dir.con commit read-cache: fix untracked cache invalidation when split-index is used (ffcc9ba)
   1/*
   2 * This handles recursive filename detection with exclude
   3 * files, index knowledge etc..
   4 *
   5 * See Documentation/technical/api-directory-listing.txt
   6 *
   7 * Copyright (C) Linus Torvalds, 2005-2006
   8 *               Junio Hamano, 2005-2006
   9 */
  10#include "cache.h"
  11#include "dir.h"
  12#include "refs.h"
  13#include "wildmatch.h"
  14#include "pathspec.h"
  15#include "varint.h"
  16#include "ewah/ewok.h"
  17
  18struct path_simplify {
  19        int len;
  20        const char *path;
  21};
  22
  23/*
  24 * Tells read_directory_recursive how a file or directory should be treated.
  25 * Values are ordered by significance, e.g. if a directory contains both
  26 * excluded and untracked files, it is listed as untracked because
  27 * path_untracked > path_excluded.
  28 */
  29enum path_treatment {
  30        path_none = 0,
  31        path_recurse,
  32        path_excluded,
  33        path_untracked
  34};
  35
  36/*
  37 * Support data structure for our opendir/readdir/closedir wrappers
  38 */
  39struct cached_dir {
  40        DIR *fdir;
  41        struct untracked_cache_dir *untracked;
  42        int nr_files;
  43        int nr_dirs;
  44
  45        struct dirent *de;
  46        const char *file;
  47        struct untracked_cache_dir *ucd;
  48};
  49
  50static enum path_treatment read_directory_recursive(struct dir_struct *dir,
  51        const char *path, int len, struct untracked_cache_dir *untracked,
  52        int check_only, const struct path_simplify *simplify);
  53static int get_dtype(struct dirent *de, const char *path, int len);
  54
  55/* helper string functions with support for the ignore_case flag */
  56int strcmp_icase(const char *a, const char *b)
  57{
  58        return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
  59}
  60
  61int strncmp_icase(const char *a, const char *b, size_t count)
  62{
  63        return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
  64}
  65
  66int fnmatch_icase(const char *pattern, const char *string, int flags)
  67{
  68        return wildmatch(pattern, string,
  69                         flags | (ignore_case ? WM_CASEFOLD : 0),
  70                         NULL);
  71}
  72
  73int git_fnmatch(const struct pathspec_item *item,
  74                const char *pattern, const char *string,
  75                int prefix)
  76{
  77        if (prefix > 0) {
  78                if (ps_strncmp(item, pattern, string, prefix))
  79                        return WM_NOMATCH;
  80                pattern += prefix;
  81                string += prefix;
  82        }
  83        if (item->flags & PATHSPEC_ONESTAR) {
  84                int pattern_len = strlen(++pattern);
  85                int string_len = strlen(string);
  86                return string_len < pattern_len ||
  87                        ps_strcmp(item, pattern,
  88                                  string + string_len - pattern_len);
  89        }
  90        if (item->magic & PATHSPEC_GLOB)
  91                return wildmatch(pattern, string,
  92                                 WM_PATHNAME |
  93                                 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
  94                                 NULL);
  95        else
  96                /* wildmatch has not learned no FNM_PATHNAME mode yet */
  97                return wildmatch(pattern, string,
  98                                 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
  99                                 NULL);
 100}
 101
 102static int fnmatch_icase_mem(const char *pattern, int patternlen,
 103                             const char *string, int stringlen,
 104                             int flags)
 105{
 106        int match_status;
 107        struct strbuf pat_buf = STRBUF_INIT;
 108        struct strbuf str_buf = STRBUF_INIT;
 109        const char *use_pat = pattern;
 110        const char *use_str = string;
 111
 112        if (pattern[patternlen]) {
 113                strbuf_add(&pat_buf, pattern, patternlen);
 114                use_pat = pat_buf.buf;
 115        }
 116        if (string[stringlen]) {
 117                strbuf_add(&str_buf, string, stringlen);
 118                use_str = str_buf.buf;
 119        }
 120
 121        if (ignore_case)
 122                flags |= WM_CASEFOLD;
 123        match_status = wildmatch(use_pat, use_str, flags, NULL);
 124
 125        strbuf_release(&pat_buf);
 126        strbuf_release(&str_buf);
 127
 128        return match_status;
 129}
 130
 131static size_t common_prefix_len(const struct pathspec *pathspec)
 132{
 133        int n;
 134        size_t max = 0;
 135
 136        /*
 137         * ":(icase)path" is treated as a pathspec full of
 138         * wildcard. In other words, only prefix is considered common
 139         * prefix. If the pathspec is abc/foo abc/bar, running in
 140         * subdir xyz, the common prefix is still xyz, not xuz/abc as
 141         * in non-:(icase).
 142         */
 143        GUARD_PATHSPEC(pathspec,
 144                       PATHSPEC_FROMTOP |
 145                       PATHSPEC_MAXDEPTH |
 146                       PATHSPEC_LITERAL |
 147                       PATHSPEC_GLOB |
 148                       PATHSPEC_ICASE |
 149                       PATHSPEC_EXCLUDE);
 150
 151        for (n = 0; n < pathspec->nr; n++) {
 152                size_t i = 0, len = 0, item_len;
 153                if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
 154                        continue;
 155                if (pathspec->items[n].magic & PATHSPEC_ICASE)
 156                        item_len = pathspec->items[n].prefix;
 157                else
 158                        item_len = pathspec->items[n].nowildcard_len;
 159                while (i < item_len && (n == 0 || i < max)) {
 160                        char c = pathspec->items[n].match[i];
 161                        if (c != pathspec->items[0].match[i])
 162                                break;
 163                        if (c == '/')
 164                                len = i + 1;
 165                        i++;
 166                }
 167                if (n == 0 || len < max) {
 168                        max = len;
 169                        if (!max)
 170                                break;
 171                }
 172        }
 173        return max;
 174}
 175
 176/*
 177 * Returns a copy of the longest leading path common among all
 178 * pathspecs.
 179 */
 180char *common_prefix(const struct pathspec *pathspec)
 181{
 182        unsigned long len = common_prefix_len(pathspec);
 183
 184        return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
 185}
 186
 187int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
 188{
 189        size_t len;
 190
 191        /*
 192         * Calculate common prefix for the pathspec, and
 193         * use that to optimize the directory walk
 194         */
 195        len = common_prefix_len(pathspec);
 196
 197        /* Read the directory and prune it */
 198        read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec);
 199        return len;
 200}
 201
 202int within_depth(const char *name, int namelen,
 203                        int depth, int max_depth)
 204{
 205        const char *cp = name, *cpe = name + namelen;
 206
 207        while (cp < cpe) {
 208                if (*cp++ != '/')
 209                        continue;
 210                depth++;
 211                if (depth > max_depth)
 212                        return 0;
 213        }
 214        return 1;
 215}
 216
 217#define DO_MATCH_EXCLUDE   1
 218#define DO_MATCH_DIRECTORY 2
 219
 220/*
 221 * Does 'match' match the given name?
 222 * A match is found if
 223 *
 224 * (1) the 'match' string is leading directory of 'name', or
 225 * (2) the 'match' string is a wildcard and matches 'name', or
 226 * (3) the 'match' string is exactly the same as 'name'.
 227 *
 228 * and the return value tells which case it was.
 229 *
 230 * It returns 0 when there is no match.
 231 */
 232static int match_pathspec_item(const struct pathspec_item *item, int prefix,
 233                               const char *name, int namelen, unsigned flags)
 234{
 235        /* name/namelen has prefix cut off by caller */
 236        const char *match = item->match + prefix;
 237        int matchlen = item->len - prefix;
 238
 239        /*
 240         * The normal call pattern is:
 241         * 1. prefix = common_prefix_len(ps);
 242         * 2. prune something, or fill_directory
 243         * 3. match_pathspec()
 244         *
 245         * 'prefix' at #1 may be shorter than the command's prefix and
 246         * it's ok for #2 to match extra files. Those extras will be
 247         * trimmed at #3.
 248         *
 249         * Suppose the pathspec is 'foo' and '../bar' running from
 250         * subdir 'xyz'. The common prefix at #1 will be empty, thanks
 251         * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
 252         * user does not want XYZ/foo, only the "foo" part should be
 253         * case-insensitive. We need to filter out XYZ/foo here. In
 254         * other words, we do not trust the caller on comparing the
 255         * prefix part when :(icase) is involved. We do exact
 256         * comparison ourselves.
 257         *
 258         * Normally the caller (common_prefix_len() in fact) does
 259         * _exact_ matching on name[-prefix+1..-1] and we do not need
 260         * to check that part. Be defensive and check it anyway, in
 261         * case common_prefix_len is changed, or a new caller is
 262         * introduced that does not use common_prefix_len.
 263         *
 264         * If the penalty turns out too high when prefix is really
 265         * long, maybe change it to
 266         * strncmp(match, name, item->prefix - prefix)
 267         */
 268        if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
 269            strncmp(item->match, name - prefix, item->prefix))
 270                return 0;
 271
 272        /* If the match was just the prefix, we matched */
 273        if (!*match)
 274                return MATCHED_RECURSIVELY;
 275
 276        if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
 277                if (matchlen == namelen)
 278                        return MATCHED_EXACTLY;
 279
 280                if (match[matchlen-1] == '/' || name[matchlen] == '/')
 281                        return MATCHED_RECURSIVELY;
 282        } else if ((flags & DO_MATCH_DIRECTORY) &&
 283                   match[matchlen - 1] == '/' &&
 284                   namelen == matchlen - 1 &&
 285                   !ps_strncmp(item, match, name, namelen))
 286                return MATCHED_EXACTLY;
 287
 288        if (item->nowildcard_len < item->len &&
 289            !git_fnmatch(item, match, name,
 290                         item->nowildcard_len - prefix))
 291                return MATCHED_FNMATCH;
 292
 293        return 0;
 294}
 295
 296/*
 297 * Given a name and a list of pathspecs, returns the nature of the
 298 * closest (i.e. most specific) match of the name to any of the
 299 * pathspecs.
 300 *
 301 * The caller typically calls this multiple times with the same
 302 * pathspec and seen[] array but with different name/namelen
 303 * (e.g. entries from the index) and is interested in seeing if and
 304 * how each pathspec matches all the names it calls this function
 305 * with.  A mark is left in the seen[] array for each pathspec element
 306 * indicating the closest type of match that element achieved, so if
 307 * seen[n] remains zero after multiple invocations, that means the nth
 308 * pathspec did not match any names, which could indicate that the
 309 * user mistyped the nth pathspec.
 310 */
 311static int do_match_pathspec(const struct pathspec *ps,
 312                             const char *name, int namelen,
 313                             int prefix, char *seen,
 314                             unsigned flags)
 315{
 316        int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
 317
 318        GUARD_PATHSPEC(ps,
 319                       PATHSPEC_FROMTOP |
 320                       PATHSPEC_MAXDEPTH |
 321                       PATHSPEC_LITERAL |
 322                       PATHSPEC_GLOB |
 323                       PATHSPEC_ICASE |
 324                       PATHSPEC_EXCLUDE);
 325
 326        if (!ps->nr) {
 327                if (!ps->recursive ||
 328                    !(ps->magic & PATHSPEC_MAXDEPTH) ||
 329                    ps->max_depth == -1)
 330                        return MATCHED_RECURSIVELY;
 331
 332                if (within_depth(name, namelen, 0, ps->max_depth))
 333                        return MATCHED_EXACTLY;
 334                else
 335                        return 0;
 336        }
 337
 338        name += prefix;
 339        namelen -= prefix;
 340
 341        for (i = ps->nr - 1; i >= 0; i--) {
 342                int how;
 343
 344                if ((!exclude &&   ps->items[i].magic & PATHSPEC_EXCLUDE) ||
 345                    ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
 346                        continue;
 347
 348                if (seen && seen[i] == MATCHED_EXACTLY)
 349                        continue;
 350                /*
 351                 * Make exclude patterns optional and never report
 352                 * "pathspec ':(exclude)foo' matches no files"
 353                 */
 354                if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
 355                        seen[i] = MATCHED_FNMATCH;
 356                how = match_pathspec_item(ps->items+i, prefix, name,
 357                                          namelen, flags);
 358                if (ps->recursive &&
 359                    (ps->magic & PATHSPEC_MAXDEPTH) &&
 360                    ps->max_depth != -1 &&
 361                    how && how != MATCHED_FNMATCH) {
 362                        int len = ps->items[i].len;
 363                        if (name[len] == '/')
 364                                len++;
 365                        if (within_depth(name+len, namelen-len, 0, ps->max_depth))
 366                                how = MATCHED_EXACTLY;
 367                        else
 368                                how = 0;
 369                }
 370                if (how) {
 371                        if (retval < how)
 372                                retval = how;
 373                        if (seen && seen[i] < how)
 374                                seen[i] = how;
 375                }
 376        }
 377        return retval;
 378}
 379
 380int match_pathspec(const struct pathspec *ps,
 381                   const char *name, int namelen,
 382                   int prefix, char *seen, int is_dir)
 383{
 384        int positive, negative;
 385        unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
 386        positive = do_match_pathspec(ps, name, namelen,
 387                                     prefix, seen, flags);
 388        if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
 389                return positive;
 390        negative = do_match_pathspec(ps, name, namelen,
 391                                     prefix, seen,
 392                                     flags | DO_MATCH_EXCLUDE);
 393        return negative ? 0 : positive;
 394}
 395
 396/*
 397 * Return the length of the "simple" part of a path match limiter.
 398 */
 399int simple_length(const char *match)
 400{
 401        int len = -1;
 402
 403        for (;;) {
 404                unsigned char c = *match++;
 405                len++;
 406                if (c == '\0' || is_glob_special(c))
 407                        return len;
 408        }
 409}
 410
 411int no_wildcard(const char *string)
 412{
 413        return string[simple_length(string)] == '\0';
 414}
 415
 416void parse_exclude_pattern(const char **pattern,
 417                           int *patternlen,
 418                           int *flags,
 419                           int *nowildcardlen)
 420{
 421        const char *p = *pattern;
 422        size_t i, len;
 423
 424        *flags = 0;
 425        if (*p == '!') {
 426                *flags |= EXC_FLAG_NEGATIVE;
 427                p++;
 428        }
 429        len = strlen(p);
 430        if (len && p[len - 1] == '/') {
 431                len--;
 432                *flags |= EXC_FLAG_MUSTBEDIR;
 433        }
 434        for (i = 0; i < len; i++) {
 435                if (p[i] == '/')
 436                        break;
 437        }
 438        if (i == len)
 439                *flags |= EXC_FLAG_NODIR;
 440        *nowildcardlen = simple_length(p);
 441        /*
 442         * we should have excluded the trailing slash from 'p' too,
 443         * but that's one more allocation. Instead just make sure
 444         * nowildcardlen does not exceed real patternlen
 445         */
 446        if (*nowildcardlen > len)
 447                *nowildcardlen = len;
 448        if (*p == '*' && no_wildcard(p + 1))
 449                *flags |= EXC_FLAG_ENDSWITH;
 450        *pattern = p;
 451        *patternlen = len;
 452}
 453
 454void add_exclude(const char *string, const char *base,
 455                 int baselen, struct exclude_list *el, int srcpos)
 456{
 457        struct exclude *x;
 458        int patternlen;
 459        int flags;
 460        int nowildcardlen;
 461
 462        parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
 463        if (flags & EXC_FLAG_MUSTBEDIR) {
 464                char *s;
 465                x = xmalloc(sizeof(*x) + patternlen + 1);
 466                s = (char *)(x+1);
 467                memcpy(s, string, patternlen);
 468                s[patternlen] = '\0';
 469                x->pattern = s;
 470        } else {
 471                x = xmalloc(sizeof(*x));
 472                x->pattern = string;
 473        }
 474        x->patternlen = patternlen;
 475        x->nowildcardlen = nowildcardlen;
 476        x->base = base;
 477        x->baselen = baselen;
 478        x->flags = flags;
 479        x->srcpos = srcpos;
 480        ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
 481        el->excludes[el->nr++] = x;
 482        x->el = el;
 483}
 484
 485static void *read_skip_worktree_file_from_index(const char *path, size_t *size,
 486                                                struct sha1_stat *sha1_stat)
 487{
 488        int pos, len;
 489        unsigned long sz;
 490        enum object_type type;
 491        void *data;
 492
 493        len = strlen(path);
 494        pos = cache_name_pos(path, len);
 495        if (pos < 0)
 496                return NULL;
 497        if (!ce_skip_worktree(active_cache[pos]))
 498                return NULL;
 499        data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
 500        if (!data || type != OBJ_BLOB) {
 501                free(data);
 502                return NULL;
 503        }
 504        *size = xsize_t(sz);
 505        if (sha1_stat) {
 506                memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
 507                hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
 508        }
 509        return data;
 510}
 511
 512/*
 513 * Frees memory within el which was allocated for exclude patterns and
 514 * the file buffer.  Does not free el itself.
 515 */
 516void clear_exclude_list(struct exclude_list *el)
 517{
 518        int i;
 519
 520        for (i = 0; i < el->nr; i++)
 521                free(el->excludes[i]);
 522        free(el->excludes);
 523        free(el->filebuf);
 524
 525        el->nr = 0;
 526        el->excludes = NULL;
 527        el->filebuf = NULL;
 528}
 529
 530static void trim_trailing_spaces(char *buf)
 531{
 532        char *p, *last_space = NULL;
 533
 534        for (p = buf; *p; p++)
 535                switch (*p) {
 536                case ' ':
 537                        if (!last_space)
 538                                last_space = p;
 539                        break;
 540                case '\\':
 541                        p++;
 542                        if (!*p)
 543                                return;
 544                        /* fallthrough */
 545                default:
 546                        last_space = NULL;
 547                }
 548
 549        if (last_space)
 550                *last_space = '\0';
 551}
 552
 553/*
 554 * Given a subdirectory name and "dir" of the current directory,
 555 * search the subdir in "dir" and return it, or create a new one if it
 556 * does not exist in "dir".
 557 *
 558 * If "name" has the trailing slash, it'll be excluded in the search.
 559 */
 560static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
 561                                                    struct untracked_cache_dir *dir,
 562                                                    const char *name, int len)
 563{
 564        int first, last;
 565        struct untracked_cache_dir *d;
 566        if (!dir)
 567                return NULL;
 568        if (len && name[len - 1] == '/')
 569                len--;
 570        first = 0;
 571        last = dir->dirs_nr;
 572        while (last > first) {
 573                int cmp, next = (last + first) >> 1;
 574                d = dir->dirs[next];
 575                cmp = strncmp(name, d->name, len);
 576                if (!cmp && strlen(d->name) > len)
 577                        cmp = -1;
 578                if (!cmp)
 579                        return d;
 580                if (cmp < 0) {
 581                        last = next;
 582                        continue;
 583                }
 584                first = next+1;
 585        }
 586
 587        uc->dir_created++;
 588        d = xmalloc(sizeof(*d) + len + 1);
 589        memset(d, 0, sizeof(*d));
 590        memcpy(d->name, name, len);
 591        d->name[len] = '\0';
 592
 593        ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
 594        memmove(dir->dirs + first + 1, dir->dirs + first,
 595                (dir->dirs_nr - first) * sizeof(*dir->dirs));
 596        dir->dirs_nr++;
 597        dir->dirs[first] = d;
 598        return d;
 599}
 600
 601static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
 602{
 603        int i;
 604        dir->valid = 0;
 605        dir->untracked_nr = 0;
 606        for (i = 0; i < dir->dirs_nr; i++)
 607                do_invalidate_gitignore(dir->dirs[i]);
 608}
 609
 610static void invalidate_gitignore(struct untracked_cache *uc,
 611                                 struct untracked_cache_dir *dir)
 612{
 613        uc->gitignore_invalidated++;
 614        do_invalidate_gitignore(dir);
 615}
 616
 617static void invalidate_directory(struct untracked_cache *uc,
 618                                 struct untracked_cache_dir *dir)
 619{
 620        int i;
 621        uc->dir_invalidated++;
 622        dir->valid = 0;
 623        dir->untracked_nr = 0;
 624        for (i = 0; i < dir->dirs_nr; i++)
 625                dir->dirs[i]->recurse = 0;
 626}
 627
 628/*
 629 * Given a file with name "fname", read it (either from disk, or from
 630 * the index if "check_index" is non-zero), parse it and store the
 631 * exclude rules in "el".
 632 *
 633 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
 634 * stat data from disk (only valid if add_excludes returns zero). If
 635 * ss_valid is non-zero, "ss" must contain good value as input.
 636 */
 637static int add_excludes(const char *fname, const char *base, int baselen,
 638                        struct exclude_list *el, int check_index,
 639                        struct sha1_stat *sha1_stat)
 640{
 641        struct stat st;
 642        int fd, i, lineno = 1;
 643        size_t size = 0;
 644        char *buf, *entry;
 645
 646        fd = open(fname, O_RDONLY);
 647        if (fd < 0 || fstat(fd, &st) < 0) {
 648                if (errno != ENOENT)
 649                        warn_on_inaccessible(fname);
 650                if (0 <= fd)
 651                        close(fd);
 652                if (!check_index ||
 653                    (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL)
 654                        return -1;
 655                if (size == 0) {
 656                        free(buf);
 657                        return 0;
 658                }
 659                if (buf[size-1] != '\n') {
 660                        buf = xrealloc(buf, size+1);
 661                        buf[size++] = '\n';
 662                }
 663        } else {
 664                size = xsize_t(st.st_size);
 665                if (size == 0) {
 666                        if (sha1_stat) {
 667                                fill_stat_data(&sha1_stat->stat, &st);
 668                                hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
 669                                sha1_stat->valid = 1;
 670                        }
 671                        close(fd);
 672                        return 0;
 673                }
 674                buf = xmalloc(size+1);
 675                if (read_in_full(fd, buf, size) != size) {
 676                        free(buf);
 677                        close(fd);
 678                        return -1;
 679                }
 680                buf[size++] = '\n';
 681                close(fd);
 682                if (sha1_stat) {
 683                        int pos;
 684                        if (sha1_stat->valid &&
 685                            !match_stat_data_racy(&the_index, &sha1_stat->stat, &st))
 686                                ; /* no content change, ss->sha1 still good */
 687                        else if (check_index &&
 688                                 (pos = cache_name_pos(fname, strlen(fname))) >= 0 &&
 689                                 !ce_stage(active_cache[pos]) &&
 690                                 ce_uptodate(active_cache[pos]) &&
 691                                 !would_convert_to_git(fname))
 692                                hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
 693                        else
 694                                hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
 695                        fill_stat_data(&sha1_stat->stat, &st);
 696                        sha1_stat->valid = 1;
 697                }
 698        }
 699
 700        el->filebuf = buf;
 701        entry = buf;
 702        for (i = 0; i < size; i++) {
 703                if (buf[i] == '\n') {
 704                        if (entry != buf + i && entry[0] != '#') {
 705                                buf[i - (i && buf[i-1] == '\r')] = 0;
 706                                trim_trailing_spaces(entry);
 707                                add_exclude(entry, base, baselen, el, lineno);
 708                        }
 709                        lineno++;
 710                        entry = buf + i + 1;
 711                }
 712        }
 713        return 0;
 714}
 715
 716int add_excludes_from_file_to_list(const char *fname, const char *base,
 717                                   int baselen, struct exclude_list *el,
 718                                   int check_index)
 719{
 720        return add_excludes(fname, base, baselen, el, check_index, NULL);
 721}
 722
 723struct exclude_list *add_exclude_list(struct dir_struct *dir,
 724                                      int group_type, const char *src)
 725{
 726        struct exclude_list *el;
 727        struct exclude_list_group *group;
 728
 729        group = &dir->exclude_list_group[group_type];
 730        ALLOC_GROW(group->el, group->nr + 1, group->alloc);
 731        el = &group->el[group->nr++];
 732        memset(el, 0, sizeof(*el));
 733        el->src = src;
 734        return el;
 735}
 736
 737/*
 738 * Used to set up core.excludesfile and .git/info/exclude lists.
 739 */
 740static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
 741                                     struct sha1_stat *sha1_stat)
 742{
 743        struct exclude_list *el;
 744        /*
 745         * catch setup_standard_excludes() that's called before
 746         * dir->untracked is assigned. That function behaves
 747         * differently when dir->untracked is non-NULL.
 748         */
 749        if (!dir->untracked)
 750                dir->unmanaged_exclude_files++;
 751        el = add_exclude_list(dir, EXC_FILE, fname);
 752        if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0)
 753                die("cannot use %s as an exclude file", fname);
 754}
 755
 756void add_excludes_from_file(struct dir_struct *dir, const char *fname)
 757{
 758        dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
 759        add_excludes_from_file_1(dir, fname, NULL);
 760}
 761
 762int match_basename(const char *basename, int basenamelen,
 763                   const char *pattern, int prefix, int patternlen,
 764                   int flags)
 765{
 766        if (prefix == patternlen) {
 767                if (patternlen == basenamelen &&
 768                    !strncmp_icase(pattern, basename, basenamelen))
 769                        return 1;
 770        } else if (flags & EXC_FLAG_ENDSWITH) {
 771                /* "*literal" matching against "fooliteral" */
 772                if (patternlen - 1 <= basenamelen &&
 773                    !strncmp_icase(pattern + 1,
 774                                   basename + basenamelen - (patternlen - 1),
 775                                   patternlen - 1))
 776                        return 1;
 777        } else {
 778                if (fnmatch_icase_mem(pattern, patternlen,
 779                                      basename, basenamelen,
 780                                      0) == 0)
 781                        return 1;
 782        }
 783        return 0;
 784}
 785
 786int match_pathname(const char *pathname, int pathlen,
 787                   const char *base, int baselen,
 788                   const char *pattern, int prefix, int patternlen,
 789                   int flags)
 790{
 791        const char *name;
 792        int namelen;
 793
 794        /*
 795         * match with FNM_PATHNAME; the pattern has base implicitly
 796         * in front of it.
 797         */
 798        if (*pattern == '/') {
 799                pattern++;
 800                patternlen--;
 801                prefix--;
 802        }
 803
 804        /*
 805         * baselen does not count the trailing slash. base[] may or
 806         * may not end with a trailing slash though.
 807         */
 808        if (pathlen < baselen + 1 ||
 809            (baselen && pathname[baselen] != '/') ||
 810            strncmp_icase(pathname, base, baselen))
 811                return 0;
 812
 813        namelen = baselen ? pathlen - baselen - 1 : pathlen;
 814        name = pathname + pathlen - namelen;
 815
 816        if (prefix) {
 817                /*
 818                 * if the non-wildcard part is longer than the
 819                 * remaining pathname, surely it cannot match.
 820                 */
 821                if (prefix > namelen)
 822                        return 0;
 823
 824                if (strncmp_icase(pattern, name, prefix))
 825                        return 0;
 826                pattern += prefix;
 827                patternlen -= prefix;
 828                name    += prefix;
 829                namelen -= prefix;
 830
 831                /*
 832                 * If the whole pattern did not have a wildcard,
 833                 * then our prefix match is all we need; we
 834                 * do not need to call fnmatch at all.
 835                 */
 836                if (!patternlen && !namelen)
 837                        return 1;
 838        }
 839
 840        return fnmatch_icase_mem(pattern, patternlen,
 841                                 name, namelen,
 842                                 WM_PATHNAME) == 0;
 843}
 844
 845/*
 846 * Scan the given exclude list in reverse to see whether pathname
 847 * should be ignored.  The first match (i.e. the last on the list), if
 848 * any, determines the fate.  Returns the exclude_list element which
 849 * matched, or NULL for undecided.
 850 */
 851static struct exclude *last_exclude_matching_from_list(const char *pathname,
 852                                                       int pathlen,
 853                                                       const char *basename,
 854                                                       int *dtype,
 855                                                       struct exclude_list *el)
 856{
 857        int i;
 858
 859        if (!el->nr)
 860                return NULL;    /* undefined */
 861
 862        for (i = el->nr - 1; 0 <= i; i--) {
 863                struct exclude *x = el->excludes[i];
 864                const char *exclude = x->pattern;
 865                int prefix = x->nowildcardlen;
 866
 867                if (x->flags & EXC_FLAG_MUSTBEDIR) {
 868                        if (*dtype == DT_UNKNOWN)
 869                                *dtype = get_dtype(NULL, pathname, pathlen);
 870                        if (*dtype != DT_DIR)
 871                                continue;
 872                }
 873
 874                if (x->flags & EXC_FLAG_NODIR) {
 875                        if (match_basename(basename,
 876                                           pathlen - (basename - pathname),
 877                                           exclude, prefix, x->patternlen,
 878                                           x->flags))
 879                                return x;
 880                        continue;
 881                }
 882
 883                assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
 884                if (match_pathname(pathname, pathlen,
 885                                   x->base, x->baselen ? x->baselen - 1 : 0,
 886                                   exclude, prefix, x->patternlen, x->flags))
 887                        return x;
 888        }
 889        return NULL; /* undecided */
 890}
 891
 892/*
 893 * Scan the list and let the last match determine the fate.
 894 * Return 1 for exclude, 0 for include and -1 for undecided.
 895 */
 896int is_excluded_from_list(const char *pathname,
 897                          int pathlen, const char *basename, int *dtype,
 898                          struct exclude_list *el)
 899{
 900        struct exclude *exclude;
 901        exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
 902        if (exclude)
 903                return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
 904        return -1; /* undecided */
 905}
 906
 907static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
 908                const char *pathname, int pathlen, const char *basename,
 909                int *dtype_p)
 910{
 911        int i, j;
 912        struct exclude_list_group *group;
 913        struct exclude *exclude;
 914        for (i = EXC_CMDL; i <= EXC_FILE; i++) {
 915                group = &dir->exclude_list_group[i];
 916                for (j = group->nr - 1; j >= 0; j--) {
 917                        exclude = last_exclude_matching_from_list(
 918                                pathname, pathlen, basename, dtype_p,
 919                                &group->el[j]);
 920                        if (exclude)
 921                                return exclude;
 922                }
 923        }
 924        return NULL;
 925}
 926
 927/*
 928 * Loads the per-directory exclude list for the substring of base
 929 * which has a char length of baselen.
 930 */
 931static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
 932{
 933        struct exclude_list_group *group;
 934        struct exclude_list *el;
 935        struct exclude_stack *stk = NULL;
 936        struct untracked_cache_dir *untracked;
 937        int current;
 938
 939        group = &dir->exclude_list_group[EXC_DIRS];
 940
 941        /*
 942         * Pop the exclude lists from the EXCL_DIRS exclude_list_group
 943         * which originate from directories not in the prefix of the
 944         * path being checked.
 945         */
 946        while ((stk = dir->exclude_stack) != NULL) {
 947                if (stk->baselen <= baselen &&
 948                    !strncmp(dir->basebuf.buf, base, stk->baselen))
 949                        break;
 950                el = &group->el[dir->exclude_stack->exclude_ix];
 951                dir->exclude_stack = stk->prev;
 952                dir->exclude = NULL;
 953                free((char *)el->src); /* see strbuf_detach() below */
 954                clear_exclude_list(el);
 955                free(stk);
 956                group->nr--;
 957        }
 958
 959        /* Skip traversing into sub directories if the parent is excluded */
 960        if (dir->exclude)
 961                return;
 962
 963        /*
 964         * Lazy initialization. All call sites currently just
 965         * memset(dir, 0, sizeof(*dir)) before use. Changing all of
 966         * them seems lots of work for little benefit.
 967         */
 968        if (!dir->basebuf.buf)
 969                strbuf_init(&dir->basebuf, PATH_MAX);
 970
 971        /* Read from the parent directories and push them down. */
 972        current = stk ? stk->baselen : -1;
 973        strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
 974        if (dir->untracked)
 975                untracked = stk ? stk->ucd : dir->untracked->root;
 976        else
 977                untracked = NULL;
 978
 979        while (current < baselen) {
 980                const char *cp;
 981                struct sha1_stat sha1_stat;
 982
 983                stk = xcalloc(1, sizeof(*stk));
 984                if (current < 0) {
 985                        cp = base;
 986                        current = 0;
 987                } else {
 988                        cp = strchr(base + current + 1, '/');
 989                        if (!cp)
 990                                die("oops in prep_exclude");
 991                        cp++;
 992                        untracked =
 993                                lookup_untracked(dir->untracked, untracked,
 994                                                 base + current,
 995                                                 cp - base - current);
 996                }
 997                stk->prev = dir->exclude_stack;
 998                stk->baselen = cp - base;
 999                stk->exclude_ix = group->nr;
1000                stk->ucd = untracked;
1001                el = add_exclude_list(dir, EXC_DIRS, NULL);
1002                strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1003                assert(stk->baselen == dir->basebuf.len);
1004
1005                /* Abort if the directory is excluded */
1006                if (stk->baselen) {
1007                        int dt = DT_DIR;
1008                        dir->basebuf.buf[stk->baselen - 1] = 0;
1009                        dir->exclude = last_exclude_matching_from_lists(dir,
1010                                dir->basebuf.buf, stk->baselen - 1,
1011                                dir->basebuf.buf + current, &dt);
1012                        dir->basebuf.buf[stk->baselen - 1] = '/';
1013                        if (dir->exclude &&
1014                            dir->exclude->flags & EXC_FLAG_NEGATIVE)
1015                                dir->exclude = NULL;
1016                        if (dir->exclude) {
1017                                dir->exclude_stack = stk;
1018                                return;
1019                        }
1020                }
1021
1022                /* Try to read per-directory file */
1023                hashclr(sha1_stat.sha1);
1024                sha1_stat.valid = 0;
1025                if (dir->exclude_per_dir &&
1026                    /*
1027                     * If we know that no files have been added in
1028                     * this directory (i.e. valid_cached_dir() has
1029                     * been executed and set untracked->valid) ..
1030                     */
1031                    (!untracked || !untracked->valid ||
1032                     /*
1033                      * .. and .gitignore does not exist before
1034                      * (i.e. null exclude_sha1 and skip_worktree is
1035                      * not set). Then we can skip loading .gitignore,
1036                      * which would result in ENOENT anyway.
1037                      * skip_worktree is taken care in read_directory()
1038                      */
1039                     !is_null_sha1(untracked->exclude_sha1))) {
1040                        /*
1041                         * dir->basebuf gets reused by the traversal, but we
1042                         * need fname to remain unchanged to ensure the src
1043                         * member of each struct exclude correctly
1044                         * back-references its source file.  Other invocations
1045                         * of add_exclude_list provide stable strings, so we
1046                         * strbuf_detach() and free() here in the caller.
1047                         */
1048                        struct strbuf sb = STRBUF_INIT;
1049                        strbuf_addbuf(&sb, &dir->basebuf);
1050                        strbuf_addstr(&sb, dir->exclude_per_dir);
1051                        el->src = strbuf_detach(&sb, NULL);
1052                        add_excludes(el->src, el->src, stk->baselen, el, 1,
1053                                     untracked ? &sha1_stat : NULL);
1054                }
1055                /*
1056                 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1057                 * will first be called in valid_cached_dir() then maybe many
1058                 * times more in last_exclude_matching(). When the cache is
1059                 * used, last_exclude_matching() will not be called and
1060                 * reading .gitignore content will be a waste.
1061                 *
1062                 * So when it's called by valid_cached_dir() and we can get
1063                 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1064                 * modified on work tree), we could delay reading the
1065                 * .gitignore content until we absolutely need it in
1066                 * last_exclude_matching(). Be careful about ignore rule
1067                 * order, though, if you do that.
1068                 */
1069                if (untracked &&
1070                    hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1071                        invalidate_gitignore(dir->untracked, untracked);
1072                        hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1073                }
1074                dir->exclude_stack = stk;
1075                current = stk->baselen;
1076        }
1077        strbuf_setlen(&dir->basebuf, baselen);
1078}
1079
1080/*
1081 * Loads the exclude lists for the directory containing pathname, then
1082 * scans all exclude lists to determine whether pathname is excluded.
1083 * Returns the exclude_list element which matched, or NULL for
1084 * undecided.
1085 */
1086struct exclude *last_exclude_matching(struct dir_struct *dir,
1087                                             const char *pathname,
1088                                             int *dtype_p)
1089{
1090        int pathlen = strlen(pathname);
1091        const char *basename = strrchr(pathname, '/');
1092        basename = (basename) ? basename+1 : pathname;
1093
1094        prep_exclude(dir, pathname, basename-pathname);
1095
1096        if (dir->exclude)
1097                return dir->exclude;
1098
1099        return last_exclude_matching_from_lists(dir, pathname, pathlen,
1100                        basename, dtype_p);
1101}
1102
1103/*
1104 * Loads the exclude lists for the directory containing pathname, then
1105 * scans all exclude lists to determine whether pathname is excluded.
1106 * Returns 1 if true, otherwise 0.
1107 */
1108int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1109{
1110        struct exclude *exclude =
1111                last_exclude_matching(dir, pathname, dtype_p);
1112        if (exclude)
1113                return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1114        return 0;
1115}
1116
1117static struct dir_entry *dir_entry_new(const char *pathname, int len)
1118{
1119        struct dir_entry *ent;
1120
1121        ent = xmalloc(sizeof(*ent) + len + 1);
1122        ent->len = len;
1123        memcpy(ent->name, pathname, len);
1124        ent->name[len] = 0;
1125        return ent;
1126}
1127
1128static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1129{
1130        if (cache_file_exists(pathname, len, ignore_case))
1131                return NULL;
1132
1133        ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1134        return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1135}
1136
1137struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1138{
1139        if (!cache_name_is_other(pathname, len))
1140                return NULL;
1141
1142        ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1143        return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1144}
1145
1146enum exist_status {
1147        index_nonexistent = 0,
1148        index_directory,
1149        index_gitdir
1150};
1151
1152/*
1153 * Do not use the alphabetically sorted index to look up
1154 * the directory name; instead, use the case insensitive
1155 * directory hash.
1156 */
1157static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1158{
1159        const struct cache_entry *ce = cache_dir_exists(dirname, len);
1160        unsigned char endchar;
1161
1162        if (!ce)
1163                return index_nonexistent;
1164        endchar = ce->name[len];
1165
1166        /*
1167         * The cache_entry structure returned will contain this dirname
1168         * and possibly additional path components.
1169         */
1170        if (endchar == '/')
1171                return index_directory;
1172
1173        /*
1174         * If there are no additional path components, then this cache_entry
1175         * represents a submodule.  Submodules, despite being directories,
1176         * are stored in the cache without a closing slash.
1177         */
1178        if (!endchar && S_ISGITLINK(ce->ce_mode))
1179                return index_gitdir;
1180
1181        /* This should never be hit, but it exists just in case. */
1182        return index_nonexistent;
1183}
1184
1185/*
1186 * The index sorts alphabetically by entry name, which
1187 * means that a gitlink sorts as '\0' at the end, while
1188 * a directory (which is defined not as an entry, but as
1189 * the files it contains) will sort with the '/' at the
1190 * end.
1191 */
1192static enum exist_status directory_exists_in_index(const char *dirname, int len)
1193{
1194        int pos;
1195
1196        if (ignore_case)
1197                return directory_exists_in_index_icase(dirname, len);
1198
1199        pos = cache_name_pos(dirname, len);
1200        if (pos < 0)
1201                pos = -pos-1;
1202        while (pos < active_nr) {
1203                const struct cache_entry *ce = active_cache[pos++];
1204                unsigned char endchar;
1205
1206                if (strncmp(ce->name, dirname, len))
1207                        break;
1208                endchar = ce->name[len];
1209                if (endchar > '/')
1210                        break;
1211                if (endchar == '/')
1212                        return index_directory;
1213                if (!endchar && S_ISGITLINK(ce->ce_mode))
1214                        return index_gitdir;
1215        }
1216        return index_nonexistent;
1217}
1218
1219/*
1220 * When we find a directory when traversing the filesystem, we
1221 * have three distinct cases:
1222 *
1223 *  - ignore it
1224 *  - see it as a directory
1225 *  - recurse into it
1226 *
1227 * and which one we choose depends on a combination of existing
1228 * git index contents and the flags passed into the directory
1229 * traversal routine.
1230 *
1231 * Case 1: If we *already* have entries in the index under that
1232 * directory name, we always recurse into the directory to see
1233 * all the files.
1234 *
1235 * Case 2: If we *already* have that directory name as a gitlink,
1236 * we always continue to see it as a gitlink, regardless of whether
1237 * there is an actual git directory there or not (it might not
1238 * be checked out as a subproject!)
1239 *
1240 * Case 3: if we didn't have it in the index previously, we
1241 * have a few sub-cases:
1242 *
1243 *  (a) if "show_other_directories" is true, we show it as
1244 *      just a directory, unless "hide_empty_directories" is
1245 *      also true, in which case we need to check if it contains any
1246 *      untracked and / or ignored files.
1247 *  (b) if it looks like a git directory, and we don't have
1248 *      'no_gitlinks' set we treat it as a gitlink, and show it
1249 *      as a directory.
1250 *  (c) otherwise, we recurse into it.
1251 */
1252static enum path_treatment treat_directory(struct dir_struct *dir,
1253        struct untracked_cache_dir *untracked,
1254        const char *dirname, int len, int exclude,
1255        const struct path_simplify *simplify)
1256{
1257        /* The "len-1" is to strip the final '/' */
1258        switch (directory_exists_in_index(dirname, len-1)) {
1259        case index_directory:
1260                return path_recurse;
1261
1262        case index_gitdir:
1263                return path_none;
1264
1265        case index_nonexistent:
1266                if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1267                        break;
1268                if (!(dir->flags & DIR_NO_GITLINKS)) {
1269                        unsigned char sha1[20];
1270                        if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1271                                return path_untracked;
1272                }
1273                return path_recurse;
1274        }
1275
1276        /* This is the "show_other_directories" case */
1277
1278        if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1279                return exclude ? path_excluded : path_untracked;
1280
1281        untracked = lookup_untracked(dir->untracked, untracked, dirname, len);
1282        return read_directory_recursive(dir, dirname, len,
1283                                        untracked, 1, simplify);
1284}
1285
1286/*
1287 * This is an inexact early pruning of any recursive directory
1288 * reading - if the path cannot possibly be in the pathspec,
1289 * return true, and we'll skip it early.
1290 */
1291static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1292{
1293        if (simplify) {
1294                for (;;) {
1295                        const char *match = simplify->path;
1296                        int len = simplify->len;
1297
1298                        if (!match)
1299                                break;
1300                        if (len > pathlen)
1301                                len = pathlen;
1302                        if (!memcmp(path, match, len))
1303                                return 0;
1304                        simplify++;
1305                }
1306                return 1;
1307        }
1308        return 0;
1309}
1310
1311/*
1312 * This function tells us whether an excluded path matches a
1313 * list of "interesting" pathspecs. That is, whether a path matched
1314 * by any of the pathspecs could possibly be ignored by excluding
1315 * the specified path. This can happen if:
1316 *
1317 *   1. the path is mentioned explicitly in the pathspec
1318 *
1319 *   2. the path is a directory prefix of some element in the
1320 *      pathspec
1321 */
1322static int exclude_matches_pathspec(const char *path, int len,
1323                const struct path_simplify *simplify)
1324{
1325        if (simplify) {
1326                for (; simplify->path; simplify++) {
1327                        if (len == simplify->len
1328                            && !memcmp(path, simplify->path, len))
1329                                return 1;
1330                        if (len < simplify->len
1331                            && simplify->path[len] == '/'
1332                            && !memcmp(path, simplify->path, len))
1333                                return 1;
1334                }
1335        }
1336        return 0;
1337}
1338
1339static int get_index_dtype(const char *path, int len)
1340{
1341        int pos;
1342        const struct cache_entry *ce;
1343
1344        ce = cache_file_exists(path, len, 0);
1345        if (ce) {
1346                if (!ce_uptodate(ce))
1347                        return DT_UNKNOWN;
1348                if (S_ISGITLINK(ce->ce_mode))
1349                        return DT_DIR;
1350                /*
1351                 * Nobody actually cares about the
1352                 * difference between DT_LNK and DT_REG
1353                 */
1354                return DT_REG;
1355        }
1356
1357        /* Try to look it up as a directory */
1358        pos = cache_name_pos(path, len);
1359        if (pos >= 0)
1360                return DT_UNKNOWN;
1361        pos = -pos-1;
1362        while (pos < active_nr) {
1363                ce = active_cache[pos++];
1364                if (strncmp(ce->name, path, len))
1365                        break;
1366                if (ce->name[len] > '/')
1367                        break;
1368                if (ce->name[len] < '/')
1369                        continue;
1370                if (!ce_uptodate(ce))
1371                        break;  /* continue? */
1372                return DT_DIR;
1373        }
1374        return DT_UNKNOWN;
1375}
1376
1377static int get_dtype(struct dirent *de, const char *path, int len)
1378{
1379        int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1380        struct stat st;
1381
1382        if (dtype != DT_UNKNOWN)
1383                return dtype;
1384        dtype = get_index_dtype(path, len);
1385        if (dtype != DT_UNKNOWN)
1386                return dtype;
1387        if (lstat(path, &st))
1388                return dtype;
1389        if (S_ISREG(st.st_mode))
1390                return DT_REG;
1391        if (S_ISDIR(st.st_mode))
1392                return DT_DIR;
1393        if (S_ISLNK(st.st_mode))
1394                return DT_LNK;
1395        return dtype;
1396}
1397
1398static enum path_treatment treat_one_path(struct dir_struct *dir,
1399                                          struct untracked_cache_dir *untracked,
1400                                          struct strbuf *path,
1401                                          const struct path_simplify *simplify,
1402                                          int dtype, struct dirent *de)
1403{
1404        int exclude;
1405        int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1406
1407        if (dtype == DT_UNKNOWN)
1408                dtype = get_dtype(de, path->buf, path->len);
1409
1410        /* Always exclude indexed files */
1411        if (dtype != DT_DIR && has_path_in_index)
1412                return path_none;
1413
1414        /*
1415         * When we are looking at a directory P in the working tree,
1416         * there are three cases:
1417         *
1418         * (1) P exists in the index.  Everything inside the directory P in
1419         * the working tree needs to go when P is checked out from the
1420         * index.
1421         *
1422         * (2) P does not exist in the index, but there is P/Q in the index.
1423         * We know P will stay a directory when we check out the contents
1424         * of the index, but we do not know yet if there is a directory
1425         * P/Q in the working tree to be killed, so we need to recurse.
1426         *
1427         * (3) P does not exist in the index, and there is no P/Q in the index
1428         * to require P to be a directory, either.  Only in this case, we
1429         * know that everything inside P will not be killed without
1430         * recursing.
1431         */
1432        if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1433            (dtype == DT_DIR) &&
1434            !has_path_in_index &&
1435            (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1436                return path_none;
1437
1438        exclude = is_excluded(dir, path->buf, &dtype);
1439
1440        /*
1441         * Excluded? If we don't explicitly want to show
1442         * ignored files, ignore it
1443         */
1444        if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1445                return path_excluded;
1446
1447        switch (dtype) {
1448        default:
1449                return path_none;
1450        case DT_DIR:
1451                strbuf_addch(path, '/');
1452                return treat_directory(dir, untracked, path->buf, path->len, exclude,
1453                        simplify);
1454        case DT_REG:
1455        case DT_LNK:
1456                return exclude ? path_excluded : path_untracked;
1457        }
1458}
1459
1460static enum path_treatment treat_path_fast(struct dir_struct *dir,
1461                                           struct untracked_cache_dir *untracked,
1462                                           struct cached_dir *cdir,
1463                                           struct strbuf *path,
1464                                           int baselen,
1465                                           const struct path_simplify *simplify)
1466{
1467        strbuf_setlen(path, baselen);
1468        if (!cdir->ucd) {
1469                strbuf_addstr(path, cdir->file);
1470                return path_untracked;
1471        }
1472        strbuf_addstr(path, cdir->ucd->name);
1473        /* treat_one_path() does this before it calls treat_directory() */
1474        if (path->buf[path->len - 1] != '/')
1475                strbuf_addch(path, '/');
1476        if (cdir->ucd->check_only)
1477                /*
1478                 * check_only is set as a result of treat_directory() getting
1479                 * to its bottom. Verify again the same set of directories
1480                 * with check_only set.
1481                 */
1482                return read_directory_recursive(dir, path->buf, path->len,
1483                                                cdir->ucd, 1, simplify);
1484        /*
1485         * We get path_recurse in the first run when
1486         * directory_exists_in_index() returns index_nonexistent. We
1487         * are sure that new changes in the index does not impact the
1488         * outcome. Return now.
1489         */
1490        return path_recurse;
1491}
1492
1493static enum path_treatment treat_path(struct dir_struct *dir,
1494                                      struct untracked_cache_dir *untracked,
1495                                      struct cached_dir *cdir,
1496                                      struct strbuf *path,
1497                                      int baselen,
1498                                      const struct path_simplify *simplify)
1499{
1500        int dtype;
1501        struct dirent *de = cdir->de;
1502
1503        if (!de)
1504                return treat_path_fast(dir, untracked, cdir, path,
1505                                       baselen, simplify);
1506        if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1507                return path_none;
1508        strbuf_setlen(path, baselen);
1509        strbuf_addstr(path, de->d_name);
1510        if (simplify_away(path->buf, path->len, simplify))
1511                return path_none;
1512
1513        dtype = DTYPE(de);
1514        return treat_one_path(dir, untracked, path, simplify, dtype, de);
1515}
1516
1517static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1518{
1519        if (!dir)
1520                return;
1521        ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1522                   dir->untracked_alloc);
1523        dir->untracked[dir->untracked_nr++] = xstrdup(name);
1524}
1525
1526static int valid_cached_dir(struct dir_struct *dir,
1527                            struct untracked_cache_dir *untracked,
1528                            struct strbuf *path,
1529                            int check_only)
1530{
1531        struct stat st;
1532
1533        if (!untracked)
1534                return 0;
1535
1536        if (stat(path->len ? path->buf : ".", &st)) {
1537                invalidate_directory(dir->untracked, untracked);
1538                memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1539                return 0;
1540        }
1541        if (!untracked->valid ||
1542            match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
1543                if (untracked->valid)
1544                        invalidate_directory(dir->untracked, untracked);
1545                fill_stat_data(&untracked->stat_data, &st);
1546                return 0;
1547        }
1548
1549        if (untracked->check_only != !!check_only) {
1550                invalidate_directory(dir->untracked, untracked);
1551                return 0;
1552        }
1553
1554        /*
1555         * prep_exclude will be called eventually on this directory,
1556         * but it's called much later in last_exclude_matching(). We
1557         * need it now to determine the validity of the cache for this
1558         * path. The next calls will be nearly no-op, the way
1559         * prep_exclude() is designed.
1560         */
1561        if (path->len && path->buf[path->len - 1] != '/') {
1562                strbuf_addch(path, '/');
1563                prep_exclude(dir, path->buf, path->len);
1564                strbuf_setlen(path, path->len - 1);
1565        } else
1566                prep_exclude(dir, path->buf, path->len);
1567
1568        /* hopefully prep_exclude() haven't invalidated this entry... */
1569        return untracked->valid;
1570}
1571
1572static int open_cached_dir(struct cached_dir *cdir,
1573                           struct dir_struct *dir,
1574                           struct untracked_cache_dir *untracked,
1575                           struct strbuf *path,
1576                           int check_only)
1577{
1578        memset(cdir, 0, sizeof(*cdir));
1579        cdir->untracked = untracked;
1580        if (valid_cached_dir(dir, untracked, path, check_only))
1581                return 0;
1582        cdir->fdir = opendir(path->len ? path->buf : ".");
1583        if (dir->untracked)
1584                dir->untracked->dir_opened++;
1585        if (!cdir->fdir)
1586                return -1;
1587        return 0;
1588}
1589
1590static int read_cached_dir(struct cached_dir *cdir)
1591{
1592        if (cdir->fdir) {
1593                cdir->de = readdir(cdir->fdir);
1594                if (!cdir->de)
1595                        return -1;
1596                return 0;
1597        }
1598        while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1599                struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1600                if (!d->recurse) {
1601                        cdir->nr_dirs++;
1602                        continue;
1603                }
1604                cdir->ucd = d;
1605                cdir->nr_dirs++;
1606                return 0;
1607        }
1608        cdir->ucd = NULL;
1609        if (cdir->nr_files < cdir->untracked->untracked_nr) {
1610                struct untracked_cache_dir *d = cdir->untracked;
1611                cdir->file = d->untracked[cdir->nr_files++];
1612                return 0;
1613        }
1614        return -1;
1615}
1616
1617static void close_cached_dir(struct cached_dir *cdir)
1618{
1619        if (cdir->fdir)
1620                closedir(cdir->fdir);
1621        /*
1622         * We have gone through this directory and found no untracked
1623         * entries. Mark it valid.
1624         */
1625        if (cdir->untracked) {
1626                cdir->untracked->valid = 1;
1627                cdir->untracked->recurse = 1;
1628        }
1629}
1630
1631/*
1632 * Read a directory tree. We currently ignore anything but
1633 * directories, regular files and symlinks. That's because git
1634 * doesn't handle them at all yet. Maybe that will change some
1635 * day.
1636 *
1637 * Also, we ignore the name ".git" (even if it is not a directory).
1638 * That likely will not change.
1639 *
1640 * Returns the most significant path_treatment value encountered in the scan.
1641 */
1642static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1643                                    const char *base, int baselen,
1644                                    struct untracked_cache_dir *untracked, int check_only,
1645                                    const struct path_simplify *simplify)
1646{
1647        struct cached_dir cdir;
1648        enum path_treatment state, subdir_state, dir_state = path_none;
1649        struct strbuf path = STRBUF_INIT;
1650
1651        strbuf_add(&path, base, baselen);
1652
1653        if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1654                goto out;
1655
1656        if (untracked)
1657                untracked->check_only = !!check_only;
1658
1659        while (!read_cached_dir(&cdir)) {
1660                /* check how the file or directory should be treated */
1661                state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);
1662
1663                if (state > dir_state)
1664                        dir_state = state;
1665
1666                /* recurse into subdir if instructed by treat_path */
1667                if (state == path_recurse) {
1668                        struct untracked_cache_dir *ud;
1669                        ud = lookup_untracked(dir->untracked, untracked,
1670                                              path.buf + baselen,
1671                                              path.len - baselen);
1672                        subdir_state =
1673                                read_directory_recursive(dir, path.buf, path.len,
1674                                                         ud, check_only, simplify);
1675                        if (subdir_state > dir_state)
1676                                dir_state = subdir_state;
1677                }
1678
1679                if (check_only) {
1680                        /* abort early if maximum state has been reached */
1681                        if (dir_state == path_untracked) {
1682                                if (cdir.fdir)
1683                                        add_untracked(untracked, path.buf + baselen);
1684                                break;
1685                        }
1686                        /* skip the dir_add_* part */
1687                        continue;
1688                }
1689
1690                /* add the path to the appropriate result list */
1691                switch (state) {
1692                case path_excluded:
1693                        if (dir->flags & DIR_SHOW_IGNORED)
1694                                dir_add_name(dir, path.buf, path.len);
1695                        else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1696                                ((dir->flags & DIR_COLLECT_IGNORED) &&
1697                                exclude_matches_pathspec(path.buf, path.len,
1698                                        simplify)))
1699                                dir_add_ignored(dir, path.buf, path.len);
1700                        break;
1701
1702                case path_untracked:
1703                        if (dir->flags & DIR_SHOW_IGNORED)
1704                                break;
1705                        dir_add_name(dir, path.buf, path.len);
1706                        if (cdir.fdir)
1707                                add_untracked(untracked, path.buf + baselen);
1708                        break;
1709
1710                default:
1711                        break;
1712                }
1713        }
1714        close_cached_dir(&cdir);
1715 out:
1716        strbuf_release(&path);
1717
1718        return dir_state;
1719}
1720
1721static int cmp_name(const void *p1, const void *p2)
1722{
1723        const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1724        const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1725
1726        return name_compare(e1->name, e1->len, e2->name, e2->len);
1727}
1728
1729static struct path_simplify *create_simplify(const char **pathspec)
1730{
1731        int nr, alloc = 0;
1732        struct path_simplify *simplify = NULL;
1733
1734        if (!pathspec)
1735                return NULL;
1736
1737        for (nr = 0 ; ; nr++) {
1738                const char *match;
1739                ALLOC_GROW(simplify, nr + 1, alloc);
1740                match = *pathspec++;
1741                if (!match)
1742                        break;
1743                simplify[nr].path = match;
1744                simplify[nr].len = simple_length(match);
1745        }
1746        simplify[nr].path = NULL;
1747        simplify[nr].len = 0;
1748        return simplify;
1749}
1750
1751static void free_simplify(struct path_simplify *simplify)
1752{
1753        free(simplify);
1754}
1755
1756static int treat_leading_path(struct dir_struct *dir,
1757                              const char *path, int len,
1758                              const struct path_simplify *simplify)
1759{
1760        struct strbuf sb = STRBUF_INIT;
1761        int baselen, rc = 0;
1762        const char *cp;
1763        int old_flags = dir->flags;
1764
1765        while (len && path[len - 1] == '/')
1766                len--;
1767        if (!len)
1768                return 1;
1769        baselen = 0;
1770        dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1771        while (1) {
1772                cp = path + baselen + !!baselen;
1773                cp = memchr(cp, '/', path + len - cp);
1774                if (!cp)
1775                        baselen = len;
1776                else
1777                        baselen = cp - path;
1778                strbuf_setlen(&sb, 0);
1779                strbuf_add(&sb, path, baselen);
1780                if (!is_directory(sb.buf))
1781                        break;
1782                if (simplify_away(sb.buf, sb.len, simplify))
1783                        break;
1784                if (treat_one_path(dir, NULL, &sb, simplify,
1785                                   DT_DIR, NULL) == path_none)
1786                        break; /* do not recurse into it */
1787                if (len <= baselen) {
1788                        rc = 1;
1789                        break; /* finished checking */
1790                }
1791        }
1792        strbuf_release(&sb);
1793        dir->flags = old_flags;
1794        return rc;
1795}
1796
1797static const char *get_ident_string(void)
1798{
1799        static struct strbuf sb = STRBUF_INIT;
1800        struct utsname uts;
1801
1802        if (sb.len)
1803                return sb.buf;
1804        if (uname(&uts))
1805                die_errno(_("failed to get kernel name and information"));
1806        strbuf_addf(&sb, "Location %s, system %s %s %s", get_git_work_tree(),
1807                    uts.sysname, uts.release, uts.version);
1808        return sb.buf;
1809}
1810
1811static int ident_in_untracked(const struct untracked_cache *uc)
1812{
1813        const char *end = uc->ident.buf + uc->ident.len;
1814        const char *p   = uc->ident.buf;
1815
1816        for (p = uc->ident.buf; p < end; p += strlen(p) + 1)
1817                if (!strcmp(p, get_ident_string()))
1818                        return 1;
1819        return 0;
1820}
1821
1822void add_untracked_ident(struct untracked_cache *uc)
1823{
1824        if (ident_in_untracked(uc))
1825                return;
1826        strbuf_addstr(&uc->ident, get_ident_string());
1827        /* this strbuf contains a list of strings, save NUL too */
1828        strbuf_addch(&uc->ident, 0);
1829}
1830
1831static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1832                                                      int base_len,
1833                                                      const struct pathspec *pathspec)
1834{
1835        struct untracked_cache_dir *root;
1836        int i;
1837
1838        if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
1839                return NULL;
1840
1841        /*
1842         * We only support $GIT_DIR/info/exclude and core.excludesfile
1843         * as the global ignore rule files. Any other additions
1844         * (e.g. from command line) invalidate the cache. This
1845         * condition also catches running setup_standard_excludes()
1846         * before setting dir->untracked!
1847         */
1848        if (dir->unmanaged_exclude_files)
1849                return NULL;
1850
1851        /*
1852         * Optimize for the main use case only: whole-tree git
1853         * status. More work involved in treat_leading_path() if we
1854         * use cache on just a subset of the worktree. pathspec
1855         * support could make the matter even worse.
1856         */
1857        if (base_len || (pathspec && pathspec->nr))
1858                return NULL;
1859
1860        /* Different set of flags may produce different results */
1861        if (dir->flags != dir->untracked->dir_flags ||
1862            /*
1863             * See treat_directory(), case index_nonexistent. Without
1864             * this flag, we may need to also cache .git file content
1865             * for the resolve_gitlink_ref() call, which we don't.
1866             */
1867            !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1868            /* We don't support collecting ignore files */
1869            (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1870                           DIR_COLLECT_IGNORED)))
1871                return NULL;
1872
1873        /*
1874         * If we use .gitignore in the cache and now you change it to
1875         * .gitexclude, everything will go wrong.
1876         */
1877        if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
1878            strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
1879                return NULL;
1880
1881        /*
1882         * EXC_CMDL is not considered in the cache. If people set it,
1883         * skip the cache.
1884         */
1885        if (dir->exclude_list_group[EXC_CMDL].nr)
1886                return NULL;
1887
1888        /*
1889         * An optimization in prep_exclude() does not play well with
1890         * CE_SKIP_WORKTREE. It's a rare case anyway, if a single
1891         * entry has that bit set, disable the whole untracked cache.
1892         */
1893        for (i = 0; i < active_nr; i++)
1894                if (ce_skip_worktree(active_cache[i]))
1895                        return NULL;
1896
1897        if (!ident_in_untracked(dir->untracked)) {
1898                warning(_("Untracked cache is disabled on this system."));
1899                return NULL;
1900        }
1901
1902        if (!dir->untracked->root) {
1903                const int len = sizeof(*dir->untracked->root);
1904                dir->untracked->root = xmalloc(len);
1905                memset(dir->untracked->root, 0, len);
1906        }
1907
1908        /* Validate $GIT_DIR/info/exclude and core.excludesfile */
1909        root = dir->untracked->root;
1910        if (hashcmp(dir->ss_info_exclude.sha1,
1911                    dir->untracked->ss_info_exclude.sha1)) {
1912                invalidate_gitignore(dir->untracked, root);
1913                dir->untracked->ss_info_exclude = dir->ss_info_exclude;
1914        }
1915        if (hashcmp(dir->ss_excludes_file.sha1,
1916                    dir->untracked->ss_excludes_file.sha1)) {
1917                invalidate_gitignore(dir->untracked, root);
1918                dir->untracked->ss_excludes_file = dir->ss_excludes_file;
1919        }
1920
1921        /* Make sure this directory is not dropped out at saving phase */
1922        root->recurse = 1;
1923        return root;
1924}
1925
1926int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1927{
1928        struct path_simplify *simplify;
1929        struct untracked_cache_dir *untracked;
1930
1931        /*
1932         * Check out create_simplify()
1933         */
1934        if (pathspec)
1935                GUARD_PATHSPEC(pathspec,
1936                               PATHSPEC_FROMTOP |
1937                               PATHSPEC_MAXDEPTH |
1938                               PATHSPEC_LITERAL |
1939                               PATHSPEC_GLOB |
1940                               PATHSPEC_ICASE |
1941                               PATHSPEC_EXCLUDE);
1942
1943        if (has_symlink_leading_path(path, len))
1944                return dir->nr;
1945
1946        /*
1947         * exclude patterns are treated like positive ones in
1948         * create_simplify. Usually exclude patterns should be a
1949         * subset of positive ones, which has no impacts on
1950         * create_simplify().
1951         */
1952        simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
1953        untracked = validate_untracked_cache(dir, len, pathspec);
1954        if (!untracked)
1955                /*
1956                 * make sure untracked cache code path is disabled,
1957                 * e.g. prep_exclude()
1958                 */
1959                dir->untracked = NULL;
1960        if (!len || treat_leading_path(dir, path, len, simplify))
1961                read_directory_recursive(dir, path, len, untracked, 0, simplify);
1962        free_simplify(simplify);
1963        qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1964        qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1965        if (dir->untracked) {
1966                static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
1967                trace_printf_key(&trace_untracked_stats,
1968                                 "node creation: %u\n"
1969                                 "gitignore invalidation: %u\n"
1970                                 "directory invalidation: %u\n"
1971                                 "opendir: %u\n",
1972                                 dir->untracked->dir_created,
1973                                 dir->untracked->gitignore_invalidated,
1974                                 dir->untracked->dir_invalidated,
1975                                 dir->untracked->dir_opened);
1976                if (dir->untracked == the_index.untracked &&
1977                    (dir->untracked->dir_opened ||
1978                     dir->untracked->gitignore_invalidated ||
1979                     dir->untracked->dir_invalidated))
1980                        the_index.cache_changed |= UNTRACKED_CHANGED;
1981                if (dir->untracked != the_index.untracked) {
1982                        free(dir->untracked);
1983                        dir->untracked = NULL;
1984                }
1985        }
1986        return dir->nr;
1987}
1988
1989int file_exists(const char *f)
1990{
1991        struct stat sb;
1992        return lstat(f, &sb) == 0;
1993}
1994
1995/*
1996 * Given two normalized paths (a trailing slash is ok), if subdir is
1997 * outside dir, return -1.  Otherwise return the offset in subdir that
1998 * can be used as relative path to dir.
1999 */
2000int dir_inside_of(const char *subdir, const char *dir)
2001{
2002        int offset = 0;
2003
2004        assert(dir && subdir && *dir && *subdir);
2005
2006        while (*dir && *subdir && *dir == *subdir) {
2007                dir++;
2008                subdir++;
2009                offset++;
2010        }
2011
2012        /* hel[p]/me vs hel[l]/yeah */
2013        if (*dir && *subdir)
2014                return -1;
2015
2016        if (!*subdir)
2017                return !*dir ? offset : -1; /* same dir */
2018
2019        /* foo/[b]ar vs foo/[] */
2020        if (is_dir_sep(dir[-1]))
2021                return is_dir_sep(subdir[-1]) ? offset : -1;
2022
2023        /* foo[/]bar vs foo[] */
2024        return is_dir_sep(*subdir) ? offset + 1 : -1;
2025}
2026
2027int is_inside_dir(const char *dir)
2028{
2029        char *cwd;
2030        int rc;
2031
2032        if (!dir)
2033                return 0;
2034
2035        cwd = xgetcwd();
2036        rc = (dir_inside_of(cwd, dir) >= 0);
2037        free(cwd);
2038        return rc;
2039}
2040
2041int is_empty_dir(const char *path)
2042{
2043        DIR *dir = opendir(path);
2044        struct dirent *e;
2045        int ret = 1;
2046
2047        if (!dir)
2048                return 0;
2049
2050        while ((e = readdir(dir)) != NULL)
2051                if (!is_dot_or_dotdot(e->d_name)) {
2052                        ret = 0;
2053                        break;
2054                }
2055
2056        closedir(dir);
2057        return ret;
2058}
2059
2060static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2061{
2062        DIR *dir;
2063        struct dirent *e;
2064        int ret = 0, original_len = path->len, len, kept_down = 0;
2065        int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2066        int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2067        unsigned char submodule_head[20];
2068
2069        if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2070            !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2071                /* Do not descend and nuke a nested git work tree. */
2072                if (kept_up)
2073                        *kept_up = 1;
2074                return 0;
2075        }
2076
2077        flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2078        dir = opendir(path->buf);
2079        if (!dir) {
2080                if (errno == ENOENT)
2081                        return keep_toplevel ? -1 : 0;
2082                else if (errno == EACCES && !keep_toplevel)
2083                        /*
2084                         * An empty dir could be removable even if it
2085                         * is unreadable:
2086                         */
2087                        return rmdir(path->buf);
2088                else
2089                        return -1;
2090        }
2091        if (path->buf[original_len - 1] != '/')
2092                strbuf_addch(path, '/');
2093
2094        len = path->len;
2095        while ((e = readdir(dir)) != NULL) {
2096                struct stat st;
2097                if (is_dot_or_dotdot(e->d_name))
2098                        continue;
2099
2100                strbuf_setlen(path, len);
2101                strbuf_addstr(path, e->d_name);
2102                if (lstat(path->buf, &st)) {
2103                        if (errno == ENOENT)
2104                                /*
2105                                 * file disappeared, which is what we
2106                                 * wanted anyway
2107                                 */
2108                                continue;
2109                        /* fall thru */
2110                } else if (S_ISDIR(st.st_mode)) {
2111                        if (!remove_dir_recurse(path, flag, &kept_down))
2112                                continue; /* happy */
2113                } else if (!only_empty &&
2114                           (!unlink(path->buf) || errno == ENOENT)) {
2115                        continue; /* happy, too */
2116                }
2117
2118                /* path too long, stat fails, or non-directory still exists */
2119                ret = -1;
2120                break;
2121        }
2122        closedir(dir);
2123
2124        strbuf_setlen(path, original_len);
2125        if (!ret && !keep_toplevel && !kept_down)
2126                ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2127        else if (kept_up)
2128                /*
2129                 * report the uplevel that it is not an error that we
2130                 * did not rmdir() our directory.
2131                 */
2132                *kept_up = !ret;
2133        return ret;
2134}
2135
2136int remove_dir_recursively(struct strbuf *path, int flag)
2137{
2138        return remove_dir_recurse(path, flag, NULL);
2139}
2140
2141void setup_standard_excludes(struct dir_struct *dir)
2142{
2143        const char *path;
2144        char *xdg_path;
2145
2146        dir->exclude_per_dir = ".gitignore";
2147        path = git_path("info/exclude");
2148        if (!excludes_file) {
2149                home_config_paths(NULL, &xdg_path, "ignore");
2150                excludes_file = xdg_path;
2151        }
2152        if (!access_or_warn(path, R_OK, 0))
2153                add_excludes_from_file_1(dir, path,
2154                                         dir->untracked ? &dir->ss_info_exclude : NULL);
2155        if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2156                add_excludes_from_file_1(dir, excludes_file,
2157                                         dir->untracked ? &dir->ss_excludes_file : NULL);
2158}
2159
2160int remove_path(const char *name)
2161{
2162        char *slash;
2163
2164        if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
2165                return -1;
2166
2167        slash = strrchr(name, '/');
2168        if (slash) {
2169                char *dirs = xstrdup(name);
2170                slash = dirs + (slash - name);
2171                do {
2172                        *slash = '\0';
2173                } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2174                free(dirs);
2175        }
2176        return 0;
2177}
2178
2179/*
2180 * Frees memory within dir which was allocated for exclude lists and
2181 * the exclude_stack.  Does not free dir itself.
2182 */
2183void clear_directory(struct dir_struct *dir)
2184{
2185        int i, j;
2186        struct exclude_list_group *group;
2187        struct exclude_list *el;
2188        struct exclude_stack *stk;
2189
2190        for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2191                group = &dir->exclude_list_group[i];
2192                for (j = 0; j < group->nr; j++) {
2193                        el = &group->el[j];
2194                        if (i == EXC_DIRS)
2195                                free((char *)el->src);
2196                        clear_exclude_list(el);
2197                }
2198                free(group->el);
2199        }
2200
2201        stk = dir->exclude_stack;
2202        while (stk) {
2203                struct exclude_stack *prev = stk->prev;
2204                free(stk);
2205                stk = prev;
2206        }
2207        strbuf_release(&dir->basebuf);
2208}
2209
2210struct ondisk_untracked_cache {
2211        struct stat_data info_exclude_stat;
2212        struct stat_data excludes_file_stat;
2213        uint32_t dir_flags;
2214        unsigned char info_exclude_sha1[20];
2215        unsigned char excludes_file_sha1[20];
2216        char exclude_per_dir[FLEX_ARRAY];
2217};
2218
2219#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2220
2221struct write_data {
2222        int index;         /* number of written untracked_cache_dir */
2223        struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2224        struct ewah_bitmap *valid;      /* from untracked_cache_dir */
2225        struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2226        struct strbuf out;
2227        struct strbuf sb_stat;
2228        struct strbuf sb_sha1;
2229};
2230
2231static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2232{
2233        to->sd_ctime.sec  = htonl(from->sd_ctime.sec);
2234        to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2235        to->sd_mtime.sec  = htonl(from->sd_mtime.sec);
2236        to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2237        to->sd_dev        = htonl(from->sd_dev);
2238        to->sd_ino        = htonl(from->sd_ino);
2239        to->sd_uid        = htonl(from->sd_uid);
2240        to->sd_gid        = htonl(from->sd_gid);
2241        to->sd_size       = htonl(from->sd_size);
2242}
2243
2244static void write_one_dir(struct untracked_cache_dir *untracked,
2245                          struct write_data *wd)
2246{
2247        struct stat_data stat_data;
2248        struct strbuf *out = &wd->out;
2249        unsigned char intbuf[16];
2250        unsigned int intlen, value;
2251        int i = wd->index++;
2252
2253        /*
2254         * untracked_nr should be reset whenever valid is clear, but
2255         * for safety..
2256         */
2257        if (!untracked->valid) {
2258                untracked->untracked_nr = 0;
2259                untracked->check_only = 0;
2260        }
2261
2262        if (untracked->check_only)
2263                ewah_set(wd->check_only, i);
2264        if (untracked->valid) {
2265                ewah_set(wd->valid, i);
2266                stat_data_to_disk(&stat_data, &untracked->stat_data);
2267                strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2268        }
2269        if (!is_null_sha1(untracked->exclude_sha1)) {
2270                ewah_set(wd->sha1_valid, i);
2271                strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2272        }
2273
2274        intlen = encode_varint(untracked->untracked_nr, intbuf);
2275        strbuf_add(out, intbuf, intlen);
2276
2277        /* skip non-recurse directories */
2278        for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2279                if (untracked->dirs[i]->recurse)
2280                        value++;
2281        intlen = encode_varint(value, intbuf);
2282        strbuf_add(out, intbuf, intlen);
2283
2284        strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2285
2286        for (i = 0; i < untracked->untracked_nr; i++)
2287                strbuf_add(out, untracked->untracked[i],
2288                           strlen(untracked->untracked[i]) + 1);
2289
2290        for (i = 0; i < untracked->dirs_nr; i++)
2291                if (untracked->dirs[i]->recurse)
2292                        write_one_dir(untracked->dirs[i], wd);
2293}
2294
2295void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2296{
2297        struct ondisk_untracked_cache *ouc;
2298        struct write_data wd;
2299        unsigned char varbuf[16];
2300        int len = 0, varint_len;
2301        if (untracked->exclude_per_dir)
2302                len = strlen(untracked->exclude_per_dir);
2303        ouc = xmalloc(sizeof(*ouc) + len + 1);
2304        stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2305        stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2306        hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2307        hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2308        ouc->dir_flags = htonl(untracked->dir_flags);
2309        memcpy(ouc->exclude_per_dir, untracked->exclude_per_dir, len + 1);
2310
2311        varint_len = encode_varint(untracked->ident.len, varbuf);
2312        strbuf_add(out, varbuf, varint_len);
2313        strbuf_add(out, untracked->ident.buf, untracked->ident.len);
2314
2315        strbuf_add(out, ouc, ouc_size(len));
2316        free(ouc);
2317        ouc = NULL;
2318
2319        if (!untracked->root) {
2320                varint_len = encode_varint(0, varbuf);
2321                strbuf_add(out, varbuf, varint_len);
2322                return;
2323        }
2324
2325        wd.index      = 0;
2326        wd.check_only = ewah_new();
2327        wd.valid      = ewah_new();
2328        wd.sha1_valid = ewah_new();
2329        strbuf_init(&wd.out, 1024);
2330        strbuf_init(&wd.sb_stat, 1024);
2331        strbuf_init(&wd.sb_sha1, 1024);
2332        write_one_dir(untracked->root, &wd);
2333
2334        varint_len = encode_varint(wd.index, varbuf);
2335        strbuf_add(out, varbuf, varint_len);
2336        strbuf_addbuf(out, &wd.out);
2337        ewah_serialize_strbuf(wd.valid, out);
2338        ewah_serialize_strbuf(wd.check_only, out);
2339        ewah_serialize_strbuf(wd.sha1_valid, out);
2340        strbuf_addbuf(out, &wd.sb_stat);
2341        strbuf_addbuf(out, &wd.sb_sha1);
2342        strbuf_addch(out, '\0'); /* safe guard for string lists */
2343
2344        ewah_free(wd.valid);
2345        ewah_free(wd.check_only);
2346        ewah_free(wd.sha1_valid);
2347        strbuf_release(&wd.out);
2348        strbuf_release(&wd.sb_stat);
2349        strbuf_release(&wd.sb_sha1);
2350}
2351
2352static void free_untracked(struct untracked_cache_dir *ucd)
2353{
2354        int i;
2355        if (!ucd)
2356                return;
2357        for (i = 0; i < ucd->dirs_nr; i++)
2358                free_untracked(ucd->dirs[i]);
2359        for (i = 0; i < ucd->untracked_nr; i++)
2360                free(ucd->untracked[i]);
2361        free(ucd->untracked);
2362        free(ucd->dirs);
2363        free(ucd);
2364}
2365
2366void free_untracked_cache(struct untracked_cache *uc)
2367{
2368        if (uc)
2369                free_untracked(uc->root);
2370        free(uc);
2371}
2372
2373struct read_data {
2374        int index;
2375        struct untracked_cache_dir **ucd;
2376        struct ewah_bitmap *check_only;
2377        struct ewah_bitmap *valid;
2378        struct ewah_bitmap *sha1_valid;
2379        const unsigned char *data;
2380        const unsigned char *end;
2381};
2382
2383static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2384{
2385        to->sd_ctime.sec  = get_be32(&from->sd_ctime.sec);
2386        to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2387        to->sd_mtime.sec  = get_be32(&from->sd_mtime.sec);
2388        to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2389        to->sd_dev        = get_be32(&from->sd_dev);
2390        to->sd_ino        = get_be32(&from->sd_ino);
2391        to->sd_uid        = get_be32(&from->sd_uid);
2392        to->sd_gid        = get_be32(&from->sd_gid);
2393        to->sd_size       = get_be32(&from->sd_size);
2394}
2395
2396static int read_one_dir(struct untracked_cache_dir **untracked_,
2397                        struct read_data *rd)
2398{
2399        struct untracked_cache_dir ud, *untracked;
2400        const unsigned char *next, *data = rd->data, *end = rd->end;
2401        unsigned int value;
2402        int i, len;
2403
2404        memset(&ud, 0, sizeof(ud));
2405
2406        next = data;
2407        value = decode_varint(&next);
2408        if (next > end)
2409                return -1;
2410        ud.recurse         = 1;
2411        ud.untracked_alloc = value;
2412        ud.untracked_nr    = value;
2413        if (ud.untracked_nr)
2414                ud.untracked = xmalloc(sizeof(*ud.untracked) * ud.untracked_nr);
2415        data = next;
2416
2417        next = data;
2418        ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2419        if (next > end)
2420                return -1;
2421        ud.dirs = xmalloc(sizeof(*ud.dirs) * ud.dirs_nr);
2422        data = next;
2423
2424        len = strlen((const char *)data);
2425        next = data + len + 1;
2426        if (next > rd->end)
2427                return -1;
2428        *untracked_ = untracked = xmalloc(sizeof(*untracked) + len);
2429        memcpy(untracked, &ud, sizeof(ud));
2430        memcpy(untracked->name, data, len + 1);
2431        data = next;
2432
2433        for (i = 0; i < untracked->untracked_nr; i++) {
2434                len = strlen((const char *)data);
2435                next = data + len + 1;
2436                if (next > rd->end)
2437                        return -1;
2438                untracked->untracked[i] = xstrdup((const char*)data);
2439                data = next;
2440        }
2441
2442        rd->ucd[rd->index++] = untracked;
2443        rd->data = data;
2444
2445        for (i = 0; i < untracked->dirs_nr; i++) {
2446                len = read_one_dir(untracked->dirs + i, rd);
2447                if (len < 0)
2448                        return -1;
2449        }
2450        return 0;
2451}
2452
2453static void set_check_only(size_t pos, void *cb)
2454{
2455        struct read_data *rd = cb;
2456        struct untracked_cache_dir *ud = rd->ucd[pos];
2457        ud->check_only = 1;
2458}
2459
2460static void read_stat(size_t pos, void *cb)
2461{
2462        struct read_data *rd = cb;
2463        struct untracked_cache_dir *ud = rd->ucd[pos];
2464        if (rd->data + sizeof(struct stat_data) > rd->end) {
2465                rd->data = rd->end + 1;
2466                return;
2467        }
2468        stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2469        rd->data += sizeof(struct stat_data);
2470        ud->valid = 1;
2471}
2472
2473static void read_sha1(size_t pos, void *cb)
2474{
2475        struct read_data *rd = cb;
2476        struct untracked_cache_dir *ud = rd->ucd[pos];
2477        if (rd->data + 20 > rd->end) {
2478                rd->data = rd->end + 1;
2479                return;
2480        }
2481        hashcpy(ud->exclude_sha1, rd->data);
2482        rd->data += 20;
2483}
2484
2485static void load_sha1_stat(struct sha1_stat *sha1_stat,
2486                           const struct stat_data *stat,
2487                           const unsigned char *sha1)
2488{
2489        stat_data_from_disk(&sha1_stat->stat, stat);
2490        hashcpy(sha1_stat->sha1, sha1);
2491        sha1_stat->valid = 1;
2492}
2493
2494struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2495{
2496        const struct ondisk_untracked_cache *ouc;
2497        struct untracked_cache *uc;
2498        struct read_data rd;
2499        const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2500        const char *ident;
2501        int ident_len, len;
2502
2503        if (sz <= 1 || end[-1] != '\0')
2504                return NULL;
2505        end--;
2506
2507        ident_len = decode_varint(&next);
2508        if (next + ident_len > end)
2509                return NULL;
2510        ident = (const char *)next;
2511        next += ident_len;
2512
2513        ouc = (const struct ondisk_untracked_cache *)next;
2514        if (next + ouc_size(0) > end)
2515                return NULL;
2516
2517        uc = xcalloc(1, sizeof(*uc));
2518        strbuf_init(&uc->ident, ident_len);
2519        strbuf_add(&uc->ident, ident, ident_len);
2520        load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2521                       ouc->info_exclude_sha1);
2522        load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2523                       ouc->excludes_file_sha1);
2524        uc->dir_flags = get_be32(&ouc->dir_flags);
2525        uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2526        /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2527        next += ouc_size(strlen(ouc->exclude_per_dir));
2528        if (next >= end)
2529                goto done2;
2530
2531        len = decode_varint(&next);
2532        if (next > end || len == 0)
2533                goto done2;
2534
2535        rd.valid      = ewah_new();
2536        rd.check_only = ewah_new();
2537        rd.sha1_valid = ewah_new();
2538        rd.data       = next;
2539        rd.end        = end;
2540        rd.index      = 0;
2541        rd.ucd        = xmalloc(sizeof(*rd.ucd) * len);
2542
2543        if (read_one_dir(&uc->root, &rd) || rd.index != len)
2544                goto done;
2545
2546        next = rd.data;
2547        len = ewah_read_mmap(rd.valid, next, end - next);
2548        if (len < 0)
2549                goto done;
2550
2551        next += len;
2552        len = ewah_read_mmap(rd.check_only, next, end - next);
2553        if (len < 0)
2554                goto done;
2555
2556        next += len;
2557        len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2558        if (len < 0)
2559                goto done;
2560
2561        ewah_each_bit(rd.check_only, set_check_only, &rd);
2562        rd.data = next + len;
2563        ewah_each_bit(rd.valid, read_stat, &rd);
2564        ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2565        next = rd.data;
2566
2567done:
2568        free(rd.ucd);
2569        ewah_free(rd.valid);
2570        ewah_free(rd.check_only);
2571        ewah_free(rd.sha1_valid);
2572done2:
2573        if (next != end) {
2574                free_untracked_cache(uc);
2575                uc = NULL;
2576        }
2577        return uc;
2578}
2579
2580void untracked_cache_invalidate_path(struct index_state *istate,
2581                                     const char *path)
2582{
2583        const char *sep;
2584        struct untracked_cache_dir *d;
2585        if (!istate->untracked || !istate->untracked->root)
2586                return;
2587        sep = strrchr(path, '/');
2588        if (sep)
2589                d = lookup_untracked(istate->untracked,
2590                                     istate->untracked->root,
2591                                     path, sep - path);
2592        else
2593                d = istate->untracked->root;
2594        istate->untracked->dir_invalidated++;
2595        d->valid = 0;
2596        d->untracked_nr = 0;
2597}
2598
2599void untracked_cache_remove_from_index(struct index_state *istate,
2600                                       const char *path)
2601{
2602        untracked_cache_invalidate_path(istate, path);
2603}
2604
2605void untracked_cache_add_to_index(struct index_state *istate,
2606                                  const char *path)
2607{
2608        untracked_cache_invalidate_path(istate, path);
2609}