commit-reach(paint_down_to_common): plug two memory leaks
[git.git] / commit.c
blobef679a0b939046c4aac15567cdf3c0ae8c079d29
1 #include "git-compat-util.h"
2 #include "tag.h"
3 #include "commit.h"
4 #include "commit-graph.h"
5 #include "environment.h"
6 #include "gettext.h"
7 #include "hex.h"
8 #include "repository.h"
9 #include "object-name.h"
10 #include "object-store-ll.h"
11 #include "utf8.h"
12 #include "diff.h"
13 #include "revision.h"
14 #include "notes.h"
15 #include "alloc.h"
16 #include "gpg-interface.h"
17 #include "mergesort.h"
18 #include "commit-slab.h"
19 #include "prio-queue.h"
20 #include "hash-lookup.h"
21 #include "wt-status.h"
22 #include "advice.h"
23 #include "refs.h"
24 #include "commit-reach.h"
25 #include "setup.h"
26 #include "shallow.h"
27 #include "tree.h"
28 #include "hook.h"
29 #include "parse.h"
31 static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
33 int save_commit_buffer = 1;
34 int no_graft_file_deprecated_advice;
36 const char *commit_type = "commit";
38 struct commit *lookup_commit_reference_gently(struct repository *r,
39 const struct object_id *oid, int quiet)
41 struct object *obj = deref_tag(r,
42 parse_object(r, oid),
43 NULL, 0);
45 if (!obj)
46 return NULL;
47 return object_as_type(obj, OBJ_COMMIT, quiet);
50 struct commit *lookup_commit_reference(struct repository *r, const struct object_id *oid)
52 return lookup_commit_reference_gently(r, oid, 0);
55 struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
57 struct commit *c = lookup_commit_reference(the_repository, oid);
58 if (!c)
59 die(_("could not parse %s"), ref_name);
60 if (!oideq(oid, &c->object.oid)) {
61 warning(_("%s %s is not a commit!"),
62 ref_name, oid_to_hex(oid));
64 return c;
67 struct commit *lookup_commit_object(struct repository *r,
68 const struct object_id *oid)
70 struct object *obj = parse_object(r, oid);
71 return obj ? object_as_type(obj, OBJ_COMMIT, 0) : NULL;
75 struct commit *lookup_commit(struct repository *r, const struct object_id *oid)
77 struct object *obj = lookup_object(r, oid);
78 if (!obj)
79 return create_object(r, oid, alloc_commit_node(r));
80 return object_as_type(obj, OBJ_COMMIT, 0);
83 struct commit *lookup_commit_reference_by_name(const char *name)
85 struct object_id oid;
86 struct commit *commit;
88 if (repo_get_oid_committish(the_repository, name, &oid))
89 return NULL;
90 commit = lookup_commit_reference(the_repository, &oid);
91 if (repo_parse_commit(the_repository, commit))
92 return NULL;
93 return commit;
96 static timestamp_t parse_commit_date(const char *buf, const char *tail)
98 const char *dateptr;
99 const char *eol;
101 if (buf + 6 >= tail)
102 return 0;
103 if (memcmp(buf, "author", 6))
104 return 0;
105 while (buf < tail && *buf++ != '\n')
106 /* nada */;
107 if (buf + 9 >= tail)
108 return 0;
109 if (memcmp(buf, "committer", 9))
110 return 0;
113 * Jump to end-of-line so that we can walk backwards to find the
114 * end-of-email ">". This is more forgiving of malformed cases
115 * because unexpected characters tend to be in the name and email
116 * fields.
118 eol = memchr(buf, '\n', tail - buf);
119 if (!eol)
120 return 0;
121 dateptr = eol;
122 while (dateptr > buf && dateptr[-1] != '>')
123 dateptr--;
124 if (dateptr == buf)
125 return 0;
128 * Trim leading whitespace, but make sure we have at least one
129 * non-whitespace character, as parse_timestamp() will otherwise walk
130 * right past the newline we found in "eol" when skipping whitespace
131 * itself.
133 * In theory it would be sufficient to allow any character not matched
134 * by isspace(), but there's a catch: our isspace() does not
135 * necessarily match the behavior of parse_timestamp(), as the latter
136 * is implemented by system routines which match more exotic control
137 * codes, or even locale-dependent sequences.
139 * Since we expect the timestamp to be a number, we can check for that.
140 * Anything else (e.g., a non-numeric token like "foo") would just
141 * cause parse_timestamp() to return 0 anyway.
143 while (dateptr < eol && isspace(*dateptr))
144 dateptr++;
145 if (!isdigit(*dateptr) && *dateptr != '-')
146 return 0;
149 * We know there is at least one digit (or dash), so we'll begin
150 * parsing there and stop at worst case at eol.
152 * Note that we may feed parse_timestamp() extra characters here if the
153 * commit is malformed, and it will parse as far as it can. For
154 * example, "123foo456" would return "123". That might be questionable
155 * (versus returning "0"), but it would help in a hypothetical case
156 * like "123456+0100", where the whitespace from the timezone is
157 * missing. Since such syntactic errors may be baked into history and
158 * hard to correct now, let's err on trying to make our best guess
159 * here, rather than insist on perfect syntax.
161 return parse_timestamp(dateptr, NULL, 10);
164 static const struct object_id *commit_graft_oid_access(size_t index, const void *table)
166 const struct commit_graft * const *commit_graft_table = table;
167 return &commit_graft_table[index]->oid;
170 int commit_graft_pos(struct repository *r, const struct object_id *oid)
172 return oid_pos(oid, r->parsed_objects->grafts,
173 r->parsed_objects->grafts_nr,
174 commit_graft_oid_access);
177 static void unparse_commit(struct repository *r, const struct object_id *oid)
179 struct commit *c = lookup_commit(r, oid);
181 if (!c->object.parsed)
182 return;
183 free_commit_list(c->parents);
184 c->parents = NULL;
185 c->object.parsed = 0;
188 int register_commit_graft(struct repository *r, struct commit_graft *graft,
189 int ignore_dups)
191 int pos = commit_graft_pos(r, &graft->oid);
193 if (0 <= pos) {
194 if (ignore_dups)
195 free(graft);
196 else {
197 free(r->parsed_objects->grafts[pos]);
198 r->parsed_objects->grafts[pos] = graft;
200 return 1;
202 pos = -pos - 1;
203 ALLOC_GROW(r->parsed_objects->grafts,
204 r->parsed_objects->grafts_nr + 1,
205 r->parsed_objects->grafts_alloc);
206 r->parsed_objects->grafts_nr++;
207 if (pos < r->parsed_objects->grafts_nr)
208 memmove(r->parsed_objects->grafts + pos + 1,
209 r->parsed_objects->grafts + pos,
210 (r->parsed_objects->grafts_nr - pos - 1) *
211 sizeof(*r->parsed_objects->grafts));
212 r->parsed_objects->grafts[pos] = graft;
213 unparse_commit(r, &graft->oid);
214 return 0;
217 struct commit_graft *read_graft_line(struct strbuf *line)
219 /* The format is just "Commit Parent1 Parent2 ...\n" */
220 int i, phase;
221 const char *tail = NULL;
222 struct commit_graft *graft = NULL;
223 struct object_id dummy_oid, *oid;
225 strbuf_rtrim(line);
226 if (!line->len || line->buf[0] == '#')
227 return NULL;
229 * phase 0 verifies line, counts hashes in line and allocates graft
230 * phase 1 fills graft
232 for (phase = 0; phase < 2; phase++) {
233 oid = graft ? &graft->oid : &dummy_oid;
234 if (parse_oid_hex(line->buf, oid, &tail))
235 goto bad_graft_data;
236 for (i = 0; *tail != '\0'; i++) {
237 oid = graft ? &graft->parent[i] : &dummy_oid;
238 if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail))
239 goto bad_graft_data;
241 if (!graft) {
242 graft = xmalloc(st_add(sizeof(*graft),
243 st_mult(sizeof(struct object_id), i)));
244 graft->nr_parent = i;
247 return graft;
249 bad_graft_data:
250 error("bad graft data: %s", line->buf);
251 assert(!graft);
252 return NULL;
255 static int read_graft_file(struct repository *r, const char *graft_file)
257 FILE *fp = fopen_or_warn(graft_file, "r");
258 struct strbuf buf = STRBUF_INIT;
259 if (!fp)
260 return -1;
261 if (!no_graft_file_deprecated_advice &&
262 advice_enabled(ADVICE_GRAFT_FILE_DEPRECATED))
263 advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n"
264 "and will be removed in a future Git version.\n"
265 "\n"
266 "Please use \"git replace --convert-graft-file\"\n"
267 "to convert the grafts into replace refs.\n"
268 "\n"
269 "Turn this message off by running\n"
270 "\"git config advice.graftFileDeprecated false\""));
271 while (!strbuf_getwholeline(&buf, fp, '\n')) {
272 /* The format is just "Commit Parent1 Parent2 ...\n" */
273 struct commit_graft *graft = read_graft_line(&buf);
274 if (!graft)
275 continue;
276 if (register_commit_graft(r, graft, 1))
277 error("duplicate graft data: %s", buf.buf);
279 fclose(fp);
280 strbuf_release(&buf);
281 return 0;
284 void prepare_commit_graft(struct repository *r)
286 char *graft_file;
288 if (r->parsed_objects->commit_graft_prepared)
289 return;
290 if (!startup_info->have_repository)
291 return;
293 graft_file = get_graft_file(r);
294 read_graft_file(r, graft_file);
295 /* make sure shallows are read */
296 is_repository_shallow(r);
297 r->parsed_objects->commit_graft_prepared = 1;
300 struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid)
302 int pos;
303 prepare_commit_graft(r);
304 pos = commit_graft_pos(r, oid);
305 if (pos < 0)
306 return NULL;
307 return r->parsed_objects->grafts[pos];
310 int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
312 int i, ret;
313 for (i = ret = 0; i < the_repository->parsed_objects->grafts_nr && !ret; i++)
314 ret = fn(the_repository->parsed_objects->grafts[i], cb_data);
315 return ret;
318 void reset_commit_grafts(struct repository *r)
320 int i;
322 for (i = 0; i < r->parsed_objects->grafts_nr; i++) {
323 unparse_commit(r, &r->parsed_objects->grafts[i]->oid);
324 free(r->parsed_objects->grafts[i]);
326 r->parsed_objects->grafts_nr = 0;
327 r->parsed_objects->commit_graft_prepared = 0;
330 struct commit_buffer {
331 void *buffer;
332 unsigned long size;
334 define_commit_slab(buffer_slab, struct commit_buffer);
336 struct buffer_slab *allocate_commit_buffer_slab(void)
338 struct buffer_slab *bs = xmalloc(sizeof(*bs));
339 init_buffer_slab(bs);
340 return bs;
343 void free_commit_buffer_slab(struct buffer_slab *bs)
345 clear_buffer_slab(bs);
346 free(bs);
349 void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size)
351 struct commit_buffer *v = buffer_slab_at(
352 r->parsed_objects->buffer_slab, commit);
353 v->buffer = buffer;
354 v->size = size;
357 const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep)
359 struct commit_buffer *v = buffer_slab_peek(
360 r->parsed_objects->buffer_slab, commit);
361 if (!v) {
362 if (sizep)
363 *sizep = 0;
364 return NULL;
366 if (sizep)
367 *sizep = v->size;
368 return v->buffer;
371 const void *repo_get_commit_buffer(struct repository *r,
372 const struct commit *commit,
373 unsigned long *sizep)
375 const void *ret = get_cached_commit_buffer(r, commit, sizep);
376 if (!ret) {
377 enum object_type type;
378 unsigned long size;
379 ret = repo_read_object_file(r, &commit->object.oid, &type, &size);
380 if (!ret)
381 die("cannot read commit object %s",
382 oid_to_hex(&commit->object.oid));
383 if (type != OBJ_COMMIT)
384 die("expected commit for %s, got %s",
385 oid_to_hex(&commit->object.oid), type_name(type));
386 if (sizep)
387 *sizep = size;
389 return ret;
392 void repo_unuse_commit_buffer(struct repository *r,
393 const struct commit *commit,
394 const void *buffer)
396 struct commit_buffer *v = buffer_slab_peek(
397 r->parsed_objects->buffer_slab, commit);
398 if (!(v && v->buffer == buffer))
399 free((void *)buffer);
402 void free_commit_buffer(struct parsed_object_pool *pool, struct commit *commit)
404 struct commit_buffer *v = buffer_slab_peek(
405 pool->buffer_slab, commit);
406 if (v) {
407 FREE_AND_NULL(v->buffer);
408 v->size = 0;
412 static inline void set_commit_tree(struct commit *c, struct tree *t)
414 c->maybe_tree = t;
417 struct tree *repo_get_commit_tree(struct repository *r,
418 const struct commit *commit)
420 if (commit->maybe_tree || !commit->object.parsed)
421 return commit->maybe_tree;
423 if (commit_graph_position(commit) != COMMIT_NOT_FROM_GRAPH)
424 return get_commit_tree_in_graph(r, commit);
426 return NULL;
429 struct object_id *get_commit_tree_oid(const struct commit *commit)
431 struct tree *tree = repo_get_commit_tree(the_repository, commit);
432 return tree ? &tree->object.oid : NULL;
435 void release_commit_memory(struct parsed_object_pool *pool, struct commit *c)
437 set_commit_tree(c, NULL);
438 free_commit_buffer(pool, c);
439 c->index = 0;
440 free_commit_list(c->parents);
442 c->object.parsed = 0;
445 const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
447 struct commit_buffer *v = buffer_slab_peek(
448 the_repository->parsed_objects->buffer_slab, commit);
449 void *ret;
451 if (!v) {
452 if (sizep)
453 *sizep = 0;
454 return NULL;
456 ret = v->buffer;
457 if (sizep)
458 *sizep = v->size;
460 v->buffer = NULL;
461 v->size = 0;
462 return ret;
465 int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
467 const char *tail = buffer;
468 const char *bufptr = buffer;
469 struct object_id parent;
470 struct commit_list **pptr;
471 struct commit_graft *graft;
472 const int tree_entry_len = the_hash_algo->hexsz + 5;
473 const int parent_entry_len = the_hash_algo->hexsz + 7;
474 struct tree *tree;
476 if (item->object.parsed)
477 return 0;
479 * Presumably this is leftover from an earlier failed parse;
480 * clear it out in preparation for us re-parsing (we'll hit the
481 * same error, but that's good, since it lets our caller know
482 * the result cannot be trusted.
484 free_commit_list(item->parents);
485 item->parents = NULL;
487 tail += size;
488 if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
489 bufptr[tree_entry_len] != '\n')
490 return error("bogus commit object %s", oid_to_hex(&item->object.oid));
491 if (get_oid_hex(bufptr + 5, &parent) < 0)
492 return error("bad tree pointer in commit %s",
493 oid_to_hex(&item->object.oid));
494 tree = lookup_tree(r, &parent);
495 if (!tree)
496 return error("bad tree pointer %s in commit %s",
497 oid_to_hex(&parent),
498 oid_to_hex(&item->object.oid));
499 set_commit_tree(item, tree);
500 bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
501 pptr = &item->parents;
503 graft = lookup_commit_graft(r, &item->object.oid);
504 if (graft)
505 r->parsed_objects->substituted_parent = 1;
506 while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
507 struct commit *new_parent;
509 if (tail <= bufptr + parent_entry_len + 1 ||
510 get_oid_hex(bufptr + 7, &parent) ||
511 bufptr[parent_entry_len] != '\n')
512 return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
513 bufptr += parent_entry_len + 1;
515 * The clone is shallow if nr_parent < 0, and we must
516 * not traverse its real parents even when we unhide them.
518 if (graft && (graft->nr_parent < 0 || !grafts_keep_true_parents))
519 continue;
520 new_parent = lookup_commit(r, &parent);
521 if (!new_parent)
522 return error("bad parent %s in commit %s",
523 oid_to_hex(&parent),
524 oid_to_hex(&item->object.oid));
525 pptr = &commit_list_insert(new_parent, pptr)->next;
527 if (graft) {
528 int i;
529 struct commit *new_parent;
530 for (i = 0; i < graft->nr_parent; i++) {
531 new_parent = lookup_commit(r,
532 &graft->parent[i]);
533 if (!new_parent)
534 return error("bad graft parent %s in commit %s",
535 oid_to_hex(&graft->parent[i]),
536 oid_to_hex(&item->object.oid));
537 pptr = &commit_list_insert(new_parent, pptr)->next;
540 item->date = parse_commit_date(bufptr, tail);
542 if (check_graph)
543 load_commit_graph_info(r, item);
545 item->object.parsed = 1;
546 return 0;
549 int repo_parse_commit_internal(struct repository *r,
550 struct commit *item,
551 int quiet_on_missing,
552 int use_commit_graph)
554 enum object_type type;
555 void *buffer;
556 unsigned long size;
557 struct object_info oi = {
558 .typep = &type,
559 .sizep = &size,
560 .contentp = &buffer,
563 * Git does not support partial clones that exclude commits, so set
564 * OBJECT_INFO_SKIP_FETCH_OBJECT to fail fast when an object is missing.
566 int flags = OBJECT_INFO_LOOKUP_REPLACE | OBJECT_INFO_SKIP_FETCH_OBJECT |
567 OBJECT_INFO_DIE_IF_CORRUPT;
568 int ret;
570 if (!item)
571 return -1;
572 if (item->object.parsed)
573 return 0;
574 if (use_commit_graph && parse_commit_in_graph(r, item)) {
575 static int commit_graph_paranoia = -1;
577 if (commit_graph_paranoia == -1)
578 commit_graph_paranoia = git_env_bool(GIT_COMMIT_GRAPH_PARANOIA, 0);
580 if (commit_graph_paranoia && !has_object(r, &item->object.oid, 0)) {
581 unparse_commit(r, &item->object.oid);
582 return quiet_on_missing ? -1 :
583 error(_("commit %s exists in commit-graph but not in the object database"),
584 oid_to_hex(&item->object.oid));
587 return 0;
590 if (oid_object_info_extended(r, &item->object.oid, &oi, flags) < 0)
591 return quiet_on_missing ? -1 :
592 error("Could not read %s",
593 oid_to_hex(&item->object.oid));
594 if (type != OBJ_COMMIT) {
595 free(buffer);
596 return error("Object %s not a commit",
597 oid_to_hex(&item->object.oid));
600 ret = parse_commit_buffer(r, item, buffer, size, 0);
601 if (save_commit_buffer && !ret) {
602 set_commit_buffer(r, item, buffer, size);
603 return 0;
605 free(buffer);
606 return ret;
609 int repo_parse_commit_gently(struct repository *r,
610 struct commit *item, int quiet_on_missing)
612 return repo_parse_commit_internal(r, item, quiet_on_missing, 1);
615 void parse_commit_or_die(struct commit *item)
617 if (repo_parse_commit(the_repository, item))
618 die("unable to parse commit %s",
619 item ? oid_to_hex(&item->object.oid) : "(null)");
622 int find_commit_subject(const char *commit_buffer, const char **subject)
624 const char *eol;
625 const char *p = commit_buffer;
627 while (*p && (*p != '\n' || p[1] != '\n'))
628 p++;
629 if (*p) {
630 p = skip_blank_lines(p + 2);
631 eol = strchrnul(p, '\n');
632 } else
633 eol = p;
635 *subject = p;
637 return eol - p;
640 size_t commit_subject_length(const char *body)
642 const char *p = body;
643 while (*p) {
644 const char *next = skip_blank_lines(p);
645 if (next != p)
646 break;
647 p = strchrnul(p, '\n');
648 if (*p)
649 p++;
651 return p - body;
654 struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
656 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
657 new_list->item = item;
658 new_list->next = *list_p;
659 *list_p = new_list;
660 return new_list;
663 int commit_list_contains(struct commit *item, struct commit_list *list)
665 while (list) {
666 if (list->item == item)
667 return 1;
668 list = list->next;
671 return 0;
674 unsigned commit_list_count(const struct commit_list *l)
676 unsigned c = 0;
677 for (; l; l = l->next )
678 c++;
679 return c;
682 struct commit_list *copy_commit_list(struct commit_list *list)
684 struct commit_list *head = NULL;
685 struct commit_list **pp = &head;
686 while (list) {
687 pp = commit_list_append(list->item, pp);
688 list = list->next;
690 return head;
693 struct commit_list *reverse_commit_list(struct commit_list *list)
695 struct commit_list *next = NULL, *current, *backup;
696 for (current = list; current; current = backup) {
697 backup = current->next;
698 current->next = next;
699 next = current;
701 return next;
704 void free_commit_list(struct commit_list *list)
706 while (list)
707 pop_commit(&list);
710 struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
712 struct commit_list **pp = list;
713 struct commit_list *p;
714 while ((p = *pp) != NULL) {
715 if (p->item->date < item->date) {
716 break;
718 pp = &p->next;
720 return commit_list_insert(item, pp);
723 static int commit_list_compare_by_date(const struct commit_list *a,
724 const struct commit_list *b)
726 timestamp_t a_date = a->item->date;
727 timestamp_t b_date = b->item->date;
728 if (a_date < b_date)
729 return 1;
730 if (a_date > b_date)
731 return -1;
732 return 0;
735 DEFINE_LIST_SORT(static, commit_list_sort, struct commit_list, next);
737 void commit_list_sort_by_date(struct commit_list **list)
739 commit_list_sort(list, commit_list_compare_by_date);
742 struct commit *pop_most_recent_commit(struct commit_list **list,
743 unsigned int mark)
745 struct commit *ret = pop_commit(list);
746 struct commit_list *parents = ret->parents;
748 while (parents) {
749 struct commit *commit = parents->item;
750 if (!repo_parse_commit(the_repository, commit) && !(commit->object.flags & mark)) {
751 commit->object.flags |= mark;
752 commit_list_insert_by_date(commit, list);
754 parents = parents->next;
756 return ret;
759 static void clear_commit_marks_1(struct commit_list **plist,
760 struct commit *commit, unsigned int mark)
762 while (commit) {
763 struct commit_list *parents;
765 if (!(mark & commit->object.flags))
766 return;
768 commit->object.flags &= ~mark;
770 parents = commit->parents;
771 if (!parents)
772 return;
774 while ((parents = parents->next)) {
775 if (parents->item->object.flags & mark)
776 commit_list_insert(parents->item, plist);
779 commit = commit->parents->item;
783 void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
785 struct commit_list *list = NULL;
787 while (nr--) {
788 clear_commit_marks_1(&list, *commit, mark);
789 commit++;
791 while (list)
792 clear_commit_marks_1(&list, pop_commit(&list), mark);
795 void clear_commit_marks(struct commit *commit, unsigned int mark)
797 clear_commit_marks_many(1, &commit, mark);
800 struct commit *pop_commit(struct commit_list **stack)
802 struct commit_list *top = *stack;
803 struct commit *item = top ? top->item : NULL;
805 if (top) {
806 *stack = top->next;
807 free(top);
809 return item;
813 * Topological sort support
816 /* count number of children that have not been emitted */
817 define_commit_slab(indegree_slab, int);
819 define_commit_slab(author_date_slab, timestamp_t);
821 void record_author_date(struct author_date_slab *author_date,
822 struct commit *commit)
824 const char *buffer = repo_get_commit_buffer(the_repository, commit,
825 NULL);
826 struct ident_split ident;
827 const char *ident_line;
828 size_t ident_len;
829 char *date_end;
830 timestamp_t date;
832 ident_line = find_commit_header(buffer, "author", &ident_len);
833 if (!ident_line)
834 goto fail_exit; /* no author line */
835 if (split_ident_line(&ident, ident_line, ident_len) ||
836 !ident.date_begin || !ident.date_end)
837 goto fail_exit; /* malformed "author" line */
839 date = parse_timestamp(ident.date_begin, &date_end, 10);
840 if (date_end != ident.date_end)
841 goto fail_exit; /* malformed date */
842 *(author_date_slab_at(author_date, commit)) = date;
844 fail_exit:
845 repo_unuse_commit_buffer(the_repository, commit, buffer);
848 int compare_commits_by_author_date(const void *a_, const void *b_,
849 void *cb_data)
851 const struct commit *a = a_, *b = b_;
852 struct author_date_slab *author_date = cb_data;
853 timestamp_t a_date = *(author_date_slab_at(author_date, a));
854 timestamp_t b_date = *(author_date_slab_at(author_date, b));
856 /* newer commits with larger date first */
857 if (a_date < b_date)
858 return 1;
859 else if (a_date > b_date)
860 return -1;
861 return 0;
864 int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_,
865 void *unused UNUSED)
867 const struct commit *a = a_, *b = b_;
868 const timestamp_t generation_a = commit_graph_generation(a),
869 generation_b = commit_graph_generation(b);
871 /* newer commits first */
872 if (generation_a < generation_b)
873 return 1;
874 else if (generation_a > generation_b)
875 return -1;
877 /* use date as a heuristic when generations are equal */
878 if (a->date < b->date)
879 return 1;
880 else if (a->date > b->date)
881 return -1;
882 return 0;
885 int compare_commits_by_commit_date(const void *a_, const void *b_,
886 void *unused UNUSED)
888 const struct commit *a = a_, *b = b_;
889 /* newer commits with larger date first */
890 if (a->date < b->date)
891 return 1;
892 else if (a->date > b->date)
893 return -1;
894 return 0;
898 * Performs an in-place topological sort on the list supplied.
900 void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
902 struct commit_list *next, *orig = *list;
903 struct commit_list **pptr;
904 struct indegree_slab indegree;
905 struct prio_queue queue;
906 struct commit *commit;
907 struct author_date_slab author_date;
909 if (!orig)
910 return;
911 *list = NULL;
913 init_indegree_slab(&indegree);
914 memset(&queue, '\0', sizeof(queue));
916 switch (sort_order) {
917 default: /* REV_SORT_IN_GRAPH_ORDER */
918 queue.compare = NULL;
919 break;
920 case REV_SORT_BY_COMMIT_DATE:
921 queue.compare = compare_commits_by_commit_date;
922 break;
923 case REV_SORT_BY_AUTHOR_DATE:
924 init_author_date_slab(&author_date);
925 queue.compare = compare_commits_by_author_date;
926 queue.cb_data = &author_date;
927 break;
930 /* Mark them and clear the indegree */
931 for (next = orig; next; next = next->next) {
932 struct commit *commit = next->item;
933 *(indegree_slab_at(&indegree, commit)) = 1;
934 /* also record the author dates, if needed */
935 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
936 record_author_date(&author_date, commit);
939 /* update the indegree */
940 for (next = orig; next; next = next->next) {
941 struct commit_list *parents = next->item->parents;
942 while (parents) {
943 struct commit *parent = parents->item;
944 int *pi = indegree_slab_at(&indegree, parent);
946 if (*pi)
947 (*pi)++;
948 parents = parents->next;
953 * find the tips
955 * tips are nodes not reachable from any other node in the list
957 * the tips serve as a starting set for the work queue.
959 for (next = orig; next; next = next->next) {
960 struct commit *commit = next->item;
962 if (*(indegree_slab_at(&indegree, commit)) == 1)
963 prio_queue_put(&queue, commit);
967 * This is unfortunate; the initial tips need to be shown
968 * in the order given from the revision traversal machinery.
970 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
971 prio_queue_reverse(&queue);
973 /* We no longer need the commit list */
974 free_commit_list(orig);
976 pptr = list;
977 *list = NULL;
978 while ((commit = prio_queue_get(&queue)) != NULL) {
979 struct commit_list *parents;
981 for (parents = commit->parents; parents ; parents = parents->next) {
982 struct commit *parent = parents->item;
983 int *pi = indegree_slab_at(&indegree, parent);
985 if (!*pi)
986 continue;
989 * parents are only enqueued for emission
990 * when all their children have been emitted thereby
991 * guaranteeing topological order.
993 if (--(*pi) == 1)
994 prio_queue_put(&queue, parent);
997 * all children of commit have already been
998 * emitted. we can emit it now.
1000 *(indegree_slab_at(&indegree, commit)) = 0;
1002 pptr = &commit_list_insert(commit, pptr)->next;
1005 clear_indegree_slab(&indegree);
1006 clear_prio_queue(&queue);
1007 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
1008 clear_author_date_slab(&author_date);
1011 struct rev_collect {
1012 struct commit **commit;
1013 int nr;
1014 int alloc;
1015 unsigned int initial : 1;
1018 static void add_one_commit(struct object_id *oid, struct rev_collect *revs)
1020 struct commit *commit;
1022 if (is_null_oid(oid))
1023 return;
1025 commit = lookup_commit(the_repository, oid);
1026 if (!commit ||
1027 (commit->object.flags & TMP_MARK) ||
1028 repo_parse_commit(the_repository, commit))
1029 return;
1031 ALLOC_GROW(revs->commit, revs->nr + 1, revs->alloc);
1032 revs->commit[revs->nr++] = commit;
1033 commit->object.flags |= TMP_MARK;
1036 static int collect_one_reflog_ent(struct object_id *ooid, struct object_id *noid,
1037 const char *ident UNUSED,
1038 timestamp_t timestamp UNUSED, int tz UNUSED,
1039 const char *message UNUSED, void *cbdata)
1041 struct rev_collect *revs = cbdata;
1043 if (revs->initial) {
1044 revs->initial = 0;
1045 add_one_commit(ooid, revs);
1047 add_one_commit(noid, revs);
1048 return 0;
1051 struct commit *get_fork_point(const char *refname, struct commit *commit)
1053 struct object_id oid;
1054 struct rev_collect revs;
1055 struct commit_list *bases;
1056 int i;
1057 struct commit *ret = NULL;
1058 char *full_refname;
1060 switch (repo_dwim_ref(the_repository, refname, strlen(refname), &oid,
1061 &full_refname, 0)) {
1062 case 0:
1063 die("No such ref: '%s'", refname);
1064 case 1:
1065 break; /* good */
1066 default:
1067 die("Ambiguous refname: '%s'", refname);
1070 memset(&revs, 0, sizeof(revs));
1071 revs.initial = 1;
1072 for_each_reflog_ent(full_refname, collect_one_reflog_ent, &revs);
1074 if (!revs.nr)
1075 add_one_commit(&oid, &revs);
1077 for (i = 0; i < revs.nr; i++)
1078 revs.commit[i]->object.flags &= ~TMP_MARK;
1080 bases = repo_get_merge_bases_many(the_repository, commit, revs.nr,
1081 revs.commit);
1084 * There should be one and only one merge base, when we found
1085 * a common ancestor among reflog entries.
1087 if (!bases || bases->next)
1088 goto cleanup_return;
1090 /* And the found one must be one of the reflog entries */
1091 for (i = 0; i < revs.nr; i++)
1092 if (&bases->item->object == &revs.commit[i]->object)
1093 break; /* found */
1094 if (revs.nr <= i)
1095 goto cleanup_return;
1097 ret = bases->item;
1099 cleanup_return:
1100 free(revs.commit);
1101 free_commit_list(bases);
1102 free(full_refname);
1103 return ret;
1107 * Indexed by hash algorithm identifier.
1109 static const char *gpg_sig_headers[] = {
1110 NULL,
1111 "gpgsig",
1112 "gpgsig-sha256",
1115 int sign_with_header(struct strbuf *buf, const char *keyid)
1117 struct strbuf sig = STRBUF_INIT;
1118 int inspos, copypos;
1119 const char *eoh;
1120 const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(the_hash_algo)];
1121 int gpg_sig_header_len = strlen(gpg_sig_header);
1123 /* find the end of the header */
1124 eoh = strstr(buf->buf, "\n\n");
1125 if (!eoh)
1126 inspos = buf->len;
1127 else
1128 inspos = eoh - buf->buf + 1;
1130 if (!keyid || !*keyid)
1131 keyid = get_signing_key();
1132 if (sign_buffer(buf, &sig, keyid)) {
1133 strbuf_release(&sig);
1134 return -1;
1137 for (copypos = 0; sig.buf[copypos]; ) {
1138 const char *bol = sig.buf + copypos;
1139 const char *eol = strchrnul(bol, '\n');
1140 int len = (eol - bol) + !!*eol;
1142 if (!copypos) {
1143 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
1144 inspos += gpg_sig_header_len;
1146 strbuf_insertstr(buf, inspos++, " ");
1147 strbuf_insert(buf, inspos, bol, len);
1148 inspos += len;
1149 copypos += len;
1151 strbuf_release(&sig);
1152 return 0;
1157 int parse_signed_commit(const struct commit *commit,
1158 struct strbuf *payload, struct strbuf *signature,
1159 const struct git_hash_algo *algop)
1161 unsigned long size;
1162 const char *buffer = repo_get_commit_buffer(the_repository, commit,
1163 &size);
1164 int ret = parse_buffer_signed_by_header(buffer, size, payload, signature, algop);
1166 repo_unuse_commit_buffer(the_repository, commit, buffer);
1167 return ret;
1170 int parse_buffer_signed_by_header(const char *buffer,
1171 unsigned long size,
1172 struct strbuf *payload,
1173 struct strbuf *signature,
1174 const struct git_hash_algo *algop)
1176 int in_signature = 0, saw_signature = 0, other_signature = 0;
1177 const char *line, *tail, *p;
1178 const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(algop)];
1180 line = buffer;
1181 tail = buffer + size;
1182 while (line < tail) {
1183 const char *sig = NULL;
1184 const char *next = memchr(line, '\n', tail - line);
1186 next = next ? next + 1 : tail;
1187 if (in_signature && line[0] == ' ')
1188 sig = line + 1;
1189 else if (skip_prefix(line, gpg_sig_header, &p) &&
1190 *p == ' ') {
1191 sig = line + strlen(gpg_sig_header) + 1;
1192 other_signature = 0;
1194 else if (starts_with(line, "gpgsig"))
1195 other_signature = 1;
1196 else if (other_signature && line[0] != ' ')
1197 other_signature = 0;
1198 if (sig) {
1199 strbuf_add(signature, sig, next - sig);
1200 saw_signature = 1;
1201 in_signature = 1;
1202 } else {
1203 if (*line == '\n')
1204 /* dump the whole remainder of the buffer */
1205 next = tail;
1206 if (!other_signature)
1207 strbuf_add(payload, line, next - line);
1208 in_signature = 0;
1210 line = next;
1212 return saw_signature;
1215 int remove_signature(struct strbuf *buf)
1217 const char *line = buf->buf;
1218 const char *tail = buf->buf + buf->len;
1219 int in_signature = 0;
1220 struct sigbuf {
1221 const char *start;
1222 const char *end;
1223 } sigs[2], *sigp = &sigs[0];
1224 int i;
1225 const char *orig_buf = buf->buf;
1227 memset(sigs, 0, sizeof(sigs));
1229 while (line < tail) {
1230 const char *next = memchr(line, '\n', tail - line);
1231 next = next ? next + 1 : tail;
1233 if (in_signature && line[0] == ' ')
1234 sigp->end = next;
1235 else if (starts_with(line, "gpgsig")) {
1236 int i;
1237 for (i = 1; i < GIT_HASH_NALGOS; i++) {
1238 const char *p;
1239 if (skip_prefix(line, gpg_sig_headers[i], &p) &&
1240 *p == ' ') {
1241 sigp->start = line;
1242 sigp->end = next;
1243 in_signature = 1;
1246 } else {
1247 if (*line == '\n')
1248 /* dump the whole remainder of the buffer */
1249 next = tail;
1250 if (in_signature && sigp - sigs != ARRAY_SIZE(sigs))
1251 sigp++;
1252 in_signature = 0;
1254 line = next;
1257 for (i = ARRAY_SIZE(sigs) - 1; i >= 0; i--)
1258 if (sigs[i].start)
1259 strbuf_remove(buf, sigs[i].start - orig_buf, sigs[i].end - sigs[i].start);
1261 return sigs[0].start != NULL;
1264 static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
1266 struct merge_remote_desc *desc;
1267 struct commit_extra_header *mergetag;
1268 char *buf;
1269 unsigned long size;
1270 enum object_type type;
1271 struct strbuf payload = STRBUF_INIT;
1272 struct strbuf signature = STRBUF_INIT;
1274 desc = merge_remote_util(parent);
1275 if (!desc || !desc->obj)
1276 return;
1277 buf = repo_read_object_file(the_repository, &desc->obj->oid, &type,
1278 &size);
1279 if (!buf || type != OBJ_TAG)
1280 goto free_return;
1281 if (!parse_signature(buf, size, &payload, &signature))
1282 goto free_return;
1284 * We could verify this signature and either omit the tag when
1285 * it does not validate, but the integrator may not have the
1286 * public key of the signer of the tag being merged, while a
1287 * later auditor may have it while auditing, so let's not run
1288 * verify-signed-buffer here for now...
1290 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
1291 * warn("warning: signed tag unverified.");
1293 CALLOC_ARRAY(mergetag, 1);
1294 mergetag->key = xstrdup("mergetag");
1295 mergetag->value = buf;
1296 mergetag->len = size;
1298 **tail = mergetag;
1299 *tail = &mergetag->next;
1300 strbuf_release(&payload);
1301 strbuf_release(&signature);
1302 return;
1304 free_return:
1305 free(buf);
1308 int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1310 struct strbuf payload = STRBUF_INIT;
1311 struct strbuf signature = STRBUF_INIT;
1312 int ret = 1;
1314 sigc->result = 'N';
1316 if (parse_signed_commit(commit, &payload, &signature, the_hash_algo) <= 0)
1317 goto out;
1319 sigc->payload_type = SIGNATURE_PAYLOAD_COMMIT;
1320 sigc->payload = strbuf_detach(&payload, &sigc->payload_len);
1321 ret = check_signature(sigc, signature.buf, signature.len);
1323 out:
1324 strbuf_release(&payload);
1325 strbuf_release(&signature);
1327 return ret;
1330 void verify_merge_signature(struct commit *commit, int verbosity,
1331 int check_trust)
1333 char hex[GIT_MAX_HEXSZ + 1];
1334 struct signature_check signature_check;
1335 int ret;
1336 memset(&signature_check, 0, sizeof(signature_check));
1338 ret = check_commit_signature(commit, &signature_check);
1340 repo_find_unique_abbrev_r(the_repository, hex, &commit->object.oid,
1341 DEFAULT_ABBREV);
1342 switch (signature_check.result) {
1343 case 'G':
1344 if (ret || (check_trust && signature_check.trust_level < TRUST_MARGINAL))
1345 die(_("Commit %s has an untrusted GPG signature, "
1346 "allegedly by %s."), hex, signature_check.signer);
1347 break;
1348 case 'B':
1349 die(_("Commit %s has a bad GPG signature "
1350 "allegedly by %s."), hex, signature_check.signer);
1351 default: /* 'N' */
1352 die(_("Commit %s does not have a GPG signature."), hex);
1354 if (verbosity >= 0 && signature_check.result == 'G')
1355 printf(_("Commit %s has a good GPG signature by %s\n"),
1356 hex, signature_check.signer);
1358 signature_check_clear(&signature_check);
1361 void append_merge_tag_headers(struct commit_list *parents,
1362 struct commit_extra_header ***tail)
1364 while (parents) {
1365 struct commit *parent = parents->item;
1366 handle_signed_tag(parent, tail);
1367 parents = parents->next;
1371 static void add_extra_header(struct strbuf *buffer,
1372 struct commit_extra_header *extra)
1374 strbuf_addstr(buffer, extra->key);
1375 if (extra->len)
1376 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1377 else
1378 strbuf_addch(buffer, '\n');
1381 struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1382 const char **exclude)
1384 struct commit_extra_header *extra = NULL;
1385 unsigned long size;
1386 const char *buffer = repo_get_commit_buffer(the_repository, commit,
1387 &size);
1388 extra = read_commit_extra_header_lines(buffer, size, exclude);
1389 repo_unuse_commit_buffer(the_repository, commit, buffer);
1390 return extra;
1393 int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1395 struct commit_extra_header *extra, *to_free;
1396 int res = 0;
1398 to_free = read_commit_extra_headers(commit, NULL);
1399 for (extra = to_free; !res && extra; extra = extra->next) {
1400 if (strcmp(extra->key, "mergetag"))
1401 continue; /* not a merge tag */
1402 res = fn(commit, extra, data);
1404 free_commit_extra_headers(to_free);
1405 return res;
1408 static inline int standard_header_field(const char *field, size_t len)
1410 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1411 (len == 6 && !memcmp(field, "parent", 6)) ||
1412 (len == 6 && !memcmp(field, "author", 6)) ||
1413 (len == 9 && !memcmp(field, "committer", 9)) ||
1414 (len == 8 && !memcmp(field, "encoding", 8)));
1417 static int excluded_header_field(const char *field, size_t len, const char **exclude)
1419 if (!exclude)
1420 return 0;
1422 while (*exclude) {
1423 size_t xlen = strlen(*exclude);
1424 if (len == xlen && !memcmp(field, *exclude, xlen))
1425 return 1;
1426 exclude++;
1428 return 0;
1431 static struct commit_extra_header *read_commit_extra_header_lines(
1432 const char *buffer, size_t size,
1433 const char **exclude)
1435 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1436 const char *line, *next, *eof, *eob;
1437 struct strbuf buf = STRBUF_INIT;
1439 for (line = buffer, eob = line + size;
1440 line < eob && *line != '\n';
1441 line = next) {
1442 next = memchr(line, '\n', eob - line);
1443 next = next ? next + 1 : eob;
1444 if (*line == ' ') {
1445 /* continuation */
1446 if (it)
1447 strbuf_add(&buf, line + 1, next - (line + 1));
1448 continue;
1450 if (it)
1451 it->value = strbuf_detach(&buf, &it->len);
1452 strbuf_reset(&buf);
1453 it = NULL;
1455 eof = memchr(line, ' ', next - line);
1456 if (!eof)
1457 eof = next;
1458 else if (standard_header_field(line, eof - line) ||
1459 excluded_header_field(line, eof - line, exclude))
1460 continue;
1462 CALLOC_ARRAY(it, 1);
1463 it->key = xmemdupz(line, eof-line);
1464 *tail = it;
1465 tail = &it->next;
1466 if (eof + 1 < next)
1467 strbuf_add(&buf, eof + 1, next - (eof + 1));
1469 if (it)
1470 it->value = strbuf_detach(&buf, &it->len);
1471 return extra;
1474 void free_commit_extra_headers(struct commit_extra_header *extra)
1476 while (extra) {
1477 struct commit_extra_header *next = extra->next;
1478 free(extra->key);
1479 free(extra->value);
1480 free(extra);
1481 extra = next;
1485 int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1486 struct commit_list *parents, struct object_id *ret,
1487 const char *author, const char *sign_commit)
1489 struct commit_extra_header *extra = NULL, **tail = &extra;
1490 int result;
1492 append_merge_tag_headers(parents, &tail);
1493 result = commit_tree_extended(msg, msg_len, tree, parents, ret, author,
1494 NULL, sign_commit, extra);
1495 free_commit_extra_headers(extra);
1496 return result;
1499 static int find_invalid_utf8(const char *buf, int len)
1501 int offset = 0;
1502 static const unsigned int max_codepoint[] = {
1503 0x7f, 0x7ff, 0xffff, 0x10ffff
1506 while (len) {
1507 unsigned char c = *buf++;
1508 int bytes, bad_offset;
1509 unsigned int codepoint;
1510 unsigned int min_val, max_val;
1512 len--;
1513 offset++;
1515 /* Simple US-ASCII? No worries. */
1516 if (c < 0x80)
1517 continue;
1519 bad_offset = offset-1;
1522 * Count how many more high bits set: that's how
1523 * many more bytes this sequence should have.
1525 bytes = 0;
1526 while (c & 0x40) {
1527 c <<= 1;
1528 bytes++;
1532 * Must be between 1 and 3 more bytes. Longer sequences result in
1533 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1535 if (bytes < 1 || 3 < bytes)
1536 return bad_offset;
1538 /* Do we *have* that many bytes? */
1539 if (len < bytes)
1540 return bad_offset;
1543 * Place the encoded bits at the bottom of the value and compute the
1544 * valid range.
1546 codepoint = (c & 0x7f) >> bytes;
1547 min_val = max_codepoint[bytes-1] + 1;
1548 max_val = max_codepoint[bytes];
1550 offset += bytes;
1551 len -= bytes;
1553 /* And verify that they are good continuation bytes */
1554 do {
1555 codepoint <<= 6;
1556 codepoint |= *buf & 0x3f;
1557 if ((*buf++ & 0xc0) != 0x80)
1558 return bad_offset;
1559 } while (--bytes);
1561 /* Reject codepoints that are out of range for the sequence length. */
1562 if (codepoint < min_val || codepoint > max_val)
1563 return bad_offset;
1564 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1565 if ((codepoint & 0x1ff800) == 0xd800)
1566 return bad_offset;
1567 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1568 if ((codepoint & 0xfffe) == 0xfffe)
1569 return bad_offset;
1570 /* So are anything in the range U+FDD0..U+FDEF. */
1571 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1572 return bad_offset;
1574 return -1;
1578 * This verifies that the buffer is in proper utf8 format.
1580 * If it isn't, it assumes any non-utf8 characters are Latin1,
1581 * and does the conversion.
1583 static int verify_utf8(struct strbuf *buf)
1585 int ok = 1;
1586 long pos = 0;
1588 for (;;) {
1589 int bad;
1590 unsigned char c;
1591 unsigned char replace[2];
1593 bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1594 if (bad < 0)
1595 return ok;
1596 pos += bad;
1597 ok = 0;
1598 c = buf->buf[pos];
1599 strbuf_remove(buf, pos, 1);
1601 /* We know 'c' must be in the range 128-255 */
1602 replace[0] = 0xc0 + (c >> 6);
1603 replace[1] = 0x80 + (c & 0x3f);
1604 strbuf_insert(buf, pos, replace, 2);
1605 pos += 2;
1609 static const char commit_utf8_warn[] =
1610 N_("Warning: commit message did not conform to UTF-8.\n"
1611 "You may want to amend it after fixing the message, or set the config\n"
1612 "variable i18n.commitEncoding to the encoding your project uses.\n");
1614 int commit_tree_extended(const char *msg, size_t msg_len,
1615 const struct object_id *tree,
1616 struct commit_list *parents, struct object_id *ret,
1617 const char *author, const char *committer,
1618 const char *sign_commit,
1619 struct commit_extra_header *extra)
1621 int result;
1622 int encoding_is_utf8;
1623 struct strbuf buffer;
1625 assert_oid_type(tree, OBJ_TREE);
1627 if (memchr(msg, '\0', msg_len))
1628 return error("a NUL byte in commit log message not allowed.");
1630 /* Not having i18n.commitencoding is the same as having utf-8 */
1631 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1633 strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1634 strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1637 * NOTE! This ordering means that the same exact tree merged with a
1638 * different order of parents will be a _different_ changeset even
1639 * if everything else stays the same.
1641 while (parents) {
1642 struct commit *parent = pop_commit(&parents);
1643 strbuf_addf(&buffer, "parent %s\n",
1644 oid_to_hex(&parent->object.oid));
1647 /* Person/date information */
1648 if (!author)
1649 author = git_author_info(IDENT_STRICT);
1650 strbuf_addf(&buffer, "author %s\n", author);
1651 if (!committer)
1652 committer = git_committer_info(IDENT_STRICT);
1653 strbuf_addf(&buffer, "committer %s\n", committer);
1654 if (!encoding_is_utf8)
1655 strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1657 while (extra) {
1658 add_extra_header(&buffer, extra);
1659 extra = extra->next;
1661 strbuf_addch(&buffer, '\n');
1663 /* And add the comment */
1664 strbuf_add(&buffer, msg, msg_len);
1666 /* And check the encoding */
1667 if (encoding_is_utf8 && !verify_utf8(&buffer))
1668 fprintf(stderr, _(commit_utf8_warn));
1670 if (sign_commit && sign_with_header(&buffer, sign_commit)) {
1671 result = -1;
1672 goto out;
1675 result = write_object_file(buffer.buf, buffer.len, OBJ_COMMIT, ret);
1676 out:
1677 strbuf_release(&buffer);
1678 return result;
1681 define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1682 static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1684 struct merge_remote_desc *merge_remote_util(struct commit *commit)
1686 return *merge_desc_slab_at(&merge_desc_slab, commit);
1689 void set_merge_remote_desc(struct commit *commit,
1690 const char *name, struct object *obj)
1692 struct merge_remote_desc *desc;
1693 FLEX_ALLOC_STR(desc, name, name);
1694 desc->obj = obj;
1695 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1698 struct commit *get_merge_parent(const char *name)
1700 struct object *obj;
1701 struct commit *commit;
1702 struct object_id oid;
1703 if (repo_get_oid(the_repository, name, &oid))
1704 return NULL;
1705 obj = parse_object(the_repository, &oid);
1706 commit = (struct commit *)repo_peel_to_type(the_repository, name, 0,
1707 obj, OBJ_COMMIT);
1708 if (commit && !merge_remote_util(commit))
1709 set_merge_remote_desc(commit, name, obj);
1710 return commit;
1714 * Append a commit to the end of the commit_list.
1716 * next starts by pointing to the variable that holds the head of an
1717 * empty commit_list, and is updated to point to the "next" field of
1718 * the last item on the list as new commits are appended.
1720 * Usage example:
1722 * struct commit_list *list;
1723 * struct commit_list **next = &list;
1725 * next = commit_list_append(c1, next);
1726 * next = commit_list_append(c2, next);
1727 * assert(commit_list_count(list) == 2);
1728 * return list;
1730 struct commit_list **commit_list_append(struct commit *commit,
1731 struct commit_list **next)
1733 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1734 new_commit->item = commit;
1735 *next = new_commit;
1736 new_commit->next = NULL;
1737 return &new_commit->next;
1740 const char *find_header_mem(const char *msg, size_t len,
1741 const char *key, size_t *out_len)
1743 int key_len = strlen(key);
1744 const char *line = msg;
1747 * NEEDSWORK: It's possible for strchrnul() to scan beyond the range
1748 * given by len. However, current callers are safe because they compute
1749 * len by scanning a NUL-terminated block of memory starting at msg.
1750 * Nonetheless, it would be better to ensure the function does not look
1751 * at msg beyond the len provided by the caller.
1753 while (line && line < msg + len) {
1754 const char *eol = strchrnul(line, '\n');
1756 if (line == eol)
1757 return NULL;
1759 if (eol - line > key_len &&
1760 !strncmp(line, key, key_len) &&
1761 line[key_len] == ' ') {
1762 *out_len = eol - line - key_len - 1;
1763 return line + key_len + 1;
1765 line = *eol ? eol + 1 : NULL;
1767 return NULL;
1770 const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1772 return find_header_mem(msg, strlen(msg), key, out_len);
1775 * Inspect the given string and determine the true "end" of the log message, in
1776 * order to find where to put a new Signed-off-by trailer. Ignored are
1777 * trailing comment lines and blank lines. To support "git commit -s
1778 * --amend" on an existing commit, we also ignore "Conflicts:". To
1779 * support "git commit -v", we truncate at cut lines.
1781 * Returns the number of bytes from the tail to ignore, to be fed as
1782 * the second parameter to append_signoff().
1784 size_t ignored_log_message_bytes(const char *buf, size_t len)
1786 size_t boc = 0;
1787 size_t bol = 0;
1788 int in_old_conflicts_block = 0;
1789 size_t cutoff = wt_status_locate_end(buf, len);
1791 while (bol < cutoff) {
1792 const char *next_line = memchr(buf + bol, '\n', len - bol);
1794 if (!next_line)
1795 next_line = buf + len;
1796 else
1797 next_line++;
1799 if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1800 /* is this the first of the run of comments? */
1801 if (!boc)
1802 boc = bol;
1803 /* otherwise, it is just continuing */
1804 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1805 in_old_conflicts_block = 1;
1806 if (!boc)
1807 boc = bol;
1808 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1809 ; /* a pathname in the conflicts block */
1810 } else if (boc) {
1811 /* the previous was not trailing comment */
1812 boc = 0;
1813 in_old_conflicts_block = 0;
1815 bol = next_line - buf;
1817 return boc ? len - boc : len - cutoff;
1820 int run_commit_hook(int editor_is_used, const char *index_file,
1821 int *invoked_hook, const char *name, ...)
1823 struct run_hooks_opt opt = RUN_HOOKS_OPT_INIT;
1824 va_list args;
1825 const char *arg;
1827 strvec_pushf(&opt.env, "GIT_INDEX_FILE=%s", index_file);
1830 * Let the hook know that no editor will be launched.
1832 if (!editor_is_used)
1833 strvec_push(&opt.env, "GIT_EDITOR=:");
1835 va_start(args, name);
1836 while ((arg = va_arg(args, const char *)))
1837 strvec_push(&opt.args, arg);
1838 va_end(args);
1840 opt.invoked_hook = invoked_hook;
1841 return run_hooks_opt(name, &opt);