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#define NO_THE_INDEX_COMPATIBILITY_MACROS 11#include "cache.h" 12#include "dir.h" 13#include "attr.h" 14#include "refs.h" 15#include "wildmatch.h" 16#include "pathspec.h" 17#include "utf8.h" 18#include "varint.h" 19#include "ewah/ewok.h" 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 34/* 35 * Support data structure for our opendir/readdir/closedir wrappers 36 */ 37struct cached_dir { 38 DIR *fdir; 39 struct untracked_cache_dir *untracked; 40 int nr_files; 41 int nr_dirs; 42 43 struct dirent *de; 44 const char *file; 45 struct untracked_cache_dir *ucd; 46}; 47 48static enum path_treatment read_directory_recursive(struct dir_struct *dir, 49 struct index_state *istate, const char *path, int len, 50 struct untracked_cache_dir *untracked, 51 int check_only, const struct pathspec *pathspec); 52static int get_dtype(struct dirent *de, struct index_state *istate, 53 const char *path, int len); 54 55int count_slashes(const char *s) 56{ 57 int cnt = 0; 58 while (*s) 59 if (*s++ == '/') 60 cnt++; 61 return cnt; 62} 63 64int fspathcmp(const char *a, const char *b) 65{ 66 return ignore_case ? strcasecmp(a, b) : strcmp(a, b); 67} 68 69int fspathncmp(const char *a, const char *b, size_t count) 70{ 71 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count); 72} 73 74int git_fnmatch(const struct pathspec_item *item, 75 const char *pattern, const char *string, 76 int prefix) 77{ 78 if (prefix > 0) { 79 if (ps_strncmp(item, pattern, string, prefix)) 80 return WM_NOMATCH; 81 pattern += prefix; 82 string += prefix; 83 } 84 if (item->flags & PATHSPEC_ONESTAR) { 85 int pattern_len = strlen(++pattern); 86 int string_len = strlen(string); 87 return string_len < pattern_len || 88 ps_strcmp(item, pattern, 89 string + string_len - pattern_len); 90 } 91 if (item->magic & PATHSPEC_GLOB) 92 return wildmatch(pattern, string, 93 WM_PATHNAME | 94 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0), 95 NULL); 96 else 97 /* wildmatch has not learned no FNM_PATHNAME mode yet */ 98 return wildmatch(pattern, string, 99 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0, 100 NULL); 101} 102 103static int fnmatch_icase_mem(const char *pattern, int patternlen, 104 const char *string, int stringlen, 105 int flags) 106{ 107 int match_status; 108 struct strbuf pat_buf = STRBUF_INIT; 109 struct strbuf str_buf = STRBUF_INIT; 110 const char *use_pat = pattern; 111 const char *use_str = string; 112 113 if (pattern[patternlen]) { 114 strbuf_add(&pat_buf, pattern, patternlen); 115 use_pat = pat_buf.buf; 116 } 117 if (string[stringlen]) { 118 strbuf_add(&str_buf, string, stringlen); 119 use_str = str_buf.buf; 120 } 121 122 if (ignore_case) 123 flags |= WM_CASEFOLD; 124 match_status = wildmatch(use_pat, use_str, flags, NULL); 125 126 strbuf_release(&pat_buf); 127 strbuf_release(&str_buf); 128 129 return match_status; 130} 131 132static size_t common_prefix_len(const struct pathspec *pathspec) 133{ 134 int n; 135 size_t max = 0; 136 137 /* 138 * ":(icase)path" is treated as a pathspec full of 139 * wildcard. In other words, only prefix is considered common 140 * prefix. If the pathspec is abc/foo abc/bar, running in 141 * subdir xyz, the common prefix is still xyz, not xuz/abc as 142 * in non-:(icase). 143 */ 144 GUARD_PATHSPEC(pathspec, 145 PATHSPEC_FROMTOP | 146 PATHSPEC_MAXDEPTH | 147 PATHSPEC_LITERAL | 148 PATHSPEC_GLOB | 149 PATHSPEC_ICASE | 150 PATHSPEC_EXCLUDE | 151 PATHSPEC_ATTR); 152 153 for (n = 0; n < pathspec->nr; n++) { 154 size_t i = 0, len = 0, item_len; 155 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE) 156 continue; 157 if (pathspec->items[n].magic & PATHSPEC_ICASE) 158 item_len = pathspec->items[n].prefix; 159 else 160 item_len = pathspec->items[n].nowildcard_len; 161 while (i < item_len && (n == 0 || i < max)) { 162 char c = pathspec->items[n].match[i]; 163 if (c != pathspec->items[0].match[i]) 164 break; 165 if (c == '/') 166 len = i + 1; 167 i++; 168 } 169 if (n == 0 || len < max) { 170 max = len; 171 if (!max) 172 break; 173 } 174 } 175 return max; 176} 177 178/* 179 * Returns a copy of the longest leading path common among all 180 * pathspecs. 181 */ 182char *common_prefix(const struct pathspec *pathspec) 183{ 184 unsigned long len = common_prefix_len(pathspec); 185 186 return len ? xmemdupz(pathspec->items[0].match, len) : NULL; 187} 188 189int fill_directory(struct dir_struct *dir, 190 struct index_state *istate, 191 const struct pathspec *pathspec) 192{ 193 const char *prefix; 194 size_t prefix_len; 195 196 /* 197 * Calculate common prefix for the pathspec, and 198 * use that to optimize the directory walk 199 */ 200 prefix_len = common_prefix_len(pathspec); 201 prefix = prefix_len ? pathspec->items[0].match : ""; 202 203 /* Read the directory and prune it */ 204 read_directory(dir, istate, prefix, prefix_len, pathspec); 205 206 return prefix_len; 207} 208 209int within_depth(const char *name, int namelen, 210 int depth, int max_depth) 211{ 212 const char *cp = name, *cpe = name + namelen; 213 214 while (cp < cpe) { 215 if (*cp++ != '/') 216 continue; 217 depth++; 218 if (depth > max_depth) 219 return 0; 220 } 221 return 1; 222} 223 224#define DO_MATCH_EXCLUDE (1<<0) 225#define DO_MATCH_DIRECTORY (1<<1) 226#define DO_MATCH_SUBMODULE (1<<2) 227 228static int match_attrs(const char *name, int namelen, 229 const struct pathspec_item *item) 230{ 231 int i; 232 233 git_check_attr(name, item->attr_check); 234 for (i = 0; i < item->attr_match_nr; i++) { 235 const char *value; 236 int matched; 237 enum attr_match_mode match_mode; 238 239 value = item->attr_check->items[i].value; 240 match_mode = item->attr_match[i].match_mode; 241 242 if (ATTR_TRUE(value)) 243 matched = (match_mode == MATCH_SET); 244 else if (ATTR_FALSE(value)) 245 matched = (match_mode == MATCH_UNSET); 246 else if (ATTR_UNSET(value)) 247 matched = (match_mode == MATCH_UNSPECIFIED); 248 else 249 matched = (match_mode == MATCH_VALUE && 250 !strcmp(item->attr_match[i].value, value)); 251 if (!matched) 252 return 0; 253 } 254 255 return 1; 256} 257 258/* 259 * Does 'match' match the given name? 260 * A match is found if 261 * 262 * (1) the 'match' string is leading directory of 'name', or 263 * (2) the 'match' string is a wildcard and matches 'name', or 264 * (3) the 'match' string is exactly the same as 'name'. 265 * 266 * and the return value tells which case it was. 267 * 268 * It returns 0 when there is no match. 269 */ 270static int match_pathspec_item(const struct pathspec_item *item, int prefix, 271 const char *name, int namelen, unsigned flags) 272{ 273 /* name/namelen has prefix cut off by caller */ 274 const char *match = item->match + prefix; 275 int matchlen = item->len - prefix; 276 277 /* 278 * The normal call pattern is: 279 * 1. prefix = common_prefix_len(ps); 280 * 2. prune something, or fill_directory 281 * 3. match_pathspec() 282 * 283 * 'prefix' at #1 may be shorter than the command's prefix and 284 * it's ok for #2 to match extra files. Those extras will be 285 * trimmed at #3. 286 * 287 * Suppose the pathspec is 'foo' and '../bar' running from 288 * subdir 'xyz'. The common prefix at #1 will be empty, thanks 289 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The 290 * user does not want XYZ/foo, only the "foo" part should be 291 * case-insensitive. We need to filter out XYZ/foo here. In 292 * other words, we do not trust the caller on comparing the 293 * prefix part when :(icase) is involved. We do exact 294 * comparison ourselves. 295 * 296 * Normally the caller (common_prefix_len() in fact) does 297 * _exact_ matching on name[-prefix+1..-1] and we do not need 298 * to check that part. Be defensive and check it anyway, in 299 * case common_prefix_len is changed, or a new caller is 300 * introduced that does not use common_prefix_len. 301 * 302 * If the penalty turns out too high when prefix is really 303 * long, maybe change it to 304 * strncmp(match, name, item->prefix - prefix) 305 */ 306 if (item->prefix && (item->magic & PATHSPEC_ICASE) && 307 strncmp(item->match, name - prefix, item->prefix)) 308 return 0; 309 310 if (item->attr_match_nr && !match_attrs(name, namelen, item)) 311 return 0; 312 313 /* If the match was just the prefix, we matched */ 314 if (!*match) 315 return MATCHED_RECURSIVELY; 316 317 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) { 318 if (matchlen == namelen) 319 return MATCHED_EXACTLY; 320 321 if (match[matchlen-1] == '/' || name[matchlen] == '/') 322 return MATCHED_RECURSIVELY; 323 } else if ((flags & DO_MATCH_DIRECTORY) && 324 match[matchlen - 1] == '/' && 325 namelen == matchlen - 1 && 326 !ps_strncmp(item, match, name, namelen)) 327 return MATCHED_EXACTLY; 328 329 if (item->nowildcard_len < item->len && 330 !git_fnmatch(item, match, name, 331 item->nowildcard_len - prefix)) 332 return MATCHED_FNMATCH; 333 334 /* Perform checks to see if "name" is a super set of the pathspec */ 335 if (flags & DO_MATCH_SUBMODULE) { 336 /* name is a literal prefix of the pathspec */ 337 if ((namelen < matchlen) && 338 (match[namelen] == '/') && 339 !ps_strncmp(item, match, name, namelen)) 340 return MATCHED_RECURSIVELY; 341 342 /* name" doesn't match up to the first wild character */ 343 if (item->nowildcard_len < item->len && 344 ps_strncmp(item, match, name, 345 item->nowildcard_len - prefix)) 346 return 0; 347 348 /* 349 * Here is where we would perform a wildmatch to check if 350 * "name" can be matched as a directory (or a prefix) against 351 * the pathspec. Since wildmatch doesn't have this capability 352 * at the present we have to punt and say that it is a match, 353 * potentially returning a false positive 354 * The submodules themselves will be able to perform more 355 * accurate matching to determine if the pathspec matches. 356 */ 357 return MATCHED_RECURSIVELY; 358 } 359 360 return 0; 361} 362 363/* 364 * Given a name and a list of pathspecs, returns the nature of the 365 * closest (i.e. most specific) match of the name to any of the 366 * pathspecs. 367 * 368 * The caller typically calls this multiple times with the same 369 * pathspec and seen[] array but with different name/namelen 370 * (e.g. entries from the index) and is interested in seeing if and 371 * how each pathspec matches all the names it calls this function 372 * with. A mark is left in the seen[] array for each pathspec element 373 * indicating the closest type of match that element achieved, so if 374 * seen[n] remains zero after multiple invocations, that means the nth 375 * pathspec did not match any names, which could indicate that the 376 * user mistyped the nth pathspec. 377 */ 378static int do_match_pathspec(const struct pathspec *ps, 379 const char *name, int namelen, 380 int prefix, char *seen, 381 unsigned flags) 382{ 383 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE; 384 385 GUARD_PATHSPEC(ps, 386 PATHSPEC_FROMTOP | 387 PATHSPEC_MAXDEPTH | 388 PATHSPEC_LITERAL | 389 PATHSPEC_GLOB | 390 PATHSPEC_ICASE | 391 PATHSPEC_EXCLUDE | 392 PATHSPEC_ATTR); 393 394 if (!ps->nr) { 395 if (!ps->recursive || 396 !(ps->magic & PATHSPEC_MAXDEPTH) || 397 ps->max_depth == -1) 398 return MATCHED_RECURSIVELY; 399 400 if (within_depth(name, namelen, 0, ps->max_depth)) 401 return MATCHED_EXACTLY; 402 else 403 return 0; 404 } 405 406 name += prefix; 407 namelen -= prefix; 408 409 for (i = ps->nr - 1; i >= 0; i--) { 410 int how; 411 412 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) || 413 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE))) 414 continue; 415 416 if (seen && seen[i] == MATCHED_EXACTLY) 417 continue; 418 /* 419 * Make exclude patterns optional and never report 420 * "pathspec ':(exclude)foo' matches no files" 421 */ 422 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE) 423 seen[i] = MATCHED_FNMATCH; 424 how = match_pathspec_item(ps->items+i, prefix, name, 425 namelen, flags); 426 if (ps->recursive && 427 (ps->magic & PATHSPEC_MAXDEPTH) && 428 ps->max_depth != -1 && 429 how && how != MATCHED_FNMATCH) { 430 int len = ps->items[i].len; 431 if (name[len] == '/') 432 len++; 433 if (within_depth(name+len, namelen-len, 0, ps->max_depth)) 434 how = MATCHED_EXACTLY; 435 else 436 how = 0; 437 } 438 if (how) { 439 if (retval < how) 440 retval = how; 441 if (seen && seen[i] < how) 442 seen[i] = how; 443 } 444 } 445 return retval; 446} 447 448int match_pathspec(const struct pathspec *ps, 449 const char *name, int namelen, 450 int prefix, char *seen, int is_dir) 451{ 452 int positive, negative; 453 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0; 454 positive = do_match_pathspec(ps, name, namelen, 455 prefix, seen, flags); 456 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive) 457 return positive; 458 negative = do_match_pathspec(ps, name, namelen, 459 prefix, seen, 460 flags | DO_MATCH_EXCLUDE); 461 return negative ? 0 : positive; 462} 463 464/** 465 * Check if a submodule is a superset of the pathspec 466 */ 467int submodule_path_match(const struct pathspec *ps, 468 const char *submodule_name, 469 char *seen) 470{ 471 int matched = do_match_pathspec(ps, submodule_name, 472 strlen(submodule_name), 473 0, seen, 474 DO_MATCH_DIRECTORY | 475 DO_MATCH_SUBMODULE); 476 return matched; 477} 478 479int report_path_error(const char *ps_matched, 480 const struct pathspec *pathspec, 481 const char *prefix) 482{ 483 /* 484 * Make sure all pathspec matched; otherwise it is an error. 485 */ 486 int num, errors = 0; 487 for (num = 0; num < pathspec->nr; num++) { 488 int other, found_dup; 489 490 if (ps_matched[num]) 491 continue; 492 /* 493 * The caller might have fed identical pathspec 494 * twice. Do not barf on such a mistake. 495 * FIXME: parse_pathspec should have eliminated 496 * duplicate pathspec. 497 */ 498 for (found_dup = other = 0; 499 !found_dup && other < pathspec->nr; 500 other++) { 501 if (other == num || !ps_matched[other]) 502 continue; 503 if (!strcmp(pathspec->items[other].original, 504 pathspec->items[num].original)) 505 /* 506 * Ok, we have a match already. 507 */ 508 found_dup = 1; 509 } 510 if (found_dup) 511 continue; 512 513 error("pathspec '%s' did not match any file(s) known to git.", 514 pathspec->items[num].original); 515 errors++; 516 } 517 return errors; 518} 519 520/* 521 * Return the length of the "simple" part of a path match limiter. 522 */ 523int simple_length(const char *match) 524{ 525 int len = -1; 526 527 for (;;) { 528 unsigned char c = *match++; 529 len++; 530 if (c == '\0' || is_glob_special(c)) 531 return len; 532 } 533} 534 535int no_wildcard(const char *string) 536{ 537 return string[simple_length(string)] == '\0'; 538} 539 540void parse_exclude_pattern(const char **pattern, 541 int *patternlen, 542 unsigned *flags, 543 int *nowildcardlen) 544{ 545 const char *p = *pattern; 546 size_t i, len; 547 548 *flags = 0; 549 if (*p == '!') { 550 *flags |= EXC_FLAG_NEGATIVE; 551 p++; 552 } 553 len = strlen(p); 554 if (len && p[len - 1] == '/') { 555 len--; 556 *flags |= EXC_FLAG_MUSTBEDIR; 557 } 558 for (i = 0; i < len; i++) { 559 if (p[i] == '/') 560 break; 561 } 562 if (i == len) 563 *flags |= EXC_FLAG_NODIR; 564 *nowildcardlen = simple_length(p); 565 /* 566 * we should have excluded the trailing slash from 'p' too, 567 * but that's one more allocation. Instead just make sure 568 * nowildcardlen does not exceed real patternlen 569 */ 570 if (*nowildcardlen > len) 571 *nowildcardlen = len; 572 if (*p == '*' && no_wildcard(p + 1)) 573 *flags |= EXC_FLAG_ENDSWITH; 574 *pattern = p; 575 *patternlen = len; 576} 577 578void add_exclude(const char *string, const char *base, 579 int baselen, struct exclude_list *el, int srcpos) 580{ 581 struct exclude *x; 582 int patternlen; 583 unsigned flags; 584 int nowildcardlen; 585 586 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen); 587 if (flags & EXC_FLAG_MUSTBEDIR) { 588 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen); 589 } else { 590 x = xmalloc(sizeof(*x)); 591 x->pattern = string; 592 } 593 x->patternlen = patternlen; 594 x->nowildcardlen = nowildcardlen; 595 x->base = base; 596 x->baselen = baselen; 597 x->flags = flags; 598 x->srcpos = srcpos; 599 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc); 600 el->excludes[el->nr++] = x; 601 x->el = el; 602} 603 604static void *read_skip_worktree_file_from_index(const struct index_state *istate, 605 const char *path, size_t *size, 606 struct sha1_stat *sha1_stat) 607{ 608 int pos, len; 609 unsigned long sz; 610 enum object_type type; 611 void *data; 612 613 len = strlen(path); 614 pos = index_name_pos(istate, path, len); 615 if (pos < 0) 616 return NULL; 617 if (!ce_skip_worktree(istate->cache[pos])) 618 return NULL; 619 data = read_sha1_file(istate->cache[pos]->oid.hash, &type, &sz); 620 if (!data || type != OBJ_BLOB) { 621 free(data); 622 return NULL; 623 } 624 *size = xsize_t(sz); 625 if (sha1_stat) { 626 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat)); 627 hashcpy(sha1_stat->sha1, istate->cache[pos]->oid.hash); 628 } 629 return data; 630} 631 632/* 633 * Frees memory within el which was allocated for exclude patterns and 634 * the file buffer. Does not free el itself. 635 */ 636void clear_exclude_list(struct exclude_list *el) 637{ 638 int i; 639 640 for (i = 0; i < el->nr; i++) 641 free(el->excludes[i]); 642 free(el->excludes); 643 free(el->filebuf); 644 645 memset(el, 0, sizeof(*el)); 646} 647 648static void trim_trailing_spaces(char *buf) 649{ 650 char *p, *last_space = NULL; 651 652 for (p = buf; *p; p++) 653 switch (*p) { 654 case ' ': 655 if (!last_space) 656 last_space = p; 657 break; 658 case '\\': 659 p++; 660 if (!*p) 661 return; 662 /* fallthrough */ 663 default: 664 last_space = NULL; 665 } 666 667 if (last_space) 668 *last_space = '\0'; 669} 670 671/* 672 * Given a subdirectory name and "dir" of the current directory, 673 * search the subdir in "dir" and return it, or create a new one if it 674 * does not exist in "dir". 675 * 676 * If "name" has the trailing slash, it'll be excluded in the search. 677 */ 678static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc, 679 struct untracked_cache_dir *dir, 680 const char *name, int len) 681{ 682 int first, last; 683 struct untracked_cache_dir *d; 684 if (!dir) 685 return NULL; 686 if (len && name[len - 1] == '/') 687 len--; 688 first = 0; 689 last = dir->dirs_nr; 690 while (last > first) { 691 int cmp, next = (last + first) >> 1; 692 d = dir->dirs[next]; 693 cmp = strncmp(name, d->name, len); 694 if (!cmp && strlen(d->name) > len) 695 cmp = -1; 696 if (!cmp) 697 return d; 698 if (cmp < 0) { 699 last = next; 700 continue; 701 } 702 first = next+1; 703 } 704 705 uc->dir_created++; 706 FLEX_ALLOC_MEM(d, name, name, len); 707 708 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc); 709 memmove(dir->dirs + first + 1, dir->dirs + first, 710 (dir->dirs_nr - first) * sizeof(*dir->dirs)); 711 dir->dirs_nr++; 712 dir->dirs[first] = d; 713 return d; 714} 715 716static void do_invalidate_gitignore(struct untracked_cache_dir *dir) 717{ 718 int i; 719 dir->valid = 0; 720 dir->untracked_nr = 0; 721 for (i = 0; i < dir->dirs_nr; i++) 722 do_invalidate_gitignore(dir->dirs[i]); 723} 724 725static void invalidate_gitignore(struct untracked_cache *uc, 726 struct untracked_cache_dir *dir) 727{ 728 uc->gitignore_invalidated++; 729 do_invalidate_gitignore(dir); 730} 731 732static void invalidate_directory(struct untracked_cache *uc, 733 struct untracked_cache_dir *dir) 734{ 735 int i; 736 uc->dir_invalidated++; 737 dir->valid = 0; 738 dir->untracked_nr = 0; 739 for (i = 0; i < dir->dirs_nr; i++) 740 dir->dirs[i]->recurse = 0; 741} 742 743/* 744 * Given a file with name "fname", read it (either from disk, or from 745 * an index if 'istate' is non-null), parse it and store the 746 * exclude rules in "el". 747 * 748 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill 749 * stat data from disk (only valid if add_excludes returns zero). If 750 * ss_valid is non-zero, "ss" must contain good value as input. 751 */ 752static int add_excludes(const char *fname, const char *base, int baselen, 753 struct exclude_list *el, 754 struct index_state *istate, 755 struct sha1_stat *sha1_stat) 756{ 757 struct stat st; 758 int fd, i, lineno = 1; 759 size_t size = 0; 760 char *buf, *entry; 761 762 fd = open(fname, O_RDONLY); 763 if (fd < 0 || fstat(fd, &st) < 0) { 764 if (fd < 0) 765 warn_on_fopen_errors(fname); 766 else 767 close(fd); 768 if (!istate || 769 (buf = read_skip_worktree_file_from_index(istate, fname, &size, sha1_stat)) == NULL) 770 return -1; 771 if (size == 0) { 772 free(buf); 773 return 0; 774 } 775 if (buf[size-1] != '\n') { 776 buf = xrealloc(buf, st_add(size, 1)); 777 buf[size++] = '\n'; 778 } 779 } else { 780 size = xsize_t(st.st_size); 781 if (size == 0) { 782 if (sha1_stat) { 783 fill_stat_data(&sha1_stat->stat, &st); 784 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN); 785 sha1_stat->valid = 1; 786 } 787 close(fd); 788 return 0; 789 } 790 buf = xmallocz(size); 791 if (read_in_full(fd, buf, size) != size) { 792 free(buf); 793 close(fd); 794 return -1; 795 } 796 buf[size++] = '\n'; 797 close(fd); 798 if (sha1_stat) { 799 int pos; 800 if (sha1_stat->valid && 801 !match_stat_data_racy(istate, &sha1_stat->stat, &st)) 802 ; /* no content change, ss->sha1 still good */ 803 else if (istate && 804 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 && 805 !ce_stage(istate->cache[pos]) && 806 ce_uptodate(istate->cache[pos]) && 807 !would_convert_to_git(fname)) 808 hashcpy(sha1_stat->sha1, 809 istate->cache[pos]->oid.hash); 810 else 811 hash_sha1_file(buf, size, "blob", sha1_stat->sha1); 812 fill_stat_data(&sha1_stat->stat, &st); 813 sha1_stat->valid = 1; 814 } 815 } 816 817 el->filebuf = buf; 818 819 if (skip_utf8_bom(&buf, size)) 820 size -= buf - el->filebuf; 821 822 entry = buf; 823 824 for (i = 0; i < size; i++) { 825 if (buf[i] == '\n') { 826 if (entry != buf + i && entry[0] != '#') { 827 buf[i - (i && buf[i-1] == '\r')] = 0; 828 trim_trailing_spaces(entry); 829 add_exclude(entry, base, baselen, el, lineno); 830 } 831 lineno++; 832 entry = buf + i + 1; 833 } 834 } 835 return 0; 836} 837 838int add_excludes_from_file_to_list(const char *fname, const char *base, 839 int baselen, struct exclude_list *el, 840 struct index_state *istate) 841{ 842 return add_excludes(fname, base, baselen, el, istate, NULL); 843} 844 845struct exclude_list *add_exclude_list(struct dir_struct *dir, 846 int group_type, const char *src) 847{ 848 struct exclude_list *el; 849 struct exclude_list_group *group; 850 851 group = &dir->exclude_list_group[group_type]; 852 ALLOC_GROW(group->el, group->nr + 1, group->alloc); 853 el = &group->el[group->nr++]; 854 memset(el, 0, sizeof(*el)); 855 el->src = src; 856 return el; 857} 858 859/* 860 * Used to set up core.excludesfile and .git/info/exclude lists. 861 */ 862static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname, 863 struct sha1_stat *sha1_stat) 864{ 865 struct exclude_list *el; 866 /* 867 * catch setup_standard_excludes() that's called before 868 * dir->untracked is assigned. That function behaves 869 * differently when dir->untracked is non-NULL. 870 */ 871 if (!dir->untracked) 872 dir->unmanaged_exclude_files++; 873 el = add_exclude_list(dir, EXC_FILE, fname); 874 if (add_excludes(fname, "", 0, el, NULL, sha1_stat) < 0) 875 die("cannot use %s as an exclude file", fname); 876} 877 878void add_excludes_from_file(struct dir_struct *dir, const char *fname) 879{ 880 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */ 881 add_excludes_from_file_1(dir, fname, NULL); 882} 883 884int match_basename(const char *basename, int basenamelen, 885 const char *pattern, int prefix, int patternlen, 886 unsigned flags) 887{ 888 if (prefix == patternlen) { 889 if (patternlen == basenamelen && 890 !fspathncmp(pattern, basename, basenamelen)) 891 return 1; 892 } else if (flags & EXC_FLAG_ENDSWITH) { 893 /* "*literal" matching against "fooliteral" */ 894 if (patternlen - 1 <= basenamelen && 895 !fspathncmp(pattern + 1, 896 basename + basenamelen - (patternlen - 1), 897 patternlen - 1)) 898 return 1; 899 } else { 900 if (fnmatch_icase_mem(pattern, patternlen, 901 basename, basenamelen, 902 0) == 0) 903 return 1; 904 } 905 return 0; 906} 907 908int match_pathname(const char *pathname, int pathlen, 909 const char *base, int baselen, 910 const char *pattern, int prefix, int patternlen, 911 unsigned flags) 912{ 913 const char *name; 914 int namelen; 915 916 /* 917 * match with FNM_PATHNAME; the pattern has base implicitly 918 * in front of it. 919 */ 920 if (*pattern == '/') { 921 pattern++; 922 patternlen--; 923 prefix--; 924 } 925 926 /* 927 * baselen does not count the trailing slash. base[] may or 928 * may not end with a trailing slash though. 929 */ 930 if (pathlen < baselen + 1 || 931 (baselen && pathname[baselen] != '/') || 932 fspathncmp(pathname, base, baselen)) 933 return 0; 934 935 namelen = baselen ? pathlen - baselen - 1 : pathlen; 936 name = pathname + pathlen - namelen; 937 938 if (prefix) { 939 /* 940 * if the non-wildcard part is longer than the 941 * remaining pathname, surely it cannot match. 942 */ 943 if (prefix > namelen) 944 return 0; 945 946 if (fspathncmp(pattern, name, prefix)) 947 return 0; 948 pattern += prefix; 949 patternlen -= prefix; 950 name += prefix; 951 namelen -= prefix; 952 953 /* 954 * If the whole pattern did not have a wildcard, 955 * then our prefix match is all we need; we 956 * do not need to call fnmatch at all. 957 */ 958 if (!patternlen && !namelen) 959 return 1; 960 } 961 962 return fnmatch_icase_mem(pattern, patternlen, 963 name, namelen, 964 WM_PATHNAME) == 0; 965} 966 967/* 968 * Scan the given exclude list in reverse to see whether pathname 969 * should be ignored. The first match (i.e. the last on the list), if 970 * any, determines the fate. Returns the exclude_list element which 971 * matched, or NULL for undecided. 972 */ 973static struct exclude *last_exclude_matching_from_list(const char *pathname, 974 int pathlen, 975 const char *basename, 976 int *dtype, 977 struct exclude_list *el, 978 struct index_state *istate) 979{ 980 struct exclude *exc = NULL; /* undecided */ 981 int i; 982 983 if (!el->nr) 984 return NULL; /* undefined */ 985 986 for (i = el->nr - 1; 0 <= i; i--) { 987 struct exclude *x = el->excludes[i]; 988 const char *exclude = x->pattern; 989 int prefix = x->nowildcardlen; 990 991 if (x->flags & EXC_FLAG_MUSTBEDIR) { 992 if (*dtype == DT_UNKNOWN) 993 *dtype = get_dtype(NULL, istate, pathname, pathlen); 994 if (*dtype != DT_DIR) 995 continue; 996 } 997 998 if (x->flags & EXC_FLAG_NODIR) { 999 if (match_basename(basename,1000 pathlen - (basename - pathname),1001 exclude, prefix, x->patternlen,1002 x->flags)) {1003 exc = x;1004 break;1005 }1006 continue;1007 }10081009 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');1010 if (match_pathname(pathname, pathlen,1011 x->base, x->baselen ? x->baselen - 1 : 0,1012 exclude, prefix, x->patternlen, x->flags)) {1013 exc = x;1014 break;1015 }1016 }1017 return exc;1018}10191020/*1021 * Scan the list and let the last match determine the fate.1022 * Return 1 for exclude, 0 for include and -1 for undecided.1023 */1024int is_excluded_from_list(const char *pathname,1025 int pathlen, const char *basename, int *dtype,1026 struct exclude_list *el, struct index_state *istate)1027{1028 struct exclude *exclude;1029 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,1030 dtype, el, istate);1031 if (exclude)1032 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1033 return -1; /* undecided */1034}10351036static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,1037 struct index_state *istate,1038 const char *pathname, int pathlen, const char *basename,1039 int *dtype_p)1040{1041 int i, j;1042 struct exclude_list_group *group;1043 struct exclude *exclude;1044 for (i = EXC_CMDL; i <= EXC_FILE; i++) {1045 group = &dir->exclude_list_group[i];1046 for (j = group->nr - 1; j >= 0; j--) {1047 exclude = last_exclude_matching_from_list(1048 pathname, pathlen, basename, dtype_p,1049 &group->el[j], istate);1050 if (exclude)1051 return exclude;1052 }1053 }1054 return NULL;1055}10561057/*1058 * Loads the per-directory exclude list for the substring of base1059 * which has a char length of baselen.1060 */1061static void prep_exclude(struct dir_struct *dir,1062 struct index_state *istate,1063 const char *base, int baselen)1064{1065 struct exclude_list_group *group;1066 struct exclude_list *el;1067 struct exclude_stack *stk = NULL;1068 struct untracked_cache_dir *untracked;1069 int current;10701071 group = &dir->exclude_list_group[EXC_DIRS];10721073 /*1074 * Pop the exclude lists from the EXCL_DIRS exclude_list_group1075 * which originate from directories not in the prefix of the1076 * path being checked.1077 */1078 while ((stk = dir->exclude_stack) != NULL) {1079 if (stk->baselen <= baselen &&1080 !strncmp(dir->basebuf.buf, base, stk->baselen))1081 break;1082 el = &group->el[dir->exclude_stack->exclude_ix];1083 dir->exclude_stack = stk->prev;1084 dir->exclude = NULL;1085 free((char *)el->src); /* see strbuf_detach() below */1086 clear_exclude_list(el);1087 free(stk);1088 group->nr--;1089 }10901091 /* Skip traversing into sub directories if the parent is excluded */1092 if (dir->exclude)1093 return;10941095 /*1096 * Lazy initialization. All call sites currently just1097 * memset(dir, 0, sizeof(*dir)) before use. Changing all of1098 * them seems lots of work for little benefit.1099 */1100 if (!dir->basebuf.buf)1101 strbuf_init(&dir->basebuf, PATH_MAX);11021103 /* Read from the parent directories and push them down. */1104 current = stk ? stk->baselen : -1;1105 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);1106 if (dir->untracked)1107 untracked = stk ? stk->ucd : dir->untracked->root;1108 else1109 untracked = NULL;11101111 while (current < baselen) {1112 const char *cp;1113 struct sha1_stat sha1_stat;11141115 stk = xcalloc(1, sizeof(*stk));1116 if (current < 0) {1117 cp = base;1118 current = 0;1119 } else {1120 cp = strchr(base + current + 1, '/');1121 if (!cp)1122 die("oops in prep_exclude");1123 cp++;1124 untracked =1125 lookup_untracked(dir->untracked, untracked,1126 base + current,1127 cp - base - current);1128 }1129 stk->prev = dir->exclude_stack;1130 stk->baselen = cp - base;1131 stk->exclude_ix = group->nr;1132 stk->ucd = untracked;1133 el = add_exclude_list(dir, EXC_DIRS, NULL);1134 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);1135 assert(stk->baselen == dir->basebuf.len);11361137 /* Abort if the directory is excluded */1138 if (stk->baselen) {1139 int dt = DT_DIR;1140 dir->basebuf.buf[stk->baselen - 1] = 0;1141 dir->exclude = last_exclude_matching_from_lists(dir,1142 istate,1143 dir->basebuf.buf, stk->baselen - 1,1144 dir->basebuf.buf + current, &dt);1145 dir->basebuf.buf[stk->baselen - 1] = '/';1146 if (dir->exclude &&1147 dir->exclude->flags & EXC_FLAG_NEGATIVE)1148 dir->exclude = NULL;1149 if (dir->exclude) {1150 dir->exclude_stack = stk;1151 return;1152 }1153 }11541155 /* Try to read per-directory file */1156 hashclr(sha1_stat.sha1);1157 sha1_stat.valid = 0;1158 if (dir->exclude_per_dir &&1159 /*1160 * If we know that no files have been added in1161 * this directory (i.e. valid_cached_dir() has1162 * been executed and set untracked->valid) ..1163 */1164 (!untracked || !untracked->valid ||1165 /*1166 * .. and .gitignore does not exist before1167 * (i.e. null exclude_sha1). Then we can skip1168 * loading .gitignore, which would result in1169 * ENOENT anyway.1170 */1171 !is_null_sha1(untracked->exclude_sha1))) {1172 /*1173 * dir->basebuf gets reused by the traversal, but we1174 * need fname to remain unchanged to ensure the src1175 * member of each struct exclude correctly1176 * back-references its source file. Other invocations1177 * of add_exclude_list provide stable strings, so we1178 * strbuf_detach() and free() here in the caller.1179 */1180 struct strbuf sb = STRBUF_INIT;1181 strbuf_addbuf(&sb, &dir->basebuf);1182 strbuf_addstr(&sb, dir->exclude_per_dir);1183 el->src = strbuf_detach(&sb, NULL);1184 add_excludes(el->src, el->src, stk->baselen, el, istate,1185 untracked ? &sha1_stat : NULL);1186 }1187 /*1188 * NEEDSWORK: when untracked cache is enabled, prep_exclude()1189 * will first be called in valid_cached_dir() then maybe many1190 * times more in last_exclude_matching(). When the cache is1191 * used, last_exclude_matching() will not be called and1192 * reading .gitignore content will be a waste.1193 *1194 * So when it's called by valid_cached_dir() and we can get1195 * .gitignore SHA-1 from the index (i.e. .gitignore is not1196 * modified on work tree), we could delay reading the1197 * .gitignore content until we absolutely need it in1198 * last_exclude_matching(). Be careful about ignore rule1199 * order, though, if you do that.1200 */1201 if (untracked &&1202 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {1203 invalidate_gitignore(dir->untracked, untracked);1204 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);1205 }1206 dir->exclude_stack = stk;1207 current = stk->baselen;1208 }1209 strbuf_setlen(&dir->basebuf, baselen);1210}12111212/*1213 * Loads the exclude lists for the directory containing pathname, then1214 * scans all exclude lists to determine whether pathname is excluded.1215 * Returns the exclude_list element which matched, or NULL for1216 * undecided.1217 */1218struct exclude *last_exclude_matching(struct dir_struct *dir,1219 struct index_state *istate,1220 const char *pathname,1221 int *dtype_p)1222{1223 int pathlen = strlen(pathname);1224 const char *basename = strrchr(pathname, '/');1225 basename = (basename) ? basename+1 : pathname;12261227 prep_exclude(dir, istate, pathname, basename-pathname);12281229 if (dir->exclude)1230 return dir->exclude;12311232 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,1233 basename, dtype_p);1234}12351236/*1237 * Loads the exclude lists for the directory containing pathname, then1238 * scans all exclude lists to determine whether pathname is excluded.1239 * Returns 1 if true, otherwise 0.1240 */1241int is_excluded(struct dir_struct *dir, struct index_state *istate,1242 const char *pathname, int *dtype_p)1243{1244 struct exclude *exclude =1245 last_exclude_matching(dir, istate, pathname, dtype_p);1246 if (exclude)1247 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1248 return 0;1249}12501251static struct dir_entry *dir_entry_new(const char *pathname, int len)1252{1253 struct dir_entry *ent;12541255 FLEX_ALLOC_MEM(ent, name, pathname, len);1256 ent->len = len;1257 return ent;1258}12591260static struct dir_entry *dir_add_name(struct dir_struct *dir,1261 struct index_state *istate,1262 const char *pathname, int len)1263{1264 if (index_file_exists(istate, pathname, len, ignore_case))1265 return NULL;12661267 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);1268 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);1269}12701271struct dir_entry *dir_add_ignored(struct dir_struct *dir,1272 struct index_state *istate,1273 const char *pathname, int len)1274{1275 if (!index_name_is_other(istate, pathname, len))1276 return NULL;12771278 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);1279 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);1280}12811282enum exist_status {1283 index_nonexistent = 0,1284 index_directory,1285 index_gitdir1286};12871288/*1289 * Do not use the alphabetically sorted index to look up1290 * the directory name; instead, use the case insensitive1291 * directory hash.1292 */1293static enum exist_status directory_exists_in_index_icase(struct index_state *istate,1294 const char *dirname, int len)1295{1296 struct cache_entry *ce;12971298 if (index_dir_exists(istate, dirname, len))1299 return index_directory;13001301 ce = index_file_exists(istate, dirname, len, ignore_case);1302 if (ce && S_ISGITLINK(ce->ce_mode))1303 return index_gitdir;13041305 return index_nonexistent;1306}13071308/*1309 * The index sorts alphabetically by entry name, which1310 * means that a gitlink sorts as '\0' at the end, while1311 * a directory (which is defined not as an entry, but as1312 * the files it contains) will sort with the '/' at the1313 * end.1314 */1315static enum exist_status directory_exists_in_index(struct index_state *istate,1316 const char *dirname, int len)1317{1318 int pos;13191320 if (ignore_case)1321 return directory_exists_in_index_icase(istate, dirname, len);13221323 pos = index_name_pos(istate, dirname, len);1324 if (pos < 0)1325 pos = -pos-1;1326 while (pos < istate->cache_nr) {1327 const struct cache_entry *ce = istate->cache[pos++];1328 unsigned char endchar;13291330 if (strncmp(ce->name, dirname, len))1331 break;1332 endchar = ce->name[len];1333 if (endchar > '/')1334 break;1335 if (endchar == '/')1336 return index_directory;1337 if (!endchar && S_ISGITLINK(ce->ce_mode))1338 return index_gitdir;1339 }1340 return index_nonexistent;1341}13421343/*1344 * When we find a directory when traversing the filesystem, we1345 * have three distinct cases:1346 *1347 * - ignore it1348 * - see it as a directory1349 * - recurse into it1350 *1351 * and which one we choose depends on a combination of existing1352 * git index contents and the flags passed into the directory1353 * traversal routine.1354 *1355 * Case 1: If we *already* have entries in the index under that1356 * directory name, we always recurse into the directory to see1357 * all the files.1358 *1359 * Case 2: If we *already* have that directory name as a gitlink,1360 * we always continue to see it as a gitlink, regardless of whether1361 * there is an actual git directory there or not (it might not1362 * be checked out as a subproject!)1363 *1364 * Case 3: if we didn't have it in the index previously, we1365 * have a few sub-cases:1366 *1367 * (a) if "show_other_directories" is true, we show it as1368 * just a directory, unless "hide_empty_directories" is1369 * also true, in which case we need to check if it contains any1370 * untracked and / or ignored files.1371 * (b) if it looks like a git directory, and we don't have1372 * 'no_gitlinks' set we treat it as a gitlink, and show it1373 * as a directory.1374 * (c) otherwise, we recurse into it.1375 */1376static enum path_treatment treat_directory(struct dir_struct *dir,1377 struct index_state *istate,1378 struct untracked_cache_dir *untracked,1379 const char *dirname, int len, int baselen, int exclude,1380 const struct pathspec *pathspec)1381{1382 /* The "len-1" is to strip the final '/' */1383 switch (directory_exists_in_index(istate, dirname, len-1)) {1384 case index_directory:1385 return path_recurse;13861387 case index_gitdir:1388 return path_none;13891390 case index_nonexistent:1391 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)1392 break;1393 if (!(dir->flags & DIR_NO_GITLINKS)) {1394 unsigned char sha1[20];1395 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)1396 return path_untracked;1397 }1398 return path_recurse;1399 }14001401 /* This is the "show_other_directories" case */14021403 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))1404 return exclude ? path_excluded : path_untracked;14051406 untracked = lookup_untracked(dir->untracked, untracked,1407 dirname + baselen, len - baselen);1408 return read_directory_recursive(dir, istate, dirname, len,1409 untracked, 1, pathspec);1410}14111412/*1413 * This is an inexact early pruning of any recursive directory1414 * reading - if the path cannot possibly be in the pathspec,1415 * return true, and we'll skip it early.1416 */1417static int simplify_away(const char *path, int pathlen,1418 const struct pathspec *pathspec)1419{1420 int i;14211422 if (!pathspec || !pathspec->nr)1423 return 0;14241425 GUARD_PATHSPEC(pathspec,1426 PATHSPEC_FROMTOP |1427 PATHSPEC_MAXDEPTH |1428 PATHSPEC_LITERAL |1429 PATHSPEC_GLOB |1430 PATHSPEC_ICASE |1431 PATHSPEC_EXCLUDE |1432 PATHSPEC_ATTR);14331434 for (i = 0; i < pathspec->nr; i++) {1435 const struct pathspec_item *item = &pathspec->items[i];1436 int len = item->nowildcard_len;14371438 if (len > pathlen)1439 len = pathlen;1440 if (!ps_strncmp(item, item->match, path, len))1441 return 0;1442 }14431444 return 1;1445}14461447/*1448 * This function tells us whether an excluded path matches a1449 * list of "interesting" pathspecs. That is, whether a path matched1450 * by any of the pathspecs could possibly be ignored by excluding1451 * the specified path. This can happen if:1452 *1453 * 1. the path is mentioned explicitly in the pathspec1454 *1455 * 2. the path is a directory prefix of some element in the1456 * pathspec1457 */1458static int exclude_matches_pathspec(const char *path, int pathlen,1459 const struct pathspec *pathspec)1460{1461 int i;14621463 if (!pathspec || !pathspec->nr)1464 return 0;14651466 GUARD_PATHSPEC(pathspec,1467 PATHSPEC_FROMTOP |1468 PATHSPEC_MAXDEPTH |1469 PATHSPEC_LITERAL |1470 PATHSPEC_GLOB |1471 PATHSPEC_ICASE |1472 PATHSPEC_EXCLUDE);14731474 for (i = 0; i < pathspec->nr; i++) {1475 const struct pathspec_item *item = &pathspec->items[i];1476 int len = item->nowildcard_len;14771478 if (len == pathlen &&1479 !ps_strncmp(item, item->match, path, pathlen))1480 return 1;1481 if (len > pathlen &&1482 item->match[pathlen] == '/' &&1483 !ps_strncmp(item, item->match, path, pathlen))1484 return 1;1485 }1486 return 0;1487}14881489static int get_index_dtype(struct index_state *istate,1490 const char *path, int len)1491{1492 int pos;1493 const struct cache_entry *ce;14941495 ce = index_file_exists(istate, path, len, 0);1496 if (ce) {1497 if (!ce_uptodate(ce))1498 return DT_UNKNOWN;1499 if (S_ISGITLINK(ce->ce_mode))1500 return DT_DIR;1501 /*1502 * Nobody actually cares about the1503 * difference between DT_LNK and DT_REG1504 */1505 return DT_REG;1506 }15071508 /* Try to look it up as a directory */1509 pos = index_name_pos(istate, path, len);1510 if (pos >= 0)1511 return DT_UNKNOWN;1512 pos = -pos-1;1513 while (pos < istate->cache_nr) {1514 ce = istate->cache[pos++];1515 if (strncmp(ce->name, path, len))1516 break;1517 if (ce->name[len] > '/')1518 break;1519 if (ce->name[len] < '/')1520 continue;1521 if (!ce_uptodate(ce))1522 break; /* continue? */1523 return DT_DIR;1524 }1525 return DT_UNKNOWN;1526}15271528static int get_dtype(struct dirent *de, struct index_state *istate,1529 const char *path, int len)1530{1531 int dtype = de ? DTYPE(de) : DT_UNKNOWN;1532 struct stat st;15331534 if (dtype != DT_UNKNOWN)1535 return dtype;1536 dtype = get_index_dtype(istate, path, len);1537 if (dtype != DT_UNKNOWN)1538 return dtype;1539 if (lstat(path, &st))1540 return dtype;1541 if (S_ISREG(st.st_mode))1542 return DT_REG;1543 if (S_ISDIR(st.st_mode))1544 return DT_DIR;1545 if (S_ISLNK(st.st_mode))1546 return DT_LNK;1547 return dtype;1548}15491550static enum path_treatment treat_one_path(struct dir_struct *dir,1551 struct untracked_cache_dir *untracked,1552 struct index_state *istate,1553 struct strbuf *path,1554 int baselen,1555 const struct pathspec *pathspec,1556 int dtype, struct dirent *de)1557{1558 int exclude;1559 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);15601561 if (dtype == DT_UNKNOWN)1562 dtype = get_dtype(de, istate, path->buf, path->len);15631564 /* Always exclude indexed files */1565 if (dtype != DT_DIR && has_path_in_index)1566 return path_none;15671568 /*1569 * When we are looking at a directory P in the working tree,1570 * there are three cases:1571 *1572 * (1) P exists in the index. Everything inside the directory P in1573 * the working tree needs to go when P is checked out from the1574 * index.1575 *1576 * (2) P does not exist in the index, but there is P/Q in the index.1577 * We know P will stay a directory when we check out the contents1578 * of the index, but we do not know yet if there is a directory1579 * P/Q in the working tree to be killed, so we need to recurse.1580 *1581 * (3) P does not exist in the index, and there is no P/Q in the index1582 * to require P to be a directory, either. Only in this case, we1583 * know that everything inside P will not be killed without1584 * recursing.1585 */1586 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&1587 (dtype == DT_DIR) &&1588 !has_path_in_index &&1589 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))1590 return path_none;15911592 exclude = is_excluded(dir, istate, path->buf, &dtype);15931594 /*1595 * Excluded? If we don't explicitly want to show1596 * ignored files, ignore it1597 */1598 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))1599 return path_excluded;16001601 switch (dtype) {1602 default:1603 return path_none;1604 case DT_DIR:1605 strbuf_addch(path, '/');1606 return treat_directory(dir, istate, untracked, path->buf, path->len,1607 baselen, exclude, pathspec);1608 case DT_REG:1609 case DT_LNK:1610 return exclude ? path_excluded : path_untracked;1611 }1612}16131614static enum path_treatment treat_path_fast(struct dir_struct *dir,1615 struct untracked_cache_dir *untracked,1616 struct cached_dir *cdir,1617 struct index_state *istate,1618 struct strbuf *path,1619 int baselen,1620 const struct pathspec *pathspec)1621{1622 strbuf_setlen(path, baselen);1623 if (!cdir->ucd) {1624 strbuf_addstr(path, cdir->file);1625 return path_untracked;1626 }1627 strbuf_addstr(path, cdir->ucd->name);1628 /* treat_one_path() does this before it calls treat_directory() */1629 strbuf_complete(path, '/');1630 if (cdir->ucd->check_only)1631 /*1632 * check_only is set as a result of treat_directory() getting1633 * to its bottom. Verify again the same set of directories1634 * with check_only set.1635 */1636 return read_directory_recursive(dir, istate, path->buf, path->len,1637 cdir->ucd, 1, pathspec);1638 /*1639 * We get path_recurse in the first run when1640 * directory_exists_in_index() returns index_nonexistent. We1641 * are sure that new changes in the index does not impact the1642 * outcome. Return now.1643 */1644 return path_recurse;1645}16461647static enum path_treatment treat_path(struct dir_struct *dir,1648 struct untracked_cache_dir *untracked,1649 struct cached_dir *cdir,1650 struct index_state *istate,1651 struct strbuf *path,1652 int baselen,1653 const struct pathspec *pathspec)1654{1655 int dtype;1656 struct dirent *de = cdir->de;16571658 if (!de)1659 return treat_path_fast(dir, untracked, cdir, istate, path,1660 baselen, pathspec);1661 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))1662 return path_none;1663 strbuf_setlen(path, baselen);1664 strbuf_addstr(path, de->d_name);1665 if (simplify_away(path->buf, path->len, pathspec))1666 return path_none;16671668 dtype = DTYPE(de);1669 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);1670}16711672static void add_untracked(struct untracked_cache_dir *dir, const char *name)1673{1674 if (!dir)1675 return;1676 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,1677 dir->untracked_alloc);1678 dir->untracked[dir->untracked_nr++] = xstrdup(name);1679}16801681static int valid_cached_dir(struct dir_struct *dir,1682 struct untracked_cache_dir *untracked,1683 struct index_state *istate,1684 struct strbuf *path,1685 int check_only)1686{1687 struct stat st;16881689 if (!untracked)1690 return 0;16911692 if (stat(path->len ? path->buf : ".", &st)) {1693 invalidate_directory(dir->untracked, untracked);1694 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));1695 return 0;1696 }1697 if (!untracked->valid ||1698 match_stat_data_racy(istate, &untracked->stat_data, &st)) {1699 if (untracked->valid)1700 invalidate_directory(dir->untracked, untracked);1701 fill_stat_data(&untracked->stat_data, &st);1702 return 0;1703 }17041705 if (untracked->check_only != !!check_only) {1706 invalidate_directory(dir->untracked, untracked);1707 return 0;1708 }17091710 /*1711 * prep_exclude will be called eventually on this directory,1712 * but it's called much later in last_exclude_matching(). We1713 * need it now to determine the validity of the cache for this1714 * path. The next calls will be nearly no-op, the way1715 * prep_exclude() is designed.1716 */1717 if (path->len && path->buf[path->len - 1] != '/') {1718 strbuf_addch(path, '/');1719 prep_exclude(dir, istate, path->buf, path->len);1720 strbuf_setlen(path, path->len - 1);1721 } else1722 prep_exclude(dir, istate, path->buf, path->len);17231724 /* hopefully prep_exclude() haven't invalidated this entry... */1725 return untracked->valid;1726}17271728static int open_cached_dir(struct cached_dir *cdir,1729 struct dir_struct *dir,1730 struct untracked_cache_dir *untracked,1731 struct index_state *istate,1732 struct strbuf *path,1733 int check_only)1734{1735 memset(cdir, 0, sizeof(*cdir));1736 cdir->untracked = untracked;1737 if (valid_cached_dir(dir, untracked, istate, path, check_only))1738 return 0;1739 cdir->fdir = opendir(path->len ? path->buf : ".");1740 if (dir->untracked)1741 dir->untracked->dir_opened++;1742 if (!cdir->fdir)1743 return -1;1744 return 0;1745}17461747static int read_cached_dir(struct cached_dir *cdir)1748{1749 if (cdir->fdir) {1750 cdir->de = readdir(cdir->fdir);1751 if (!cdir->de)1752 return -1;1753 return 0;1754 }1755 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {1756 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];1757 if (!d->recurse) {1758 cdir->nr_dirs++;1759 continue;1760 }1761 cdir->ucd = d;1762 cdir->nr_dirs++;1763 return 0;1764 }1765 cdir->ucd = NULL;1766 if (cdir->nr_files < cdir->untracked->untracked_nr) {1767 struct untracked_cache_dir *d = cdir->untracked;1768 cdir->file = d->untracked[cdir->nr_files++];1769 return 0;1770 }1771 return -1;1772}17731774static void close_cached_dir(struct cached_dir *cdir)1775{1776 if (cdir->fdir)1777 closedir(cdir->fdir);1778 /*1779 * We have gone through this directory and found no untracked1780 * entries. Mark it valid.1781 */1782 if (cdir->untracked) {1783 cdir->untracked->valid = 1;1784 cdir->untracked->recurse = 1;1785 }1786}17871788/*1789 * Read a directory tree. We currently ignore anything but1790 * directories, regular files and symlinks. That's because git1791 * doesn't handle them at all yet. Maybe that will change some1792 * day.1793 *1794 * Also, we ignore the name ".git" (even if it is not a directory).1795 * That likely will not change.1796 *1797 * Returns the most significant path_treatment value encountered in the scan.1798 */1799static enum path_treatment read_directory_recursive(struct dir_struct *dir,1800 struct index_state *istate, const char *base, int baselen,1801 struct untracked_cache_dir *untracked, int check_only,1802 const struct pathspec *pathspec)1803{1804 struct cached_dir cdir;1805 enum path_treatment state, subdir_state, dir_state = path_none;1806 struct strbuf path = STRBUF_INIT;18071808 strbuf_add(&path, base, baselen);18091810 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))1811 goto out;18121813 if (untracked)1814 untracked->check_only = !!check_only;18151816 while (!read_cached_dir(&cdir)) {1817 /* check how the file or directory should be treated */1818 state = treat_path(dir, untracked, &cdir, istate, &path,1819 baselen, pathspec);18201821 if (state > dir_state)1822 dir_state = state;18231824 /* recurse into subdir if instructed by treat_path */1825 if ((state == path_recurse) ||1826 ((state == path_untracked) &&1827 (dir->flags & DIR_SHOW_IGNORED_TOO) &&1828 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {1829 struct untracked_cache_dir *ud;1830 ud = lookup_untracked(dir->untracked, untracked,1831 path.buf + baselen,1832 path.len - baselen);1833 subdir_state =1834 read_directory_recursive(dir, istate, path.buf,1835 path.len, ud,1836 check_only, pathspec);1837 if (subdir_state > dir_state)1838 dir_state = subdir_state;1839 }18401841 if (check_only) {1842 /* abort early if maximum state has been reached */1843 if (dir_state == path_untracked) {1844 if (cdir.fdir)1845 add_untracked(untracked, path.buf + baselen);1846 break;1847 }1848 /* skip the dir_add_* part */1849 continue;1850 }18511852 /* add the path to the appropriate result list */1853 switch (state) {1854 case path_excluded:1855 if (dir->flags & DIR_SHOW_IGNORED)1856 dir_add_name(dir, istate, path.buf, path.len);1857 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||1858 ((dir->flags & DIR_COLLECT_IGNORED) &&1859 exclude_matches_pathspec(path.buf, path.len,1860 pathspec)))1861 dir_add_ignored(dir, istate, path.buf, path.len);1862 break;18631864 case path_untracked:1865 if (dir->flags & DIR_SHOW_IGNORED)1866 break;1867 dir_add_name(dir, istate, path.buf, path.len);1868 if (cdir.fdir)1869 add_untracked(untracked, path.buf + baselen);1870 break;18711872 default:1873 break;1874 }1875 }1876 close_cached_dir(&cdir);1877 out:1878 strbuf_release(&path);18791880 return dir_state;1881}18821883int cmp_dir_entry(const void *p1, const void *p2)1884{1885 const struct dir_entry *e1 = *(const struct dir_entry **)p1;1886 const struct dir_entry *e2 = *(const struct dir_entry **)p2;18871888 return name_compare(e1->name, e1->len, e2->name, e2->len);1889}18901891/* check if *out lexically strictly contains *in */1892int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)1893{1894 return (out->len < in->len) &&1895 (out->name[out->len - 1] == '/') &&1896 !memcmp(out->name, in->name, out->len);1897}18981899static int treat_leading_path(struct dir_struct *dir,1900 struct index_state *istate,1901 const char *path, int len,1902 const struct pathspec *pathspec)1903{1904 struct strbuf sb = STRBUF_INIT;1905 int baselen, rc = 0;1906 const char *cp;1907 int old_flags = dir->flags;19081909 while (len && path[len - 1] == '/')1910 len--;1911 if (!len)1912 return 1;1913 baselen = 0;1914 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;1915 while (1) {1916 cp = path + baselen + !!baselen;1917 cp = memchr(cp, '/', path + len - cp);1918 if (!cp)1919 baselen = len;1920 else1921 baselen = cp - path;1922 strbuf_setlen(&sb, 0);1923 strbuf_add(&sb, path, baselen);1924 if (!is_directory(sb.buf))1925 break;1926 if (simplify_away(sb.buf, sb.len, pathspec))1927 break;1928 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,1929 DT_DIR, NULL) == path_none)1930 break; /* do not recurse into it */1931 if (len <= baselen) {1932 rc = 1;1933 break; /* finished checking */1934 }1935 }1936 strbuf_release(&sb);1937 dir->flags = old_flags;1938 return rc;1939}19401941static const char *get_ident_string(void)1942{1943 static struct strbuf sb = STRBUF_INIT;1944 struct utsname uts;19451946 if (sb.len)1947 return sb.buf;1948 if (uname(&uts) < 0)1949 die_errno(_("failed to get kernel name and information"));1950 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),1951 uts.sysname);1952 return sb.buf;1953}19541955static int ident_in_untracked(const struct untracked_cache *uc)1956{1957 /*1958 * Previous git versions may have saved many NUL separated1959 * strings in the "ident" field, but it is insane to manage1960 * many locations, so just take care of the first one.1961 */19621963 return !strcmp(uc->ident.buf, get_ident_string());1964}19651966static void set_untracked_ident(struct untracked_cache *uc)1967{1968 strbuf_reset(&uc->ident);1969 strbuf_addstr(&uc->ident, get_ident_string());19701971 /*1972 * This strbuf used to contain a list of NUL separated1973 * strings, so save NUL too for backward compatibility.1974 */1975 strbuf_addch(&uc->ident, 0);1976}19771978static void new_untracked_cache(struct index_state *istate)1979{1980 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));1981 strbuf_init(&uc->ident, 100);1982 uc->exclude_per_dir = ".gitignore";1983 /* should be the same flags used by git-status */1984 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;1985 set_untracked_ident(uc);1986 istate->untracked = uc;1987 istate->cache_changed |= UNTRACKED_CHANGED;1988}19891990void add_untracked_cache(struct index_state *istate)1991{1992 if (!istate->untracked) {1993 new_untracked_cache(istate);1994 } else {1995 if (!ident_in_untracked(istate->untracked)) {1996 free_untracked_cache(istate->untracked);1997 new_untracked_cache(istate);1998 }1999 }2000}20012002void remove_untracked_cache(struct index_state *istate)2003{2004 if (istate->untracked) {2005 free_untracked_cache(istate->untracked);2006 istate->untracked = NULL;2007 istate->cache_changed |= UNTRACKED_CHANGED;2008 }2009}20102011static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,2012 int base_len,2013 const struct pathspec *pathspec)2014{2015 struct untracked_cache_dir *root;20162017 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))2018 return NULL;20192020 /*2021 * We only support $GIT_DIR/info/exclude and core.excludesfile2022 * as the global ignore rule files. Any other additions2023 * (e.g. from command line) invalidate the cache. This2024 * condition also catches running setup_standard_excludes()2025 * before setting dir->untracked!2026 */2027 if (dir->unmanaged_exclude_files)2028 return NULL;20292030 /*2031 * Optimize for the main use case only: whole-tree git2032 * status. More work involved in treat_leading_path() if we2033 * use cache on just a subset of the worktree. pathspec2034 * support could make the matter even worse.2035 */2036 if (base_len || (pathspec && pathspec->nr))2037 return NULL;20382039 /* Different set of flags may produce different results */2040 if (dir->flags != dir->untracked->dir_flags ||2041 /*2042 * See treat_directory(), case index_nonexistent. Without2043 * this flag, we may need to also cache .git file content2044 * for the resolve_gitlink_ref() call, which we don't.2045 */2046 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||2047 /* We don't support collecting ignore files */2048 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |2049 DIR_COLLECT_IGNORED)))2050 return NULL;20512052 /*2053 * If we use .gitignore in the cache and now you change it to2054 * .gitexclude, everything will go wrong.2055 */2056 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&2057 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))2058 return NULL;20592060 /*2061 * EXC_CMDL is not considered in the cache. If people set it,2062 * skip the cache.2063 */2064 if (dir->exclude_list_group[EXC_CMDL].nr)2065 return NULL;20662067 if (!ident_in_untracked(dir->untracked)) {2068 warning(_("Untracked cache is disabled on this system or location."));2069 return NULL;2070 }20712072 if (!dir->untracked->root) {2073 const int len = sizeof(*dir->untracked->root);2074 dir->untracked->root = xmalloc(len);2075 memset(dir->untracked->root, 0, len);2076 }20772078 /* Validate $GIT_DIR/info/exclude and core.excludesfile */2079 root = dir->untracked->root;2080 if (hashcmp(dir->ss_info_exclude.sha1,2081 dir->untracked->ss_info_exclude.sha1)) {2082 invalidate_gitignore(dir->untracked, root);2083 dir->untracked->ss_info_exclude = dir->ss_info_exclude;2084 }2085 if (hashcmp(dir->ss_excludes_file.sha1,2086 dir->untracked->ss_excludes_file.sha1)) {2087 invalidate_gitignore(dir->untracked, root);2088 dir->untracked->ss_excludes_file = dir->ss_excludes_file;2089 }20902091 /* Make sure this directory is not dropped out at saving phase */2092 root->recurse = 1;2093 return root;2094}20952096int read_directory(struct dir_struct *dir, struct index_state *istate,2097 const char *path, int len, const struct pathspec *pathspec)2098{2099 struct untracked_cache_dir *untracked;21002101 if (has_symlink_leading_path(path, len))2102 return dir->nr;21032104 untracked = validate_untracked_cache(dir, len, pathspec);2105 if (!untracked)2106 /*2107 * make sure untracked cache code path is disabled,2108 * e.g. prep_exclude()2109 */2110 dir->untracked = NULL;2111 if (!len || treat_leading_path(dir, istate, path, len, pathspec))2112 read_directory_recursive(dir, istate, path, len, untracked, 0, pathspec);2113 QSORT(dir->entries, dir->nr, cmp_dir_entry);2114 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);21152116 /*2117 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will2118 * also pick up untracked contents of untracked dirs; by default2119 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.2120 */2121 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&2122 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {2123 int i, j;21242125 /* remove from dir->entries untracked contents of untracked dirs */2126 for (i = j = 0; j < dir->nr; j++) {2127 if (i &&2128 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {2129 free(dir->entries[j]);2130 dir->entries[j] = NULL;2131 } else {2132 dir->entries[i++] = dir->entries[j];2133 }2134 }21352136 dir->nr = i;2137 }21382139 if (dir->untracked) {2140 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);2141 trace_printf_key(&trace_untracked_stats,2142 "node creation: %u\n"2143 "gitignore invalidation: %u\n"2144 "directory invalidation: %u\n"2145 "opendir: %u\n",2146 dir->untracked->dir_created,2147 dir->untracked->gitignore_invalidated,2148 dir->untracked->dir_invalidated,2149 dir->untracked->dir_opened);2150 if (dir->untracked == istate->untracked &&2151 (dir->untracked->dir_opened ||2152 dir->untracked->gitignore_invalidated ||2153 dir->untracked->dir_invalidated))2154 istate->cache_changed |= UNTRACKED_CHANGED;2155 if (dir->untracked != istate->untracked) {2156 free(dir->untracked);2157 dir->untracked = NULL;2158 }2159 }2160 return dir->nr;2161}21622163int file_exists(const char *f)2164{2165 struct stat sb;2166 return lstat(f, &sb) == 0;2167}21682169static int cmp_icase(char a, char b)2170{2171 if (a == b)2172 return 0;2173 if (ignore_case)2174 return toupper(a) - toupper(b);2175 return a - b;2176}21772178/*2179 * Given two normalized paths (a trailing slash is ok), if subdir is2180 * outside dir, return -1. Otherwise return the offset in subdir that2181 * can be used as relative path to dir.2182 */2183int dir_inside_of(const char *subdir, const char *dir)2184{2185 int offset = 0;21862187 assert(dir && subdir && *dir && *subdir);21882189 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {2190 dir++;2191 subdir++;2192 offset++;2193 }21942195 /* hel[p]/me vs hel[l]/yeah */2196 if (*dir && *subdir)2197 return -1;21982199 if (!*subdir)2200 return !*dir ? offset : -1; /* same dir */22012202 /* foo/[b]ar vs foo/[] */2203 if (is_dir_sep(dir[-1]))2204 return is_dir_sep(subdir[-1]) ? offset : -1;22052206 /* foo[/]bar vs foo[] */2207 return is_dir_sep(*subdir) ? offset + 1 : -1;2208}22092210int is_inside_dir(const char *dir)2211{2212 char *cwd;2213 int rc;22142215 if (!dir)2216 return 0;22172218 cwd = xgetcwd();2219 rc = (dir_inside_of(cwd, dir) >= 0);2220 free(cwd);2221 return rc;2222}22232224int is_empty_dir(const char *path)2225{2226 DIR *dir = opendir(path);2227 struct dirent *e;2228 int ret = 1;22292230 if (!dir)2231 return 0;22322233 while ((e = readdir(dir)) != NULL)2234 if (!is_dot_or_dotdot(e->d_name)) {2235 ret = 0;2236 break;2237 }22382239 closedir(dir);2240 return ret;2241}22422243static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)2244{2245 DIR *dir;2246 struct dirent *e;2247 int ret = 0, original_len = path->len, len, kept_down = 0;2248 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);2249 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);2250 unsigned char submodule_head[20];22512252 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&2253 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {2254 /* Do not descend and nuke a nested git work tree. */2255 if (kept_up)2256 *kept_up = 1;2257 return 0;2258 }22592260 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;2261 dir = opendir(path->buf);2262 if (!dir) {2263 if (errno == ENOENT)2264 return keep_toplevel ? -1 : 0;2265 else if (errno == EACCES && !keep_toplevel)2266 /*2267 * An empty dir could be removable even if it2268 * is unreadable:2269 */2270 return rmdir(path->buf);2271 else2272 return -1;2273 }2274 strbuf_complete(path, '/');22752276 len = path->len;2277 while ((e = readdir(dir)) != NULL) {2278 struct stat st;2279 if (is_dot_or_dotdot(e->d_name))2280 continue;22812282 strbuf_setlen(path, len);2283 strbuf_addstr(path, e->d_name);2284 if (lstat(path->buf, &st)) {2285 if (errno == ENOENT)2286 /*2287 * file disappeared, which is what we2288 * wanted anyway2289 */2290 continue;2291 /* fall thru */2292 } else if (S_ISDIR(st.st_mode)) {2293 if (!remove_dir_recurse(path, flag, &kept_down))2294 continue; /* happy */2295 } else if (!only_empty &&2296 (!unlink(path->buf) || errno == ENOENT)) {2297 continue; /* happy, too */2298 }22992300 /* path too long, stat fails, or non-directory still exists */2301 ret = -1;2302 break;2303 }2304 closedir(dir);23052306 strbuf_setlen(path, original_len);2307 if (!ret && !keep_toplevel && !kept_down)2308 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;2309 else if (kept_up)2310 /*2311 * report the uplevel that it is not an error that we2312 * did not rmdir() our directory.2313 */2314 *kept_up = !ret;2315 return ret;2316}23172318int remove_dir_recursively(struct strbuf *path, int flag)2319{2320 return remove_dir_recurse(path, flag, NULL);2321}23222323static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")23242325void setup_standard_excludes(struct dir_struct *dir)2326{2327 dir->exclude_per_dir = ".gitignore";23282329 /* core.excludefile defaulting to $XDG_HOME/git/ignore */2330 if (!excludes_file)2331 excludes_file = xdg_config_home("ignore");2332 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))2333 add_excludes_from_file_1(dir, excludes_file,2334 dir->untracked ? &dir->ss_excludes_file : NULL);23352336 /* per repository user preference */2337 if (startup_info->have_repository) {2338 const char *path = git_path_info_exclude();2339 if (!access_or_warn(path, R_OK, 0))2340 add_excludes_from_file_1(dir, path,2341 dir->untracked ? &dir->ss_info_exclude : NULL);2342 }2343}23442345int remove_path(const char *name)2346{2347 char *slash;23482349 if (unlink(name) && !is_missing_file_error(errno))2350 return -1;23512352 slash = strrchr(name, '/');2353 if (slash) {2354 char *dirs = xstrdup(name);2355 slash = dirs + (slash - name);2356 do {2357 *slash = '\0';2358 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));2359 free(dirs);2360 }2361 return 0;2362}23632364/*2365 * Frees memory within dir which was allocated for exclude lists and2366 * the exclude_stack. Does not free dir itself.2367 */2368void clear_directory(struct dir_struct *dir)2369{2370 int i, j;2371 struct exclude_list_group *group;2372 struct exclude_list *el;2373 struct exclude_stack *stk;23742375 for (i = EXC_CMDL; i <= EXC_FILE; i++) {2376 group = &dir->exclude_list_group[i];2377 for (j = 0; j < group->nr; j++) {2378 el = &group->el[j];2379 if (i == EXC_DIRS)2380 free((char *)el->src);2381 clear_exclude_list(el);2382 }2383 free(group->el);2384 }23852386 stk = dir->exclude_stack;2387 while (stk) {2388 struct exclude_stack *prev = stk->prev;2389 free(stk);2390 stk = prev;2391 }2392 strbuf_release(&dir->basebuf);2393}23942395struct ondisk_untracked_cache {2396 struct stat_data info_exclude_stat;2397 struct stat_data excludes_file_stat;2398 uint32_t dir_flags;2399 unsigned char info_exclude_sha1[20];2400 unsigned char excludes_file_sha1[20];2401 char exclude_per_dir[FLEX_ARRAY];2402};24032404#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)24052406struct write_data {2407 int index; /* number of written untracked_cache_dir */2408 struct ewah_bitmap *check_only; /* from untracked_cache_dir */2409 struct ewah_bitmap *valid; /* from untracked_cache_dir */2410 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */2411 struct strbuf out;2412 struct strbuf sb_stat;2413 struct strbuf sb_sha1;2414};24152416static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)2417{2418 to->sd_ctime.sec = htonl(from->sd_ctime.sec);2419 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);2420 to->sd_mtime.sec = htonl(from->sd_mtime.sec);2421 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);2422 to->sd_dev = htonl(from->sd_dev);2423 to->sd_ino = htonl(from->sd_ino);2424 to->sd_uid = htonl(from->sd_uid);2425 to->sd_gid = htonl(from->sd_gid);2426 to->sd_size = htonl(from->sd_size);2427}24282429static void write_one_dir(struct untracked_cache_dir *untracked,2430 struct write_data *wd)2431{2432 struct stat_data stat_data;2433 struct strbuf *out = &wd->out;2434 unsigned char intbuf[16];2435 unsigned int intlen, value;2436 int i = wd->index++;24372438 /*2439 * untracked_nr should be reset whenever valid is clear, but2440 * for safety..2441 */2442 if (!untracked->valid) {2443 untracked->untracked_nr = 0;2444 untracked->check_only = 0;2445 }24462447 if (untracked->check_only)2448 ewah_set(wd->check_only, i);2449 if (untracked->valid) {2450 ewah_set(wd->valid, i);2451 stat_data_to_disk(&stat_data, &untracked->stat_data);2452 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));2453 }2454 if (!is_null_sha1(untracked->exclude_sha1)) {2455 ewah_set(wd->sha1_valid, i);2456 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);2457 }24582459 intlen = encode_varint(untracked->untracked_nr, intbuf);2460 strbuf_add(out, intbuf, intlen);24612462 /* skip non-recurse directories */2463 for (i = 0, value = 0; i < untracked->dirs_nr; i++)2464 if (untracked->dirs[i]->recurse)2465 value++;2466 intlen = encode_varint(value, intbuf);2467 strbuf_add(out, intbuf, intlen);24682469 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);24702471 for (i = 0; i < untracked->untracked_nr; i++)2472 strbuf_add(out, untracked->untracked[i],2473 strlen(untracked->untracked[i]) + 1);24742475 for (i = 0; i < untracked->dirs_nr; i++)2476 if (untracked->dirs[i]->recurse)2477 write_one_dir(untracked->dirs[i], wd);2478}24792480void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)2481{2482 struct ondisk_untracked_cache *ouc;2483 struct write_data wd;2484 unsigned char varbuf[16];2485 int varint_len;2486 size_t len = strlen(untracked->exclude_per_dir);24872488 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);2489 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);2490 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);2491 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);2492 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);2493 ouc->dir_flags = htonl(untracked->dir_flags);24942495 varint_len = encode_varint(untracked->ident.len, varbuf);2496 strbuf_add(out, varbuf, varint_len);2497 strbuf_addbuf(out, &untracked->ident);24982499 strbuf_add(out, ouc, ouc_size(len));2500 free(ouc);2501 ouc = NULL;25022503 if (!untracked->root) {2504 varint_len = encode_varint(0, varbuf);2505 strbuf_add(out, varbuf, varint_len);2506 return;2507 }25082509 wd.index = 0;2510 wd.check_only = ewah_new();2511 wd.valid = ewah_new();2512 wd.sha1_valid = ewah_new();2513 strbuf_init(&wd.out, 1024);2514 strbuf_init(&wd.sb_stat, 1024);2515 strbuf_init(&wd.sb_sha1, 1024);2516 write_one_dir(untracked->root, &wd);25172518 varint_len = encode_varint(wd.index, varbuf);2519 strbuf_add(out, varbuf, varint_len);2520 strbuf_addbuf(out, &wd.out);2521 ewah_serialize_strbuf(wd.valid, out);2522 ewah_serialize_strbuf(wd.check_only, out);2523 ewah_serialize_strbuf(wd.sha1_valid, out);2524 strbuf_addbuf(out, &wd.sb_stat);2525 strbuf_addbuf(out, &wd.sb_sha1);2526 strbuf_addch(out, '\0'); /* safe guard for string lists */25272528 ewah_free(wd.valid);2529 ewah_free(wd.check_only);2530 ewah_free(wd.sha1_valid);2531 strbuf_release(&wd.out);2532 strbuf_release(&wd.sb_stat);2533 strbuf_release(&wd.sb_sha1);2534}25352536static void free_untracked(struct untracked_cache_dir *ucd)2537{2538 int i;2539 if (!ucd)2540 return;2541 for (i = 0; i < ucd->dirs_nr; i++)2542 free_untracked(ucd->dirs[i]);2543 for (i = 0; i < ucd->untracked_nr; i++)2544 free(ucd->untracked[i]);2545 free(ucd->untracked);2546 free(ucd->dirs);2547 free(ucd);2548}25492550void free_untracked_cache(struct untracked_cache *uc)2551{2552 if (uc)2553 free_untracked(uc->root);2554 free(uc);2555}25562557struct read_data {2558 int index;2559 struct untracked_cache_dir **ucd;2560 struct ewah_bitmap *check_only;2561 struct ewah_bitmap *valid;2562 struct ewah_bitmap *sha1_valid;2563 const unsigned char *data;2564 const unsigned char *end;2565};25662567static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)2568{2569 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);2570 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);2571 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);2572 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);2573 to->sd_dev = get_be32(&from->sd_dev);2574 to->sd_ino = get_be32(&from->sd_ino);2575 to->sd_uid = get_be32(&from->sd_uid);2576 to->sd_gid = get_be32(&from->sd_gid);2577 to->sd_size = get_be32(&from->sd_size);2578}25792580static int read_one_dir(struct untracked_cache_dir **untracked_,2581 struct read_data *rd)2582{2583 struct untracked_cache_dir ud, *untracked;2584 const unsigned char *next, *data = rd->data, *end = rd->end;2585 unsigned int value;2586 int i, len;25872588 memset(&ud, 0, sizeof(ud));25892590 next = data;2591 value = decode_varint(&next);2592 if (next > end)2593 return -1;2594 ud.recurse = 1;2595 ud.untracked_alloc = value;2596 ud.untracked_nr = value;2597 if (ud.untracked_nr)2598 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);2599 data = next;26002601 next = data;2602 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);2603 if (next > end)2604 return -1;2605 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);2606 data = next;26072608 len = strlen((const char *)data);2609 next = data + len + 1;2610 if (next > rd->end)2611 return -1;2612 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));2613 memcpy(untracked, &ud, sizeof(ud));2614 memcpy(untracked->name, data, len + 1);2615 data = next;26162617 for (i = 0; i < untracked->untracked_nr; i++) {2618 len = strlen((const char *)data);2619 next = data + len + 1;2620 if (next > rd->end)2621 return -1;2622 untracked->untracked[i] = xstrdup((const char*)data);2623 data = next;2624 }26252626 rd->ucd[rd->index++] = untracked;2627 rd->data = data;26282629 for (i = 0; i < untracked->dirs_nr; i++) {2630 len = read_one_dir(untracked->dirs + i, rd);2631 if (len < 0)2632 return -1;2633 }2634 return 0;2635}26362637static void set_check_only(size_t pos, void *cb)2638{2639 struct read_data *rd = cb;2640 struct untracked_cache_dir *ud = rd->ucd[pos];2641 ud->check_only = 1;2642}26432644static void read_stat(size_t pos, void *cb)2645{2646 struct read_data *rd = cb;2647 struct untracked_cache_dir *ud = rd->ucd[pos];2648 if (rd->data + sizeof(struct stat_data) > rd->end) {2649 rd->data = rd->end + 1;2650 return;2651 }2652 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);2653 rd->data += sizeof(struct stat_data);2654 ud->valid = 1;2655}26562657static void read_sha1(size_t pos, void *cb)2658{2659 struct read_data *rd = cb;2660 struct untracked_cache_dir *ud = rd->ucd[pos];2661 if (rd->data + 20 > rd->end) {2662 rd->data = rd->end + 1;2663 return;2664 }2665 hashcpy(ud->exclude_sha1, rd->data);2666 rd->data += 20;2667}26682669static void load_sha1_stat(struct sha1_stat *sha1_stat,2670 const struct stat_data *stat,2671 const unsigned char *sha1)2672{2673 stat_data_from_disk(&sha1_stat->stat, stat);2674 hashcpy(sha1_stat->sha1, sha1);2675 sha1_stat->valid = 1;2676}26772678struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)2679{2680 const struct ondisk_untracked_cache *ouc;2681 struct untracked_cache *uc;2682 struct read_data rd;2683 const unsigned char *next = data, *end = (const unsigned char *)data + sz;2684 const char *ident;2685 int ident_len, len;26862687 if (sz <= 1 || end[-1] != '\0')2688 return NULL;2689 end--;26902691 ident_len = decode_varint(&next);2692 if (next + ident_len > end)2693 return NULL;2694 ident = (const char *)next;2695 next += ident_len;26962697 ouc = (const struct ondisk_untracked_cache *)next;2698 if (next + ouc_size(0) > end)2699 return NULL;27002701 uc = xcalloc(1, sizeof(*uc));2702 strbuf_init(&uc->ident, ident_len);2703 strbuf_add(&uc->ident, ident, ident_len);2704 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,2705 ouc->info_exclude_sha1);2706 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,2707 ouc->excludes_file_sha1);2708 uc->dir_flags = get_be32(&ouc->dir_flags);2709 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);2710 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */2711 next += ouc_size(strlen(ouc->exclude_per_dir));2712 if (next >= end)2713 goto done2;27142715 len = decode_varint(&next);2716 if (next > end || len == 0)2717 goto done2;27182719 rd.valid = ewah_new();2720 rd.check_only = ewah_new();2721 rd.sha1_valid = ewah_new();2722 rd.data = next;2723 rd.end = end;2724 rd.index = 0;2725 ALLOC_ARRAY(rd.ucd, len);27262727 if (read_one_dir(&uc->root, &rd) || rd.index != len)2728 goto done;27292730 next = rd.data;2731 len = ewah_read_mmap(rd.valid, next, end - next);2732 if (len < 0)2733 goto done;27342735 next += len;2736 len = ewah_read_mmap(rd.check_only, next, end - next);2737 if (len < 0)2738 goto done;27392740 next += len;2741 len = ewah_read_mmap(rd.sha1_valid, next, end - next);2742 if (len < 0)2743 goto done;27442745 ewah_each_bit(rd.check_only, set_check_only, &rd);2746 rd.data = next + len;2747 ewah_each_bit(rd.valid, read_stat, &rd);2748 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);2749 next = rd.data;27502751done:2752 free(rd.ucd);2753 ewah_free(rd.valid);2754 ewah_free(rd.check_only);2755 ewah_free(rd.sha1_valid);2756done2:2757 if (next != end) {2758 free_untracked_cache(uc);2759 uc = NULL;2760 }2761 return uc;2762}27632764static void invalidate_one_directory(struct untracked_cache *uc,2765 struct untracked_cache_dir *ucd)2766{2767 uc->dir_invalidated++;2768 ucd->valid = 0;2769 ucd->untracked_nr = 0;2770}27712772/*2773 * Normally when an entry is added or removed from a directory,2774 * invalidating that directory is enough. No need to touch its2775 * ancestors. When a directory is shown as "foo/bar/" in git-status2776 * however, deleting or adding an entry may have cascading effect.2777 *2778 * Say the "foo/bar/file" has become untracked, we need to tell the2779 * untracked_cache_dir of "foo" that "bar/" is not an untracked2780 * directory any more (because "bar" is managed by foo as an untracked2781 * "file").2782 *2783 * Similarly, if "foo/bar/file" moves from untracked to tracked and it2784 * was the last untracked entry in the entire "foo", we should show2785 * "foo/" instead. Which means we have to invalidate past "bar" up to2786 * "foo".2787 *2788 * This function traverses all directories from root to leaf. If there2789 * is a chance of one of the above cases happening, we invalidate back2790 * to root. Otherwise we just invalidate the leaf. There may be a more2791 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to2792 * detect these cases and avoid unnecessary invalidation, for example,2793 * checking for the untracked entry named "bar/" in "foo", but for now2794 * stick to something safe and simple.2795 */2796static int invalidate_one_component(struct untracked_cache *uc,2797 struct untracked_cache_dir *dir,2798 const char *path, int len)2799{2800 const char *rest = strchr(path, '/');28012802 if (rest) {2803 int component_len = rest - path;2804 struct untracked_cache_dir *d =2805 lookup_untracked(uc, dir, path, component_len);2806 int ret =2807 invalidate_one_component(uc, d, rest + 1,2808 len - (component_len + 1));2809 if (ret)2810 invalidate_one_directory(uc, dir);2811 return ret;2812 }28132814 invalidate_one_directory(uc, dir);2815 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;2816}28172818void untracked_cache_invalidate_path(struct index_state *istate,2819 const char *path)2820{2821 if (!istate->untracked || !istate->untracked->root)2822 return;2823 invalidate_one_component(istate->untracked, istate->untracked->root,2824 path, strlen(path));2825}28262827void untracked_cache_remove_from_index(struct index_state *istate,2828 const char *path)2829{2830 untracked_cache_invalidate_path(istate, path);2831}28322833void untracked_cache_add_to_index(struct index_state *istate,2834 const char *path)2835{2836 untracked_cache_invalidate_path(istate, path);2837}28382839/* Update gitfile and core.worktree setting to connect work tree and git dir */2840void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)2841{2842 struct strbuf gitfile_sb = STRBUF_INIT;2843 struct strbuf cfg_sb = STRBUF_INIT;2844 struct strbuf rel_path = STRBUF_INIT;2845 char *git_dir, *work_tree;28462847 /* Prepare .git file */2848 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);2849 if (safe_create_leading_directories_const(gitfile_sb.buf))2850 die(_("could not create directories for %s"), gitfile_sb.buf);28512852 /* Prepare config file */2853 strbuf_addf(&cfg_sb, "%s/config", git_dir_);2854 if (safe_create_leading_directories_const(cfg_sb.buf))2855 die(_("could not create directories for %s"), cfg_sb.buf);28562857 git_dir = real_pathdup(git_dir_, 1);2858 work_tree = real_pathdup(work_tree_, 1);28592860 /* Write .git file */2861 write_file(gitfile_sb.buf, "gitdir: %s",2862 relative_path(git_dir, work_tree, &rel_path));2863 /* Update core.worktree setting */2864 git_config_set_in_file(cfg_sb.buf, "core.worktree",2865 relative_path(work_tree, git_dir, &rel_path));28662867 strbuf_release(&gitfile_sb);2868 strbuf_release(&cfg_sb);2869 strbuf_release(&rel_path);2870 free(work_tree);2871 free(git_dir);2872}28732874/*2875 * Migrate the git directory of the given path from old_git_dir to new_git_dir.2876 */2877void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)2878{2879 if (rename(old_git_dir, new_git_dir) < 0)2880 die_errno(_("could not migrate git directory from '%s' to '%s'"),2881 old_git_dir, new_git_dir);28822883 connect_work_tree_and_git_dir(path, new_git_dir);2884}