]> git.ipfire.org Git - thirdparty/git.git/blob - commit.c
Merge branch 'en/header-cleanup'
[thirdparty/git.git] / commit.c
1 #include "cache.h"
2 #include "tag.h"
3 #include "commit.h"
4 #include "commit-graph.h"
5 #include "hex.h"
6 #include "repository.h"
7 #include "object-store.h"
8 #include "pkt-line.h"
9 #include "utf8.h"
10 #include "diff.h"
11 #include "revision.h"
12 #include "notes.h"
13 #include "alloc.h"
14 #include "gpg-interface.h"
15 #include "mergesort.h"
16 #include "commit-slab.h"
17 #include "prio-queue.h"
18 #include "hash-lookup.h"
19 #include "wt-status.h"
20 #include "advice.h"
21 #include "refs.h"
22 #include "commit-reach.h"
23 #include "run-command.h"
24 #include "shallow.h"
25 #include "hook.h"
26
27 static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
28
29 int save_commit_buffer = 1;
30 int no_graft_file_deprecated_advice;
31
32 const char *commit_type = "commit";
33
34 struct commit *lookup_commit_reference_gently(struct repository *r,
35 const struct object_id *oid, int quiet)
36 {
37 struct object *obj = deref_tag(r,
38 parse_object(r, oid),
39 NULL, 0);
40
41 if (!obj)
42 return NULL;
43 return object_as_type(obj, OBJ_COMMIT, quiet);
44 }
45
46 struct commit *lookup_commit_reference(struct repository *r, const struct object_id *oid)
47 {
48 return lookup_commit_reference_gently(r, oid, 0);
49 }
50
51 struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
52 {
53 struct commit *c = lookup_commit_reference(the_repository, oid);
54 if (!c)
55 die(_("could not parse %s"), ref_name);
56 if (!oideq(oid, &c->object.oid)) {
57 warning(_("%s %s is not a commit!"),
58 ref_name, oid_to_hex(oid));
59 }
60 return c;
61 }
62
63 struct commit *lookup_commit_object(struct repository *r,
64 const struct object_id *oid)
65 {
66 struct object *obj = parse_object(r, oid);
67 return obj ? object_as_type(obj, OBJ_COMMIT, 0) : NULL;
68
69 }
70
71 struct commit *lookup_commit(struct repository *r, const struct object_id *oid)
72 {
73 struct object *obj = lookup_object(r, oid);
74 if (!obj)
75 return create_object(r, oid, alloc_commit_node(r));
76 return object_as_type(obj, OBJ_COMMIT, 0);
77 }
78
79 struct commit *lookup_commit_reference_by_name(const char *name)
80 {
81 struct object_id oid;
82 struct commit *commit;
83
84 if (get_oid_committish(name, &oid))
85 return NULL;
86 commit = lookup_commit_reference(the_repository, &oid);
87 if (parse_commit(commit))
88 return NULL;
89 return commit;
90 }
91
92 static timestamp_t parse_commit_date(const char *buf, const char *tail)
93 {
94 const char *dateptr;
95
96 if (buf + 6 >= tail)
97 return 0;
98 if (memcmp(buf, "author", 6))
99 return 0;
100 while (buf < tail && *buf++ != '\n')
101 /* nada */;
102 if (buf + 9 >= tail)
103 return 0;
104 if (memcmp(buf, "committer", 9))
105 return 0;
106 while (buf < tail && *buf++ != '>')
107 /* nada */;
108 if (buf >= tail)
109 return 0;
110 dateptr = buf;
111 while (buf < tail && *buf++ != '\n')
112 /* nada */;
113 if (buf >= tail)
114 return 0;
115 /* dateptr < buf && buf[-1] == '\n', so parsing will stop at buf-1 */
116 return parse_timestamp(dateptr, NULL, 10);
117 }
118
119 static const struct object_id *commit_graft_oid_access(size_t index, const void *table)
120 {
121 const struct commit_graft * const *commit_graft_table = table;
122 return &commit_graft_table[index]->oid;
123 }
124
125 int commit_graft_pos(struct repository *r, const struct object_id *oid)
126 {
127 return oid_pos(oid, r->parsed_objects->grafts,
128 r->parsed_objects->grafts_nr,
129 commit_graft_oid_access);
130 }
131
132 static void unparse_commit(struct repository *r, const struct object_id *oid)
133 {
134 struct commit *c = lookup_commit(r, oid);
135
136 if (!c->object.parsed)
137 return;
138 free_commit_list(c->parents);
139 c->parents = NULL;
140 c->object.parsed = 0;
141 }
142
143 int register_commit_graft(struct repository *r, struct commit_graft *graft,
144 int ignore_dups)
145 {
146 int pos = commit_graft_pos(r, &graft->oid);
147
148 if (0 <= pos) {
149 if (ignore_dups)
150 free(graft);
151 else {
152 free(r->parsed_objects->grafts[pos]);
153 r->parsed_objects->grafts[pos] = graft;
154 }
155 return 1;
156 }
157 pos = -pos - 1;
158 ALLOC_GROW(r->parsed_objects->grafts,
159 r->parsed_objects->grafts_nr + 1,
160 r->parsed_objects->grafts_alloc);
161 r->parsed_objects->grafts_nr++;
162 if (pos < r->parsed_objects->grafts_nr)
163 memmove(r->parsed_objects->grafts + pos + 1,
164 r->parsed_objects->grafts + pos,
165 (r->parsed_objects->grafts_nr - pos - 1) *
166 sizeof(*r->parsed_objects->grafts));
167 r->parsed_objects->grafts[pos] = graft;
168 unparse_commit(r, &graft->oid);
169 return 0;
170 }
171
172 struct commit_graft *read_graft_line(struct strbuf *line)
173 {
174 /* The format is just "Commit Parent1 Parent2 ...\n" */
175 int i, phase;
176 const char *tail = NULL;
177 struct commit_graft *graft = NULL;
178 struct object_id dummy_oid, *oid;
179
180 strbuf_rtrim(line);
181 if (!line->len || line->buf[0] == '#')
182 return NULL;
183 /*
184 * phase 0 verifies line, counts hashes in line and allocates graft
185 * phase 1 fills graft
186 */
187 for (phase = 0; phase < 2; phase++) {
188 oid = graft ? &graft->oid : &dummy_oid;
189 if (parse_oid_hex(line->buf, oid, &tail))
190 goto bad_graft_data;
191 for (i = 0; *tail != '\0'; i++) {
192 oid = graft ? &graft->parent[i] : &dummy_oid;
193 if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail))
194 goto bad_graft_data;
195 }
196 if (!graft) {
197 graft = xmalloc(st_add(sizeof(*graft),
198 st_mult(sizeof(struct object_id), i)));
199 graft->nr_parent = i;
200 }
201 }
202 return graft;
203
204 bad_graft_data:
205 error("bad graft data: %s", line->buf);
206 assert(!graft);
207 return NULL;
208 }
209
210 static int read_graft_file(struct repository *r, const char *graft_file)
211 {
212 FILE *fp = fopen_or_warn(graft_file, "r");
213 struct strbuf buf = STRBUF_INIT;
214 if (!fp)
215 return -1;
216 if (!no_graft_file_deprecated_advice &&
217 advice_enabled(ADVICE_GRAFT_FILE_DEPRECATED))
218 advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n"
219 "and will be removed in a future Git version.\n"
220 "\n"
221 "Please use \"git replace --convert-graft-file\"\n"
222 "to convert the grafts into replace refs.\n"
223 "\n"
224 "Turn this message off by running\n"
225 "\"git config advice.graftFileDeprecated false\""));
226 while (!strbuf_getwholeline(&buf, fp, '\n')) {
227 /* The format is just "Commit Parent1 Parent2 ...\n" */
228 struct commit_graft *graft = read_graft_line(&buf);
229 if (!graft)
230 continue;
231 if (register_commit_graft(r, graft, 1))
232 error("duplicate graft data: %s", buf.buf);
233 }
234 fclose(fp);
235 strbuf_release(&buf);
236 return 0;
237 }
238
239 void prepare_commit_graft(struct repository *r)
240 {
241 char *graft_file;
242
243 if (r->parsed_objects->commit_graft_prepared)
244 return;
245 if (!startup_info->have_repository)
246 return;
247
248 graft_file = get_graft_file(r);
249 read_graft_file(r, graft_file);
250 /* make sure shallows are read */
251 is_repository_shallow(r);
252 r->parsed_objects->commit_graft_prepared = 1;
253 }
254
255 struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid)
256 {
257 int pos;
258 prepare_commit_graft(r);
259 pos = commit_graft_pos(r, oid);
260 if (pos < 0)
261 return NULL;
262 return r->parsed_objects->grafts[pos];
263 }
264
265 int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
266 {
267 int i, ret;
268 for (i = ret = 0; i < the_repository->parsed_objects->grafts_nr && !ret; i++)
269 ret = fn(the_repository->parsed_objects->grafts[i], cb_data);
270 return ret;
271 }
272
273 void reset_commit_grafts(struct repository *r)
274 {
275 int i;
276
277 for (i = 0; i < r->parsed_objects->grafts_nr; i++) {
278 unparse_commit(r, &r->parsed_objects->grafts[i]->oid);
279 free(r->parsed_objects->grafts[i]);
280 }
281 r->parsed_objects->grafts_nr = 0;
282 r->parsed_objects->commit_graft_prepared = 0;
283 }
284
285 struct commit_buffer {
286 void *buffer;
287 unsigned long size;
288 };
289 define_commit_slab(buffer_slab, struct commit_buffer);
290
291 struct buffer_slab *allocate_commit_buffer_slab(void)
292 {
293 struct buffer_slab *bs = xmalloc(sizeof(*bs));
294 init_buffer_slab(bs);
295 return bs;
296 }
297
298 void free_commit_buffer_slab(struct buffer_slab *bs)
299 {
300 clear_buffer_slab(bs);
301 free(bs);
302 }
303
304 void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size)
305 {
306 struct commit_buffer *v = buffer_slab_at(
307 r->parsed_objects->buffer_slab, commit);
308 v->buffer = buffer;
309 v->size = size;
310 }
311
312 const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep)
313 {
314 struct commit_buffer *v = buffer_slab_peek(
315 r->parsed_objects->buffer_slab, commit);
316 if (!v) {
317 if (sizep)
318 *sizep = 0;
319 return NULL;
320 }
321 if (sizep)
322 *sizep = v->size;
323 return v->buffer;
324 }
325
326 const void *repo_get_commit_buffer(struct repository *r,
327 const struct commit *commit,
328 unsigned long *sizep)
329 {
330 const void *ret = get_cached_commit_buffer(r, commit, sizep);
331 if (!ret) {
332 enum object_type type;
333 unsigned long size;
334 ret = repo_read_object_file(r, &commit->object.oid, &type, &size);
335 if (!ret)
336 die("cannot read commit object %s",
337 oid_to_hex(&commit->object.oid));
338 if (type != OBJ_COMMIT)
339 die("expected commit for %s, got %s",
340 oid_to_hex(&commit->object.oid), type_name(type));
341 if (sizep)
342 *sizep = size;
343 }
344 return ret;
345 }
346
347 void repo_unuse_commit_buffer(struct repository *r,
348 const struct commit *commit,
349 const void *buffer)
350 {
351 struct commit_buffer *v = buffer_slab_peek(
352 r->parsed_objects->buffer_slab, commit);
353 if (!(v && v->buffer == buffer))
354 free((void *)buffer);
355 }
356
357 void free_commit_buffer(struct parsed_object_pool *pool, struct commit *commit)
358 {
359 struct commit_buffer *v = buffer_slab_peek(
360 pool->buffer_slab, commit);
361 if (v) {
362 FREE_AND_NULL(v->buffer);
363 v->size = 0;
364 }
365 }
366
367 static inline void set_commit_tree(struct commit *c, struct tree *t)
368 {
369 c->maybe_tree = t;
370 }
371
372 struct tree *repo_get_commit_tree(struct repository *r,
373 const struct commit *commit)
374 {
375 if (commit->maybe_tree || !commit->object.parsed)
376 return commit->maybe_tree;
377
378 if (commit_graph_position(commit) != COMMIT_NOT_FROM_GRAPH)
379 return get_commit_tree_in_graph(r, commit);
380
381 return NULL;
382 }
383
384 struct object_id *get_commit_tree_oid(const struct commit *commit)
385 {
386 struct tree *tree = get_commit_tree(commit);
387 return tree ? &tree->object.oid : NULL;
388 }
389
390 void release_commit_memory(struct parsed_object_pool *pool, struct commit *c)
391 {
392 set_commit_tree(c, NULL);
393 free_commit_buffer(pool, c);
394 c->index = 0;
395 free_commit_list(c->parents);
396
397 c->object.parsed = 0;
398 }
399
400 const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
401 {
402 struct commit_buffer *v = buffer_slab_peek(
403 the_repository->parsed_objects->buffer_slab, commit);
404 void *ret;
405
406 if (!v) {
407 if (sizep)
408 *sizep = 0;
409 return NULL;
410 }
411 ret = v->buffer;
412 if (sizep)
413 *sizep = v->size;
414
415 v->buffer = NULL;
416 v->size = 0;
417 return ret;
418 }
419
420 int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
421 {
422 const char *tail = buffer;
423 const char *bufptr = buffer;
424 struct object_id parent;
425 struct commit_list **pptr;
426 struct commit_graft *graft;
427 const int tree_entry_len = the_hash_algo->hexsz + 5;
428 const int parent_entry_len = the_hash_algo->hexsz + 7;
429 struct tree *tree;
430
431 if (item->object.parsed)
432 return 0;
433 /*
434 * Presumably this is leftover from an earlier failed parse;
435 * clear it out in preparation for us re-parsing (we'll hit the
436 * same error, but that's good, since it lets our caller know
437 * the result cannot be trusted.
438 */
439 free_commit_list(item->parents);
440 item->parents = NULL;
441
442 tail += size;
443 if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
444 bufptr[tree_entry_len] != '\n')
445 return error("bogus commit object %s", oid_to_hex(&item->object.oid));
446 if (get_oid_hex(bufptr + 5, &parent) < 0)
447 return error("bad tree pointer in commit %s",
448 oid_to_hex(&item->object.oid));
449 tree = lookup_tree(r, &parent);
450 if (!tree)
451 return error("bad tree pointer %s in commit %s",
452 oid_to_hex(&parent),
453 oid_to_hex(&item->object.oid));
454 set_commit_tree(item, tree);
455 bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
456 pptr = &item->parents;
457
458 graft = lookup_commit_graft(r, &item->object.oid);
459 if (graft)
460 r->parsed_objects->substituted_parent = 1;
461 while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
462 struct commit *new_parent;
463
464 if (tail <= bufptr + parent_entry_len + 1 ||
465 get_oid_hex(bufptr + 7, &parent) ||
466 bufptr[parent_entry_len] != '\n')
467 return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
468 bufptr += parent_entry_len + 1;
469 /*
470 * The clone is shallow if nr_parent < 0, and we must
471 * not traverse its real parents even when we unhide them.
472 */
473 if (graft && (graft->nr_parent < 0 || grafts_replace_parents))
474 continue;
475 new_parent = lookup_commit(r, &parent);
476 if (!new_parent)
477 return error("bad parent %s in commit %s",
478 oid_to_hex(&parent),
479 oid_to_hex(&item->object.oid));
480 pptr = &commit_list_insert(new_parent, pptr)->next;
481 }
482 if (graft) {
483 int i;
484 struct commit *new_parent;
485 for (i = 0; i < graft->nr_parent; i++) {
486 new_parent = lookup_commit(r,
487 &graft->parent[i]);
488 if (!new_parent)
489 return error("bad graft parent %s in commit %s",
490 oid_to_hex(&graft->parent[i]),
491 oid_to_hex(&item->object.oid));
492 pptr = &commit_list_insert(new_parent, pptr)->next;
493 }
494 }
495 item->date = parse_commit_date(bufptr, tail);
496
497 if (check_graph)
498 load_commit_graph_info(r, item);
499
500 item->object.parsed = 1;
501 return 0;
502 }
503
504 int repo_parse_commit_internal(struct repository *r,
505 struct commit *item,
506 int quiet_on_missing,
507 int use_commit_graph)
508 {
509 enum object_type type;
510 void *buffer;
511 unsigned long size;
512 struct object_info oi = {
513 .typep = &type,
514 .sizep = &size,
515 .contentp = &buffer,
516 };
517 /*
518 * Git does not support partial clones that exclude commits, so set
519 * OBJECT_INFO_SKIP_FETCH_OBJECT to fail fast when an object is missing.
520 */
521 int flags = OBJECT_INFO_LOOKUP_REPLACE | OBJECT_INFO_SKIP_FETCH_OBJECT |
522 OBJECT_INFO_DIE_IF_CORRUPT;
523 int ret;
524
525 if (!item)
526 return -1;
527 if (item->object.parsed)
528 return 0;
529 if (use_commit_graph && parse_commit_in_graph(r, item))
530 return 0;
531
532 if (oid_object_info_extended(r, &item->object.oid, &oi, flags) < 0)
533 return quiet_on_missing ? -1 :
534 error("Could not read %s",
535 oid_to_hex(&item->object.oid));
536 if (type != OBJ_COMMIT) {
537 free(buffer);
538 return error("Object %s not a commit",
539 oid_to_hex(&item->object.oid));
540 }
541
542 ret = parse_commit_buffer(r, item, buffer, size, 0);
543 if (save_commit_buffer && !ret) {
544 set_commit_buffer(r, item, buffer, size);
545 return 0;
546 }
547 free(buffer);
548 return ret;
549 }
550
551 int repo_parse_commit_gently(struct repository *r,
552 struct commit *item, int quiet_on_missing)
553 {
554 return repo_parse_commit_internal(r, item, quiet_on_missing, 1);
555 }
556
557 void parse_commit_or_die(struct commit *item)
558 {
559 if (parse_commit(item))
560 die("unable to parse commit %s",
561 item ? oid_to_hex(&item->object.oid) : "(null)");
562 }
563
564 int find_commit_subject(const char *commit_buffer, const char **subject)
565 {
566 const char *eol;
567 const char *p = commit_buffer;
568
569 while (*p && (*p != '\n' || p[1] != '\n'))
570 p++;
571 if (*p) {
572 p = skip_blank_lines(p + 2);
573 eol = strchrnul(p, '\n');
574 } else
575 eol = p;
576
577 *subject = p;
578
579 return eol - p;
580 }
581
582 size_t commit_subject_length(const char *body)
583 {
584 const char *p = body;
585 while (*p) {
586 const char *next = skip_blank_lines(p);
587 if (next != p)
588 break;
589 p = strchrnul(p, '\n');
590 if (*p)
591 p++;
592 }
593 return p - body;
594 }
595
596 struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
597 {
598 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
599 new_list->item = item;
600 new_list->next = *list_p;
601 *list_p = new_list;
602 return new_list;
603 }
604
605 int commit_list_contains(struct commit *item, struct commit_list *list)
606 {
607 while (list) {
608 if (list->item == item)
609 return 1;
610 list = list->next;
611 }
612
613 return 0;
614 }
615
616 unsigned commit_list_count(const struct commit_list *l)
617 {
618 unsigned c = 0;
619 for (; l; l = l->next )
620 c++;
621 return c;
622 }
623
624 struct commit_list *copy_commit_list(struct commit_list *list)
625 {
626 struct commit_list *head = NULL;
627 struct commit_list **pp = &head;
628 while (list) {
629 pp = commit_list_append(list->item, pp);
630 list = list->next;
631 }
632 return head;
633 }
634
635 struct commit_list *reverse_commit_list(struct commit_list *list)
636 {
637 struct commit_list *next = NULL, *current, *backup;
638 for (current = list; current; current = backup) {
639 backup = current->next;
640 current->next = next;
641 next = current;
642 }
643 return next;
644 }
645
646 void free_commit_list(struct commit_list *list)
647 {
648 while (list)
649 pop_commit(&list);
650 }
651
652 struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
653 {
654 struct commit_list **pp = list;
655 struct commit_list *p;
656 while ((p = *pp) != NULL) {
657 if (p->item->date < item->date) {
658 break;
659 }
660 pp = &p->next;
661 }
662 return commit_list_insert(item, pp);
663 }
664
665 static int commit_list_compare_by_date(const struct commit_list *a,
666 const struct commit_list *b)
667 {
668 timestamp_t a_date = a->item->date;
669 timestamp_t b_date = b->item->date;
670 if (a_date < b_date)
671 return 1;
672 if (a_date > b_date)
673 return -1;
674 return 0;
675 }
676
677 DEFINE_LIST_SORT(static, commit_list_sort, struct commit_list, next);
678
679 void commit_list_sort_by_date(struct commit_list **list)
680 {
681 commit_list_sort(list, commit_list_compare_by_date);
682 }
683
684 struct commit *pop_most_recent_commit(struct commit_list **list,
685 unsigned int mark)
686 {
687 struct commit *ret = pop_commit(list);
688 struct commit_list *parents = ret->parents;
689
690 while (parents) {
691 struct commit *commit = parents->item;
692 if (!parse_commit(commit) && !(commit->object.flags & mark)) {
693 commit->object.flags |= mark;
694 commit_list_insert_by_date(commit, list);
695 }
696 parents = parents->next;
697 }
698 return ret;
699 }
700
701 static void clear_commit_marks_1(struct commit_list **plist,
702 struct commit *commit, unsigned int mark)
703 {
704 while (commit) {
705 struct commit_list *parents;
706
707 if (!(mark & commit->object.flags))
708 return;
709
710 commit->object.flags &= ~mark;
711
712 parents = commit->parents;
713 if (!parents)
714 return;
715
716 while ((parents = parents->next)) {
717 if (parents->item->object.flags & mark)
718 commit_list_insert(parents->item, plist);
719 }
720
721 commit = commit->parents->item;
722 }
723 }
724
725 void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
726 {
727 struct commit_list *list = NULL;
728
729 while (nr--) {
730 clear_commit_marks_1(&list, *commit, mark);
731 commit++;
732 }
733 while (list)
734 clear_commit_marks_1(&list, pop_commit(&list), mark);
735 }
736
737 void clear_commit_marks(struct commit *commit, unsigned int mark)
738 {
739 clear_commit_marks_many(1, &commit, mark);
740 }
741
742 struct commit *pop_commit(struct commit_list **stack)
743 {
744 struct commit_list *top = *stack;
745 struct commit *item = top ? top->item : NULL;
746
747 if (top) {
748 *stack = top->next;
749 free(top);
750 }
751 return item;
752 }
753
754 /*
755 * Topological sort support
756 */
757
758 /* count number of children that have not been emitted */
759 define_commit_slab(indegree_slab, int);
760
761 define_commit_slab(author_date_slab, timestamp_t);
762
763 void record_author_date(struct author_date_slab *author_date,
764 struct commit *commit)
765 {
766 const char *buffer = get_commit_buffer(commit, NULL);
767 struct ident_split ident;
768 const char *ident_line;
769 size_t ident_len;
770 char *date_end;
771 timestamp_t date;
772
773 ident_line = find_commit_header(buffer, "author", &ident_len);
774 if (!ident_line)
775 goto fail_exit; /* no author line */
776 if (split_ident_line(&ident, ident_line, ident_len) ||
777 !ident.date_begin || !ident.date_end)
778 goto fail_exit; /* malformed "author" line */
779
780 date = parse_timestamp(ident.date_begin, &date_end, 10);
781 if (date_end != ident.date_end)
782 goto fail_exit; /* malformed date */
783 *(author_date_slab_at(author_date, commit)) = date;
784
785 fail_exit:
786 unuse_commit_buffer(commit, buffer);
787 }
788
789 int compare_commits_by_author_date(const void *a_, const void *b_,
790 void *cb_data)
791 {
792 const struct commit *a = a_, *b = b_;
793 struct author_date_slab *author_date = cb_data;
794 timestamp_t a_date = *(author_date_slab_at(author_date, a));
795 timestamp_t b_date = *(author_date_slab_at(author_date, b));
796
797 /* newer commits with larger date first */
798 if (a_date < b_date)
799 return 1;
800 else if (a_date > b_date)
801 return -1;
802 return 0;
803 }
804
805 int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, void *unused)
806 {
807 const struct commit *a = a_, *b = b_;
808 const timestamp_t generation_a = commit_graph_generation(a),
809 generation_b = commit_graph_generation(b);
810
811 /* newer commits first */
812 if (generation_a < generation_b)
813 return 1;
814 else if (generation_a > generation_b)
815 return -1;
816
817 /* use date as a heuristic when generations are equal */
818 if (a->date < b->date)
819 return 1;
820 else if (a->date > b->date)
821 return -1;
822 return 0;
823 }
824
825 int compare_commits_by_commit_date(const void *a_, const void *b_, void *unused)
826 {
827 const struct commit *a = a_, *b = b_;
828 /* newer commits with larger date first */
829 if (a->date < b->date)
830 return 1;
831 else if (a->date > b->date)
832 return -1;
833 return 0;
834 }
835
836 /*
837 * Performs an in-place topological sort on the list supplied.
838 */
839 void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
840 {
841 struct commit_list *next, *orig = *list;
842 struct commit_list **pptr;
843 struct indegree_slab indegree;
844 struct prio_queue queue;
845 struct commit *commit;
846 struct author_date_slab author_date;
847
848 if (!orig)
849 return;
850 *list = NULL;
851
852 init_indegree_slab(&indegree);
853 memset(&queue, '\0', sizeof(queue));
854
855 switch (sort_order) {
856 default: /* REV_SORT_IN_GRAPH_ORDER */
857 queue.compare = NULL;
858 break;
859 case REV_SORT_BY_COMMIT_DATE:
860 queue.compare = compare_commits_by_commit_date;
861 break;
862 case REV_SORT_BY_AUTHOR_DATE:
863 init_author_date_slab(&author_date);
864 queue.compare = compare_commits_by_author_date;
865 queue.cb_data = &author_date;
866 break;
867 }
868
869 /* Mark them and clear the indegree */
870 for (next = orig; next; next = next->next) {
871 struct commit *commit = next->item;
872 *(indegree_slab_at(&indegree, commit)) = 1;
873 /* also record the author dates, if needed */
874 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
875 record_author_date(&author_date, commit);
876 }
877
878 /* update the indegree */
879 for (next = orig; next; next = next->next) {
880 struct commit_list *parents = next->item->parents;
881 while (parents) {
882 struct commit *parent = parents->item;
883 int *pi = indegree_slab_at(&indegree, parent);
884
885 if (*pi)
886 (*pi)++;
887 parents = parents->next;
888 }
889 }
890
891 /*
892 * find the tips
893 *
894 * tips are nodes not reachable from any other node in the list
895 *
896 * the tips serve as a starting set for the work queue.
897 */
898 for (next = orig; next; next = next->next) {
899 struct commit *commit = next->item;
900
901 if (*(indegree_slab_at(&indegree, commit)) == 1)
902 prio_queue_put(&queue, commit);
903 }
904
905 /*
906 * This is unfortunate; the initial tips need to be shown
907 * in the order given from the revision traversal machinery.
908 */
909 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
910 prio_queue_reverse(&queue);
911
912 /* We no longer need the commit list */
913 free_commit_list(orig);
914
915 pptr = list;
916 *list = NULL;
917 while ((commit = prio_queue_get(&queue)) != NULL) {
918 struct commit_list *parents;
919
920 for (parents = commit->parents; parents ; parents = parents->next) {
921 struct commit *parent = parents->item;
922 int *pi = indegree_slab_at(&indegree, parent);
923
924 if (!*pi)
925 continue;
926
927 /*
928 * parents are only enqueued for emission
929 * when all their children have been emitted thereby
930 * guaranteeing topological order.
931 */
932 if (--(*pi) == 1)
933 prio_queue_put(&queue, parent);
934 }
935 /*
936 * all children of commit have already been
937 * emitted. we can emit it now.
938 */
939 *(indegree_slab_at(&indegree, commit)) = 0;
940
941 pptr = &commit_list_insert(commit, pptr)->next;
942 }
943
944 clear_indegree_slab(&indegree);
945 clear_prio_queue(&queue);
946 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
947 clear_author_date_slab(&author_date);
948 }
949
950 struct rev_collect {
951 struct commit **commit;
952 int nr;
953 int alloc;
954 unsigned int initial : 1;
955 };
956
957 static void add_one_commit(struct object_id *oid, struct rev_collect *revs)
958 {
959 struct commit *commit;
960
961 if (is_null_oid(oid))
962 return;
963
964 commit = lookup_commit(the_repository, oid);
965 if (!commit ||
966 (commit->object.flags & TMP_MARK) ||
967 parse_commit(commit))
968 return;
969
970 ALLOC_GROW(revs->commit, revs->nr + 1, revs->alloc);
971 revs->commit[revs->nr++] = commit;
972 commit->object.flags |= TMP_MARK;
973 }
974
975 static int collect_one_reflog_ent(struct object_id *ooid, struct object_id *noid,
976 const char *ident UNUSED,
977 timestamp_t timestamp UNUSED, int tz UNUSED,
978 const char *message UNUSED, void *cbdata)
979 {
980 struct rev_collect *revs = cbdata;
981
982 if (revs->initial) {
983 revs->initial = 0;
984 add_one_commit(ooid, revs);
985 }
986 add_one_commit(noid, revs);
987 return 0;
988 }
989
990 struct commit *get_fork_point(const char *refname, struct commit *commit)
991 {
992 struct object_id oid;
993 struct rev_collect revs;
994 struct commit_list *bases;
995 int i;
996 struct commit *ret = NULL;
997 char *full_refname;
998
999 switch (dwim_ref(refname, strlen(refname), &oid, &full_refname, 0)) {
1000 case 0:
1001 die("No such ref: '%s'", refname);
1002 case 1:
1003 break; /* good */
1004 default:
1005 die("Ambiguous refname: '%s'", refname);
1006 }
1007
1008 memset(&revs, 0, sizeof(revs));
1009 revs.initial = 1;
1010 for_each_reflog_ent(full_refname, collect_one_reflog_ent, &revs);
1011
1012 if (!revs.nr)
1013 add_one_commit(&oid, &revs);
1014
1015 for (i = 0; i < revs.nr; i++)
1016 revs.commit[i]->object.flags &= ~TMP_MARK;
1017
1018 bases = get_merge_bases_many(commit, revs.nr, revs.commit);
1019
1020 /*
1021 * There should be one and only one merge base, when we found
1022 * a common ancestor among reflog entries.
1023 */
1024 if (!bases || bases->next)
1025 goto cleanup_return;
1026
1027 /* And the found one must be one of the reflog entries */
1028 for (i = 0; i < revs.nr; i++)
1029 if (&bases->item->object == &revs.commit[i]->object)
1030 break; /* found */
1031 if (revs.nr <= i)
1032 goto cleanup_return;
1033
1034 ret = bases->item;
1035
1036 cleanup_return:
1037 free(revs.commit);
1038 free_commit_list(bases);
1039 free(full_refname);
1040 return ret;
1041 }
1042
1043 /*
1044 * Indexed by hash algorithm identifier.
1045 */
1046 static const char *gpg_sig_headers[] = {
1047 NULL,
1048 "gpgsig",
1049 "gpgsig-sha256",
1050 };
1051
1052 int sign_with_header(struct strbuf *buf, const char *keyid)
1053 {
1054 struct strbuf sig = STRBUF_INIT;
1055 int inspos, copypos;
1056 const char *eoh;
1057 const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(the_hash_algo)];
1058 int gpg_sig_header_len = strlen(gpg_sig_header);
1059
1060 /* find the end of the header */
1061 eoh = strstr(buf->buf, "\n\n");
1062 if (!eoh)
1063 inspos = buf->len;
1064 else
1065 inspos = eoh - buf->buf + 1;
1066
1067 if (!keyid || !*keyid)
1068 keyid = get_signing_key();
1069 if (sign_buffer(buf, &sig, keyid)) {
1070 strbuf_release(&sig);
1071 return -1;
1072 }
1073
1074 for (copypos = 0; sig.buf[copypos]; ) {
1075 const char *bol = sig.buf + copypos;
1076 const char *eol = strchrnul(bol, '\n');
1077 int len = (eol - bol) + !!*eol;
1078
1079 if (!copypos) {
1080 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
1081 inspos += gpg_sig_header_len;
1082 }
1083 strbuf_insertstr(buf, inspos++, " ");
1084 strbuf_insert(buf, inspos, bol, len);
1085 inspos += len;
1086 copypos += len;
1087 }
1088 strbuf_release(&sig);
1089 return 0;
1090 }
1091
1092
1093
1094 int parse_signed_commit(const struct commit *commit,
1095 struct strbuf *payload, struct strbuf *signature,
1096 const struct git_hash_algo *algop)
1097 {
1098 unsigned long size;
1099 const char *buffer = get_commit_buffer(commit, &size);
1100 int ret = parse_buffer_signed_by_header(buffer, size, payload, signature, algop);
1101
1102 unuse_commit_buffer(commit, buffer);
1103 return ret;
1104 }
1105
1106 int parse_buffer_signed_by_header(const char *buffer,
1107 unsigned long size,
1108 struct strbuf *payload,
1109 struct strbuf *signature,
1110 const struct git_hash_algo *algop)
1111 {
1112 int in_signature = 0, saw_signature = 0, other_signature = 0;
1113 const char *line, *tail, *p;
1114 const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(algop)];
1115
1116 line = buffer;
1117 tail = buffer + size;
1118 while (line < tail) {
1119 const char *sig = NULL;
1120 const char *next = memchr(line, '\n', tail - line);
1121
1122 next = next ? next + 1 : tail;
1123 if (in_signature && line[0] == ' ')
1124 sig = line + 1;
1125 else if (skip_prefix(line, gpg_sig_header, &p) &&
1126 *p == ' ') {
1127 sig = line + strlen(gpg_sig_header) + 1;
1128 other_signature = 0;
1129 }
1130 else if (starts_with(line, "gpgsig"))
1131 other_signature = 1;
1132 else if (other_signature && line[0] != ' ')
1133 other_signature = 0;
1134 if (sig) {
1135 strbuf_add(signature, sig, next - sig);
1136 saw_signature = 1;
1137 in_signature = 1;
1138 } else {
1139 if (*line == '\n')
1140 /* dump the whole remainder of the buffer */
1141 next = tail;
1142 if (!other_signature)
1143 strbuf_add(payload, line, next - line);
1144 in_signature = 0;
1145 }
1146 line = next;
1147 }
1148 return saw_signature;
1149 }
1150
1151 int remove_signature(struct strbuf *buf)
1152 {
1153 const char *line = buf->buf;
1154 const char *tail = buf->buf + buf->len;
1155 int in_signature = 0;
1156 struct sigbuf {
1157 const char *start;
1158 const char *end;
1159 } sigs[2], *sigp = &sigs[0];
1160 int i;
1161 const char *orig_buf = buf->buf;
1162
1163 memset(sigs, 0, sizeof(sigs));
1164
1165 while (line < tail) {
1166 const char *next = memchr(line, '\n', tail - line);
1167 next = next ? next + 1 : tail;
1168
1169 if (in_signature && line[0] == ' ')
1170 sigp->end = next;
1171 else if (starts_with(line, "gpgsig")) {
1172 int i;
1173 for (i = 1; i < GIT_HASH_NALGOS; i++) {
1174 const char *p;
1175 if (skip_prefix(line, gpg_sig_headers[i], &p) &&
1176 *p == ' ') {
1177 sigp->start = line;
1178 sigp->end = next;
1179 in_signature = 1;
1180 }
1181 }
1182 } else {
1183 if (*line == '\n')
1184 /* dump the whole remainder of the buffer */
1185 next = tail;
1186 if (in_signature && sigp - sigs != ARRAY_SIZE(sigs))
1187 sigp++;
1188 in_signature = 0;
1189 }
1190 line = next;
1191 }
1192
1193 for (i = ARRAY_SIZE(sigs) - 1; i >= 0; i--)
1194 if (sigs[i].start)
1195 strbuf_remove(buf, sigs[i].start - orig_buf, sigs[i].end - sigs[i].start);
1196
1197 return sigs[0].start != NULL;
1198 }
1199
1200 static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
1201 {
1202 struct merge_remote_desc *desc;
1203 struct commit_extra_header *mergetag;
1204 char *buf;
1205 unsigned long size;
1206 enum object_type type;
1207 struct strbuf payload = STRBUF_INIT;
1208 struct strbuf signature = STRBUF_INIT;
1209
1210 desc = merge_remote_util(parent);
1211 if (!desc || !desc->obj)
1212 return;
1213 buf = read_object_file(&desc->obj->oid, &type, &size);
1214 if (!buf || type != OBJ_TAG)
1215 goto free_return;
1216 if (!parse_signature(buf, size, &payload, &signature))
1217 goto free_return;
1218 /*
1219 * We could verify this signature and either omit the tag when
1220 * it does not validate, but the integrator may not have the
1221 * public key of the signer of the tag being merged, while a
1222 * later auditor may have it while auditing, so let's not run
1223 * verify-signed-buffer here for now...
1224 *
1225 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
1226 * warn("warning: signed tag unverified.");
1227 */
1228 CALLOC_ARRAY(mergetag, 1);
1229 mergetag->key = xstrdup("mergetag");
1230 mergetag->value = buf;
1231 mergetag->len = size;
1232
1233 **tail = mergetag;
1234 *tail = &mergetag->next;
1235 strbuf_release(&payload);
1236 strbuf_release(&signature);
1237 return;
1238
1239 free_return:
1240 free(buf);
1241 }
1242
1243 int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1244 {
1245 struct strbuf payload = STRBUF_INIT;
1246 struct strbuf signature = STRBUF_INIT;
1247 int ret = 1;
1248
1249 sigc->result = 'N';
1250
1251 if (parse_signed_commit(commit, &payload, &signature, the_hash_algo) <= 0)
1252 goto out;
1253
1254 sigc->payload_type = SIGNATURE_PAYLOAD_COMMIT;
1255 sigc->payload = strbuf_detach(&payload, &sigc->payload_len);
1256 ret = check_signature(sigc, signature.buf, signature.len);
1257
1258 out:
1259 strbuf_release(&payload);
1260 strbuf_release(&signature);
1261
1262 return ret;
1263 }
1264
1265 void verify_merge_signature(struct commit *commit, int verbosity,
1266 int check_trust)
1267 {
1268 char hex[GIT_MAX_HEXSZ + 1];
1269 struct signature_check signature_check;
1270 int ret;
1271 memset(&signature_check, 0, sizeof(signature_check));
1272
1273 ret = check_commit_signature(commit, &signature_check);
1274
1275 find_unique_abbrev_r(hex, &commit->object.oid, DEFAULT_ABBREV);
1276 switch (signature_check.result) {
1277 case 'G':
1278 if (ret || (check_trust && signature_check.trust_level < TRUST_MARGINAL))
1279 die(_("Commit %s has an untrusted GPG signature, "
1280 "allegedly by %s."), hex, signature_check.signer);
1281 break;
1282 case 'B':
1283 die(_("Commit %s has a bad GPG signature "
1284 "allegedly by %s."), hex, signature_check.signer);
1285 default: /* 'N' */
1286 die(_("Commit %s does not have a GPG signature."), hex);
1287 }
1288 if (verbosity >= 0 && signature_check.result == 'G')
1289 printf(_("Commit %s has a good GPG signature by %s\n"),
1290 hex, signature_check.signer);
1291
1292 signature_check_clear(&signature_check);
1293 }
1294
1295 void append_merge_tag_headers(struct commit_list *parents,
1296 struct commit_extra_header ***tail)
1297 {
1298 while (parents) {
1299 struct commit *parent = parents->item;
1300 handle_signed_tag(parent, tail);
1301 parents = parents->next;
1302 }
1303 }
1304
1305 static void add_extra_header(struct strbuf *buffer,
1306 struct commit_extra_header *extra)
1307 {
1308 strbuf_addstr(buffer, extra->key);
1309 if (extra->len)
1310 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1311 else
1312 strbuf_addch(buffer, '\n');
1313 }
1314
1315 struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1316 const char **exclude)
1317 {
1318 struct commit_extra_header *extra = NULL;
1319 unsigned long size;
1320 const char *buffer = get_commit_buffer(commit, &size);
1321 extra = read_commit_extra_header_lines(buffer, size, exclude);
1322 unuse_commit_buffer(commit, buffer);
1323 return extra;
1324 }
1325
1326 int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1327 {
1328 struct commit_extra_header *extra, *to_free;
1329 int res = 0;
1330
1331 to_free = read_commit_extra_headers(commit, NULL);
1332 for (extra = to_free; !res && extra; extra = extra->next) {
1333 if (strcmp(extra->key, "mergetag"))
1334 continue; /* not a merge tag */
1335 res = fn(commit, extra, data);
1336 }
1337 free_commit_extra_headers(to_free);
1338 return res;
1339 }
1340
1341 static inline int standard_header_field(const char *field, size_t len)
1342 {
1343 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1344 (len == 6 && !memcmp(field, "parent", 6)) ||
1345 (len == 6 && !memcmp(field, "author", 6)) ||
1346 (len == 9 && !memcmp(field, "committer", 9)) ||
1347 (len == 8 && !memcmp(field, "encoding", 8)));
1348 }
1349
1350 static int excluded_header_field(const char *field, size_t len, const char **exclude)
1351 {
1352 if (!exclude)
1353 return 0;
1354
1355 while (*exclude) {
1356 size_t xlen = strlen(*exclude);
1357 if (len == xlen && !memcmp(field, *exclude, xlen))
1358 return 1;
1359 exclude++;
1360 }
1361 return 0;
1362 }
1363
1364 static struct commit_extra_header *read_commit_extra_header_lines(
1365 const char *buffer, size_t size,
1366 const char **exclude)
1367 {
1368 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1369 const char *line, *next, *eof, *eob;
1370 struct strbuf buf = STRBUF_INIT;
1371
1372 for (line = buffer, eob = line + size;
1373 line < eob && *line != '\n';
1374 line = next) {
1375 next = memchr(line, '\n', eob - line);
1376 next = next ? next + 1 : eob;
1377 if (*line == ' ') {
1378 /* continuation */
1379 if (it)
1380 strbuf_add(&buf, line + 1, next - (line + 1));
1381 continue;
1382 }
1383 if (it)
1384 it->value = strbuf_detach(&buf, &it->len);
1385 strbuf_reset(&buf);
1386 it = NULL;
1387
1388 eof = memchr(line, ' ', next - line);
1389 if (!eof)
1390 eof = next;
1391 else if (standard_header_field(line, eof - line) ||
1392 excluded_header_field(line, eof - line, exclude))
1393 continue;
1394
1395 CALLOC_ARRAY(it, 1);
1396 it->key = xmemdupz(line, eof-line);
1397 *tail = it;
1398 tail = &it->next;
1399 if (eof + 1 < next)
1400 strbuf_add(&buf, eof + 1, next - (eof + 1));
1401 }
1402 if (it)
1403 it->value = strbuf_detach(&buf, &it->len);
1404 return extra;
1405 }
1406
1407 void free_commit_extra_headers(struct commit_extra_header *extra)
1408 {
1409 while (extra) {
1410 struct commit_extra_header *next = extra->next;
1411 free(extra->key);
1412 free(extra->value);
1413 free(extra);
1414 extra = next;
1415 }
1416 }
1417
1418 int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1419 struct commit_list *parents, struct object_id *ret,
1420 const char *author, const char *sign_commit)
1421 {
1422 struct commit_extra_header *extra = NULL, **tail = &extra;
1423 int result;
1424
1425 append_merge_tag_headers(parents, &tail);
1426 result = commit_tree_extended(msg, msg_len, tree, parents, ret, author,
1427 NULL, sign_commit, extra);
1428 free_commit_extra_headers(extra);
1429 return result;
1430 }
1431
1432 static int find_invalid_utf8(const char *buf, int len)
1433 {
1434 int offset = 0;
1435 static const unsigned int max_codepoint[] = {
1436 0x7f, 0x7ff, 0xffff, 0x10ffff
1437 };
1438
1439 while (len) {
1440 unsigned char c = *buf++;
1441 int bytes, bad_offset;
1442 unsigned int codepoint;
1443 unsigned int min_val, max_val;
1444
1445 len--;
1446 offset++;
1447
1448 /* Simple US-ASCII? No worries. */
1449 if (c < 0x80)
1450 continue;
1451
1452 bad_offset = offset-1;
1453
1454 /*
1455 * Count how many more high bits set: that's how
1456 * many more bytes this sequence should have.
1457 */
1458 bytes = 0;
1459 while (c & 0x40) {
1460 c <<= 1;
1461 bytes++;
1462 }
1463
1464 /*
1465 * Must be between 1 and 3 more bytes. Longer sequences result in
1466 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1467 */
1468 if (bytes < 1 || 3 < bytes)
1469 return bad_offset;
1470
1471 /* Do we *have* that many bytes? */
1472 if (len < bytes)
1473 return bad_offset;
1474
1475 /*
1476 * Place the encoded bits at the bottom of the value and compute the
1477 * valid range.
1478 */
1479 codepoint = (c & 0x7f) >> bytes;
1480 min_val = max_codepoint[bytes-1] + 1;
1481 max_val = max_codepoint[bytes];
1482
1483 offset += bytes;
1484 len -= bytes;
1485
1486 /* And verify that they are good continuation bytes */
1487 do {
1488 codepoint <<= 6;
1489 codepoint |= *buf & 0x3f;
1490 if ((*buf++ & 0xc0) != 0x80)
1491 return bad_offset;
1492 } while (--bytes);
1493
1494 /* Reject codepoints that are out of range for the sequence length. */
1495 if (codepoint < min_val || codepoint > max_val)
1496 return bad_offset;
1497 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1498 if ((codepoint & 0x1ff800) == 0xd800)
1499 return bad_offset;
1500 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1501 if ((codepoint & 0xfffe) == 0xfffe)
1502 return bad_offset;
1503 /* So are anything in the range U+FDD0..U+FDEF. */
1504 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1505 return bad_offset;
1506 }
1507 return -1;
1508 }
1509
1510 /*
1511 * This verifies that the buffer is in proper utf8 format.
1512 *
1513 * If it isn't, it assumes any non-utf8 characters are Latin1,
1514 * and does the conversion.
1515 */
1516 static int verify_utf8(struct strbuf *buf)
1517 {
1518 int ok = 1;
1519 long pos = 0;
1520
1521 for (;;) {
1522 int bad;
1523 unsigned char c;
1524 unsigned char replace[2];
1525
1526 bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1527 if (bad < 0)
1528 return ok;
1529 pos += bad;
1530 ok = 0;
1531 c = buf->buf[pos];
1532 strbuf_remove(buf, pos, 1);
1533
1534 /* We know 'c' must be in the range 128-255 */
1535 replace[0] = 0xc0 + (c >> 6);
1536 replace[1] = 0x80 + (c & 0x3f);
1537 strbuf_insert(buf, pos, replace, 2);
1538 pos += 2;
1539 }
1540 }
1541
1542 static const char commit_utf8_warn[] =
1543 N_("Warning: commit message did not conform to UTF-8.\n"
1544 "You may want to amend it after fixing the message, or set the config\n"
1545 "variable i18n.commitEncoding to the encoding your project uses.\n");
1546
1547 int commit_tree_extended(const char *msg, size_t msg_len,
1548 const struct object_id *tree,
1549 struct commit_list *parents, struct object_id *ret,
1550 const char *author, const char *committer,
1551 const char *sign_commit,
1552 struct commit_extra_header *extra)
1553 {
1554 int result;
1555 int encoding_is_utf8;
1556 struct strbuf buffer;
1557
1558 assert_oid_type(tree, OBJ_TREE);
1559
1560 if (memchr(msg, '\0', msg_len))
1561 return error("a NUL byte in commit log message not allowed.");
1562
1563 /* Not having i18n.commitencoding is the same as having utf-8 */
1564 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1565
1566 strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1567 strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1568
1569 /*
1570 * NOTE! This ordering means that the same exact tree merged with a
1571 * different order of parents will be a _different_ changeset even
1572 * if everything else stays the same.
1573 */
1574 while (parents) {
1575 struct commit *parent = pop_commit(&parents);
1576 strbuf_addf(&buffer, "parent %s\n",
1577 oid_to_hex(&parent->object.oid));
1578 }
1579
1580 /* Person/date information */
1581 if (!author)
1582 author = git_author_info(IDENT_STRICT);
1583 strbuf_addf(&buffer, "author %s\n", author);
1584 if (!committer)
1585 committer = git_committer_info(IDENT_STRICT);
1586 strbuf_addf(&buffer, "committer %s\n", committer);
1587 if (!encoding_is_utf8)
1588 strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1589
1590 while (extra) {
1591 add_extra_header(&buffer, extra);
1592 extra = extra->next;
1593 }
1594 strbuf_addch(&buffer, '\n');
1595
1596 /* And add the comment */
1597 strbuf_add(&buffer, msg, msg_len);
1598
1599 /* And check the encoding */
1600 if (encoding_is_utf8 && !verify_utf8(&buffer))
1601 fprintf(stderr, _(commit_utf8_warn));
1602
1603 if (sign_commit && sign_with_header(&buffer, sign_commit)) {
1604 result = -1;
1605 goto out;
1606 }
1607
1608 result = write_object_file(buffer.buf, buffer.len, OBJ_COMMIT, ret);
1609 out:
1610 strbuf_release(&buffer);
1611 return result;
1612 }
1613
1614 define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1615 static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1616
1617 struct merge_remote_desc *merge_remote_util(struct commit *commit)
1618 {
1619 return *merge_desc_slab_at(&merge_desc_slab, commit);
1620 }
1621
1622 void set_merge_remote_desc(struct commit *commit,
1623 const char *name, struct object *obj)
1624 {
1625 struct merge_remote_desc *desc;
1626 FLEX_ALLOC_STR(desc, name, name);
1627 desc->obj = obj;
1628 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1629 }
1630
1631 struct commit *get_merge_parent(const char *name)
1632 {
1633 struct object *obj;
1634 struct commit *commit;
1635 struct object_id oid;
1636 if (get_oid(name, &oid))
1637 return NULL;
1638 obj = parse_object(the_repository, &oid);
1639 commit = (struct commit *)peel_to_type(name, 0, obj, OBJ_COMMIT);
1640 if (commit && !merge_remote_util(commit))
1641 set_merge_remote_desc(commit, name, obj);
1642 return commit;
1643 }
1644
1645 /*
1646 * Append a commit to the end of the commit_list.
1647 *
1648 * next starts by pointing to the variable that holds the head of an
1649 * empty commit_list, and is updated to point to the "next" field of
1650 * the last item on the list as new commits are appended.
1651 *
1652 * Usage example:
1653 *
1654 * struct commit_list *list;
1655 * struct commit_list **next = &list;
1656 *
1657 * next = commit_list_append(c1, next);
1658 * next = commit_list_append(c2, next);
1659 * assert(commit_list_count(list) == 2);
1660 * return list;
1661 */
1662 struct commit_list **commit_list_append(struct commit *commit,
1663 struct commit_list **next)
1664 {
1665 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1666 new_commit->item = commit;
1667 *next = new_commit;
1668 new_commit->next = NULL;
1669 return &new_commit->next;
1670 }
1671
1672 const char *find_header_mem(const char *msg, size_t len,
1673 const char *key, size_t *out_len)
1674 {
1675 int key_len = strlen(key);
1676 const char *line = msg;
1677
1678 /*
1679 * NEEDSWORK: It's possible for strchrnul() to scan beyond the range
1680 * given by len. However, current callers are safe because they compute
1681 * len by scanning a NUL-terminated block of memory starting at msg.
1682 * Nonetheless, it would be better to ensure the function does not look
1683 * at msg beyond the len provided by the caller.
1684 */
1685 while (line && line < msg + len) {
1686 const char *eol = strchrnul(line, '\n');
1687
1688 if (line == eol)
1689 return NULL;
1690
1691 if (eol - line > key_len &&
1692 !strncmp(line, key, key_len) &&
1693 line[key_len] == ' ') {
1694 *out_len = eol - line - key_len - 1;
1695 return line + key_len + 1;
1696 }
1697 line = *eol ? eol + 1 : NULL;
1698 }
1699 return NULL;
1700 }
1701
1702 const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1703 {
1704 return find_header_mem(msg, strlen(msg), key, out_len);
1705 }
1706 /*
1707 * Inspect the given string and determine the true "end" of the log message, in
1708 * order to find where to put a new Signed-off-by trailer. Ignored are
1709 * trailing comment lines and blank lines. To support "git commit -s
1710 * --amend" on an existing commit, we also ignore "Conflicts:". To
1711 * support "git commit -v", we truncate at cut lines.
1712 *
1713 * Returns the number of bytes from the tail to ignore, to be fed as
1714 * the second parameter to append_signoff().
1715 */
1716 size_t ignore_non_trailer(const char *buf, size_t len)
1717 {
1718 size_t boc = 0;
1719 size_t bol = 0;
1720 int in_old_conflicts_block = 0;
1721 size_t cutoff = wt_status_locate_end(buf, len);
1722
1723 while (bol < cutoff) {
1724 const char *next_line = memchr(buf + bol, '\n', len - bol);
1725
1726 if (!next_line)
1727 next_line = buf + len;
1728 else
1729 next_line++;
1730
1731 if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1732 /* is this the first of the run of comments? */
1733 if (!boc)
1734 boc = bol;
1735 /* otherwise, it is just continuing */
1736 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1737 in_old_conflicts_block = 1;
1738 if (!boc)
1739 boc = bol;
1740 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1741 ; /* a pathname in the conflicts block */
1742 } else if (boc) {
1743 /* the previous was not trailing comment */
1744 boc = 0;
1745 in_old_conflicts_block = 0;
1746 }
1747 bol = next_line - buf;
1748 }
1749 return boc ? len - boc : len - cutoff;
1750 }
1751
1752 int run_commit_hook(int editor_is_used, const char *index_file,
1753 int *invoked_hook, const char *name, ...)
1754 {
1755 struct run_hooks_opt opt = RUN_HOOKS_OPT_INIT;
1756 va_list args;
1757 const char *arg;
1758
1759 strvec_pushf(&opt.env, "GIT_INDEX_FILE=%s", index_file);
1760
1761 /*
1762 * Let the hook know that no editor will be launched.
1763 */
1764 if (!editor_is_used)
1765 strvec_push(&opt.env, "GIT_EDITOR=:");
1766
1767 va_start(args, name);
1768 while ((arg = va_arg(args, const char *)))
1769 strvec_push(&opt.args, arg);
1770 va_end(args);
1771
1772 opt.invoked_hook = invoked_hook;
1773 return run_hooks_opt(name, &opt);
1774 }