]> git.ipfire.org Git - thirdparty/git.git/blob - commit.c
object-name.h: move declarations for object-name.c functions from cache.h
[thirdparty/git.git] / commit.c
1 #include "cache.h"
2 #include "tag.h"
3 #include "commit.h"
4 #include "commit-graph.h"
5 #include "environment.h"
6 #include "gettext.h"
7 #include "hex.h"
8 #include "repository.h"
9 #include "object-name.h"
10 #include "object-store.h"
11 #include "pkt-line.h"
12 #include "utf8.h"
13 #include "diff.h"
14 #include "revision.h"
15 #include "notes.h"
16 #include "alloc.h"
17 #include "gpg-interface.h"
18 #include "mergesort.h"
19 #include "commit-slab.h"
20 #include "prio-queue.h"
21 #include "hash-lookup.h"
22 #include "wt-status.h"
23 #include "advice.h"
24 #include "refs.h"
25 #include "commit-reach.h"
26 #include "run-command.h"
27 #include "setup.h"
28 #include "shallow.h"
29 #include "hook.h"
30
31 static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
32
33 int save_commit_buffer = 1;
34 int no_graft_file_deprecated_advice;
35
36 const char *commit_type = "commit";
37
38 struct commit *lookup_commit_reference_gently(struct repository *r,
39 const struct object_id *oid, int quiet)
40 {
41 struct object *obj = deref_tag(r,
42 parse_object(r, oid),
43 NULL, 0);
44
45 if (!obj)
46 return NULL;
47 return object_as_type(obj, OBJ_COMMIT, quiet);
48 }
49
50 struct commit *lookup_commit_reference(struct repository *r, const struct object_id *oid)
51 {
52 return lookup_commit_reference_gently(r, oid, 0);
53 }
54
55 struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
56 {
57 struct commit *c = lookup_commit_reference(the_repository, oid);
58 if (!c)
59 die(_("could not parse %s"), ref_name);
60 if (!oideq(oid, &c->object.oid)) {
61 warning(_("%s %s is not a commit!"),
62 ref_name, oid_to_hex(oid));
63 }
64 return c;
65 }
66
67 struct commit *lookup_commit_object(struct repository *r,
68 const struct object_id *oid)
69 {
70 struct object *obj = parse_object(r, oid);
71 return obj ? object_as_type(obj, OBJ_COMMIT, 0) : NULL;
72
73 }
74
75 struct commit *lookup_commit(struct repository *r, const struct object_id *oid)
76 {
77 struct object *obj = lookup_object(r, oid);
78 if (!obj)
79 return create_object(r, oid, alloc_commit_node(r));
80 return object_as_type(obj, OBJ_COMMIT, 0);
81 }
82
83 struct commit *lookup_commit_reference_by_name(const char *name)
84 {
85 struct object_id oid;
86 struct commit *commit;
87
88 if (repo_get_oid_committish(the_repository, name, &oid))
89 return NULL;
90 commit = lookup_commit_reference(the_repository, &oid);
91 if (repo_parse_commit(the_repository, commit))
92 return NULL;
93 return commit;
94 }
95
96 static timestamp_t parse_commit_date(const char *buf, const char *tail)
97 {
98 const char *dateptr;
99
100 if (buf + 6 >= tail)
101 return 0;
102 if (memcmp(buf, "author", 6))
103 return 0;
104 while (buf < tail && *buf++ != '\n')
105 /* nada */;
106 if (buf + 9 >= tail)
107 return 0;
108 if (memcmp(buf, "committer", 9))
109 return 0;
110 while (buf < tail && *buf++ != '>')
111 /* nada */;
112 if (buf >= tail)
113 return 0;
114 dateptr = buf;
115 while (buf < tail && *buf++ != '\n')
116 /* nada */;
117 if (buf >= tail)
118 return 0;
119 /* dateptr < buf && buf[-1] == '\n', so parsing will stop at buf-1 */
120 return parse_timestamp(dateptr, NULL, 10);
121 }
122
123 static const struct object_id *commit_graft_oid_access(size_t index, const void *table)
124 {
125 const struct commit_graft * const *commit_graft_table = table;
126 return &commit_graft_table[index]->oid;
127 }
128
129 int commit_graft_pos(struct repository *r, const struct object_id *oid)
130 {
131 return oid_pos(oid, r->parsed_objects->grafts,
132 r->parsed_objects->grafts_nr,
133 commit_graft_oid_access);
134 }
135
136 static void unparse_commit(struct repository *r, const struct object_id *oid)
137 {
138 struct commit *c = lookup_commit(r, oid);
139
140 if (!c->object.parsed)
141 return;
142 free_commit_list(c->parents);
143 c->parents = NULL;
144 c->object.parsed = 0;
145 }
146
147 int register_commit_graft(struct repository *r, struct commit_graft *graft,
148 int ignore_dups)
149 {
150 int pos = commit_graft_pos(r, &graft->oid);
151
152 if (0 <= pos) {
153 if (ignore_dups)
154 free(graft);
155 else {
156 free(r->parsed_objects->grafts[pos]);
157 r->parsed_objects->grafts[pos] = graft;
158 }
159 return 1;
160 }
161 pos = -pos - 1;
162 ALLOC_GROW(r->parsed_objects->grafts,
163 r->parsed_objects->grafts_nr + 1,
164 r->parsed_objects->grafts_alloc);
165 r->parsed_objects->grafts_nr++;
166 if (pos < r->parsed_objects->grafts_nr)
167 memmove(r->parsed_objects->grafts + pos + 1,
168 r->parsed_objects->grafts + pos,
169 (r->parsed_objects->grafts_nr - pos - 1) *
170 sizeof(*r->parsed_objects->grafts));
171 r->parsed_objects->grafts[pos] = graft;
172 unparse_commit(r, &graft->oid);
173 return 0;
174 }
175
176 struct commit_graft *read_graft_line(struct strbuf *line)
177 {
178 /* The format is just "Commit Parent1 Parent2 ...\n" */
179 int i, phase;
180 const char *tail = NULL;
181 struct commit_graft *graft = NULL;
182 struct object_id dummy_oid, *oid;
183
184 strbuf_rtrim(line);
185 if (!line->len || line->buf[0] == '#')
186 return NULL;
187 /*
188 * phase 0 verifies line, counts hashes in line and allocates graft
189 * phase 1 fills graft
190 */
191 for (phase = 0; phase < 2; phase++) {
192 oid = graft ? &graft->oid : &dummy_oid;
193 if (parse_oid_hex(line->buf, oid, &tail))
194 goto bad_graft_data;
195 for (i = 0; *tail != '\0'; i++) {
196 oid = graft ? &graft->parent[i] : &dummy_oid;
197 if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail))
198 goto bad_graft_data;
199 }
200 if (!graft) {
201 graft = xmalloc(st_add(sizeof(*graft),
202 st_mult(sizeof(struct object_id), i)));
203 graft->nr_parent = i;
204 }
205 }
206 return graft;
207
208 bad_graft_data:
209 error("bad graft data: %s", line->buf);
210 assert(!graft);
211 return NULL;
212 }
213
214 static int read_graft_file(struct repository *r, const char *graft_file)
215 {
216 FILE *fp = fopen_or_warn(graft_file, "r");
217 struct strbuf buf = STRBUF_INIT;
218 if (!fp)
219 return -1;
220 if (!no_graft_file_deprecated_advice &&
221 advice_enabled(ADVICE_GRAFT_FILE_DEPRECATED))
222 advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n"
223 "and will be removed in a future Git version.\n"
224 "\n"
225 "Please use \"git replace --convert-graft-file\"\n"
226 "to convert the grafts into replace refs.\n"
227 "\n"
228 "Turn this message off by running\n"
229 "\"git config advice.graftFileDeprecated false\""));
230 while (!strbuf_getwholeline(&buf, fp, '\n')) {
231 /* The format is just "Commit Parent1 Parent2 ...\n" */
232 struct commit_graft *graft = read_graft_line(&buf);
233 if (!graft)
234 continue;
235 if (register_commit_graft(r, graft, 1))
236 error("duplicate graft data: %s", buf.buf);
237 }
238 fclose(fp);
239 strbuf_release(&buf);
240 return 0;
241 }
242
243 void prepare_commit_graft(struct repository *r)
244 {
245 char *graft_file;
246
247 if (r->parsed_objects->commit_graft_prepared)
248 return;
249 if (!startup_info->have_repository)
250 return;
251
252 graft_file = get_graft_file(r);
253 read_graft_file(r, graft_file);
254 /* make sure shallows are read */
255 is_repository_shallow(r);
256 r->parsed_objects->commit_graft_prepared = 1;
257 }
258
259 struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid)
260 {
261 int pos;
262 prepare_commit_graft(r);
263 pos = commit_graft_pos(r, oid);
264 if (pos < 0)
265 return NULL;
266 return r->parsed_objects->grafts[pos];
267 }
268
269 int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
270 {
271 int i, ret;
272 for (i = ret = 0; i < the_repository->parsed_objects->grafts_nr && !ret; i++)
273 ret = fn(the_repository->parsed_objects->grafts[i], cb_data);
274 return ret;
275 }
276
277 void reset_commit_grafts(struct repository *r)
278 {
279 int i;
280
281 for (i = 0; i < r->parsed_objects->grafts_nr; i++) {
282 unparse_commit(r, &r->parsed_objects->grafts[i]->oid);
283 free(r->parsed_objects->grafts[i]);
284 }
285 r->parsed_objects->grafts_nr = 0;
286 r->parsed_objects->commit_graft_prepared = 0;
287 }
288
289 struct commit_buffer {
290 void *buffer;
291 unsigned long size;
292 };
293 define_commit_slab(buffer_slab, struct commit_buffer);
294
295 struct buffer_slab *allocate_commit_buffer_slab(void)
296 {
297 struct buffer_slab *bs = xmalloc(sizeof(*bs));
298 init_buffer_slab(bs);
299 return bs;
300 }
301
302 void free_commit_buffer_slab(struct buffer_slab *bs)
303 {
304 clear_buffer_slab(bs);
305 free(bs);
306 }
307
308 void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size)
309 {
310 struct commit_buffer *v = buffer_slab_at(
311 r->parsed_objects->buffer_slab, commit);
312 v->buffer = buffer;
313 v->size = size;
314 }
315
316 const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep)
317 {
318 struct commit_buffer *v = buffer_slab_peek(
319 r->parsed_objects->buffer_slab, commit);
320 if (!v) {
321 if (sizep)
322 *sizep = 0;
323 return NULL;
324 }
325 if (sizep)
326 *sizep = v->size;
327 return v->buffer;
328 }
329
330 const void *repo_get_commit_buffer(struct repository *r,
331 const struct commit *commit,
332 unsigned long *sizep)
333 {
334 const void *ret = get_cached_commit_buffer(r, commit, sizep);
335 if (!ret) {
336 enum object_type type;
337 unsigned long size;
338 ret = repo_read_object_file(r, &commit->object.oid, &type, &size);
339 if (!ret)
340 die("cannot read commit object %s",
341 oid_to_hex(&commit->object.oid));
342 if (type != OBJ_COMMIT)
343 die("expected commit for %s, got %s",
344 oid_to_hex(&commit->object.oid), type_name(type));
345 if (sizep)
346 *sizep = size;
347 }
348 return ret;
349 }
350
351 void repo_unuse_commit_buffer(struct repository *r,
352 const struct commit *commit,
353 const void *buffer)
354 {
355 struct commit_buffer *v = buffer_slab_peek(
356 r->parsed_objects->buffer_slab, commit);
357 if (!(v && v->buffer == buffer))
358 free((void *)buffer);
359 }
360
361 void free_commit_buffer(struct parsed_object_pool *pool, struct commit *commit)
362 {
363 struct commit_buffer *v = buffer_slab_peek(
364 pool->buffer_slab, commit);
365 if (v) {
366 FREE_AND_NULL(v->buffer);
367 v->size = 0;
368 }
369 }
370
371 static inline void set_commit_tree(struct commit *c, struct tree *t)
372 {
373 c->maybe_tree = t;
374 }
375
376 struct tree *repo_get_commit_tree(struct repository *r,
377 const struct commit *commit)
378 {
379 if (commit->maybe_tree || !commit->object.parsed)
380 return commit->maybe_tree;
381
382 if (commit_graph_position(commit) != COMMIT_NOT_FROM_GRAPH)
383 return get_commit_tree_in_graph(r, commit);
384
385 return NULL;
386 }
387
388 struct object_id *get_commit_tree_oid(const struct commit *commit)
389 {
390 struct tree *tree = repo_get_commit_tree(the_repository, commit);
391 return tree ? &tree->object.oid : NULL;
392 }
393
394 void release_commit_memory(struct parsed_object_pool *pool, struct commit *c)
395 {
396 set_commit_tree(c, NULL);
397 free_commit_buffer(pool, c);
398 c->index = 0;
399 free_commit_list(c->parents);
400
401 c->object.parsed = 0;
402 }
403
404 const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
405 {
406 struct commit_buffer *v = buffer_slab_peek(
407 the_repository->parsed_objects->buffer_slab, commit);
408 void *ret;
409
410 if (!v) {
411 if (sizep)
412 *sizep = 0;
413 return NULL;
414 }
415 ret = v->buffer;
416 if (sizep)
417 *sizep = v->size;
418
419 v->buffer = NULL;
420 v->size = 0;
421 return ret;
422 }
423
424 int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
425 {
426 const char *tail = buffer;
427 const char *bufptr = buffer;
428 struct object_id parent;
429 struct commit_list **pptr;
430 struct commit_graft *graft;
431 const int tree_entry_len = the_hash_algo->hexsz + 5;
432 const int parent_entry_len = the_hash_algo->hexsz + 7;
433 struct tree *tree;
434
435 if (item->object.parsed)
436 return 0;
437 /*
438 * Presumably this is leftover from an earlier failed parse;
439 * clear it out in preparation for us re-parsing (we'll hit the
440 * same error, but that's good, since it lets our caller know
441 * the result cannot be trusted.
442 */
443 free_commit_list(item->parents);
444 item->parents = NULL;
445
446 tail += size;
447 if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
448 bufptr[tree_entry_len] != '\n')
449 return error("bogus commit object %s", oid_to_hex(&item->object.oid));
450 if (get_oid_hex(bufptr + 5, &parent) < 0)
451 return error("bad tree pointer in commit %s",
452 oid_to_hex(&item->object.oid));
453 tree = lookup_tree(r, &parent);
454 if (!tree)
455 return error("bad tree pointer %s in commit %s",
456 oid_to_hex(&parent),
457 oid_to_hex(&item->object.oid));
458 set_commit_tree(item, tree);
459 bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
460 pptr = &item->parents;
461
462 graft = lookup_commit_graft(r, &item->object.oid);
463 if (graft)
464 r->parsed_objects->substituted_parent = 1;
465 while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
466 struct commit *new_parent;
467
468 if (tail <= bufptr + parent_entry_len + 1 ||
469 get_oid_hex(bufptr + 7, &parent) ||
470 bufptr[parent_entry_len] != '\n')
471 return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
472 bufptr += parent_entry_len + 1;
473 /*
474 * The clone is shallow if nr_parent < 0, and we must
475 * not traverse its real parents even when we unhide them.
476 */
477 if (graft && (graft->nr_parent < 0 || grafts_replace_parents))
478 continue;
479 new_parent = lookup_commit(r, &parent);
480 if (!new_parent)
481 return error("bad parent %s in commit %s",
482 oid_to_hex(&parent),
483 oid_to_hex(&item->object.oid));
484 pptr = &commit_list_insert(new_parent, pptr)->next;
485 }
486 if (graft) {
487 int i;
488 struct commit *new_parent;
489 for (i = 0; i < graft->nr_parent; i++) {
490 new_parent = lookup_commit(r,
491 &graft->parent[i]);
492 if (!new_parent)
493 return error("bad graft parent %s in commit %s",
494 oid_to_hex(&graft->parent[i]),
495 oid_to_hex(&item->object.oid));
496 pptr = &commit_list_insert(new_parent, pptr)->next;
497 }
498 }
499 item->date = parse_commit_date(bufptr, tail);
500
501 if (check_graph)
502 load_commit_graph_info(r, item);
503
504 item->object.parsed = 1;
505 return 0;
506 }
507
508 int repo_parse_commit_internal(struct repository *r,
509 struct commit *item,
510 int quiet_on_missing,
511 int use_commit_graph)
512 {
513 enum object_type type;
514 void *buffer;
515 unsigned long size;
516 struct object_info oi = {
517 .typep = &type,
518 .sizep = &size,
519 .contentp = &buffer,
520 };
521 /*
522 * Git does not support partial clones that exclude commits, so set
523 * OBJECT_INFO_SKIP_FETCH_OBJECT to fail fast when an object is missing.
524 */
525 int flags = OBJECT_INFO_LOOKUP_REPLACE | OBJECT_INFO_SKIP_FETCH_OBJECT |
526 OBJECT_INFO_DIE_IF_CORRUPT;
527 int ret;
528
529 if (!item)
530 return -1;
531 if (item->object.parsed)
532 return 0;
533 if (use_commit_graph && parse_commit_in_graph(r, item))
534 return 0;
535
536 if (oid_object_info_extended(r, &item->object.oid, &oi, flags) < 0)
537 return quiet_on_missing ? -1 :
538 error("Could not read %s",
539 oid_to_hex(&item->object.oid));
540 if (type != OBJ_COMMIT) {
541 free(buffer);
542 return error("Object %s not a commit",
543 oid_to_hex(&item->object.oid));
544 }
545
546 ret = parse_commit_buffer(r, item, buffer, size, 0);
547 if (save_commit_buffer && !ret) {
548 set_commit_buffer(r, item, buffer, size);
549 return 0;
550 }
551 free(buffer);
552 return ret;
553 }
554
555 int repo_parse_commit_gently(struct repository *r,
556 struct commit *item, int quiet_on_missing)
557 {
558 return repo_parse_commit_internal(r, item, quiet_on_missing, 1);
559 }
560
561 void parse_commit_or_die(struct commit *item)
562 {
563 if (repo_parse_commit(the_repository, item))
564 die("unable to parse commit %s",
565 item ? oid_to_hex(&item->object.oid) : "(null)");
566 }
567
568 int find_commit_subject(const char *commit_buffer, const char **subject)
569 {
570 const char *eol;
571 const char *p = commit_buffer;
572
573 while (*p && (*p != '\n' || p[1] != '\n'))
574 p++;
575 if (*p) {
576 p = skip_blank_lines(p + 2);
577 eol = strchrnul(p, '\n');
578 } else
579 eol = p;
580
581 *subject = p;
582
583 return eol - p;
584 }
585
586 size_t commit_subject_length(const char *body)
587 {
588 const char *p = body;
589 while (*p) {
590 const char *next = skip_blank_lines(p);
591 if (next != p)
592 break;
593 p = strchrnul(p, '\n');
594 if (*p)
595 p++;
596 }
597 return p - body;
598 }
599
600 struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
601 {
602 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
603 new_list->item = item;
604 new_list->next = *list_p;
605 *list_p = new_list;
606 return new_list;
607 }
608
609 int commit_list_contains(struct commit *item, struct commit_list *list)
610 {
611 while (list) {
612 if (list->item == item)
613 return 1;
614 list = list->next;
615 }
616
617 return 0;
618 }
619
620 unsigned commit_list_count(const struct commit_list *l)
621 {
622 unsigned c = 0;
623 for (; l; l = l->next )
624 c++;
625 return c;
626 }
627
628 struct commit_list *copy_commit_list(struct commit_list *list)
629 {
630 struct commit_list *head = NULL;
631 struct commit_list **pp = &head;
632 while (list) {
633 pp = commit_list_append(list->item, pp);
634 list = list->next;
635 }
636 return head;
637 }
638
639 struct commit_list *reverse_commit_list(struct commit_list *list)
640 {
641 struct commit_list *next = NULL, *current, *backup;
642 for (current = list; current; current = backup) {
643 backup = current->next;
644 current->next = next;
645 next = current;
646 }
647 return next;
648 }
649
650 void free_commit_list(struct commit_list *list)
651 {
652 while (list)
653 pop_commit(&list);
654 }
655
656 struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
657 {
658 struct commit_list **pp = list;
659 struct commit_list *p;
660 while ((p = *pp) != NULL) {
661 if (p->item->date < item->date) {
662 break;
663 }
664 pp = &p->next;
665 }
666 return commit_list_insert(item, pp);
667 }
668
669 static int commit_list_compare_by_date(const struct commit_list *a,
670 const struct commit_list *b)
671 {
672 timestamp_t a_date = a->item->date;
673 timestamp_t b_date = b->item->date;
674 if (a_date < b_date)
675 return 1;
676 if (a_date > b_date)
677 return -1;
678 return 0;
679 }
680
681 DEFINE_LIST_SORT(static, commit_list_sort, struct commit_list, next);
682
683 void commit_list_sort_by_date(struct commit_list **list)
684 {
685 commit_list_sort(list, commit_list_compare_by_date);
686 }
687
688 struct commit *pop_most_recent_commit(struct commit_list **list,
689 unsigned int mark)
690 {
691 struct commit *ret = pop_commit(list);
692 struct commit_list *parents = ret->parents;
693
694 while (parents) {
695 struct commit *commit = parents->item;
696 if (!repo_parse_commit(the_repository, commit) && !(commit->object.flags & mark)) {
697 commit->object.flags |= mark;
698 commit_list_insert_by_date(commit, list);
699 }
700 parents = parents->next;
701 }
702 return ret;
703 }
704
705 static void clear_commit_marks_1(struct commit_list **plist,
706 struct commit *commit, unsigned int mark)
707 {
708 while (commit) {
709 struct commit_list *parents;
710
711 if (!(mark & commit->object.flags))
712 return;
713
714 commit->object.flags &= ~mark;
715
716 parents = commit->parents;
717 if (!parents)
718 return;
719
720 while ((parents = parents->next)) {
721 if (parents->item->object.flags & mark)
722 commit_list_insert(parents->item, plist);
723 }
724
725 commit = commit->parents->item;
726 }
727 }
728
729 void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
730 {
731 struct commit_list *list = NULL;
732
733 while (nr--) {
734 clear_commit_marks_1(&list, *commit, mark);
735 commit++;
736 }
737 while (list)
738 clear_commit_marks_1(&list, pop_commit(&list), mark);
739 }
740
741 void clear_commit_marks(struct commit *commit, unsigned int mark)
742 {
743 clear_commit_marks_many(1, &commit, mark);
744 }
745
746 struct commit *pop_commit(struct commit_list **stack)
747 {
748 struct commit_list *top = *stack;
749 struct commit *item = top ? top->item : NULL;
750
751 if (top) {
752 *stack = top->next;
753 free(top);
754 }
755 return item;
756 }
757
758 /*
759 * Topological sort support
760 */
761
762 /* count number of children that have not been emitted */
763 define_commit_slab(indegree_slab, int);
764
765 define_commit_slab(author_date_slab, timestamp_t);
766
767 void record_author_date(struct author_date_slab *author_date,
768 struct commit *commit)
769 {
770 const char *buffer = repo_get_commit_buffer(the_repository, commit,
771 NULL);
772 struct ident_split ident;
773 const char *ident_line;
774 size_t ident_len;
775 char *date_end;
776 timestamp_t date;
777
778 ident_line = find_commit_header(buffer, "author", &ident_len);
779 if (!ident_line)
780 goto fail_exit; /* no author line */
781 if (split_ident_line(&ident, ident_line, ident_len) ||
782 !ident.date_begin || !ident.date_end)
783 goto fail_exit; /* malformed "author" line */
784
785 date = parse_timestamp(ident.date_begin, &date_end, 10);
786 if (date_end != ident.date_end)
787 goto fail_exit; /* malformed date */
788 *(author_date_slab_at(author_date, commit)) = date;
789
790 fail_exit:
791 repo_unuse_commit_buffer(the_repository, commit, buffer);
792 }
793
794 int compare_commits_by_author_date(const void *a_, const void *b_,
795 void *cb_data)
796 {
797 const struct commit *a = a_, *b = b_;
798 struct author_date_slab *author_date = cb_data;
799 timestamp_t a_date = *(author_date_slab_at(author_date, a));
800 timestamp_t b_date = *(author_date_slab_at(author_date, b));
801
802 /* newer commits with larger date first */
803 if (a_date < b_date)
804 return 1;
805 else if (a_date > b_date)
806 return -1;
807 return 0;
808 }
809
810 int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_,
811 void *unused UNUSED)
812 {
813 const struct commit *a = a_, *b = b_;
814 const timestamp_t generation_a = commit_graph_generation(a),
815 generation_b = commit_graph_generation(b);
816
817 /* newer commits first */
818 if (generation_a < generation_b)
819 return 1;
820 else if (generation_a > generation_b)
821 return -1;
822
823 /* use date as a heuristic when generations are equal */
824 if (a->date < b->date)
825 return 1;
826 else if (a->date > b->date)
827 return -1;
828 return 0;
829 }
830
831 int compare_commits_by_commit_date(const void *a_, const void *b_,
832 void *unused UNUSED)
833 {
834 const struct commit *a = a_, *b = b_;
835 /* newer commits with larger date first */
836 if (a->date < b->date)
837 return 1;
838 else if (a->date > b->date)
839 return -1;
840 return 0;
841 }
842
843 /*
844 * Performs an in-place topological sort on the list supplied.
845 */
846 void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
847 {
848 struct commit_list *next, *orig = *list;
849 struct commit_list **pptr;
850 struct indegree_slab indegree;
851 struct prio_queue queue;
852 struct commit *commit;
853 struct author_date_slab author_date;
854
855 if (!orig)
856 return;
857 *list = NULL;
858
859 init_indegree_slab(&indegree);
860 memset(&queue, '\0', sizeof(queue));
861
862 switch (sort_order) {
863 default: /* REV_SORT_IN_GRAPH_ORDER */
864 queue.compare = NULL;
865 break;
866 case REV_SORT_BY_COMMIT_DATE:
867 queue.compare = compare_commits_by_commit_date;
868 break;
869 case REV_SORT_BY_AUTHOR_DATE:
870 init_author_date_slab(&author_date);
871 queue.compare = compare_commits_by_author_date;
872 queue.cb_data = &author_date;
873 break;
874 }
875
876 /* Mark them and clear the indegree */
877 for (next = orig; next; next = next->next) {
878 struct commit *commit = next->item;
879 *(indegree_slab_at(&indegree, commit)) = 1;
880 /* also record the author dates, if needed */
881 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
882 record_author_date(&author_date, commit);
883 }
884
885 /* update the indegree */
886 for (next = orig; next; next = next->next) {
887 struct commit_list *parents = next->item->parents;
888 while (parents) {
889 struct commit *parent = parents->item;
890 int *pi = indegree_slab_at(&indegree, parent);
891
892 if (*pi)
893 (*pi)++;
894 parents = parents->next;
895 }
896 }
897
898 /*
899 * find the tips
900 *
901 * tips are nodes not reachable from any other node in the list
902 *
903 * the tips serve as a starting set for the work queue.
904 */
905 for (next = orig; next; next = next->next) {
906 struct commit *commit = next->item;
907
908 if (*(indegree_slab_at(&indegree, commit)) == 1)
909 prio_queue_put(&queue, commit);
910 }
911
912 /*
913 * This is unfortunate; the initial tips need to be shown
914 * in the order given from the revision traversal machinery.
915 */
916 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
917 prio_queue_reverse(&queue);
918
919 /* We no longer need the commit list */
920 free_commit_list(orig);
921
922 pptr = list;
923 *list = NULL;
924 while ((commit = prio_queue_get(&queue)) != NULL) {
925 struct commit_list *parents;
926
927 for (parents = commit->parents; parents ; parents = parents->next) {
928 struct commit *parent = parents->item;
929 int *pi = indegree_slab_at(&indegree, parent);
930
931 if (!*pi)
932 continue;
933
934 /*
935 * parents are only enqueued for emission
936 * when all their children have been emitted thereby
937 * guaranteeing topological order.
938 */
939 if (--(*pi) == 1)
940 prio_queue_put(&queue, parent);
941 }
942 /*
943 * all children of commit have already been
944 * emitted. we can emit it now.
945 */
946 *(indegree_slab_at(&indegree, commit)) = 0;
947
948 pptr = &commit_list_insert(commit, pptr)->next;
949 }
950
951 clear_indegree_slab(&indegree);
952 clear_prio_queue(&queue);
953 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
954 clear_author_date_slab(&author_date);
955 }
956
957 struct rev_collect {
958 struct commit **commit;
959 int nr;
960 int alloc;
961 unsigned int initial : 1;
962 };
963
964 static void add_one_commit(struct object_id *oid, struct rev_collect *revs)
965 {
966 struct commit *commit;
967
968 if (is_null_oid(oid))
969 return;
970
971 commit = lookup_commit(the_repository, oid);
972 if (!commit ||
973 (commit->object.flags & TMP_MARK) ||
974 repo_parse_commit(the_repository, commit))
975 return;
976
977 ALLOC_GROW(revs->commit, revs->nr + 1, revs->alloc);
978 revs->commit[revs->nr++] = commit;
979 commit->object.flags |= TMP_MARK;
980 }
981
982 static int collect_one_reflog_ent(struct object_id *ooid, struct object_id *noid,
983 const char *ident UNUSED,
984 timestamp_t timestamp UNUSED, int tz UNUSED,
985 const char *message UNUSED, void *cbdata)
986 {
987 struct rev_collect *revs = cbdata;
988
989 if (revs->initial) {
990 revs->initial = 0;
991 add_one_commit(ooid, revs);
992 }
993 add_one_commit(noid, revs);
994 return 0;
995 }
996
997 struct commit *get_fork_point(const char *refname, struct commit *commit)
998 {
999 struct object_id oid;
1000 struct rev_collect revs;
1001 struct commit_list *bases;
1002 int i;
1003 struct commit *ret = NULL;
1004 char *full_refname;
1005
1006 switch (repo_dwim_ref(the_repository, refname, strlen(refname), &oid,
1007 &full_refname, 0)) {
1008 case 0:
1009 die("No such ref: '%s'", refname);
1010 case 1:
1011 break; /* good */
1012 default:
1013 die("Ambiguous refname: '%s'", refname);
1014 }
1015
1016 memset(&revs, 0, sizeof(revs));
1017 revs.initial = 1;
1018 for_each_reflog_ent(full_refname, collect_one_reflog_ent, &revs);
1019
1020 if (!revs.nr)
1021 add_one_commit(&oid, &revs);
1022
1023 for (i = 0; i < revs.nr; i++)
1024 revs.commit[i]->object.flags &= ~TMP_MARK;
1025
1026 bases = repo_get_merge_bases_many(the_repository, commit, revs.nr,
1027 revs.commit);
1028
1029 /*
1030 * There should be one and only one merge base, when we found
1031 * a common ancestor among reflog entries.
1032 */
1033 if (!bases || bases->next)
1034 goto cleanup_return;
1035
1036 /* And the found one must be one of the reflog entries */
1037 for (i = 0; i < revs.nr; i++)
1038 if (&bases->item->object == &revs.commit[i]->object)
1039 break; /* found */
1040 if (revs.nr <= i)
1041 goto cleanup_return;
1042
1043 ret = bases->item;
1044
1045 cleanup_return:
1046 free(revs.commit);
1047 free_commit_list(bases);
1048 free(full_refname);
1049 return ret;
1050 }
1051
1052 /*
1053 * Indexed by hash algorithm identifier.
1054 */
1055 static const char *gpg_sig_headers[] = {
1056 NULL,
1057 "gpgsig",
1058 "gpgsig-sha256",
1059 };
1060
1061 int sign_with_header(struct strbuf *buf, const char *keyid)
1062 {
1063 struct strbuf sig = STRBUF_INIT;
1064 int inspos, copypos;
1065 const char *eoh;
1066 const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(the_hash_algo)];
1067 int gpg_sig_header_len = strlen(gpg_sig_header);
1068
1069 /* find the end of the header */
1070 eoh = strstr(buf->buf, "\n\n");
1071 if (!eoh)
1072 inspos = buf->len;
1073 else
1074 inspos = eoh - buf->buf + 1;
1075
1076 if (!keyid || !*keyid)
1077 keyid = get_signing_key();
1078 if (sign_buffer(buf, &sig, keyid)) {
1079 strbuf_release(&sig);
1080 return -1;
1081 }
1082
1083 for (copypos = 0; sig.buf[copypos]; ) {
1084 const char *bol = sig.buf + copypos;
1085 const char *eol = strchrnul(bol, '\n');
1086 int len = (eol - bol) + !!*eol;
1087
1088 if (!copypos) {
1089 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
1090 inspos += gpg_sig_header_len;
1091 }
1092 strbuf_insertstr(buf, inspos++, " ");
1093 strbuf_insert(buf, inspos, bol, len);
1094 inspos += len;
1095 copypos += len;
1096 }
1097 strbuf_release(&sig);
1098 return 0;
1099 }
1100
1101
1102
1103 int parse_signed_commit(const struct commit *commit,
1104 struct strbuf *payload, struct strbuf *signature,
1105 const struct git_hash_algo *algop)
1106 {
1107 unsigned long size;
1108 const char *buffer = repo_get_commit_buffer(the_repository, commit,
1109 &size);
1110 int ret = parse_buffer_signed_by_header(buffer, size, payload, signature, algop);
1111
1112 repo_unuse_commit_buffer(the_repository, commit, buffer);
1113 return ret;
1114 }
1115
1116 int parse_buffer_signed_by_header(const char *buffer,
1117 unsigned long size,
1118 struct strbuf *payload,
1119 struct strbuf *signature,
1120 const struct git_hash_algo *algop)
1121 {
1122 int in_signature = 0, saw_signature = 0, other_signature = 0;
1123 const char *line, *tail, *p;
1124 const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(algop)];
1125
1126 line = buffer;
1127 tail = buffer + size;
1128 while (line < tail) {
1129 const char *sig = NULL;
1130 const char *next = memchr(line, '\n', tail - line);
1131
1132 next = next ? next + 1 : tail;
1133 if (in_signature && line[0] == ' ')
1134 sig = line + 1;
1135 else if (skip_prefix(line, gpg_sig_header, &p) &&
1136 *p == ' ') {
1137 sig = line + strlen(gpg_sig_header) + 1;
1138 other_signature = 0;
1139 }
1140 else if (starts_with(line, "gpgsig"))
1141 other_signature = 1;
1142 else if (other_signature && line[0] != ' ')
1143 other_signature = 0;
1144 if (sig) {
1145 strbuf_add(signature, sig, next - sig);
1146 saw_signature = 1;
1147 in_signature = 1;
1148 } else {
1149 if (*line == '\n')
1150 /* dump the whole remainder of the buffer */
1151 next = tail;
1152 if (!other_signature)
1153 strbuf_add(payload, line, next - line);
1154 in_signature = 0;
1155 }
1156 line = next;
1157 }
1158 return saw_signature;
1159 }
1160
1161 int remove_signature(struct strbuf *buf)
1162 {
1163 const char *line = buf->buf;
1164 const char *tail = buf->buf + buf->len;
1165 int in_signature = 0;
1166 struct sigbuf {
1167 const char *start;
1168 const char *end;
1169 } sigs[2], *sigp = &sigs[0];
1170 int i;
1171 const char *orig_buf = buf->buf;
1172
1173 memset(sigs, 0, sizeof(sigs));
1174
1175 while (line < tail) {
1176 const char *next = memchr(line, '\n', tail - line);
1177 next = next ? next + 1 : tail;
1178
1179 if (in_signature && line[0] == ' ')
1180 sigp->end = next;
1181 else if (starts_with(line, "gpgsig")) {
1182 int i;
1183 for (i = 1; i < GIT_HASH_NALGOS; i++) {
1184 const char *p;
1185 if (skip_prefix(line, gpg_sig_headers[i], &p) &&
1186 *p == ' ') {
1187 sigp->start = line;
1188 sigp->end = next;
1189 in_signature = 1;
1190 }
1191 }
1192 } else {
1193 if (*line == '\n')
1194 /* dump the whole remainder of the buffer */
1195 next = tail;
1196 if (in_signature && sigp - sigs != ARRAY_SIZE(sigs))
1197 sigp++;
1198 in_signature = 0;
1199 }
1200 line = next;
1201 }
1202
1203 for (i = ARRAY_SIZE(sigs) - 1; i >= 0; i--)
1204 if (sigs[i].start)
1205 strbuf_remove(buf, sigs[i].start - orig_buf, sigs[i].end - sigs[i].start);
1206
1207 return sigs[0].start != NULL;
1208 }
1209
1210 static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
1211 {
1212 struct merge_remote_desc *desc;
1213 struct commit_extra_header *mergetag;
1214 char *buf;
1215 unsigned long size;
1216 enum object_type type;
1217 struct strbuf payload = STRBUF_INIT;
1218 struct strbuf signature = STRBUF_INIT;
1219
1220 desc = merge_remote_util(parent);
1221 if (!desc || !desc->obj)
1222 return;
1223 buf = repo_read_object_file(the_repository, &desc->obj->oid, &type,
1224 &size);
1225 if (!buf || type != OBJ_TAG)
1226 goto free_return;
1227 if (!parse_signature(buf, size, &payload, &signature))
1228 goto free_return;
1229 /*
1230 * We could verify this signature and either omit the tag when
1231 * it does not validate, but the integrator may not have the
1232 * public key of the signer of the tag being merged, while a
1233 * later auditor may have it while auditing, so let's not run
1234 * verify-signed-buffer here for now...
1235 *
1236 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
1237 * warn("warning: signed tag unverified.");
1238 */
1239 CALLOC_ARRAY(mergetag, 1);
1240 mergetag->key = xstrdup("mergetag");
1241 mergetag->value = buf;
1242 mergetag->len = size;
1243
1244 **tail = mergetag;
1245 *tail = &mergetag->next;
1246 strbuf_release(&payload);
1247 strbuf_release(&signature);
1248 return;
1249
1250 free_return:
1251 free(buf);
1252 }
1253
1254 int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1255 {
1256 struct strbuf payload = STRBUF_INIT;
1257 struct strbuf signature = STRBUF_INIT;
1258 int ret = 1;
1259
1260 sigc->result = 'N';
1261
1262 if (parse_signed_commit(commit, &payload, &signature, the_hash_algo) <= 0)
1263 goto out;
1264
1265 sigc->payload_type = SIGNATURE_PAYLOAD_COMMIT;
1266 sigc->payload = strbuf_detach(&payload, &sigc->payload_len);
1267 ret = check_signature(sigc, signature.buf, signature.len);
1268
1269 out:
1270 strbuf_release(&payload);
1271 strbuf_release(&signature);
1272
1273 return ret;
1274 }
1275
1276 void verify_merge_signature(struct commit *commit, int verbosity,
1277 int check_trust)
1278 {
1279 char hex[GIT_MAX_HEXSZ + 1];
1280 struct signature_check signature_check;
1281 int ret;
1282 memset(&signature_check, 0, sizeof(signature_check));
1283
1284 ret = check_commit_signature(commit, &signature_check);
1285
1286 repo_find_unique_abbrev_r(the_repository, hex, &commit->object.oid,
1287 DEFAULT_ABBREV);
1288 switch (signature_check.result) {
1289 case 'G':
1290 if (ret || (check_trust && signature_check.trust_level < TRUST_MARGINAL))
1291 die(_("Commit %s has an untrusted GPG signature, "
1292 "allegedly by %s."), hex, signature_check.signer);
1293 break;
1294 case 'B':
1295 die(_("Commit %s has a bad GPG signature "
1296 "allegedly by %s."), hex, signature_check.signer);
1297 default: /* 'N' */
1298 die(_("Commit %s does not have a GPG signature."), hex);
1299 }
1300 if (verbosity >= 0 && signature_check.result == 'G')
1301 printf(_("Commit %s has a good GPG signature by %s\n"),
1302 hex, signature_check.signer);
1303
1304 signature_check_clear(&signature_check);
1305 }
1306
1307 void append_merge_tag_headers(struct commit_list *parents,
1308 struct commit_extra_header ***tail)
1309 {
1310 while (parents) {
1311 struct commit *parent = parents->item;
1312 handle_signed_tag(parent, tail);
1313 parents = parents->next;
1314 }
1315 }
1316
1317 static void add_extra_header(struct strbuf *buffer,
1318 struct commit_extra_header *extra)
1319 {
1320 strbuf_addstr(buffer, extra->key);
1321 if (extra->len)
1322 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1323 else
1324 strbuf_addch(buffer, '\n');
1325 }
1326
1327 struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1328 const char **exclude)
1329 {
1330 struct commit_extra_header *extra = NULL;
1331 unsigned long size;
1332 const char *buffer = repo_get_commit_buffer(the_repository, commit,
1333 &size);
1334 extra = read_commit_extra_header_lines(buffer, size, exclude);
1335 repo_unuse_commit_buffer(the_repository, commit, buffer);
1336 return extra;
1337 }
1338
1339 int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1340 {
1341 struct commit_extra_header *extra, *to_free;
1342 int res = 0;
1343
1344 to_free = read_commit_extra_headers(commit, NULL);
1345 for (extra = to_free; !res && extra; extra = extra->next) {
1346 if (strcmp(extra->key, "mergetag"))
1347 continue; /* not a merge tag */
1348 res = fn(commit, extra, data);
1349 }
1350 free_commit_extra_headers(to_free);
1351 return res;
1352 }
1353
1354 static inline int standard_header_field(const char *field, size_t len)
1355 {
1356 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1357 (len == 6 && !memcmp(field, "parent", 6)) ||
1358 (len == 6 && !memcmp(field, "author", 6)) ||
1359 (len == 9 && !memcmp(field, "committer", 9)) ||
1360 (len == 8 && !memcmp(field, "encoding", 8)));
1361 }
1362
1363 static int excluded_header_field(const char *field, size_t len, const char **exclude)
1364 {
1365 if (!exclude)
1366 return 0;
1367
1368 while (*exclude) {
1369 size_t xlen = strlen(*exclude);
1370 if (len == xlen && !memcmp(field, *exclude, xlen))
1371 return 1;
1372 exclude++;
1373 }
1374 return 0;
1375 }
1376
1377 static struct commit_extra_header *read_commit_extra_header_lines(
1378 const char *buffer, size_t size,
1379 const char **exclude)
1380 {
1381 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1382 const char *line, *next, *eof, *eob;
1383 struct strbuf buf = STRBUF_INIT;
1384
1385 for (line = buffer, eob = line + size;
1386 line < eob && *line != '\n';
1387 line = next) {
1388 next = memchr(line, '\n', eob - line);
1389 next = next ? next + 1 : eob;
1390 if (*line == ' ') {
1391 /* continuation */
1392 if (it)
1393 strbuf_add(&buf, line + 1, next - (line + 1));
1394 continue;
1395 }
1396 if (it)
1397 it->value = strbuf_detach(&buf, &it->len);
1398 strbuf_reset(&buf);
1399 it = NULL;
1400
1401 eof = memchr(line, ' ', next - line);
1402 if (!eof)
1403 eof = next;
1404 else if (standard_header_field(line, eof - line) ||
1405 excluded_header_field(line, eof - line, exclude))
1406 continue;
1407
1408 CALLOC_ARRAY(it, 1);
1409 it->key = xmemdupz(line, eof-line);
1410 *tail = it;
1411 tail = &it->next;
1412 if (eof + 1 < next)
1413 strbuf_add(&buf, eof + 1, next - (eof + 1));
1414 }
1415 if (it)
1416 it->value = strbuf_detach(&buf, &it->len);
1417 return extra;
1418 }
1419
1420 void free_commit_extra_headers(struct commit_extra_header *extra)
1421 {
1422 while (extra) {
1423 struct commit_extra_header *next = extra->next;
1424 free(extra->key);
1425 free(extra->value);
1426 free(extra);
1427 extra = next;
1428 }
1429 }
1430
1431 int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1432 struct commit_list *parents, struct object_id *ret,
1433 const char *author, const char *sign_commit)
1434 {
1435 struct commit_extra_header *extra = NULL, **tail = &extra;
1436 int result;
1437
1438 append_merge_tag_headers(parents, &tail);
1439 result = commit_tree_extended(msg, msg_len, tree, parents, ret, author,
1440 NULL, sign_commit, extra);
1441 free_commit_extra_headers(extra);
1442 return result;
1443 }
1444
1445 static int find_invalid_utf8(const char *buf, int len)
1446 {
1447 int offset = 0;
1448 static const unsigned int max_codepoint[] = {
1449 0x7f, 0x7ff, 0xffff, 0x10ffff
1450 };
1451
1452 while (len) {
1453 unsigned char c = *buf++;
1454 int bytes, bad_offset;
1455 unsigned int codepoint;
1456 unsigned int min_val, max_val;
1457
1458 len--;
1459 offset++;
1460
1461 /* Simple US-ASCII? No worries. */
1462 if (c < 0x80)
1463 continue;
1464
1465 bad_offset = offset-1;
1466
1467 /*
1468 * Count how many more high bits set: that's how
1469 * many more bytes this sequence should have.
1470 */
1471 bytes = 0;
1472 while (c & 0x40) {
1473 c <<= 1;
1474 bytes++;
1475 }
1476
1477 /*
1478 * Must be between 1 and 3 more bytes. Longer sequences result in
1479 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1480 */
1481 if (bytes < 1 || 3 < bytes)
1482 return bad_offset;
1483
1484 /* Do we *have* that many bytes? */
1485 if (len < bytes)
1486 return bad_offset;
1487
1488 /*
1489 * Place the encoded bits at the bottom of the value and compute the
1490 * valid range.
1491 */
1492 codepoint = (c & 0x7f) >> bytes;
1493 min_val = max_codepoint[bytes-1] + 1;
1494 max_val = max_codepoint[bytes];
1495
1496 offset += bytes;
1497 len -= bytes;
1498
1499 /* And verify that they are good continuation bytes */
1500 do {
1501 codepoint <<= 6;
1502 codepoint |= *buf & 0x3f;
1503 if ((*buf++ & 0xc0) != 0x80)
1504 return bad_offset;
1505 } while (--bytes);
1506
1507 /* Reject codepoints that are out of range for the sequence length. */
1508 if (codepoint < min_val || codepoint > max_val)
1509 return bad_offset;
1510 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1511 if ((codepoint & 0x1ff800) == 0xd800)
1512 return bad_offset;
1513 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1514 if ((codepoint & 0xfffe) == 0xfffe)
1515 return bad_offset;
1516 /* So are anything in the range U+FDD0..U+FDEF. */
1517 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1518 return bad_offset;
1519 }
1520 return -1;
1521 }
1522
1523 /*
1524 * This verifies that the buffer is in proper utf8 format.
1525 *
1526 * If it isn't, it assumes any non-utf8 characters are Latin1,
1527 * and does the conversion.
1528 */
1529 static int verify_utf8(struct strbuf *buf)
1530 {
1531 int ok = 1;
1532 long pos = 0;
1533
1534 for (;;) {
1535 int bad;
1536 unsigned char c;
1537 unsigned char replace[2];
1538
1539 bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1540 if (bad < 0)
1541 return ok;
1542 pos += bad;
1543 ok = 0;
1544 c = buf->buf[pos];
1545 strbuf_remove(buf, pos, 1);
1546
1547 /* We know 'c' must be in the range 128-255 */
1548 replace[0] = 0xc0 + (c >> 6);
1549 replace[1] = 0x80 + (c & 0x3f);
1550 strbuf_insert(buf, pos, replace, 2);
1551 pos += 2;
1552 }
1553 }
1554
1555 static const char commit_utf8_warn[] =
1556 N_("Warning: commit message did not conform to UTF-8.\n"
1557 "You may want to amend it after fixing the message, or set the config\n"
1558 "variable i18n.commitEncoding to the encoding your project uses.\n");
1559
1560 int commit_tree_extended(const char *msg, size_t msg_len,
1561 const struct object_id *tree,
1562 struct commit_list *parents, struct object_id *ret,
1563 const char *author, const char *committer,
1564 const char *sign_commit,
1565 struct commit_extra_header *extra)
1566 {
1567 int result;
1568 int encoding_is_utf8;
1569 struct strbuf buffer;
1570
1571 assert_oid_type(tree, OBJ_TREE);
1572
1573 if (memchr(msg, '\0', msg_len))
1574 return error("a NUL byte in commit log message not allowed.");
1575
1576 /* Not having i18n.commitencoding is the same as having utf-8 */
1577 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1578
1579 strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1580 strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1581
1582 /*
1583 * NOTE! This ordering means that the same exact tree merged with a
1584 * different order of parents will be a _different_ changeset even
1585 * if everything else stays the same.
1586 */
1587 while (parents) {
1588 struct commit *parent = pop_commit(&parents);
1589 strbuf_addf(&buffer, "parent %s\n",
1590 oid_to_hex(&parent->object.oid));
1591 }
1592
1593 /* Person/date information */
1594 if (!author)
1595 author = git_author_info(IDENT_STRICT);
1596 strbuf_addf(&buffer, "author %s\n", author);
1597 if (!committer)
1598 committer = git_committer_info(IDENT_STRICT);
1599 strbuf_addf(&buffer, "committer %s\n", committer);
1600 if (!encoding_is_utf8)
1601 strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1602
1603 while (extra) {
1604 add_extra_header(&buffer, extra);
1605 extra = extra->next;
1606 }
1607 strbuf_addch(&buffer, '\n');
1608
1609 /* And add the comment */
1610 strbuf_add(&buffer, msg, msg_len);
1611
1612 /* And check the encoding */
1613 if (encoding_is_utf8 && !verify_utf8(&buffer))
1614 fprintf(stderr, _(commit_utf8_warn));
1615
1616 if (sign_commit && sign_with_header(&buffer, sign_commit)) {
1617 result = -1;
1618 goto out;
1619 }
1620
1621 result = write_object_file(buffer.buf, buffer.len, OBJ_COMMIT, ret);
1622 out:
1623 strbuf_release(&buffer);
1624 return result;
1625 }
1626
1627 define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1628 static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1629
1630 struct merge_remote_desc *merge_remote_util(struct commit *commit)
1631 {
1632 return *merge_desc_slab_at(&merge_desc_slab, commit);
1633 }
1634
1635 void set_merge_remote_desc(struct commit *commit,
1636 const char *name, struct object *obj)
1637 {
1638 struct merge_remote_desc *desc;
1639 FLEX_ALLOC_STR(desc, name, name);
1640 desc->obj = obj;
1641 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1642 }
1643
1644 struct commit *get_merge_parent(const char *name)
1645 {
1646 struct object *obj;
1647 struct commit *commit;
1648 struct object_id oid;
1649 if (repo_get_oid(the_repository, name, &oid))
1650 return NULL;
1651 obj = parse_object(the_repository, &oid);
1652 commit = (struct commit *)repo_peel_to_type(the_repository, name, 0,
1653 obj, OBJ_COMMIT);
1654 if (commit && !merge_remote_util(commit))
1655 set_merge_remote_desc(commit, name, obj);
1656 return commit;
1657 }
1658
1659 /*
1660 * Append a commit to the end of the commit_list.
1661 *
1662 * next starts by pointing to the variable that holds the head of an
1663 * empty commit_list, and is updated to point to the "next" field of
1664 * the last item on the list as new commits are appended.
1665 *
1666 * Usage example:
1667 *
1668 * struct commit_list *list;
1669 * struct commit_list **next = &list;
1670 *
1671 * next = commit_list_append(c1, next);
1672 * next = commit_list_append(c2, next);
1673 * assert(commit_list_count(list) == 2);
1674 * return list;
1675 */
1676 struct commit_list **commit_list_append(struct commit *commit,
1677 struct commit_list **next)
1678 {
1679 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1680 new_commit->item = commit;
1681 *next = new_commit;
1682 new_commit->next = NULL;
1683 return &new_commit->next;
1684 }
1685
1686 const char *find_header_mem(const char *msg, size_t len,
1687 const char *key, size_t *out_len)
1688 {
1689 int key_len = strlen(key);
1690 const char *line = msg;
1691
1692 /*
1693 * NEEDSWORK: It's possible for strchrnul() to scan beyond the range
1694 * given by len. However, current callers are safe because they compute
1695 * len by scanning a NUL-terminated block of memory starting at msg.
1696 * Nonetheless, it would be better to ensure the function does not look
1697 * at msg beyond the len provided by the caller.
1698 */
1699 while (line && line < msg + len) {
1700 const char *eol = strchrnul(line, '\n');
1701
1702 if (line == eol)
1703 return NULL;
1704
1705 if (eol - line > key_len &&
1706 !strncmp(line, key, key_len) &&
1707 line[key_len] == ' ') {
1708 *out_len = eol - line - key_len - 1;
1709 return line + key_len + 1;
1710 }
1711 line = *eol ? eol + 1 : NULL;
1712 }
1713 return NULL;
1714 }
1715
1716 const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1717 {
1718 return find_header_mem(msg, strlen(msg), key, out_len);
1719 }
1720 /*
1721 * Inspect the given string and determine the true "end" of the log message, in
1722 * order to find where to put a new Signed-off-by trailer. Ignored are
1723 * trailing comment lines and blank lines. To support "git commit -s
1724 * --amend" on an existing commit, we also ignore "Conflicts:". To
1725 * support "git commit -v", we truncate at cut lines.
1726 *
1727 * Returns the number of bytes from the tail to ignore, to be fed as
1728 * the second parameter to append_signoff().
1729 */
1730 size_t ignore_non_trailer(const char *buf, size_t len)
1731 {
1732 size_t boc = 0;
1733 size_t bol = 0;
1734 int in_old_conflicts_block = 0;
1735 size_t cutoff = wt_status_locate_end(buf, len);
1736
1737 while (bol < cutoff) {
1738 const char *next_line = memchr(buf + bol, '\n', len - bol);
1739
1740 if (!next_line)
1741 next_line = buf + len;
1742 else
1743 next_line++;
1744
1745 if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1746 /* is this the first of the run of comments? */
1747 if (!boc)
1748 boc = bol;
1749 /* otherwise, it is just continuing */
1750 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1751 in_old_conflicts_block = 1;
1752 if (!boc)
1753 boc = bol;
1754 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1755 ; /* a pathname in the conflicts block */
1756 } else if (boc) {
1757 /* the previous was not trailing comment */
1758 boc = 0;
1759 in_old_conflicts_block = 0;
1760 }
1761 bol = next_line - buf;
1762 }
1763 return boc ? len - boc : len - cutoff;
1764 }
1765
1766 int run_commit_hook(int editor_is_used, const char *index_file,
1767 int *invoked_hook, const char *name, ...)
1768 {
1769 struct run_hooks_opt opt = RUN_HOOKS_OPT_INIT;
1770 va_list args;
1771 const char *arg;
1772
1773 strvec_pushf(&opt.env, "GIT_INDEX_FILE=%s", index_file);
1774
1775 /*
1776 * Let the hook know that no editor will be launched.
1777 */
1778 if (!editor_is_used)
1779 strvec_push(&opt.env, "GIT_EDITOR=:");
1780
1781 va_start(args, name);
1782 while ((arg = va_arg(args, const char *)))
1783 strvec_push(&opt.args, arg);
1784 va_end(args);
1785
1786 opt.invoked_hook = invoked_hook;
1787 return run_hooks_opt(name, &opt);
1788 }