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 * the index if "check_index" is non-zero), 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, int check_index, 742 struct sha1_stat *sha1_stat) 743{ 744 struct stat st; 745 int fd, i, lineno = 1; 746 size_t size = 0; 747 char *buf, *entry; 748 749 fd = open(fname, O_RDONLY); 750 if (fd < 0 || fstat(fd, &st) < 0) { 751 if (errno != ENOENT) 752 warn_on_inaccessible(fname); 753 if (0 <= fd) 754 close(fd); 755 if (!check_index || 756 (buf = read_skip_worktree_file_from_index(&the_index, fname, &size, sha1_stat)) == NULL) 757 return -1; 758 if (size == 0) { 759 free(buf); 760 return 0; 761 } 762 if (buf[size-1] != '\n') { 763 buf = xrealloc(buf, st_add(size, 1)); 764 buf[size++] = '\n'; 765 } 766 } else { 767 size = xsize_t(st.st_size); 768 if (size == 0) { 769 if (sha1_stat) { 770 fill_stat_data(&sha1_stat->stat, &st); 771 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN); 772 sha1_stat->valid = 1; 773 } 774 close(fd); 775 return 0; 776 } 777 buf = xmallocz(size); 778 if (read_in_full(fd, buf, size) != size) { 779 free(buf); 780 close(fd); 781 return -1; 782 } 783 buf[size++] = '\n'; 784 close(fd); 785 if (sha1_stat) { 786 int pos; 787 if (sha1_stat->valid && 788 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st)) 789 ; /* no content change, ss->sha1 still good */ 790 else if (check_index && 791 (pos = index_name_pos(&the_index, fname, strlen(fname))) >= 0 && 792 !ce_stage(the_index.cache[pos]) && 793 ce_uptodate(the_index.cache[pos]) && 794 !would_convert_to_git(fname)) 795 hashcpy(sha1_stat->sha1, 796 the_index.cache[pos]->oid.hash); 797 else 798 hash_sha1_file(buf, size, "blob", sha1_stat->sha1); 799 fill_stat_data(&sha1_stat->stat, &st); 800 sha1_stat->valid = 1; 801 } 802 } 803 804 el->filebuf = buf; 805 806 if (skip_utf8_bom(&buf, size)) 807 size -= buf - el->filebuf; 808 809 entry = buf; 810 811 for (i = 0; i < size; i++) { 812 if (buf[i] == '\n') { 813 if (entry != buf + i && entry[0] != '#') { 814 buf[i - (i && buf[i-1] == '\r')] = 0; 815 trim_trailing_spaces(entry); 816 add_exclude(entry, base, baselen, el, lineno); 817 } 818 lineno++; 819 entry = buf + i + 1; 820 } 821 } 822 return 0; 823} 824 825int add_excludes_from_file_to_list(const char *fname, const char *base, 826 int baselen, struct exclude_list *el, 827 int check_index) 828{ 829 return add_excludes(fname, base, baselen, el, check_index, NULL); 830} 831 832struct exclude_list *add_exclude_list(struct dir_struct *dir, 833 int group_type, const char *src) 834{ 835 struct exclude_list *el; 836 struct exclude_list_group *group; 837 838 group = &dir->exclude_list_group[group_type]; 839 ALLOC_GROW(group->el, group->nr + 1, group->alloc); 840 el = &group->el[group->nr++]; 841 memset(el, 0, sizeof(*el)); 842 el->src = src; 843 return el; 844} 845 846/* 847 * Used to set up core.excludesfile and .git/info/exclude lists. 848 */ 849static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname, 850 struct sha1_stat *sha1_stat) 851{ 852 struct exclude_list *el; 853 /* 854 * catch setup_standard_excludes() that's called before 855 * dir->untracked is assigned. That function behaves 856 * differently when dir->untracked is non-NULL. 857 */ 858 if (!dir->untracked) 859 dir->unmanaged_exclude_files++; 860 el = add_exclude_list(dir, EXC_FILE, fname); 861 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0) 862 die("cannot use %s as an exclude file", fname); 863} 864 865void add_excludes_from_file(struct dir_struct *dir, const char *fname) 866{ 867 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */ 868 add_excludes_from_file_1(dir, fname, NULL); 869} 870 871int match_basename(const char *basename, int basenamelen, 872 const char *pattern, int prefix, int patternlen, 873 unsigned flags) 874{ 875 if (prefix == patternlen) { 876 if (patternlen == basenamelen && 877 !fspathncmp(pattern, basename, basenamelen)) 878 return 1; 879 } else if (flags & EXC_FLAG_ENDSWITH) { 880 /* "*literal" matching against "fooliteral" */ 881 if (patternlen - 1 <= basenamelen && 882 !fspathncmp(pattern + 1, 883 basename + basenamelen - (patternlen - 1), 884 patternlen - 1)) 885 return 1; 886 } else { 887 if (fnmatch_icase_mem(pattern, patternlen, 888 basename, basenamelen, 889 0) == 0) 890 return 1; 891 } 892 return 0; 893} 894 895int match_pathname(const char *pathname, int pathlen, 896 const char *base, int baselen, 897 const char *pattern, int prefix, int patternlen, 898 unsigned flags) 899{ 900 const char *name; 901 int namelen; 902 903 /* 904 * match with FNM_PATHNAME; the pattern has base implicitly 905 * in front of it. 906 */ 907 if (*pattern == '/') { 908 pattern++; 909 patternlen--; 910 prefix--; 911 } 912 913 /* 914 * baselen does not count the trailing slash. base[] may or 915 * may not end with a trailing slash though. 916 */ 917 if (pathlen < baselen + 1 || 918 (baselen && pathname[baselen] != '/') || 919 fspathncmp(pathname, base, baselen)) 920 return 0; 921 922 namelen = baselen ? pathlen - baselen - 1 : pathlen; 923 name = pathname + pathlen - namelen; 924 925 if (prefix) { 926 /* 927 * if the non-wildcard part is longer than the 928 * remaining pathname, surely it cannot match. 929 */ 930 if (prefix > namelen) 931 return 0; 932 933 if (fspathncmp(pattern, name, prefix)) 934 return 0; 935 pattern += prefix; 936 patternlen -= prefix; 937 name += prefix; 938 namelen -= prefix; 939 940 /* 941 * If the whole pattern did not have a wildcard, 942 * then our prefix match is all we need; we 943 * do not need to call fnmatch at all. 944 */ 945 if (!patternlen && !namelen) 946 return 1; 947 } 948 949 return fnmatch_icase_mem(pattern, patternlen, 950 name, namelen, 951 WM_PATHNAME) == 0; 952} 953 954/* 955 * Scan the given exclude list in reverse to see whether pathname 956 * should be ignored. The first match (i.e. the last on the list), if 957 * any, determines the fate. Returns the exclude_list element which 958 * matched, or NULL for undecided. 959 */ 960static struct exclude *last_exclude_matching_from_list(const char *pathname, 961 int pathlen, 962 const char *basename, 963 int *dtype, 964 struct exclude_list *el, 965 struct index_state *istate) 966{ 967 struct exclude *exc = NULL; /* undecided */ 968 int i; 969 970 if (!el->nr) 971 return NULL; /* undefined */ 972 973 for (i = el->nr - 1; 0 <= i; i--) { 974 struct exclude *x = el->excludes[i]; 975 const char *exclude = x->pattern; 976 int prefix = x->nowildcardlen; 977 978 if (x->flags & EXC_FLAG_MUSTBEDIR) { 979 if (*dtype == DT_UNKNOWN) 980 *dtype = get_dtype(NULL, istate, pathname, pathlen); 981 if (*dtype != DT_DIR) 982 continue; 983 } 984 985 if (x->flags & EXC_FLAG_NODIR) { 986 if (match_basename(basename, 987 pathlen - (basename - pathname), 988 exclude, prefix, x->patternlen, 989 x->flags)) { 990 exc = x; 991 break; 992 } 993 continue; 994 } 995 996 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/'); 997 if (match_pathname(pathname, pathlen, 998 x->base, x->baselen ? x->baselen - 1 : 0, 999 exclude, prefix, x->patternlen, x->flags)) {1000 exc = x;1001 break;1002 }1003 }1004 return exc;1005}10061007/*1008 * Scan the list and let the last match determine the fate.1009 * Return 1 for exclude, 0 for include and -1 for undecided.1010 */1011int is_excluded_from_list(const char *pathname,1012 int pathlen, const char *basename, int *dtype,1013 struct exclude_list *el, struct index_state *istate)1014{1015 struct exclude *exclude;1016 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,1017 dtype, el, istate);1018 if (exclude)1019 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1020 return -1; /* undecided */1021}10221023static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,1024 struct index_state *istate,1025 const char *pathname, int pathlen, const char *basename,1026 int *dtype_p)1027{1028 int i, j;1029 struct exclude_list_group *group;1030 struct exclude *exclude;1031 for (i = EXC_CMDL; i <= EXC_FILE; i++) {1032 group = &dir->exclude_list_group[i];1033 for (j = group->nr - 1; j >= 0; j--) {1034 exclude = last_exclude_matching_from_list(1035 pathname, pathlen, basename, dtype_p,1036 &group->el[j], istate);1037 if (exclude)1038 return exclude;1039 }1040 }1041 return NULL;1042}10431044/*1045 * Loads the per-directory exclude list for the substring of base1046 * which has a char length of baselen.1047 */1048static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)1049{1050 struct exclude_list_group *group;1051 struct exclude_list *el;1052 struct exclude_stack *stk = NULL;1053 struct untracked_cache_dir *untracked;1054 int current;10551056 group = &dir->exclude_list_group[EXC_DIRS];10571058 /*1059 * Pop the exclude lists from the EXCL_DIRS exclude_list_group1060 * which originate from directories not in the prefix of the1061 * path being checked.1062 */1063 while ((stk = dir->exclude_stack) != NULL) {1064 if (stk->baselen <= baselen &&1065 !strncmp(dir->basebuf.buf, base, stk->baselen))1066 break;1067 el = &group->el[dir->exclude_stack->exclude_ix];1068 dir->exclude_stack = stk->prev;1069 dir->exclude = NULL;1070 free((char *)el->src); /* see strbuf_detach() below */1071 clear_exclude_list(el);1072 free(stk);1073 group->nr--;1074 }10751076 /* Skip traversing into sub directories if the parent is excluded */1077 if (dir->exclude)1078 return;10791080 /*1081 * Lazy initialization. All call sites currently just1082 * memset(dir, 0, sizeof(*dir)) before use. Changing all of1083 * them seems lots of work for little benefit.1084 */1085 if (!dir->basebuf.buf)1086 strbuf_init(&dir->basebuf, PATH_MAX);10871088 /* Read from the parent directories and push them down. */1089 current = stk ? stk->baselen : -1;1090 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);1091 if (dir->untracked)1092 untracked = stk ? stk->ucd : dir->untracked->root;1093 else1094 untracked = NULL;10951096 while (current < baselen) {1097 const char *cp;1098 struct sha1_stat sha1_stat;10991100 stk = xcalloc(1, sizeof(*stk));1101 if (current < 0) {1102 cp = base;1103 current = 0;1104 } else {1105 cp = strchr(base + current + 1, '/');1106 if (!cp)1107 die("oops in prep_exclude");1108 cp++;1109 untracked =1110 lookup_untracked(dir->untracked, untracked,1111 base + current,1112 cp - base - current);1113 }1114 stk->prev = dir->exclude_stack;1115 stk->baselen = cp - base;1116 stk->exclude_ix = group->nr;1117 stk->ucd = untracked;1118 el = add_exclude_list(dir, EXC_DIRS, NULL);1119 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);1120 assert(stk->baselen == dir->basebuf.len);11211122 /* Abort if the directory is excluded */1123 if (stk->baselen) {1124 int dt = DT_DIR;1125 dir->basebuf.buf[stk->baselen - 1] = 0;1126 dir->exclude = last_exclude_matching_from_lists(dir,1127 &the_index,1128 dir->basebuf.buf, stk->baselen - 1,1129 dir->basebuf.buf + current, &dt);1130 dir->basebuf.buf[stk->baselen - 1] = '/';1131 if (dir->exclude &&1132 dir->exclude->flags & EXC_FLAG_NEGATIVE)1133 dir->exclude = NULL;1134 if (dir->exclude) {1135 dir->exclude_stack = stk;1136 return;1137 }1138 }11391140 /* Try to read per-directory file */1141 hashclr(sha1_stat.sha1);1142 sha1_stat.valid = 0;1143 if (dir->exclude_per_dir &&1144 /*1145 * If we know that no files have been added in1146 * this directory (i.e. valid_cached_dir() has1147 * been executed and set untracked->valid) ..1148 */1149 (!untracked || !untracked->valid ||1150 /*1151 * .. and .gitignore does not exist before1152 * (i.e. null exclude_sha1). Then we can skip1153 * loading .gitignore, which would result in1154 * ENOENT anyway.1155 */1156 !is_null_sha1(untracked->exclude_sha1))) {1157 /*1158 * dir->basebuf gets reused by the traversal, but we1159 * need fname to remain unchanged to ensure the src1160 * member of each struct exclude correctly1161 * back-references its source file. Other invocations1162 * of add_exclude_list provide stable strings, so we1163 * strbuf_detach() and free() here in the caller.1164 */1165 struct strbuf sb = STRBUF_INIT;1166 strbuf_addbuf(&sb, &dir->basebuf);1167 strbuf_addstr(&sb, dir->exclude_per_dir);1168 el->src = strbuf_detach(&sb, NULL);1169 add_excludes(el->src, el->src, stk->baselen, el, 1,1170 untracked ? &sha1_stat : NULL);1171 }1172 /*1173 * NEEDSWORK: when untracked cache is enabled, prep_exclude()1174 * will first be called in valid_cached_dir() then maybe many1175 * times more in last_exclude_matching(). When the cache is1176 * used, last_exclude_matching() will not be called and1177 * reading .gitignore content will be a waste.1178 *1179 * So when it's called by valid_cached_dir() and we can get1180 * .gitignore SHA-1 from the index (i.e. .gitignore is not1181 * modified on work tree), we could delay reading the1182 * .gitignore content until we absolutely need it in1183 * last_exclude_matching(). Be careful about ignore rule1184 * order, though, if you do that.1185 */1186 if (untracked &&1187 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {1188 invalidate_gitignore(dir->untracked, untracked);1189 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);1190 }1191 dir->exclude_stack = stk;1192 current = stk->baselen;1193 }1194 strbuf_setlen(&dir->basebuf, baselen);1195}11961197/*1198 * Loads the exclude lists for the directory containing pathname, then1199 * scans all exclude lists to determine whether pathname is excluded.1200 * Returns the exclude_list element which matched, or NULL for1201 * undecided.1202 */1203struct exclude *last_exclude_matching(struct dir_struct *dir,1204 const char *pathname,1205 int *dtype_p)1206{1207 int pathlen = strlen(pathname);1208 const char *basename = strrchr(pathname, '/');1209 basename = (basename) ? basename+1 : pathname;12101211 prep_exclude(dir, pathname, basename-pathname);12121213 if (dir->exclude)1214 return dir->exclude;12151216 return last_exclude_matching_from_lists(dir, &the_index, pathname, pathlen,1217 basename, dtype_p);1218}12191220/*1221 * Loads the exclude lists for the directory containing pathname, then1222 * scans all exclude lists to determine whether pathname is excluded.1223 * Returns 1 if true, otherwise 0.1224 */1225int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)1226{1227 struct exclude *exclude =1228 last_exclude_matching(dir, pathname, dtype_p);1229 if (exclude)1230 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1231 return 0;1232}12331234static struct dir_entry *dir_entry_new(const char *pathname, int len)1235{1236 struct dir_entry *ent;12371238 FLEX_ALLOC_MEM(ent, name, pathname, len);1239 ent->len = len;1240 return ent;1241}12421243static struct dir_entry *dir_add_name(struct dir_struct *dir,1244 struct index_state *istate,1245 const char *pathname, int len)1246{1247 if (index_file_exists(istate, pathname, len, ignore_case))1248 return NULL;12491250 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);1251 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);1252}12531254struct dir_entry *dir_add_ignored(struct dir_struct *dir,1255 struct index_state *istate,1256 const char *pathname, int len)1257{1258 if (!index_name_is_other(istate, pathname, len))1259 return NULL;12601261 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);1262 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);1263}12641265enum exist_status {1266 index_nonexistent = 0,1267 index_directory,1268 index_gitdir1269};12701271/*1272 * Do not use the alphabetically sorted index to look up1273 * the directory name; instead, use the case insensitive1274 * directory hash.1275 */1276static enum exist_status directory_exists_in_index_icase(struct index_state *istate,1277 const char *dirname, int len)1278{1279 struct cache_entry *ce;12801281 if (index_dir_exists(istate, dirname, len))1282 return index_directory;12831284 ce = index_file_exists(istate, dirname, len, ignore_case);1285 if (ce && S_ISGITLINK(ce->ce_mode))1286 return index_gitdir;12871288 return index_nonexistent;1289}12901291/*1292 * The index sorts alphabetically by entry name, which1293 * means that a gitlink sorts as '\0' at the end, while1294 * a directory (which is defined not as an entry, but as1295 * the files it contains) will sort with the '/' at the1296 * end.1297 */1298static enum exist_status directory_exists_in_index(struct index_state *istate,1299 const char *dirname, int len)1300{1301 int pos;13021303 if (ignore_case)1304 return directory_exists_in_index_icase(istate, dirname, len);13051306 pos = index_name_pos(istate, dirname, len);1307 if (pos < 0)1308 pos = -pos-1;1309 while (pos < istate->cache_nr) {1310 const struct cache_entry *ce = istate->cache[pos++];1311 unsigned char endchar;13121313 if (strncmp(ce->name, dirname, len))1314 break;1315 endchar = ce->name[len];1316 if (endchar > '/')1317 break;1318 if (endchar == '/')1319 return index_directory;1320 if (!endchar && S_ISGITLINK(ce->ce_mode))1321 return index_gitdir;1322 }1323 return index_nonexistent;1324}13251326/*1327 * When we find a directory when traversing the filesystem, we1328 * have three distinct cases:1329 *1330 * - ignore it1331 * - see it as a directory1332 * - recurse into it1333 *1334 * and which one we choose depends on a combination of existing1335 * git index contents and the flags passed into the directory1336 * traversal routine.1337 *1338 * Case 1: If we *already* have entries in the index under that1339 * directory name, we always recurse into the directory to see1340 * all the files.1341 *1342 * Case 2: If we *already* have that directory name as a gitlink,1343 * we always continue to see it as a gitlink, regardless of whether1344 * there is an actual git directory there or not (it might not1345 * be checked out as a subproject!)1346 *1347 * Case 3: if we didn't have it in the index previously, we1348 * have a few sub-cases:1349 *1350 * (a) if "show_other_directories" is true, we show it as1351 * just a directory, unless "hide_empty_directories" is1352 * also true, in which case we need to check if it contains any1353 * untracked and / or ignored files.1354 * (b) if it looks like a git directory, and we don't have1355 * 'no_gitlinks' set we treat it as a gitlink, and show it1356 * as a directory.1357 * (c) otherwise, we recurse into it.1358 */1359static enum path_treatment treat_directory(struct dir_struct *dir,1360 struct untracked_cache_dir *untracked,1361 const char *dirname, int len, int baselen, int exclude,1362 const struct pathspec *pathspec)1363{1364 /* The "len-1" is to strip the final '/' */1365 switch (directory_exists_in_index(&the_index, dirname, len-1)) {1366 case index_directory:1367 return path_recurse;13681369 case index_gitdir:1370 return path_none;13711372 case index_nonexistent:1373 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)1374 break;1375 if (!(dir->flags & DIR_NO_GITLINKS)) {1376 unsigned char sha1[20];1377 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)1378 return path_untracked;1379 }1380 return path_recurse;1381 }13821383 /* This is the "show_other_directories" case */13841385 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))1386 return exclude ? path_excluded : path_untracked;13871388 untracked = lookup_untracked(dir->untracked, untracked,1389 dirname + baselen, len - baselen);1390 return read_directory_recursive(dir, dirname, len,1391 untracked, 1, pathspec);1392}13931394/*1395 * This is an inexact early pruning of any recursive directory1396 * reading - if the path cannot possibly be in the pathspec,1397 * return true, and we'll skip it early.1398 */1399static int simplify_away(const char *path, int pathlen,1400 const struct pathspec *pathspec)1401{1402 int i;14031404 if (!pathspec || !pathspec->nr)1405 return 0;14061407 GUARD_PATHSPEC(pathspec,1408 PATHSPEC_FROMTOP |1409 PATHSPEC_MAXDEPTH |1410 PATHSPEC_LITERAL |1411 PATHSPEC_GLOB |1412 PATHSPEC_ICASE |1413 PATHSPEC_EXCLUDE |1414 PATHSPEC_ATTR);14151416 for (i = 0; i < pathspec->nr; i++) {1417 const struct pathspec_item *item = &pathspec->items[i];1418 int len = item->nowildcard_len;14191420 if (len > pathlen)1421 len = pathlen;1422 if (!ps_strncmp(item, item->match, path, len))1423 return 0;1424 }14251426 return 1;1427}14281429/*1430 * This function tells us whether an excluded path matches a1431 * list of "interesting" pathspecs. That is, whether a path matched1432 * by any of the pathspecs could possibly be ignored by excluding1433 * the specified path. This can happen if:1434 *1435 * 1. the path is mentioned explicitly in the pathspec1436 *1437 * 2. the path is a directory prefix of some element in the1438 * pathspec1439 */1440static int exclude_matches_pathspec(const char *path, int pathlen,1441 const struct pathspec *pathspec)1442{1443 int i;14441445 if (!pathspec || !pathspec->nr)1446 return 0;14471448 GUARD_PATHSPEC(pathspec,1449 PATHSPEC_FROMTOP |1450 PATHSPEC_MAXDEPTH |1451 PATHSPEC_LITERAL |1452 PATHSPEC_GLOB |1453 PATHSPEC_ICASE |1454 PATHSPEC_EXCLUDE);14551456 for (i = 0; i < pathspec->nr; i++) {1457 const struct pathspec_item *item = &pathspec->items[i];1458 int len = item->nowildcard_len;14591460 if (len == pathlen &&1461 !ps_strncmp(item, item->match, path, pathlen))1462 return 1;1463 if (len > pathlen &&1464 item->match[pathlen] == '/' &&1465 !ps_strncmp(item, item->match, path, pathlen))1466 return 1;1467 }1468 return 0;1469}14701471static int get_index_dtype(struct index_state *istate,1472 const char *path, int len)1473{1474 int pos;1475 const struct cache_entry *ce;14761477 ce = index_file_exists(istate, path, len, 0);1478 if (ce) {1479 if (!ce_uptodate(ce))1480 return DT_UNKNOWN;1481 if (S_ISGITLINK(ce->ce_mode))1482 return DT_DIR;1483 /*1484 * Nobody actually cares about the1485 * difference between DT_LNK and DT_REG1486 */1487 return DT_REG;1488 }14891490 /* Try to look it up as a directory */1491 pos = index_name_pos(istate, path, len);1492 if (pos >= 0)1493 return DT_UNKNOWN;1494 pos = -pos-1;1495 while (pos < istate->cache_nr) {1496 ce = istate->cache[pos++];1497 if (strncmp(ce->name, path, len))1498 break;1499 if (ce->name[len] > '/')1500 break;1501 if (ce->name[len] < '/')1502 continue;1503 if (!ce_uptodate(ce))1504 break; /* continue? */1505 return DT_DIR;1506 }1507 return DT_UNKNOWN;1508}15091510static int get_dtype(struct dirent *de, struct index_state *istate,1511 const char *path, int len)1512{1513 int dtype = de ? DTYPE(de) : DT_UNKNOWN;1514 struct stat st;15151516 if (dtype != DT_UNKNOWN)1517 return dtype;1518 dtype = get_index_dtype(istate, path, len);1519 if (dtype != DT_UNKNOWN)1520 return dtype;1521 if (lstat(path, &st))1522 return dtype;1523 if (S_ISREG(st.st_mode))1524 return DT_REG;1525 if (S_ISDIR(st.st_mode))1526 return DT_DIR;1527 if (S_ISLNK(st.st_mode))1528 return DT_LNK;1529 return dtype;1530}15311532static enum path_treatment treat_one_path(struct dir_struct *dir,1533 struct untracked_cache_dir *untracked,1534 struct strbuf *path,1535 int baselen,1536 const struct pathspec *pathspec,1537 int dtype, struct dirent *de)1538{1539 int exclude;1540 int has_path_in_index = !!index_file_exists(&the_index, path->buf, path->len, ignore_case);15411542 if (dtype == DT_UNKNOWN)1543 dtype = get_dtype(de, &the_index, path->buf, path->len);15441545 /* Always exclude indexed files */1546 if (dtype != DT_DIR && has_path_in_index)1547 return path_none;15481549 /*1550 * When we are looking at a directory P in the working tree,1551 * there are three cases:1552 *1553 * (1) P exists in the index. Everything inside the directory P in1554 * the working tree needs to go when P is checked out from the1555 * index.1556 *1557 * (2) P does not exist in the index, but there is P/Q in the index.1558 * We know P will stay a directory when we check out the contents1559 * of the index, but we do not know yet if there is a directory1560 * P/Q in the working tree to be killed, so we need to recurse.1561 *1562 * (3) P does not exist in the index, and there is no P/Q in the index1563 * to require P to be a directory, either. Only in this case, we1564 * know that everything inside P will not be killed without1565 * recursing.1566 */1567 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&1568 (dtype == DT_DIR) &&1569 !has_path_in_index &&1570 (directory_exists_in_index(&the_index, path->buf, path->len) == index_nonexistent))1571 return path_none;15721573 exclude = is_excluded(dir, path->buf, &dtype);15741575 /*1576 * Excluded? If we don't explicitly want to show1577 * ignored files, ignore it1578 */1579 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))1580 return path_excluded;15811582 switch (dtype) {1583 default:1584 return path_none;1585 case DT_DIR:1586 strbuf_addch(path, '/');1587 return treat_directory(dir, untracked, path->buf, path->len,1588 baselen, exclude, pathspec);1589 case DT_REG:1590 case DT_LNK:1591 return exclude ? path_excluded : path_untracked;1592 }1593}15941595static enum path_treatment treat_path_fast(struct dir_struct *dir,1596 struct untracked_cache_dir *untracked,1597 struct cached_dir *cdir,1598 struct strbuf *path,1599 int baselen,1600 const struct pathspec *pathspec)1601{1602 strbuf_setlen(path, baselen);1603 if (!cdir->ucd) {1604 strbuf_addstr(path, cdir->file);1605 return path_untracked;1606 }1607 strbuf_addstr(path, cdir->ucd->name);1608 /* treat_one_path() does this before it calls treat_directory() */1609 strbuf_complete(path, '/');1610 if (cdir->ucd->check_only)1611 /*1612 * check_only is set as a result of treat_directory() getting1613 * to its bottom. Verify again the same set of directories1614 * with check_only set.1615 */1616 return read_directory_recursive(dir, path->buf, path->len,1617 cdir->ucd, 1, pathspec);1618 /*1619 * We get path_recurse in the first run when1620 * directory_exists_in_index() returns index_nonexistent. We1621 * are sure that new changes in the index does not impact the1622 * outcome. Return now.1623 */1624 return path_recurse;1625}16261627static enum path_treatment treat_path(struct dir_struct *dir,1628 struct untracked_cache_dir *untracked,1629 struct cached_dir *cdir,1630 struct strbuf *path,1631 int baselen,1632 const struct pathspec *pathspec)1633{1634 int dtype;1635 struct dirent *de = cdir->de;16361637 if (!de)1638 return treat_path_fast(dir, untracked, cdir, path,1639 baselen, pathspec);1640 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))1641 return path_none;1642 strbuf_setlen(path, baselen);1643 strbuf_addstr(path, de->d_name);1644 if (simplify_away(path->buf, path->len, pathspec))1645 return path_none;16461647 dtype = DTYPE(de);1648 return treat_one_path(dir, untracked, path, baselen, pathspec, dtype, de);1649}16501651static void add_untracked(struct untracked_cache_dir *dir, const char *name)1652{1653 if (!dir)1654 return;1655 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,1656 dir->untracked_alloc);1657 dir->untracked[dir->untracked_nr++] = xstrdup(name);1658}16591660static int valid_cached_dir(struct dir_struct *dir,1661 struct untracked_cache_dir *untracked,1662 struct strbuf *path,1663 int check_only)1664{1665 struct stat st;16661667 if (!untracked)1668 return 0;16691670 if (stat(path->len ? path->buf : ".", &st)) {1671 invalidate_directory(dir->untracked, untracked);1672 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));1673 return 0;1674 }1675 if (!untracked->valid ||1676 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {1677 if (untracked->valid)1678 invalidate_directory(dir->untracked, untracked);1679 fill_stat_data(&untracked->stat_data, &st);1680 return 0;1681 }16821683 if (untracked->check_only != !!check_only) {1684 invalidate_directory(dir->untracked, untracked);1685 return 0;1686 }16871688 /*1689 * prep_exclude will be called eventually on this directory,1690 * but it's called much later in last_exclude_matching(). We1691 * need it now to determine the validity of the cache for this1692 * path. The next calls will be nearly no-op, the way1693 * prep_exclude() is designed.1694 */1695 if (path->len && path->buf[path->len - 1] != '/') {1696 strbuf_addch(path, '/');1697 prep_exclude(dir, path->buf, path->len);1698 strbuf_setlen(path, path->len - 1);1699 } else1700 prep_exclude(dir, path->buf, path->len);17011702 /* hopefully prep_exclude() haven't invalidated this entry... */1703 return untracked->valid;1704}17051706static int open_cached_dir(struct cached_dir *cdir,1707 struct dir_struct *dir,1708 struct untracked_cache_dir *untracked,1709 struct strbuf *path,1710 int check_only)1711{1712 memset(cdir, 0, sizeof(*cdir));1713 cdir->untracked = untracked;1714 if (valid_cached_dir(dir, untracked, path, check_only))1715 return 0;1716 cdir->fdir = opendir(path->len ? path->buf : ".");1717 if (dir->untracked)1718 dir->untracked->dir_opened++;1719 if (!cdir->fdir)1720 return -1;1721 return 0;1722}17231724static int read_cached_dir(struct cached_dir *cdir)1725{1726 if (cdir->fdir) {1727 cdir->de = readdir(cdir->fdir);1728 if (!cdir->de)1729 return -1;1730 return 0;1731 }1732 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {1733 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];1734 if (!d->recurse) {1735 cdir->nr_dirs++;1736 continue;1737 }1738 cdir->ucd = d;1739 cdir->nr_dirs++;1740 return 0;1741 }1742 cdir->ucd = NULL;1743 if (cdir->nr_files < cdir->untracked->untracked_nr) {1744 struct untracked_cache_dir *d = cdir->untracked;1745 cdir->file = d->untracked[cdir->nr_files++];1746 return 0;1747 }1748 return -1;1749}17501751static void close_cached_dir(struct cached_dir *cdir)1752{1753 if (cdir->fdir)1754 closedir(cdir->fdir);1755 /*1756 * We have gone through this directory and found no untracked1757 * entries. Mark it valid.1758 */1759 if (cdir->untracked) {1760 cdir->untracked->valid = 1;1761 cdir->untracked->recurse = 1;1762 }1763}17641765/*1766 * Read a directory tree. We currently ignore anything but1767 * directories, regular files and symlinks. That's because git1768 * doesn't handle them at all yet. Maybe that will change some1769 * day.1770 *1771 * Also, we ignore the name ".git" (even if it is not a directory).1772 * That likely will not change.1773 *1774 * Returns the most significant path_treatment value encountered in the scan.1775 */1776static enum path_treatment read_directory_recursive(struct dir_struct *dir,1777 const char *base, int baselen,1778 struct untracked_cache_dir *untracked, int check_only,1779 const struct pathspec *pathspec)1780{1781 struct cached_dir cdir;1782 enum path_treatment state, subdir_state, dir_state = path_none;1783 struct strbuf path = STRBUF_INIT;17841785 strbuf_add(&path, base, baselen);17861787 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))1788 goto out;17891790 if (untracked)1791 untracked->check_only = !!check_only;17921793 while (!read_cached_dir(&cdir)) {1794 /* check how the file or directory should be treated */1795 state = treat_path(dir, untracked, &cdir, &path,1796 baselen, pathspec);17971798 if (state > dir_state)1799 dir_state = state;18001801 /* recurse into subdir if instructed by treat_path */1802 if (state == path_recurse) {1803 struct untracked_cache_dir *ud;1804 ud = lookup_untracked(dir->untracked, untracked,1805 path.buf + baselen,1806 path.len - baselen);1807 subdir_state =1808 read_directory_recursive(dir, path.buf,1809 path.len, ud,1810 check_only, pathspec);1811 if (subdir_state > dir_state)1812 dir_state = subdir_state;1813 }18141815 if (check_only) {1816 /* abort early if maximum state has been reached */1817 if (dir_state == path_untracked) {1818 if (cdir.fdir)1819 add_untracked(untracked, path.buf + baselen);1820 break;1821 }1822 /* skip the dir_add_* part */1823 continue;1824 }18251826 /* add the path to the appropriate result list */1827 switch (state) {1828 case path_excluded:1829 if (dir->flags & DIR_SHOW_IGNORED)1830 dir_add_name(dir, &the_index, path.buf, path.len);1831 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||1832 ((dir->flags & DIR_COLLECT_IGNORED) &&1833 exclude_matches_pathspec(path.buf, path.len,1834 pathspec)))1835 dir_add_ignored(dir, &the_index, path.buf, path.len);1836 break;18371838 case path_untracked:1839 if (dir->flags & DIR_SHOW_IGNORED)1840 break;1841 dir_add_name(dir, &the_index, path.buf, path.len);1842 if (cdir.fdir)1843 add_untracked(untracked, path.buf + baselen);1844 break;18451846 default:1847 break;1848 }1849 }1850 close_cached_dir(&cdir);1851 out:1852 strbuf_release(&path);18531854 return dir_state;1855}18561857static int cmp_name(const void *p1, const void *p2)1858{1859 const struct dir_entry *e1 = *(const struct dir_entry **)p1;1860 const struct dir_entry *e2 = *(const struct dir_entry **)p2;18611862 return name_compare(e1->name, e1->len, e2->name, e2->len);1863}18641865static int treat_leading_path(struct dir_struct *dir,1866 const char *path, int len,1867 const struct pathspec *pathspec)1868{1869 struct strbuf sb = STRBUF_INIT;1870 int baselen, rc = 0;1871 const char *cp;1872 int old_flags = dir->flags;18731874 while (len && path[len - 1] == '/')1875 len--;1876 if (!len)1877 return 1;1878 baselen = 0;1879 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;1880 while (1) {1881 cp = path + baselen + !!baselen;1882 cp = memchr(cp, '/', path + len - cp);1883 if (!cp)1884 baselen = len;1885 else1886 baselen = cp - path;1887 strbuf_setlen(&sb, 0);1888 strbuf_add(&sb, path, baselen);1889 if (!is_directory(sb.buf))1890 break;1891 if (simplify_away(sb.buf, sb.len, pathspec))1892 break;1893 if (treat_one_path(dir, NULL, &sb, baselen, pathspec,1894 DT_DIR, NULL) == path_none)1895 break; /* do not recurse into it */1896 if (len <= baselen) {1897 rc = 1;1898 break; /* finished checking */1899 }1900 }1901 strbuf_release(&sb);1902 dir->flags = old_flags;1903 return rc;1904}19051906static const char *get_ident_string(void)1907{1908 static struct strbuf sb = STRBUF_INIT;1909 struct utsname uts;19101911 if (sb.len)1912 return sb.buf;1913 if (uname(&uts) < 0)1914 die_errno(_("failed to get kernel name and information"));1915 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),1916 uts.sysname);1917 return sb.buf;1918}19191920static int ident_in_untracked(const struct untracked_cache *uc)1921{1922 /*1923 * Previous git versions may have saved many NUL separated1924 * strings in the "ident" field, but it is insane to manage1925 * many locations, so just take care of the first one.1926 */19271928 return !strcmp(uc->ident.buf, get_ident_string());1929}19301931static void set_untracked_ident(struct untracked_cache *uc)1932{1933 strbuf_reset(&uc->ident);1934 strbuf_addstr(&uc->ident, get_ident_string());19351936 /*1937 * This strbuf used to contain a list of NUL separated1938 * strings, so save NUL too for backward compatibility.1939 */1940 strbuf_addch(&uc->ident, 0);1941}19421943static void new_untracked_cache(struct index_state *istate)1944{1945 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));1946 strbuf_init(&uc->ident, 100);1947 uc->exclude_per_dir = ".gitignore";1948 /* should be the same flags used by git-status */1949 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;1950 set_untracked_ident(uc);1951 istate->untracked = uc;1952 istate->cache_changed |= UNTRACKED_CHANGED;1953}19541955void add_untracked_cache(struct index_state *istate)1956{1957 if (!istate->untracked) {1958 new_untracked_cache(istate);1959 } else {1960 if (!ident_in_untracked(istate->untracked)) {1961 free_untracked_cache(istate->untracked);1962 new_untracked_cache(istate);1963 }1964 }1965}19661967void remove_untracked_cache(struct index_state *istate)1968{1969 if (istate->untracked) {1970 free_untracked_cache(istate->untracked);1971 istate->untracked = NULL;1972 istate->cache_changed |= UNTRACKED_CHANGED;1973 }1974}19751976static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,1977 int base_len,1978 const struct pathspec *pathspec)1979{1980 struct untracked_cache_dir *root;19811982 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))1983 return NULL;19841985 /*1986 * We only support $GIT_DIR/info/exclude and core.excludesfile1987 * as the global ignore rule files. Any other additions1988 * (e.g. from command line) invalidate the cache. This1989 * condition also catches running setup_standard_excludes()1990 * before setting dir->untracked!1991 */1992 if (dir->unmanaged_exclude_files)1993 return NULL;19941995 /*1996 * Optimize for the main use case only: whole-tree git1997 * status. More work involved in treat_leading_path() if we1998 * use cache on just a subset of the worktree. pathspec1999 * support could make the matter even worse.2000 */2001 if (base_len || (pathspec && pathspec->nr))2002 return NULL;20032004 /* Different set of flags may produce different results */2005 if (dir->flags != dir->untracked->dir_flags ||2006 /*2007 * See treat_directory(), case index_nonexistent. Without2008 * this flag, we may need to also cache .git file content2009 * for the resolve_gitlink_ref() call, which we don't.2010 */2011 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||2012 /* We don't support collecting ignore files */2013 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |2014 DIR_COLLECT_IGNORED)))2015 return NULL;20162017 /*2018 * If we use .gitignore in the cache and now you change it to2019 * .gitexclude, everything will go wrong.2020 */2021 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&2022 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))2023 return NULL;20242025 /*2026 * EXC_CMDL is not considered in the cache. If people set it,2027 * skip the cache.2028 */2029 if (dir->exclude_list_group[EXC_CMDL].nr)2030 return NULL;20312032 if (!ident_in_untracked(dir->untracked)) {2033 warning(_("Untracked cache is disabled on this system or location."));2034 return NULL;2035 }20362037 if (!dir->untracked->root) {2038 const int len = sizeof(*dir->untracked->root);2039 dir->untracked->root = xmalloc(len);2040 memset(dir->untracked->root, 0, len);2041 }20422043 /* Validate $GIT_DIR/info/exclude and core.excludesfile */2044 root = dir->untracked->root;2045 if (hashcmp(dir->ss_info_exclude.sha1,2046 dir->untracked->ss_info_exclude.sha1)) {2047 invalidate_gitignore(dir->untracked, root);2048 dir->untracked->ss_info_exclude = dir->ss_info_exclude;2049 }2050 if (hashcmp(dir->ss_excludes_file.sha1,2051 dir->untracked->ss_excludes_file.sha1)) {2052 invalidate_gitignore(dir->untracked, root);2053 dir->untracked->ss_excludes_file = dir->ss_excludes_file;2054 }20552056 /* Make sure this directory is not dropped out at saving phase */2057 root->recurse = 1;2058 return root;2059}20602061int read_directory(struct dir_struct *dir, const char *path,2062 int len, const struct pathspec *pathspec)2063{2064 struct untracked_cache_dir *untracked;20652066 if (has_symlink_leading_path(path, len))2067 return dir->nr;20682069 untracked = validate_untracked_cache(dir, len, pathspec);2070 if (!untracked)2071 /*2072 * make sure untracked cache code path is disabled,2073 * e.g. prep_exclude()2074 */2075 dir->untracked = NULL;2076 if (!len || treat_leading_path(dir, path, len, pathspec))2077 read_directory_recursive(dir, path, len, untracked, 0, pathspec);2078 QSORT(dir->entries, dir->nr, cmp_name);2079 QSORT(dir->ignored, dir->ignored_nr, cmp_name);2080 if (dir->untracked) {2081 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);2082 trace_printf_key(&trace_untracked_stats,2083 "node creation: %u\n"2084 "gitignore invalidation: %u\n"2085 "directory invalidation: %u\n"2086 "opendir: %u\n",2087 dir->untracked->dir_created,2088 dir->untracked->gitignore_invalidated,2089 dir->untracked->dir_invalidated,2090 dir->untracked->dir_opened);2091 if (dir->untracked == the_index.untracked &&2092 (dir->untracked->dir_opened ||2093 dir->untracked->gitignore_invalidated ||2094 dir->untracked->dir_invalidated))2095 the_index.cache_changed |= UNTRACKED_CHANGED;2096 if (dir->untracked != the_index.untracked) {2097 free(dir->untracked);2098 dir->untracked = NULL;2099 }2100 }2101 return dir->nr;2102}21032104int file_exists(const char *f)2105{2106 struct stat sb;2107 return lstat(f, &sb) == 0;2108}21092110static int cmp_icase(char a, char b)2111{2112 if (a == b)2113 return 0;2114 if (ignore_case)2115 return toupper(a) - toupper(b);2116 return a - b;2117}21182119/*2120 * Given two normalized paths (a trailing slash is ok), if subdir is2121 * outside dir, return -1. Otherwise return the offset in subdir that2122 * can be used as relative path to dir.2123 */2124int dir_inside_of(const char *subdir, const char *dir)2125{2126 int offset = 0;21272128 assert(dir && subdir && *dir && *subdir);21292130 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {2131 dir++;2132 subdir++;2133 offset++;2134 }21352136 /* hel[p]/me vs hel[l]/yeah */2137 if (*dir && *subdir)2138 return -1;21392140 if (!*subdir)2141 return !*dir ? offset : -1; /* same dir */21422143 /* foo/[b]ar vs foo/[] */2144 if (is_dir_sep(dir[-1]))2145 return is_dir_sep(subdir[-1]) ? offset : -1;21462147 /* foo[/]bar vs foo[] */2148 return is_dir_sep(*subdir) ? offset + 1 : -1;2149}21502151int is_inside_dir(const char *dir)2152{2153 char *cwd;2154 int rc;21552156 if (!dir)2157 return 0;21582159 cwd = xgetcwd();2160 rc = (dir_inside_of(cwd, dir) >= 0);2161 free(cwd);2162 return rc;2163}21642165int is_empty_dir(const char *path)2166{2167 DIR *dir = opendir(path);2168 struct dirent *e;2169 int ret = 1;21702171 if (!dir)2172 return 0;21732174 while ((e = readdir(dir)) != NULL)2175 if (!is_dot_or_dotdot(e->d_name)) {2176 ret = 0;2177 break;2178 }21792180 closedir(dir);2181 return ret;2182}21832184static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)2185{2186 DIR *dir;2187 struct dirent *e;2188 int ret = 0, original_len = path->len, len, kept_down = 0;2189 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);2190 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);2191 unsigned char submodule_head[20];21922193 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&2194 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {2195 /* Do not descend and nuke a nested git work tree. */2196 if (kept_up)2197 *kept_up = 1;2198 return 0;2199 }22002201 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;2202 dir = opendir(path->buf);2203 if (!dir) {2204 if (errno == ENOENT)2205 return keep_toplevel ? -1 : 0;2206 else if (errno == EACCES && !keep_toplevel)2207 /*2208 * An empty dir could be removable even if it2209 * is unreadable:2210 */2211 return rmdir(path->buf);2212 else2213 return -1;2214 }2215 strbuf_complete(path, '/');22162217 len = path->len;2218 while ((e = readdir(dir)) != NULL) {2219 struct stat st;2220 if (is_dot_or_dotdot(e->d_name))2221 continue;22222223 strbuf_setlen(path, len);2224 strbuf_addstr(path, e->d_name);2225 if (lstat(path->buf, &st)) {2226 if (errno == ENOENT)2227 /*2228 * file disappeared, which is what we2229 * wanted anyway2230 */2231 continue;2232 /* fall thru */2233 } else if (S_ISDIR(st.st_mode)) {2234 if (!remove_dir_recurse(path, flag, &kept_down))2235 continue; /* happy */2236 } else if (!only_empty &&2237 (!unlink(path->buf) || errno == ENOENT)) {2238 continue; /* happy, too */2239 }22402241 /* path too long, stat fails, or non-directory still exists */2242 ret = -1;2243 break;2244 }2245 closedir(dir);22462247 strbuf_setlen(path, original_len);2248 if (!ret && !keep_toplevel && !kept_down)2249 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;2250 else if (kept_up)2251 /*2252 * report the uplevel that it is not an error that we2253 * did not rmdir() our directory.2254 */2255 *kept_up = !ret;2256 return ret;2257}22582259int remove_dir_recursively(struct strbuf *path, int flag)2260{2261 return remove_dir_recurse(path, flag, NULL);2262}22632264static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")22652266void setup_standard_excludes(struct dir_struct *dir)2267{2268 dir->exclude_per_dir = ".gitignore";22692270 /* core.excludefile defaulting to $XDG_HOME/git/ignore */2271 if (!excludes_file)2272 excludes_file = xdg_config_home("ignore");2273 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))2274 add_excludes_from_file_1(dir, excludes_file,2275 dir->untracked ? &dir->ss_excludes_file : NULL);22762277 /* per repository user preference */2278 if (startup_info->have_repository) {2279 const char *path = git_path_info_exclude();2280 if (!access_or_warn(path, R_OK, 0))2281 add_excludes_from_file_1(dir, path,2282 dir->untracked ? &dir->ss_info_exclude : NULL);2283 }2284}22852286int remove_path(const char *name)2287{2288 char *slash;22892290 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)2291 return -1;22922293 slash = strrchr(name, '/');2294 if (slash) {2295 char *dirs = xstrdup(name);2296 slash = dirs + (slash - name);2297 do {2298 *slash = '\0';2299 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));2300 free(dirs);2301 }2302 return 0;2303}23042305/*2306 * Frees memory within dir which was allocated for exclude lists and2307 * the exclude_stack. Does not free dir itself.2308 */2309void clear_directory(struct dir_struct *dir)2310{2311 int i, j;2312 struct exclude_list_group *group;2313 struct exclude_list *el;2314 struct exclude_stack *stk;23152316 for (i = EXC_CMDL; i <= EXC_FILE; i++) {2317 group = &dir->exclude_list_group[i];2318 for (j = 0; j < group->nr; j++) {2319 el = &group->el[j];2320 if (i == EXC_DIRS)2321 free((char *)el->src);2322 clear_exclude_list(el);2323 }2324 free(group->el);2325 }23262327 stk = dir->exclude_stack;2328 while (stk) {2329 struct exclude_stack *prev = stk->prev;2330 free(stk);2331 stk = prev;2332 }2333 strbuf_release(&dir->basebuf);2334}23352336struct ondisk_untracked_cache {2337 struct stat_data info_exclude_stat;2338 struct stat_data excludes_file_stat;2339 uint32_t dir_flags;2340 unsigned char info_exclude_sha1[20];2341 unsigned char excludes_file_sha1[20];2342 char exclude_per_dir[FLEX_ARRAY];2343};23442345#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)23462347struct write_data {2348 int index; /* number of written untracked_cache_dir */2349 struct ewah_bitmap *check_only; /* from untracked_cache_dir */2350 struct ewah_bitmap *valid; /* from untracked_cache_dir */2351 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */2352 struct strbuf out;2353 struct strbuf sb_stat;2354 struct strbuf sb_sha1;2355};23562357static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)2358{2359 to->sd_ctime.sec = htonl(from->sd_ctime.sec);2360 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);2361 to->sd_mtime.sec = htonl(from->sd_mtime.sec);2362 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);2363 to->sd_dev = htonl(from->sd_dev);2364 to->sd_ino = htonl(from->sd_ino);2365 to->sd_uid = htonl(from->sd_uid);2366 to->sd_gid = htonl(from->sd_gid);2367 to->sd_size = htonl(from->sd_size);2368}23692370static void write_one_dir(struct untracked_cache_dir *untracked,2371 struct write_data *wd)2372{2373 struct stat_data stat_data;2374 struct strbuf *out = &wd->out;2375 unsigned char intbuf[16];2376 unsigned int intlen, value;2377 int i = wd->index++;23782379 /*2380 * untracked_nr should be reset whenever valid is clear, but2381 * for safety..2382 */2383 if (!untracked->valid) {2384 untracked->untracked_nr = 0;2385 untracked->check_only = 0;2386 }23872388 if (untracked->check_only)2389 ewah_set(wd->check_only, i);2390 if (untracked->valid) {2391 ewah_set(wd->valid, i);2392 stat_data_to_disk(&stat_data, &untracked->stat_data);2393 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));2394 }2395 if (!is_null_sha1(untracked->exclude_sha1)) {2396 ewah_set(wd->sha1_valid, i);2397 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);2398 }23992400 intlen = encode_varint(untracked->untracked_nr, intbuf);2401 strbuf_add(out, intbuf, intlen);24022403 /* skip non-recurse directories */2404 for (i = 0, value = 0; i < untracked->dirs_nr; i++)2405 if (untracked->dirs[i]->recurse)2406 value++;2407 intlen = encode_varint(value, intbuf);2408 strbuf_add(out, intbuf, intlen);24092410 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);24112412 for (i = 0; i < untracked->untracked_nr; i++)2413 strbuf_add(out, untracked->untracked[i],2414 strlen(untracked->untracked[i]) + 1);24152416 for (i = 0; i < untracked->dirs_nr; i++)2417 if (untracked->dirs[i]->recurse)2418 write_one_dir(untracked->dirs[i], wd);2419}24202421void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)2422{2423 struct ondisk_untracked_cache *ouc;2424 struct write_data wd;2425 unsigned char varbuf[16];2426 int varint_len;2427 size_t len = strlen(untracked->exclude_per_dir);24282429 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);2430 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);2431 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);2432 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);2433 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);2434 ouc->dir_flags = htonl(untracked->dir_flags);24352436 varint_len = encode_varint(untracked->ident.len, varbuf);2437 strbuf_add(out, varbuf, varint_len);2438 strbuf_addbuf(out, &untracked->ident);24392440 strbuf_add(out, ouc, ouc_size(len));2441 free(ouc);2442 ouc = NULL;24432444 if (!untracked->root) {2445 varint_len = encode_varint(0, varbuf);2446 strbuf_add(out, varbuf, varint_len);2447 return;2448 }24492450 wd.index = 0;2451 wd.check_only = ewah_new();2452 wd.valid = ewah_new();2453 wd.sha1_valid = ewah_new();2454 strbuf_init(&wd.out, 1024);2455 strbuf_init(&wd.sb_stat, 1024);2456 strbuf_init(&wd.sb_sha1, 1024);2457 write_one_dir(untracked->root, &wd);24582459 varint_len = encode_varint(wd.index, varbuf);2460 strbuf_add(out, varbuf, varint_len);2461 strbuf_addbuf(out, &wd.out);2462 ewah_serialize_strbuf(wd.valid, out);2463 ewah_serialize_strbuf(wd.check_only, out);2464 ewah_serialize_strbuf(wd.sha1_valid, out);2465 strbuf_addbuf(out, &wd.sb_stat);2466 strbuf_addbuf(out, &wd.sb_sha1);2467 strbuf_addch(out, '\0'); /* safe guard for string lists */24682469 ewah_free(wd.valid);2470 ewah_free(wd.check_only);2471 ewah_free(wd.sha1_valid);2472 strbuf_release(&wd.out);2473 strbuf_release(&wd.sb_stat);2474 strbuf_release(&wd.sb_sha1);2475}24762477static void free_untracked(struct untracked_cache_dir *ucd)2478{2479 int i;2480 if (!ucd)2481 return;2482 for (i = 0; i < ucd->dirs_nr; i++)2483 free_untracked(ucd->dirs[i]);2484 for (i = 0; i < ucd->untracked_nr; i++)2485 free(ucd->untracked[i]);2486 free(ucd->untracked);2487 free(ucd->dirs);2488 free(ucd);2489}24902491void free_untracked_cache(struct untracked_cache *uc)2492{2493 if (uc)2494 free_untracked(uc->root);2495 free(uc);2496}24972498struct read_data {2499 int index;2500 struct untracked_cache_dir **ucd;2501 struct ewah_bitmap *check_only;2502 struct ewah_bitmap *valid;2503 struct ewah_bitmap *sha1_valid;2504 const unsigned char *data;2505 const unsigned char *end;2506};25072508static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)2509{2510 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);2511 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);2512 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);2513 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);2514 to->sd_dev = get_be32(&from->sd_dev);2515 to->sd_ino = get_be32(&from->sd_ino);2516 to->sd_uid = get_be32(&from->sd_uid);2517 to->sd_gid = get_be32(&from->sd_gid);2518 to->sd_size = get_be32(&from->sd_size);2519}25202521static int read_one_dir(struct untracked_cache_dir **untracked_,2522 struct read_data *rd)2523{2524 struct untracked_cache_dir ud, *untracked;2525 const unsigned char *next, *data = rd->data, *end = rd->end;2526 unsigned int value;2527 int i, len;25282529 memset(&ud, 0, sizeof(ud));25302531 next = data;2532 value = decode_varint(&next);2533 if (next > end)2534 return -1;2535 ud.recurse = 1;2536 ud.untracked_alloc = value;2537 ud.untracked_nr = value;2538 if (ud.untracked_nr)2539 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);2540 data = next;25412542 next = data;2543 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);2544 if (next > end)2545 return -1;2546 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);2547 data = next;25482549 len = strlen((const char *)data);2550 next = data + len + 1;2551 if (next > rd->end)2552 return -1;2553 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));2554 memcpy(untracked, &ud, sizeof(ud));2555 memcpy(untracked->name, data, len + 1);2556 data = next;25572558 for (i = 0; i < untracked->untracked_nr; i++) {2559 len = strlen((const char *)data);2560 next = data + len + 1;2561 if (next > rd->end)2562 return -1;2563 untracked->untracked[i] = xstrdup((const char*)data);2564 data = next;2565 }25662567 rd->ucd[rd->index++] = untracked;2568 rd->data = data;25692570 for (i = 0; i < untracked->dirs_nr; i++) {2571 len = read_one_dir(untracked->dirs + i, rd);2572 if (len < 0)2573 return -1;2574 }2575 return 0;2576}25772578static void set_check_only(size_t pos, void *cb)2579{2580 struct read_data *rd = cb;2581 struct untracked_cache_dir *ud = rd->ucd[pos];2582 ud->check_only = 1;2583}25842585static void read_stat(size_t pos, void *cb)2586{2587 struct read_data *rd = cb;2588 struct untracked_cache_dir *ud = rd->ucd[pos];2589 if (rd->data + sizeof(struct stat_data) > rd->end) {2590 rd->data = rd->end + 1;2591 return;2592 }2593 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);2594 rd->data += sizeof(struct stat_data);2595 ud->valid = 1;2596}25972598static void read_sha1(size_t pos, void *cb)2599{2600 struct read_data *rd = cb;2601 struct untracked_cache_dir *ud = rd->ucd[pos];2602 if (rd->data + 20 > rd->end) {2603 rd->data = rd->end + 1;2604 return;2605 }2606 hashcpy(ud->exclude_sha1, rd->data);2607 rd->data += 20;2608}26092610static void load_sha1_stat(struct sha1_stat *sha1_stat,2611 const struct stat_data *stat,2612 const unsigned char *sha1)2613{2614 stat_data_from_disk(&sha1_stat->stat, stat);2615 hashcpy(sha1_stat->sha1, sha1);2616 sha1_stat->valid = 1;2617}26182619struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)2620{2621 const struct ondisk_untracked_cache *ouc;2622 struct untracked_cache *uc;2623 struct read_data rd;2624 const unsigned char *next = data, *end = (const unsigned char *)data + sz;2625 const char *ident;2626 int ident_len, len;26272628 if (sz <= 1 || end[-1] != '\0')2629 return NULL;2630 end--;26312632 ident_len = decode_varint(&next);2633 if (next + ident_len > end)2634 return NULL;2635 ident = (const char *)next;2636 next += ident_len;26372638 ouc = (const struct ondisk_untracked_cache *)next;2639 if (next + ouc_size(0) > end)2640 return NULL;26412642 uc = xcalloc(1, sizeof(*uc));2643 strbuf_init(&uc->ident, ident_len);2644 strbuf_add(&uc->ident, ident, ident_len);2645 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,2646 ouc->info_exclude_sha1);2647 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,2648 ouc->excludes_file_sha1);2649 uc->dir_flags = get_be32(&ouc->dir_flags);2650 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);2651 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */2652 next += ouc_size(strlen(ouc->exclude_per_dir));2653 if (next >= end)2654 goto done2;26552656 len = decode_varint(&next);2657 if (next > end || len == 0)2658 goto done2;26592660 rd.valid = ewah_new();2661 rd.check_only = ewah_new();2662 rd.sha1_valid = ewah_new();2663 rd.data = next;2664 rd.end = end;2665 rd.index = 0;2666 ALLOC_ARRAY(rd.ucd, len);26672668 if (read_one_dir(&uc->root, &rd) || rd.index != len)2669 goto done;26702671 next = rd.data;2672 len = ewah_read_mmap(rd.valid, next, end - next);2673 if (len < 0)2674 goto done;26752676 next += len;2677 len = ewah_read_mmap(rd.check_only, next, end - next);2678 if (len < 0)2679 goto done;26802681 next += len;2682 len = ewah_read_mmap(rd.sha1_valid, next, end - next);2683 if (len < 0)2684 goto done;26852686 ewah_each_bit(rd.check_only, set_check_only, &rd);2687 rd.data = next + len;2688 ewah_each_bit(rd.valid, read_stat, &rd);2689 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);2690 next = rd.data;26912692done:2693 free(rd.ucd);2694 ewah_free(rd.valid);2695 ewah_free(rd.check_only);2696 ewah_free(rd.sha1_valid);2697done2:2698 if (next != end) {2699 free_untracked_cache(uc);2700 uc = NULL;2701 }2702 return uc;2703}27042705static void invalidate_one_directory(struct untracked_cache *uc,2706 struct untracked_cache_dir *ucd)2707{2708 uc->dir_invalidated++;2709 ucd->valid = 0;2710 ucd->untracked_nr = 0;2711}27122713/*2714 * Normally when an entry is added or removed from a directory,2715 * invalidating that directory is enough. No need to touch its2716 * ancestors. When a directory is shown as "foo/bar/" in git-status2717 * however, deleting or adding an entry may have cascading effect.2718 *2719 * Say the "foo/bar/file" has become untracked, we need to tell the2720 * untracked_cache_dir of "foo" that "bar/" is not an untracked2721 * directory any more (because "bar" is managed by foo as an untracked2722 * "file").2723 *2724 * Similarly, if "foo/bar/file" moves from untracked to tracked and it2725 * was the last untracked entry in the entire "foo", we should show2726 * "foo/" instead. Which means we have to invalidate past "bar" up to2727 * "foo".2728 *2729 * This function traverses all directories from root to leaf. If there2730 * is a chance of one of the above cases happening, we invalidate back2731 * to root. Otherwise we just invalidate the leaf. There may be a more2732 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to2733 * detect these cases and avoid unnecessary invalidation, for example,2734 * checking for the untracked entry named "bar/" in "foo", but for now2735 * stick to something safe and simple.2736 */2737static int invalidate_one_component(struct untracked_cache *uc,2738 struct untracked_cache_dir *dir,2739 const char *path, int len)2740{2741 const char *rest = strchr(path, '/');27422743 if (rest) {2744 int component_len = rest - path;2745 struct untracked_cache_dir *d =2746 lookup_untracked(uc, dir, path, component_len);2747 int ret =2748 invalidate_one_component(uc, d, rest + 1,2749 len - (component_len + 1));2750 if (ret)2751 invalidate_one_directory(uc, dir);2752 return ret;2753 }27542755 invalidate_one_directory(uc, dir);2756 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;2757}27582759void untracked_cache_invalidate_path(struct index_state *istate,2760 const char *path)2761{2762 if (!istate->untracked || !istate->untracked->root)2763 return;2764 invalidate_one_component(istate->untracked, istate->untracked->root,2765 path, strlen(path));2766}27672768void untracked_cache_remove_from_index(struct index_state *istate,2769 const char *path)2770{2771 untracked_cache_invalidate_path(istate, path);2772}27732774void untracked_cache_add_to_index(struct index_state *istate,2775 const char *path)2776{2777 untracked_cache_invalidate_path(istate, path);2778}27792780/* Update gitfile and core.worktree setting to connect work tree and git dir */2781void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)2782{2783 struct strbuf gitfile_sb = STRBUF_INIT;2784 struct strbuf cfg_sb = STRBUF_INIT;2785 struct strbuf rel_path = STRBUF_INIT;2786 char *git_dir, *work_tree;27872788 /* Prepare .git file */2789 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);2790 if (safe_create_leading_directories_const(gitfile_sb.buf))2791 die(_("could not create directories for %s"), gitfile_sb.buf);27922793 /* Prepare config file */2794 strbuf_addf(&cfg_sb, "%s/config", git_dir_);2795 if (safe_create_leading_directories_const(cfg_sb.buf))2796 die(_("could not create directories for %s"), cfg_sb.buf);27972798 git_dir = real_pathdup(git_dir_, 1);2799 work_tree = real_pathdup(work_tree_, 1);28002801 /* Write .git file */2802 write_file(gitfile_sb.buf, "gitdir: %s",2803 relative_path(git_dir, work_tree, &rel_path));2804 /* Update core.worktree setting */2805 git_config_set_in_file(cfg_sb.buf, "core.worktree",2806 relative_path(work_tree, git_dir, &rel_path));28072808 strbuf_release(&gitfile_sb);2809 strbuf_release(&cfg_sb);2810 strbuf_release(&rel_path);2811 free(work_tree);2812 free(git_dir);2813}28142815/*2816 * Migrate the git directory of the given path from old_git_dir to new_git_dir.2817 */2818void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)2819{2820 if (rename(old_git_dir, new_git_dir) < 0)2821 die_errno(_("could not migrate git directory from '%s' to '%s'"),2822 old_git_dir, new_git_dir);28232824 connect_work_tree_and_git_dir(path, new_git_dir);2825}