dir.con commit Merge branch 'rr/test-make-sure-we-have-git' into maint (5178ee1)
   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
  21/* helper string functions with support for the ignore_case flag */
  22int strcmp_icase(const char *a, const char *b)
  23{
  24        return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
  25}
  26
  27int strncmp_icase(const char *a, const char *b, size_t count)
  28{
  29        return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
  30}
  31
  32int fnmatch_icase(const char *pattern, const char *string, int flags)
  33{
  34        return fnmatch(pattern, string, flags | (ignore_case ? FNM_CASEFOLD : 0));
  35}
  36
  37static size_t common_prefix_len(const char **pathspec)
  38{
  39        const char *n, *first;
  40        size_t max = 0;
  41
  42        if (!pathspec)
  43                return max;
  44
  45        first = *pathspec;
  46        while ((n = *pathspec++)) {
  47                size_t i, len = 0;
  48                for (i = 0; first == n || i < max; i++) {
  49                        char c = n[i];
  50                        if (!c || c != first[i] || is_glob_special(c))
  51                                break;
  52                        if (c == '/')
  53                                len = i + 1;
  54                }
  55                if (first == n || len < max) {
  56                        max = len;
  57                        if (!max)
  58                                break;
  59                }
  60        }
  61        return max;
  62}
  63
  64/*
  65 * Returns a copy of the longest leading path common among all
  66 * pathspecs.
  67 */
  68char *common_prefix(const char **pathspec)
  69{
  70        unsigned long len = common_prefix_len(pathspec);
  71
  72        return len ? xmemdupz(*pathspec, len) : NULL;
  73}
  74
  75int fill_directory(struct dir_struct *dir, const char **pathspec)
  76{
  77        size_t len;
  78
  79        /*
  80         * Calculate common prefix for the pathspec, and
  81         * use that to optimize the directory walk
  82         */
  83        len = common_prefix_len(pathspec);
  84
  85        /* Read the directory and prune it */
  86        read_directory(dir, pathspec ? *pathspec : "", len, pathspec);
  87        return len;
  88}
  89
  90int within_depth(const char *name, int namelen,
  91                        int depth, int max_depth)
  92{
  93        const char *cp = name, *cpe = name + namelen;
  94
  95        while (cp < cpe) {
  96                if (*cp++ != '/')
  97                        continue;
  98                depth++;
  99                if (depth > max_depth)
 100                        return 0;
 101        }
 102        return 1;
 103}
 104
 105/*
 106 * Does 'match' match the given name?
 107 * A match is found if
 108 *
 109 * (1) the 'match' string is leading directory of 'name', or
 110 * (2) the 'match' string is a wildcard and matches 'name', or
 111 * (3) the 'match' string is exactly the same as 'name'.
 112 *
 113 * and the return value tells which case it was.
 114 *
 115 * It returns 0 when there is no match.
 116 */
 117static int match_one(const char *match, const char *name, int namelen)
 118{
 119        int matchlen;
 120
 121        /* If the match was just the prefix, we matched */
 122        if (!*match)
 123                return MATCHED_RECURSIVELY;
 124
 125        if (ignore_case) {
 126                for (;;) {
 127                        unsigned char c1 = tolower(*match);
 128                        unsigned char c2 = tolower(*name);
 129                        if (c1 == '\0' || is_glob_special(c1))
 130                                break;
 131                        if (c1 != c2)
 132                                return 0;
 133                        match++;
 134                        name++;
 135                        namelen--;
 136                }
 137        } else {
 138                for (;;) {
 139                        unsigned char c1 = *match;
 140                        unsigned char c2 = *name;
 141                        if (c1 == '\0' || is_glob_special(c1))
 142                                break;
 143                        if (c1 != c2)
 144                                return 0;
 145                        match++;
 146                        name++;
 147                        namelen--;
 148                }
 149        }
 150
 151
 152        /*
 153         * If we don't match the matchstring exactly,
 154         * we need to match by fnmatch
 155         */
 156        matchlen = strlen(match);
 157        if (strncmp_icase(match, name, matchlen))
 158                return !fnmatch_icase(match, name, 0) ? MATCHED_FNMATCH : 0;
 159
 160        if (namelen == matchlen)
 161                return MATCHED_EXACTLY;
 162        if (match[matchlen-1] == '/' || name[matchlen] == '/')
 163                return MATCHED_RECURSIVELY;
 164        return 0;
 165}
 166
 167/*
 168 * Given a name and a list of pathspecs, see if the name matches
 169 * any of the pathspecs.  The caller is also interested in seeing
 170 * all pathspec matches some names it calls this function with
 171 * (otherwise the user could have mistyped the unmatched pathspec),
 172 * and a mark is left in seen[] array for pathspec element that
 173 * actually matched anything.
 174 */
 175int match_pathspec(const char **pathspec, const char *name, int namelen,
 176                int prefix, char *seen)
 177{
 178        int i, retval = 0;
 179
 180        if (!pathspec)
 181                return 1;
 182
 183        name += prefix;
 184        namelen -= prefix;
 185
 186        for (i = 0; pathspec[i] != NULL; i++) {
 187                int how;
 188                const char *match = pathspec[i] + prefix;
 189                if (seen && seen[i] == MATCHED_EXACTLY)
 190                        continue;
 191                how = match_one(match, name, namelen);
 192                if (how) {
 193                        if (retval < how)
 194                                retval = how;
 195                        if (seen && seen[i] < how)
 196                                seen[i] = how;
 197                }
 198        }
 199        return retval;
 200}
 201
 202/*
 203 * Does 'match' match the given name?
 204 * A match is found if
 205 *
 206 * (1) the 'match' string is leading directory of 'name', or
 207 * (2) the 'match' string is a wildcard and matches 'name', or
 208 * (3) the 'match' string is exactly the same as 'name'.
 209 *
 210 * and the return value tells which case it was.
 211 *
 212 * It returns 0 when there is no match.
 213 */
 214static int match_pathspec_item(const struct pathspec_item *item, int prefix,
 215                               const char *name, int namelen)
 216{
 217        /* name/namelen has prefix cut off by caller */
 218        const char *match = item->match + prefix;
 219        int matchlen = item->len - prefix;
 220
 221        /* If the match was just the prefix, we matched */
 222        if (!*match)
 223                return MATCHED_RECURSIVELY;
 224
 225        if (matchlen <= namelen && !strncmp(match, name, matchlen)) {
 226                if (matchlen == namelen)
 227                        return MATCHED_EXACTLY;
 228
 229                if (match[matchlen-1] == '/' || name[matchlen] == '/')
 230                        return MATCHED_RECURSIVELY;
 231        }
 232
 233        if (item->use_wildcard && !fnmatch(match, name, 0))
 234                return MATCHED_FNMATCH;
 235
 236        return 0;
 237}
 238
 239/*
 240 * Given a name and a list of pathspecs, see if the name matches
 241 * any of the pathspecs.  The caller is also interested in seeing
 242 * all pathspec matches some names it calls this function with
 243 * (otherwise the user could have mistyped the unmatched pathspec),
 244 * and a mark is left in seen[] array for pathspec element that
 245 * actually matched anything.
 246 */
 247int match_pathspec_depth(const struct pathspec *ps,
 248                         const char *name, int namelen,
 249                         int prefix, char *seen)
 250{
 251        int i, retval = 0;
 252
 253        if (!ps->nr) {
 254                if (!ps->recursive || ps->max_depth == -1)
 255                        return MATCHED_RECURSIVELY;
 256
 257                if (within_depth(name, namelen, 0, ps->max_depth))
 258                        return MATCHED_EXACTLY;
 259                else
 260                        return 0;
 261        }
 262
 263        name += prefix;
 264        namelen -= prefix;
 265
 266        for (i = ps->nr - 1; i >= 0; i--) {
 267                int how;
 268                if (seen && seen[i] == MATCHED_EXACTLY)
 269                        continue;
 270                how = match_pathspec_item(ps->items+i, prefix, name, namelen);
 271                if (ps->recursive && ps->max_depth != -1 &&
 272                    how && how != MATCHED_FNMATCH) {
 273                        int len = ps->items[i].len;
 274                        if (name[len] == '/')
 275                                len++;
 276                        if (within_depth(name+len, namelen-len, 0, ps->max_depth))
 277                                how = MATCHED_EXACTLY;
 278                        else
 279                                how = 0;
 280                }
 281                if (how) {
 282                        if (retval < how)
 283                                retval = how;
 284                        if (seen && seen[i] < how)
 285                                seen[i] = how;
 286                }
 287        }
 288        return retval;
 289}
 290
 291/*
 292 * Return the length of the "simple" part of a path match limiter.
 293 */
 294static int simple_length(const char *match)
 295{
 296        int len = -1;
 297
 298        for (;;) {
 299                unsigned char c = *match++;
 300                len++;
 301                if (c == '\0' || is_glob_special(c))
 302                        return len;
 303        }
 304}
 305
 306static int no_wildcard(const char *string)
 307{
 308        return string[simple_length(string)] == '\0';
 309}
 310
 311void add_exclude(const char *string, const char *base,
 312                 int baselen, struct exclude_list *which)
 313{
 314        struct exclude *x;
 315        size_t len;
 316        int to_exclude = 1;
 317        int flags = 0;
 318
 319        if (*string == '!') {
 320                to_exclude = 0;
 321                string++;
 322        }
 323        len = strlen(string);
 324        if (len && string[len - 1] == '/') {
 325                char *s;
 326                x = xmalloc(sizeof(*x) + len);
 327                s = (char *)(x+1);
 328                memcpy(s, string, len - 1);
 329                s[len - 1] = '\0';
 330                string = s;
 331                x->pattern = s;
 332                flags = EXC_FLAG_MUSTBEDIR;
 333        } else {
 334                x = xmalloc(sizeof(*x));
 335                x->pattern = string;
 336        }
 337        x->to_exclude = to_exclude;
 338        x->patternlen = strlen(string);
 339        x->base = base;
 340        x->baselen = baselen;
 341        x->flags = flags;
 342        if (!strchr(string, '/'))
 343                x->flags |= EXC_FLAG_NODIR;
 344        x->nowildcardlen = simple_length(string);
 345        if (*string == '*' && no_wildcard(string+1))
 346                x->flags |= EXC_FLAG_ENDSWITH;
 347        ALLOC_GROW(which->excludes, which->nr + 1, which->alloc);
 348        which->excludes[which->nr++] = x;
 349}
 350
 351static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
 352{
 353        int pos, len;
 354        unsigned long sz;
 355        enum object_type type;
 356        void *data;
 357        struct index_state *istate = &the_index;
 358
 359        len = strlen(path);
 360        pos = index_name_pos(istate, path, len);
 361        if (pos < 0)
 362                return NULL;
 363        if (!ce_skip_worktree(istate->cache[pos]))
 364                return NULL;
 365        data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
 366        if (!data || type != OBJ_BLOB) {
 367                free(data);
 368                return NULL;
 369        }
 370        *size = xsize_t(sz);
 371        return data;
 372}
 373
 374void free_excludes(struct exclude_list *el)
 375{
 376        int i;
 377
 378        for (i = 0; i < el->nr; i++)
 379                free(el->excludes[i]);
 380        free(el->excludes);
 381
 382        el->nr = 0;
 383        el->excludes = NULL;
 384}
 385
 386int add_excludes_from_file_to_list(const char *fname,
 387                                   const char *base,
 388                                   int baselen,
 389                                   char **buf_p,
 390                                   struct exclude_list *which,
 391                                   int check_index)
 392{
 393        struct stat st;
 394        int fd, i;
 395        size_t size = 0;
 396        char *buf, *entry;
 397
 398        fd = open(fname, O_RDONLY);
 399        if (fd < 0 || fstat(fd, &st) < 0) {
 400                if (errno != ENOENT)
 401                        warn_on_inaccessible(fname);
 402                if (0 <= fd)
 403                        close(fd);
 404                if (!check_index ||
 405                    (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
 406                        return -1;
 407                if (size == 0) {
 408                        free(buf);
 409                        return 0;
 410                }
 411                if (buf[size-1] != '\n') {
 412                        buf = xrealloc(buf, size+1);
 413                        buf[size++] = '\n';
 414                }
 415        }
 416        else {
 417                size = xsize_t(st.st_size);
 418                if (size == 0) {
 419                        close(fd);
 420                        return 0;
 421                }
 422                buf = xmalloc(size+1);
 423                if (read_in_full(fd, buf, size) != size) {
 424                        free(buf);
 425                        close(fd);
 426                        return -1;
 427                }
 428                buf[size++] = '\n';
 429                close(fd);
 430        }
 431
 432        if (buf_p)
 433                *buf_p = buf;
 434        entry = buf;
 435        for (i = 0; i < size; i++) {
 436                if (buf[i] == '\n') {
 437                        if (entry != buf + i && entry[0] != '#') {
 438                                buf[i - (i && buf[i-1] == '\r')] = 0;
 439                                add_exclude(entry, base, baselen, which);
 440                        }
 441                        entry = buf + i + 1;
 442                }
 443        }
 444        return 0;
 445}
 446
 447void add_excludes_from_file(struct dir_struct *dir, const char *fname)
 448{
 449        if (add_excludes_from_file_to_list(fname, "", 0, NULL,
 450                                           &dir->exclude_list[EXC_FILE], 0) < 0)
 451                die("cannot use %s as an exclude file", fname);
 452}
 453
 454static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
 455{
 456        struct exclude_list *el;
 457        struct exclude_stack *stk = NULL;
 458        int current;
 459
 460        if ((!dir->exclude_per_dir) ||
 461            (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
 462                return; /* too long a path -- ignore */
 463
 464        /* Pop the ones that are not the prefix of the path being checked. */
 465        el = &dir->exclude_list[EXC_DIRS];
 466        while ((stk = dir->exclude_stack) != NULL) {
 467                if (stk->baselen <= baselen &&
 468                    !strncmp(dir->basebuf, base, stk->baselen))
 469                        break;
 470                dir->exclude_stack = stk->prev;
 471                while (stk->exclude_ix < el->nr)
 472                        free(el->excludes[--el->nr]);
 473                free(stk->filebuf);
 474                free(stk);
 475        }
 476
 477        /* Read from the parent directories and push them down. */
 478        current = stk ? stk->baselen : -1;
 479        while (current < baselen) {
 480                struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
 481                const char *cp;
 482
 483                if (current < 0) {
 484                        cp = base;
 485                        current = 0;
 486                }
 487                else {
 488                        cp = strchr(base + current + 1, '/');
 489                        if (!cp)
 490                                die("oops in prep_exclude");
 491                        cp++;
 492                }
 493                stk->prev = dir->exclude_stack;
 494                stk->baselen = cp - base;
 495                stk->exclude_ix = el->nr;
 496                memcpy(dir->basebuf + current, base + current,
 497                       stk->baselen - current);
 498                strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
 499                add_excludes_from_file_to_list(dir->basebuf,
 500                                               dir->basebuf, stk->baselen,
 501                                               &stk->filebuf, el, 1);
 502                dir->exclude_stack = stk;
 503                current = stk->baselen;
 504        }
 505        dir->basebuf[baselen] = '\0';
 506}
 507
 508/* Scan the list and let the last match determine the fate.
 509 * Return 1 for exclude, 0 for include and -1 for undecided.
 510 */
 511int excluded_from_list(const char *pathname,
 512                       int pathlen, const char *basename, int *dtype,
 513                       struct exclude_list *el)
 514{
 515        int i;
 516
 517        if (!el->nr)
 518                return -1;      /* undefined */
 519
 520        for (i = el->nr - 1; 0 <= i; i--) {
 521                struct exclude *x = el->excludes[i];
 522                const char *name, *exclude = x->pattern;
 523                int to_exclude = x->to_exclude;
 524                int namelen, prefix = x->nowildcardlen;
 525
 526                if (x->flags & EXC_FLAG_MUSTBEDIR) {
 527                        if (*dtype == DT_UNKNOWN)
 528                                *dtype = get_dtype(NULL, pathname, pathlen);
 529                        if (*dtype != DT_DIR)
 530                                continue;
 531                }
 532
 533                if (x->flags & EXC_FLAG_NODIR) {
 534                        /* match basename */
 535                        if (prefix == x->patternlen) {
 536                                if (!strcmp_icase(exclude, basename))
 537                                        return to_exclude;
 538                        } else if (x->flags & EXC_FLAG_ENDSWITH) {
 539                                if (x->patternlen - 1 <= pathlen &&
 540                                    !strcmp_icase(exclude + 1, pathname + pathlen - x->patternlen + 1))
 541                                        return to_exclude;
 542                        } else {
 543                                if (fnmatch_icase(exclude, basename, 0) == 0)
 544                                        return to_exclude;
 545                        }
 546                        continue;
 547                }
 548
 549                /* match with FNM_PATHNAME:
 550                 * exclude has base (baselen long) implicitly in front of it.
 551                 */
 552                if (*exclude == '/') {
 553                        exclude++;
 554                        prefix--;
 555                }
 556
 557                if (pathlen < x->baselen ||
 558                    (x->baselen && pathname[x->baselen-1] != '/') ||
 559                    strncmp_icase(pathname, x->base, x->baselen))
 560                        continue;
 561
 562                namelen = x->baselen ? pathlen - x->baselen : pathlen;
 563                name = pathname + pathlen  - namelen;
 564
 565                /* if the non-wildcard part is longer than the
 566                   remaining pathname, surely it cannot match */
 567                if (prefix > namelen)
 568                        continue;
 569
 570                if (prefix) {
 571                        if (strncmp_icase(exclude, name, prefix))
 572                                continue;
 573                        exclude += prefix;
 574                        name    += prefix;
 575                        namelen -= prefix;
 576                }
 577
 578                if (!namelen || !fnmatch_icase(exclude, name, FNM_PATHNAME))
 579                        return to_exclude;
 580        }
 581        return -1; /* undecided */
 582}
 583
 584static int excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
 585{
 586        int pathlen = strlen(pathname);
 587        int st;
 588        const char *basename = strrchr(pathname, '/');
 589        basename = (basename) ? basename+1 : pathname;
 590
 591        prep_exclude(dir, pathname, basename-pathname);
 592        for (st = EXC_CMDL; st <= EXC_FILE; st++) {
 593                switch (excluded_from_list(pathname, pathlen, basename,
 594                                           dtype_p, &dir->exclude_list[st])) {
 595                case 0:
 596                        return 0;
 597                case 1:
 598                        return 1;
 599                }
 600        }
 601        return 0;
 602}
 603
 604void path_exclude_check_init(struct path_exclude_check *check,
 605                             struct dir_struct *dir)
 606{
 607        check->dir = dir;
 608        strbuf_init(&check->path, 256);
 609}
 610
 611void path_exclude_check_clear(struct path_exclude_check *check)
 612{
 613        strbuf_release(&check->path);
 614}
 615
 616/*
 617 * Is this name excluded?  This is for a caller like show_files() that
 618 * do not honor directory hierarchy and iterate through paths that are
 619 * possibly in an ignored directory.
 620 *
 621 * A path to a directory known to be excluded is left in check->path to
 622 * optimize for repeated checks for files in the same excluded directory.
 623 */
 624int path_excluded(struct path_exclude_check *check,
 625                  const char *name, int namelen, int *dtype)
 626{
 627        int i;
 628        struct strbuf *path = &check->path;
 629
 630        /*
 631         * we allow the caller to pass namelen as an optimization; it
 632         * must match the length of the name, as we eventually call
 633         * excluded() on the whole name string.
 634         */
 635        if (namelen < 0)
 636                namelen = strlen(name);
 637
 638        if (path->len &&
 639            path->len <= namelen &&
 640            !memcmp(name, path->buf, path->len) &&
 641            (!name[path->len] || name[path->len] == '/'))
 642                return 1;
 643
 644        strbuf_setlen(path, 0);
 645        for (i = 0; name[i]; i++) {
 646                int ch = name[i];
 647
 648                if (ch == '/') {
 649                        int dt = DT_DIR;
 650                        if (excluded(check->dir, path->buf, &dt))
 651                                return 1;
 652                }
 653                strbuf_addch(path, ch);
 654        }
 655
 656        /* An entry in the index; cannot be a directory with subentries */
 657        strbuf_setlen(path, 0);
 658
 659        return excluded(check->dir, name, dtype);
 660}
 661
 662static struct dir_entry *dir_entry_new(const char *pathname, int len)
 663{
 664        struct dir_entry *ent;
 665
 666        ent = xmalloc(sizeof(*ent) + len + 1);
 667        ent->len = len;
 668        memcpy(ent->name, pathname, len);
 669        ent->name[len] = 0;
 670        return ent;
 671}
 672
 673static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
 674{
 675        if (cache_name_exists(pathname, len, ignore_case))
 676                return NULL;
 677
 678        ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
 679        return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
 680}
 681
 682struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
 683{
 684        if (!cache_name_is_other(pathname, len))
 685                return NULL;
 686
 687        ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
 688        return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
 689}
 690
 691enum exist_status {
 692        index_nonexistent = 0,
 693        index_directory,
 694        index_gitdir
 695};
 696
 697/*
 698 * Do not use the alphabetically stored index to look up
 699 * the directory name; instead, use the case insensitive
 700 * name hash.
 701 */
 702static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
 703{
 704        struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
 705        unsigned char endchar;
 706
 707        if (!ce)
 708                return index_nonexistent;
 709        endchar = ce->name[len];
 710
 711        /*
 712         * The cache_entry structure returned will contain this dirname
 713         * and possibly additional path components.
 714         */
 715        if (endchar == '/')
 716                return index_directory;
 717
 718        /*
 719         * If there are no additional path components, then this cache_entry
 720         * represents a submodule.  Submodules, despite being directories,
 721         * are stored in the cache without a closing slash.
 722         */
 723        if (!endchar && S_ISGITLINK(ce->ce_mode))
 724                return index_gitdir;
 725
 726        /* This should never be hit, but it exists just in case. */
 727        return index_nonexistent;
 728}
 729
 730/*
 731 * The index sorts alphabetically by entry name, which
 732 * means that a gitlink sorts as '\0' at the end, while
 733 * a directory (which is defined not as an entry, but as
 734 * the files it contains) will sort with the '/' at the
 735 * end.
 736 */
 737static enum exist_status directory_exists_in_index(const char *dirname, int len)
 738{
 739        int pos;
 740
 741        if (ignore_case)
 742                return directory_exists_in_index_icase(dirname, len);
 743
 744        pos = cache_name_pos(dirname, len);
 745        if (pos < 0)
 746                pos = -pos-1;
 747        while (pos < active_nr) {
 748                struct cache_entry *ce = active_cache[pos++];
 749                unsigned char endchar;
 750
 751                if (strncmp(ce->name, dirname, len))
 752                        break;
 753                endchar = ce->name[len];
 754                if (endchar > '/')
 755                        break;
 756                if (endchar == '/')
 757                        return index_directory;
 758                if (!endchar && S_ISGITLINK(ce->ce_mode))
 759                        return index_gitdir;
 760        }
 761        return index_nonexistent;
 762}
 763
 764/*
 765 * When we find a directory when traversing the filesystem, we
 766 * have three distinct cases:
 767 *
 768 *  - ignore it
 769 *  - see it as a directory
 770 *  - recurse into it
 771 *
 772 * and which one we choose depends on a combination of existing
 773 * git index contents and the flags passed into the directory
 774 * traversal routine.
 775 *
 776 * Case 1: If we *already* have entries in the index under that
 777 * directory name, we always recurse into the directory to see
 778 * all the files.
 779 *
 780 * Case 2: If we *already* have that directory name as a gitlink,
 781 * we always continue to see it as a gitlink, regardless of whether
 782 * there is an actual git directory there or not (it might not
 783 * be checked out as a subproject!)
 784 *
 785 * Case 3: if we didn't have it in the index previously, we
 786 * have a few sub-cases:
 787 *
 788 *  (a) if "show_other_directories" is true, we show it as
 789 *      just a directory, unless "hide_empty_directories" is
 790 *      also true and the directory is empty, in which case
 791 *      we just ignore it entirely.
 792 *  (b) if it looks like a git directory, and we don't have
 793 *      'no_gitlinks' set we treat it as a gitlink, and show it
 794 *      as a directory.
 795 *  (c) otherwise, we recurse into it.
 796 */
 797enum directory_treatment {
 798        show_directory,
 799        ignore_directory,
 800        recurse_into_directory
 801};
 802
 803static enum directory_treatment treat_directory(struct dir_struct *dir,
 804        const char *dirname, int len,
 805        const struct path_simplify *simplify)
 806{
 807        /* The "len-1" is to strip the final '/' */
 808        switch (directory_exists_in_index(dirname, len-1)) {
 809        case index_directory:
 810                return recurse_into_directory;
 811
 812        case index_gitdir:
 813                if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
 814                        return ignore_directory;
 815                return show_directory;
 816
 817        case index_nonexistent:
 818                if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
 819                        break;
 820                if (!(dir->flags & DIR_NO_GITLINKS)) {
 821                        unsigned char sha1[20];
 822                        if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
 823                                return show_directory;
 824                }
 825                return recurse_into_directory;
 826        }
 827
 828        /* This is the "show_other_directories" case */
 829        if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
 830                return show_directory;
 831        if (!read_directory_recursive(dir, dirname, len, 1, simplify))
 832                return ignore_directory;
 833        return show_directory;
 834}
 835
 836/*
 837 * This is an inexact early pruning of any recursive directory
 838 * reading - if the path cannot possibly be in the pathspec,
 839 * return true, and we'll skip it early.
 840 */
 841static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
 842{
 843        if (simplify) {
 844                for (;;) {
 845                        const char *match = simplify->path;
 846                        int len = simplify->len;
 847
 848                        if (!match)
 849                                break;
 850                        if (len > pathlen)
 851                                len = pathlen;
 852                        if (!memcmp(path, match, len))
 853                                return 0;
 854                        simplify++;
 855                }
 856                return 1;
 857        }
 858        return 0;
 859}
 860
 861/*
 862 * This function tells us whether an excluded path matches a
 863 * list of "interesting" pathspecs. That is, whether a path matched
 864 * by any of the pathspecs could possibly be ignored by excluding
 865 * the specified path. This can happen if:
 866 *
 867 *   1. the path is mentioned explicitly in the pathspec
 868 *
 869 *   2. the path is a directory prefix of some element in the
 870 *      pathspec
 871 */
 872static int exclude_matches_pathspec(const char *path, int len,
 873                const struct path_simplify *simplify)
 874{
 875        if (simplify) {
 876                for (; simplify->path; simplify++) {
 877                        if (len == simplify->len
 878                            && !memcmp(path, simplify->path, len))
 879                                return 1;
 880                        if (len < simplify->len
 881                            && simplify->path[len] == '/'
 882                            && !memcmp(path, simplify->path, len))
 883                                return 1;
 884                }
 885        }
 886        return 0;
 887}
 888
 889static int get_index_dtype(const char *path, int len)
 890{
 891        int pos;
 892        struct cache_entry *ce;
 893
 894        ce = cache_name_exists(path, len, 0);
 895        if (ce) {
 896                if (!ce_uptodate(ce))
 897                        return DT_UNKNOWN;
 898                if (S_ISGITLINK(ce->ce_mode))
 899                        return DT_DIR;
 900                /*
 901                 * Nobody actually cares about the
 902                 * difference between DT_LNK and DT_REG
 903                 */
 904                return DT_REG;
 905        }
 906
 907        /* Try to look it up as a directory */
 908        pos = cache_name_pos(path, len);
 909        if (pos >= 0)
 910                return DT_UNKNOWN;
 911        pos = -pos-1;
 912        while (pos < active_nr) {
 913                ce = active_cache[pos++];
 914                if (strncmp(ce->name, path, len))
 915                        break;
 916                if (ce->name[len] > '/')
 917                        break;
 918                if (ce->name[len] < '/')
 919                        continue;
 920                if (!ce_uptodate(ce))
 921                        break;  /* continue? */
 922                return DT_DIR;
 923        }
 924        return DT_UNKNOWN;
 925}
 926
 927static int get_dtype(struct dirent *de, const char *path, int len)
 928{
 929        int dtype = de ? DTYPE(de) : DT_UNKNOWN;
 930        struct stat st;
 931
 932        if (dtype != DT_UNKNOWN)
 933                return dtype;
 934        dtype = get_index_dtype(path, len);
 935        if (dtype != DT_UNKNOWN)
 936                return dtype;
 937        if (lstat(path, &st))
 938                return dtype;
 939        if (S_ISREG(st.st_mode))
 940                return DT_REG;
 941        if (S_ISDIR(st.st_mode))
 942                return DT_DIR;
 943        if (S_ISLNK(st.st_mode))
 944                return DT_LNK;
 945        return dtype;
 946}
 947
 948enum path_treatment {
 949        path_ignored,
 950        path_handled,
 951        path_recurse
 952};
 953
 954static enum path_treatment treat_one_path(struct dir_struct *dir,
 955                                          struct strbuf *path,
 956                                          const struct path_simplify *simplify,
 957                                          int dtype, struct dirent *de)
 958{
 959        int exclude = excluded(dir, path->buf, &dtype);
 960        if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
 961            && exclude_matches_pathspec(path->buf, path->len, simplify))
 962                dir_add_ignored(dir, path->buf, path->len);
 963
 964        /*
 965         * Excluded? If we don't explicitly want to show
 966         * ignored files, ignore it
 967         */
 968        if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
 969                return path_ignored;
 970
 971        if (dtype == DT_UNKNOWN)
 972                dtype = get_dtype(de, path->buf, path->len);
 973
 974        /*
 975         * Do we want to see just the ignored files?
 976         * We still need to recurse into directories,
 977         * even if we don't ignore them, since the
 978         * directory may contain files that we do..
 979         */
 980        if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
 981                if (dtype != DT_DIR)
 982                        return path_ignored;
 983        }
 984
 985        switch (dtype) {
 986        default:
 987                return path_ignored;
 988        case DT_DIR:
 989                strbuf_addch(path, '/');
 990                switch (treat_directory(dir, path->buf, path->len, simplify)) {
 991                case show_directory:
 992                        if (exclude != !!(dir->flags
 993                                          & DIR_SHOW_IGNORED))
 994                                return path_ignored;
 995                        break;
 996                case recurse_into_directory:
 997                        return path_recurse;
 998                case ignore_directory:
 999                        return path_ignored;
1000                }
1001                break;
1002        case DT_REG:
1003        case DT_LNK:
1004                break;
1005        }
1006        return path_handled;
1007}
1008
1009static enum path_treatment treat_path(struct dir_struct *dir,
1010                                      struct dirent *de,
1011                                      struct strbuf *path,
1012                                      int baselen,
1013                                      const struct path_simplify *simplify)
1014{
1015        int dtype;
1016
1017        if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1018                return path_ignored;
1019        strbuf_setlen(path, baselen);
1020        strbuf_addstr(path, de->d_name);
1021        if (simplify_away(path->buf, path->len, simplify))
1022                return path_ignored;
1023
1024        dtype = DTYPE(de);
1025        return treat_one_path(dir, path, simplify, dtype, de);
1026}
1027
1028/*
1029 * Read a directory tree. We currently ignore anything but
1030 * directories, regular files and symlinks. That's because git
1031 * doesn't handle them at all yet. Maybe that will change some
1032 * day.
1033 *
1034 * Also, we ignore the name ".git" (even if it is not a directory).
1035 * That likely will not change.
1036 */
1037static int read_directory_recursive(struct dir_struct *dir,
1038                                    const char *base, int baselen,
1039                                    int check_only,
1040                                    const struct path_simplify *simplify)
1041{
1042        DIR *fdir;
1043        int contents = 0;
1044        struct dirent *de;
1045        struct strbuf path = STRBUF_INIT;
1046
1047        strbuf_add(&path, base, baselen);
1048
1049        fdir = opendir(path.len ? path.buf : ".");
1050        if (!fdir)
1051                goto out;
1052
1053        while ((de = readdir(fdir)) != NULL) {
1054                switch (treat_path(dir, de, &path, baselen, simplify)) {
1055                case path_recurse:
1056                        contents += read_directory_recursive(dir, path.buf,
1057                                                             path.len, 0,
1058                                                             simplify);
1059                        continue;
1060                case path_ignored:
1061                        continue;
1062                case path_handled:
1063                        break;
1064                }
1065                contents++;
1066                if (check_only)
1067                        break;
1068                dir_add_name(dir, path.buf, path.len);
1069        }
1070        closedir(fdir);
1071 out:
1072        strbuf_release(&path);
1073
1074        return contents;
1075}
1076
1077static int cmp_name(const void *p1, const void *p2)
1078{
1079        const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1080        const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1081
1082        return cache_name_compare(e1->name, e1->len,
1083                                  e2->name, e2->len);
1084}
1085
1086static struct path_simplify *create_simplify(const char **pathspec)
1087{
1088        int nr, alloc = 0;
1089        struct path_simplify *simplify = NULL;
1090
1091        if (!pathspec)
1092                return NULL;
1093
1094        for (nr = 0 ; ; nr++) {
1095                const char *match;
1096                if (nr >= alloc) {
1097                        alloc = alloc_nr(alloc);
1098                        simplify = xrealloc(simplify, alloc * sizeof(*simplify));
1099                }
1100                match = *pathspec++;
1101                if (!match)
1102                        break;
1103                simplify[nr].path = match;
1104                simplify[nr].len = simple_length(match);
1105        }
1106        simplify[nr].path = NULL;
1107        simplify[nr].len = 0;
1108        return simplify;
1109}
1110
1111static void free_simplify(struct path_simplify *simplify)
1112{
1113        free(simplify);
1114}
1115
1116static int treat_leading_path(struct dir_struct *dir,
1117                              const char *path, int len,
1118                              const struct path_simplify *simplify)
1119{
1120        struct strbuf sb = STRBUF_INIT;
1121        int baselen, rc = 0;
1122        const char *cp;
1123
1124        while (len && path[len - 1] == '/')
1125                len--;
1126        if (!len)
1127                return 1;
1128        baselen = 0;
1129        while (1) {
1130                cp = path + baselen + !!baselen;
1131                cp = memchr(cp, '/', path + len - cp);
1132                if (!cp)
1133                        baselen = len;
1134                else
1135                        baselen = cp - path;
1136                strbuf_setlen(&sb, 0);
1137                strbuf_add(&sb, path, baselen);
1138                if (!is_directory(sb.buf))
1139                        break;
1140                if (simplify_away(sb.buf, sb.len, simplify))
1141                        break;
1142                if (treat_one_path(dir, &sb, simplify,
1143                                   DT_DIR, NULL) == path_ignored)
1144                        break; /* do not recurse into it */
1145                if (len <= baselen) {
1146                        rc = 1;
1147                        break; /* finished checking */
1148                }
1149        }
1150        strbuf_release(&sb);
1151        return rc;
1152}
1153
1154int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
1155{
1156        struct path_simplify *simplify;
1157
1158        if (has_symlink_leading_path(path, len))
1159                return dir->nr;
1160
1161        simplify = create_simplify(pathspec);
1162        if (!len || treat_leading_path(dir, path, len, simplify))
1163                read_directory_recursive(dir, path, len, 0, simplify);
1164        free_simplify(simplify);
1165        qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1166        qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1167        return dir->nr;
1168}
1169
1170int file_exists(const char *f)
1171{
1172        struct stat sb;
1173        return lstat(f, &sb) == 0;
1174}
1175
1176/*
1177 * Given two normalized paths (a trailing slash is ok), if subdir is
1178 * outside dir, return -1.  Otherwise return the offset in subdir that
1179 * can be used as relative path to dir.
1180 */
1181int dir_inside_of(const char *subdir, const char *dir)
1182{
1183        int offset = 0;
1184
1185        assert(dir && subdir && *dir && *subdir);
1186
1187        while (*dir && *subdir && *dir == *subdir) {
1188                dir++;
1189                subdir++;
1190                offset++;
1191        }
1192
1193        /* hel[p]/me vs hel[l]/yeah */
1194        if (*dir && *subdir)
1195                return -1;
1196
1197        if (!*subdir)
1198                return !*dir ? offset : -1; /* same dir */
1199
1200        /* foo/[b]ar vs foo/[] */
1201        if (is_dir_sep(dir[-1]))
1202                return is_dir_sep(subdir[-1]) ? offset : -1;
1203
1204        /* foo[/]bar vs foo[] */
1205        return is_dir_sep(*subdir) ? offset + 1 : -1;
1206}
1207
1208int is_inside_dir(const char *dir)
1209{
1210        char cwd[PATH_MAX];
1211        if (!dir)
1212                return 0;
1213        if (!getcwd(cwd, sizeof(cwd)))
1214                die_errno("can't find the current directory");
1215        return dir_inside_of(cwd, dir) >= 0;
1216}
1217
1218int is_empty_dir(const char *path)
1219{
1220        DIR *dir = opendir(path);
1221        struct dirent *e;
1222        int ret = 1;
1223
1224        if (!dir)
1225                return 0;
1226
1227        while ((e = readdir(dir)) != NULL)
1228                if (!is_dot_or_dotdot(e->d_name)) {
1229                        ret = 0;
1230                        break;
1231                }
1232
1233        closedir(dir);
1234        return ret;
1235}
1236
1237static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1238{
1239        DIR *dir;
1240        struct dirent *e;
1241        int ret = 0, original_len = path->len, len, kept_down = 0;
1242        int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1243        int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1244        unsigned char submodule_head[20];
1245
1246        if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1247            !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1248                /* Do not descend and nuke a nested git work tree. */
1249                if (kept_up)
1250                        *kept_up = 1;
1251                return 0;
1252        }
1253
1254        flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1255        dir = opendir(path->buf);
1256        if (!dir) {
1257                /* an empty dir could be removed even if it is unreadble */
1258                if (!keep_toplevel)
1259                        return rmdir(path->buf);
1260                else
1261                        return -1;
1262        }
1263        if (path->buf[original_len - 1] != '/')
1264                strbuf_addch(path, '/');
1265
1266        len = path->len;
1267        while ((e = readdir(dir)) != NULL) {
1268                struct stat st;
1269                if (is_dot_or_dotdot(e->d_name))
1270                        continue;
1271
1272                strbuf_setlen(path, len);
1273                strbuf_addstr(path, e->d_name);
1274                if (lstat(path->buf, &st))
1275                        ; /* fall thru */
1276                else if (S_ISDIR(st.st_mode)) {
1277                        if (!remove_dir_recurse(path, flag, &kept_down))
1278                                continue; /* happy */
1279                } else if (!only_empty && !unlink(path->buf))
1280                        continue; /* happy, too */
1281
1282                /* path too long, stat fails, or non-directory still exists */
1283                ret = -1;
1284                break;
1285        }
1286        closedir(dir);
1287
1288        strbuf_setlen(path, original_len);
1289        if (!ret && !keep_toplevel && !kept_down)
1290                ret = rmdir(path->buf);
1291        else if (kept_up)
1292                /*
1293                 * report the uplevel that it is not an error that we
1294                 * did not rmdir() our directory.
1295                 */
1296                *kept_up = !ret;
1297        return ret;
1298}
1299
1300int remove_dir_recursively(struct strbuf *path, int flag)
1301{
1302        return remove_dir_recurse(path, flag, NULL);
1303}
1304
1305void setup_standard_excludes(struct dir_struct *dir)
1306{
1307        const char *path;
1308        char *xdg_path;
1309
1310        dir->exclude_per_dir = ".gitignore";
1311        path = git_path("info/exclude");
1312        if (!excludes_file) {
1313                home_config_paths(NULL, &xdg_path, "ignore");
1314                excludes_file = xdg_path;
1315        }
1316        if (!access_or_warn(path, R_OK))
1317                add_excludes_from_file(dir, path);
1318        if (excludes_file && !access_or_warn(excludes_file, R_OK))
1319                add_excludes_from_file(dir, excludes_file);
1320}
1321
1322int remove_path(const char *name)
1323{
1324        char *slash;
1325
1326        if (unlink(name) && errno != ENOENT)
1327                return -1;
1328
1329        slash = strrchr(name, '/');
1330        if (slash) {
1331                char *dirs = xstrdup(name);
1332                slash = dirs + (slash - name);
1333                do {
1334                        *slash = '\0';
1335                } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1336                free(dirs);
1337        }
1338        return 0;
1339}
1340
1341static int pathspec_item_cmp(const void *a_, const void *b_)
1342{
1343        struct pathspec_item *a, *b;
1344
1345        a = (struct pathspec_item *)a_;
1346        b = (struct pathspec_item *)b_;
1347        return strcmp(a->match, b->match);
1348}
1349
1350int init_pathspec(struct pathspec *pathspec, const char **paths)
1351{
1352        const char **p = paths;
1353        int i;
1354
1355        memset(pathspec, 0, sizeof(*pathspec));
1356        if (!p)
1357                return 0;
1358        while (*p)
1359                p++;
1360        pathspec->raw = paths;
1361        pathspec->nr = p - paths;
1362        if (!pathspec->nr)
1363                return 0;
1364
1365        pathspec->items = xmalloc(sizeof(struct pathspec_item)*pathspec->nr);
1366        for (i = 0; i < pathspec->nr; i++) {
1367                struct pathspec_item *item = pathspec->items+i;
1368                const char *path = paths[i];
1369
1370                item->match = path;
1371                item->len = strlen(path);
1372                item->use_wildcard = !no_wildcard(path);
1373                if (item->use_wildcard)
1374                        pathspec->has_wildcard = 1;
1375        }
1376
1377        qsort(pathspec->items, pathspec->nr,
1378              sizeof(struct pathspec_item), pathspec_item_cmp);
1379
1380        return 0;
1381}
1382
1383void free_pathspec(struct pathspec *pathspec)
1384{
1385        free(pathspec->items);
1386        pathspec->items = NULL;
1387}