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 const char *path, int len, struct untracked_cache_dir *untracked, 50 int check_only, const struct pathspec *pathspec); 51static int get_dtype(struct dirent *de, struct index_state *istate, 52 const char *path, int len); 53 54int fspathcmp(const char *a, const char *b) 55{ 56 return ignore_case ? strcasecmp(a, b) : strcmp(a, b); 57} 58 59int fspathncmp(const char *a, const char *b, size_t count) 60{ 61 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count); 62} 63 64int git_fnmatch(const struct pathspec_item *item, 65 const char *pattern, const char *string, 66 int prefix) 67{ 68 if (prefix > 0) { 69 if (ps_strncmp(item, pattern, string, prefix)) 70 return WM_NOMATCH; 71 pattern += prefix; 72 string += prefix; 73 } 74 if (item->flags & PATHSPEC_ONESTAR) { 75 int pattern_len = strlen(++pattern); 76 int string_len = strlen(string); 77 return string_len < pattern_len || 78 ps_strcmp(item, pattern, 79 string + string_len - pattern_len); 80 } 81 if (item->magic & PATHSPEC_GLOB) 82 return wildmatch(pattern, string, 83 WM_PATHNAME | 84 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0), 85 NULL); 86 else 87 /* wildmatch has not learned no FNM_PATHNAME mode yet */ 88 return wildmatch(pattern, string, 89 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0, 90 NULL); 91} 92 93static int fnmatch_icase_mem(const char *pattern, int patternlen, 94 const char *string, int stringlen, 95 int flags) 96{ 97 int match_status; 98 struct strbuf pat_buf = STRBUF_INIT; 99 struct strbuf str_buf = STRBUF_INIT; 100 const char *use_pat = pattern; 101 const char *use_str = string; 102 103 if (pattern[patternlen]) { 104 strbuf_add(&pat_buf, pattern, patternlen); 105 use_pat = pat_buf.buf; 106 } 107 if (string[stringlen]) { 108 strbuf_add(&str_buf, string, stringlen); 109 use_str = str_buf.buf; 110 } 111 112 if (ignore_case) 113 flags |= WM_CASEFOLD; 114 match_status = wildmatch(use_pat, use_str, flags, NULL); 115 116 strbuf_release(&pat_buf); 117 strbuf_release(&str_buf); 118 119 return match_status; 120} 121 122static size_t common_prefix_len(const struct pathspec *pathspec) 123{ 124 int n; 125 size_t max = 0; 126 127 /* 128 * ":(icase)path" is treated as a pathspec full of 129 * wildcard. In other words, only prefix is considered common 130 * prefix. If the pathspec is abc/foo abc/bar, running in 131 * subdir xyz, the common prefix is still xyz, not xuz/abc as 132 * in non-:(icase). 133 */ 134 GUARD_PATHSPEC(pathspec, 135 PATHSPEC_FROMTOP | 136 PATHSPEC_MAXDEPTH | 137 PATHSPEC_LITERAL | 138 PATHSPEC_GLOB | 139 PATHSPEC_ICASE | 140 PATHSPEC_EXCLUDE | 141 PATHSPEC_ATTR); 142 143 for (n = 0; n < pathspec->nr; n++) { 144 size_t i = 0, len = 0, item_len; 145 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE) 146 continue; 147 if (pathspec->items[n].magic & PATHSPEC_ICASE) 148 item_len = pathspec->items[n].prefix; 149 else 150 item_len = pathspec->items[n].nowildcard_len; 151 while (i < item_len && (n == 0 || i < max)) { 152 char c = pathspec->items[n].match[i]; 153 if (c != pathspec->items[0].match[i]) 154 break; 155 if (c == '/') 156 len = i + 1; 157 i++; 158 } 159 if (n == 0 || len < max) { 160 max = len; 161 if (!max) 162 break; 163 } 164 } 165 return max; 166} 167 168/* 169 * Returns a copy of the longest leading path common among all 170 * pathspecs. 171 */ 172char *common_prefix(const struct pathspec *pathspec) 173{ 174 unsigned long len = common_prefix_len(pathspec); 175 176 return len ? xmemdupz(pathspec->items[0].match, len) : NULL; 177} 178 179int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec) 180{ 181 const char *prefix; 182 size_t prefix_len; 183 184 /* 185 * Calculate common prefix for the pathspec, and 186 * use that to optimize the directory walk 187 */ 188 prefix_len = common_prefix_len(pathspec); 189 prefix = prefix_len ? pathspec->items[0].match : ""; 190 191 /* Read the directory and prune it */ 192 read_directory(dir, prefix, prefix_len, pathspec); 193 194 return prefix_len; 195} 196 197int within_depth(const char *name, int namelen, 198 int depth, int max_depth) 199{ 200 const char *cp = name, *cpe = name + namelen; 201 202 while (cp < cpe) { 203 if (*cp++ != '/') 204 continue; 205 depth++; 206 if (depth > max_depth) 207 return 0; 208 } 209 return 1; 210} 211 212#define DO_MATCH_EXCLUDE (1<<0) 213#define DO_MATCH_DIRECTORY (1<<1) 214#define DO_MATCH_SUBMODULE (1<<2) 215 216static int match_attrs(const char *name, int namelen, 217 const struct pathspec_item *item) 218{ 219 int i; 220 221 git_check_attr(name, item->attr_check); 222 for (i = 0; i < item->attr_match_nr; i++) { 223 const char *value; 224 int matched; 225 enum attr_match_mode match_mode; 226 227 value = item->attr_check->items[i].value; 228 match_mode = item->attr_match[i].match_mode; 229 230 if (ATTR_TRUE(value)) 231 matched = (match_mode == MATCH_SET); 232 else if (ATTR_FALSE(value)) 233 matched = (match_mode == MATCH_UNSET); 234 else if (ATTR_UNSET(value)) 235 matched = (match_mode == MATCH_UNSPECIFIED); 236 else 237 matched = (match_mode == MATCH_VALUE && 238 !strcmp(item->attr_match[i].value, value)); 239 if (!matched) 240 return 0; 241 } 242 243 return 1; 244} 245 246/* 247 * Does 'match' match the given name? 248 * A match is found if 249 * 250 * (1) the 'match' string is leading directory of 'name', or 251 * (2) the 'match' string is a wildcard and matches 'name', or 252 * (3) the 'match' string is exactly the same as 'name'. 253 * 254 * and the return value tells which case it was. 255 * 256 * It returns 0 when there is no match. 257 */ 258static int match_pathspec_item(const struct pathspec_item *item, int prefix, 259 const char *name, int namelen, unsigned flags) 260{ 261 /* name/namelen has prefix cut off by caller */ 262 const char *match = item->match + prefix; 263 int matchlen = item->len - prefix; 264 265 /* 266 * The normal call pattern is: 267 * 1. prefix = common_prefix_len(ps); 268 * 2. prune something, or fill_directory 269 * 3. match_pathspec() 270 * 271 * 'prefix' at #1 may be shorter than the command's prefix and 272 * it's ok for #2 to match extra files. Those extras will be 273 * trimmed at #3. 274 * 275 * Suppose the pathspec is 'foo' and '../bar' running from 276 * subdir 'xyz'. The common prefix at #1 will be empty, thanks 277 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The 278 * user does not want XYZ/foo, only the "foo" part should be 279 * case-insensitive. We need to filter out XYZ/foo here. In 280 * other words, we do not trust the caller on comparing the 281 * prefix part when :(icase) is involved. We do exact 282 * comparison ourselves. 283 * 284 * Normally the caller (common_prefix_len() in fact) does 285 * _exact_ matching on name[-prefix+1..-1] and we do not need 286 * to check that part. Be defensive and check it anyway, in 287 * case common_prefix_len is changed, or a new caller is 288 * introduced that does not use common_prefix_len. 289 * 290 * If the penalty turns out too high when prefix is really 291 * long, maybe change it to 292 * strncmp(match, name, item->prefix - prefix) 293 */ 294 if (item->prefix && (item->magic & PATHSPEC_ICASE) && 295 strncmp(item->match, name - prefix, item->prefix)) 296 return 0; 297 298 if (item->attr_match_nr && !match_attrs(name, namelen, item)) 299 return 0; 300 301 /* If the match was just the prefix, we matched */ 302 if (!*match) 303 return MATCHED_RECURSIVELY; 304 305 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) { 306 if (matchlen == namelen) 307 return MATCHED_EXACTLY; 308 309 if (match[matchlen-1] == '/' || name[matchlen] == '/') 310 return MATCHED_RECURSIVELY; 311 } else if ((flags & DO_MATCH_DIRECTORY) && 312 match[matchlen - 1] == '/' && 313 namelen == matchlen - 1 && 314 !ps_strncmp(item, match, name, namelen)) 315 return MATCHED_EXACTLY; 316 317 if (item->nowildcard_len < item->len && 318 !git_fnmatch(item, match, name, 319 item->nowildcard_len - prefix)) 320 return MATCHED_FNMATCH; 321 322 /* Perform checks to see if "name" is a super set of the pathspec */ 323 if (flags & DO_MATCH_SUBMODULE) { 324 /* name is a literal prefix of the pathspec */ 325 if ((namelen < matchlen) && 326 (match[namelen] == '/') && 327 !ps_strncmp(item, match, name, namelen)) 328 return MATCHED_RECURSIVELY; 329 330 /* name" doesn't match up to the first wild character */ 331 if (item->nowildcard_len < item->len && 332 ps_strncmp(item, match, name, 333 item->nowildcard_len - prefix)) 334 return 0; 335 336 /* 337 * Here is where we would perform a wildmatch to check if 338 * "name" can be matched as a directory (or a prefix) against 339 * the pathspec. Since wildmatch doesn't have this capability 340 * at the present we have to punt and say that it is a match, 341 * potentially returning a false positive 342 * The submodules themselves will be able to perform more 343 * accurate matching to determine if the pathspec matches. 344 */ 345 return MATCHED_RECURSIVELY; 346 } 347 348 return 0; 349} 350 351/* 352 * Given a name and a list of pathspecs, returns the nature of the 353 * closest (i.e. most specific) match of the name to any of the 354 * pathspecs. 355 * 356 * The caller typically calls this multiple times with the same 357 * pathspec and seen[] array but with different name/namelen 358 * (e.g. entries from the index) and is interested in seeing if and 359 * how each pathspec matches all the names it calls this function 360 * with. A mark is left in the seen[] array for each pathspec element 361 * indicating the closest type of match that element achieved, so if 362 * seen[n] remains zero after multiple invocations, that means the nth 363 * pathspec did not match any names, which could indicate that the 364 * user mistyped the nth pathspec. 365 */ 366static int do_match_pathspec(const struct pathspec *ps, 367 const char *name, int namelen, 368 int prefix, char *seen, 369 unsigned flags) 370{ 371 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE; 372 373 GUARD_PATHSPEC(ps, 374 PATHSPEC_FROMTOP | 375 PATHSPEC_MAXDEPTH | 376 PATHSPEC_LITERAL | 377 PATHSPEC_GLOB | 378 PATHSPEC_ICASE | 379 PATHSPEC_EXCLUDE | 380 PATHSPEC_ATTR); 381 382 if (!ps->nr) { 383 if (!ps->recursive || 384 !(ps->magic & PATHSPEC_MAXDEPTH) || 385 ps->max_depth == -1) 386 return MATCHED_RECURSIVELY; 387 388 if (within_depth(name, namelen, 0, ps->max_depth)) 389 return MATCHED_EXACTLY; 390 else 391 return 0; 392 } 393 394 name += prefix; 395 namelen -= prefix; 396 397 for (i = ps->nr - 1; i >= 0; i--) { 398 int how; 399 400 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) || 401 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE))) 402 continue; 403 404 if (seen && seen[i] == MATCHED_EXACTLY) 405 continue; 406 /* 407 * Make exclude patterns optional and never report 408 * "pathspec ':(exclude)foo' matches no files" 409 */ 410 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE) 411 seen[i] = MATCHED_FNMATCH; 412 how = match_pathspec_item(ps->items+i, prefix, name, 413 namelen, flags); 414 if (ps->recursive && 415 (ps->magic & PATHSPEC_MAXDEPTH) && 416 ps->max_depth != -1 && 417 how && how != MATCHED_FNMATCH) { 418 int len = ps->items[i].len; 419 if (name[len] == '/') 420 len++; 421 if (within_depth(name+len, namelen-len, 0, ps->max_depth)) 422 how = MATCHED_EXACTLY; 423 else 424 how = 0; 425 } 426 if (how) { 427 if (retval < how) 428 retval = how; 429 if (seen && seen[i] < how) 430 seen[i] = how; 431 } 432 } 433 return retval; 434} 435 436int match_pathspec(const struct pathspec *ps, 437 const char *name, int namelen, 438 int prefix, char *seen, int is_dir) 439{ 440 int positive, negative; 441 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0; 442 positive = do_match_pathspec(ps, name, namelen, 443 prefix, seen, flags); 444 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive) 445 return positive; 446 negative = do_match_pathspec(ps, name, namelen, 447 prefix, seen, 448 flags | DO_MATCH_EXCLUDE); 449 return negative ? 0 : positive; 450} 451 452/** 453 * Check if a submodule is a superset of the pathspec 454 */ 455int submodule_path_match(const struct pathspec *ps, 456 const char *submodule_name, 457 char *seen) 458{ 459 int matched = do_match_pathspec(ps, submodule_name, 460 strlen(submodule_name), 461 0, seen, 462 DO_MATCH_DIRECTORY | 463 DO_MATCH_SUBMODULE); 464 return matched; 465} 466 467int report_path_error(const char *ps_matched, 468 const struct pathspec *pathspec, 469 const char *prefix) 470{ 471 /* 472 * Make sure all pathspec matched; otherwise it is an error. 473 */ 474 int num, errors = 0; 475 for (num = 0; num < pathspec->nr; num++) { 476 int other, found_dup; 477 478 if (ps_matched[num]) 479 continue; 480 /* 481 * The caller might have fed identical pathspec 482 * twice. Do not barf on such a mistake. 483 * FIXME: parse_pathspec should have eliminated 484 * duplicate pathspec. 485 */ 486 for (found_dup = other = 0; 487 !found_dup && other < pathspec->nr; 488 other++) { 489 if (other == num || !ps_matched[other]) 490 continue; 491 if (!strcmp(pathspec->items[other].original, 492 pathspec->items[num].original)) 493 /* 494 * Ok, we have a match already. 495 */ 496 found_dup = 1; 497 } 498 if (found_dup) 499 continue; 500 501 error("pathspec '%s' did not match any file(s) known to git.", 502 pathspec->items[num].original); 503 errors++; 504 } 505 return errors; 506} 507 508/* 509 * Return the length of the "simple" part of a path match limiter. 510 */ 511int simple_length(const char *match) 512{ 513 int len = -1; 514 515 for (;;) { 516 unsigned char c = *match++; 517 len++; 518 if (c == '\0' || is_glob_special(c)) 519 return len; 520 } 521} 522 523int no_wildcard(const char *string) 524{ 525 return string[simple_length(string)] == '\0'; 526} 527 528void parse_exclude_pattern(const char **pattern, 529 int *patternlen, 530 unsigned *flags, 531 int *nowildcardlen) 532{ 533 const char *p = *pattern; 534 size_t i, len; 535 536 *flags = 0; 537 if (*p == '!') { 538 *flags |= EXC_FLAG_NEGATIVE; 539 p++; 540 } 541 len = strlen(p); 542 if (len && p[len - 1] == '/') { 543 len--; 544 *flags |= EXC_FLAG_MUSTBEDIR; 545 } 546 for (i = 0; i < len; i++) { 547 if (p[i] == '/') 548 break; 549 } 550 if (i == len) 551 *flags |= EXC_FLAG_NODIR; 552 *nowildcardlen = simple_length(p); 553 /* 554 * we should have excluded the trailing slash from 'p' too, 555 * but that's one more allocation. Instead just make sure 556 * nowildcardlen does not exceed real patternlen 557 */ 558 if (*nowildcardlen > len) 559 *nowildcardlen = len; 560 if (*p == '*' && no_wildcard(p + 1)) 561 *flags |= EXC_FLAG_ENDSWITH; 562 *pattern = p; 563 *patternlen = len; 564} 565 566void add_exclude(const char *string, const char *base, 567 int baselen, struct exclude_list *el, int srcpos) 568{ 569 struct exclude *x; 570 int patternlen; 571 unsigned flags; 572 int nowildcardlen; 573 574 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen); 575 if (flags & EXC_FLAG_MUSTBEDIR) { 576 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen); 577 } else { 578 x = xmalloc(sizeof(*x)); 579 x->pattern = string; 580 } 581 x->patternlen = patternlen; 582 x->nowildcardlen = nowildcardlen; 583 x->base = base; 584 x->baselen = baselen; 585 x->flags = flags; 586 x->srcpos = srcpos; 587 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc); 588 el->excludes[el->nr++] = x; 589 x->el = el; 590} 591 592static void *read_skip_worktree_file_from_index(const struct index_state *istate, 593 const char *path, size_t *size, 594 struct sha1_stat *sha1_stat) 595{ 596 int pos, len; 597 unsigned long sz; 598 enum object_type type; 599 void *data; 600 601 len = strlen(path); 602 pos = index_name_pos(istate, path, len); 603 if (pos < 0) 604 return NULL; 605 if (!ce_skip_worktree(istate->cache[pos])) 606 return NULL; 607 data = read_sha1_file(istate->cache[pos]->oid.hash, &type, &sz); 608 if (!data || type != OBJ_BLOB) { 609 free(data); 610 return NULL; 611 } 612 *size = xsize_t(sz); 613 if (sha1_stat) { 614 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat)); 615 hashcpy(sha1_stat->sha1, istate->cache[pos]->oid.hash); 616 } 617 return data; 618} 619 620/* 621 * Frees memory within el which was allocated for exclude patterns and 622 * the file buffer. Does not free el itself. 623 */ 624void clear_exclude_list(struct exclude_list *el) 625{ 626 int i; 627 628 for (i = 0; i < el->nr; i++) 629 free(el->excludes[i]); 630 free(el->excludes); 631 free(el->filebuf); 632 633 memset(el, 0, sizeof(*el)); 634} 635 636static void trim_trailing_spaces(char *buf) 637{ 638 char *p, *last_space = NULL; 639 640 for (p = buf; *p; p++) 641 switch (*p) { 642 case ' ': 643 if (!last_space) 644 last_space = p; 645 break; 646 case '\\': 647 p++; 648 if (!*p) 649 return; 650 /* fallthrough */ 651 default: 652 last_space = NULL; 653 } 654 655 if (last_space) 656 *last_space = '\0'; 657} 658 659/* 660 * Given a subdirectory name and "dir" of the current directory, 661 * search the subdir in "dir" and return it, or create a new one if it 662 * does not exist in "dir". 663 * 664 * If "name" has the trailing slash, it'll be excluded in the search. 665 */ 666static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc, 667 struct untracked_cache_dir *dir, 668 const char *name, int len) 669{ 670 int first, last; 671 struct untracked_cache_dir *d; 672 if (!dir) 673 return NULL; 674 if (len && name[len - 1] == '/') 675 len--; 676 first = 0; 677 last = dir->dirs_nr; 678 while (last > first) { 679 int cmp, next = (last + first) >> 1; 680 d = dir->dirs[next]; 681 cmp = strncmp(name, d->name, len); 682 if (!cmp && strlen(d->name) > len) 683 cmp = -1; 684 if (!cmp) 685 return d; 686 if (cmp < 0) { 687 last = next; 688 continue; 689 } 690 first = next+1; 691 } 692 693 uc->dir_created++; 694 FLEX_ALLOC_MEM(d, name, name, len); 695 696 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc); 697 memmove(dir->dirs + first + 1, dir->dirs + first, 698 (dir->dirs_nr - first) * sizeof(*dir->dirs)); 699 dir->dirs_nr++; 700 dir->dirs[first] = d; 701 return d; 702} 703 704static void do_invalidate_gitignore(struct untracked_cache_dir *dir) 705{ 706 int i; 707 dir->valid = 0; 708 dir->untracked_nr = 0; 709 for (i = 0; i < dir->dirs_nr; i++) 710 do_invalidate_gitignore(dir->dirs[i]); 711} 712 713static void invalidate_gitignore(struct untracked_cache *uc, 714 struct untracked_cache_dir *dir) 715{ 716 uc->gitignore_invalidated++; 717 do_invalidate_gitignore(dir); 718} 719 720static void invalidate_directory(struct untracked_cache *uc, 721 struct untracked_cache_dir *dir) 722{ 723 int i; 724 uc->dir_invalidated++; 725 dir->valid = 0; 726 dir->untracked_nr = 0; 727 for (i = 0; i < dir->dirs_nr; i++) 728 dir->dirs[i]->recurse = 0; 729} 730 731/* 732 * Given a file with name "fname", read it (either from disk, or from 733 * an index if 'istate' is non-null), parse it and store the 734 * exclude rules in "el". 735 * 736 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill 737 * stat data from disk (only valid if add_excludes returns zero). If 738 * ss_valid is non-zero, "ss" must contain good value as input. 739 */ 740static int add_excludes(const char *fname, const char *base, int baselen, 741 struct exclude_list *el, 742 struct index_state *istate, 743 struct sha1_stat *sha1_stat) 744{ 745 struct stat st; 746 int fd, i, lineno = 1; 747 size_t size = 0; 748 char *buf, *entry; 749 750 fd = open(fname, O_RDONLY); 751 if (fd < 0 || fstat(fd, &st) < 0) { 752 if (errno != ENOENT) 753 warn_on_inaccessible(fname); 754 if (0 <= fd) 755 close(fd); 756 if (!istate || 757 (buf = read_skip_worktree_file_from_index(istate, fname, &size, sha1_stat)) == NULL) 758 return -1; 759 if (size == 0) { 760 free(buf); 761 return 0; 762 } 763 if (buf[size-1] != '\n') { 764 buf = xrealloc(buf, st_add(size, 1)); 765 buf[size++] = '\n'; 766 } 767 } else { 768 size = xsize_t(st.st_size); 769 if (size == 0) { 770 if (sha1_stat) { 771 fill_stat_data(&sha1_stat->stat, &st); 772 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN); 773 sha1_stat->valid = 1; 774 } 775 close(fd); 776 return 0; 777 } 778 buf = xmallocz(size); 779 if (read_in_full(fd, buf, size) != size) { 780 free(buf); 781 close(fd); 782 return -1; 783 } 784 buf[size++] = '\n'; 785 close(fd); 786 if (sha1_stat) { 787 int pos; 788 if (sha1_stat->valid && 789 !match_stat_data_racy(istate, &sha1_stat->stat, &st)) 790 ; /* no content change, ss->sha1 still good */ 791 else if (istate && 792 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 && 793 !ce_stage(istate->cache[pos]) && 794 ce_uptodate(istate->cache[pos]) && 795 !would_convert_to_git(fname)) 796 hashcpy(sha1_stat->sha1, 797 istate->cache[pos]->oid.hash); 798 else 799 hash_sha1_file(buf, size, "blob", sha1_stat->sha1); 800 fill_stat_data(&sha1_stat->stat, &st); 801 sha1_stat->valid = 1; 802 } 803 } 804 805 el->filebuf = buf; 806 807 if (skip_utf8_bom(&buf, size)) 808 size -= buf - el->filebuf; 809 810 entry = buf; 811 812 for (i = 0; i < size; i++) { 813 if (buf[i] == '\n') { 814 if (entry != buf + i && entry[0] != '#') { 815 buf[i - (i && buf[i-1] == '\r')] = 0; 816 trim_trailing_spaces(entry); 817 add_exclude(entry, base, baselen, el, lineno); 818 } 819 lineno++; 820 entry = buf + i + 1; 821 } 822 } 823 return 0; 824} 825 826int add_excludes_from_file_to_list(const char *fname, const char *base, 827 int baselen, struct exclude_list *el, 828 struct index_state *istate) 829{ 830 return add_excludes(fname, base, baselen, el, istate, NULL); 831} 832 833struct exclude_list *add_exclude_list(struct dir_struct *dir, 834 int group_type, const char *src) 835{ 836 struct exclude_list *el; 837 struct exclude_list_group *group; 838 839 group = &dir->exclude_list_group[group_type]; 840 ALLOC_GROW(group->el, group->nr + 1, group->alloc); 841 el = &group->el[group->nr++]; 842 memset(el, 0, sizeof(*el)); 843 el->src = src; 844 return el; 845} 846 847/* 848 * Used to set up core.excludesfile and .git/info/exclude lists. 849 */ 850static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname, 851 struct sha1_stat *sha1_stat) 852{ 853 struct exclude_list *el; 854 /* 855 * catch setup_standard_excludes() that's called before 856 * dir->untracked is assigned. That function behaves 857 * differently when dir->untracked is non-NULL. 858 */ 859 if (!dir->untracked) 860 dir->unmanaged_exclude_files++; 861 el = add_exclude_list(dir, EXC_FILE, fname); 862 if (add_excludes(fname, "", 0, el, NULL, sha1_stat) < 0) 863 die("cannot use %s as an exclude file", fname); 864} 865 866void add_excludes_from_file(struct dir_struct *dir, const char *fname) 867{ 868 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */ 869 add_excludes_from_file_1(dir, fname, NULL); 870} 871 872int match_basename(const char *basename, int basenamelen, 873 const char *pattern, int prefix, int patternlen, 874 unsigned flags) 875{ 876 if (prefix == patternlen) { 877 if (patternlen == basenamelen && 878 !fspathncmp(pattern, basename, basenamelen)) 879 return 1; 880 } else if (flags & EXC_FLAG_ENDSWITH) { 881 /* "*literal" matching against "fooliteral" */ 882 if (patternlen - 1 <= basenamelen && 883 !fspathncmp(pattern + 1, 884 basename + basenamelen - (patternlen - 1), 885 patternlen - 1)) 886 return 1; 887 } else { 888 if (fnmatch_icase_mem(pattern, patternlen, 889 basename, basenamelen, 890 0) == 0) 891 return 1; 892 } 893 return 0; 894} 895 896int match_pathname(const char *pathname, int pathlen, 897 const char *base, int baselen, 898 const char *pattern, int prefix, int patternlen, 899 unsigned flags) 900{ 901 const char *name; 902 int namelen; 903 904 /* 905 * match with FNM_PATHNAME; the pattern has base implicitly 906 * in front of it. 907 */ 908 if (*pattern == '/') { 909 pattern++; 910 patternlen--; 911 prefix--; 912 } 913 914 /* 915 * baselen does not count the trailing slash. base[] may or 916 * may not end with a trailing slash though. 917 */ 918 if (pathlen < baselen + 1 || 919 (baselen && pathname[baselen] != '/') || 920 fspathncmp(pathname, base, baselen)) 921 return 0; 922 923 namelen = baselen ? pathlen - baselen - 1 : pathlen; 924 name = pathname + pathlen - namelen; 925 926 if (prefix) { 927 /* 928 * if the non-wildcard part is longer than the 929 * remaining pathname, surely it cannot match. 930 */ 931 if (prefix > namelen) 932 return 0; 933 934 if (fspathncmp(pattern, name, prefix)) 935 return 0; 936 pattern += prefix; 937 patternlen -= prefix; 938 name += prefix; 939 namelen -= prefix; 940 941 /* 942 * If the whole pattern did not have a wildcard, 943 * then our prefix match is all we need; we 944 * do not need to call fnmatch at all. 945 */ 946 if (!patternlen && !namelen) 947 return 1; 948 } 949 950 return fnmatch_icase_mem(pattern, patternlen, 951 name, namelen, 952 WM_PATHNAME) == 0; 953} 954 955/* 956 * Scan the given exclude list in reverse to see whether pathname 957 * should be ignored. The first match (i.e. the last on the list), if 958 * any, determines the fate. Returns the exclude_list element which 959 * matched, or NULL for undecided. 960 */ 961static struct exclude *last_exclude_matching_from_list(const char *pathname, 962 int pathlen, 963 const char *basename, 964 int *dtype, 965 struct exclude_list *el, 966 struct index_state *istate) 967{ 968 struct exclude *exc = NULL; /* undecided */ 969 int i; 970 971 if (!el->nr) 972 return NULL; /* undefined */ 973 974 for (i = el->nr - 1; 0 <= i; i--) { 975 struct exclude *x = el->excludes[i]; 976 const char *exclude = x->pattern; 977 int prefix = x->nowildcardlen; 978 979 if (x->flags & EXC_FLAG_MUSTBEDIR) { 980 if (*dtype == DT_UNKNOWN) 981 *dtype = get_dtype(NULL, istate, pathname, pathlen); 982 if (*dtype != DT_DIR) 983 continue; 984 } 985 986 if (x->flags & EXC_FLAG_NODIR) { 987 if (match_basename(basename, 988 pathlen - (basename - pathname), 989 exclude, prefix, x->patternlen, 990 x->flags)) { 991 exc = x; 992 break; 993 } 994 continue; 995 } 996 997 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/'); 998 if (match_pathname(pathname, pathlen, 999 x->base, x->baselen ? x->baselen - 1 : 0,1000 exclude, prefix, x->patternlen, x->flags)) {1001 exc = x;1002 break;1003 }1004 }1005 return exc;1006}10071008/*1009 * Scan the list and let the last match determine the fate.1010 * Return 1 for exclude, 0 for include and -1 for undecided.1011 */1012int is_excluded_from_list(const char *pathname,1013 int pathlen, const char *basename, int *dtype,1014 struct exclude_list *el, struct index_state *istate)1015{1016 struct exclude *exclude;1017 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,1018 dtype, el, istate);1019 if (exclude)1020 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1021 return -1; /* undecided */1022}10231024static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,1025 struct index_state *istate,1026 const char *pathname, int pathlen, const char *basename,1027 int *dtype_p)1028{1029 int i, j;1030 struct exclude_list_group *group;1031 struct exclude *exclude;1032 for (i = EXC_CMDL; i <= EXC_FILE; i++) {1033 group = &dir->exclude_list_group[i];1034 for (j = group->nr - 1; j >= 0; j--) {1035 exclude = last_exclude_matching_from_list(1036 pathname, pathlen, basename, dtype_p,1037 &group->el[j], istate);1038 if (exclude)1039 return exclude;1040 }1041 }1042 return NULL;1043}10441045/*1046 * Loads the per-directory exclude list for the substring of base1047 * which has a char length of baselen.1048 */1049static void prep_exclude(struct dir_struct *dir,1050 struct index_state *istate,1051 const char *base, int baselen)1052{1053 struct exclude_list_group *group;1054 struct exclude_list *el;1055 struct exclude_stack *stk = NULL;1056 struct untracked_cache_dir *untracked;1057 int current;10581059 group = &dir->exclude_list_group[EXC_DIRS];10601061 /*1062 * Pop the exclude lists from the EXCL_DIRS exclude_list_group1063 * which originate from directories not in the prefix of the1064 * path being checked.1065 */1066 while ((stk = dir->exclude_stack) != NULL) {1067 if (stk->baselen <= baselen &&1068 !strncmp(dir->basebuf.buf, base, stk->baselen))1069 break;1070 el = &group->el[dir->exclude_stack->exclude_ix];1071 dir->exclude_stack = stk->prev;1072 dir->exclude = NULL;1073 free((char *)el->src); /* see strbuf_detach() below */1074 clear_exclude_list(el);1075 free(stk);1076 group->nr--;1077 }10781079 /* Skip traversing into sub directories if the parent is excluded */1080 if (dir->exclude)1081 return;10821083 /*1084 * Lazy initialization. All call sites currently just1085 * memset(dir, 0, sizeof(*dir)) before use. Changing all of1086 * them seems lots of work for little benefit.1087 */1088 if (!dir->basebuf.buf)1089 strbuf_init(&dir->basebuf, PATH_MAX);10901091 /* Read from the parent directories and push them down. */1092 current = stk ? stk->baselen : -1;1093 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);1094 if (dir->untracked)1095 untracked = stk ? stk->ucd : dir->untracked->root;1096 else1097 untracked = NULL;10981099 while (current < baselen) {1100 const char *cp;1101 struct sha1_stat sha1_stat;11021103 stk = xcalloc(1, sizeof(*stk));1104 if (current < 0) {1105 cp = base;1106 current = 0;1107 } else {1108 cp = strchr(base + current + 1, '/');1109 if (!cp)1110 die("oops in prep_exclude");1111 cp++;1112 untracked =1113 lookup_untracked(dir->untracked, untracked,1114 base + current,1115 cp - base - current);1116 }1117 stk->prev = dir->exclude_stack;1118 stk->baselen = cp - base;1119 stk->exclude_ix = group->nr;1120 stk->ucd = untracked;1121 el = add_exclude_list(dir, EXC_DIRS, NULL);1122 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);1123 assert(stk->baselen == dir->basebuf.len);11241125 /* Abort if the directory is excluded */1126 if (stk->baselen) {1127 int dt = DT_DIR;1128 dir->basebuf.buf[stk->baselen - 1] = 0;1129 dir->exclude = last_exclude_matching_from_lists(dir,1130 istate,1131 dir->basebuf.buf, stk->baselen - 1,1132 dir->basebuf.buf + current, &dt);1133 dir->basebuf.buf[stk->baselen - 1] = '/';1134 if (dir->exclude &&1135 dir->exclude->flags & EXC_FLAG_NEGATIVE)1136 dir->exclude = NULL;1137 if (dir->exclude) {1138 dir->exclude_stack = stk;1139 return;1140 }1141 }11421143 /* Try to read per-directory file */1144 hashclr(sha1_stat.sha1);1145 sha1_stat.valid = 0;1146 if (dir->exclude_per_dir &&1147 /*1148 * If we know that no files have been added in1149 * this directory (i.e. valid_cached_dir() has1150 * been executed and set untracked->valid) ..1151 */1152 (!untracked || !untracked->valid ||1153 /*1154 * .. and .gitignore does not exist before1155 * (i.e. null exclude_sha1). Then we can skip1156 * loading .gitignore, which would result in1157 * ENOENT anyway.1158 */1159 !is_null_sha1(untracked->exclude_sha1))) {1160 /*1161 * dir->basebuf gets reused by the traversal, but we1162 * need fname to remain unchanged to ensure the src1163 * member of each struct exclude correctly1164 * back-references its source file. Other invocations1165 * of add_exclude_list provide stable strings, so we1166 * strbuf_detach() and free() here in the caller.1167 */1168 struct strbuf sb = STRBUF_INIT;1169 strbuf_addbuf(&sb, &dir->basebuf);1170 strbuf_addstr(&sb, dir->exclude_per_dir);1171 el->src = strbuf_detach(&sb, NULL);1172 add_excludes(el->src, el->src, stk->baselen, el, istate,1173 untracked ? &sha1_stat : NULL);1174 }1175 /*1176 * NEEDSWORK: when untracked cache is enabled, prep_exclude()1177 * will first be called in valid_cached_dir() then maybe many1178 * times more in last_exclude_matching(). When the cache is1179 * used, last_exclude_matching() will not be called and1180 * reading .gitignore content will be a waste.1181 *1182 * So when it's called by valid_cached_dir() and we can get1183 * .gitignore SHA-1 from the index (i.e. .gitignore is not1184 * modified on work tree), we could delay reading the1185 * .gitignore content until we absolutely need it in1186 * last_exclude_matching(). Be careful about ignore rule1187 * order, though, if you do that.1188 */1189 if (untracked &&1190 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {1191 invalidate_gitignore(dir->untracked, untracked);1192 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);1193 }1194 dir->exclude_stack = stk;1195 current = stk->baselen;1196 }1197 strbuf_setlen(&dir->basebuf, baselen);1198}11991200/*1201 * Loads the exclude lists for the directory containing pathname, then1202 * scans all exclude lists to determine whether pathname is excluded.1203 * Returns the exclude_list element which matched, or NULL for1204 * undecided.1205 */1206struct exclude *last_exclude_matching(struct dir_struct *dir,1207 struct index_state *istate,1208 const char *pathname,1209 int *dtype_p)1210{1211 int pathlen = strlen(pathname);1212 const char *basename = strrchr(pathname, '/');1213 basename = (basename) ? basename+1 : pathname;12141215 prep_exclude(dir, istate, pathname, basename-pathname);12161217 if (dir->exclude)1218 return dir->exclude;12191220 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,1221 basename, dtype_p);1222}12231224/*1225 * Loads the exclude lists for the directory containing pathname, then1226 * scans all exclude lists to determine whether pathname is excluded.1227 * Returns 1 if true, otherwise 0.1228 */1229int is_excluded(struct dir_struct *dir, struct index_state *istate,1230 const char *pathname, int *dtype_p)1231{1232 struct exclude *exclude =1233 last_exclude_matching(dir, istate, pathname, dtype_p);1234 if (exclude)1235 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1236 return 0;1237}12381239static struct dir_entry *dir_entry_new(const char *pathname, int len)1240{1241 struct dir_entry *ent;12421243 FLEX_ALLOC_MEM(ent, name, pathname, len);1244 ent->len = len;1245 return ent;1246}12471248static struct dir_entry *dir_add_name(struct dir_struct *dir,1249 struct index_state *istate,1250 const char *pathname, int len)1251{1252 if (index_file_exists(istate, pathname, len, ignore_case))1253 return NULL;12541255 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);1256 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);1257}12581259struct dir_entry *dir_add_ignored(struct dir_struct *dir,1260 struct index_state *istate,1261 const char *pathname, int len)1262{1263 if (!index_name_is_other(istate, pathname, len))1264 return NULL;12651266 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);1267 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);1268}12691270enum exist_status {1271 index_nonexistent = 0,1272 index_directory,1273 index_gitdir1274};12751276/*1277 * Do not use the alphabetically sorted index to look up1278 * the directory name; instead, use the case insensitive1279 * directory hash.1280 */1281static enum exist_status directory_exists_in_index_icase(struct index_state *istate,1282 const char *dirname, int len)1283{1284 struct cache_entry *ce;12851286 if (index_dir_exists(istate, dirname, len))1287 return index_directory;12881289 ce = index_file_exists(istate, dirname, len, ignore_case);1290 if (ce && S_ISGITLINK(ce->ce_mode))1291 return index_gitdir;12921293 return index_nonexistent;1294}12951296/*1297 * The index sorts alphabetically by entry name, which1298 * means that a gitlink sorts as '\0' at the end, while1299 * a directory (which is defined not as an entry, but as1300 * the files it contains) will sort with the '/' at the1301 * end.1302 */1303static enum exist_status directory_exists_in_index(struct index_state *istate,1304 const char *dirname, int len)1305{1306 int pos;13071308 if (ignore_case)1309 return directory_exists_in_index_icase(istate, dirname, len);13101311 pos = index_name_pos(istate, dirname, len);1312 if (pos < 0)1313 pos = -pos-1;1314 while (pos < istate->cache_nr) {1315 const struct cache_entry *ce = istate->cache[pos++];1316 unsigned char endchar;13171318 if (strncmp(ce->name, dirname, len))1319 break;1320 endchar = ce->name[len];1321 if (endchar > '/')1322 break;1323 if (endchar == '/')1324 return index_directory;1325 if (!endchar && S_ISGITLINK(ce->ce_mode))1326 return index_gitdir;1327 }1328 return index_nonexistent;1329}13301331/*1332 * When we find a directory when traversing the filesystem, we1333 * have three distinct cases:1334 *1335 * - ignore it1336 * - see it as a directory1337 * - recurse into it1338 *1339 * and which one we choose depends on a combination of existing1340 * git index contents and the flags passed into the directory1341 * traversal routine.1342 *1343 * Case 1: If we *already* have entries in the index under that1344 * directory name, we always recurse into the directory to see1345 * all the files.1346 *1347 * Case 2: If we *already* have that directory name as a gitlink,1348 * we always continue to see it as a gitlink, regardless of whether1349 * there is an actual git directory there or not (it might not1350 * be checked out as a subproject!)1351 *1352 * Case 3: if we didn't have it in the index previously, we1353 * have a few sub-cases:1354 *1355 * (a) if "show_other_directories" is true, we show it as1356 * just a directory, unless "hide_empty_directories" is1357 * also true, in which case we need to check if it contains any1358 * untracked and / or ignored files.1359 * (b) if it looks like a git directory, and we don't have1360 * 'no_gitlinks' set we treat it as a gitlink, and show it1361 * as a directory.1362 * (c) otherwise, we recurse into it.1363 */1364static enum path_treatment treat_directory(struct dir_struct *dir,1365 struct untracked_cache_dir *untracked,1366 const char *dirname, int len, int baselen, int exclude,1367 const struct pathspec *pathspec)1368{1369 /* The "len-1" is to strip the final '/' */1370 switch (directory_exists_in_index(&the_index, dirname, len-1)) {1371 case index_directory:1372 return path_recurse;13731374 case index_gitdir:1375 return path_none;13761377 case index_nonexistent:1378 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)1379 break;1380 if (!(dir->flags & DIR_NO_GITLINKS)) {1381 unsigned char sha1[20];1382 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)1383 return path_untracked;1384 }1385 return path_recurse;1386 }13871388 /* This is the "show_other_directories" case */13891390 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))1391 return exclude ? path_excluded : path_untracked;13921393 untracked = lookup_untracked(dir->untracked, untracked,1394 dirname + baselen, len - baselen);1395 return read_directory_recursive(dir, dirname, len,1396 untracked, 1, pathspec);1397}13981399/*1400 * This is an inexact early pruning of any recursive directory1401 * reading - if the path cannot possibly be in the pathspec,1402 * return true, and we'll skip it early.1403 */1404static int simplify_away(const char *path, int pathlen,1405 const struct pathspec *pathspec)1406{1407 int i;14081409 if (!pathspec || !pathspec->nr)1410 return 0;14111412 GUARD_PATHSPEC(pathspec,1413 PATHSPEC_FROMTOP |1414 PATHSPEC_MAXDEPTH |1415 PATHSPEC_LITERAL |1416 PATHSPEC_GLOB |1417 PATHSPEC_ICASE |1418 PATHSPEC_EXCLUDE |1419 PATHSPEC_ATTR);14201421 for (i = 0; i < pathspec->nr; i++) {1422 const struct pathspec_item *item = &pathspec->items[i];1423 int len = item->nowildcard_len;14241425 if (len > pathlen)1426 len = pathlen;1427 if (!ps_strncmp(item, item->match, path, len))1428 return 0;1429 }14301431 return 1;1432}14331434/*1435 * This function tells us whether an excluded path matches a1436 * list of "interesting" pathspecs. That is, whether a path matched1437 * by any of the pathspecs could possibly be ignored by excluding1438 * the specified path. This can happen if:1439 *1440 * 1. the path is mentioned explicitly in the pathspec1441 *1442 * 2. the path is a directory prefix of some element in the1443 * pathspec1444 */1445static int exclude_matches_pathspec(const char *path, int pathlen,1446 const struct pathspec *pathspec)1447{1448 int i;14491450 if (!pathspec || !pathspec->nr)1451 return 0;14521453 GUARD_PATHSPEC(pathspec,1454 PATHSPEC_FROMTOP |1455 PATHSPEC_MAXDEPTH |1456 PATHSPEC_LITERAL |1457 PATHSPEC_GLOB |1458 PATHSPEC_ICASE |1459 PATHSPEC_EXCLUDE);14601461 for (i = 0; i < pathspec->nr; i++) {1462 const struct pathspec_item *item = &pathspec->items[i];1463 int len = item->nowildcard_len;14641465 if (len == pathlen &&1466 !ps_strncmp(item, item->match, path, pathlen))1467 return 1;1468 if (len > pathlen &&1469 item->match[pathlen] == '/' &&1470 !ps_strncmp(item, item->match, path, pathlen))1471 return 1;1472 }1473 return 0;1474}14751476static int get_index_dtype(struct index_state *istate,1477 const char *path, int len)1478{1479 int pos;1480 const struct cache_entry *ce;14811482 ce = index_file_exists(istate, path, len, 0);1483 if (ce) {1484 if (!ce_uptodate(ce))1485 return DT_UNKNOWN;1486 if (S_ISGITLINK(ce->ce_mode))1487 return DT_DIR;1488 /*1489 * Nobody actually cares about the1490 * difference between DT_LNK and DT_REG1491 */1492 return DT_REG;1493 }14941495 /* Try to look it up as a directory */1496 pos = index_name_pos(istate, path, len);1497 if (pos >= 0)1498 return DT_UNKNOWN;1499 pos = -pos-1;1500 while (pos < istate->cache_nr) {1501 ce = istate->cache[pos++];1502 if (strncmp(ce->name, path, len))1503 break;1504 if (ce->name[len] > '/')1505 break;1506 if (ce->name[len] < '/')1507 continue;1508 if (!ce_uptodate(ce))1509 break; /* continue? */1510 return DT_DIR;1511 }1512 return DT_UNKNOWN;1513}15141515static int get_dtype(struct dirent *de, struct index_state *istate,1516 const char *path, int len)1517{1518 int dtype = de ? DTYPE(de) : DT_UNKNOWN;1519 struct stat st;15201521 if (dtype != DT_UNKNOWN)1522 return dtype;1523 dtype = get_index_dtype(istate, path, len);1524 if (dtype != DT_UNKNOWN)1525 return dtype;1526 if (lstat(path, &st))1527 return dtype;1528 if (S_ISREG(st.st_mode))1529 return DT_REG;1530 if (S_ISDIR(st.st_mode))1531 return DT_DIR;1532 if (S_ISLNK(st.st_mode))1533 return DT_LNK;1534 return dtype;1535}15361537static enum path_treatment treat_one_path(struct dir_struct *dir,1538 struct untracked_cache_dir *untracked,1539 struct strbuf *path,1540 int baselen,1541 const struct pathspec *pathspec,1542 int dtype, struct dirent *de)1543{1544 int exclude;1545 int has_path_in_index = !!index_file_exists(&the_index, path->buf, path->len, ignore_case);15461547 if (dtype == DT_UNKNOWN)1548 dtype = get_dtype(de, &the_index, path->buf, path->len);15491550 /* Always exclude indexed files */1551 if (dtype != DT_DIR && has_path_in_index)1552 return path_none;15531554 /*1555 * When we are looking at a directory P in the working tree,1556 * there are three cases:1557 *1558 * (1) P exists in the index. Everything inside the directory P in1559 * the working tree needs to go when P is checked out from the1560 * index.1561 *1562 * (2) P does not exist in the index, but there is P/Q in the index.1563 * We know P will stay a directory when we check out the contents1564 * of the index, but we do not know yet if there is a directory1565 * P/Q in the working tree to be killed, so we need to recurse.1566 *1567 * (3) P does not exist in the index, and there is no P/Q in the index1568 * to require P to be a directory, either. Only in this case, we1569 * know that everything inside P will not be killed without1570 * recursing.1571 */1572 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&1573 (dtype == DT_DIR) &&1574 !has_path_in_index &&1575 (directory_exists_in_index(&the_index, path->buf, path->len) == index_nonexistent))1576 return path_none;15771578 exclude = is_excluded(dir, &the_index, path->buf, &dtype);15791580 /*1581 * Excluded? If we don't explicitly want to show1582 * ignored files, ignore it1583 */1584 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))1585 return path_excluded;15861587 switch (dtype) {1588 default:1589 return path_none;1590 case DT_DIR:1591 strbuf_addch(path, '/');1592 return treat_directory(dir, untracked, path->buf, path->len,1593 baselen, exclude, pathspec);1594 case DT_REG:1595 case DT_LNK:1596 return exclude ? path_excluded : path_untracked;1597 }1598}15991600static enum path_treatment treat_path_fast(struct dir_struct *dir,1601 struct untracked_cache_dir *untracked,1602 struct cached_dir *cdir,1603 struct strbuf *path,1604 int baselen,1605 const struct pathspec *pathspec)1606{1607 strbuf_setlen(path, baselen);1608 if (!cdir->ucd) {1609 strbuf_addstr(path, cdir->file);1610 return path_untracked;1611 }1612 strbuf_addstr(path, cdir->ucd->name);1613 /* treat_one_path() does this before it calls treat_directory() */1614 strbuf_complete(path, '/');1615 if (cdir->ucd->check_only)1616 /*1617 * check_only is set as a result of treat_directory() getting1618 * to its bottom. Verify again the same set of directories1619 * with check_only set.1620 */1621 return read_directory_recursive(dir, path->buf, path->len,1622 cdir->ucd, 1, pathspec);1623 /*1624 * We get path_recurse in the first run when1625 * directory_exists_in_index() returns index_nonexistent. We1626 * are sure that new changes in the index does not impact the1627 * outcome. Return now.1628 */1629 return path_recurse;1630}16311632static enum path_treatment treat_path(struct dir_struct *dir,1633 struct untracked_cache_dir *untracked,1634 struct cached_dir *cdir,1635 struct strbuf *path,1636 int baselen,1637 const struct pathspec *pathspec)1638{1639 int dtype;1640 struct dirent *de = cdir->de;16411642 if (!de)1643 return treat_path_fast(dir, untracked, cdir, path,1644 baselen, pathspec);1645 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))1646 return path_none;1647 strbuf_setlen(path, baselen);1648 strbuf_addstr(path, de->d_name);1649 if (simplify_away(path->buf, path->len, pathspec))1650 return path_none;16511652 dtype = DTYPE(de);1653 return treat_one_path(dir, untracked, path, baselen, pathspec, dtype, de);1654}16551656static void add_untracked(struct untracked_cache_dir *dir, const char *name)1657{1658 if (!dir)1659 return;1660 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,1661 dir->untracked_alloc);1662 dir->untracked[dir->untracked_nr++] = xstrdup(name);1663}16641665static int valid_cached_dir(struct dir_struct *dir,1666 struct untracked_cache_dir *untracked,1667 struct index_state *istate,1668 struct strbuf *path,1669 int check_only)1670{1671 struct stat st;16721673 if (!untracked)1674 return 0;16751676 if (stat(path->len ? path->buf : ".", &st)) {1677 invalidate_directory(dir->untracked, untracked);1678 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));1679 return 0;1680 }1681 if (!untracked->valid ||1682 match_stat_data_racy(istate, &untracked->stat_data, &st)) {1683 if (untracked->valid)1684 invalidate_directory(dir->untracked, untracked);1685 fill_stat_data(&untracked->stat_data, &st);1686 return 0;1687 }16881689 if (untracked->check_only != !!check_only) {1690 invalidate_directory(dir->untracked, untracked);1691 return 0;1692 }16931694 /*1695 * prep_exclude will be called eventually on this directory,1696 * but it's called much later in last_exclude_matching(). We1697 * need it now to determine the validity of the cache for this1698 * path. The next calls will be nearly no-op, the way1699 * prep_exclude() is designed.1700 */1701 if (path->len && path->buf[path->len - 1] != '/') {1702 strbuf_addch(path, '/');1703 prep_exclude(dir, istate, path->buf, path->len);1704 strbuf_setlen(path, path->len - 1);1705 } else1706 prep_exclude(dir, istate, path->buf, path->len);17071708 /* hopefully prep_exclude() haven't invalidated this entry... */1709 return untracked->valid;1710}17111712static int open_cached_dir(struct cached_dir *cdir,1713 struct dir_struct *dir,1714 struct untracked_cache_dir *untracked,1715 struct index_state *istate,1716 struct strbuf *path,1717 int check_only)1718{1719 memset(cdir, 0, sizeof(*cdir));1720 cdir->untracked = untracked;1721 if (valid_cached_dir(dir, untracked, istate, path, check_only))1722 return 0;1723 cdir->fdir = opendir(path->len ? path->buf : ".");1724 if (dir->untracked)1725 dir->untracked->dir_opened++;1726 if (!cdir->fdir)1727 return -1;1728 return 0;1729}17301731static int read_cached_dir(struct cached_dir *cdir)1732{1733 if (cdir->fdir) {1734 cdir->de = readdir(cdir->fdir);1735 if (!cdir->de)1736 return -1;1737 return 0;1738 }1739 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {1740 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];1741 if (!d->recurse) {1742 cdir->nr_dirs++;1743 continue;1744 }1745 cdir->ucd = d;1746 cdir->nr_dirs++;1747 return 0;1748 }1749 cdir->ucd = NULL;1750 if (cdir->nr_files < cdir->untracked->untracked_nr) {1751 struct untracked_cache_dir *d = cdir->untracked;1752 cdir->file = d->untracked[cdir->nr_files++];1753 return 0;1754 }1755 return -1;1756}17571758static void close_cached_dir(struct cached_dir *cdir)1759{1760 if (cdir->fdir)1761 closedir(cdir->fdir);1762 /*1763 * We have gone through this directory and found no untracked1764 * entries. Mark it valid.1765 */1766 if (cdir->untracked) {1767 cdir->untracked->valid = 1;1768 cdir->untracked->recurse = 1;1769 }1770}17711772/*1773 * Read a directory tree. We currently ignore anything but1774 * directories, regular files and symlinks. That's because git1775 * doesn't handle them at all yet. Maybe that will change some1776 * day.1777 *1778 * Also, we ignore the name ".git" (even if it is not a directory).1779 * That likely will not change.1780 *1781 * Returns the most significant path_treatment value encountered in the scan.1782 */1783static enum path_treatment read_directory_recursive(struct dir_struct *dir,1784 const char *base, int baselen,1785 struct untracked_cache_dir *untracked, int check_only,1786 const struct pathspec *pathspec)1787{1788 struct cached_dir cdir;1789 enum path_treatment state, subdir_state, dir_state = path_none;1790 struct strbuf path = STRBUF_INIT;17911792 strbuf_add(&path, base, baselen);17931794 if (open_cached_dir(&cdir, dir, untracked, &the_index, &path, check_only))1795 goto out;17961797 if (untracked)1798 untracked->check_only = !!check_only;17991800 while (!read_cached_dir(&cdir)) {1801 /* check how the file or directory should be treated */1802 state = treat_path(dir, untracked, &cdir, &path,1803 baselen, pathspec);18041805 if (state > dir_state)1806 dir_state = state;18071808 /* recurse into subdir if instructed by treat_path */1809 if (state == path_recurse) {1810 struct untracked_cache_dir *ud;1811 ud = lookup_untracked(dir->untracked, untracked,1812 path.buf + baselen,1813 path.len - baselen);1814 subdir_state =1815 read_directory_recursive(dir, path.buf,1816 path.len, ud,1817 check_only, pathspec);1818 if (subdir_state > dir_state)1819 dir_state = subdir_state;1820 }18211822 if (check_only) {1823 /* abort early if maximum state has been reached */1824 if (dir_state == path_untracked) {1825 if (cdir.fdir)1826 add_untracked(untracked, path.buf + baselen);1827 break;1828 }1829 /* skip the dir_add_* part */1830 continue;1831 }18321833 /* add the path to the appropriate result list */1834 switch (state) {1835 case path_excluded:1836 if (dir->flags & DIR_SHOW_IGNORED)1837 dir_add_name(dir, &the_index, path.buf, path.len);1838 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||1839 ((dir->flags & DIR_COLLECT_IGNORED) &&1840 exclude_matches_pathspec(path.buf, path.len,1841 pathspec)))1842 dir_add_ignored(dir, &the_index, path.buf, path.len);1843 break;18441845 case path_untracked:1846 if (dir->flags & DIR_SHOW_IGNORED)1847 break;1848 dir_add_name(dir, &the_index, path.buf, path.len);1849 if (cdir.fdir)1850 add_untracked(untracked, path.buf + baselen);1851 break;18521853 default:1854 break;1855 }1856 }1857 close_cached_dir(&cdir);1858 out:1859 strbuf_release(&path);18601861 return dir_state;1862}18631864static int cmp_name(const void *p1, const void *p2)1865{1866 const struct dir_entry *e1 = *(const struct dir_entry **)p1;1867 const struct dir_entry *e2 = *(const struct dir_entry **)p2;18681869 return name_compare(e1->name, e1->len, e2->name, e2->len);1870}18711872static int treat_leading_path(struct dir_struct *dir,1873 const char *path, int len,1874 const struct pathspec *pathspec)1875{1876 struct strbuf sb = STRBUF_INIT;1877 int baselen, rc = 0;1878 const char *cp;1879 int old_flags = dir->flags;18801881 while (len && path[len - 1] == '/')1882 len--;1883 if (!len)1884 return 1;1885 baselen = 0;1886 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;1887 while (1) {1888 cp = path + baselen + !!baselen;1889 cp = memchr(cp, '/', path + len - cp);1890 if (!cp)1891 baselen = len;1892 else1893 baselen = cp - path;1894 strbuf_setlen(&sb, 0);1895 strbuf_add(&sb, path, baselen);1896 if (!is_directory(sb.buf))1897 break;1898 if (simplify_away(sb.buf, sb.len, pathspec))1899 break;1900 if (treat_one_path(dir, NULL, &sb, baselen, pathspec,1901 DT_DIR, NULL) == path_none)1902 break; /* do not recurse into it */1903 if (len <= baselen) {1904 rc = 1;1905 break; /* finished checking */1906 }1907 }1908 strbuf_release(&sb);1909 dir->flags = old_flags;1910 return rc;1911}19121913static const char *get_ident_string(void)1914{1915 static struct strbuf sb = STRBUF_INIT;1916 struct utsname uts;19171918 if (sb.len)1919 return sb.buf;1920 if (uname(&uts) < 0)1921 die_errno(_("failed to get kernel name and information"));1922 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),1923 uts.sysname);1924 return sb.buf;1925}19261927static int ident_in_untracked(const struct untracked_cache *uc)1928{1929 /*1930 * Previous git versions may have saved many NUL separated1931 * strings in the "ident" field, but it is insane to manage1932 * many locations, so just take care of the first one.1933 */19341935 return !strcmp(uc->ident.buf, get_ident_string());1936}19371938static void set_untracked_ident(struct untracked_cache *uc)1939{1940 strbuf_reset(&uc->ident);1941 strbuf_addstr(&uc->ident, get_ident_string());19421943 /*1944 * This strbuf used to contain a list of NUL separated1945 * strings, so save NUL too for backward compatibility.1946 */1947 strbuf_addch(&uc->ident, 0);1948}19491950static void new_untracked_cache(struct index_state *istate)1951{1952 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));1953 strbuf_init(&uc->ident, 100);1954 uc->exclude_per_dir = ".gitignore";1955 /* should be the same flags used by git-status */1956 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;1957 set_untracked_ident(uc);1958 istate->untracked = uc;1959 istate->cache_changed |= UNTRACKED_CHANGED;1960}19611962void add_untracked_cache(struct index_state *istate)1963{1964 if (!istate->untracked) {1965 new_untracked_cache(istate);1966 } else {1967 if (!ident_in_untracked(istate->untracked)) {1968 free_untracked_cache(istate->untracked);1969 new_untracked_cache(istate);1970 }1971 }1972}19731974void remove_untracked_cache(struct index_state *istate)1975{1976 if (istate->untracked) {1977 free_untracked_cache(istate->untracked);1978 istate->untracked = NULL;1979 istate->cache_changed |= UNTRACKED_CHANGED;1980 }1981}19821983static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,1984 int base_len,1985 const struct pathspec *pathspec)1986{1987 struct untracked_cache_dir *root;19881989 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))1990 return NULL;19911992 /*1993 * We only support $GIT_DIR/info/exclude and core.excludesfile1994 * as the global ignore rule files. Any other additions1995 * (e.g. from command line) invalidate the cache. This1996 * condition also catches running setup_standard_excludes()1997 * before setting dir->untracked!1998 */1999 if (dir->unmanaged_exclude_files)2000 return NULL;20012002 /*2003 * Optimize for the main use case only: whole-tree git2004 * status. More work involved in treat_leading_path() if we2005 * use cache on just a subset of the worktree. pathspec2006 * support could make the matter even worse.2007 */2008 if (base_len || (pathspec && pathspec->nr))2009 return NULL;20102011 /* Different set of flags may produce different results */2012 if (dir->flags != dir->untracked->dir_flags ||2013 /*2014 * See treat_directory(), case index_nonexistent. Without2015 * this flag, we may need to also cache .git file content2016 * for the resolve_gitlink_ref() call, which we don't.2017 */2018 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||2019 /* We don't support collecting ignore files */2020 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |2021 DIR_COLLECT_IGNORED)))2022 return NULL;20232024 /*2025 * If we use .gitignore in the cache and now you change it to2026 * .gitexclude, everything will go wrong.2027 */2028 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&2029 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))2030 return NULL;20312032 /*2033 * EXC_CMDL is not considered in the cache. If people set it,2034 * skip the cache.2035 */2036 if (dir->exclude_list_group[EXC_CMDL].nr)2037 return NULL;20382039 if (!ident_in_untracked(dir->untracked)) {2040 warning(_("Untracked cache is disabled on this system or location."));2041 return NULL;2042 }20432044 if (!dir->untracked->root) {2045 const int len = sizeof(*dir->untracked->root);2046 dir->untracked->root = xmalloc(len);2047 memset(dir->untracked->root, 0, len);2048 }20492050 /* Validate $GIT_DIR/info/exclude and core.excludesfile */2051 root = dir->untracked->root;2052 if (hashcmp(dir->ss_info_exclude.sha1,2053 dir->untracked->ss_info_exclude.sha1)) {2054 invalidate_gitignore(dir->untracked, root);2055 dir->untracked->ss_info_exclude = dir->ss_info_exclude;2056 }2057 if (hashcmp(dir->ss_excludes_file.sha1,2058 dir->untracked->ss_excludes_file.sha1)) {2059 invalidate_gitignore(dir->untracked, root);2060 dir->untracked->ss_excludes_file = dir->ss_excludes_file;2061 }20622063 /* Make sure this directory is not dropped out at saving phase */2064 root->recurse = 1;2065 return root;2066}20672068int read_directory(struct dir_struct *dir, const char *path,2069 int len, const struct pathspec *pathspec)2070{2071 struct untracked_cache_dir *untracked;20722073 if (has_symlink_leading_path(path, len))2074 return dir->nr;20752076 untracked = validate_untracked_cache(dir, len, pathspec);2077 if (!untracked)2078 /*2079 * make sure untracked cache code path is disabled,2080 * e.g. prep_exclude()2081 */2082 dir->untracked = NULL;2083 if (!len || treat_leading_path(dir, path, len, pathspec))2084 read_directory_recursive(dir, path, len, untracked, 0, pathspec);2085 QSORT(dir->entries, dir->nr, cmp_name);2086 QSORT(dir->ignored, dir->ignored_nr, cmp_name);2087 if (dir->untracked) {2088 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);2089 trace_printf_key(&trace_untracked_stats,2090 "node creation: %u\n"2091 "gitignore invalidation: %u\n"2092 "directory invalidation: %u\n"2093 "opendir: %u\n",2094 dir->untracked->dir_created,2095 dir->untracked->gitignore_invalidated,2096 dir->untracked->dir_invalidated,2097 dir->untracked->dir_opened);2098 if (dir->untracked == the_index.untracked &&2099 (dir->untracked->dir_opened ||2100 dir->untracked->gitignore_invalidated ||2101 dir->untracked->dir_invalidated))2102 the_index.cache_changed |= UNTRACKED_CHANGED;2103 if (dir->untracked != the_index.untracked) {2104 free(dir->untracked);2105 dir->untracked = NULL;2106 }2107 }2108 return dir->nr;2109}21102111int file_exists(const char *f)2112{2113 struct stat sb;2114 return lstat(f, &sb) == 0;2115}21162117static int cmp_icase(char a, char b)2118{2119 if (a == b)2120 return 0;2121 if (ignore_case)2122 return toupper(a) - toupper(b);2123 return a - b;2124}21252126/*2127 * Given two normalized paths (a trailing slash is ok), if subdir is2128 * outside dir, return -1. Otherwise return the offset in subdir that2129 * can be used as relative path to dir.2130 */2131int dir_inside_of(const char *subdir, const char *dir)2132{2133 int offset = 0;21342135 assert(dir && subdir && *dir && *subdir);21362137 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {2138 dir++;2139 subdir++;2140 offset++;2141 }21422143 /* hel[p]/me vs hel[l]/yeah */2144 if (*dir && *subdir)2145 return -1;21462147 if (!*subdir)2148 return !*dir ? offset : -1; /* same dir */21492150 /* foo/[b]ar vs foo/[] */2151 if (is_dir_sep(dir[-1]))2152 return is_dir_sep(subdir[-1]) ? offset : -1;21532154 /* foo[/]bar vs foo[] */2155 return is_dir_sep(*subdir) ? offset + 1 : -1;2156}21572158int is_inside_dir(const char *dir)2159{2160 char *cwd;2161 int rc;21622163 if (!dir)2164 return 0;21652166 cwd = xgetcwd();2167 rc = (dir_inside_of(cwd, dir) >= 0);2168 free(cwd);2169 return rc;2170}21712172int is_empty_dir(const char *path)2173{2174 DIR *dir = opendir(path);2175 struct dirent *e;2176 int ret = 1;21772178 if (!dir)2179 return 0;21802181 while ((e = readdir(dir)) != NULL)2182 if (!is_dot_or_dotdot(e->d_name)) {2183 ret = 0;2184 break;2185 }21862187 closedir(dir);2188 return ret;2189}21902191static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)2192{2193 DIR *dir;2194 struct dirent *e;2195 int ret = 0, original_len = path->len, len, kept_down = 0;2196 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);2197 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);2198 unsigned char submodule_head[20];21992200 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&2201 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {2202 /* Do not descend and nuke a nested git work tree. */2203 if (kept_up)2204 *kept_up = 1;2205 return 0;2206 }22072208 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;2209 dir = opendir(path->buf);2210 if (!dir) {2211 if (errno == ENOENT)2212 return keep_toplevel ? -1 : 0;2213 else if (errno == EACCES && !keep_toplevel)2214 /*2215 * An empty dir could be removable even if it2216 * is unreadable:2217 */2218 return rmdir(path->buf);2219 else2220 return -1;2221 }2222 strbuf_complete(path, '/');22232224 len = path->len;2225 while ((e = readdir(dir)) != NULL) {2226 struct stat st;2227 if (is_dot_or_dotdot(e->d_name))2228 continue;22292230 strbuf_setlen(path, len);2231 strbuf_addstr(path, e->d_name);2232 if (lstat(path->buf, &st)) {2233 if (errno == ENOENT)2234 /*2235 * file disappeared, which is what we2236 * wanted anyway2237 */2238 continue;2239 /* fall thru */2240 } else if (S_ISDIR(st.st_mode)) {2241 if (!remove_dir_recurse(path, flag, &kept_down))2242 continue; /* happy */2243 } else if (!only_empty &&2244 (!unlink(path->buf) || errno == ENOENT)) {2245 continue; /* happy, too */2246 }22472248 /* path too long, stat fails, or non-directory still exists */2249 ret = -1;2250 break;2251 }2252 closedir(dir);22532254 strbuf_setlen(path, original_len);2255 if (!ret && !keep_toplevel && !kept_down)2256 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;2257 else if (kept_up)2258 /*2259 * report the uplevel that it is not an error that we2260 * did not rmdir() our directory.2261 */2262 *kept_up = !ret;2263 return ret;2264}22652266int remove_dir_recursively(struct strbuf *path, int flag)2267{2268 return remove_dir_recurse(path, flag, NULL);2269}22702271static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")22722273void setup_standard_excludes(struct dir_struct *dir)2274{2275 dir->exclude_per_dir = ".gitignore";22762277 /* core.excludefile defaulting to $XDG_HOME/git/ignore */2278 if (!excludes_file)2279 excludes_file = xdg_config_home("ignore");2280 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))2281 add_excludes_from_file_1(dir, excludes_file,2282 dir->untracked ? &dir->ss_excludes_file : NULL);22832284 /* per repository user preference */2285 if (startup_info->have_repository) {2286 const char *path = git_path_info_exclude();2287 if (!access_or_warn(path, R_OK, 0))2288 add_excludes_from_file_1(dir, path,2289 dir->untracked ? &dir->ss_info_exclude : NULL);2290 }2291}22922293int remove_path(const char *name)2294{2295 char *slash;22962297 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)2298 return -1;22992300 slash = strrchr(name, '/');2301 if (slash) {2302 char *dirs = xstrdup(name);2303 slash = dirs + (slash - name);2304 do {2305 *slash = '\0';2306 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));2307 free(dirs);2308 }2309 return 0;2310}23112312/*2313 * Frees memory within dir which was allocated for exclude lists and2314 * the exclude_stack. Does not free dir itself.2315 */2316void clear_directory(struct dir_struct *dir)2317{2318 int i, j;2319 struct exclude_list_group *group;2320 struct exclude_list *el;2321 struct exclude_stack *stk;23222323 for (i = EXC_CMDL; i <= EXC_FILE; i++) {2324 group = &dir->exclude_list_group[i];2325 for (j = 0; j < group->nr; j++) {2326 el = &group->el[j];2327 if (i == EXC_DIRS)2328 free((char *)el->src);2329 clear_exclude_list(el);2330 }2331 free(group->el);2332 }23332334 stk = dir->exclude_stack;2335 while (stk) {2336 struct exclude_stack *prev = stk->prev;2337 free(stk);2338 stk = prev;2339 }2340 strbuf_release(&dir->basebuf);2341}23422343struct ondisk_untracked_cache {2344 struct stat_data info_exclude_stat;2345 struct stat_data excludes_file_stat;2346 uint32_t dir_flags;2347 unsigned char info_exclude_sha1[20];2348 unsigned char excludes_file_sha1[20];2349 char exclude_per_dir[FLEX_ARRAY];2350};23512352#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)23532354struct write_data {2355 int index; /* number of written untracked_cache_dir */2356 struct ewah_bitmap *check_only; /* from untracked_cache_dir */2357 struct ewah_bitmap *valid; /* from untracked_cache_dir */2358 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */2359 struct strbuf out;2360 struct strbuf sb_stat;2361 struct strbuf sb_sha1;2362};23632364static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)2365{2366 to->sd_ctime.sec = htonl(from->sd_ctime.sec);2367 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);2368 to->sd_mtime.sec = htonl(from->sd_mtime.sec);2369 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);2370 to->sd_dev = htonl(from->sd_dev);2371 to->sd_ino = htonl(from->sd_ino);2372 to->sd_uid = htonl(from->sd_uid);2373 to->sd_gid = htonl(from->sd_gid);2374 to->sd_size = htonl(from->sd_size);2375}23762377static void write_one_dir(struct untracked_cache_dir *untracked,2378 struct write_data *wd)2379{2380 struct stat_data stat_data;2381 struct strbuf *out = &wd->out;2382 unsigned char intbuf[16];2383 unsigned int intlen, value;2384 int i = wd->index++;23852386 /*2387 * untracked_nr should be reset whenever valid is clear, but2388 * for safety..2389 */2390 if (!untracked->valid) {2391 untracked->untracked_nr = 0;2392 untracked->check_only = 0;2393 }23942395 if (untracked->check_only)2396 ewah_set(wd->check_only, i);2397 if (untracked->valid) {2398 ewah_set(wd->valid, i);2399 stat_data_to_disk(&stat_data, &untracked->stat_data);2400 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));2401 }2402 if (!is_null_sha1(untracked->exclude_sha1)) {2403 ewah_set(wd->sha1_valid, i);2404 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);2405 }24062407 intlen = encode_varint(untracked->untracked_nr, intbuf);2408 strbuf_add(out, intbuf, intlen);24092410 /* skip non-recurse directories */2411 for (i = 0, value = 0; i < untracked->dirs_nr; i++)2412 if (untracked->dirs[i]->recurse)2413 value++;2414 intlen = encode_varint(value, intbuf);2415 strbuf_add(out, intbuf, intlen);24162417 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);24182419 for (i = 0; i < untracked->untracked_nr; i++)2420 strbuf_add(out, untracked->untracked[i],2421 strlen(untracked->untracked[i]) + 1);24222423 for (i = 0; i < untracked->dirs_nr; i++)2424 if (untracked->dirs[i]->recurse)2425 write_one_dir(untracked->dirs[i], wd);2426}24272428void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)2429{2430 struct ondisk_untracked_cache *ouc;2431 struct write_data wd;2432 unsigned char varbuf[16];2433 int varint_len;2434 size_t len = strlen(untracked->exclude_per_dir);24352436 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);2437 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);2438 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);2439 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);2440 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);2441 ouc->dir_flags = htonl(untracked->dir_flags);24422443 varint_len = encode_varint(untracked->ident.len, varbuf);2444 strbuf_add(out, varbuf, varint_len);2445 strbuf_addbuf(out, &untracked->ident);24462447 strbuf_add(out, ouc, ouc_size(len));2448 free(ouc);2449 ouc = NULL;24502451 if (!untracked->root) {2452 varint_len = encode_varint(0, varbuf);2453 strbuf_add(out, varbuf, varint_len);2454 return;2455 }24562457 wd.index = 0;2458 wd.check_only = ewah_new();2459 wd.valid = ewah_new();2460 wd.sha1_valid = ewah_new();2461 strbuf_init(&wd.out, 1024);2462 strbuf_init(&wd.sb_stat, 1024);2463 strbuf_init(&wd.sb_sha1, 1024);2464 write_one_dir(untracked->root, &wd);24652466 varint_len = encode_varint(wd.index, varbuf);2467 strbuf_add(out, varbuf, varint_len);2468 strbuf_addbuf(out, &wd.out);2469 ewah_serialize_strbuf(wd.valid, out);2470 ewah_serialize_strbuf(wd.check_only, out);2471 ewah_serialize_strbuf(wd.sha1_valid, out);2472 strbuf_addbuf(out, &wd.sb_stat);2473 strbuf_addbuf(out, &wd.sb_sha1);2474 strbuf_addch(out, '\0'); /* safe guard for string lists */24752476 ewah_free(wd.valid);2477 ewah_free(wd.check_only);2478 ewah_free(wd.sha1_valid);2479 strbuf_release(&wd.out);2480 strbuf_release(&wd.sb_stat);2481 strbuf_release(&wd.sb_sha1);2482}24832484static void free_untracked(struct untracked_cache_dir *ucd)2485{2486 int i;2487 if (!ucd)2488 return;2489 for (i = 0; i < ucd->dirs_nr; i++)2490 free_untracked(ucd->dirs[i]);2491 for (i = 0; i < ucd->untracked_nr; i++)2492 free(ucd->untracked[i]);2493 free(ucd->untracked);2494 free(ucd->dirs);2495 free(ucd);2496}24972498void free_untracked_cache(struct untracked_cache *uc)2499{2500 if (uc)2501 free_untracked(uc->root);2502 free(uc);2503}25042505struct read_data {2506 int index;2507 struct untracked_cache_dir **ucd;2508 struct ewah_bitmap *check_only;2509 struct ewah_bitmap *valid;2510 struct ewah_bitmap *sha1_valid;2511 const unsigned char *data;2512 const unsigned char *end;2513};25142515static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)2516{2517 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);2518 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);2519 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);2520 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);2521 to->sd_dev = get_be32(&from->sd_dev);2522 to->sd_ino = get_be32(&from->sd_ino);2523 to->sd_uid = get_be32(&from->sd_uid);2524 to->sd_gid = get_be32(&from->sd_gid);2525 to->sd_size = get_be32(&from->sd_size);2526}25272528static int read_one_dir(struct untracked_cache_dir **untracked_,2529 struct read_data *rd)2530{2531 struct untracked_cache_dir ud, *untracked;2532 const unsigned char *next, *data = rd->data, *end = rd->end;2533 unsigned int value;2534 int i, len;25352536 memset(&ud, 0, sizeof(ud));25372538 next = data;2539 value = decode_varint(&next);2540 if (next > end)2541 return -1;2542 ud.recurse = 1;2543 ud.untracked_alloc = value;2544 ud.untracked_nr = value;2545 if (ud.untracked_nr)2546 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);2547 data = next;25482549 next = data;2550 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);2551 if (next > end)2552 return -1;2553 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);2554 data = next;25552556 len = strlen((const char *)data);2557 next = data + len + 1;2558 if (next > rd->end)2559 return -1;2560 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));2561 memcpy(untracked, &ud, sizeof(ud));2562 memcpy(untracked->name, data, len + 1);2563 data = next;25642565 for (i = 0; i < untracked->untracked_nr; i++) {2566 len = strlen((const char *)data);2567 next = data + len + 1;2568 if (next > rd->end)2569 return -1;2570 untracked->untracked[i] = xstrdup((const char*)data);2571 data = next;2572 }25732574 rd->ucd[rd->index++] = untracked;2575 rd->data = data;25762577 for (i = 0; i < untracked->dirs_nr; i++) {2578 len = read_one_dir(untracked->dirs + i, rd);2579 if (len < 0)2580 return -1;2581 }2582 return 0;2583}25842585static void set_check_only(size_t pos, void *cb)2586{2587 struct read_data *rd = cb;2588 struct untracked_cache_dir *ud = rd->ucd[pos];2589 ud->check_only = 1;2590}25912592static void read_stat(size_t pos, void *cb)2593{2594 struct read_data *rd = cb;2595 struct untracked_cache_dir *ud = rd->ucd[pos];2596 if (rd->data + sizeof(struct stat_data) > rd->end) {2597 rd->data = rd->end + 1;2598 return;2599 }2600 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);2601 rd->data += sizeof(struct stat_data);2602 ud->valid = 1;2603}26042605static void read_sha1(size_t pos, void *cb)2606{2607 struct read_data *rd = cb;2608 struct untracked_cache_dir *ud = rd->ucd[pos];2609 if (rd->data + 20 > rd->end) {2610 rd->data = rd->end + 1;2611 return;2612 }2613 hashcpy(ud->exclude_sha1, rd->data);2614 rd->data += 20;2615}26162617static void load_sha1_stat(struct sha1_stat *sha1_stat,2618 const struct stat_data *stat,2619 const unsigned char *sha1)2620{2621 stat_data_from_disk(&sha1_stat->stat, stat);2622 hashcpy(sha1_stat->sha1, sha1);2623 sha1_stat->valid = 1;2624}26252626struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)2627{2628 const struct ondisk_untracked_cache *ouc;2629 struct untracked_cache *uc;2630 struct read_data rd;2631 const unsigned char *next = data, *end = (const unsigned char *)data + sz;2632 const char *ident;2633 int ident_len, len;26342635 if (sz <= 1 || end[-1] != '\0')2636 return NULL;2637 end--;26382639 ident_len = decode_varint(&next);2640 if (next + ident_len > end)2641 return NULL;2642 ident = (const char *)next;2643 next += ident_len;26442645 ouc = (const struct ondisk_untracked_cache *)next;2646 if (next + ouc_size(0) > end)2647 return NULL;26482649 uc = xcalloc(1, sizeof(*uc));2650 strbuf_init(&uc->ident, ident_len);2651 strbuf_add(&uc->ident, ident, ident_len);2652 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,2653 ouc->info_exclude_sha1);2654 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,2655 ouc->excludes_file_sha1);2656 uc->dir_flags = get_be32(&ouc->dir_flags);2657 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);2658 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */2659 next += ouc_size(strlen(ouc->exclude_per_dir));2660 if (next >= end)2661 goto done2;26622663 len = decode_varint(&next);2664 if (next > end || len == 0)2665 goto done2;26662667 rd.valid = ewah_new();2668 rd.check_only = ewah_new();2669 rd.sha1_valid = ewah_new();2670 rd.data = next;2671 rd.end = end;2672 rd.index = 0;2673 ALLOC_ARRAY(rd.ucd, len);26742675 if (read_one_dir(&uc->root, &rd) || rd.index != len)2676 goto done;26772678 next = rd.data;2679 len = ewah_read_mmap(rd.valid, next, end - next);2680 if (len < 0)2681 goto done;26822683 next += len;2684 len = ewah_read_mmap(rd.check_only, next, end - next);2685 if (len < 0)2686 goto done;26872688 next += len;2689 len = ewah_read_mmap(rd.sha1_valid, next, end - next);2690 if (len < 0)2691 goto done;26922693 ewah_each_bit(rd.check_only, set_check_only, &rd);2694 rd.data = next + len;2695 ewah_each_bit(rd.valid, read_stat, &rd);2696 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);2697 next = rd.data;26982699done:2700 free(rd.ucd);2701 ewah_free(rd.valid);2702 ewah_free(rd.check_only);2703 ewah_free(rd.sha1_valid);2704done2:2705 if (next != end) {2706 free_untracked_cache(uc);2707 uc = NULL;2708 }2709 return uc;2710}27112712static void invalidate_one_directory(struct untracked_cache *uc,2713 struct untracked_cache_dir *ucd)2714{2715 uc->dir_invalidated++;2716 ucd->valid = 0;2717 ucd->untracked_nr = 0;2718}27192720/*2721 * Normally when an entry is added or removed from a directory,2722 * invalidating that directory is enough. No need to touch its2723 * ancestors. When a directory is shown as "foo/bar/" in git-status2724 * however, deleting or adding an entry may have cascading effect.2725 *2726 * Say the "foo/bar/file" has become untracked, we need to tell the2727 * untracked_cache_dir of "foo" that "bar/" is not an untracked2728 * directory any more (because "bar" is managed by foo as an untracked2729 * "file").2730 *2731 * Similarly, if "foo/bar/file" moves from untracked to tracked and it2732 * was the last untracked entry in the entire "foo", we should show2733 * "foo/" instead. Which means we have to invalidate past "bar" up to2734 * "foo".2735 *2736 * This function traverses all directories from root to leaf. If there2737 * is a chance of one of the above cases happening, we invalidate back2738 * to root. Otherwise we just invalidate the leaf. There may be a more2739 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to2740 * detect these cases and avoid unnecessary invalidation, for example,2741 * checking for the untracked entry named "bar/" in "foo", but for now2742 * stick to something safe and simple.2743 */2744static int invalidate_one_component(struct untracked_cache *uc,2745 struct untracked_cache_dir *dir,2746 const char *path, int len)2747{2748 const char *rest = strchr(path, '/');27492750 if (rest) {2751 int component_len = rest - path;2752 struct untracked_cache_dir *d =2753 lookup_untracked(uc, dir, path, component_len);2754 int ret =2755 invalidate_one_component(uc, d, rest + 1,2756 len - (component_len + 1));2757 if (ret)2758 invalidate_one_directory(uc, dir);2759 return ret;2760 }27612762 invalidate_one_directory(uc, dir);2763 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;2764}27652766void untracked_cache_invalidate_path(struct index_state *istate,2767 const char *path)2768{2769 if (!istate->untracked || !istate->untracked->root)2770 return;2771 invalidate_one_component(istate->untracked, istate->untracked->root,2772 path, strlen(path));2773}27742775void untracked_cache_remove_from_index(struct index_state *istate,2776 const char *path)2777{2778 untracked_cache_invalidate_path(istate, path);2779}27802781void untracked_cache_add_to_index(struct index_state *istate,2782 const char *path)2783{2784 untracked_cache_invalidate_path(istate, path);2785}27862787/* Update gitfile and core.worktree setting to connect work tree and git dir */2788void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)2789{2790 struct strbuf gitfile_sb = STRBUF_INIT;2791 struct strbuf cfg_sb = STRBUF_INIT;2792 struct strbuf rel_path = STRBUF_INIT;2793 char *git_dir, *work_tree;27942795 /* Prepare .git file */2796 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);2797 if (safe_create_leading_directories_const(gitfile_sb.buf))2798 die(_("could not create directories for %s"), gitfile_sb.buf);27992800 /* Prepare config file */2801 strbuf_addf(&cfg_sb, "%s/config", git_dir_);2802 if (safe_create_leading_directories_const(cfg_sb.buf))2803 die(_("could not create directories for %s"), cfg_sb.buf);28042805 git_dir = real_pathdup(git_dir_, 1);2806 work_tree = real_pathdup(work_tree_, 1);28072808 /* Write .git file */2809 write_file(gitfile_sb.buf, "gitdir: %s",2810 relative_path(git_dir, work_tree, &rel_path));2811 /* Update core.worktree setting */2812 git_config_set_in_file(cfg_sb.buf, "core.worktree",2813 relative_path(work_tree, git_dir, &rel_path));28142815 strbuf_release(&gitfile_sb);2816 strbuf_release(&cfg_sb);2817 strbuf_release(&rel_path);2818 free(work_tree);2819 free(git_dir);2820}28212822/*2823 * Migrate the git directory of the given path from old_git_dir to new_git_dir.2824 */2825void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)2826{2827 if (rename(old_git_dir, new_git_dir) < 0)2828 die_errno(_("could not migrate git directory from '%s' to '%s'"),2829 old_git_dir, new_git_dir);28302831 connect_work_tree_and_git_dir(path, new_git_dir);2832}