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, const char *path, int len); 52 53int fspathcmp(const char *a, const char *b) 54{ 55 return ignore_case ? strcasecmp(a, b) : strcmp(a, b); 56} 57 58int fspathncmp(const char *a, const char *b, size_t count) 59{ 60 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count); 61} 62 63int git_fnmatch(const struct pathspec_item *item, 64 const char *pattern, const char *string, 65 int prefix) 66{ 67 if (prefix > 0) { 68 if (ps_strncmp(item, pattern, string, prefix)) 69 return WM_NOMATCH; 70 pattern += prefix; 71 string += prefix; 72 } 73 if (item->flags & PATHSPEC_ONESTAR) { 74 int pattern_len = strlen(++pattern); 75 int string_len = strlen(string); 76 return string_len < pattern_len || 77 ps_strcmp(item, pattern, 78 string + string_len - pattern_len); 79 } 80 if (item->magic & PATHSPEC_GLOB) 81 return wildmatch(pattern, string, 82 WM_PATHNAME | 83 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0), 84 NULL); 85 else 86 /* wildmatch has not learned no FNM_PATHNAME mode yet */ 87 return wildmatch(pattern, string, 88 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0, 89 NULL); 90} 91 92static int fnmatch_icase_mem(const char *pattern, int patternlen, 93 const char *string, int stringlen, 94 int flags) 95{ 96 int match_status; 97 struct strbuf pat_buf = STRBUF_INIT; 98 struct strbuf str_buf = STRBUF_INIT; 99 const char *use_pat = pattern; 100 const char *use_str = string; 101 102 if (pattern[patternlen]) { 103 strbuf_add(&pat_buf, pattern, patternlen); 104 use_pat = pat_buf.buf; 105 } 106 if (string[stringlen]) { 107 strbuf_add(&str_buf, string, stringlen); 108 use_str = str_buf.buf; 109 } 110 111 if (ignore_case) 112 flags |= WM_CASEFOLD; 113 match_status = wildmatch(use_pat, use_str, flags, NULL); 114 115 strbuf_release(&pat_buf); 116 strbuf_release(&str_buf); 117 118 return match_status; 119} 120 121static size_t common_prefix_len(const struct pathspec *pathspec) 122{ 123 int n; 124 size_t max = 0; 125 126 /* 127 * ":(icase)path" is treated as a pathspec full of 128 * wildcard. In other words, only prefix is considered common 129 * prefix. If the pathspec is abc/foo abc/bar, running in 130 * subdir xyz, the common prefix is still xyz, not xuz/abc as 131 * in non-:(icase). 132 */ 133 GUARD_PATHSPEC(pathspec, 134 PATHSPEC_FROMTOP | 135 PATHSPEC_MAXDEPTH | 136 PATHSPEC_LITERAL | 137 PATHSPEC_GLOB | 138 PATHSPEC_ICASE | 139 PATHSPEC_EXCLUDE | 140 PATHSPEC_ATTR); 141 142 for (n = 0; n < pathspec->nr; n++) { 143 size_t i = 0, len = 0, item_len; 144 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE) 145 continue; 146 if (pathspec->items[n].magic & PATHSPEC_ICASE) 147 item_len = pathspec->items[n].prefix; 148 else 149 item_len = pathspec->items[n].nowildcard_len; 150 while (i < item_len && (n == 0 || i < max)) { 151 char c = pathspec->items[n].match[i]; 152 if (c != pathspec->items[0].match[i]) 153 break; 154 if (c == '/') 155 len = i + 1; 156 i++; 157 } 158 if (n == 0 || len < max) { 159 max = len; 160 if (!max) 161 break; 162 } 163 } 164 return max; 165} 166 167/* 168 * Returns a copy of the longest leading path common among all 169 * pathspecs. 170 */ 171char *common_prefix(const struct pathspec *pathspec) 172{ 173 unsigned long len = common_prefix_len(pathspec); 174 175 return len ? xmemdupz(pathspec->items[0].match, len) : NULL; 176} 177 178int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec) 179{ 180 const char *prefix; 181 size_t prefix_len; 182 183 /* 184 * Calculate common prefix for the pathspec, and 185 * use that to optimize the directory walk 186 */ 187 prefix_len = common_prefix_len(pathspec); 188 prefix = prefix_len ? pathspec->items[0].match : ""; 189 190 /* Read the directory and prune it */ 191 read_directory(dir, prefix, prefix_len, pathspec); 192 193 return prefix_len; 194} 195 196int within_depth(const char *name, int namelen, 197 int depth, int max_depth) 198{ 199 const char *cp = name, *cpe = name + namelen; 200 201 while (cp < cpe) { 202 if (*cp++ != '/') 203 continue; 204 depth++; 205 if (depth > max_depth) 206 return 0; 207 } 208 return 1; 209} 210 211#define DO_MATCH_EXCLUDE (1<<0) 212#define DO_MATCH_DIRECTORY (1<<1) 213#define DO_MATCH_SUBMODULE (1<<2) 214 215static int match_attrs(const char *name, int namelen, 216 const struct pathspec_item *item) 217{ 218 int i; 219 220 git_check_attr(name, item->attr_check); 221 for (i = 0; i < item->attr_match_nr; i++) { 222 const char *value; 223 int matched; 224 enum attr_match_mode match_mode; 225 226 value = item->attr_check->items[i].value; 227 match_mode = item->attr_match[i].match_mode; 228 229 if (ATTR_TRUE(value)) 230 matched = (match_mode == MATCH_SET); 231 else if (ATTR_FALSE(value)) 232 matched = (match_mode == MATCH_UNSET); 233 else if (ATTR_UNSET(value)) 234 matched = (match_mode == MATCH_UNSPECIFIED); 235 else 236 matched = (match_mode == MATCH_VALUE && 237 !strcmp(item->attr_match[i].value, value)); 238 if (!matched) 239 return 0; 240 } 241 242 return 1; 243} 244 245/* 246 * Does 'match' match the given name? 247 * A match is found if 248 * 249 * (1) the 'match' string is leading directory of 'name', or 250 * (2) the 'match' string is a wildcard and matches 'name', or 251 * (3) the 'match' string is exactly the same as 'name'. 252 * 253 * and the return value tells which case it was. 254 * 255 * It returns 0 when there is no match. 256 */ 257static int match_pathspec_item(const struct pathspec_item *item, int prefix, 258 const char *name, int namelen, unsigned flags) 259{ 260 /* name/namelen has prefix cut off by caller */ 261 const char *match = item->match + prefix; 262 int matchlen = item->len - prefix; 263 264 /* 265 * The normal call pattern is: 266 * 1. prefix = common_prefix_len(ps); 267 * 2. prune something, or fill_directory 268 * 3. match_pathspec() 269 * 270 * 'prefix' at #1 may be shorter than the command's prefix and 271 * it's ok for #2 to match extra files. Those extras will be 272 * trimmed at #3. 273 * 274 * Suppose the pathspec is 'foo' and '../bar' running from 275 * subdir 'xyz'. The common prefix at #1 will be empty, thanks 276 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The 277 * user does not want XYZ/foo, only the "foo" part should be 278 * case-insensitive. We need to filter out XYZ/foo here. In 279 * other words, we do not trust the caller on comparing the 280 * prefix part when :(icase) is involved. We do exact 281 * comparison ourselves. 282 * 283 * Normally the caller (common_prefix_len() in fact) does 284 * _exact_ matching on name[-prefix+1..-1] and we do not need 285 * to check that part. Be defensive and check it anyway, in 286 * case common_prefix_len is changed, or a new caller is 287 * introduced that does not use common_prefix_len. 288 * 289 * If the penalty turns out too high when prefix is really 290 * long, maybe change it to 291 * strncmp(match, name, item->prefix - prefix) 292 */ 293 if (item->prefix && (item->magic & PATHSPEC_ICASE) && 294 strncmp(item->match, name - prefix, item->prefix)) 295 return 0; 296 297 if (item->attr_match_nr && !match_attrs(name, namelen, item)) 298 return 0; 299 300 /* If the match was just the prefix, we matched */ 301 if (!*match) 302 return MATCHED_RECURSIVELY; 303 304 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) { 305 if (matchlen == namelen) 306 return MATCHED_EXACTLY; 307 308 if (match[matchlen-1] == '/' || name[matchlen] == '/') 309 return MATCHED_RECURSIVELY; 310 } else if ((flags & DO_MATCH_DIRECTORY) && 311 match[matchlen - 1] == '/' && 312 namelen == matchlen - 1 && 313 !ps_strncmp(item, match, name, namelen)) 314 return MATCHED_EXACTLY; 315 316 if (item->nowildcard_len < item->len && 317 !git_fnmatch(item, match, name, 318 item->nowildcard_len - prefix)) 319 return MATCHED_FNMATCH; 320 321 /* Perform checks to see if "name" is a super set of the pathspec */ 322 if (flags & DO_MATCH_SUBMODULE) { 323 /* name is a literal prefix of the pathspec */ 324 if ((namelen < matchlen) && 325 (match[namelen] == '/') && 326 !ps_strncmp(item, match, name, namelen)) 327 return MATCHED_RECURSIVELY; 328 329 /* name" doesn't match up to the first wild character */ 330 if (item->nowildcard_len < item->len && 331 ps_strncmp(item, match, name, 332 item->nowildcard_len - prefix)) 333 return 0; 334 335 /* 336 * Here is where we would perform a wildmatch to check if 337 * "name" can be matched as a directory (or a prefix) against 338 * the pathspec. Since wildmatch doesn't have this capability 339 * at the present we have to punt and say that it is a match, 340 * potentially returning a false positive 341 * The submodules themselves will be able to perform more 342 * accurate matching to determine if the pathspec matches. 343 */ 344 return MATCHED_RECURSIVELY; 345 } 346 347 return 0; 348} 349 350/* 351 * Given a name and a list of pathspecs, returns the nature of the 352 * closest (i.e. most specific) match of the name to any of the 353 * pathspecs. 354 * 355 * The caller typically calls this multiple times with the same 356 * pathspec and seen[] array but with different name/namelen 357 * (e.g. entries from the index) and is interested in seeing if and 358 * how each pathspec matches all the names it calls this function 359 * with. A mark is left in the seen[] array for each pathspec element 360 * indicating the closest type of match that element achieved, so if 361 * seen[n] remains zero after multiple invocations, that means the nth 362 * pathspec did not match any names, which could indicate that the 363 * user mistyped the nth pathspec. 364 */ 365static int do_match_pathspec(const struct pathspec *ps, 366 const char *name, int namelen, 367 int prefix, char *seen, 368 unsigned flags) 369{ 370 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE; 371 372 GUARD_PATHSPEC(ps, 373 PATHSPEC_FROMTOP | 374 PATHSPEC_MAXDEPTH | 375 PATHSPEC_LITERAL | 376 PATHSPEC_GLOB | 377 PATHSPEC_ICASE | 378 PATHSPEC_EXCLUDE | 379 PATHSPEC_ATTR); 380 381 if (!ps->nr) { 382 if (!ps->recursive || 383 !(ps->magic & PATHSPEC_MAXDEPTH) || 384 ps->max_depth == -1) 385 return MATCHED_RECURSIVELY; 386 387 if (within_depth(name, namelen, 0, ps->max_depth)) 388 return MATCHED_EXACTLY; 389 else 390 return 0; 391 } 392 393 name += prefix; 394 namelen -= prefix; 395 396 for (i = ps->nr - 1; i >= 0; i--) { 397 int how; 398 399 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) || 400 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE))) 401 continue; 402 403 if (seen && seen[i] == MATCHED_EXACTLY) 404 continue; 405 /* 406 * Make exclude patterns optional and never report 407 * "pathspec ':(exclude)foo' matches no files" 408 */ 409 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE) 410 seen[i] = MATCHED_FNMATCH; 411 how = match_pathspec_item(ps->items+i, prefix, name, 412 namelen, flags); 413 if (ps->recursive && 414 (ps->magic & PATHSPEC_MAXDEPTH) && 415 ps->max_depth != -1 && 416 how && how != MATCHED_FNMATCH) { 417 int len = ps->items[i].len; 418 if (name[len] == '/') 419 len++; 420 if (within_depth(name+len, namelen-len, 0, ps->max_depth)) 421 how = MATCHED_EXACTLY; 422 else 423 how = 0; 424 } 425 if (how) { 426 if (retval < how) 427 retval = how; 428 if (seen && seen[i] < how) 429 seen[i] = how; 430 } 431 } 432 return retval; 433} 434 435int match_pathspec(const struct pathspec *ps, 436 const char *name, int namelen, 437 int prefix, char *seen, int is_dir) 438{ 439 int positive, negative; 440 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0; 441 positive = do_match_pathspec(ps, name, namelen, 442 prefix, seen, flags); 443 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive) 444 return positive; 445 negative = do_match_pathspec(ps, name, namelen, 446 prefix, seen, 447 flags | DO_MATCH_EXCLUDE); 448 return negative ? 0 : positive; 449} 450 451/** 452 * Check if a submodule is a superset of the pathspec 453 */ 454int submodule_path_match(const struct pathspec *ps, 455 const char *submodule_name, 456 char *seen) 457{ 458 int matched = do_match_pathspec(ps, submodule_name, 459 strlen(submodule_name), 460 0, seen, 461 DO_MATCH_DIRECTORY | 462 DO_MATCH_SUBMODULE); 463 return matched; 464} 465 466int report_path_error(const char *ps_matched, 467 const struct pathspec *pathspec, 468 const char *prefix) 469{ 470 /* 471 * Make sure all pathspec matched; otherwise it is an error. 472 */ 473 int num, errors = 0; 474 for (num = 0; num < pathspec->nr; num++) { 475 int other, found_dup; 476 477 if (ps_matched[num]) 478 continue; 479 /* 480 * The caller might have fed identical pathspec 481 * twice. Do not barf on such a mistake. 482 * FIXME: parse_pathspec should have eliminated 483 * duplicate pathspec. 484 */ 485 for (found_dup = other = 0; 486 !found_dup && other < pathspec->nr; 487 other++) { 488 if (other == num || !ps_matched[other]) 489 continue; 490 if (!strcmp(pathspec->items[other].original, 491 pathspec->items[num].original)) 492 /* 493 * Ok, we have a match already. 494 */ 495 found_dup = 1; 496 } 497 if (found_dup) 498 continue; 499 500 error("pathspec '%s' did not match any file(s) known to git.", 501 pathspec->items[num].original); 502 errors++; 503 } 504 return errors; 505} 506 507/* 508 * Return the length of the "simple" part of a path match limiter. 509 */ 510int simple_length(const char *match) 511{ 512 int len = -1; 513 514 for (;;) { 515 unsigned char c = *match++; 516 len++; 517 if (c == '\0' || is_glob_special(c)) 518 return len; 519 } 520} 521 522int no_wildcard(const char *string) 523{ 524 return string[simple_length(string)] == '\0'; 525} 526 527void parse_exclude_pattern(const char **pattern, 528 int *patternlen, 529 unsigned *flags, 530 int *nowildcardlen) 531{ 532 const char *p = *pattern; 533 size_t i, len; 534 535 *flags = 0; 536 if (*p == '!') { 537 *flags |= EXC_FLAG_NEGATIVE; 538 p++; 539 } 540 len = strlen(p); 541 if (len && p[len - 1] == '/') { 542 len--; 543 *flags |= EXC_FLAG_MUSTBEDIR; 544 } 545 for (i = 0; i < len; i++) { 546 if (p[i] == '/') 547 break; 548 } 549 if (i == len) 550 *flags |= EXC_FLAG_NODIR; 551 *nowildcardlen = simple_length(p); 552 /* 553 * we should have excluded the trailing slash from 'p' too, 554 * but that's one more allocation. Instead just make sure 555 * nowildcardlen does not exceed real patternlen 556 */ 557 if (*nowildcardlen > len) 558 *nowildcardlen = len; 559 if (*p == '*' && no_wildcard(p + 1)) 560 *flags |= EXC_FLAG_ENDSWITH; 561 *pattern = p; 562 *patternlen = len; 563} 564 565void add_exclude(const char *string, const char *base, 566 int baselen, struct exclude_list *el, int srcpos) 567{ 568 struct exclude *x; 569 int patternlen; 570 unsigned flags; 571 int nowildcardlen; 572 573 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen); 574 if (flags & EXC_FLAG_MUSTBEDIR) { 575 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen); 576 } else { 577 x = xmalloc(sizeof(*x)); 578 x->pattern = string; 579 } 580 x->patternlen = patternlen; 581 x->nowildcardlen = nowildcardlen; 582 x->base = base; 583 x->baselen = baselen; 584 x->flags = flags; 585 x->srcpos = srcpos; 586 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc); 587 el->excludes[el->nr++] = x; 588 x->el = el; 589} 590 591static void *read_skip_worktree_file_from_index(const char *path, size_t *size, 592 struct sha1_stat *sha1_stat) 593{ 594 int pos, len; 595 unsigned long sz; 596 enum object_type type; 597 void *data; 598 599 len = strlen(path); 600 pos = index_name_pos(&the_index, path, len); 601 if (pos < 0) 602 return NULL; 603 if (!ce_skip_worktree(the_index.cache[pos])) 604 return NULL; 605 data = read_sha1_file(the_index.cache[pos]->oid.hash, &type, &sz); 606 if (!data || type != OBJ_BLOB) { 607 free(data); 608 return NULL; 609 } 610 *size = xsize_t(sz); 611 if (sha1_stat) { 612 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat)); 613 hashcpy(sha1_stat->sha1, the_index.cache[pos]->oid.hash); 614 } 615 return data; 616} 617 618/* 619 * Frees memory within el which was allocated for exclude patterns and 620 * the file buffer. Does not free el itself. 621 */ 622void clear_exclude_list(struct exclude_list *el) 623{ 624 int i; 625 626 for (i = 0; i < el->nr; i++) 627 free(el->excludes[i]); 628 free(el->excludes); 629 free(el->filebuf); 630 631 memset(el, 0, sizeof(*el)); 632} 633 634static void trim_trailing_spaces(char *buf) 635{ 636 char *p, *last_space = NULL; 637 638 for (p = buf; *p; p++) 639 switch (*p) { 640 case ' ': 641 if (!last_space) 642 last_space = p; 643 break; 644 case '\\': 645 p++; 646 if (!*p) 647 return; 648 /* fallthrough */ 649 default: 650 last_space = NULL; 651 } 652 653 if (last_space) 654 *last_space = '\0'; 655} 656 657/* 658 * Given a subdirectory name and "dir" of the current directory, 659 * search the subdir in "dir" and return it, or create a new one if it 660 * does not exist in "dir". 661 * 662 * If "name" has the trailing slash, it'll be excluded in the search. 663 */ 664static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc, 665 struct untracked_cache_dir *dir, 666 const char *name, int len) 667{ 668 int first, last; 669 struct untracked_cache_dir *d; 670 if (!dir) 671 return NULL; 672 if (len && name[len - 1] == '/') 673 len--; 674 first = 0; 675 last = dir->dirs_nr; 676 while (last > first) { 677 int cmp, next = (last + first) >> 1; 678 d = dir->dirs[next]; 679 cmp = strncmp(name, d->name, len); 680 if (!cmp && strlen(d->name) > len) 681 cmp = -1; 682 if (!cmp) 683 return d; 684 if (cmp < 0) { 685 last = next; 686 continue; 687 } 688 first = next+1; 689 } 690 691 uc->dir_created++; 692 FLEX_ALLOC_MEM(d, name, name, len); 693 694 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc); 695 memmove(dir->dirs + first + 1, dir->dirs + first, 696 (dir->dirs_nr - first) * sizeof(*dir->dirs)); 697 dir->dirs_nr++; 698 dir->dirs[first] = d; 699 return d; 700} 701 702static void do_invalidate_gitignore(struct untracked_cache_dir *dir) 703{ 704 int i; 705 dir->valid = 0; 706 dir->untracked_nr = 0; 707 for (i = 0; i < dir->dirs_nr; i++) 708 do_invalidate_gitignore(dir->dirs[i]); 709} 710 711static void invalidate_gitignore(struct untracked_cache *uc, 712 struct untracked_cache_dir *dir) 713{ 714 uc->gitignore_invalidated++; 715 do_invalidate_gitignore(dir); 716} 717 718static void invalidate_directory(struct untracked_cache *uc, 719 struct untracked_cache_dir *dir) 720{ 721 int i; 722 uc->dir_invalidated++; 723 dir->valid = 0; 724 dir->untracked_nr = 0; 725 for (i = 0; i < dir->dirs_nr; i++) 726 dir->dirs[i]->recurse = 0; 727} 728 729/* 730 * Given a file with name "fname", read it (either from disk, or from 731 * the index if "check_index" is non-zero), parse it and store the 732 * exclude rules in "el". 733 * 734 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill 735 * stat data from disk (only valid if add_excludes returns zero). If 736 * ss_valid is non-zero, "ss" must contain good value as input. 737 */ 738static int add_excludes(const char *fname, const char *base, int baselen, 739 struct exclude_list *el, int check_index, 740 struct sha1_stat *sha1_stat) 741{ 742 struct stat st; 743 int fd, i, lineno = 1; 744 size_t size = 0; 745 char *buf, *entry; 746 747 fd = open(fname, O_RDONLY); 748 if (fd < 0 || fstat(fd, &st) < 0) { 749 if (errno != ENOENT) 750 warn_on_inaccessible(fname); 751 if (0 <= fd) 752 close(fd); 753 if (!check_index || 754 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL) 755 return -1; 756 if (size == 0) { 757 free(buf); 758 return 0; 759 } 760 if (buf[size-1] != '\n') { 761 buf = xrealloc(buf, st_add(size, 1)); 762 buf[size++] = '\n'; 763 } 764 } else { 765 size = xsize_t(st.st_size); 766 if (size == 0) { 767 if (sha1_stat) { 768 fill_stat_data(&sha1_stat->stat, &st); 769 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN); 770 sha1_stat->valid = 1; 771 } 772 close(fd); 773 return 0; 774 } 775 buf = xmallocz(size); 776 if (read_in_full(fd, buf, size) != size) { 777 free(buf); 778 close(fd); 779 return -1; 780 } 781 buf[size++] = '\n'; 782 close(fd); 783 if (sha1_stat) { 784 int pos; 785 if (sha1_stat->valid && 786 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st)) 787 ; /* no content change, ss->sha1 still good */ 788 else if (check_index && 789 (pos = index_name_pos(&the_index, fname, strlen(fname))) >= 0 && 790 !ce_stage(the_index.cache[pos]) && 791 ce_uptodate(the_index.cache[pos]) && 792 !would_convert_to_git(fname)) 793 hashcpy(sha1_stat->sha1, 794 the_index.cache[pos]->oid.hash); 795 else 796 hash_sha1_file(buf, size, "blob", sha1_stat->sha1); 797 fill_stat_data(&sha1_stat->stat, &st); 798 sha1_stat->valid = 1; 799 } 800 } 801 802 el->filebuf = buf; 803 804 if (skip_utf8_bom(&buf, size)) 805 size -= buf - el->filebuf; 806 807 entry = buf; 808 809 for (i = 0; i < size; i++) { 810 if (buf[i] == '\n') { 811 if (entry != buf + i && entry[0] != '#') { 812 buf[i - (i && buf[i-1] == '\r')] = 0; 813 trim_trailing_spaces(entry); 814 add_exclude(entry, base, baselen, el, lineno); 815 } 816 lineno++; 817 entry = buf + i + 1; 818 } 819 } 820 return 0; 821} 822 823int add_excludes_from_file_to_list(const char *fname, const char *base, 824 int baselen, struct exclude_list *el, 825 int check_index) 826{ 827 return add_excludes(fname, base, baselen, el, check_index, NULL); 828} 829 830struct exclude_list *add_exclude_list(struct dir_struct *dir, 831 int group_type, const char *src) 832{ 833 struct exclude_list *el; 834 struct exclude_list_group *group; 835 836 group = &dir->exclude_list_group[group_type]; 837 ALLOC_GROW(group->el, group->nr + 1, group->alloc); 838 el = &group->el[group->nr++]; 839 memset(el, 0, sizeof(*el)); 840 el->src = src; 841 return el; 842} 843 844/* 845 * Used to set up core.excludesfile and .git/info/exclude lists. 846 */ 847static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname, 848 struct sha1_stat *sha1_stat) 849{ 850 struct exclude_list *el; 851 /* 852 * catch setup_standard_excludes() that's called before 853 * dir->untracked is assigned. That function behaves 854 * differently when dir->untracked is non-NULL. 855 */ 856 if (!dir->untracked) 857 dir->unmanaged_exclude_files++; 858 el = add_exclude_list(dir, EXC_FILE, fname); 859 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0) 860 die("cannot use %s as an exclude file", fname); 861} 862 863void add_excludes_from_file(struct dir_struct *dir, const char *fname) 864{ 865 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */ 866 add_excludes_from_file_1(dir, fname, NULL); 867} 868 869int match_basename(const char *basename, int basenamelen, 870 const char *pattern, int prefix, int patternlen, 871 unsigned flags) 872{ 873 if (prefix == patternlen) { 874 if (patternlen == basenamelen && 875 !fspathncmp(pattern, basename, basenamelen)) 876 return 1; 877 } else if (flags & EXC_FLAG_ENDSWITH) { 878 /* "*literal" matching against "fooliteral" */ 879 if (patternlen - 1 <= basenamelen && 880 !fspathncmp(pattern + 1, 881 basename + basenamelen - (patternlen - 1), 882 patternlen - 1)) 883 return 1; 884 } else { 885 if (fnmatch_icase_mem(pattern, patternlen, 886 basename, basenamelen, 887 0) == 0) 888 return 1; 889 } 890 return 0; 891} 892 893int match_pathname(const char *pathname, int pathlen, 894 const char *base, int baselen, 895 const char *pattern, int prefix, int patternlen, 896 unsigned flags) 897{ 898 const char *name; 899 int namelen; 900 901 /* 902 * match with FNM_PATHNAME; the pattern has base implicitly 903 * in front of it. 904 */ 905 if (*pattern == '/') { 906 pattern++; 907 patternlen--; 908 prefix--; 909 } 910 911 /* 912 * baselen does not count the trailing slash. base[] may or 913 * may not end with a trailing slash though. 914 */ 915 if (pathlen < baselen + 1 || 916 (baselen && pathname[baselen] != '/') || 917 fspathncmp(pathname, base, baselen)) 918 return 0; 919 920 namelen = baselen ? pathlen - baselen - 1 : pathlen; 921 name = pathname + pathlen - namelen; 922 923 if (prefix) { 924 /* 925 * if the non-wildcard part is longer than the 926 * remaining pathname, surely it cannot match. 927 */ 928 if (prefix > namelen) 929 return 0; 930 931 if (fspathncmp(pattern, name, prefix)) 932 return 0; 933 pattern += prefix; 934 patternlen -= prefix; 935 name += prefix; 936 namelen -= prefix; 937 938 /* 939 * If the whole pattern did not have a wildcard, 940 * then our prefix match is all we need; we 941 * do not need to call fnmatch at all. 942 */ 943 if (!patternlen && !namelen) 944 return 1; 945 } 946 947 return fnmatch_icase_mem(pattern, patternlen, 948 name, namelen, 949 WM_PATHNAME) == 0; 950} 951 952/* 953 * Scan the given exclude list in reverse to see whether pathname 954 * should be ignored. The first match (i.e. the last on the list), if 955 * any, determines the fate. Returns the exclude_list element which 956 * matched, or NULL for undecided. 957 */ 958static struct exclude *last_exclude_matching_from_list(const char *pathname, 959 int pathlen, 960 const char *basename, 961 int *dtype, 962 struct exclude_list *el) 963{ 964 struct exclude *exc = NULL; /* undecided */ 965 int i; 966 967 if (!el->nr) 968 return NULL; /* undefined */ 969 970 for (i = el->nr - 1; 0 <= i; i--) { 971 struct exclude *x = el->excludes[i]; 972 const char *exclude = x->pattern; 973 int prefix = x->nowildcardlen; 974 975 if (x->flags & EXC_FLAG_MUSTBEDIR) { 976 if (*dtype == DT_UNKNOWN) 977 *dtype = get_dtype(NULL, pathname, pathlen); 978 if (*dtype != DT_DIR) 979 continue; 980 } 981 982 if (x->flags & EXC_FLAG_NODIR) { 983 if (match_basename(basename, 984 pathlen - (basename - pathname), 985 exclude, prefix, x->patternlen, 986 x->flags)) { 987 exc = x; 988 break; 989 } 990 continue; 991 } 992 993 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/'); 994 if (match_pathname(pathname, pathlen, 995 x->base, x->baselen ? x->baselen - 1 : 0, 996 exclude, prefix, x->patternlen, x->flags)) { 997 exc = x; 998 break; 999 }1000 }1001 return exc;1002}10031004/*1005 * Scan the list and let the last match determine the fate.1006 * Return 1 for exclude, 0 for include and -1 for undecided.1007 */1008int is_excluded_from_list(const char *pathname,1009 int pathlen, const char *basename, int *dtype,1010 struct exclude_list *el)1011{1012 struct exclude *exclude;1013 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);1014 if (exclude)1015 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1016 return -1; /* undecided */1017}10181019static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,1020 const char *pathname, int pathlen, const char *basename,1021 int *dtype_p)1022{1023 int i, j;1024 struct exclude_list_group *group;1025 struct exclude *exclude;1026 for (i = EXC_CMDL; i <= EXC_FILE; i++) {1027 group = &dir->exclude_list_group[i];1028 for (j = group->nr - 1; j >= 0; j--) {1029 exclude = last_exclude_matching_from_list(1030 pathname, pathlen, basename, dtype_p,1031 &group->el[j]);1032 if (exclude)1033 return exclude;1034 }1035 }1036 return NULL;1037}10381039/*1040 * Loads the per-directory exclude list for the substring of base1041 * which has a char length of baselen.1042 */1043static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)1044{1045 struct exclude_list_group *group;1046 struct exclude_list *el;1047 struct exclude_stack *stk = NULL;1048 struct untracked_cache_dir *untracked;1049 int current;10501051 group = &dir->exclude_list_group[EXC_DIRS];10521053 /*1054 * Pop the exclude lists from the EXCL_DIRS exclude_list_group1055 * which originate from directories not in the prefix of the1056 * path being checked.1057 */1058 while ((stk = dir->exclude_stack) != NULL) {1059 if (stk->baselen <= baselen &&1060 !strncmp(dir->basebuf.buf, base, stk->baselen))1061 break;1062 el = &group->el[dir->exclude_stack->exclude_ix];1063 dir->exclude_stack = stk->prev;1064 dir->exclude = NULL;1065 free((char *)el->src); /* see strbuf_detach() below */1066 clear_exclude_list(el);1067 free(stk);1068 group->nr--;1069 }10701071 /* Skip traversing into sub directories if the parent is excluded */1072 if (dir->exclude)1073 return;10741075 /*1076 * Lazy initialization. All call sites currently just1077 * memset(dir, 0, sizeof(*dir)) before use. Changing all of1078 * them seems lots of work for little benefit.1079 */1080 if (!dir->basebuf.buf)1081 strbuf_init(&dir->basebuf, PATH_MAX);10821083 /* Read from the parent directories and push them down. */1084 current = stk ? stk->baselen : -1;1085 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);1086 if (dir->untracked)1087 untracked = stk ? stk->ucd : dir->untracked->root;1088 else1089 untracked = NULL;10901091 while (current < baselen) {1092 const char *cp;1093 struct sha1_stat sha1_stat;10941095 stk = xcalloc(1, sizeof(*stk));1096 if (current < 0) {1097 cp = base;1098 current = 0;1099 } else {1100 cp = strchr(base + current + 1, '/');1101 if (!cp)1102 die("oops in prep_exclude");1103 cp++;1104 untracked =1105 lookup_untracked(dir->untracked, untracked,1106 base + current,1107 cp - base - current);1108 }1109 stk->prev = dir->exclude_stack;1110 stk->baselen = cp - base;1111 stk->exclude_ix = group->nr;1112 stk->ucd = untracked;1113 el = add_exclude_list(dir, EXC_DIRS, NULL);1114 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);1115 assert(stk->baselen == dir->basebuf.len);11161117 /* Abort if the directory is excluded */1118 if (stk->baselen) {1119 int dt = DT_DIR;1120 dir->basebuf.buf[stk->baselen - 1] = 0;1121 dir->exclude = last_exclude_matching_from_lists(dir,1122 dir->basebuf.buf, stk->baselen - 1,1123 dir->basebuf.buf + current, &dt);1124 dir->basebuf.buf[stk->baselen - 1] = '/';1125 if (dir->exclude &&1126 dir->exclude->flags & EXC_FLAG_NEGATIVE)1127 dir->exclude = NULL;1128 if (dir->exclude) {1129 dir->exclude_stack = stk;1130 return;1131 }1132 }11331134 /* Try to read per-directory file */1135 hashclr(sha1_stat.sha1);1136 sha1_stat.valid = 0;1137 if (dir->exclude_per_dir &&1138 /*1139 * If we know that no files have been added in1140 * this directory (i.e. valid_cached_dir() has1141 * been executed and set untracked->valid) ..1142 */1143 (!untracked || !untracked->valid ||1144 /*1145 * .. and .gitignore does not exist before1146 * (i.e. null exclude_sha1). Then we can skip1147 * loading .gitignore, which would result in1148 * ENOENT anyway.1149 */1150 !is_null_sha1(untracked->exclude_sha1))) {1151 /*1152 * dir->basebuf gets reused by the traversal, but we1153 * need fname to remain unchanged to ensure the src1154 * member of each struct exclude correctly1155 * back-references its source file. Other invocations1156 * of add_exclude_list provide stable strings, so we1157 * strbuf_detach() and free() here in the caller.1158 */1159 struct strbuf sb = STRBUF_INIT;1160 strbuf_addbuf(&sb, &dir->basebuf);1161 strbuf_addstr(&sb, dir->exclude_per_dir);1162 el->src = strbuf_detach(&sb, NULL);1163 add_excludes(el->src, el->src, stk->baselen, el, 1,1164 untracked ? &sha1_stat : NULL);1165 }1166 /*1167 * NEEDSWORK: when untracked cache is enabled, prep_exclude()1168 * will first be called in valid_cached_dir() then maybe many1169 * times more in last_exclude_matching(). When the cache is1170 * used, last_exclude_matching() will not be called and1171 * reading .gitignore content will be a waste.1172 *1173 * So when it's called by valid_cached_dir() and we can get1174 * .gitignore SHA-1 from the index (i.e. .gitignore is not1175 * modified on work tree), we could delay reading the1176 * .gitignore content until we absolutely need it in1177 * last_exclude_matching(). Be careful about ignore rule1178 * order, though, if you do that.1179 */1180 if (untracked &&1181 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {1182 invalidate_gitignore(dir->untracked, untracked);1183 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);1184 }1185 dir->exclude_stack = stk;1186 current = stk->baselen;1187 }1188 strbuf_setlen(&dir->basebuf, baselen);1189}11901191/*1192 * Loads the exclude lists for the directory containing pathname, then1193 * scans all exclude lists to determine whether pathname is excluded.1194 * Returns the exclude_list element which matched, or NULL for1195 * undecided.1196 */1197struct exclude *last_exclude_matching(struct dir_struct *dir,1198 const char *pathname,1199 int *dtype_p)1200{1201 int pathlen = strlen(pathname);1202 const char *basename = strrchr(pathname, '/');1203 basename = (basename) ? basename+1 : pathname;12041205 prep_exclude(dir, pathname, basename-pathname);12061207 if (dir->exclude)1208 return dir->exclude;12091210 return last_exclude_matching_from_lists(dir, pathname, pathlen,1211 basename, dtype_p);1212}12131214/*1215 * Loads the exclude lists for the directory containing pathname, then1216 * scans all exclude lists to determine whether pathname is excluded.1217 * Returns 1 if true, otherwise 0.1218 */1219int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)1220{1221 struct exclude *exclude =1222 last_exclude_matching(dir, pathname, dtype_p);1223 if (exclude)1224 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1225 return 0;1226}12271228static struct dir_entry *dir_entry_new(const char *pathname, int len)1229{1230 struct dir_entry *ent;12311232 FLEX_ALLOC_MEM(ent, name, pathname, len);1233 ent->len = len;1234 return ent;1235}12361237static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)1238{1239 if (index_file_exists(&the_index, pathname, len, ignore_case))1240 return NULL;12411242 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);1243 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);1244}12451246struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)1247{1248 if (!index_name_is_other(&the_index, pathname, len))1249 return NULL;12501251 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);1252 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);1253}12541255enum exist_status {1256 index_nonexistent = 0,1257 index_directory,1258 index_gitdir1259};12601261/*1262 * Do not use the alphabetically sorted index to look up1263 * the directory name; instead, use the case insensitive1264 * directory hash.1265 */1266static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)1267{1268 struct cache_entry *ce;12691270 if (index_dir_exists(&the_index, dirname, len))1271 return index_directory;12721273 ce = index_file_exists(&the_index, dirname, len, ignore_case);1274 if (ce && S_ISGITLINK(ce->ce_mode))1275 return index_gitdir;12761277 return index_nonexistent;1278}12791280/*1281 * The index sorts alphabetically by entry name, which1282 * means that a gitlink sorts as '\0' at the end, while1283 * a directory (which is defined not as an entry, but as1284 * the files it contains) will sort with the '/' at the1285 * end.1286 */1287static enum exist_status directory_exists_in_index(const char *dirname, int len)1288{1289 int pos;12901291 if (ignore_case)1292 return directory_exists_in_index_icase(dirname, len);12931294 pos = index_name_pos(&the_index, dirname, len);1295 if (pos < 0)1296 pos = -pos-1;1297 while (pos < the_index.cache_nr) {1298 const struct cache_entry *ce = the_index.cache[pos++];1299 unsigned char endchar;13001301 if (strncmp(ce->name, dirname, len))1302 break;1303 endchar = ce->name[len];1304 if (endchar > '/')1305 break;1306 if (endchar == '/')1307 return index_directory;1308 if (!endchar && S_ISGITLINK(ce->ce_mode))1309 return index_gitdir;1310 }1311 return index_nonexistent;1312}13131314/*1315 * When we find a directory when traversing the filesystem, we1316 * have three distinct cases:1317 *1318 * - ignore it1319 * - see it as a directory1320 * - recurse into it1321 *1322 * and which one we choose depends on a combination of existing1323 * git index contents and the flags passed into the directory1324 * traversal routine.1325 *1326 * Case 1: If we *already* have entries in the index under that1327 * directory name, we always recurse into the directory to see1328 * all the files.1329 *1330 * Case 2: If we *already* have that directory name as a gitlink,1331 * we always continue to see it as a gitlink, regardless of whether1332 * there is an actual git directory there or not (it might not1333 * be checked out as a subproject!)1334 *1335 * Case 3: if we didn't have it in the index previously, we1336 * have a few sub-cases:1337 *1338 * (a) if "show_other_directories" is true, we show it as1339 * just a directory, unless "hide_empty_directories" is1340 * also true, in which case we need to check if it contains any1341 * untracked and / or ignored files.1342 * (b) if it looks like a git directory, and we don't have1343 * 'no_gitlinks' set we treat it as a gitlink, and show it1344 * as a directory.1345 * (c) otherwise, we recurse into it.1346 */1347static enum path_treatment treat_directory(struct dir_struct *dir,1348 struct untracked_cache_dir *untracked,1349 const char *dirname, int len, int baselen, int exclude,1350 const struct pathspec *pathspec)1351{1352 /* The "len-1" is to strip the final '/' */1353 switch (directory_exists_in_index(dirname, len-1)) {1354 case index_directory:1355 return path_recurse;13561357 case index_gitdir:1358 return path_none;13591360 case index_nonexistent:1361 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)1362 break;1363 if (!(dir->flags & DIR_NO_GITLINKS)) {1364 unsigned char sha1[20];1365 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)1366 return path_untracked;1367 }1368 return path_recurse;1369 }13701371 /* This is the "show_other_directories" case */13721373 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))1374 return exclude ? path_excluded : path_untracked;13751376 untracked = lookup_untracked(dir->untracked, untracked,1377 dirname + baselen, len - baselen);1378 return read_directory_recursive(dir, dirname, len,1379 untracked, 1, pathspec);1380}13811382/*1383 * This is an inexact early pruning of any recursive directory1384 * reading - if the path cannot possibly be in the pathspec,1385 * return true, and we'll skip it early.1386 */1387static int simplify_away(const char *path, int pathlen,1388 const struct pathspec *pathspec)1389{1390 int i;13911392 if (!pathspec || !pathspec->nr)1393 return 0;13941395 GUARD_PATHSPEC(pathspec,1396 PATHSPEC_FROMTOP |1397 PATHSPEC_MAXDEPTH |1398 PATHSPEC_LITERAL |1399 PATHSPEC_GLOB |1400 PATHSPEC_ICASE |1401 PATHSPEC_EXCLUDE |1402 PATHSPEC_ATTR);14031404 for (i = 0; i < pathspec->nr; i++) {1405 const struct pathspec_item *item = &pathspec->items[i];1406 int len = item->nowildcard_len;14071408 if (len > pathlen)1409 len = pathlen;1410 if (!ps_strncmp(item, item->match, path, len))1411 return 0;1412 }14131414 return 1;1415}14161417/*1418 * This function tells us whether an excluded path matches a1419 * list of "interesting" pathspecs. That is, whether a path matched1420 * by any of the pathspecs could possibly be ignored by excluding1421 * the specified path. This can happen if:1422 *1423 * 1. the path is mentioned explicitly in the pathspec1424 *1425 * 2. the path is a directory prefix of some element in the1426 * pathspec1427 */1428static int exclude_matches_pathspec(const char *path, int pathlen,1429 const struct pathspec *pathspec)1430{1431 int i;14321433 if (!pathspec || !pathspec->nr)1434 return 0;14351436 GUARD_PATHSPEC(pathspec,1437 PATHSPEC_FROMTOP |1438 PATHSPEC_MAXDEPTH |1439 PATHSPEC_LITERAL |1440 PATHSPEC_GLOB |1441 PATHSPEC_ICASE |1442 PATHSPEC_EXCLUDE);14431444 for (i = 0; i < pathspec->nr; i++) {1445 const struct pathspec_item *item = &pathspec->items[i];1446 int len = item->nowildcard_len;14471448 if (len == pathlen &&1449 !ps_strncmp(item, item->match, path, pathlen))1450 return 1;1451 if (len > pathlen &&1452 item->match[pathlen] == '/' &&1453 !ps_strncmp(item, item->match, path, pathlen))1454 return 1;1455 }1456 return 0;1457}14581459static int get_index_dtype(const char *path, int len)1460{1461 int pos;1462 const struct cache_entry *ce;14631464 ce = index_file_exists(&the_index, path, len, 0);1465 if (ce) {1466 if (!ce_uptodate(ce))1467 return DT_UNKNOWN;1468 if (S_ISGITLINK(ce->ce_mode))1469 return DT_DIR;1470 /*1471 * Nobody actually cares about the1472 * difference between DT_LNK and DT_REG1473 */1474 return DT_REG;1475 }14761477 /* Try to look it up as a directory */1478 pos = index_name_pos(&the_index, path, len);1479 if (pos >= 0)1480 return DT_UNKNOWN;1481 pos = -pos-1;1482 while (pos < the_index.cache_nr) {1483 ce = the_index.cache[pos++];1484 if (strncmp(ce->name, path, len))1485 break;1486 if (ce->name[len] > '/')1487 break;1488 if (ce->name[len] < '/')1489 continue;1490 if (!ce_uptodate(ce))1491 break; /* continue? */1492 return DT_DIR;1493 }1494 return DT_UNKNOWN;1495}14961497static int get_dtype(struct dirent *de, const char *path, int len)1498{1499 int dtype = de ? DTYPE(de) : DT_UNKNOWN;1500 struct stat st;15011502 if (dtype != DT_UNKNOWN)1503 return dtype;1504 dtype = get_index_dtype(path, len);1505 if (dtype != DT_UNKNOWN)1506 return dtype;1507 if (lstat(path, &st))1508 return dtype;1509 if (S_ISREG(st.st_mode))1510 return DT_REG;1511 if (S_ISDIR(st.st_mode))1512 return DT_DIR;1513 if (S_ISLNK(st.st_mode))1514 return DT_LNK;1515 return dtype;1516}15171518static enum path_treatment treat_one_path(struct dir_struct *dir,1519 struct untracked_cache_dir *untracked,1520 struct strbuf *path,1521 int baselen,1522 const struct pathspec *pathspec,1523 int dtype, struct dirent *de)1524{1525 int exclude;1526 int has_path_in_index = !!index_file_exists(&the_index, path->buf, path->len, ignore_case);15271528 if (dtype == DT_UNKNOWN)1529 dtype = get_dtype(de, path->buf, path->len);15301531 /* Always exclude indexed files */1532 if (dtype != DT_DIR && has_path_in_index)1533 return path_none;15341535 /*1536 * When we are looking at a directory P in the working tree,1537 * there are three cases:1538 *1539 * (1) P exists in the index. Everything inside the directory P in1540 * the working tree needs to go when P is checked out from the1541 * index.1542 *1543 * (2) P does not exist in the index, but there is P/Q in the index.1544 * We know P will stay a directory when we check out the contents1545 * of the index, but we do not know yet if there is a directory1546 * P/Q in the working tree to be killed, so we need to recurse.1547 *1548 * (3) P does not exist in the index, and there is no P/Q in the index1549 * to require P to be a directory, either. Only in this case, we1550 * know that everything inside P will not be killed without1551 * recursing.1552 */1553 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&1554 (dtype == DT_DIR) &&1555 !has_path_in_index &&1556 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))1557 return path_none;15581559 exclude = is_excluded(dir, path->buf, &dtype);15601561 /*1562 * Excluded? If we don't explicitly want to show1563 * ignored files, ignore it1564 */1565 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))1566 return path_excluded;15671568 switch (dtype) {1569 default:1570 return path_none;1571 case DT_DIR:1572 strbuf_addch(path, '/');1573 return treat_directory(dir, untracked, path->buf, path->len,1574 baselen, exclude, pathspec);1575 case DT_REG:1576 case DT_LNK:1577 return exclude ? path_excluded : path_untracked;1578 }1579}15801581static enum path_treatment treat_path_fast(struct dir_struct *dir,1582 struct untracked_cache_dir *untracked,1583 struct cached_dir *cdir,1584 struct strbuf *path,1585 int baselen,1586 const struct pathspec *pathspec)1587{1588 strbuf_setlen(path, baselen);1589 if (!cdir->ucd) {1590 strbuf_addstr(path, cdir->file);1591 return path_untracked;1592 }1593 strbuf_addstr(path, cdir->ucd->name);1594 /* treat_one_path() does this before it calls treat_directory() */1595 strbuf_complete(path, '/');1596 if (cdir->ucd->check_only)1597 /*1598 * check_only is set as a result of treat_directory() getting1599 * to its bottom. Verify again the same set of directories1600 * with check_only set.1601 */1602 return read_directory_recursive(dir, path->buf, path->len,1603 cdir->ucd, 1, pathspec);1604 /*1605 * We get path_recurse in the first run when1606 * directory_exists_in_index() returns index_nonexistent. We1607 * are sure that new changes in the index does not impact the1608 * outcome. Return now.1609 */1610 return path_recurse;1611}16121613static enum path_treatment treat_path(struct dir_struct *dir,1614 struct untracked_cache_dir *untracked,1615 struct cached_dir *cdir,1616 struct strbuf *path,1617 int baselen,1618 const struct pathspec *pathspec)1619{1620 int dtype;1621 struct dirent *de = cdir->de;16221623 if (!de)1624 return treat_path_fast(dir, untracked, cdir, path,1625 baselen, pathspec);1626 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))1627 return path_none;1628 strbuf_setlen(path, baselen);1629 strbuf_addstr(path, de->d_name);1630 if (simplify_away(path->buf, path->len, pathspec))1631 return path_none;16321633 dtype = DTYPE(de);1634 return treat_one_path(dir, untracked, path, baselen, pathspec, dtype, de);1635}16361637static void add_untracked(struct untracked_cache_dir *dir, const char *name)1638{1639 if (!dir)1640 return;1641 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,1642 dir->untracked_alloc);1643 dir->untracked[dir->untracked_nr++] = xstrdup(name);1644}16451646static int valid_cached_dir(struct dir_struct *dir,1647 struct untracked_cache_dir *untracked,1648 struct strbuf *path,1649 int check_only)1650{1651 struct stat st;16521653 if (!untracked)1654 return 0;16551656 if (stat(path->len ? path->buf : ".", &st)) {1657 invalidate_directory(dir->untracked, untracked);1658 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));1659 return 0;1660 }1661 if (!untracked->valid ||1662 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {1663 if (untracked->valid)1664 invalidate_directory(dir->untracked, untracked);1665 fill_stat_data(&untracked->stat_data, &st);1666 return 0;1667 }16681669 if (untracked->check_only != !!check_only) {1670 invalidate_directory(dir->untracked, untracked);1671 return 0;1672 }16731674 /*1675 * prep_exclude will be called eventually on this directory,1676 * but it's called much later in last_exclude_matching(). We1677 * need it now to determine the validity of the cache for this1678 * path. The next calls will be nearly no-op, the way1679 * prep_exclude() is designed.1680 */1681 if (path->len && path->buf[path->len - 1] != '/') {1682 strbuf_addch(path, '/');1683 prep_exclude(dir, path->buf, path->len);1684 strbuf_setlen(path, path->len - 1);1685 } else1686 prep_exclude(dir, path->buf, path->len);16871688 /* hopefully prep_exclude() haven't invalidated this entry... */1689 return untracked->valid;1690}16911692static int open_cached_dir(struct cached_dir *cdir,1693 struct dir_struct *dir,1694 struct untracked_cache_dir *untracked,1695 struct strbuf *path,1696 int check_only)1697{1698 memset(cdir, 0, sizeof(*cdir));1699 cdir->untracked = untracked;1700 if (valid_cached_dir(dir, untracked, path, check_only))1701 return 0;1702 cdir->fdir = opendir(path->len ? path->buf : ".");1703 if (dir->untracked)1704 dir->untracked->dir_opened++;1705 if (!cdir->fdir)1706 return -1;1707 return 0;1708}17091710static int read_cached_dir(struct cached_dir *cdir)1711{1712 if (cdir->fdir) {1713 cdir->de = readdir(cdir->fdir);1714 if (!cdir->de)1715 return -1;1716 return 0;1717 }1718 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {1719 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];1720 if (!d->recurse) {1721 cdir->nr_dirs++;1722 continue;1723 }1724 cdir->ucd = d;1725 cdir->nr_dirs++;1726 return 0;1727 }1728 cdir->ucd = NULL;1729 if (cdir->nr_files < cdir->untracked->untracked_nr) {1730 struct untracked_cache_dir *d = cdir->untracked;1731 cdir->file = d->untracked[cdir->nr_files++];1732 return 0;1733 }1734 return -1;1735}17361737static void close_cached_dir(struct cached_dir *cdir)1738{1739 if (cdir->fdir)1740 closedir(cdir->fdir);1741 /*1742 * We have gone through this directory and found no untracked1743 * entries. Mark it valid.1744 */1745 if (cdir->untracked) {1746 cdir->untracked->valid = 1;1747 cdir->untracked->recurse = 1;1748 }1749}17501751/*1752 * Read a directory tree. We currently ignore anything but1753 * directories, regular files and symlinks. That's because git1754 * doesn't handle them at all yet. Maybe that will change some1755 * day.1756 *1757 * Also, we ignore the name ".git" (even if it is not a directory).1758 * That likely will not change.1759 *1760 * Returns the most significant path_treatment value encountered in the scan.1761 */1762static enum path_treatment read_directory_recursive(struct dir_struct *dir,1763 const char *base, int baselen,1764 struct untracked_cache_dir *untracked, int check_only,1765 const struct pathspec *pathspec)1766{1767 struct cached_dir cdir;1768 enum path_treatment state, subdir_state, dir_state = path_none;1769 struct strbuf path = STRBUF_INIT;17701771 strbuf_add(&path, base, baselen);17721773 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))1774 goto out;17751776 if (untracked)1777 untracked->check_only = !!check_only;17781779 while (!read_cached_dir(&cdir)) {1780 /* check how the file or directory should be treated */1781 state = treat_path(dir, untracked, &cdir, &path,1782 baselen, pathspec);17831784 if (state > dir_state)1785 dir_state = state;17861787 /* recurse into subdir if instructed by treat_path */1788 if (state == path_recurse) {1789 struct untracked_cache_dir *ud;1790 ud = lookup_untracked(dir->untracked, untracked,1791 path.buf + baselen,1792 path.len - baselen);1793 subdir_state =1794 read_directory_recursive(dir, path.buf,1795 path.len, ud,1796 check_only, pathspec);1797 if (subdir_state > dir_state)1798 dir_state = subdir_state;1799 }18001801 if (check_only) {1802 /* abort early if maximum state has been reached */1803 if (dir_state == path_untracked) {1804 if (cdir.fdir)1805 add_untracked(untracked, path.buf + baselen);1806 break;1807 }1808 /* skip the dir_add_* part */1809 continue;1810 }18111812 /* add the path to the appropriate result list */1813 switch (state) {1814 case path_excluded:1815 if (dir->flags & DIR_SHOW_IGNORED)1816 dir_add_name(dir, path.buf, path.len);1817 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||1818 ((dir->flags & DIR_COLLECT_IGNORED) &&1819 exclude_matches_pathspec(path.buf, path.len,1820 pathspec)))1821 dir_add_ignored(dir, path.buf, path.len);1822 break;18231824 case path_untracked:1825 if (dir->flags & DIR_SHOW_IGNORED)1826 break;1827 dir_add_name(dir, path.buf, path.len);1828 if (cdir.fdir)1829 add_untracked(untracked, path.buf + baselen);1830 break;18311832 default:1833 break;1834 }1835 }1836 close_cached_dir(&cdir);1837 out:1838 strbuf_release(&path);18391840 return dir_state;1841}18421843static int cmp_name(const void *p1, const void *p2)1844{1845 const struct dir_entry *e1 = *(const struct dir_entry **)p1;1846 const struct dir_entry *e2 = *(const struct dir_entry **)p2;18471848 return name_compare(e1->name, e1->len, e2->name, e2->len);1849}18501851static int treat_leading_path(struct dir_struct *dir,1852 const char *path, int len,1853 const struct pathspec *pathspec)1854{1855 struct strbuf sb = STRBUF_INIT;1856 int baselen, rc = 0;1857 const char *cp;1858 int old_flags = dir->flags;18591860 while (len && path[len - 1] == '/')1861 len--;1862 if (!len)1863 return 1;1864 baselen = 0;1865 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;1866 while (1) {1867 cp = path + baselen + !!baselen;1868 cp = memchr(cp, '/', path + len - cp);1869 if (!cp)1870 baselen = len;1871 else1872 baselen = cp - path;1873 strbuf_setlen(&sb, 0);1874 strbuf_add(&sb, path, baselen);1875 if (!is_directory(sb.buf))1876 break;1877 if (simplify_away(sb.buf, sb.len, pathspec))1878 break;1879 if (treat_one_path(dir, NULL, &sb, baselen, pathspec,1880 DT_DIR, NULL) == path_none)1881 break; /* do not recurse into it */1882 if (len <= baselen) {1883 rc = 1;1884 break; /* finished checking */1885 }1886 }1887 strbuf_release(&sb);1888 dir->flags = old_flags;1889 return rc;1890}18911892static const char *get_ident_string(void)1893{1894 static struct strbuf sb = STRBUF_INIT;1895 struct utsname uts;18961897 if (sb.len)1898 return sb.buf;1899 if (uname(&uts) < 0)1900 die_errno(_("failed to get kernel name and information"));1901 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),1902 uts.sysname);1903 return sb.buf;1904}19051906static int ident_in_untracked(const struct untracked_cache *uc)1907{1908 /*1909 * Previous git versions may have saved many NUL separated1910 * strings in the "ident" field, but it is insane to manage1911 * many locations, so just take care of the first one.1912 */19131914 return !strcmp(uc->ident.buf, get_ident_string());1915}19161917static void set_untracked_ident(struct untracked_cache *uc)1918{1919 strbuf_reset(&uc->ident);1920 strbuf_addstr(&uc->ident, get_ident_string());19211922 /*1923 * This strbuf used to contain a list of NUL separated1924 * strings, so save NUL too for backward compatibility.1925 */1926 strbuf_addch(&uc->ident, 0);1927}19281929static void new_untracked_cache(struct index_state *istate)1930{1931 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));1932 strbuf_init(&uc->ident, 100);1933 uc->exclude_per_dir = ".gitignore";1934 /* should be the same flags used by git-status */1935 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;1936 set_untracked_ident(uc);1937 istate->untracked = uc;1938 istate->cache_changed |= UNTRACKED_CHANGED;1939}19401941void add_untracked_cache(struct index_state *istate)1942{1943 if (!istate->untracked) {1944 new_untracked_cache(istate);1945 } else {1946 if (!ident_in_untracked(istate->untracked)) {1947 free_untracked_cache(istate->untracked);1948 new_untracked_cache(istate);1949 }1950 }1951}19521953void remove_untracked_cache(struct index_state *istate)1954{1955 if (istate->untracked) {1956 free_untracked_cache(istate->untracked);1957 istate->untracked = NULL;1958 istate->cache_changed |= UNTRACKED_CHANGED;1959 }1960}19611962static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,1963 int base_len,1964 const struct pathspec *pathspec)1965{1966 struct untracked_cache_dir *root;19671968 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))1969 return NULL;19701971 /*1972 * We only support $GIT_DIR/info/exclude and core.excludesfile1973 * as the global ignore rule files. Any other additions1974 * (e.g. from command line) invalidate the cache. This1975 * condition also catches running setup_standard_excludes()1976 * before setting dir->untracked!1977 */1978 if (dir->unmanaged_exclude_files)1979 return NULL;19801981 /*1982 * Optimize for the main use case only: whole-tree git1983 * status. More work involved in treat_leading_path() if we1984 * use cache on just a subset of the worktree. pathspec1985 * support could make the matter even worse.1986 */1987 if (base_len || (pathspec && pathspec->nr))1988 return NULL;19891990 /* Different set of flags may produce different results */1991 if (dir->flags != dir->untracked->dir_flags ||1992 /*1993 * See treat_directory(), case index_nonexistent. Without1994 * this flag, we may need to also cache .git file content1995 * for the resolve_gitlink_ref() call, which we don't.1996 */1997 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||1998 /* We don't support collecting ignore files */1999 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |2000 DIR_COLLECT_IGNORED)))2001 return NULL;20022003 /*2004 * If we use .gitignore in the cache and now you change it to2005 * .gitexclude, everything will go wrong.2006 */2007 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&2008 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))2009 return NULL;20102011 /*2012 * EXC_CMDL is not considered in the cache. If people set it,2013 * skip the cache.2014 */2015 if (dir->exclude_list_group[EXC_CMDL].nr)2016 return NULL;20172018 if (!ident_in_untracked(dir->untracked)) {2019 warning(_("Untracked cache is disabled on this system or location."));2020 return NULL;2021 }20222023 if (!dir->untracked->root) {2024 const int len = sizeof(*dir->untracked->root);2025 dir->untracked->root = xmalloc(len);2026 memset(dir->untracked->root, 0, len);2027 }20282029 /* Validate $GIT_DIR/info/exclude and core.excludesfile */2030 root = dir->untracked->root;2031 if (hashcmp(dir->ss_info_exclude.sha1,2032 dir->untracked->ss_info_exclude.sha1)) {2033 invalidate_gitignore(dir->untracked, root);2034 dir->untracked->ss_info_exclude = dir->ss_info_exclude;2035 }2036 if (hashcmp(dir->ss_excludes_file.sha1,2037 dir->untracked->ss_excludes_file.sha1)) {2038 invalidate_gitignore(dir->untracked, root);2039 dir->untracked->ss_excludes_file = dir->ss_excludes_file;2040 }20412042 /* Make sure this directory is not dropped out at saving phase */2043 root->recurse = 1;2044 return root;2045}20462047int read_directory(struct dir_struct *dir, const char *path,2048 int len, const struct pathspec *pathspec)2049{2050 struct untracked_cache_dir *untracked;20512052 if (has_symlink_leading_path(path, len))2053 return dir->nr;20542055 untracked = validate_untracked_cache(dir, len, pathspec);2056 if (!untracked)2057 /*2058 * make sure untracked cache code path is disabled,2059 * e.g. prep_exclude()2060 */2061 dir->untracked = NULL;2062 if (!len || treat_leading_path(dir, path, len, pathspec))2063 read_directory_recursive(dir, path, len, untracked, 0, pathspec);2064 QSORT(dir->entries, dir->nr, cmp_name);2065 QSORT(dir->ignored, dir->ignored_nr, cmp_name);2066 if (dir->untracked) {2067 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);2068 trace_printf_key(&trace_untracked_stats,2069 "node creation: %u\n"2070 "gitignore invalidation: %u\n"2071 "directory invalidation: %u\n"2072 "opendir: %u\n",2073 dir->untracked->dir_created,2074 dir->untracked->gitignore_invalidated,2075 dir->untracked->dir_invalidated,2076 dir->untracked->dir_opened);2077 if (dir->untracked == the_index.untracked &&2078 (dir->untracked->dir_opened ||2079 dir->untracked->gitignore_invalidated ||2080 dir->untracked->dir_invalidated))2081 the_index.cache_changed |= UNTRACKED_CHANGED;2082 if (dir->untracked != the_index.untracked) {2083 free(dir->untracked);2084 dir->untracked = NULL;2085 }2086 }2087 return dir->nr;2088}20892090int file_exists(const char *f)2091{2092 struct stat sb;2093 return lstat(f, &sb) == 0;2094}20952096static int cmp_icase(char a, char b)2097{2098 if (a == b)2099 return 0;2100 if (ignore_case)2101 return toupper(a) - toupper(b);2102 return a - b;2103}21042105/*2106 * Given two normalized paths (a trailing slash is ok), if subdir is2107 * outside dir, return -1. Otherwise return the offset in subdir that2108 * can be used as relative path to dir.2109 */2110int dir_inside_of(const char *subdir, const char *dir)2111{2112 int offset = 0;21132114 assert(dir && subdir && *dir && *subdir);21152116 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {2117 dir++;2118 subdir++;2119 offset++;2120 }21212122 /* hel[p]/me vs hel[l]/yeah */2123 if (*dir && *subdir)2124 return -1;21252126 if (!*subdir)2127 return !*dir ? offset : -1; /* same dir */21282129 /* foo/[b]ar vs foo/[] */2130 if (is_dir_sep(dir[-1]))2131 return is_dir_sep(subdir[-1]) ? offset : -1;21322133 /* foo[/]bar vs foo[] */2134 return is_dir_sep(*subdir) ? offset + 1 : -1;2135}21362137int is_inside_dir(const char *dir)2138{2139 char *cwd;2140 int rc;21412142 if (!dir)2143 return 0;21442145 cwd = xgetcwd();2146 rc = (dir_inside_of(cwd, dir) >= 0);2147 free(cwd);2148 return rc;2149}21502151int is_empty_dir(const char *path)2152{2153 DIR *dir = opendir(path);2154 struct dirent *e;2155 int ret = 1;21562157 if (!dir)2158 return 0;21592160 while ((e = readdir(dir)) != NULL)2161 if (!is_dot_or_dotdot(e->d_name)) {2162 ret = 0;2163 break;2164 }21652166 closedir(dir);2167 return ret;2168}21692170static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)2171{2172 DIR *dir;2173 struct dirent *e;2174 int ret = 0, original_len = path->len, len, kept_down = 0;2175 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);2176 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);2177 unsigned char submodule_head[20];21782179 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&2180 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {2181 /* Do not descend and nuke a nested git work tree. */2182 if (kept_up)2183 *kept_up = 1;2184 return 0;2185 }21862187 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;2188 dir = opendir(path->buf);2189 if (!dir) {2190 if (errno == ENOENT)2191 return keep_toplevel ? -1 : 0;2192 else if (errno == EACCES && !keep_toplevel)2193 /*2194 * An empty dir could be removable even if it2195 * is unreadable:2196 */2197 return rmdir(path->buf);2198 else2199 return -1;2200 }2201 strbuf_complete(path, '/');22022203 len = path->len;2204 while ((e = readdir(dir)) != NULL) {2205 struct stat st;2206 if (is_dot_or_dotdot(e->d_name))2207 continue;22082209 strbuf_setlen(path, len);2210 strbuf_addstr(path, e->d_name);2211 if (lstat(path->buf, &st)) {2212 if (errno == ENOENT)2213 /*2214 * file disappeared, which is what we2215 * wanted anyway2216 */2217 continue;2218 /* fall thru */2219 } else if (S_ISDIR(st.st_mode)) {2220 if (!remove_dir_recurse(path, flag, &kept_down))2221 continue; /* happy */2222 } else if (!only_empty &&2223 (!unlink(path->buf) || errno == ENOENT)) {2224 continue; /* happy, too */2225 }22262227 /* path too long, stat fails, or non-directory still exists */2228 ret = -1;2229 break;2230 }2231 closedir(dir);22322233 strbuf_setlen(path, original_len);2234 if (!ret && !keep_toplevel && !kept_down)2235 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;2236 else if (kept_up)2237 /*2238 * report the uplevel that it is not an error that we2239 * did not rmdir() our directory.2240 */2241 *kept_up = !ret;2242 return ret;2243}22442245int remove_dir_recursively(struct strbuf *path, int flag)2246{2247 return remove_dir_recurse(path, flag, NULL);2248}22492250static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")22512252void setup_standard_excludes(struct dir_struct *dir)2253{2254 dir->exclude_per_dir = ".gitignore";22552256 /* core.excludefile defaulting to $XDG_HOME/git/ignore */2257 if (!excludes_file)2258 excludes_file = xdg_config_home("ignore");2259 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))2260 add_excludes_from_file_1(dir, excludes_file,2261 dir->untracked ? &dir->ss_excludes_file : NULL);22622263 /* per repository user preference */2264 if (startup_info->have_repository) {2265 const char *path = git_path_info_exclude();2266 if (!access_or_warn(path, R_OK, 0))2267 add_excludes_from_file_1(dir, path,2268 dir->untracked ? &dir->ss_info_exclude : NULL);2269 }2270}22712272int remove_path(const char *name)2273{2274 char *slash;22752276 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)2277 return -1;22782279 slash = strrchr(name, '/');2280 if (slash) {2281 char *dirs = xstrdup(name);2282 slash = dirs + (slash - name);2283 do {2284 *slash = '\0';2285 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));2286 free(dirs);2287 }2288 return 0;2289}22902291/*2292 * Frees memory within dir which was allocated for exclude lists and2293 * the exclude_stack. Does not free dir itself.2294 */2295void clear_directory(struct dir_struct *dir)2296{2297 int i, j;2298 struct exclude_list_group *group;2299 struct exclude_list *el;2300 struct exclude_stack *stk;23012302 for (i = EXC_CMDL; i <= EXC_FILE; i++) {2303 group = &dir->exclude_list_group[i];2304 for (j = 0; j < group->nr; j++) {2305 el = &group->el[j];2306 if (i == EXC_DIRS)2307 free((char *)el->src);2308 clear_exclude_list(el);2309 }2310 free(group->el);2311 }23122313 stk = dir->exclude_stack;2314 while (stk) {2315 struct exclude_stack *prev = stk->prev;2316 free(stk);2317 stk = prev;2318 }2319 strbuf_release(&dir->basebuf);2320}23212322struct ondisk_untracked_cache {2323 struct stat_data info_exclude_stat;2324 struct stat_data excludes_file_stat;2325 uint32_t dir_flags;2326 unsigned char info_exclude_sha1[20];2327 unsigned char excludes_file_sha1[20];2328 char exclude_per_dir[FLEX_ARRAY];2329};23302331#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)23322333struct write_data {2334 int index; /* number of written untracked_cache_dir */2335 struct ewah_bitmap *check_only; /* from untracked_cache_dir */2336 struct ewah_bitmap *valid; /* from untracked_cache_dir */2337 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */2338 struct strbuf out;2339 struct strbuf sb_stat;2340 struct strbuf sb_sha1;2341};23422343static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)2344{2345 to->sd_ctime.sec = htonl(from->sd_ctime.sec);2346 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);2347 to->sd_mtime.sec = htonl(from->sd_mtime.sec);2348 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);2349 to->sd_dev = htonl(from->sd_dev);2350 to->sd_ino = htonl(from->sd_ino);2351 to->sd_uid = htonl(from->sd_uid);2352 to->sd_gid = htonl(from->sd_gid);2353 to->sd_size = htonl(from->sd_size);2354}23552356static void write_one_dir(struct untracked_cache_dir *untracked,2357 struct write_data *wd)2358{2359 struct stat_data stat_data;2360 struct strbuf *out = &wd->out;2361 unsigned char intbuf[16];2362 unsigned int intlen, value;2363 int i = wd->index++;23642365 /*2366 * untracked_nr should be reset whenever valid is clear, but2367 * for safety..2368 */2369 if (!untracked->valid) {2370 untracked->untracked_nr = 0;2371 untracked->check_only = 0;2372 }23732374 if (untracked->check_only)2375 ewah_set(wd->check_only, i);2376 if (untracked->valid) {2377 ewah_set(wd->valid, i);2378 stat_data_to_disk(&stat_data, &untracked->stat_data);2379 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));2380 }2381 if (!is_null_sha1(untracked->exclude_sha1)) {2382 ewah_set(wd->sha1_valid, i);2383 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);2384 }23852386 intlen = encode_varint(untracked->untracked_nr, intbuf);2387 strbuf_add(out, intbuf, intlen);23882389 /* skip non-recurse directories */2390 for (i = 0, value = 0; i < untracked->dirs_nr; i++)2391 if (untracked->dirs[i]->recurse)2392 value++;2393 intlen = encode_varint(value, intbuf);2394 strbuf_add(out, intbuf, intlen);23952396 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);23972398 for (i = 0; i < untracked->untracked_nr; i++)2399 strbuf_add(out, untracked->untracked[i],2400 strlen(untracked->untracked[i]) + 1);24012402 for (i = 0; i < untracked->dirs_nr; i++)2403 if (untracked->dirs[i]->recurse)2404 write_one_dir(untracked->dirs[i], wd);2405}24062407void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)2408{2409 struct ondisk_untracked_cache *ouc;2410 struct write_data wd;2411 unsigned char varbuf[16];2412 int varint_len;2413 size_t len = strlen(untracked->exclude_per_dir);24142415 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);2416 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);2417 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);2418 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);2419 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);2420 ouc->dir_flags = htonl(untracked->dir_flags);24212422 varint_len = encode_varint(untracked->ident.len, varbuf);2423 strbuf_add(out, varbuf, varint_len);2424 strbuf_addbuf(out, &untracked->ident);24252426 strbuf_add(out, ouc, ouc_size(len));2427 free(ouc);2428 ouc = NULL;24292430 if (!untracked->root) {2431 varint_len = encode_varint(0, varbuf);2432 strbuf_add(out, varbuf, varint_len);2433 return;2434 }24352436 wd.index = 0;2437 wd.check_only = ewah_new();2438 wd.valid = ewah_new();2439 wd.sha1_valid = ewah_new();2440 strbuf_init(&wd.out, 1024);2441 strbuf_init(&wd.sb_stat, 1024);2442 strbuf_init(&wd.sb_sha1, 1024);2443 write_one_dir(untracked->root, &wd);24442445 varint_len = encode_varint(wd.index, varbuf);2446 strbuf_add(out, varbuf, varint_len);2447 strbuf_addbuf(out, &wd.out);2448 ewah_serialize_strbuf(wd.valid, out);2449 ewah_serialize_strbuf(wd.check_only, out);2450 ewah_serialize_strbuf(wd.sha1_valid, out);2451 strbuf_addbuf(out, &wd.sb_stat);2452 strbuf_addbuf(out, &wd.sb_sha1);2453 strbuf_addch(out, '\0'); /* safe guard for string lists */24542455 ewah_free(wd.valid);2456 ewah_free(wd.check_only);2457 ewah_free(wd.sha1_valid);2458 strbuf_release(&wd.out);2459 strbuf_release(&wd.sb_stat);2460 strbuf_release(&wd.sb_sha1);2461}24622463static void free_untracked(struct untracked_cache_dir *ucd)2464{2465 int i;2466 if (!ucd)2467 return;2468 for (i = 0; i < ucd->dirs_nr; i++)2469 free_untracked(ucd->dirs[i]);2470 for (i = 0; i < ucd->untracked_nr; i++)2471 free(ucd->untracked[i]);2472 free(ucd->untracked);2473 free(ucd->dirs);2474 free(ucd);2475}24762477void free_untracked_cache(struct untracked_cache *uc)2478{2479 if (uc)2480 free_untracked(uc->root);2481 free(uc);2482}24832484struct read_data {2485 int index;2486 struct untracked_cache_dir **ucd;2487 struct ewah_bitmap *check_only;2488 struct ewah_bitmap *valid;2489 struct ewah_bitmap *sha1_valid;2490 const unsigned char *data;2491 const unsigned char *end;2492};24932494static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)2495{2496 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);2497 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);2498 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);2499 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);2500 to->sd_dev = get_be32(&from->sd_dev);2501 to->sd_ino = get_be32(&from->sd_ino);2502 to->sd_uid = get_be32(&from->sd_uid);2503 to->sd_gid = get_be32(&from->sd_gid);2504 to->sd_size = get_be32(&from->sd_size);2505}25062507static int read_one_dir(struct untracked_cache_dir **untracked_,2508 struct read_data *rd)2509{2510 struct untracked_cache_dir ud, *untracked;2511 const unsigned char *next, *data = rd->data, *end = rd->end;2512 unsigned int value;2513 int i, len;25142515 memset(&ud, 0, sizeof(ud));25162517 next = data;2518 value = decode_varint(&next);2519 if (next > end)2520 return -1;2521 ud.recurse = 1;2522 ud.untracked_alloc = value;2523 ud.untracked_nr = value;2524 if (ud.untracked_nr)2525 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);2526 data = next;25272528 next = data;2529 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);2530 if (next > end)2531 return -1;2532 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);2533 data = next;25342535 len = strlen((const char *)data);2536 next = data + len + 1;2537 if (next > rd->end)2538 return -1;2539 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));2540 memcpy(untracked, &ud, sizeof(ud));2541 memcpy(untracked->name, data, len + 1);2542 data = next;25432544 for (i = 0; i < untracked->untracked_nr; i++) {2545 len = strlen((const char *)data);2546 next = data + len + 1;2547 if (next > rd->end)2548 return -1;2549 untracked->untracked[i] = xstrdup((const char*)data);2550 data = next;2551 }25522553 rd->ucd[rd->index++] = untracked;2554 rd->data = data;25552556 for (i = 0; i < untracked->dirs_nr; i++) {2557 len = read_one_dir(untracked->dirs + i, rd);2558 if (len < 0)2559 return -1;2560 }2561 return 0;2562}25632564static void set_check_only(size_t pos, void *cb)2565{2566 struct read_data *rd = cb;2567 struct untracked_cache_dir *ud = rd->ucd[pos];2568 ud->check_only = 1;2569}25702571static void read_stat(size_t pos, void *cb)2572{2573 struct read_data *rd = cb;2574 struct untracked_cache_dir *ud = rd->ucd[pos];2575 if (rd->data + sizeof(struct stat_data) > rd->end) {2576 rd->data = rd->end + 1;2577 return;2578 }2579 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);2580 rd->data += sizeof(struct stat_data);2581 ud->valid = 1;2582}25832584static void read_sha1(size_t pos, void *cb)2585{2586 struct read_data *rd = cb;2587 struct untracked_cache_dir *ud = rd->ucd[pos];2588 if (rd->data + 20 > rd->end) {2589 rd->data = rd->end + 1;2590 return;2591 }2592 hashcpy(ud->exclude_sha1, rd->data);2593 rd->data += 20;2594}25952596static void load_sha1_stat(struct sha1_stat *sha1_stat,2597 const struct stat_data *stat,2598 const unsigned char *sha1)2599{2600 stat_data_from_disk(&sha1_stat->stat, stat);2601 hashcpy(sha1_stat->sha1, sha1);2602 sha1_stat->valid = 1;2603}26042605struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)2606{2607 const struct ondisk_untracked_cache *ouc;2608 struct untracked_cache *uc;2609 struct read_data rd;2610 const unsigned char *next = data, *end = (const unsigned char *)data + sz;2611 const char *ident;2612 int ident_len, len;26132614 if (sz <= 1 || end[-1] != '\0')2615 return NULL;2616 end--;26172618 ident_len = decode_varint(&next);2619 if (next + ident_len > end)2620 return NULL;2621 ident = (const char *)next;2622 next += ident_len;26232624 ouc = (const struct ondisk_untracked_cache *)next;2625 if (next + ouc_size(0) > end)2626 return NULL;26272628 uc = xcalloc(1, sizeof(*uc));2629 strbuf_init(&uc->ident, ident_len);2630 strbuf_add(&uc->ident, ident, ident_len);2631 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,2632 ouc->info_exclude_sha1);2633 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,2634 ouc->excludes_file_sha1);2635 uc->dir_flags = get_be32(&ouc->dir_flags);2636 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);2637 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */2638 next += ouc_size(strlen(ouc->exclude_per_dir));2639 if (next >= end)2640 goto done2;26412642 len = decode_varint(&next);2643 if (next > end || len == 0)2644 goto done2;26452646 rd.valid = ewah_new();2647 rd.check_only = ewah_new();2648 rd.sha1_valid = ewah_new();2649 rd.data = next;2650 rd.end = end;2651 rd.index = 0;2652 ALLOC_ARRAY(rd.ucd, len);26532654 if (read_one_dir(&uc->root, &rd) || rd.index != len)2655 goto done;26562657 next = rd.data;2658 len = ewah_read_mmap(rd.valid, next, end - next);2659 if (len < 0)2660 goto done;26612662 next += len;2663 len = ewah_read_mmap(rd.check_only, next, end - next);2664 if (len < 0)2665 goto done;26662667 next += len;2668 len = ewah_read_mmap(rd.sha1_valid, next, end - next);2669 if (len < 0)2670 goto done;26712672 ewah_each_bit(rd.check_only, set_check_only, &rd);2673 rd.data = next + len;2674 ewah_each_bit(rd.valid, read_stat, &rd);2675 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);2676 next = rd.data;26772678done:2679 free(rd.ucd);2680 ewah_free(rd.valid);2681 ewah_free(rd.check_only);2682 ewah_free(rd.sha1_valid);2683done2:2684 if (next != end) {2685 free_untracked_cache(uc);2686 uc = NULL;2687 }2688 return uc;2689}26902691static void invalidate_one_directory(struct untracked_cache *uc,2692 struct untracked_cache_dir *ucd)2693{2694 uc->dir_invalidated++;2695 ucd->valid = 0;2696 ucd->untracked_nr = 0;2697}26982699/*2700 * Normally when an entry is added or removed from a directory,2701 * invalidating that directory is enough. No need to touch its2702 * ancestors. When a directory is shown as "foo/bar/" in git-status2703 * however, deleting or adding an entry may have cascading effect.2704 *2705 * Say the "foo/bar/file" has become untracked, we need to tell the2706 * untracked_cache_dir of "foo" that "bar/" is not an untracked2707 * directory any more (because "bar" is managed by foo as an untracked2708 * "file").2709 *2710 * Similarly, if "foo/bar/file" moves from untracked to tracked and it2711 * was the last untracked entry in the entire "foo", we should show2712 * "foo/" instead. Which means we have to invalidate past "bar" up to2713 * "foo".2714 *2715 * This function traverses all directories from root to leaf. If there2716 * is a chance of one of the above cases happening, we invalidate back2717 * to root. Otherwise we just invalidate the leaf. There may be a more2718 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to2719 * detect these cases and avoid unnecessary invalidation, for example,2720 * checking for the untracked entry named "bar/" in "foo", but for now2721 * stick to something safe and simple.2722 */2723static int invalidate_one_component(struct untracked_cache *uc,2724 struct untracked_cache_dir *dir,2725 const char *path, int len)2726{2727 const char *rest = strchr(path, '/');27282729 if (rest) {2730 int component_len = rest - path;2731 struct untracked_cache_dir *d =2732 lookup_untracked(uc, dir, path, component_len);2733 int ret =2734 invalidate_one_component(uc, d, rest + 1,2735 len - (component_len + 1));2736 if (ret)2737 invalidate_one_directory(uc, dir);2738 return ret;2739 }27402741 invalidate_one_directory(uc, dir);2742 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;2743}27442745void untracked_cache_invalidate_path(struct index_state *istate,2746 const char *path)2747{2748 if (!istate->untracked || !istate->untracked->root)2749 return;2750 invalidate_one_component(istate->untracked, istate->untracked->root,2751 path, strlen(path));2752}27532754void untracked_cache_remove_from_index(struct index_state *istate,2755 const char *path)2756{2757 untracked_cache_invalidate_path(istate, path);2758}27592760void untracked_cache_add_to_index(struct index_state *istate,2761 const char *path)2762{2763 untracked_cache_invalidate_path(istate, path);2764}27652766/* Update gitfile and core.worktree setting to connect work tree and git dir */2767void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)2768{2769 struct strbuf gitfile_sb = STRBUF_INIT;2770 struct strbuf cfg_sb = STRBUF_INIT;2771 struct strbuf rel_path = STRBUF_INIT;2772 char *git_dir, *work_tree;27732774 /* Prepare .git file */2775 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);2776 if (safe_create_leading_directories_const(gitfile_sb.buf))2777 die(_("could not create directories for %s"), gitfile_sb.buf);27782779 /* Prepare config file */2780 strbuf_addf(&cfg_sb, "%s/config", git_dir_);2781 if (safe_create_leading_directories_const(cfg_sb.buf))2782 die(_("could not create directories for %s"), cfg_sb.buf);27832784 git_dir = real_pathdup(git_dir_, 1);2785 work_tree = real_pathdup(work_tree_, 1);27862787 /* Write .git file */2788 write_file(gitfile_sb.buf, "gitdir: %s",2789 relative_path(git_dir, work_tree, &rel_path));2790 /* Update core.worktree setting */2791 git_config_set_in_file(cfg_sb.buf, "core.worktree",2792 relative_path(work_tree, git_dir, &rel_path));27932794 strbuf_release(&gitfile_sb);2795 strbuf_release(&cfg_sb);2796 strbuf_release(&rel_path);2797 free(work_tree);2798 free(git_dir);2799}28002801/*2802 * Migrate the git directory of the given path from old_git_dir to new_git_dir.2803 */2804void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)2805{2806 if (rename(old_git_dir, new_git_dir) < 0)2807 die_errno(_("could not migrate git directory from '%s' to '%s'"),2808 old_git_dir, new_git_dir);28092810 connect_work_tree_and_git_dir(path, new_git_dir);2811}