Raw
1 #define USE_THE_REPOSITORY_VARIABLE
2
3 #include "git-compat-util.h"
4 #include "tag.h"
5 #include "commit.h"
6 #include "commit-graph.h"
7 #include "environment.h"
8 #include "gettext.h"
9 #include "hex.h"
10 #include "repository.h"
11 #include "object-name.h"
12 #include "odb.h"
13 #include "utf8.h"
14 #include "diff.h"
15 #include "revision.h"
16 #include "notes.h"
17 #include "alloc.h"
18 #include "gpg-interface.h"
19 #include "mergesort.h"
20 #include "commit-slab.h"
21 #include "prio-queue.h"
22 #include "hash-lookup.h"
23 #include "wt-status.h"
24 #include "advice.h"
25 #include "refs.h"
26 #include "commit-reach.h"
27 #include "setup.h"
28 #include "shallow.h"
29 #include "tree.h"
30 #include "hook.h"
31 #include "parse.h"
32 #include "object-file.h"
33 #include "object-file-convert.h"
34
35 static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
36
37 int save_commit_buffer = 1;
38 int no_graft_file_deprecated_advice;
39
40 const char *commit_type = "commit";
41
42 struct commit *lookup_commit_reference_gently(struct repository *r,
43 const struct object_id *oid, int quiet)
44 {
45 const struct object_id *maybe_peeled;
46 struct object_id peeled_oid;
47 struct commit *commit;
48 enum object_type type;
49
50 switch (peel_object_ext(r, oid, &peeled_oid, 0, &type)) {
51 case PEEL_NON_TAG:
52 maybe_peeled = oid;
53 break;
54 case PEEL_PEELED:
55 maybe_peeled = &peeled_oid;
56 break;
57 default:
58 return NULL;
59 }
60
61 if (type != OBJ_COMMIT) {
62 if (!quiet)
63 error(_("object %s is a %s, not a %s"),
64 oid_to_hex(oid), type_name(type),
65 type_name(OBJ_COMMIT));
66 return NULL;
67 }
68
69 commit = lookup_commit(r, maybe_peeled);
70 if (!commit || repo_parse_commit_gently(r, commit, quiet) < 0)
71 return NULL;
72
73 return commit;
74 }
75
76 struct commit *lookup_commit_reference(struct repository *r, const struct object_id *oid)
77 {
78 return lookup_commit_reference_gently(r, oid, 0);
79 }
80
81 struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
82 {
83 struct commit *c = lookup_commit_reference(the_repository, oid);
84 if (!c)
85 die(_("could not parse %s"), ref_name);
86 if (!oideq(oid, &c->object.oid)) {
87 warning(_("%s %s is not a commit!"),
88 ref_name, oid_to_hex(oid));
89 }
90 return c;
91 }
92
93 struct commit *lookup_commit_object(struct repository *r,
94 const struct object_id *oid)
95 {
96 struct object *obj = parse_object(r, oid);
97 return obj ? object_as_type(obj, OBJ_COMMIT, 0) : NULL;
98
99 }
100
101 struct commit *lookup_commit(struct repository *r, const struct object_id *oid)
102 {
103 struct object *obj = lookup_object(r, oid);
104 if (!obj)
105 return create_object(r, oid, alloc_commit_node(r));
106 return object_as_type(obj, OBJ_COMMIT, 0);
107 }
108
109 struct commit *lookup_commit_reference_by_name(const char *name)
110 {
111 return lookup_commit_reference_by_name_gently(name, 0);
112 }
113
114 struct commit *lookup_commit_reference_by_name_gently(const char *name,
115 int quiet)
116 {
117 struct object_id oid;
118 struct commit *commit;
119
120 if (repo_get_oid_committish(the_repository, name, &oid))
121 return NULL;
122 commit = lookup_commit_reference_gently(the_repository, &oid, quiet);
123 if (repo_parse_commit(the_repository, commit))
124 return NULL;
125 return commit;
126 }
127
128 static timestamp_t parse_commit_date(const char *buf, const char *tail)
129 {
130 const char *dateptr;
131 const char *eol;
132
133 if (buf + 6 >= tail)
134 return 0;
135 if (memcmp(buf, "author", 6))
136 return 0;
137 while (buf < tail && *buf++ != '\n')
138 /* nada */;
139 if (buf + 9 >= tail)
140 return 0;
141 if (memcmp(buf, "committer", 9))
142 return 0;
143
144 /*
145 * Jump to end-of-line so that we can walk backwards to find the
146 * end-of-email ">". This is more forgiving of malformed cases
147 * because unexpected characters tend to be in the name and email
148 * fields.
149 */
150 eol = memchr(buf, '\n', tail - buf);
151 if (!eol)
152 return 0;
153 dateptr = eol;
154 while (dateptr > buf && dateptr[-1] != '>')
155 dateptr--;
156 if (dateptr == buf)
157 return 0;
158
159 /*
160 * Trim leading whitespace, but make sure we have at least one
161 * non-whitespace character, as parse_timestamp() will otherwise walk
162 * right past the newline we found in "eol" when skipping whitespace
163 * itself.
164 *
165 * In theory it would be sufficient to allow any character not matched
166 * by isspace(), but there's a catch: our isspace() does not
167 * necessarily match the behavior of parse_timestamp(), as the latter
168 * is implemented by system routines which match more exotic control
169 * codes, or even locale-dependent sequences.
170 *
171 * Since we expect the timestamp to be a number, we can check for that.
172 * Anything else (e.g., a non-numeric token like "foo") would just
173 * cause parse_timestamp() to return 0 anyway.
174 */
175 while (dateptr < eol && isspace(*dateptr))
176 dateptr++;
177 if (!isdigit(*dateptr) && *dateptr != '-')
178 return 0;
179
180 /*
181 * We know there is at least one digit (or dash), so we'll begin
182 * parsing there and stop at worst case at eol.
183 *
184 * Note that we may feed parse_timestamp() extra characters here if the
185 * commit is malformed, and it will parse as far as it can. For
186 * example, "123foo456" would return "123". That might be questionable
187 * (versus returning "0"), but it would help in a hypothetical case
188 * like "123456+0100", where the whitespace from the timezone is
189 * missing. Since such syntactic errors may be baked into history and
190 * hard to correct now, let's err on trying to make our best guess
191 * here, rather than insist on perfect syntax.
192 */
193 return parse_timestamp(dateptr, NULL, 10);
194 }
195
196 static const struct object_id *commit_graft_oid_access(size_t index, const void *table)
197 {
198 const struct commit_graft * const *commit_graft_table = table;
199 return &commit_graft_table[index]->oid;
200 }
201
202 int commit_graft_pos(struct repository *r, const struct object_id *oid)
203 {
204 return oid_pos(oid, r->parsed_objects->grafts,
205 r->parsed_objects->grafts_nr,
206 commit_graft_oid_access);
207 }
208
209 void unparse_commit(struct repository *r, const struct object_id *oid)
210 {
211 struct commit *c = lookup_commit(r, oid);
212
213 if (!c->object.parsed)
214 return;
215 commit_list_free(c->parents);
216 c->parents = NULL;
217 c->object.parsed = 0;
218 }
219
220 int register_commit_graft(struct repository *r, struct commit_graft *graft,
221 int ignore_dups)
222 {
223 int pos = commit_graft_pos(r, &graft->oid);
224
225 if (0 <= pos) {
226 if (ignore_dups)
227 free(graft);
228 else {
229 free(r->parsed_objects->grafts[pos]);
230 r->parsed_objects->grafts[pos] = graft;
231 }
232 return 1;
233 }
234 pos = -pos - 1;
235 ALLOC_GROW(r->parsed_objects->grafts,
236 r->parsed_objects->grafts_nr + 1,
237 r->parsed_objects->grafts_alloc);
238 r->parsed_objects->grafts_nr++;
239 if (pos < r->parsed_objects->grafts_nr)
240 memmove(r->parsed_objects->grafts + pos + 1,
241 r->parsed_objects->grafts + pos,
242 (r->parsed_objects->grafts_nr - pos - 1) *
243 sizeof(*r->parsed_objects->grafts));
244 r->parsed_objects->grafts[pos] = graft;
245 unparse_commit(r, &graft->oid);
246 return 0;
247 }
248
249 struct commit_graft *read_graft_line(struct strbuf *line)
250 {
251 /* The format is just "Commit Parent1 Parent2 ...\n" */
252 int i, phase;
253 const char *tail = NULL;
254 struct commit_graft *graft = NULL;
255 struct object_id dummy_oid, *oid;
256
257 strbuf_rtrim(line);
258 if (!line->len || line->buf[0] == '#')
259 return NULL;
260 /*
261 * phase 0 verifies line, counts hashes in line and allocates graft
262 * phase 1 fills graft
263 */
264 for (phase = 0; phase < 2; phase++) {
265 oid = graft ? &graft->oid : &dummy_oid;
266 if (parse_oid_hex(line->buf, oid, &tail))
267 goto bad_graft_data;
268 for (i = 0; *tail != '\0'; i++) {
269 oid = graft ? &graft->parent[i] : &dummy_oid;
270 if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail))
271 goto bad_graft_data;
272 }
273 if (!graft) {
274 graft = xmalloc(st_add(sizeof(*graft),
275 st_mult(sizeof(struct object_id), i)));
276 graft->nr_parent = i;
277 }
278 }
279 return graft;
280
281 bad_graft_data:
282 error("bad graft data: %s", line->buf);
283 assert(!graft);
284 return NULL;
285 }
286
287 static int read_graft_file(struct repository *r, const char *graft_file)
288 {
289 FILE *fp = fopen_or_warn(graft_file, "r");
290 struct strbuf buf = STRBUF_INIT;
291 if (!fp)
292 return -1;
293 if (!no_graft_file_deprecated_advice &&
294 advice_enabled(ADVICE_GRAFT_FILE_DEPRECATED))
295 advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n"
296 "and will be removed in a future Git version.\n"
297 "\n"
298 "Please use \"git replace --convert-graft-file\"\n"
299 "to convert the grafts into replace refs.\n"
300 "\n"
301 "Turn this message off by running\n"
302 "\"git config set advice.graftFileDeprecated false\""));
303 while (!strbuf_getwholeline(&buf, fp, '\n')) {
304 /* The format is just "Commit Parent1 Parent2 ...\n" */
305 struct commit_graft *graft = read_graft_line(&buf);
306 if (!graft)
307 continue;
308 if (register_commit_graft(r, graft, 1))
309 error("duplicate graft data: %s", buf.buf);
310 }
311 fclose(fp);
312 strbuf_release(&buf);
313 return 0;
314 }
315
316 void prepare_commit_graft(struct repository *r)
317 {
318 const char *graft_file;
319
320 if (r->parsed_objects->commit_graft_prepared)
321 return;
322 if (!startup_info->have_repository)
323 return;
324
325 graft_file = repo_get_graft_file(r);
326 read_graft_file(r, graft_file);
327 /* make sure shallows are read */
328 is_repository_shallow(r);
329 r->parsed_objects->commit_graft_prepared = 1;
330 }
331
332 struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid)
333 {
334 int pos;
335 prepare_commit_graft(r);
336 pos = commit_graft_pos(r, oid);
337 if (pos < 0)
338 return NULL;
339 return r->parsed_objects->grafts[pos];
340 }
341
342 int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
343 {
344 int i, ret;
345 for (i = ret = 0; i < the_repository->parsed_objects->grafts_nr && !ret; i++)
346 ret = fn(the_repository->parsed_objects->grafts[i], cb_data);
347 return ret;
348 }
349
350 struct commit_buffer {
351 void *buffer;
352 unsigned long size;
353 };
354 define_commit_slab(buffer_slab, struct commit_buffer);
355
356 struct buffer_slab *allocate_commit_buffer_slab(void)
357 {
358 struct buffer_slab *bs = xmalloc(sizeof(*bs));
359 init_buffer_slab(bs);
360 return bs;
361 }
362
363 void free_commit_buffer_slab(struct buffer_slab *bs)
364 {
365 clear_buffer_slab(bs);
366 free(bs);
367 }
368
369 void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size)
370 {
371 struct commit_buffer *v = buffer_slab_at(
372 r->parsed_objects->buffer_slab, commit);
373 v->buffer = buffer;
374 v->size = size;
375 }
376
377 const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep)
378 {
379 struct commit_buffer *v = buffer_slab_peek(
380 r->parsed_objects->buffer_slab, commit);
381 if (!v) {
382 if (sizep)
383 *sizep = 0;
384 return NULL;
385 }
386 if (sizep)
387 *sizep = v->size;
388 return v->buffer;
389 }
390
391 const void *repo_get_commit_buffer(struct repository *r,
392 const struct commit *commit,
393 unsigned long *sizep)
394 {
395 const void *ret = get_cached_commit_buffer(r, commit, sizep);
396 if (!ret) {
397 enum object_type type;
398 unsigned long size;
399 ret = odb_read_object(r->objects, &commit->object.oid, &type, &size);
400 if (!ret)
401 die("cannot read commit object %s",
402 oid_to_hex(&commit->object.oid));
403 if (type != OBJ_COMMIT)
404 die("expected commit for %s, got %s",
405 oid_to_hex(&commit->object.oid), type_name(type));
406 if (sizep)
407 *sizep = size;
408 }
409 return ret;
410 }
411
412 void repo_unuse_commit_buffer(struct repository *r,
413 const struct commit *commit,
414 const void *buffer)
415 {
416 struct commit_buffer *v = buffer_slab_peek(
417 r->parsed_objects->buffer_slab, commit);
418 if (!(v && v->buffer == buffer))
419 free((void *)buffer);
420 }
421
422 void free_commit_buffer(struct parsed_object_pool *pool, struct commit *commit)
423 {
424 struct commit_buffer *v = buffer_slab_peek(
425 pool->buffer_slab, commit);
426 if (v) {
427 FREE_AND_NULL(v->buffer);
428 v->size = 0;
429 }
430 }
431
432 static inline void set_commit_tree(struct commit *c, struct tree *t)
433 {
434 c->maybe_tree = t;
435 }
436
437 static void load_tree_from_commit_contents(struct repository *r, struct commit *commit)
438 {
439 enum object_type type;
440 unsigned long size;
441 char *buf;
442 const char *p;
443 struct object_id tree_oid;
444
445 buf = odb_read_object(r->objects, &commit->object.oid, &type, &size);
446 if (!buf)
447 return;
448
449 if (type == OBJ_COMMIT &&
450 skip_prefix(buf, "tree ", &p) &&
451 !parse_oid_hex_algop(p, &tree_oid, &p, r->hash_algo) &&
452 *p == '\n')
453 set_commit_tree(commit, lookup_tree(r, &tree_oid));
454
455 free(buf);
456 }
457
458 struct tree *repo_get_commit_tree(struct repository *r,
459 const struct commit *commit)
460 {
461 if (commit->maybe_tree || !commit->object.parsed)
462 return commit->maybe_tree;
463
464 if (commit_graph_position(commit) != COMMIT_NOT_FROM_GRAPH)
465 return get_commit_tree_in_graph(r, commit);
466
467 /*
468 * This is either a corrupt commit, or one which we partially loaded
469 * from a graph file but then subsequently threw away the graph data.
470 *
471 * Optimistically assume it's the latter and try to reload from
472 * scratch. This gives a performance penalty if it really is a corrupt
473 * commit, but presumably that happens rarely (and only once per
474 * process).
475 */
476 load_tree_from_commit_contents(r, (struct commit *)commit);
477 return commit->maybe_tree;
478 }
479
480 struct object_id *get_commit_tree_oid(const struct commit *commit)
481 {
482 struct tree *tree = repo_get_commit_tree(the_repository, commit);
483 return tree ? &tree->object.oid : NULL;
484 }
485
486 void release_commit_memory(struct parsed_object_pool *pool, struct commit *c)
487 {
488 set_commit_tree(c, NULL);
489 free_commit_buffer(pool, c);
490 c->index = 0;
491 commit_list_free(c->parents);
492
493 c->object.parsed = 0;
494 }
495
496 const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
497 {
498 struct commit_buffer *v = buffer_slab_peek(
499 the_repository->parsed_objects->buffer_slab, commit);
500 void *ret;
501
502 if (!v) {
503 if (sizep)
504 *sizep = 0;
505 return NULL;
506 }
507 ret = v->buffer;
508 if (sizep)
509 *sizep = v->size;
510
511 v->buffer = NULL;
512 v->size = 0;
513 return ret;
514 }
515
516 int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph)
517 {
518 const char *tail = buffer;
519 const char *bufptr = buffer;
520 struct object_id parent;
521 struct commit_list **pptr;
522 struct commit_graft *graft;
523 const int tree_entry_len = the_hash_algo->hexsz + 5;
524 const int parent_entry_len = the_hash_algo->hexsz + 7;
525 struct tree *tree;
526
527 if (item->object.parsed)
528 return 0;
529 /*
530 * Presumably this is leftover from an earlier failed parse;
531 * clear it out in preparation for us re-parsing (we'll hit the
532 * same error, but that's good, since it lets our caller know
533 * the result cannot be trusted.
534 */
535 commit_list_free(item->parents);
536 item->parents = NULL;
537
538 tail += size;
539 if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
540 bufptr[tree_entry_len] != '\n')
541 return error("bogus commit object %s", oid_to_hex(&item->object.oid));
542 if (get_oid_hex(bufptr + 5, &parent) < 0)
543 return error("bad tree pointer in commit %s",
544 oid_to_hex(&item->object.oid));
545 tree = lookup_tree(r, &parent);
546 if (!tree)
547 return error("bad tree pointer %s in commit %s",
548 oid_to_hex(&parent),
549 oid_to_hex(&item->object.oid));
550 set_commit_tree(item, tree);
551 bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
552 pptr = &item->parents;
553
554 graft = lookup_commit_graft(r, &item->object.oid);
555 if (graft)
556 r->parsed_objects->substituted_parent = 1;
557 while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
558 struct commit *new_parent;
559
560 if (tail <= bufptr + parent_entry_len + 1 ||
561 get_oid_hex(bufptr + 7, &parent) ||
562 bufptr[parent_entry_len] != '\n')
563 return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
564 bufptr += parent_entry_len + 1;
565 /*
566 * The clone is shallow if nr_parent < 0, and we must
567 * not traverse its real parents even when we unhide them.
568 */
569 if (graft && (graft->nr_parent < 0 || !grafts_keep_true_parents))
570 continue;
571 new_parent = lookup_commit(r, &parent);
572 if (!new_parent)
573 return error("bad parent %s in commit %s",
574 oid_to_hex(&parent),
575 oid_to_hex(&item->object.oid));
576 pptr = &commit_list_insert(new_parent, pptr)->next;
577 }
578 if (graft) {
579 int i;
580 struct commit *new_parent;
581 for (i = 0; i < graft->nr_parent; i++) {
582 new_parent = lookup_commit(r,
583 &graft->parent[i]);
584 if (!new_parent)
585 return error("bad graft parent %s in commit %s",
586 oid_to_hex(&graft->parent[i]),
587 oid_to_hex(&item->object.oid));
588 pptr = &commit_list_insert(new_parent, pptr)->next;
589 }
590 }
591 item->date = parse_commit_date(bufptr, tail);
592
593 if (check_graph)
594 load_commit_graph_info(r, item);
595
596 item->object.parsed = 1;
597 return 0;
598 }
599
600 int repo_parse_commit_internal(struct repository *r,
601 struct commit *item,
602 int quiet_on_missing,
603 int use_commit_graph)
604 {
605 enum object_type type;
606 void *buffer;
607 unsigned long size;
608 struct object_info oi = {
609 .typep = &type,
610 .sizep = &size,
611 .contentp = &buffer,
612 };
613 /*
614 * Git does not support partial clones that exclude commits, so set
615 * OBJECT_INFO_SKIP_FETCH_OBJECT to fail fast when an object is missing.
616 */
617 int flags = OBJECT_INFO_LOOKUP_REPLACE | OBJECT_INFO_SKIP_FETCH_OBJECT |
618 OBJECT_INFO_DIE_IF_CORRUPT;
619 int ret;
620
621 if (!item)
622 return -1;
623 if (item->object.parsed)
624 return 0;
625 if (use_commit_graph && parse_commit_in_graph(r, item)) {
626 static int commit_graph_paranoia = -1;
627
628 if (commit_graph_paranoia == -1)
629 commit_graph_paranoia = git_env_bool(GIT_COMMIT_GRAPH_PARANOIA, 0);
630
631 if (commit_graph_paranoia && !odb_has_object(r->objects, &item->object.oid, 0)) {
632 unparse_commit(r, &item->object.oid);
633 return quiet_on_missing ? -1 :
634 error(_("commit %s exists in commit-graph but not in the object database"),
635 oid_to_hex(&item->object.oid));
636 }
637
638 return 0;
639 }
640
641 if (odb_read_object_info_extended(r->objects, &item->object.oid,
642 &oi, flags) < 0)
643 return quiet_on_missing ? -1 :
644 error("Could not read %s",
645 oid_to_hex(&item->object.oid));
646 if (type != OBJ_COMMIT) {
647 free(buffer);
648 return error("Object %s not a commit",
649 oid_to_hex(&item->object.oid));
650 }
651
652 ret = parse_commit_buffer(r, item, buffer, size, 0);
653 if (save_commit_buffer && !ret &&
654 !get_cached_commit_buffer(r, item, NULL)) {
655 set_commit_buffer(r, item, buffer, size);
656 return 0;
657 }
658 free(buffer);
659 return ret;
660 }
661
662 int repo_parse_commit_gently(struct repository *r,
663 struct commit *item, int quiet_on_missing)
664 {
665 return repo_parse_commit_internal(r, item, quiet_on_missing, 1);
666 }
667
668 void parse_commit_or_die(struct commit *item)
669 {
670 if (repo_parse_commit(the_repository, item))
671 die("unable to parse commit %s",
672 item ? oid_to_hex(&item->object.oid) : "(null)");
673 }
674
675 int find_commit_subject(const char *commit_buffer, const char **subject)
676 {
677 const char *eol;
678 const char *p = commit_buffer;
679
680 while (*p && (*p != '\n' || p[1] != '\n'))
681 p++;
682 if (*p) {
683 p = skip_blank_lines(p + 2);
684 eol = strchrnul(p, '\n');
685 } else
686 eol = p;
687
688 *subject = p;
689
690 return eol - p;
691 }
692
693 size_t commit_subject_length(const char *body)
694 {
695 const char *p = body;
696 while (*p) {
697 const char *next = skip_blank_lines(p);
698 if (next != p)
699 break;
700 p = strchrnul(p, '\n');
701 if (*p)
702 p++;
703 }
704 return p - body;
705 }
706
707 struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
708 {
709 struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
710 new_list->item = item;
711 new_list->next = *list_p;
712 *list_p = new_list;
713 return new_list;
714 }
715
716 int commit_list_contains(struct commit *item, struct commit_list *list)
717 {
718 while (list) {
719 if (list->item == item)
720 return 1;
721 list = list->next;
722 }
723
724 return 0;
725 }
726
727 unsigned commit_list_count(const struct commit_list *l)
728 {
729 unsigned c = 0;
730 for (; l; l = l->next )
731 c++;
732 return c;
733 }
734
735 struct commit_list *commit_list_copy(const struct commit_list *list)
736 {
737 struct commit_list *head = NULL;
738 struct commit_list **pp = &head;
739 while (list) {
740 pp = commit_list_append(list->item, pp);
741 list = list->next;
742 }
743 return head;
744 }
745
746 struct commit_list *commit_list_reverse(struct commit_list *list)
747 {
748 struct commit_list *next = NULL, *current, *backup;
749 for (current = list; current; current = backup) {
750 backup = current->next;
751 current->next = next;
752 next = current;
753 }
754 return next;
755 }
756
757 void commit_list_free(struct commit_list *list)
758 {
759 while (list)
760 pop_commit(&list);
761 }
762
763 static int commit_list_compare_by_date(const struct commit_list *a,
764 const struct commit_list *b)
765 {
766 timestamp_t a_date = a->item->date;
767 timestamp_t b_date = b->item->date;
768 if (a_date < b_date)
769 return 1;
770 if (a_date > b_date)
771 return -1;
772 return 0;
773 }
774
775 DEFINE_LIST_SORT(static, commit_list_sort, struct commit_list, next);
776
777 void commit_list_sort_by_date(struct commit_list **list)
778 {
779 commit_list_sort(list, commit_list_compare_by_date);
780 }
781
782 struct commit *pop_most_recent_commit(struct prio_queue *queue,
783 unsigned int mark)
784 {
785 struct commit *ret = prio_queue_peek(queue);
786 int get_pending = 1;
787 struct commit_list *parents = ret->parents;
788
789 while (parents) {
790 struct commit *commit = parents->item;
791 if (!repo_parse_commit(the_repository, commit) && !(commit->object.flags & mark)) {
792 commit->object.flags |= mark;
793 if (get_pending)
794 prio_queue_replace(queue, commit);
795 else
796 prio_queue_put(queue, commit);
797 get_pending = 0;
798 }
799 parents = parents->next;
800 }
801 if (get_pending)
802 prio_queue_get(queue);
803 return ret;
804 }
805
806 static void clear_commit_marks_1(struct commit_list **plist,
807 struct commit *commit, unsigned int mark)
808 {
809 while (commit) {
810 struct commit_list *parents;
811
812 if (!(mark & commit->object.flags))
813 return;
814
815 commit->object.flags &= ~mark;
816
817 parents = commit->parents;
818 if (!parents)
819 return;
820
821 while ((parents = parents->next)) {
822 if (parents->item->object.flags & mark)
823 commit_list_insert(parents->item, plist);
824 }
825
826 commit = commit->parents->item;
827 }
828 }
829
830 void clear_commit_marks_many(size_t nr, struct commit **commit, unsigned int mark)
831 {
832 for (size_t i = 0; i < nr; i++)
833 clear_commit_marks(commit[i], mark);
834 }
835
836 void clear_commit_marks(struct commit *commit, unsigned int mark)
837 {
838 struct commit_list *list = NULL;
839
840 clear_commit_marks_1(&list, commit, mark);
841 while (list)
842 clear_commit_marks_1(&list, pop_commit(&list), mark);
843 }
844
845 struct commit *pop_commit(struct commit_list **stack)
846 {
847 struct commit_list *top = *stack;
848 struct commit *item = top ? top->item : NULL;
849
850 if (top) {
851 *stack = top->next;
852 free(top);
853 }
854 return item;
855 }
856
857 /*
858 * Topological sort support
859 */
860
861 /* count number of children that have not been emitted */
862 define_commit_slab(indegree_slab, int);
863
864 define_commit_slab(author_date_slab, timestamp_t);
865
866 void record_author_date(struct author_date_slab *author_date,
867 struct commit *commit)
868 {
869 const char *buffer = repo_get_commit_buffer(the_repository, commit,
870 NULL);
871 struct ident_split ident;
872 const char *ident_line;
873 size_t ident_len;
874 char *date_end;
875 timestamp_t date;
876
877 ident_line = find_commit_header(buffer, "author", &ident_len);
878 if (!ident_line)
879 goto fail_exit; /* no author line */
880 if (split_ident_line(&ident, ident_line, ident_len) ||
881 !ident.date_begin || !ident.date_end)
882 goto fail_exit; /* malformed "author" line */
883
884 date = parse_timestamp(ident.date_begin, &date_end, 10);
885 if (date_end != ident.date_end)
886 goto fail_exit; /* malformed date */
887 *(author_date_slab_at(author_date, commit)) = date;
888
889 fail_exit:
890 repo_unuse_commit_buffer(the_repository, commit, buffer);
891 }
892
893 int compare_commits_by_author_date(const void *a_, const void *b_,
894 void *cb_data)
895 {
896 const struct commit *a = a_, *b = b_;
897 struct author_date_slab *author_date = cb_data;
898 timestamp_t a_date = *(author_date_slab_at(author_date, a));
899 timestamp_t b_date = *(author_date_slab_at(author_date, b));
900
901 /* newer commits with larger date first */
902 if (a_date < b_date)
903 return 1;
904 else if (a_date > b_date)
905 return -1;
906 return 0;
907 }
908
909 int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_,
910 void *unused UNUSED)
911 {
912 const struct commit *a = a_, *b = b_;
913 const timestamp_t generation_a = commit_graph_generation(a),
914 generation_b = commit_graph_generation(b);
915
916 /* newer commits first */
917 if (generation_a < generation_b)
918 return 1;
919 else if (generation_a > generation_b)
920 return -1;
921
922 /* use date as a heuristic when generations are equal */
923 if (a->date < b->date)
924 return 1;
925 else if (a->date > b->date)
926 return -1;
927 return 0;
928 }
929
930 int compare_commits_by_commit_date(const void *a_, const void *b_,
931 void *unused UNUSED)
932 {
933 const struct commit *a = a_, *b = b_;
934 /* newer commits with larger date first */
935 if (a->date < b->date)
936 return 1;
937 else if (a->date > b->date)
938 return -1;
939 return 0;
940 }
941
942 /*
943 * Performs an in-place topological sort on the list supplied.
944 */
945 void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
946 {
947 struct commit_list *next, *orig = *list;
948 struct commit_list **pptr;
949 struct indegree_slab indegree;
950 struct prio_queue queue;
951 struct commit *commit;
952 struct author_date_slab author_date;
953
954 if (!orig)
955 return;
956 *list = NULL;
957
958 init_indegree_slab(&indegree);
959 memset(&queue, '\0', sizeof(queue));
960
961 switch (sort_order) {
962 default: /* REV_SORT_IN_GRAPH_ORDER */
963 queue.compare = NULL;
964 break;
965 case REV_SORT_BY_COMMIT_DATE:
966 queue.compare = compare_commits_by_commit_date;
967 break;
968 case REV_SORT_BY_AUTHOR_DATE:
969 init_author_date_slab(&author_date);
970 queue.compare = compare_commits_by_author_date;
971 queue.cb_data = &author_date;
972 break;
973 }
974
975 /* Mark them and clear the indegree */
976 for (next = orig; next; next = next->next) {
977 struct commit *commit = next->item;
978 *(indegree_slab_at(&indegree, commit)) = 1;
979 /* also record the author dates, if needed */
980 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
981 record_author_date(&author_date, commit);
982 }
983
984 /* update the indegree */
985 for (next = orig; next; next = next->next) {
986 struct commit_list *parents = next->item->parents;
987 while (parents) {
988 struct commit *parent = parents->item;
989 int *pi = indegree_slab_at(&indegree, parent);
990
991 if (*pi)
992 (*pi)++;
993 parents = parents->next;
994 }
995 }
996
997 /*
998 * find the tips
999 *
1000 * tips are nodes not reachable from any other node in the list
1001 *
1002 * the tips serve as a starting set for the work queue.
1003 */
1004 for (next = orig; next; next = next->next) {
1005 struct commit *commit = next->item;
1006
1007 if (*(indegree_slab_at(&indegree, commit)) == 1)
1008 prio_queue_put(&queue, commit);
1009 }
1010
1011 /*
1012 * This is unfortunate; the initial tips need to be shown
1013 * in the order given from the revision traversal machinery.
1014 */
1015 if (sort_order == REV_SORT_IN_GRAPH_ORDER)
1016 prio_queue_reverse(&queue);
1017
1018 /* We no longer need the commit list */
1019 commit_list_free(orig);
1020
1021 pptr = list;
1022 *list = NULL;
1023 while ((commit = prio_queue_get(&queue)) != NULL) {
1024 struct commit_list *parents;
1025
1026 for (parents = commit->parents; parents ; parents = parents->next) {
1027 struct commit *parent = parents->item;
1028 int *pi = indegree_slab_at(&indegree, parent);
1029
1030 if (!*pi)
1031 continue;
1032
1033 /*
1034 * parents are only enqueued for emission
1035 * when all their children have been emitted thereby
1036 * guaranteeing topological order.
1037 */
1038 if (--(*pi) == 1)
1039 prio_queue_put(&queue, parent);
1040 }
1041 /*
1042 * all children of commit have already been
1043 * emitted. we can emit it now.
1044 */
1045 *(indegree_slab_at(&indegree, commit)) = 0;
1046
1047 pptr = &commit_list_insert(commit, pptr)->next;
1048 }
1049
1050 clear_indegree_slab(&indegree);
1051 clear_prio_queue(&queue);
1052 if (sort_order == REV_SORT_BY_AUTHOR_DATE)
1053 clear_author_date_slab(&author_date);
1054 }
1055
1056 struct rev_collect {
1057 struct commit_stack stack;
1058 unsigned int initial : 1;
1059 };
1060
1061 static void add_one_commit(struct object_id *oid, struct rev_collect *revs)
1062 {
1063 struct commit *commit;
1064
1065 if (is_null_oid(oid))
1066 return;
1067
1068 commit = lookup_commit(the_repository, oid);
1069 if (!commit ||
1070 (commit->object.flags & TMP_MARK) ||
1071 repo_parse_commit(the_repository, commit))
1072 return;
1073
1074 commit_stack_push(&revs->stack, commit);
1075 commit->object.flags |= TMP_MARK;
1076 }
1077
1078 static int collect_one_reflog_ent(const char *refname UNUSED,
1079 struct object_id *ooid, struct object_id *noid,
1080 const char *ident UNUSED,
1081 timestamp_t timestamp UNUSED, int tz UNUSED,
1082 const char *message UNUSED, void *cbdata)
1083 {
1084 struct rev_collect *revs = cbdata;
1085
1086 if (revs->initial) {
1087 revs->initial = 0;
1088 add_one_commit(ooid, revs);
1089 }
1090 add_one_commit(noid, revs);
1091 return 0;
1092 }
1093
1094 struct commit *get_fork_point(const char *refname, struct commit *commit)
1095 {
1096 struct object_id oid;
1097 struct rev_collect revs;
1098 struct commit_list *bases = NULL;
1099 size_t i;
1100 struct commit *ret = NULL;
1101 char *full_refname;
1102
1103 switch (repo_dwim_ref(the_repository, refname, strlen(refname), &oid,
1104 &full_refname, 0)) {
1105 case 0:
1106 die("No such ref: '%s'", refname);
1107 case 1:
1108 break; /* good */
1109 default:
1110 die("Ambiguous refname: '%s'", refname);
1111 }
1112
1113 commit_stack_init(&revs.stack);
1114 revs.initial = 1;
1115 refs_for_each_reflog_ent(get_main_ref_store(the_repository),
1116 full_refname, collect_one_reflog_ent, &revs);
1117
1118 if (!revs.stack.nr)
1119 add_one_commit(&oid, &revs);
1120
1121 for (i = 0; i < revs.stack.nr; i++)
1122 revs.stack.items[i]->object.flags &= ~TMP_MARK;
1123
1124 if (repo_get_merge_bases_many(the_repository, commit, revs.stack.nr,
1125 revs.stack.items, &bases) < 0)
1126 exit(128);
1127
1128 /*
1129 * There should be one and only one merge base, when we found
1130 * a common ancestor among reflog entries.
1131 */
1132 if (!bases || bases->next)
1133 goto cleanup_return;
1134
1135 /* And the found one must be one of the reflog entries */
1136 for (i = 0; i < revs.stack.nr; i++)
1137 if (&bases->item->object == &revs.stack.items[i]->object)
1138 break; /* found */
1139 if (revs.stack.nr <= i)
1140 goto cleanup_return;
1141
1142 ret = bases->item;
1143
1144 cleanup_return:
1145 commit_stack_clear(&revs.stack);
1146 commit_list_free(bases);
1147 free(full_refname);
1148 return ret;
1149 }
1150
1151 /*
1152 * Indexed by hash algorithm identifier.
1153 */
1154 static const char *gpg_sig_headers[] = {
1155 NULL,
1156 "gpgsig",
1157 "gpgsig-sha256",
1158 };
1159
1160 int add_header_signature(struct strbuf *buf, struct strbuf *sig, const struct git_hash_algo *algo)
1161 {
1162 int inspos, copypos;
1163 const char *eoh;
1164 const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(algo)];
1165 int gpg_sig_header_len = strlen(gpg_sig_header);
1166
1167 /* find the end of the header */
1168 eoh = strstr(buf->buf, "\n\n");
1169 if (!eoh)
1170 inspos = buf->len;
1171 else
1172 inspos = eoh - buf->buf + 1;
1173
1174 for (copypos = 0; sig->buf[copypos]; ) {
1175 const char *bol = sig->buf + copypos;
1176 const char *eol = strchrnul(bol, '\n');
1177 int len = (eol - bol) + !!*eol;
1178
1179 if (!copypos) {
1180 strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
1181 inspos += gpg_sig_header_len;
1182 }
1183 strbuf_insertstr(buf, inspos++, " ");
1184 strbuf_insert(buf, inspos, bol, len);
1185 inspos += len;
1186 copypos += len;
1187 }
1188 return 0;
1189 }
1190
1191 int parse_signed_commit(const struct commit *commit,
1192 struct strbuf *payload, struct strbuf *signature,
1193 const struct git_hash_algo *algop)
1194 {
1195 unsigned long size;
1196 const char *buffer = repo_get_commit_buffer(the_repository, commit,
1197 &size);
1198 int ret = parse_buffer_signed_by_header(buffer, size, payload, signature, algop);
1199
1200 repo_unuse_commit_buffer(the_repository, commit, buffer);
1201 return ret;
1202 }
1203
1204 int parse_buffer_signed_by_header(const char *buffer,
1205 unsigned long size,
1206 struct strbuf *payload,
1207 struct strbuf *signature,
1208 const struct git_hash_algo *algop)
1209 {
1210 int in_signature = 0, saw_signature = 0, other_signature = 0;
1211 const char *line, *tail, *p;
1212 const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(algop)];
1213
1214 line = buffer;
1215 tail = buffer + size;
1216 while (line < tail) {
1217 const char *sig = NULL;
1218 const char *next = memchr(line, '\n', tail - line);
1219
1220 next = next ? next + 1 : tail;
1221 if (in_signature && line[0] == ' ')
1222 sig = line + 1;
1223 else if (skip_prefix(line, gpg_sig_header, &p) &&
1224 *p == ' ') {
1225 sig = line + strlen(gpg_sig_header) + 1;
1226 other_signature = 0;
1227 }
1228 else if (starts_with(line, "gpgsig"))
1229 other_signature = 1;
1230 else if (other_signature && line[0] != ' ')
1231 other_signature = 0;
1232 if (sig) {
1233 strbuf_add(signature, sig, next - sig);
1234 saw_signature = 1;
1235 in_signature = 1;
1236 } else {
1237 if (*line == '\n')
1238 /* dump the whole remainder of the buffer */
1239 next = tail;
1240 if (!other_signature)
1241 strbuf_add(payload, line, next - line);
1242 in_signature = 0;
1243 }
1244 line = next;
1245 }
1246 return saw_signature;
1247 }
1248
1249 int remove_signature(struct strbuf *buf)
1250 {
1251 const char *line = buf->buf;
1252 const char *tail = buf->buf + buf->len;
1253 int in_signature = 0;
1254 struct sigbuf {
1255 const char *start;
1256 const char *end;
1257 } sigs[2], *sigp = &sigs[0];
1258 int i;
1259 const char *orig_buf = buf->buf;
1260
1261 memset(sigs, 0, sizeof(sigs));
1262
1263 while (line < tail) {
1264 const char *next = memchr(line, '\n', tail - line);
1265 next = next ? next + 1 : tail;
1266
1267 if (in_signature && line[0] == ' ')
1268 sigp->end = next;
1269 else if (starts_with(line, "gpgsig")) {
1270 int i;
1271 for (i = 1; i < GIT_HASH_NALGOS; i++) {
1272 const char *p;
1273 if (skip_prefix(line, gpg_sig_headers[i], &p) &&
1274 *p == ' ') {
1275 sigp->start = line;
1276 sigp->end = next;
1277 in_signature = 1;
1278 }
1279 }
1280 } else {
1281 if (*line == '\n')
1282 /* dump the whole remainder of the buffer */
1283 next = tail;
1284 if (in_signature && sigp - sigs != ARRAY_SIZE(sigs))
1285 sigp++;
1286 in_signature = 0;
1287 }
1288 line = next;
1289 }
1290
1291 for (i = ARRAY_SIZE(sigs) - 1; i >= 0; i--)
1292 if (sigs[i].start)
1293 strbuf_remove(buf, sigs[i].start - orig_buf, sigs[i].end - sigs[i].start);
1294
1295 return sigs[0].start != NULL;
1296 }
1297
1298 static void handle_signed_tag(const struct commit *parent, struct commit_extra_header ***tail)
1299 {
1300 struct merge_remote_desc *desc;
1301 struct commit_extra_header *mergetag;
1302 char *buf;
1303 unsigned long size;
1304 enum object_type type;
1305 struct strbuf payload = STRBUF_INIT;
1306 struct strbuf signature = STRBUF_INIT;
1307
1308 desc = merge_remote_util(parent);
1309 if (!desc || !desc->obj)
1310 return;
1311 buf = odb_read_object(the_repository->objects, &desc->obj->oid,
1312 &type, &size);
1313 if (!buf || type != OBJ_TAG)
1314 goto free_return;
1315 if (!parse_signature(buf, size, &payload, &signature))
1316 goto free_return;
1317 /*
1318 * We could verify this signature and either omit the tag when
1319 * it does not validate, but the integrator may not have the
1320 * public key of the signer of the tag being merged, while a
1321 * later auditor may have it while auditing, so let's not run
1322 * verify-signed-buffer here for now...
1323 *
1324 * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
1325 * warn("warning: signed tag unverified.");
1326 */
1327 CALLOC_ARRAY(mergetag, 1);
1328 mergetag->key = xstrdup("mergetag");
1329 mergetag->value = buf;
1330 mergetag->len = size;
1331
1332 **tail = mergetag;
1333 *tail = &mergetag->next;
1334 strbuf_release(&payload);
1335 strbuf_release(&signature);
1336 return;
1337
1338 free_return:
1339 free(buf);
1340 }
1341
1342 int verify_commit_buffer(const char *buffer, size_t size,
1343 struct signature_check *sigc)
1344 {
1345 struct strbuf payload = STRBUF_INIT;
1346 struct strbuf signature = STRBUF_INIT;
1347 int ret = 1;
1348
1349 sigc->result = 'N';
1350
1351 if (parse_buffer_signed_by_header(buffer, size, &payload,
1352 &signature, the_hash_algo) <= 0)
1353 goto out;
1354
1355 sigc->payload_type = SIGNATURE_PAYLOAD_COMMIT;
1356 sigc->payload = strbuf_detach(&payload, &sigc->payload_len);
1357 ret = check_signature(sigc, signature.buf, signature.len);
1358
1359 out:
1360 strbuf_release(&payload);
1361 strbuf_release(&signature);
1362
1363 return ret;
1364 }
1365
1366 int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1367 {
1368 unsigned long size;
1369 const char *buffer = repo_get_commit_buffer(the_repository, commit, &size);
1370 int ret = verify_commit_buffer(buffer, size, sigc);
1371
1372 repo_unuse_commit_buffer(the_repository, commit, buffer);
1373
1374 return ret;
1375 }
1376
1377 void verify_merge_signature(struct commit *commit, int verbosity,
1378 int check_trust)
1379 {
1380 char hex[GIT_MAX_HEXSZ + 1];
1381 struct signature_check signature_check;
1382 int ret;
1383 memset(&signature_check, 0, sizeof(signature_check));
1384
1385 ret = check_commit_signature(commit, &signature_check);
1386
1387 repo_find_unique_abbrev_r(the_repository, hex, &commit->object.oid,
1388 DEFAULT_ABBREV);
1389 switch (signature_check.result) {
1390 case 'G':
1391 if (ret || (check_trust && signature_check.trust_level < TRUST_MARGINAL))
1392 die(_("Commit %s has an untrusted GPG signature, "
1393 "allegedly by %s."), hex, signature_check.signer);
1394 break;
1395 case 'B':
1396 die(_("Commit %s has a bad GPG signature "
1397 "allegedly by %s."), hex, signature_check.signer);
1398 default: /* 'N' */
1399 die(_("Commit %s does not have a GPG signature."), hex);
1400 }
1401 if (verbosity >= 0 && signature_check.result == 'G')
1402 printf(_("Commit %s has a good GPG signature by %s\n"),
1403 hex, signature_check.signer);
1404
1405 signature_check_clear(&signature_check);
1406 }
1407
1408 void append_merge_tag_headers(const struct commit_list *parents,
1409 struct commit_extra_header ***tail)
1410 {
1411 while (parents) {
1412 const struct commit *parent = parents->item;
1413 handle_signed_tag(parent, tail);
1414 parents = parents->next;
1415 }
1416 }
1417
1418 static int convert_commit_extra_headers(const struct commit_extra_header *orig,
1419 struct commit_extra_header **result)
1420 {
1421 const struct git_hash_algo *compat = the_repository->compat_hash_algo;
1422 const struct git_hash_algo *algo = the_repository->hash_algo;
1423 struct commit_extra_header *extra = NULL, **tail = &extra;
1424 struct strbuf out = STRBUF_INIT;
1425 while (orig) {
1426 struct commit_extra_header *new;
1427 CALLOC_ARRAY(new, 1);
1428 if (!strcmp(orig->key, "mergetag")) {
1429 if (convert_object_file(the_repository, &out, algo, compat,
1430 orig->value, orig->len,
1431 OBJ_TAG, 1)) {
1432 free(new);
1433 free_commit_extra_headers(extra);
1434 return -1;
1435 }
1436 new->key = xstrdup("mergetag");
1437 new->value = strbuf_detach(&out, &new->len);
1438 } else {
1439 new->key = xstrdup(orig->key);
1440 new->len = orig->len;
1441 new->value = xmemdupz(orig->value, orig->len);
1442 }
1443 *tail = new;
1444 tail = &new->next;
1445 orig = orig->next;
1446 }
1447 *result = extra;
1448 return 0;
1449 }
1450
1451 static void add_extra_header(struct strbuf *buffer,
1452 const struct commit_extra_header *extra)
1453 {
1454 strbuf_addstr(buffer, extra->key);
1455 if (extra->len)
1456 strbuf_add_lines(buffer, " ", extra->value, extra->len);
1457 else
1458 strbuf_addch(buffer, '\n');
1459 }
1460
1461 struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1462 const char **exclude)
1463 {
1464 struct commit_extra_header *extra = NULL;
1465 unsigned long size;
1466 const char *buffer = repo_get_commit_buffer(the_repository, commit,
1467 &size);
1468 extra = read_commit_extra_header_lines(buffer, size, exclude);
1469 repo_unuse_commit_buffer(the_repository, commit, buffer);
1470 return extra;
1471 }
1472
1473 int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1474 {
1475 struct commit_extra_header *extra, *to_free;
1476 int res = 0;
1477
1478 to_free = read_commit_extra_headers(commit, NULL);
1479 for (extra = to_free; !res && extra; extra = extra->next) {
1480 if (strcmp(extra->key, "mergetag"))
1481 continue; /* not a merge tag */
1482 res = fn(commit, extra, data);
1483 }
1484 free_commit_extra_headers(to_free);
1485 return res;
1486 }
1487
1488 static inline int standard_header_field(const char *field, size_t len)
1489 {
1490 return ((len == 4 && !memcmp(field, "tree", 4)) ||
1491 (len == 6 && !memcmp(field, "parent", 6)) ||
1492 (len == 6 && !memcmp(field, "author", 6)) ||
1493 (len == 9 && !memcmp(field, "committer", 9)) ||
1494 (len == 8 && !memcmp(field, "encoding", 8)));
1495 }
1496
1497 static int excluded_header_field(const char *field, size_t len, const char **exclude)
1498 {
1499 if (!exclude)
1500 return 0;
1501
1502 while (*exclude) {
1503 size_t xlen = strlen(*exclude);
1504 if (len == xlen && !memcmp(field, *exclude, xlen))
1505 return 1;
1506 exclude++;
1507 }
1508 return 0;
1509 }
1510
1511 static struct commit_extra_header *read_commit_extra_header_lines(
1512 const char *buffer, size_t size,
1513 const char **exclude)
1514 {
1515 struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1516 const char *line, *next, *eof, *eob;
1517 struct strbuf buf = STRBUF_INIT;
1518
1519 for (line = buffer, eob = line + size;
1520 line < eob && *line != '\n';
1521 line = next) {
1522 next = memchr(line, '\n', eob - line);
1523 next = next ? next + 1 : eob;
1524 if (*line == ' ') {
1525 /* continuation */
1526 if (it)
1527 strbuf_add(&buf, line + 1, next - (line + 1));
1528 continue;
1529 }
1530 if (it)
1531 it->value = strbuf_detach(&buf, &it->len);
1532 strbuf_reset(&buf);
1533 it = NULL;
1534
1535 eof = memchr(line, ' ', next - line);
1536 if (!eof)
1537 eof = next;
1538 else if (standard_header_field(line, eof - line) ||
1539 excluded_header_field(line, eof - line, exclude))
1540 continue;
1541
1542 CALLOC_ARRAY(it, 1);
1543 it->key = xmemdupz(line, eof-line);
1544 *tail = it;
1545 tail = &it->next;
1546 if (eof + 1 < next)
1547 strbuf_add(&buf, eof + 1, next - (eof + 1));
1548 }
1549 if (it)
1550 it->value = strbuf_detach(&buf, &it->len);
1551 return extra;
1552 }
1553
1554 void free_commit_extra_headers(struct commit_extra_header *extra)
1555 {
1556 while (extra) {
1557 struct commit_extra_header *next = extra->next;
1558 free(extra->key);
1559 free(extra->value);
1560 free(extra);
1561 extra = next;
1562 }
1563 }
1564
1565 int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1566 const struct commit_list *parents, struct object_id *ret,
1567 const char *author, const char *sign_commit)
1568 {
1569 struct commit_extra_header *extra = NULL, **tail = &extra;
1570 int result;
1571
1572 append_merge_tag_headers(parents, &tail);
1573 result = commit_tree_extended(msg, msg_len, tree, parents, ret, author,
1574 NULL, sign_commit, extra);
1575 free_commit_extra_headers(extra);
1576 return result;
1577 }
1578
1579 static bool has_invalid_utf8(const char *buf, size_t len, size_t *bad_offset)
1580 {
1581 size_t offset = 0;
1582 static const unsigned int max_codepoint[] = {
1583 0x7f, 0x7ff, 0xffff, 0x10ffff
1584 };
1585
1586 while (len) {
1587 unsigned char c = *buf++;
1588 unsigned bytes;
1589 unsigned int codepoint;
1590 unsigned int min_val, max_val;
1591
1592 len--;
1593 offset++;
1594
1595 /* Simple US-ASCII? No worries. */
1596 if (c < 0x80)
1597 continue;
1598
1599 *bad_offset = offset-1;
1600
1601 /*
1602 * Count how many more high bits set: that's how
1603 * many more bytes this sequence should have.
1604 */
1605 bytes = 0;
1606 while (c & 0x40) {
1607 c <<= 1;
1608 bytes++;
1609 }
1610
1611 /*
1612 * Must be between 1 and 3 more bytes. Longer sequences result in
1613 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1614 */
1615 if (bytes < 1 || 3 < bytes)
1616 return true;
1617
1618 /* Do we *have* that many bytes? */
1619 if (len < bytes)
1620 return true;
1621
1622 /*
1623 * Place the encoded bits at the bottom of the value and compute the
1624 * valid range.
1625 */
1626 codepoint = (c & 0x7f) >> bytes;
1627 min_val = max_codepoint[bytes-1] + 1;
1628 max_val = max_codepoint[bytes];
1629
1630 offset += bytes;
1631 len -= bytes;
1632
1633 /* And verify that they are good continuation bytes */
1634 do {
1635 codepoint <<= 6;
1636 codepoint |= *buf & 0x3f;
1637 if ((*buf++ & 0xc0) != 0x80)
1638 return true;
1639 } while (--bytes);
1640
1641 /* Reject codepoints that are out of range for the sequence length. */
1642 if (codepoint < min_val || codepoint > max_val)
1643 return true;
1644 /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1645 if ((codepoint & 0x1ff800) == 0xd800)
1646 return true;
1647 /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1648 if ((codepoint & 0xfffe) == 0xfffe)
1649 return true;
1650 /* So are anything in the range U+FDD0..U+FDEF. */
1651 if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1652 return true;
1653 }
1654 return false;
1655 }
1656
1657 /*
1658 * This ensures that the buffer is in proper utf8 format.
1659 *
1660 * If it isn't, it assumes any non-utf8 characters are Latin1,
1661 * and does the conversion.
1662 */
1663 static int ensure_utf8(struct strbuf *buf)
1664 {
1665 int ok = 1;
1666 size_t pos = 0;
1667
1668 for (;;) {
1669 size_t bad;
1670 unsigned char c;
1671 unsigned char replace[2];
1672
1673 if (!has_invalid_utf8(buf->buf + pos, buf->len - pos, &bad))
1674 return ok;
1675 pos += bad;
1676 ok = 0;
1677 c = buf->buf[pos];
1678 strbuf_remove(buf, pos, 1);
1679
1680 /* We know 'c' must be in the range 128-255 */
1681 replace[0] = 0xc0 + (c >> 6);
1682 replace[1] = 0x80 + (c & 0x3f);
1683 strbuf_insert(buf, pos, replace, 2);
1684 pos += 2;
1685 }
1686 }
1687
1688 static const char commit_utf8_warn[] =
1689 N_("Warning: commit message did not conform to UTF-8.\n"
1690 "You may want to amend it after fixing the message, or set the config\n"
1691 "variable i18n.commitEncoding to the encoding your project uses.\n");
1692
1693 static void write_commit_tree(struct strbuf *buffer, const char *msg, size_t msg_len,
1694 const struct object_id *tree,
1695 const struct object_id *parents, size_t parents_len,
1696 const char *author, const char *committer,
1697 const struct commit_extra_header *extra)
1698 {
1699 int encoding_is_utf8;
1700 size_t i;
1701
1702 /* Not having i18n.commitencoding is the same as having utf-8 */
1703 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1704
1705 strbuf_grow(buffer, 8192); /* should avoid reallocs for the headers */
1706 strbuf_addf(buffer, "tree %s\n", oid_to_hex(tree));
1707
1708 /*
1709 * NOTE! This ordering means that the same exact tree merged with a
1710 * different order of parents will be a _different_ changeset even
1711 * if everything else stays the same.
1712 */
1713 for (i = 0; i < parents_len; i++)
1714 strbuf_addf(buffer, "parent %s\n", oid_to_hex(&parents[i]));
1715
1716 /* Person/date information */
1717 if (!author)
1718 author = git_author_info(IDENT_STRICT);
1719 strbuf_addf(buffer, "author %s\n", author);
1720 if (!committer)
1721 committer = git_committer_info(IDENT_STRICT);
1722 strbuf_addf(buffer, "committer %s\n", committer);
1723 if (!encoding_is_utf8)
1724 strbuf_addf(buffer, "encoding %s\n", git_commit_encoding);
1725
1726 while (extra) {
1727 add_extra_header(buffer, extra);
1728 extra = extra->next;
1729 }
1730 strbuf_addch(buffer, '\n');
1731
1732 /* And add the comment */
1733 strbuf_add(buffer, msg, msg_len);
1734 }
1735
1736 int commit_tree_extended(const char *msg, size_t msg_len,
1737 const struct object_id *tree,
1738 const struct commit_list *parents, struct object_id *ret,
1739 const char *author, const char *committer,
1740 const char *sign_commit,
1741 const struct commit_extra_header *extra)
1742 {
1743 struct repository *r = the_repository;
1744 int result = 0;
1745 int encoding_is_utf8;
1746 bool warned = false;
1747 struct strbuf buffer = STRBUF_INIT, compat_buffer = STRBUF_INIT;
1748 struct strbuf sig = STRBUF_INIT, compat_sig = STRBUF_INIT;
1749 struct object_id *parent_buf = NULL, *compat_oid = NULL;
1750 struct object_id compat_oid_buf;
1751 size_t i, nparents;
1752
1753 /* Not having i18n.commitencoding is the same as having utf-8 */
1754 encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1755
1756 odb_assert_oid_type(the_repository->objects, tree, OBJ_TREE);
1757
1758 if (memchr(msg, '\0', msg_len))
1759 return error("a NUL byte in commit log message not allowed.");
1760
1761 nparents = commit_list_count(parents);
1762 CALLOC_ARRAY(parent_buf, nparents);
1763 i = 0;
1764 for (const struct commit_list *p = parents; p; p = p->next)
1765 oidcpy(&parent_buf[i++], &p->item->object.oid);
1766
1767 write_commit_tree(&buffer, msg, msg_len, tree, parent_buf, nparents, author, committer, extra);
1768
1769 /* And check the encoding. */
1770 if (encoding_is_utf8 && !ensure_utf8(&buffer)) {
1771 fprintf(stderr, _(commit_utf8_warn));
1772 warned = true;
1773 }
1774
1775 if (sign_commit && sign_buffer(&buffer, &sig, sign_commit,
1776 SIGN_BUFFER_USE_DEFAULT_KEY)) {
1777 result = -1;
1778 goto out;
1779 }
1780 if (r->compat_hash_algo) {
1781 struct commit_extra_header *compat_extra = NULL;
1782 struct object_id mapped_tree;
1783 struct object_id *mapped_parents;
1784
1785 CALLOC_ARRAY(mapped_parents, nparents);
1786
1787 if (repo_oid_to_algop(r, tree, r->compat_hash_algo, &mapped_tree)) {
1788 result = -1;
1789 free(mapped_parents);
1790 goto out;
1791 }
1792 for (i = 0; i < nparents; i++)
1793 if (repo_oid_to_algop(r, &parent_buf[i], r->compat_hash_algo, &mapped_parents[i])) {
1794 result = -1;
1795 free(mapped_parents);
1796 goto out;
1797 }
1798 if (convert_commit_extra_headers(extra, &compat_extra)) {
1799 result = -1;
1800 free(mapped_parents);
1801 goto out;
1802 }
1803 write_commit_tree(&compat_buffer, msg, msg_len, &mapped_tree,
1804 mapped_parents, nparents, author, committer, compat_extra);
1805 free_commit_extra_headers(compat_extra);
1806 free(mapped_parents);
1807
1808 if (encoding_is_utf8 && !ensure_utf8(&compat_buffer) && !warned)
1809 fprintf(stderr, _(commit_utf8_warn));
1810
1811 if (sign_commit && sign_buffer(&compat_buffer, &compat_sig,
1812 sign_commit,
1813 SIGN_BUFFER_USE_DEFAULT_KEY)) {
1814 result = -1;
1815 goto out;
1816 }
1817 }
1818
1819 if (sign_commit) {
1820 struct sig_pairs {
1821 struct strbuf *sig;
1822 const struct git_hash_algo *algo;
1823 } bufs [2] = {
1824 { &compat_sig, r->compat_hash_algo },
1825 { &sig, r->hash_algo },
1826 };
1827
1828 /*
1829 * We write algorithms in the order they were implemented in
1830 * Git to produce a stable hash when multiple algorithms are
1831 * used.
1832 */
1833 if (r->compat_hash_algo && hash_algo_by_ptr(bufs[0].algo) > hash_algo_by_ptr(bufs[1].algo))
1834 SWAP(bufs[0], bufs[1]);
1835
1836 /*
1837 * We traverse each algorithm in order, and apply the signature
1838 * to each buffer.
1839 */
1840 for (size_t i = 0; i < ARRAY_SIZE(bufs); i++) {
1841 if (!bufs[i].algo)
1842 continue;
1843 add_header_signature(&buffer, bufs[i].sig, bufs[i].algo);
1844 if (r->compat_hash_algo)
1845 add_header_signature(&compat_buffer, bufs[i].sig, bufs[i].algo);
1846 }
1847 }
1848
1849 if (r->compat_hash_algo) {
1850 hash_object_file(r->compat_hash_algo, compat_buffer.buf, compat_buffer.len,
1851 OBJ_COMMIT, &compat_oid_buf);
1852 compat_oid = &compat_oid_buf;
1853 }
1854
1855 result = odb_write_object_ext(the_repository->objects, buffer.buf, buffer.len,
1856 OBJ_COMMIT, ret, compat_oid, 0);
1857 out:
1858 free(parent_buf);
1859 strbuf_release(&buffer);
1860 strbuf_release(&compat_buffer);
1861 strbuf_release(&sig);
1862 strbuf_release(&compat_sig);
1863 return result;
1864 }
1865
1866 define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1867 static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1868
1869 struct merge_remote_desc *merge_remote_util(const struct commit *commit)
1870 {
1871 return *merge_desc_slab_at(&merge_desc_slab, commit);
1872 }
1873
1874 void set_merge_remote_desc(struct commit *commit,
1875 const char *name, struct object *obj)
1876 {
1877 struct merge_remote_desc *desc;
1878 FLEX_ALLOC_STR(desc, name, name);
1879 desc->obj = obj;
1880 *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1881 }
1882
1883 struct commit *get_merge_parent(const char *name)
1884 {
1885 struct object *obj;
1886 struct commit *commit;
1887 struct object_id oid;
1888 if (repo_get_oid(the_repository, name, &oid))
1889 return NULL;
1890 obj = parse_object(the_repository, &oid);
1891 commit = (struct commit *)repo_peel_to_type(the_repository, name, 0,
1892 obj, OBJ_COMMIT);
1893 if (commit && !merge_remote_util(commit))
1894 set_merge_remote_desc(commit, name, obj);
1895 return commit;
1896 }
1897
1898 /*
1899 * Append a commit to the end of the commit_list.
1900 *
1901 * next starts by pointing to the variable that holds the head of an
1902 * empty commit_list, and is updated to point to the "next" field of
1903 * the last item on the list as new commits are appended.
1904 *
1905 * Usage example:
1906 *
1907 * struct commit_list *list;
1908 * struct commit_list **next = &list;
1909 *
1910 * next = commit_list_append(c1, next);
1911 * next = commit_list_append(c2, next);
1912 * assert(commit_list_count(list) == 2);
1913 * return list;
1914 */
1915 struct commit_list **commit_list_append(struct commit *commit,
1916 struct commit_list **next)
1917 {
1918 struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1919 new_commit->item = commit;
1920 *next = new_commit;
1921 new_commit->next = NULL;
1922 return &new_commit->next;
1923 }
1924
1925 const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1926 {
1927 int key_len = strlen(key);
1928 const char *line = msg;
1929
1930 while (line) {
1931 const char *eol = strchrnul(line, '\n');
1932
1933 if (line == eol)
1934 return NULL;
1935
1936 if (eol - line > key_len &&
1937 !strncmp(line, key, key_len) &&
1938 line[key_len] == ' ') {
1939 *out_len = eol - line - key_len - 1;
1940 return line + key_len + 1;
1941 }
1942 line = *eol ? eol + 1 : NULL;
1943 }
1944 return NULL;
1945 }
1946
1947 /*
1948 * Inspect the given string and determine the true "end" of the log message, in
1949 * order to find where to put a new Signed-off-by trailer. Ignored are
1950 * trailing comment lines and blank lines. To support "git commit -s
1951 * --amend" on an existing commit, we also ignore "Conflicts:". To
1952 * support "git commit -v", we truncate at cut lines.
1953 *
1954 * Returns the number of bytes from the tail to ignore, to be fed as
1955 * the second parameter to append_signoff().
1956 */
1957 size_t ignored_log_message_bytes(const char *buf, size_t len)
1958 {
1959 size_t boc = 0;
1960 size_t bol = 0;
1961 int in_old_conflicts_block = 0;
1962 size_t cutoff = wt_status_locate_end(buf, len);
1963
1964 while (bol < cutoff) {
1965 const char *next_line = memchr(buf + bol, '\n', len - bol);
1966
1967 if (!next_line)
1968 next_line = buf + len;
1969 else
1970 next_line++;
1971
1972 if (starts_with_mem(buf + bol, cutoff - bol, comment_line_str) ||
1973 buf[bol] == '\n') {
1974 /* is this the first of the run of comments? */
1975 if (!boc)
1976 boc = bol;
1977 /* otherwise, it is just continuing */
1978 } else if (starts_with(buf + bol, "Conflicts:\n")) {
1979 in_old_conflicts_block = 1;
1980 if (!boc)
1981 boc = bol;
1982 } else if (in_old_conflicts_block && buf[bol] == '\t') {
1983 ; /* a pathname in the conflicts block */
1984 } else if (boc) {
1985 /* the previous was not trailing comment */
1986 boc = 0;
1987 in_old_conflicts_block = 0;
1988 }
1989 bol = next_line - buf;
1990 }
1991 return boc ? len - boc : len - cutoff;
1992 }
1993
1994 int run_commit_hook(int editor_is_used, const char *index_file,
1995 int *invoked_hook, const char *name, ...)
1996 {
1997 struct run_hooks_opt opt = RUN_HOOKS_OPT_INIT_FORCE_SERIAL;
1998 va_list args;
1999 const char *arg;
2000
2001 strvec_pushf(&opt.env, "GIT_INDEX_FILE=%s", index_file);
2002
2003 /*
2004 * Let the hook know that no editor will be launched.
2005 */
2006 if (!editor_is_used)
2007 strvec_push(&opt.env, "GIT_EDITOR=:");
2008
2009 va_start(args, name);
2010 while ((arg = va_arg(args, const char *)))
2011 strvec_push(&opt.args, arg);
2012 va_end(args);
2013
2014 opt.invoked_hook = invoked_hook;
2015 return run_hooks_opt(the_repository, name, &opt);
2016 }
2017
2018 void commit_stack_init(struct commit_stack *stack)
2019 {
2020 stack->items = NULL;
2021 stack->nr = stack->alloc = 0;
2022 }
2023
2024 void commit_stack_grow(struct commit_stack *stack, size_t extra)
2025 {
2026 ALLOC_GROW(stack->items, st_add(stack->nr, extra), stack->alloc);
2027 }
2028
2029 void commit_stack_push(struct commit_stack *stack, struct commit *commit)
2030 {
2031 commit_stack_grow(stack, 1);
2032 stack->items[stack->nr++] = commit;
2033 }
2034
2035 struct commit *commit_stack_pop(struct commit_stack *stack)
2036 {
2037 return stack->nr ? stack->items[--stack->nr] : NULL;
2038 }
2039
2040 void commit_stack_clear(struct commit_stack *stack)
2041 {
2042 free(stack->items);
2043 commit_stack_init(stack);
2044 }