]>
Commit | Line | Data |
---|---|---|
1 | #include "cache.h" | |
2 | #include "tag.h" | |
3 | #include "commit.h" | |
4 | #include "commit-graph.h" | |
5 | #include "repository.h" | |
6 | #include "object-store.h" | |
7 | #include "pkt-line.h" | |
8 | #include "utf8.h" | |
9 | #include "diff.h" | |
10 | #include "revision.h" | |
11 | #include "notes.h" | |
12 | #include "alloc.h" | |
13 | #include "gpg-interface.h" | |
14 | #include "mergesort.h" | |
15 | #include "commit-slab.h" | |
16 | #include "prio-queue.h" | |
17 | #include "hash-lookup.h" | |
18 | #include "wt-status.h" | |
19 | #include "advice.h" | |
20 | #include "refs.h" | |
21 | #include "commit-reach.h" | |
22 | #include "run-command.h" | |
23 | #include "shallow.h" | |
24 | #include "hook.h" | |
25 | ||
26 | static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **); | |
27 | ||
28 | int save_commit_buffer = 1; | |
29 | int no_graft_file_deprecated_advice; | |
30 | ||
31 | const char *commit_type = "commit"; | |
32 | ||
33 | struct commit *lookup_commit_reference_gently(struct repository *r, | |
34 | const struct object_id *oid, int quiet) | |
35 | { | |
36 | struct object *obj = deref_tag(r, | |
37 | parse_object(r, oid), | |
38 | NULL, 0); | |
39 | ||
40 | if (!obj) | |
41 | return NULL; | |
42 | return object_as_type(obj, OBJ_COMMIT, quiet); | |
43 | } | |
44 | ||
45 | struct commit *lookup_commit_reference(struct repository *r, const struct object_id *oid) | |
46 | { | |
47 | return lookup_commit_reference_gently(r, oid, 0); | |
48 | } | |
49 | ||
50 | struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name) | |
51 | { | |
52 | struct commit *c = lookup_commit_reference(the_repository, oid); | |
53 | if (!c) | |
54 | die(_("could not parse %s"), ref_name); | |
55 | if (!oideq(oid, &c->object.oid)) { | |
56 | warning(_("%s %s is not a commit!"), | |
57 | ref_name, oid_to_hex(oid)); | |
58 | } | |
59 | return c; | |
60 | } | |
61 | ||
62 | struct commit *lookup_commit_object(struct repository *r, | |
63 | const struct object_id *oid) | |
64 | { | |
65 | struct object *obj = parse_object(r, oid); | |
66 | return obj ? object_as_type(obj, OBJ_COMMIT, 0) : NULL; | |
67 | ||
68 | } | |
69 | ||
70 | struct commit *lookup_commit(struct repository *r, const struct object_id *oid) | |
71 | { | |
72 | struct object *obj = lookup_object(r, oid); | |
73 | if (!obj) | |
74 | return create_object(r, oid, alloc_commit_node(r)); | |
75 | return object_as_type(obj, OBJ_COMMIT, 0); | |
76 | } | |
77 | ||
78 | struct commit *lookup_commit_reference_by_name(const char *name) | |
79 | { | |
80 | struct object_id oid; | |
81 | struct commit *commit; | |
82 | ||
83 | if (get_oid_committish(name, &oid)) | |
84 | return NULL; | |
85 | commit = lookup_commit_reference(the_repository, &oid); | |
86 | if (parse_commit(commit)) | |
87 | return NULL; | |
88 | return commit; | |
89 | } | |
90 | ||
91 | static timestamp_t parse_commit_date(const char *buf, const char *tail) | |
92 | { | |
93 | const char *dateptr; | |
94 | const char *eol; | |
95 | ||
96 | if (buf + 6 >= tail) | |
97 | return 0; | |
98 | if (memcmp(buf, "author", 6)) | |
99 | return 0; | |
100 | while (buf < tail && *buf++ != '\n') | |
101 | /* nada */; | |
102 | if (buf + 9 >= tail) | |
103 | return 0; | |
104 | if (memcmp(buf, "committer", 9)) | |
105 | return 0; | |
106 | ||
107 | /* | |
108 | * Jump to end-of-line so that we can walk backwards to find the | |
109 | * end-of-email ">". This is more forgiving of malformed cases | |
110 | * because unexpected characters tend to be in the name and email | |
111 | * fields. | |
112 | */ | |
113 | eol = memchr(buf, '\n', tail - buf); | |
114 | if (!eol) | |
115 | return 0; | |
116 | dateptr = eol; | |
117 | while (dateptr > buf && dateptr[-1] != '>') | |
118 | dateptr--; | |
119 | if (dateptr == buf) | |
120 | return 0; | |
121 | ||
122 | /* | |
123 | * Trim leading whitespace, but make sure we have at least one | |
124 | * non-whitespace character, as parse_timestamp() will otherwise walk | |
125 | * right past the newline we found in "eol" when skipping whitespace | |
126 | * itself. | |
127 | * | |
128 | * In theory it would be sufficient to allow any character not matched | |
129 | * by isspace(), but there's a catch: our isspace() does not | |
130 | * necessarily match the behavior of parse_timestamp(), as the latter | |
131 | * is implemented by system routines which match more exotic control | |
132 | * codes, or even locale-dependent sequences. | |
133 | * | |
134 | * Since we expect the timestamp to be a number, we can check for that. | |
135 | * Anything else (e.g., a non-numeric token like "foo") would just | |
136 | * cause parse_timestamp() to return 0 anyway. | |
137 | */ | |
138 | while (dateptr < eol && isspace(*dateptr)) | |
139 | dateptr++; | |
140 | if (!isdigit(*dateptr) && *dateptr != '-') | |
141 | return 0; | |
142 | ||
143 | /* | |
144 | * We know there is at least one digit (or dash), so we'll begin | |
145 | * parsing there and stop at worst case at eol. | |
146 | */ | |
147 | return parse_timestamp(dateptr, NULL, 10); | |
148 | } | |
149 | ||
150 | static const struct object_id *commit_graft_oid_access(size_t index, const void *table) | |
151 | { | |
152 | const struct commit_graft * const *commit_graft_table = table; | |
153 | return &commit_graft_table[index]->oid; | |
154 | } | |
155 | ||
156 | int commit_graft_pos(struct repository *r, const struct object_id *oid) | |
157 | { | |
158 | return oid_pos(oid, r->parsed_objects->grafts, | |
159 | r->parsed_objects->grafts_nr, | |
160 | commit_graft_oid_access); | |
161 | } | |
162 | ||
163 | static void unparse_commit(struct repository *r, const struct object_id *oid) | |
164 | { | |
165 | struct commit *c = lookup_commit(r, oid); | |
166 | ||
167 | if (!c->object.parsed) | |
168 | return; | |
169 | free_commit_list(c->parents); | |
170 | c->parents = NULL; | |
171 | c->object.parsed = 0; | |
172 | } | |
173 | ||
174 | int register_commit_graft(struct repository *r, struct commit_graft *graft, | |
175 | int ignore_dups) | |
176 | { | |
177 | int pos = commit_graft_pos(r, &graft->oid); | |
178 | ||
179 | if (0 <= pos) { | |
180 | if (ignore_dups) | |
181 | free(graft); | |
182 | else { | |
183 | free(r->parsed_objects->grafts[pos]); | |
184 | r->parsed_objects->grafts[pos] = graft; | |
185 | } | |
186 | return 1; | |
187 | } | |
188 | pos = -pos - 1; | |
189 | ALLOC_GROW(r->parsed_objects->grafts, | |
190 | r->parsed_objects->grafts_nr + 1, | |
191 | r->parsed_objects->grafts_alloc); | |
192 | r->parsed_objects->grafts_nr++; | |
193 | if (pos < r->parsed_objects->grafts_nr) | |
194 | memmove(r->parsed_objects->grafts + pos + 1, | |
195 | r->parsed_objects->grafts + pos, | |
196 | (r->parsed_objects->grafts_nr - pos - 1) * | |
197 | sizeof(*r->parsed_objects->grafts)); | |
198 | r->parsed_objects->grafts[pos] = graft; | |
199 | unparse_commit(r, &graft->oid); | |
200 | return 0; | |
201 | } | |
202 | ||
203 | struct commit_graft *read_graft_line(struct strbuf *line) | |
204 | { | |
205 | /* The format is just "Commit Parent1 Parent2 ...\n" */ | |
206 | int i, phase; | |
207 | const char *tail = NULL; | |
208 | struct commit_graft *graft = NULL; | |
209 | struct object_id dummy_oid, *oid; | |
210 | ||
211 | strbuf_rtrim(line); | |
212 | if (!line->len || line->buf[0] == '#') | |
213 | return NULL; | |
214 | /* | |
215 | * phase 0 verifies line, counts hashes in line and allocates graft | |
216 | * phase 1 fills graft | |
217 | */ | |
218 | for (phase = 0; phase < 2; phase++) { | |
219 | oid = graft ? &graft->oid : &dummy_oid; | |
220 | if (parse_oid_hex(line->buf, oid, &tail)) | |
221 | goto bad_graft_data; | |
222 | for (i = 0; *tail != '\0'; i++) { | |
223 | oid = graft ? &graft->parent[i] : &dummy_oid; | |
224 | if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail)) | |
225 | goto bad_graft_data; | |
226 | } | |
227 | if (!graft) { | |
228 | graft = xmalloc(st_add(sizeof(*graft), | |
229 | st_mult(sizeof(struct object_id), i))); | |
230 | graft->nr_parent = i; | |
231 | } | |
232 | } | |
233 | return graft; | |
234 | ||
235 | bad_graft_data: | |
236 | error("bad graft data: %s", line->buf); | |
237 | assert(!graft); | |
238 | return NULL; | |
239 | } | |
240 | ||
241 | static int read_graft_file(struct repository *r, const char *graft_file) | |
242 | { | |
243 | FILE *fp = fopen_or_warn(graft_file, "r"); | |
244 | struct strbuf buf = STRBUF_INIT; | |
245 | if (!fp) | |
246 | return -1; | |
247 | if (!no_graft_file_deprecated_advice && | |
248 | advice_enabled(ADVICE_GRAFT_FILE_DEPRECATED)) | |
249 | advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n" | |
250 | "and will be removed in a future Git version.\n" | |
251 | "\n" | |
252 | "Please use \"git replace --convert-graft-file\"\n" | |
253 | "to convert the grafts into replace refs.\n" | |
254 | "\n" | |
255 | "Turn this message off by running\n" | |
256 | "\"git config advice.graftFileDeprecated false\"")); | |
257 | while (!strbuf_getwholeline(&buf, fp, '\n')) { | |
258 | /* The format is just "Commit Parent1 Parent2 ...\n" */ | |
259 | struct commit_graft *graft = read_graft_line(&buf); | |
260 | if (!graft) | |
261 | continue; | |
262 | if (register_commit_graft(r, graft, 1)) | |
263 | error("duplicate graft data: %s", buf.buf); | |
264 | } | |
265 | fclose(fp); | |
266 | strbuf_release(&buf); | |
267 | return 0; | |
268 | } | |
269 | ||
270 | void prepare_commit_graft(struct repository *r) | |
271 | { | |
272 | char *graft_file; | |
273 | ||
274 | if (r->parsed_objects->commit_graft_prepared) | |
275 | return; | |
276 | if (!startup_info->have_repository) | |
277 | return; | |
278 | ||
279 | graft_file = get_graft_file(r); | |
280 | read_graft_file(r, graft_file); | |
281 | /* make sure shallows are read */ | |
282 | is_repository_shallow(r); | |
283 | r->parsed_objects->commit_graft_prepared = 1; | |
284 | } | |
285 | ||
286 | struct commit_graft *lookup_commit_graft(struct repository *r, const struct object_id *oid) | |
287 | { | |
288 | int pos; | |
289 | prepare_commit_graft(r); | |
290 | pos = commit_graft_pos(r, oid); | |
291 | if (pos < 0) | |
292 | return NULL; | |
293 | return r->parsed_objects->grafts[pos]; | |
294 | } | |
295 | ||
296 | int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data) | |
297 | { | |
298 | int i, ret; | |
299 | for (i = ret = 0; i < the_repository->parsed_objects->grafts_nr && !ret; i++) | |
300 | ret = fn(the_repository->parsed_objects->grafts[i], cb_data); | |
301 | return ret; | |
302 | } | |
303 | ||
304 | void reset_commit_grafts(struct repository *r) | |
305 | { | |
306 | int i; | |
307 | ||
308 | for (i = 0; i < r->parsed_objects->grafts_nr; i++) { | |
309 | unparse_commit(r, &r->parsed_objects->grafts[i]->oid); | |
310 | free(r->parsed_objects->grafts[i]); | |
311 | } | |
312 | r->parsed_objects->grafts_nr = 0; | |
313 | r->parsed_objects->commit_graft_prepared = 0; | |
314 | } | |
315 | ||
316 | struct commit_buffer { | |
317 | void *buffer; | |
318 | unsigned long size; | |
319 | }; | |
320 | define_commit_slab(buffer_slab, struct commit_buffer); | |
321 | ||
322 | struct buffer_slab *allocate_commit_buffer_slab(void) | |
323 | { | |
324 | struct buffer_slab *bs = xmalloc(sizeof(*bs)); | |
325 | init_buffer_slab(bs); | |
326 | return bs; | |
327 | } | |
328 | ||
329 | void free_commit_buffer_slab(struct buffer_slab *bs) | |
330 | { | |
331 | clear_buffer_slab(bs); | |
332 | free(bs); | |
333 | } | |
334 | ||
335 | void set_commit_buffer(struct repository *r, struct commit *commit, void *buffer, unsigned long size) | |
336 | { | |
337 | struct commit_buffer *v = buffer_slab_at( | |
338 | r->parsed_objects->buffer_slab, commit); | |
339 | v->buffer = buffer; | |
340 | v->size = size; | |
341 | } | |
342 | ||
343 | const void *get_cached_commit_buffer(struct repository *r, const struct commit *commit, unsigned long *sizep) | |
344 | { | |
345 | struct commit_buffer *v = buffer_slab_peek( | |
346 | r->parsed_objects->buffer_slab, commit); | |
347 | if (!v) { | |
348 | if (sizep) | |
349 | *sizep = 0; | |
350 | return NULL; | |
351 | } | |
352 | if (sizep) | |
353 | *sizep = v->size; | |
354 | return v->buffer; | |
355 | } | |
356 | ||
357 | const void *repo_get_commit_buffer(struct repository *r, | |
358 | const struct commit *commit, | |
359 | unsigned long *sizep) | |
360 | { | |
361 | const void *ret = get_cached_commit_buffer(r, commit, sizep); | |
362 | if (!ret) { | |
363 | enum object_type type; | |
364 | unsigned long size; | |
365 | ret = repo_read_object_file(r, &commit->object.oid, &type, &size); | |
366 | if (!ret) | |
367 | die("cannot read commit object %s", | |
368 | oid_to_hex(&commit->object.oid)); | |
369 | if (type != OBJ_COMMIT) | |
370 | die("expected commit for %s, got %s", | |
371 | oid_to_hex(&commit->object.oid), type_name(type)); | |
372 | if (sizep) | |
373 | *sizep = size; | |
374 | } | |
375 | return ret; | |
376 | } | |
377 | ||
378 | void repo_unuse_commit_buffer(struct repository *r, | |
379 | const struct commit *commit, | |
380 | const void *buffer) | |
381 | { | |
382 | struct commit_buffer *v = buffer_slab_peek( | |
383 | r->parsed_objects->buffer_slab, commit); | |
384 | if (!(v && v->buffer == buffer)) | |
385 | free((void *)buffer); | |
386 | } | |
387 | ||
388 | void free_commit_buffer(struct parsed_object_pool *pool, struct commit *commit) | |
389 | { | |
390 | struct commit_buffer *v = buffer_slab_peek( | |
391 | pool->buffer_slab, commit); | |
392 | if (v) { | |
393 | FREE_AND_NULL(v->buffer); | |
394 | v->size = 0; | |
395 | } | |
396 | } | |
397 | ||
398 | static inline void set_commit_tree(struct commit *c, struct tree *t) | |
399 | { | |
400 | c->maybe_tree = t; | |
401 | } | |
402 | ||
403 | struct tree *repo_get_commit_tree(struct repository *r, | |
404 | const struct commit *commit) | |
405 | { | |
406 | if (commit->maybe_tree || !commit->object.parsed) | |
407 | return commit->maybe_tree; | |
408 | ||
409 | if (commit_graph_position(commit) != COMMIT_NOT_FROM_GRAPH) | |
410 | return get_commit_tree_in_graph(r, commit); | |
411 | ||
412 | return NULL; | |
413 | } | |
414 | ||
415 | struct object_id *get_commit_tree_oid(const struct commit *commit) | |
416 | { | |
417 | struct tree *tree = get_commit_tree(commit); | |
418 | return tree ? &tree->object.oid : NULL; | |
419 | } | |
420 | ||
421 | void release_commit_memory(struct parsed_object_pool *pool, struct commit *c) | |
422 | { | |
423 | set_commit_tree(c, NULL); | |
424 | free_commit_buffer(pool, c); | |
425 | c->index = 0; | |
426 | free_commit_list(c->parents); | |
427 | ||
428 | c->object.parsed = 0; | |
429 | } | |
430 | ||
431 | const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep) | |
432 | { | |
433 | struct commit_buffer *v = buffer_slab_peek( | |
434 | the_repository->parsed_objects->buffer_slab, commit); | |
435 | void *ret; | |
436 | ||
437 | if (!v) { | |
438 | if (sizep) | |
439 | *sizep = 0; | |
440 | return NULL; | |
441 | } | |
442 | ret = v->buffer; | |
443 | if (sizep) | |
444 | *sizep = v->size; | |
445 | ||
446 | v->buffer = NULL; | |
447 | v->size = 0; | |
448 | return ret; | |
449 | } | |
450 | ||
451 | int parse_commit_buffer(struct repository *r, struct commit *item, const void *buffer, unsigned long size, int check_graph) | |
452 | { | |
453 | const char *tail = buffer; | |
454 | const char *bufptr = buffer; | |
455 | struct object_id parent; | |
456 | struct commit_list **pptr; | |
457 | struct commit_graft *graft; | |
458 | const int tree_entry_len = the_hash_algo->hexsz + 5; | |
459 | const int parent_entry_len = the_hash_algo->hexsz + 7; | |
460 | struct tree *tree; | |
461 | ||
462 | if (item->object.parsed) | |
463 | return 0; | |
464 | /* | |
465 | * Presumably this is leftover from an earlier failed parse; | |
466 | * clear it out in preparation for us re-parsing (we'll hit the | |
467 | * same error, but that's good, since it lets our caller know | |
468 | * the result cannot be trusted. | |
469 | */ | |
470 | free_commit_list(item->parents); | |
471 | item->parents = NULL; | |
472 | ||
473 | tail += size; | |
474 | if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) || | |
475 | bufptr[tree_entry_len] != '\n') | |
476 | return error("bogus commit object %s", oid_to_hex(&item->object.oid)); | |
477 | if (get_oid_hex(bufptr + 5, &parent) < 0) | |
478 | return error("bad tree pointer in commit %s", | |
479 | oid_to_hex(&item->object.oid)); | |
480 | tree = lookup_tree(r, &parent); | |
481 | if (!tree) | |
482 | return error("bad tree pointer %s in commit %s", | |
483 | oid_to_hex(&parent), | |
484 | oid_to_hex(&item->object.oid)); | |
485 | set_commit_tree(item, tree); | |
486 | bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */ | |
487 | pptr = &item->parents; | |
488 | ||
489 | graft = lookup_commit_graft(r, &item->object.oid); | |
490 | if (graft) | |
491 | r->parsed_objects->substituted_parent = 1; | |
492 | while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) { | |
493 | struct commit *new_parent; | |
494 | ||
495 | if (tail <= bufptr + parent_entry_len + 1 || | |
496 | get_oid_hex(bufptr + 7, &parent) || | |
497 | bufptr[parent_entry_len] != '\n') | |
498 | return error("bad parents in commit %s", oid_to_hex(&item->object.oid)); | |
499 | bufptr += parent_entry_len + 1; | |
500 | /* | |
501 | * The clone is shallow if nr_parent < 0, and we must | |
502 | * not traverse its real parents even when we unhide them. | |
503 | */ | |
504 | if (graft && (graft->nr_parent < 0 || grafts_replace_parents)) | |
505 | continue; | |
506 | new_parent = lookup_commit(r, &parent); | |
507 | if (!new_parent) | |
508 | return error("bad parent %s in commit %s", | |
509 | oid_to_hex(&parent), | |
510 | oid_to_hex(&item->object.oid)); | |
511 | pptr = &commit_list_insert(new_parent, pptr)->next; | |
512 | } | |
513 | if (graft) { | |
514 | int i; | |
515 | struct commit *new_parent; | |
516 | for (i = 0; i < graft->nr_parent; i++) { | |
517 | new_parent = lookup_commit(r, | |
518 | &graft->parent[i]); | |
519 | if (!new_parent) | |
520 | return error("bad graft parent %s in commit %s", | |
521 | oid_to_hex(&graft->parent[i]), | |
522 | oid_to_hex(&item->object.oid)); | |
523 | pptr = &commit_list_insert(new_parent, pptr)->next; | |
524 | } | |
525 | } | |
526 | item->date = parse_commit_date(bufptr, tail); | |
527 | ||
528 | if (check_graph) | |
529 | load_commit_graph_info(r, item); | |
530 | ||
531 | item->object.parsed = 1; | |
532 | return 0; | |
533 | } | |
534 | ||
535 | int repo_parse_commit_internal(struct repository *r, | |
536 | struct commit *item, | |
537 | int quiet_on_missing, | |
538 | int use_commit_graph) | |
539 | { | |
540 | enum object_type type; | |
541 | void *buffer; | |
542 | unsigned long size; | |
543 | struct object_info oi = { | |
544 | .typep = &type, | |
545 | .sizep = &size, | |
546 | .contentp = &buffer, | |
547 | }; | |
548 | /* | |
549 | * Git does not support partial clones that exclude commits, so set | |
550 | * OBJECT_INFO_SKIP_FETCH_OBJECT to fail fast when an object is missing. | |
551 | */ | |
552 | int flags = OBJECT_INFO_LOOKUP_REPLACE | OBJECT_INFO_SKIP_FETCH_OBJECT | | |
553 | OBJECT_INFO_DIE_IF_CORRUPT; | |
554 | int ret; | |
555 | ||
556 | if (!item) | |
557 | return -1; | |
558 | if (item->object.parsed) | |
559 | return 0; | |
560 | if (use_commit_graph && parse_commit_in_graph(r, item)) | |
561 | return 0; | |
562 | ||
563 | if (oid_object_info_extended(r, &item->object.oid, &oi, flags) < 0) | |
564 | return quiet_on_missing ? -1 : | |
565 | error("Could not read %s", | |
566 | oid_to_hex(&item->object.oid)); | |
567 | if (type != OBJ_COMMIT) { | |
568 | free(buffer); | |
569 | return error("Object %s not a commit", | |
570 | oid_to_hex(&item->object.oid)); | |
571 | } | |
572 | ||
573 | ret = parse_commit_buffer(r, item, buffer, size, 0); | |
574 | if (save_commit_buffer && !ret) { | |
575 | set_commit_buffer(r, item, buffer, size); | |
576 | return 0; | |
577 | } | |
578 | free(buffer); | |
579 | return ret; | |
580 | } | |
581 | ||
582 | int repo_parse_commit_gently(struct repository *r, | |
583 | struct commit *item, int quiet_on_missing) | |
584 | { | |
585 | return repo_parse_commit_internal(r, item, quiet_on_missing, 1); | |
586 | } | |
587 | ||
588 | void parse_commit_or_die(struct commit *item) | |
589 | { | |
590 | if (parse_commit(item)) | |
591 | die("unable to parse commit %s", | |
592 | item ? oid_to_hex(&item->object.oid) : "(null)"); | |
593 | } | |
594 | ||
595 | int find_commit_subject(const char *commit_buffer, const char **subject) | |
596 | { | |
597 | const char *eol; | |
598 | const char *p = commit_buffer; | |
599 | ||
600 | while (*p && (*p != '\n' || p[1] != '\n')) | |
601 | p++; | |
602 | if (*p) { | |
603 | p = skip_blank_lines(p + 2); | |
604 | eol = strchrnul(p, '\n'); | |
605 | } else | |
606 | eol = p; | |
607 | ||
608 | *subject = p; | |
609 | ||
610 | return eol - p; | |
611 | } | |
612 | ||
613 | size_t commit_subject_length(const char *body) | |
614 | { | |
615 | const char *p = body; | |
616 | while (*p) { | |
617 | const char *next = skip_blank_lines(p); | |
618 | if (next != p) | |
619 | break; | |
620 | p = strchrnul(p, '\n'); | |
621 | if (*p) | |
622 | p++; | |
623 | } | |
624 | return p - body; | |
625 | } | |
626 | ||
627 | struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p) | |
628 | { | |
629 | struct commit_list *new_list = xmalloc(sizeof(struct commit_list)); | |
630 | new_list->item = item; | |
631 | new_list->next = *list_p; | |
632 | *list_p = new_list; | |
633 | return new_list; | |
634 | } | |
635 | ||
636 | int commit_list_contains(struct commit *item, struct commit_list *list) | |
637 | { | |
638 | while (list) { | |
639 | if (list->item == item) | |
640 | return 1; | |
641 | list = list->next; | |
642 | } | |
643 | ||
644 | return 0; | |
645 | } | |
646 | ||
647 | unsigned commit_list_count(const struct commit_list *l) | |
648 | { | |
649 | unsigned c = 0; | |
650 | for (; l; l = l->next ) | |
651 | c++; | |
652 | return c; | |
653 | } | |
654 | ||
655 | struct commit_list *copy_commit_list(struct commit_list *list) | |
656 | { | |
657 | struct commit_list *head = NULL; | |
658 | struct commit_list **pp = &head; | |
659 | while (list) { | |
660 | pp = commit_list_append(list->item, pp); | |
661 | list = list->next; | |
662 | } | |
663 | return head; | |
664 | } | |
665 | ||
666 | struct commit_list *reverse_commit_list(struct commit_list *list) | |
667 | { | |
668 | struct commit_list *next = NULL, *current, *backup; | |
669 | for (current = list; current; current = backup) { | |
670 | backup = current->next; | |
671 | current->next = next; | |
672 | next = current; | |
673 | } | |
674 | return next; | |
675 | } | |
676 | ||
677 | void free_commit_list(struct commit_list *list) | |
678 | { | |
679 | while (list) | |
680 | pop_commit(&list); | |
681 | } | |
682 | ||
683 | struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list) | |
684 | { | |
685 | struct commit_list **pp = list; | |
686 | struct commit_list *p; | |
687 | while ((p = *pp) != NULL) { | |
688 | if (p->item->date < item->date) { | |
689 | break; | |
690 | } | |
691 | pp = &p->next; | |
692 | } | |
693 | return commit_list_insert(item, pp); | |
694 | } | |
695 | ||
696 | static int commit_list_compare_by_date(const struct commit_list *a, | |
697 | const struct commit_list *b) | |
698 | { | |
699 | timestamp_t a_date = a->item->date; | |
700 | timestamp_t b_date = b->item->date; | |
701 | if (a_date < b_date) | |
702 | return 1; | |
703 | if (a_date > b_date) | |
704 | return -1; | |
705 | return 0; | |
706 | } | |
707 | ||
708 | DEFINE_LIST_SORT(static, commit_list_sort, struct commit_list, next); | |
709 | ||
710 | void commit_list_sort_by_date(struct commit_list **list) | |
711 | { | |
712 | commit_list_sort(list, commit_list_compare_by_date); | |
713 | } | |
714 | ||
715 | struct commit *pop_most_recent_commit(struct commit_list **list, | |
716 | unsigned int mark) | |
717 | { | |
718 | struct commit *ret = pop_commit(list); | |
719 | struct commit_list *parents = ret->parents; | |
720 | ||
721 | while (parents) { | |
722 | struct commit *commit = parents->item; | |
723 | if (!parse_commit(commit) && !(commit->object.flags & mark)) { | |
724 | commit->object.flags |= mark; | |
725 | commit_list_insert_by_date(commit, list); | |
726 | } | |
727 | parents = parents->next; | |
728 | } | |
729 | return ret; | |
730 | } | |
731 | ||
732 | static void clear_commit_marks_1(struct commit_list **plist, | |
733 | struct commit *commit, unsigned int mark) | |
734 | { | |
735 | while (commit) { | |
736 | struct commit_list *parents; | |
737 | ||
738 | if (!(mark & commit->object.flags)) | |
739 | return; | |
740 | ||
741 | commit->object.flags &= ~mark; | |
742 | ||
743 | parents = commit->parents; | |
744 | if (!parents) | |
745 | return; | |
746 | ||
747 | while ((parents = parents->next)) { | |
748 | if (parents->item->object.flags & mark) | |
749 | commit_list_insert(parents->item, plist); | |
750 | } | |
751 | ||
752 | commit = commit->parents->item; | |
753 | } | |
754 | } | |
755 | ||
756 | void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark) | |
757 | { | |
758 | struct commit_list *list = NULL; | |
759 | ||
760 | while (nr--) { | |
761 | clear_commit_marks_1(&list, *commit, mark); | |
762 | commit++; | |
763 | } | |
764 | while (list) | |
765 | clear_commit_marks_1(&list, pop_commit(&list), mark); | |
766 | } | |
767 | ||
768 | void clear_commit_marks(struct commit *commit, unsigned int mark) | |
769 | { | |
770 | clear_commit_marks_many(1, &commit, mark); | |
771 | } | |
772 | ||
773 | struct commit *pop_commit(struct commit_list **stack) | |
774 | { | |
775 | struct commit_list *top = *stack; | |
776 | struct commit *item = top ? top->item : NULL; | |
777 | ||
778 | if (top) { | |
779 | *stack = top->next; | |
780 | free(top); | |
781 | } | |
782 | return item; | |
783 | } | |
784 | ||
785 | /* | |
786 | * Topological sort support | |
787 | */ | |
788 | ||
789 | /* count number of children that have not been emitted */ | |
790 | define_commit_slab(indegree_slab, int); | |
791 | ||
792 | define_commit_slab(author_date_slab, timestamp_t); | |
793 | ||
794 | void record_author_date(struct author_date_slab *author_date, | |
795 | struct commit *commit) | |
796 | { | |
797 | const char *buffer = get_commit_buffer(commit, NULL); | |
798 | struct ident_split ident; | |
799 | const char *ident_line; | |
800 | size_t ident_len; | |
801 | char *date_end; | |
802 | timestamp_t date; | |
803 | ||
804 | ident_line = find_commit_header(buffer, "author", &ident_len); | |
805 | if (!ident_line) | |
806 | goto fail_exit; /* no author line */ | |
807 | if (split_ident_line(&ident, ident_line, ident_len) || | |
808 | !ident.date_begin || !ident.date_end) | |
809 | goto fail_exit; /* malformed "author" line */ | |
810 | ||
811 | date = parse_timestamp(ident.date_begin, &date_end, 10); | |
812 | if (date_end != ident.date_end) | |
813 | goto fail_exit; /* malformed date */ | |
814 | *(author_date_slab_at(author_date, commit)) = date; | |
815 | ||
816 | fail_exit: | |
817 | unuse_commit_buffer(commit, buffer); | |
818 | } | |
819 | ||
820 | int compare_commits_by_author_date(const void *a_, const void *b_, | |
821 | void *cb_data) | |
822 | { | |
823 | const struct commit *a = a_, *b = b_; | |
824 | struct author_date_slab *author_date = cb_data; | |
825 | timestamp_t a_date = *(author_date_slab_at(author_date, a)); | |
826 | timestamp_t b_date = *(author_date_slab_at(author_date, b)); | |
827 | ||
828 | /* newer commits with larger date first */ | |
829 | if (a_date < b_date) | |
830 | return 1; | |
831 | else if (a_date > b_date) | |
832 | return -1; | |
833 | return 0; | |
834 | } | |
835 | ||
836 | int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, void *unused) | |
837 | { | |
838 | const struct commit *a = a_, *b = b_; | |
839 | const timestamp_t generation_a = commit_graph_generation(a), | |
840 | generation_b = commit_graph_generation(b); | |
841 | ||
842 | /* newer commits first */ | |
843 | if (generation_a < generation_b) | |
844 | return 1; | |
845 | else if (generation_a > generation_b) | |
846 | return -1; | |
847 | ||
848 | /* use date as a heuristic when generations are equal */ | |
849 | if (a->date < b->date) | |
850 | return 1; | |
851 | else if (a->date > b->date) | |
852 | return -1; | |
853 | return 0; | |
854 | } | |
855 | ||
856 | int compare_commits_by_commit_date(const void *a_, const void *b_, void *unused) | |
857 | { | |
858 | const struct commit *a = a_, *b = b_; | |
859 | /* newer commits with larger date first */ | |
860 | if (a->date < b->date) | |
861 | return 1; | |
862 | else if (a->date > b->date) | |
863 | return -1; | |
864 | return 0; | |
865 | } | |
866 | ||
867 | /* | |
868 | * Performs an in-place topological sort on the list supplied. | |
869 | */ | |
870 | void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order) | |
871 | { | |
872 | struct commit_list *next, *orig = *list; | |
873 | struct commit_list **pptr; | |
874 | struct indegree_slab indegree; | |
875 | struct prio_queue queue; | |
876 | struct commit *commit; | |
877 | struct author_date_slab author_date; | |
878 | ||
879 | if (!orig) | |
880 | return; | |
881 | *list = NULL; | |
882 | ||
883 | init_indegree_slab(&indegree); | |
884 | memset(&queue, '\0', sizeof(queue)); | |
885 | ||
886 | switch (sort_order) { | |
887 | default: /* REV_SORT_IN_GRAPH_ORDER */ | |
888 | queue.compare = NULL; | |
889 | break; | |
890 | case REV_SORT_BY_COMMIT_DATE: | |
891 | queue.compare = compare_commits_by_commit_date; | |
892 | break; | |
893 | case REV_SORT_BY_AUTHOR_DATE: | |
894 | init_author_date_slab(&author_date); | |
895 | queue.compare = compare_commits_by_author_date; | |
896 | queue.cb_data = &author_date; | |
897 | break; | |
898 | } | |
899 | ||
900 | /* Mark them and clear the indegree */ | |
901 | for (next = orig; next; next = next->next) { | |
902 | struct commit *commit = next->item; | |
903 | *(indegree_slab_at(&indegree, commit)) = 1; | |
904 | /* also record the author dates, if needed */ | |
905 | if (sort_order == REV_SORT_BY_AUTHOR_DATE) | |
906 | record_author_date(&author_date, commit); | |
907 | } | |
908 | ||
909 | /* update the indegree */ | |
910 | for (next = orig; next; next = next->next) { | |
911 | struct commit_list *parents = next->item->parents; | |
912 | while (parents) { | |
913 | struct commit *parent = parents->item; | |
914 | int *pi = indegree_slab_at(&indegree, parent); | |
915 | ||
916 | if (*pi) | |
917 | (*pi)++; | |
918 | parents = parents->next; | |
919 | } | |
920 | } | |
921 | ||
922 | /* | |
923 | * find the tips | |
924 | * | |
925 | * tips are nodes not reachable from any other node in the list | |
926 | * | |
927 | * the tips serve as a starting set for the work queue. | |
928 | */ | |
929 | for (next = orig; next; next = next->next) { | |
930 | struct commit *commit = next->item; | |
931 | ||
932 | if (*(indegree_slab_at(&indegree, commit)) == 1) | |
933 | prio_queue_put(&queue, commit); | |
934 | } | |
935 | ||
936 | /* | |
937 | * This is unfortunate; the initial tips need to be shown | |
938 | * in the order given from the revision traversal machinery. | |
939 | */ | |
940 | if (sort_order == REV_SORT_IN_GRAPH_ORDER) | |
941 | prio_queue_reverse(&queue); | |
942 | ||
943 | /* We no longer need the commit list */ | |
944 | free_commit_list(orig); | |
945 | ||
946 | pptr = list; | |
947 | *list = NULL; | |
948 | while ((commit = prio_queue_get(&queue)) != NULL) { | |
949 | struct commit_list *parents; | |
950 | ||
951 | for (parents = commit->parents; parents ; parents = parents->next) { | |
952 | struct commit *parent = parents->item; | |
953 | int *pi = indegree_slab_at(&indegree, parent); | |
954 | ||
955 | if (!*pi) | |
956 | continue; | |
957 | ||
958 | /* | |
959 | * parents are only enqueued for emission | |
960 | * when all their children have been emitted thereby | |
961 | * guaranteeing topological order. | |
962 | */ | |
963 | if (--(*pi) == 1) | |
964 | prio_queue_put(&queue, parent); | |
965 | } | |
966 | /* | |
967 | * all children of commit have already been | |
968 | * emitted. we can emit it now. | |
969 | */ | |
970 | *(indegree_slab_at(&indegree, commit)) = 0; | |
971 | ||
972 | pptr = &commit_list_insert(commit, pptr)->next; | |
973 | } | |
974 | ||
975 | clear_indegree_slab(&indegree); | |
976 | clear_prio_queue(&queue); | |
977 | if (sort_order == REV_SORT_BY_AUTHOR_DATE) | |
978 | clear_author_date_slab(&author_date); | |
979 | } | |
980 | ||
981 | struct rev_collect { | |
982 | struct commit **commit; | |
983 | int nr; | |
984 | int alloc; | |
985 | unsigned int initial : 1; | |
986 | }; | |
987 | ||
988 | static void add_one_commit(struct object_id *oid, struct rev_collect *revs) | |
989 | { | |
990 | struct commit *commit; | |
991 | ||
992 | if (is_null_oid(oid)) | |
993 | return; | |
994 | ||
995 | commit = lookup_commit(the_repository, oid); | |
996 | if (!commit || | |
997 | (commit->object.flags & TMP_MARK) || | |
998 | parse_commit(commit)) | |
999 | return; | |
1000 | ||
1001 | ALLOC_GROW(revs->commit, revs->nr + 1, revs->alloc); | |
1002 | revs->commit[revs->nr++] = commit; | |
1003 | commit->object.flags |= TMP_MARK; | |
1004 | } | |
1005 | ||
1006 | static int collect_one_reflog_ent(struct object_id *ooid, struct object_id *noid, | |
1007 | const char *ident UNUSED, | |
1008 | timestamp_t timestamp UNUSED, int tz UNUSED, | |
1009 | const char *message UNUSED, void *cbdata) | |
1010 | { | |
1011 | struct rev_collect *revs = cbdata; | |
1012 | ||
1013 | if (revs->initial) { | |
1014 | revs->initial = 0; | |
1015 | add_one_commit(ooid, revs); | |
1016 | } | |
1017 | add_one_commit(noid, revs); | |
1018 | return 0; | |
1019 | } | |
1020 | ||
1021 | struct commit *get_fork_point(const char *refname, struct commit *commit) | |
1022 | { | |
1023 | struct object_id oid; | |
1024 | struct rev_collect revs; | |
1025 | struct commit_list *bases; | |
1026 | int i; | |
1027 | struct commit *ret = NULL; | |
1028 | char *full_refname; | |
1029 | ||
1030 | switch (dwim_ref(refname, strlen(refname), &oid, &full_refname, 0)) { | |
1031 | case 0: | |
1032 | die("No such ref: '%s'", refname); | |
1033 | case 1: | |
1034 | break; /* good */ | |
1035 | default: | |
1036 | die("Ambiguous refname: '%s'", refname); | |
1037 | } | |
1038 | ||
1039 | memset(&revs, 0, sizeof(revs)); | |
1040 | revs.initial = 1; | |
1041 | for_each_reflog_ent(full_refname, collect_one_reflog_ent, &revs); | |
1042 | ||
1043 | if (!revs.nr) | |
1044 | add_one_commit(&oid, &revs); | |
1045 | ||
1046 | for (i = 0; i < revs.nr; i++) | |
1047 | revs.commit[i]->object.flags &= ~TMP_MARK; | |
1048 | ||
1049 | bases = get_merge_bases_many(commit, revs.nr, revs.commit); | |
1050 | ||
1051 | /* | |
1052 | * There should be one and only one merge base, when we found | |
1053 | * a common ancestor among reflog entries. | |
1054 | */ | |
1055 | if (!bases || bases->next) | |
1056 | goto cleanup_return; | |
1057 | ||
1058 | /* And the found one must be one of the reflog entries */ | |
1059 | for (i = 0; i < revs.nr; i++) | |
1060 | if (&bases->item->object == &revs.commit[i]->object) | |
1061 | break; /* found */ | |
1062 | if (revs.nr <= i) | |
1063 | goto cleanup_return; | |
1064 | ||
1065 | ret = bases->item; | |
1066 | ||
1067 | cleanup_return: | |
1068 | free(revs.commit); | |
1069 | free_commit_list(bases); | |
1070 | free(full_refname); | |
1071 | return ret; | |
1072 | } | |
1073 | ||
1074 | /* | |
1075 | * Indexed by hash algorithm identifier. | |
1076 | */ | |
1077 | static const char *gpg_sig_headers[] = { | |
1078 | NULL, | |
1079 | "gpgsig", | |
1080 | "gpgsig-sha256", | |
1081 | }; | |
1082 | ||
1083 | int sign_with_header(struct strbuf *buf, const char *keyid) | |
1084 | { | |
1085 | struct strbuf sig = STRBUF_INIT; | |
1086 | int inspos, copypos; | |
1087 | const char *eoh; | |
1088 | const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(the_hash_algo)]; | |
1089 | int gpg_sig_header_len = strlen(gpg_sig_header); | |
1090 | ||
1091 | /* find the end of the header */ | |
1092 | eoh = strstr(buf->buf, "\n\n"); | |
1093 | if (!eoh) | |
1094 | inspos = buf->len; | |
1095 | else | |
1096 | inspos = eoh - buf->buf + 1; | |
1097 | ||
1098 | if (!keyid || !*keyid) | |
1099 | keyid = get_signing_key(); | |
1100 | if (sign_buffer(buf, &sig, keyid)) { | |
1101 | strbuf_release(&sig); | |
1102 | return -1; | |
1103 | } | |
1104 | ||
1105 | for (copypos = 0; sig.buf[copypos]; ) { | |
1106 | const char *bol = sig.buf + copypos; | |
1107 | const char *eol = strchrnul(bol, '\n'); | |
1108 | int len = (eol - bol) + !!*eol; | |
1109 | ||
1110 | if (!copypos) { | |
1111 | strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len); | |
1112 | inspos += gpg_sig_header_len; | |
1113 | } | |
1114 | strbuf_insertstr(buf, inspos++, " "); | |
1115 | strbuf_insert(buf, inspos, bol, len); | |
1116 | inspos += len; | |
1117 | copypos += len; | |
1118 | } | |
1119 | strbuf_release(&sig); | |
1120 | return 0; | |
1121 | } | |
1122 | ||
1123 | ||
1124 | ||
1125 | int parse_signed_commit(const struct commit *commit, | |
1126 | struct strbuf *payload, struct strbuf *signature, | |
1127 | const struct git_hash_algo *algop) | |
1128 | { | |
1129 | unsigned long size; | |
1130 | const char *buffer = get_commit_buffer(commit, &size); | |
1131 | int ret = parse_buffer_signed_by_header(buffer, size, payload, signature, algop); | |
1132 | ||
1133 | unuse_commit_buffer(commit, buffer); | |
1134 | return ret; | |
1135 | } | |
1136 | ||
1137 | int parse_buffer_signed_by_header(const char *buffer, | |
1138 | unsigned long size, | |
1139 | struct strbuf *payload, | |
1140 | struct strbuf *signature, | |
1141 | const struct git_hash_algo *algop) | |
1142 | { | |
1143 | int in_signature = 0, saw_signature = 0, other_signature = 0; | |
1144 | const char *line, *tail, *p; | |
1145 | const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(algop)]; | |
1146 | ||
1147 | line = buffer; | |
1148 | tail = buffer + size; | |
1149 | while (line < tail) { | |
1150 | const char *sig = NULL; | |
1151 | const char *next = memchr(line, '\n', tail - line); | |
1152 | ||
1153 | next = next ? next + 1 : tail; | |
1154 | if (in_signature && line[0] == ' ') | |
1155 | sig = line + 1; | |
1156 | else if (skip_prefix(line, gpg_sig_header, &p) && | |
1157 | *p == ' ') { | |
1158 | sig = line + strlen(gpg_sig_header) + 1; | |
1159 | other_signature = 0; | |
1160 | } | |
1161 | else if (starts_with(line, "gpgsig")) | |
1162 | other_signature = 1; | |
1163 | else if (other_signature && line[0] != ' ') | |
1164 | other_signature = 0; | |
1165 | if (sig) { | |
1166 | strbuf_add(signature, sig, next - sig); | |
1167 | saw_signature = 1; | |
1168 | in_signature = 1; | |
1169 | } else { | |
1170 | if (*line == '\n') | |
1171 | /* dump the whole remainder of the buffer */ | |
1172 | next = tail; | |
1173 | if (!other_signature) | |
1174 | strbuf_add(payload, line, next - line); | |
1175 | in_signature = 0; | |
1176 | } | |
1177 | line = next; | |
1178 | } | |
1179 | return saw_signature; | |
1180 | } | |
1181 | ||
1182 | int remove_signature(struct strbuf *buf) | |
1183 | { | |
1184 | const char *line = buf->buf; | |
1185 | const char *tail = buf->buf + buf->len; | |
1186 | int in_signature = 0; | |
1187 | struct sigbuf { | |
1188 | const char *start; | |
1189 | const char *end; | |
1190 | } sigs[2], *sigp = &sigs[0]; | |
1191 | int i; | |
1192 | const char *orig_buf = buf->buf; | |
1193 | ||
1194 | memset(sigs, 0, sizeof(sigs)); | |
1195 | ||
1196 | while (line < tail) { | |
1197 | const char *next = memchr(line, '\n', tail - line); | |
1198 | next = next ? next + 1 : tail; | |
1199 | ||
1200 | if (in_signature && line[0] == ' ') | |
1201 | sigp->end = next; | |
1202 | else if (starts_with(line, "gpgsig")) { | |
1203 | int i; | |
1204 | for (i = 1; i < GIT_HASH_NALGOS; i++) { | |
1205 | const char *p; | |
1206 | if (skip_prefix(line, gpg_sig_headers[i], &p) && | |
1207 | *p == ' ') { | |
1208 | sigp->start = line; | |
1209 | sigp->end = next; | |
1210 | in_signature = 1; | |
1211 | } | |
1212 | } | |
1213 | } else { | |
1214 | if (*line == '\n') | |
1215 | /* dump the whole remainder of the buffer */ | |
1216 | next = tail; | |
1217 | if (in_signature && sigp - sigs != ARRAY_SIZE(sigs)) | |
1218 | sigp++; | |
1219 | in_signature = 0; | |
1220 | } | |
1221 | line = next; | |
1222 | } | |
1223 | ||
1224 | for (i = ARRAY_SIZE(sigs) - 1; i >= 0; i--) | |
1225 | if (sigs[i].start) | |
1226 | strbuf_remove(buf, sigs[i].start - orig_buf, sigs[i].end - sigs[i].start); | |
1227 | ||
1228 | return sigs[0].start != NULL; | |
1229 | } | |
1230 | ||
1231 | static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail) | |
1232 | { | |
1233 | struct merge_remote_desc *desc; | |
1234 | struct commit_extra_header *mergetag; | |
1235 | char *buf; | |
1236 | unsigned long size; | |
1237 | enum object_type type; | |
1238 | struct strbuf payload = STRBUF_INIT; | |
1239 | struct strbuf signature = STRBUF_INIT; | |
1240 | ||
1241 | desc = merge_remote_util(parent); | |
1242 | if (!desc || !desc->obj) | |
1243 | return; | |
1244 | buf = read_object_file(&desc->obj->oid, &type, &size); | |
1245 | if (!buf || type != OBJ_TAG) | |
1246 | goto free_return; | |
1247 | if (!parse_signature(buf, size, &payload, &signature)) | |
1248 | goto free_return; | |
1249 | /* | |
1250 | * We could verify this signature and either omit the tag when | |
1251 | * it does not validate, but the integrator may not have the | |
1252 | * public key of the signer of the tag being merged, while a | |
1253 | * later auditor may have it while auditing, so let's not run | |
1254 | * verify-signed-buffer here for now... | |
1255 | * | |
1256 | * if (verify_signed_buffer(buf, len, buf + len, size - len, ...)) | |
1257 | * warn("warning: signed tag unverified."); | |
1258 | */ | |
1259 | CALLOC_ARRAY(mergetag, 1); | |
1260 | mergetag->key = xstrdup("mergetag"); | |
1261 | mergetag->value = buf; | |
1262 | mergetag->len = size; | |
1263 | ||
1264 | **tail = mergetag; | |
1265 | *tail = &mergetag->next; | |
1266 | strbuf_release(&payload); | |
1267 | strbuf_release(&signature); | |
1268 | return; | |
1269 | ||
1270 | free_return: | |
1271 | free(buf); | |
1272 | } | |
1273 | ||
1274 | int check_commit_signature(const struct commit *commit, struct signature_check *sigc) | |
1275 | { | |
1276 | struct strbuf payload = STRBUF_INIT; | |
1277 | struct strbuf signature = STRBUF_INIT; | |
1278 | int ret = 1; | |
1279 | ||
1280 | sigc->result = 'N'; | |
1281 | ||
1282 | if (parse_signed_commit(commit, &payload, &signature, the_hash_algo) <= 0) | |
1283 | goto out; | |
1284 | ||
1285 | sigc->payload_type = SIGNATURE_PAYLOAD_COMMIT; | |
1286 | sigc->payload = strbuf_detach(&payload, &sigc->payload_len); | |
1287 | ret = check_signature(sigc, signature.buf, signature.len); | |
1288 | ||
1289 | out: | |
1290 | strbuf_release(&payload); | |
1291 | strbuf_release(&signature); | |
1292 | ||
1293 | return ret; | |
1294 | } | |
1295 | ||
1296 | void verify_merge_signature(struct commit *commit, int verbosity, | |
1297 | int check_trust) | |
1298 | { | |
1299 | char hex[GIT_MAX_HEXSZ + 1]; | |
1300 | struct signature_check signature_check; | |
1301 | int ret; | |
1302 | memset(&signature_check, 0, sizeof(signature_check)); | |
1303 | ||
1304 | ret = check_commit_signature(commit, &signature_check); | |
1305 | ||
1306 | find_unique_abbrev_r(hex, &commit->object.oid, DEFAULT_ABBREV); | |
1307 | switch (signature_check.result) { | |
1308 | case 'G': | |
1309 | if (ret || (check_trust && signature_check.trust_level < TRUST_MARGINAL)) | |
1310 | die(_("Commit %s has an untrusted GPG signature, " | |
1311 | "allegedly by %s."), hex, signature_check.signer); | |
1312 | break; | |
1313 | case 'B': | |
1314 | die(_("Commit %s has a bad GPG signature " | |
1315 | "allegedly by %s."), hex, signature_check.signer); | |
1316 | default: /* 'N' */ | |
1317 | die(_("Commit %s does not have a GPG signature."), hex); | |
1318 | } | |
1319 | if (verbosity >= 0 && signature_check.result == 'G') | |
1320 | printf(_("Commit %s has a good GPG signature by %s\n"), | |
1321 | hex, signature_check.signer); | |
1322 | ||
1323 | signature_check_clear(&signature_check); | |
1324 | } | |
1325 | ||
1326 | void append_merge_tag_headers(struct commit_list *parents, | |
1327 | struct commit_extra_header ***tail) | |
1328 | { | |
1329 | while (parents) { | |
1330 | struct commit *parent = parents->item; | |
1331 | handle_signed_tag(parent, tail); | |
1332 | parents = parents->next; | |
1333 | } | |
1334 | } | |
1335 | ||
1336 | static void add_extra_header(struct strbuf *buffer, | |
1337 | struct commit_extra_header *extra) | |
1338 | { | |
1339 | strbuf_addstr(buffer, extra->key); | |
1340 | if (extra->len) | |
1341 | strbuf_add_lines(buffer, " ", extra->value, extra->len); | |
1342 | else | |
1343 | strbuf_addch(buffer, '\n'); | |
1344 | } | |
1345 | ||
1346 | struct commit_extra_header *read_commit_extra_headers(struct commit *commit, | |
1347 | const char **exclude) | |
1348 | { | |
1349 | struct commit_extra_header *extra = NULL; | |
1350 | unsigned long size; | |
1351 | const char *buffer = get_commit_buffer(commit, &size); | |
1352 | extra = read_commit_extra_header_lines(buffer, size, exclude); | |
1353 | unuse_commit_buffer(commit, buffer); | |
1354 | return extra; | |
1355 | } | |
1356 | ||
1357 | int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data) | |
1358 | { | |
1359 | struct commit_extra_header *extra, *to_free; | |
1360 | int res = 0; | |
1361 | ||
1362 | to_free = read_commit_extra_headers(commit, NULL); | |
1363 | for (extra = to_free; !res && extra; extra = extra->next) { | |
1364 | if (strcmp(extra->key, "mergetag")) | |
1365 | continue; /* not a merge tag */ | |
1366 | res = fn(commit, extra, data); | |
1367 | } | |
1368 | free_commit_extra_headers(to_free); | |
1369 | return res; | |
1370 | } | |
1371 | ||
1372 | static inline int standard_header_field(const char *field, size_t len) | |
1373 | { | |
1374 | return ((len == 4 && !memcmp(field, "tree", 4)) || | |
1375 | (len == 6 && !memcmp(field, "parent", 6)) || | |
1376 | (len == 6 && !memcmp(field, "author", 6)) || | |
1377 | (len == 9 && !memcmp(field, "committer", 9)) || | |
1378 | (len == 8 && !memcmp(field, "encoding", 8))); | |
1379 | } | |
1380 | ||
1381 | static int excluded_header_field(const char *field, size_t len, const char **exclude) | |
1382 | { | |
1383 | if (!exclude) | |
1384 | return 0; | |
1385 | ||
1386 | while (*exclude) { | |
1387 | size_t xlen = strlen(*exclude); | |
1388 | if (len == xlen && !memcmp(field, *exclude, xlen)) | |
1389 | return 1; | |
1390 | exclude++; | |
1391 | } | |
1392 | return 0; | |
1393 | } | |
1394 | ||
1395 | static struct commit_extra_header *read_commit_extra_header_lines( | |
1396 | const char *buffer, size_t size, | |
1397 | const char **exclude) | |
1398 | { | |
1399 | struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL; | |
1400 | const char *line, *next, *eof, *eob; | |
1401 | struct strbuf buf = STRBUF_INIT; | |
1402 | ||
1403 | for (line = buffer, eob = line + size; | |
1404 | line < eob && *line != '\n'; | |
1405 | line = next) { | |
1406 | next = memchr(line, '\n', eob - line); | |
1407 | next = next ? next + 1 : eob; | |
1408 | if (*line == ' ') { | |
1409 | /* continuation */ | |
1410 | if (it) | |
1411 | strbuf_add(&buf, line + 1, next - (line + 1)); | |
1412 | continue; | |
1413 | } | |
1414 | if (it) | |
1415 | it->value = strbuf_detach(&buf, &it->len); | |
1416 | strbuf_reset(&buf); | |
1417 | it = NULL; | |
1418 | ||
1419 | eof = memchr(line, ' ', next - line); | |
1420 | if (!eof) | |
1421 | eof = next; | |
1422 | else if (standard_header_field(line, eof - line) || | |
1423 | excluded_header_field(line, eof - line, exclude)) | |
1424 | continue; | |
1425 | ||
1426 | CALLOC_ARRAY(it, 1); | |
1427 | it->key = xmemdupz(line, eof-line); | |
1428 | *tail = it; | |
1429 | tail = &it->next; | |
1430 | if (eof + 1 < next) | |
1431 | strbuf_add(&buf, eof + 1, next - (eof + 1)); | |
1432 | } | |
1433 | if (it) | |
1434 | it->value = strbuf_detach(&buf, &it->len); | |
1435 | return extra; | |
1436 | } | |
1437 | ||
1438 | void free_commit_extra_headers(struct commit_extra_header *extra) | |
1439 | { | |
1440 | while (extra) { | |
1441 | struct commit_extra_header *next = extra->next; | |
1442 | free(extra->key); | |
1443 | free(extra->value); | |
1444 | free(extra); | |
1445 | extra = next; | |
1446 | } | |
1447 | } | |
1448 | ||
1449 | int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree, | |
1450 | struct commit_list *parents, struct object_id *ret, | |
1451 | const char *author, const char *sign_commit) | |
1452 | { | |
1453 | struct commit_extra_header *extra = NULL, **tail = &extra; | |
1454 | int result; | |
1455 | ||
1456 | append_merge_tag_headers(parents, &tail); | |
1457 | result = commit_tree_extended(msg, msg_len, tree, parents, ret, author, | |
1458 | NULL, sign_commit, extra); | |
1459 | free_commit_extra_headers(extra); | |
1460 | return result; | |
1461 | } | |
1462 | ||
1463 | static int find_invalid_utf8(const char *buf, int len) | |
1464 | { | |
1465 | int offset = 0; | |
1466 | static const unsigned int max_codepoint[] = { | |
1467 | 0x7f, 0x7ff, 0xffff, 0x10ffff | |
1468 | }; | |
1469 | ||
1470 | while (len) { | |
1471 | unsigned char c = *buf++; | |
1472 | int bytes, bad_offset; | |
1473 | unsigned int codepoint; | |
1474 | unsigned int min_val, max_val; | |
1475 | ||
1476 | len--; | |
1477 | offset++; | |
1478 | ||
1479 | /* Simple US-ASCII? No worries. */ | |
1480 | if (c < 0x80) | |
1481 | continue; | |
1482 | ||
1483 | bad_offset = offset-1; | |
1484 | ||
1485 | /* | |
1486 | * Count how many more high bits set: that's how | |
1487 | * many more bytes this sequence should have. | |
1488 | */ | |
1489 | bytes = 0; | |
1490 | while (c & 0x40) { | |
1491 | c <<= 1; | |
1492 | bytes++; | |
1493 | } | |
1494 | ||
1495 | /* | |
1496 | * Must be between 1 and 3 more bytes. Longer sequences result in | |
1497 | * codepoints beyond U+10FFFF, which are guaranteed never to exist. | |
1498 | */ | |
1499 | if (bytes < 1 || 3 < bytes) | |
1500 | return bad_offset; | |
1501 | ||
1502 | /* Do we *have* that many bytes? */ | |
1503 | if (len < bytes) | |
1504 | return bad_offset; | |
1505 | ||
1506 | /* | |
1507 | * Place the encoded bits at the bottom of the value and compute the | |
1508 | * valid range. | |
1509 | */ | |
1510 | codepoint = (c & 0x7f) >> bytes; | |
1511 | min_val = max_codepoint[bytes-1] + 1; | |
1512 | max_val = max_codepoint[bytes]; | |
1513 | ||
1514 | offset += bytes; | |
1515 | len -= bytes; | |
1516 | ||
1517 | /* And verify that they are good continuation bytes */ | |
1518 | do { | |
1519 | codepoint <<= 6; | |
1520 | codepoint |= *buf & 0x3f; | |
1521 | if ((*buf++ & 0xc0) != 0x80) | |
1522 | return bad_offset; | |
1523 | } while (--bytes); | |
1524 | ||
1525 | /* Reject codepoints that are out of range for the sequence length. */ | |
1526 | if (codepoint < min_val || codepoint > max_val) | |
1527 | return bad_offset; | |
1528 | /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */ | |
1529 | if ((codepoint & 0x1ff800) == 0xd800) | |
1530 | return bad_offset; | |
1531 | /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */ | |
1532 | if ((codepoint & 0xfffe) == 0xfffe) | |
1533 | return bad_offset; | |
1534 | /* So are anything in the range U+FDD0..U+FDEF. */ | |
1535 | if (codepoint >= 0xfdd0 && codepoint <= 0xfdef) | |
1536 | return bad_offset; | |
1537 | } | |
1538 | return -1; | |
1539 | } | |
1540 | ||
1541 | /* | |
1542 | * This verifies that the buffer is in proper utf8 format. | |
1543 | * | |
1544 | * If it isn't, it assumes any non-utf8 characters are Latin1, | |
1545 | * and does the conversion. | |
1546 | */ | |
1547 | static int verify_utf8(struct strbuf *buf) | |
1548 | { | |
1549 | int ok = 1; | |
1550 | long pos = 0; | |
1551 | ||
1552 | for (;;) { | |
1553 | int bad; | |
1554 | unsigned char c; | |
1555 | unsigned char replace[2]; | |
1556 | ||
1557 | bad = find_invalid_utf8(buf->buf + pos, buf->len - pos); | |
1558 | if (bad < 0) | |
1559 | return ok; | |
1560 | pos += bad; | |
1561 | ok = 0; | |
1562 | c = buf->buf[pos]; | |
1563 | strbuf_remove(buf, pos, 1); | |
1564 | ||
1565 | /* We know 'c' must be in the range 128-255 */ | |
1566 | replace[0] = 0xc0 + (c >> 6); | |
1567 | replace[1] = 0x80 + (c & 0x3f); | |
1568 | strbuf_insert(buf, pos, replace, 2); | |
1569 | pos += 2; | |
1570 | } | |
1571 | } | |
1572 | ||
1573 | static const char commit_utf8_warn[] = | |
1574 | N_("Warning: commit message did not conform to UTF-8.\n" | |
1575 | "You may want to amend it after fixing the message, or set the config\n" | |
1576 | "variable i18n.commitEncoding to the encoding your project uses.\n"); | |
1577 | ||
1578 | int commit_tree_extended(const char *msg, size_t msg_len, | |
1579 | const struct object_id *tree, | |
1580 | struct commit_list *parents, struct object_id *ret, | |
1581 | const char *author, const char *committer, | |
1582 | const char *sign_commit, | |
1583 | struct commit_extra_header *extra) | |
1584 | { | |
1585 | int result; | |
1586 | int encoding_is_utf8; | |
1587 | struct strbuf buffer; | |
1588 | ||
1589 | assert_oid_type(tree, OBJ_TREE); | |
1590 | ||
1591 | if (memchr(msg, '\0', msg_len)) | |
1592 | return error("a NUL byte in commit log message not allowed."); | |
1593 | ||
1594 | /* Not having i18n.commitencoding is the same as having utf-8 */ | |
1595 | encoding_is_utf8 = is_encoding_utf8(git_commit_encoding); | |
1596 | ||
1597 | strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */ | |
1598 | strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree)); | |
1599 | ||
1600 | /* | |
1601 | * NOTE! This ordering means that the same exact tree merged with a | |
1602 | * different order of parents will be a _different_ changeset even | |
1603 | * if everything else stays the same. | |
1604 | */ | |
1605 | while (parents) { | |
1606 | struct commit *parent = pop_commit(&parents); | |
1607 | strbuf_addf(&buffer, "parent %s\n", | |
1608 | oid_to_hex(&parent->object.oid)); | |
1609 | } | |
1610 | ||
1611 | /* Person/date information */ | |
1612 | if (!author) | |
1613 | author = git_author_info(IDENT_STRICT); | |
1614 | strbuf_addf(&buffer, "author %s\n", author); | |
1615 | if (!committer) | |
1616 | committer = git_committer_info(IDENT_STRICT); | |
1617 | strbuf_addf(&buffer, "committer %s\n", committer); | |
1618 | if (!encoding_is_utf8) | |
1619 | strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding); | |
1620 | ||
1621 | while (extra) { | |
1622 | add_extra_header(&buffer, extra); | |
1623 | extra = extra->next; | |
1624 | } | |
1625 | strbuf_addch(&buffer, '\n'); | |
1626 | ||
1627 | /* And add the comment */ | |
1628 | strbuf_add(&buffer, msg, msg_len); | |
1629 | ||
1630 | /* And check the encoding */ | |
1631 | if (encoding_is_utf8 && !verify_utf8(&buffer)) | |
1632 | fprintf(stderr, _(commit_utf8_warn)); | |
1633 | ||
1634 | if (sign_commit && sign_with_header(&buffer, sign_commit)) { | |
1635 | result = -1; | |
1636 | goto out; | |
1637 | } | |
1638 | ||
1639 | result = write_object_file(buffer.buf, buffer.len, OBJ_COMMIT, ret); | |
1640 | out: | |
1641 | strbuf_release(&buffer); | |
1642 | return result; | |
1643 | } | |
1644 | ||
1645 | define_commit_slab(merge_desc_slab, struct merge_remote_desc *); | |
1646 | static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab); | |
1647 | ||
1648 | struct merge_remote_desc *merge_remote_util(struct commit *commit) | |
1649 | { | |
1650 | return *merge_desc_slab_at(&merge_desc_slab, commit); | |
1651 | } | |
1652 | ||
1653 | void set_merge_remote_desc(struct commit *commit, | |
1654 | const char *name, struct object *obj) | |
1655 | { | |
1656 | struct merge_remote_desc *desc; | |
1657 | FLEX_ALLOC_STR(desc, name, name); | |
1658 | desc->obj = obj; | |
1659 | *merge_desc_slab_at(&merge_desc_slab, commit) = desc; | |
1660 | } | |
1661 | ||
1662 | struct commit *get_merge_parent(const char *name) | |
1663 | { | |
1664 | struct object *obj; | |
1665 | struct commit *commit; | |
1666 | struct object_id oid; | |
1667 | if (get_oid(name, &oid)) | |
1668 | return NULL; | |
1669 | obj = parse_object(the_repository, &oid); | |
1670 | commit = (struct commit *)peel_to_type(name, 0, obj, OBJ_COMMIT); | |
1671 | if (commit && !merge_remote_util(commit)) | |
1672 | set_merge_remote_desc(commit, name, obj); | |
1673 | return commit; | |
1674 | } | |
1675 | ||
1676 | /* | |
1677 | * Append a commit to the end of the commit_list. | |
1678 | * | |
1679 | * next starts by pointing to the variable that holds the head of an | |
1680 | * empty commit_list, and is updated to point to the "next" field of | |
1681 | * the last item on the list as new commits are appended. | |
1682 | * | |
1683 | * Usage example: | |
1684 | * | |
1685 | * struct commit_list *list; | |
1686 | * struct commit_list **next = &list; | |
1687 | * | |
1688 | * next = commit_list_append(c1, next); | |
1689 | * next = commit_list_append(c2, next); | |
1690 | * assert(commit_list_count(list) == 2); | |
1691 | * return list; | |
1692 | */ | |
1693 | struct commit_list **commit_list_append(struct commit *commit, | |
1694 | struct commit_list **next) | |
1695 | { | |
1696 | struct commit_list *new_commit = xmalloc(sizeof(struct commit_list)); | |
1697 | new_commit->item = commit; | |
1698 | *next = new_commit; | |
1699 | new_commit->next = NULL; | |
1700 | return &new_commit->next; | |
1701 | } | |
1702 | ||
1703 | const char *find_header_mem(const char *msg, size_t len, | |
1704 | const char *key, size_t *out_len) | |
1705 | { | |
1706 | int key_len = strlen(key); | |
1707 | const char *line = msg; | |
1708 | ||
1709 | /* | |
1710 | * NEEDSWORK: It's possible for strchrnul() to scan beyond the range | |
1711 | * given by len. However, current callers are safe because they compute | |
1712 | * len by scanning a NUL-terminated block of memory starting at msg. | |
1713 | * Nonetheless, it would be better to ensure the function does not look | |
1714 | * at msg beyond the len provided by the caller. | |
1715 | */ | |
1716 | while (line && line < msg + len) { | |
1717 | const char *eol = strchrnul(line, '\n'); | |
1718 | ||
1719 | if (line == eol) | |
1720 | return NULL; | |
1721 | ||
1722 | if (eol - line > key_len && | |
1723 | !strncmp(line, key, key_len) && | |
1724 | line[key_len] == ' ') { | |
1725 | *out_len = eol - line - key_len - 1; | |
1726 | return line + key_len + 1; | |
1727 | } | |
1728 | line = *eol ? eol + 1 : NULL; | |
1729 | } | |
1730 | return NULL; | |
1731 | } | |
1732 | ||
1733 | const char *find_commit_header(const char *msg, const char *key, size_t *out_len) | |
1734 | { | |
1735 | return find_header_mem(msg, strlen(msg), key, out_len); | |
1736 | } | |
1737 | /* | |
1738 | * Inspect the given string and determine the true "end" of the log message, in | |
1739 | * order to find where to put a new Signed-off-by trailer. Ignored are | |
1740 | * trailing comment lines and blank lines. To support "git commit -s | |
1741 | * --amend" on an existing commit, we also ignore "Conflicts:". To | |
1742 | * support "git commit -v", we truncate at cut lines. | |
1743 | * | |
1744 | * Returns the number of bytes from the tail to ignore, to be fed as | |
1745 | * the second parameter to append_signoff(). | |
1746 | */ | |
1747 | size_t ignore_non_trailer(const char *buf, size_t len) | |
1748 | { | |
1749 | size_t boc = 0; | |
1750 | size_t bol = 0; | |
1751 | int in_old_conflicts_block = 0; | |
1752 | size_t cutoff = wt_status_locate_end(buf, len); | |
1753 | ||
1754 | while (bol < cutoff) { | |
1755 | const char *next_line = memchr(buf + bol, '\n', len - bol); | |
1756 | ||
1757 | if (!next_line) | |
1758 | next_line = buf + len; | |
1759 | else | |
1760 | next_line++; | |
1761 | ||
1762 | if (buf[bol] == comment_line_char || buf[bol] == '\n') { | |
1763 | /* is this the first of the run of comments? */ | |
1764 | if (!boc) | |
1765 | boc = bol; | |
1766 | /* otherwise, it is just continuing */ | |
1767 | } else if (starts_with(buf + bol, "Conflicts:\n")) { | |
1768 | in_old_conflicts_block = 1; | |
1769 | if (!boc) | |
1770 | boc = bol; | |
1771 | } else if (in_old_conflicts_block && buf[bol] == '\t') { | |
1772 | ; /* a pathname in the conflicts block */ | |
1773 | } else if (boc) { | |
1774 | /* the previous was not trailing comment */ | |
1775 | boc = 0; | |
1776 | in_old_conflicts_block = 0; | |
1777 | } | |
1778 | bol = next_line - buf; | |
1779 | } | |
1780 | return boc ? len - boc : len - cutoff; | |
1781 | } | |
1782 | ||
1783 | int run_commit_hook(int editor_is_used, const char *index_file, | |
1784 | int *invoked_hook, const char *name, ...) | |
1785 | { | |
1786 | struct run_hooks_opt opt = RUN_HOOKS_OPT_INIT; | |
1787 | va_list args; | |
1788 | const char *arg; | |
1789 | ||
1790 | strvec_pushf(&opt.env, "GIT_INDEX_FILE=%s", index_file); | |
1791 | ||
1792 | /* | |
1793 | * Let the hook know that no editor will be launched. | |
1794 | */ | |
1795 | if (!editor_is_used) | |
1796 | strvec_push(&opt.env, "GIT_EDITOR=:"); | |
1797 | ||
1798 | va_start(args, name); | |
1799 | while ((arg = va_arg(args, const char *))) | |
1800 | strvec_push(&opt.args, arg); | |
1801 | va_end(args); | |
1802 | ||
1803 | opt.invoked_hook = invoked_hook; | |
1804 | return run_hooks_opt(name, &opt); | |
1805 | } |