0e411c08021d5bf3e1ceb799e0af4a7949d5dc93
   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
  16struct path_simplify {
  17        int len;
  18        const char *path;
  19};
  20
  21/*
  22 * Tells read_directory_recursive how a file or directory should be treated.
  23 * Values are ordered by significance, e.g. if a directory contains both
  24 * excluded and untracked files, it is listed as untracked because
  25 * path_untracked > path_excluded.
  26 */
  27enum path_treatment {
  28        path_none = 0,
  29        path_recurse,
  30        path_excluded,
  31        path_untracked
  32};
  33
  34static enum path_treatment read_directory_recursive(struct dir_struct *dir,
  35        const char *path, int len, struct untracked_cache_dir *untracked,
  36        int check_only, const struct path_simplify *simplify);
  37static int get_dtype(struct dirent *de, const char *path, int len);
  38
  39/* helper string functions with support for the ignore_case flag */
  40int strcmp_icase(const char *a, const char *b)
  41{
  42        return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
  43}
  44
  45int strncmp_icase(const char *a, const char *b, size_t count)
  46{
  47        return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
  48}
  49
  50int fnmatch_icase(const char *pattern, const char *string, int flags)
  51{
  52        return wildmatch(pattern, string,
  53                         flags | (ignore_case ? WM_CASEFOLD : 0),
  54                         NULL);
  55}
  56
  57int git_fnmatch(const struct pathspec_item *item,
  58                const char *pattern, const char *string,
  59                int prefix)
  60{
  61        if (prefix > 0) {
  62                if (ps_strncmp(item, pattern, string, prefix))
  63                        return WM_NOMATCH;
  64                pattern += prefix;
  65                string += prefix;
  66        }
  67        if (item->flags & PATHSPEC_ONESTAR) {
  68                int pattern_len = strlen(++pattern);
  69                int string_len = strlen(string);
  70                return string_len < pattern_len ||
  71                        ps_strcmp(item, pattern,
  72                                  string + string_len - pattern_len);
  73        }
  74        if (item->magic & PATHSPEC_GLOB)
  75                return wildmatch(pattern, string,
  76                                 WM_PATHNAME |
  77                                 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
  78                                 NULL);
  79        else
  80                /* wildmatch has not learned no FNM_PATHNAME mode yet */
  81                return wildmatch(pattern, string,
  82                                 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
  83                                 NULL);
  84}
  85
  86static int fnmatch_icase_mem(const char *pattern, int patternlen,
  87                             const char *string, int stringlen,
  88                             int flags)
  89{
  90        int match_status;
  91        struct strbuf pat_buf = STRBUF_INIT;
  92        struct strbuf str_buf = STRBUF_INIT;
  93        const char *use_pat = pattern;
  94        const char *use_str = string;
  95
  96        if (pattern[patternlen]) {
  97                strbuf_add(&pat_buf, pattern, patternlen);
  98                use_pat = pat_buf.buf;
  99        }
 100        if (string[stringlen]) {
 101                strbuf_add(&str_buf, string, stringlen);
 102                use_str = str_buf.buf;
 103        }
 104
 105        if (ignore_case)
 106                flags |= WM_CASEFOLD;
 107        match_status = wildmatch(use_pat, use_str, flags, NULL);
 108
 109        strbuf_release(&pat_buf);
 110        strbuf_release(&str_buf);
 111
 112        return match_status;
 113}
 114
 115static size_t common_prefix_len(const struct pathspec *pathspec)
 116{
 117        int n;
 118        size_t max = 0;
 119
 120        /*
 121         * ":(icase)path" is treated as a pathspec full of
 122         * wildcard. In other words, only prefix is considered common
 123         * prefix. If the pathspec is abc/foo abc/bar, running in
 124         * subdir xyz, the common prefix is still xyz, not xuz/abc as
 125         * in non-:(icase).
 126         */
 127        GUARD_PATHSPEC(pathspec,
 128                       PATHSPEC_FROMTOP |
 129                       PATHSPEC_MAXDEPTH |
 130                       PATHSPEC_LITERAL |
 131                       PATHSPEC_GLOB |
 132                       PATHSPEC_ICASE |
 133                       PATHSPEC_EXCLUDE);
 134
 135        for (n = 0; n < pathspec->nr; n++) {
 136                size_t i = 0, len = 0, item_len;
 137                if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
 138                        continue;
 139                if (pathspec->items[n].magic & PATHSPEC_ICASE)
 140                        item_len = pathspec->items[n].prefix;
 141                else
 142                        item_len = pathspec->items[n].nowildcard_len;
 143                while (i < item_len && (n == 0 || i < max)) {
 144                        char c = pathspec->items[n].match[i];
 145                        if (c != pathspec->items[0].match[i])
 146                                break;
 147                        if (c == '/')
 148                                len = i + 1;
 149                        i++;
 150                }
 151                if (n == 0 || len < max) {
 152                        max = len;
 153                        if (!max)
 154                                break;
 155                }
 156        }
 157        return max;
 158}
 159
 160/*
 161 * Returns a copy of the longest leading path common among all
 162 * pathspecs.
 163 */
 164char *common_prefix(const struct pathspec *pathspec)
 165{
 166        unsigned long len = common_prefix_len(pathspec);
 167
 168        return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
 169}
 170
 171int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
 172{
 173        size_t len;
 174
 175        /*
 176         * Calculate common prefix for the pathspec, and
 177         * use that to optimize the directory walk
 178         */
 179        len = common_prefix_len(pathspec);
 180
 181        /* Read the directory and prune it */
 182        read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec);
 183        return len;
 184}
 185
 186int within_depth(const char *name, int namelen,
 187                        int depth, int max_depth)
 188{
 189        const char *cp = name, *cpe = name + namelen;
 190
 191        while (cp < cpe) {
 192                if (*cp++ != '/')
 193                        continue;
 194                depth++;
 195                if (depth > max_depth)
 196                        return 0;
 197        }
 198        return 1;
 199}
 200
 201#define DO_MATCH_EXCLUDE   1
 202#define DO_MATCH_DIRECTORY 2
 203
 204/*
 205 * Does 'match' match the given name?
 206 * A match is found if
 207 *
 208 * (1) the 'match' string is leading directory of 'name', or
 209 * (2) the 'match' string is a wildcard and matches 'name', or
 210 * (3) the 'match' string is exactly the same as 'name'.
 211 *
 212 * and the return value tells which case it was.
 213 *
 214 * It returns 0 when there is no match.
 215 */
 216static int match_pathspec_item(const struct pathspec_item *item, int prefix,
 217                               const char *name, int namelen, unsigned flags)
 218{
 219        /* name/namelen has prefix cut off by caller */
 220        const char *match = item->match + prefix;
 221        int matchlen = item->len - prefix;
 222
 223        /*
 224         * The normal call pattern is:
 225         * 1. prefix = common_prefix_len(ps);
 226         * 2. prune something, or fill_directory
 227         * 3. match_pathspec()
 228         *
 229         * 'prefix' at #1 may be shorter than the command's prefix and
 230         * it's ok for #2 to match extra files. Those extras will be
 231         * trimmed at #3.
 232         *
 233         * Suppose the pathspec is 'foo' and '../bar' running from
 234         * subdir 'xyz'. The common prefix at #1 will be empty, thanks
 235         * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
 236         * user does not want XYZ/foo, only the "foo" part should be
 237         * case-insensitive. We need to filter out XYZ/foo here. In
 238         * other words, we do not trust the caller on comparing the
 239         * prefix part when :(icase) is involved. We do exact
 240         * comparison ourselves.
 241         *
 242         * Normally the caller (common_prefix_len() in fact) does
 243         * _exact_ matching on name[-prefix+1..-1] and we do not need
 244         * to check that part. Be defensive and check it anyway, in
 245         * case common_prefix_len is changed, or a new caller is
 246         * introduced that does not use common_prefix_len.
 247         *
 248         * If the penalty turns out too high when prefix is really
 249         * long, maybe change it to
 250         * strncmp(match, name, item->prefix - prefix)
 251         */
 252        if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
 253            strncmp(item->match, name - prefix, item->prefix))
 254                return 0;
 255
 256        /* If the match was just the prefix, we matched */
 257        if (!*match)
 258                return MATCHED_RECURSIVELY;
 259
 260        if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
 261                if (matchlen == namelen)
 262                        return MATCHED_EXACTLY;
 263
 264                if (match[matchlen-1] == '/' || name[matchlen] == '/')
 265                        return MATCHED_RECURSIVELY;
 266        } else if ((flags & DO_MATCH_DIRECTORY) &&
 267                   match[matchlen - 1] == '/' &&
 268                   namelen == matchlen - 1 &&
 269                   !ps_strncmp(item, match, name, namelen))
 270                return MATCHED_EXACTLY;
 271
 272        if (item->nowildcard_len < item->len &&
 273            !git_fnmatch(item, match, name,
 274                         item->nowildcard_len - prefix))
 275                return MATCHED_FNMATCH;
 276
 277        return 0;
 278}
 279
 280/*
 281 * Given a name and a list of pathspecs, returns the nature of the
 282 * closest (i.e. most specific) match of the name to any of the
 283 * pathspecs.
 284 *
 285 * The caller typically calls this multiple times with the same
 286 * pathspec and seen[] array but with different name/namelen
 287 * (e.g. entries from the index) and is interested in seeing if and
 288 * how each pathspec matches all the names it calls this function
 289 * with.  A mark is left in the seen[] array for each pathspec element
 290 * indicating the closest type of match that element achieved, so if
 291 * seen[n] remains zero after multiple invocations, that means the nth
 292 * pathspec did not match any names, which could indicate that the
 293 * user mistyped the nth pathspec.
 294 */
 295static int do_match_pathspec(const struct pathspec *ps,
 296                             const char *name, int namelen,
 297                             int prefix, char *seen,
 298                             unsigned flags)
 299{
 300        int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
 301
 302        GUARD_PATHSPEC(ps,
 303                       PATHSPEC_FROMTOP |
 304                       PATHSPEC_MAXDEPTH |
 305                       PATHSPEC_LITERAL |
 306                       PATHSPEC_GLOB |
 307                       PATHSPEC_ICASE |
 308                       PATHSPEC_EXCLUDE);
 309
 310        if (!ps->nr) {
 311                if (!ps->recursive ||
 312                    !(ps->magic & PATHSPEC_MAXDEPTH) ||
 313                    ps->max_depth == -1)
 314                        return MATCHED_RECURSIVELY;
 315
 316                if (within_depth(name, namelen, 0, ps->max_depth))
 317                        return MATCHED_EXACTLY;
 318                else
 319                        return 0;
 320        }
 321
 322        name += prefix;
 323        namelen -= prefix;
 324
 325        for (i = ps->nr - 1; i >= 0; i--) {
 326                int how;
 327
 328                if ((!exclude &&   ps->items[i].magic & PATHSPEC_EXCLUDE) ||
 329                    ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
 330                        continue;
 331
 332                if (seen && seen[i] == MATCHED_EXACTLY)
 333                        continue;
 334                /*
 335                 * Make exclude patterns optional and never report
 336                 * "pathspec ':(exclude)foo' matches no files"
 337                 */
 338                if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
 339                        seen[i] = MATCHED_FNMATCH;
 340                how = match_pathspec_item(ps->items+i, prefix, name,
 341                                          namelen, flags);
 342                if (ps->recursive &&
 343                    (ps->magic & PATHSPEC_MAXDEPTH) &&
 344                    ps->max_depth != -1 &&
 345                    how && how != MATCHED_FNMATCH) {
 346                        int len = ps->items[i].len;
 347                        if (name[len] == '/')
 348                                len++;
 349                        if (within_depth(name+len, namelen-len, 0, ps->max_depth))
 350                                how = MATCHED_EXACTLY;
 351                        else
 352                                how = 0;
 353                }
 354                if (how) {
 355                        if (retval < how)
 356                                retval = how;
 357                        if (seen && seen[i] < how)
 358                                seen[i] = how;
 359                }
 360        }
 361        return retval;
 362}
 363
 364int match_pathspec(const struct pathspec *ps,
 365                   const char *name, int namelen,
 366                   int prefix, char *seen, int is_dir)
 367{
 368        int positive, negative;
 369        unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
 370        positive = do_match_pathspec(ps, name, namelen,
 371                                     prefix, seen, flags);
 372        if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
 373                return positive;
 374        negative = do_match_pathspec(ps, name, namelen,
 375                                     prefix, seen,
 376                                     flags | DO_MATCH_EXCLUDE);
 377        return negative ? 0 : positive;
 378}
 379
 380/*
 381 * Return the length of the "simple" part of a path match limiter.
 382 */
 383int simple_length(const char *match)
 384{
 385        int len = -1;
 386
 387        for (;;) {
 388                unsigned char c = *match++;
 389                len++;
 390                if (c == '\0' || is_glob_special(c))
 391                        return len;
 392        }
 393}
 394
 395int no_wildcard(const char *string)
 396{
 397        return string[simple_length(string)] == '\0';
 398}
 399
 400void parse_exclude_pattern(const char **pattern,
 401                           int *patternlen,
 402                           int *flags,
 403                           int *nowildcardlen)
 404{
 405        const char *p = *pattern;
 406        size_t i, len;
 407
 408        *flags = 0;
 409        if (*p == '!') {
 410                *flags |= EXC_FLAG_NEGATIVE;
 411                p++;
 412        }
 413        len = strlen(p);
 414        if (len && p[len - 1] == '/') {
 415                len--;
 416                *flags |= EXC_FLAG_MUSTBEDIR;
 417        }
 418        for (i = 0; i < len; i++) {
 419                if (p[i] == '/')
 420                        break;
 421        }
 422        if (i == len)
 423                *flags |= EXC_FLAG_NODIR;
 424        *nowildcardlen = simple_length(p);
 425        /*
 426         * we should have excluded the trailing slash from 'p' too,
 427         * but that's one more allocation. Instead just make sure
 428         * nowildcardlen does not exceed real patternlen
 429         */
 430        if (*nowildcardlen > len)
 431                *nowildcardlen = len;
 432        if (*p == '*' && no_wildcard(p + 1))
 433                *flags |= EXC_FLAG_ENDSWITH;
 434        *pattern = p;
 435        *patternlen = len;
 436}
 437
 438void add_exclude(const char *string, const char *base,
 439                 int baselen, struct exclude_list *el, int srcpos)
 440{
 441        struct exclude *x;
 442        int patternlen;
 443        int flags;
 444        int nowildcardlen;
 445
 446        parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
 447        if (flags & EXC_FLAG_MUSTBEDIR) {
 448                char *s;
 449                x = xmalloc(sizeof(*x) + patternlen + 1);
 450                s = (char *)(x+1);
 451                memcpy(s, string, patternlen);
 452                s[patternlen] = '\0';
 453                x->pattern = s;
 454        } else {
 455                x = xmalloc(sizeof(*x));
 456                x->pattern = string;
 457        }
 458        x->patternlen = patternlen;
 459        x->nowildcardlen = nowildcardlen;
 460        x->base = base;
 461        x->baselen = baselen;
 462        x->flags = flags;
 463        x->srcpos = srcpos;
 464        ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
 465        el->excludes[el->nr++] = x;
 466        x->el = el;
 467}
 468
 469static void *read_skip_worktree_file_from_index(const char *path, size_t *size,
 470                                                struct sha1_stat *sha1_stat)
 471{
 472        int pos, len;
 473        unsigned long sz;
 474        enum object_type type;
 475        void *data;
 476
 477        len = strlen(path);
 478        pos = cache_name_pos(path, len);
 479        if (pos < 0)
 480                return NULL;
 481        if (!ce_skip_worktree(active_cache[pos]))
 482                return NULL;
 483        data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
 484        if (!data || type != OBJ_BLOB) {
 485                free(data);
 486                return NULL;
 487        }
 488        *size = xsize_t(sz);
 489        if (sha1_stat) {
 490                memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
 491                hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
 492        }
 493        return data;
 494}
 495
 496/*
 497 * Frees memory within el which was allocated for exclude patterns and
 498 * the file buffer.  Does not free el itself.
 499 */
 500void clear_exclude_list(struct exclude_list *el)
 501{
 502        int i;
 503
 504        for (i = 0; i < el->nr; i++)
 505                free(el->excludes[i]);
 506        free(el->excludes);
 507        free(el->filebuf);
 508
 509        el->nr = 0;
 510        el->excludes = NULL;
 511        el->filebuf = NULL;
 512}
 513
 514static void trim_trailing_spaces(char *buf)
 515{
 516        char *p, *last_space = NULL;
 517
 518        for (p = buf; *p; p++)
 519                switch (*p) {
 520                case ' ':
 521                        if (!last_space)
 522                                last_space = p;
 523                        break;
 524                case '\\':
 525                        p++;
 526                        if (!*p)
 527                                return;
 528                        /* fallthrough */
 529                default:
 530                        last_space = NULL;
 531                }
 532
 533        if (last_space)
 534                *last_space = '\0';
 535}
 536
 537/*
 538 * Given a subdirectory name and "dir" of the current directory,
 539 * search the subdir in "dir" and return it, or create a new one if it
 540 * does not exist in "dir".
 541 *
 542 * If "name" has the trailing slash, it'll be excluded in the search.
 543 */
 544static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
 545                                                    struct untracked_cache_dir *dir,
 546                                                    const char *name, int len)
 547{
 548        int first, last;
 549        struct untracked_cache_dir *d;
 550        if (!dir)
 551                return NULL;
 552        if (len && name[len - 1] == '/')
 553                len--;
 554        first = 0;
 555        last = dir->dirs_nr;
 556        while (last > first) {
 557                int cmp, next = (last + first) >> 1;
 558                d = dir->dirs[next];
 559                cmp = strncmp(name, d->name, len);
 560                if (!cmp && strlen(d->name) > len)
 561                        cmp = -1;
 562                if (!cmp)
 563                        return d;
 564                if (cmp < 0) {
 565                        last = next;
 566                        continue;
 567                }
 568                first = next+1;
 569        }
 570
 571        uc->dir_created++;
 572        d = xmalloc(sizeof(*d) + len + 1);
 573        memset(d, 0, sizeof(*d));
 574        memcpy(d->name, name, len);
 575        d->name[len] = '\0';
 576
 577        ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
 578        memmove(dir->dirs + first + 1, dir->dirs + first,
 579                (dir->dirs_nr - first) * sizeof(*dir->dirs));
 580        dir->dirs_nr++;
 581        dir->dirs[first] = d;
 582        return d;
 583}
 584
 585/*
 586 * Given a file with name "fname", read it (either from disk, or from
 587 * the index if "check_index" is non-zero), parse it and store the
 588 * exclude rules in "el".
 589 *
 590 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
 591 * stat data from disk (only valid if add_excludes returns zero). If
 592 * ss_valid is non-zero, "ss" must contain good value as input.
 593 */
 594static int add_excludes(const char *fname, const char *base, int baselen,
 595                        struct exclude_list *el, int check_index,
 596                        struct sha1_stat *sha1_stat)
 597{
 598        struct stat st;
 599        int fd, i, lineno = 1;
 600        size_t size = 0;
 601        char *buf, *entry;
 602
 603        fd = open(fname, O_RDONLY);
 604        if (fd < 0 || fstat(fd, &st) < 0) {
 605                if (errno != ENOENT)
 606                        warn_on_inaccessible(fname);
 607                if (0 <= fd)
 608                        close(fd);
 609                if (!check_index ||
 610                    (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL)
 611                        return -1;
 612                if (size == 0) {
 613                        free(buf);
 614                        return 0;
 615                }
 616                if (buf[size-1] != '\n') {
 617                        buf = xrealloc(buf, size+1);
 618                        buf[size++] = '\n';
 619                }
 620        } else {
 621                size = xsize_t(st.st_size);
 622                if (size == 0) {
 623                        if (sha1_stat) {
 624                                fill_stat_data(&sha1_stat->stat, &st);
 625                                hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
 626                                sha1_stat->valid = 1;
 627                        }
 628                        close(fd);
 629                        return 0;
 630                }
 631                buf = xmalloc(size+1);
 632                if (read_in_full(fd, buf, size) != size) {
 633                        free(buf);
 634                        close(fd);
 635                        return -1;
 636                }
 637                buf[size++] = '\n';
 638                close(fd);
 639                if (sha1_stat) {
 640                        int pos;
 641                        if (sha1_stat->valid &&
 642                            !match_stat_data(&sha1_stat->stat, &st))
 643                                ; /* no content change, ss->sha1 still good */
 644                        else if (check_index &&
 645                                 (pos = cache_name_pos(fname, strlen(fname))) >= 0 &&
 646                                 !ce_stage(active_cache[pos]) &&
 647                                 ce_uptodate(active_cache[pos]) &&
 648                                 !would_convert_to_git(fname))
 649                                hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
 650                        else
 651                                hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
 652                        fill_stat_data(&sha1_stat->stat, &st);
 653                        sha1_stat->valid = 1;
 654                }
 655        }
 656
 657        el->filebuf = buf;
 658        entry = buf;
 659        for (i = 0; i < size; i++) {
 660                if (buf[i] == '\n') {
 661                        if (entry != buf + i && entry[0] != '#') {
 662                                buf[i - (i && buf[i-1] == '\r')] = 0;
 663                                trim_trailing_spaces(entry);
 664                                add_exclude(entry, base, baselen, el, lineno);
 665                        }
 666                        lineno++;
 667                        entry = buf + i + 1;
 668                }
 669        }
 670        return 0;
 671}
 672
 673int add_excludes_from_file_to_list(const char *fname, const char *base,
 674                                   int baselen, struct exclude_list *el,
 675                                   int check_index)
 676{
 677        return add_excludes(fname, base, baselen, el, check_index, NULL);
 678}
 679
 680struct exclude_list *add_exclude_list(struct dir_struct *dir,
 681                                      int group_type, const char *src)
 682{
 683        struct exclude_list *el;
 684        struct exclude_list_group *group;
 685
 686        group = &dir->exclude_list_group[group_type];
 687        ALLOC_GROW(group->el, group->nr + 1, group->alloc);
 688        el = &group->el[group->nr++];
 689        memset(el, 0, sizeof(*el));
 690        el->src = src;
 691        return el;
 692}
 693
 694/*
 695 * Used to set up core.excludesfile and .git/info/exclude lists.
 696 */
 697static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
 698                                     struct sha1_stat *sha1_stat)
 699{
 700        struct exclude_list *el;
 701        el = add_exclude_list(dir, EXC_FILE, fname);
 702        if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0)
 703                die("cannot use %s as an exclude file", fname);
 704}
 705
 706void add_excludes_from_file(struct dir_struct *dir, const char *fname)
 707{
 708        add_excludes_from_file_1(dir, fname, NULL);
 709}
 710
 711int match_basename(const char *basename, int basenamelen,
 712                   const char *pattern, int prefix, int patternlen,
 713                   int flags)
 714{
 715        if (prefix == patternlen) {
 716                if (patternlen == basenamelen &&
 717                    !strncmp_icase(pattern, basename, basenamelen))
 718                        return 1;
 719        } else if (flags & EXC_FLAG_ENDSWITH) {
 720                /* "*literal" matching against "fooliteral" */
 721                if (patternlen - 1 <= basenamelen &&
 722                    !strncmp_icase(pattern + 1,
 723                                   basename + basenamelen - (patternlen - 1),
 724                                   patternlen - 1))
 725                        return 1;
 726        } else {
 727                if (fnmatch_icase_mem(pattern, patternlen,
 728                                      basename, basenamelen,
 729                                      0) == 0)
 730                        return 1;
 731        }
 732        return 0;
 733}
 734
 735int match_pathname(const char *pathname, int pathlen,
 736                   const char *base, int baselen,
 737                   const char *pattern, int prefix, int patternlen,
 738                   int flags)
 739{
 740        const char *name;
 741        int namelen;
 742
 743        /*
 744         * match with FNM_PATHNAME; the pattern has base implicitly
 745         * in front of it.
 746         */
 747        if (*pattern == '/') {
 748                pattern++;
 749                patternlen--;
 750                prefix--;
 751        }
 752
 753        /*
 754         * baselen does not count the trailing slash. base[] may or
 755         * may not end with a trailing slash though.
 756         */
 757        if (pathlen < baselen + 1 ||
 758            (baselen && pathname[baselen] != '/') ||
 759            strncmp_icase(pathname, base, baselen))
 760                return 0;
 761
 762        namelen = baselen ? pathlen - baselen - 1 : pathlen;
 763        name = pathname + pathlen - namelen;
 764
 765        if (prefix) {
 766                /*
 767                 * if the non-wildcard part is longer than the
 768                 * remaining pathname, surely it cannot match.
 769                 */
 770                if (prefix > namelen)
 771                        return 0;
 772
 773                if (strncmp_icase(pattern, name, prefix))
 774                        return 0;
 775                pattern += prefix;
 776                patternlen -= prefix;
 777                name    += prefix;
 778                namelen -= prefix;
 779
 780                /*
 781                 * If the whole pattern did not have a wildcard,
 782                 * then our prefix match is all we need; we
 783                 * do not need to call fnmatch at all.
 784                 */
 785                if (!patternlen && !namelen)
 786                        return 1;
 787        }
 788
 789        return fnmatch_icase_mem(pattern, patternlen,
 790                                 name, namelen,
 791                                 WM_PATHNAME) == 0;
 792}
 793
 794/*
 795 * Scan the given exclude list in reverse to see whether pathname
 796 * should be ignored.  The first match (i.e. the last on the list), if
 797 * any, determines the fate.  Returns the exclude_list element which
 798 * matched, or NULL for undecided.
 799 */
 800static struct exclude *last_exclude_matching_from_list(const char *pathname,
 801                                                       int pathlen,
 802                                                       const char *basename,
 803                                                       int *dtype,
 804                                                       struct exclude_list *el)
 805{
 806        int i;
 807
 808        if (!el->nr)
 809                return NULL;    /* undefined */
 810
 811        for (i = el->nr - 1; 0 <= i; i--) {
 812                struct exclude *x = el->excludes[i];
 813                const char *exclude = x->pattern;
 814                int prefix = x->nowildcardlen;
 815
 816                if (x->flags & EXC_FLAG_MUSTBEDIR) {
 817                        if (*dtype == DT_UNKNOWN)
 818                                *dtype = get_dtype(NULL, pathname, pathlen);
 819                        if (*dtype != DT_DIR)
 820                                continue;
 821                }
 822
 823                if (x->flags & EXC_FLAG_NODIR) {
 824                        if (match_basename(basename,
 825                                           pathlen - (basename - pathname),
 826                                           exclude, prefix, x->patternlen,
 827                                           x->flags))
 828                                return x;
 829                        continue;
 830                }
 831
 832                assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
 833                if (match_pathname(pathname, pathlen,
 834                                   x->base, x->baselen ? x->baselen - 1 : 0,
 835                                   exclude, prefix, x->patternlen, x->flags))
 836                        return x;
 837        }
 838        return NULL; /* undecided */
 839}
 840
 841/*
 842 * Scan the list and let the last match determine the fate.
 843 * Return 1 for exclude, 0 for include and -1 for undecided.
 844 */
 845int is_excluded_from_list(const char *pathname,
 846                          int pathlen, const char *basename, int *dtype,
 847                          struct exclude_list *el)
 848{
 849        struct exclude *exclude;
 850        exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
 851        if (exclude)
 852                return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
 853        return -1; /* undecided */
 854}
 855
 856static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
 857                const char *pathname, int pathlen, const char *basename,
 858                int *dtype_p)
 859{
 860        int i, j;
 861        struct exclude_list_group *group;
 862        struct exclude *exclude;
 863        for (i = EXC_CMDL; i <= EXC_FILE; i++) {
 864                group = &dir->exclude_list_group[i];
 865                for (j = group->nr - 1; j >= 0; j--) {
 866                        exclude = last_exclude_matching_from_list(
 867                                pathname, pathlen, basename, dtype_p,
 868                                &group->el[j]);
 869                        if (exclude)
 870                                return exclude;
 871                }
 872        }
 873        return NULL;
 874}
 875
 876/*
 877 * Loads the per-directory exclude list for the substring of base
 878 * which has a char length of baselen.
 879 */
 880static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
 881{
 882        struct exclude_list_group *group;
 883        struct exclude_list *el;
 884        struct exclude_stack *stk = NULL;
 885        struct untracked_cache_dir *untracked;
 886        int current;
 887
 888        group = &dir->exclude_list_group[EXC_DIRS];
 889
 890        /*
 891         * Pop the exclude lists from the EXCL_DIRS exclude_list_group
 892         * which originate from directories not in the prefix of the
 893         * path being checked.
 894         */
 895        while ((stk = dir->exclude_stack) != NULL) {
 896                if (stk->baselen <= baselen &&
 897                    !strncmp(dir->basebuf.buf, base, stk->baselen))
 898                        break;
 899                el = &group->el[dir->exclude_stack->exclude_ix];
 900                dir->exclude_stack = stk->prev;
 901                dir->exclude = NULL;
 902                free((char *)el->src); /* see strbuf_detach() below */
 903                clear_exclude_list(el);
 904                free(stk);
 905                group->nr--;
 906        }
 907
 908        /* Skip traversing into sub directories if the parent is excluded */
 909        if (dir->exclude)
 910                return;
 911
 912        /*
 913         * Lazy initialization. All call sites currently just
 914         * memset(dir, 0, sizeof(*dir)) before use. Changing all of
 915         * them seems lots of work for little benefit.
 916         */
 917        if (!dir->basebuf.buf)
 918                strbuf_init(&dir->basebuf, PATH_MAX);
 919
 920        /* Read from the parent directories and push them down. */
 921        current = stk ? stk->baselen : -1;
 922        strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
 923        if (dir->untracked)
 924                untracked = stk ? stk->ucd : dir->untracked->root;
 925        else
 926                untracked = NULL;
 927
 928        while (current < baselen) {
 929                const char *cp;
 930                struct sha1_stat sha1_stat;
 931
 932                stk = xcalloc(1, sizeof(*stk));
 933                if (current < 0) {
 934                        cp = base;
 935                        current = 0;
 936                } else {
 937                        cp = strchr(base + current + 1, '/');
 938                        if (!cp)
 939                                die("oops in prep_exclude");
 940                        cp++;
 941                        untracked =
 942                                lookup_untracked(dir->untracked, untracked,
 943                                                 base + current,
 944                                                 cp - base - current);
 945                }
 946                stk->prev = dir->exclude_stack;
 947                stk->baselen = cp - base;
 948                stk->exclude_ix = group->nr;
 949                stk->ucd = untracked;
 950                el = add_exclude_list(dir, EXC_DIRS, NULL);
 951                strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
 952                assert(stk->baselen == dir->basebuf.len);
 953
 954                /* Abort if the directory is excluded */
 955                if (stk->baselen) {
 956                        int dt = DT_DIR;
 957                        dir->basebuf.buf[stk->baselen - 1] = 0;
 958                        dir->exclude = last_exclude_matching_from_lists(dir,
 959                                dir->basebuf.buf, stk->baselen - 1,
 960                                dir->basebuf.buf + current, &dt);
 961                        dir->basebuf.buf[stk->baselen - 1] = '/';
 962                        if (dir->exclude &&
 963                            dir->exclude->flags & EXC_FLAG_NEGATIVE)
 964                                dir->exclude = NULL;
 965                        if (dir->exclude) {
 966                                dir->exclude_stack = stk;
 967                                return;
 968                        }
 969                }
 970
 971                /* Try to read per-directory file */
 972                hashclr(sha1_stat.sha1);
 973                sha1_stat.valid = 0;
 974                if (dir->exclude_per_dir) {
 975                        /*
 976                         * dir->basebuf gets reused by the traversal, but we
 977                         * need fname to remain unchanged to ensure the src
 978                         * member of each struct exclude correctly
 979                         * back-references its source file.  Other invocations
 980                         * of add_exclude_list provide stable strings, so we
 981                         * strbuf_detach() and free() here in the caller.
 982                         */
 983                        struct strbuf sb = STRBUF_INIT;
 984                        strbuf_addbuf(&sb, &dir->basebuf);
 985                        strbuf_addstr(&sb, dir->exclude_per_dir);
 986                        el->src = strbuf_detach(&sb, NULL);
 987                        add_excludes(el->src, el->src, stk->baselen, el, 1,
 988                                     untracked ? &sha1_stat : NULL);
 989                }
 990                if (untracked) {
 991                        hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
 992                }
 993                dir->exclude_stack = stk;
 994                current = stk->baselen;
 995        }
 996        strbuf_setlen(&dir->basebuf, baselen);
 997}
 998
 999/*
1000 * Loads the exclude lists for the directory containing pathname, then
1001 * scans all exclude lists to determine whether pathname is excluded.
1002 * Returns the exclude_list element which matched, or NULL for
1003 * undecided.
1004 */
1005struct exclude *last_exclude_matching(struct dir_struct *dir,
1006                                             const char *pathname,
1007                                             int *dtype_p)
1008{
1009        int pathlen = strlen(pathname);
1010        const char *basename = strrchr(pathname, '/');
1011        basename = (basename) ? basename+1 : pathname;
1012
1013        prep_exclude(dir, pathname, basename-pathname);
1014
1015        if (dir->exclude)
1016                return dir->exclude;
1017
1018        return last_exclude_matching_from_lists(dir, pathname, pathlen,
1019                        basename, dtype_p);
1020}
1021
1022/*
1023 * Loads the exclude lists for the directory containing pathname, then
1024 * scans all exclude lists to determine whether pathname is excluded.
1025 * Returns 1 if true, otherwise 0.
1026 */
1027int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1028{
1029        struct exclude *exclude =
1030                last_exclude_matching(dir, pathname, dtype_p);
1031        if (exclude)
1032                return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1033        return 0;
1034}
1035
1036static struct dir_entry *dir_entry_new(const char *pathname, int len)
1037{
1038        struct dir_entry *ent;
1039
1040        ent = xmalloc(sizeof(*ent) + len + 1);
1041        ent->len = len;
1042        memcpy(ent->name, pathname, len);
1043        ent->name[len] = 0;
1044        return ent;
1045}
1046
1047static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1048{
1049        if (cache_file_exists(pathname, len, ignore_case))
1050                return NULL;
1051
1052        ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1053        return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1054}
1055
1056struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1057{
1058        if (!cache_name_is_other(pathname, len))
1059                return NULL;
1060
1061        ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1062        return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1063}
1064
1065enum exist_status {
1066        index_nonexistent = 0,
1067        index_directory,
1068        index_gitdir
1069};
1070
1071/*
1072 * Do not use the alphabetically sorted index to look up
1073 * the directory name; instead, use the case insensitive
1074 * directory hash.
1075 */
1076static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1077{
1078        const struct cache_entry *ce = cache_dir_exists(dirname, len);
1079        unsigned char endchar;
1080
1081        if (!ce)
1082                return index_nonexistent;
1083        endchar = ce->name[len];
1084
1085        /*
1086         * The cache_entry structure returned will contain this dirname
1087         * and possibly additional path components.
1088         */
1089        if (endchar == '/')
1090                return index_directory;
1091
1092        /*
1093         * If there are no additional path components, then this cache_entry
1094         * represents a submodule.  Submodules, despite being directories,
1095         * are stored in the cache without a closing slash.
1096         */
1097        if (!endchar && S_ISGITLINK(ce->ce_mode))
1098                return index_gitdir;
1099
1100        /* This should never be hit, but it exists just in case. */
1101        return index_nonexistent;
1102}
1103
1104/*
1105 * The index sorts alphabetically by entry name, which
1106 * means that a gitlink sorts as '\0' at the end, while
1107 * a directory (which is defined not as an entry, but as
1108 * the files it contains) will sort with the '/' at the
1109 * end.
1110 */
1111static enum exist_status directory_exists_in_index(const char *dirname, int len)
1112{
1113        int pos;
1114
1115        if (ignore_case)
1116                return directory_exists_in_index_icase(dirname, len);
1117
1118        pos = cache_name_pos(dirname, len);
1119        if (pos < 0)
1120                pos = -pos-1;
1121        while (pos < active_nr) {
1122                const struct cache_entry *ce = active_cache[pos++];
1123                unsigned char endchar;
1124
1125                if (strncmp(ce->name, dirname, len))
1126                        break;
1127                endchar = ce->name[len];
1128                if (endchar > '/')
1129                        break;
1130                if (endchar == '/')
1131                        return index_directory;
1132                if (!endchar && S_ISGITLINK(ce->ce_mode))
1133                        return index_gitdir;
1134        }
1135        return index_nonexistent;
1136}
1137
1138/*
1139 * When we find a directory when traversing the filesystem, we
1140 * have three distinct cases:
1141 *
1142 *  - ignore it
1143 *  - see it as a directory
1144 *  - recurse into it
1145 *
1146 * and which one we choose depends on a combination of existing
1147 * git index contents and the flags passed into the directory
1148 * traversal routine.
1149 *
1150 * Case 1: If we *already* have entries in the index under that
1151 * directory name, we always recurse into the directory to see
1152 * all the files.
1153 *
1154 * Case 2: If we *already* have that directory name as a gitlink,
1155 * we always continue to see it as a gitlink, regardless of whether
1156 * there is an actual git directory there or not (it might not
1157 * be checked out as a subproject!)
1158 *
1159 * Case 3: if we didn't have it in the index previously, we
1160 * have a few sub-cases:
1161 *
1162 *  (a) if "show_other_directories" is true, we show it as
1163 *      just a directory, unless "hide_empty_directories" is
1164 *      also true, in which case we need to check if it contains any
1165 *      untracked and / or ignored files.
1166 *  (b) if it looks like a git directory, and we don't have
1167 *      'no_gitlinks' set we treat it as a gitlink, and show it
1168 *      as a directory.
1169 *  (c) otherwise, we recurse into it.
1170 */
1171static enum path_treatment treat_directory(struct dir_struct *dir,
1172        struct untracked_cache_dir *untracked,
1173        const char *dirname, int len, int exclude,
1174        const struct path_simplify *simplify)
1175{
1176        /* The "len-1" is to strip the final '/' */
1177        switch (directory_exists_in_index(dirname, len-1)) {
1178        case index_directory:
1179                return path_recurse;
1180
1181        case index_gitdir:
1182                return path_none;
1183
1184        case index_nonexistent:
1185                if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1186                        break;
1187                if (!(dir->flags & DIR_NO_GITLINKS)) {
1188                        unsigned char sha1[20];
1189                        if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1190                                return path_untracked;
1191                }
1192                return path_recurse;
1193        }
1194
1195        /* This is the "show_other_directories" case */
1196
1197        if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1198                return exclude ? path_excluded : path_untracked;
1199
1200        untracked = lookup_untracked(dir->untracked, untracked, dirname, len);
1201        return read_directory_recursive(dir, dirname, len,
1202                                        untracked, 1, simplify);
1203}
1204
1205/*
1206 * This is an inexact early pruning of any recursive directory
1207 * reading - if the path cannot possibly be in the pathspec,
1208 * return true, and we'll skip it early.
1209 */
1210static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1211{
1212        if (simplify) {
1213                for (;;) {
1214                        const char *match = simplify->path;
1215                        int len = simplify->len;
1216
1217                        if (!match)
1218                                break;
1219                        if (len > pathlen)
1220                                len = pathlen;
1221                        if (!memcmp(path, match, len))
1222                                return 0;
1223                        simplify++;
1224                }
1225                return 1;
1226        }
1227        return 0;
1228}
1229
1230/*
1231 * This function tells us whether an excluded path matches a
1232 * list of "interesting" pathspecs. That is, whether a path matched
1233 * by any of the pathspecs could possibly be ignored by excluding
1234 * the specified path. This can happen if:
1235 *
1236 *   1. the path is mentioned explicitly in the pathspec
1237 *
1238 *   2. the path is a directory prefix of some element in the
1239 *      pathspec
1240 */
1241static int exclude_matches_pathspec(const char *path, int len,
1242                const struct path_simplify *simplify)
1243{
1244        if (simplify) {
1245                for (; simplify->path; simplify++) {
1246                        if (len == simplify->len
1247                            && !memcmp(path, simplify->path, len))
1248                                return 1;
1249                        if (len < simplify->len
1250                            && simplify->path[len] == '/'
1251                            && !memcmp(path, simplify->path, len))
1252                                return 1;
1253                }
1254        }
1255        return 0;
1256}
1257
1258static int get_index_dtype(const char *path, int len)
1259{
1260        int pos;
1261        const struct cache_entry *ce;
1262
1263        ce = cache_file_exists(path, len, 0);
1264        if (ce) {
1265                if (!ce_uptodate(ce))
1266                        return DT_UNKNOWN;
1267                if (S_ISGITLINK(ce->ce_mode))
1268                        return DT_DIR;
1269                /*
1270                 * Nobody actually cares about the
1271                 * difference between DT_LNK and DT_REG
1272                 */
1273                return DT_REG;
1274        }
1275
1276        /* Try to look it up as a directory */
1277        pos = cache_name_pos(path, len);
1278        if (pos >= 0)
1279                return DT_UNKNOWN;
1280        pos = -pos-1;
1281        while (pos < active_nr) {
1282                ce = active_cache[pos++];
1283                if (strncmp(ce->name, path, len))
1284                        break;
1285                if (ce->name[len] > '/')
1286                        break;
1287                if (ce->name[len] < '/')
1288                        continue;
1289                if (!ce_uptodate(ce))
1290                        break;  /* continue? */
1291                return DT_DIR;
1292        }
1293        return DT_UNKNOWN;
1294}
1295
1296static int get_dtype(struct dirent *de, const char *path, int len)
1297{
1298        int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1299        struct stat st;
1300
1301        if (dtype != DT_UNKNOWN)
1302                return dtype;
1303        dtype = get_index_dtype(path, len);
1304        if (dtype != DT_UNKNOWN)
1305                return dtype;
1306        if (lstat(path, &st))
1307                return dtype;
1308        if (S_ISREG(st.st_mode))
1309                return DT_REG;
1310        if (S_ISDIR(st.st_mode))
1311                return DT_DIR;
1312        if (S_ISLNK(st.st_mode))
1313                return DT_LNK;
1314        return dtype;
1315}
1316
1317static enum path_treatment treat_one_path(struct dir_struct *dir,
1318                                          struct untracked_cache_dir *untracked,
1319                                          struct strbuf *path,
1320                                          const struct path_simplify *simplify,
1321                                          int dtype, struct dirent *de)
1322{
1323        int exclude;
1324        int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1325
1326        if (dtype == DT_UNKNOWN)
1327                dtype = get_dtype(de, path->buf, path->len);
1328
1329        /* Always exclude indexed files */
1330        if (dtype != DT_DIR && has_path_in_index)
1331                return path_none;
1332
1333        /*
1334         * When we are looking at a directory P in the working tree,
1335         * there are three cases:
1336         *
1337         * (1) P exists in the index.  Everything inside the directory P in
1338         * the working tree needs to go when P is checked out from the
1339         * index.
1340         *
1341         * (2) P does not exist in the index, but there is P/Q in the index.
1342         * We know P will stay a directory when we check out the contents
1343         * of the index, but we do not know yet if there is a directory
1344         * P/Q in the working tree to be killed, so we need to recurse.
1345         *
1346         * (3) P does not exist in the index, and there is no P/Q in the index
1347         * to require P to be a directory, either.  Only in this case, we
1348         * know that everything inside P will not be killed without
1349         * recursing.
1350         */
1351        if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1352            (dtype == DT_DIR) &&
1353            !has_path_in_index &&
1354            (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1355                return path_none;
1356
1357        exclude = is_excluded(dir, path->buf, &dtype);
1358
1359        /*
1360         * Excluded? If we don't explicitly want to show
1361         * ignored files, ignore it
1362         */
1363        if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1364                return path_excluded;
1365
1366        switch (dtype) {
1367        default:
1368                return path_none;
1369        case DT_DIR:
1370                strbuf_addch(path, '/');
1371                return treat_directory(dir, untracked, path->buf, path->len, exclude,
1372                        simplify);
1373        case DT_REG:
1374        case DT_LNK:
1375                return exclude ? path_excluded : path_untracked;
1376        }
1377}
1378
1379static enum path_treatment treat_path(struct dir_struct *dir,
1380                                      struct untracked_cache_dir *untracked,
1381                                      struct dirent *de,
1382                                      struct strbuf *path,
1383                                      int baselen,
1384                                      const struct path_simplify *simplify)
1385{
1386        int dtype;
1387
1388        if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1389                return path_none;
1390        strbuf_setlen(path, baselen);
1391        strbuf_addstr(path, de->d_name);
1392        if (simplify_away(path->buf, path->len, simplify))
1393                return path_none;
1394
1395        dtype = DTYPE(de);
1396        return treat_one_path(dir, untracked, path, simplify, dtype, de);
1397}
1398
1399static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1400{
1401        if (!dir)
1402                return;
1403        ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1404                   dir->untracked_alloc);
1405        dir->untracked[dir->untracked_nr++] = xstrdup(name);
1406}
1407
1408/*
1409 * Read a directory tree. We currently ignore anything but
1410 * directories, regular files and symlinks. That's because git
1411 * doesn't handle them at all yet. Maybe that will change some
1412 * day.
1413 *
1414 * Also, we ignore the name ".git" (even if it is not a directory).
1415 * That likely will not change.
1416 *
1417 * Returns the most significant path_treatment value encountered in the scan.
1418 */
1419static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1420                                    const char *base, int baselen,
1421                                    struct untracked_cache_dir *untracked, int check_only,
1422                                    const struct path_simplify *simplify)
1423{
1424        DIR *fdir;
1425        enum path_treatment state, subdir_state, dir_state = path_none;
1426        struct dirent *de;
1427        struct strbuf path = STRBUF_INIT;
1428
1429        strbuf_add(&path, base, baselen);
1430
1431        fdir = opendir(path.len ? path.buf : ".");
1432        if (!fdir)
1433                goto out;
1434
1435        if (untracked)
1436                untracked->check_only = !!check_only;
1437
1438        while ((de = readdir(fdir)) != NULL) {
1439                /* check how the file or directory should be treated */
1440                state = treat_path(dir, untracked, de, &path, baselen, simplify);
1441
1442                if (state > dir_state)
1443                        dir_state = state;
1444
1445                /* recurse into subdir if instructed by treat_path */
1446                if (state == path_recurse) {
1447                        struct untracked_cache_dir *ud;
1448                        ud = lookup_untracked(dir->untracked, untracked,
1449                                              path.buf + baselen,
1450                                              path.len - baselen);
1451                        subdir_state =
1452                                read_directory_recursive(dir, path.buf, path.len,
1453                                                         ud, check_only, simplify);
1454                        if (subdir_state > dir_state)
1455                                dir_state = subdir_state;
1456                }
1457
1458                if (check_only) {
1459                        /* abort early if maximum state has been reached */
1460                        if (dir_state == path_untracked) {
1461                                if (untracked)
1462                                        add_untracked(untracked, path.buf + baselen);
1463                                break;
1464                        }
1465                        /* skip the dir_add_* part */
1466                        continue;
1467                }
1468
1469                /* add the path to the appropriate result list */
1470                switch (state) {
1471                case path_excluded:
1472                        if (dir->flags & DIR_SHOW_IGNORED)
1473                                dir_add_name(dir, path.buf, path.len);
1474                        else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1475                                ((dir->flags & DIR_COLLECT_IGNORED) &&
1476                                exclude_matches_pathspec(path.buf, path.len,
1477                                        simplify)))
1478                                dir_add_ignored(dir, path.buf, path.len);
1479                        break;
1480
1481                case path_untracked:
1482                        if (dir->flags & DIR_SHOW_IGNORED)
1483                                break;
1484                        dir_add_name(dir, path.buf, path.len);
1485                        if (untracked)
1486                                add_untracked(untracked, path.buf + baselen);
1487                        break;
1488
1489                default:
1490                        break;
1491                }
1492        }
1493        closedir(fdir);
1494 out:
1495        strbuf_release(&path);
1496
1497        return dir_state;
1498}
1499
1500static int cmp_name(const void *p1, const void *p2)
1501{
1502        const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1503        const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1504
1505        return name_compare(e1->name, e1->len, e2->name, e2->len);
1506}
1507
1508static struct path_simplify *create_simplify(const char **pathspec)
1509{
1510        int nr, alloc = 0;
1511        struct path_simplify *simplify = NULL;
1512
1513        if (!pathspec)
1514                return NULL;
1515
1516        for (nr = 0 ; ; nr++) {
1517                const char *match;
1518                ALLOC_GROW(simplify, nr + 1, alloc);
1519                match = *pathspec++;
1520                if (!match)
1521                        break;
1522                simplify[nr].path = match;
1523                simplify[nr].len = simple_length(match);
1524        }
1525        simplify[nr].path = NULL;
1526        simplify[nr].len = 0;
1527        return simplify;
1528}
1529
1530static void free_simplify(struct path_simplify *simplify)
1531{
1532        free(simplify);
1533}
1534
1535static int treat_leading_path(struct dir_struct *dir,
1536                              const char *path, int len,
1537                              const struct path_simplify *simplify)
1538{
1539        struct strbuf sb = STRBUF_INIT;
1540        int baselen, rc = 0;
1541        const char *cp;
1542        int old_flags = dir->flags;
1543
1544        while (len && path[len - 1] == '/')
1545                len--;
1546        if (!len)
1547                return 1;
1548        baselen = 0;
1549        dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1550        while (1) {
1551                cp = path + baselen + !!baselen;
1552                cp = memchr(cp, '/', path + len - cp);
1553                if (!cp)
1554                        baselen = len;
1555                else
1556                        baselen = cp - path;
1557                strbuf_setlen(&sb, 0);
1558                strbuf_add(&sb, path, baselen);
1559                if (!is_directory(sb.buf))
1560                        break;
1561                if (simplify_away(sb.buf, sb.len, simplify))
1562                        break;
1563                if (treat_one_path(dir, NULL, &sb, simplify,
1564                                   DT_DIR, NULL) == path_none)
1565                        break; /* do not recurse into it */
1566                if (len <= baselen) {
1567                        rc = 1;
1568                        break; /* finished checking */
1569                }
1570        }
1571        strbuf_release(&sb);
1572        dir->flags = old_flags;
1573        return rc;
1574}
1575
1576int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1577{
1578        struct path_simplify *simplify;
1579
1580        /*
1581         * Check out create_simplify()
1582         */
1583        if (pathspec)
1584                GUARD_PATHSPEC(pathspec,
1585                               PATHSPEC_FROMTOP |
1586                               PATHSPEC_MAXDEPTH |
1587                               PATHSPEC_LITERAL |
1588                               PATHSPEC_GLOB |
1589                               PATHSPEC_ICASE |
1590                               PATHSPEC_EXCLUDE);
1591
1592        if (has_symlink_leading_path(path, len))
1593                return dir->nr;
1594
1595        /*
1596         * exclude patterns are treated like positive ones in
1597         * create_simplify. Usually exclude patterns should be a
1598         * subset of positive ones, which has no impacts on
1599         * create_simplify().
1600         */
1601        simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
1602        if (!len || treat_leading_path(dir, path, len, simplify))
1603                read_directory_recursive(dir, path, len,
1604                                         dir->untracked ? dir->untracked->root : NULL,
1605                                         0, simplify);
1606        free_simplify(simplify);
1607        qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1608        qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1609        return dir->nr;
1610}
1611
1612int file_exists(const char *f)
1613{
1614        struct stat sb;
1615        return lstat(f, &sb) == 0;
1616}
1617
1618/*
1619 * Given two normalized paths (a trailing slash is ok), if subdir is
1620 * outside dir, return -1.  Otherwise return the offset in subdir that
1621 * can be used as relative path to dir.
1622 */
1623int dir_inside_of(const char *subdir, const char *dir)
1624{
1625        int offset = 0;
1626
1627        assert(dir && subdir && *dir && *subdir);
1628
1629        while (*dir && *subdir && *dir == *subdir) {
1630                dir++;
1631                subdir++;
1632                offset++;
1633        }
1634
1635        /* hel[p]/me vs hel[l]/yeah */
1636        if (*dir && *subdir)
1637                return -1;
1638
1639        if (!*subdir)
1640                return !*dir ? offset : -1; /* same dir */
1641
1642        /* foo/[b]ar vs foo/[] */
1643        if (is_dir_sep(dir[-1]))
1644                return is_dir_sep(subdir[-1]) ? offset : -1;
1645
1646        /* foo[/]bar vs foo[] */
1647        return is_dir_sep(*subdir) ? offset + 1 : -1;
1648}
1649
1650int is_inside_dir(const char *dir)
1651{
1652        char *cwd;
1653        int rc;
1654
1655        if (!dir)
1656                return 0;
1657
1658        cwd = xgetcwd();
1659        rc = (dir_inside_of(cwd, dir) >= 0);
1660        free(cwd);
1661        return rc;
1662}
1663
1664int is_empty_dir(const char *path)
1665{
1666        DIR *dir = opendir(path);
1667        struct dirent *e;
1668        int ret = 1;
1669
1670        if (!dir)
1671                return 0;
1672
1673        while ((e = readdir(dir)) != NULL)
1674                if (!is_dot_or_dotdot(e->d_name)) {
1675                        ret = 0;
1676                        break;
1677                }
1678
1679        closedir(dir);
1680        return ret;
1681}
1682
1683static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1684{
1685        DIR *dir;
1686        struct dirent *e;
1687        int ret = 0, original_len = path->len, len, kept_down = 0;
1688        int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1689        int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1690        unsigned char submodule_head[20];
1691
1692        if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1693            !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1694                /* Do not descend and nuke a nested git work tree. */
1695                if (kept_up)
1696                        *kept_up = 1;
1697                return 0;
1698        }
1699
1700        flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1701        dir = opendir(path->buf);
1702        if (!dir) {
1703                if (errno == ENOENT)
1704                        return keep_toplevel ? -1 : 0;
1705                else if (errno == EACCES && !keep_toplevel)
1706                        /*
1707                         * An empty dir could be removable even if it
1708                         * is unreadable:
1709                         */
1710                        return rmdir(path->buf);
1711                else
1712                        return -1;
1713        }
1714        if (path->buf[original_len - 1] != '/')
1715                strbuf_addch(path, '/');
1716
1717        len = path->len;
1718        while ((e = readdir(dir)) != NULL) {
1719                struct stat st;
1720                if (is_dot_or_dotdot(e->d_name))
1721                        continue;
1722
1723                strbuf_setlen(path, len);
1724                strbuf_addstr(path, e->d_name);
1725                if (lstat(path->buf, &st)) {
1726                        if (errno == ENOENT)
1727                                /*
1728                                 * file disappeared, which is what we
1729                                 * wanted anyway
1730                                 */
1731                                continue;
1732                        /* fall thru */
1733                } else if (S_ISDIR(st.st_mode)) {
1734                        if (!remove_dir_recurse(path, flag, &kept_down))
1735                                continue; /* happy */
1736                } else if (!only_empty &&
1737                           (!unlink(path->buf) || errno == ENOENT)) {
1738                        continue; /* happy, too */
1739                }
1740
1741                /* path too long, stat fails, or non-directory still exists */
1742                ret = -1;
1743                break;
1744        }
1745        closedir(dir);
1746
1747        strbuf_setlen(path, original_len);
1748        if (!ret && !keep_toplevel && !kept_down)
1749                ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
1750        else if (kept_up)
1751                /*
1752                 * report the uplevel that it is not an error that we
1753                 * did not rmdir() our directory.
1754                 */
1755                *kept_up = !ret;
1756        return ret;
1757}
1758
1759int remove_dir_recursively(struct strbuf *path, int flag)
1760{
1761        return remove_dir_recurse(path, flag, NULL);
1762}
1763
1764void setup_standard_excludes(struct dir_struct *dir)
1765{
1766        const char *path;
1767        char *xdg_path;
1768
1769        dir->exclude_per_dir = ".gitignore";
1770        path = git_path("info/exclude");
1771        if (!excludes_file) {
1772                home_config_paths(NULL, &xdg_path, "ignore");
1773                excludes_file = xdg_path;
1774        }
1775        if (!access_or_warn(path, R_OK, 0))
1776                add_excludes_from_file_1(dir, path,
1777                                         dir->untracked ? &dir->ss_info_exclude : NULL);
1778        if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
1779                add_excludes_from_file_1(dir, excludes_file,
1780                                         dir->untracked ? &dir->ss_excludes_file : NULL);
1781}
1782
1783int remove_path(const char *name)
1784{
1785        char *slash;
1786
1787        if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
1788                return -1;
1789
1790        slash = strrchr(name, '/');
1791        if (slash) {
1792                char *dirs = xstrdup(name);
1793                slash = dirs + (slash - name);
1794                do {
1795                        *slash = '\0';
1796                } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1797                free(dirs);
1798        }
1799        return 0;
1800}
1801
1802/*
1803 * Frees memory within dir which was allocated for exclude lists and
1804 * the exclude_stack.  Does not free dir itself.
1805 */
1806void clear_directory(struct dir_struct *dir)
1807{
1808        int i, j;
1809        struct exclude_list_group *group;
1810        struct exclude_list *el;
1811        struct exclude_stack *stk;
1812
1813        for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1814                group = &dir->exclude_list_group[i];
1815                for (j = 0; j < group->nr; j++) {
1816                        el = &group->el[j];
1817                        if (i == EXC_DIRS)
1818                                free((char *)el->src);
1819                        clear_exclude_list(el);
1820                }
1821                free(group->el);
1822        }
1823
1824        stk = dir->exclude_stack;
1825        while (stk) {
1826                struct exclude_stack *prev = stk->prev;
1827                free(stk);
1828                stk = prev;
1829        }
1830        strbuf_release(&dir->basebuf);
1831}