1#include "cache.h"
2#include "tag.h"
3#include "commit.h"
4#include "commit-graph.h"
5#include "repository.h"
6#include "object-store.h"
7#include "pkt-line.h"
8#include "utf8.h"
9#include "diff.h"
10#include "revision.h"
11#include "notes.h"
12#include "alloc.h"
13#include "gpg-interface.h"
14#include "mergesort.h"
15#include "commit-slab.h"
16#include "prio-queue.h"
17#include "sha1-lookup.h"
18#include "wt-status.h"
19#include "advice.h"
20
21static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
22
23int save_commit_buffer = 1;
24
25const char *commit_type = "commit";
26
27struct commit *lookup_commit_reference_gently(struct repository *r,
28 const struct object_id *oid, int quiet)
29{
30 struct object *obj = deref_tag(r,
31 parse_object(r, oid),
32 NULL, 0);
33
34 if (!obj)
35 return NULL;
36 return object_as_type(r, obj, OBJ_COMMIT, quiet);
37}
38
39struct commit *lookup_commit_reference(struct repository *r, const struct object_id *oid)
40{
41 return lookup_commit_reference_gently(r, oid, 0);
42}
43
44struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
45{
46 struct commit *c = lookup_commit_reference(the_repository, oid);
47 if (!c)
48 die(_("could not parse %s"), ref_name);
49 if (!oideq(oid, &c->object.oid)) {
50 warning(_("%s %s is not a commit!"),
51 ref_name, oid_to_hex(oid));
52 }
53 return c;
54}
55
56struct commit *lookup_commit(struct repository *r, const struct object_id *oid)
57{
58 struct object *obj = lookup_object(r, oid->hash);
59 if (!obj)
60 return create_object(r, oid->hash,
61 alloc_commit_node(r));
62 return object_as_type(r, obj, OBJ_COMMIT, 0);
63}
64
65struct commit *lookup_commit_reference_by_name(const char *name)
66{
67 struct object_id oid;
68 struct commit *commit;
69
70 if (get_oid_committish(name, &oid))
71 return NULL;
72 commit = lookup_commit_reference(the_repository, &oid);
73 if (parse_commit(commit))
74 return NULL;
75 return commit;
76}
77
78static timestamp_t parse_commit_date(const char *buf, const char *tail)
79{
80 const char *dateptr;
81
82 if (buf + 6 >= tail)
83 return 0;
84 if (memcmp(buf, "author", 6))
85 return 0;
86 while (buf < tail && *buf++ != '\n')
87 /* nada */;
88 if (buf + 9 >= tail)
89 return 0;
90 if (memcmp(buf, "committer", 9))
91 return 0;
92 while (buf < tail && *buf++ != '>')
93 /* nada */;
94 if (buf >= tail)
95 return 0;
96 dateptr = buf;
97 while (buf < tail && *buf++ != '\n')
98 /* nada */;
99 if (buf >= tail)
100 return 0;
101 /* dateptr < buf && buf[-1] == '\n', so parsing will stop at buf-1 */
102 return parse_timestamp(dateptr, NULL, 10);
103}
104
105static const unsigned char *commit_graft_sha1_access(size_t index, void *table)
106{
107 struct commit_graft **commit_graft_table = table;
108 return commit_graft_table[index]->oid.hash;
109}
110
111static int commit_graft_pos(struct repository *r, const unsigned char *sha1)
112{
113 return sha1_pos(sha1, r->parsed_objects->grafts,
114 r->parsed_objects->grafts_nr,
115 commit_graft_sha1_access);
116}
117
118int register_commit_graft(struct repository *r, struct commit_graft *graft,
119 int ignore_dups)
120{
121 int pos = commit_graft_pos(r, graft->oid.hash);
122
123 if (0 <= pos) {
124 if (ignore_dups)
125 free(graft);
126 else {
127 free(r->parsed_objects->grafts[pos]);
128 r->parsed_objects->grafts[pos] = graft;
129 }
130 return 1;
131 }
132 pos = -pos - 1;
133 ALLOC_GROW(r->parsed_objects->grafts,
134 r->parsed_objects->grafts_nr + 1,
135 r->parsed_objects->grafts_alloc);
136 r->parsed_objects->grafts_nr++;
137 if (pos < r->parsed_objects->grafts_nr)
138 memmove(r->parsed_objects->grafts + pos + 1,
139 r->parsed_objects->grafts + pos,
140 (r->parsed_objects->grafts_nr - pos - 1) *
141 sizeof(*r->parsed_objects->grafts));
142 r->parsed_objects->grafts[pos] = graft;
143 return 0;
144}
145
146struct commit_graft *read_graft_line(struct strbuf *line)
147{
148 /* The format is just "Commit Parent1 Parent2 ...\n" */
149 int i, phase;
150 const char *tail = NULL;
151 struct commit_graft *graft = NULL;
152 struct object_id dummy_oid, *oid;
153
154 strbuf_rtrim(line);
155 if (!line->len || line->buf[0] == '#')
156 return NULL;
157 /*
158 * phase 0 verifies line, counts hashes in line and allocates graft
159 * phase 1 fills graft
160 */
161 for (phase = 0; phase < 2; phase++) {
162 oid = graft ? &graft->oid : &dummy_oid;
163 if (parse_oid_hex(line->buf, oid, &tail))
164 goto bad_graft_data;
165 for (i = 0; *tail != '\0'; i++) {
166 oid = graft ? &graft->parent[i] : &dummy_oid;
167 if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail))
168 goto bad_graft_data;
169 }
170 if (!graft) {
171 graft = xmalloc(st_add(sizeof(*graft),
172 st_mult(sizeof(struct object_id), i)));
173 graft->nr_parent = i;
174 }
175 }
176 return graft;
177
178bad_graft_data:
179 error("bad graft data: %s", line->buf);
180 assert(!graft);
181 return NULL;
182}
183
184static int read_graft_file(struct repository *r, const char *graft_file)
185{
186 FILE *fp = fopen_or_warn(graft_file, "r");
187 struct strbuf buf = STRBUF_INIT;
188 if (!fp)
189 return -1;
190 if (advice_graft_file_deprecated)
191 advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n"
192 "and will be removed in a future Git version.\n"
193 "\n"
194 "Please use \"git replace --convert-graft-file\"\n"
195 "to convert the grafts into replace refs.\n"
196 "\n"
197 "Turn this message off by running\n"
198 "\"git config advice.graftFileDeprecated false\""));
199 while (!strbuf_getwholeline(&buf, fp, '\n')) {
200 /* The format is just "Commit Parent1 Parent2 ...\n" */
201 struct commit_graft *graft = read_graft_line(&buf);
202 if (!graft)
203 continue;
204 if (register_commit_graft(r, graft, 1))
205 error("duplicate graft data: %s", buf.buf);
206 }
207 fclose(fp);
208 strbuf_release(&buf);
209 return 0;
210}
211
212void prepare_commit_graft(struct repository *r)
213{
214 char *graft_file;
215
216 if (r->parsed_objects->commit_graft_prepared)
217 return;
218 if (!startup_info->have_repository)
219 return;
220
221 graft_file = get_graft_file(r);
222 read_graft_file(r, graft_file);
223 /* make sure shallows are read */
224 is_repository_shallow(r);
225 r->parsed_objects->commit_graft_prepared = 1;
226}
227
228struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid)
229{
230 int pos;
231 prepare_commit_graft(r);
232 pos = commit_graft_pos(r, oid->hash);
233 if (pos < 0)
234 return NULL;
235 return r->parsed_objects->grafts[pos];
236}
237
238int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
239{
240 int i, ret;
241 for (i = ret = 0; i < the_repository->parsed_objects->grafts_nr && !ret; i++)
242 ret = fn(the_repository->parsed_objects->grafts[i], cb_data);
243 return ret;
244}
245
246int unregister_shallow(const struct object_id *oid)
247{
248 int pos = commit_graft_pos(the_repository, oid->hash);
249 if (pos < 0)
250 return -1;
251 if (pos + 1 < the_repository->parsed_objects->grafts_nr)
252 MOVE_ARRAY(the_repository->parsed_objects->grafts + pos,
253 the_repository->parsed_objects->grafts + pos + 1,
254 the_repository->parsed_objects->grafts_nr - pos - 1);
255 the_repository->parsed_objects->grafts_nr--;
256 return 0;
257}
258
259struct commit_buffer {
260 void *buffer;
261 unsigned long size;
262};
263define_commit_slab(buffer_slab, struct commit_buffer);
264
265struct buffer_slab *allocate_commit_buffer_slab(void)
266{
267 struct buffer_slab *bs = xmalloc(sizeof(*bs));
268 init_buffer_slab(bs);
269 return bs;
270}
271
272void free_commit_buffer_slab(struct buffer_slab *bs)
273{
274 clear_buffer_slab(bs);
275 free(bs);
276}
277
278void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size)
279{
280 struct commit_buffer *v = buffer_slab_at(
281 r->parsed_objects->buffer_slab, commit);
282 v->buffer = buffer;
283 v->size = size;
284}
285
286const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep)
287{
288 struct commit_buffer *v = buffer_slab_peek(
289 r->parsed_objects->buffer_slab, commit);
290 if (!v) {
291 if (sizep)
292 *sizep = 0;
293 return NULL;
294 }
295 if (sizep)
296 *sizep = v->size;
297 return v->buffer;
298}
299
300const void *repo_get_commit_buffer(struct repository *r,
301 const struct commit *commit,
302 unsigned long *sizep)
303{
304 const void *ret = get_cached_commit_buffer(r, commit, sizep);
305 if (!ret) {
306 enum object_type type;
307 unsigned long size;
308 ret = repo_read_object_file(r, &commit->object.oid, &type, &size);
309 if (!ret)
310 die("cannot read commit object %s",
311 oid_to_hex(&commit->object.oid));
312 if (type != OBJ_COMMIT)
313 die("expected commit for %s, got %s",
314 oid_to_hex(&commit->object.oid), type_name(type));
315 if (sizep)
316 *sizep = size;
317 }
318 return ret;
319}
320
321void repo_unuse_commit_buffer(struct repository *r,
322 const struct commit *commit,
323 const void *buffer)
324{
325 struct commit_buffer *v = buffer_slab_peek(
326 r->parsed_objects->buffer_slab, commit);
327 if (!(v && v->buffer == buffer))
328 free((void *)buffer);
329}
330
331void free_commit_buffer(struct parsed_object_pool *pool, struct commit *commit)
332{
333 struct commit_buffer *v = buffer_slab_peek(
334 pool->buffer_slab, commit);
335 if (v) {
336 FREE_AND_NULL(v->buffer);
337 v->size = 0;
338 }
339}
340
341struct tree *get_commit_tree(const struct commit *commit)
342{
343 if (commit->maybe_tree || !commit->object.parsed)
344 return commit->maybe_tree;
345
346 if (commit->graph_pos == COMMIT_NOT_FROM_GRAPH)
347 BUG("commit has NULL tree, but was not loaded from commit-graph");
348
349 return get_commit_tree_in_graph(the_repository, commit);
350}
351
352struct object_id *get_commit_tree_oid(const struct commit *commit)
353{
354 return &get_commit_tree(commit)->object.oid;
355}
356
357void release_commit_memory(struct parsed_object_pool *pool, struct commit *c)
358{
359 c->maybe_tree = NULL;
360 c->index = 0;
361 free_commit_buffer(pool, c);
362 free_commit_list(c->parents);
363
364 c->object.parsed = 0;
365}
366
367const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
368{
369 struct commit_buffer *v = buffer_slab_peek(
370 the_repository->parsed_objects->buffer_slab, commit);
371 void *ret;
372
373 if (!v) {
374 if (sizep)
375 *sizep = 0;
376 return NULL;
377 }
378 ret = v->buffer;
379 if (sizep)
380 *sizep = v->size;
381
382 v->buffer = NULL;
383 v->size = 0;
384 return ret;
385}
386
387int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
388{
389 const char *tail = buffer;
390 const char *bufptr = buffer;
391 struct object_id parent;
392 struct commit_list **pptr;
393 struct commit_graft *graft;
394 const int tree_entry_len = the_hash_algo->hexsz + 5;
395 const int parent_entry_len = the_hash_algo->hexsz + 7;
396
397 if (item->object.parsed)
398 return 0;
399 item->object.parsed = 1;
400 tail += size;
401 if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
402 bufptr[tree_entry_len] != '\n')
403 return error("bogus commit object %s", oid_to_hex(&item->object.oid));
404 if (get_oid_hex(bufptr + 5, &parent) < 0)
405 return error("bad tree pointer in commit %s",
406 oid_to_hex(&item->object.oid));
407 item->maybe_tree = lookup_tree(r, &parent);
408 bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
409 pptr = &item->parents;
410
411 graft = lookup_commit_graft(r, &item->object.oid);
412 while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
413 struct commit *new_parent;
414
415 if (tail <= bufptr + parent_entry_len + 1 ||
416 get_oid_hex(bufptr + 7, &parent) ||
417 bufptr[parent_entry_len] != '\n')
418 return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
419 bufptr += parent_entry_len + 1;
420 /*
421 * The clone is shallow if nr_parent < 0, and we must
422 * not traverse its real parents even when we unhide them.
423 */
424 if (graft && (graft->nr_parent < 0 || grafts_replace_parents))
425 continue;
426 new_parent = lookup_commit(r, &parent);
427 if (new_parent)
428 pptr = &commit_list_insert(new_parent, pptr)->next;
429 }
430 if (graft) {
431 int i;
432 struct commit *new_parent;
433 for (i = 0; i < graft->nr_parent; i++) {
434 new_parent = lookup_commit(r,
435 &graft->parent[i]);
436 if (!new_parent)
437 continue;
438 pptr = &commit_list_insert(new_parent, pptr)->next;
439 }
440 }
441 item->date = parse_commit_date(bufptr, tail);
442
443 if (check_graph)
444 load_commit_graph_info(the_repository, item);
445
446 return 0;
447}
448
449int repo_parse_commit_internal(struct repository *r,
450 struct commit *item,
451 int quiet_on_missing,
452 int use_commit_graph)
453{
454 enum object_type type;
455 void *buffer;
456 unsigned long size;
457 int ret;
458
459 if (!item)
460 return -1;
461 if (item->object.parsed)
462 return 0;
463 if (use_commit_graph && parse_commit_in_graph(r, item))
464 return 0;
465 buffer = repo_read_object_file(r, &item->object.oid, &type, &size);
466 if (!buffer)
467 return quiet_on_missing ? -1 :
468 error("Could not read %s",
469 oid_to_hex(&item->object.oid));
470 if (type != OBJ_COMMIT) {
471 free(buffer);
472 return error("Object %s not a commit",
473 oid_to_hex(&item->object.oid));
474 }
475
476 ret = parse_commit_buffer(r, item, buffer, size, 0);
477 if (save_commit_buffer && !ret) {
478 set_commit_buffer(r, item, buffer, size);
479 return 0;
480 }
481 free(buffer);
482 return ret;
483}
484
485int repo_parse_commit_gently(struct repository *r,
486 struct commit *item, int quiet_on_missing)
487{
488 return repo_parse_commit_internal(r, item, quiet_on_missing, 1);
489}
490
491void parse_commit_or_die(struct commit *item)
492{
493 if (parse_commit(item))
494 die("unable to parse commit %s",
495 item ? oid_to_hex(&item->object.oid) : "(null)");
496}
497
498int find_commit_subject(const char *commit_buffer, const char **subject)
499{
500 const char *eol;
501 const char *p = commit_buffer;
502
503 while (*p && (*p != '\n' || p[1] != '\n'))
504 p++;
505 if (*p) {
506 p = skip_blank_lines(p + 2);
507 eol = strchrnul(p, '\n');
508 } else
509 eol = p;
510
511 *subject = p;
512
513 return eol - p;
514}
515
516struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
517{
518 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
519 new_list->item = item;
520 new_list->next = *list_p;
521 *list_p = new_list;
522 return new_list;
523}
524
525unsigned commit_list_count(const struct commit_list *l)
526{
527 unsigned c = 0;
528 for (; l; l = l->next )
529 c++;
530 return c;
531}
532
533struct commit_list *copy_commit_list(struct commit_list *list)
534{
535 struct commit_list *head = NULL;
536 struct commit_list **pp = &head;
537 while (list) {
538 pp = commit_list_append(list->item, pp);
539 list = list->next;
540 }
541 return head;
542}
543
544void free_commit_list(struct commit_list *list)
545{
546 while (list)
547 pop_commit(&list);
548}
549
550struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
551{
552 struct commit_list **pp = list;
553 struct commit_list *p;
554 while ((p = *pp) != NULL) {
555 if (p->item->date < item->date) {
556 break;
557 }
558 pp = &p->next;
559 }
560 return commit_list_insert(item, pp);
561}
562
563static int commit_list_compare_by_date(const void *a, const void *b)
564{
565 timestamp_t a_date = ((const struct commit_list *)a)->item->date;
566 timestamp_t b_date = ((const struct commit_list *)b)->item->date;
567 if (a_date < b_date)
568 return 1;
569 if (a_date > b_date)
570 return -1;
571 return 0;
572}
573
574static void *commit_list_get_next(const void *a)
575{
576 return ((const struct commit_list *)a)->next;
577}
578
579static void commit_list_set_next(void *a, void *next)
580{
581 ((struct commit_list *)a)->next = next;
582}
583
584void commit_list_sort_by_date(struct commit_list **list)
585{
586 *list = llist_mergesort(*list, commit_list_get_next, commit_list_set_next,
587 commit_list_compare_by_date);
588}
589
590struct commit *pop_most_recent_commit(struct commit_list **list,
591 unsigned int mark)
592{
593 struct commit *ret = pop_commit(list);
594 struct commit_list *parents = ret->parents;
595
596 while (parents) {
597 struct commit *commit = parents->item;
598 if (!parse_commit(commit) && !(commit->object.flags & mark)) {
599 commit->object.flags |= mark;
600 commit_list_insert_by_date(commit, list);
601 }
602 parents = parents->next;
603 }
604 return ret;
605}
606
607static void clear_commit_marks_1(struct commit_list **plist,
608 struct commit *commit, unsigned int mark)
609{
610 while (commit) {
611 struct commit_list *parents;
612
613 if (!(mark & commit->object.flags))
614 return;
615
616 commit->object.flags &= ~mark;
617
618 parents = commit->parents;
619 if (!parents)
620 return;
621
622 while ((parents = parents->next))
623 commit_list_insert(parents->item, plist);
624
625 commit = commit->parents->item;
626 }
627}
628
629void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
630{
631 struct commit_list *list = NULL;
632
633 while (nr--) {
634 clear_commit_marks_1(&list, *commit, mark);
635 commit++;
636 }
637 while (list)
638 clear_commit_marks_1(&list, pop_commit(&list), mark);
639}
640
641void clear_commit_marks(struct commit *commit, unsigned int mark)
642{
643 clear_commit_marks_many(1, &commit, mark);
644}
645
646struct commit *pop_commit(struct commit_list **stack)
647{
648 struct commit_list *top = *stack;
649 struct commit *item = top ? top->item : NULL;
650
651 if (top) {
652 *stack = top->next;
653 free(top);
654 }
655 return item;
656}
657
658/*
659 * Topological sort support
660 */
661
662/* count number of children that have not been emitted */
663define_commit_slab(indegree_slab, int);
664
665/* record author-date for each commit object */
666define_commit_slab(author_date_slab, timestamp_t);
667
668static void record_author_date(struct author_date_slab *author_date,
669 struct commit *commit)
670{
671 const char *buffer = get_commit_buffer(commit, NULL);
672 struct ident_split ident;
673 const char *ident_line;
674 size_t ident_len;
675 char *date_end;
676 timestamp_t date;
677
678 ident_line = find_commit_header(buffer, "author", &ident_len);
679 if (!ident_line)
680 goto fail_exit; /* no author line */
681 if (split_ident_line(&ident, ident_line, ident_len) ||
682 !ident.date_begin || !ident.date_end)
683 goto fail_exit; /* malformed "author" line */
684
685 date = parse_timestamp(ident.date_begin, &date_end, 10);
686 if (date_end != ident.date_end)
687 goto fail_exit; /* malformed date */
688 *(author_date_slab_at(author_date, commit)) = date;
689
690fail_exit:
691 unuse_commit_buffer(commit, buffer);
692}
693
694static int compare_commits_by_author_date(const void *a_, const void *b_,
695 void *cb_data)
696{
697 const struct commit *a = a_, *b = b_;
698 struct author_date_slab *author_date = cb_data;
699 timestamp_t a_date = *(author_date_slab_at(author_date, a));
700 timestamp_t b_date = *(author_date_slab_at(author_date, b));
701
702 /* newer commits with larger date first */
703 if (a_date < b_date)
704 return 1;
705 else if (a_date > b_date)
706 return -1;
707 return 0;
708}
709
710int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, void *unused)
711{
712 const struct commit *a = a_, *b = b_;
713
714 /* newer commits first */
715 if (a->generation < b->generation)
716 return 1;
717 else if (a->generation > b->generation)
718 return -1;
719
720 /* use date as a heuristic when generations are equal */
721 if (a->date < b->date)
722 return 1;
723 else if (a->date > b->date)
724 return -1;
725 return 0;
726}
727
728int compare_commits_by_commit_date(const void *a_, const void *b_, void *unused)
729{
730 const struct commit *a = a_, *b = b_;
731 /* newer commits with larger date first */
732 if (a->date < b->date)
733 return 1;
734 else if (a->date > b->date)
735 return -1;
736 return 0;
737}
738
739/*
740 * Performs an in-place topological sort on the list supplied.
741 */
742void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
743{
744 struct commit_list *next, *orig = *list;
745 struct commit_list **pptr;
746 struct indegree_slab indegree;
747 struct prio_queue queue;
748 struct commit *commit;
749 struct author_date_slab author_date;
750
751 if (!orig)
752 return;
753 *list = NULL;
754
755 init_indegree_slab(&indegree);
756 memset(&queue, '\0', sizeof(queue));
757
758 switch (sort_order) {
759 default: /* REV_SORT_IN_GRAPH_ORDER */
760 queue.compare = NULL;
761 break;
762 case REV_SORT_BY_COMMIT_DATE:
763 queue.compare = compare_commits_by_commit_date;
764 break;
765 case REV_SORT_BY_AUTHOR_DATE:
766 init_author_date_slab(&author_date);
767 queue.compare = compare_commits_by_author_date;
768 queue.cb_data = &author_date;
769 break;
770 }
771
772 /* Mark them and clear the indegree */
773 for (next = orig; next; next = next->next) {
774 struct commit *commit = next->item;
775 *(indegree_slab_at(&indegree, commit)) = 1;
776 /* also record the author dates, if needed */
777 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
778 record_author_date(&author_date, commit);
779 }
780
781 /* update the indegree */
782 for (next = orig; next; next = next->next) {
783 struct commit_list *parents = next->item->parents;
784 while (parents) {
785 struct commit *parent = parents->item;
786 int *pi = indegree_slab_at(&indegree, parent);
787
788 if (*pi)
789 (*pi)++;
790 parents = parents->next;
791 }
792 }
793
794 /*
795 * find the tips
796 *
797 * tips are nodes not reachable from any other node in the list
798 *
799 * the tips serve as a starting set for the work queue.
800 */
801 for (next = orig; next; next = next->next) {
802 struct commit *commit = next->item;
803
804 if (*(indegree_slab_at(&indegree, commit)) == 1)
805 prio_queue_put(&queue, commit);
806 }
807
808 /*
809 * This is unfortunate; the initial tips need to be shown
810 * in the order given from the revision traversal machinery.
811 */
812 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
813 prio_queue_reverse(&queue);
814
815 /* We no longer need the commit list */
816 free_commit_list(orig);
817
818 pptr = list;
819 *list = NULL;
820 while ((commit = prio_queue_get(&queue)) != NULL) {
821 struct commit_list *parents;
822
823 for (parents = commit->parents; parents ; parents = parents->next) {
824 struct commit *parent = parents->item;
825 int *pi = indegree_slab_at(&indegree, parent);
826
827 if (!*pi)
828 continue;
829
830 /*
831 * parents are only enqueued for emission
832 * when all their children have been emitted thereby
833 * guaranteeing topological order.
834 */
835 if (--(*pi) == 1)
836 prio_queue_put(&queue, parent);
837 }
838 /*
839 * all children of commit have already been
840 * emitted. we can emit it now.
841 */
842 *(indegree_slab_at(&indegree, commit)) = 0;
843
844 pptr = &commit_list_insert(commit, pptr)->next;
845 }
846
847 clear_indegree_slab(&indegree);
848 clear_prio_queue(&queue);
849 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
850 clear_author_date_slab(&author_date);
851}
852
853static const char gpg_sig_header[] = "gpgsig";
854static const int gpg_sig_header_len = sizeof(gpg_sig_header) - 1;
855
856static int do_sign_commit(struct strbuf *buf, const char *keyid)
857{
858 struct strbuf sig = STRBUF_INIT;
859 int inspos, copypos;
860 const char *eoh;
861
862 /* find the end of the header */
863 eoh = strstr(buf->buf, "\n\n");
864 if (!eoh)
865 inspos = buf->len;
866 else
867 inspos = eoh - buf->buf + 1;
868
869 if (!keyid || !*keyid)
870 keyid = get_signing_key();
871 if (sign_buffer(buf, &sig, keyid)) {
872 strbuf_release(&sig);
873 return -1;
874 }
875
876 for (copypos = 0; sig.buf[copypos]; ) {
877 const char *bol = sig.buf + copypos;
878 const char *eol = strchrnul(bol, '\n');
879 int len = (eol - bol) + !!*eol;
880
881 if (!copypos) {
882 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
883 inspos += gpg_sig_header_len;
884 }
885 strbuf_insert(buf, inspos++, " ", 1);
886 strbuf_insert(buf, inspos, bol, len);
887 inspos += len;
888 copypos += len;
889 }
890 strbuf_release(&sig);
891 return 0;
892}
893
894int parse_signed_commit(const struct commit *commit,
895 struct strbuf *payload, struct strbuf *signature)
896{
897
898 unsigned long size;
899 const char *buffer = get_commit_buffer(commit, &size);
900 int in_signature, saw_signature = -1;
901 const char *line, *tail;
902
903 line = buffer;
904 tail = buffer + size;
905 in_signature = 0;
906 saw_signature = 0;
907 while (line < tail) {
908 const char *sig = NULL;
909 const char *next = memchr(line, '\n', tail - line);
910
911 next = next ? next + 1 : tail;
912 if (in_signature && line[0] == ' ')
913 sig = line + 1;
914 else if (starts_with(line, gpg_sig_header) &&
915 line[gpg_sig_header_len] == ' ')
916 sig = line + gpg_sig_header_len + 1;
917 if (sig) {
918 strbuf_add(signature, sig, next - sig);
919 saw_signature = 1;
920 in_signature = 1;
921 } else {
922 if (*line == '\n')
923 /* dump the whole remainder of the buffer */
924 next = tail;
925 strbuf_add(payload, line, next - line);
926 in_signature = 0;
927 }
928 line = next;
929 }
930 unuse_commit_buffer(commit, buffer);
931 return saw_signature;
932}
933
934int remove_signature(struct strbuf *buf)
935{
936 const char *line = buf->buf;
937 const char *tail = buf->buf + buf->len;
938 int in_signature = 0;
939 const char *sig_start = NULL;
940 const char *sig_end = NULL;
941
942 while (line < tail) {
943 const char *next = memchr(line, '\n', tail - line);
944 next = next ? next + 1 : tail;
945
946 if (in_signature && line[0] == ' ')
947 sig_end = next;
948 else if (starts_with(line, gpg_sig_header) &&
949 line[gpg_sig_header_len] == ' ') {
950 sig_start = line;
951 sig_end = next;
952 in_signature = 1;
953 } else {
954 if (*line == '\n')
955 /* dump the whole remainder of the buffer */
956 next = tail;
957 in_signature = 0;
958 }
959 line = next;
960 }
961
962 if (sig_start)
963 strbuf_remove(buf, sig_start - buf->buf, sig_end - sig_start);
964
965 return sig_start != NULL;
966}
967
968static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
969{
970 struct merge_remote_desc *desc;
971 struct commit_extra_header *mergetag;
972 char *buf;
973 unsigned long size, len;
974 enum object_type type;
975
976 desc = merge_remote_util(parent);
977 if (!desc || !desc->obj)
978 return;
979 buf = read_object_file(&desc->obj->oid, &type, &size);
980 if (!buf || type != OBJ_TAG)
981 goto free_return;
982 len = parse_signature(buf, size);
983 if (size == len)
984 goto free_return;
985 /*
986 * We could verify this signature and either omit the tag when
987 * it does not validate, but the integrator may not have the
988 * public key of the signer of the tag he is merging, while a
989 * later auditor may have it while auditing, so let's not run
990 * verify-signed-buffer here for now...
991 *
992 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
993 * warn("warning: signed tag unverified.");
994 */
995 mergetag = xcalloc(1, sizeof(*mergetag));
996 mergetag->key = xstrdup("mergetag");
997 mergetag->value = buf;
998 mergetag->len = size;
999
1000 **tail = mergetag;
1001 *tail = &mergetag->next;
1002 return;
1003
1004free_return:
1005 free(buf);
1006}
1007
1008int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1009{
1010 struct strbuf payload = STRBUF_INIT;
1011 struct strbuf signature = STRBUF_INIT;
1012 int ret = 1;
1013
1014 sigc->result = 'N';
1015
1016 if (parse_signed_commit(commit, &payload, &signature) <= 0)
1017 goto out;
1018 ret = check_signature(payload.buf, payload.len, signature.buf,
1019 signature.len, sigc);
1020
1021 out:
1022 strbuf_release(&payload);
1023 strbuf_release(&signature);
1024
1025 return ret;
1026}
1027
1028
1029
1030void append_merge_tag_headers(struct commit_list *parents,
1031 struct commit_extra_header ***tail)
1032{
1033 while (parents) {
1034 struct commit *parent = parents->item;
1035 handle_signed_tag(parent, tail);
1036 parents = parents->next;
1037 }
1038}
1039
1040static void add_extra_header(struct strbuf *buffer,
1041 struct commit_extra_header *extra)
1042{
1043 strbuf_addstr(buffer, extra->key);
1044 if (extra->len)
1045 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1046 else
1047 strbuf_addch(buffer, '\n');
1048}
1049
1050struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1051 const char **exclude)
1052{
1053 struct commit_extra_header *extra = NULL;
1054 unsigned long size;
1055 const char *buffer = get_commit_buffer(commit, &size);
1056 extra = read_commit_extra_header_lines(buffer, size, exclude);
1057 unuse_commit_buffer(commit, buffer);
1058 return extra;
1059}
1060
1061int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1062{
1063 struct commit_extra_header *extra, *to_free;
1064 int res = 0;
1065
1066 to_free = read_commit_extra_headers(commit, NULL);
1067 for (extra = to_free; !res && extra; extra = extra->next) {
1068 if (strcmp(extra->key, "mergetag"))
1069 continue; /* not a merge tag */
1070 res = fn(commit, extra, data);
1071 }
1072 free_commit_extra_headers(to_free);
1073 return res;
1074}
1075
1076static inline int standard_header_field(const char *field, size_t len)
1077{
1078 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1079 (len == 6 && !memcmp(field, "parent", 6)) ||
1080 (len == 6 && !memcmp(field, "author", 6)) ||
1081 (len == 9 && !memcmp(field, "committer", 9)) ||
1082 (len == 8 && !memcmp(field, "encoding", 8)));
1083}
1084
1085static int excluded_header_field(const char *field, size_t len, const char **exclude)
1086{
1087 if (!exclude)
1088 return 0;
1089
1090 while (*exclude) {
1091 size_t xlen = strlen(*exclude);
1092 if (len == xlen && !memcmp(field, *exclude, xlen))
1093 return 1;
1094 exclude++;
1095 }
1096 return 0;
1097}
1098
1099static struct commit_extra_header *read_commit_extra_header_lines(
1100 const char *buffer, size_t size,
1101 const char **exclude)
1102{
1103 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1104 const char *line, *next, *eof, *eob;
1105 struct strbuf buf = STRBUF_INIT;
1106
1107 for (line = buffer, eob = line + size;
1108 line < eob && *line != '\n';
1109 line = next) {
1110 next = memchr(line, '\n', eob - line);
1111 next = next ? next + 1 : eob;
1112 if (*line == ' ') {
1113 /* continuation */
1114 if (it)
1115 strbuf_add(&buf, line + 1, next - (line + 1));
1116 continue;
1117 }
1118 if (it)
1119 it->value = strbuf_detach(&buf, &it->len);
1120 strbuf_reset(&buf);
1121 it = NULL;
1122
1123 eof = memchr(line, ' ', next - line);
1124 if (!eof)
1125 eof = next;
1126 else if (standard_header_field(line, eof - line) ||
1127 excluded_header_field(line, eof - line, exclude))
1128 continue;
1129
1130 it = xcalloc(1, sizeof(*it));
1131 it->key = xmemdupz(line, eof-line);
1132 *tail = it;
1133 tail = &it->next;
1134 if (eof + 1 < next)
1135 strbuf_add(&buf, eof + 1, next - (eof + 1));
1136 }
1137 if (it)
1138 it->value = strbuf_detach(&buf, &it->len);
1139 return extra;
1140}
1141
1142void free_commit_extra_headers(struct commit_extra_header *extra)
1143{
1144 while (extra) {
1145 struct commit_extra_header *next = extra->next;
1146 free(extra->key);
1147 free(extra->value);
1148 free(extra);
1149 extra = next;
1150 }
1151}
1152
1153int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1154 struct commit_list *parents, struct object_id *ret,
1155 const char *author, const char *sign_commit)
1156{
1157 struct commit_extra_header *extra = NULL, **tail = &extra;
1158 int result;
1159
1160 append_merge_tag_headers(parents, &tail);
1161 result = commit_tree_extended(msg, msg_len, tree, parents, ret,
1162 author, sign_commit, extra);
1163 free_commit_extra_headers(extra);
1164 return result;
1165}
1166
1167static int find_invalid_utf8(const char *buf, int len)
1168{
1169 int offset = 0;
1170 static const unsigned int max_codepoint[] = {
1171 0x7f, 0x7ff, 0xffff, 0x10ffff
1172 };
1173
1174 while (len) {
1175 unsigned char c = *buf++;
1176 int bytes, bad_offset;
1177 unsigned int codepoint;
1178 unsigned int min_val, max_val;
1179
1180 len--;
1181 offset++;
1182
1183 /* Simple US-ASCII? No worries. */
1184 if (c < 0x80)
1185 continue;
1186
1187 bad_offset = offset-1;
1188
1189 /*
1190 * Count how many more high bits set: that's how
1191 * many more bytes this sequence should have.
1192 */
1193 bytes = 0;
1194 while (c & 0x40) {
1195 c <<= 1;
1196 bytes++;
1197 }
1198
1199 /*
1200 * Must be between 1 and 3 more bytes. Longer sequences result in
1201 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1202 */
1203 if (bytes < 1 || 3 < bytes)
1204 return bad_offset;
1205
1206 /* Do we *have* that many bytes? */
1207 if (len < bytes)
1208 return bad_offset;
1209
1210 /*
1211 * Place the encoded bits at the bottom of the value and compute the
1212 * valid range.
1213 */
1214 codepoint = (c & 0x7f) >> bytes;
1215 min_val = max_codepoint[bytes-1] + 1;
1216 max_val = max_codepoint[bytes];
1217
1218 offset += bytes;
1219 len -= bytes;
1220
1221 /* And verify that they are good continuation bytes */
1222 do {
1223 codepoint <<= 6;
1224 codepoint |= *buf & 0x3f;
1225 if ((*buf++ & 0xc0) != 0x80)
1226 return bad_offset;
1227 } while (--bytes);
1228
1229 /* Reject codepoints that are out of range for the sequence length. */
1230 if (codepoint < min_val || codepoint > max_val)
1231 return bad_offset;
1232 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1233 if ((codepoint & 0x1ff800) == 0xd800)
1234 return bad_offset;
1235 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1236 if ((codepoint & 0xfffe) == 0xfffe)
1237 return bad_offset;
1238 /* So are anything in the range U+FDD0..U+FDEF. */
1239 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1240 return bad_offset;
1241 }
1242 return -1;
1243}
1244
1245/*
1246 * This verifies that the buffer is in proper utf8 format.
1247 *
1248 * If it isn't, it assumes any non-utf8 characters are Latin1,
1249 * and does the conversion.
1250 */
1251static int verify_utf8(struct strbuf *buf)
1252{
1253 int ok = 1;
1254 long pos = 0;
1255
1256 for (;;) {
1257 int bad;
1258 unsigned char c;
1259 unsigned char replace[2];
1260
1261 bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1262 if (bad < 0)
1263 return ok;
1264 pos += bad;
1265 ok = 0;
1266 c = buf->buf[pos];
1267 strbuf_remove(buf, pos, 1);
1268
1269 /* We know 'c' must be in the range 128-255 */
1270 replace[0] = 0xc0 + (c >> 6);
1271 replace[1] = 0x80 + (c & 0x3f);
1272 strbuf_insert(buf, pos, replace, 2);
1273 pos += 2;
1274 }
1275}
1276
1277static const char commit_utf8_warn[] =
1278N_("Warning: commit message did not conform to UTF-8.\n"
1279 "You may want to amend it after fixing the message, or set the config\n"
1280 "variable i18n.commitencoding to the encoding your project uses.\n");
1281
1282int commit_tree_extended(const char *msg, size_t msg_len,
1283 const struct object_id *tree,
1284 struct commit_list *parents, struct object_id *ret,
1285 const char *author, const char *sign_commit,
1286 struct commit_extra_header *extra)
1287{
1288 int result;
1289 int encoding_is_utf8;
1290 struct strbuf buffer;
1291
1292 assert_oid_type(tree, OBJ_TREE);
1293
1294 if (memchr(msg, '\0', msg_len))
1295 return error("a NUL byte in commit log message not allowed.");
1296
1297 /* Not having i18n.commitencoding is the same as having utf-8 */
1298 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1299
1300 strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1301 strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1302
1303 /*
1304 * NOTE! This ordering means that the same exact tree merged with a
1305 * different order of parents will be a _different_ changeset even
1306 * if everything else stays the same.
1307 */
1308 while (parents) {
1309 struct commit *parent = pop_commit(&parents);
1310 strbuf_addf(&buffer, "parent %s\n",
1311 oid_to_hex(&parent->object.oid));
1312 }
1313
1314 /* Person/date information */
1315 if (!author)
1316 author = git_author_info(IDENT_STRICT);
1317 strbuf_addf(&buffer, "author %s\n", author);
1318 strbuf_addf(&buffer, "committer %s\n", git_committer_info(IDENT_STRICT));
1319 if (!encoding_is_utf8)
1320 strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1321
1322 while (extra) {
1323 add_extra_header(&buffer, extra);
1324 extra = extra->next;
1325 }
1326 strbuf_addch(&buffer, '\n');
1327
1328 /* And add the comment */
1329 strbuf_add(&buffer, msg, msg_len);
1330
1331 /* And check the encoding */
1332 if (encoding_is_utf8 && !verify_utf8(&buffer))
1333 fprintf(stderr, _(commit_utf8_warn));
1334
1335 if (sign_commit && do_sign_commit(&buffer, sign_commit)) {
1336 result = -1;
1337 goto out;
1338 }
1339
1340 result = write_object_file(buffer.buf, buffer.len, commit_type, ret);
1341out:
1342 strbuf_release(&buffer);
1343 return result;
1344}
1345
1346define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1347static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1348
1349struct merge_remote_desc *merge_remote_util(struct commit *commit)
1350{
1351 return *merge_desc_slab_at(&merge_desc_slab, commit);
1352}
1353
1354void set_merge_remote_desc(struct commit *commit,
1355 const char *name, struct object *obj)
1356{
1357 struct merge_remote_desc *desc;
1358 FLEX_ALLOC_STR(desc, name, name);
1359 desc->obj = obj;
1360 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1361}
1362
1363struct commit *get_merge_parent(const char *name)
1364{
1365 struct object *obj;
1366 struct commit *commit;
1367 struct object_id oid;
1368 if (get_oid(name, &oid))
1369 return NULL;
1370 obj = parse_object(the_repository, &oid);
1371 commit = (struct commit *)peel_to_type(name, 0, obj, OBJ_COMMIT);
1372 if (commit && !merge_remote_util(commit))
1373 set_merge_remote_desc(commit, name, obj);
1374 return commit;
1375}
1376
1377/*
1378 * Append a commit to the end of the commit_list.
1379 *
1380 * next starts by pointing to the variable that holds the head of an
1381 * empty commit_list, and is updated to point to the "next" field of
1382 * the last item on the list as new commits are appended.
1383 *
1384 * Usage example:
1385 *
1386 * struct commit_list *list;
1387 * struct commit_list **next = &list;
1388 *
1389 * next = commit_list_append(c1, next);
1390 * next = commit_list_append(c2, next);
1391 * assert(commit_list_count(list) == 2);
1392 * return list;
1393 */
1394struct commit_list **commit_list_append(struct commit *commit,
1395 struct commit_list **next)
1396{
1397 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1398 new_commit->item = commit;
1399 *next = new_commit;
1400 new_commit->next = NULL;
1401 return &new_commit->next;
1402}
1403
1404const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1405{
1406 int key_len = strlen(key);
1407 const char *line = msg;
1408
1409 while (line) {
1410 const char *eol = strchrnul(line, '\n');
1411
1412 if (line == eol)
1413 return NULL;
1414
1415 if (eol - line > key_len &&
1416 !strncmp(line, key, key_len) &&
1417 line[key_len] == ' ') {
1418 *out_len = eol - line - key_len - 1;
1419 return line + key_len + 1;
1420 }
1421 line = *eol ? eol + 1 : NULL;
1422 }
1423 return NULL;
1424}
1425
1426/*
1427 * Inspect the given string and determine the true "end" of the log message, in
1428 * order to find where to put a new Signed-off-by: line. Ignored are
1429 * trailing comment lines and blank lines. To support "git commit -s
1430 * --amend" on an existing commit, we also ignore "Conflicts:". To
1431 * support "git commit -v", we truncate at cut lines.
1432 *
1433 * Returns the number of bytes from the tail to ignore, to be fed as
1434 * the second parameter to append_signoff().
1435 */
1436size_t ignore_non_trailer(const char *buf, size_t len)
1437{
1438 size_t boc = 0;
1439 size_t bol = 0;
1440 int in_old_conflicts_block = 0;
1441 size_t cutoff = wt_status_locate_end(buf, len);
1442
1443 while (bol < cutoff) {
1444 const char *next_line = memchr(buf + bol, '\n', len - bol);
1445
1446 if (!next_line)
1447 next_line = buf + len;
1448 else
1449 next_line++;
1450
1451 if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1452 /* is this the first of the run of comments? */
1453 if (!boc)
1454 boc = bol;
1455 /* otherwise, it is just continuing */
1456 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1457 in_old_conflicts_block = 1;
1458 if (!boc)
1459 boc = bol;
1460 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1461 ; /* a pathname in the conflicts block */
1462 } else if (boc) {
1463 /* the previous was not trailing comment */
1464 boc = 0;
1465 in_old_conflicts_block = 0;
1466 }
1467 bol = next_line - buf;
1468 }
1469 return boc ? len - boc : len - cutoff;
1470}