commit.con commit chainlint: match arbitrary here-docs tags rather than hard-coded names (c2c29cc)
   1#include "cache.h"
   2#include "tag.h"
   3#include "commit.h"
   4#include "commit-graph.h"
   5#include "pkt-line.h"
   6#include "utf8.h"
   7#include "diff.h"
   8#include "revision.h"
   9#include "notes.h"
  10#include "alloc.h"
  11#include "gpg-interface.h"
  12#include "mergesort.h"
  13#include "commit-slab.h"
  14#include "prio-queue.h"
  15#include "sha1-lookup.h"
  16#include "wt-status.h"
  17#include "advice.h"
  18
  19static struct commit_extra_header *read_commit_extra_header_lines(const char *buf, size_t len, const char **);
  20
  21int save_commit_buffer = 1;
  22
  23const char *commit_type = "commit";
  24
  25struct commit *lookup_commit_reference_gently(const struct object_id *oid,
  26                                              int quiet)
  27{
  28        struct object *obj = deref_tag(parse_object(oid), NULL, 0);
  29
  30        if (!obj)
  31                return NULL;
  32        return object_as_type(obj, OBJ_COMMIT, quiet);
  33}
  34
  35struct commit *lookup_commit_reference(const struct object_id *oid)
  36{
  37        return lookup_commit_reference_gently(oid, 0);
  38}
  39
  40struct commit *lookup_commit_or_die(const struct object_id *oid, const char *ref_name)
  41{
  42        struct commit *c = lookup_commit_reference(oid);
  43        if (!c)
  44                die(_("could not parse %s"), ref_name);
  45        if (oidcmp(oid, &c->object.oid)) {
  46                warning(_("%s %s is not a commit!"),
  47                        ref_name, oid_to_hex(oid));
  48        }
  49        return c;
  50}
  51
  52struct commit *lookup_commit(const struct object_id *oid)
  53{
  54        struct object *obj = lookup_object(oid->hash);
  55        if (!obj)
  56                return create_object(the_repository, oid->hash,
  57                                     alloc_commit_node(the_repository));
  58        return object_as_type(obj, OBJ_COMMIT, 0);
  59}
  60
  61struct commit *lookup_commit_reference_by_name(const char *name)
  62{
  63        struct object_id oid;
  64        struct commit *commit;
  65
  66        if (get_oid_committish(name, &oid))
  67                return NULL;
  68        commit = lookup_commit_reference(&oid);
  69        if (parse_commit(commit))
  70                return NULL;
  71        return commit;
  72}
  73
  74static timestamp_t parse_commit_date(const char *buf, const char *tail)
  75{
  76        const char *dateptr;
  77
  78        if (buf + 6 >= tail)
  79                return 0;
  80        if (memcmp(buf, "author", 6))
  81                return 0;
  82        while (buf < tail && *buf++ != '\n')
  83                /* nada */;
  84        if (buf + 9 >= tail)
  85                return 0;
  86        if (memcmp(buf, "committer", 9))
  87                return 0;
  88        while (buf < tail && *buf++ != '>')
  89                /* nada */;
  90        if (buf >= tail)
  91                return 0;
  92        dateptr = buf;
  93        while (buf < tail && *buf++ != '\n')
  94                /* nada */;
  95        if (buf >= tail)
  96                return 0;
  97        /* dateptr < buf && buf[-1] == '\n', so parsing will stop at buf-1 */
  98        return parse_timestamp(dateptr, NULL, 10);
  99}
 100
 101static struct commit_graft **commit_graft;
 102static int commit_graft_alloc, commit_graft_nr;
 103
 104static const unsigned char *commit_graft_sha1_access(size_t index, void *table)
 105{
 106        struct commit_graft **commit_graft_table = table;
 107        return commit_graft_table[index]->oid.hash;
 108}
 109
 110static int commit_graft_pos(const unsigned char *sha1)
 111{
 112        return sha1_pos(sha1, commit_graft, commit_graft_nr,
 113                        commit_graft_sha1_access);
 114}
 115
 116int register_commit_graft(struct commit_graft *graft, int ignore_dups)
 117{
 118        int pos = commit_graft_pos(graft->oid.hash);
 119
 120        if (0 <= pos) {
 121                if (ignore_dups)
 122                        free(graft);
 123                else {
 124                        free(commit_graft[pos]);
 125                        commit_graft[pos] = graft;
 126                }
 127                return 1;
 128        }
 129        pos = -pos - 1;
 130        ALLOC_GROW(commit_graft, commit_graft_nr + 1, commit_graft_alloc);
 131        commit_graft_nr++;
 132        if (pos < commit_graft_nr)
 133                MOVE_ARRAY(commit_graft + pos + 1, commit_graft + pos,
 134                           commit_graft_nr - pos - 1);
 135        commit_graft[pos] = graft;
 136        return 0;
 137}
 138
 139struct commit_graft *read_graft_line(struct strbuf *line)
 140{
 141        /* The format is just "Commit Parent1 Parent2 ...\n" */
 142        int i, phase;
 143        const char *tail = NULL;
 144        struct commit_graft *graft = NULL;
 145        struct object_id dummy_oid, *oid;
 146
 147        strbuf_rtrim(line);
 148        if (!line->len || line->buf[0] == '#')
 149                return NULL;
 150        /*
 151         * phase 0 verifies line, counts hashes in line and allocates graft
 152         * phase 1 fills graft
 153         */
 154        for (phase = 0; phase < 2; phase++) {
 155                oid = graft ? &graft->oid : &dummy_oid;
 156                if (parse_oid_hex(line->buf, oid, &tail))
 157                        goto bad_graft_data;
 158                for (i = 0; *tail != '\0'; i++) {
 159                        oid = graft ? &graft->parent[i] : &dummy_oid;
 160                        if (!isspace(*tail++) || parse_oid_hex(tail, oid, &tail))
 161                                goto bad_graft_data;
 162                }
 163                if (!graft) {
 164                        graft = xmalloc(st_add(sizeof(*graft),
 165                                               st_mult(sizeof(struct object_id), i)));
 166                        graft->nr_parent = i;
 167                }
 168        }
 169        return graft;
 170
 171bad_graft_data:
 172        error("bad graft data: %s", line->buf);
 173        assert(!graft);
 174        return NULL;
 175}
 176
 177static int read_graft_file(const char *graft_file)
 178{
 179        FILE *fp = fopen_or_warn(graft_file, "r");
 180        struct strbuf buf = STRBUF_INIT;
 181        if (!fp)
 182                return -1;
 183        if (advice_graft_file_deprecated)
 184                advise(_("Support for <GIT_DIR>/info/grafts is deprecated\n"
 185                         "and will be removed in a future Git version.\n"
 186                         "\n"
 187                         "Please use \"git replace --convert-graft-file\"\n"
 188                         "to convert the grafts into replace refs.\n"
 189                         "\n"
 190                         "Turn this message off by running\n"
 191                         "\"git config advice.graftFileDeprecated false\""));
 192        while (!strbuf_getwholeline(&buf, fp, '\n')) {
 193                /* The format is just "Commit Parent1 Parent2 ...\n" */
 194                struct commit_graft *graft = read_graft_line(&buf);
 195                if (!graft)
 196                        continue;
 197                if (register_commit_graft(graft, 1))
 198                        error("duplicate graft data: %s", buf.buf);
 199        }
 200        fclose(fp);
 201        strbuf_release(&buf);
 202        return 0;
 203}
 204
 205static void prepare_commit_graft(void)
 206{
 207        static int commit_graft_prepared;
 208        char *graft_file;
 209
 210        if (commit_graft_prepared)
 211                return;
 212        if (!startup_info->have_repository)
 213                return;
 214
 215        graft_file = get_graft_file();
 216        read_graft_file(graft_file);
 217        /* make sure shallows are read */
 218        is_repository_shallow();
 219        commit_graft_prepared = 1;
 220}
 221
 222struct commit_graft *lookup_commit_graft(const struct object_id *oid)
 223{
 224        int pos;
 225        prepare_commit_graft();
 226        pos = commit_graft_pos(oid->hash);
 227        if (pos < 0)
 228                return NULL;
 229        return commit_graft[pos];
 230}
 231
 232int for_each_commit_graft(each_commit_graft_fn fn, void *cb_data)
 233{
 234        int i, ret;
 235        for (i = ret = 0; i < commit_graft_nr && !ret; i++)
 236                ret = fn(commit_graft[i], cb_data);
 237        return ret;
 238}
 239
 240int unregister_shallow(const struct object_id *oid)
 241{
 242        int pos = commit_graft_pos(oid->hash);
 243        if (pos < 0)
 244                return -1;
 245        if (pos + 1 < commit_graft_nr)
 246                MOVE_ARRAY(commit_graft + pos, commit_graft + pos + 1,
 247                           commit_graft_nr - pos - 1);
 248        commit_graft_nr--;
 249        return 0;
 250}
 251
 252struct commit_buffer {
 253        void *buffer;
 254        unsigned long size;
 255};
 256define_commit_slab(buffer_slab, struct commit_buffer);
 257static struct buffer_slab buffer_slab = COMMIT_SLAB_INIT(1, buffer_slab);
 258
 259void set_commit_buffer(struct commit *commit, void *buffer, unsigned long size)
 260{
 261        struct commit_buffer *v = buffer_slab_at(&buffer_slab, commit);
 262        v->buffer = buffer;
 263        v->size = size;
 264}
 265
 266const void *get_cached_commit_buffer(const struct commit *commit, unsigned long *sizep)
 267{
 268        struct commit_buffer *v = buffer_slab_peek(&buffer_slab, commit);
 269        if (!v) {
 270                if (sizep)
 271                        *sizep = 0;
 272                return NULL;
 273        }
 274        if (sizep)
 275                *sizep = v->size;
 276        return v->buffer;
 277}
 278
 279const void *get_commit_buffer(const struct commit *commit, unsigned long *sizep)
 280{
 281        const void *ret = get_cached_commit_buffer(commit, sizep);
 282        if (!ret) {
 283                enum object_type type;
 284                unsigned long size;
 285                ret = read_object_file(&commit->object.oid, &type, &size);
 286                if (!ret)
 287                        die("cannot read commit object %s",
 288                            oid_to_hex(&commit->object.oid));
 289                if (type != OBJ_COMMIT)
 290                        die("expected commit for %s, got %s",
 291                            oid_to_hex(&commit->object.oid), type_name(type));
 292                if (sizep)
 293                        *sizep = size;
 294        }
 295        return ret;
 296}
 297
 298void unuse_commit_buffer(const struct commit *commit, const void *buffer)
 299{
 300        struct commit_buffer *v = buffer_slab_peek(&buffer_slab, commit);
 301        if (!(v && v->buffer == buffer))
 302                free((void *)buffer);
 303}
 304
 305void free_commit_buffer(struct commit *commit)
 306{
 307        struct commit_buffer *v = buffer_slab_peek(&buffer_slab, commit);
 308        if (v) {
 309                FREE_AND_NULL(v->buffer);
 310                v->size = 0;
 311        }
 312}
 313
 314struct tree *get_commit_tree(const struct commit *commit)
 315{
 316        if (commit->maybe_tree || !commit->object.parsed)
 317                return commit->maybe_tree;
 318
 319        if (commit->graph_pos == COMMIT_NOT_FROM_GRAPH)
 320                BUG("commit has NULL tree, but was not loaded from commit-graph");
 321
 322        return get_commit_tree_in_graph(commit);
 323}
 324
 325struct object_id *get_commit_tree_oid(const struct commit *commit)
 326{
 327        return &get_commit_tree(commit)->object.oid;
 328}
 329
 330void release_commit_memory(struct commit *c)
 331{
 332        c->maybe_tree = NULL;
 333        c->index = 0;
 334        free_commit_buffer(c);
 335        free_commit_list(c->parents);
 336        /* TODO: what about commit->util? */
 337
 338        c->object.parsed = 0;
 339}
 340
 341const void *detach_commit_buffer(struct commit *commit, unsigned long *sizep)
 342{
 343        struct commit_buffer *v = buffer_slab_peek(&buffer_slab, commit);
 344        void *ret;
 345
 346        if (!v) {
 347                if (sizep)
 348                        *sizep = 0;
 349                return NULL;
 350        }
 351        ret = v->buffer;
 352        if (sizep)
 353                *sizep = v->size;
 354
 355        v->buffer = NULL;
 356        v->size = 0;
 357        return ret;
 358}
 359
 360int parse_commit_buffer(struct commit *item, const void *buffer, unsigned long size, int check_graph)
 361{
 362        const char *tail = buffer;
 363        const char *bufptr = buffer;
 364        struct object_id parent;
 365        struct commit_list **pptr;
 366        struct commit_graft *graft;
 367        const int tree_entry_len = GIT_SHA1_HEXSZ + 5;
 368        const int parent_entry_len = GIT_SHA1_HEXSZ + 7;
 369
 370        if (item->object.parsed)
 371                return 0;
 372        item->object.parsed = 1;
 373        tail += size;
 374        if (tail <= bufptr + tree_entry_len + 1 || memcmp(bufptr, "tree ", 5) ||
 375                        bufptr[tree_entry_len] != '\n')
 376                return error("bogus commit object %s", oid_to_hex(&item->object.oid));
 377        if (get_oid_hex(bufptr + 5, &parent) < 0)
 378                return error("bad tree pointer in commit %s",
 379                             oid_to_hex(&item->object.oid));
 380        item->maybe_tree = lookup_tree(&parent);
 381        bufptr += tree_entry_len + 1; /* "tree " + "hex sha1" + "\n" */
 382        pptr = &item->parents;
 383
 384        graft = lookup_commit_graft(&item->object.oid);
 385        while (bufptr + parent_entry_len < tail && !memcmp(bufptr, "parent ", 7)) {
 386                struct commit *new_parent;
 387
 388                if (tail <= bufptr + parent_entry_len + 1 ||
 389                    get_oid_hex(bufptr + 7, &parent) ||
 390                    bufptr[parent_entry_len] != '\n')
 391                        return error("bad parents in commit %s", oid_to_hex(&item->object.oid));
 392                bufptr += parent_entry_len + 1;
 393                /*
 394                 * The clone is shallow if nr_parent < 0, and we must
 395                 * not traverse its real parents even when we unhide them.
 396                 */
 397                if (graft && (graft->nr_parent < 0 || grafts_replace_parents))
 398                        continue;
 399                new_parent = lookup_commit(&parent);
 400                if (new_parent)
 401                        pptr = &commit_list_insert(new_parent, pptr)->next;
 402        }
 403        if (graft) {
 404                int i;
 405                struct commit *new_parent;
 406                for (i = 0; i < graft->nr_parent; i++) {
 407                        new_parent = lookup_commit(&graft->parent[i]);
 408                        if (!new_parent)
 409                                continue;
 410                        pptr = &commit_list_insert(new_parent, pptr)->next;
 411                }
 412        }
 413        item->date = parse_commit_date(bufptr, tail);
 414
 415        if (check_graph)
 416                load_commit_graph_info(item);
 417
 418        return 0;
 419}
 420
 421int parse_commit_gently(struct commit *item, int quiet_on_missing)
 422{
 423        enum object_type type;
 424        void *buffer;
 425        unsigned long size;
 426        int ret;
 427
 428        if (!item)
 429                return -1;
 430        if (item->object.parsed)
 431                return 0;
 432        if (parse_commit_in_graph(item))
 433                return 0;
 434        buffer = read_object_file(&item->object.oid, &type, &size);
 435        if (!buffer)
 436                return quiet_on_missing ? -1 :
 437                        error("Could not read %s",
 438                             oid_to_hex(&item->object.oid));
 439        if (type != OBJ_COMMIT) {
 440                free(buffer);
 441                return error("Object %s not a commit",
 442                             oid_to_hex(&item->object.oid));
 443        }
 444        ret = parse_commit_buffer(item, buffer, size, 0);
 445        if (save_commit_buffer && !ret) {
 446                set_commit_buffer(item, buffer, size);
 447                return 0;
 448        }
 449        free(buffer);
 450        return ret;
 451}
 452
 453void parse_commit_or_die(struct commit *item)
 454{
 455        if (parse_commit(item))
 456                die("unable to parse commit %s",
 457                    item ? oid_to_hex(&item->object.oid) : "(null)");
 458}
 459
 460int find_commit_subject(const char *commit_buffer, const char **subject)
 461{
 462        const char *eol;
 463        const char *p = commit_buffer;
 464
 465        while (*p && (*p != '\n' || p[1] != '\n'))
 466                p++;
 467        if (*p) {
 468                p = skip_blank_lines(p + 2);
 469                eol = strchrnul(p, '\n');
 470        } else
 471                eol = p;
 472
 473        *subject = p;
 474
 475        return eol - p;
 476}
 477
 478struct commit_list *commit_list_insert(struct commit *item, struct commit_list **list_p)
 479{
 480        struct commit_list *new_list = xmalloc(sizeof(struct commit_list));
 481        new_list->item = item;
 482        new_list->next = *list_p;
 483        *list_p = new_list;
 484        return new_list;
 485}
 486
 487unsigned commit_list_count(const struct commit_list *l)
 488{
 489        unsigned c = 0;
 490        for (; l; l = l->next )
 491                c++;
 492        return c;
 493}
 494
 495struct commit_list *copy_commit_list(struct commit_list *list)
 496{
 497        struct commit_list *head = NULL;
 498        struct commit_list **pp = &head;
 499        while (list) {
 500                pp = commit_list_append(list->item, pp);
 501                list = list->next;
 502        }
 503        return head;
 504}
 505
 506void free_commit_list(struct commit_list *list)
 507{
 508        while (list)
 509                pop_commit(&list);
 510}
 511
 512struct commit_list * commit_list_insert_by_date(struct commit *item, struct commit_list **list)
 513{
 514        struct commit_list **pp = list;
 515        struct commit_list *p;
 516        while ((p = *pp) != NULL) {
 517                if (p->item->date < item->date) {
 518                        break;
 519                }
 520                pp = &p->next;
 521        }
 522        return commit_list_insert(item, pp);
 523}
 524
 525static int commit_list_compare_by_date(const void *a, const void *b)
 526{
 527        timestamp_t a_date = ((const struct commit_list *)a)->item->date;
 528        timestamp_t b_date = ((const struct commit_list *)b)->item->date;
 529        if (a_date < b_date)
 530                return 1;
 531        if (a_date > b_date)
 532                return -1;
 533        return 0;
 534}
 535
 536static void *commit_list_get_next(const void *a)
 537{
 538        return ((const struct commit_list *)a)->next;
 539}
 540
 541static void commit_list_set_next(void *a, void *next)
 542{
 543        ((struct commit_list *)a)->next = next;
 544}
 545
 546void commit_list_sort_by_date(struct commit_list **list)
 547{
 548        *list = llist_mergesort(*list, commit_list_get_next, commit_list_set_next,
 549                                commit_list_compare_by_date);
 550}
 551
 552struct commit *pop_most_recent_commit(struct commit_list **list,
 553                                      unsigned int mark)
 554{
 555        struct commit *ret = pop_commit(list);
 556        struct commit_list *parents = ret->parents;
 557
 558        while (parents) {
 559                struct commit *commit = parents->item;
 560                if (!parse_commit(commit) && !(commit->object.flags & mark)) {
 561                        commit->object.flags |= mark;
 562                        commit_list_insert_by_date(commit, list);
 563                }
 564                parents = parents->next;
 565        }
 566        return ret;
 567}
 568
 569static void clear_commit_marks_1(struct commit_list **plist,
 570                                 struct commit *commit, unsigned int mark)
 571{
 572        while (commit) {
 573                struct commit_list *parents;
 574
 575                if (!(mark & commit->object.flags))
 576                        return;
 577
 578                commit->object.flags &= ~mark;
 579
 580                parents = commit->parents;
 581                if (!parents)
 582                        return;
 583
 584                while ((parents = parents->next))
 585                        commit_list_insert(parents->item, plist);
 586
 587                commit = commit->parents->item;
 588        }
 589}
 590
 591void clear_commit_marks_many(int nr, struct commit **commit, unsigned int mark)
 592{
 593        struct commit_list *list = NULL;
 594
 595        while (nr--) {
 596                clear_commit_marks_1(&list, *commit, mark);
 597                commit++;
 598        }
 599        while (list)
 600                clear_commit_marks_1(&list, pop_commit(&list), mark);
 601}
 602
 603void clear_commit_marks(struct commit *commit, unsigned int mark)
 604{
 605        clear_commit_marks_many(1, &commit, mark);
 606}
 607
 608struct commit *pop_commit(struct commit_list **stack)
 609{
 610        struct commit_list *top = *stack;
 611        struct commit *item = top ? top->item : NULL;
 612
 613        if (top) {
 614                *stack = top->next;
 615                free(top);
 616        }
 617        return item;
 618}
 619
 620/*
 621 * Topological sort support
 622 */
 623
 624/* count number of children that have not been emitted */
 625define_commit_slab(indegree_slab, int);
 626
 627/* record author-date for each commit object */
 628define_commit_slab(author_date_slab, unsigned long);
 629
 630static void record_author_date(struct author_date_slab *author_date,
 631                               struct commit *commit)
 632{
 633        const char *buffer = get_commit_buffer(commit, NULL);
 634        struct ident_split ident;
 635        const char *ident_line;
 636        size_t ident_len;
 637        char *date_end;
 638        timestamp_t date;
 639
 640        ident_line = find_commit_header(buffer, "author", &ident_len);
 641        if (!ident_line)
 642                goto fail_exit; /* no author line */
 643        if (split_ident_line(&ident, ident_line, ident_len) ||
 644            !ident.date_begin || !ident.date_end)
 645                goto fail_exit; /* malformed "author" line */
 646
 647        date = parse_timestamp(ident.date_begin, &date_end, 10);
 648        if (date_end != ident.date_end)
 649                goto fail_exit; /* malformed date */
 650        *(author_date_slab_at(author_date, commit)) = date;
 651
 652fail_exit:
 653        unuse_commit_buffer(commit, buffer);
 654}
 655
 656static int compare_commits_by_author_date(const void *a_, const void *b_,
 657                                          void *cb_data)
 658{
 659        const struct commit *a = a_, *b = b_;
 660        struct author_date_slab *author_date = cb_data;
 661        timestamp_t a_date = *(author_date_slab_at(author_date, a));
 662        timestamp_t b_date = *(author_date_slab_at(author_date, b));
 663
 664        /* newer commits with larger date first */
 665        if (a_date < b_date)
 666                return 1;
 667        else if (a_date > b_date)
 668                return -1;
 669        return 0;
 670}
 671
 672int compare_commits_by_gen_then_commit_date(const void *a_, const void *b_, void *unused)
 673{
 674        const struct commit *a = a_, *b = b_;
 675
 676        /* newer commits first */
 677        if (a->generation < b->generation)
 678                return 1;
 679        else if (a->generation > b->generation)
 680                return -1;
 681
 682        /* use date as a heuristic when generations are equal */
 683        if (a->date < b->date)
 684                return 1;
 685        else if (a->date > b->date)
 686                return -1;
 687        return 0;
 688}
 689
 690int compare_commits_by_commit_date(const void *a_, const void *b_, void *unused)
 691{
 692        const struct commit *a = a_, *b = b_;
 693        /* newer commits with larger date first */
 694        if (a->date < b->date)
 695                return 1;
 696        else if (a->date > b->date)
 697                return -1;
 698        return 0;
 699}
 700
 701/*
 702 * Performs an in-place topological sort on the list supplied.
 703 */
 704void sort_in_topological_order(struct commit_list **list, enum rev_sort_order sort_order)
 705{
 706        struct commit_list *next, *orig = *list;
 707        struct commit_list **pptr;
 708        struct indegree_slab indegree;
 709        struct prio_queue queue;
 710        struct commit *commit;
 711        struct author_date_slab author_date;
 712
 713        if (!orig)
 714                return;
 715        *list = NULL;
 716
 717        init_indegree_slab(&indegree);
 718        memset(&queue, '\0', sizeof(queue));
 719
 720        switch (sort_order) {
 721        default: /* REV_SORT_IN_GRAPH_ORDER */
 722                queue.compare = NULL;
 723                break;
 724        case REV_SORT_BY_COMMIT_DATE:
 725                queue.compare = compare_commits_by_commit_date;
 726                break;
 727        case REV_SORT_BY_AUTHOR_DATE:
 728                init_author_date_slab(&author_date);
 729                queue.compare = compare_commits_by_author_date;
 730                queue.cb_data = &author_date;
 731                break;
 732        }
 733
 734        /* Mark them and clear the indegree */
 735        for (next = orig; next; next = next->next) {
 736                struct commit *commit = next->item;
 737                *(indegree_slab_at(&indegree, commit)) = 1;
 738                /* also record the author dates, if needed */
 739                if (sort_order == REV_SORT_BY_AUTHOR_DATE)
 740                        record_author_date(&author_date, commit);
 741        }
 742
 743        /* update the indegree */
 744        for (next = orig; next; next = next->next) {
 745                struct commit_list *parents = next->item->parents;
 746                while (parents) {
 747                        struct commit *parent = parents->item;
 748                        int *pi = indegree_slab_at(&indegree, parent);
 749
 750                        if (*pi)
 751                                (*pi)++;
 752                        parents = parents->next;
 753                }
 754        }
 755
 756        /*
 757         * find the tips
 758         *
 759         * tips are nodes not reachable from any other node in the list
 760         *
 761         * the tips serve as a starting set for the work queue.
 762         */
 763        for (next = orig; next; next = next->next) {
 764                struct commit *commit = next->item;
 765
 766                if (*(indegree_slab_at(&indegree, commit)) == 1)
 767                        prio_queue_put(&queue, commit);
 768        }
 769
 770        /*
 771         * This is unfortunate; the initial tips need to be shown
 772         * in the order given from the revision traversal machinery.
 773         */
 774        if (sort_order == REV_SORT_IN_GRAPH_ORDER)
 775                prio_queue_reverse(&queue);
 776
 777        /* We no longer need the commit list */
 778        free_commit_list(orig);
 779
 780        pptr = list;
 781        *list = NULL;
 782        while ((commit = prio_queue_get(&queue)) != NULL) {
 783                struct commit_list *parents;
 784
 785                for (parents = commit->parents; parents ; parents = parents->next) {
 786                        struct commit *parent = parents->item;
 787                        int *pi = indegree_slab_at(&indegree, parent);
 788
 789                        if (!*pi)
 790                                continue;
 791
 792                        /*
 793                         * parents are only enqueued for emission
 794                         * when all their children have been emitted thereby
 795                         * guaranteeing topological order.
 796                         */
 797                        if (--(*pi) == 1)
 798                                prio_queue_put(&queue, parent);
 799                }
 800                /*
 801                 * all children of commit have already been
 802                 * emitted. we can emit it now.
 803                 */
 804                *(indegree_slab_at(&indegree, commit)) = 0;
 805
 806                pptr = &commit_list_insert(commit, pptr)->next;
 807        }
 808
 809        clear_indegree_slab(&indegree);
 810        clear_prio_queue(&queue);
 811        if (sort_order == REV_SORT_BY_AUTHOR_DATE)
 812                clear_author_date_slab(&author_date);
 813}
 814
 815/* merge-base stuff */
 816
 817/* Remember to update object flag allocation in object.h */
 818#define PARENT1         (1u<<16)
 819#define PARENT2         (1u<<17)
 820#define STALE           (1u<<18)
 821#define RESULT          (1u<<19)
 822
 823static const unsigned all_flags = (PARENT1 | PARENT2 | STALE | RESULT);
 824
 825static int queue_has_nonstale(struct prio_queue *queue)
 826{
 827        int i;
 828        for (i = 0; i < queue->nr; i++) {
 829                struct commit *commit = queue->array[i].data;
 830                if (!(commit->object.flags & STALE))
 831                        return 1;
 832        }
 833        return 0;
 834}
 835
 836/* all input commits in one and twos[] must have been parsed! */
 837static struct commit_list *paint_down_to_common(struct commit *one, int n,
 838                                                struct commit **twos,
 839                                                int min_generation)
 840{
 841        struct prio_queue queue = { compare_commits_by_gen_then_commit_date };
 842        struct commit_list *result = NULL;
 843        int i;
 844        uint32_t last_gen = GENERATION_NUMBER_INFINITY;
 845
 846        one->object.flags |= PARENT1;
 847        if (!n) {
 848                commit_list_append(one, &result);
 849                return result;
 850        }
 851        prio_queue_put(&queue, one);
 852
 853        for (i = 0; i < n; i++) {
 854                twos[i]->object.flags |= PARENT2;
 855                prio_queue_put(&queue, twos[i]);
 856        }
 857
 858        while (queue_has_nonstale(&queue)) {
 859                struct commit *commit = prio_queue_get(&queue);
 860                struct commit_list *parents;
 861                int flags;
 862
 863                if (commit->generation > last_gen)
 864                        BUG("bad generation skip %8x > %8x at %s",
 865                            commit->generation, last_gen,
 866                            oid_to_hex(&commit->object.oid));
 867                last_gen = commit->generation;
 868
 869                if (commit->generation < min_generation)
 870                        break;
 871
 872                flags = commit->object.flags & (PARENT1 | PARENT2 | STALE);
 873                if (flags == (PARENT1 | PARENT2)) {
 874                        if (!(commit->object.flags & RESULT)) {
 875                                commit->object.flags |= RESULT;
 876                                commit_list_insert_by_date(commit, &result);
 877                        }
 878                        /* Mark parents of a found merge stale */
 879                        flags |= STALE;
 880                }
 881                parents = commit->parents;
 882                while (parents) {
 883                        struct commit *p = parents->item;
 884                        parents = parents->next;
 885                        if ((p->object.flags & flags) == flags)
 886                                continue;
 887                        if (parse_commit(p))
 888                                return NULL;
 889                        p->object.flags |= flags;
 890                        prio_queue_put(&queue, p);
 891                }
 892        }
 893
 894        clear_prio_queue(&queue);
 895        return result;
 896}
 897
 898static struct commit_list *merge_bases_many(struct commit *one, int n, struct commit **twos)
 899{
 900        struct commit_list *list = NULL;
 901        struct commit_list *result = NULL;
 902        int i;
 903
 904        for (i = 0; i < n; i++) {
 905                if (one == twos[i])
 906                        /*
 907                         * We do not mark this even with RESULT so we do not
 908                         * have to clean it up.
 909                         */
 910                        return commit_list_insert(one, &result);
 911        }
 912
 913        if (parse_commit(one))
 914                return NULL;
 915        for (i = 0; i < n; i++) {
 916                if (parse_commit(twos[i]))
 917                        return NULL;
 918        }
 919
 920        list = paint_down_to_common(one, n, twos, 0);
 921
 922        while (list) {
 923                struct commit *commit = pop_commit(&list);
 924                if (!(commit->object.flags & STALE))
 925                        commit_list_insert_by_date(commit, &result);
 926        }
 927        return result;
 928}
 929
 930struct commit_list *get_octopus_merge_bases(struct commit_list *in)
 931{
 932        struct commit_list *i, *j, *k, *ret = NULL;
 933
 934        if (!in)
 935                return ret;
 936
 937        commit_list_insert(in->item, &ret);
 938
 939        for (i = in->next; i; i = i->next) {
 940                struct commit_list *new_commits = NULL, *end = NULL;
 941
 942                for (j = ret; j; j = j->next) {
 943                        struct commit_list *bases;
 944                        bases = get_merge_bases(i->item, j->item);
 945                        if (!new_commits)
 946                                new_commits = bases;
 947                        else
 948                                end->next = bases;
 949                        for (k = bases; k; k = k->next)
 950                                end = k;
 951                }
 952                ret = new_commits;
 953        }
 954        return ret;
 955}
 956
 957static int remove_redundant(struct commit **array, int cnt)
 958{
 959        /*
 960         * Some commit in the array may be an ancestor of
 961         * another commit.  Move such commit to the end of
 962         * the array, and return the number of commits that
 963         * are independent from each other.
 964         */
 965        struct commit **work;
 966        unsigned char *redundant;
 967        int *filled_index;
 968        int i, j, filled;
 969
 970        work = xcalloc(cnt, sizeof(*work));
 971        redundant = xcalloc(cnt, 1);
 972        ALLOC_ARRAY(filled_index, cnt - 1);
 973
 974        for (i = 0; i < cnt; i++)
 975                parse_commit(array[i]);
 976        for (i = 0; i < cnt; i++) {
 977                struct commit_list *common;
 978                uint32_t min_generation = array[i]->generation;
 979
 980                if (redundant[i])
 981                        continue;
 982                for (j = filled = 0; j < cnt; j++) {
 983                        if (i == j || redundant[j])
 984                                continue;
 985                        filled_index[filled] = j;
 986                        work[filled++] = array[j];
 987
 988                        if (array[j]->generation < min_generation)
 989                                min_generation = array[j]->generation;
 990                }
 991                common = paint_down_to_common(array[i], filled, work,
 992                                              min_generation);
 993                if (array[i]->object.flags & PARENT2)
 994                        redundant[i] = 1;
 995                for (j = 0; j < filled; j++)
 996                        if (work[j]->object.flags & PARENT1)
 997                                redundant[filled_index[j]] = 1;
 998                clear_commit_marks(array[i], all_flags);
 999                clear_commit_marks_many(filled, work, all_flags);
1000                free_commit_list(common);
1001        }
1002
1003        /* Now collect the result */
1004        COPY_ARRAY(work, array, cnt);
1005        for (i = filled = 0; i < cnt; i++)
1006                if (!redundant[i])
1007                        array[filled++] = work[i];
1008        for (j = filled, i = 0; i < cnt; i++)
1009                if (redundant[i])
1010                        array[j++] = work[i];
1011        free(work);
1012        free(redundant);
1013        free(filled_index);
1014        return filled;
1015}
1016
1017static struct commit_list *get_merge_bases_many_0(struct commit *one,
1018                                                  int n,
1019                                                  struct commit **twos,
1020                                                  int cleanup)
1021{
1022        struct commit_list *list;
1023        struct commit **rslt;
1024        struct commit_list *result;
1025        int cnt, i;
1026
1027        result = merge_bases_many(one, n, twos);
1028        for (i = 0; i < n; i++) {
1029                if (one == twos[i])
1030                        return result;
1031        }
1032        if (!result || !result->next) {
1033                if (cleanup) {
1034                        clear_commit_marks(one, all_flags);
1035                        clear_commit_marks_many(n, twos, all_flags);
1036                }
1037                return result;
1038        }
1039
1040        /* There are more than one */
1041        cnt = commit_list_count(result);
1042        rslt = xcalloc(cnt, sizeof(*rslt));
1043        for (list = result, i = 0; list; list = list->next)
1044                rslt[i++] = list->item;
1045        free_commit_list(result);
1046
1047        clear_commit_marks(one, all_flags);
1048        clear_commit_marks_many(n, twos, all_flags);
1049
1050        cnt = remove_redundant(rslt, cnt);
1051        result = NULL;
1052        for (i = 0; i < cnt; i++)
1053                commit_list_insert_by_date(rslt[i], &result);
1054        free(rslt);
1055        return result;
1056}
1057
1058struct commit_list *get_merge_bases_many(struct commit *one,
1059                                         int n,
1060                                         struct commit **twos)
1061{
1062        return get_merge_bases_many_0(one, n, twos, 1);
1063}
1064
1065struct commit_list *get_merge_bases_many_dirty(struct commit *one,
1066                                               int n,
1067                                               struct commit **twos)
1068{
1069        return get_merge_bases_many_0(one, n, twos, 0);
1070}
1071
1072struct commit_list *get_merge_bases(struct commit *one, struct commit *two)
1073{
1074        return get_merge_bases_many_0(one, 1, &two, 1);
1075}
1076
1077/*
1078 * Is "commit" a descendant of one of the elements on the "with_commit" list?
1079 */
1080int is_descendant_of(struct commit *commit, struct commit_list *with_commit)
1081{
1082        if (!with_commit)
1083                return 1;
1084        while (with_commit) {
1085                struct commit *other;
1086
1087                other = with_commit->item;
1088                with_commit = with_commit->next;
1089                if (in_merge_bases(other, commit))
1090                        return 1;
1091        }
1092        return 0;
1093}
1094
1095/*
1096 * Is "commit" an ancestor of one of the "references"?
1097 */
1098int in_merge_bases_many(struct commit *commit, int nr_reference, struct commit **reference)
1099{
1100        struct commit_list *bases;
1101        int ret = 0, i;
1102        uint32_t min_generation = GENERATION_NUMBER_INFINITY;
1103
1104        if (parse_commit(commit))
1105                return ret;
1106        for (i = 0; i < nr_reference; i++) {
1107                if (parse_commit(reference[i]))
1108                        return ret;
1109                if (reference[i]->generation < min_generation)
1110                        min_generation = reference[i]->generation;
1111        }
1112
1113        if (commit->generation > min_generation)
1114                return ret;
1115
1116        bases = paint_down_to_common(commit, nr_reference, reference, commit->generation);
1117        if (commit->object.flags & PARENT2)
1118                ret = 1;
1119        clear_commit_marks(commit, all_flags);
1120        clear_commit_marks_many(nr_reference, reference, all_flags);
1121        free_commit_list(bases);
1122        return ret;
1123}
1124
1125/*
1126 * Is "commit" an ancestor of (i.e. reachable from) the "reference"?
1127 */
1128int in_merge_bases(struct commit *commit, struct commit *reference)
1129{
1130        return in_merge_bases_many(commit, 1, &reference);
1131}
1132
1133struct commit_list *reduce_heads(struct commit_list *heads)
1134{
1135        struct commit_list *p;
1136        struct commit_list *result = NULL, **tail = &result;
1137        struct commit **array;
1138        int num_head, i;
1139
1140        if (!heads)
1141                return NULL;
1142
1143        /* Uniquify */
1144        for (p = heads; p; p = p->next)
1145                p->item->object.flags &= ~STALE;
1146        for (p = heads, num_head = 0; p; p = p->next) {
1147                if (p->item->object.flags & STALE)
1148                        continue;
1149                p->item->object.flags |= STALE;
1150                num_head++;
1151        }
1152        array = xcalloc(num_head, sizeof(*array));
1153        for (p = heads, i = 0; p; p = p->next) {
1154                if (p->item->object.flags & STALE) {
1155                        array[i++] = p->item;
1156                        p->item->object.flags &= ~STALE;
1157                }
1158        }
1159        num_head = remove_redundant(array, num_head);
1160        for (i = 0; i < num_head; i++)
1161                tail = &commit_list_insert(array[i], tail)->next;
1162        free(array);
1163        return result;
1164}
1165
1166void reduce_heads_replace(struct commit_list **heads)
1167{
1168        struct commit_list *result = reduce_heads(*heads);
1169        free_commit_list(*heads);
1170        *heads = result;
1171}
1172
1173static const char gpg_sig_header[] = "gpgsig";
1174static const int gpg_sig_header_len = sizeof(gpg_sig_header) - 1;
1175
1176static int do_sign_commit(struct strbuf *buf, const char *keyid)
1177{
1178        struct strbuf sig = STRBUF_INIT;
1179        int inspos, copypos;
1180        const char *eoh;
1181
1182        /* find the end of the header */
1183        eoh = strstr(buf->buf, "\n\n");
1184        if (!eoh)
1185                inspos = buf->len;
1186        else
1187                inspos = eoh - buf->buf + 1;
1188
1189        if (!keyid || !*keyid)
1190                keyid = get_signing_key();
1191        if (sign_buffer(buf, &sig, keyid)) {
1192                strbuf_release(&sig);
1193                return -1;
1194        }
1195
1196        for (copypos = 0; sig.buf[copypos]; ) {
1197                const char *bol = sig.buf + copypos;
1198                const char *eol = strchrnul(bol, '\n');
1199                int len = (eol - bol) + !!*eol;
1200
1201                if (!copypos) {
1202                        strbuf_insert(buf, inspos, gpg_sig_header, gpg_sig_header_len);
1203                        inspos += gpg_sig_header_len;
1204                }
1205                strbuf_insert(buf, inspos++, " ", 1);
1206                strbuf_insert(buf, inspos, bol, len);
1207                inspos += len;
1208                copypos += len;
1209        }
1210        strbuf_release(&sig);
1211        return 0;
1212}
1213
1214int parse_signed_commit(const struct commit *commit,
1215                        struct strbuf *payload, struct strbuf *signature)
1216{
1217
1218        unsigned long size;
1219        const char *buffer = get_commit_buffer(commit, &size);
1220        int in_signature, saw_signature = -1;
1221        const char *line, *tail;
1222
1223        line = buffer;
1224        tail = buffer + size;
1225        in_signature = 0;
1226        saw_signature = 0;
1227        while (line < tail) {
1228                const char *sig = NULL;
1229                const char *next = memchr(line, '\n', tail - line);
1230
1231                next = next ? next + 1 : tail;
1232                if (in_signature && line[0] == ' ')
1233                        sig = line + 1;
1234                else if (starts_with(line, gpg_sig_header) &&
1235                         line[gpg_sig_header_len] == ' ')
1236                        sig = line + gpg_sig_header_len + 1;
1237                if (sig) {
1238                        strbuf_add(signature, sig, next - sig);
1239                        saw_signature = 1;
1240                        in_signature = 1;
1241                } else {
1242                        if (*line == '\n')
1243                                /* dump the whole remainder of the buffer */
1244                                next = tail;
1245                        strbuf_add(payload, line, next - line);
1246                        in_signature = 0;
1247                }
1248                line = next;
1249        }
1250        unuse_commit_buffer(commit, buffer);
1251        return saw_signature;
1252}
1253
1254int remove_signature(struct strbuf *buf)
1255{
1256        const char *line = buf->buf;
1257        const char *tail = buf->buf + buf->len;
1258        int in_signature = 0;
1259        const char *sig_start = NULL;
1260        const char *sig_end = NULL;
1261
1262        while (line < tail) {
1263                const char *next = memchr(line, '\n', tail - line);
1264                next = next ? next + 1 : tail;
1265
1266                if (in_signature && line[0] == ' ')
1267                        sig_end = next;
1268                else if (starts_with(line, gpg_sig_header) &&
1269                         line[gpg_sig_header_len] == ' ') {
1270                        sig_start = line;
1271                        sig_end = next;
1272                        in_signature = 1;
1273                } else {
1274                        if (*line == '\n')
1275                                /* dump the whole remainder of the buffer */
1276                                next = tail;
1277                        in_signature = 0;
1278                }
1279                line = next;
1280        }
1281
1282        if (sig_start)
1283                strbuf_remove(buf, sig_start - buf->buf, sig_end - sig_start);
1284
1285        return sig_start != NULL;
1286}
1287
1288static void handle_signed_tag(struct commit *parent, struct commit_extra_header ***tail)
1289{
1290        struct merge_remote_desc *desc;
1291        struct commit_extra_header *mergetag;
1292        char *buf;
1293        unsigned long size, len;
1294        enum object_type type;
1295
1296        desc = merge_remote_util(parent);
1297        if (!desc || !desc->obj)
1298                return;
1299        buf = read_object_file(&desc->obj->oid, &type, &size);
1300        if (!buf || type != OBJ_TAG)
1301                goto free_return;
1302        len = parse_signature(buf, size);
1303        if (size == len)
1304                goto free_return;
1305        /*
1306         * We could verify this signature and either omit the tag when
1307         * it does not validate, but the integrator may not have the
1308         * public key of the signer of the tag he is merging, while a
1309         * later auditor may have it while auditing, so let's not run
1310         * verify-signed-buffer here for now...
1311         *
1312         * if (verify_signed_buffer(buf, len, buf + len, size - len, ...))
1313         *      warn("warning: signed tag unverified.");
1314         */
1315        mergetag = xcalloc(1, sizeof(*mergetag));
1316        mergetag->key = xstrdup("mergetag");
1317        mergetag->value = buf;
1318        mergetag->len = size;
1319
1320        **tail = mergetag;
1321        *tail = &mergetag->next;
1322        return;
1323
1324free_return:
1325        free(buf);
1326}
1327
1328int check_commit_signature(const struct commit *commit, struct signature_check *sigc)
1329{
1330        struct strbuf payload = STRBUF_INIT;
1331        struct strbuf signature = STRBUF_INIT;
1332        int ret = 1;
1333
1334        sigc->result = 'N';
1335
1336        if (parse_signed_commit(commit, &payload, &signature) <= 0)
1337                goto out;
1338        ret = check_signature(payload.buf, payload.len, signature.buf,
1339                signature.len, sigc);
1340
1341 out:
1342        strbuf_release(&payload);
1343        strbuf_release(&signature);
1344
1345        return ret;
1346}
1347
1348
1349
1350void append_merge_tag_headers(struct commit_list *parents,
1351                              struct commit_extra_header ***tail)
1352{
1353        while (parents) {
1354                struct commit *parent = parents->item;
1355                handle_signed_tag(parent, tail);
1356                parents = parents->next;
1357        }
1358}
1359
1360static void add_extra_header(struct strbuf *buffer,
1361                             struct commit_extra_header *extra)
1362{
1363        strbuf_addstr(buffer, extra->key);
1364        if (extra->len)
1365                strbuf_add_lines(buffer, " ", extra->value, extra->len);
1366        else
1367                strbuf_addch(buffer, '\n');
1368}
1369
1370struct commit_extra_header *read_commit_extra_headers(struct commit *commit,
1371                                                      const char **exclude)
1372{
1373        struct commit_extra_header *extra = NULL;
1374        unsigned long size;
1375        const char *buffer = get_commit_buffer(commit, &size);
1376        extra = read_commit_extra_header_lines(buffer, size, exclude);
1377        unuse_commit_buffer(commit, buffer);
1378        return extra;
1379}
1380
1381int for_each_mergetag(each_mergetag_fn fn, struct commit *commit, void *data)
1382{
1383        struct commit_extra_header *extra, *to_free;
1384        int res = 0;
1385
1386        to_free = read_commit_extra_headers(commit, NULL);
1387        for (extra = to_free; !res && extra; extra = extra->next) {
1388                if (strcmp(extra->key, "mergetag"))
1389                        continue; /* not a merge tag */
1390                res = fn(commit, extra, data);
1391        }
1392        free_commit_extra_headers(to_free);
1393        return res;
1394}
1395
1396static inline int standard_header_field(const char *field, size_t len)
1397{
1398        return ((len == 4 && !memcmp(field, "tree", 4)) ||
1399                (len == 6 && !memcmp(field, "parent", 6)) ||
1400                (len == 6 && !memcmp(field, "author", 6)) ||
1401                (len == 9 && !memcmp(field, "committer", 9)) ||
1402                (len == 8 && !memcmp(field, "encoding", 8)));
1403}
1404
1405static int excluded_header_field(const char *field, size_t len, const char **exclude)
1406{
1407        if (!exclude)
1408                return 0;
1409
1410        while (*exclude) {
1411                size_t xlen = strlen(*exclude);
1412                if (len == xlen && !memcmp(field, *exclude, xlen))
1413                        return 1;
1414                exclude++;
1415        }
1416        return 0;
1417}
1418
1419static struct commit_extra_header *read_commit_extra_header_lines(
1420        const char *buffer, size_t size,
1421        const char **exclude)
1422{
1423        struct commit_extra_header *extra = NULL, **tail = &extra, *it = NULL;
1424        const char *line, *next, *eof, *eob;
1425        struct strbuf buf = STRBUF_INIT;
1426
1427        for (line = buffer, eob = line + size;
1428             line < eob && *line != '\n';
1429             line = next) {
1430                next = memchr(line, '\n', eob - line);
1431                next = next ? next + 1 : eob;
1432                if (*line == ' ') {
1433                        /* continuation */
1434                        if (it)
1435                                strbuf_add(&buf, line + 1, next - (line + 1));
1436                        continue;
1437                }
1438                if (it)
1439                        it->value = strbuf_detach(&buf, &it->len);
1440                strbuf_reset(&buf);
1441                it = NULL;
1442
1443                eof = memchr(line, ' ', next - line);
1444                if (!eof)
1445                        eof = next;
1446                else if (standard_header_field(line, eof - line) ||
1447                         excluded_header_field(line, eof - line, exclude))
1448                        continue;
1449
1450                it = xcalloc(1, sizeof(*it));
1451                it->key = xmemdupz(line, eof-line);
1452                *tail = it;
1453                tail = &it->next;
1454                if (eof + 1 < next)
1455                        strbuf_add(&buf, eof + 1, next - (eof + 1));
1456        }
1457        if (it)
1458                it->value = strbuf_detach(&buf, &it->len);
1459        return extra;
1460}
1461
1462void free_commit_extra_headers(struct commit_extra_header *extra)
1463{
1464        while (extra) {
1465                struct commit_extra_header *next = extra->next;
1466                free(extra->key);
1467                free(extra->value);
1468                free(extra);
1469                extra = next;
1470        }
1471}
1472
1473int commit_tree(const char *msg, size_t msg_len, const struct object_id *tree,
1474                struct commit_list *parents, struct object_id *ret,
1475                const char *author, const char *sign_commit)
1476{
1477        struct commit_extra_header *extra = NULL, **tail = &extra;
1478        int result;
1479
1480        append_merge_tag_headers(parents, &tail);
1481        result = commit_tree_extended(msg, msg_len, tree, parents, ret,
1482                                      author, sign_commit, extra);
1483        free_commit_extra_headers(extra);
1484        return result;
1485}
1486
1487static int find_invalid_utf8(const char *buf, int len)
1488{
1489        int offset = 0;
1490        static const unsigned int max_codepoint[] = {
1491                0x7f, 0x7ff, 0xffff, 0x10ffff
1492        };
1493
1494        while (len) {
1495                unsigned char c = *buf++;
1496                int bytes, bad_offset;
1497                unsigned int codepoint;
1498                unsigned int min_val, max_val;
1499
1500                len--;
1501                offset++;
1502
1503                /* Simple US-ASCII? No worries. */
1504                if (c < 0x80)
1505                        continue;
1506
1507                bad_offset = offset-1;
1508
1509                /*
1510                 * Count how many more high bits set: that's how
1511                 * many more bytes this sequence should have.
1512                 */
1513                bytes = 0;
1514                while (c & 0x40) {
1515                        c <<= 1;
1516                        bytes++;
1517                }
1518
1519                /*
1520                 * Must be between 1 and 3 more bytes.  Longer sequences result in
1521                 * codepoints beyond U+10FFFF, which are guaranteed never to exist.
1522                 */
1523                if (bytes < 1 || 3 < bytes)
1524                        return bad_offset;
1525
1526                /* Do we *have* that many bytes? */
1527                if (len < bytes)
1528                        return bad_offset;
1529
1530                /*
1531                 * Place the encoded bits at the bottom of the value and compute the
1532                 * valid range.
1533                 */
1534                codepoint = (c & 0x7f) >> bytes;
1535                min_val = max_codepoint[bytes-1] + 1;
1536                max_val = max_codepoint[bytes];
1537
1538                offset += bytes;
1539                len -= bytes;
1540
1541                /* And verify that they are good continuation bytes */
1542                do {
1543                        codepoint <<= 6;
1544                        codepoint |= *buf & 0x3f;
1545                        if ((*buf++ & 0xc0) != 0x80)
1546                                return bad_offset;
1547                } while (--bytes);
1548
1549                /* Reject codepoints that are out of range for the sequence length. */
1550                if (codepoint < min_val || codepoint > max_val)
1551                        return bad_offset;
1552                /* Surrogates are only for UTF-16 and cannot be encoded in UTF-8. */
1553                if ((codepoint & 0x1ff800) == 0xd800)
1554                        return bad_offset;
1555                /* U+xxFFFE and U+xxFFFF are guaranteed non-characters. */
1556                if ((codepoint & 0xfffe) == 0xfffe)
1557                        return bad_offset;
1558                /* So are anything in the range U+FDD0..U+FDEF. */
1559                if (codepoint >= 0xfdd0 && codepoint <= 0xfdef)
1560                        return bad_offset;
1561        }
1562        return -1;
1563}
1564
1565/*
1566 * This verifies that the buffer is in proper utf8 format.
1567 *
1568 * If it isn't, it assumes any non-utf8 characters are Latin1,
1569 * and does the conversion.
1570 */
1571static int verify_utf8(struct strbuf *buf)
1572{
1573        int ok = 1;
1574        long pos = 0;
1575
1576        for (;;) {
1577                int bad;
1578                unsigned char c;
1579                unsigned char replace[2];
1580
1581                bad = find_invalid_utf8(buf->buf + pos, buf->len - pos);
1582                if (bad < 0)
1583                        return ok;
1584                pos += bad;
1585                ok = 0;
1586                c = buf->buf[pos];
1587                strbuf_remove(buf, pos, 1);
1588
1589                /* We know 'c' must be in the range 128-255 */
1590                replace[0] = 0xc0 + (c >> 6);
1591                replace[1] = 0x80 + (c & 0x3f);
1592                strbuf_insert(buf, pos, replace, 2);
1593                pos += 2;
1594        }
1595}
1596
1597static const char commit_utf8_warn[] =
1598N_("Warning: commit message did not conform to UTF-8.\n"
1599   "You may want to amend it after fixing the message, or set the config\n"
1600   "variable i18n.commitencoding to the encoding your project uses.\n");
1601
1602int commit_tree_extended(const char *msg, size_t msg_len,
1603                         const struct object_id *tree,
1604                         struct commit_list *parents, struct object_id *ret,
1605                         const char *author, const char *sign_commit,
1606                         struct commit_extra_header *extra)
1607{
1608        int result;
1609        int encoding_is_utf8;
1610        struct strbuf buffer;
1611
1612        assert_oid_type(tree, OBJ_TREE);
1613
1614        if (memchr(msg, '\0', msg_len))
1615                return error("a NUL byte in commit log message not allowed.");
1616
1617        /* Not having i18n.commitencoding is the same as having utf-8 */
1618        encoding_is_utf8 = is_encoding_utf8(git_commit_encoding);
1619
1620        strbuf_init(&buffer, 8192); /* should avoid reallocs for the headers */
1621        strbuf_addf(&buffer, "tree %s\n", oid_to_hex(tree));
1622
1623        /*
1624         * NOTE! This ordering means that the same exact tree merged with a
1625         * different order of parents will be a _different_ changeset even
1626         * if everything else stays the same.
1627         */
1628        while (parents) {
1629                struct commit *parent = pop_commit(&parents);
1630                strbuf_addf(&buffer, "parent %s\n",
1631                            oid_to_hex(&parent->object.oid));
1632        }
1633
1634        /* Person/date information */
1635        if (!author)
1636                author = git_author_info(IDENT_STRICT);
1637        strbuf_addf(&buffer, "author %s\n", author);
1638        strbuf_addf(&buffer, "committer %s\n", git_committer_info(IDENT_STRICT));
1639        if (!encoding_is_utf8)
1640                strbuf_addf(&buffer, "encoding %s\n", git_commit_encoding);
1641
1642        while (extra) {
1643                add_extra_header(&buffer, extra);
1644                extra = extra->next;
1645        }
1646        strbuf_addch(&buffer, '\n');
1647
1648        /* And add the comment */
1649        strbuf_add(&buffer, msg, msg_len);
1650
1651        /* And check the encoding */
1652        if (encoding_is_utf8 && !verify_utf8(&buffer))
1653                fprintf(stderr, _(commit_utf8_warn));
1654
1655        if (sign_commit && do_sign_commit(&buffer, sign_commit)) {
1656                result = -1;
1657                goto out;
1658        }
1659
1660        result = write_object_file(buffer.buf, buffer.len, commit_type, ret);
1661out:
1662        strbuf_release(&buffer);
1663        return result;
1664}
1665
1666define_commit_slab(merge_desc_slab, struct merge_remote_desc *);
1667static struct merge_desc_slab merge_desc_slab = COMMIT_SLAB_INIT(1, merge_desc_slab);
1668
1669struct merge_remote_desc *merge_remote_util(struct commit *commit)
1670{
1671        return *merge_desc_slab_at(&merge_desc_slab, commit);
1672}
1673
1674void set_merge_remote_desc(struct commit *commit,
1675                           const char *name, struct object *obj)
1676{
1677        struct merge_remote_desc *desc;
1678        FLEX_ALLOC_STR(desc, name, name);
1679        desc->obj = obj;
1680        *merge_desc_slab_at(&merge_desc_slab, commit) = desc;
1681}
1682
1683struct commit *get_merge_parent(const char *name)
1684{
1685        struct object *obj;
1686        struct commit *commit;
1687        struct object_id oid;
1688        if (get_oid(name, &oid))
1689                return NULL;
1690        obj = parse_object(&oid);
1691        commit = (struct commit *)peel_to_type(name, 0, obj, OBJ_COMMIT);
1692        if (commit && !merge_remote_util(commit))
1693                set_merge_remote_desc(commit, name, obj);
1694        return commit;
1695}
1696
1697/*
1698 * Append a commit to the end of the commit_list.
1699 *
1700 * next starts by pointing to the variable that holds the head of an
1701 * empty commit_list, and is updated to point to the "next" field of
1702 * the last item on the list as new commits are appended.
1703 *
1704 * Usage example:
1705 *
1706 *     struct commit_list *list;
1707 *     struct commit_list **next = &list;
1708 *
1709 *     next = commit_list_append(c1, next);
1710 *     next = commit_list_append(c2, next);
1711 *     assert(commit_list_count(list) == 2);
1712 *     return list;
1713 */
1714struct commit_list **commit_list_append(struct commit *commit,
1715                                        struct commit_list **next)
1716{
1717        struct commit_list *new_commit = xmalloc(sizeof(struct commit_list));
1718        new_commit->item = commit;
1719        *next = new_commit;
1720        new_commit->next = NULL;
1721        return &new_commit->next;
1722}
1723
1724const char *find_commit_header(const char *msg, const char *key, size_t *out_len)
1725{
1726        int key_len = strlen(key);
1727        const char *line = msg;
1728
1729        while (line) {
1730                const char *eol = strchrnul(line, '\n');
1731
1732                if (line == eol)
1733                        return NULL;
1734
1735                if (eol - line > key_len &&
1736                    !strncmp(line, key, key_len) &&
1737                    line[key_len] == ' ') {
1738                        *out_len = eol - line - key_len - 1;
1739                        return line + key_len + 1;
1740                }
1741                line = *eol ? eol + 1 : NULL;
1742        }
1743        return NULL;
1744}
1745
1746/*
1747 * Inspect the given string and determine the true "end" of the log message, in
1748 * order to find where to put a new Signed-off-by: line.  Ignored are
1749 * trailing comment lines and blank lines.  To support "git commit -s
1750 * --amend" on an existing commit, we also ignore "Conflicts:".  To
1751 * support "git commit -v", we truncate at cut lines.
1752 *
1753 * Returns the number of bytes from the tail to ignore, to be fed as
1754 * the second parameter to append_signoff().
1755 */
1756int ignore_non_trailer(const char *buf, size_t len)
1757{
1758        int boc = 0;
1759        int bol = 0;
1760        int in_old_conflicts_block = 0;
1761        size_t cutoff = wt_status_locate_end(buf, len);
1762
1763        while (bol < cutoff) {
1764                const char *next_line = memchr(buf + bol, '\n', len - bol);
1765
1766                if (!next_line)
1767                        next_line = buf + len;
1768                else
1769                        next_line++;
1770
1771                if (buf[bol] == comment_line_char || buf[bol] == '\n') {
1772                        /* is this the first of the run of comments? */
1773                        if (!boc)
1774                                boc = bol;
1775                        /* otherwise, it is just continuing */
1776                } else if (starts_with(buf + bol, "Conflicts:\n")) {
1777                        in_old_conflicts_block = 1;
1778                        if (!boc)
1779                                boc = bol;
1780                } else if (in_old_conflicts_block && buf[bol] == '\t') {
1781                        ; /* a pathname in the conflicts block */
1782                } else if (boc) {
1783                        /* the previous was not trailing comment */
1784                        boc = 0;
1785                        in_old_conflicts_block = 0;
1786                }
1787                bol = next_line - buf;
1788        }
1789        return boc ? len - boc : len - cutoff;
1790}