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