Ninth batch
[git/debian.git] / commit.c
blob40890ae7ce8a48a2da06bcde6fc5a9db77260707
1 #include "cache.h"
2 #include "tag.h"
3 #include "commit.h"
4 #include "commit-graph.h"
5 #include "repository.h"
6 #include "object-store.h"
7 #include "pkt-line.h"
8 #include "utf8.h"
9 #include "diff.h"
10 #include "revision.h"
11 #include "notes.h"
12 #include "alloc.h"
13 #include "gpg-interface.h"
14 #include "mergesort.h"
15 #include "commit-slab.h"
16 #include "prio-queue.h"
17 #include "sha1-lookup.h"
18 #include "wt-status.h"
19 #include "advice.h"
20 #include "refs.h"
21 #include "commit-reach.h"
23 static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
25 int save_commit_buffer = 1;
27 const char *commit_type = "commit";
29 struct commit *lookup_commit_reference_gently(struct repository *r,
30 const struct object_id *oid, int quiet)
32 struct object *obj = deref_tag(r,
33 parse_object(r, oid),
34 NULL, 0);
36 if (!obj)
37 return NULL;
38 return object_as_type(r, obj, OBJ_COMMIT, quiet);
41 struct commit *lookup_commit_reference(struct repository *r, const struct object_id *oid)
43 return lookup_commit_reference_gently(r, oid, 0);
46 struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
48 struct commit *c = lookup_commit_reference(the_repository, oid);
49 if (!c)
50 die(_("could not parse %s"), ref_name);
51 if (!oideq(oid, &c->object.oid)) {
52 warning(_("%s %s is not a commit!"),
53 ref_name, oid_to_hex(oid));
55 return c;
58 struct commit *lookup_commit(struct repository *r, const struct object_id *oid)
60 struct object *obj = lookup_object(r, oid);
61 if (!obj)
62 return create_object(r, oid, alloc_commit_node(r));
63 return object_as_type(r, obj, OBJ_COMMIT, 0);
66 struct commit *lookup_commit_reference_by_name(const char *name)
68 struct object_id oid;
69 struct commit *commit;
71 if (get_oid_committish(name, &oid))
72 return NULL;
73 commit = lookup_commit_reference(the_repository, &oid);
74 if (parse_commit(commit))
75 return NULL;
76 return commit;
79 static timestamp_t parse_commit_date(const char *buf, const char *tail)
81 const char *dateptr;
83 if (buf + 6 >= tail)
84 return 0;
85 if (memcmp(buf, "author", 6))
86 return 0;
87 while (buf < tail && *buf++ != '\n')
88 /* nada */;
89 if (buf + 9 >= tail)
90 return 0;
91 if (memcmp(buf, "committer", 9))
92 return 0;
93 while (buf < tail && *buf++ != '>')
94 /* nada */;
95 if (buf >= tail)
96 return 0;
97 dateptr = buf;
98 while (buf < tail && *buf++ != '\n')
99 /* nada */;
100 if (buf >= tail)
101 return 0;
102 /* dateptr < buf && buf[-1] == '\n', so parsing will stop at buf-1 */
103 return parse_timestamp(dateptr, NULL, 10);
106 static const unsigned char *commit_graft_sha1_access(size_t index, void *table)
108 struct commit_graft **commit_graft_table = table;
109 return commit_graft_table[index]->oid.hash;
112 static int commit_graft_pos(struct repository *r, const unsigned char *sha1)
114 return sha1_pos(sha1, r->parsed_objects->grafts,
115 r->parsed_objects->grafts_nr,
116 commit_graft_sha1_access);
119 int register_commit_graft(struct repository *r, struct commit_graft *graft,
120 int ignore_dups)
122 int pos = commit_graft_pos(r, graft->oid.hash);
124 if (0 <= pos) {
125 if (ignore_dups)
126 free(graft);
127 else {
128 free(r->parsed_objects->grafts[pos]);
129 r->parsed_objects->grafts[pos] = graft;
131 return 1;
133 pos = -pos - 1;
134 ALLOC_GROW(r->parsed_objects->grafts,
135 r->parsed_objects->grafts_nr + 1,
136 r->parsed_objects->grafts_alloc);
137 r->parsed_objects->grafts_nr++;
138 if (pos < r->parsed_objects->grafts_nr)
139 memmove(r->parsed_objects->grafts + pos + 1,
140 r->parsed_objects->grafts + pos,
141 (r->parsed_objects->grafts_nr - pos - 1) *
142 sizeof(*r->parsed_objects->grafts));
143 r->parsed_objects->grafts[pos] = graft;
144 return 0;
147 struct commit_graft *read_graft_line(struct strbuf *line)
149 /* The format is just "Commit Parent1 Parent2 ...\n" */
150 int i, phase;
151 const char *tail = NULL;
152 struct commit_graft *graft = NULL;
153 struct object_id dummy_oid, *oid;
155 strbuf_rtrim(line);
156 if (!line->len || line->buf[0] == '#')
157 return NULL;
159 * phase 0 verifies line, counts hashes in line and allocates graft
160 * phase 1 fills graft
162 for (phase = 0; phase < 2; phase++) {
163 oid = graft ? &graft->oid : &dummy_oid;
164 if (parse_oid_hex(line->buf, oid, &tail))
165 goto bad_graft_data;
166 for (i = 0; *tail != '\0'; i++) {
167 oid = graft ? &graft->parent[i] : &dummy_oid;
168 if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail))
169 goto bad_graft_data;
171 if (!graft) {
172 graft = xmalloc(st_add(sizeof(*graft),
173 st_mult(sizeof(struct object_id), i)));
174 graft->nr_parent = i;
177 return graft;
179 bad_graft_data:
180 error("bad graft data: %s", line->buf);
181 assert(!graft);
182 return NULL;
185 static int read_graft_file(struct repository *r, const char *graft_file)
187 FILE *fp = fopen_or_warn(graft_file, "r");
188 struct strbuf buf = STRBUF_INIT;
189 if (!fp)
190 return -1;
191 if (advice_graft_file_deprecated)
192 advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n"
193 "and will be removed in a future Git version.\n"
194 "\n"
195 "Please use \"git replace --convert-graft-file\"\n"
196 "to convert the grafts into replace refs.\n"
197 "\n"
198 "Turn this message off by running\n"
199 "\"git config advice.graftFileDeprecated false\""));
200 while (!strbuf_getwholeline(&buf, fp, '\n')) {
201 /* The format is just "Commit Parent1 Parent2 ...\n" */
202 struct commit_graft *graft = read_graft_line(&buf);
203 if (!graft)
204 continue;
205 if (register_commit_graft(r, graft, 1))
206 error("duplicate graft data: %s", buf.buf);
208 fclose(fp);
209 strbuf_release(&buf);
210 return 0;
213 void prepare_commit_graft(struct repository *r)
215 char *graft_file;
217 if (r->parsed_objects->commit_graft_prepared)
218 return;
219 if (!startup_info->have_repository)
220 return;
222 graft_file = get_graft_file(r);
223 read_graft_file(r, graft_file);
224 /* make sure shallows are read */
225 is_repository_shallow(r);
226 r->parsed_objects->commit_graft_prepared = 1;
229 struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid)
231 int pos;
232 prepare_commit_graft(r);
233 pos = commit_graft_pos(r, oid->hash);
234 if (pos < 0)
235 return NULL;
236 return r->parsed_objects->grafts[pos];
239 int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
241 int i, ret;
242 for (i = ret = 0; i < the_repository->parsed_objects->grafts_nr && !ret; i++)
243 ret = fn(the_repository->parsed_objects->grafts[i], cb_data);
244 return ret;
247 int unregister_shallow(const struct object_id *oid)
249 int pos = commit_graft_pos(the_repository, oid->hash);
250 if (pos < 0)
251 return -1;
252 if (pos + 1 < the_repository->parsed_objects->grafts_nr)
253 MOVE_ARRAY(the_repository->parsed_objects->grafts + pos,
254 the_repository->parsed_objects->grafts + pos + 1,
255 the_repository->parsed_objects->grafts_nr - pos - 1);
256 the_repository->parsed_objects->grafts_nr--;
257 return 0;
260 struct commit_buffer {
261 void *buffer;
262 unsigned long size;
264 define_commit_slab(buffer_slab, struct commit_buffer);
266 struct buffer_slab *allocate_commit_buffer_slab(void)
268 struct buffer_slab *bs = xmalloc(sizeof(*bs));
269 init_buffer_slab(bs);
270 return bs;
273 void free_commit_buffer_slab(struct buffer_slab *bs)
275 clear_buffer_slab(bs);
276 free(bs);
279 void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size)
281 struct commit_buffer *v = buffer_slab_at(
282 r->parsed_objects->buffer_slab, commit);
283 v->buffer = buffer;
284 v->size = size;
287 const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep)
289 struct commit_buffer *v = buffer_slab_peek(
290 r->parsed_objects->buffer_slab, commit);
291 if (!v) {
292 if (sizep)
293 *sizep = 0;
294 return NULL;
296 if (sizep)
297 *sizep = v->size;
298 return v->buffer;
301 const void *repo_get_commit_buffer(struct repository *r,
302 const struct commit *commit,
303 unsigned long *sizep)
305 const void *ret = get_cached_commit_buffer(r, commit, sizep);
306 if (!ret) {
307 enum object_type type;
308 unsigned long size;
309 ret = repo_read_object_file(r, &commit->object.oid, &type, &size);
310 if (!ret)
311 die("cannot read commit object %s",
312 oid_to_hex(&commit->object.oid));
313 if (type != OBJ_COMMIT)
314 die("expected commit for %s, got %s",
315 oid_to_hex(&commit->object.oid), type_name(type));
316 if (sizep)
317 *sizep = size;
319 return ret;
322 void repo_unuse_commit_buffer(struct repository *r,
323 const struct commit *commit,
324 const void *buffer)
326 struct commit_buffer *v = buffer_slab_peek(
327 r->parsed_objects->buffer_slab, commit);
328 if (!(v && v->buffer == buffer))
329 free((void *)buffer);
332 void free_commit_buffer(struct parsed_object_pool *pool, struct commit *commit)
334 struct commit_buffer *v = buffer_slab_peek(
335 pool->buffer_slab, commit);
336 if (v) {
337 FREE_AND_NULL(v->buffer);
338 v->size = 0;
342 static inline void set_commit_tree(struct commit *c, struct tree *t)
344 c->maybe_tree = t;
347 struct tree *repo_get_commit_tree(struct repository *r,
348 const struct commit *commit)
350 if (commit->maybe_tree || !commit->object.parsed)
351 return commit->maybe_tree;
353 if (commit->graph_pos != COMMIT_NOT_FROM_GRAPH)
354 return get_commit_tree_in_graph(r, commit);
356 return NULL;
359 struct object_id *get_commit_tree_oid(const struct commit *commit)
361 struct tree *tree = get_commit_tree(commit);
362 return tree ? &tree->object.oid : NULL;
365 void release_commit_memory(struct parsed_object_pool *pool, struct commit *c)
367 set_commit_tree(c, NULL);
368 free_commit_buffer(pool, c);
369 c->index = 0;
370 free_commit_list(c->parents);
372 c->object.parsed = 0;
375 const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
377 struct commit_buffer *v = buffer_slab_peek(
378 the_repository->parsed_objects->buffer_slab, commit);
379 void *ret;
381 if (!v) {
382 if (sizep)
383 *sizep = 0;
384 return NULL;
386 ret = v->buffer;
387 if (sizep)
388 *sizep = v->size;
390 v->buffer = NULL;
391 v->size = 0;
392 return ret;
395 int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
397 const char *tail = buffer;
398 const char *bufptr = buffer;
399 struct object_id parent;
400 struct commit_list **pptr;
401 struct commit_graft *graft;
402 const int tree_entry_len = the_hash_algo->hexsz + 5;
403 const int parent_entry_len = the_hash_algo->hexsz + 7;
405 if (item->object.parsed)
406 return 0;
407 item->object.parsed = 1;
408 tail += size;
409 if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
410 bufptr[tree_entry_len] != '\n')
411 return error("bogus commit object %s", oid_to_hex(&item->object.oid));
412 if (get_oid_hex(bufptr + 5, &parent) < 0)
413 return error("bad tree pointer in commit %s",
414 oid_to_hex(&item->object.oid));
415 set_commit_tree(item, lookup_tree(r, &parent));
416 bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
417 pptr = &item->parents;
419 graft = lookup_commit_graft(r, &item->object.oid);
420 while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
421 struct commit *new_parent;
423 if (tail <= bufptr + parent_entry_len + 1 ||
424 get_oid_hex(bufptr + 7, &parent) ||
425 bufptr[parent_entry_len] != '\n')
426 return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
427 bufptr += parent_entry_len + 1;
429 * The clone is shallow if nr_parent < 0, and we must
430 * not traverse its real parents even when we unhide them.
432 if (graft && (graft->nr_parent < 0 || grafts_replace_parents))
433 continue;
434 new_parent = lookup_commit(r, &parent);
435 if (new_parent)
436 pptr = &commit_list_insert(new_parent, pptr)->next;
438 if (graft) {
439 int i;
440 struct commit *new_parent;
441 for (i = 0; i < graft->nr_parent; i++) {
442 new_parent = lookup_commit(r,
443 &graft->parent[i]);
444 if (!new_parent)
445 continue;
446 pptr = &commit_list_insert(new_parent, pptr)->next;
449 item->date = parse_commit_date(bufptr, tail);
451 if (check_graph)
452 load_commit_graph_info(r, item);
454 return 0;
457 int repo_parse_commit_internal(struct repository *r,
458 struct commit *item,
459 int quiet_on_missing,
460 int use_commit_graph)
462 enum object_type type;
463 void *buffer;
464 unsigned long size;
465 int ret;
467 if (!item)
468 return -1;
469 if (item->object.parsed)
470 return 0;
471 if (use_commit_graph && parse_commit_in_graph(r, item))
472 return 0;
473 buffer = repo_read_object_file(r, &item->object.oid, &type, &size);
474 if (!buffer)
475 return quiet_on_missing ? -1 :
476 error("Could not read %s",
477 oid_to_hex(&item->object.oid));
478 if (type != OBJ_COMMIT) {
479 free(buffer);
480 return error("Object %s not a commit",
481 oid_to_hex(&item->object.oid));
484 ret = parse_commit_buffer(r, item, buffer, size, 0);
485 if (save_commit_buffer && !ret) {
486 set_commit_buffer(r, item, buffer, size);
487 return 0;
489 free(buffer);
490 return ret;
493 int repo_parse_commit_gently(struct repository *r,
494 struct commit *item, int quiet_on_missing)
496 return repo_parse_commit_internal(r, item, quiet_on_missing, 1);
499 void parse_commit_or_die(struct commit *item)
501 if (parse_commit(item))
502 die("unable to parse commit %s",
503 item ? oid_to_hex(&item->object.oid) : "(null)");
506 int find_commit_subject(const char *commit_buffer, const char **subject)
508 const char *eol;
509 const char *p = commit_buffer;
511 while (*p && (*p != '\n' || p[1] != '\n'))
512 p++;
513 if (*p) {
514 p = skip_blank_lines(p + 2);
515 eol = strchrnul(p, '\n');
516 } else
517 eol = p;
519 *subject = p;
521 return eol - p;
524 struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
526 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
527 new_list->item = item;
528 new_list->next = *list_p;
529 *list_p = new_list;
530 return new_list;
533 unsigned commit_list_count(const struct commit_list *l)
535 unsigned c = 0;
536 for (; l; l = l->next )
537 c++;
538 return c;
541 struct commit_list *copy_commit_list(struct commit_list *list)
543 struct commit_list *head = NULL;
544 struct commit_list **pp = &head;
545 while (list) {
546 pp = commit_list_append(list->item, pp);
547 list = list->next;
549 return head;
552 void free_commit_list(struct commit_list *list)
554 while (list)
555 pop_commit(&list);
558 struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
560 struct commit_list **pp = list;
561 struct commit_list *p;
562 while ((p = *pp) != NULL) {
563 if (p->item->date < item->date) {
564 break;
566 pp = &p->next;
568 return commit_list_insert(item, pp);
571 static int commit_list_compare_by_date(const void *a, const void *b)
573 timestamp_t a_date = ((const struct commit_list *)a)->item->date;
574 timestamp_t b_date = ((const struct commit_list *)b)->item->date;
575 if (a_date < b_date)
576 return 1;
577 if (a_date > b_date)
578 return -1;
579 return 0;
582 static void *commit_list_get_next(const void *a)
584 return ((const struct commit_list *)a)->next;
587 static void commit_list_set_next(void *a, void *next)
589 ((struct commit_list *)a)->next = next;
592 void commit_list_sort_by_date(struct commit_list **list)
594 *list = llist_mergesort(*list, commit_list_get_next, commit_list_set_next,
595 commit_list_compare_by_date);
598 struct commit *pop_most_recent_commit(struct commit_list **list,
599 unsigned int mark)
601 struct commit *ret = pop_commit(list);
602 struct commit_list *parents = ret->parents;
604 while (parents) {
605 struct commit *commit = parents->item;
606 if (!parse_commit(commit) && !(commit->object.flags & mark)) {
607 commit->object.flags |= mark;
608 commit_list_insert_by_date(commit, list);
610 parents = parents->next;
612 return ret;
615 static void clear_commit_marks_1(struct commit_list **plist,
616 struct commit *commit, unsigned int mark)
618 while (commit) {
619 struct commit_list *parents;
621 if (!(mark & commit->object.flags))
622 return;
624 commit->object.flags &= ~mark;
626 parents = commit->parents;
627 if (!parents)
628 return;
630 while ((parents = parents->next))
631 commit_list_insert(parents->item, plist);
633 commit = commit->parents->item;
637 void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
639 struct commit_list *list = NULL;
641 while (nr--) {
642 clear_commit_marks_1(&list, *commit, mark);
643 commit++;
645 while (list)
646 clear_commit_marks_1(&list, pop_commit(&list), mark);
649 void clear_commit_marks(struct commit *commit, unsigned int mark)
651 clear_commit_marks_many(1, &commit, mark);
654 struct commit *pop_commit(struct commit_list **stack)
656 struct commit_list *top = *stack;
657 struct commit *item = top ? top->item : NULL;
659 if (top) {
660 *stack = top->next;
661 free(top);
663 return item;
667 * Topological sort support
670 /* count number of children that have not been emitted */
671 define_commit_slab(indegree_slab, int);
673 define_commit_slab(author_date_slab, timestamp_t);
675 void record_author_date(struct author_date_slab *author_date,
676 struct commit *commit)
678 const char *buffer = get_commit_buffer(commit, NULL);
679 struct ident_split ident;
680 const char *ident_line;
681 size_t ident_len;
682 char *date_end;
683 timestamp_t date;
685 ident_line = find_commit_header(buffer, "author", &ident_len);
686 if (!ident_line)
687 goto fail_exit; /* no author line */
688 if (split_ident_line(&ident, ident_line, ident_len) ||
689 !ident.date_begin || !ident.date_end)
690 goto fail_exit; /* malformed "author" line */
692 date = parse_timestamp(ident.date_begin, &date_end, 10);
693 if (date_end != ident.date_end)
694 goto fail_exit; /* malformed date */
695 *(author_date_slab_at(author_date, commit)) = date;
697 fail_exit:
698 unuse_commit_buffer(commit, buffer);
701 int compare_commits_by_author_date(const void *a_, const void *b_,
702 void *cb_data)
704 const struct commit *a = a_, *b = b_;
705 struct author_date_slab *author_date = cb_data;
706 timestamp_t a_date = *(author_date_slab_at(author_date, a));
707 timestamp_t b_date = *(author_date_slab_at(author_date, b));
709 /* newer commits with larger date first */
710 if (a_date < b_date)
711 return 1;
712 else if (a_date > b_date)
713 return -1;
714 return 0;
717 int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, void *unused)
719 const struct commit *a = a_, *b = b_;
721 /* newer commits first */
722 if (a->generation < b->generation)
723 return 1;
724 else if (a->generation > b->generation)
725 return -1;
727 /* use date as a heuristic when generations are equal */
728 if (a->date < b->date)
729 return 1;
730 else if (a->date > b->date)
731 return -1;
732 return 0;
735 int compare_commits_by_commit_date(const void *a_, const void *b_, void *unused)
737 const struct commit *a = a_, *b = b_;
738 /* newer commits with larger date first */
739 if (a->date < b->date)
740 return 1;
741 else if (a->date > b->date)
742 return -1;
743 return 0;
747 * Performs an in-place topological sort on the list supplied.
749 void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
751 struct commit_list *next, *orig = *list;
752 struct commit_list **pptr;
753 struct indegree_slab indegree;
754 struct prio_queue queue;
755 struct commit *commit;
756 struct author_date_slab author_date;
758 if (!orig)
759 return;
760 *list = NULL;
762 init_indegree_slab(&indegree);
763 memset(&queue, '\0', sizeof(queue));
765 switch (sort_order) {
766 default: /* REV_SORT_IN_GRAPH_ORDER */
767 queue.compare = NULL;
768 break;
769 case REV_SORT_BY_COMMIT_DATE:
770 queue.compare = compare_commits_by_commit_date;
771 break;
772 case REV_SORT_BY_AUTHOR_DATE:
773 init_author_date_slab(&author_date);
774 queue.compare = compare_commits_by_author_date;
775 queue.cb_data = &author_date;
776 break;
779 /* Mark them and clear the indegree */
780 for (next = orig; next; next = next->next) {
781 struct commit *commit = next->item;
782 *(indegree_slab_at(&indegree, commit)) = 1;
783 /* also record the author dates, if needed */
784 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
785 record_author_date(&author_date, commit);
788 /* update the indegree */
789 for (next = orig; next; next = next->next) {
790 struct commit_list *parents = next->item->parents;
791 while (parents) {
792 struct commit *parent = parents->item;
793 int *pi = indegree_slab_at(&indegree, parent);
795 if (*pi)
796 (*pi)++;
797 parents = parents->next;
802 * find the tips
804 * tips are nodes not reachable from any other node in the list
806 * the tips serve as a starting set for the work queue.
808 for (next = orig; next; next = next->next) {
809 struct commit *commit = next->item;
811 if (*(indegree_slab_at(&indegree, commit)) == 1)
812 prio_queue_put(&queue, commit);
816 * This is unfortunate; the initial tips need to be shown
817 * in the order given from the revision traversal machinery.
819 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
820 prio_queue_reverse(&queue);
822 /* We no longer need the commit list */
823 free_commit_list(orig);
825 pptr = list;
826 *list = NULL;
827 while ((commit = prio_queue_get(&queue)) != NULL) {
828 struct commit_list *parents;
830 for (parents = commit->parents; parents ; parents = parents->next) {
831 struct commit *parent = parents->item;
832 int *pi = indegree_slab_at(&indegree, parent);
834 if (!*pi)
835 continue;
838 * parents are only enqueued for emission
839 * when all their children have been emitted thereby
840 * guaranteeing topological order.
842 if (--(*pi) == 1)
843 prio_queue_put(&queue, parent);
846 * all children of commit have already been
847 * emitted. we can emit it now.
849 *(indegree_slab_at(&indegree, commit)) = 0;
851 pptr = &commit_list_insert(commit, pptr)->next;
854 clear_indegree_slab(&indegree);
855 clear_prio_queue(&queue);
856 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
857 clear_author_date_slab(&author_date);
860 struct rev_collect {
861 struct commit **commit;
862 int nr;
863 int alloc;
864 unsigned int initial : 1;
867 static void add_one_commit(struct object_id *oid, struct rev_collect *revs)
869 struct commit *commit;
871 if (is_null_oid(oid))
872 return;
874 commit = lookup_commit(the_repository, oid);
875 if (!commit ||
876 (commit->object.flags & TMP_MARK) ||
877 parse_commit(commit))
878 return;
880 ALLOC_GROW(revs->commit, revs->nr + 1, revs->alloc);
881 revs->commit[revs->nr++] = commit;
882 commit->object.flags |= TMP_MARK;
885 static int collect_one_reflog_ent(struct object_id *ooid, struct object_id *noid,
886 const char *ident, timestamp_t timestamp,
887 int tz, const char *message, void *cbdata)
889 struct rev_collect *revs = cbdata;
891 if (revs->initial) {
892 revs->initial = 0;
893 add_one_commit(ooid, revs);
895 add_one_commit(noid, revs);
896 return 0;
899 struct commit *get_fork_point(const char *refname, struct commit *commit)
901 struct object_id oid;
902 struct rev_collect revs;
903 struct commit_list *bases;
904 int i;
905 struct commit *ret = NULL;
907 memset(&revs, 0, sizeof(revs));
908 revs.initial = 1;
909 for_each_reflog_ent(refname, collect_one_reflog_ent, &revs);
911 if (!revs.nr && !get_oid(refname, &oid))
912 add_one_commit(&oid, &revs);
914 for (i = 0; i < revs.nr; i++)
915 revs.commit[i]->object.flags &= ~TMP_MARK;
917 bases = get_merge_bases_many(commit, revs.nr, revs.commit);
920 * There should be one and only one merge base, when we found
921 * a common ancestor among reflog entries.
923 if (!bases || bases->next)
924 goto cleanup_return;
926 /* And the found one must be one of the reflog entries */
927 for (i = 0; i < revs.nr; i++)
928 if (&bases->item->object == &revs.commit[i]->object)
929 break; /* found */
930 if (revs.nr <= i)
931 goto cleanup_return;
933 ret = bases->item;
935 cleanup_return:
936 free_commit_list(bases);
937 return ret;
940 static const char gpg_sig_header[] = "gpgsig";
941 static const int gpg_sig_header_len = sizeof(gpg_sig_header) - 1;
943 static int do_sign_commit(struct strbuf *buf, const char *keyid)
945 struct strbuf sig = STRBUF_INIT;
946 int inspos, copypos;
947 const char *eoh;
949 /* find the end of the header */
950 eoh = strstr(buf->buf, "\n\n");
951 if (!eoh)
952 inspos = buf->len;
953 else
954 inspos = eoh - buf->buf + 1;
956 if (!keyid || !*keyid)
957 keyid = get_signing_key();
958 if (sign_buffer(buf, &sig, keyid)) {
959 strbuf_release(&sig);
960 return -1;
963 for (copypos = 0; sig.buf[copypos]; ) {
964 const char *bol = sig.buf + copypos;
965 const char *eol = strchrnul(bol, '\n');
966 int len = (eol - bol) + !!*eol;
968 if (!copypos) {
969 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
970 inspos += gpg_sig_header_len;
972 strbuf_insert(buf, inspos++, " ", 1);
973 strbuf_insert(buf, inspos, bol, len);
974 inspos += len;
975 copypos += len;
977 strbuf_release(&sig);
978 return 0;
981 int parse_signed_commit(const struct commit *commit,
982 struct strbuf *payload, struct strbuf *signature)
985 unsigned long size;
986 const char *buffer = get_commit_buffer(commit, &size);
987 int in_signature, saw_signature = -1;
988 const char *line, *tail;
990 line = buffer;
991 tail = buffer + size;
992 in_signature = 0;
993 saw_signature = 0;
994 while (line < tail) {
995 const char *sig = NULL;
996 const char *next = memchr(line, '\n', tail - line);
998 next = next ? next + 1 : tail;
999 if (in_signature && line[0] == ' ')
1000 sig = line + 1;
1001 else if (starts_with(line, gpg_sig_header) &&
1002 line[gpg_sig_header_len] == ' ')
1003 sig = line + gpg_sig_header_len + 1;
1004 if (sig) {
1005 strbuf_add(signature, sig, next - sig);
1006 saw_signature = 1;
1007 in_signature = 1;
1008 } else {
1009 if (*line == '\n')
1010 /* dump the whole remainder of the buffer */
1011 next = tail;
1012 strbuf_add(payload, line, next - line);
1013 in_signature = 0;
1015 line = next;
1017 unuse_commit_buffer(commit, buffer);
1018 return saw_signature;
1021 int remove_signature(struct strbuf *buf)
1023 const char *line = buf->buf;
1024 const char *tail = buf->buf + buf->len;
1025 int in_signature = 0;
1026 const char *sig_start = NULL;
1027 const char *sig_end = NULL;
1029 while (line < tail) {
1030 const char *next = memchr(line, '\n', tail - line);
1031 next = next ? next + 1 : tail;
1033 if (in_signature && line[0] == ' ')
1034 sig_end = next;
1035 else if (starts_with(line, gpg_sig_header) &&
1036 line[gpg_sig_header_len] == ' ') {
1037 sig_start = line;
1038 sig_end = next;
1039 in_signature = 1;
1040 } else {
1041 if (*line == '\n')
1042 /* dump the whole remainder of the buffer */
1043 next = tail;
1044 in_signature = 0;
1046 line = next;
1049 if (sig_start)
1050 strbuf_remove(buf, sig_start - buf->buf, sig_end - sig_start);
1052 return sig_start != NULL;
1055 static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
1057 struct merge_remote_desc *desc;
1058 struct commit_extra_header *mergetag;
1059 char *buf;
1060 unsigned long size, len;
1061 enum object_type type;
1063 desc = merge_remote_util(parent);
1064 if (!desc || !desc->obj)
1065 return;
1066 buf = read_object_file(&desc->obj->oid, &type, &size);
1067 if (!buf || type != OBJ_TAG)
1068 goto free_return;
1069 len = parse_signature(buf, size);
1070 if (size == len)
1071 goto free_return;
1073 * We could verify this signature and either omit the tag when
1074 * it does not validate, but the integrator may not have the
1075 * public key of the signer of the tag he is merging, while a
1076 * later auditor may have it while auditing, so let's not run
1077 * verify-signed-buffer here for now...
1079 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
1080 * warn("warning: signed tag unverified.");
1082 mergetag = xcalloc(1, sizeof(*mergetag));
1083 mergetag->key = xstrdup("mergetag");
1084 mergetag->value = buf;
1085 mergetag->len = size;
1087 **tail = mergetag;
1088 *tail = &mergetag->next;
1089 return;
1091 free_return:
1092 free(buf);
1095 int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1097 struct strbuf payload = STRBUF_INIT;
1098 struct strbuf signature = STRBUF_INIT;
1099 int ret = 1;
1101 sigc->result = 'N';
1103 if (parse_signed_commit(commit, &payload, &signature) <= 0)
1104 goto out;
1105 ret = check_signature(payload.buf, payload.len, signature.buf,
1106 signature.len, sigc);
1108 out:
1109 strbuf_release(&payload);
1110 strbuf_release(&signature);
1112 return ret;
1115 void verify_merge_signature(struct commit *commit, int verbosity)
1117 char hex[GIT_MAX_HEXSZ + 1];
1118 struct signature_check signature_check;
1119 memset(&signature_check, 0, sizeof(signature_check));
1121 check_commit_signature(commit, &signature_check);
1123 find_unique_abbrev_r(hex, &commit->object.oid, DEFAULT_ABBREV);
1124 switch (signature_check.result) {
1125 case 'G':
1126 break;
1127 case 'U':
1128 die(_("Commit %s has an untrusted GPG signature, "
1129 "allegedly by %s."), hex, signature_check.signer);
1130 case 'B':
1131 die(_("Commit %s has a bad GPG signature "
1132 "allegedly by %s."), hex, signature_check.signer);
1133 default: /* 'N' */
1134 die(_("Commit %s does not have a GPG signature."), hex);
1136 if (verbosity >= 0 && signature_check.result == 'G')
1137 printf(_("Commit %s has a good GPG signature by %s\n"),
1138 hex, signature_check.signer);
1140 signature_check_clear(&signature_check);
1143 void append_merge_tag_headers(struct commit_list *parents,
1144 struct commit_extra_header ***tail)
1146 while (parents) {
1147 struct commit *parent = parents->item;
1148 handle_signed_tag(parent, tail);
1149 parents = parents->next;
1153 static void add_extra_header(struct strbuf *buffer,
1154 struct commit_extra_header *extra)
1156 strbuf_addstr(buffer, extra->key);
1157 if (extra->len)
1158 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1159 else
1160 strbuf_addch(buffer, '\n');
1163 struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1164 const char **exclude)
1166 struct commit_extra_header *extra = NULL;
1167 unsigned long size;
1168 const char *buffer = get_commit_buffer(commit, &size);
1169 extra = read_commit_extra_header_lines(buffer, size, exclude);
1170 unuse_commit_buffer(commit, buffer);
1171 return extra;
1174 int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1176 struct commit_extra_header *extra, *to_free;
1177 int res = 0;
1179 to_free = read_commit_extra_headers(commit, NULL);
1180 for (extra = to_free; !res && extra; extra = extra->next) {
1181 if (strcmp(extra->key, "mergetag"))
1182 continue; /* not a merge tag */
1183 res = fn(commit, extra, data);
1185 free_commit_extra_headers(to_free);
1186 return res;
1189 static inline int standard_header_field(const char *field, size_t len)
1191 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1192 (len == 6 && !memcmp(field, "parent", 6)) ||
1193 (len == 6 && !memcmp(field, "author", 6)) ||
1194 (len == 9 && !memcmp(field, "committer", 9)) ||
1195 (len == 8 && !memcmp(field, "encoding", 8)));
1198 static int excluded_header_field(const char *field, size_t len, const char **exclude)
1200 if (!exclude)
1201 return 0;
1203 while (*exclude) {
1204 size_t xlen = strlen(*exclude);
1205 if (len == xlen && !memcmp(field, *exclude, xlen))
1206 return 1;
1207 exclude++;
1209 return 0;
1212 static struct commit_extra_header *read_commit_extra_header_lines(
1213 const char *buffer, size_t size,
1214 const char **exclude)
1216 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1217 const char *line, *next, *eof, *eob;
1218 struct strbuf buf = STRBUF_INIT;
1220 for (line = buffer, eob = line + size;
1221 line < eob && *line != '\n';
1222 line = next) {
1223 next = memchr(line, '\n', eob - line);
1224 next = next ? next + 1 : eob;
1225 if (*line == ' ') {
1226 /* continuation */
1227 if (it)
1228 strbuf_add(&buf, line + 1, next - (line + 1));
1229 continue;
1231 if (it)
1232 it->value = strbuf_detach(&buf, &it->len);
1233 strbuf_reset(&buf);
1234 it = NULL;
1236 eof = memchr(line, ' ', next - line);
1237 if (!eof)
1238 eof = next;
1239 else if (standard_header_field(line, eof - line) ||
1240 excluded_header_field(line, eof - line, exclude))
1241 continue;
1243 it = xcalloc(1, sizeof(*it));
1244 it->key = xmemdupz(line, eof-line);
1245 *tail = it;
1246 tail = &it->next;
1247 if (eof + 1 < next)
1248 strbuf_add(&buf, eof + 1, next - (eof + 1));
1250 if (it)
1251 it->value = strbuf_detach(&buf, &it->len);
1252 return extra;
1255 void free_commit_extra_headers(struct commit_extra_header *extra)
1257 while (extra) {
1258 struct commit_extra_header *next = extra->next;
1259 free(extra->key);
1260 free(extra->value);
1261 free(extra);
1262 extra = next;
1266 int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1267 struct commit_list *parents, struct object_id *ret,
1268 const char *author, const char *sign_commit)
1270 struct commit_extra_header *extra = NULL, **tail = &extra;
1271 int result;
1273 append_merge_tag_headers(parents, &tail);
1274 result = commit_tree_extended(msg, msg_len, tree, parents, ret,
1275 author, sign_commit, extra);
1276 free_commit_extra_headers(extra);
1277 return result;
1280 static int find_invalid_utf8(const char *buf, int len)
1282 int offset = 0;
1283 static const unsigned int max_codepoint[] = {
1284 0x7f, 0x7ff, 0xffff, 0x10ffff
1287 while (len) {
1288 unsigned char c = *buf++;
1289 int bytes, bad_offset;
1290 unsigned int codepoint;
1291 unsigned int min_val, max_val;
1293 len--;
1294 offset++;
1296 /* Simple US-ASCII? No worries. */
1297 if (c < 0x80)
1298 continue;
1300 bad_offset = offset-1;
1303 * Count how many more high bits set: that's how
1304 * many more bytes this sequence should have.
1306 bytes = 0;
1307 while (c & 0x40) {
1308 c <<= 1;
1309 bytes++;
1313 * Must be between 1 and 3 more bytes. Longer sequences result in
1314 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1316 if (bytes < 1 || 3 < bytes)
1317 return bad_offset;
1319 /* Do we *have* that many bytes? */
1320 if (len < bytes)
1321 return bad_offset;
1324 * Place the encoded bits at the bottom of the value and compute the
1325 * valid range.
1327 codepoint = (c & 0x7f) >> bytes;
1328 min_val = max_codepoint[bytes-1] + 1;
1329 max_val = max_codepoint[bytes];
1331 offset += bytes;
1332 len -= bytes;
1334 /* And verify that they are good continuation bytes */
1335 do {
1336 codepoint <<= 6;
1337 codepoint |= *buf & 0x3f;
1338 if ((*buf++ & 0xc0) != 0x80)
1339 return bad_offset;
1340 } while (--bytes);
1342 /* Reject codepoints that are out of range for the sequence length. */
1343 if (codepoint < min_val || codepoint > max_val)
1344 return bad_offset;
1345 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1346 if ((codepoint & 0x1ff800) == 0xd800)
1347 return bad_offset;
1348 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1349 if ((codepoint & 0xfffe) == 0xfffe)
1350 return bad_offset;
1351 /* So are anything in the range U+FDD0..U+FDEF. */
1352 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1353 return bad_offset;
1355 return -1;
1359 * This verifies that the buffer is in proper utf8 format.
1361 * If it isn't, it assumes any non-utf8 characters are Latin1,
1362 * and does the conversion.
1364 static int verify_utf8(struct strbuf *buf)
1366 int ok = 1;
1367 long pos = 0;
1369 for (;;) {
1370 int bad;
1371 unsigned char c;
1372 unsigned char replace[2];
1374 bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1375 if (bad < 0)
1376 return ok;
1377 pos += bad;
1378 ok = 0;
1379 c = buf->buf[pos];
1380 strbuf_remove(buf, pos, 1);
1382 /* We know 'c' must be in the range 128-255 */
1383 replace[0] = 0xc0 + (c >> 6);
1384 replace[1] = 0x80 + (c & 0x3f);
1385 strbuf_insert(buf, pos, replace, 2);
1386 pos += 2;
1390 static const char commit_utf8_warn[] =
1391 N_("Warning: commit message did not conform to UTF-8.\n"
1392 "You may want to amend it after fixing the message, or set the config\n"
1393 "variable i18n.commitencoding to the encoding your project uses.\n");
1395 int commit_tree_extended(const char *msg, size_t msg_len,
1396 const struct object_id *tree,
1397 struct commit_list *parents, struct object_id *ret,
1398 const char *author, const char *sign_commit,
1399 struct commit_extra_header *extra)
1401 int result;
1402 int encoding_is_utf8;
1403 struct strbuf buffer;
1405 assert_oid_type(tree, OBJ_TREE);
1407 if (memchr(msg, '\0', msg_len))
1408 return error("a NUL byte in commit log message not allowed.");
1410 /* Not having i18n.commitencoding is the same as having utf-8 */
1411 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1413 strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1414 strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1417 * NOTE! This ordering means that the same exact tree merged with a
1418 * different order of parents will be a _different_ changeset even
1419 * if everything else stays the same.
1421 while (parents) {
1422 struct commit *parent = pop_commit(&parents);
1423 strbuf_addf(&buffer, "parent %s\n",
1424 oid_to_hex(&parent->object.oid));
1427 /* Person/date information */
1428 if (!author)
1429 author = git_author_info(IDENT_STRICT);
1430 strbuf_addf(&buffer, "author %s\n", author);
1431 strbuf_addf(&buffer, "committer %s\n", git_committer_info(IDENT_STRICT));
1432 if (!encoding_is_utf8)
1433 strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1435 while (extra) {
1436 add_extra_header(&buffer, extra);
1437 extra = extra->next;
1439 strbuf_addch(&buffer, '\n');
1441 /* And add the comment */
1442 strbuf_add(&buffer, msg, msg_len);
1444 /* And check the encoding */
1445 if (encoding_is_utf8 && !verify_utf8(&buffer))
1446 fprintf(stderr, _(commit_utf8_warn));
1448 if (sign_commit && do_sign_commit(&buffer, sign_commit)) {
1449 result = -1;
1450 goto out;
1453 result = write_object_file(buffer.buf, buffer.len, commit_type, ret);
1454 out:
1455 strbuf_release(&buffer);
1456 return result;
1459 define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1460 static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1462 struct merge_remote_desc *merge_remote_util(struct commit *commit)
1464 return *merge_desc_slab_at(&merge_desc_slab, commit);
1467 void set_merge_remote_desc(struct commit *commit,
1468 const char *name, struct object *obj)
1470 struct merge_remote_desc *desc;
1471 FLEX_ALLOC_STR(desc, name, name);
1472 desc->obj = obj;
1473 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1476 struct commit *get_merge_parent(const char *name)
1478 struct object *obj;
1479 struct commit *commit;
1480 struct object_id oid;
1481 if (get_oid(name, &oid))
1482 return NULL;
1483 obj = parse_object(the_repository, &oid);
1484 commit = (struct commit *)peel_to_type(name, 0, obj, OBJ_COMMIT);
1485 if (commit && !merge_remote_util(commit))
1486 set_merge_remote_desc(commit, name, obj);
1487 return commit;
1491 * Append a commit to the end of the commit_list.
1493 * next starts by pointing to the variable that holds the head of an
1494 * empty commit_list, and is updated to point to the "next" field of
1495 * the last item on the list as new commits are appended.
1497 * Usage example:
1499 * struct commit_list *list;
1500 * struct commit_list **next = &list;
1502 * next = commit_list_append(c1, next);
1503 * next = commit_list_append(c2, next);
1504 * assert(commit_list_count(list) == 2);
1505 * return list;
1507 struct commit_list **commit_list_append(struct commit *commit,
1508 struct commit_list **next)
1510 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1511 new_commit->item = commit;
1512 *next = new_commit;
1513 new_commit->next = NULL;
1514 return &new_commit->next;
1517 const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1519 int key_len = strlen(key);
1520 const char *line = msg;
1522 while (line) {
1523 const char *eol = strchrnul(line, '\n');
1525 if (line == eol)
1526 return NULL;
1528 if (eol - line > key_len &&
1529 !strncmp(line, key, key_len) &&
1530 line[key_len] == ' ') {
1531 *out_len = eol - line - key_len - 1;
1532 return line + key_len + 1;
1534 line = *eol ? eol + 1 : NULL;
1536 return NULL;
1540 * Inspect the given string and determine the true "end" of the log message, in
1541 * order to find where to put a new Signed-off-by: line. Ignored are
1542 * trailing comment lines and blank lines. To support "git commit -s
1543 * --amend" on an existing commit, we also ignore "Conflicts:". To
1544 * support "git commit -v", we truncate at cut lines.
1546 * Returns the number of bytes from the tail to ignore, to be fed as
1547 * the second parameter to append_signoff().
1549 size_t ignore_non_trailer(const char *buf, size_t len)
1551 size_t boc = 0;
1552 size_t bol = 0;
1553 int in_old_conflicts_block = 0;
1554 size_t cutoff = wt_status_locate_end(buf, len);
1556 while (bol < cutoff) {
1557 const char *next_line = memchr(buf + bol, '\n', len - bol);
1559 if (!next_line)
1560 next_line = buf + len;
1561 else
1562 next_line++;
1564 if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1565 /* is this the first of the run of comments? */
1566 if (!boc)
1567 boc = bol;
1568 /* otherwise, it is just continuing */
1569 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1570 in_old_conflicts_block = 1;
1571 if (!boc)
1572 boc = bol;
1573 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1574 ; /* a pathname in the conflicts block */
1575 } else if (boc) {
1576 /* the previous was not trailing comment */
1577 boc = 0;
1578 in_old_conflicts_block = 0;
1580 bol = next_line - buf;
1582 return boc ? len - boc : len - cutoff;