| 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 | size_t 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 = cast_size_t_to_ulong(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 | size_t 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 | size_t 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_get(queue); |
| 786 | struct commit_list *parents = ret->parents; |
| 787 | |
| 788 | while (parents) { |
| 789 | struct commit *commit = parents->item; |
| 790 | if (!repo_parse_commit(the_repository, commit) && !(commit->object.flags & mark)) { |
| 791 | commit->object.flags |= mark; |
| 792 | prio_queue_put(queue, commit); |
| 793 | } |
| 794 | parents = parents->next; |
| 795 | } |
| 796 | return ret; |
| 797 | } |
| 798 | |
| 799 | static void clear_commit_marks_1(struct commit_list **plist, |
| 800 | struct commit *commit, unsigned int mark) |
| 801 | { |
| 802 | while (commit) { |
| 803 | struct commit_list *parents; |
| 804 | |
| 805 | if (!(mark & commit->object.flags)) |
| 806 | return; |
| 807 | |
| 808 | commit->object.flags &= ~mark; |
| 809 | |
| 810 | parents = commit->parents; |
| 811 | if (!parents) |
| 812 | return; |
| 813 | |
| 814 | while ((parents = parents->next)) { |
| 815 | if (parents->item->object.flags & mark) |
| 816 | commit_list_insert(parents->item, plist); |
| 817 | } |
| 818 | |
| 819 | commit = commit->parents->item; |
| 820 | } |
| 821 | } |
| 822 | |
| 823 | void clear_commit_marks_many(size_t nr, struct commit **commit, unsigned int mark) |
| 824 | { |
| 825 | for (size_t i = 0; i < nr; i++) |
| 826 | clear_commit_marks(commit[i], mark); |
| 827 | } |
| 828 | |
| 829 | void clear_commit_marks(struct commit *commit, unsigned int mark) |
| 830 | { |
| 831 | struct commit_list *list = NULL; |
| 832 | |
| 833 | clear_commit_marks_1(&list, commit, mark); |
| 834 | while (list) |
| 835 | clear_commit_marks_1(&list, pop_commit(&list), mark); |
| 836 | } |
| 837 | |
| 838 | struct commit *pop_commit(struct commit_list **stack) |
| 839 | { |
| 840 | struct commit_list *top = *stack; |
| 841 | struct commit *item = top ? top->item : NULL; |
| 842 | |
| 843 | if (top) { |
| 844 | *stack = top->next; |
| 845 | free(top); |
| 846 | } |
| 847 | return item; |
| 848 | } |
| 849 | |
| 850 | /* |
| 851 | * Topological sort support |
| 852 | */ |
| 853 | |
| 854 | /* count number of children that have not been emitted */ |
| 855 | define_commit_slab(indegree_slab, int); |
| 856 | |
| 857 | define_commit_slab(author_date_slab, timestamp_t); |
| 858 | |
| 859 | void record_author_date(struct author_date_slab *author_date, |
| 860 | struct commit *commit) |
| 861 | { |
| 862 | const char *buffer = repo_get_commit_buffer(the_repository, commit, |
| 863 | NULL); |
| 864 | struct ident_split ident; |
| 865 | const char *ident_line; |
| 866 | size_t ident_len; |
| 867 | char *date_end; |
| 868 | timestamp_t date; |
| 869 | |
| 870 | ident_line = find_commit_header(buffer, "author", &ident_len); |
| 871 | if (!ident_line) |
| 872 | goto fail_exit; /* no author line */ |
| 873 | if (split_ident_line(&ident, ident_line, ident_len) || |
| 874 | !ident.date_begin || !ident.date_end) |
| 875 | goto fail_exit; /* malformed "author" line */ |
| 876 | |
| 877 | date = parse_timestamp(ident.date_begin, &date_end, 10); |
| 878 | if (date_end != ident.date_end) |
| 879 | goto fail_exit; /* malformed date */ |
| 880 | *(author_date_slab_at(author_date, commit)) = date; |
| 881 | |
| 882 | fail_exit: |
| 883 | repo_unuse_commit_buffer(the_repository, commit, buffer); |
| 884 | } |
| 885 | |
| 886 | int compare_commits_by_author_date(const void *a_, const void *b_, |
| 887 | void *cb_data) |
| 888 | { |
| 889 | const struct commit *a = a_, *b = b_; |
| 890 | struct author_date_slab *author_date = cb_data; |
| 891 | timestamp_t a_date = *(author_date_slab_at(author_date, a)); |
| 892 | timestamp_t b_date = *(author_date_slab_at(author_date, b)); |
| 893 | |
| 894 | /* newer commits with larger date first */ |
| 895 | if (a_date < b_date) |
| 896 | return 1; |
| 897 | else if (a_date > b_date) |
| 898 | return -1; |
| 899 | return 0; |
| 900 | } |
| 901 | |
| 902 | int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, |
| 903 | void *unused UNUSED) |
| 904 | { |
| 905 | const struct commit *a = a_, *b = b_; |
| 906 | const timestamp_t generation_a = commit_graph_generation(a), |
| 907 | generation_b = commit_graph_generation(b); |
| 908 | |
| 909 | /* newer commits first */ |
| 910 | if (generation_a < generation_b) |
| 911 | return 1; |
| 912 | else if (generation_a > generation_b) |
| 913 | return -1; |
| 914 | |
| 915 | /* use date as a heuristic when generations are equal */ |
| 916 | if (a->date < b->date) |
| 917 | return 1; |
| 918 | else if (a->date > b->date) |
| 919 | return -1; |
| 920 | return 0; |
| 921 | } |
| 922 | |
| 923 | int compare_commits_by_commit_date(const void *a_, const void *b_, |
| 924 | void *unused UNUSED) |
| 925 | { |
| 926 | const struct commit *a = a_, *b = b_; |
| 927 | /* newer commits with larger date first */ |
| 928 | if (a->date < b->date) |
| 929 | return 1; |
| 930 | else if (a->date > b->date) |
| 931 | return -1; |
| 932 | return 0; |
| 933 | } |
| 934 | |
| 935 | /* |
| 936 | * Performs an in-place topological sort on the list supplied. |
| 937 | */ |
| 938 | void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order) |
| 939 | { |
| 940 | struct commit_list *next, *orig = *list; |
| 941 | struct commit_list **pptr; |
| 942 | struct indegree_slab indegree; |
| 943 | struct prio_queue queue; |
| 944 | struct commit *commit; |
| 945 | struct author_date_slab author_date; |
| 946 | |
| 947 | if (!orig) |
| 948 | return; |
| 949 | *list = NULL; |
| 950 | |
| 951 | init_indegree_slab(&indegree); |
| 952 | memset(&queue, '\0', sizeof(queue)); |
| 953 | |
| 954 | switch (sort_order) { |
| 955 | default: /* REV_SORT_IN_GRAPH_ORDER */ |
| 956 | queue.compare = NULL; |
| 957 | break; |
| 958 | case REV_SORT_BY_COMMIT_DATE: |
| 959 | queue.compare = compare_commits_by_commit_date; |
| 960 | break; |
| 961 | case REV_SORT_BY_AUTHOR_DATE: |
| 962 | init_author_date_slab(&author_date); |
| 963 | queue.compare = compare_commits_by_author_date; |
| 964 | queue.cb_data = &author_date; |
| 965 | break; |
| 966 | } |
| 967 | |
| 968 | /* Mark them and clear the indegree */ |
| 969 | for (next = orig; next; next = next->next) { |
| 970 | struct commit *commit = next->item; |
| 971 | *(indegree_slab_at(&indegree, commit)) = 1; |
| 972 | /* also record the author dates, if needed */ |
| 973 | if (sort_order == REV_SORT_BY_AUTHOR_DATE) |
| 974 | record_author_date(&author_date, commit); |
| 975 | } |
| 976 | |
| 977 | /* update the indegree */ |
| 978 | for (next = orig; next; next = next->next) { |
| 979 | struct commit_list *parents = next->item->parents; |
| 980 | while (parents) { |
| 981 | struct commit *parent = parents->item; |
| 982 | int *pi = indegree_slab_at(&indegree, parent); |
| 983 | |
| 984 | if (*pi) |
| 985 | (*pi)++; |
| 986 | parents = parents->next; |
| 987 | } |
| 988 | } |
| 989 | |
| 990 | /* |
| 991 | * find the tips |
| 992 | * |
| 993 | * tips are nodes not reachable from any other node in the list |
| 994 | * |
| 995 | * the tips serve as a starting set for the work queue. |
| 996 | */ |
| 997 | for (next = orig; next; next = next->next) { |
| 998 | struct commit *commit = next->item; |
| 999 | |
| 1000 | if (*(indegree_slab_at(&indegree, commit)) == 1) |
| 1001 | prio_queue_put(&queue, commit); |
| 1002 | } |
| 1003 | |
| 1004 | /* |
| 1005 | * This is unfortunate; the initial tips need to be shown |
| 1006 | * in the order given from the revision traversal machinery. |
| 1007 | */ |
| 1008 | if (sort_order == REV_SORT_IN_GRAPH_ORDER) |
| 1009 | prio_queue_reverse(&queue); |
| 1010 | |
| 1011 | /* We no longer need the commit list */ |
| 1012 | commit_list_free(orig); |
| 1013 | |
| 1014 | pptr = list; |
| 1015 | *list = NULL; |
| 1016 | while ((commit = prio_queue_get(&queue)) != NULL) { |
| 1017 | struct commit_list *parents; |
| 1018 | |
| 1019 | for (parents = commit->parents; parents ; parents = parents->next) { |
| 1020 | struct commit *parent = parents->item; |
| 1021 | int *pi = indegree_slab_at(&indegree, parent); |
| 1022 | |
| 1023 | if (!*pi) |
| 1024 | continue; |
| 1025 | |
| 1026 | /* |
| 1027 | * parents are only enqueued for emission |
| 1028 | * when all their children have been emitted thereby |
| 1029 | * guaranteeing topological order. |
| 1030 | */ |
| 1031 | if (--(*pi) == 1) |
| 1032 | prio_queue_put(&queue, parent); |
| 1033 | } |
| 1034 | /* |
| 1035 | * all children of commit have already been |
| 1036 | * emitted. we can emit it now. |
| 1037 | */ |
| 1038 | *(indegree_slab_at(&indegree, commit)) = 0; |
| 1039 | |
| 1040 | pptr = &commit_list_insert(commit, pptr)->next; |
| 1041 | } |
| 1042 | |
| 1043 | clear_indegree_slab(&indegree); |
| 1044 | clear_prio_queue(&queue); |
| 1045 | if (sort_order == REV_SORT_BY_AUTHOR_DATE) |
| 1046 | clear_author_date_slab(&author_date); |
| 1047 | } |
| 1048 | |
| 1049 | struct rev_collect { |
| 1050 | struct commit_stack stack; |
| 1051 | unsigned int initial : 1; |
| 1052 | }; |
| 1053 | |
| 1054 | static void add_one_commit(struct object_id *oid, struct rev_collect *revs) |
| 1055 | { |
| 1056 | struct commit *commit; |
| 1057 | |
| 1058 | if (is_null_oid(oid)) |
| 1059 | return; |
| 1060 | |
| 1061 | commit = lookup_commit(the_repository, oid); |
| 1062 | if (!commit || |
| 1063 | (commit->object.flags & TMP_MARK) || |
| 1064 | repo_parse_commit(the_repository, commit)) |
| 1065 | return; |
| 1066 | |
| 1067 | commit_stack_push(&revs->stack, commit); |
| 1068 | commit->object.flags |= TMP_MARK; |
| 1069 | } |
| 1070 | |
| 1071 | static int collect_one_reflog_ent(const char *refname UNUSED, |
| 1072 | struct object_id *ooid, struct object_id *noid, |
| 1073 | const char *ident UNUSED, |
| 1074 | timestamp_t timestamp UNUSED, int tz UNUSED, |
| 1075 | const char *message UNUSED, void *cbdata) |
| 1076 | { |
| 1077 | struct rev_collect *revs = cbdata; |
| 1078 | |
| 1079 | if (revs->initial) { |
| 1080 | revs->initial = 0; |
| 1081 | add_one_commit(ooid, revs); |
| 1082 | } |
| 1083 | add_one_commit(noid, revs); |
| 1084 | return 0; |
| 1085 | } |
| 1086 | |
| 1087 | struct commit *get_fork_point(const char *refname, struct commit *commit) |
| 1088 | { |
| 1089 | struct object_id oid; |
| 1090 | struct rev_collect revs; |
| 1091 | struct commit_list *bases = NULL; |
| 1092 | size_t i; |
| 1093 | struct commit *ret = NULL; |
| 1094 | char *full_refname; |
| 1095 | |
| 1096 | switch (repo_dwim_ref(the_repository, refname, strlen(refname), &oid, |
| 1097 | &full_refname, 0)) { |
| 1098 | case 0: |
| 1099 | die("No such ref: '%s'", refname); |
| 1100 | case 1: |
| 1101 | break; /* good */ |
| 1102 | default: |
| 1103 | die("Ambiguous refname: '%s'", refname); |
| 1104 | } |
| 1105 | |
| 1106 | commit_stack_init(&revs.stack); |
| 1107 | revs.initial = 1; |
| 1108 | refs_for_each_reflog_ent(get_main_ref_store(the_repository), |
| 1109 | full_refname, collect_one_reflog_ent, &revs); |
| 1110 | |
| 1111 | if (!revs.stack.nr) |
| 1112 | add_one_commit(&oid, &revs); |
| 1113 | |
| 1114 | for (i = 0; i < revs.stack.nr; i++) |
| 1115 | revs.stack.items[i]->object.flags &= ~TMP_MARK; |
| 1116 | |
| 1117 | if (repo_get_merge_bases_many(the_repository, commit, revs.stack.nr, |
| 1118 | revs.stack.items, &bases) < 0) |
| 1119 | exit(128); |
| 1120 | |
| 1121 | /* |
| 1122 | * There should be one and only one merge base, when we found |
| 1123 | * a common ancestor among reflog entries. |
| 1124 | */ |
| 1125 | if (!bases || bases->next) |
| 1126 | goto cleanup_return; |
| 1127 | |
| 1128 | /* And the found one must be one of the reflog entries */ |
| 1129 | for (i = 0; i < revs.stack.nr; i++) |
| 1130 | if (&bases->item->object == &revs.stack.items[i]->object) |
| 1131 | break; /* found */ |
| 1132 | if (revs.stack.nr <= i) |
| 1133 | goto cleanup_return; |
| 1134 | |
| 1135 | ret = bases->item; |
| 1136 | |
| 1137 | cleanup_return: |
| 1138 | commit_stack_clear(&revs.stack); |
| 1139 | commit_list_free(bases); |
| 1140 | free(full_refname); |
| 1141 | return ret; |
| 1142 | } |
| 1143 | |
| 1144 | /* |
| 1145 | * Indexed by hash algorithm identifier. |
| 1146 | */ |
| 1147 | static const char *gpg_sig_headers[] = { |
| 1148 | NULL, |
| 1149 | "gpgsig", |
| 1150 | "gpgsig-sha256", |
| 1151 | }; |
| 1152 | |
| 1153 | int add_header_signature(struct strbuf *buf, struct strbuf *sig, const struct git_hash_algo *algo) |
| 1154 | { |
| 1155 | int inspos, copypos; |
| 1156 | const char *eoh; |
| 1157 | const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(algo)]; |
| 1158 | int gpg_sig_header_len = strlen(gpg_sig_header); |
| 1159 | |
| 1160 | /* find the end of the header */ |
| 1161 | eoh = strstr(buf->buf, "\n\n"); |
| 1162 | if (!eoh) |
| 1163 | inspos = buf->len; |
| 1164 | else |
| 1165 | inspos = eoh - buf->buf + 1; |
| 1166 | |
| 1167 | for (copypos = 0; sig->buf[copypos]; ) { |
| 1168 | const char *bol = sig->buf + copypos; |
| 1169 | const char *eol = strchrnul(bol, '\n'); |
| 1170 | int len = (eol - bol) + !!*eol; |
| 1171 | |
| 1172 | if (!copypos) { |
| 1173 | strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len); |
| 1174 | inspos += gpg_sig_header_len; |
| 1175 | } |
| 1176 | strbuf_insertstr(buf, inspos++, " "); |
| 1177 | strbuf_insert(buf, inspos, bol, len); |
| 1178 | inspos += len; |
| 1179 | copypos += len; |
| 1180 | } |
| 1181 | return 0; |
| 1182 | } |
| 1183 | |
| 1184 | int parse_signed_commit(const struct commit *commit, |
| 1185 | struct strbuf *payload, struct strbuf *signature, |
| 1186 | const struct git_hash_algo *algop) |
| 1187 | { |
| 1188 | unsigned long size; |
| 1189 | const char *buffer = repo_get_commit_buffer(the_repository, commit, |
| 1190 | &size); |
| 1191 | int ret = parse_buffer_signed_by_header(buffer, size, payload, signature, algop); |
| 1192 | |
| 1193 | repo_unuse_commit_buffer(the_repository, commit, buffer); |
| 1194 | return ret; |
| 1195 | } |
| 1196 | |
| 1197 | int parse_buffer_signed_by_header(const char *buffer, |
| 1198 | unsigned long size, |
| 1199 | struct strbuf *payload, |
| 1200 | struct strbuf *signature, |
| 1201 | const struct git_hash_algo *algop) |
| 1202 | { |
| 1203 | int in_signature = 0, saw_signature = 0, other_signature = 0; |
| 1204 | const char *line, *tail, *p; |
| 1205 | const char *gpg_sig_header = gpg_sig_headers[hash_algo_by_ptr(algop)]; |
| 1206 | |
| 1207 | line = buffer; |
| 1208 | tail = buffer + size; |
| 1209 | while (line < tail) { |
| 1210 | const char *sig = NULL; |
| 1211 | const char *next = memchr(line, '\n', tail - line); |
| 1212 | |
| 1213 | next = next ? next + 1 : tail; |
| 1214 | if (in_signature && line[0] == ' ') |
| 1215 | sig = line + 1; |
| 1216 | else if (skip_prefix(line, gpg_sig_header, &p) && |
| 1217 | *p == ' ') { |
| 1218 | sig = line + strlen(gpg_sig_header) + 1; |
| 1219 | other_signature = 0; |
| 1220 | } |
| 1221 | else if (starts_with(line, "gpgsig")) |
| 1222 | other_signature = 1; |
| 1223 | else if (other_signature && line[0] != ' ') |
| 1224 | other_signature = 0; |
| 1225 | if (sig) { |
| 1226 | strbuf_add(signature, sig, next - sig); |
| 1227 | saw_signature = 1; |
| 1228 | in_signature = 1; |
| 1229 | } else { |
| 1230 | if (*line == '\n') |
| 1231 | /* dump the whole remainder of the buffer */ |
| 1232 | next = tail; |
| 1233 | if (!other_signature) |
| 1234 | strbuf_add(payload, line, next - line); |
| 1235 | in_signature = 0; |
| 1236 | } |
| 1237 | line = next; |
| 1238 | } |
| 1239 | return saw_signature; |
| 1240 | } |
| 1241 | |
| 1242 | int remove_signature(struct strbuf *buf) |
| 1243 | { |
| 1244 | const char *line = buf->buf; |
| 1245 | const char *tail = buf->buf + buf->len; |
| 1246 | int in_signature = 0; |
| 1247 | struct sigbuf { |
| 1248 | const char *start; |
| 1249 | const char *end; |
| 1250 | } sigs[2], *sigp = &sigs[0]; |
| 1251 | int i; |
| 1252 | const char *orig_buf = buf->buf; |
| 1253 | |
| 1254 | memset(sigs, 0, sizeof(sigs)); |
| 1255 | |
| 1256 | while (line < tail) { |
| 1257 | const char *next = memchr(line, '\n', tail - line); |
| 1258 | next = next ? next + 1 : tail; |
| 1259 | |
| 1260 | if (in_signature && line[0] == ' ') |
| 1261 | sigp->end = next; |
| 1262 | else if (starts_with(line, "gpgsig")) { |
| 1263 | int i; |
| 1264 | for (i = 1; i < GIT_HASH_NALGOS; i++) { |
| 1265 | const char *p; |
| 1266 | if (skip_prefix(line, gpg_sig_headers[i], &p) && |
| 1267 | *p == ' ') { |
| 1268 | sigp->start = line; |
| 1269 | sigp->end = next; |
| 1270 | in_signature = 1; |
| 1271 | } |
| 1272 | } |
| 1273 | } else { |
| 1274 | if (*line == '\n') |
| 1275 | /* dump the whole remainder of the buffer */ |
| 1276 | next = tail; |
| 1277 | if (in_signature && sigp - sigs != ARRAY_SIZE(sigs)) |
| 1278 | sigp++; |
| 1279 | in_signature = 0; |
| 1280 | } |
| 1281 | line = next; |
| 1282 | } |
| 1283 | |
| 1284 | for (i = ARRAY_SIZE(sigs) - 1; i >= 0; i--) |
| 1285 | if (sigs[i].start) |
| 1286 | strbuf_remove(buf, sigs[i].start - orig_buf, sigs[i].end - sigs[i].start); |
| 1287 | |
| 1288 | return sigs[0].start != NULL; |
| 1289 | } |
| 1290 | |
| 1291 | static void handle_signed_tag(const struct commit *parent, struct commit_extra_header ***tail) |
| 1292 | { |
| 1293 | struct merge_remote_desc *desc; |
| 1294 | struct commit_extra_header *mergetag; |
| 1295 | char *buf; |
| 1296 | size_t size; |
| 1297 | enum object_type type; |
| 1298 | struct strbuf payload = STRBUF_INIT; |
| 1299 | struct strbuf signature = STRBUF_INIT; |
| 1300 | |
| 1301 | desc = merge_remote_util(parent); |
| 1302 | if (!desc || !desc->obj) |
| 1303 | return; |
| 1304 | buf = odb_read_object(the_repository->objects, &desc->obj->oid, |
| 1305 | &type, &size); |
| 1306 | if (!buf || type != OBJ_TAG) |
| 1307 | goto free_return; |
| 1308 | if (!parse_signature(buf, size, &payload, &signature)) |
| 1309 | goto free_return; |
| 1310 | /* |
| 1311 | * We could verify this signature and either omit the tag when |
| 1312 | * it does not validate, but the integrator may not have the |
| 1313 | * public key of the signer of the tag being merged, while a |
| 1314 | * later auditor may have it while auditing, so let's not run |
| 1315 | * verify-signed-buffer here for now... |
| 1316 | * |
| 1317 | * if (verify_signed_buffer(buf, len, buf + len, size - len, ...)) |
| 1318 | * warn("warning: signed tag unverified."); |
| 1319 | */ |
| 1320 | CALLOC_ARRAY(mergetag, 1); |
| 1321 | mergetag->key = xstrdup("mergetag"); |
| 1322 | mergetag->value = buf; |
| 1323 | mergetag->len = size; |
| 1324 | |
| 1325 | **tail = mergetag; |
| 1326 | *tail = &mergetag->next; |
| 1327 | strbuf_release(&payload); |
| 1328 | strbuf_release(&signature); |
| 1329 | return; |
| 1330 | |
| 1331 | free_return: |
| 1332 | free(buf); |
| 1333 | } |
| 1334 | |
| 1335 | int verify_commit_buffer(const char *buffer, size_t size, |
| 1336 | struct signature_check *sigc) |
| 1337 | { |
| 1338 | struct strbuf payload = STRBUF_INIT; |
| 1339 | struct strbuf signature = STRBUF_INIT; |
| 1340 | int ret = 1; |
| 1341 | |
| 1342 | sigc->result = 'N'; |
| 1343 | |
| 1344 | if (parse_buffer_signed_by_header(buffer, size, &payload, |
| 1345 | &signature, the_hash_algo) <= 0) |
| 1346 | goto out; |
| 1347 | |
| 1348 | sigc->payload_type = SIGNATURE_PAYLOAD_COMMIT; |
| 1349 | sigc->payload = strbuf_detach(&payload, &sigc->payload_len); |
| 1350 | ret = check_signature(sigc, signature.buf, signature.len); |
| 1351 | |
| 1352 | out: |
| 1353 | strbuf_release(&payload); |
| 1354 | strbuf_release(&signature); |
| 1355 | |
| 1356 | return ret; |
| 1357 | } |
| 1358 | |
| 1359 | int check_commit_signature(const struct commit *commit, struct signature_check *sigc) |
| 1360 | { |
| 1361 | unsigned long size; |
| 1362 | const char *buffer = repo_get_commit_buffer(the_repository, commit, &size); |
| 1363 | int ret = verify_commit_buffer(buffer, size, sigc); |
| 1364 | |
| 1365 | repo_unuse_commit_buffer(the_repository, commit, buffer); |
| 1366 | |
| 1367 | return ret; |
| 1368 | } |
| 1369 | |
| 1370 | void verify_merge_signature(struct commit *commit, int verbosity, |
| 1371 | int check_trust) |
| 1372 | { |
| 1373 | char hex[GIT_MAX_HEXSZ + 1]; |
| 1374 | struct signature_check signature_check; |
| 1375 | int ret; |
| 1376 | memset(&signature_check, 0, sizeof(signature_check)); |
| 1377 | |
| 1378 | ret = check_commit_signature(commit, &signature_check); |
| 1379 | |
| 1380 | repo_find_unique_abbrev_r(the_repository, hex, &commit->object.oid, |
| 1381 | DEFAULT_ABBREV); |
| 1382 | switch (signature_check.result) { |
| 1383 | case 'G': |
| 1384 | if (ret || (check_trust && signature_check.trust_level < TRUST_MARGINAL)) |
| 1385 | die(_("Commit %s has an untrusted GPG signature, " |
| 1386 | "allegedly by %s."), hex, signature_check.signer); |
| 1387 | break; |
| 1388 | case 'B': |
| 1389 | die(_("Commit %s has a bad GPG signature " |
| 1390 | "allegedly by %s."), hex, signature_check.signer); |
| 1391 | default: /* 'N' */ |
| 1392 | die(_("Commit %s does not have a GPG signature."), hex); |
| 1393 | } |
| 1394 | if (verbosity >= 0 && signature_check.result == 'G') |
| 1395 | printf(_("Commit %s has a good GPG signature by %s\n"), |
| 1396 | hex, signature_check.signer); |
| 1397 | |
| 1398 | signature_check_clear(&signature_check); |
| 1399 | } |
| 1400 | |
| 1401 | void append_merge_tag_headers(const struct commit_list *parents, |
| 1402 | struct commit_extra_header ***tail) |
| 1403 | { |
| 1404 | while (parents) { |
| 1405 | const struct commit *parent = parents->item; |
| 1406 | handle_signed_tag(parent, tail); |
| 1407 | parents = parents->next; |
| 1408 | } |
| 1409 | } |
| 1410 | |
| 1411 | static int convert_commit_extra_headers(const struct commit_extra_header *orig, |
| 1412 | struct commit_extra_header **result) |
| 1413 | { |
| 1414 | const struct git_hash_algo *compat = the_repository->compat_hash_algo; |
| 1415 | const struct git_hash_algo *algo = the_repository->hash_algo; |
| 1416 | struct commit_extra_header *extra = NULL, **tail = &extra; |
| 1417 | struct strbuf out = STRBUF_INIT; |
| 1418 | while (orig) { |
| 1419 | struct commit_extra_header *new; |
| 1420 | CALLOC_ARRAY(new, 1); |
| 1421 | if (!strcmp(orig->key, "mergetag")) { |
| 1422 | if (convert_object_file(the_repository, &out, algo, compat, |
| 1423 | orig->value, orig->len, |
| 1424 | OBJ_TAG, 1)) { |
| 1425 | free(new); |
| 1426 | free_commit_extra_headers(extra); |
| 1427 | return -1; |
| 1428 | } |
| 1429 | new->key = xstrdup("mergetag"); |
| 1430 | new->value = strbuf_detach(&out, &new->len); |
| 1431 | } else { |
| 1432 | new->key = xstrdup(orig->key); |
| 1433 | new->len = orig->len; |
| 1434 | new->value = xmemdupz(orig->value, orig->len); |
| 1435 | } |
| 1436 | *tail = new; |
| 1437 | tail = &new->next; |
| 1438 | orig = orig->next; |
| 1439 | } |
| 1440 | *result = extra; |
| 1441 | return 0; |
| 1442 | } |
| 1443 | |
| 1444 | static void add_extra_header(struct strbuf *buffer, |
| 1445 | const struct commit_extra_header *extra) |
| 1446 | { |
| 1447 | strbuf_addstr(buffer, extra->key); |
| 1448 | if (extra->len) |
| 1449 | strbuf_add_lines(buffer, " ", extra->value, extra->len); |
| 1450 | else |
| 1451 | strbuf_addch(buffer, '\n'); |
| 1452 | } |
| 1453 | |
| 1454 | struct commit_extra_header *read_commit_extra_headers(struct commit *commit, |
| 1455 | const char **exclude) |
| 1456 | { |
| 1457 | struct commit_extra_header *extra = NULL; |
| 1458 | unsigned long size; |
| 1459 | const char *buffer = repo_get_commit_buffer(the_repository, commit, |
| 1460 | &size); |
| 1461 | extra = read_commit_extra_header_lines(buffer, size, exclude); |
| 1462 | repo_unuse_commit_buffer(the_repository, commit, buffer); |
| 1463 | return extra; |
| 1464 | } |
| 1465 | |
| 1466 | int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data) |
| 1467 | { |
| 1468 | struct commit_extra_header *extra, *to_free; |
| 1469 | int res = 0; |
| 1470 | |
| 1471 | to_free = read_commit_extra_headers(commit, NULL); |
| 1472 | for (extra = to_free; !res && extra; extra = extra->next) { |
| 1473 | if (strcmp(extra->key, "mergetag")) |
| 1474 | continue; /* not a merge tag */ |
| 1475 | res = fn(commit, extra, data); |
| 1476 | } |
| 1477 | free_commit_extra_headers(to_free); |
| 1478 | return res; |
| 1479 | } |
| 1480 | |
| 1481 | static inline int standard_header_field(const char *field, size_t len) |
| 1482 | { |
| 1483 | return ((len == 4 && !memcmp(field, "tree", 4)) || |
| 1484 | (len == 6 && !memcmp(field, "parent", 6)) || |
| 1485 | (len == 6 && !memcmp(field, "author", 6)) || |
| 1486 | (len == 9 && !memcmp(field, "committer", 9)) || |
| 1487 | (len == 8 && !memcmp(field, "encoding", 8))); |
| 1488 | } |
| 1489 | |
| 1490 | static int excluded_header_field(const char *field, size_t len, const char **exclude) |
| 1491 | { |
| 1492 | if (!exclude) |
| 1493 | return 0; |
| 1494 | |
| 1495 | while (*exclude) { |
| 1496 | size_t xlen = strlen(*exclude); |
| 1497 | if (len == xlen && !memcmp(field, *exclude, xlen)) |
| 1498 | return 1; |
| 1499 | exclude++; |
| 1500 | } |
| 1501 | return 0; |
| 1502 | } |
| 1503 | |
| 1504 | static struct commit_extra_header *read_commit_extra_header_lines( |
| 1505 | const char *buffer, size_t size, |
| 1506 | const char **exclude) |
| 1507 | { |
| 1508 | struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL; |
| 1509 | const char *line, *next, *eof, *eob; |
| 1510 | struct strbuf buf = STRBUF_INIT; |
| 1511 | |
| 1512 | for (line = buffer, eob = line + size; |
| 1513 | line < eob && *line != '\n'; |
| 1514 | line = next) { |
| 1515 | next = memchr(line, '\n', eob - line); |
| 1516 | next = next ? next + 1 : eob; |
| 1517 | if (*line == ' ') { |
| 1518 | /* continuation */ |
| 1519 | if (it) |
| 1520 | strbuf_add(&buf, line + 1, next - (line + 1)); |
| 1521 | continue; |
| 1522 | } |
| 1523 | if (it) |
| 1524 | it->value = strbuf_detach(&buf, &it->len); |
| 1525 | strbuf_reset(&buf); |
| 1526 | it = NULL; |
| 1527 | |
| 1528 | eof = memchr(line, ' ', next - line); |
| 1529 | if (!eof) |
| 1530 | eof = next; |
| 1531 | else if (standard_header_field(line, eof - line) || |
| 1532 | excluded_header_field(line, eof - line, exclude)) |
| 1533 | continue; |
| 1534 | |
| 1535 | CALLOC_ARRAY(it, 1); |
| 1536 | it->key = xmemdupz(line, eof-line); |
| 1537 | *tail = it; |
| 1538 | tail = &it->next; |
| 1539 | if (eof + 1 < next) |
| 1540 | strbuf_add(&buf, eof + 1, next - (eof + 1)); |
| 1541 | } |
| 1542 | if (it) |
| 1543 | it->value = strbuf_detach(&buf, &it->len); |
| 1544 | return extra; |
| 1545 | } |
| 1546 | |
| 1547 | void free_commit_extra_headers(struct commit_extra_header *extra) |
| 1548 | { |
| 1549 | while (extra) { |
| 1550 | struct commit_extra_header *next = extra->next; |
| 1551 | free(extra->key); |
| 1552 | free(extra->value); |
| 1553 | free(extra); |
| 1554 | extra = next; |
| 1555 | } |
| 1556 | } |
| 1557 | |
| 1558 | int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree, |
| 1559 | const struct commit_list *parents, struct object_id *ret, |
| 1560 | const char *author, const char *sign_commit) |
| 1561 | { |
| 1562 | struct commit_extra_header *extra = NULL, **tail = &extra; |
| 1563 | int result; |
| 1564 | |
| 1565 | append_merge_tag_headers(parents, &tail); |
| 1566 | result = commit_tree_extended(msg, msg_len, tree, parents, ret, author, |
| 1567 | NULL, sign_commit, extra); |
| 1568 | free_commit_extra_headers(extra); |
| 1569 | return result; |
| 1570 | } |
| 1571 | |
| 1572 | static bool has_invalid_utf8(const char *buf, size_t len, size_t *bad_offset) |
| 1573 | { |
| 1574 | size_t offset = 0; |
| 1575 | static const unsigned int max_codepoint[] = { |
| 1576 | 0x7f, 0x7ff, 0xffff, 0x10ffff |
| 1577 | }; |
| 1578 | |
| 1579 | while (len) { |
| 1580 | unsigned char c = *buf++; |
| 1581 | unsigned bytes; |
| 1582 | unsigned int codepoint; |
| 1583 | unsigned int min_val, max_val; |
| 1584 | |
| 1585 | len--; |
| 1586 | offset++; |
| 1587 | |
| 1588 | /* Simple US-ASCII? No worries. */ |
| 1589 | if (c < 0x80) |
| 1590 | continue; |
| 1591 | |
| 1592 | *bad_offset = offset-1; |
| 1593 | |
| 1594 | /* |
| 1595 | * Count how many more high bits set: that's how |
| 1596 | * many more bytes this sequence should have. |
| 1597 | */ |
| 1598 | bytes = 0; |
| 1599 | while (c & 0x40) { |
| 1600 | c <<= 1; |
| 1601 | bytes++; |
| 1602 | } |
| 1603 | |
| 1604 | /* |
| 1605 | * Must be between 1 and 3 more bytes. Longer sequences result in |
| 1606 | * codepoints beyond U+10FFFF, which are guaranteed never to exist. |
| 1607 | */ |
| 1608 | if (bytes < 1 || 3 < bytes) |
| 1609 | return true; |
| 1610 | |
| 1611 | /* Do we *have* that many bytes? */ |
| 1612 | if (len < bytes) |
| 1613 | return true; |
| 1614 | |
| 1615 | /* |
| 1616 | * Place the encoded bits at the bottom of the value and compute the |
| 1617 | * valid range. |
| 1618 | */ |
| 1619 | codepoint = (c & 0x7f) >> bytes; |
| 1620 | min_val = max_codepoint[bytes-1] + 1; |
| 1621 | max_val = max_codepoint[bytes]; |
| 1622 | |
| 1623 | offset += bytes; |
| 1624 | len -= bytes; |
| 1625 | |
| 1626 | /* And verify that they are good continuation bytes */ |
| 1627 | do { |
| 1628 | codepoint <<= 6; |
| 1629 | codepoint |= *buf & 0x3f; |
| 1630 | if ((*buf++ & 0xc0) != 0x80) |
| 1631 | return true; |
| 1632 | } while (--bytes); |
| 1633 | |
| 1634 | /* Reject codepoints that are out of range for the sequence length. */ |
| 1635 | if (codepoint < min_val || codepoint > max_val) |
| 1636 | return true; |
| 1637 | /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */ |
| 1638 | if ((codepoint & 0x1ff800) == 0xd800) |
| 1639 | return true; |
| 1640 | /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */ |
| 1641 | if ((codepoint & 0xfffe) == 0xfffe) |
| 1642 | return true; |
| 1643 | /* So are anything in the range U+FDD0..U+FDEF. */ |
| 1644 | if (codepoint >= 0xfdd0 && codepoint <= 0xfdef) |
| 1645 | return true; |
| 1646 | } |
| 1647 | return false; |
| 1648 | } |
| 1649 | |
| 1650 | /* |
| 1651 | * This ensures that the buffer is in proper utf8 format. |
| 1652 | * |
| 1653 | * If it isn't, it assumes any non-utf8 characters are Latin1, |
| 1654 | * and does the conversion. |
| 1655 | */ |
| 1656 | static int ensure_utf8(struct strbuf *buf) |
| 1657 | { |
| 1658 | int ok = 1; |
| 1659 | size_t pos = 0; |
| 1660 | |
| 1661 | for (;;) { |
| 1662 | size_t bad; |
| 1663 | unsigned char c; |
| 1664 | unsigned char replace[2]; |
| 1665 | |
| 1666 | if (!has_invalid_utf8(buf->buf + pos, buf->len - pos, &bad)) |
| 1667 | return ok; |
| 1668 | pos += bad; |
| 1669 | ok = 0; |
| 1670 | c = buf->buf[pos]; |
| 1671 | strbuf_remove(buf, pos, 1); |
| 1672 | |
| 1673 | /* We know 'c' must be in the range 128-255 */ |
| 1674 | replace[0] = 0xc0 + (c >> 6); |
| 1675 | replace[1] = 0x80 + (c & 0x3f); |
| 1676 | strbuf_insert(buf, pos, replace, 2); |
| 1677 | pos += 2; |
| 1678 | } |
| 1679 | } |
| 1680 | |
| 1681 | static const char commit_utf8_warn[] = |
| 1682 | N_("Warning: commit message did not conform to UTF-8.\n" |
| 1683 | "You may want to amend it after fixing the message, or set the config\n" |
| 1684 | "variable i18n.commitEncoding to the encoding your project uses.\n"); |
| 1685 | |
| 1686 | static void write_commit_tree(struct strbuf *buffer, const char *msg, size_t msg_len, |
| 1687 | const struct object_id *tree, |
| 1688 | const struct object_id *parents, size_t parents_len, |
| 1689 | const char *author, const char *committer, |
| 1690 | const struct commit_extra_header *extra) |
| 1691 | { |
| 1692 | int encoding_is_utf8; |
| 1693 | size_t i; |
| 1694 | |
| 1695 | /* Not having i18n.commitencoding is the same as having utf-8 */ |
| 1696 | encoding_is_utf8 = is_encoding_utf8(git_commit_encoding); |
| 1697 | |
| 1698 | strbuf_grow(buffer, 8192); /* should avoid reallocs for the headers */ |
| 1699 | strbuf_addf(buffer, "tree %s\n", oid_to_hex(tree)); |
| 1700 | |
| 1701 | /* |
| 1702 | * NOTE! This ordering means that the same exact tree merged with a |
| 1703 | * different order of parents will be a _different_ changeset even |
| 1704 | * if everything else stays the same. |
| 1705 | */ |
| 1706 | for (i = 0; i < parents_len; i++) |
| 1707 | strbuf_addf(buffer, "parent %s\n", oid_to_hex(&parents[i])); |
| 1708 | |
| 1709 | /* Person/date information */ |
| 1710 | if (!author) |
| 1711 | author = git_author_info(IDENT_STRICT); |
| 1712 | strbuf_addf(buffer, "author %s\n", author); |
| 1713 | if (!committer) |
| 1714 | committer = git_committer_info(IDENT_STRICT); |
| 1715 | strbuf_addf(buffer, "committer %s\n", committer); |
| 1716 | if (!encoding_is_utf8) |
| 1717 | strbuf_addf(buffer, "encoding %s\n", git_commit_encoding); |
| 1718 | |
| 1719 | while (extra) { |
| 1720 | add_extra_header(buffer, extra); |
| 1721 | extra = extra->next; |
| 1722 | } |
| 1723 | strbuf_addch(buffer, '\n'); |
| 1724 | |
| 1725 | /* And add the comment */ |
| 1726 | strbuf_add(buffer, msg, msg_len); |
| 1727 | } |
| 1728 | |
| 1729 | int commit_tree_extended(const char *msg, size_t msg_len, |
| 1730 | const struct object_id *tree, |
| 1731 | const struct commit_list *parents, struct object_id *ret, |
| 1732 | const char *author, const char *committer, |
| 1733 | const char *sign_commit, |
| 1734 | const struct commit_extra_header *extra) |
| 1735 | { |
| 1736 | struct repository *r = the_repository; |
| 1737 | int result = 0; |
| 1738 | int encoding_is_utf8; |
| 1739 | bool warned = false; |
| 1740 | struct strbuf buffer = STRBUF_INIT, compat_buffer = STRBUF_INIT; |
| 1741 | struct strbuf sig = STRBUF_INIT, compat_sig = STRBUF_INIT; |
| 1742 | struct object_id *parent_buf = NULL, *compat_oid = NULL; |
| 1743 | struct object_id compat_oid_buf; |
| 1744 | size_t i, nparents; |
| 1745 | |
| 1746 | /* Not having i18n.commitencoding is the same as having utf-8 */ |
| 1747 | encoding_is_utf8 = is_encoding_utf8(git_commit_encoding); |
| 1748 | |
| 1749 | odb_assert_oid_type(the_repository->objects, tree, OBJ_TREE); |
| 1750 | |
| 1751 | if (memchr(msg, '\0', msg_len)) |
| 1752 | return error("a NUL byte in commit log message not allowed."); |
| 1753 | |
| 1754 | nparents = commit_list_count(parents); |
| 1755 | CALLOC_ARRAY(parent_buf, nparents); |
| 1756 | i = 0; |
| 1757 | for (const struct commit_list *p = parents; p; p = p->next) |
| 1758 | oidcpy(&parent_buf[i++], &p->item->object.oid); |
| 1759 | |
| 1760 | write_commit_tree(&buffer, msg, msg_len, tree, parent_buf, nparents, author, committer, extra); |
| 1761 | |
| 1762 | /* And check the encoding. */ |
| 1763 | if (encoding_is_utf8 && !ensure_utf8(&buffer)) { |
| 1764 | fprintf(stderr, _(commit_utf8_warn)); |
| 1765 | warned = true; |
| 1766 | } |
| 1767 | |
| 1768 | if (sign_commit && sign_buffer(&buffer, &sig, sign_commit, |
| 1769 | SIGN_BUFFER_USE_DEFAULT_KEY)) { |
| 1770 | result = -1; |
| 1771 | goto out; |
| 1772 | } |
| 1773 | if (r->compat_hash_algo) { |
| 1774 | struct commit_extra_header *compat_extra = NULL; |
| 1775 | struct object_id mapped_tree; |
| 1776 | struct object_id *mapped_parents; |
| 1777 | |
| 1778 | CALLOC_ARRAY(mapped_parents, nparents); |
| 1779 | |
| 1780 | if (repo_oid_to_algop(r, tree, r->compat_hash_algo, &mapped_tree)) { |
| 1781 | result = -1; |
| 1782 | free(mapped_parents); |
| 1783 | goto out; |
| 1784 | } |
| 1785 | for (i = 0; i < nparents; i++) |
| 1786 | if (repo_oid_to_algop(r, &parent_buf[i], r->compat_hash_algo, &mapped_parents[i])) { |
| 1787 | result = -1; |
| 1788 | free(mapped_parents); |
| 1789 | goto out; |
| 1790 | } |
| 1791 | if (convert_commit_extra_headers(extra, &compat_extra)) { |
| 1792 | result = -1; |
| 1793 | free(mapped_parents); |
| 1794 | goto out; |
| 1795 | } |
| 1796 | write_commit_tree(&compat_buffer, msg, msg_len, &mapped_tree, |
| 1797 | mapped_parents, nparents, author, committer, compat_extra); |
| 1798 | free_commit_extra_headers(compat_extra); |
| 1799 | free(mapped_parents); |
| 1800 | |
| 1801 | if (encoding_is_utf8 && !ensure_utf8(&compat_buffer) && !warned) |
| 1802 | fprintf(stderr, _(commit_utf8_warn)); |
| 1803 | |
| 1804 | if (sign_commit && sign_buffer(&compat_buffer, &compat_sig, |
| 1805 | sign_commit, |
| 1806 | SIGN_BUFFER_USE_DEFAULT_KEY)) { |
| 1807 | result = -1; |
| 1808 | goto out; |
| 1809 | } |
| 1810 | } |
| 1811 | |
| 1812 | if (sign_commit) { |
| 1813 | struct sig_pairs { |
| 1814 | struct strbuf *sig; |
| 1815 | const struct git_hash_algo *algo; |
| 1816 | } bufs [2] = { |
| 1817 | { &compat_sig, r->compat_hash_algo }, |
| 1818 | { &sig, r->hash_algo }, |
| 1819 | }; |
| 1820 | |
| 1821 | /* |
| 1822 | * We write algorithms in the order they were implemented in |
| 1823 | * Git to produce a stable hash when multiple algorithms are |
| 1824 | * used. |
| 1825 | */ |
| 1826 | if (r->compat_hash_algo && hash_algo_by_ptr(bufs[0].algo) > hash_algo_by_ptr(bufs[1].algo)) |
| 1827 | SWAP(bufs[0], bufs[1]); |
| 1828 | |
| 1829 | /* |
| 1830 | * We traverse each algorithm in order, and apply the signature |
| 1831 | * to each buffer. |
| 1832 | */ |
| 1833 | for (size_t i = 0; i < ARRAY_SIZE(bufs); i++) { |
| 1834 | if (!bufs[i].algo) |
| 1835 | continue; |
| 1836 | add_header_signature(&buffer, bufs[i].sig, bufs[i].algo); |
| 1837 | if (r->compat_hash_algo) |
| 1838 | add_header_signature(&compat_buffer, bufs[i].sig, bufs[i].algo); |
| 1839 | } |
| 1840 | } |
| 1841 | |
| 1842 | if (r->compat_hash_algo) { |
| 1843 | hash_object_file(r->compat_hash_algo, compat_buffer.buf, compat_buffer.len, |
| 1844 | OBJ_COMMIT, &compat_oid_buf); |
| 1845 | compat_oid = &compat_oid_buf; |
| 1846 | } |
| 1847 | |
| 1848 | result = odb_write_object_ext(the_repository->objects, buffer.buf, buffer.len, |
| 1849 | OBJ_COMMIT, ret, compat_oid, 0); |
| 1850 | out: |
| 1851 | free(parent_buf); |
| 1852 | strbuf_release(&buffer); |
| 1853 | strbuf_release(&compat_buffer); |
| 1854 | strbuf_release(&sig); |
| 1855 | strbuf_release(&compat_sig); |
| 1856 | return result; |
| 1857 | } |
| 1858 | |
| 1859 | define_commit_slab(merge_desc_slab, struct merge_remote_desc *); |
| 1860 | static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab); |
| 1861 | |
| 1862 | struct merge_remote_desc *merge_remote_util(const struct commit *commit) |
| 1863 | { |
| 1864 | return *merge_desc_slab_at(&merge_desc_slab, commit); |
| 1865 | } |
| 1866 | |
| 1867 | void set_merge_remote_desc(struct commit *commit, |
| 1868 | const char *name, struct object *obj) |
| 1869 | { |
| 1870 | struct merge_remote_desc *desc; |
| 1871 | FLEX_ALLOC_STR(desc, name, name); |
| 1872 | desc->obj = obj; |
| 1873 | *merge_desc_slab_at(&merge_desc_slab, commit) = desc; |
| 1874 | } |
| 1875 | |
| 1876 | struct commit *get_merge_parent(const char *name) |
| 1877 | { |
| 1878 | struct object *obj; |
| 1879 | struct commit *commit; |
| 1880 | struct object_id oid; |
| 1881 | if (repo_get_oid(the_repository, name, &oid)) |
| 1882 | return NULL; |
| 1883 | obj = parse_object(the_repository, &oid); |
| 1884 | commit = (struct commit *)repo_peel_to_type(the_repository, name, 0, |
| 1885 | obj, OBJ_COMMIT); |
| 1886 | if (commit && !merge_remote_util(commit)) |
| 1887 | set_merge_remote_desc(commit, name, obj); |
| 1888 | return commit; |
| 1889 | } |
| 1890 | |
| 1891 | /* |
| 1892 | * Append a commit to the end of the commit_list. |
| 1893 | * |
| 1894 | * next starts by pointing to the variable that holds the head of an |
| 1895 | * empty commit_list, and is updated to point to the "next" field of |
| 1896 | * the last item on the list as new commits are appended. |
| 1897 | * |
| 1898 | * Usage example: |
| 1899 | * |
| 1900 | * struct commit_list *list; |
| 1901 | * struct commit_list **next = &list; |
| 1902 | * |
| 1903 | * next = commit_list_append(c1, next); |
| 1904 | * next = commit_list_append(c2, next); |
| 1905 | * assert(commit_list_count(list) == 2); |
| 1906 | * return list; |
| 1907 | */ |
| 1908 | struct commit_list **commit_list_append(struct commit *commit, |
| 1909 | struct commit_list **next) |
| 1910 | { |
| 1911 | struct commit_list *new_commit = xmalloc(sizeof(struct commit_list)); |
| 1912 | new_commit->item = commit; |
| 1913 | *next = new_commit; |
| 1914 | new_commit->next = NULL; |
| 1915 | return &new_commit->next; |
| 1916 | } |
| 1917 | |
| 1918 | const char *find_commit_header(const char *msg, const char *key, size_t *out_len) |
| 1919 | { |
| 1920 | int key_len = strlen(key); |
| 1921 | const char *line = msg; |
| 1922 | |
| 1923 | while (line) { |
| 1924 | const char *eol = strchrnul(line, '\n'); |
| 1925 | |
| 1926 | if (line == eol) |
| 1927 | return NULL; |
| 1928 | |
| 1929 | if (eol - line > key_len && |
| 1930 | !strncmp(line, key, key_len) && |
| 1931 | line[key_len] == ' ') { |
| 1932 | *out_len = eol - line - key_len - 1; |
| 1933 | return line + key_len + 1; |
| 1934 | } |
| 1935 | line = *eol ? eol + 1 : NULL; |
| 1936 | } |
| 1937 | return NULL; |
| 1938 | } |
| 1939 | |
| 1940 | /* |
| 1941 | * Inspect the given string and determine the true "end" of the log message, in |
| 1942 | * order to find where to put a new Signed-off-by trailer. Ignored are |
| 1943 | * trailing comment lines and blank lines. To support "git commit -s |
| 1944 | * --amend" on an existing commit, we also ignore "Conflicts:". To |
| 1945 | * support "git commit -v", we truncate at cut lines. |
| 1946 | * |
| 1947 | * Returns the number of bytes from the tail to ignore, to be fed as |
| 1948 | * the second parameter to append_signoff(). |
| 1949 | */ |
| 1950 | size_t ignored_log_message_bytes(const char *buf, size_t len) |
| 1951 | { |
| 1952 | size_t boc = 0; |
| 1953 | size_t bol = 0; |
| 1954 | int in_old_conflicts_block = 0; |
| 1955 | size_t cutoff = wt_status_locate_end(buf, len); |
| 1956 | |
| 1957 | while (bol < cutoff) { |
| 1958 | const char *next_line = memchr(buf + bol, '\n', len - bol); |
| 1959 | |
| 1960 | if (!next_line) |
| 1961 | next_line = buf + len; |
| 1962 | else |
| 1963 | next_line++; |
| 1964 | |
| 1965 | if (starts_with_mem(buf + bol, cutoff - bol, comment_line_str) || |
| 1966 | buf[bol] == '\n') { |
| 1967 | /* is this the first of the run of comments? */ |
| 1968 | if (!boc) |
| 1969 | boc = bol; |
| 1970 | /* otherwise, it is just continuing */ |
| 1971 | } else if (starts_with(buf + bol, "Conflicts:\n")) { |
| 1972 | in_old_conflicts_block = 1; |
| 1973 | if (!boc) |
| 1974 | boc = bol; |
| 1975 | } else if (in_old_conflicts_block && buf[bol] == '\t') { |
| 1976 | ; /* a pathname in the conflicts block */ |
| 1977 | } else if (boc) { |
| 1978 | /* the previous was not trailing comment */ |
| 1979 | boc = 0; |
| 1980 | in_old_conflicts_block = 0; |
| 1981 | } |
| 1982 | bol = next_line - buf; |
| 1983 | } |
| 1984 | return boc ? len - boc : len - cutoff; |
| 1985 | } |
| 1986 | |
| 1987 | int run_commit_hook(int editor_is_used, const char *index_file, |
| 1988 | int *invoked_hook, const char *name, ...) |
| 1989 | { |
| 1990 | struct run_hooks_opt opt = RUN_HOOKS_OPT_INIT_FORCE_SERIAL; |
| 1991 | va_list args; |
| 1992 | const char *arg; |
| 1993 | |
| 1994 | strvec_pushf(&opt.env, "GIT_INDEX_FILE=%s", index_file); |
| 1995 | |
| 1996 | /* |
| 1997 | * Let the hook know that no editor will be launched. |
| 1998 | */ |
| 1999 | if (!editor_is_used) |
| 2000 | strvec_push(&opt.env, "GIT_EDITOR=:"); |
| 2001 | |
| 2002 | va_start(args, name); |
| 2003 | while ((arg = va_arg(args, const char *))) |
| 2004 | strvec_push(&opt.args, arg); |
| 2005 | va_end(args); |
| 2006 | |
| 2007 | opt.invoked_hook = invoked_hook; |
| 2008 | return run_hooks_opt(the_repository, name, &opt); |
| 2009 | } |
| 2010 | |
| 2011 | void commit_stack_init(struct commit_stack *stack) |
| 2012 | { |
| 2013 | stack->items = NULL; |
| 2014 | stack->nr = stack->alloc = 0; |
| 2015 | } |
| 2016 | |
| 2017 | void commit_stack_grow(struct commit_stack *stack, size_t extra) |
| 2018 | { |
| 2019 | ALLOC_GROW(stack->items, st_add(stack->nr, extra), stack->alloc); |
| 2020 | } |
| 2021 | |
| 2022 | void commit_stack_push(struct commit_stack *stack, struct commit *commit) |
| 2023 | { |
| 2024 | commit_stack_grow(stack, 1); |
| 2025 | stack->items[stack->nr++] = commit; |
| 2026 | } |
| 2027 | |
| 2028 | struct commit *commit_stack_pop(struct commit_stack *stack) |
| 2029 | { |
| 2030 | return stack->nr ? stack->items[--stack->nr] : NULL; |
| 2031 | } |
| 2032 | |
| 2033 | void commit_stack_clear(struct commit_stack *stack) |
| 2034 | { |
| 2035 | free(stack->items); |
| 2036 | commit_stack_init(stack); |
| 2037 | } |