commit: allow parse_commit* to handle any repo
[git.git] / commit.c
blob7a931d7fd49fe6c5149d7e42898744d5778dc8d2
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 *get_commit_buffer(const struct commit *commit, unsigned long *sizep)
302 const void *ret = get_cached_commit_buffer(the_repository, commit, sizep);
303 if (!ret) {
304 enum object_type type;
305 unsigned long size;
306 ret = read_object_file(&commit->object.oid, &type, &size);
307 if (!ret)
308 die("cannot read commit object %s",
309 oid_to_hex(&commit->object.oid));
310 if (type != OBJ_COMMIT)
311 die("expected commit for %s, got %s",
312 oid_to_hex(&commit->object.oid), type_name(type));
313 if (sizep)
314 *sizep = size;
316 return ret;
319 void unuse_commit_buffer(const struct commit *commit, const void *buffer)
321 struct commit_buffer *v = buffer_slab_peek(
322 the_repository->parsed_objects->buffer_slab, commit);
323 if (!(v && v->buffer == buffer))
324 free((void *)buffer);
327 void free_commit_buffer(struct commit *commit)
329 struct commit_buffer *v = buffer_slab_peek(
330 the_repository->parsed_objects->buffer_slab, commit);
331 if (v) {
332 FREE_AND_NULL(v->buffer);
333 v->size = 0;
337 struct tree *get_commit_tree(const struct commit *commit)
339 if (commit->maybe_tree || !commit->object.parsed)
340 return commit->maybe_tree;
342 if (commit->graph_pos == COMMIT_NOT_FROM_GRAPH)
343 BUG("commit has NULL tree, but was not loaded from commit-graph");
345 return get_commit_tree_in_graph(the_repository, commit);
348 struct object_id *get_commit_tree_oid(const struct commit *commit)
350 return &get_commit_tree(commit)->object.oid;
353 void release_commit_memory(struct commit *c)
355 c->maybe_tree = NULL;
356 c->index = 0;
357 free_commit_buffer(c);
358 free_commit_list(c->parents);
359 /* TODO: what about commit->util? */
361 c->object.parsed = 0;
364 const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
366 struct commit_buffer *v = buffer_slab_peek(
367 the_repository->parsed_objects->buffer_slab, commit);
368 void *ret;
370 if (!v) {
371 if (sizep)
372 *sizep = 0;
373 return NULL;
375 ret = v->buffer;
376 if (sizep)
377 *sizep = v->size;
379 v->buffer = NULL;
380 v->size = 0;
381 return ret;
384 int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
386 const char *tail = buffer;
387 const char *bufptr = buffer;
388 struct object_id parent;
389 struct commit_list **pptr;
390 struct commit_graft *graft;
391 const int tree_entry_len = the_hash_algo->hexsz + 5;
392 const int parent_entry_len = the_hash_algo->hexsz + 7;
394 if (item->object.parsed)
395 return 0;
396 item->object.parsed = 1;
397 tail += size;
398 if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
399 bufptr[tree_entry_len] != '\n')
400 return error("bogus commit object %s", oid_to_hex(&item->object.oid));
401 if (get_oid_hex(bufptr + 5, &parent) < 0)
402 return error("bad tree pointer in commit %s",
403 oid_to_hex(&item->object.oid));
404 item->maybe_tree = lookup_tree(r, &parent);
405 bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
406 pptr = &item->parents;
408 graft = lookup_commit_graft(r, &item->object.oid);
409 while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
410 struct commit *new_parent;
412 if (tail <= bufptr + parent_entry_len + 1 ||
413 get_oid_hex(bufptr + 7, &parent) ||
414 bufptr[parent_entry_len] != '\n')
415 return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
416 bufptr += parent_entry_len + 1;
418 * The clone is shallow if nr_parent < 0, and we must
419 * not traverse its real parents even when we unhide them.
421 if (graft && (graft->nr_parent < 0 || grafts_replace_parents))
422 continue;
423 new_parent = lookup_commit(r, &parent);
424 if (new_parent)
425 pptr = &commit_list_insert(new_parent, pptr)->next;
427 if (graft) {
428 int i;
429 struct commit *new_parent;
430 for (i = 0; i < graft->nr_parent; i++) {
431 new_parent = lookup_commit(r,
432 &graft->parent[i]);
433 if (!new_parent)
434 continue;
435 pptr = &commit_list_insert(new_parent, pptr)->next;
438 item->date = parse_commit_date(bufptr, tail);
440 if (check_graph)
441 load_commit_graph_info(the_repository, item);
443 return 0;
446 int repo_parse_commit_internal(struct repository *r,
447 struct commit *item,
448 int quiet_on_missing,
449 int use_commit_graph)
451 enum object_type type;
452 void *buffer;
453 unsigned long size;
454 int ret;
456 if (!item)
457 return -1;
458 if (item->object.parsed)
459 return 0;
460 if (use_commit_graph && parse_commit_in_graph(r, item))
461 return 0;
462 buffer = repo_read_object_file(r, &item->object.oid, &type, &size);
463 if (!buffer)
464 return quiet_on_missing ? -1 :
465 error("Could not read %s",
466 oid_to_hex(&item->object.oid));
467 if (type != OBJ_COMMIT) {
468 free(buffer);
469 return error("Object %s not a commit",
470 oid_to_hex(&item->object.oid));
473 ret = parse_commit_buffer(r, item, buffer, size, 0);
474 if (save_commit_buffer && !ret) {
475 set_commit_buffer(r, item, buffer, size);
476 return 0;
478 free(buffer);
479 return ret;
482 int repo_parse_commit_gently(struct repository *r,
483 struct commit *item, int quiet_on_missing)
485 return repo_parse_commit_internal(r, item, quiet_on_missing, 1);
488 void parse_commit_or_die(struct commit *item)
490 if (parse_commit(item))
491 die("unable to parse commit %s",
492 item ? oid_to_hex(&item->object.oid) : "(null)");
495 int find_commit_subject(const char *commit_buffer, const char **subject)
497 const char *eol;
498 const char *p = commit_buffer;
500 while (*p && (*p != '\n' || p[1] != '\n'))
501 p++;
502 if (*p) {
503 p = skip_blank_lines(p + 2);
504 eol = strchrnul(p, '\n');
505 } else
506 eol = p;
508 *subject = p;
510 return eol - p;
513 struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
515 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
516 new_list->item = item;
517 new_list->next = *list_p;
518 *list_p = new_list;
519 return new_list;
522 unsigned commit_list_count(const struct commit_list *l)
524 unsigned c = 0;
525 for (; l; l = l->next )
526 c++;
527 return c;
530 struct commit_list *copy_commit_list(struct commit_list *list)
532 struct commit_list *head = NULL;
533 struct commit_list **pp = &head;
534 while (list) {
535 pp = commit_list_append(list->item, pp);
536 list = list->next;
538 return head;
541 void free_commit_list(struct commit_list *list)
543 while (list)
544 pop_commit(&list);
547 struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
549 struct commit_list **pp = list;
550 struct commit_list *p;
551 while ((p = *pp) != NULL) {
552 if (p->item->date < item->date) {
553 break;
555 pp = &p->next;
557 return commit_list_insert(item, pp);
560 static int commit_list_compare_by_date(const void *a, const void *b)
562 timestamp_t a_date = ((const struct commit_list *)a)->item->date;
563 timestamp_t b_date = ((const struct commit_list *)b)->item->date;
564 if (a_date < b_date)
565 return 1;
566 if (a_date > b_date)
567 return -1;
568 return 0;
571 static void *commit_list_get_next(const void *a)
573 return ((const struct commit_list *)a)->next;
576 static void commit_list_set_next(void *a, void *next)
578 ((struct commit_list *)a)->next = next;
581 void commit_list_sort_by_date(struct commit_list **list)
583 *list = llist_mergesort(*list, commit_list_get_next, commit_list_set_next,
584 commit_list_compare_by_date);
587 struct commit *pop_most_recent_commit(struct commit_list **list,
588 unsigned int mark)
590 struct commit *ret = pop_commit(list);
591 struct commit_list *parents = ret->parents;
593 while (parents) {
594 struct commit *commit = parents->item;
595 if (!parse_commit(commit) && !(commit->object.flags & mark)) {
596 commit->object.flags |= mark;
597 commit_list_insert_by_date(commit, list);
599 parents = parents->next;
601 return ret;
604 static void clear_commit_marks_1(struct commit_list **plist,
605 struct commit *commit, unsigned int mark)
607 while (commit) {
608 struct commit_list *parents;
610 if (!(mark & commit->object.flags))
611 return;
613 commit->object.flags &= ~mark;
615 parents = commit->parents;
616 if (!parents)
617 return;
619 while ((parents = parents->next))
620 commit_list_insert(parents->item, plist);
622 commit = commit->parents->item;
626 void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
628 struct commit_list *list = NULL;
630 while (nr--) {
631 clear_commit_marks_1(&list, *commit, mark);
632 commit++;
634 while (list)
635 clear_commit_marks_1(&list, pop_commit(&list), mark);
638 void clear_commit_marks(struct commit *commit, unsigned int mark)
640 clear_commit_marks_many(1, &commit, mark);
643 struct commit *pop_commit(struct commit_list **stack)
645 struct commit_list *top = *stack;
646 struct commit *item = top ? top->item : NULL;
648 if (top) {
649 *stack = top->next;
650 free(top);
652 return item;
656 * Topological sort support
659 /* count number of children that have not been emitted */
660 define_commit_slab(indegree_slab, int);
662 /* record author-date for each commit object */
663 define_commit_slab(author_date_slab, timestamp_t);
665 static void record_author_date(struct author_date_slab *author_date,
666 struct commit *commit)
668 const char *buffer = get_commit_buffer(commit, NULL);
669 struct ident_split ident;
670 const char *ident_line;
671 size_t ident_len;
672 char *date_end;
673 timestamp_t date;
675 ident_line = find_commit_header(buffer, "author", &ident_len);
676 if (!ident_line)
677 goto fail_exit; /* no author line */
678 if (split_ident_line(&ident, ident_line, ident_len) ||
679 !ident.date_begin || !ident.date_end)
680 goto fail_exit; /* malformed "author" line */
682 date = parse_timestamp(ident.date_begin, &date_end, 10);
683 if (date_end != ident.date_end)
684 goto fail_exit; /* malformed date */
685 *(author_date_slab_at(author_date, commit)) = date;
687 fail_exit:
688 unuse_commit_buffer(commit, buffer);
691 static int compare_commits_by_author_date(const void *a_, const void *b_,
692 void *cb_data)
694 const struct commit *a = a_, *b = b_;
695 struct author_date_slab *author_date = cb_data;
696 timestamp_t a_date = *(author_date_slab_at(author_date, a));
697 timestamp_t b_date = *(author_date_slab_at(author_date, b));
699 /* newer commits with larger date first */
700 if (a_date < b_date)
701 return 1;
702 else if (a_date > b_date)
703 return -1;
704 return 0;
707 int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, void *unused)
709 const struct commit *a = a_, *b = b_;
711 /* newer commits first */
712 if (a->generation < b->generation)
713 return 1;
714 else if (a->generation > b->generation)
715 return -1;
717 /* use date as a heuristic when generations are equal */
718 if (a->date < b->date)
719 return 1;
720 else if (a->date > b->date)
721 return -1;
722 return 0;
725 int compare_commits_by_commit_date(const void *a_, const void *b_, void *unused)
727 const struct commit *a = a_, *b = b_;
728 /* newer commits with larger date first */
729 if (a->date < b->date)
730 return 1;
731 else if (a->date > b->date)
732 return -1;
733 return 0;
737 * Performs an in-place topological sort on the list supplied.
739 void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
741 struct commit_list *next, *orig = *list;
742 struct commit_list **pptr;
743 struct indegree_slab indegree;
744 struct prio_queue queue;
745 struct commit *commit;
746 struct author_date_slab author_date;
748 if (!orig)
749 return;
750 *list = NULL;
752 init_indegree_slab(&indegree);
753 memset(&queue, '\0', sizeof(queue));
755 switch (sort_order) {
756 default: /* REV_SORT_IN_GRAPH_ORDER */
757 queue.compare = NULL;
758 break;
759 case REV_SORT_BY_COMMIT_DATE:
760 queue.compare = compare_commits_by_commit_date;
761 break;
762 case REV_SORT_BY_AUTHOR_DATE:
763 init_author_date_slab(&author_date);
764 queue.compare = compare_commits_by_author_date;
765 queue.cb_data = &author_date;
766 break;
769 /* Mark them and clear the indegree */
770 for (next = orig; next; next = next->next) {
771 struct commit *commit = next->item;
772 *(indegree_slab_at(&indegree, commit)) = 1;
773 /* also record the author dates, if needed */
774 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
775 record_author_date(&author_date, commit);
778 /* update the indegree */
779 for (next = orig; next; next = next->next) {
780 struct commit_list *parents = next->item->parents;
781 while (parents) {
782 struct commit *parent = parents->item;
783 int *pi = indegree_slab_at(&indegree, parent);
785 if (*pi)
786 (*pi)++;
787 parents = parents->next;
792 * find the tips
794 * tips are nodes not reachable from any other node in the list
796 * the tips serve as a starting set for the work queue.
798 for (next = orig; next; next = next->next) {
799 struct commit *commit = next->item;
801 if (*(indegree_slab_at(&indegree, commit)) == 1)
802 prio_queue_put(&queue, commit);
806 * This is unfortunate; the initial tips need to be shown
807 * in the order given from the revision traversal machinery.
809 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
810 prio_queue_reverse(&queue);
812 /* We no longer need the commit list */
813 free_commit_list(orig);
815 pptr = list;
816 *list = NULL;
817 while ((commit = prio_queue_get(&queue)) != NULL) {
818 struct commit_list *parents;
820 for (parents = commit->parents; parents ; parents = parents->next) {
821 struct commit *parent = parents->item;
822 int *pi = indegree_slab_at(&indegree, parent);
824 if (!*pi)
825 continue;
828 * parents are only enqueued for emission
829 * when all their children have been emitted thereby
830 * guaranteeing topological order.
832 if (--(*pi) == 1)
833 prio_queue_put(&queue, parent);
836 * all children of commit have already been
837 * emitted. we can emit it now.
839 *(indegree_slab_at(&indegree, commit)) = 0;
841 pptr = &commit_list_insert(commit, pptr)->next;
844 clear_indegree_slab(&indegree);
845 clear_prio_queue(&queue);
846 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
847 clear_author_date_slab(&author_date);
850 static const char gpg_sig_header[] = "gpgsig";
851 static const int gpg_sig_header_len = sizeof(gpg_sig_header) - 1;
853 static int do_sign_commit(struct strbuf *buf, const char *keyid)
855 struct strbuf sig = STRBUF_INIT;
856 int inspos, copypos;
857 const char *eoh;
859 /* find the end of the header */
860 eoh = strstr(buf->buf, "\n\n");
861 if (!eoh)
862 inspos = buf->len;
863 else
864 inspos = eoh - buf->buf + 1;
866 if (!keyid || !*keyid)
867 keyid = get_signing_key();
868 if (sign_buffer(buf, &sig, keyid)) {
869 strbuf_release(&sig);
870 return -1;
873 for (copypos = 0; sig.buf[copypos]; ) {
874 const char *bol = sig.buf + copypos;
875 const char *eol = strchrnul(bol, '\n');
876 int len = (eol - bol) + !!*eol;
878 if (!copypos) {
879 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
880 inspos += gpg_sig_header_len;
882 strbuf_insert(buf, inspos++, " ", 1);
883 strbuf_insert(buf, inspos, bol, len);
884 inspos += len;
885 copypos += len;
887 strbuf_release(&sig);
888 return 0;
891 int parse_signed_commit(const struct commit *commit,
892 struct strbuf *payload, struct strbuf *signature)
895 unsigned long size;
896 const char *buffer = get_commit_buffer(commit, &size);
897 int in_signature, saw_signature = -1;
898 const char *line, *tail;
900 line = buffer;
901 tail = buffer + size;
902 in_signature = 0;
903 saw_signature = 0;
904 while (line < tail) {
905 const char *sig = NULL;
906 const char *next = memchr(line, '\n', tail - line);
908 next = next ? next + 1 : tail;
909 if (in_signature && line[0] == ' ')
910 sig = line + 1;
911 else if (starts_with(line, gpg_sig_header) &&
912 line[gpg_sig_header_len] == ' ')
913 sig = line + gpg_sig_header_len + 1;
914 if (sig) {
915 strbuf_add(signature, sig, next - sig);
916 saw_signature = 1;
917 in_signature = 1;
918 } else {
919 if (*line == '\n')
920 /* dump the whole remainder of the buffer */
921 next = tail;
922 strbuf_add(payload, line, next - line);
923 in_signature = 0;
925 line = next;
927 unuse_commit_buffer(commit, buffer);
928 return saw_signature;
931 int remove_signature(struct strbuf *buf)
933 const char *line = buf->buf;
934 const char *tail = buf->buf + buf->len;
935 int in_signature = 0;
936 const char *sig_start = NULL;
937 const char *sig_end = NULL;
939 while (line < tail) {
940 const char *next = memchr(line, '\n', tail - line);
941 next = next ? next + 1 : tail;
943 if (in_signature && line[0] == ' ')
944 sig_end = next;
945 else if (starts_with(line, gpg_sig_header) &&
946 line[gpg_sig_header_len] == ' ') {
947 sig_start = line;
948 sig_end = next;
949 in_signature = 1;
950 } else {
951 if (*line == '\n')
952 /* dump the whole remainder of the buffer */
953 next = tail;
954 in_signature = 0;
956 line = next;
959 if (sig_start)
960 strbuf_remove(buf, sig_start - buf->buf, sig_end - sig_start);
962 return sig_start != NULL;
965 static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
967 struct merge_remote_desc *desc;
968 struct commit_extra_header *mergetag;
969 char *buf;
970 unsigned long size, len;
971 enum object_type type;
973 desc = merge_remote_util(parent);
974 if (!desc || !desc->obj)
975 return;
976 buf = read_object_file(&desc->obj->oid, &type, &size);
977 if (!buf || type != OBJ_TAG)
978 goto free_return;
979 len = parse_signature(buf, size);
980 if (size == len)
981 goto free_return;
983 * We could verify this signature and either omit the tag when
984 * it does not validate, but the integrator may not have the
985 * public key of the signer of the tag he is merging, while a
986 * later auditor may have it while auditing, so let's not run
987 * verify-signed-buffer here for now...
989 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
990 * warn("warning: signed tag unverified.");
992 mergetag = xcalloc(1, sizeof(*mergetag));
993 mergetag->key = xstrdup("mergetag");
994 mergetag->value = buf;
995 mergetag->len = size;
997 **tail = mergetag;
998 *tail = &mergetag->next;
999 return;
1001 free_return:
1002 free(buf);
1005 int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1007 struct strbuf payload = STRBUF_INIT;
1008 struct strbuf signature = STRBUF_INIT;
1009 int ret = 1;
1011 sigc->result = 'N';
1013 if (parse_signed_commit(commit, &payload, &signature) <= 0)
1014 goto out;
1015 ret = check_signature(payload.buf, payload.len, signature.buf,
1016 signature.len, sigc);
1018 out:
1019 strbuf_release(&payload);
1020 strbuf_release(&signature);
1022 return ret;
1027 void append_merge_tag_headers(struct commit_list *parents,
1028 struct commit_extra_header ***tail)
1030 while (parents) {
1031 struct commit *parent = parents->item;
1032 handle_signed_tag(parent, tail);
1033 parents = parents->next;
1037 static void add_extra_header(struct strbuf *buffer,
1038 struct commit_extra_header *extra)
1040 strbuf_addstr(buffer, extra->key);
1041 if (extra->len)
1042 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1043 else
1044 strbuf_addch(buffer, '\n');
1047 struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1048 const char **exclude)
1050 struct commit_extra_header *extra = NULL;
1051 unsigned long size;
1052 const char *buffer = get_commit_buffer(commit, &size);
1053 extra = read_commit_extra_header_lines(buffer, size, exclude);
1054 unuse_commit_buffer(commit, buffer);
1055 return extra;
1058 int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1060 struct commit_extra_header *extra, *to_free;
1061 int res = 0;
1063 to_free = read_commit_extra_headers(commit, NULL);
1064 for (extra = to_free; !res && extra; extra = extra->next) {
1065 if (strcmp(extra->key, "mergetag"))
1066 continue; /* not a merge tag */
1067 res = fn(commit, extra, data);
1069 free_commit_extra_headers(to_free);
1070 return res;
1073 static inline int standard_header_field(const char *field, size_t len)
1075 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1076 (len == 6 && !memcmp(field, "parent", 6)) ||
1077 (len == 6 && !memcmp(field, "author", 6)) ||
1078 (len == 9 && !memcmp(field, "committer", 9)) ||
1079 (len == 8 && !memcmp(field, "encoding", 8)));
1082 static int excluded_header_field(const char *field, size_t len, const char **exclude)
1084 if (!exclude)
1085 return 0;
1087 while (*exclude) {
1088 size_t xlen = strlen(*exclude);
1089 if (len == xlen && !memcmp(field, *exclude, xlen))
1090 return 1;
1091 exclude++;
1093 return 0;
1096 static struct commit_extra_header *read_commit_extra_header_lines(
1097 const char *buffer, size_t size,
1098 const char **exclude)
1100 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1101 const char *line, *next, *eof, *eob;
1102 struct strbuf buf = STRBUF_INIT;
1104 for (line = buffer, eob = line + size;
1105 line < eob && *line != '\n';
1106 line = next) {
1107 next = memchr(line, '\n', eob - line);
1108 next = next ? next + 1 : eob;
1109 if (*line == ' ') {
1110 /* continuation */
1111 if (it)
1112 strbuf_add(&buf, line + 1, next - (line + 1));
1113 continue;
1115 if (it)
1116 it->value = strbuf_detach(&buf, &it->len);
1117 strbuf_reset(&buf);
1118 it = NULL;
1120 eof = memchr(line, ' ', next - line);
1121 if (!eof)
1122 eof = next;
1123 else if (standard_header_field(line, eof - line) ||
1124 excluded_header_field(line, eof - line, exclude))
1125 continue;
1127 it = xcalloc(1, sizeof(*it));
1128 it->key = xmemdupz(line, eof-line);
1129 *tail = it;
1130 tail = &it->next;
1131 if (eof + 1 < next)
1132 strbuf_add(&buf, eof + 1, next - (eof + 1));
1134 if (it)
1135 it->value = strbuf_detach(&buf, &it->len);
1136 return extra;
1139 void free_commit_extra_headers(struct commit_extra_header *extra)
1141 while (extra) {
1142 struct commit_extra_header *next = extra->next;
1143 free(extra->key);
1144 free(extra->value);
1145 free(extra);
1146 extra = next;
1150 int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1151 struct commit_list *parents, struct object_id *ret,
1152 const char *author, const char *sign_commit)
1154 struct commit_extra_header *extra = NULL, **tail = &extra;
1155 int result;
1157 append_merge_tag_headers(parents, &tail);
1158 result = commit_tree_extended(msg, msg_len, tree, parents, ret,
1159 author, sign_commit, extra);
1160 free_commit_extra_headers(extra);
1161 return result;
1164 static int find_invalid_utf8(const char *buf, int len)
1166 int offset = 0;
1167 static const unsigned int max_codepoint[] = {
1168 0x7f, 0x7ff, 0xffff, 0x10ffff
1171 while (len) {
1172 unsigned char c = *buf++;
1173 int bytes, bad_offset;
1174 unsigned int codepoint;
1175 unsigned int min_val, max_val;
1177 len--;
1178 offset++;
1180 /* Simple US-ASCII? No worries. */
1181 if (c < 0x80)
1182 continue;
1184 bad_offset = offset-1;
1187 * Count how many more high bits set: that's how
1188 * many more bytes this sequence should have.
1190 bytes = 0;
1191 while (c & 0x40) {
1192 c <<= 1;
1193 bytes++;
1197 * Must be between 1 and 3 more bytes. Longer sequences result in
1198 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1200 if (bytes < 1 || 3 < bytes)
1201 return bad_offset;
1203 /* Do we *have* that many bytes? */
1204 if (len < bytes)
1205 return bad_offset;
1208 * Place the encoded bits at the bottom of the value and compute the
1209 * valid range.
1211 codepoint = (c & 0x7f) >> bytes;
1212 min_val = max_codepoint[bytes-1] + 1;
1213 max_val = max_codepoint[bytes];
1215 offset += bytes;
1216 len -= bytes;
1218 /* And verify that they are good continuation bytes */
1219 do {
1220 codepoint <<= 6;
1221 codepoint |= *buf & 0x3f;
1222 if ((*buf++ & 0xc0) != 0x80)
1223 return bad_offset;
1224 } while (--bytes);
1226 /* Reject codepoints that are out of range for the sequence length. */
1227 if (codepoint < min_val || codepoint > max_val)
1228 return bad_offset;
1229 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1230 if ((codepoint & 0x1ff800) == 0xd800)
1231 return bad_offset;
1232 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1233 if ((codepoint & 0xfffe) == 0xfffe)
1234 return bad_offset;
1235 /* So are anything in the range U+FDD0..U+FDEF. */
1236 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1237 return bad_offset;
1239 return -1;
1243 * This verifies that the buffer is in proper utf8 format.
1245 * If it isn't, it assumes any non-utf8 characters are Latin1,
1246 * and does the conversion.
1248 static int verify_utf8(struct strbuf *buf)
1250 int ok = 1;
1251 long pos = 0;
1253 for (;;) {
1254 int bad;
1255 unsigned char c;
1256 unsigned char replace[2];
1258 bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1259 if (bad < 0)
1260 return ok;
1261 pos += bad;
1262 ok = 0;
1263 c = buf->buf[pos];
1264 strbuf_remove(buf, pos, 1);
1266 /* We know 'c' must be in the range 128-255 */
1267 replace[0] = 0xc0 + (c >> 6);
1268 replace[1] = 0x80 + (c & 0x3f);
1269 strbuf_insert(buf, pos, replace, 2);
1270 pos += 2;
1274 static const char commit_utf8_warn[] =
1275 N_("Warning: commit message did not conform to UTF-8.\n"
1276 "You may want to amend it after fixing the message, or set the config\n"
1277 "variable i18n.commitencoding to the encoding your project uses.\n");
1279 int commit_tree_extended(const char *msg, size_t msg_len,
1280 const struct object_id *tree,
1281 struct commit_list *parents, struct object_id *ret,
1282 const char *author, const char *sign_commit,
1283 struct commit_extra_header *extra)
1285 int result;
1286 int encoding_is_utf8;
1287 struct strbuf buffer;
1289 assert_oid_type(tree, OBJ_TREE);
1291 if (memchr(msg, '\0', msg_len))
1292 return error("a NUL byte in commit log message not allowed.");
1294 /* Not having i18n.commitencoding is the same as having utf-8 */
1295 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1297 strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1298 strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1301 * NOTE! This ordering means that the same exact tree merged with a
1302 * different order of parents will be a _different_ changeset even
1303 * if everything else stays the same.
1305 while (parents) {
1306 struct commit *parent = pop_commit(&parents);
1307 strbuf_addf(&buffer, "parent %s\n",
1308 oid_to_hex(&parent->object.oid));
1311 /* Person/date information */
1312 if (!author)
1313 author = git_author_info(IDENT_STRICT);
1314 strbuf_addf(&buffer, "author %s\n", author);
1315 strbuf_addf(&buffer, "committer %s\n", git_committer_info(IDENT_STRICT));
1316 if (!encoding_is_utf8)
1317 strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1319 while (extra) {
1320 add_extra_header(&buffer, extra);
1321 extra = extra->next;
1323 strbuf_addch(&buffer, '\n');
1325 /* And add the comment */
1326 strbuf_add(&buffer, msg, msg_len);
1328 /* And check the encoding */
1329 if (encoding_is_utf8 && !verify_utf8(&buffer))
1330 fprintf(stderr, _(commit_utf8_warn));
1332 if (sign_commit && do_sign_commit(&buffer, sign_commit)) {
1333 result = -1;
1334 goto out;
1337 result = write_object_file(buffer.buf, buffer.len, commit_type, ret);
1338 out:
1339 strbuf_release(&buffer);
1340 return result;
1343 define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1344 static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1346 struct merge_remote_desc *merge_remote_util(struct commit *commit)
1348 return *merge_desc_slab_at(&merge_desc_slab, commit);
1351 void set_merge_remote_desc(struct commit *commit,
1352 const char *name, struct object *obj)
1354 struct merge_remote_desc *desc;
1355 FLEX_ALLOC_STR(desc, name, name);
1356 desc->obj = obj;
1357 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1360 struct commit *get_merge_parent(const char *name)
1362 struct object *obj;
1363 struct commit *commit;
1364 struct object_id oid;
1365 if (get_oid(name, &oid))
1366 return NULL;
1367 obj = parse_object(the_repository, &oid);
1368 commit = (struct commit *)peel_to_type(name, 0, obj, OBJ_COMMIT);
1369 if (commit && !merge_remote_util(commit))
1370 set_merge_remote_desc(commit, name, obj);
1371 return commit;
1375 * Append a commit to the end of the commit_list.
1377 * next starts by pointing to the variable that holds the head of an
1378 * empty commit_list, and is updated to point to the "next" field of
1379 * the last item on the list as new commits are appended.
1381 * Usage example:
1383 * struct commit_list *list;
1384 * struct commit_list **next = &list;
1386 * next = commit_list_append(c1, next);
1387 * next = commit_list_append(c2, next);
1388 * assert(commit_list_count(list) == 2);
1389 * return list;
1391 struct commit_list **commit_list_append(struct commit *commit,
1392 struct commit_list **next)
1394 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1395 new_commit->item = commit;
1396 *next = new_commit;
1397 new_commit->next = NULL;
1398 return &new_commit->next;
1401 const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1403 int key_len = strlen(key);
1404 const char *line = msg;
1406 while (line) {
1407 const char *eol = strchrnul(line, '\n');
1409 if (line == eol)
1410 return NULL;
1412 if (eol - line > key_len &&
1413 !strncmp(line, key, key_len) &&
1414 line[key_len] == ' ') {
1415 *out_len = eol - line - key_len - 1;
1416 return line + key_len + 1;
1418 line = *eol ? eol + 1 : NULL;
1420 return NULL;
1424 * Inspect the given string and determine the true "end" of the log message, in
1425 * order to find where to put a new Signed-off-by: line. Ignored are
1426 * trailing comment lines and blank lines. To support "git commit -s
1427 * --amend" on an existing commit, we also ignore "Conflicts:". To
1428 * support "git commit -v", we truncate at cut lines.
1430 * Returns the number of bytes from the tail to ignore, to be fed as
1431 * the second parameter to append_signoff().
1433 size_t ignore_non_trailer(const char *buf, size_t len)
1435 size_t boc = 0;
1436 size_t bol = 0;
1437 int in_old_conflicts_block = 0;
1438 size_t cutoff = wt_status_locate_end(buf, len);
1440 while (bol < cutoff) {
1441 const char *next_line = memchr(buf + bol, '\n', len - bol);
1443 if (!next_line)
1444 next_line = buf + len;
1445 else
1446 next_line++;
1448 if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1449 /* is this the first of the run of comments? */
1450 if (!boc)
1451 boc = bol;
1452 /* otherwise, it is just continuing */
1453 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1454 in_old_conflicts_block = 1;
1455 if (!boc)
1456 boc = bol;
1457 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1458 ; /* a pathname in the conflicts block */
1459 } else if (boc) {
1460 /* the previous was not trailing comment */
1461 boc = 0;
1462 in_old_conflicts_block = 0;
1464 bol = next_line - buf;
1466 return boc ? len - boc : len - cutoff;