1/* 2 * This handles recursive filename detection with exclude 3 * files, index knowledge etc.. 4 * 5 * See Documentation/technical/api-directory-listing.txt 6 * 7 * Copyright (C) Linus Torvalds, 2005-2006 8 * Junio Hamano, 2005-2006 9 */ 10#include "cache.h" 11#include "dir.h" 12#include "refs.h" 13#include "wildmatch.h" 14#include "pathspec.h" 15#include "utf8.h" 16#include "varint.h" 17#include "ewah/ewok.h" 18 19struct path_simplify { 20 int len; 21 const char *path; 22}; 23 24/* 25 * Tells read_directory_recursive how a file or directory should be treated. 26 * Values are ordered by significance, e.g. if a directory contains both 27 * excluded and untracked files, it is listed as untracked because 28 * path_untracked > path_excluded. 29 */ 30enum path_treatment { 31 path_none = 0, 32 path_recurse, 33 path_excluded, 34 path_untracked 35}; 36 37/* 38 * Support data structure for our opendir/readdir/closedir wrappers 39 */ 40struct cached_dir { 41 DIR *fdir; 42 struct untracked_cache_dir *untracked; 43 int nr_files; 44 int nr_dirs; 45 46 struct dirent *de; 47 const char *file; 48 struct untracked_cache_dir *ucd; 49}; 50 51static enum path_treatment read_directory_recursive(struct dir_struct *dir, 52 const char *path, int len, struct untracked_cache_dir *untracked, 53 int check_only, const struct path_simplify *simplify); 54static int get_dtype(struct dirent *de, const char *path, int len); 55 56/* helper string functions with support for the ignore_case flag */ 57int strcmp_icase(const char *a, const char *b) 58{ 59 return ignore_case ? strcasecmp(a, b) : strcmp(a, b); 60} 61 62int strncmp_icase(const char *a, const char *b, size_t count) 63{ 64 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count); 65} 66 67int fnmatch_icase(const char *pattern, const char *string, int flags) 68{ 69 return wildmatch(pattern, string, 70 flags | (ignore_case ? WM_CASEFOLD : 0), 71 NULL); 72} 73 74int git_fnmatch(const struct pathspec_item *item, 75 const char *pattern, const char *string, 76 int prefix) 77{ 78 if (prefix > 0) { 79 if (ps_strncmp(item, pattern, string, prefix)) 80 return WM_NOMATCH; 81 pattern += prefix; 82 string += prefix; 83 } 84 if (item->flags & PATHSPEC_ONESTAR) { 85 int pattern_len = strlen(++pattern); 86 int string_len = strlen(string); 87 return string_len < pattern_len || 88 ps_strcmp(item, pattern, 89 string + string_len - pattern_len); 90 } 91 if (item->magic & PATHSPEC_GLOB) 92 return wildmatch(pattern, string, 93 WM_PATHNAME | 94 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0), 95 NULL); 96 else 97 /* wildmatch has not learned no FNM_PATHNAME mode yet */ 98 return wildmatch(pattern, string, 99 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0, 100 NULL); 101} 102 103static int fnmatch_icase_mem(const char *pattern, int patternlen, 104 const char *string, int stringlen, 105 int flags) 106{ 107 int match_status; 108 struct strbuf pat_buf = STRBUF_INIT; 109 struct strbuf str_buf = STRBUF_INIT; 110 const char *use_pat = pattern; 111 const char *use_str = string; 112 113 if (pattern[patternlen]) { 114 strbuf_add(&pat_buf, pattern, patternlen); 115 use_pat = pat_buf.buf; 116 } 117 if (string[stringlen]) { 118 strbuf_add(&str_buf, string, stringlen); 119 use_str = str_buf.buf; 120 } 121 122 if (ignore_case) 123 flags |= WM_CASEFOLD; 124 match_status = wildmatch(use_pat, use_str, flags, NULL); 125 126 strbuf_release(&pat_buf); 127 strbuf_release(&str_buf); 128 129 return match_status; 130} 131 132static size_t common_prefix_len(const struct pathspec *pathspec) 133{ 134 int n; 135 size_t max = 0; 136 137 /* 138 * ":(icase)path" is treated as a pathspec full of 139 * wildcard. In other words, only prefix is considered common 140 * prefix. If the pathspec is abc/foo abc/bar, running in 141 * subdir xyz, the common prefix is still xyz, not xuz/abc as 142 * in non-:(icase). 143 */ 144 GUARD_PATHSPEC(pathspec, 145 PATHSPEC_FROMTOP | 146 PATHSPEC_MAXDEPTH | 147 PATHSPEC_LITERAL | 148 PATHSPEC_GLOB | 149 PATHSPEC_ICASE | 150 PATHSPEC_EXCLUDE); 151 152 for (n = 0; n < pathspec->nr; n++) { 153 size_t i = 0, len = 0, item_len; 154 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE) 155 continue; 156 if (pathspec->items[n].magic & PATHSPEC_ICASE) 157 item_len = pathspec->items[n].prefix; 158 else 159 item_len = pathspec->items[n].nowildcard_len; 160 while (i < item_len && (n == 0 || i < max)) { 161 char c = pathspec->items[n].match[i]; 162 if (c != pathspec->items[0].match[i]) 163 break; 164 if (c == '/') 165 len = i + 1; 166 i++; 167 } 168 if (n == 0 || len < max) { 169 max = len; 170 if (!max) 171 break; 172 } 173 } 174 return max; 175} 176 177/* 178 * Returns a copy of the longest leading path common among all 179 * pathspecs. 180 */ 181char *common_prefix(const struct pathspec *pathspec) 182{ 183 unsigned long len = common_prefix_len(pathspec); 184 185 return len ? xmemdupz(pathspec->items[0].match, len) : NULL; 186} 187 188int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec) 189{ 190 size_t len; 191 192 /* 193 * Calculate common prefix for the pathspec, and 194 * use that to optimize the directory walk 195 */ 196 len = common_prefix_len(pathspec); 197 198 /* Read the directory and prune it */ 199 read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec); 200 return len; 201} 202 203int within_depth(const char *name, int namelen, 204 int depth, int max_depth) 205{ 206 const char *cp = name, *cpe = name + namelen; 207 208 while (cp < cpe) { 209 if (*cp++ != '/') 210 continue; 211 depth++; 212 if (depth > max_depth) 213 return 0; 214 } 215 return 1; 216} 217 218#define DO_MATCH_EXCLUDE 1 219#define DO_MATCH_DIRECTORY 2 220 221/* 222 * Does 'match' match the given name? 223 * A match is found if 224 * 225 * (1) the 'match' string is leading directory of 'name', or 226 * (2) the 'match' string is a wildcard and matches 'name', or 227 * (3) the 'match' string is exactly the same as 'name'. 228 * 229 * and the return value tells which case it was. 230 * 231 * It returns 0 when there is no match. 232 */ 233static int match_pathspec_item(const struct pathspec_item *item, int prefix, 234 const char *name, int namelen, unsigned flags) 235{ 236 /* name/namelen has prefix cut off by caller */ 237 const char *match = item->match + prefix; 238 int matchlen = item->len - prefix; 239 240 /* 241 * The normal call pattern is: 242 * 1. prefix = common_prefix_len(ps); 243 * 2. prune something, or fill_directory 244 * 3. match_pathspec() 245 * 246 * 'prefix' at #1 may be shorter than the command's prefix and 247 * it's ok for #2 to match extra files. Those extras will be 248 * trimmed at #3. 249 * 250 * Suppose the pathspec is 'foo' and '../bar' running from 251 * subdir 'xyz'. The common prefix at #1 will be empty, thanks 252 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The 253 * user does not want XYZ/foo, only the "foo" part should be 254 * case-insensitive. We need to filter out XYZ/foo here. In 255 * other words, we do not trust the caller on comparing the 256 * prefix part when :(icase) is involved. We do exact 257 * comparison ourselves. 258 * 259 * Normally the caller (common_prefix_len() in fact) does 260 * _exact_ matching on name[-prefix+1..-1] and we do not need 261 * to check that part. Be defensive and check it anyway, in 262 * case common_prefix_len is changed, or a new caller is 263 * introduced that does not use common_prefix_len. 264 * 265 * If the penalty turns out too high when prefix is really 266 * long, maybe change it to 267 * strncmp(match, name, item->prefix - prefix) 268 */ 269 if (item->prefix && (item->magic & PATHSPEC_ICASE) && 270 strncmp(item->match, name - prefix, item->prefix)) 271 return 0; 272 273 /* If the match was just the prefix, we matched */ 274 if (!*match) 275 return MATCHED_RECURSIVELY; 276 277 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) { 278 if (matchlen == namelen) 279 return MATCHED_EXACTLY; 280 281 if (match[matchlen-1] == '/' || name[matchlen] == '/') 282 return MATCHED_RECURSIVELY; 283 } else if ((flags & DO_MATCH_DIRECTORY) && 284 match[matchlen - 1] == '/' && 285 namelen == matchlen - 1 && 286 !ps_strncmp(item, match, name, namelen)) 287 return MATCHED_EXACTLY; 288 289 if (item->nowildcard_len < item->len && 290 !git_fnmatch(item, match, name, 291 item->nowildcard_len - prefix)) 292 return MATCHED_FNMATCH; 293 294 return 0; 295} 296 297/* 298 * Given a name and a list of pathspecs, returns the nature of the 299 * closest (i.e. most specific) match of the name to any of the 300 * pathspecs. 301 * 302 * The caller typically calls this multiple times with the same 303 * pathspec and seen[] array but with different name/namelen 304 * (e.g. entries from the index) and is interested in seeing if and 305 * how each pathspec matches all the names it calls this function 306 * with. A mark is left in the seen[] array for each pathspec element 307 * indicating the closest type of match that element achieved, so if 308 * seen[n] remains zero after multiple invocations, that means the nth 309 * pathspec did not match any names, which could indicate that the 310 * user mistyped the nth pathspec. 311 */ 312static int do_match_pathspec(const struct pathspec *ps, 313 const char *name, int namelen, 314 int prefix, char *seen, 315 unsigned flags) 316{ 317 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE; 318 319 GUARD_PATHSPEC(ps, 320 PATHSPEC_FROMTOP | 321 PATHSPEC_MAXDEPTH | 322 PATHSPEC_LITERAL | 323 PATHSPEC_GLOB | 324 PATHSPEC_ICASE | 325 PATHSPEC_EXCLUDE); 326 327 if (!ps->nr) { 328 if (!ps->recursive || 329 !(ps->magic & PATHSPEC_MAXDEPTH) || 330 ps->max_depth == -1) 331 return MATCHED_RECURSIVELY; 332 333 if (within_depth(name, namelen, 0, ps->max_depth)) 334 return MATCHED_EXACTLY; 335 else 336 return 0; 337 } 338 339 name += prefix; 340 namelen -= prefix; 341 342 for (i = ps->nr - 1; i >= 0; i--) { 343 int how; 344 345 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) || 346 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE))) 347 continue; 348 349 if (seen && seen[i] == MATCHED_EXACTLY) 350 continue; 351 /* 352 * Make exclude patterns optional and never report 353 * "pathspec ':(exclude)foo' matches no files" 354 */ 355 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE) 356 seen[i] = MATCHED_FNMATCH; 357 how = match_pathspec_item(ps->items+i, prefix, name, 358 namelen, flags); 359 if (ps->recursive && 360 (ps->magic & PATHSPEC_MAXDEPTH) && 361 ps->max_depth != -1 && 362 how && how != MATCHED_FNMATCH) { 363 int len = ps->items[i].len; 364 if (name[len] == '/') 365 len++; 366 if (within_depth(name+len, namelen-len, 0, ps->max_depth)) 367 how = MATCHED_EXACTLY; 368 else 369 how = 0; 370 } 371 if (how) { 372 if (retval < how) 373 retval = how; 374 if (seen && seen[i] < how) 375 seen[i] = how; 376 } 377 } 378 return retval; 379} 380 381int match_pathspec(const struct pathspec *ps, 382 const char *name, int namelen, 383 int prefix, char *seen, int is_dir) 384{ 385 int positive, negative; 386 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0; 387 positive = do_match_pathspec(ps, name, namelen, 388 prefix, seen, flags); 389 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive) 390 return positive; 391 negative = do_match_pathspec(ps, name, namelen, 392 prefix, seen, 393 flags | DO_MATCH_EXCLUDE); 394 return negative ? 0 : positive; 395} 396 397int report_path_error(const char *ps_matched, 398 const struct pathspec *pathspec, 399 const char *prefix) 400{ 401 /* 402 * Make sure all pathspec matched; otherwise it is an error. 403 */ 404 int num, errors = 0; 405 for (num = 0; num < pathspec->nr; num++) { 406 int other, found_dup; 407 408 if (ps_matched[num]) 409 continue; 410 /* 411 * The caller might have fed identical pathspec 412 * twice. Do not barf on such a mistake. 413 * FIXME: parse_pathspec should have eliminated 414 * duplicate pathspec. 415 */ 416 for (found_dup = other = 0; 417 !found_dup && other < pathspec->nr; 418 other++) { 419 if (other == num || !ps_matched[other]) 420 continue; 421 if (!strcmp(pathspec->items[other].original, 422 pathspec->items[num].original)) 423 /* 424 * Ok, we have a match already. 425 */ 426 found_dup = 1; 427 } 428 if (found_dup) 429 continue; 430 431 error("pathspec '%s' did not match any file(s) known to git.", 432 pathspec->items[num].original); 433 errors++; 434 } 435 return errors; 436} 437 438/* 439 * Return the length of the "simple" part of a path match limiter. 440 */ 441int simple_length(const char *match) 442{ 443 int len = -1; 444 445 for (;;) { 446 unsigned char c = *match++; 447 len++; 448 if (c == '\0' || is_glob_special(c)) 449 return len; 450 } 451} 452 453int no_wildcard(const char *string) 454{ 455 return string[simple_length(string)] == '\0'; 456} 457 458void parse_exclude_pattern(const char **pattern, 459 int *patternlen, 460 int *flags, 461 int *nowildcardlen) 462{ 463 const char *p = *pattern; 464 size_t i, len; 465 466 *flags = 0; 467 if (*p == '!') { 468 *flags |= EXC_FLAG_NEGATIVE; 469 p++; 470 } 471 len = strlen(p); 472 if (len && p[len - 1] == '/') { 473 len--; 474 *flags |= EXC_FLAG_MUSTBEDIR; 475 } 476 for (i = 0; i < len; i++) { 477 if (p[i] == '/') 478 break; 479 } 480 if (i == len) 481 *flags |= EXC_FLAG_NODIR; 482 *nowildcardlen = simple_length(p); 483 /* 484 * we should have excluded the trailing slash from 'p' too, 485 * but that's one more allocation. Instead just make sure 486 * nowildcardlen does not exceed real patternlen 487 */ 488 if (*nowildcardlen > len) 489 *nowildcardlen = len; 490 if (*p == '*' && no_wildcard(p + 1)) 491 *flags |= EXC_FLAG_ENDSWITH; 492 *pattern = p; 493 *patternlen = len; 494} 495 496void add_exclude(const char *string, const char *base, 497 int baselen, struct exclude_list *el, int srcpos) 498{ 499 struct exclude *x; 500 int patternlen; 501 int flags; 502 int nowildcardlen; 503 504 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen); 505 if (flags & EXC_FLAG_MUSTBEDIR) { 506 char *s; 507 x = xmalloc(sizeof(*x) + patternlen + 1); 508 s = (char *)(x+1); 509 memcpy(s, string, patternlen); 510 s[patternlen] = '\0'; 511 x->pattern = s; 512 } else { 513 x = xmalloc(sizeof(*x)); 514 x->pattern = string; 515 } 516 x->patternlen = patternlen; 517 x->nowildcardlen = nowildcardlen; 518 x->base = base; 519 x->baselen = baselen; 520 x->flags = flags; 521 x->srcpos = srcpos; 522 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc); 523 el->excludes[el->nr++] = x; 524 x->el = el; 525} 526 527static void *read_skip_worktree_file_from_index(const char *path, size_t *size, 528 struct sha1_stat *sha1_stat) 529{ 530 int pos, len; 531 unsigned long sz; 532 enum object_type type; 533 void *data; 534 535 len = strlen(path); 536 pos = cache_name_pos(path, len); 537 if (pos < 0) 538 return NULL; 539 if (!ce_skip_worktree(active_cache[pos])) 540 return NULL; 541 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz); 542 if (!data || type != OBJ_BLOB) { 543 free(data); 544 return NULL; 545 } 546 *size = xsize_t(sz); 547 if (sha1_stat) { 548 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat)); 549 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1); 550 } 551 return data; 552} 553 554/* 555 * Frees memory within el which was allocated for exclude patterns and 556 * the file buffer. Does not free el itself. 557 */ 558void clear_exclude_list(struct exclude_list *el) 559{ 560 int i; 561 562 for (i = 0; i < el->nr; i++) 563 free(el->excludes[i]); 564 free(el->excludes); 565 free(el->filebuf); 566 567 el->nr = 0; 568 el->excludes = NULL; 569 el->filebuf = NULL; 570} 571 572static void trim_trailing_spaces(char *buf) 573{ 574 char *p, *last_space = NULL; 575 576 for (p = buf; *p; p++) 577 switch (*p) { 578 case ' ': 579 if (!last_space) 580 last_space = p; 581 break; 582 case '\\': 583 p++; 584 if (!*p) 585 return; 586 /* fallthrough */ 587 default: 588 last_space = NULL; 589 } 590 591 if (last_space) 592 *last_space = '\0'; 593} 594 595/* 596 * Given a subdirectory name and "dir" of the current directory, 597 * search the subdir in "dir" and return it, or create a new one if it 598 * does not exist in "dir". 599 * 600 * If "name" has the trailing slash, it'll be excluded in the search. 601 */ 602static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc, 603 struct untracked_cache_dir *dir, 604 const char *name, int len) 605{ 606 int first, last; 607 struct untracked_cache_dir *d; 608 if (!dir) 609 return NULL; 610 if (len && name[len - 1] == '/') 611 len--; 612 first = 0; 613 last = dir->dirs_nr; 614 while (last > first) { 615 int cmp, next = (last + first) >> 1; 616 d = dir->dirs[next]; 617 cmp = strncmp(name, d->name, len); 618 if (!cmp && strlen(d->name) > len) 619 cmp = -1; 620 if (!cmp) 621 return d; 622 if (cmp < 0) { 623 last = next; 624 continue; 625 } 626 first = next+1; 627 } 628 629 uc->dir_created++; 630 d = xmalloc(sizeof(*d) + len + 1); 631 memset(d, 0, sizeof(*d)); 632 memcpy(d->name, name, len); 633 d->name[len] = '\0'; 634 635 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc); 636 memmove(dir->dirs + first + 1, dir->dirs + first, 637 (dir->dirs_nr - first) * sizeof(*dir->dirs)); 638 dir->dirs_nr++; 639 dir->dirs[first] = d; 640 return d; 641} 642 643static void do_invalidate_gitignore(struct untracked_cache_dir *dir) 644{ 645 int i; 646 dir->valid = 0; 647 dir->untracked_nr = 0; 648 for (i = 0; i < dir->dirs_nr; i++) 649 do_invalidate_gitignore(dir->dirs[i]); 650} 651 652static void invalidate_gitignore(struct untracked_cache *uc, 653 struct untracked_cache_dir *dir) 654{ 655 uc->gitignore_invalidated++; 656 do_invalidate_gitignore(dir); 657} 658 659static void invalidate_directory(struct untracked_cache *uc, 660 struct untracked_cache_dir *dir) 661{ 662 int i; 663 uc->dir_invalidated++; 664 dir->valid = 0; 665 dir->untracked_nr = 0; 666 for (i = 0; i < dir->dirs_nr; i++) 667 dir->dirs[i]->recurse = 0; 668} 669 670/* 671 * Given a file with name "fname", read it (either from disk, or from 672 * the index if "check_index" is non-zero), parse it and store the 673 * exclude rules in "el". 674 * 675 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill 676 * stat data from disk (only valid if add_excludes returns zero). If 677 * ss_valid is non-zero, "ss" must contain good value as input. 678 */ 679static int add_excludes(const char *fname, const char *base, int baselen, 680 struct exclude_list *el, int check_index, 681 struct sha1_stat *sha1_stat) 682{ 683 struct stat st; 684 int fd, i, lineno = 1; 685 size_t size = 0; 686 char *buf, *entry; 687 688 fd = open(fname, O_RDONLY); 689 if (fd < 0 || fstat(fd, &st) < 0) { 690 if (errno != ENOENT) 691 warn_on_inaccessible(fname); 692 if (0 <= fd) 693 close(fd); 694 if (!check_index || 695 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL) 696 return -1; 697 if (size == 0) { 698 free(buf); 699 return 0; 700 } 701 if (buf[size-1] != '\n') { 702 buf = xrealloc(buf, size+1); 703 buf[size++] = '\n'; 704 } 705 } else { 706 size = xsize_t(st.st_size); 707 if (size == 0) { 708 if (sha1_stat) { 709 fill_stat_data(&sha1_stat->stat, &st); 710 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN); 711 sha1_stat->valid = 1; 712 } 713 close(fd); 714 return 0; 715 } 716 buf = xmalloc(size+1); 717 if (read_in_full(fd, buf, size) != size) { 718 free(buf); 719 close(fd); 720 return -1; 721 } 722 buf[size++] = '\n'; 723 close(fd); 724 if (sha1_stat) { 725 int pos; 726 if (sha1_stat->valid && 727 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st)) 728 ; /* no content change, ss->sha1 still good */ 729 else if (check_index && 730 (pos = cache_name_pos(fname, strlen(fname))) >= 0 && 731 !ce_stage(active_cache[pos]) && 732 ce_uptodate(active_cache[pos]) && 733 !would_convert_to_git(fname)) 734 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1); 735 else 736 hash_sha1_file(buf, size, "blob", sha1_stat->sha1); 737 fill_stat_data(&sha1_stat->stat, &st); 738 sha1_stat->valid = 1; 739 } 740 } 741 742 el->filebuf = buf; 743 744 if (skip_utf8_bom(&buf, size)) 745 size -= buf - el->filebuf; 746 747 entry = buf; 748 749 for (i = 0; i < size; i++) { 750 if (buf[i] == '\n') { 751 if (entry != buf + i && entry[0] != '#') { 752 buf[i - (i && buf[i-1] == '\r')] = 0; 753 trim_trailing_spaces(entry); 754 add_exclude(entry, base, baselen, el, lineno); 755 } 756 lineno++; 757 entry = buf + i + 1; 758 } 759 } 760 return 0; 761} 762 763int add_excludes_from_file_to_list(const char *fname, const char *base, 764 int baselen, struct exclude_list *el, 765 int check_index) 766{ 767 return add_excludes(fname, base, baselen, el, check_index, NULL); 768} 769 770struct exclude_list *add_exclude_list(struct dir_struct *dir, 771 int group_type, const char *src) 772{ 773 struct exclude_list *el; 774 struct exclude_list_group *group; 775 776 group = &dir->exclude_list_group[group_type]; 777 ALLOC_GROW(group->el, group->nr + 1, group->alloc); 778 el = &group->el[group->nr++]; 779 memset(el, 0, sizeof(*el)); 780 el->src = src; 781 return el; 782} 783 784/* 785 * Used to set up core.excludesfile and .git/info/exclude lists. 786 */ 787static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname, 788 struct sha1_stat *sha1_stat) 789{ 790 struct exclude_list *el; 791 /* 792 * catch setup_standard_excludes() that's called before 793 * dir->untracked is assigned. That function behaves 794 * differently when dir->untracked is non-NULL. 795 */ 796 if (!dir->untracked) 797 dir->unmanaged_exclude_files++; 798 el = add_exclude_list(dir, EXC_FILE, fname); 799 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0) 800 die("cannot use %s as an exclude file", fname); 801} 802 803void add_excludes_from_file(struct dir_struct *dir, const char *fname) 804{ 805 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */ 806 add_excludes_from_file_1(dir, fname, NULL); 807} 808 809int match_basename(const char *basename, int basenamelen, 810 const char *pattern, int prefix, int patternlen, 811 int flags) 812{ 813 if (prefix == patternlen) { 814 if (patternlen == basenamelen && 815 !strncmp_icase(pattern, basename, basenamelen)) 816 return 1; 817 } else if (flags & EXC_FLAG_ENDSWITH) { 818 /* "*literal" matching against "fooliteral" */ 819 if (patternlen - 1 <= basenamelen && 820 !strncmp_icase(pattern + 1, 821 basename + basenamelen - (patternlen - 1), 822 patternlen - 1)) 823 return 1; 824 } else { 825 if (fnmatch_icase_mem(pattern, patternlen, 826 basename, basenamelen, 827 0) == 0) 828 return 1; 829 } 830 return 0; 831} 832 833int match_pathname(const char *pathname, int pathlen, 834 const char *base, int baselen, 835 const char *pattern, int prefix, int patternlen, 836 int flags) 837{ 838 const char *name; 839 int namelen; 840 841 /* 842 * match with FNM_PATHNAME; the pattern has base implicitly 843 * in front of it. 844 */ 845 if (*pattern == '/') { 846 pattern++; 847 patternlen--; 848 prefix--; 849 } 850 851 /* 852 * baselen does not count the trailing slash. base[] may or 853 * may not end with a trailing slash though. 854 */ 855 if (pathlen < baselen + 1 || 856 (baselen && pathname[baselen] != '/') || 857 strncmp_icase(pathname, base, baselen)) 858 return 0; 859 860 namelen = baselen ? pathlen - baselen - 1 : pathlen; 861 name = pathname + pathlen - namelen; 862 863 if (prefix) { 864 /* 865 * if the non-wildcard part is longer than the 866 * remaining pathname, surely it cannot match. 867 */ 868 if (prefix > namelen) 869 return 0; 870 871 if (strncmp_icase(pattern, name, prefix)) 872 return 0; 873 pattern += prefix; 874 patternlen -= prefix; 875 name += prefix; 876 namelen -= prefix; 877 878 /* 879 * If the whole pattern did not have a wildcard, 880 * then our prefix match is all we need; we 881 * do not need to call fnmatch at all. 882 */ 883 if (!patternlen && !namelen) 884 return 1; 885 } 886 887 return fnmatch_icase_mem(pattern, patternlen, 888 name, namelen, 889 WM_PATHNAME) == 0; 890} 891 892/* 893 * Scan the given exclude list in reverse to see whether pathname 894 * should be ignored. The first match (i.e. the last on the list), if 895 * any, determines the fate. Returns the exclude_list element which 896 * matched, or NULL for undecided. 897 */ 898static struct exclude *last_exclude_matching_from_list(const char *pathname, 899 int pathlen, 900 const char *basename, 901 int *dtype, 902 struct exclude_list *el) 903{ 904 int i; 905 906 if (!el->nr) 907 return NULL; /* undefined */ 908 909 for (i = el->nr - 1; 0 <= i; i--) { 910 struct exclude *x = el->excludes[i]; 911 const char *exclude = x->pattern; 912 int prefix = x->nowildcardlen; 913 914 if (x->flags & EXC_FLAG_MUSTBEDIR) { 915 if (*dtype == DT_UNKNOWN) 916 *dtype = get_dtype(NULL, pathname, pathlen); 917 if (*dtype != DT_DIR) 918 continue; 919 } 920 921 if (x->flags & EXC_FLAG_NODIR) { 922 if (match_basename(basename, 923 pathlen - (basename - pathname), 924 exclude, prefix, x->patternlen, 925 x->flags)) 926 return x; 927 continue; 928 } 929 930 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/'); 931 if (match_pathname(pathname, pathlen, 932 x->base, x->baselen ? x->baselen - 1 : 0, 933 exclude, prefix, x->patternlen, x->flags)) 934 return x; 935 } 936 return NULL; /* undecided */ 937} 938 939/* 940 * Scan the list and let the last match determine the fate. 941 * Return 1 for exclude, 0 for include and -1 for undecided. 942 */ 943int is_excluded_from_list(const char *pathname, 944 int pathlen, const char *basename, int *dtype, 945 struct exclude_list *el) 946{ 947 struct exclude *exclude; 948 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el); 949 if (exclude) 950 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1; 951 return -1; /* undecided */ 952} 953 954static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir, 955 const char *pathname, int pathlen, const char *basename, 956 int *dtype_p) 957{ 958 int i, j; 959 struct exclude_list_group *group; 960 struct exclude *exclude; 961 for (i = EXC_CMDL; i <= EXC_FILE; i++) { 962 group = &dir->exclude_list_group[i]; 963 for (j = group->nr - 1; j >= 0; j--) { 964 exclude = last_exclude_matching_from_list( 965 pathname, pathlen, basename, dtype_p, 966 &group->el[j]); 967 if (exclude) 968 return exclude; 969 } 970 } 971 return NULL; 972} 973 974/* 975 * Loads the per-directory exclude list for the substring of base 976 * which has a char length of baselen. 977 */ 978static void prep_exclude(struct dir_struct *dir, const char *base, int baselen) 979{ 980 struct exclude_list_group *group; 981 struct exclude_list *el; 982 struct exclude_stack *stk = NULL; 983 struct untracked_cache_dir *untracked; 984 int current; 985 986 group = &dir->exclude_list_group[EXC_DIRS]; 987 988 /* 989 * Pop the exclude lists from the EXCL_DIRS exclude_list_group 990 * which originate from directories not in the prefix of the 991 * path being checked. 992 */ 993 while ((stk = dir->exclude_stack) != NULL) { 994 if (stk->baselen <= baselen && 995 !strncmp(dir->basebuf.buf, base, stk->baselen)) 996 break; 997 el = &group->el[dir->exclude_stack->exclude_ix]; 998 dir->exclude_stack = stk->prev; 999 dir->exclude = NULL;1000 free((char *)el->src); /* see strbuf_detach() below */1001 clear_exclude_list(el);1002 free(stk);1003 group->nr--;1004 }10051006 /* Skip traversing into sub directories if the parent is excluded */1007 if (dir->exclude)1008 return;10091010 /*1011 * Lazy initialization. All call sites currently just1012 * memset(dir, 0, sizeof(*dir)) before use. Changing all of1013 * them seems lots of work for little benefit.1014 */1015 if (!dir->basebuf.buf)1016 strbuf_init(&dir->basebuf, PATH_MAX);10171018 /* Read from the parent directories and push them down. */1019 current = stk ? stk->baselen : -1;1020 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);1021 if (dir->untracked)1022 untracked = stk ? stk->ucd : dir->untracked->root;1023 else1024 untracked = NULL;10251026 while (current < baselen) {1027 const char *cp;1028 struct sha1_stat sha1_stat;10291030 stk = xcalloc(1, sizeof(*stk));1031 if (current < 0) {1032 cp = base;1033 current = 0;1034 } else {1035 cp = strchr(base + current + 1, '/');1036 if (!cp)1037 die("oops in prep_exclude");1038 cp++;1039 untracked =1040 lookup_untracked(dir->untracked, untracked,1041 base + current,1042 cp - base - current);1043 }1044 stk->prev = dir->exclude_stack;1045 stk->baselen = cp - base;1046 stk->exclude_ix = group->nr;1047 stk->ucd = untracked;1048 el = add_exclude_list(dir, EXC_DIRS, NULL);1049 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);1050 assert(stk->baselen == dir->basebuf.len);10511052 /* Abort if the directory is excluded */1053 if (stk->baselen) {1054 int dt = DT_DIR;1055 dir->basebuf.buf[stk->baselen - 1] = 0;1056 dir->exclude = last_exclude_matching_from_lists(dir,1057 dir->basebuf.buf, stk->baselen - 1,1058 dir->basebuf.buf + current, &dt);1059 dir->basebuf.buf[stk->baselen - 1] = '/';1060 if (dir->exclude &&1061 dir->exclude->flags & EXC_FLAG_NEGATIVE)1062 dir->exclude = NULL;1063 if (dir->exclude) {1064 dir->exclude_stack = stk;1065 return;1066 }1067 }10681069 /* Try to read per-directory file */1070 hashclr(sha1_stat.sha1);1071 sha1_stat.valid = 0;1072 if (dir->exclude_per_dir &&1073 /*1074 * If we know that no files have been added in1075 * this directory (i.e. valid_cached_dir() has1076 * been executed and set untracked->valid) ..1077 */1078 (!untracked || !untracked->valid ||1079 /*1080 * .. and .gitignore does not exist before1081 * (i.e. null exclude_sha1 and skip_worktree is1082 * not set). Then we can skip loading .gitignore,1083 * which would result in ENOENT anyway.1084 * skip_worktree is taken care in read_directory()1085 */1086 !is_null_sha1(untracked->exclude_sha1))) {1087 /*1088 * dir->basebuf gets reused by the traversal, but we1089 * need fname to remain unchanged to ensure the src1090 * member of each struct exclude correctly1091 * back-references its source file. Other invocations1092 * of add_exclude_list provide stable strings, so we1093 * strbuf_detach() and free() here in the caller.1094 */1095 struct strbuf sb = STRBUF_INIT;1096 strbuf_addbuf(&sb, &dir->basebuf);1097 strbuf_addstr(&sb, dir->exclude_per_dir);1098 el->src = strbuf_detach(&sb, NULL);1099 add_excludes(el->src, el->src, stk->baselen, el, 1,1100 untracked ? &sha1_stat : NULL);1101 }1102 /*1103 * NEEDSWORK: when untracked cache is enabled, prep_exclude()1104 * will first be called in valid_cached_dir() then maybe many1105 * times more in last_exclude_matching(). When the cache is1106 * used, last_exclude_matching() will not be called and1107 * reading .gitignore content will be a waste.1108 *1109 * So when it's called by valid_cached_dir() and we can get1110 * .gitignore SHA-1 from the index (i.e. .gitignore is not1111 * modified on work tree), we could delay reading the1112 * .gitignore content until we absolutely need it in1113 * last_exclude_matching(). Be careful about ignore rule1114 * order, though, if you do that.1115 */1116 if (untracked &&1117 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {1118 invalidate_gitignore(dir->untracked, untracked);1119 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);1120 }1121 dir->exclude_stack = stk;1122 current = stk->baselen;1123 }1124 strbuf_setlen(&dir->basebuf, baselen);1125}11261127/*1128 * Loads the exclude lists for the directory containing pathname, then1129 * scans all exclude lists to determine whether pathname is excluded.1130 * Returns the exclude_list element which matched, or NULL for1131 * undecided.1132 */1133struct exclude *last_exclude_matching(struct dir_struct *dir,1134 const char *pathname,1135 int *dtype_p)1136{1137 int pathlen = strlen(pathname);1138 const char *basename = strrchr(pathname, '/');1139 basename = (basename) ? basename+1 : pathname;11401141 prep_exclude(dir, pathname, basename-pathname);11421143 if (dir->exclude)1144 return dir->exclude;11451146 return last_exclude_matching_from_lists(dir, pathname, pathlen,1147 basename, dtype_p);1148}11491150/*1151 * Loads the exclude lists for the directory containing pathname, then1152 * scans all exclude lists to determine whether pathname is excluded.1153 * Returns 1 if true, otherwise 0.1154 */1155int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)1156{1157 struct exclude *exclude =1158 last_exclude_matching(dir, pathname, dtype_p);1159 if (exclude)1160 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;1161 return 0;1162}11631164static struct dir_entry *dir_entry_new(const char *pathname, int len)1165{1166 struct dir_entry *ent;11671168 ent = xmalloc(sizeof(*ent) + len + 1);1169 ent->len = len;1170 memcpy(ent->name, pathname, len);1171 ent->name[len] = 0;1172 return ent;1173}11741175static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)1176{1177 if (cache_file_exists(pathname, len, ignore_case))1178 return NULL;11791180 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);1181 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);1182}11831184struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)1185{1186 if (!cache_name_is_other(pathname, len))1187 return NULL;11881189 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);1190 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);1191}11921193enum exist_status {1194 index_nonexistent = 0,1195 index_directory,1196 index_gitdir1197};11981199/*1200 * Do not use the alphabetically sorted index to look up1201 * the directory name; instead, use the case insensitive1202 * directory hash.1203 */1204static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)1205{1206 const struct cache_entry *ce = cache_dir_exists(dirname, len);1207 unsigned char endchar;12081209 if (!ce)1210 return index_nonexistent;1211 endchar = ce->name[len];12121213 /*1214 * The cache_entry structure returned will contain this dirname1215 * and possibly additional path components.1216 */1217 if (endchar == '/')1218 return index_directory;12191220 /*1221 * If there are no additional path components, then this cache_entry1222 * represents a submodule. Submodules, despite being directories,1223 * are stored in the cache without a closing slash.1224 */1225 if (!endchar && S_ISGITLINK(ce->ce_mode))1226 return index_gitdir;12271228 /* This should never be hit, but it exists just in case. */1229 return index_nonexistent;1230}12311232/*1233 * The index sorts alphabetically by entry name, which1234 * means that a gitlink sorts as '\0' at the end, while1235 * a directory (which is defined not as an entry, but as1236 * the files it contains) will sort with the '/' at the1237 * end.1238 */1239static enum exist_status directory_exists_in_index(const char *dirname, int len)1240{1241 int pos;12421243 if (ignore_case)1244 return directory_exists_in_index_icase(dirname, len);12451246 pos = cache_name_pos(dirname, len);1247 if (pos < 0)1248 pos = -pos-1;1249 while (pos < active_nr) {1250 const struct cache_entry *ce = active_cache[pos++];1251 unsigned char endchar;12521253 if (strncmp(ce->name, dirname, len))1254 break;1255 endchar = ce->name[len];1256 if (endchar > '/')1257 break;1258 if (endchar == '/')1259 return index_directory;1260 if (!endchar && S_ISGITLINK(ce->ce_mode))1261 return index_gitdir;1262 }1263 return index_nonexistent;1264}12651266/*1267 * When we find a directory when traversing the filesystem, we1268 * have three distinct cases:1269 *1270 * - ignore it1271 * - see it as a directory1272 * - recurse into it1273 *1274 * and which one we choose depends on a combination of existing1275 * git index contents and the flags passed into the directory1276 * traversal routine.1277 *1278 * Case 1: If we *already* have entries in the index under that1279 * directory name, we always recurse into the directory to see1280 * all the files.1281 *1282 * Case 2: If we *already* have that directory name as a gitlink,1283 * we always continue to see it as a gitlink, regardless of whether1284 * there is an actual git directory there or not (it might not1285 * be checked out as a subproject!)1286 *1287 * Case 3: if we didn't have it in the index previously, we1288 * have a few sub-cases:1289 *1290 * (a) if "show_other_directories" is true, we show it as1291 * just a directory, unless "hide_empty_directories" is1292 * also true, in which case we need to check if it contains any1293 * untracked and / or ignored files.1294 * (b) if it looks like a git directory, and we don't have1295 * 'no_gitlinks' set we treat it as a gitlink, and show it1296 * as a directory.1297 * (c) otherwise, we recurse into it.1298 */1299static enum path_treatment treat_directory(struct dir_struct *dir,1300 struct untracked_cache_dir *untracked,1301 const char *dirname, int len, int exclude,1302 const struct path_simplify *simplify)1303{1304 /* The "len-1" is to strip the final '/' */1305 switch (directory_exists_in_index(dirname, len-1)) {1306 case index_directory:1307 return path_recurse;13081309 case index_gitdir:1310 return path_none;13111312 case index_nonexistent:1313 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)1314 break;1315 if (!(dir->flags & DIR_NO_GITLINKS)) {1316 unsigned char sha1[20];1317 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)1318 return path_untracked;1319 }1320 return path_recurse;1321 }13221323 /* This is the "show_other_directories" case */13241325 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))1326 return exclude ? path_excluded : path_untracked;13271328 untracked = lookup_untracked(dir->untracked, untracked, dirname, len);1329 return read_directory_recursive(dir, dirname, len,1330 untracked, 1, simplify);1331}13321333/*1334 * This is an inexact early pruning of any recursive directory1335 * reading - if the path cannot possibly be in the pathspec,1336 * return true, and we'll skip it early.1337 */1338static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)1339{1340 if (simplify) {1341 for (;;) {1342 const char *match = simplify->path;1343 int len = simplify->len;13441345 if (!match)1346 break;1347 if (len > pathlen)1348 len = pathlen;1349 if (!memcmp(path, match, len))1350 return 0;1351 simplify++;1352 }1353 return 1;1354 }1355 return 0;1356}13571358/*1359 * This function tells us whether an excluded path matches a1360 * list of "interesting" pathspecs. That is, whether a path matched1361 * by any of the pathspecs could possibly be ignored by excluding1362 * the specified path. This can happen if:1363 *1364 * 1. the path is mentioned explicitly in the pathspec1365 *1366 * 2. the path is a directory prefix of some element in the1367 * pathspec1368 */1369static int exclude_matches_pathspec(const char *path, int len,1370 const struct path_simplify *simplify)1371{1372 if (simplify) {1373 for (; simplify->path; simplify++) {1374 if (len == simplify->len1375 && !memcmp(path, simplify->path, len))1376 return 1;1377 if (len < simplify->len1378 && simplify->path[len] == '/'1379 && !memcmp(path, simplify->path, len))1380 return 1;1381 }1382 }1383 return 0;1384}13851386static int get_index_dtype(const char *path, int len)1387{1388 int pos;1389 const struct cache_entry *ce;13901391 ce = cache_file_exists(path, len, 0);1392 if (ce) {1393 if (!ce_uptodate(ce))1394 return DT_UNKNOWN;1395 if (S_ISGITLINK(ce->ce_mode))1396 return DT_DIR;1397 /*1398 * Nobody actually cares about the1399 * difference between DT_LNK and DT_REG1400 */1401 return DT_REG;1402 }14031404 /* Try to look it up as a directory */1405 pos = cache_name_pos(path, len);1406 if (pos >= 0)1407 return DT_UNKNOWN;1408 pos = -pos-1;1409 while (pos < active_nr) {1410 ce = active_cache[pos++];1411 if (strncmp(ce->name, path, len))1412 break;1413 if (ce->name[len] > '/')1414 break;1415 if (ce->name[len] < '/')1416 continue;1417 if (!ce_uptodate(ce))1418 break; /* continue? */1419 return DT_DIR;1420 }1421 return DT_UNKNOWN;1422}14231424static int get_dtype(struct dirent *de, const char *path, int len)1425{1426 int dtype = de ? DTYPE(de) : DT_UNKNOWN;1427 struct stat st;14281429 if (dtype != DT_UNKNOWN)1430 return dtype;1431 dtype = get_index_dtype(path, len);1432 if (dtype != DT_UNKNOWN)1433 return dtype;1434 if (lstat(path, &st))1435 return dtype;1436 if (S_ISREG(st.st_mode))1437 return DT_REG;1438 if (S_ISDIR(st.st_mode))1439 return DT_DIR;1440 if (S_ISLNK(st.st_mode))1441 return DT_LNK;1442 return dtype;1443}14441445static enum path_treatment treat_one_path(struct dir_struct *dir,1446 struct untracked_cache_dir *untracked,1447 struct strbuf *path,1448 const struct path_simplify *simplify,1449 int dtype, struct dirent *de)1450{1451 int exclude;1452 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);14531454 if (dtype == DT_UNKNOWN)1455 dtype = get_dtype(de, path->buf, path->len);14561457 /* Always exclude indexed files */1458 if (dtype != DT_DIR && has_path_in_index)1459 return path_none;14601461 /*1462 * When we are looking at a directory P in the working tree,1463 * there are three cases:1464 *1465 * (1) P exists in the index. Everything inside the directory P in1466 * the working tree needs to go when P is checked out from the1467 * index.1468 *1469 * (2) P does not exist in the index, but there is P/Q in the index.1470 * We know P will stay a directory when we check out the contents1471 * of the index, but we do not know yet if there is a directory1472 * P/Q in the working tree to be killed, so we need to recurse.1473 *1474 * (3) P does not exist in the index, and there is no P/Q in the index1475 * to require P to be a directory, either. Only in this case, we1476 * know that everything inside P will not be killed without1477 * recursing.1478 */1479 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&1480 (dtype == DT_DIR) &&1481 !has_path_in_index &&1482 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))1483 return path_none;14841485 exclude = is_excluded(dir, path->buf, &dtype);14861487 /*1488 * Excluded? If we don't explicitly want to show1489 * ignored files, ignore it1490 */1491 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))1492 return path_excluded;14931494 switch (dtype) {1495 default:1496 return path_none;1497 case DT_DIR:1498 strbuf_addch(path, '/');1499 return treat_directory(dir, untracked, path->buf, path->len, exclude,1500 simplify);1501 case DT_REG:1502 case DT_LNK:1503 return exclude ? path_excluded : path_untracked;1504 }1505}15061507static enum path_treatment treat_path_fast(struct dir_struct *dir,1508 struct untracked_cache_dir *untracked,1509 struct cached_dir *cdir,1510 struct strbuf *path,1511 int baselen,1512 const struct path_simplify *simplify)1513{1514 strbuf_setlen(path, baselen);1515 if (!cdir->ucd) {1516 strbuf_addstr(path, cdir->file);1517 return path_untracked;1518 }1519 strbuf_addstr(path, cdir->ucd->name);1520 /* treat_one_path() does this before it calls treat_directory() */1521 if (path->buf[path->len - 1] != '/')1522 strbuf_addch(path, '/');1523 if (cdir->ucd->check_only)1524 /*1525 * check_only is set as a result of treat_directory() getting1526 * to its bottom. Verify again the same set of directories1527 * with check_only set.1528 */1529 return read_directory_recursive(dir, path->buf, path->len,1530 cdir->ucd, 1, simplify);1531 /*1532 * We get path_recurse in the first run when1533 * directory_exists_in_index() returns index_nonexistent. We1534 * are sure that new changes in the index does not impact the1535 * outcome. Return now.1536 */1537 return path_recurse;1538}15391540static enum path_treatment treat_path(struct dir_struct *dir,1541 struct untracked_cache_dir *untracked,1542 struct cached_dir *cdir,1543 struct strbuf *path,1544 int baselen,1545 const struct path_simplify *simplify)1546{1547 int dtype;1548 struct dirent *de = cdir->de;15491550 if (!de)1551 return treat_path_fast(dir, untracked, cdir, path,1552 baselen, simplify);1553 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))1554 return path_none;1555 strbuf_setlen(path, baselen);1556 strbuf_addstr(path, de->d_name);1557 if (simplify_away(path->buf, path->len, simplify))1558 return path_none;15591560 dtype = DTYPE(de);1561 return treat_one_path(dir, untracked, path, simplify, dtype, de);1562}15631564static void add_untracked(struct untracked_cache_dir *dir, const char *name)1565{1566 if (!dir)1567 return;1568 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,1569 dir->untracked_alloc);1570 dir->untracked[dir->untracked_nr++] = xstrdup(name);1571}15721573static int valid_cached_dir(struct dir_struct *dir,1574 struct untracked_cache_dir *untracked,1575 struct strbuf *path,1576 int check_only)1577{1578 struct stat st;15791580 if (!untracked)1581 return 0;15821583 if (stat(path->len ? path->buf : ".", &st)) {1584 invalidate_directory(dir->untracked, untracked);1585 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));1586 return 0;1587 }1588 if (!untracked->valid ||1589 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {1590 if (untracked->valid)1591 invalidate_directory(dir->untracked, untracked);1592 fill_stat_data(&untracked->stat_data, &st);1593 return 0;1594 }15951596 if (untracked->check_only != !!check_only) {1597 invalidate_directory(dir->untracked, untracked);1598 return 0;1599 }16001601 /*1602 * prep_exclude will be called eventually on this directory,1603 * but it's called much later in last_exclude_matching(). We1604 * need it now to determine the validity of the cache for this1605 * path. The next calls will be nearly no-op, the way1606 * prep_exclude() is designed.1607 */1608 if (path->len && path->buf[path->len - 1] != '/') {1609 strbuf_addch(path, '/');1610 prep_exclude(dir, path->buf, path->len);1611 strbuf_setlen(path, path->len - 1);1612 } else1613 prep_exclude(dir, path->buf, path->len);16141615 /* hopefully prep_exclude() haven't invalidated this entry... */1616 return untracked->valid;1617}16181619static int open_cached_dir(struct cached_dir *cdir,1620 struct dir_struct *dir,1621 struct untracked_cache_dir *untracked,1622 struct strbuf *path,1623 int check_only)1624{1625 memset(cdir, 0, sizeof(*cdir));1626 cdir->untracked = untracked;1627 if (valid_cached_dir(dir, untracked, path, check_only))1628 return 0;1629 cdir->fdir = opendir(path->len ? path->buf : ".");1630 if (dir->untracked)1631 dir->untracked->dir_opened++;1632 if (!cdir->fdir)1633 return -1;1634 return 0;1635}16361637static int read_cached_dir(struct cached_dir *cdir)1638{1639 if (cdir->fdir) {1640 cdir->de = readdir(cdir->fdir);1641 if (!cdir->de)1642 return -1;1643 return 0;1644 }1645 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {1646 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];1647 if (!d->recurse) {1648 cdir->nr_dirs++;1649 continue;1650 }1651 cdir->ucd = d;1652 cdir->nr_dirs++;1653 return 0;1654 }1655 cdir->ucd = NULL;1656 if (cdir->nr_files < cdir->untracked->untracked_nr) {1657 struct untracked_cache_dir *d = cdir->untracked;1658 cdir->file = d->untracked[cdir->nr_files++];1659 return 0;1660 }1661 return -1;1662}16631664static void close_cached_dir(struct cached_dir *cdir)1665{1666 if (cdir->fdir)1667 closedir(cdir->fdir);1668 /*1669 * We have gone through this directory and found no untracked1670 * entries. Mark it valid.1671 */1672 if (cdir->untracked) {1673 cdir->untracked->valid = 1;1674 cdir->untracked->recurse = 1;1675 }1676}16771678/*1679 * Read a directory tree. We currently ignore anything but1680 * directories, regular files and symlinks. That's because git1681 * doesn't handle them at all yet. Maybe that will change some1682 * day.1683 *1684 * Also, we ignore the name ".git" (even if it is not a directory).1685 * That likely will not change.1686 *1687 * Returns the most significant path_treatment value encountered in the scan.1688 */1689static enum path_treatment read_directory_recursive(struct dir_struct *dir,1690 const char *base, int baselen,1691 struct untracked_cache_dir *untracked, int check_only,1692 const struct path_simplify *simplify)1693{1694 struct cached_dir cdir;1695 enum path_treatment state, subdir_state, dir_state = path_none;1696 struct strbuf path = STRBUF_INIT;16971698 strbuf_add(&path, base, baselen);16991700 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))1701 goto out;17021703 if (untracked)1704 untracked->check_only = !!check_only;17051706 while (!read_cached_dir(&cdir)) {1707 /* check how the file or directory should be treated */1708 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);17091710 if (state > dir_state)1711 dir_state = state;17121713 /* recurse into subdir if instructed by treat_path */1714 if (state == path_recurse) {1715 struct untracked_cache_dir *ud;1716 ud = lookup_untracked(dir->untracked, untracked,1717 path.buf + baselen,1718 path.len - baselen);1719 subdir_state =1720 read_directory_recursive(dir, path.buf, path.len,1721 ud, check_only, simplify);1722 if (subdir_state > dir_state)1723 dir_state = subdir_state;1724 }17251726 if (check_only) {1727 /* abort early if maximum state has been reached */1728 if (dir_state == path_untracked) {1729 if (cdir.fdir)1730 add_untracked(untracked, path.buf + baselen);1731 break;1732 }1733 /* skip the dir_add_* part */1734 continue;1735 }17361737 /* add the path to the appropriate result list */1738 switch (state) {1739 case path_excluded:1740 if (dir->flags & DIR_SHOW_IGNORED)1741 dir_add_name(dir, path.buf, path.len);1742 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||1743 ((dir->flags & DIR_COLLECT_IGNORED) &&1744 exclude_matches_pathspec(path.buf, path.len,1745 simplify)))1746 dir_add_ignored(dir, path.buf, path.len);1747 break;17481749 case path_untracked:1750 if (dir->flags & DIR_SHOW_IGNORED)1751 break;1752 dir_add_name(dir, path.buf, path.len);1753 if (cdir.fdir)1754 add_untracked(untracked, path.buf + baselen);1755 break;17561757 default:1758 break;1759 }1760 }1761 close_cached_dir(&cdir);1762 out:1763 strbuf_release(&path);17641765 return dir_state;1766}17671768static int cmp_name(const void *p1, const void *p2)1769{1770 const struct dir_entry *e1 = *(const struct dir_entry **)p1;1771 const struct dir_entry *e2 = *(const struct dir_entry **)p2;17721773 return name_compare(e1->name, e1->len, e2->name, e2->len);1774}17751776static struct path_simplify *create_simplify(const char **pathspec)1777{1778 int nr, alloc = 0;1779 struct path_simplify *simplify = NULL;17801781 if (!pathspec)1782 return NULL;17831784 for (nr = 0 ; ; nr++) {1785 const char *match;1786 ALLOC_GROW(simplify, nr + 1, alloc);1787 match = *pathspec++;1788 if (!match)1789 break;1790 simplify[nr].path = match;1791 simplify[nr].len = simple_length(match);1792 }1793 simplify[nr].path = NULL;1794 simplify[nr].len = 0;1795 return simplify;1796}17971798static void free_simplify(struct path_simplify *simplify)1799{1800 free(simplify);1801}18021803static int treat_leading_path(struct dir_struct *dir,1804 const char *path, int len,1805 const struct path_simplify *simplify)1806{1807 struct strbuf sb = STRBUF_INIT;1808 int baselen, rc = 0;1809 const char *cp;1810 int old_flags = dir->flags;18111812 while (len && path[len - 1] == '/')1813 len--;1814 if (!len)1815 return 1;1816 baselen = 0;1817 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;1818 while (1) {1819 cp = path + baselen + !!baselen;1820 cp = memchr(cp, '/', path + len - cp);1821 if (!cp)1822 baselen = len;1823 else1824 baselen = cp - path;1825 strbuf_setlen(&sb, 0);1826 strbuf_add(&sb, path, baselen);1827 if (!is_directory(sb.buf))1828 break;1829 if (simplify_away(sb.buf, sb.len, simplify))1830 break;1831 if (treat_one_path(dir, NULL, &sb, simplify,1832 DT_DIR, NULL) == path_none)1833 break; /* do not recurse into it */1834 if (len <= baselen) {1835 rc = 1;1836 break; /* finished checking */1837 }1838 }1839 strbuf_release(&sb);1840 dir->flags = old_flags;1841 return rc;1842}18431844static const char *get_ident_string(void)1845{1846 static struct strbuf sb = STRBUF_INIT;1847 struct utsname uts;18481849 if (sb.len)1850 return sb.buf;1851 if (uname(&uts))1852 die_errno(_("failed to get kernel name and information"));1853 strbuf_addf(&sb, "Location %s, system %s %s %s", get_git_work_tree(),1854 uts.sysname, uts.release, uts.version);1855 return sb.buf;1856}18571858static int ident_in_untracked(const struct untracked_cache *uc)1859{1860 const char *end = uc->ident.buf + uc->ident.len;1861 const char *p = uc->ident.buf;18621863 for (p = uc->ident.buf; p < end; p += strlen(p) + 1)1864 if (!strcmp(p, get_ident_string()))1865 return 1;1866 return 0;1867}18681869void add_untracked_ident(struct untracked_cache *uc)1870{1871 if (ident_in_untracked(uc))1872 return;1873 strbuf_addstr(&uc->ident, get_ident_string());1874 /* this strbuf contains a list of strings, save NUL too */1875 strbuf_addch(&uc->ident, 0);1876}18771878static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,1879 int base_len,1880 const struct pathspec *pathspec)1881{1882 struct untracked_cache_dir *root;1883 int i;18841885 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))1886 return NULL;18871888 /*1889 * We only support $GIT_DIR/info/exclude and core.excludesfile1890 * as the global ignore rule files. Any other additions1891 * (e.g. from command line) invalidate the cache. This1892 * condition also catches running setup_standard_excludes()1893 * before setting dir->untracked!1894 */1895 if (dir->unmanaged_exclude_files)1896 return NULL;18971898 /*1899 * Optimize for the main use case only: whole-tree git1900 * status. More work involved in treat_leading_path() if we1901 * use cache on just a subset of the worktree. pathspec1902 * support could make the matter even worse.1903 */1904 if (base_len || (pathspec && pathspec->nr))1905 return NULL;19061907 /* Different set of flags may produce different results */1908 if (dir->flags != dir->untracked->dir_flags ||1909 /*1910 * See treat_directory(), case index_nonexistent. Without1911 * this flag, we may need to also cache .git file content1912 * for the resolve_gitlink_ref() call, which we don't.1913 */1914 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||1915 /* We don't support collecting ignore files */1916 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |1917 DIR_COLLECT_IGNORED)))1918 return NULL;19191920 /*1921 * If we use .gitignore in the cache and now you change it to1922 * .gitexclude, everything will go wrong.1923 */1924 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&1925 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))1926 return NULL;19271928 /*1929 * EXC_CMDL is not considered in the cache. If people set it,1930 * skip the cache.1931 */1932 if (dir->exclude_list_group[EXC_CMDL].nr)1933 return NULL;19341935 /*1936 * An optimization in prep_exclude() does not play well with1937 * CE_SKIP_WORKTREE. It's a rare case anyway, if a single1938 * entry has that bit set, disable the whole untracked cache.1939 */1940 for (i = 0; i < active_nr; i++)1941 if (ce_skip_worktree(active_cache[i]))1942 return NULL;19431944 if (!ident_in_untracked(dir->untracked)) {1945 warning(_("Untracked cache is disabled on this system."));1946 return NULL;1947 }19481949 if (!dir->untracked->root) {1950 const int len = sizeof(*dir->untracked->root);1951 dir->untracked->root = xmalloc(len);1952 memset(dir->untracked->root, 0, len);1953 }19541955 /* Validate $GIT_DIR/info/exclude and core.excludesfile */1956 root = dir->untracked->root;1957 if (hashcmp(dir->ss_info_exclude.sha1,1958 dir->untracked->ss_info_exclude.sha1)) {1959 invalidate_gitignore(dir->untracked, root);1960 dir->untracked->ss_info_exclude = dir->ss_info_exclude;1961 }1962 if (hashcmp(dir->ss_excludes_file.sha1,1963 dir->untracked->ss_excludes_file.sha1)) {1964 invalidate_gitignore(dir->untracked, root);1965 dir->untracked->ss_excludes_file = dir->ss_excludes_file;1966 }19671968 /* Make sure this directory is not dropped out at saving phase */1969 root->recurse = 1;1970 return root;1971}19721973int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)1974{1975 struct path_simplify *simplify;1976 struct untracked_cache_dir *untracked;19771978 /*1979 * Check out create_simplify()1980 */1981 if (pathspec)1982 GUARD_PATHSPEC(pathspec,1983 PATHSPEC_FROMTOP |1984 PATHSPEC_MAXDEPTH |1985 PATHSPEC_LITERAL |1986 PATHSPEC_GLOB |1987 PATHSPEC_ICASE |1988 PATHSPEC_EXCLUDE);19891990 if (has_symlink_leading_path(path, len))1991 return dir->nr;19921993 /*1994 * exclude patterns are treated like positive ones in1995 * create_simplify. Usually exclude patterns should be a1996 * subset of positive ones, which has no impacts on1997 * create_simplify().1998 */1999 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);2000 untracked = validate_untracked_cache(dir, len, pathspec);2001 if (!untracked)2002 /*2003 * make sure untracked cache code path is disabled,2004 * e.g. prep_exclude()2005 */2006 dir->untracked = NULL;2007 if (!len || treat_leading_path(dir, path, len, simplify))2008 read_directory_recursive(dir, path, len, untracked, 0, simplify);2009 free_simplify(simplify);2010 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);2011 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);2012 if (dir->untracked) {2013 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);2014 trace_printf_key(&trace_untracked_stats,2015 "node creation: %u\n"2016 "gitignore invalidation: %u\n"2017 "directory invalidation: %u\n"2018 "opendir: %u\n",2019 dir->untracked->dir_created,2020 dir->untracked->gitignore_invalidated,2021 dir->untracked->dir_invalidated,2022 dir->untracked->dir_opened);2023 if (dir->untracked == the_index.untracked &&2024 (dir->untracked->dir_opened ||2025 dir->untracked->gitignore_invalidated ||2026 dir->untracked->dir_invalidated))2027 the_index.cache_changed |= UNTRACKED_CHANGED;2028 if (dir->untracked != the_index.untracked) {2029 free(dir->untracked);2030 dir->untracked = NULL;2031 }2032 }2033 return dir->nr;2034}20352036int file_exists(const char *f)2037{2038 struct stat sb;2039 return lstat(f, &sb) == 0;2040}20412042/*2043 * Given two normalized paths (a trailing slash is ok), if subdir is2044 * outside dir, return -1. Otherwise return the offset in subdir that2045 * can be used as relative path to dir.2046 */2047int dir_inside_of(const char *subdir, const char *dir)2048{2049 int offset = 0;20502051 assert(dir && subdir && *dir && *subdir);20522053 while (*dir && *subdir && *dir == *subdir) {2054 dir++;2055 subdir++;2056 offset++;2057 }20582059 /* hel[p]/me vs hel[l]/yeah */2060 if (*dir && *subdir)2061 return -1;20622063 if (!*subdir)2064 return !*dir ? offset : -1; /* same dir */20652066 /* foo/[b]ar vs foo/[] */2067 if (is_dir_sep(dir[-1]))2068 return is_dir_sep(subdir[-1]) ? offset : -1;20692070 /* foo[/]bar vs foo[] */2071 return is_dir_sep(*subdir) ? offset + 1 : -1;2072}20732074int is_inside_dir(const char *dir)2075{2076 char *cwd;2077 int rc;20782079 if (!dir)2080 return 0;20812082 cwd = xgetcwd();2083 rc = (dir_inside_of(cwd, dir) >= 0);2084 free(cwd);2085 return rc;2086}20872088int is_empty_dir(const char *path)2089{2090 DIR *dir = opendir(path);2091 struct dirent *e;2092 int ret = 1;20932094 if (!dir)2095 return 0;20962097 while ((e = readdir(dir)) != NULL)2098 if (!is_dot_or_dotdot(e->d_name)) {2099 ret = 0;2100 break;2101 }21022103 closedir(dir);2104 return ret;2105}21062107static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)2108{2109 DIR *dir;2110 struct dirent *e;2111 int ret = 0, original_len = path->len, len, kept_down = 0;2112 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);2113 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);2114 unsigned char submodule_head[20];21152116 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&2117 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {2118 /* Do not descend and nuke a nested git work tree. */2119 if (kept_up)2120 *kept_up = 1;2121 return 0;2122 }21232124 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;2125 dir = opendir(path->buf);2126 if (!dir) {2127 if (errno == ENOENT)2128 return keep_toplevel ? -1 : 0;2129 else if (errno == EACCES && !keep_toplevel)2130 /*2131 * An empty dir could be removable even if it2132 * is unreadable:2133 */2134 return rmdir(path->buf);2135 else2136 return -1;2137 }2138 if (path->buf[original_len - 1] != '/')2139 strbuf_addch(path, '/');21402141 len = path->len;2142 while ((e = readdir(dir)) != NULL) {2143 struct stat st;2144 if (is_dot_or_dotdot(e->d_name))2145 continue;21462147 strbuf_setlen(path, len);2148 strbuf_addstr(path, e->d_name);2149 if (lstat(path->buf, &st)) {2150 if (errno == ENOENT)2151 /*2152 * file disappeared, which is what we2153 * wanted anyway2154 */2155 continue;2156 /* fall thru */2157 } else if (S_ISDIR(st.st_mode)) {2158 if (!remove_dir_recurse(path, flag, &kept_down))2159 continue; /* happy */2160 } else if (!only_empty &&2161 (!unlink(path->buf) || errno == ENOENT)) {2162 continue; /* happy, too */2163 }21642165 /* path too long, stat fails, or non-directory still exists */2166 ret = -1;2167 break;2168 }2169 closedir(dir);21702171 strbuf_setlen(path, original_len);2172 if (!ret && !keep_toplevel && !kept_down)2173 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;2174 else if (kept_up)2175 /*2176 * report the uplevel that it is not an error that we2177 * did not rmdir() our directory.2178 */2179 *kept_up = !ret;2180 return ret;2181}21822183int remove_dir_recursively(struct strbuf *path, int flag)2184{2185 return remove_dir_recurse(path, flag, NULL);2186}21872188void setup_standard_excludes(struct dir_struct *dir)2189{2190 const char *path;21912192 dir->exclude_per_dir = ".gitignore";21932194 /* core.excludefile defaulting to $XDG_HOME/git/ignore */2195 if (!excludes_file)2196 excludes_file = xdg_config_home("ignore");2197 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))2198 add_excludes_from_file_1(dir, excludes_file,2199 dir->untracked ? &dir->ss_excludes_file : NULL);22002201 /* per repository user preference */2202 path = git_path("info/exclude");2203 if (!access_or_warn(path, R_OK, 0))2204 add_excludes_from_file_1(dir, path,2205 dir->untracked ? &dir->ss_info_exclude : NULL);2206}22072208int remove_path(const char *name)2209{2210 char *slash;22112212 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)2213 return -1;22142215 slash = strrchr(name, '/');2216 if (slash) {2217 char *dirs = xstrdup(name);2218 slash = dirs + (slash - name);2219 do {2220 *slash = '\0';2221 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));2222 free(dirs);2223 }2224 return 0;2225}22262227/*2228 * Frees memory within dir which was allocated for exclude lists and2229 * the exclude_stack. Does not free dir itself.2230 */2231void clear_directory(struct dir_struct *dir)2232{2233 int i, j;2234 struct exclude_list_group *group;2235 struct exclude_list *el;2236 struct exclude_stack *stk;22372238 for (i = EXC_CMDL; i <= EXC_FILE; i++) {2239 group = &dir->exclude_list_group[i];2240 for (j = 0; j < group->nr; j++) {2241 el = &group->el[j];2242 if (i == EXC_DIRS)2243 free((char *)el->src);2244 clear_exclude_list(el);2245 }2246 free(group->el);2247 }22482249 stk = dir->exclude_stack;2250 while (stk) {2251 struct exclude_stack *prev = stk->prev;2252 free(stk);2253 stk = prev;2254 }2255 strbuf_release(&dir->basebuf);2256}22572258struct ondisk_untracked_cache {2259 struct stat_data info_exclude_stat;2260 struct stat_data excludes_file_stat;2261 uint32_t dir_flags;2262 unsigned char info_exclude_sha1[20];2263 unsigned char excludes_file_sha1[20];2264 char exclude_per_dir[FLEX_ARRAY];2265};22662267#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)22682269struct write_data {2270 int index; /* number of written untracked_cache_dir */2271 struct ewah_bitmap *check_only; /* from untracked_cache_dir */2272 struct ewah_bitmap *valid; /* from untracked_cache_dir */2273 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */2274 struct strbuf out;2275 struct strbuf sb_stat;2276 struct strbuf sb_sha1;2277};22782279static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)2280{2281 to->sd_ctime.sec = htonl(from->sd_ctime.sec);2282 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);2283 to->sd_mtime.sec = htonl(from->sd_mtime.sec);2284 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);2285 to->sd_dev = htonl(from->sd_dev);2286 to->sd_ino = htonl(from->sd_ino);2287 to->sd_uid = htonl(from->sd_uid);2288 to->sd_gid = htonl(from->sd_gid);2289 to->sd_size = htonl(from->sd_size);2290}22912292static void write_one_dir(struct untracked_cache_dir *untracked,2293 struct write_data *wd)2294{2295 struct stat_data stat_data;2296 struct strbuf *out = &wd->out;2297 unsigned char intbuf[16];2298 unsigned int intlen, value;2299 int i = wd->index++;23002301 /*2302 * untracked_nr should be reset whenever valid is clear, but2303 * for safety..2304 */2305 if (!untracked->valid) {2306 untracked->untracked_nr = 0;2307 untracked->check_only = 0;2308 }23092310 if (untracked->check_only)2311 ewah_set(wd->check_only, i);2312 if (untracked->valid) {2313 ewah_set(wd->valid, i);2314 stat_data_to_disk(&stat_data, &untracked->stat_data);2315 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));2316 }2317 if (!is_null_sha1(untracked->exclude_sha1)) {2318 ewah_set(wd->sha1_valid, i);2319 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);2320 }23212322 intlen = encode_varint(untracked->untracked_nr, intbuf);2323 strbuf_add(out, intbuf, intlen);23242325 /* skip non-recurse directories */2326 for (i = 0, value = 0; i < untracked->dirs_nr; i++)2327 if (untracked->dirs[i]->recurse)2328 value++;2329 intlen = encode_varint(value, intbuf);2330 strbuf_add(out, intbuf, intlen);23312332 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);23332334 for (i = 0; i < untracked->untracked_nr; i++)2335 strbuf_add(out, untracked->untracked[i],2336 strlen(untracked->untracked[i]) + 1);23372338 for (i = 0; i < untracked->dirs_nr; i++)2339 if (untracked->dirs[i]->recurse)2340 write_one_dir(untracked->dirs[i], wd);2341}23422343void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)2344{2345 struct ondisk_untracked_cache *ouc;2346 struct write_data wd;2347 unsigned char varbuf[16];2348 int len = 0, varint_len;2349 if (untracked->exclude_per_dir)2350 len = strlen(untracked->exclude_per_dir);2351 ouc = xmalloc(sizeof(*ouc) + len + 1);2352 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);2353 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);2354 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);2355 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);2356 ouc->dir_flags = htonl(untracked->dir_flags);2357 memcpy(ouc->exclude_per_dir, untracked->exclude_per_dir, len + 1);23582359 varint_len = encode_varint(untracked->ident.len, varbuf);2360 strbuf_add(out, varbuf, varint_len);2361 strbuf_add(out, untracked->ident.buf, untracked->ident.len);23622363 strbuf_add(out, ouc, ouc_size(len));2364 free(ouc);2365 ouc = NULL;23662367 if (!untracked->root) {2368 varint_len = encode_varint(0, varbuf);2369 strbuf_add(out, varbuf, varint_len);2370 return;2371 }23722373 wd.index = 0;2374 wd.check_only = ewah_new();2375 wd.valid = ewah_new();2376 wd.sha1_valid = ewah_new();2377 strbuf_init(&wd.out, 1024);2378 strbuf_init(&wd.sb_stat, 1024);2379 strbuf_init(&wd.sb_sha1, 1024);2380 write_one_dir(untracked->root, &wd);23812382 varint_len = encode_varint(wd.index, varbuf);2383 strbuf_add(out, varbuf, varint_len);2384 strbuf_addbuf(out, &wd.out);2385 ewah_serialize_strbuf(wd.valid, out);2386 ewah_serialize_strbuf(wd.check_only, out);2387 ewah_serialize_strbuf(wd.sha1_valid, out);2388 strbuf_addbuf(out, &wd.sb_stat);2389 strbuf_addbuf(out, &wd.sb_sha1);2390 strbuf_addch(out, '\0'); /* safe guard for string lists */23912392 ewah_free(wd.valid);2393 ewah_free(wd.check_only);2394 ewah_free(wd.sha1_valid);2395 strbuf_release(&wd.out);2396 strbuf_release(&wd.sb_stat);2397 strbuf_release(&wd.sb_sha1);2398}23992400static void free_untracked(struct untracked_cache_dir *ucd)2401{2402 int i;2403 if (!ucd)2404 return;2405 for (i = 0; i < ucd->dirs_nr; i++)2406 free_untracked(ucd->dirs[i]);2407 for (i = 0; i < ucd->untracked_nr; i++)2408 free(ucd->untracked[i]);2409 free(ucd->untracked);2410 free(ucd->dirs);2411 free(ucd);2412}24132414void free_untracked_cache(struct untracked_cache *uc)2415{2416 if (uc)2417 free_untracked(uc->root);2418 free(uc);2419}24202421struct read_data {2422 int index;2423 struct untracked_cache_dir **ucd;2424 struct ewah_bitmap *check_only;2425 struct ewah_bitmap *valid;2426 struct ewah_bitmap *sha1_valid;2427 const unsigned char *data;2428 const unsigned char *end;2429};24302431static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)2432{2433 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);2434 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);2435 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);2436 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);2437 to->sd_dev = get_be32(&from->sd_dev);2438 to->sd_ino = get_be32(&from->sd_ino);2439 to->sd_uid = get_be32(&from->sd_uid);2440 to->sd_gid = get_be32(&from->sd_gid);2441 to->sd_size = get_be32(&from->sd_size);2442}24432444static int read_one_dir(struct untracked_cache_dir **untracked_,2445 struct read_data *rd)2446{2447 struct untracked_cache_dir ud, *untracked;2448 const unsigned char *next, *data = rd->data, *end = rd->end;2449 unsigned int value;2450 int i, len;24512452 memset(&ud, 0, sizeof(ud));24532454 next = data;2455 value = decode_varint(&next);2456 if (next > end)2457 return -1;2458 ud.recurse = 1;2459 ud.untracked_alloc = value;2460 ud.untracked_nr = value;2461 if (ud.untracked_nr)2462 ud.untracked = xmalloc(sizeof(*ud.untracked) * ud.untracked_nr);2463 data = next;24642465 next = data;2466 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);2467 if (next > end)2468 return -1;2469 ud.dirs = xmalloc(sizeof(*ud.dirs) * ud.dirs_nr);2470 data = next;24712472 len = strlen((const char *)data);2473 next = data + len + 1;2474 if (next > rd->end)2475 return -1;2476 *untracked_ = untracked = xmalloc(sizeof(*untracked) + len);2477 memcpy(untracked, &ud, sizeof(ud));2478 memcpy(untracked->name, data, len + 1);2479 data = next;24802481 for (i = 0; i < untracked->untracked_nr; i++) {2482 len = strlen((const char *)data);2483 next = data + len + 1;2484 if (next > rd->end)2485 return -1;2486 untracked->untracked[i] = xstrdup((const char*)data);2487 data = next;2488 }24892490 rd->ucd[rd->index++] = untracked;2491 rd->data = data;24922493 for (i = 0; i < untracked->dirs_nr; i++) {2494 len = read_one_dir(untracked->dirs + i, rd);2495 if (len < 0)2496 return -1;2497 }2498 return 0;2499}25002501static void set_check_only(size_t pos, void *cb)2502{2503 struct read_data *rd = cb;2504 struct untracked_cache_dir *ud = rd->ucd[pos];2505 ud->check_only = 1;2506}25072508static void read_stat(size_t pos, void *cb)2509{2510 struct read_data *rd = cb;2511 struct untracked_cache_dir *ud = rd->ucd[pos];2512 if (rd->data + sizeof(struct stat_data) > rd->end) {2513 rd->data = rd->end + 1;2514 return;2515 }2516 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);2517 rd->data += sizeof(struct stat_data);2518 ud->valid = 1;2519}25202521static void read_sha1(size_t pos, void *cb)2522{2523 struct read_data *rd = cb;2524 struct untracked_cache_dir *ud = rd->ucd[pos];2525 if (rd->data + 20 > rd->end) {2526 rd->data = rd->end + 1;2527 return;2528 }2529 hashcpy(ud->exclude_sha1, rd->data);2530 rd->data += 20;2531}25322533static void load_sha1_stat(struct sha1_stat *sha1_stat,2534 const struct stat_data *stat,2535 const unsigned char *sha1)2536{2537 stat_data_from_disk(&sha1_stat->stat, stat);2538 hashcpy(sha1_stat->sha1, sha1);2539 sha1_stat->valid = 1;2540}25412542struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)2543{2544 const struct ondisk_untracked_cache *ouc;2545 struct untracked_cache *uc;2546 struct read_data rd;2547 const unsigned char *next = data, *end = (const unsigned char *)data + sz;2548 const char *ident;2549 int ident_len, len;25502551 if (sz <= 1 || end[-1] != '\0')2552 return NULL;2553 end--;25542555 ident_len = decode_varint(&next);2556 if (next + ident_len > end)2557 return NULL;2558 ident = (const char *)next;2559 next += ident_len;25602561 ouc = (const struct ondisk_untracked_cache *)next;2562 if (next + ouc_size(0) > end)2563 return NULL;25642565 uc = xcalloc(1, sizeof(*uc));2566 strbuf_init(&uc->ident, ident_len);2567 strbuf_add(&uc->ident, ident, ident_len);2568 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,2569 ouc->info_exclude_sha1);2570 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,2571 ouc->excludes_file_sha1);2572 uc->dir_flags = get_be32(&ouc->dir_flags);2573 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);2574 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */2575 next += ouc_size(strlen(ouc->exclude_per_dir));2576 if (next >= end)2577 goto done2;25782579 len = decode_varint(&next);2580 if (next > end || len == 0)2581 goto done2;25822583 rd.valid = ewah_new();2584 rd.check_only = ewah_new();2585 rd.sha1_valid = ewah_new();2586 rd.data = next;2587 rd.end = end;2588 rd.index = 0;2589 rd.ucd = xmalloc(sizeof(*rd.ucd) * len);25902591 if (read_one_dir(&uc->root, &rd) || rd.index != len)2592 goto done;25932594 next = rd.data;2595 len = ewah_read_mmap(rd.valid, next, end - next);2596 if (len < 0)2597 goto done;25982599 next += len;2600 len = ewah_read_mmap(rd.check_only, next, end - next);2601 if (len < 0)2602 goto done;26032604 next += len;2605 len = ewah_read_mmap(rd.sha1_valid, next, end - next);2606 if (len < 0)2607 goto done;26082609 ewah_each_bit(rd.check_only, set_check_only, &rd);2610 rd.data = next + len;2611 ewah_each_bit(rd.valid, read_stat, &rd);2612 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);2613 next = rd.data;26142615done:2616 free(rd.ucd);2617 ewah_free(rd.valid);2618 ewah_free(rd.check_only);2619 ewah_free(rd.sha1_valid);2620done2:2621 if (next != end) {2622 free_untracked_cache(uc);2623 uc = NULL;2624 }2625 return uc;2626}26272628void untracked_cache_invalidate_path(struct index_state *istate,2629 const char *path)2630{2631 const char *sep;2632 struct untracked_cache_dir *d;2633 if (!istate->untracked || !istate->untracked->root)2634 return;2635 sep = strrchr(path, '/');2636 if (sep)2637 d = lookup_untracked(istate->untracked,2638 istate->untracked->root,2639 path, sep - path);2640 else2641 d = istate->untracked->root;2642 istate->untracked->dir_invalidated++;2643 d->valid = 0;2644 d->untracked_nr = 0;2645}26462647void untracked_cache_remove_from_index(struct index_state *istate,2648 const char *path)2649{2650 untracked_cache_invalidate_path(istate, path);2651}26522653void untracked_cache_add_to_index(struct index_state *istate,2654 const char *path)2655{2656 untracked_cache_invalidate_path(istate, path);2657}