *buf)
1050 {
1051 /* we have extra data, which might need further processing */
1052 struct strbuf tmp = STRBUF_INIT;
1053 int used = buf->len;
1054 int ret;
1055
1056 strbuf_add(buf, name + len, namelen - len);
1057 ret = interpret_branch_name(buf->buf, buf->len, &tmp);
1058 /* that data was not interpreted, remove our cruft */
1059 if (ret < 0) {
1060 strbuf_setlen(buf, used);
1061 return len;
1062 }
1063 strbuf_reset(buf);
1064 strbuf_addbuf(buf, &tmp);
1065 strbuf_release(&tmp);
1066 /* tweak for size of {-N} versus expanded ref name */
1067 return ret - used + len;
1068 }
1069
1070 static void set_shortened_ref(struct strbuf *buf, const char *ref)
1071 {
1072 char *s = shorten_unambiguous_ref(ref, 0);
1073 strbuf_reset(buf);
1074 strbuf_addstr(buf, s);
1075 free(s);
1076 }
1077
1078 static int interpret_branch_mark(const char *name, int namelen,
1079 int at, struct strbuf *buf,
1080 int (*get_mark)(const char *, int),
1081 const char *(*get_data)(struct branch *,
1082 struct strbuf *))
1083 {
1084 int len;
1085 struct branch *branch;
1086 struct strbuf err = STRBUF_INIT;
1087 const char *value;
1088
1089 len = get_mark(name + at, namelen - at);
1090 if (!len)
1091 return -1;
1092
1093 if (memchr(name, ':', at))
1094 return -1;
1095
1096 if (at) {
1097 char *name_str = xmemdupz(name, at);
1098 branch = branch_get(name_str);
1099 free(name_str);
1100 } else
1101 branch = branch_get(NULL);
1102
1103 value = get_data(branch, &err);
1104 if (!value)
1105 die("%s", err.buf);
1106
1107 set_shortened_ref(buf, value);
1108 return len + at;
1109 }
1110
1111 /*
1112 * This reads short-hand syntax that not only evaluates to a commit
1113 * object name, but also can act as if the end user spelled the name
1114 * of the branch from the command line.
1115 *
1116 * - "@{-N}" finds the name of the Nth previous branch we were on, and
1117 * places the name of the branch in the given buf and returns the
1118 * number of characters parsed if successful.
1119 *
1120 * - "<branch>@{upstream}" finds the name of the other ref that
1121 * <branch> is configured to merge with (missing <branch> defaults
1122 * to the current branch), and places the name of the branch in the
1123 * given buf and returns the number of characters parsed if
1124 * successful.
1125 *
1126 * If the input is not of the accepted format, it returns a negative
1127 * number to signal an error.
1128 *
1129 * If the input was ok but there are not N branch switches in the
1130 * reflog, it returns 0.
1131 */
1132 int interpret_branch_name(const char *name, int namelen, struct strbuf *buf)
1133 {
1134 char *at;
1135 const char *start;
1136 int len = interpret_nth_prior_checkout(name, namelen, buf);
1137
1138 if (!namelen)
1139 namelen = strlen(name);
1140
1141 if (!len) {
1142 return len; /* syntax Ok, not enough switches */
1143 } else if (len > 0) {
1144 if (len == namelen)
1145 return len; /* consumed all */
1146 else
1147 return reinterpret(name, namelen, len, buf);
1148 }
1149
1150 for (start = name;
1151 (at = memchr(start, '@', namelen - (start - name)));
1152 start = at + 1) {
1153
1154 len = interpret_empty_at(name, namelen, at - name, buf);
1155 if (len > 0)
1156 return reinterpret(name, namelen, len, buf);
1157
1158 len = interpret_branch_mark(name, namelen, at - name, buf,
1159 upstream_mark, branch_get_upstream);
1160 if (len > 0)
1161 return len;
1162
1163 len = interpret_branch_mark(name, namelen, at - name, buf,
1164 push_mark, branch_get_push);
1165 if (len > 0)
1166 return len;
1167 }
1168
1169 return -1;
1170 }
1171
1172 int strbuf_branchname(struct strbuf *sb, const char *name)
1173 {
1174 int len = strlen(name);
1175 int used = interpret_branch_name(name, len, sb);
1176
1177 if (used == len)
1178 return 0;
1179 if (used < 0)
1180 used = 0;
1181 strbuf_add(sb, name + used, len - used);
1182 return len;
1183 }
1184
1185 int strbuf_check_branch_ref(struct strbuf *sb, const char *name)
1186 {
1187 strbuf_branchname(sb, name);
1188 if (name[0] == '-')
1189 return -1;
1190 strbuf_splice(sb, 0, 0, "refs/heads/", 11);
1191 return check_refname_format(sb->buf, 0);
1192 }
1193
1194 /*
1195 * This is like "get_sha1_basic()", except it allows "sha1 expressions",
1196 * notably "xyz^" for "parent of xyz"
1197 */
1198 int get_sha1(const char *name, unsigned char *sha1)
1199 {
1200 struct object_context unused;
1201 return get_sha1_with_context(name, 0, sha1, &unused);
1202 }
1203
1204 /*
1205 * Many callers know that the user meant to name a commit-ish by
1206 * syntactical positions where the object name appears. Calling this
1207 * function allows the machinery to disambiguate shorter-than-unique
1208 * abbreviated object names between commit-ish and others.
1209 *
1210 * Note that this does NOT error out when the named object is not a
1211 * commit-ish. It is merely to give a hint to the disambiguation
1212 * machinery.
1213 */
1214 int get_sha1_committish(const char *name, unsigned char *sha1)
1215 {
1216 struct object_context unused;
1217 return get_sha1_with_context(name, GET_SHA1_COMMITTISH,
1218 sha1, &unused);
1219 }
1220
1221 int get_sha1_treeish(const char *name, unsigned char *sha1)
1222 {
1223 struct object_context unused;
1224 return get_sha1_with_context(name, GET_SHA1_TREEISH,
1225 sha1, &unused);
1226 }
1227
1228 int get_sha1_commit(const char *name, unsigned char *sha1)
1229 {
1230 struct object_context unused;
1231 return get_sha1_with_context(name, GET_SHA1_COMMIT,
1232 sha1, &unused);
1233 }
1234
1235 int get_sha1_tree(const char *name, unsigned char *sha1)
1236 {
1237 struct object_context unused;
1238 return get_sha1_with_context(name, GET_SHA1_TREE,
1239 sha1, &unused);
1240 }
1241
1242 int get_sha1_blob(const char *name, unsigned char *sha1)
1243 {
1244 struct object_context unused;
1245 return get_sha1_with_context(name, GET_SHA1_BLOB,
1246 sha1, &unused);
1247 }
1248
1249 /* Must be called only when object_name:filename doesn't exist. */
1250 static void diagnose_invalid_sha1_path(const char *prefix,
1251 const char *filename,
1252 const unsigned char *tree_sha1,
1253 const char *object_name,
1254 int object_name_len)
1255 {
1256 unsigned char sha1[20];
1257 unsigned mode;
1258
1259 if (!prefix)
1260 prefix = "";
1261
1262 if (file_exists(filename))
1263 die("Path '%s' exists on disk, but not in '%.*s'.",
1264 filename, object_name_len, object_name);
1265 if (errno == ENOENT || errno == ENOTDIR) {
1266 char *fullname = xstrfmt("%s%s", prefix, filename);
1267
1268 if (!get_tree_entry(tree_sha1, fullname,
1269 sha1, &mode)) {
1270 die("Path '%s' exists, but not '%s'.\n"
1271 "Did you mean '%.*s:%s' aka '%.*s:./%s'?",
1272 fullname,
1273 filename,
1274 object_name_len, object_name,
1275 fullname,
1276 object_name_len, object_name,
1277 filename);
1278 }
1279 die("Path '%s' does not exist in '%.*s'",
1280 filename, object_name_len, object_name);
1281 }
1282 }
1283
1284 /* Must be called only when :stage:filename doesn't exist. */
1285 static void diagnose_invalid_index_path(int stage,
1286 const char *prefix,
1287 const char *filename)
1288 {
1289 const struct cache_entry *ce;
1290 int pos;
1291 unsigned namelen = strlen(filename);
1292 unsigned fullnamelen;
1293 char *fullname;
1294
1295 if (!prefix)
1296 prefix = "";
1297
1298 /* Wrong stage number? */
1299 pos = cache_name_pos(filename, namelen);
1300 if (pos < 0)
1301 pos = -pos - 1;
1302 if (pos < active_nr) {
1303 ce = active_cache[pos];
1304 if (ce_namelen(ce) == namelen &&
1305 !memcmp(ce->name, filename, namelen))
1306 die("Path '%s' is in the index, but not at stage %d.\n"
1307 "Did you mean ':%d:%s'?",
1308 filename, stage,
1309 ce_stage(ce), filename);
1310 }
1311
1312 /* Confusion between relative and absolute filenames? */
1313 fullnamelen = namelen + strlen(prefix);
1314 fullname = xmalloc(fullnamelen + 1);
1315 strcpy(fullname, prefix);
1316 strcat(fullname, filename);
1317 pos = cache_name_pos(fullname, fullnamelen);
1318 if (pos < 0)
1319 pos = -pos - 1;
1320 if (pos < active_nr) {
1321 ce = active_cache[pos];
1322 if (ce_namelen(ce) == fullnamelen &&
1323 !memcmp(ce->name, fullname, fullnamelen))
1324 die("Path '%s' is in the index, but not '%s'.\n"
1325 "Did you mean ':%d:%s' aka ':%d:./%s'?",
1326 fullname, filename,
1327 ce_stage(ce), fullname,
1328 ce_stage(ce), filename);
1329 }
1330
1331 if (file_exists(filename))
1332 die("Path '%s' exists on disk, but not in the index.", filename);
1333 if (errno == ENOENT || errno == ENOTDIR)
1334 die("Path '%s' does not exist (neither on disk nor in the index).",
1335 filename);
1336
1337 free(fullname);
1338 }
1339
1340
1341 static char *resolve_relative_path(const char *rel)
1342 {
1343 if (!starts_with(rel, "./") && !starts_with(rel, "../"))
1344 return NULL;
1345
1346 if (!startup_info)
1347 die("BUG: startup_info struct is not initialized.");
1348
1349 if (!is_inside_work_tree())
1350 die("relative path syntax can't be used outside working tree.");
1351
1352 /* die() inside prefix_path() if resolved path is outside worktree */
1353 return prefix_path(startup_info->prefix,
1354 startup_info->prefix ? strlen(startup_info->prefix) : 0,
1355 rel);
1356 }
1357
1358 static int get_sha1_with_context_1(const char *name,
1359 unsigned flags,
1360 const char *prefix,
1361 unsigned char *sha1,
1362 struct object_context *oc)
1363 {
1364 int ret, bracket_depth;
1365 int namelen = strlen(name);
1366 const char *cp;
1367 int only_to_die = flags & GET_SHA1_ONLY_TO_DIE;
1368
1369 memset(oc, 0, sizeof(*oc));
1370 oc->mode = S_IFINVALID;
1371 ret = get_sha1_1(name, namelen, sha1, flags);
1372 if (!ret)
1373 return ret;
1374 /*
1375 * sha1:path --> object name of path in ent sha1
1376 * :path -> object name of absolute path in index
1377 * :./path -> object name of path relative to cwd in index
1378 * :[0-3]:path -> object name of path in index at stage
1379 * :/foo -> recent commit matching foo
1380 */
1381 if (name[0] == ':') {
1382 int stage = 0;
1383 const struct cache_entry *ce;
1384 char *new_path = NULL;
1385 int pos;
1386 if (!only_to_die && namelen > 2 && name[1] == '/') {
1387 struct commit_list *list = NULL;
1388
1389 for_each_ref(handle_one_ref, &list);
1390 commit_list_sort_by_date(&list);
1391 return get_sha1_oneline(name + 2, sha1, list);
1392 }
1393 if (namelen < 3 ||
1394 name[2] != ':' ||
1395 name[1] < '0' || '3' < name[1])
1396 cp = name + 1;
1397 else {
1398 stage = name[1] - '0';
1399 cp = name + 3;
1400 }
1401 new_path = resolve_relative_path(cp);
1402 if (!new_path) {
1403 namelen = namelen - (cp - name);
1404 } else {
1405 cp = new_path;
1406 namelen = strlen(cp);
1407 }
1408
1409 strlcpy(oc->path, cp, sizeof(oc->path));
1410
1411 if (!active_cache)
1412 read_cache();
1413 pos = cache_name_pos(cp, namelen);
1414 if (pos < 0)
1415 pos = -pos - 1;
1416 while (pos < active_nr) {
1417 ce = active_cache[pos];
1418 if (ce_namelen(ce) != namelen ||
1419 memcmp(ce->name, cp, namelen))
1420 break;
1421 if (ce_stage(ce) == stage) {
1422 hashcpy(sha1, ce->sha1);
1423 oc->mode = ce->ce_mode;
1424 free(new_path);
1425 return 0;
1426 }
1427 pos++;
1428 }
1429 if (only_to_die && name[1] && name[1] != '/')
1430 diagnose_invalid_index_path(stage, prefix, cp);
1431 free(new_path);
1432 return -1;
1433 }
1434 for (cp = name, bracket_depth = 0; *cp; cp++) {
1435 if (*cp == '{')
1436 bracket_depth++;
1437 else if (bracket_depth && *cp == '}')
1438 bracket_depth--;
1439 else if (!bracket_depth && *cp == ':')
1440 break;
1441 }
1442 if (*cp == ':') {
1443 unsigned char tree_sha1[20];
1444 int len = cp - name;
1445 if (!get_sha1_1(name, len, tree_sha1, GET_SHA1_TREEISH)) {
1446 const char *filename = cp+1;
1447 char *new_filename = NULL;
1448
1449 new_filename = resolve_relative_path(filename);
1450 if (new_filename)
1451 filename = new_filename;
1452 if (flags & GET_SHA1_FOLLOW_SYMLINKS) {
1453 ret = get_tree_entry_follow_symlinks(tree_sha1,
1454 filename, sha1, &oc->symlink_path,
1455 &oc->mode);
1456 } else {
1457 ret = get_tree_entry(tree_sha1, filename,
1458 sha1, &oc->mode);
1459 if (ret && only_to_die) {
1460 diagnose_invalid_sha1_path(prefix,
1461 filename,
1462 tree_sha1,
1463 name, len);
1464 }
1465 }
1466 hashcpy(oc->tree, tree_sha1);
1467 strlcpy(oc->path, filename, sizeof(oc->path));
1468
1469 free(new_filename);
1470 return ret;
1471 } else {
1472 if (only_to_die)
1473 die("Invalid object name '%.*s'.", len, name);
1474 }
1475 }
1476 return ret;
1477 }
1478
1479 /*
1480 * Call this function when you know "name" given by the end user must
1481 * name an object but it doesn't; the function _may_ die with a better
1482 * diagnostic message than "no such object 'name'", e.g. "Path 'doc' does not
1483 * exist in 'HEAD'" when given "HEAD:doc", or it may return in which case
1484 * you have a chance to diagnose the error further.
1485 */
1486 void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
1487 {
1488 struct object_context oc;
1489 unsigned char sha1[20];
1490 get_sha1_with_context_1(name, GET_SHA1_ONLY_TO_DIE, prefix, sha1, &oc);
1491 }
1492
1493 int get_sha1_with_context(const char *str, unsigned flags, unsigned char *sha1, struct object_context *orc)
1494 {
1495 if (flags & GET_SHA1_FOLLOW_SYMLINKS && flags & GET_SHA1_ONLY_TO_DIE)
1496 die("BUG: incompatible flags for get_sha1_with_context");
1497 return get_sha1_with_context_1(str, flags, NULL, sha1, orc);
1498 }