commit.c: allow get_cached_commit_buffer to handle arbitrary repositories
[git/git.git] / commit.c
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
21 static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
22
23 int save_commit_buffer = 1;
24
25 const char *commit_type = "commit";
26
27 struct commit *lookup_commit_reference_gently_the_repository(
28 const struct object_id *oid, int quiet)
29 {
30 struct object *obj = deref_tag(the_repository,
31 parse_object(the_repository, oid),
32 NULL, 0);
33
34 if (!obj)
35 return NULL;
36 return object_as_type(the_repository, obj, OBJ_COMMIT, quiet);
37 }
38
39 struct commit *lookup_commit_reference_the_repository(const struct object_id *oid)
40 {
41 return lookup_commit_reference_gently(the_repository, oid, 0);
42 }
43
44 struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
45 {
46 struct commit *c = lookup_commit_reference(the_repository, oid);
47 if (!c)
48 die(_("could not parse %s"), ref_name);
49 if (oidcmp(oid, &c->object.oid)) {
50 warning(_("%s %s is not a commit!"),
51 ref_name, oid_to_hex(oid));
52 }
53 return c;
54 }
55
56 struct commit *lookup_commit(struct repository *r, const struct object_id *oid)
57 {
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);
63 }
64
65 struct commit *lookup_commit_reference_by_name(const char *name)
66 {
67 struct object_id oid;
68 struct commit *commit;
69
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;
76 }
77
78 static timestamp_t parse_commit_date(const char *buf, const char *tail)
79 {
80 const char *dateptr;
81
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);
103 }
104
105 static const unsigned char *commit_graft_sha1_access(size_t index, void *table)
106 {
107 struct commit_graft **commit_graft_table = table;
108 return commit_graft_table[index]->oid.hash;
109 }
110
111 static int commit_graft_pos(struct repository *r, const unsigned char *sha1)
112 {
113 return sha1_pos(sha1, r->parsed_objects->grafts,
114 r->parsed_objects->grafts_nr,
115 commit_graft_sha1_access);
116 }
117
118 int register_commit_graft(struct repository *r, struct commit_graft *graft,
119 int ignore_dups)
120 {
121 int pos = commit_graft_pos(r, graft->oid.hash);
122
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;
129 }
130 return 1;
131 }
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;
144 }
145
146 struct commit_graft *read_graft_line(struct strbuf *line)
147 {
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;
153
154 strbuf_rtrim(line);
155 if (!line->len || line->buf[0] == '#')
156 return NULL;
157 /*
158 * phase 0 verifies line, counts hashes in line and allocates graft
159 * phase 1 fills graft
160 */
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;
169 }
170 if (!graft) {
171 graft = xmalloc(st_add(sizeof(*graft),
172 st_mult(sizeof(struct object_id), i)));
173 graft->nr_parent = i;
174 }
175 }
176 return graft;
177
178 bad_graft_data:
179 error("bad graft data: %s", line->buf);
180 assert(!graft);
181 return NULL;
182 }
183
184 static int read_graft_file(struct repository *r, const char *graft_file)
185 {
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);
206 }
207 fclose(fp);
208 strbuf_release(&buf);
209 return 0;
210 }
211
212 static void prepare_commit_graft(struct repository *r)
213 {
214 char *graft_file;
215
216 if (r->parsed_objects->commit_graft_prepared)
217 return;
218 if (!startup_info->have_repository)
219 return;
220
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;
226 }
227
228 struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid)
229 {
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];
236 }
237
238 int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
239 {
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;
244 }
245
246 int unregister_shallow(const struct object_id *oid)
247 {
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;
257 }
258
259 struct commit_buffer {
260 void *buffer;
261 unsigned long size;
262 };
263 define_commit_slab(buffer_slab, struct commit_buffer);
264
265 struct buffer_slab *allocate_commit_buffer_slab(void)
266 {
267 struct buffer_slab *bs = xmalloc(sizeof(*bs));
268 init_buffer_slab(bs);
269 return bs;
270 }
271
272 void free_commit_buffer_slab(struct buffer_slab *bs)
273 {
274 clear_buffer_slab(bs);
275 free(bs);
276 }
277
278 void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size)
279 {
280 struct commit_buffer *v = buffer_slab_at(
281 r->parsed_objects->buffer_slab, commit);
282 v->buffer = buffer;
283 v->size = size;
284 }
285
286 const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep)
287 {
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;
294 }
295 if (sizep)
296 *sizep = v->size;
297 return v->buffer;
298 }
299
300 const void *get_commit_buffer(const struct commit *commit, unsigned long *sizep)
301 {
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;
315 }
316 return ret;
317 }
318
319 void unuse_commit_buffer(const struct commit *commit, const void *buffer)
320 {
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);
325 }
326
327 void free_commit_buffer(struct commit *commit)
328 {
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;
334 }
335 }
336
337 struct tree *get_commit_tree(const struct commit *commit)
338 {
339 if (commit->maybe_tree || !commit->object.parsed)
340 return commit->maybe_tree;
341
342 if (commit->graph_pos == COMMIT_NOT_FROM_GRAPH)
343 BUG("commit has NULL tree, but was not loaded from commit-graph");
344
345 return get_commit_tree_in_graph(commit);
346 }
347
348 struct object_id *get_commit_tree_oid(const struct commit *commit)
349 {
350 return &get_commit_tree(commit)->object.oid;
351 }
352
353 void release_commit_memory(struct commit *c)
354 {
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? */
360
361 c->object.parsed = 0;
362 }
363
364 const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
365 {
366 struct commit_buffer *v = buffer_slab_peek(
367 the_repository->parsed_objects->buffer_slab, commit);
368 void *ret;
369
370 if (!v) {
371 if (sizep)
372 *sizep = 0;
373 return NULL;
374 }
375 ret = v->buffer;
376 if (sizep)
377 *sizep = v->size;
378
379 v->buffer = NULL;
380 v->size = 0;
381 return ret;
382 }
383
384 int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
385 {
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 = GIT_SHA1_HEXSZ + 5;
392 const int parent_entry_len = GIT_SHA1_HEXSZ + 7;
393
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;
407
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;
411
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;
417 /*
418 * The clone is shallow if nr_parent < 0, and we must
419 * not traverse its real parents even when we unhide them.
420 */
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;
426 }
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;
436 }
437 }
438 item->date = parse_commit_date(bufptr, tail);
439
440 if (check_graph)
441 load_commit_graph_info(item);
442
443 return 0;
444 }
445
446 int parse_commit_gently(struct commit *item, int quiet_on_missing)
447 {
448 enum object_type type;
449 void *buffer;
450 unsigned long size;
451 int ret;
452
453 if (!item)
454 return -1;
455 if (item->object.parsed)
456 return 0;
457 if (parse_commit_in_graph(item))
458 return 0;
459 buffer = read_object_file(&item->object.oid, &type, &size);
460 if (!buffer)
461 return quiet_on_missing ? -1 :
462 error("Could not read %s",
463 oid_to_hex(&item->object.oid));
464 if (type != OBJ_COMMIT) {
465 free(buffer);
466 return error("Object %s not a commit",
467 oid_to_hex(&item->object.oid));
468 }
469 ret = parse_commit_buffer(the_repository, item, buffer, size, 0);
470 if (save_commit_buffer && !ret) {
471 set_commit_buffer(the_repository, item, buffer, size);
472 return 0;
473 }
474 free(buffer);
475 return ret;
476 }
477
478 void parse_commit_or_die(struct commit *item)
479 {
480 if (parse_commit(item))
481 die("unable to parse commit %s",
482 item ? oid_to_hex(&item->object.oid) : "(null)");
483 }
484
485 int find_commit_subject(const char *commit_buffer, const char **subject)
486 {
487 const char *eol;
488 const char *p = commit_buffer;
489
490 while (*p && (*p != '\n' || p[1] != '\n'))
491 p++;
492 if (*p) {
493 p = skip_blank_lines(p + 2);
494 eol = strchrnul(p, '\n');
495 } else
496 eol = p;
497
498 *subject = p;
499
500 return eol - p;
501 }
502
503 struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
504 {
505 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
506 new_list->item = item;
507 new_list->next = *list_p;
508 *list_p = new_list;
509 return new_list;
510 }
511
512 unsigned commit_list_count(const struct commit_list *l)
513 {
514 unsigned c = 0;
515 for (; l; l = l->next )
516 c++;
517 return c;
518 }
519
520 struct commit_list *copy_commit_list(struct commit_list *list)
521 {
522 struct commit_list *head = NULL;
523 struct commit_list **pp = &head;
524 while (list) {
525 pp = commit_list_append(list->item, pp);
526 list = list->next;
527 }
528 return head;
529 }
530
531 void free_commit_list(struct commit_list *list)
532 {
533 while (list)
534 pop_commit(&list);
535 }
536
537 struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
538 {
539 struct commit_list **pp = list;
540 struct commit_list *p;
541 while ((p = *pp) != NULL) {
542 if (p->item->date < item->date) {
543 break;
544 }
545 pp = &p->next;
546 }
547 return commit_list_insert(item, pp);
548 }
549
550 static int commit_list_compare_by_date(const void *a, const void *b)
551 {
552 timestamp_t a_date = ((const struct commit_list *)a)->item->date;
553 timestamp_t b_date = ((const struct commit_list *)b)->item->date;
554 if (a_date < b_date)
555 return 1;
556 if (a_date > b_date)
557 return -1;
558 return 0;
559 }
560
561 static void *commit_list_get_next(const void *a)
562 {
563 return ((const struct commit_list *)a)->next;
564 }
565
566 static void commit_list_set_next(void *a, void *next)
567 {
568 ((struct commit_list *)a)->next = next;
569 }
570
571 void commit_list_sort_by_date(struct commit_list **list)
572 {
573 *list = llist_mergesort(*list, commit_list_get_next, commit_list_set_next,
574 commit_list_compare_by_date);
575 }
576
577 struct commit *pop_most_recent_commit(struct commit_list **list,
578 unsigned int mark)
579 {
580 struct commit *ret = pop_commit(list);
581 struct commit_list *parents = ret->parents;
582
583 while (parents) {
584 struct commit *commit = parents->item;
585 if (!parse_commit(commit) && !(commit->object.flags & mark)) {
586 commit->object.flags |= mark;
587 commit_list_insert_by_date(commit, list);
588 }
589 parents = parents->next;
590 }
591 return ret;
592 }
593
594 static void clear_commit_marks_1(struct commit_list **plist,
595 struct commit *commit, unsigned int mark)
596 {
597 while (commit) {
598 struct commit_list *parents;
599
600 if (!(mark & commit->object.flags))
601 return;
602
603 commit->object.flags &= ~mark;
604
605 parents = commit->parents;
606 if (!parents)
607 return;
608
609 while ((parents = parents->next))
610 commit_list_insert(parents->item, plist);
611
612 commit = commit->parents->item;
613 }
614 }
615
616 void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
617 {
618 struct commit_list *list = NULL;
619
620 while (nr--) {
621 clear_commit_marks_1(&list, *commit, mark);
622 commit++;
623 }
624 while (list)
625 clear_commit_marks_1(&list, pop_commit(&list), mark);
626 }
627
628 void clear_commit_marks(struct commit *commit, unsigned int mark)
629 {
630 clear_commit_marks_many(1, &commit, mark);
631 }
632
633 struct commit *pop_commit(struct commit_list **stack)
634 {
635 struct commit_list *top = *stack;
636 struct commit *item = top ? top->item : NULL;
637
638 if (top) {
639 *stack = top->next;
640 free(top);
641 }
642 return item;
643 }
644
645 /*
646 * Topological sort support
647 */
648
649 /* count number of children that have not been emitted */
650 define_commit_slab(indegree_slab, int);
651
652 /* record author-date for each commit object */
653 define_commit_slab(author_date_slab, unsigned long);
654
655 static void record_author_date(struct author_date_slab *author_date,
656 struct commit *commit)
657 {
658 const char *buffer = get_commit_buffer(commit, NULL);
659 struct ident_split ident;
660 const char *ident_line;
661 size_t ident_len;
662 char *date_end;
663 timestamp_t date;
664
665 ident_line = find_commit_header(buffer, "author", &ident_len);
666 if (!ident_line)
667 goto fail_exit; /* no author line */
668 if (split_ident_line(&ident, ident_line, ident_len) ||
669 !ident.date_begin || !ident.date_end)
670 goto fail_exit; /* malformed "author" line */
671
672 date = parse_timestamp(ident.date_begin, &date_end, 10);
673 if (date_end != ident.date_end)
674 goto fail_exit; /* malformed date */
675 *(author_date_slab_at(author_date, commit)) = date;
676
677 fail_exit:
678 unuse_commit_buffer(commit, buffer);
679 }
680
681 static int compare_commits_by_author_date(const void *a_, const void *b_,
682 void *cb_data)
683 {
684 const struct commit *a = a_, *b = b_;
685 struct author_date_slab *author_date = cb_data;
686 timestamp_t a_date = *(author_date_slab_at(author_date, a));
687 timestamp_t b_date = *(author_date_slab_at(author_date, b));
688
689 /* newer commits with larger date first */
690 if (a_date < b_date)
691 return 1;
692 else if (a_date > b_date)
693 return -1;
694 return 0;
695 }
696
697 int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, void *unused)
698 {
699 const struct commit *a = a_, *b = b_;
700
701 /* newer commits first */
702 if (a->generation < b->generation)
703 return 1;
704 else if (a->generation > b->generation)
705 return -1;
706
707 /* use date as a heuristic when generations are equal */
708 if (a->date < b->date)
709 return 1;
710 else if (a->date > b->date)
711 return -1;
712 return 0;
713 }
714
715 int compare_commits_by_commit_date(const void *a_, const void *b_, void *unused)
716 {
717 const struct commit *a = a_, *b = b_;
718 /* newer commits with larger date first */
719 if (a->date < b->date)
720 return 1;
721 else if (a->date > b->date)
722 return -1;
723 return 0;
724 }
725
726 /*
727 * Performs an in-place topological sort on the list supplied.
728 */
729 void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
730 {
731 struct commit_list *next, *orig = *list;
732 struct commit_list **pptr;
733 struct indegree_slab indegree;
734 struct prio_queue queue;
735 struct commit *commit;
736 struct author_date_slab author_date;
737
738 if (!orig)
739 return;
740 *list = NULL;
741
742 init_indegree_slab(&indegree);
743 memset(&queue, '\0', sizeof(queue));
744
745 switch (sort_order) {
746 default: /* REV_SORT_IN_GRAPH_ORDER */
747 queue.compare = NULL;
748 break;
749 case REV_SORT_BY_COMMIT_DATE:
750 queue.compare = compare_commits_by_commit_date;
751 break;
752 case REV_SORT_BY_AUTHOR_DATE:
753 init_author_date_slab(&author_date);
754 queue.compare = compare_commits_by_author_date;
755 queue.cb_data = &author_date;
756 break;
757 }
758
759 /* Mark them and clear the indegree */
760 for (next = orig; next; next = next->next) {
761 struct commit *commit = next->item;
762 *(indegree_slab_at(&indegree, commit)) = 1;
763 /* also record the author dates, if needed */
764 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
765 record_author_date(&author_date, commit);
766 }
767
768 /* update the indegree */
769 for (next = orig; next; next = next->next) {
770 struct commit_list *parents = next->item->parents;
771 while (parents) {
772 struct commit *parent = parents->item;
773 int *pi = indegree_slab_at(&indegree, parent);
774
775 if (*pi)
776 (*pi)++;
777 parents = parents->next;
778 }
779 }
780
781 /*
782 * find the tips
783 *
784 * tips are nodes not reachable from any other node in the list
785 *
786 * the tips serve as a starting set for the work queue.
787 */
788 for (next = orig; next; next = next->next) {
789 struct commit *commit = next->item;
790
791 if (*(indegree_slab_at(&indegree, commit)) == 1)
792 prio_queue_put(&queue, commit);
793 }
794
795 /*
796 * This is unfortunate; the initial tips need to be shown
797 * in the order given from the revision traversal machinery.
798 */
799 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
800 prio_queue_reverse(&queue);
801
802 /* We no longer need the commit list */
803 free_commit_list(orig);
804
805 pptr = list;
806 *list = NULL;
807 while ((commit = prio_queue_get(&queue)) != NULL) {
808 struct commit_list *parents;
809
810 for (parents = commit->parents; parents ; parents = parents->next) {
811 struct commit *parent = parents->item;
812 int *pi = indegree_slab_at(&indegree, parent);
813
814 if (!*pi)
815 continue;
816
817 /*
818 * parents are only enqueued for emission
819 * when all their children have been emitted thereby
820 * guaranteeing topological order.
821 */
822 if (--(*pi) == 1)
823 prio_queue_put(&queue, parent);
824 }
825 /*
826 * all children of commit have already been
827 * emitted. we can emit it now.
828 */
829 *(indegree_slab_at(&indegree, commit)) = 0;
830
831 pptr = &commit_list_insert(commit, pptr)->next;
832 }
833
834 clear_indegree_slab(&indegree);
835 clear_prio_queue(&queue);
836 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
837 clear_author_date_slab(&author_date);
838 }
839
840 /* merge-base stuff */
841
842 /* Remember to update object flag allocation in object.h */
843 #define PARENT1 (1u<<16)
844 #define PARENT2 (1u<<17)
845 #define STALE (1u<<18)
846 #define RESULT (1u<<19)
847
848 static const unsigned all_flags = (PARENT1 | PARENT2 | STALE | RESULT);
849
850 static int queue_has_nonstale(struct prio_queue *queue)
851 {
852 int i;
853 for (i = 0; i < queue->nr; i++) {
854 struct commit *commit = queue->array[i].data;
855 if (!(commit->object.flags & STALE))
856 return 1;
857 }
858 return 0;
859 }
860
861 /* all input commits in one and twos[] must have been parsed! */
862 static struct commit_list *paint_down_to_common(struct commit *one, int n,
863 struct commit **twos,
864 int min_generation)
865 {
866 struct prio_queue queue = { compare_commits_by_gen_then_commit_date };
867 struct commit_list *result = NULL;
868 int i;
869 uint32_t last_gen = GENERATION_NUMBER_INFINITY;
870
871 one->object.flags |= PARENT1;
872 if (!n) {
873 commit_list_append(one, &result);
874 return result;
875 }
876 prio_queue_put(&queue, one);
877
878 for (i = 0; i < n; i++) {
879 twos[i]->object.flags |= PARENT2;
880 prio_queue_put(&queue, twos[i]);
881 }
882
883 while (queue_has_nonstale(&queue)) {
884 struct commit *commit = prio_queue_get(&queue);
885 struct commit_list *parents;
886 int flags;
887
888 if (commit->generation > last_gen)
889 BUG("bad generation skip %8x > %8x at %s",
890 commit->generation, last_gen,
891 oid_to_hex(&commit->object.oid));
892 last_gen = commit->generation;
893
894 if (commit->generation < min_generation)
895 break;
896
897 flags = commit->object.flags & (PARENT1 | PARENT2 | STALE);
898 if (flags == (PARENT1 | PARENT2)) {
899 if (!(commit->object.flags & RESULT)) {
900 commit->object.flags |= RESULT;
901 commit_list_insert_by_date(commit, &result);
902 }
903 /* Mark parents of a found merge stale */
904 flags |= STALE;
905 }
906 parents = commit->parents;
907 while (parents) {
908 struct commit *p = parents->item;
909 parents = parents->next;
910 if ((p->object.flags & flags) == flags)
911 continue;
912 if (parse_commit(p))
913 return NULL;
914 p->object.flags |= flags;
915 prio_queue_put(&queue, p);
916 }
917 }
918
919 clear_prio_queue(&queue);
920 return result;
921 }
922
923 static struct commit_list *merge_bases_many(struct commit *one, int n, struct commit **twos)
924 {
925 struct commit_list *list = NULL;
926 struct commit_list *result = NULL;
927 int i;
928
929 for (i = 0; i < n; i++) {
930 if (one == twos[i])
931 /*
932 * We do not mark this even with RESULT so we do not
933 * have to clean it up.
934 */
935 return commit_list_insert(one, &result);
936 }
937
938 if (parse_commit(one))
939 return NULL;
940 for (i = 0; i < n; i++) {
941 if (parse_commit(twos[i]))
942 return NULL;
943 }
944
945 list = paint_down_to_common(one, n, twos, 0);
946
947 while (list) {
948 struct commit *commit = pop_commit(&list);
949 if (!(commit->object.flags & STALE))
950 commit_list_insert_by_date(commit, &result);
951 }
952 return result;
953 }
954
955 struct commit_list *get_octopus_merge_bases(struct commit_list *in)
956 {
957 struct commit_list *i, *j, *k, *ret = NULL;
958
959 if (!in)
960 return ret;
961
962 commit_list_insert(in->item, &ret);
963
964 for (i = in->next; i; i = i->next) {
965 struct commit_list *new_commits = NULL, *end = NULL;
966
967 for (j = ret; j; j = j->next) {
968 struct commit_list *bases;
969 bases = get_merge_bases(i->item, j->item);
970 if (!new_commits)
971 new_commits = bases;
972 else
973 end->next = bases;
974 for (k = bases; k; k = k->next)
975 end = k;
976 }
977 ret = new_commits;
978 }
979 return ret;
980 }
981
982 static int remove_redundant(struct commit **array, int cnt)
983 {
984 /*
985 * Some commit in the array may be an ancestor of
986 * another commit. Move such commit to the end of
987 * the array, and return the number of commits that
988 * are independent from each other.
989 */
990 struct commit **work;
991 unsigned char *redundant;
992 int *filled_index;
993 int i, j, filled;
994
995 work = xcalloc(cnt, sizeof(*work));
996 redundant = xcalloc(cnt, 1);
997 ALLOC_ARRAY(filled_index, cnt - 1);
998
999 for (i = 0; i < cnt; i++)
1000 parse_commit(array[i]);
1001 for (i = 0; i < cnt; i++) {
1002 struct commit_list *common;
1003 uint32_t min_generation = array[i]->generation;
1004
1005 if (redundant[i])
1006 continue;
1007 for (j = filled = 0; j < cnt; j++) {
1008 if (i == j || redundant[j])
1009 continue;
1010 filled_index[filled] = j;
1011 work[filled++] = array[j];
1012
1013 if (array[j]->generation < min_generation)
1014 min_generation = array[j]->generation;
1015 }
1016 common = paint_down_to_common(array[i], filled, work,
1017 min_generation);
1018 if (array[i]->object.flags & PARENT2)
1019 redundant[i] = 1;
1020 for (j = 0; j < filled; j++)
1021 if (work[j]->object.flags & PARENT1)
1022 redundant[filled_index[j]] = 1;
1023 clear_commit_marks(array[i], all_flags);
1024 clear_commit_marks_many(filled, work, all_flags);
1025 free_commit_list(common);
1026 }
1027
1028 /* Now collect the result */
1029 COPY_ARRAY(work, array, cnt);
1030 for (i = filled = 0; i < cnt; i++)
1031 if (!redundant[i])
1032 array[filled++] = work[i];
1033 for (j = filled, i = 0; i < cnt; i++)
1034 if (redundant[i])
1035 array[j++] = work[i];
1036 free(work);
1037 free(redundant);
1038 free(filled_index);
1039 return filled;
1040 }
1041
1042 static struct commit_list *get_merge_bases_many_0(struct commit *one,
1043 int n,
1044 struct commit **twos,
1045 int cleanup)
1046 {
1047 struct commit_list *list;
1048 struct commit **rslt;
1049 struct commit_list *result;
1050 int cnt, i;
1051
1052 result = merge_bases_many(one, n, twos);
1053 for (i = 0; i < n; i++) {
1054 if (one == twos[i])
1055 return result;
1056 }
1057 if (!result || !result->next) {
1058 if (cleanup) {
1059 clear_commit_marks(one, all_flags);
1060 clear_commit_marks_many(n, twos, all_flags);
1061 }
1062 return result;
1063 }
1064
1065 /* There are more than one */
1066 cnt = commit_list_count(result);
1067 rslt = xcalloc(cnt, sizeof(*rslt));
1068 for (list = result, i = 0; list; list = list->next)
1069 rslt[i++] = list->item;
1070 free_commit_list(result);
1071
1072 clear_commit_marks(one, all_flags);
1073 clear_commit_marks_many(n, twos, all_flags);
1074
1075 cnt = remove_redundant(rslt, cnt);
1076 result = NULL;
1077 for (i = 0; i < cnt; i++)
1078 commit_list_insert_by_date(rslt[i], &result);
1079 free(rslt);
1080 return result;
1081 }
1082
1083 struct commit_list *get_merge_bases_many(struct commit *one,
1084 int n,
1085 struct commit **twos)
1086 {
1087 return get_merge_bases_many_0(one, n, twos, 1);
1088 }
1089
1090 struct commit_list *get_merge_bases_many_dirty(struct commit *one,
1091 int n,
1092 struct commit **twos)
1093 {
1094 return get_merge_bases_many_0(one, n, twos, 0);
1095 }
1096
1097 struct commit_list *get_merge_bases(struct commit *one, struct commit *two)
1098 {
1099 return get_merge_bases_many_0(one, 1, &two, 1);
1100 }
1101
1102 /*
1103 * Is "commit" a descendant of one of the elements on the "with_commit" list?
1104 */
1105 int is_descendant_of(struct commit *commit, struct commit_list *with_commit)
1106 {
1107 if (!with_commit)
1108 return 1;
1109 while (with_commit) {
1110 struct commit *other;
1111
1112 other = with_commit->item;
1113 with_commit = with_commit->next;
1114 if (in_merge_bases(other, commit))
1115 return 1;
1116 }
1117 return 0;
1118 }
1119
1120 /*
1121 * Is "commit" an ancestor of one of the "references"?
1122 */
1123 int in_merge_bases_many(struct commit *commit, int nr_reference, struct commit **reference)
1124 {
1125 struct commit_list *bases;
1126 int ret = 0, i;
1127 uint32_t min_generation = GENERATION_NUMBER_INFINITY;
1128
1129 if (parse_commit(commit))
1130 return ret;
1131 for (i = 0; i < nr_reference; i++) {
1132 if (parse_commit(reference[i]))
1133 return ret;
1134 if (reference[i]->generation < min_generation)
1135 min_generation = reference[i]->generation;
1136 }
1137
1138 if (commit->generation > min_generation)
1139 return ret;
1140
1141 bases = paint_down_to_common(commit, nr_reference, reference, commit->generation);
1142 if (commit->object.flags & PARENT2)
1143 ret = 1;
1144 clear_commit_marks(commit, all_flags);
1145 clear_commit_marks_many(nr_reference, reference, all_flags);
1146 free_commit_list(bases);
1147 return ret;
1148 }
1149
1150 /*
1151 * Is "commit" an ancestor of (i.e. reachable from) the "reference"?
1152 */
1153 int in_merge_bases(struct commit *commit, struct commit *reference)
1154 {
1155 return in_merge_bases_many(commit, 1, &reference);
1156 }
1157
1158 struct commit_list *reduce_heads(struct commit_list *heads)
1159 {
1160 struct commit_list *p;
1161 struct commit_list *result = NULL, **tail = &result;
1162 struct commit **array;
1163 int num_head, i;
1164
1165 if (!heads)
1166 return NULL;
1167
1168 /* Uniquify */
1169 for (p = heads; p; p = p->next)
1170 p->item->object.flags &= ~STALE;
1171 for (p = heads, num_head = 0; p; p = p->next) {
1172 if (p->item->object.flags & STALE)
1173 continue;
1174 p->item->object.flags |= STALE;
1175 num_head++;
1176 }
1177 array = xcalloc(num_head, sizeof(*array));
1178 for (p = heads, i = 0; p; p = p->next) {
1179 if (p->item->object.flags & STALE) {
1180 array[i++] = p->item;
1181 p->item->object.flags &= ~STALE;
1182 }
1183 }
1184 num_head = remove_redundant(array, num_head);
1185 for (i = 0; i < num_head; i++)
1186 tail = &commit_list_insert(array[i], tail)->next;
1187 free(array);
1188 return result;
1189 }
1190
1191 void reduce_heads_replace(struct commit_list **heads)
1192 {
1193 struct commit_list *result = reduce_heads(*heads);
1194 free_commit_list(*heads);
1195 *heads = result;
1196 }
1197
1198 static const char gpg_sig_header[] = "gpgsig";
1199 static const int gpg_sig_header_len = sizeof(gpg_sig_header) - 1;
1200
1201 static int do_sign_commit(struct strbuf *buf, const char *keyid)
1202 {
1203 struct strbuf sig = STRBUF_INIT;
1204 int inspos, copypos;
1205 const char *eoh;
1206
1207 /* find the end of the header */
1208 eoh = strstr(buf->buf, "\n\n");
1209 if (!eoh)
1210 inspos = buf->len;
1211 else
1212 inspos = eoh - buf->buf + 1;
1213
1214 if (!keyid || !*keyid)
1215 keyid = get_signing_key();
1216 if (sign_buffer(buf, &sig, keyid)) {
1217 strbuf_release(&sig);
1218 return -1;
1219 }
1220
1221 for (copypos = 0; sig.buf[copypos]; ) {
1222 const char *bol = sig.buf + copypos;
1223 const char *eol = strchrnul(bol, '\n');
1224 int len = (eol - bol) + !!*eol;
1225
1226 if (!copypos) {
1227 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
1228 inspos += gpg_sig_header_len;
1229 }
1230 strbuf_insert(buf, inspos++, " ", 1);
1231 strbuf_insert(buf, inspos, bol, len);
1232 inspos += len;
1233 copypos += len;
1234 }
1235 strbuf_release(&sig);
1236 return 0;
1237 }
1238
1239 int parse_signed_commit(const struct commit *commit,
1240 struct strbuf *payload, struct strbuf *signature)
1241 {
1242
1243 unsigned long size;
1244 const char *buffer = get_commit_buffer(commit, &size);
1245 int in_signature, saw_signature = -1;
1246 const char *line, *tail;
1247
1248 line = buffer;
1249 tail = buffer + size;
1250 in_signature = 0;
1251 saw_signature = 0;
1252 while (line < tail) {
1253 const char *sig = NULL;
1254 const char *next = memchr(line, '\n', tail - line);
1255
1256 next = next ? next + 1 : tail;
1257 if (in_signature && line[0] == ' ')
1258 sig = line + 1;
1259 else if (starts_with(line, gpg_sig_header) &&
1260 line[gpg_sig_header_len] == ' ')
1261 sig = line + gpg_sig_header_len + 1;
1262 if (sig) {
1263 strbuf_add(signature, sig, next - sig);
1264 saw_signature = 1;
1265 in_signature = 1;
1266 } else {
1267 if (*line == '\n')
1268 /* dump the whole remainder of the buffer */
1269 next = tail;
1270 strbuf_add(payload, line, next - line);
1271 in_signature = 0;
1272 }
1273 line = next;
1274 }
1275 unuse_commit_buffer(commit, buffer);
1276 return saw_signature;
1277 }
1278
1279 int remove_signature(struct strbuf *buf)
1280 {
1281 const char *line = buf->buf;
1282 const char *tail = buf->buf + buf->len;
1283 int in_signature = 0;
1284 const char *sig_start = NULL;
1285 const char *sig_end = NULL;
1286
1287 while (line < tail) {
1288 const char *next = memchr(line, '\n', tail - line);
1289 next = next ? next + 1 : tail;
1290
1291 if (in_signature && line[0] == ' ')
1292 sig_end = next;
1293 else if (starts_with(line, gpg_sig_header) &&
1294 line[gpg_sig_header_len] == ' ') {
1295 sig_start = line;
1296 sig_end = next;
1297 in_signature = 1;
1298 } else {
1299 if (*line == '\n')
1300 /* dump the whole remainder of the buffer */
1301 next = tail;
1302 in_signature = 0;
1303 }
1304 line = next;
1305 }
1306
1307 if (sig_start)
1308 strbuf_remove(buf, sig_start - buf->buf, sig_end - sig_start);
1309
1310 return sig_start != NULL;
1311 }
1312
1313 static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
1314 {
1315 struct merge_remote_desc *desc;
1316 struct commit_extra_header *mergetag;
1317 char *buf;
1318 unsigned long size, len;
1319 enum object_type type;
1320
1321 desc = merge_remote_util(parent);
1322 if (!desc || !desc->obj)
1323 return;
1324 buf = read_object_file(&desc->obj->oid, &type, &size);
1325 if (!buf || type != OBJ_TAG)
1326 goto free_return;
1327 len = parse_signature(buf, size);
1328 if (size == len)
1329 goto free_return;
1330 /*
1331 * We could verify this signature and either omit the tag when
1332 * it does not validate, but the integrator may not have the
1333 * public key of the signer of the tag he is merging, while a
1334 * later auditor may have it while auditing, so let's not run
1335 * verify-signed-buffer here for now...
1336 *
1337 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
1338 * warn("warning: signed tag unverified.");
1339 */
1340 mergetag = xcalloc(1, sizeof(*mergetag));
1341 mergetag->key = xstrdup("mergetag");
1342 mergetag->value = buf;
1343 mergetag->len = size;
1344
1345 **tail = mergetag;
1346 *tail = &mergetag->next;
1347 return;
1348
1349 free_return:
1350 free(buf);
1351 }
1352
1353 int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1354 {
1355 struct strbuf payload = STRBUF_INIT;
1356 struct strbuf signature = STRBUF_INIT;
1357 int ret = 1;
1358
1359 sigc->result = 'N';
1360
1361 if (parse_signed_commit(commit, &payload, &signature) <= 0)
1362 goto out;
1363 ret = check_signature(payload.buf, payload.len, signature.buf,
1364 signature.len, sigc);
1365
1366 out:
1367 strbuf_release(&payload);
1368 strbuf_release(&signature);
1369
1370 return ret;
1371 }
1372
1373
1374
1375 void append_merge_tag_headers(struct commit_list *parents,
1376 struct commit_extra_header ***tail)
1377 {
1378 while (parents) {
1379 struct commit *parent = parents->item;
1380 handle_signed_tag(parent, tail);
1381 parents = parents->next;
1382 }
1383 }
1384
1385 static void add_extra_header(struct strbuf *buffer,
1386 struct commit_extra_header *extra)
1387 {
1388 strbuf_addstr(buffer, extra->key);
1389 if (extra->len)
1390 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1391 else
1392 strbuf_addch(buffer, '\n');
1393 }
1394
1395 struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1396 const char **exclude)
1397 {
1398 struct commit_extra_header *extra = NULL;
1399 unsigned long size;
1400 const char *buffer = get_commit_buffer(commit, &size);
1401 extra = read_commit_extra_header_lines(buffer, size, exclude);
1402 unuse_commit_buffer(commit, buffer);
1403 return extra;
1404 }
1405
1406 int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1407 {
1408 struct commit_extra_header *extra, *to_free;
1409 int res = 0;
1410
1411 to_free = read_commit_extra_headers(commit, NULL);
1412 for (extra = to_free; !res && extra; extra = extra->next) {
1413 if (strcmp(extra->key, "mergetag"))
1414 continue; /* not a merge tag */
1415 res = fn(commit, extra, data);
1416 }
1417 free_commit_extra_headers(to_free);
1418 return res;
1419 }
1420
1421 static inline int standard_header_field(const char *field, size_t len)
1422 {
1423 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1424 (len == 6 && !memcmp(field, "parent", 6)) ||
1425 (len == 6 && !memcmp(field, "author", 6)) ||
1426 (len == 9 && !memcmp(field, "committer", 9)) ||
1427 (len == 8 && !memcmp(field, "encoding", 8)));
1428 }
1429
1430 static int excluded_header_field(const char *field, size_t len, const char **exclude)
1431 {
1432 if (!exclude)
1433 return 0;
1434
1435 while (*exclude) {
1436 size_t xlen = strlen(*exclude);
1437 if (len == xlen && !memcmp(field, *exclude, xlen))
1438 return 1;
1439 exclude++;
1440 }
1441 return 0;
1442 }
1443
1444 static struct commit_extra_header *read_commit_extra_header_lines(
1445 const char *buffer, size_t size,
1446 const char **exclude)
1447 {
1448 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1449 const char *line, *next, *eof, *eob;
1450 struct strbuf buf = STRBUF_INIT;
1451
1452 for (line = buffer, eob = line + size;
1453 line < eob && *line != '\n';
1454 line = next) {
1455 next = memchr(line, '\n', eob - line);
1456 next = next ? next + 1 : eob;
1457 if (*line == ' ') {
1458 /* continuation */
1459 if (it)
1460 strbuf_add(&buf, line + 1, next - (line + 1));
1461 continue;
1462 }
1463 if (it)
1464 it->value = strbuf_detach(&buf, &it->len);
1465 strbuf_reset(&buf);
1466 it = NULL;
1467
1468 eof = memchr(line, ' ', next - line);
1469 if (!eof)
1470 eof = next;
1471 else if (standard_header_field(line, eof - line) ||
1472 excluded_header_field(line, eof - line, exclude))
1473 continue;
1474
1475 it = xcalloc(1, sizeof(*it));
1476 it->key = xmemdupz(line, eof-line);
1477 *tail = it;
1478 tail = &it->next;
1479 if (eof + 1 < next)
1480 strbuf_add(&buf, eof + 1, next - (eof + 1));
1481 }
1482 if (it)
1483 it->value = strbuf_detach(&buf, &it->len);
1484 return extra;
1485 }
1486
1487 void free_commit_extra_headers(struct commit_extra_header *extra)
1488 {
1489 while (extra) {
1490 struct commit_extra_header *next = extra->next;
1491 free(extra->key);
1492 free(extra->value);
1493 free(extra);
1494 extra = next;
1495 }
1496 }
1497
1498 int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1499 struct commit_list *parents, struct object_id *ret,
1500 const char *author, const char *sign_commit)
1501 {
1502 struct commit_extra_header *extra = NULL, **tail = &extra;
1503 int result;
1504
1505 append_merge_tag_headers(parents, &tail);
1506 result = commit_tree_extended(msg, msg_len, tree, parents, ret,
1507 author, sign_commit, extra);
1508 free_commit_extra_headers(extra);
1509 return result;
1510 }
1511
1512 static int find_invalid_utf8(const char *buf, int len)
1513 {
1514 int offset = 0;
1515 static const unsigned int max_codepoint[] = {
1516 0x7f, 0x7ff, 0xffff, 0x10ffff
1517 };
1518
1519 while (len) {
1520 unsigned char c = *buf++;
1521 int bytes, bad_offset;
1522 unsigned int codepoint;
1523 unsigned int min_val, max_val;
1524
1525 len--;
1526 offset++;
1527
1528 /* Simple US-ASCII? No worries. */
1529 if (c < 0x80)
1530 continue;
1531
1532 bad_offset = offset-1;
1533
1534 /*
1535 * Count how many more high bits set: that's how
1536 * many more bytes this sequence should have.
1537 */
1538 bytes = 0;
1539 while (c & 0x40) {
1540 c <<= 1;
1541 bytes++;
1542 }
1543
1544 /*
1545 * Must be between 1 and 3 more bytes. Longer sequences result in
1546 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1547 */
1548 if (bytes < 1 || 3 < bytes)
1549 return bad_offset;
1550
1551 /* Do we *have* that many bytes? */
1552 if (len < bytes)
1553 return bad_offset;
1554
1555 /*
1556 * Place the encoded bits at the bottom of the value and compute the
1557 * valid range.
1558 */
1559 codepoint = (c & 0x7f) >> bytes;
1560 min_val = max_codepoint[bytes-1] + 1;
1561 max_val = max_codepoint[bytes];
1562
1563 offset += bytes;
1564 len -= bytes;
1565
1566 /* And verify that they are good continuation bytes */
1567 do {
1568 codepoint <<= 6;
1569 codepoint |= *buf & 0x3f;
1570 if ((*buf++ & 0xc0) != 0x80)
1571 return bad_offset;
1572 } while (--bytes);
1573
1574 /* Reject codepoints that are out of range for the sequence length. */
1575 if (codepoint < min_val || codepoint > max_val)
1576 return bad_offset;
1577 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1578 if ((codepoint & 0x1ff800) == 0xd800)
1579 return bad_offset;
1580 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1581 if ((codepoint & 0xfffe) == 0xfffe)
1582 return bad_offset;
1583 /* So are anything in the range U+FDD0..U+FDEF. */
1584 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1585 return bad_offset;
1586 }
1587 return -1;
1588 }
1589
1590 /*
1591 * This verifies that the buffer is in proper utf8 format.
1592 *
1593 * If it isn't, it assumes any non-utf8 characters are Latin1,
1594 * and does the conversion.
1595 */
1596 static int verify_utf8(struct strbuf *buf)
1597 {
1598 int ok = 1;
1599 long pos = 0;
1600
1601 for (;;) {
1602 int bad;
1603 unsigned char c;
1604 unsigned char replace[2];
1605
1606 bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1607 if (bad < 0)
1608 return ok;
1609 pos += bad;
1610 ok = 0;
1611 c = buf->buf[pos];
1612 strbuf_remove(buf, pos, 1);
1613
1614 /* We know 'c' must be in the range 128-255 */
1615 replace[0] = 0xc0 + (c >> 6);
1616 replace[1] = 0x80 + (c & 0x3f);
1617 strbuf_insert(buf, pos, replace, 2);
1618 pos += 2;
1619 }
1620 }
1621
1622 static const char commit_utf8_warn[] =
1623 N_("Warning: commit message did not conform to UTF-8.\n"
1624 "You may want to amend it after fixing the message, or set the config\n"
1625 "variable i18n.commitencoding to the encoding your project uses.\n");
1626
1627 int commit_tree_extended(const char *msg, size_t msg_len,
1628 const struct object_id *tree,
1629 struct commit_list *parents, struct object_id *ret,
1630 const char *author, const char *sign_commit,
1631 struct commit_extra_header *extra)
1632 {
1633 int result;
1634 int encoding_is_utf8;
1635 struct strbuf buffer;
1636
1637 assert_oid_type(tree, OBJ_TREE);
1638
1639 if (memchr(msg, '\0', msg_len))
1640 return error("a NUL byte in commit log message not allowed.");
1641
1642 /* Not having i18n.commitencoding is the same as having utf-8 */
1643 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1644
1645 strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1646 strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1647
1648 /*
1649 * NOTE! This ordering means that the same exact tree merged with a
1650 * different order of parents will be a _different_ changeset even
1651 * if everything else stays the same.
1652 */
1653 while (parents) {
1654 struct commit *parent = pop_commit(&parents);
1655 strbuf_addf(&buffer, "parent %s\n",
1656 oid_to_hex(&parent->object.oid));
1657 }
1658
1659 /* Person/date information */
1660 if (!author)
1661 author = git_author_info(IDENT_STRICT);
1662 strbuf_addf(&buffer, "author %s\n", author);
1663 strbuf_addf(&buffer, "committer %s\n", git_committer_info(IDENT_STRICT));
1664 if (!encoding_is_utf8)
1665 strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1666
1667 while (extra) {
1668 add_extra_header(&buffer, extra);
1669 extra = extra->next;
1670 }
1671 strbuf_addch(&buffer, '\n');
1672
1673 /* And add the comment */
1674 strbuf_add(&buffer, msg, msg_len);
1675
1676 /* And check the encoding */
1677 if (encoding_is_utf8 && !verify_utf8(&buffer))
1678 fprintf(stderr, _(commit_utf8_warn));
1679
1680 if (sign_commit && do_sign_commit(&buffer, sign_commit)) {
1681 result = -1;
1682 goto out;
1683 }
1684
1685 result = write_object_file(buffer.buf, buffer.len, commit_type, ret);
1686 out:
1687 strbuf_release(&buffer);
1688 return result;
1689 }
1690
1691 define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1692 static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1693
1694 struct merge_remote_desc *merge_remote_util(struct commit *commit)
1695 {
1696 return *merge_desc_slab_at(&merge_desc_slab, commit);
1697 }
1698
1699 void set_merge_remote_desc(struct commit *commit,
1700 const char *name, struct object *obj)
1701 {
1702 struct merge_remote_desc *desc;
1703 FLEX_ALLOC_STR(desc, name, name);
1704 desc->obj = obj;
1705 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1706 }
1707
1708 struct commit *get_merge_parent(const char *name)
1709 {
1710 struct object *obj;
1711 struct commit *commit;
1712 struct object_id oid;
1713 if (get_oid(name, &oid))
1714 return NULL;
1715 obj = parse_object(the_repository, &oid);
1716 commit = (struct commit *)peel_to_type(name, 0, obj, OBJ_COMMIT);
1717 if (commit && !merge_remote_util(commit))
1718 set_merge_remote_desc(commit, name, obj);
1719 return commit;
1720 }
1721
1722 /*
1723 * Append a commit to the end of the commit_list.
1724 *
1725 * next starts by pointing to the variable that holds the head of an
1726 * empty commit_list, and is updated to point to the "next" field of
1727 * the last item on the list as new commits are appended.
1728 *
1729 * Usage example:
1730 *
1731 * struct commit_list *list;
1732 * struct commit_list **next = &list;
1733 *
1734 * next = commit_list_append(c1, next);
1735 * next = commit_list_append(c2, next);
1736 * assert(commit_list_count(list) == 2);
1737 * return list;
1738 */
1739 struct commit_list **commit_list_append(struct commit *commit,
1740 struct commit_list **next)
1741 {
1742 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1743 new_commit->item = commit;
1744 *next = new_commit;
1745 new_commit->next = NULL;
1746 return &new_commit->next;
1747 }
1748
1749 const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1750 {
1751 int key_len = strlen(key);
1752 const char *line = msg;
1753
1754 while (line) {
1755 const char *eol = strchrnul(line, '\n');
1756
1757 if (line == eol)
1758 return NULL;
1759
1760 if (eol - line > key_len &&
1761 !strncmp(line, key, key_len) &&
1762 line[key_len] == ' ') {
1763 *out_len = eol - line - key_len - 1;
1764 return line + key_len + 1;
1765 }
1766 line = *eol ? eol + 1 : NULL;
1767 }
1768 return NULL;
1769 }
1770
1771 /*
1772 * Inspect the given string and determine the true "end" of the log message, in
1773 * order to find where to put a new Signed-off-by: line. Ignored are
1774 * trailing comment lines and blank lines. To support "git commit -s
1775 * --amend" on an existing commit, we also ignore "Conflicts:". To
1776 * support "git commit -v", we truncate at cut lines.
1777 *
1778 * Returns the number of bytes from the tail to ignore, to be fed as
1779 * the second parameter to append_signoff().
1780 */
1781 int ignore_non_trailer(const char *buf, size_t len)
1782 {
1783 int boc = 0;
1784 int bol = 0;
1785 int in_old_conflicts_block = 0;
1786 size_t cutoff = wt_status_locate_end(buf, len);
1787
1788 while (bol < cutoff) {
1789 const char *next_line = memchr(buf + bol, '\n', len - bol);
1790
1791 if (!next_line)
1792 next_line = buf + len;
1793 else
1794 next_line++;
1795
1796 if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1797 /* is this the first of the run of comments? */
1798 if (!boc)
1799 boc = bol;
1800 /* otherwise, it is just continuing */
1801 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1802 in_old_conflicts_block = 1;
1803 if (!boc)
1804 boc = bol;
1805 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1806 ; /* a pathname in the conflicts block */
1807 } else if (boc) {
1808 /* the previous was not trailing comment */
1809 boc = 0;
1810 in_old_conflicts_block = 0;
1811 }
1812 bol = next_line - buf;
1813 }
1814 return boc ? len - boc : len - cutoff;
1815 }