dir.con commit index-pack: smarter memory usage when appending objects (7734d7f)
   1/*
   2 * This handles recursive filename detection with exclude
   3 * files, index knowledge etc..
   4 *
   5 * Copyright (C) Linus Torvalds, 2005-2006
   6 *               Junio Hamano, 2005-2006
   7 */
   8#include "cache.h"
   9#include "dir.h"
  10#include "refs.h"
  11
  12struct path_simplify {
  13        int len;
  14        const char *path;
  15};
  16
  17static int read_directory_recursive(struct dir_struct *dir, const char *path, int len,
  18        int check_only, const struct path_simplify *simplify);
  19static int get_dtype(struct dirent *de, const char *path, int len);
  20
  21static int common_prefix(const char **pathspec)
  22{
  23        const char *path, *slash, *next;
  24        int prefix;
  25
  26        if (!pathspec)
  27                return 0;
  28
  29        path = *pathspec;
  30        slash = strrchr(path, '/');
  31        if (!slash)
  32                return 0;
  33
  34        prefix = slash - path + 1;
  35        while ((next = *++pathspec) != NULL) {
  36                int len = strlen(next);
  37                if (len >= prefix && !memcmp(path, next, prefix))
  38                        continue;
  39                len = prefix - 1;
  40                for (;;) {
  41                        if (!len)
  42                                return 0;
  43                        if (next[--len] != '/')
  44                                continue;
  45                        if (memcmp(path, next, len+1))
  46                                continue;
  47                        prefix = len + 1;
  48                        break;
  49                }
  50        }
  51        return prefix;
  52}
  53
  54int fill_directory(struct dir_struct *dir, const char **pathspec)
  55{
  56        const char *path;
  57        int len;
  58
  59        /*
  60         * Calculate common prefix for the pathspec, and
  61         * use that to optimize the directory walk
  62         */
  63        len = common_prefix(pathspec);
  64        path = "";
  65
  66        if (len)
  67                path = xmemdupz(*pathspec, len);
  68
  69        /* Read the directory and prune it */
  70        read_directory(dir, path, len, pathspec);
  71        return len;
  72}
  73
  74/*
  75 * Does 'match' match the given name?
  76 * A match is found if
  77 *
  78 * (1) the 'match' string is leading directory of 'name', or
  79 * (2) the 'match' string is a wildcard and matches 'name', or
  80 * (3) the 'match' string is exactly the same as 'name'.
  81 *
  82 * and the return value tells which case it was.
  83 *
  84 * It returns 0 when there is no match.
  85 */
  86static int match_one(const char *match, const char *name, int namelen)
  87{
  88        int matchlen;
  89
  90        /* If the match was just the prefix, we matched */
  91        if (!*match)
  92                return MATCHED_RECURSIVELY;
  93
  94        for (;;) {
  95                unsigned char c1 = *match;
  96                unsigned char c2 = *name;
  97                if (c1 == '\0' || is_glob_special(c1))
  98                        break;
  99                if (c1 != c2)
 100                        return 0;
 101                match++;
 102                name++;
 103                namelen--;
 104        }
 105
 106
 107        /*
 108         * If we don't match the matchstring exactly,
 109         * we need to match by fnmatch
 110         */
 111        matchlen = strlen(match);
 112        if (strncmp(match, name, matchlen))
 113                return !fnmatch(match, name, 0) ? MATCHED_FNMATCH : 0;
 114
 115        if (namelen == matchlen)
 116                return MATCHED_EXACTLY;
 117        if (match[matchlen-1] == '/' || name[matchlen] == '/')
 118                return MATCHED_RECURSIVELY;
 119        return 0;
 120}
 121
 122/*
 123 * Given a name and a list of pathspecs, see if the name matches
 124 * any of the pathspecs.  The caller is also interested in seeing
 125 * all pathspec matches some names it calls this function with
 126 * (otherwise the user could have mistyped the unmatched pathspec),
 127 * and a mark is left in seen[] array for pathspec element that
 128 * actually matched anything.
 129 */
 130int match_pathspec(const char **pathspec, const char *name, int namelen,
 131                int prefix, char *seen)
 132{
 133        int i, retval = 0;
 134
 135        if (!pathspec)
 136                return 1;
 137
 138        name += prefix;
 139        namelen -= prefix;
 140
 141        for (i = 0; pathspec[i] != NULL; i++) {
 142                int how;
 143                const char *match = pathspec[i] + prefix;
 144                if (seen && seen[i] == MATCHED_EXACTLY)
 145                        continue;
 146                how = match_one(match, name, namelen);
 147                if (how) {
 148                        if (retval < how)
 149                                retval = how;
 150                        if (seen && seen[i] < how)
 151                                seen[i] = how;
 152                }
 153        }
 154        return retval;
 155}
 156
 157static int no_wildcard(const char *string)
 158{
 159        return string[strcspn(string, "*?[{\\")] == '\0';
 160}
 161
 162void add_exclude(const char *string, const char *base,
 163                 int baselen, struct exclude_list *which)
 164{
 165        struct exclude *x;
 166        size_t len;
 167        int to_exclude = 1;
 168        int flags = 0;
 169
 170        if (*string == '!') {
 171                to_exclude = 0;
 172                string++;
 173        }
 174        len = strlen(string);
 175        if (len && string[len - 1] == '/') {
 176                char *s;
 177                x = xmalloc(sizeof(*x) + len);
 178                s = (char *)(x+1);
 179                memcpy(s, string, len - 1);
 180                s[len - 1] = '\0';
 181                string = s;
 182                x->pattern = s;
 183                flags = EXC_FLAG_MUSTBEDIR;
 184        } else {
 185                x = xmalloc(sizeof(*x));
 186                x->pattern = string;
 187        }
 188        x->to_exclude = to_exclude;
 189        x->patternlen = strlen(string);
 190        x->base = base;
 191        x->baselen = baselen;
 192        x->flags = flags;
 193        if (!strchr(string, '/'))
 194                x->flags |= EXC_FLAG_NODIR;
 195        if (no_wildcard(string))
 196                x->flags |= EXC_FLAG_NOWILDCARD;
 197        if (*string == '*' && no_wildcard(string+1))
 198                x->flags |= EXC_FLAG_ENDSWITH;
 199        ALLOC_GROW(which->excludes, which->nr + 1, which->alloc);
 200        which->excludes[which->nr++] = x;
 201}
 202
 203static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
 204{
 205        int pos, len;
 206        unsigned long sz;
 207        enum object_type type;
 208        void *data;
 209        struct index_state *istate = &the_index;
 210
 211        len = strlen(path);
 212        pos = index_name_pos(istate, path, len);
 213        if (pos < 0)
 214                return NULL;
 215        if (!ce_skip_worktree(istate->cache[pos]))
 216                return NULL;
 217        data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
 218        if (!data || type != OBJ_BLOB) {
 219                free(data);
 220                return NULL;
 221        }
 222        *size = xsize_t(sz);
 223        return data;
 224}
 225
 226int add_excludes_from_file_to_list(const char *fname,
 227                                   const char *base,
 228                                   int baselen,
 229                                   char **buf_p,
 230                                   struct exclude_list *which,
 231                                   int check_index)
 232{
 233        struct stat st;
 234        int fd, i;
 235        size_t size;
 236        char *buf, *entry;
 237
 238        fd = open(fname, O_RDONLY);
 239        if (fd < 0 || fstat(fd, &st) < 0) {
 240                if (0 <= fd)
 241                        close(fd);
 242                if (!check_index ||
 243                    (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
 244                        return -1;
 245                if (size == 0) {
 246                        free(buf);
 247                        return 0;
 248                }
 249                if (buf[size-1] != '\n') {
 250                        buf = xrealloc(buf, size+1);
 251                        buf[size++] = '\n';
 252                }
 253        }
 254        else {
 255                size = xsize_t(st.st_size);
 256                if (size == 0) {
 257                        close(fd);
 258                        return 0;
 259                }
 260                buf = xmalloc(size+1);
 261                if (read_in_full(fd, buf, size) != size) {
 262                        free(buf);
 263                        close(fd);
 264                        return -1;
 265                }
 266                buf[size++] = '\n';
 267                close(fd);
 268        }
 269
 270        if (buf_p)
 271                *buf_p = buf;
 272        entry = buf;
 273        for (i = 0; i < size; i++) {
 274                if (buf[i] == '\n') {
 275                        if (entry != buf + i && entry[0] != '#') {
 276                                buf[i - (i && buf[i-1] == '\r')] = 0;
 277                                add_exclude(entry, base, baselen, which);
 278                        }
 279                        entry = buf + i + 1;
 280                }
 281        }
 282        return 0;
 283}
 284
 285void add_excludes_from_file(struct dir_struct *dir, const char *fname)
 286{
 287        if (add_excludes_from_file_to_list(fname, "", 0, NULL,
 288                                           &dir->exclude_list[EXC_FILE], 0) < 0)
 289                die("cannot use %s as an exclude file", fname);
 290}
 291
 292static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
 293{
 294        struct exclude_list *el;
 295        struct exclude_stack *stk = NULL;
 296        int current;
 297
 298        if ((!dir->exclude_per_dir) ||
 299            (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
 300                return; /* too long a path -- ignore */
 301
 302        /* Pop the ones that are not the prefix of the path being checked. */
 303        el = &dir->exclude_list[EXC_DIRS];
 304        while ((stk = dir->exclude_stack) != NULL) {
 305                if (stk->baselen <= baselen &&
 306                    !strncmp(dir->basebuf, base, stk->baselen))
 307                        break;
 308                dir->exclude_stack = stk->prev;
 309                while (stk->exclude_ix < el->nr)
 310                        free(el->excludes[--el->nr]);
 311                free(stk->filebuf);
 312                free(stk);
 313        }
 314
 315        /* Read from the parent directories and push them down. */
 316        current = stk ? stk->baselen : -1;
 317        while (current < baselen) {
 318                struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
 319                const char *cp;
 320
 321                if (current < 0) {
 322                        cp = base;
 323                        current = 0;
 324                }
 325                else {
 326                        cp = strchr(base + current + 1, '/');
 327                        if (!cp)
 328                                die("oops in prep_exclude");
 329                        cp++;
 330                }
 331                stk->prev = dir->exclude_stack;
 332                stk->baselen = cp - base;
 333                stk->exclude_ix = el->nr;
 334                memcpy(dir->basebuf + current, base + current,
 335                       stk->baselen - current);
 336                strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
 337                add_excludes_from_file_to_list(dir->basebuf,
 338                                               dir->basebuf, stk->baselen,
 339                                               &stk->filebuf, el, 1);
 340                dir->exclude_stack = stk;
 341                current = stk->baselen;
 342        }
 343        dir->basebuf[baselen] = '\0';
 344}
 345
 346/* Scan the list and let the last match determine the fate.
 347 * Return 1 for exclude, 0 for include and -1 for undecided.
 348 */
 349int excluded_from_list(const char *pathname,
 350                       int pathlen, const char *basename, int *dtype,
 351                       struct exclude_list *el)
 352{
 353        int i;
 354
 355        if (el->nr) {
 356                for (i = el->nr - 1; 0 <= i; i--) {
 357                        struct exclude *x = el->excludes[i];
 358                        const char *exclude = x->pattern;
 359                        int to_exclude = x->to_exclude;
 360
 361                        if (x->flags & EXC_FLAG_MUSTBEDIR) {
 362                                if (!dtype) {
 363                                        if (!prefixcmp(pathname, exclude))
 364                                                return to_exclude;
 365                                        else
 366                                                continue;
 367                                }
 368                                if (*dtype == DT_UNKNOWN)
 369                                        *dtype = get_dtype(NULL, pathname, pathlen);
 370                                if (*dtype != DT_DIR)
 371                                        continue;
 372                        }
 373
 374                        if (x->flags & EXC_FLAG_NODIR) {
 375                                /* match basename */
 376                                if (x->flags & EXC_FLAG_NOWILDCARD) {
 377                                        if (!strcmp(exclude, basename))
 378                                                return to_exclude;
 379                                } else if (x->flags & EXC_FLAG_ENDSWITH) {
 380                                        if (x->patternlen - 1 <= pathlen &&
 381                                            !strcmp(exclude + 1, pathname + pathlen - x->patternlen + 1))
 382                                                return to_exclude;
 383                                } else {
 384                                        if (fnmatch(exclude, basename, 0) == 0)
 385                                                return to_exclude;
 386                                }
 387                        }
 388                        else {
 389                                /* match with FNM_PATHNAME:
 390                                 * exclude has base (baselen long) implicitly
 391                                 * in front of it.
 392                                 */
 393                                int baselen = x->baselen;
 394                                if (*exclude == '/')
 395                                        exclude++;
 396
 397                                if (pathlen < baselen ||
 398                                    (baselen && pathname[baselen-1] != '/') ||
 399                                    strncmp(pathname, x->base, baselen))
 400                                    continue;
 401
 402                                if (x->flags & EXC_FLAG_NOWILDCARD) {
 403                                        if (!strcmp(exclude, pathname + baselen))
 404                                                return to_exclude;
 405                                } else {
 406                                        if (fnmatch(exclude, pathname+baselen,
 407                                                    FNM_PATHNAME) == 0)
 408                                            return to_exclude;
 409                                }
 410                        }
 411                }
 412        }
 413        return -1; /* undecided */
 414}
 415
 416int excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
 417{
 418        int pathlen = strlen(pathname);
 419        int st;
 420        const char *basename = strrchr(pathname, '/');
 421        basename = (basename) ? basename+1 : pathname;
 422
 423        prep_exclude(dir, pathname, basename-pathname);
 424        for (st = EXC_CMDL; st <= EXC_FILE; st++) {
 425                switch (excluded_from_list(pathname, pathlen, basename,
 426                                           dtype_p, &dir->exclude_list[st])) {
 427                case 0:
 428                        return 0;
 429                case 1:
 430                        return 1;
 431                }
 432        }
 433        return 0;
 434}
 435
 436static struct dir_entry *dir_entry_new(const char *pathname, int len)
 437{
 438        struct dir_entry *ent;
 439
 440        ent = xmalloc(sizeof(*ent) + len + 1);
 441        ent->len = len;
 442        memcpy(ent->name, pathname, len);
 443        ent->name[len] = 0;
 444        return ent;
 445}
 446
 447static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
 448{
 449        if (cache_name_exists(pathname, len, ignore_case))
 450                return NULL;
 451
 452        ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
 453        return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
 454}
 455
 456static struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
 457{
 458        if (!cache_name_is_other(pathname, len))
 459                return NULL;
 460
 461        ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
 462        return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
 463}
 464
 465enum exist_status {
 466        index_nonexistent = 0,
 467        index_directory,
 468        index_gitdir,
 469};
 470
 471/*
 472 * The index sorts alphabetically by entry name, which
 473 * means that a gitlink sorts as '\0' at the end, while
 474 * a directory (which is defined not as an entry, but as
 475 * the files it contains) will sort with the '/' at the
 476 * end.
 477 */
 478static enum exist_status directory_exists_in_index(const char *dirname, int len)
 479{
 480        int pos = cache_name_pos(dirname, len);
 481        if (pos < 0)
 482                pos = -pos-1;
 483        while (pos < active_nr) {
 484                struct cache_entry *ce = active_cache[pos++];
 485                unsigned char endchar;
 486
 487                if (strncmp(ce->name, dirname, len))
 488                        break;
 489                endchar = ce->name[len];
 490                if (endchar > '/')
 491                        break;
 492                if (endchar == '/')
 493                        return index_directory;
 494                if (!endchar && S_ISGITLINK(ce->ce_mode))
 495                        return index_gitdir;
 496        }
 497        return index_nonexistent;
 498}
 499
 500/*
 501 * When we find a directory when traversing the filesystem, we
 502 * have three distinct cases:
 503 *
 504 *  - ignore it
 505 *  - see it as a directory
 506 *  - recurse into it
 507 *
 508 * and which one we choose depends on a combination of existing
 509 * git index contents and the flags passed into the directory
 510 * traversal routine.
 511 *
 512 * Case 1: If we *already* have entries in the index under that
 513 * directory name, we always recurse into the directory to see
 514 * all the files.
 515 *
 516 * Case 2: If we *already* have that directory name as a gitlink,
 517 * we always continue to see it as a gitlink, regardless of whether
 518 * there is an actual git directory there or not (it might not
 519 * be checked out as a subproject!)
 520 *
 521 * Case 3: if we didn't have it in the index previously, we
 522 * have a few sub-cases:
 523 *
 524 *  (a) if "show_other_directories" is true, we show it as
 525 *      just a directory, unless "hide_empty_directories" is
 526 *      also true and the directory is empty, in which case
 527 *      we just ignore it entirely.
 528 *  (b) if it looks like a git directory, and we don't have
 529 *      'no_gitlinks' set we treat it as a gitlink, and show it
 530 *      as a directory.
 531 *  (c) otherwise, we recurse into it.
 532 */
 533enum directory_treatment {
 534        show_directory,
 535        ignore_directory,
 536        recurse_into_directory,
 537};
 538
 539static enum directory_treatment treat_directory(struct dir_struct *dir,
 540        const char *dirname, int len,
 541        const struct path_simplify *simplify)
 542{
 543        /* The "len-1" is to strip the final '/' */
 544        switch (directory_exists_in_index(dirname, len-1)) {
 545        case index_directory:
 546                return recurse_into_directory;
 547
 548        case index_gitdir:
 549                if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
 550                        return ignore_directory;
 551                return show_directory;
 552
 553        case index_nonexistent:
 554                if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
 555                        break;
 556                if (!(dir->flags & DIR_NO_GITLINKS)) {
 557                        unsigned char sha1[20];
 558                        if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
 559                                return show_directory;
 560                }
 561                return recurse_into_directory;
 562        }
 563
 564        /* This is the "show_other_directories" case */
 565        if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
 566                return show_directory;
 567        if (!read_directory_recursive(dir, dirname, len, 1, simplify))
 568                return ignore_directory;
 569        return show_directory;
 570}
 571
 572/*
 573 * This is an inexact early pruning of any recursive directory
 574 * reading - if the path cannot possibly be in the pathspec,
 575 * return true, and we'll skip it early.
 576 */
 577static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
 578{
 579        if (simplify) {
 580                for (;;) {
 581                        const char *match = simplify->path;
 582                        int len = simplify->len;
 583
 584                        if (!match)
 585                                break;
 586                        if (len > pathlen)
 587                                len = pathlen;
 588                        if (!memcmp(path, match, len))
 589                                return 0;
 590                        simplify++;
 591                }
 592                return 1;
 593        }
 594        return 0;
 595}
 596
 597static int in_pathspec(const char *path, int len, const struct path_simplify *simplify)
 598{
 599        if (simplify) {
 600                for (; simplify->path; simplify++) {
 601                        if (len == simplify->len
 602                            && !memcmp(path, simplify->path, len))
 603                                return 1;
 604                }
 605        }
 606        return 0;
 607}
 608
 609static int get_index_dtype(const char *path, int len)
 610{
 611        int pos;
 612        struct cache_entry *ce;
 613
 614        ce = cache_name_exists(path, len, 0);
 615        if (ce) {
 616                if (!ce_uptodate(ce))
 617                        return DT_UNKNOWN;
 618                if (S_ISGITLINK(ce->ce_mode))
 619                        return DT_DIR;
 620                /*
 621                 * Nobody actually cares about the
 622                 * difference between DT_LNK and DT_REG
 623                 */
 624                return DT_REG;
 625        }
 626
 627        /* Try to look it up as a directory */
 628        pos = cache_name_pos(path, len);
 629        if (pos >= 0)
 630                return DT_UNKNOWN;
 631        pos = -pos-1;
 632        while (pos < active_nr) {
 633                ce = active_cache[pos++];
 634                if (strncmp(ce->name, path, len))
 635                        break;
 636                if (ce->name[len] > '/')
 637                        break;
 638                if (ce->name[len] < '/')
 639                        continue;
 640                if (!ce_uptodate(ce))
 641                        break;  /* continue? */
 642                return DT_DIR;
 643        }
 644        return DT_UNKNOWN;
 645}
 646
 647static int get_dtype(struct dirent *de, const char *path, int len)
 648{
 649        int dtype = de ? DTYPE(de) : DT_UNKNOWN;
 650        struct stat st;
 651
 652        if (dtype != DT_UNKNOWN)
 653                return dtype;
 654        dtype = get_index_dtype(path, len);
 655        if (dtype != DT_UNKNOWN)
 656                return dtype;
 657        if (lstat(path, &st))
 658                return dtype;
 659        if (S_ISREG(st.st_mode))
 660                return DT_REG;
 661        if (S_ISDIR(st.st_mode))
 662                return DT_DIR;
 663        if (S_ISLNK(st.st_mode))
 664                return DT_LNK;
 665        return dtype;
 666}
 667
 668enum path_treatment {
 669        path_ignored,
 670        path_handled,
 671        path_recurse,
 672};
 673
 674static enum path_treatment treat_one_path(struct dir_struct *dir,
 675                                          char *path, int *len,
 676                                          const struct path_simplify *simplify,
 677                                          int dtype, struct dirent *de)
 678{
 679        int exclude = excluded(dir, path, &dtype);
 680        if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
 681            && in_pathspec(path, *len, simplify))
 682                dir_add_ignored(dir, path, *len);
 683
 684        /*
 685         * Excluded? If we don't explicitly want to show
 686         * ignored files, ignore it
 687         */
 688        if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
 689                return path_ignored;
 690
 691        if (dtype == DT_UNKNOWN)
 692                dtype = get_dtype(de, path, *len);
 693
 694        /*
 695         * Do we want to see just the ignored files?
 696         * We still need to recurse into directories,
 697         * even if we don't ignore them, since the
 698         * directory may contain files that we do..
 699         */
 700        if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
 701                if (dtype != DT_DIR)
 702                        return path_ignored;
 703        }
 704
 705        switch (dtype) {
 706        default:
 707                return path_ignored;
 708        case DT_DIR:
 709                memcpy(path + *len, "/", 2);
 710                (*len)++;
 711                switch (treat_directory(dir, path, *len, simplify)) {
 712                case show_directory:
 713                        if (exclude != !!(dir->flags
 714                                          & DIR_SHOW_IGNORED))
 715                                return path_ignored;
 716                        break;
 717                case recurse_into_directory:
 718                        return path_recurse;
 719                case ignore_directory:
 720                        return path_ignored;
 721                }
 722                break;
 723        case DT_REG:
 724        case DT_LNK:
 725                break;
 726        }
 727        return path_handled;
 728}
 729
 730static enum path_treatment treat_path(struct dir_struct *dir,
 731                                      struct dirent *de,
 732                                      char *path, int path_max,
 733                                      int baselen,
 734                                      const struct path_simplify *simplify,
 735                                      int *len)
 736{
 737        int dtype;
 738
 739        if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
 740                return path_ignored;
 741        *len = strlen(de->d_name);
 742        /* Ignore overly long pathnames! */
 743        if (*len + baselen + 8 > path_max)
 744                return path_ignored;
 745        memcpy(path + baselen, de->d_name, *len + 1);
 746        *len += baselen;
 747        if (simplify_away(path, *len, simplify))
 748                return path_ignored;
 749
 750        dtype = DTYPE(de);
 751        return treat_one_path(dir, path, len, simplify, dtype, de);
 752}
 753
 754/*
 755 * Read a directory tree. We currently ignore anything but
 756 * directories, regular files and symlinks. That's because git
 757 * doesn't handle them at all yet. Maybe that will change some
 758 * day.
 759 *
 760 * Also, we ignore the name ".git" (even if it is not a directory).
 761 * That likely will not change.
 762 */
 763static int read_directory_recursive(struct dir_struct *dir,
 764                                    const char *base, int baselen,
 765                                    int check_only,
 766                                    const struct path_simplify *simplify)
 767{
 768        DIR *fdir = opendir(*base ? base : ".");
 769        int contents = 0;
 770
 771        if (fdir) {
 772                struct dirent *de;
 773                char path[PATH_MAX + 1];
 774                memcpy(path, base, baselen);
 775
 776                while ((de = readdir(fdir)) != NULL) {
 777                        int len;
 778                        switch (treat_path(dir, de, path, sizeof(path),
 779                                           baselen, simplify, &len)) {
 780                        case path_recurse:
 781                                contents += read_directory_recursive
 782                                        (dir, path, len, 0, simplify);
 783                                continue;
 784                        case path_ignored:
 785                                continue;
 786                        case path_handled:
 787                                break;
 788                        }
 789                        contents++;
 790                        if (check_only)
 791                                goto exit_early;
 792                        else
 793                                dir_add_name(dir, path, len);
 794                }
 795exit_early:
 796                closedir(fdir);
 797        }
 798
 799        return contents;
 800}
 801
 802static int cmp_name(const void *p1, const void *p2)
 803{
 804        const struct dir_entry *e1 = *(const struct dir_entry **)p1;
 805        const struct dir_entry *e2 = *(const struct dir_entry **)p2;
 806
 807        return cache_name_compare(e1->name, e1->len,
 808                                  e2->name, e2->len);
 809}
 810
 811/*
 812 * Return the length of the "simple" part of a path match limiter.
 813 */
 814static int simple_length(const char *match)
 815{
 816        int len = -1;
 817
 818        for (;;) {
 819                unsigned char c = *match++;
 820                len++;
 821                if (c == '\0' || is_glob_special(c))
 822                        return len;
 823        }
 824}
 825
 826static struct path_simplify *create_simplify(const char **pathspec)
 827{
 828        int nr, alloc = 0;
 829        struct path_simplify *simplify = NULL;
 830
 831        if (!pathspec)
 832                return NULL;
 833
 834        for (nr = 0 ; ; nr++) {
 835                const char *match;
 836                if (nr >= alloc) {
 837                        alloc = alloc_nr(alloc);
 838                        simplify = xrealloc(simplify, alloc * sizeof(*simplify));
 839                }
 840                match = *pathspec++;
 841                if (!match)
 842                        break;
 843                simplify[nr].path = match;
 844                simplify[nr].len = simple_length(match);
 845        }
 846        simplify[nr].path = NULL;
 847        simplify[nr].len = 0;
 848        return simplify;
 849}
 850
 851static void free_simplify(struct path_simplify *simplify)
 852{
 853        free(simplify);
 854}
 855
 856static int treat_leading_path(struct dir_struct *dir,
 857                              const char *path, int len,
 858                              const struct path_simplify *simplify)
 859{
 860        char pathbuf[PATH_MAX];
 861        int baselen, blen;
 862        const char *cp;
 863
 864        while (len && path[len - 1] == '/')
 865                len--;
 866        if (!len)
 867                return 1;
 868        baselen = 0;
 869        while (1) {
 870                cp = path + baselen + !!baselen;
 871                cp = memchr(cp, '/', path + len - cp);
 872                if (!cp)
 873                        baselen = len;
 874                else
 875                        baselen = cp - path;
 876                memcpy(pathbuf, path, baselen);
 877                pathbuf[baselen] = '\0';
 878                if (!is_directory(pathbuf))
 879                        return 0;
 880                if (simplify_away(pathbuf, baselen, simplify))
 881                        return 0;
 882                blen = baselen;
 883                if (treat_one_path(dir, pathbuf, &blen, simplify,
 884                                   DT_DIR, NULL) == path_ignored)
 885                        return 0; /* do not recurse into it */
 886                if (len <= baselen)
 887                        return 1; /* finished checking */
 888        }
 889}
 890
 891int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
 892{
 893        struct path_simplify *simplify;
 894
 895        if (has_symlink_leading_path(path, len))
 896                return dir->nr;
 897
 898        simplify = create_simplify(pathspec);
 899        if (!len || treat_leading_path(dir, path, len, simplify))
 900                read_directory_recursive(dir, path, len, 0, simplify);
 901        free_simplify(simplify);
 902        qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
 903        qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
 904        return dir->nr;
 905}
 906
 907int file_exists(const char *f)
 908{
 909        struct stat sb;
 910        return lstat(f, &sb) == 0;
 911}
 912
 913/*
 914 * get_relative_cwd() gets the prefix of the current working directory
 915 * relative to 'dir'.  If we are not inside 'dir', it returns NULL.
 916 *
 917 * As a convenience, it also returns NULL if 'dir' is already NULL.  The
 918 * reason for this behaviour is that it is natural for functions returning
 919 * directory names to return NULL to say "this directory does not exist"
 920 * or "this directory is invalid".  These cases are usually handled the
 921 * same as if the cwd is not inside 'dir' at all, so get_relative_cwd()
 922 * returns NULL for both of them.
 923 *
 924 * Most notably, get_relative_cwd(buffer, size, get_git_work_tree())
 925 * unifies the handling of "outside work tree" with "no work tree at all".
 926 */
 927char *get_relative_cwd(char *buffer, int size, const char *dir)
 928{
 929        char *cwd = buffer;
 930
 931        if (!dir)
 932                return NULL;
 933        if (!getcwd(buffer, size))
 934                die_errno("can't find the current directory");
 935
 936        if (!is_absolute_path(dir))
 937                dir = make_absolute_path(dir);
 938
 939        while (*dir && *dir == *cwd) {
 940                dir++;
 941                cwd++;
 942        }
 943        if (*dir)
 944                return NULL;
 945        if (*cwd == '/')
 946                return cwd + 1;
 947        return cwd;
 948}
 949
 950int is_inside_dir(const char *dir)
 951{
 952        char buffer[PATH_MAX];
 953        return get_relative_cwd(buffer, sizeof(buffer), dir) != NULL;
 954}
 955
 956int is_empty_dir(const char *path)
 957{
 958        DIR *dir = opendir(path);
 959        struct dirent *e;
 960        int ret = 1;
 961
 962        if (!dir)
 963                return 0;
 964
 965        while ((e = readdir(dir)) != NULL)
 966                if (!is_dot_or_dotdot(e->d_name)) {
 967                        ret = 0;
 968                        break;
 969                }
 970
 971        closedir(dir);
 972        return ret;
 973}
 974
 975int remove_dir_recursively(struct strbuf *path, int flag)
 976{
 977        DIR *dir;
 978        struct dirent *e;
 979        int ret = 0, original_len = path->len, len;
 980        int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
 981        unsigned char submodule_head[20];
 982
 983        if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
 984            !resolve_gitlink_ref(path->buf, "HEAD", submodule_head))
 985                /* Do not descend and nuke a nested git work tree. */
 986                return 0;
 987
 988        dir = opendir(path->buf);
 989        if (!dir)
 990                return -1;
 991        if (path->buf[original_len - 1] != '/')
 992                strbuf_addch(path, '/');
 993
 994        len = path->len;
 995        while ((e = readdir(dir)) != NULL) {
 996                struct stat st;
 997                if (is_dot_or_dotdot(e->d_name))
 998                        continue;
 999
1000                strbuf_setlen(path, len);
1001                strbuf_addstr(path, e->d_name);
1002                if (lstat(path->buf, &st))
1003                        ; /* fall thru */
1004                else if (S_ISDIR(st.st_mode)) {
1005                        if (!remove_dir_recursively(path, only_empty))
1006                                continue; /* happy */
1007                } else if (!only_empty && !unlink(path->buf))
1008                        continue; /* happy, too */
1009
1010                /* path too long, stat fails, or non-directory still exists */
1011                ret = -1;
1012                break;
1013        }
1014        closedir(dir);
1015
1016        strbuf_setlen(path, original_len);
1017        if (!ret)
1018                ret = rmdir(path->buf);
1019        return ret;
1020}
1021
1022void setup_standard_excludes(struct dir_struct *dir)
1023{
1024        const char *path;
1025
1026        dir->exclude_per_dir = ".gitignore";
1027        path = git_path("info/exclude");
1028        if (!access(path, R_OK))
1029                add_excludes_from_file(dir, path);
1030        if (excludes_file && !access(excludes_file, R_OK))
1031                add_excludes_from_file(dir, excludes_file);
1032}
1033
1034int remove_path(const char *name)
1035{
1036        char *slash;
1037
1038        if (unlink(name) && errno != ENOENT)
1039                return -1;
1040
1041        slash = strrchr(name, '/');
1042        if (slash) {
1043                char *dirs = xstrdup(name);
1044                slash = dirs + (slash - name);
1045                do {
1046                        *slash = '\0';
1047                } while (rmdir(dirs) && (slash = strrchr(dirs, '/')));
1048                free(dirs);
1049        }
1050        return 0;
1051}
1052