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 "utf8.h" 16#include "varint.h" 17#include "ewah/ewok.h" 18 19struct path_simplify { 20 int len; 21 const char *path; 22}; 23 24/* 25 * Tells read_directory_recursive how a file or directory should be treated. 26 * Values are ordered by significance, e.g. if a directory contains both 27 * excluded and untracked files, it is listed as untracked because 28 * path_untracked > path_excluded. 29 */ 30enum path_treatment { 31 path_none = 0, 32 path_recurse, 33 path_excluded, 34 path_untracked 35}; 36 37/* 38 * Support data structure for our opendir/readdir/closedir wrappers 39 */ 40struct cached_dir { 41 DIR *fdir; 42 struct untracked_cache_dir *untracked; 43 int nr_files; 44 int nr_dirs; 45 46 struct dirent *de; 47 const char *file; 48 struct untracked_cache_dir *ucd; 49}; 50 51static enum path_treatment read_directory_recursive(struct dir_struct *dir, 52 const char *path, int len, struct untracked_cache_dir *untracked, 53 int check_only, const struct path_simplify *simplify); 54static int get_dtype(struct dirent *de, const char *path, int len); 55 56static struct trace_key trace_exclude = TRACE_KEY_INIT(EXCLUDE); 57 58/* helper string functions with support for the ignore_case flag */ 59int strcmp_icase(const char *a, const char *b) 60{ 61 return ignore_case ? strcasecmp(a, b) : strcmp(a, b); 62} 63 64int strncmp_icase(const char *a, const char *b, size_t count) 65{ 66 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count); 67} 68 69int fnmatch_icase(const char *pattern, const char *string, int flags) 70{ 71 return wildmatch(pattern, string, 72 flags | (ignore_case ? WM_CASEFOLD : 0), 73 NULL); 74} 75 76int git_fnmatch(const struct pathspec_item *item, 77 const char *pattern, const char *string, 78 int prefix) 79{ 80 if (prefix > 0) { 81 if (ps_strncmp(item, pattern, string, prefix)) 82 return WM_NOMATCH; 83 pattern += prefix; 84 string += prefix; 85 } 86 if (item->flags & PATHSPEC_ONESTAR) { 87 int pattern_len = strlen(++pattern); 88 int string_len = strlen(string); 89 return string_len < pattern_len || 90 ps_strcmp(item, pattern, 91 string + string_len - pattern_len); 92 } 93 if (item->magic & PATHSPEC_GLOB) 94 return wildmatch(pattern, string, 95 WM_PATHNAME | 96 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0), 97 NULL); 98 else 99 /* wildmatch has not learned no FNM_PATHNAME mode yet */ 100 return wildmatch(pattern, string, 101 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0, 102 NULL); 103} 104 105static int fnmatch_icase_mem(const char *pattern, int patternlen, 106 const char *string, int stringlen, 107 int flags) 108{ 109 int match_status; 110 struct strbuf pat_buf = STRBUF_INIT; 111 struct strbuf str_buf = STRBUF_INIT; 112 const char *use_pat = pattern; 113 const char *use_str = string; 114 115 if (pattern[patternlen]) { 116 strbuf_add(&pat_buf, pattern, patternlen); 117 use_pat = pat_buf.buf; 118 } 119 if (string[stringlen]) { 120 strbuf_add(&str_buf, string, stringlen); 121 use_str = str_buf.buf; 122 } 123 124 if (ignore_case) 125 flags |= WM_CASEFOLD; 126 match_status = wildmatch(use_pat, use_str, flags, NULL); 127 128 strbuf_release(&pat_buf); 129 strbuf_release(&str_buf); 130 131 return match_status; 132} 133 134static size_t common_prefix_len(const struct pathspec *pathspec) 135{ 136 int n; 137 size_t max = 0; 138 139 /* 140 * ":(icase)path" is treated as a pathspec full of 141 * wildcard. In other words, only prefix is considered common 142 * prefix. If the pathspec is abc/foo abc/bar, running in 143 * subdir xyz, the common prefix is still xyz, not xuz/abc as 144 * in non-:(icase). 145 */ 146 GUARD_PATHSPEC(pathspec, 147 PATHSPEC_FROMTOP | 148 PATHSPEC_MAXDEPTH | 149 PATHSPEC_LITERAL | 150 PATHSPEC_GLOB | 151 PATHSPEC_ICASE | 152 PATHSPEC_EXCLUDE); 153 154 for (n = 0; n < pathspec->nr; n++) { 155 size_t i = 0, len = 0, item_len; 156 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE) 157 continue; 158 if (pathspec->items[n].magic & PATHSPEC_ICASE) 159 item_len = pathspec->items[n].prefix; 160 else 161 item_len = pathspec->items[n].nowildcard_len; 162 while (i < item_len && (n == 0 || i < max)) { 163 char c = pathspec->items[n].match[i]; 164 if (c != pathspec->items[0].match[i]) 165 break; 166 if (c == '/') 167 len = i + 1; 168 i++; 169 } 170 if (n == 0 || len < max) { 171 max = len; 172 if (!max) 173 break; 174 } 175 } 176 return max; 177} 178 179/* 180 * Returns a copy of the longest leading path common among all 181 * pathspecs. 182 */ 183char *common_prefix(const struct pathspec *pathspec) 184{ 185 unsigned long len = common_prefix_len(pathspec); 186 187 return len ? xmemdupz(pathspec->items[0].match, len) : NULL; 188} 189 190int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec) 191{ 192 size_t len; 193 194 /* 195 * Calculate common prefix for the pathspec, and 196 * use that to optimize the directory walk 197 */ 198 len = common_prefix_len(pathspec); 199 200 /* Read the directory and prune it */ 201 read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec); 202 return len; 203} 204 205int within_depth(const char *name, int namelen, 206 int depth, int max_depth) 207{ 208 const char *cp = name, *cpe = name + namelen; 209 210 while (cp < cpe) { 211 if (*cp++ != '/') 212 continue; 213 depth++; 214 if (depth > max_depth) 215 return 0; 216 } 217 return 1; 218} 219 220#define DO_MATCH_EXCLUDE 1 221#define DO_MATCH_DIRECTORY 2 222 223/* 224 * Does 'match' match the given name? 225 * A match is found if 226 * 227 * (1) the 'match' string is leading directory of 'name', or 228 * (2) the 'match' string is a wildcard and matches 'name', or 229 * (3) the 'match' string is exactly the same as 'name'. 230 * 231 * and the return value tells which case it was. 232 * 233 * It returns 0 when there is no match. 234 */ 235static int match_pathspec_item(const struct pathspec_item *item, int prefix, 236 const char *name, int namelen, unsigned flags) 237{ 238 /* name/namelen has prefix cut off by caller */ 239 const char *match = item->match + prefix; 240 int matchlen = item->len - prefix; 241 242 /* 243 * The normal call pattern is: 244 * 1. prefix = common_prefix_len(ps); 245 * 2. prune something, or fill_directory 246 * 3. match_pathspec() 247 * 248 * 'prefix' at #1 may be shorter than the command's prefix and 249 * it's ok for #2 to match extra files. Those extras will be 250 * trimmed at #3. 251 * 252 * Suppose the pathspec is 'foo' and '../bar' running from 253 * subdir 'xyz'. The common prefix at #1 will be empty, thanks 254 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The 255 * user does not want XYZ/foo, only the "foo" part should be 256 * case-insensitive. We need to filter out XYZ/foo here. In 257 * other words, we do not trust the caller on comparing the 258 * prefix part when :(icase) is involved. We do exact 259 * comparison ourselves. 260 * 261 * Normally the caller (common_prefix_len() in fact) does 262 * _exact_ matching on name[-prefix+1..-1] and we do not need 263 * to check that part. Be defensive and check it anyway, in 264 * case common_prefix_len is changed, or a new caller is 265 * introduced that does not use common_prefix_len. 266 * 267 * If the penalty turns out too high when prefix is really 268 * long, maybe change it to 269 * strncmp(match, name, item->prefix - prefix) 270 */ 271 if (item->prefix && (item->magic & PATHSPEC_ICASE) && 272 strncmp(item->match, name - prefix, item->prefix)) 273 return 0; 274 275 /* If the match was just the prefix, we matched */ 276 if (!*match) 277 return MATCHED_RECURSIVELY; 278 279 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) { 280 if (matchlen == namelen) 281 return MATCHED_EXACTLY; 282 283 if (match[matchlen-1] == '/' || name[matchlen] == '/') 284 return MATCHED_RECURSIVELY; 285 } else if ((flags & DO_MATCH_DIRECTORY) && 286 match[matchlen - 1] == '/' && 287 namelen == matchlen - 1 && 288 !ps_strncmp(item, match, name, namelen)) 289 return MATCHED_EXACTLY; 290 291 if (item->nowildcard_len < item->len && 292 !git_fnmatch(item, match, name, 293 item->nowildcard_len - prefix)) 294 return MATCHED_FNMATCH; 295 296 return 0; 297} 298 299/* 300 * Given a name and a list of pathspecs, returns the nature of the 301 * closest (i.e. most specific) match of the name to any of the 302 * pathspecs. 303 * 304 * The caller typically calls this multiple times with the same 305 * pathspec and seen[] array but with different name/namelen 306 * (e.g. entries from the index) and is interested in seeing if and 307 * how each pathspec matches all the names it calls this function 308 * with. A mark is left in the seen[] array for each pathspec element 309 * indicating the closest type of match that element achieved, so if 310 * seen[n] remains zero after multiple invocations, that means the nth 311 * pathspec did not match any names, which could indicate that the 312 * user mistyped the nth pathspec. 313 */ 314static int do_match_pathspec(const struct pathspec *ps, 315 const char *name, int namelen, 316 int prefix, char *seen, 317 unsigned flags) 318{ 319 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE; 320 321 GUARD_PATHSPEC(ps, 322 PATHSPEC_FROMTOP | 323 PATHSPEC_MAXDEPTH | 324 PATHSPEC_LITERAL | 325 PATHSPEC_GLOB | 326 PATHSPEC_ICASE | 327 PATHSPEC_EXCLUDE); 328 329 if (!ps->nr) { 330 if (!ps->recursive || 331 !(ps->magic & PATHSPEC_MAXDEPTH) || 332 ps->max_depth == -1) 333 return MATCHED_RECURSIVELY; 334 335 if (within_depth(name, namelen, 0, ps->max_depth)) 336 return MATCHED_EXACTLY; 337 else 338 return 0; 339 } 340 341 name += prefix; 342 namelen -= prefix; 343 344 for (i = ps->nr - 1; i >= 0; i--) { 345 int how; 346 347 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) || 348 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE))) 349 continue; 350 351 if (seen && seen[i] == MATCHED_EXACTLY) 352 continue; 353 /* 354 * Make exclude patterns optional and never report 355 * "pathspec ':(exclude)foo' matches no files" 356 */ 357 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE) 358 seen[i] = MATCHED_FNMATCH; 359 how = match_pathspec_item(ps->items+i, prefix, name, 360 namelen, flags); 361 if (ps->recursive && 362 (ps->magic & PATHSPEC_MAXDEPTH) && 363 ps->max_depth != -1 && 364 how && how != MATCHED_FNMATCH) { 365 int len = ps->items[i].len; 366 if (name[len] == '/') 367 len++; 368 if (within_depth(name+len, namelen-len, 0, ps->max_depth)) 369 how = MATCHED_EXACTLY; 370 else 371 how = 0; 372 } 373 if (how) { 374 if (retval < how) 375 retval = how; 376 if (seen && seen[i] < how) 377 seen[i] = how; 378 } 379 } 380 return retval; 381} 382 383int match_pathspec(const struct pathspec *ps, 384 const char *name, int namelen, 385 int prefix, char *seen, int is_dir) 386{ 387 int positive, negative; 388 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0; 389 positive = do_match_pathspec(ps, name, namelen, 390 prefix, seen, flags); 391 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive) 392 return positive; 393 negative = do_match_pathspec(ps, name, namelen, 394 prefix, seen, 395 flags | DO_MATCH_EXCLUDE); 396 return negative ? 0 : positive; 397} 398 399int report_path_error(const char *ps_matched, 400 const struct pathspec *pathspec, 401 const char *prefix) 402{ 403 /* 404 * Make sure all pathspec matched; otherwise it is an error. 405 */ 406 int num, errors = 0; 407 for (num = 0; num < pathspec->nr; num++) { 408 int other, found_dup; 409 410 if (ps_matched[num]) 411 continue; 412 /* 413 * The caller might have fed identical pathspec 414 * twice. Do not barf on such a mistake. 415 * FIXME: parse_pathspec should have eliminated 416 * duplicate pathspec. 417 */ 418 for (found_dup = other = 0; 419 !found_dup && other < pathspec->nr; 420 other++) { 421 if (other == num || !ps_matched[other]) 422 continue; 423 if (!strcmp(pathspec->items[other].original, 424 pathspec->items[num].original)) 425 /* 426 * Ok, we have a match already. 427 */ 428 found_dup = 1; 429 } 430 if (found_dup) 431 continue; 432 433 error("pathspec '%s' did not match any file(s) known to git.", 434 pathspec->items[num].original); 435 errors++; 436 } 437 return errors; 438} 439 440/* 441 * Return the length of the "simple" part of a path match limiter. 442 */ 443int simple_length(const char *match) 444{ 445 int len = -1; 446 447 for (;;) { 448 unsigned char c = *match++; 449 len++; 450 if (c == '\0' || is_glob_special(c)) 451 return len; 452 } 453} 454 455int no_wildcard(const char *string) 456{ 457 return string[simple_length(string)] == '\0'; 458} 459 460void parse_exclude_pattern(const char **pattern, 461 int *patternlen, 462 int *flags, 463 int *nowildcardlen) 464{ 465 const char *p = *pattern; 466 size_t i, len; 467 468 *flags = 0; 469 if (*p == '!') { 470 *flags |= EXC_FLAG_NEGATIVE; 471 p++; 472 } 473 len = strlen(p); 474 if (len && p[len - 1] == '/') { 475 len--; 476 *flags |= EXC_FLAG_MUSTBEDIR; 477 } 478 for (i = 0; i < len; i++) { 479 if (p[i] == '/') 480 break; 481 } 482 if (i == len) 483 *flags |= EXC_FLAG_NODIR; 484 *nowildcardlen = simple_length(p); 485 /* 486 * we should have excluded the trailing slash from 'p' too, 487 * but that's one more allocation. Instead just make sure 488 * nowildcardlen does not exceed real patternlen 489 */ 490 if (*nowildcardlen > len) 491 *nowildcardlen = len; 492 if (*p == '*' && no_wildcard(p + 1)) 493 *flags |= EXC_FLAG_ENDSWITH; 494 *pattern = p; 495 *patternlen = len; 496} 497 498void add_exclude(const char *string, const char *base, 499 int baselen, struct exclude_list *el, int srcpos) 500{ 501 struct exclude *x; 502 int patternlen; 503 int flags; 504 int nowildcardlen; 505 506 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen); 507 if (flags & EXC_FLAG_MUSTBEDIR) { 508 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen); 509 } else { 510 x = xmalloc(sizeof(*x)); 511 x->pattern = string; 512 } 513 x->patternlen = patternlen; 514 x->nowildcardlen = nowildcardlen; 515 x->base = base; 516 x->baselen = baselen; 517 x->flags = flags; 518 x->srcpos = srcpos; 519 string_list_init(&x->sticky_paths, 1); 520 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc); 521 el->excludes[el->nr++] = x; 522 x->el = el; 523} 524 525static void *read_skip_worktree_file_from_index(const char *path, size_t *size, 526 struct sha1_stat *sha1_stat) 527{ 528 int pos, len; 529 unsigned long sz; 530 enum object_type type; 531 void *data; 532 533 len = strlen(path); 534 pos = cache_name_pos(path, len); 535 if (pos < 0) 536 return NULL; 537 if (!ce_skip_worktree(active_cache[pos])) 538 return NULL; 539 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz); 540 if (!data || type != OBJ_BLOB) { 541 free(data); 542 return NULL; 543 } 544 *size = xsize_t(sz); 545 if (sha1_stat) { 546 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat)); 547 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1); 548 } 549 return data; 550} 551 552/* 553 * Frees memory within el which was allocated for exclude patterns and 554 * the file buffer. Does not free el itself. 555 */ 556void clear_exclude_list(struct exclude_list *el) 557{ 558 int i; 559 560 for (i = 0; i < el->nr; i++) { 561 string_list_clear(&el->excludes[i]->sticky_paths, 0); 562 free(el->excludes[i]); 563 } 564 free(el->excludes); 565 free(el->filebuf); 566 567 memset(el, 0, sizeof(*el)); 568} 569 570static void trim_trailing_spaces(char *buf) 571{ 572 char *p, *last_space = NULL; 573 574 for (p = buf; *p; p++) 575 switch (*p) { 576 case ' ': 577 if (!last_space) 578 last_space = p; 579 break; 580 case '\\': 581 p++; 582 if (!*p) 583 return; 584 /* fallthrough */ 585 default: 586 last_space = NULL; 587 } 588 589 if (last_space) 590 *last_space = '\0'; 591} 592 593/* 594 * Given a subdirectory name and "dir" of the current directory, 595 * search the subdir in "dir" and return it, or create a new one if it 596 * does not exist in "dir". 597 * 598 * If "name" has the trailing slash, it'll be excluded in the search. 599 */ 600static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc, 601 struct untracked_cache_dir *dir, 602 const char *name, int len) 603{ 604 int first, last; 605 struct untracked_cache_dir *d; 606 if (!dir) 607 return NULL; 608 if (len && name[len - 1] == '/') 609 len--; 610 first = 0; 611 last = dir->dirs_nr; 612 while (last > first) { 613 int cmp, next = (last + first) >> 1; 614 d = dir->dirs[next]; 615 cmp = strncmp(name, d->name, len); 616 if (!cmp && strlen(d->name) > len) 617 cmp = -1; 618 if (!cmp) 619 return d; 620 if (cmp < 0) { 621 last = next; 622 continue; 623 } 624 first = next+1; 625 } 626 627 uc->dir_created++; 628 FLEX_ALLOC_MEM(d, name, name, len); 629 630 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc); 631 memmove(dir->dirs + first + 1, dir->dirs + first, 632 (dir->dirs_nr - first) * sizeof(*dir->dirs)); 633 dir->dirs_nr++; 634 dir->dirs[first] = d; 635 return d; 636} 637 638static void do_invalidate_gitignore(struct untracked_cache_dir *dir) 639{ 640 int i; 641 dir->valid = 0; 642 dir->untracked_nr = 0; 643 for (i = 0; i < dir->dirs_nr; i++) 644 do_invalidate_gitignore(dir->dirs[i]); 645} 646 647static void invalidate_gitignore(struct untracked_cache *uc, 648 struct untracked_cache_dir *dir) 649{ 650 uc->gitignore_invalidated++; 651 do_invalidate_gitignore(dir); 652} 653 654static void invalidate_directory(struct untracked_cache *uc, 655 struct untracked_cache_dir *dir) 656{ 657 int i; 658 uc->dir_invalidated++; 659 dir->valid = 0; 660 dir->untracked_nr = 0; 661 for (i = 0; i < dir->dirs_nr; i++) 662 dir->dirs[i]->recurse = 0; 663} 664 665/* 666 * Given a file with name "fname", read it (either from disk, or from 667 * the index if "check_index" is non-zero), parse it and store the 668 * exclude rules in "el". 669 * 670 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill 671 * stat data from disk (only valid if add_excludes returns zero). If 672 * ss_valid is non-zero, "ss" must contain good value as input. 673 */ 674static int add_excludes(const char *fname, const char *base, int baselen, 675 struct exclude_list *el, int check_index, 676 struct sha1_stat *sha1_stat) 677{ 678 struct stat st; 679 int fd, i, lineno = 1; 680 size_t size = 0; 681 char *buf, *entry; 682 683 fd = open(fname, O_RDONLY); 684 if (fd < 0 || fstat(fd, &st) < 0) { 685 if (errno != ENOENT) 686 warn_on_inaccessible(fname); 687 if (0 <= fd) 688 close(fd); 689 if (!check_index || 690 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL) 691 return -1; 692 if (size == 0) { 693 free(buf); 694 return 0; 695 } 696 if (buf[size-1] != '\n') { 697 buf = xrealloc(buf, st_add(size, 1)); 698 buf[size++] = '\n'; 699 } 700 } else { 701 size = xsize_t(st.st_size); 702 if (size == 0) { 703 if (sha1_stat) { 704 fill_stat_data(&sha1_stat->stat, &st); 705 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN); 706 sha1_stat->valid = 1; 707 } 708 close(fd); 709 return 0; 710 } 711 buf = xmallocz(size); 712 if (read_in_full(fd, buf, size) != size) { 713 free(buf); 714 close(fd); 715 return -1; 716 } 717 buf[size++] = '\n'; 718 close(fd); 719 if (sha1_stat) { 720 int pos; 721 if (sha1_stat->valid && 722 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st)) 723 ; /* no content change, ss->sha1 still good */ 724 else if (check_index && 725 (pos = cache_name_pos(fname, strlen(fname))) >= 0 && 726 !ce_stage(active_cache[pos]) && 727 ce_uptodate(active_cache[pos]) && 728 !would_convert_to_git(fname)) 729 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1); 730 else 731 hash_sha1_file(buf, size, "blob", sha1_stat->sha1); 732 fill_stat_data(&sha1_stat->stat, &st); 733 sha1_stat->valid = 1; 734 } 735 } 736 737 el->filebuf = buf; 738 739 if (skip_utf8_bom(&buf, size)) 740 size -= buf - el->filebuf; 741 742 entry = buf; 743 744 for (i = 0; i < size; i++) { 745 if (buf[i] == '\n') { 746 if (entry != buf + i && entry[0] != '#') { 747 buf[i - (i && buf[i-1] == '\r')] = 0; 748 trim_trailing_spaces(entry); 749 add_exclude(entry, base, baselen, el, lineno); 750 } 751 lineno++; 752 entry = buf + i + 1; 753 } 754 } 755 return 0; 756} 757 758int add_excludes_from_file_to_list(const char *fname, const char *base, 759 int baselen, struct exclude_list *el, 760 int check_index) 761{ 762 return add_excludes(fname, base, baselen, el, check_index, NULL); 763} 764 765struct exclude_list *add_exclude_list(struct dir_struct *dir, 766 int group_type, const char *src) 767{ 768 struct exclude_list *el; 769 struct exclude_list_group *group; 770 771 group = &dir->exclude_list_group[group_type]; 772 ALLOC_GROW(group->el, group->nr + 1, group->alloc); 773 el = &group->el[group->nr++]; 774 memset(el, 0, sizeof(*el)); 775 el->src = src; 776 return el; 777} 778 779/* 780 * Used to set up core.excludesfile and .git/info/exclude lists. 781 */ 782static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname, 783 struct sha1_stat *sha1_stat) 784{ 785 struct exclude_list *el; 786 /* 787 * catch setup_standard_excludes() that's called before 788 * dir->untracked is assigned. That function behaves 789 * differently when dir->untracked is non-NULL. 790 */ 791 if (!dir->untracked) 792 dir->unmanaged_exclude_files++; 793 el = add_exclude_list(dir, EXC_FILE, fname); 794 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0) 795 die("cannot use %s as an exclude file", fname); 796} 797 798void add_excludes_from_file(struct dir_struct *dir, const char *fname) 799{ 800 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */ 801 add_excludes_from_file_1(dir, fname, NULL); 802} 803 804int match_basename(const char *basename, int basenamelen, 805 const char *pattern, int prefix, int patternlen, 806 int flags) 807{ 808 if (prefix == patternlen) { 809 if (patternlen == basenamelen && 810 !strncmp_icase(pattern, basename, basenamelen)) 811 return 1; 812 } else if (flags & EXC_FLAG_ENDSWITH) { 813 /* "*literal" matching against "fooliteral" */ 814 if (patternlen - 1 <= basenamelen && 815 !strncmp_icase(pattern + 1, 816 basename + basenamelen - (patternlen - 1), 817 patternlen - 1)) 818 return 1; 819 } else { 820 if (fnmatch_icase_mem(pattern, patternlen, 821 basename, basenamelen, 822 0) == 0) 823 return 1; 824 } 825 return 0; 826} 827 828int match_pathname(const char *pathname, int pathlen, 829 const char *base, int baselen, 830 const char *pattern, int prefix, int patternlen, 831 int flags) 832{ 833 const char *name; 834 int namelen; 835 836 /* 837 * match with FNM_PATHNAME; the pattern has base implicitly 838 * in front of it. 839 */ 840 if (*pattern == '/') { 841 pattern++; 842 patternlen--; 843 prefix--; 844 } 845 846 /* 847 * baselen does not count the trailing slash. base[] may or 848 * may not end with a trailing slash though. 849 */ 850 if (pathlen < baselen + 1 || 851 (baselen && pathname[baselen] != '/') || 852 strncmp_icase(pathname, base, baselen)) 853 return 0; 854 855 namelen = baselen ? pathlen - baselen - 1 : pathlen; 856 name = pathname + pathlen - namelen; 857 858 if (prefix) { 859 /* 860 * if the non-wildcard part is longer than the 861 * remaining pathname, surely it cannot match. 862 */ 863 if (prefix > namelen) 864 return 0; 865 866 if (strncmp_icase(pattern, name, prefix)) 867 return 0; 868 pattern += prefix; 869 patternlen -= prefix; 870 name += prefix; 871 namelen -= prefix; 872 873 /* 874 * If the whole pattern did not have a wildcard, 875 * then our prefix match is all we need; we 876 * do not need to call fnmatch at all. 877 */ 878 if (!patternlen && (!namelen || *name == '/')) 879 return 1; 880 } 881 882 return fnmatch_icase_mem(pattern, patternlen, 883 name, namelen, 884 WM_PATHNAME) == 0; 885} 886 887static void add_sticky(struct exclude *exc, const char *pathname, int pathlen) 888{ 889 struct strbuf sb = STRBUF_INIT; 890 int i; 891 892 for (i = exc->sticky_paths.nr - 1; i >= 0; i--) { 893 const char *sticky = exc->sticky_paths.items[i].string; 894 int len = strlen(sticky); 895 896 if (pathlen < len && sticky[pathlen] == '/' && 897 !strncmp(pathname, sticky, pathlen)) 898 return; 899 } 900 901 strbuf_add(&sb, pathname, pathlen); 902 string_list_append_nodup(&exc->sticky_paths, strbuf_detach(&sb, NULL)); 903} 904 905static int match_sticky(struct exclude *exc, const char *pathname, int pathlen, int dtype) 906{ 907 int i; 908 909 for (i = exc->sticky_paths.nr - 1; i >= 0; i--) { 910 const char *sticky = exc->sticky_paths.items[i].string; 911 int len = strlen(sticky); 912 913 if (pathlen == len && dtype == DT_DIR && 914 !strncmp(pathname, sticky, len)) 915 return 1; 916 917 if (pathlen > len && pathname[len] == '/' && 918 !strncmp(pathname, sticky, len)) 919 return 1; 920 } 921 922 return 0; 923} 924 925static inline int different_decisions(const struct exclude *a, 926 const struct exclude *b) 927{ 928 return (a->flags & EXC_FLAG_NEGATIVE) != (b->flags & EXC_FLAG_NEGATIVE); 929} 930 931/* 932 * Return non-zero if pathname is a directory and an ancestor of the 933 * literal path in a pattern. 934 */ 935static int match_directory_part(const char *pathname, int pathlen, 936 int *dtype, struct exclude *x) 937{ 938 const char *base = x->base; 939 int baselen = x->baselen ? x->baselen - 1 : 0; 940 const char *pattern = x->pattern; 941 int prefix = x->nowildcardlen; 942 int patternlen = x->patternlen; 943 944 if (*dtype == DT_UNKNOWN) 945 *dtype = get_dtype(NULL, pathname, pathlen); 946 if (*dtype != DT_DIR) 947 return 0; 948 949 if (*pattern == '/') { 950 pattern++; 951 patternlen--; 952 prefix--; 953 } 954 955 if (baselen) { 956 if (((pathlen < baselen && base[pathlen] == '/') || 957 pathlen == baselen) && 958 !strncmp_icase(pathname, base, pathlen)) 959 return 1; 960 pathname += baselen + 1; 961 pathlen -= baselen + 1; 962 } 963 964 965 if (prefix && 966 (((pathlen < prefix && pattern[pathlen] == '/') || 967 pathlen == prefix) && 968 !strncmp_icase(pathname, pattern, pathlen))) 969 return 1; 970 971 return 0; 972} 973 974static struct exclude *should_descend(const char *pathname, int pathlen, 975 int *dtype, struct exclude_list *el, 976 struct exclude *exc) 977{ 978 int i; 979 980 for (i = el->nr - 1; 0 <= i; i--) { 981 struct exclude *x = el->excludes[i]; 982 983 if (x == exc) 984 break; 985 986 if (!(x->flags & EXC_FLAG_NODIR) && 987 different_decisions(x, exc) && 988 match_directory_part(pathname, pathlen, dtype, x)) 989 return x; 990 } 991 return NULL; 992} 993 994/* 995 * Scan the given exclude list in reverse to see whether pathname 996 * should be ignored. The first match (i.e. the last on the list), if 997 * any, determines the fate. Returns the exclude_list element which 998 * matched, or NULL for undecided. 999 */1000static struct exclude *last_exclude_matching_from_list(const char *pathname,1001 int pathlen,1002 const char *basename,1003 int *dtype,1004 struct exclude_list *el)1005{1006 struct exclude *exc = NULL; /* undecided */1007 int i, maybe_descend = 0;10081009 if (!el->nr)1010 return NULL; /* undefined */10111012 trace_printf_key(&trace_exclude, "exclude: from %s\n", el->src);10131014 for (i = el->nr - 1; 0 <= i; i--) {1015 struct exclude *x = el->excludes[i];1016 const char *exclude = x->pattern;1017 int prefix = x->nowildcardlen;10181019 if (!maybe_descend && i < el->nr - 1 &&1020 different_decisions(x, el->excludes[i+1]))1021 maybe_descend = 1;10221023 if (x->sticky_paths.nr) {1024 if (*dtype == DT_UNKNOWN)1025 *dtype = get_dtype(NULL, pathname, pathlen);1026 if (match_sticky(x, pathname, pathlen, *dtype)) {1027 exc = x;1028 break;1029 }1030 continue;1031 }10321033 if (x->flags & EXC_FLAG_MUSTBEDIR) {1034 if (*dtype == DT_UNKNOWN)1035 *dtype = get_dtype(NULL, pathname, pathlen);1036 if (*dtype != DT_DIR)1037 continue;1038 }10391040 if (x->flags & EXC_FLAG_NODIR) {1041 if (match_basename(basename,1042 pathlen - (basename - pathname),1043 exclude, prefix, x->patternlen,1044 x->flags)) {1045 exc = x;1046 break;1047 }1048 continue;1049 }10501051 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');1052 if (match_pathname(pathname, pathlen,1053 x->base, x->baselen ? x->baselen - 1 : 0,1054 exclude, prefix, x->patternlen, x->flags)) {1055 exc = x;1056 break;1057 }1058 }10591060 if (!exc) {1061 trace_printf_key(&trace_exclude, "exclude: %.*s => n/a\n",1062 pathlen, pathname);1063 return NULL;1064 }10651066 /*1067 * We have found a matching pattern "exc" that may exclude whole1068 * directory. We also found that there may be a pattern that matches1069 * something inside the directory and reincludes stuff.1070 *1071 * Go through the patterns again, find that pattern and double check.1072 * If it's true, return "undecided" and keep descending in. "exc" is1073 * marked sticky so that it continues to match inside the directory.1074 */1075 if (!(exc->flags & EXC_FLAG_NEGATIVE) && maybe_descend) {1076 struct exclude *x;10771078 if (*dtype == DT_UNKNOWN)1079 *dtype = get_dtype(NULL, pathname, pathlen);10801081 if (*dtype == DT_DIR &&1082 (x = should_descend(pathname, pathlen, dtype, el, exc))) {1083 add_sticky(exc, pathname, pathlen);1084 trace_printf_key(&trace_exclude,1085 "exclude: %.*s vs %s at line %d => %s,"1086 " forced open by %s at line %d => n/a\n",1087 pathlen, pathname, exc->pattern, exc->srcpos,1088 exc->flags & EXC_FLAG_NEGATIVE ? "no" : "yes",1089 x->pattern, x->srcpos);1090 return NULL;1091 }1092 }10931094 trace_printf_key(&trace_exclude, "exclude: %.*s vs %s at line %d => %s%s\n",1095 pathlen, pathname, exc->pattern, exc->srcpos,1096 exc->flags & EXC_FLAG_NEGATIVE ? "no" : "yes",1097 exc->sticky_paths.nr ? " (stuck)" : "");1098 return exc;1099}11001101/*1102 * Scan the list and let the last match determine the fate.1103 * Return 1 for exclude, 0 for include and -1 for undecided.1104 */1105int is_excluded_from_list(const char *pathname,1106 int pathlen, const char *basename, int *dtype,1107 struct exclude_list *el)1108{1109 struct exclude *exclude;1110 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);1111 if (exclude)1112 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1113 return -1; /* undecided */1114}11151116static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,1117 const char *pathname, int pathlen, const char *basename,1118 int *dtype_p)1119{1120 int i, j;1121 struct exclude_list_group *group;1122 struct exclude *exclude;1123 for (i = EXC_CMDL; i <= EXC_FILE; i++) {1124 group = &dir->exclude_list_group[i];1125 for (j = group->nr - 1; j >= 0; j--) {1126 exclude = last_exclude_matching_from_list(1127 pathname, pathlen, basename, dtype_p,1128 &group->el[j]);1129 if (exclude)1130 return exclude;1131 }1132 }1133 return NULL;1134}11351136/*1137 * Loads the per-directory exclude list for the substring of base1138 * which has a char length of baselen.1139 */1140static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)1141{1142 struct exclude_list_group *group;1143 struct exclude_list *el;1144 struct exclude_stack *stk = NULL;1145 struct untracked_cache_dir *untracked;1146 int current;11471148 group = &dir->exclude_list_group[EXC_DIRS];11491150 /*1151 * Pop the exclude lists from the EXCL_DIRS exclude_list_group1152 * which originate from directories not in the prefix of the1153 * path being checked.1154 */1155 while ((stk = dir->exclude_stack) != NULL) {1156 if (stk->baselen <= baselen &&1157 !strncmp(dir->basebuf.buf, base, stk->baselen))1158 break;1159 el = &group->el[dir->exclude_stack->exclude_ix];1160 dir->exclude_stack = stk->prev;1161 dir->exclude = NULL;1162 free((char *)el->src); /* see strbuf_detach() below */1163 clear_exclude_list(el);1164 free(stk);1165 group->nr--;1166 }11671168 /* Skip traversing into sub directories if the parent is excluded */1169 if (dir->exclude)1170 return;11711172 /*1173 * Lazy initialization. All call sites currently just1174 * memset(dir, 0, sizeof(*dir)) before use. Changing all of1175 * them seems lots of work for little benefit.1176 */1177 if (!dir->basebuf.buf)1178 strbuf_init(&dir->basebuf, PATH_MAX);11791180 /* Read from the parent directories and push them down. */1181 current = stk ? stk->baselen : -1;1182 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);1183 if (dir->untracked)1184 untracked = stk ? stk->ucd : dir->untracked->root;1185 else1186 untracked = NULL;11871188 while (current < baselen) {1189 const char *cp;1190 struct sha1_stat sha1_stat;11911192 stk = xcalloc(1, sizeof(*stk));1193 if (current < 0) {1194 cp = base;1195 current = 0;1196 } else {1197 cp = strchr(base + current + 1, '/');1198 if (!cp)1199 die("oops in prep_exclude");1200 cp++;1201 untracked =1202 lookup_untracked(dir->untracked, untracked,1203 base + current,1204 cp - base - current);1205 }1206 stk->prev = dir->exclude_stack;1207 stk->baselen = cp - base;1208 stk->exclude_ix = group->nr;1209 stk->ucd = untracked;1210 el = add_exclude_list(dir, EXC_DIRS, NULL);1211 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);1212 assert(stk->baselen == dir->basebuf.len);12131214 /* Abort if the directory is excluded */1215 if (stk->baselen) {1216 int dt = DT_DIR;1217 dir->basebuf.buf[stk->baselen - 1] = 0;1218 dir->exclude = last_exclude_matching_from_lists(dir,1219 dir->basebuf.buf, stk->baselen - 1,1220 dir->basebuf.buf + current, &dt);1221 dir->basebuf.buf[stk->baselen - 1] = '/';1222 if (dir->exclude &&1223 dir->exclude->flags & EXC_FLAG_NEGATIVE)1224 dir->exclude = NULL;1225 if (dir->exclude) {1226 dir->exclude_stack = stk;1227 return;1228 }1229 }12301231 /* Try to read per-directory file */1232 hashclr(sha1_stat.sha1);1233 sha1_stat.valid = 0;1234 if (dir->exclude_per_dir &&1235 /*1236 * If we know that no files have been added in1237 * this directory (i.e. valid_cached_dir() has1238 * been executed and set untracked->valid) ..1239 */1240 (!untracked || !untracked->valid ||1241 /*1242 * .. and .gitignore does not exist before1243 * (i.e. null exclude_sha1). Then we can skip1244 * loading .gitignore, which would result in1245 * ENOENT anyway.1246 */1247 !is_null_sha1(untracked->exclude_sha1))) {1248 /*1249 * dir->basebuf gets reused by the traversal, but we1250 * need fname to remain unchanged to ensure the src1251 * member of each struct exclude correctly1252 * back-references its source file. Other invocations1253 * of add_exclude_list provide stable strings, so we1254 * strbuf_detach() and free() here in the caller.1255 */1256 struct strbuf sb = STRBUF_INIT;1257 strbuf_addbuf(&sb, &dir->basebuf);1258 strbuf_addstr(&sb, dir->exclude_per_dir);1259 el->src = strbuf_detach(&sb, NULL);1260 add_excludes(el->src, el->src, stk->baselen, el, 1,1261 untracked ? &sha1_stat : NULL);1262 }1263 /*1264 * NEEDSWORK: when untracked cache is enabled, prep_exclude()1265 * will first be called in valid_cached_dir() then maybe many1266 * times more in last_exclude_matching(). When the cache is1267 * used, last_exclude_matching() will not be called and1268 * reading .gitignore content will be a waste.1269 *1270 * So when it's called by valid_cached_dir() and we can get1271 * .gitignore SHA-1 from the index (i.e. .gitignore is not1272 * modified on work tree), we could delay reading the1273 * .gitignore content until we absolutely need it in1274 * last_exclude_matching(). Be careful about ignore rule1275 * order, though, if you do that.1276 */1277 if (untracked &&1278 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {1279 invalidate_gitignore(dir->untracked, untracked);1280 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);1281 }1282 dir->exclude_stack = stk;1283 current = stk->baselen;1284 }1285 strbuf_setlen(&dir->basebuf, baselen);1286}12871288/*1289 * Loads the exclude lists for the directory containing pathname, then1290 * scans all exclude lists to determine whether pathname is excluded.1291 * Returns the exclude_list element which matched, or NULL for1292 * undecided.1293 */1294struct exclude *last_exclude_matching(struct dir_struct *dir,1295 const char *pathname,1296 int *dtype_p)1297{1298 int pathlen = strlen(pathname);1299 const char *basename = strrchr(pathname, '/');1300 basename = (basename) ? basename+1 : pathname;13011302 prep_exclude(dir, pathname, basename-pathname);13031304 if (dir->exclude)1305 return dir->exclude;13061307 return last_exclude_matching_from_lists(dir, pathname, pathlen,1308 basename, dtype_p);1309}13101311/*1312 * Loads the exclude lists for the directory containing pathname, then1313 * scans all exclude lists to determine whether pathname is excluded.1314 * Returns 1 if true, otherwise 0.1315 */1316int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)1317{1318 struct exclude *exclude =1319 last_exclude_matching(dir, pathname, dtype_p);1320 if (exclude)1321 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1322 return 0;1323}13241325static struct dir_entry *dir_entry_new(const char *pathname, int len)1326{1327 struct dir_entry *ent;13281329 FLEX_ALLOC_MEM(ent, name, pathname, len);1330 ent->len = len;1331 return ent;1332}13331334static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)1335{1336 if (cache_file_exists(pathname, len, ignore_case))1337 return NULL;13381339 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);1340 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);1341}13421343struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)1344{1345 if (!cache_name_is_other(pathname, len))1346 return NULL;13471348 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);1349 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);1350}13511352enum exist_status {1353 index_nonexistent = 0,1354 index_directory,1355 index_gitdir1356};13571358/*1359 * Do not use the alphabetically sorted index to look up1360 * the directory name; instead, use the case insensitive1361 * directory hash.1362 */1363static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)1364{1365 struct cache_entry *ce;13661367 if (cache_dir_exists(dirname, len))1368 return index_directory;13691370 ce = cache_file_exists(dirname, len, ignore_case);1371 if (ce && S_ISGITLINK(ce->ce_mode))1372 return index_gitdir;13731374 return index_nonexistent;1375}13761377/*1378 * The index sorts alphabetically by entry name, which1379 * means that a gitlink sorts as '\0' at the end, while1380 * a directory (which is defined not as an entry, but as1381 * the files it contains) will sort with the '/' at the1382 * end.1383 */1384static enum exist_status directory_exists_in_index(const char *dirname, int len)1385{1386 int pos;13871388 if (ignore_case)1389 return directory_exists_in_index_icase(dirname, len);13901391 pos = cache_name_pos(dirname, len);1392 if (pos < 0)1393 pos = -pos-1;1394 while (pos < active_nr) {1395 const struct cache_entry *ce = active_cache[pos++];1396 unsigned char endchar;13971398 if (strncmp(ce->name, dirname, len))1399 break;1400 endchar = ce->name[len];1401 if (endchar > '/')1402 break;1403 if (endchar == '/')1404 return index_directory;1405 if (!endchar && S_ISGITLINK(ce->ce_mode))1406 return index_gitdir;1407 }1408 return index_nonexistent;1409}14101411/*1412 * When we find a directory when traversing the filesystem, we1413 * have three distinct cases:1414 *1415 * - ignore it1416 * - see it as a directory1417 * - recurse into it1418 *1419 * and which one we choose depends on a combination of existing1420 * git index contents and the flags passed into the directory1421 * traversal routine.1422 *1423 * Case 1: If we *already* have entries in the index under that1424 * directory name, we always recurse into the directory to see1425 * all the files.1426 *1427 * Case 2: If we *already* have that directory name as a gitlink,1428 * we always continue to see it as a gitlink, regardless of whether1429 * there is an actual git directory there or not (it might not1430 * be checked out as a subproject!)1431 *1432 * Case 3: if we didn't have it in the index previously, we1433 * have a few sub-cases:1434 *1435 * (a) if "show_other_directories" is true, we show it as1436 * just a directory, unless "hide_empty_directories" is1437 * also true, in which case we need to check if it contains any1438 * untracked and / or ignored files.1439 * (b) if it looks like a git directory, and we don't have1440 * 'no_gitlinks' set we treat it as a gitlink, and show it1441 * as a directory.1442 * (c) otherwise, we recurse into it.1443 */1444static enum path_treatment treat_directory(struct dir_struct *dir,1445 struct untracked_cache_dir *untracked,1446 const char *dirname, int len, int baselen, int exclude,1447 const struct path_simplify *simplify)1448{1449 /* The "len-1" is to strip the final '/' */1450 switch (directory_exists_in_index(dirname, len-1)) {1451 case index_directory:1452 return path_recurse;14531454 case index_gitdir:1455 return path_none;14561457 case index_nonexistent:1458 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)1459 break;1460 if (!(dir->flags & DIR_NO_GITLINKS)) {1461 unsigned char sha1[20];1462 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)1463 return path_untracked;1464 }1465 return path_recurse;1466 }14671468 /* This is the "show_other_directories" case */14691470 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))1471 return exclude ? path_excluded : path_untracked;14721473 untracked = lookup_untracked(dir->untracked, untracked,1474 dirname + baselen, len - baselen);1475 return read_directory_recursive(dir, dirname, len,1476 untracked, 1, simplify);1477}14781479/*1480 * This is an inexact early pruning of any recursive directory1481 * reading - if the path cannot possibly be in the pathspec,1482 * return true, and we'll skip it early.1483 */1484static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)1485{1486 if (simplify) {1487 for (;;) {1488 const char *match = simplify->path;1489 int len = simplify->len;14901491 if (!match)1492 break;1493 if (len > pathlen)1494 len = pathlen;1495 if (!memcmp(path, match, len))1496 return 0;1497 simplify++;1498 }1499 return 1;1500 }1501 return 0;1502}15031504/*1505 * This function tells us whether an excluded path matches a1506 * list of "interesting" pathspecs. That is, whether a path matched1507 * by any of the pathspecs could possibly be ignored by excluding1508 * the specified path. This can happen if:1509 *1510 * 1. the path is mentioned explicitly in the pathspec1511 *1512 * 2. the path is a directory prefix of some element in the1513 * pathspec1514 */1515static int exclude_matches_pathspec(const char *path, int len,1516 const struct path_simplify *simplify)1517{1518 if (simplify) {1519 for (; simplify->path; simplify++) {1520 if (len == simplify->len1521 && !memcmp(path, simplify->path, len))1522 return 1;1523 if (len < simplify->len1524 && simplify->path[len] == '/'1525 && !memcmp(path, simplify->path, len))1526 return 1;1527 }1528 }1529 return 0;1530}15311532static int get_index_dtype(const char *path, int len)1533{1534 int pos;1535 const struct cache_entry *ce;15361537 ce = cache_file_exists(path, len, 0);1538 if (ce) {1539 if (!ce_uptodate(ce))1540 return DT_UNKNOWN;1541 if (S_ISGITLINK(ce->ce_mode))1542 return DT_DIR;1543 /*1544 * Nobody actually cares about the1545 * difference between DT_LNK and DT_REG1546 */1547 return DT_REG;1548 }15491550 /* Try to look it up as a directory */1551 pos = cache_name_pos(path, len);1552 if (pos >= 0)1553 return DT_UNKNOWN;1554 pos = -pos-1;1555 while (pos < active_nr) {1556 ce = active_cache[pos++];1557 if (strncmp(ce->name, path, len))1558 break;1559 if (ce->name[len] > '/')1560 break;1561 if (ce->name[len] < '/')1562 continue;1563 if (!ce_uptodate(ce))1564 break; /* continue? */1565 return DT_DIR;1566 }1567 return DT_UNKNOWN;1568}15691570static int get_dtype(struct dirent *de, const char *path, int len)1571{1572 int dtype = de ? DTYPE(de) : DT_UNKNOWN;1573 struct stat st;15741575 if (dtype != DT_UNKNOWN)1576 return dtype;1577 dtype = get_index_dtype(path, len);1578 if (dtype != DT_UNKNOWN)1579 return dtype;1580 if (lstat(path, &st))1581 return dtype;1582 if (S_ISREG(st.st_mode))1583 return DT_REG;1584 if (S_ISDIR(st.st_mode))1585 return DT_DIR;1586 if (S_ISLNK(st.st_mode))1587 return DT_LNK;1588 return dtype;1589}15901591static enum path_treatment treat_one_path(struct dir_struct *dir,1592 struct untracked_cache_dir *untracked,1593 struct strbuf *path,1594 int baselen,1595 const struct path_simplify *simplify,1596 int dtype, struct dirent *de)1597{1598 int exclude;1599 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);16001601 if (dtype == DT_UNKNOWN)1602 dtype = get_dtype(de, path->buf, path->len);16031604 /* Always exclude indexed files */1605 if (dtype != DT_DIR && has_path_in_index)1606 return path_none;16071608 /*1609 * When we are looking at a directory P in the working tree,1610 * there are three cases:1611 *1612 * (1) P exists in the index. Everything inside the directory P in1613 * the working tree needs to go when P is checked out from the1614 * index.1615 *1616 * (2) P does not exist in the index, but there is P/Q in the index.1617 * We know P will stay a directory when we check out the contents1618 * of the index, but we do not know yet if there is a directory1619 * P/Q in the working tree to be killed, so we need to recurse.1620 *1621 * (3) P does not exist in the index, and there is no P/Q in the index1622 * to require P to be a directory, either. Only in this case, we1623 * know that everything inside P will not be killed without1624 * recursing.1625 */1626 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&1627 (dtype == DT_DIR) &&1628 !has_path_in_index &&1629 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))1630 return path_none;16311632 exclude = is_excluded(dir, path->buf, &dtype);16331634 /*1635 * Excluded? If we don't explicitly want to show1636 * ignored files, ignore it1637 */1638 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))1639 return path_excluded;16401641 switch (dtype) {1642 default:1643 return path_none;1644 case DT_DIR:1645 strbuf_addch(path, '/');1646 return treat_directory(dir, untracked, path->buf, path->len,1647 baselen, exclude, simplify);1648 case DT_REG:1649 case DT_LNK:1650 return exclude ? path_excluded : path_untracked;1651 }1652}16531654static enum path_treatment treat_path_fast(struct dir_struct *dir,1655 struct untracked_cache_dir *untracked,1656 struct cached_dir *cdir,1657 struct strbuf *path,1658 int baselen,1659 const struct path_simplify *simplify)1660{1661 strbuf_setlen(path, baselen);1662 if (!cdir->ucd) {1663 strbuf_addstr(path, cdir->file);1664 return path_untracked;1665 }1666 strbuf_addstr(path, cdir->ucd->name);1667 /* treat_one_path() does this before it calls treat_directory() */1668 strbuf_complete(path, '/');1669 if (cdir->ucd->check_only)1670 /*1671 * check_only is set as a result of treat_directory() getting1672 * to its bottom. Verify again the same set of directories1673 * with check_only set.1674 */1675 return read_directory_recursive(dir, path->buf, path->len,1676 cdir->ucd, 1, simplify);1677 /*1678 * We get path_recurse in the first run when1679 * directory_exists_in_index() returns index_nonexistent. We1680 * are sure that new changes in the index does not impact the1681 * outcome. Return now.1682 */1683 return path_recurse;1684}16851686static enum path_treatment treat_path(struct dir_struct *dir,1687 struct untracked_cache_dir *untracked,1688 struct cached_dir *cdir,1689 struct strbuf *path,1690 int baselen,1691 const struct path_simplify *simplify)1692{1693 int dtype;1694 struct dirent *de = cdir->de;16951696 if (!de)1697 return treat_path_fast(dir, untracked, cdir, path,1698 baselen, simplify);1699 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))1700 return path_none;1701 strbuf_setlen(path, baselen);1702 strbuf_addstr(path, de->d_name);1703 if (simplify_away(path->buf, path->len, simplify))1704 return path_none;17051706 dtype = DTYPE(de);1707 return treat_one_path(dir, untracked, path, baselen, simplify, dtype, de);1708}17091710static void add_untracked(struct untracked_cache_dir *dir, const char *name)1711{1712 if (!dir)1713 return;1714 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,1715 dir->untracked_alloc);1716 dir->untracked[dir->untracked_nr++] = xstrdup(name);1717}17181719static int valid_cached_dir(struct dir_struct *dir,1720 struct untracked_cache_dir *untracked,1721 struct strbuf *path,1722 int check_only)1723{1724 struct stat st;17251726 if (!untracked)1727 return 0;17281729 if (stat(path->len ? path->buf : ".", &st)) {1730 invalidate_directory(dir->untracked, untracked);1731 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));1732 return 0;1733 }1734 if (!untracked->valid ||1735 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {1736 if (untracked->valid)1737 invalidate_directory(dir->untracked, untracked);1738 fill_stat_data(&untracked->stat_data, &st);1739 return 0;1740 }17411742 if (untracked->check_only != !!check_only) {1743 invalidate_directory(dir->untracked, untracked);1744 return 0;1745 }17461747 /*1748 * prep_exclude will be called eventually on this directory,1749 * but it's called much later in last_exclude_matching(). We1750 * need it now to determine the validity of the cache for this1751 * path. The next calls will be nearly no-op, the way1752 * prep_exclude() is designed.1753 */1754 if (path->len && path->buf[path->len - 1] != '/') {1755 strbuf_addch(path, '/');1756 prep_exclude(dir, path->buf, path->len);1757 strbuf_setlen(path, path->len - 1);1758 } else1759 prep_exclude(dir, path->buf, path->len);17601761 /* hopefully prep_exclude() haven't invalidated this entry... */1762 return untracked->valid;1763}17641765static int open_cached_dir(struct cached_dir *cdir,1766 struct dir_struct *dir,1767 struct untracked_cache_dir *untracked,1768 struct strbuf *path,1769 int check_only)1770{1771 memset(cdir, 0, sizeof(*cdir));1772 cdir->untracked = untracked;1773 if (valid_cached_dir(dir, untracked, path, check_only))1774 return 0;1775 cdir->fdir = opendir(path->len ? path->buf : ".");1776 if (dir->untracked)1777 dir->untracked->dir_opened++;1778 if (!cdir->fdir)1779 return -1;1780 return 0;1781}17821783static int read_cached_dir(struct cached_dir *cdir)1784{1785 if (cdir->fdir) {1786 cdir->de = readdir(cdir->fdir);1787 if (!cdir->de)1788 return -1;1789 return 0;1790 }1791 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {1792 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];1793 if (!d->recurse) {1794 cdir->nr_dirs++;1795 continue;1796 }1797 cdir->ucd = d;1798 cdir->nr_dirs++;1799 return 0;1800 }1801 cdir->ucd = NULL;1802 if (cdir->nr_files < cdir->untracked->untracked_nr) {1803 struct untracked_cache_dir *d = cdir->untracked;1804 cdir->file = d->untracked[cdir->nr_files++];1805 return 0;1806 }1807 return -1;1808}18091810static void close_cached_dir(struct cached_dir *cdir)1811{1812 if (cdir->fdir)1813 closedir(cdir->fdir);1814 /*1815 * We have gone through this directory and found no untracked1816 * entries. Mark it valid.1817 */1818 if (cdir->untracked) {1819 cdir->untracked->valid = 1;1820 cdir->untracked->recurse = 1;1821 }1822}18231824/*1825 * Read a directory tree. We currently ignore anything but1826 * directories, regular files and symlinks. That's because git1827 * doesn't handle them at all yet. Maybe that will change some1828 * day.1829 *1830 * Also, we ignore the name ".git" (even if it is not a directory).1831 * That likely will not change.1832 *1833 * Returns the most significant path_treatment value encountered in the scan.1834 */1835static enum path_treatment read_directory_recursive(struct dir_struct *dir,1836 const char *base, int baselen,1837 struct untracked_cache_dir *untracked, int check_only,1838 const struct path_simplify *simplify)1839{1840 struct cached_dir cdir;1841 enum path_treatment state, subdir_state, dir_state = path_none;1842 struct strbuf path = STRBUF_INIT;1843 static int level = 0;18441845 strbuf_add(&path, base, baselen);18461847 trace_printf_key(&trace_exclude, "exclude: [%d] enter '%.*s'\n",1848 level++, baselen, base);18491850 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))1851 goto out;18521853 if (untracked)1854 untracked->check_only = !!check_only;18551856 while (!read_cached_dir(&cdir)) {1857 /* check how the file or directory should be treated */1858 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);18591860 if (state > dir_state)1861 dir_state = state;18621863 /* recurse into subdir if instructed by treat_path */1864 if (state == path_recurse) {1865 struct untracked_cache_dir *ud;1866 ud = lookup_untracked(dir->untracked, untracked,1867 path.buf + baselen,1868 path.len - baselen);1869 subdir_state =1870 read_directory_recursive(dir, path.buf, path.len,1871 ud, check_only, simplify);1872 if (subdir_state > dir_state)1873 dir_state = subdir_state;1874 }18751876 if (check_only) {1877 /* abort early if maximum state has been reached */1878 if (dir_state == path_untracked) {1879 if (cdir.fdir)1880 add_untracked(untracked, path.buf + baselen);1881 break;1882 }1883 /* skip the dir_add_* part */1884 continue;1885 }18861887 /* add the path to the appropriate result list */1888 switch (state) {1889 case path_excluded:1890 if (dir->flags & DIR_SHOW_IGNORED)1891 dir_add_name(dir, path.buf, path.len);1892 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||1893 ((dir->flags & DIR_COLLECT_IGNORED) &&1894 exclude_matches_pathspec(path.buf, path.len,1895 simplify)))1896 dir_add_ignored(dir, path.buf, path.len);1897 break;18981899 case path_untracked:1900 if (dir->flags & DIR_SHOW_IGNORED)1901 break;1902 dir_add_name(dir, path.buf, path.len);1903 if (cdir.fdir)1904 add_untracked(untracked, path.buf + baselen);1905 break;19061907 default:1908 break;1909 }1910 }1911 close_cached_dir(&cdir);1912 out:1913 trace_printf_key(&trace_exclude, "exclude: [%d] leave '%.*s'\n",1914 --level, baselen, base);1915 strbuf_release(&path);19161917 return dir_state;1918}19191920static int cmp_name(const void *p1, const void *p2)1921{1922 const struct dir_entry *e1 = *(const struct dir_entry **)p1;1923 const struct dir_entry *e2 = *(const struct dir_entry **)p2;19241925 return name_compare(e1->name, e1->len, e2->name, e2->len);1926}19271928static struct path_simplify *create_simplify(const char **pathspec)1929{1930 int nr, alloc = 0;1931 struct path_simplify *simplify = NULL;19321933 if (!pathspec)1934 return NULL;19351936 for (nr = 0 ; ; nr++) {1937 const char *match;1938 ALLOC_GROW(simplify, nr + 1, alloc);1939 match = *pathspec++;1940 if (!match)1941 break;1942 simplify[nr].path = match;1943 simplify[nr].len = simple_length(match);1944 }1945 simplify[nr].path = NULL;1946 simplify[nr].len = 0;1947 return simplify;1948}19491950static void free_simplify(struct path_simplify *simplify)1951{1952 free(simplify);1953}19541955static int treat_leading_path(struct dir_struct *dir,1956 const char *path, int len,1957 const struct path_simplify *simplify)1958{1959 struct strbuf sb = STRBUF_INIT;1960 int baselen, rc = 0;1961 const char *cp;1962 int old_flags = dir->flags;19631964 while (len && path[len - 1] == '/')1965 len--;1966 if (!len)1967 return 1;1968 baselen = 0;1969 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;1970 while (1) {1971 cp = path + baselen + !!baselen;1972 cp = memchr(cp, '/', path + len - cp);1973 if (!cp)1974 baselen = len;1975 else1976 baselen = cp - path;1977 strbuf_setlen(&sb, 0);1978 strbuf_add(&sb, path, baselen);1979 if (!is_directory(sb.buf))1980 break;1981 if (simplify_away(sb.buf, sb.len, simplify))1982 break;1983 if (treat_one_path(dir, NULL, &sb, baselen, simplify,1984 DT_DIR, NULL) == path_none)1985 break; /* do not recurse into it */1986 if (len <= baselen) {1987 rc = 1;1988 break; /* finished checking */1989 }1990 }1991 strbuf_release(&sb);1992 dir->flags = old_flags;1993 return rc;1994}19951996static const char *get_ident_string(void)1997{1998 static struct strbuf sb = STRBUF_INIT;1999 struct utsname uts;20002001 if (sb.len)2002 return sb.buf;2003 if (uname(&uts) < 0)2004 die_errno(_("failed to get kernel name and information"));2005 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),2006 uts.sysname);2007 return sb.buf;2008}20092010static int ident_in_untracked(const struct untracked_cache *uc)2011{2012 /*2013 * Previous git versions may have saved many NUL separated2014 * strings in the "ident" field, but it is insane to manage2015 * many locations, so just take care of the first one.2016 */20172018 return !strcmp(uc->ident.buf, get_ident_string());2019}20202021static void set_untracked_ident(struct untracked_cache *uc)2022{2023 strbuf_reset(&uc->ident);2024 strbuf_addstr(&uc->ident, get_ident_string());20252026 /*2027 * This strbuf used to contain a list of NUL separated2028 * strings, so save NUL too for backward compatibility.2029 */2030 strbuf_addch(&uc->ident, 0);2031}20322033static void new_untracked_cache(struct index_state *istate)2034{2035 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));2036 strbuf_init(&uc->ident, 100);2037 uc->exclude_per_dir = ".gitignore";2038 /* should be the same flags used by git-status */2039 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;2040 set_untracked_ident(uc);2041 istate->untracked = uc;2042 istate->cache_changed |= UNTRACKED_CHANGED;2043}20442045void add_untracked_cache(struct index_state *istate)2046{2047 if (!istate->untracked) {2048 new_untracked_cache(istate);2049 } else {2050 if (!ident_in_untracked(istate->untracked)) {2051 free_untracked_cache(istate->untracked);2052 new_untracked_cache(istate);2053 }2054 }2055}20562057void remove_untracked_cache(struct index_state *istate)2058{2059 if (istate->untracked) {2060 free_untracked_cache(istate->untracked);2061 istate->untracked = NULL;2062 istate->cache_changed |= UNTRACKED_CHANGED;2063 }2064}20652066static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,2067 int base_len,2068 const struct pathspec *pathspec)2069{2070 struct untracked_cache_dir *root;20712072 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))2073 return NULL;20742075 /*2076 * We only support $GIT_DIR/info/exclude and core.excludesfile2077 * as the global ignore rule files. Any other additions2078 * (e.g. from command line) invalidate the cache. This2079 * condition also catches running setup_standard_excludes()2080 * before setting dir->untracked!2081 */2082 if (dir->unmanaged_exclude_files)2083 return NULL;20842085 /*2086 * Optimize for the main use case only: whole-tree git2087 * status. More work involved in treat_leading_path() if we2088 * use cache on just a subset of the worktree. pathspec2089 * support could make the matter even worse.2090 */2091 if (base_len || (pathspec && pathspec->nr))2092 return NULL;20932094 /* Different set of flags may produce different results */2095 if (dir->flags != dir->untracked->dir_flags ||2096 /*2097 * See treat_directory(), case index_nonexistent. Without2098 * this flag, we may need to also cache .git file content2099 * for the resolve_gitlink_ref() call, which we don't.2100 */2101 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||2102 /* We don't support collecting ignore files */2103 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |2104 DIR_COLLECT_IGNORED)))2105 return NULL;21062107 /*2108 * If we use .gitignore in the cache and now you change it to2109 * .gitexclude, everything will go wrong.2110 */2111 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&2112 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))2113 return NULL;21142115 /*2116 * EXC_CMDL is not considered in the cache. If people set it,2117 * skip the cache.2118 */2119 if (dir->exclude_list_group[EXC_CMDL].nr)2120 return NULL;21212122 if (!ident_in_untracked(dir->untracked)) {2123 warning(_("Untracked cache is disabled on this system or location."));2124 return NULL;2125 }21262127 if (!dir->untracked->root) {2128 const int len = sizeof(*dir->untracked->root);2129 dir->untracked->root = xmalloc(len);2130 memset(dir->untracked->root, 0, len);2131 }21322133 /* Validate $GIT_DIR/info/exclude and core.excludesfile */2134 root = dir->untracked->root;2135 if (hashcmp(dir->ss_info_exclude.sha1,2136 dir->untracked->ss_info_exclude.sha1)) {2137 invalidate_gitignore(dir->untracked, root);2138 dir->untracked->ss_info_exclude = dir->ss_info_exclude;2139 }2140 if (hashcmp(dir->ss_excludes_file.sha1,2141 dir->untracked->ss_excludes_file.sha1)) {2142 invalidate_gitignore(dir->untracked, root);2143 dir->untracked->ss_excludes_file = dir->ss_excludes_file;2144 }21452146 /* Make sure this directory is not dropped out at saving phase */2147 root->recurse = 1;2148 return root;2149}21502151static void clear_sticky(struct dir_struct *dir)2152{2153 struct exclude_list_group *g;2154 struct exclude_list *el;2155 struct exclude *x;2156 int i, j, k;21572158 for (i = EXC_CMDL; i <= EXC_FILE; i++) {2159 g = &dir->exclude_list_group[i];2160 for (j = g->nr - 1; j >= 0; j--) {2161 el = &g->el[j];2162 for (k = el->nr - 1; 0 <= k; k--) {2163 x = el->excludes[k];2164 string_list_clear(&x->sticky_paths, 0);2165 }2166 }2167 }2168}21692170int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)2171{2172 struct path_simplify *simplify;2173 struct untracked_cache_dir *untracked;21742175 /*2176 * Check out create_simplify()2177 */2178 if (pathspec)2179 GUARD_PATHSPEC(pathspec,2180 PATHSPEC_FROMTOP |2181 PATHSPEC_MAXDEPTH |2182 PATHSPEC_LITERAL |2183 PATHSPEC_GLOB |2184 PATHSPEC_ICASE |2185 PATHSPEC_EXCLUDE);21862187 if (has_symlink_leading_path(path, len))2188 return dir->nr;21892190 /*2191 * Stay on the safe side. if read_directory() has run once on2192 * "dir", some sticky flag may have been left. Clear them all.2193 */2194 clear_sticky(dir);21952196 /*2197 * exclude patterns are treated like positive ones in2198 * create_simplify. Usually exclude patterns should be a2199 * subset of positive ones, which has no impacts on2200 * create_simplify().2201 */2202 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);2203 untracked = validate_untracked_cache(dir, len, pathspec);2204 if (!untracked)2205 /*2206 * make sure untracked cache code path is disabled,2207 * e.g. prep_exclude()2208 */2209 dir->untracked = NULL;2210 if (!len || treat_leading_path(dir, path, len, simplify))2211 read_directory_recursive(dir, path, len, untracked, 0, simplify);2212 free_simplify(simplify);2213 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);2214 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);2215 if (dir->untracked) {2216 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);2217 trace_printf_key(&trace_untracked_stats,2218 "node creation: %u\n"2219 "gitignore invalidation: %u\n"2220 "directory invalidation: %u\n"2221 "opendir: %u\n",2222 dir->untracked->dir_created,2223 dir->untracked->gitignore_invalidated,2224 dir->untracked->dir_invalidated,2225 dir->untracked->dir_opened);2226 if (dir->untracked == the_index.untracked &&2227 (dir->untracked->dir_opened ||2228 dir->untracked->gitignore_invalidated ||2229 dir->untracked->dir_invalidated))2230 the_index.cache_changed |= UNTRACKED_CHANGED;2231 if (dir->untracked != the_index.untracked) {2232 free(dir->untracked);2233 dir->untracked = NULL;2234 }2235 }2236 return dir->nr;2237}22382239int file_exists(const char *f)2240{2241 struct stat sb;2242 return lstat(f, &sb) == 0;2243}22442245static int cmp_icase(char a, char b)2246{2247 if (a == b)2248 return 0;2249 if (ignore_case)2250 return toupper(a) - toupper(b);2251 return a - b;2252}22532254/*2255 * Given two normalized paths (a trailing slash is ok), if subdir is2256 * outside dir, return -1. Otherwise return the offset in subdir that2257 * can be used as relative path to dir.2258 */2259int dir_inside_of(const char *subdir, const char *dir)2260{2261 int offset = 0;22622263 assert(dir && subdir && *dir && *subdir);22642265 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {2266 dir++;2267 subdir++;2268 offset++;2269 }22702271 /* hel[p]/me vs hel[l]/yeah */2272 if (*dir && *subdir)2273 return -1;22742275 if (!*subdir)2276 return !*dir ? offset : -1; /* same dir */22772278 /* foo/[b]ar vs foo/[] */2279 if (is_dir_sep(dir[-1]))2280 return is_dir_sep(subdir[-1]) ? offset : -1;22812282 /* foo[/]bar vs foo[] */2283 return is_dir_sep(*subdir) ? offset + 1 : -1;2284}22852286int is_inside_dir(const char *dir)2287{2288 char *cwd;2289 int rc;22902291 if (!dir)2292 return 0;22932294 cwd = xgetcwd();2295 rc = (dir_inside_of(cwd, dir) >= 0);2296 free(cwd);2297 return rc;2298}22992300int is_empty_dir(const char *path)2301{2302 DIR *dir = opendir(path);2303 struct dirent *e;2304 int ret = 1;23052306 if (!dir)2307 return 0;23082309 while ((e = readdir(dir)) != NULL)2310 if (!is_dot_or_dotdot(e->d_name)) {2311 ret = 0;2312 break;2313 }23142315 closedir(dir);2316 return ret;2317}23182319static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)2320{2321 DIR *dir;2322 struct dirent *e;2323 int ret = 0, original_len = path->len, len, kept_down = 0;2324 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);2325 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);2326 unsigned char submodule_head[20];23272328 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&2329 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {2330 /* Do not descend and nuke a nested git work tree. */2331 if (kept_up)2332 *kept_up = 1;2333 return 0;2334 }23352336 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;2337 dir = opendir(path->buf);2338 if (!dir) {2339 if (errno == ENOENT)2340 return keep_toplevel ? -1 : 0;2341 else if (errno == EACCES && !keep_toplevel)2342 /*2343 * An empty dir could be removable even if it2344 * is unreadable:2345 */2346 return rmdir(path->buf);2347 else2348 return -1;2349 }2350 strbuf_complete(path, '/');23512352 len = path->len;2353 while ((e = readdir(dir)) != NULL) {2354 struct stat st;2355 if (is_dot_or_dotdot(e->d_name))2356 continue;23572358 strbuf_setlen(path, len);2359 strbuf_addstr(path, e->d_name);2360 if (lstat(path->buf, &st)) {2361 if (errno == ENOENT)2362 /*2363 * file disappeared, which is what we2364 * wanted anyway2365 */2366 continue;2367 /* fall thru */2368 } else if (S_ISDIR(st.st_mode)) {2369 if (!remove_dir_recurse(path, flag, &kept_down))2370 continue; /* happy */2371 } else if (!only_empty &&2372 (!unlink(path->buf) || errno == ENOENT)) {2373 continue; /* happy, too */2374 }23752376 /* path too long, stat fails, or non-directory still exists */2377 ret = -1;2378 break;2379 }2380 closedir(dir);23812382 strbuf_setlen(path, original_len);2383 if (!ret && !keep_toplevel && !kept_down)2384 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;2385 else if (kept_up)2386 /*2387 * report the uplevel that it is not an error that we2388 * did not rmdir() our directory.2389 */2390 *kept_up = !ret;2391 return ret;2392}23932394int remove_dir_recursively(struct strbuf *path, int flag)2395{2396 return remove_dir_recurse(path, flag, NULL);2397}23982399static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")24002401void setup_standard_excludes(struct dir_struct *dir)2402{2403 const char *path;24042405 dir->exclude_per_dir = ".gitignore";24062407 /* core.excludefile defaulting to $XDG_HOME/git/ignore */2408 if (!excludes_file)2409 excludes_file = xdg_config_home("ignore");2410 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))2411 add_excludes_from_file_1(dir, excludes_file,2412 dir->untracked ? &dir->ss_excludes_file : NULL);24132414 /* per repository user preference */2415 path = git_path_info_exclude();2416 if (!access_or_warn(path, R_OK, 0))2417 add_excludes_from_file_1(dir, path,2418 dir->untracked ? &dir->ss_info_exclude : NULL);2419}24202421int remove_path(const char *name)2422{2423 char *slash;24242425 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)2426 return -1;24272428 slash = strrchr(name, '/');2429 if (slash) {2430 char *dirs = xstrdup(name);2431 slash = dirs + (slash - name);2432 do {2433 *slash = '\0';2434 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));2435 free(dirs);2436 }2437 return 0;2438}24392440/*2441 * Frees memory within dir which was allocated for exclude lists and2442 * the exclude_stack. Does not free dir itself.2443 */2444void clear_directory(struct dir_struct *dir)2445{2446 int i, j;2447 struct exclude_list_group *group;2448 struct exclude_list *el;2449 struct exclude_stack *stk;24502451 for (i = EXC_CMDL; i <= EXC_FILE; i++) {2452 group = &dir->exclude_list_group[i];2453 for (j = 0; j < group->nr; j++) {2454 el = &group->el[j];2455 if (i == EXC_DIRS)2456 free((char *)el->src);2457 clear_exclude_list(el);2458 }2459 free(group->el);2460 }24612462 stk = dir->exclude_stack;2463 while (stk) {2464 struct exclude_stack *prev = stk->prev;2465 free(stk);2466 stk = prev;2467 }2468 strbuf_release(&dir->basebuf);2469}24702471struct ondisk_untracked_cache {2472 struct stat_data info_exclude_stat;2473 struct stat_data excludes_file_stat;2474 uint32_t dir_flags;2475 unsigned char info_exclude_sha1[20];2476 unsigned char excludes_file_sha1[20];2477 char exclude_per_dir[FLEX_ARRAY];2478};24792480#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)24812482struct write_data {2483 int index; /* number of written untracked_cache_dir */2484 struct ewah_bitmap *check_only; /* from untracked_cache_dir */2485 struct ewah_bitmap *valid; /* from untracked_cache_dir */2486 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */2487 struct strbuf out;2488 struct strbuf sb_stat;2489 struct strbuf sb_sha1;2490};24912492static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)2493{2494 to->sd_ctime.sec = htonl(from->sd_ctime.sec);2495 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);2496 to->sd_mtime.sec = htonl(from->sd_mtime.sec);2497 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);2498 to->sd_dev = htonl(from->sd_dev);2499 to->sd_ino = htonl(from->sd_ino);2500 to->sd_uid = htonl(from->sd_uid);2501 to->sd_gid = htonl(from->sd_gid);2502 to->sd_size = htonl(from->sd_size);2503}25042505static void write_one_dir(struct untracked_cache_dir *untracked,2506 struct write_data *wd)2507{2508 struct stat_data stat_data;2509 struct strbuf *out = &wd->out;2510 unsigned char intbuf[16];2511 unsigned int intlen, value;2512 int i = wd->index++;25132514 /*2515 * untracked_nr should be reset whenever valid is clear, but2516 * for safety..2517 */2518 if (!untracked->valid) {2519 untracked->untracked_nr = 0;2520 untracked->check_only = 0;2521 }25222523 if (untracked->check_only)2524 ewah_set(wd->check_only, i);2525 if (untracked->valid) {2526 ewah_set(wd->valid, i);2527 stat_data_to_disk(&stat_data, &untracked->stat_data);2528 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));2529 }2530 if (!is_null_sha1(untracked->exclude_sha1)) {2531 ewah_set(wd->sha1_valid, i);2532 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);2533 }25342535 intlen = encode_varint(untracked->untracked_nr, intbuf);2536 strbuf_add(out, intbuf, intlen);25372538 /* skip non-recurse directories */2539 for (i = 0, value = 0; i < untracked->dirs_nr; i++)2540 if (untracked->dirs[i]->recurse)2541 value++;2542 intlen = encode_varint(value, intbuf);2543 strbuf_add(out, intbuf, intlen);25442545 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);25462547 for (i = 0; i < untracked->untracked_nr; i++)2548 strbuf_add(out, untracked->untracked[i],2549 strlen(untracked->untracked[i]) + 1);25502551 for (i = 0; i < untracked->dirs_nr; i++)2552 if (untracked->dirs[i]->recurse)2553 write_one_dir(untracked->dirs[i], wd);2554}25552556void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)2557{2558 struct ondisk_untracked_cache *ouc;2559 struct write_data wd;2560 unsigned char varbuf[16];2561 int varint_len;2562 size_t len = strlen(untracked->exclude_per_dir);25632564 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);2565 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);2566 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);2567 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);2568 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);2569 ouc->dir_flags = htonl(untracked->dir_flags);25702571 varint_len = encode_varint(untracked->ident.len, varbuf);2572 strbuf_add(out, varbuf, varint_len);2573 strbuf_add(out, untracked->ident.buf, untracked->ident.len);25742575 strbuf_add(out, ouc, ouc_size(len));2576 free(ouc);2577 ouc = NULL;25782579 if (!untracked->root) {2580 varint_len = encode_varint(0, varbuf);2581 strbuf_add(out, varbuf, varint_len);2582 return;2583 }25842585 wd.index = 0;2586 wd.check_only = ewah_new();2587 wd.valid = ewah_new();2588 wd.sha1_valid = ewah_new();2589 strbuf_init(&wd.out, 1024);2590 strbuf_init(&wd.sb_stat, 1024);2591 strbuf_init(&wd.sb_sha1, 1024);2592 write_one_dir(untracked->root, &wd);25932594 varint_len = encode_varint(wd.index, varbuf);2595 strbuf_add(out, varbuf, varint_len);2596 strbuf_addbuf(out, &wd.out);2597 ewah_serialize_strbuf(wd.valid, out);2598 ewah_serialize_strbuf(wd.check_only, out);2599 ewah_serialize_strbuf(wd.sha1_valid, out);2600 strbuf_addbuf(out, &wd.sb_stat);2601 strbuf_addbuf(out, &wd.sb_sha1);2602 strbuf_addch(out, '\0'); /* safe guard for string lists */26032604 ewah_free(wd.valid);2605 ewah_free(wd.check_only);2606 ewah_free(wd.sha1_valid);2607 strbuf_release(&wd.out);2608 strbuf_release(&wd.sb_stat);2609 strbuf_release(&wd.sb_sha1);2610}26112612static void free_untracked(struct untracked_cache_dir *ucd)2613{2614 int i;2615 if (!ucd)2616 return;2617 for (i = 0; i < ucd->dirs_nr; i++)2618 free_untracked(ucd->dirs[i]);2619 for (i = 0; i < ucd->untracked_nr; i++)2620 free(ucd->untracked[i]);2621 free(ucd->untracked);2622 free(ucd->dirs);2623 free(ucd);2624}26252626void free_untracked_cache(struct untracked_cache *uc)2627{2628 if (uc)2629 free_untracked(uc->root);2630 free(uc);2631}26322633struct read_data {2634 int index;2635 struct untracked_cache_dir **ucd;2636 struct ewah_bitmap *check_only;2637 struct ewah_bitmap *valid;2638 struct ewah_bitmap *sha1_valid;2639 const unsigned char *data;2640 const unsigned char *end;2641};26422643static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)2644{2645 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);2646 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);2647 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);2648 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);2649 to->sd_dev = get_be32(&from->sd_dev);2650 to->sd_ino = get_be32(&from->sd_ino);2651 to->sd_uid = get_be32(&from->sd_uid);2652 to->sd_gid = get_be32(&from->sd_gid);2653 to->sd_size = get_be32(&from->sd_size);2654}26552656static int read_one_dir(struct untracked_cache_dir **untracked_,2657 struct read_data *rd)2658{2659 struct untracked_cache_dir ud, *untracked;2660 const unsigned char *next, *data = rd->data, *end = rd->end;2661 unsigned int value;2662 int i, len;26632664 memset(&ud, 0, sizeof(ud));26652666 next = data;2667 value = decode_varint(&next);2668 if (next > end)2669 return -1;2670 ud.recurse = 1;2671 ud.untracked_alloc = value;2672 ud.untracked_nr = value;2673 if (ud.untracked_nr)2674 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);2675 data = next;26762677 next = data;2678 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);2679 if (next > end)2680 return -1;2681 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);2682 data = next;26832684 len = strlen((const char *)data);2685 next = data + len + 1;2686 if (next > rd->end)2687 return -1;2688 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));2689 memcpy(untracked, &ud, sizeof(ud));2690 memcpy(untracked->name, data, len + 1);2691 data = next;26922693 for (i = 0; i < untracked->untracked_nr; i++) {2694 len = strlen((const char *)data);2695 next = data + len + 1;2696 if (next > rd->end)2697 return -1;2698 untracked->untracked[i] = xstrdup((const char*)data);2699 data = next;2700 }27012702 rd->ucd[rd->index++] = untracked;2703 rd->data = data;27042705 for (i = 0; i < untracked->dirs_nr; i++) {2706 len = read_one_dir(untracked->dirs + i, rd);2707 if (len < 0)2708 return -1;2709 }2710 return 0;2711}27122713static void set_check_only(size_t pos, void *cb)2714{2715 struct read_data *rd = cb;2716 struct untracked_cache_dir *ud = rd->ucd[pos];2717 ud->check_only = 1;2718}27192720static void read_stat(size_t pos, void *cb)2721{2722 struct read_data *rd = cb;2723 struct untracked_cache_dir *ud = rd->ucd[pos];2724 if (rd->data + sizeof(struct stat_data) > rd->end) {2725 rd->data = rd->end + 1;2726 return;2727 }2728 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);2729 rd->data += sizeof(struct stat_data);2730 ud->valid = 1;2731}27322733static void read_sha1(size_t pos, void *cb)2734{2735 struct read_data *rd = cb;2736 struct untracked_cache_dir *ud = rd->ucd[pos];2737 if (rd->data + 20 > rd->end) {2738 rd->data = rd->end + 1;2739 return;2740 }2741 hashcpy(ud->exclude_sha1, rd->data);2742 rd->data += 20;2743}27442745static void load_sha1_stat(struct sha1_stat *sha1_stat,2746 const struct stat_data *stat,2747 const unsigned char *sha1)2748{2749 stat_data_from_disk(&sha1_stat->stat, stat);2750 hashcpy(sha1_stat->sha1, sha1);2751 sha1_stat->valid = 1;2752}27532754struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)2755{2756 const struct ondisk_untracked_cache *ouc;2757 struct untracked_cache *uc;2758 struct read_data rd;2759 const unsigned char *next = data, *end = (const unsigned char *)data + sz;2760 const char *ident;2761 int ident_len, len;27622763 if (sz <= 1 || end[-1] != '\0')2764 return NULL;2765 end--;27662767 ident_len = decode_varint(&next);2768 if (next + ident_len > end)2769 return NULL;2770 ident = (const char *)next;2771 next += ident_len;27722773 ouc = (const struct ondisk_untracked_cache *)next;2774 if (next + ouc_size(0) > end)2775 return NULL;27762777 uc = xcalloc(1, sizeof(*uc));2778 strbuf_init(&uc->ident, ident_len);2779 strbuf_add(&uc->ident, ident, ident_len);2780 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,2781 ouc->info_exclude_sha1);2782 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,2783 ouc->excludes_file_sha1);2784 uc->dir_flags = get_be32(&ouc->dir_flags);2785 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);2786 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */2787 next += ouc_size(strlen(ouc->exclude_per_dir));2788 if (next >= end)2789 goto done2;27902791 len = decode_varint(&next);2792 if (next > end || len == 0)2793 goto done2;27942795 rd.valid = ewah_new();2796 rd.check_only = ewah_new();2797 rd.sha1_valid = ewah_new();2798 rd.data = next;2799 rd.end = end;2800 rd.index = 0;2801 ALLOC_ARRAY(rd.ucd, len);28022803 if (read_one_dir(&uc->root, &rd) || rd.index != len)2804 goto done;28052806 next = rd.data;2807 len = ewah_read_mmap(rd.valid, next, end - next);2808 if (len < 0)2809 goto done;28102811 next += len;2812 len = ewah_read_mmap(rd.check_only, next, end - next);2813 if (len < 0)2814 goto done;28152816 next += len;2817 len = ewah_read_mmap(rd.sha1_valid, next, end - next);2818 if (len < 0)2819 goto done;28202821 ewah_each_bit(rd.check_only, set_check_only, &rd);2822 rd.data = next + len;2823 ewah_each_bit(rd.valid, read_stat, &rd);2824 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);2825 next = rd.data;28262827done:2828 free(rd.ucd);2829 ewah_free(rd.valid);2830 ewah_free(rd.check_only);2831 ewah_free(rd.sha1_valid);2832done2:2833 if (next != end) {2834 free_untracked_cache(uc);2835 uc = NULL;2836 }2837 return uc;2838}28392840static void invalidate_one_directory(struct untracked_cache *uc,2841 struct untracked_cache_dir *ucd)2842{2843 uc->dir_invalidated++;2844 ucd->valid = 0;2845 ucd->untracked_nr = 0;2846}28472848/*2849 * Normally when an entry is added or removed from a directory,2850 * invalidating that directory is enough. No need to touch its2851 * ancestors. When a directory is shown as "foo/bar/" in git-status2852 * however, deleting or adding an entry may have cascading effect.2853 *2854 * Say the "foo/bar/file" has become untracked, we need to tell the2855 * untracked_cache_dir of "foo" that "bar/" is not an untracked2856 * directory any more (because "bar" is managed by foo as an untracked2857 * "file").2858 *2859 * Similarly, if "foo/bar/file" moves from untracked to tracked and it2860 * was the last untracked entry in the entire "foo", we should show2861 * "foo/" instead. Which means we have to invalidate past "bar" up to2862 * "foo".2863 *2864 * This function traverses all directories from root to leaf. If there2865 * is a chance of one of the above cases happening, we invalidate back2866 * to root. Otherwise we just invalidate the leaf. There may be a more2867 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to2868 * detect these cases and avoid unnecessary invalidation, for example,2869 * checking for the untracked entry named "bar/" in "foo", but for now2870 * stick to something safe and simple.2871 */2872static int invalidate_one_component(struct untracked_cache *uc,2873 struct untracked_cache_dir *dir,2874 const char *path, int len)2875{2876 const char *rest = strchr(path, '/');28772878 if (rest) {2879 int component_len = rest - path;2880 struct untracked_cache_dir *d =2881 lookup_untracked(uc, dir, path, component_len);2882 int ret =2883 invalidate_one_component(uc, d, rest + 1,2884 len - (component_len + 1));2885 if (ret)2886 invalidate_one_directory(uc, dir);2887 return ret;2888 }28892890 invalidate_one_directory(uc, dir);2891 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;2892}28932894void untracked_cache_invalidate_path(struct index_state *istate,2895 const char *path)2896{2897 if (!istate->untracked || !istate->untracked->root)2898 return;2899 invalidate_one_component(istate->untracked, istate->untracked->root,2900 path, strlen(path));2901}29022903void untracked_cache_remove_from_index(struct index_state *istate,2904 const char *path)2905{2906 untracked_cache_invalidate_path(istate, path);2907}29082909void untracked_cache_add_to_index(struct index_state *istate,2910 const char *path)2911{2912 untracked_cache_invalidate_path(istate, path);2913}