submodule: use submodule repos for object lookup
[git/debian.git] / commit.c
blob7d2f3a9a934ed3fe092811265b468999a459e32c
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"
21 static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
23 int save_commit_buffer = 1;
25 const char *commit_type = "commit";
27 struct commit *lookup_commit_reference_gently(struct repository *r,
28 const struct object_id *oid, int quiet)
30 struct object *obj = deref_tag(r,
31 parse_object(r, oid),
32 NULL, 0);
34 if (!obj)
35 return NULL;
36 return object_as_type(r, obj, OBJ_COMMIT, quiet);
39 struct commit *lookup_commit_reference(struct repository *r, const struct object_id *oid)
41 return lookup_commit_reference_gently(r, oid, 0);
44 struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
46 struct commit *c = lookup_commit_reference(the_repository, oid);
47 if (!c)
48 die(_("could not parse %s"), ref_name);
49 if (!oideq(oid, &c->object.oid)) {
50 warning(_("%s %s is not a commit!"),
51 ref_name, oid_to_hex(oid));
53 return c;
56 struct commit *lookup_commit(struct repository *r, const struct object_id *oid)
58 struct object *obj = lookup_object(r, oid->hash);
59 if (!obj)
60 return create_object(r, oid->hash,
61 alloc_commit_node(r));
62 return object_as_type(r, obj, OBJ_COMMIT, 0);
65 struct commit *lookup_commit_reference_by_name(const char *name)
67 struct object_id oid;
68 struct commit *commit;
70 if (get_oid_committish(name, &oid))
71 return NULL;
72 commit = lookup_commit_reference(the_repository, &oid);
73 if (parse_commit(commit))
74 return NULL;
75 return commit;
78 static timestamp_t parse_commit_date(const char *buf, const char *tail)
80 const char *dateptr;
82 if (buf + 6 >= tail)
83 return 0;
84 if (memcmp(buf, "author", 6))
85 return 0;
86 while (buf < tail && *buf++ != '\n')
87 /* nada */;
88 if (buf + 9 >= tail)
89 return 0;
90 if (memcmp(buf, "committer", 9))
91 return 0;
92 while (buf < tail && *buf++ != '>')
93 /* nada */;
94 if (buf >= tail)
95 return 0;
96 dateptr = buf;
97 while (buf < tail && *buf++ != '\n')
98 /* nada */;
99 if (buf >= tail)
100 return 0;
101 /* dateptr < buf && buf[-1] == '\n', so parsing will stop at buf-1 */
102 return parse_timestamp(dateptr, NULL, 10);
105 static const unsigned char *commit_graft_sha1_access(size_t index, void *table)
107 struct commit_graft **commit_graft_table = table;
108 return commit_graft_table[index]->oid.hash;
111 static int commit_graft_pos(struct repository *r, const unsigned char *sha1)
113 return sha1_pos(sha1, r->parsed_objects->grafts,
114 r->parsed_objects->grafts_nr,
115 commit_graft_sha1_access);
118 int register_commit_graft(struct repository *r, struct commit_graft *graft,
119 int ignore_dups)
121 int pos = commit_graft_pos(r, graft->oid.hash);
123 if (0 <= pos) {
124 if (ignore_dups)
125 free(graft);
126 else {
127 free(r->parsed_objects->grafts[pos]);
128 r->parsed_objects->grafts[pos] = graft;
130 return 1;
132 pos = -pos - 1;
133 ALLOC_GROW(r->parsed_objects->grafts,
134 r->parsed_objects->grafts_nr + 1,
135 r->parsed_objects->grafts_alloc);
136 r->parsed_objects->grafts_nr++;
137 if (pos < r->parsed_objects->grafts_nr)
138 memmove(r->parsed_objects->grafts + pos + 1,
139 r->parsed_objects->grafts + pos,
140 (r->parsed_objects->grafts_nr - pos - 1) *
141 sizeof(*r->parsed_objects->grafts));
142 r->parsed_objects->grafts[pos] = graft;
143 return 0;
146 struct commit_graft *read_graft_line(struct strbuf *line)
148 /* The format is just "Commit Parent1 Parent2 ...\n" */
149 int i, phase;
150 const char *tail = NULL;
151 struct commit_graft *graft = NULL;
152 struct object_id dummy_oid, *oid;
154 strbuf_rtrim(line);
155 if (!line->len || line->buf[0] == '#')
156 return NULL;
158 * phase 0 verifies line, counts hashes in line and allocates graft
159 * phase 1 fills graft
161 for (phase = 0; phase < 2; phase++) {
162 oid = graft ? &graft->oid : &dummy_oid;
163 if (parse_oid_hex(line->buf, oid, &tail))
164 goto bad_graft_data;
165 for (i = 0; *tail != '\0'; i++) {
166 oid = graft ? &graft->parent[i] : &dummy_oid;
167 if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail))
168 goto bad_graft_data;
170 if (!graft) {
171 graft = xmalloc(st_add(sizeof(*graft),
172 st_mult(sizeof(struct object_id), i)));
173 graft->nr_parent = i;
176 return graft;
178 bad_graft_data:
179 error("bad graft data: %s", line->buf);
180 assert(!graft);
181 return NULL;
184 static int read_graft_file(struct repository *r, const char *graft_file)
186 FILE *fp = fopen_or_warn(graft_file, "r");
187 struct strbuf buf = STRBUF_INIT;
188 if (!fp)
189 return -1;
190 if (advice_graft_file_deprecated)
191 advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n"
192 "and will be removed in a future Git version.\n"
193 "\n"
194 "Please use \"git replace --convert-graft-file\"\n"
195 "to convert the grafts into replace refs.\n"
196 "\n"
197 "Turn this message off by running\n"
198 "\"git config advice.graftFileDeprecated false\""));
199 while (!strbuf_getwholeline(&buf, fp, '\n')) {
200 /* The format is just "Commit Parent1 Parent2 ...\n" */
201 struct commit_graft *graft = read_graft_line(&buf);
202 if (!graft)
203 continue;
204 if (register_commit_graft(r, graft, 1))
205 error("duplicate graft data: %s", buf.buf);
207 fclose(fp);
208 strbuf_release(&buf);
209 return 0;
212 void prepare_commit_graft(struct repository *r)
214 char *graft_file;
216 if (r->parsed_objects->commit_graft_prepared)
217 return;
218 if (!startup_info->have_repository)
219 return;
221 graft_file = get_graft_file(r);
222 read_graft_file(r, graft_file);
223 /* make sure shallows are read */
224 is_repository_shallow(r);
225 r->parsed_objects->commit_graft_prepared = 1;
228 struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid)
230 int pos;
231 prepare_commit_graft(r);
232 pos = commit_graft_pos(r, oid->hash);
233 if (pos < 0)
234 return NULL;
235 return r->parsed_objects->grafts[pos];
238 int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
240 int i, ret;
241 for (i = ret = 0; i < the_repository->parsed_objects->grafts_nr && !ret; i++)
242 ret = fn(the_repository->parsed_objects->grafts[i], cb_data);
243 return ret;
246 int unregister_shallow(const struct object_id *oid)
248 int pos = commit_graft_pos(the_repository, oid->hash);
249 if (pos < 0)
250 return -1;
251 if (pos + 1 < the_repository->parsed_objects->grafts_nr)
252 MOVE_ARRAY(the_repository->parsed_objects->grafts + pos,
253 the_repository->parsed_objects->grafts + pos + 1,
254 the_repository->parsed_objects->grafts_nr - pos - 1);
255 the_repository->parsed_objects->grafts_nr--;
256 return 0;
259 struct commit_buffer {
260 void *buffer;
261 unsigned long size;
263 define_commit_slab(buffer_slab, struct commit_buffer);
265 struct buffer_slab *allocate_commit_buffer_slab(void)
267 struct buffer_slab *bs = xmalloc(sizeof(*bs));
268 init_buffer_slab(bs);
269 return bs;
272 void free_commit_buffer_slab(struct buffer_slab *bs)
274 clear_buffer_slab(bs);
275 free(bs);
278 void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size)
280 struct commit_buffer *v = buffer_slab_at(
281 r->parsed_objects->buffer_slab, commit);
282 v->buffer = buffer;
283 v->size = size;
286 const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep)
288 struct commit_buffer *v = buffer_slab_peek(
289 r->parsed_objects->buffer_slab, commit);
290 if (!v) {
291 if (sizep)
292 *sizep = 0;
293 return NULL;
295 if (sizep)
296 *sizep = v->size;
297 return v->buffer;
300 const void *repo_get_commit_buffer(struct repository *r,
301 const struct commit *commit,
302 unsigned long *sizep)
304 const void *ret = get_cached_commit_buffer(r, commit, sizep);
305 if (!ret) {
306 enum object_type type;
307 unsigned long size;
308 ret = repo_read_object_file(r, &commit->object.oid, &type, &size);
309 if (!ret)
310 die("cannot read commit object %s",
311 oid_to_hex(&commit->object.oid));
312 if (type != OBJ_COMMIT)
313 die("expected commit for %s, got %s",
314 oid_to_hex(&commit->object.oid), type_name(type));
315 if (sizep)
316 *sizep = size;
318 return ret;
321 void repo_unuse_commit_buffer(struct repository *r,
322 const struct commit *commit,
323 const void *buffer)
325 struct commit_buffer *v = buffer_slab_peek(
326 r->parsed_objects->buffer_slab, commit);
327 if (!(v && v->buffer == buffer))
328 free((void *)buffer);
331 void free_commit_buffer(struct commit *commit)
333 struct commit_buffer *v = buffer_slab_peek(
334 the_repository->parsed_objects->buffer_slab, commit);
335 if (v) {
336 FREE_AND_NULL(v->buffer);
337 v->size = 0;
341 struct tree *get_commit_tree(const struct commit *commit)
343 if (commit->maybe_tree || !commit->object.parsed)
344 return commit->maybe_tree;
346 if (commit->graph_pos == COMMIT_NOT_FROM_GRAPH)
347 BUG("commit has NULL tree, but was not loaded from commit-graph");
349 return get_commit_tree_in_graph(the_repository, commit);
352 struct object_id *get_commit_tree_oid(const struct commit *commit)
354 return &get_commit_tree(commit)->object.oid;
357 void release_commit_memory(struct commit *c)
359 c->maybe_tree = NULL;
360 c->index = 0;
361 free_commit_buffer(c);
362 free_commit_list(c->parents);
363 /* TODO: what about commit->util? */
365 c->object.parsed = 0;
368 const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
370 struct commit_buffer *v = buffer_slab_peek(
371 the_repository->parsed_objects->buffer_slab, commit);
372 void *ret;
374 if (!v) {
375 if (sizep)
376 *sizep = 0;
377 return NULL;
379 ret = v->buffer;
380 if (sizep)
381 *sizep = v->size;
383 v->buffer = NULL;
384 v->size = 0;
385 return ret;
388 int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
390 const char *tail = buffer;
391 const char *bufptr = buffer;
392 struct object_id parent;
393 struct commit_list **pptr;
394 struct commit_graft *graft;
395 const int tree_entry_len = the_hash_algo->hexsz + 5;
396 const int parent_entry_len = the_hash_algo->hexsz + 7;
398 if (item->object.parsed)
399 return 0;
400 item->object.parsed = 1;
401 tail += size;
402 if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
403 bufptr[tree_entry_len] != '\n')
404 return error("bogus commit object %s", oid_to_hex(&item->object.oid));
405 if (get_oid_hex(bufptr + 5, &parent) < 0)
406 return error("bad tree pointer in commit %s",
407 oid_to_hex(&item->object.oid));
408 item->maybe_tree = lookup_tree(r, &parent);
409 bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
410 pptr = &item->parents;
412 graft = lookup_commit_graft(r, &item->object.oid);
413 while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
414 struct commit *new_parent;
416 if (tail <= bufptr + parent_entry_len + 1 ||
417 get_oid_hex(bufptr + 7, &parent) ||
418 bufptr[parent_entry_len] != '\n')
419 return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
420 bufptr += parent_entry_len + 1;
422 * The clone is shallow if nr_parent < 0, and we must
423 * not traverse its real parents even when we unhide them.
425 if (graft && (graft->nr_parent < 0 || grafts_replace_parents))
426 continue;
427 new_parent = lookup_commit(r, &parent);
428 if (new_parent)
429 pptr = &commit_list_insert(new_parent, pptr)->next;
431 if (graft) {
432 int i;
433 struct commit *new_parent;
434 for (i = 0; i < graft->nr_parent; i++) {
435 new_parent = lookup_commit(r,
436 &graft->parent[i]);
437 if (!new_parent)
438 continue;
439 pptr = &commit_list_insert(new_parent, pptr)->next;
442 item->date = parse_commit_date(bufptr, tail);
444 if (check_graph)
445 load_commit_graph_info(the_repository, item);
447 return 0;
450 int repo_parse_commit_internal(struct repository *r,
451 struct commit *item,
452 int quiet_on_missing,
453 int use_commit_graph)
455 enum object_type type;
456 void *buffer;
457 unsigned long size;
458 int ret;
460 if (!item)
461 return -1;
462 if (item->object.parsed)
463 return 0;
464 if (use_commit_graph && parse_commit_in_graph(r, item))
465 return 0;
466 buffer = repo_read_object_file(r, &item->object.oid, &type, &size);
467 if (!buffer)
468 return quiet_on_missing ? -1 :
469 error("Could not read %s",
470 oid_to_hex(&item->object.oid));
471 if (type != OBJ_COMMIT) {
472 free(buffer);
473 return error("Object %s not a commit",
474 oid_to_hex(&item->object.oid));
477 ret = parse_commit_buffer(r, item, buffer, size, 0);
478 if (save_commit_buffer && !ret) {
479 set_commit_buffer(r, item, buffer, size);
480 return 0;
482 free(buffer);
483 return ret;
486 int repo_parse_commit_gently(struct repository *r,
487 struct commit *item, int quiet_on_missing)
489 return repo_parse_commit_internal(r, item, quiet_on_missing, 1);
492 void parse_commit_or_die(struct commit *item)
494 if (parse_commit(item))
495 die("unable to parse commit %s",
496 item ? oid_to_hex(&item->object.oid) : "(null)");
499 int find_commit_subject(const char *commit_buffer, const char **subject)
501 const char *eol;
502 const char *p = commit_buffer;
504 while (*p && (*p != '\n' || p[1] != '\n'))
505 p++;
506 if (*p) {
507 p = skip_blank_lines(p + 2);
508 eol = strchrnul(p, '\n');
509 } else
510 eol = p;
512 *subject = p;
514 return eol - p;
517 struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
519 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
520 new_list->item = item;
521 new_list->next = *list_p;
522 *list_p = new_list;
523 return new_list;
526 unsigned commit_list_count(const struct commit_list *l)
528 unsigned c = 0;
529 for (; l; l = l->next )
530 c++;
531 return c;
534 struct commit_list *copy_commit_list(struct commit_list *list)
536 struct commit_list *head = NULL;
537 struct commit_list **pp = &head;
538 while (list) {
539 pp = commit_list_append(list->item, pp);
540 list = list->next;
542 return head;
545 void free_commit_list(struct commit_list *list)
547 while (list)
548 pop_commit(&list);
551 struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
553 struct commit_list **pp = list;
554 struct commit_list *p;
555 while ((p = *pp) != NULL) {
556 if (p->item->date < item->date) {
557 break;
559 pp = &p->next;
561 return commit_list_insert(item, pp);
564 static int commit_list_compare_by_date(const void *a, const void *b)
566 timestamp_t a_date = ((const struct commit_list *)a)->item->date;
567 timestamp_t b_date = ((const struct commit_list *)b)->item->date;
568 if (a_date < b_date)
569 return 1;
570 if (a_date > b_date)
571 return -1;
572 return 0;
575 static void *commit_list_get_next(const void *a)
577 return ((const struct commit_list *)a)->next;
580 static void commit_list_set_next(void *a, void *next)
582 ((struct commit_list *)a)->next = next;
585 void commit_list_sort_by_date(struct commit_list **list)
587 *list = llist_mergesort(*list, commit_list_get_next, commit_list_set_next,
588 commit_list_compare_by_date);
591 struct commit *pop_most_recent_commit(struct commit_list **list,
592 unsigned int mark)
594 struct commit *ret = pop_commit(list);
595 struct commit_list *parents = ret->parents;
597 while (parents) {
598 struct commit *commit = parents->item;
599 if (!parse_commit(commit) && !(commit->object.flags & mark)) {
600 commit->object.flags |= mark;
601 commit_list_insert_by_date(commit, list);
603 parents = parents->next;
605 return ret;
608 static void clear_commit_marks_1(struct commit_list **plist,
609 struct commit *commit, unsigned int mark)
611 while (commit) {
612 struct commit_list *parents;
614 if (!(mark & commit->object.flags))
615 return;
617 commit->object.flags &= ~mark;
619 parents = commit->parents;
620 if (!parents)
621 return;
623 while ((parents = parents->next))
624 commit_list_insert(parents->item, plist);
626 commit = commit->parents->item;
630 void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
632 struct commit_list *list = NULL;
634 while (nr--) {
635 clear_commit_marks_1(&list, *commit, mark);
636 commit++;
638 while (list)
639 clear_commit_marks_1(&list, pop_commit(&list), mark);
642 void clear_commit_marks(struct commit *commit, unsigned int mark)
644 clear_commit_marks_many(1, &commit, mark);
647 struct commit *pop_commit(struct commit_list **stack)
649 struct commit_list *top = *stack;
650 struct commit *item = top ? top->item : NULL;
652 if (top) {
653 *stack = top->next;
654 free(top);
656 return item;
660 * Topological sort support
663 /* count number of children that have not been emitted */
664 define_commit_slab(indegree_slab, int);
666 /* record author-date for each commit object */
667 define_commit_slab(author_date_slab, timestamp_t);
669 static void record_author_date(struct author_date_slab *author_date,
670 struct commit *commit)
672 const char *buffer = get_commit_buffer(commit, NULL);
673 struct ident_split ident;
674 const char *ident_line;
675 size_t ident_len;
676 char *date_end;
677 timestamp_t date;
679 ident_line = find_commit_header(buffer, "author", &ident_len);
680 if (!ident_line)
681 goto fail_exit; /* no author line */
682 if (split_ident_line(&ident, ident_line, ident_len) ||
683 !ident.date_begin || !ident.date_end)
684 goto fail_exit; /* malformed "author" line */
686 date = parse_timestamp(ident.date_begin, &date_end, 10);
687 if (date_end != ident.date_end)
688 goto fail_exit; /* malformed date */
689 *(author_date_slab_at(author_date, commit)) = date;
691 fail_exit:
692 unuse_commit_buffer(commit, buffer);
695 static int compare_commits_by_author_date(const void *a_, const void *b_,
696 void *cb_data)
698 const struct commit *a = a_, *b = b_;
699 struct author_date_slab *author_date = cb_data;
700 timestamp_t a_date = *(author_date_slab_at(author_date, a));
701 timestamp_t b_date = *(author_date_slab_at(author_date, b));
703 /* newer commits with larger date first */
704 if (a_date < b_date)
705 return 1;
706 else if (a_date > b_date)
707 return -1;
708 return 0;
711 int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, void *unused)
713 const struct commit *a = a_, *b = b_;
715 /* newer commits first */
716 if (a->generation < b->generation)
717 return 1;
718 else if (a->generation > b->generation)
719 return -1;
721 /* use date as a heuristic when generations are equal */
722 if (a->date < b->date)
723 return 1;
724 else if (a->date > b->date)
725 return -1;
726 return 0;
729 int compare_commits_by_commit_date(const void *a_, const void *b_, void *unused)
731 const struct commit *a = a_, *b = b_;
732 /* newer commits with larger date first */
733 if (a->date < b->date)
734 return 1;
735 else if (a->date > b->date)
736 return -1;
737 return 0;
741 * Performs an in-place topological sort on the list supplied.
743 void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
745 struct commit_list *next, *orig = *list;
746 struct commit_list **pptr;
747 struct indegree_slab indegree;
748 struct prio_queue queue;
749 struct commit *commit;
750 struct author_date_slab author_date;
752 if (!orig)
753 return;
754 *list = NULL;
756 init_indegree_slab(&indegree);
757 memset(&queue, '\0', sizeof(queue));
759 switch (sort_order) {
760 default: /* REV_SORT_IN_GRAPH_ORDER */
761 queue.compare = NULL;
762 break;
763 case REV_SORT_BY_COMMIT_DATE:
764 queue.compare = compare_commits_by_commit_date;
765 break;
766 case REV_SORT_BY_AUTHOR_DATE:
767 init_author_date_slab(&author_date);
768 queue.compare = compare_commits_by_author_date;
769 queue.cb_data = &author_date;
770 break;
773 /* Mark them and clear the indegree */
774 for (next = orig; next; next = next->next) {
775 struct commit *commit = next->item;
776 *(indegree_slab_at(&indegree, commit)) = 1;
777 /* also record the author dates, if needed */
778 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
779 record_author_date(&author_date, commit);
782 /* update the indegree */
783 for (next = orig; next; next = next->next) {
784 struct commit_list *parents = next->item->parents;
785 while (parents) {
786 struct commit *parent = parents->item;
787 int *pi = indegree_slab_at(&indegree, parent);
789 if (*pi)
790 (*pi)++;
791 parents = parents->next;
796 * find the tips
798 * tips are nodes not reachable from any other node in the list
800 * the tips serve as a starting set for the work queue.
802 for (next = orig; next; next = next->next) {
803 struct commit *commit = next->item;
805 if (*(indegree_slab_at(&indegree, commit)) == 1)
806 prio_queue_put(&queue, commit);
810 * This is unfortunate; the initial tips need to be shown
811 * in the order given from the revision traversal machinery.
813 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
814 prio_queue_reverse(&queue);
816 /* We no longer need the commit list */
817 free_commit_list(orig);
819 pptr = list;
820 *list = NULL;
821 while ((commit = prio_queue_get(&queue)) != NULL) {
822 struct commit_list *parents;
824 for (parents = commit->parents; parents ; parents = parents->next) {
825 struct commit *parent = parents->item;
826 int *pi = indegree_slab_at(&indegree, parent);
828 if (!*pi)
829 continue;
832 * parents are only enqueued for emission
833 * when all their children have been emitted thereby
834 * guaranteeing topological order.
836 if (--(*pi) == 1)
837 prio_queue_put(&queue, parent);
840 * all children of commit have already been
841 * emitted. we can emit it now.
843 *(indegree_slab_at(&indegree, commit)) = 0;
845 pptr = &commit_list_insert(commit, pptr)->next;
848 clear_indegree_slab(&indegree);
849 clear_prio_queue(&queue);
850 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
851 clear_author_date_slab(&author_date);
854 static const char gpg_sig_header[] = "gpgsig";
855 static const int gpg_sig_header_len = sizeof(gpg_sig_header) - 1;
857 static int do_sign_commit(struct strbuf *buf, const char *keyid)
859 struct strbuf sig = STRBUF_INIT;
860 int inspos, copypos;
861 const char *eoh;
863 /* find the end of the header */
864 eoh = strstr(buf->buf, "\n\n");
865 if (!eoh)
866 inspos = buf->len;
867 else
868 inspos = eoh - buf->buf + 1;
870 if (!keyid || !*keyid)
871 keyid = get_signing_key();
872 if (sign_buffer(buf, &sig, keyid)) {
873 strbuf_release(&sig);
874 return -1;
877 for (copypos = 0; sig.buf[copypos]; ) {
878 const char *bol = sig.buf + copypos;
879 const char *eol = strchrnul(bol, '\n');
880 int len = (eol - bol) + !!*eol;
882 if (!copypos) {
883 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
884 inspos += gpg_sig_header_len;
886 strbuf_insert(buf, inspos++, " ", 1);
887 strbuf_insert(buf, inspos, bol, len);
888 inspos += len;
889 copypos += len;
891 strbuf_release(&sig);
892 return 0;
895 int parse_signed_commit(const struct commit *commit,
896 struct strbuf *payload, struct strbuf *signature)
899 unsigned long size;
900 const char *buffer = get_commit_buffer(commit, &size);
901 int in_signature, saw_signature = -1;
902 const char *line, *tail;
904 line = buffer;
905 tail = buffer + size;
906 in_signature = 0;
907 saw_signature = 0;
908 while (line < tail) {
909 const char *sig = NULL;
910 const char *next = memchr(line, '\n', tail - line);
912 next = next ? next + 1 : tail;
913 if (in_signature && line[0] == ' ')
914 sig = line + 1;
915 else if (starts_with(line, gpg_sig_header) &&
916 line[gpg_sig_header_len] == ' ')
917 sig = line + gpg_sig_header_len + 1;
918 if (sig) {
919 strbuf_add(signature, sig, next - sig);
920 saw_signature = 1;
921 in_signature = 1;
922 } else {
923 if (*line == '\n')
924 /* dump the whole remainder of the buffer */
925 next = tail;
926 strbuf_add(payload, line, next - line);
927 in_signature = 0;
929 line = next;
931 unuse_commit_buffer(commit, buffer);
932 return saw_signature;
935 int remove_signature(struct strbuf *buf)
937 const char *line = buf->buf;
938 const char *tail = buf->buf + buf->len;
939 int in_signature = 0;
940 const char *sig_start = NULL;
941 const char *sig_end = NULL;
943 while (line < tail) {
944 const char *next = memchr(line, '\n', tail - line);
945 next = next ? next + 1 : tail;
947 if (in_signature && line[0] == ' ')
948 sig_end = next;
949 else if (starts_with(line, gpg_sig_header) &&
950 line[gpg_sig_header_len] == ' ') {
951 sig_start = line;
952 sig_end = next;
953 in_signature = 1;
954 } else {
955 if (*line == '\n')
956 /* dump the whole remainder of the buffer */
957 next = tail;
958 in_signature = 0;
960 line = next;
963 if (sig_start)
964 strbuf_remove(buf, sig_start - buf->buf, sig_end - sig_start);
966 return sig_start != NULL;
969 static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
971 struct merge_remote_desc *desc;
972 struct commit_extra_header *mergetag;
973 char *buf;
974 unsigned long size, len;
975 enum object_type type;
977 desc = merge_remote_util(parent);
978 if (!desc || !desc->obj)
979 return;
980 buf = read_object_file(&desc->obj->oid, &type, &size);
981 if (!buf || type != OBJ_TAG)
982 goto free_return;
983 len = parse_signature(buf, size);
984 if (size == len)
985 goto free_return;
987 * We could verify this signature and either omit the tag when
988 * it does not validate, but the integrator may not have the
989 * public key of the signer of the tag he is merging, while a
990 * later auditor may have it while auditing, so let's not run
991 * verify-signed-buffer here for now...
993 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
994 * warn("warning: signed tag unverified.");
996 mergetag = xcalloc(1, sizeof(*mergetag));
997 mergetag->key = xstrdup("mergetag");
998 mergetag->value = buf;
999 mergetag->len = size;
1001 **tail = mergetag;
1002 *tail = &mergetag->next;
1003 return;
1005 free_return:
1006 free(buf);
1009 int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1011 struct strbuf payload = STRBUF_INIT;
1012 struct strbuf signature = STRBUF_INIT;
1013 int ret = 1;
1015 sigc->result = 'N';
1017 if (parse_signed_commit(commit, &payload, &signature) <= 0)
1018 goto out;
1019 ret = check_signature(payload.buf, payload.len, signature.buf,
1020 signature.len, sigc);
1022 out:
1023 strbuf_release(&payload);
1024 strbuf_release(&signature);
1026 return ret;
1031 void append_merge_tag_headers(struct commit_list *parents,
1032 struct commit_extra_header ***tail)
1034 while (parents) {
1035 struct commit *parent = parents->item;
1036 handle_signed_tag(parent, tail);
1037 parents = parents->next;
1041 static void add_extra_header(struct strbuf *buffer,
1042 struct commit_extra_header *extra)
1044 strbuf_addstr(buffer, extra->key);
1045 if (extra->len)
1046 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1047 else
1048 strbuf_addch(buffer, '\n');
1051 struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1052 const char **exclude)
1054 struct commit_extra_header *extra = NULL;
1055 unsigned long size;
1056 const char *buffer = get_commit_buffer(commit, &size);
1057 extra = read_commit_extra_header_lines(buffer, size, exclude);
1058 unuse_commit_buffer(commit, buffer);
1059 return extra;
1062 int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1064 struct commit_extra_header *extra, *to_free;
1065 int res = 0;
1067 to_free = read_commit_extra_headers(commit, NULL);
1068 for (extra = to_free; !res && extra; extra = extra->next) {
1069 if (strcmp(extra->key, "mergetag"))
1070 continue; /* not a merge tag */
1071 res = fn(commit, extra, data);
1073 free_commit_extra_headers(to_free);
1074 return res;
1077 static inline int standard_header_field(const char *field, size_t len)
1079 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1080 (len == 6 && !memcmp(field, "parent", 6)) ||
1081 (len == 6 && !memcmp(field, "author", 6)) ||
1082 (len == 9 && !memcmp(field, "committer", 9)) ||
1083 (len == 8 && !memcmp(field, "encoding", 8)));
1086 static int excluded_header_field(const char *field, size_t len, const char **exclude)
1088 if (!exclude)
1089 return 0;
1091 while (*exclude) {
1092 size_t xlen = strlen(*exclude);
1093 if (len == xlen && !memcmp(field, *exclude, xlen))
1094 return 1;
1095 exclude++;
1097 return 0;
1100 static struct commit_extra_header *read_commit_extra_header_lines(
1101 const char *buffer, size_t size,
1102 const char **exclude)
1104 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1105 const char *line, *next, *eof, *eob;
1106 struct strbuf buf = STRBUF_INIT;
1108 for (line = buffer, eob = line + size;
1109 line < eob && *line != '\n';
1110 line = next) {
1111 next = memchr(line, '\n', eob - line);
1112 next = next ? next + 1 : eob;
1113 if (*line == ' ') {
1114 /* continuation */
1115 if (it)
1116 strbuf_add(&buf, line + 1, next - (line + 1));
1117 continue;
1119 if (it)
1120 it->value = strbuf_detach(&buf, &it->len);
1121 strbuf_reset(&buf);
1122 it = NULL;
1124 eof = memchr(line, ' ', next - line);
1125 if (!eof)
1126 eof = next;
1127 else if (standard_header_field(line, eof - line) ||
1128 excluded_header_field(line, eof - line, exclude))
1129 continue;
1131 it = xcalloc(1, sizeof(*it));
1132 it->key = xmemdupz(line, eof-line);
1133 *tail = it;
1134 tail = &it->next;
1135 if (eof + 1 < next)
1136 strbuf_add(&buf, eof + 1, next - (eof + 1));
1138 if (it)
1139 it->value = strbuf_detach(&buf, &it->len);
1140 return extra;
1143 void free_commit_extra_headers(struct commit_extra_header *extra)
1145 while (extra) {
1146 struct commit_extra_header *next = extra->next;
1147 free(extra->key);
1148 free(extra->value);
1149 free(extra);
1150 extra = next;
1154 int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1155 struct commit_list *parents, struct object_id *ret,
1156 const char *author, const char *sign_commit)
1158 struct commit_extra_header *extra = NULL, **tail = &extra;
1159 int result;
1161 append_merge_tag_headers(parents, &tail);
1162 result = commit_tree_extended(msg, msg_len, tree, parents, ret,
1163 author, sign_commit, extra);
1164 free_commit_extra_headers(extra);
1165 return result;
1168 static int find_invalid_utf8(const char *buf, int len)
1170 int offset = 0;
1171 static const unsigned int max_codepoint[] = {
1172 0x7f, 0x7ff, 0xffff, 0x10ffff
1175 while (len) {
1176 unsigned char c = *buf++;
1177 int bytes, bad_offset;
1178 unsigned int codepoint;
1179 unsigned int min_val, max_val;
1181 len--;
1182 offset++;
1184 /* Simple US-ASCII? No worries. */
1185 if (c < 0x80)
1186 continue;
1188 bad_offset = offset-1;
1191 * Count how many more high bits set: that's how
1192 * many more bytes this sequence should have.
1194 bytes = 0;
1195 while (c & 0x40) {
1196 c <<= 1;
1197 bytes++;
1201 * Must be between 1 and 3 more bytes. Longer sequences result in
1202 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1204 if (bytes < 1 || 3 < bytes)
1205 return bad_offset;
1207 /* Do we *have* that many bytes? */
1208 if (len < bytes)
1209 return bad_offset;
1212 * Place the encoded bits at the bottom of the value and compute the
1213 * valid range.
1215 codepoint = (c & 0x7f) >> bytes;
1216 min_val = max_codepoint[bytes-1] + 1;
1217 max_val = max_codepoint[bytes];
1219 offset += bytes;
1220 len -= bytes;
1222 /* And verify that they are good continuation bytes */
1223 do {
1224 codepoint <<= 6;
1225 codepoint |= *buf & 0x3f;
1226 if ((*buf++ & 0xc0) != 0x80)
1227 return bad_offset;
1228 } while (--bytes);
1230 /* Reject codepoints that are out of range for the sequence length. */
1231 if (codepoint < min_val || codepoint > max_val)
1232 return bad_offset;
1233 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1234 if ((codepoint & 0x1ff800) == 0xd800)
1235 return bad_offset;
1236 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1237 if ((codepoint & 0xfffe) == 0xfffe)
1238 return bad_offset;
1239 /* So are anything in the range U+FDD0..U+FDEF. */
1240 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1241 return bad_offset;
1243 return -1;
1247 * This verifies that the buffer is in proper utf8 format.
1249 * If it isn't, it assumes any non-utf8 characters are Latin1,
1250 * and does the conversion.
1252 static int verify_utf8(struct strbuf *buf)
1254 int ok = 1;
1255 long pos = 0;
1257 for (;;) {
1258 int bad;
1259 unsigned char c;
1260 unsigned char replace[2];
1262 bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1263 if (bad < 0)
1264 return ok;
1265 pos += bad;
1266 ok = 0;
1267 c = buf->buf[pos];
1268 strbuf_remove(buf, pos, 1);
1270 /* We know 'c' must be in the range 128-255 */
1271 replace[0] = 0xc0 + (c >> 6);
1272 replace[1] = 0x80 + (c & 0x3f);
1273 strbuf_insert(buf, pos, replace, 2);
1274 pos += 2;
1278 static const char commit_utf8_warn[] =
1279 N_("Warning: commit message did not conform to UTF-8.\n"
1280 "You may want to amend it after fixing the message, or set the config\n"
1281 "variable i18n.commitencoding to the encoding your project uses.\n");
1283 int commit_tree_extended(const char *msg, size_t msg_len,
1284 const struct object_id *tree,
1285 struct commit_list *parents, struct object_id *ret,
1286 const char *author, const char *sign_commit,
1287 struct commit_extra_header *extra)
1289 int result;
1290 int encoding_is_utf8;
1291 struct strbuf buffer;
1293 assert_oid_type(tree, OBJ_TREE);
1295 if (memchr(msg, '\0', msg_len))
1296 return error("a NUL byte in commit log message not allowed.");
1298 /* Not having i18n.commitencoding is the same as having utf-8 */
1299 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1301 strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1302 strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1305 * NOTE! This ordering means that the same exact tree merged with a
1306 * different order of parents will be a _different_ changeset even
1307 * if everything else stays the same.
1309 while (parents) {
1310 struct commit *parent = pop_commit(&parents);
1311 strbuf_addf(&buffer, "parent %s\n",
1312 oid_to_hex(&parent->object.oid));
1315 /* Person/date information */
1316 if (!author)
1317 author = git_author_info(IDENT_STRICT);
1318 strbuf_addf(&buffer, "author %s\n", author);
1319 strbuf_addf(&buffer, "committer %s\n", git_committer_info(IDENT_STRICT));
1320 if (!encoding_is_utf8)
1321 strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1323 while (extra) {
1324 add_extra_header(&buffer, extra);
1325 extra = extra->next;
1327 strbuf_addch(&buffer, '\n');
1329 /* And add the comment */
1330 strbuf_add(&buffer, msg, msg_len);
1332 /* And check the encoding */
1333 if (encoding_is_utf8 && !verify_utf8(&buffer))
1334 fprintf(stderr, _(commit_utf8_warn));
1336 if (sign_commit && do_sign_commit(&buffer, sign_commit)) {
1337 result = -1;
1338 goto out;
1341 result = write_object_file(buffer.buf, buffer.len, commit_type, ret);
1342 out:
1343 strbuf_release(&buffer);
1344 return result;
1347 define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1348 static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1350 struct merge_remote_desc *merge_remote_util(struct commit *commit)
1352 return *merge_desc_slab_at(&merge_desc_slab, commit);
1355 void set_merge_remote_desc(struct commit *commit,
1356 const char *name, struct object *obj)
1358 struct merge_remote_desc *desc;
1359 FLEX_ALLOC_STR(desc, name, name);
1360 desc->obj = obj;
1361 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1364 struct commit *get_merge_parent(const char *name)
1366 struct object *obj;
1367 struct commit *commit;
1368 struct object_id oid;
1369 if (get_oid(name, &oid))
1370 return NULL;
1371 obj = parse_object(the_repository, &oid);
1372 commit = (struct commit *)peel_to_type(name, 0, obj, OBJ_COMMIT);
1373 if (commit && !merge_remote_util(commit))
1374 set_merge_remote_desc(commit, name, obj);
1375 return commit;
1379 * Append a commit to the end of the commit_list.
1381 * next starts by pointing to the variable that holds the head of an
1382 * empty commit_list, and is updated to point to the "next" field of
1383 * the last item on the list as new commits are appended.
1385 * Usage example:
1387 * struct commit_list *list;
1388 * struct commit_list **next = &list;
1390 * next = commit_list_append(c1, next);
1391 * next = commit_list_append(c2, next);
1392 * assert(commit_list_count(list) == 2);
1393 * return list;
1395 struct commit_list **commit_list_append(struct commit *commit,
1396 struct commit_list **next)
1398 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1399 new_commit->item = commit;
1400 *next = new_commit;
1401 new_commit->next = NULL;
1402 return &new_commit->next;
1405 const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1407 int key_len = strlen(key);
1408 const char *line = msg;
1410 while (line) {
1411 const char *eol = strchrnul(line, '\n');
1413 if (line == eol)
1414 return NULL;
1416 if (eol - line > key_len &&
1417 !strncmp(line, key, key_len) &&
1418 line[key_len] == ' ') {
1419 *out_len = eol - line - key_len - 1;
1420 return line + key_len + 1;
1422 line = *eol ? eol + 1 : NULL;
1424 return NULL;
1428 * Inspect the given string and determine the true "end" of the log message, in
1429 * order to find where to put a new Signed-off-by: line. Ignored are
1430 * trailing comment lines and blank lines. To support "git commit -s
1431 * --amend" on an existing commit, we also ignore "Conflicts:". To
1432 * support "git commit -v", we truncate at cut lines.
1434 * Returns the number of bytes from the tail to ignore, to be fed as
1435 * the second parameter to append_signoff().
1437 size_t ignore_non_trailer(const char *buf, size_t len)
1439 size_t boc = 0;
1440 size_t bol = 0;
1441 int in_old_conflicts_block = 0;
1442 size_t cutoff = wt_status_locate_end(buf, len);
1444 while (bol < cutoff) {
1445 const char *next_line = memchr(buf + bol, '\n', len - bol);
1447 if (!next_line)
1448 next_line = buf + len;
1449 else
1450 next_line++;
1452 if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1453 /* is this the first of the run of comments? */
1454 if (!boc)
1455 boc = bol;
1456 /* otherwise, it is just continuing */
1457 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1458 in_old_conflicts_block = 1;
1459 if (!boc)
1460 boc = bol;
1461 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1462 ; /* a pathname in the conflicts block */
1463 } else if (boc) {
1464 /* the previous was not trailing comment */
1465 boc = 0;
1466 in_old_conflicts_block = 0;
1468 bol = next_line - buf;
1470 return boc ? len - boc : len - cutoff;