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 { 38DIR*fdir; 39struct untracked_cache_dir *untracked; 40int nr_files; 41int nr_dirs; 42 43struct dirent *de; 44const char*file; 45struct untracked_cache_dir *ucd; 46}; 47 48static enum path_treatment read_directory_recursive(struct dir_struct *dir, 49const char*path,int len,struct untracked_cache_dir *untracked, 50int check_only,const struct pathspec *pathspec); 51static intget_dtype(struct dirent *de,const char*path,int len); 52 53intfspathcmp(const char*a,const char*b) 54{ 55return ignore_case ?strcasecmp(a, b) :strcmp(a, b); 56} 57 58intfspathncmp(const char*a,const char*b,size_t count) 59{ 60return ignore_case ?strncasecmp(a, b, count) :strncmp(a, b, count); 61} 62 63intgit_fnmatch(const struct pathspec_item *item, 64const char*pattern,const char*string, 65int prefix) 66{ 67if(prefix >0) { 68if(ps_strncmp(item, pattern, string, prefix)) 69return WM_NOMATCH; 70 pattern += prefix; 71 string += prefix; 72} 73if(item->flags & PATHSPEC_ONESTAR) { 74int pattern_len =strlen(++pattern); 75int string_len =strlen(string); 76return string_len < pattern_len || 77ps_strcmp(item, pattern, 78 string + string_len - pattern_len); 79} 80if(item->magic & PATHSPEC_GLOB) 81returnwildmatch(pattern, string, 82 WM_PATHNAME | 83(item->magic & PATHSPEC_ICASE ? WM_CASEFOLD :0), 84 NULL); 85else 86/* wildmatch has not learned no FNM_PATHNAME mode yet */ 87returnwildmatch(pattern, string, 88 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD :0, 89 NULL); 90} 91 92static intfnmatch_icase_mem(const char*pattern,int patternlen, 93const char*string,int stringlen, 94int flags) 95{ 96int match_status; 97struct strbuf pat_buf = STRBUF_INIT; 98struct strbuf str_buf = STRBUF_INIT; 99const char*use_pat = pattern; 100const char*use_str = string; 101 102if(pattern[patternlen]) { 103strbuf_add(&pat_buf, pattern, patternlen); 104 use_pat = pat_buf.buf; 105} 106if(string[stringlen]) { 107strbuf_add(&str_buf, string, stringlen); 108 use_str = str_buf.buf; 109} 110 111if(ignore_case) 112 flags |= WM_CASEFOLD; 113 match_status =wildmatch(use_pat, use_str, flags, NULL); 114 115strbuf_release(&pat_buf); 116strbuf_release(&str_buf); 117 118return match_status; 119} 120 121static size_tcommon_prefix_len(const struct pathspec *pathspec) 122{ 123int n; 124size_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 */ 133GUARD_PATHSPEC(pathspec, 134 PATHSPEC_FROMTOP | 135 PATHSPEC_MAXDEPTH | 136 PATHSPEC_LITERAL | 137 PATHSPEC_GLOB | 138 PATHSPEC_ICASE | 139 PATHSPEC_EXCLUDE | 140 PATHSPEC_ATTR); 141 142for(n =0; n < pathspec->nr; n++) { 143size_t i =0, len =0, item_len; 144if(pathspec->items[n].magic & PATHSPEC_EXCLUDE) 145continue; 146if(pathspec->items[n].magic & PATHSPEC_ICASE) 147 item_len = pathspec->items[n].prefix; 148else 149 item_len = pathspec->items[n].nowildcard_len; 150while(i < item_len && (n ==0|| i < max)) { 151char c = pathspec->items[n].match[i]; 152if(c != pathspec->items[0].match[i]) 153break; 154if(c =='/') 155 len = i +1; 156 i++; 157} 158if(n ==0|| len < max) { 159 max = len; 160if(!max) 161break; 162} 163} 164return 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{ 173unsigned long len =common_prefix_len(pathspec); 174 175return len ?xmemdupz(pathspec->items[0].match, len) : NULL; 176} 177 178intfill_directory(struct dir_struct *dir,const struct pathspec *pathspec) 179{ 180const char*prefix; 181size_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 */ 191read_directory(dir, prefix, prefix_len, pathspec); 192 193return prefix_len; 194} 195 196intwithin_depth(const char*name,int namelen, 197int depth,int max_depth) 198{ 199const char*cp = name, *cpe = name + namelen; 200 201while(cp < cpe) { 202if(*cp++ !='/') 203continue; 204 depth++; 205if(depth > max_depth) 206return0; 207} 208return1; 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 intmatch_attrs(const char*name,int namelen, 216const struct pathspec_item *item) 217{ 218int i; 219 220git_check_attr(name, item->attr_check); 221for(i =0; i < item->attr_match_nr; i++) { 222const char*value; 223int matched; 224enum attr_match_mode match_mode; 225 226 value = item->attr_check->items[i].value; 227 match_mode = item->attr_match[i].match_mode; 228 229if(ATTR_TRUE(value)) 230 matched = (match_mode == MATCH_SET); 231else if(ATTR_FALSE(value)) 232 matched = (match_mode == MATCH_UNSET); 233else if(ATTR_UNSET(value)) 234 matched = (match_mode == MATCH_UNSPECIFIED); 235else 236 matched = (match_mode == MATCH_VALUE && 237!strcmp(item->attr_match[i].value, value)); 238if(!matched) 239return0; 240} 241 242return1; 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 intmatch_pathspec_item(const struct pathspec_item *item,int prefix, 258const char*name,int namelen,unsigned flags) 259{ 260/* name/namelen has prefix cut off by caller */ 261const char*match = item->match + prefix; 262int 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 */ 293if(item->prefix && (item->magic & PATHSPEC_ICASE) && 294strncmp(item->match, name - prefix, item->prefix)) 295return0; 296 297if(item->attr_match_nr && !match_attrs(name, namelen, item)) 298return0; 299 300/* If the match was just the prefix, we matched */ 301if(!*match) 302return MATCHED_RECURSIVELY; 303 304if(matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) { 305if(matchlen == namelen) 306return MATCHED_EXACTLY; 307 308if(match[matchlen-1] =='/'|| name[matchlen] =='/') 309return MATCHED_RECURSIVELY; 310}else if((flags & DO_MATCH_DIRECTORY) && 311 match[matchlen -1] =='/'&& 312 namelen == matchlen -1&& 313!ps_strncmp(item, match, name, namelen)) 314return MATCHED_EXACTLY; 315 316if(item->nowildcard_len < item->len && 317!git_fnmatch(item, match, name, 318 item->nowildcard_len - prefix)) 319return MATCHED_FNMATCH; 320 321/* Perform checks to see if "name" is a super set of the pathspec */ 322if(flags & DO_MATCH_SUBMODULE) { 323/* name is a literal prefix of the pathspec */ 324if((namelen < matchlen) && 325(match[namelen] =='/') && 326!ps_strncmp(item, match, name, namelen)) 327return MATCHED_RECURSIVELY; 328 329/* name" doesn't match up to the first wild character */ 330if(item->nowildcard_len < item->len && 331ps_strncmp(item, match, name, 332 item->nowildcard_len - prefix)) 333return0; 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 */ 344return MATCHED_RECURSIVELY; 345} 346 347return0; 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 intdo_match_pathspec(const struct pathspec *ps, 366const char*name,int namelen, 367int prefix,char*seen, 368unsigned flags) 369{ 370int i, retval =0, exclude = flags & DO_MATCH_EXCLUDE; 371 372GUARD_PATHSPEC(ps, 373 PATHSPEC_FROMTOP | 374 PATHSPEC_MAXDEPTH | 375 PATHSPEC_LITERAL | 376 PATHSPEC_GLOB | 377 PATHSPEC_ICASE | 378 PATHSPEC_EXCLUDE | 379 PATHSPEC_ATTR); 380 381if(!ps->nr) { 382if(!ps->recursive || 383!(ps->magic & PATHSPEC_MAXDEPTH) || 384 ps->max_depth == -1) 385return MATCHED_RECURSIVELY; 386 387if(within_depth(name, namelen,0, ps->max_depth)) 388return MATCHED_EXACTLY; 389else 390return0; 391} 392 393 name += prefix; 394 namelen -= prefix; 395 396for(i = ps->nr -1; i >=0; i--) { 397int how; 398 399if((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) || 400( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE))) 401continue; 402 403if(seen && seen[i] == MATCHED_EXACTLY) 404continue; 405/* 406 * Make exclude patterns optional and never report 407 * "pathspec ':(exclude)foo' matches no files" 408 */ 409if(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); 413if(ps->recursive && 414(ps->magic & PATHSPEC_MAXDEPTH) && 415 ps->max_depth != -1&& 416 how && how != MATCHED_FNMATCH) { 417int len = ps->items[i].len; 418if(name[len] =='/') 419 len++; 420if(within_depth(name+len, namelen-len,0, ps->max_depth)) 421 how = MATCHED_EXACTLY; 422else 423 how =0; 424} 425if(how) { 426if(retval < how) 427 retval = how; 428if(seen && seen[i] < how) 429 seen[i] = how; 430} 431} 432return retval; 433} 434 435intmatch_pathspec(const struct pathspec *ps, 436const char*name,int namelen, 437int prefix,char*seen,int is_dir) 438{ 439int positive, negative; 440unsigned flags = is_dir ? DO_MATCH_DIRECTORY :0; 441 positive =do_match_pathspec(ps, name, namelen, 442 prefix, seen, flags); 443if(!(ps->magic & PATHSPEC_EXCLUDE) || !positive) 444return positive; 445 negative =do_match_pathspec(ps, name, namelen, 446 prefix, seen, 447 flags | DO_MATCH_EXCLUDE); 448return negative ?0: positive; 449} 450 451/** 452 * Check if a submodule is a superset of the pathspec 453 */ 454intsubmodule_path_match(const struct pathspec *ps, 455const char*submodule_name, 456char*seen) 457{ 458int matched =do_match_pathspec(ps, submodule_name, 459strlen(submodule_name), 4600, seen, 461 DO_MATCH_DIRECTORY | 462 DO_MATCH_SUBMODULE); 463return matched; 464} 465 466intreport_path_error(const char*ps_matched, 467const struct pathspec *pathspec, 468const char*prefix) 469{ 470/* 471 * Make sure all pathspec matched; otherwise it is an error. 472 */ 473int num, errors =0; 474for(num =0; num < pathspec->nr; num++) { 475int other, found_dup; 476 477if(ps_matched[num]) 478continue; 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 */ 485for(found_dup = other =0; 486!found_dup && other < pathspec->nr; 487 other++) { 488if(other == num || !ps_matched[other]) 489continue; 490if(!strcmp(pathspec->items[other].original, 491 pathspec->items[num].original)) 492/* 493 * Ok, we have a match already. 494 */ 495 found_dup =1; 496} 497if(found_dup) 498continue; 499 500error("pathspec '%s' did not match any file(s) known to git.", 501 pathspec->items[num].original); 502 errors++; 503} 504return errors; 505} 506 507/* 508 * Return the length of the "simple" part of a path match limiter. 509 */ 510intsimple_length(const char*match) 511{ 512int len = -1; 513 514for(;;) { 515unsigned char c = *match++; 516 len++; 517if(c =='\0'||is_glob_special(c)) 518return len; 519} 520} 521 522intno_wildcard(const char*string) 523{ 524return string[simple_length(string)] =='\0'; 525} 526 527voidparse_exclude_pattern(const char**pattern, 528int*patternlen, 529unsigned*flags, 530int*nowildcardlen) 531{ 532const char*p = *pattern; 533size_t i, len; 534 535*flags =0; 536if(*p =='!') { 537*flags |= EXC_FLAG_NEGATIVE; 538 p++; 539} 540 len =strlen(p); 541if(len && p[len -1] =='/') { 542 len--; 543*flags |= EXC_FLAG_MUSTBEDIR; 544} 545for(i =0; i < len; i++) { 546if(p[i] =='/') 547break; 548} 549if(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 */ 557if(*nowildcardlen > len) 558*nowildcardlen = len; 559if(*p =='*'&&no_wildcard(p +1)) 560*flags |= EXC_FLAG_ENDSWITH; 561*pattern = p; 562*patternlen = len; 563} 564 565voidadd_exclude(const char*string,const char*base, 566int baselen,struct exclude_list *el,int srcpos) 567{ 568struct exclude *x; 569int patternlen; 570unsigned flags; 571int nowildcardlen; 572 573parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen); 574if(flags & EXC_FLAG_MUSTBEDIR) { 575FLEXPTR_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; 586ALLOC_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 struct index_state *istate, 592const char*path,size_t*size, 593struct sha1_stat *sha1_stat) 594{ 595int pos, len; 596unsigned long sz; 597enum object_type type; 598void*data; 599 600 len =strlen(path); 601 pos =index_name_pos(istate, path, len); 602if(pos <0) 603return NULL; 604if(!ce_skip_worktree(istate->cache[pos])) 605return NULL; 606 data =read_sha1_file(istate->cache[pos]->oid.hash, &type, &sz); 607if(!data || type != OBJ_BLOB) { 608free(data); 609return NULL; 610} 611*size =xsize_t(sz); 612if(sha1_stat) { 613memset(&sha1_stat->stat,0,sizeof(sha1_stat->stat)); 614hashcpy(sha1_stat->sha1, istate->cache[pos]->oid.hash); 615} 616return data; 617} 618 619/* 620 * Frees memory within el which was allocated for exclude patterns and 621 * the file buffer. Does not free el itself. 622 */ 623voidclear_exclude_list(struct exclude_list *el) 624{ 625int i; 626 627for(i =0; i < el->nr; i++) 628free(el->excludes[i]); 629free(el->excludes); 630free(el->filebuf); 631 632memset(el,0,sizeof(*el)); 633} 634 635static voidtrim_trailing_spaces(char*buf) 636{ 637char*p, *last_space = NULL; 638 639for(p = buf; *p; p++) 640switch(*p) { 641case' ': 642if(!last_space) 643 last_space = p; 644break; 645case'\\': 646 p++; 647if(!*p) 648return; 649/* fallthrough */ 650default: 651 last_space = NULL; 652} 653 654if(last_space) 655*last_space ='\0'; 656} 657 658/* 659 * Given a subdirectory name and "dir" of the current directory, 660 * search the subdir in "dir" and return it, or create a new one if it 661 * does not exist in "dir". 662 * 663 * If "name" has the trailing slash, it'll be excluded in the search. 664 */ 665static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc, 666struct untracked_cache_dir *dir, 667const char*name,int len) 668{ 669int first, last; 670struct untracked_cache_dir *d; 671if(!dir) 672return NULL; 673if(len && name[len -1] =='/') 674 len--; 675 first =0; 676 last = dir->dirs_nr; 677while(last > first) { 678int cmp, next = (last + first) >>1; 679 d = dir->dirs[next]; 680 cmp =strncmp(name, d->name, len); 681if(!cmp &&strlen(d->name) > len) 682 cmp = -1; 683if(!cmp) 684return d; 685if(cmp <0) { 686 last = next; 687continue; 688} 689 first = next+1; 690} 691 692 uc->dir_created++; 693FLEX_ALLOC_MEM(d, name, name, len); 694 695ALLOC_GROW(dir->dirs, dir->dirs_nr +1, dir->dirs_alloc); 696memmove(dir->dirs + first +1, dir->dirs + first, 697(dir->dirs_nr - first) *sizeof(*dir->dirs)); 698 dir->dirs_nr++; 699 dir->dirs[first] = d; 700return d; 701} 702 703static voiddo_invalidate_gitignore(struct untracked_cache_dir *dir) 704{ 705int i; 706 dir->valid =0; 707 dir->untracked_nr =0; 708for(i =0; i < dir->dirs_nr; i++) 709do_invalidate_gitignore(dir->dirs[i]); 710} 711 712static voidinvalidate_gitignore(struct untracked_cache *uc, 713struct untracked_cache_dir *dir) 714{ 715 uc->gitignore_invalidated++; 716do_invalidate_gitignore(dir); 717} 718 719static voidinvalidate_directory(struct untracked_cache *uc, 720struct untracked_cache_dir *dir) 721{ 722int i; 723 uc->dir_invalidated++; 724 dir->valid =0; 725 dir->untracked_nr =0; 726for(i =0; i < dir->dirs_nr; i++) 727 dir->dirs[i]->recurse =0; 728} 729 730/* 731 * Given a file with name "fname", read it (either from disk, or from 732 * the index if "check_index" is non-zero), parse it and store the 733 * exclude rules in "el". 734 * 735 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill 736 * stat data from disk (only valid if add_excludes returns zero). If 737 * ss_valid is non-zero, "ss" must contain good value as input. 738 */ 739static intadd_excludes(const char*fname,const char*base,int baselen, 740struct exclude_list *el,int check_index, 741struct sha1_stat *sha1_stat) 742{ 743struct stat st; 744int fd, i, lineno =1; 745size_t size =0; 746char*buf, *entry; 747 748 fd =open(fname, O_RDONLY); 749if(fd <0||fstat(fd, &st) <0) { 750if(errno != ENOENT) 751warn_on_inaccessible(fname); 752if(0<= fd) 753close(fd); 754if(!check_index || 755(buf =read_skip_worktree_file_from_index(&the_index, fname, &size, sha1_stat)) == NULL) 756return-1; 757if(size ==0) { 758free(buf); 759return0; 760} 761if(buf[size-1] !='\n') { 762 buf =xrealloc(buf,st_add(size,1)); 763 buf[size++] ='\n'; 764} 765}else{ 766 size =xsize_t(st.st_size); 767if(size ==0) { 768if(sha1_stat) { 769fill_stat_data(&sha1_stat->stat, &st); 770hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN); 771 sha1_stat->valid =1; 772} 773close(fd); 774return0; 775} 776 buf =xmallocz(size); 777if(read_in_full(fd, buf, size) != size) { 778free(buf); 779close(fd); 780return-1; 781} 782 buf[size++] ='\n'; 783close(fd); 784if(sha1_stat) { 785int pos; 786if(sha1_stat->valid && 787!match_stat_data_racy(&the_index, &sha1_stat->stat, &st)) 788;/* no content change, ss->sha1 still good */ 789else if(check_index && 790(pos =index_name_pos(&the_index, fname,strlen(fname))) >=0&& 791!ce_stage(the_index.cache[pos]) && 792ce_uptodate(the_index.cache[pos]) && 793!would_convert_to_git(fname)) 794hashcpy(sha1_stat->sha1, 795 the_index.cache[pos]->oid.hash); 796else 797hash_sha1_file(buf, size,"blob", sha1_stat->sha1); 798fill_stat_data(&sha1_stat->stat, &st); 799 sha1_stat->valid =1; 800} 801} 802 803 el->filebuf = buf; 804 805if(skip_utf8_bom(&buf, size)) 806 size -= buf - el->filebuf; 807 808 entry = buf; 809 810for(i =0; i < size; i++) { 811if(buf[i] =='\n') { 812if(entry != buf + i && entry[0] !='#') { 813 buf[i - (i && buf[i-1] =='\r')] =0; 814trim_trailing_spaces(entry); 815add_exclude(entry, base, baselen, el, lineno); 816} 817 lineno++; 818 entry = buf + i +1; 819} 820} 821return0; 822} 823 824intadd_excludes_from_file_to_list(const char*fname,const char*base, 825int baselen,struct exclude_list *el, 826int check_index) 827{ 828returnadd_excludes(fname, base, baselen, el, check_index, NULL); 829} 830 831struct exclude_list *add_exclude_list(struct dir_struct *dir, 832int group_type,const char*src) 833{ 834struct exclude_list *el; 835struct exclude_list_group *group; 836 837 group = &dir->exclude_list_group[group_type]; 838ALLOC_GROW(group->el, group->nr +1, group->alloc); 839 el = &group->el[group->nr++]; 840memset(el,0,sizeof(*el)); 841 el->src = src; 842return el; 843} 844 845/* 846 * Used to set up core.excludesfile and .git/info/exclude lists. 847 */ 848static voidadd_excludes_from_file_1(struct dir_struct *dir,const char*fname, 849struct sha1_stat *sha1_stat) 850{ 851struct exclude_list *el; 852/* 853 * catch setup_standard_excludes() that's called before 854 * dir->untracked is assigned. That function behaves 855 * differently when dir->untracked is non-NULL. 856 */ 857if(!dir->untracked) 858 dir->unmanaged_exclude_files++; 859 el =add_exclude_list(dir, EXC_FILE, fname); 860if(add_excludes(fname,"",0, el,0, sha1_stat) <0) 861die("cannot use%sas an exclude file", fname); 862} 863 864voidadd_excludes_from_file(struct dir_struct *dir,const char*fname) 865{ 866 dir->unmanaged_exclude_files++;/* see validate_untracked_cache() */ 867add_excludes_from_file_1(dir, fname, NULL); 868} 869 870intmatch_basename(const char*basename,int basenamelen, 871const char*pattern,int prefix,int patternlen, 872unsigned flags) 873{ 874if(prefix == patternlen) { 875if(patternlen == basenamelen && 876!fspathncmp(pattern, basename, basenamelen)) 877return1; 878}else if(flags & EXC_FLAG_ENDSWITH) { 879/* "*literal" matching against "fooliteral" */ 880if(patternlen -1<= basenamelen && 881!fspathncmp(pattern +1, 882 basename + basenamelen - (patternlen -1), 883 patternlen -1)) 884return1; 885}else{ 886if(fnmatch_icase_mem(pattern, patternlen, 887 basename, basenamelen, 8880) ==0) 889return1; 890} 891return0; 892} 893 894intmatch_pathname(const char*pathname,int pathlen, 895const char*base,int baselen, 896const char*pattern,int prefix,int patternlen, 897unsigned flags) 898{ 899const char*name; 900int namelen; 901 902/* 903 * match with FNM_PATHNAME; the pattern has base implicitly 904 * in front of it. 905 */ 906if(*pattern =='/') { 907 pattern++; 908 patternlen--; 909 prefix--; 910} 911 912/* 913 * baselen does not count the trailing slash. base[] may or 914 * may not end with a trailing slash though. 915 */ 916if(pathlen < baselen +1|| 917(baselen && pathname[baselen] !='/') || 918fspathncmp(pathname, base, baselen)) 919return0; 920 921 namelen = baselen ? pathlen - baselen -1: pathlen; 922 name = pathname + pathlen - namelen; 923 924if(prefix) { 925/* 926 * if the non-wildcard part is longer than the 927 * remaining pathname, surely it cannot match. 928 */ 929if(prefix > namelen) 930return0; 931 932if(fspathncmp(pattern, name, prefix)) 933return0; 934 pattern += prefix; 935 patternlen -= prefix; 936 name += prefix; 937 namelen -= prefix; 938 939/* 940 * If the whole pattern did not have a wildcard, 941 * then our prefix match is all we need; we 942 * do not need to call fnmatch at all. 943 */ 944if(!patternlen && !namelen) 945return1; 946} 947 948returnfnmatch_icase_mem(pattern, patternlen, 949 name, namelen, 950 WM_PATHNAME) ==0; 951} 952 953/* 954 * Scan the given exclude list in reverse to see whether pathname 955 * should be ignored. The first match (i.e. the last on the list), if 956 * any, determines the fate. Returns the exclude_list element which 957 * matched, or NULL for undecided. 958 */ 959static struct exclude *last_exclude_matching_from_list(const char*pathname, 960int pathlen, 961const char*basename, 962int*dtype, 963struct exclude_list *el) 964{ 965struct exclude *exc = NULL;/* undecided */ 966int i; 967 968if(!el->nr) 969return NULL;/* undefined */ 970 971for(i = el->nr -1;0<= i; i--) { 972struct exclude *x = el->excludes[i]; 973const char*exclude = x->pattern; 974int prefix = x->nowildcardlen; 975 976if(x->flags & EXC_FLAG_MUSTBEDIR) { 977if(*dtype == DT_UNKNOWN) 978*dtype =get_dtype(NULL, pathname, pathlen); 979if(*dtype != DT_DIR) 980continue; 981} 982 983if(x->flags & EXC_FLAG_NODIR) { 984if(match_basename(basename, 985 pathlen - (basename - pathname), 986 exclude, prefix, x->patternlen, 987 x->flags)) { 988 exc = x; 989break; 990} 991continue; 992} 993 994assert(x->baselen ==0|| x->base[x->baselen -1] =='/'); 995if(match_pathname(pathname, pathlen, 996 x->base, x->baselen ? x->baselen -1:0, 997 exclude, prefix, x->patternlen, x->flags)) { 998 exc = x; 999break;1000}1001}1002return exc;1003}10041005/*1006 * Scan the list and let the last match determine the fate.1007 * Return 1 for exclude, 0 for include and -1 for undecided.1008 */1009intis_excluded_from_list(const char*pathname,1010int pathlen,const char*basename,int*dtype,1011struct exclude_list *el)1012{1013struct exclude *exclude;1014 exclude =last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);1015if(exclude)1016return exclude->flags & EXC_FLAG_NEGATIVE ?0:1;1017return-1;/* undecided */1018}10191020static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,1021const char*pathname,int pathlen,const char*basename,1022int*dtype_p)1023{1024int i, j;1025struct exclude_list_group *group;1026struct exclude *exclude;1027for(i = EXC_CMDL; i <= EXC_FILE; i++) {1028 group = &dir->exclude_list_group[i];1029for(j = group->nr -1; j >=0; j--) {1030 exclude =last_exclude_matching_from_list(1031 pathname, pathlen, basename, dtype_p,1032&group->el[j]);1033if(exclude)1034return exclude;1035}1036}1037return NULL;1038}10391040/*1041 * Loads the per-directory exclude list for the substring of base1042 * which has a char length of baselen.1043 */1044static voidprep_exclude(struct dir_struct *dir,const char*base,int baselen)1045{1046struct exclude_list_group *group;1047struct exclude_list *el;1048struct exclude_stack *stk = NULL;1049struct untracked_cache_dir *untracked;1050int current;10511052 group = &dir->exclude_list_group[EXC_DIRS];10531054/*1055 * Pop the exclude lists from the EXCL_DIRS exclude_list_group1056 * which originate from directories not in the prefix of the1057 * path being checked.1058 */1059while((stk = dir->exclude_stack) != NULL) {1060if(stk->baselen <= baselen &&1061!strncmp(dir->basebuf.buf, base, stk->baselen))1062break;1063 el = &group->el[dir->exclude_stack->exclude_ix];1064 dir->exclude_stack = stk->prev;1065 dir->exclude = NULL;1066free((char*)el->src);/* see strbuf_detach() below */1067clear_exclude_list(el);1068free(stk);1069 group->nr--;1070}10711072/* Skip traversing into sub directories if the parent is excluded */1073if(dir->exclude)1074return;10751076/*1077 * Lazy initialization. All call sites currently just1078 * memset(dir, 0, sizeof(*dir)) before use. Changing all of1079 * them seems lots of work for little benefit.1080 */1081if(!dir->basebuf.buf)1082strbuf_init(&dir->basebuf, PATH_MAX);10831084/* Read from the parent directories and push them down. */1085 current = stk ? stk->baselen : -1;1086strbuf_setlen(&dir->basebuf, current <0?0: current);1087if(dir->untracked)1088 untracked = stk ? stk->ucd : dir->untracked->root;1089else1090 untracked = NULL;10911092while(current < baselen) {1093const char*cp;1094struct sha1_stat sha1_stat;10951096 stk =xcalloc(1,sizeof(*stk));1097if(current <0) {1098 cp = base;1099 current =0;1100}else{1101 cp =strchr(base + current +1,'/');1102if(!cp)1103die("oops in prep_exclude");1104 cp++;1105 untracked =1106lookup_untracked(dir->untracked, untracked,1107 base + current,1108 cp - base - current);1109}1110 stk->prev = dir->exclude_stack;1111 stk->baselen = cp - base;1112 stk->exclude_ix = group->nr;1113 stk->ucd = untracked;1114 el =add_exclude_list(dir, EXC_DIRS, NULL);1115strbuf_add(&dir->basebuf, base + current, stk->baselen - current);1116assert(stk->baselen == dir->basebuf.len);11171118/* Abort if the directory is excluded */1119if(stk->baselen) {1120int dt = DT_DIR;1121 dir->basebuf.buf[stk->baselen -1] =0;1122 dir->exclude =last_exclude_matching_from_lists(dir,1123 dir->basebuf.buf, stk->baselen -1,1124 dir->basebuf.buf + current, &dt);1125 dir->basebuf.buf[stk->baselen -1] ='/';1126if(dir->exclude &&1127 dir->exclude->flags & EXC_FLAG_NEGATIVE)1128 dir->exclude = NULL;1129if(dir->exclude) {1130 dir->exclude_stack = stk;1131return;1132}1133}11341135/* Try to read per-directory file */1136hashclr(sha1_stat.sha1);1137 sha1_stat.valid =0;1138if(dir->exclude_per_dir &&1139/*1140 * If we know that no files have been added in1141 * this directory (i.e. valid_cached_dir() has1142 * been executed and set untracked->valid) ..1143 */1144(!untracked || !untracked->valid ||1145/*1146 * .. and .gitignore does not exist before1147 * (i.e. null exclude_sha1). Then we can skip1148 * loading .gitignore, which would result in1149 * ENOENT anyway.1150 */1151!is_null_sha1(untracked->exclude_sha1))) {1152/*1153 * dir->basebuf gets reused by the traversal, but we1154 * need fname to remain unchanged to ensure the src1155 * member of each struct exclude correctly1156 * back-references its source file. Other invocations1157 * of add_exclude_list provide stable strings, so we1158 * strbuf_detach() and free() here in the caller.1159 */1160struct strbuf sb = STRBUF_INIT;1161strbuf_addbuf(&sb, &dir->basebuf);1162strbuf_addstr(&sb, dir->exclude_per_dir);1163 el->src =strbuf_detach(&sb, NULL);1164add_excludes(el->src, el->src, stk->baselen, el,1,1165 untracked ? &sha1_stat : NULL);1166}1167/*1168 * NEEDSWORK: when untracked cache is enabled, prep_exclude()1169 * will first be called in valid_cached_dir() then maybe many1170 * times more in last_exclude_matching(). When the cache is1171 * used, last_exclude_matching() will not be called and1172 * reading .gitignore content will be a waste.1173 *1174 * So when it's called by valid_cached_dir() and we can get1175 * .gitignore SHA-1 from the index (i.e. .gitignore is not1176 * modified on work tree), we could delay reading the1177 * .gitignore content until we absolutely need it in1178 * last_exclude_matching(). Be careful about ignore rule1179 * order, though, if you do that.1180 */1181if(untracked &&1182hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {1183invalidate_gitignore(dir->untracked, untracked);1184hashcpy(untracked->exclude_sha1, sha1_stat.sha1);1185}1186 dir->exclude_stack = stk;1187 current = stk->baselen;1188}1189strbuf_setlen(&dir->basebuf, baselen);1190}11911192/*1193 * Loads the exclude lists for the directory containing pathname, then1194 * scans all exclude lists to determine whether pathname is excluded.1195 * Returns the exclude_list element which matched, or NULL for1196 * undecided.1197 */1198struct exclude *last_exclude_matching(struct dir_struct *dir,1199const char*pathname,1200int*dtype_p)1201{1202int pathlen =strlen(pathname);1203const char*basename =strrchr(pathname,'/');1204 basename = (basename) ? basename+1: pathname;12051206prep_exclude(dir, pathname, basename-pathname);12071208if(dir->exclude)1209return dir->exclude;12101211returnlast_exclude_matching_from_lists(dir, pathname, pathlen,1212 basename, dtype_p);1213}12141215/*1216 * Loads the exclude lists for the directory containing pathname, then1217 * scans all exclude lists to determine whether pathname is excluded.1218 * Returns 1 if true, otherwise 0.1219 */1220intis_excluded(struct dir_struct *dir,const char*pathname,int*dtype_p)1221{1222struct exclude *exclude =1223last_exclude_matching(dir, pathname, dtype_p);1224if(exclude)1225return exclude->flags & EXC_FLAG_NEGATIVE ?0:1;1226return0;1227}12281229static struct dir_entry *dir_entry_new(const char*pathname,int len)1230{1231struct dir_entry *ent;12321233FLEX_ALLOC_MEM(ent, name, pathname, len);1234 ent->len = len;1235return ent;1236}12371238static struct dir_entry *dir_add_name(struct dir_struct *dir,const char*pathname,int len)1239{1240if(index_file_exists(&the_index, pathname, len, ignore_case))1241return NULL;12421243ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);1244return dir->entries[dir->nr++] =dir_entry_new(pathname, len);1245}12461247struct dir_entry *dir_add_ignored(struct dir_struct *dir,const char*pathname,int len)1248{1249if(!index_name_is_other(&the_index, pathname, len))1250return NULL;12511252ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);1253return dir->ignored[dir->ignored_nr++] =dir_entry_new(pathname, len);1254}12551256enum exist_status {1257 index_nonexistent =0,1258 index_directory,1259 index_gitdir1260};12611262/*1263 * Do not use the alphabetically sorted index to look up1264 * the directory name; instead, use the case insensitive1265 * directory hash.1266 */1267static enum exist_status directory_exists_in_index_icase(const char*dirname,int len)1268{1269struct cache_entry *ce;12701271if(index_dir_exists(&the_index, dirname, len))1272return index_directory;12731274 ce =index_file_exists(&the_index, dirname, len, ignore_case);1275if(ce &&S_ISGITLINK(ce->ce_mode))1276return index_gitdir;12771278return index_nonexistent;1279}12801281/*1282 * The index sorts alphabetically by entry name, which1283 * means that a gitlink sorts as '\0' at the end, while1284 * a directory (which is defined not as an entry, but as1285 * the files it contains) will sort with the '/' at the1286 * end.1287 */1288static enum exist_status directory_exists_in_index(const char*dirname,int len)1289{1290int pos;12911292if(ignore_case)1293returndirectory_exists_in_index_icase(dirname, len);12941295 pos =index_name_pos(&the_index, dirname, len);1296if(pos <0)1297 pos = -pos-1;1298while(pos < the_index.cache_nr) {1299const struct cache_entry *ce = the_index.cache[pos++];1300unsigned char endchar;13011302if(strncmp(ce->name, dirname, len))1303break;1304 endchar = ce->name[len];1305if(endchar >'/')1306break;1307if(endchar =='/')1308return index_directory;1309if(!endchar &&S_ISGITLINK(ce->ce_mode))1310return index_gitdir;1311}1312return index_nonexistent;1313}13141315/*1316 * When we find a directory when traversing the filesystem, we1317 * have three distinct cases:1318 *1319 * - ignore it1320 * - see it as a directory1321 * - recurse into it1322 *1323 * and which one we choose depends on a combination of existing1324 * git index contents and the flags passed into the directory1325 * traversal routine.1326 *1327 * Case 1: If we *already* have entries in the index under that1328 * directory name, we always recurse into the directory to see1329 * all the files.1330 *1331 * Case 2: If we *already* have that directory name as a gitlink,1332 * we always continue to see it as a gitlink, regardless of whether1333 * there is an actual git directory there or not (it might not1334 * be checked out as a subproject!)1335 *1336 * Case 3: if we didn't have it in the index previously, we1337 * have a few sub-cases:1338 *1339 * (a) if "show_other_directories" is true, we show it as1340 * just a directory, unless "hide_empty_directories" is1341 * also true, in which case we need to check if it contains any1342 * untracked and / or ignored files.1343 * (b) if it looks like a git directory, and we don't have1344 * 'no_gitlinks' set we treat it as a gitlink, and show it1345 * as a directory.1346 * (c) otherwise, we recurse into it.1347 */1348static enum path_treatment treat_directory(struct dir_struct *dir,1349struct untracked_cache_dir *untracked,1350const char*dirname,int len,int baselen,int exclude,1351const struct pathspec *pathspec)1352{1353/* The "len-1" is to strip the final '/' */1354switch(directory_exists_in_index(dirname, len-1)) {1355case index_directory:1356return path_recurse;13571358case index_gitdir:1359return path_none;13601361case index_nonexistent:1362if(dir->flags & DIR_SHOW_OTHER_DIRECTORIES)1363break;1364if(!(dir->flags & DIR_NO_GITLINKS)) {1365unsigned char sha1[20];1366if(resolve_gitlink_ref(dirname,"HEAD", sha1) ==0)1367return path_untracked;1368}1369return path_recurse;1370}13711372/* This is the "show_other_directories" case */13731374if(!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))1375return exclude ? path_excluded : path_untracked;13761377 untracked =lookup_untracked(dir->untracked, untracked,1378 dirname + baselen, len - baselen);1379returnread_directory_recursive(dir, dirname, len,1380 untracked,1, pathspec);1381}13821383/*1384 * This is an inexact early pruning of any recursive directory1385 * reading - if the path cannot possibly be in the pathspec,1386 * return true, and we'll skip it early.1387 */1388static intsimplify_away(const char*path,int pathlen,1389const struct pathspec *pathspec)1390{1391int i;13921393if(!pathspec || !pathspec->nr)1394return0;13951396GUARD_PATHSPEC(pathspec,1397 PATHSPEC_FROMTOP |1398 PATHSPEC_MAXDEPTH |1399 PATHSPEC_LITERAL |1400 PATHSPEC_GLOB |1401 PATHSPEC_ICASE |1402 PATHSPEC_EXCLUDE |1403 PATHSPEC_ATTR);14041405for(i =0; i < pathspec->nr; i++) {1406const struct pathspec_item *item = &pathspec->items[i];1407int len = item->nowildcard_len;14081409if(len > pathlen)1410 len = pathlen;1411if(!ps_strncmp(item, item->match, path, len))1412return0;1413}14141415return1;1416}14171418/*1419 * This function tells us whether an excluded path matches a1420 * list of "interesting" pathspecs. That is, whether a path matched1421 * by any of the pathspecs could possibly be ignored by excluding1422 * the specified path. This can happen if:1423 *1424 * 1. the path is mentioned explicitly in the pathspec1425 *1426 * 2. the path is a directory prefix of some element in the1427 * pathspec1428 */1429static intexclude_matches_pathspec(const char*path,int pathlen,1430const struct pathspec *pathspec)1431{1432int i;14331434if(!pathspec || !pathspec->nr)1435return0;14361437GUARD_PATHSPEC(pathspec,1438 PATHSPEC_FROMTOP |1439 PATHSPEC_MAXDEPTH |1440 PATHSPEC_LITERAL |1441 PATHSPEC_GLOB |1442 PATHSPEC_ICASE |1443 PATHSPEC_EXCLUDE);14441445for(i =0; i < pathspec->nr; i++) {1446const struct pathspec_item *item = &pathspec->items[i];1447int len = item->nowildcard_len;14481449if(len == pathlen &&1450!ps_strncmp(item, item->match, path, pathlen))1451return1;1452if(len > pathlen &&1453 item->match[pathlen] =='/'&&1454!ps_strncmp(item, item->match, path, pathlen))1455return1;1456}1457return0;1458}14591460static intget_index_dtype(const char*path,int len)1461{1462int pos;1463const struct cache_entry *ce;14641465 ce =index_file_exists(&the_index, path, len,0);1466if(ce) {1467if(!ce_uptodate(ce))1468return DT_UNKNOWN;1469if(S_ISGITLINK(ce->ce_mode))1470return DT_DIR;1471/*1472 * Nobody actually cares about the1473 * difference between DT_LNK and DT_REG1474 */1475return DT_REG;1476}14771478/* Try to look it up as a directory */1479 pos =index_name_pos(&the_index, path, len);1480if(pos >=0)1481return DT_UNKNOWN;1482 pos = -pos-1;1483while(pos < the_index.cache_nr) {1484 ce = the_index.cache[pos++];1485if(strncmp(ce->name, path, len))1486break;1487if(ce->name[len] >'/')1488break;1489if(ce->name[len] <'/')1490continue;1491if(!ce_uptodate(ce))1492break;/* continue? */1493return DT_DIR;1494}1495return DT_UNKNOWN;1496}14971498static intget_dtype(struct dirent *de,const char*path,int len)1499{1500int dtype = de ?DTYPE(de) : DT_UNKNOWN;1501struct stat st;15021503if(dtype != DT_UNKNOWN)1504return dtype;1505 dtype =get_index_dtype(path, len);1506if(dtype != DT_UNKNOWN)1507return dtype;1508if(lstat(path, &st))1509return dtype;1510if(S_ISREG(st.st_mode))1511return DT_REG;1512if(S_ISDIR(st.st_mode))1513return DT_DIR;1514if(S_ISLNK(st.st_mode))1515return DT_LNK;1516return dtype;1517}15181519static enum path_treatment treat_one_path(struct dir_struct *dir,1520struct untracked_cache_dir *untracked,1521struct strbuf *path,1522int baselen,1523const struct pathspec *pathspec,1524int dtype,struct dirent *de)1525{1526int exclude;1527int has_path_in_index = !!index_file_exists(&the_index, path->buf, path->len, ignore_case);15281529if(dtype == DT_UNKNOWN)1530 dtype =get_dtype(de, path->buf, path->len);15311532/* Always exclude indexed files */1533if(dtype != DT_DIR && has_path_in_index)1534return path_none;15351536/*1537 * When we are looking at a directory P in the working tree,1538 * there are three cases:1539 *1540 * (1) P exists in the index. Everything inside the directory P in1541 * the working tree needs to go when P is checked out from the1542 * index.1543 *1544 * (2) P does not exist in the index, but there is P/Q in the index.1545 * We know P will stay a directory when we check out the contents1546 * of the index, but we do not know yet if there is a directory1547 * P/Q in the working tree to be killed, so we need to recurse.1548 *1549 * (3) P does not exist in the index, and there is no P/Q in the index1550 * to require P to be a directory, either. Only in this case, we1551 * know that everything inside P will not be killed without1552 * recursing.1553 */1554if((dir->flags & DIR_COLLECT_KILLED_ONLY) &&1555(dtype == DT_DIR) &&1556!has_path_in_index &&1557(directory_exists_in_index(path->buf, path->len) == index_nonexistent))1558return path_none;15591560 exclude =is_excluded(dir, path->buf, &dtype);15611562/*1563 * Excluded? If we don't explicitly want to show1564 * ignored files, ignore it1565 */1566if(exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))1567return path_excluded;15681569switch(dtype) {1570default:1571return path_none;1572case DT_DIR:1573strbuf_addch(path,'/');1574returntreat_directory(dir, untracked, path->buf, path->len,1575 baselen, exclude, pathspec);1576case DT_REG:1577case DT_LNK:1578return exclude ? path_excluded : path_untracked;1579}1580}15811582static enum path_treatment treat_path_fast(struct dir_struct *dir,1583struct untracked_cache_dir *untracked,1584struct cached_dir *cdir,1585struct strbuf *path,1586int baselen,1587const struct pathspec *pathspec)1588{1589strbuf_setlen(path, baselen);1590if(!cdir->ucd) {1591strbuf_addstr(path, cdir->file);1592return path_untracked;1593}1594strbuf_addstr(path, cdir->ucd->name);1595/* treat_one_path() does this before it calls treat_directory() */1596strbuf_complete(path,'/');1597if(cdir->ucd->check_only)1598/*1599 * check_only is set as a result of treat_directory() getting1600 * to its bottom. Verify again the same set of directories1601 * with check_only set.1602 */1603returnread_directory_recursive(dir, path->buf, path->len,1604 cdir->ucd,1, pathspec);1605/*1606 * We get path_recurse in the first run when1607 * directory_exists_in_index() returns index_nonexistent. We1608 * are sure that new changes in the index does not impact the1609 * outcome. Return now.1610 */1611return path_recurse;1612}16131614static enum path_treatment treat_path(struct dir_struct *dir,1615struct untracked_cache_dir *untracked,1616struct cached_dir *cdir,1617struct strbuf *path,1618int baselen,1619const struct pathspec *pathspec)1620{1621int dtype;1622struct dirent *de = cdir->de;16231624if(!de)1625returntreat_path_fast(dir, untracked, cdir, path,1626 baselen, pathspec);1627if(is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name,".git"))1628return path_none;1629strbuf_setlen(path, baselen);1630strbuf_addstr(path, de->d_name);1631if(simplify_away(path->buf, path->len, pathspec))1632return path_none;16331634 dtype =DTYPE(de);1635returntreat_one_path(dir, untracked, path, baselen, pathspec, dtype, de);1636}16371638static voidadd_untracked(struct untracked_cache_dir *dir,const char*name)1639{1640if(!dir)1641return;1642ALLOC_GROW(dir->untracked, dir->untracked_nr +1,1643 dir->untracked_alloc);1644 dir->untracked[dir->untracked_nr++] =xstrdup(name);1645}16461647static intvalid_cached_dir(struct dir_struct *dir,1648struct untracked_cache_dir *untracked,1649struct strbuf *path,1650int check_only)1651{1652struct stat st;16531654if(!untracked)1655return0;16561657if(stat(path->len ? path->buf :".", &st)) {1658invalidate_directory(dir->untracked, untracked);1659memset(&untracked->stat_data,0,sizeof(untracked->stat_data));1660return0;1661}1662if(!untracked->valid ||1663match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {1664if(untracked->valid)1665invalidate_directory(dir->untracked, untracked);1666fill_stat_data(&untracked->stat_data, &st);1667return0;1668}16691670if(untracked->check_only != !!check_only) {1671invalidate_directory(dir->untracked, untracked);1672return0;1673}16741675/*1676 * prep_exclude will be called eventually on this directory,1677 * but it's called much later in last_exclude_matching(). We1678 * need it now to determine the validity of the cache for this1679 * path. The next calls will be nearly no-op, the way1680 * prep_exclude() is designed.1681 */1682if(path->len && path->buf[path->len -1] !='/') {1683strbuf_addch(path,'/');1684prep_exclude(dir, path->buf, path->len);1685strbuf_setlen(path, path->len -1);1686}else1687prep_exclude(dir, path->buf, path->len);16881689/* hopefully prep_exclude() haven't invalidated this entry... */1690return untracked->valid;1691}16921693static intopen_cached_dir(struct cached_dir *cdir,1694struct dir_struct *dir,1695struct untracked_cache_dir *untracked,1696struct strbuf *path,1697int check_only)1698{1699memset(cdir,0,sizeof(*cdir));1700 cdir->untracked = untracked;1701if(valid_cached_dir(dir, untracked, path, check_only))1702return0;1703 cdir->fdir =opendir(path->len ? path->buf :".");1704if(dir->untracked)1705 dir->untracked->dir_opened++;1706if(!cdir->fdir)1707return-1;1708return0;1709}17101711static intread_cached_dir(struct cached_dir *cdir)1712{1713if(cdir->fdir) {1714 cdir->de =readdir(cdir->fdir);1715if(!cdir->de)1716return-1;1717return0;1718}1719while(cdir->nr_dirs < cdir->untracked->dirs_nr) {1720struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];1721if(!d->recurse) {1722 cdir->nr_dirs++;1723continue;1724}1725 cdir->ucd = d;1726 cdir->nr_dirs++;1727return0;1728}1729 cdir->ucd = NULL;1730if(cdir->nr_files < cdir->untracked->untracked_nr) {1731struct untracked_cache_dir *d = cdir->untracked;1732 cdir->file = d->untracked[cdir->nr_files++];1733return0;1734}1735return-1;1736}17371738static voidclose_cached_dir(struct cached_dir *cdir)1739{1740if(cdir->fdir)1741closedir(cdir->fdir);1742/*1743 * We have gone through this directory and found no untracked1744 * entries. Mark it valid.1745 */1746if(cdir->untracked) {1747 cdir->untracked->valid =1;1748 cdir->untracked->recurse =1;1749}1750}17511752/*1753 * Read a directory tree. We currently ignore anything but1754 * directories, regular files and symlinks. That's because git1755 * doesn't handle them at all yet. Maybe that will change some1756 * day.1757 *1758 * Also, we ignore the name ".git" (even if it is not a directory).1759 * That likely will not change.1760 *1761 * Returns the most significant path_treatment value encountered in the scan.1762 */1763static enum path_treatment read_directory_recursive(struct dir_struct *dir,1764const char*base,int baselen,1765struct untracked_cache_dir *untracked,int check_only,1766const struct pathspec *pathspec)1767{1768struct cached_dir cdir;1769enum path_treatment state, subdir_state, dir_state = path_none;1770struct strbuf path = STRBUF_INIT;17711772strbuf_add(&path, base, baselen);17731774if(open_cached_dir(&cdir, dir, untracked, &path, check_only))1775goto out;17761777if(untracked)1778 untracked->check_only = !!check_only;17791780while(!read_cached_dir(&cdir)) {1781/* check how the file or directory should be treated */1782 state =treat_path(dir, untracked, &cdir, &path,1783 baselen, pathspec);17841785if(state > dir_state)1786 dir_state = state;17871788/* recurse into subdir if instructed by treat_path */1789if(state == path_recurse) {1790struct untracked_cache_dir *ud;1791 ud =lookup_untracked(dir->untracked, untracked,1792 path.buf + baselen,1793 path.len - baselen);1794 subdir_state =1795read_directory_recursive(dir, path.buf,1796 path.len, ud,1797 check_only, pathspec);1798if(subdir_state > dir_state)1799 dir_state = subdir_state;1800}18011802if(check_only) {1803/* abort early if maximum state has been reached */1804if(dir_state == path_untracked) {1805if(cdir.fdir)1806add_untracked(untracked, path.buf + baselen);1807break;1808}1809/* skip the dir_add_* part */1810continue;1811}18121813/* add the path to the appropriate result list */1814switch(state) {1815case path_excluded:1816if(dir->flags & DIR_SHOW_IGNORED)1817dir_add_name(dir, path.buf, path.len);1818else if((dir->flags & DIR_SHOW_IGNORED_TOO) ||1819((dir->flags & DIR_COLLECT_IGNORED) &&1820exclude_matches_pathspec(path.buf, path.len,1821 pathspec)))1822dir_add_ignored(dir, path.buf, path.len);1823break;18241825case path_untracked:1826if(dir->flags & DIR_SHOW_IGNORED)1827break;1828dir_add_name(dir, path.buf, path.len);1829if(cdir.fdir)1830add_untracked(untracked, path.buf + baselen);1831break;18321833default:1834break;1835}1836}1837close_cached_dir(&cdir);1838 out:1839strbuf_release(&path);18401841return dir_state;1842}18431844static intcmp_name(const void*p1,const void*p2)1845{1846const struct dir_entry *e1 = *(const struct dir_entry **)p1;1847const struct dir_entry *e2 = *(const struct dir_entry **)p2;18481849returnname_compare(e1->name, e1->len, e2->name, e2->len);1850}18511852static inttreat_leading_path(struct dir_struct *dir,1853const char*path,int len,1854const struct pathspec *pathspec)1855{1856struct strbuf sb = STRBUF_INIT;1857int baselen, rc =0;1858const char*cp;1859int old_flags = dir->flags;18601861while(len && path[len -1] =='/')1862 len--;1863if(!len)1864return1;1865 baselen =0;1866 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;1867while(1) {1868 cp = path + baselen + !!baselen;1869 cp =memchr(cp,'/', path + len - cp);1870if(!cp)1871 baselen = len;1872else1873 baselen = cp - path;1874strbuf_setlen(&sb,0);1875strbuf_add(&sb, path, baselen);1876if(!is_directory(sb.buf))1877break;1878if(simplify_away(sb.buf, sb.len, pathspec))1879break;1880if(treat_one_path(dir, NULL, &sb, baselen, pathspec,1881 DT_DIR, NULL) == path_none)1882break;/* do not recurse into it */1883if(len <= baselen) {1884 rc =1;1885break;/* finished checking */1886}1887}1888strbuf_release(&sb);1889 dir->flags = old_flags;1890return rc;1891}18921893static const char*get_ident_string(void)1894{1895static struct strbuf sb = STRBUF_INIT;1896struct utsname uts;18971898if(sb.len)1899return sb.buf;1900if(uname(&uts) <0)1901die_errno(_("failed to get kernel name and information"));1902strbuf_addf(&sb,"Location%s, system%s",get_git_work_tree(),1903 uts.sysname);1904return sb.buf;1905}19061907static intident_in_untracked(const struct untracked_cache *uc)1908{1909/*1910 * Previous git versions may have saved many NUL separated1911 * strings in the "ident" field, but it is insane to manage1912 * many locations, so just take care of the first one.1913 */19141915return!strcmp(uc->ident.buf,get_ident_string());1916}19171918static voidset_untracked_ident(struct untracked_cache *uc)1919{1920strbuf_reset(&uc->ident);1921strbuf_addstr(&uc->ident,get_ident_string());19221923/*1924 * This strbuf used to contain a list of NUL separated1925 * strings, so save NUL too for backward compatibility.1926 */1927strbuf_addch(&uc->ident,0);1928}19291930static voidnew_untracked_cache(struct index_state *istate)1931{1932struct untracked_cache *uc =xcalloc(1,sizeof(*uc));1933strbuf_init(&uc->ident,100);1934 uc->exclude_per_dir =".gitignore";1935/* should be the same flags used by git-status */1936 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;1937set_untracked_ident(uc);1938 istate->untracked = uc;1939 istate->cache_changed |= UNTRACKED_CHANGED;1940}19411942voidadd_untracked_cache(struct index_state *istate)1943{1944if(!istate->untracked) {1945new_untracked_cache(istate);1946}else{1947if(!ident_in_untracked(istate->untracked)) {1948free_untracked_cache(istate->untracked);1949new_untracked_cache(istate);1950}1951}1952}19531954voidremove_untracked_cache(struct index_state *istate)1955{1956if(istate->untracked) {1957free_untracked_cache(istate->untracked);1958 istate->untracked = NULL;1959 istate->cache_changed |= UNTRACKED_CHANGED;1960}1961}19621963static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,1964int base_len,1965const struct pathspec *pathspec)1966{1967struct untracked_cache_dir *root;19681969if(!dir->untracked ||getenv("GIT_DISABLE_UNTRACKED_CACHE"))1970return NULL;19711972/*1973 * We only support $GIT_DIR/info/exclude and core.excludesfile1974 * as the global ignore rule files. Any other additions1975 * (e.g. from command line) invalidate the cache. This1976 * condition also catches running setup_standard_excludes()1977 * before setting dir->untracked!1978 */1979if(dir->unmanaged_exclude_files)1980return NULL;19811982/*1983 * Optimize for the main use case only: whole-tree git1984 * status. More work involved in treat_leading_path() if we1985 * use cache on just a subset of the worktree. pathspec1986 * support could make the matter even worse.1987 */1988if(base_len || (pathspec && pathspec->nr))1989return NULL;19901991/* Different set of flags may produce different results */1992if(dir->flags != dir->untracked->dir_flags ||1993/*1994 * See treat_directory(), case index_nonexistent. Without1995 * this flag, we may need to also cache .git file content1996 * for the resolve_gitlink_ref() call, which we don't.1997 */1998!(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||1999/* We don't support collecting ignore files */2000(dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |2001 DIR_COLLECT_IGNORED)))2002return NULL;20032004/*2005 * If we use .gitignore in the cache and now you change it to2006 * .gitexclude, everything will go wrong.2007 */2008if(dir->exclude_per_dir != dir->untracked->exclude_per_dir &&2009strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))2010return NULL;20112012/*2013 * EXC_CMDL is not considered in the cache. If people set it,2014 * skip the cache.2015 */2016if(dir->exclude_list_group[EXC_CMDL].nr)2017return NULL;20182019if(!ident_in_untracked(dir->untracked)) {2020warning(_("Untracked cache is disabled on this system or location."));2021return NULL;2022}20232024if(!dir->untracked->root) {2025const int len =sizeof(*dir->untracked->root);2026 dir->untracked->root =xmalloc(len);2027memset(dir->untracked->root,0, len);2028}20292030/* Validate $GIT_DIR/info/exclude and core.excludesfile */2031 root = dir->untracked->root;2032if(hashcmp(dir->ss_info_exclude.sha1,2033 dir->untracked->ss_info_exclude.sha1)) {2034invalidate_gitignore(dir->untracked, root);2035 dir->untracked->ss_info_exclude = dir->ss_info_exclude;2036}2037if(hashcmp(dir->ss_excludes_file.sha1,2038 dir->untracked->ss_excludes_file.sha1)) {2039invalidate_gitignore(dir->untracked, root);2040 dir->untracked->ss_excludes_file = dir->ss_excludes_file;2041}20422043/* Make sure this directory is not dropped out at saving phase */2044 root->recurse =1;2045return root;2046}20472048intread_directory(struct dir_struct *dir,const char*path,2049int len,const struct pathspec *pathspec)2050{2051struct untracked_cache_dir *untracked;20522053if(has_symlink_leading_path(path, len))2054return dir->nr;20552056 untracked =validate_untracked_cache(dir, len, pathspec);2057if(!untracked)2058/*2059 * make sure untracked cache code path is disabled,2060 * e.g. prep_exclude()2061 */2062 dir->untracked = NULL;2063if(!len ||treat_leading_path(dir, path, len, pathspec))2064read_directory_recursive(dir, path, len, untracked,0, pathspec);2065QSORT(dir->entries, dir->nr, cmp_name);2066QSORT(dir->ignored, dir->ignored_nr, cmp_name);2067if(dir->untracked) {2068static struct trace_key trace_untracked_stats =TRACE_KEY_INIT(UNTRACKED_STATS);2069trace_printf_key(&trace_untracked_stats,2070"node creation:%u\n"2071"gitignore invalidation:%u\n"2072"directory invalidation:%u\n"2073"opendir:%u\n",2074 dir->untracked->dir_created,2075 dir->untracked->gitignore_invalidated,2076 dir->untracked->dir_invalidated,2077 dir->untracked->dir_opened);2078if(dir->untracked == the_index.untracked &&2079(dir->untracked->dir_opened ||2080 dir->untracked->gitignore_invalidated ||2081 dir->untracked->dir_invalidated))2082 the_index.cache_changed |= UNTRACKED_CHANGED;2083if(dir->untracked != the_index.untracked) {2084free(dir->untracked);2085 dir->untracked = NULL;2086}2087}2088return dir->nr;2089}20902091intfile_exists(const char*f)2092{2093struct stat sb;2094returnlstat(f, &sb) ==0;2095}20962097static intcmp_icase(char a,char b)2098{2099if(a == b)2100return0;2101if(ignore_case)2102returntoupper(a) -toupper(b);2103return a - b;2104}21052106/*2107 * Given two normalized paths (a trailing slash is ok), if subdir is2108 * outside dir, return -1. Otherwise return the offset in subdir that2109 * can be used as relative path to dir.2110 */2111intdir_inside_of(const char*subdir,const char*dir)2112{2113int offset =0;21142115assert(dir && subdir && *dir && *subdir);21162117while(*dir && *subdir && !cmp_icase(*dir, *subdir)) {2118 dir++;2119 subdir++;2120 offset++;2121}21222123/* hel[p]/me vs hel[l]/yeah */2124if(*dir && *subdir)2125return-1;21262127if(!*subdir)2128return!*dir ? offset : -1;/* same dir */21292130/* foo/[b]ar vs foo/[] */2131if(is_dir_sep(dir[-1]))2132returnis_dir_sep(subdir[-1]) ? offset : -1;21332134/* foo[/]bar vs foo[] */2135returnis_dir_sep(*subdir) ? offset +1: -1;2136}21372138intis_inside_dir(const char*dir)2139{2140char*cwd;2141int rc;21422143if(!dir)2144return0;21452146 cwd =xgetcwd();2147 rc = (dir_inside_of(cwd, dir) >=0);2148free(cwd);2149return rc;2150}21512152intis_empty_dir(const char*path)2153{2154DIR*dir =opendir(path);2155struct dirent *e;2156int ret =1;21572158if(!dir)2159return0;21602161while((e =readdir(dir)) != NULL)2162if(!is_dot_or_dotdot(e->d_name)) {2163 ret =0;2164break;2165}21662167closedir(dir);2168return ret;2169}21702171static intremove_dir_recurse(struct strbuf *path,int flag,int*kept_up)2172{2173DIR*dir;2174struct dirent *e;2175int ret =0, original_len = path->len, len, kept_down =0;2176int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);2177int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);2178unsigned char submodule_head[20];21792180if((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&2181!resolve_gitlink_ref(path->buf,"HEAD", submodule_head)) {2182/* Do not descend and nuke a nested git work tree. */2183if(kept_up)2184*kept_up =1;2185return0;2186}21872188 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;2189 dir =opendir(path->buf);2190if(!dir) {2191if(errno == ENOENT)2192return keep_toplevel ? -1:0;2193else if(errno == EACCES && !keep_toplevel)2194/*2195 * An empty dir could be removable even if it2196 * is unreadable:2197 */2198returnrmdir(path->buf);2199else2200return-1;2201}2202strbuf_complete(path,'/');22032204 len = path->len;2205while((e =readdir(dir)) != NULL) {2206struct stat st;2207if(is_dot_or_dotdot(e->d_name))2208continue;22092210strbuf_setlen(path, len);2211strbuf_addstr(path, e->d_name);2212if(lstat(path->buf, &st)) {2213if(errno == ENOENT)2214/*2215 * file disappeared, which is what we2216 * wanted anyway2217 */2218continue;2219/* fall thru */2220}else if(S_ISDIR(st.st_mode)) {2221if(!remove_dir_recurse(path, flag, &kept_down))2222continue;/* happy */2223}else if(!only_empty &&2224(!unlink(path->buf) || errno == ENOENT)) {2225continue;/* happy, too */2226}22272228/* path too long, stat fails, or non-directory still exists */2229 ret = -1;2230break;2231}2232closedir(dir);22332234strbuf_setlen(path, original_len);2235if(!ret && !keep_toplevel && !kept_down)2236 ret = (!rmdir(path->buf) || errno == ENOENT) ?0: -1;2237else if(kept_up)2238/*2239 * report the uplevel that it is not an error that we2240 * did not rmdir() our directory.2241 */2242*kept_up = !ret;2243return ret;2244}22452246intremove_dir_recursively(struct strbuf *path,int flag)2247{2248returnremove_dir_recurse(path, flag, NULL);2249}22502251staticGIT_PATH_FUNC(git_path_info_exclude,"info/exclude")22522253voidsetup_standard_excludes(struct dir_struct *dir)2254{2255 dir->exclude_per_dir =".gitignore";22562257/* core.excludefile defaulting to $XDG_HOME/git/ignore */2258if(!excludes_file)2259 excludes_file =xdg_config_home("ignore");2260if(excludes_file && !access_or_warn(excludes_file, R_OK,0))2261add_excludes_from_file_1(dir, excludes_file,2262 dir->untracked ? &dir->ss_excludes_file : NULL);22632264/* per repository user preference */2265if(startup_info->have_repository) {2266const char*path =git_path_info_exclude();2267if(!access_or_warn(path, R_OK,0))2268add_excludes_from_file_1(dir, path,2269 dir->untracked ? &dir->ss_info_exclude : NULL);2270}2271}22722273intremove_path(const char*name)2274{2275char*slash;22762277if(unlink(name) && errno != ENOENT && errno != ENOTDIR)2278return-1;22792280 slash =strrchr(name,'/');2281if(slash) {2282char*dirs =xstrdup(name);2283 slash = dirs + (slash - name);2284do{2285*slash ='\0';2286}while(rmdir(dirs) ==0&& (slash =strrchr(dirs,'/')));2287free(dirs);2288}2289return0;2290}22912292/*2293 * Frees memory within dir which was allocated for exclude lists and2294 * the exclude_stack. Does not free dir itself.2295 */2296voidclear_directory(struct dir_struct *dir)2297{2298int i, j;2299struct exclude_list_group *group;2300struct exclude_list *el;2301struct exclude_stack *stk;23022303for(i = EXC_CMDL; i <= EXC_FILE; i++) {2304 group = &dir->exclude_list_group[i];2305for(j =0; j < group->nr; j++) {2306 el = &group->el[j];2307if(i == EXC_DIRS)2308free((char*)el->src);2309clear_exclude_list(el);2310}2311free(group->el);2312}23132314 stk = dir->exclude_stack;2315while(stk) {2316struct exclude_stack *prev = stk->prev;2317free(stk);2318 stk = prev;2319}2320strbuf_release(&dir->basebuf);2321}23222323struct ondisk_untracked_cache {2324struct stat_data info_exclude_stat;2325struct stat_data excludes_file_stat;2326uint32_t dir_flags;2327unsigned char info_exclude_sha1[20];2328unsigned char excludes_file_sha1[20];2329char exclude_per_dir[FLEX_ARRAY];2330};23312332#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)23332334struct write_data {2335int index;/* number of written untracked_cache_dir */2336struct ewah_bitmap *check_only;/* from untracked_cache_dir */2337struct ewah_bitmap *valid;/* from untracked_cache_dir */2338struct ewah_bitmap *sha1_valid;/* set if exclude_sha1 is not null */2339struct strbuf out;2340struct strbuf sb_stat;2341struct strbuf sb_sha1;2342};23432344static voidstat_data_to_disk(struct stat_data *to,const struct stat_data *from)2345{2346 to->sd_ctime.sec =htonl(from->sd_ctime.sec);2347 to->sd_ctime.nsec =htonl(from->sd_ctime.nsec);2348 to->sd_mtime.sec =htonl(from->sd_mtime.sec);2349 to->sd_mtime.nsec =htonl(from->sd_mtime.nsec);2350 to->sd_dev =htonl(from->sd_dev);2351 to->sd_ino =htonl(from->sd_ino);2352 to->sd_uid =htonl(from->sd_uid);2353 to->sd_gid =htonl(from->sd_gid);2354 to->sd_size =htonl(from->sd_size);2355}23562357static voidwrite_one_dir(struct untracked_cache_dir *untracked,2358struct write_data *wd)2359{2360struct stat_data stat_data;2361struct strbuf *out = &wd->out;2362unsigned char intbuf[16];2363unsigned int intlen, value;2364int i = wd->index++;23652366/*2367 * untracked_nr should be reset whenever valid is clear, but2368 * for safety..2369 */2370if(!untracked->valid) {2371 untracked->untracked_nr =0;2372 untracked->check_only =0;2373}23742375if(untracked->check_only)2376ewah_set(wd->check_only, i);2377if(untracked->valid) {2378ewah_set(wd->valid, i);2379stat_data_to_disk(&stat_data, &untracked->stat_data);2380strbuf_add(&wd->sb_stat, &stat_data,sizeof(stat_data));2381}2382if(!is_null_sha1(untracked->exclude_sha1)) {2383ewah_set(wd->sha1_valid, i);2384strbuf_add(&wd->sb_sha1, untracked->exclude_sha1,20);2385}23862387 intlen =encode_varint(untracked->untracked_nr, intbuf);2388strbuf_add(out, intbuf, intlen);23892390/* skip non-recurse directories */2391for(i =0, value =0; i < untracked->dirs_nr; i++)2392if(untracked->dirs[i]->recurse)2393 value++;2394 intlen =encode_varint(value, intbuf);2395strbuf_add(out, intbuf, intlen);23962397strbuf_add(out, untracked->name,strlen(untracked->name) +1);23982399for(i =0; i < untracked->untracked_nr; i++)2400strbuf_add(out, untracked->untracked[i],2401strlen(untracked->untracked[i]) +1);24022403for(i =0; i < untracked->dirs_nr; i++)2404if(untracked->dirs[i]->recurse)2405write_one_dir(untracked->dirs[i], wd);2406}24072408voidwrite_untracked_extension(struct strbuf *out,struct untracked_cache *untracked)2409{2410struct ondisk_untracked_cache *ouc;2411struct write_data wd;2412unsigned char varbuf[16];2413int varint_len;2414size_t len =strlen(untracked->exclude_per_dir);24152416FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);2417stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);2418stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);2419hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);2420hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);2421 ouc->dir_flags =htonl(untracked->dir_flags);24222423 varint_len =encode_varint(untracked->ident.len, varbuf);2424strbuf_add(out, varbuf, varint_len);2425strbuf_addbuf(out, &untracked->ident);24262427strbuf_add(out, ouc,ouc_size(len));2428free(ouc);2429 ouc = NULL;24302431if(!untracked->root) {2432 varint_len =encode_varint(0, varbuf);2433strbuf_add(out, varbuf, varint_len);2434return;2435}24362437 wd.index =0;2438 wd.check_only =ewah_new();2439 wd.valid =ewah_new();2440 wd.sha1_valid =ewah_new();2441strbuf_init(&wd.out,1024);2442strbuf_init(&wd.sb_stat,1024);2443strbuf_init(&wd.sb_sha1,1024);2444write_one_dir(untracked->root, &wd);24452446 varint_len =encode_varint(wd.index, varbuf);2447strbuf_add(out, varbuf, varint_len);2448strbuf_addbuf(out, &wd.out);2449ewah_serialize_strbuf(wd.valid, out);2450ewah_serialize_strbuf(wd.check_only, out);2451ewah_serialize_strbuf(wd.sha1_valid, out);2452strbuf_addbuf(out, &wd.sb_stat);2453strbuf_addbuf(out, &wd.sb_sha1);2454strbuf_addch(out,'\0');/* safe guard for string lists */24552456ewah_free(wd.valid);2457ewah_free(wd.check_only);2458ewah_free(wd.sha1_valid);2459strbuf_release(&wd.out);2460strbuf_release(&wd.sb_stat);2461strbuf_release(&wd.sb_sha1);2462}24632464static voidfree_untracked(struct untracked_cache_dir *ucd)2465{2466int i;2467if(!ucd)2468return;2469for(i =0; i < ucd->dirs_nr; i++)2470free_untracked(ucd->dirs[i]);2471for(i =0; i < ucd->untracked_nr; i++)2472free(ucd->untracked[i]);2473free(ucd->untracked);2474free(ucd->dirs);2475free(ucd);2476}24772478voidfree_untracked_cache(struct untracked_cache *uc)2479{2480if(uc)2481free_untracked(uc->root);2482free(uc);2483}24842485struct read_data {2486int index;2487struct untracked_cache_dir **ucd;2488struct ewah_bitmap *check_only;2489struct ewah_bitmap *valid;2490struct ewah_bitmap *sha1_valid;2491const unsigned char*data;2492const unsigned char*end;2493};24942495static voidstat_data_from_disk(struct stat_data *to,const struct stat_data *from)2496{2497 to->sd_ctime.sec =get_be32(&from->sd_ctime.sec);2498 to->sd_ctime.nsec =get_be32(&from->sd_ctime.nsec);2499 to->sd_mtime.sec =get_be32(&from->sd_mtime.sec);2500 to->sd_mtime.nsec =get_be32(&from->sd_mtime.nsec);2501 to->sd_dev =get_be32(&from->sd_dev);2502 to->sd_ino =get_be32(&from->sd_ino);2503 to->sd_uid =get_be32(&from->sd_uid);2504 to->sd_gid =get_be32(&from->sd_gid);2505 to->sd_size =get_be32(&from->sd_size);2506}25072508static intread_one_dir(struct untracked_cache_dir **untracked_,2509struct read_data *rd)2510{2511struct untracked_cache_dir ud, *untracked;2512const unsigned char*next, *data = rd->data, *end = rd->end;2513unsigned int value;2514int i, len;25152516memset(&ud,0,sizeof(ud));25172518 next = data;2519 value =decode_varint(&next);2520if(next > end)2521return-1;2522 ud.recurse =1;2523 ud.untracked_alloc = value;2524 ud.untracked_nr = value;2525if(ud.untracked_nr)2526ALLOC_ARRAY(ud.untracked, ud.untracked_nr);2527 data = next;25282529 next = data;2530 ud.dirs_alloc = ud.dirs_nr =decode_varint(&next);2531if(next > end)2532return-1;2533ALLOC_ARRAY(ud.dirs, ud.dirs_nr);2534 data = next;25352536 len =strlen((const char*)data);2537 next = data + len +1;2538if(next > rd->end)2539return-1;2540*untracked_ = untracked =xmalloc(st_add(sizeof(*untracked), len));2541memcpy(untracked, &ud,sizeof(ud));2542memcpy(untracked->name, data, len +1);2543 data = next;25442545for(i =0; i < untracked->untracked_nr; i++) {2546 len =strlen((const char*)data);2547 next = data + len +1;2548if(next > rd->end)2549return-1;2550 untracked->untracked[i] =xstrdup((const char*)data);2551 data = next;2552}25532554 rd->ucd[rd->index++] = untracked;2555 rd->data = data;25562557for(i =0; i < untracked->dirs_nr; i++) {2558 len =read_one_dir(untracked->dirs + i, rd);2559if(len <0)2560return-1;2561}2562return0;2563}25642565static voidset_check_only(size_t pos,void*cb)2566{2567struct read_data *rd = cb;2568struct untracked_cache_dir *ud = rd->ucd[pos];2569 ud->check_only =1;2570}25712572static voidread_stat(size_t pos,void*cb)2573{2574struct read_data *rd = cb;2575struct untracked_cache_dir *ud = rd->ucd[pos];2576if(rd->data +sizeof(struct stat_data) > rd->end) {2577 rd->data = rd->end +1;2578return;2579}2580stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);2581 rd->data +=sizeof(struct stat_data);2582 ud->valid =1;2583}25842585static voidread_sha1(size_t pos,void*cb)2586{2587struct read_data *rd = cb;2588struct untracked_cache_dir *ud = rd->ucd[pos];2589if(rd->data +20> rd->end) {2590 rd->data = rd->end +1;2591return;2592}2593hashcpy(ud->exclude_sha1, rd->data);2594 rd->data +=20;2595}25962597static voidload_sha1_stat(struct sha1_stat *sha1_stat,2598const struct stat_data *stat,2599const unsigned char*sha1)2600{2601stat_data_from_disk(&sha1_stat->stat, stat);2602hashcpy(sha1_stat->sha1, sha1);2603 sha1_stat->valid =1;2604}26052606struct untracked_cache *read_untracked_extension(const void*data,unsigned long sz)2607{2608const struct ondisk_untracked_cache *ouc;2609struct untracked_cache *uc;2610struct read_data rd;2611const unsigned char*next = data, *end = (const unsigned char*)data + sz;2612const char*ident;2613int ident_len, len;26142615if(sz <=1|| end[-1] !='\0')2616return NULL;2617 end--;26182619 ident_len =decode_varint(&next);2620if(next + ident_len > end)2621return NULL;2622 ident = (const char*)next;2623 next += ident_len;26242625 ouc = (const struct ondisk_untracked_cache *)next;2626if(next +ouc_size(0) > end)2627return NULL;26282629 uc =xcalloc(1,sizeof(*uc));2630strbuf_init(&uc->ident, ident_len);2631strbuf_add(&uc->ident, ident, ident_len);2632load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,2633 ouc->info_exclude_sha1);2634load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,2635 ouc->excludes_file_sha1);2636 uc->dir_flags =get_be32(&ouc->dir_flags);2637 uc->exclude_per_dir =xstrdup(ouc->exclude_per_dir);2638/* NUL after exclude_per_dir is covered by sizeof(*ouc) */2639 next +=ouc_size(strlen(ouc->exclude_per_dir));2640if(next >= end)2641goto done2;26422643 len =decode_varint(&next);2644if(next > end || len ==0)2645goto done2;26462647 rd.valid =ewah_new();2648 rd.check_only =ewah_new();2649 rd.sha1_valid =ewah_new();2650 rd.data = next;2651 rd.end = end;2652 rd.index =0;2653ALLOC_ARRAY(rd.ucd, len);26542655if(read_one_dir(&uc->root, &rd) || rd.index != len)2656goto done;26572658 next = rd.data;2659 len =ewah_read_mmap(rd.valid, next, end - next);2660if(len <0)2661goto done;26622663 next += len;2664 len =ewah_read_mmap(rd.check_only, next, end - next);2665if(len <0)2666goto done;26672668 next += len;2669 len =ewah_read_mmap(rd.sha1_valid, next, end - next);2670if(len <0)2671goto done;26722673ewah_each_bit(rd.check_only, set_check_only, &rd);2674 rd.data = next + len;2675ewah_each_bit(rd.valid, read_stat, &rd);2676ewah_each_bit(rd.sha1_valid, read_sha1, &rd);2677 next = rd.data;26782679done:2680free(rd.ucd);2681ewah_free(rd.valid);2682ewah_free(rd.check_only);2683ewah_free(rd.sha1_valid);2684done2:2685if(next != end) {2686free_untracked_cache(uc);2687 uc = NULL;2688}2689return uc;2690}26912692static voidinvalidate_one_directory(struct untracked_cache *uc,2693struct untracked_cache_dir *ucd)2694{2695 uc->dir_invalidated++;2696 ucd->valid =0;2697 ucd->untracked_nr =0;2698}26992700/*2701 * Normally when an entry is added or removed from a directory,2702 * invalidating that directory is enough. No need to touch its2703 * ancestors. When a directory is shown as "foo/bar/" in git-status2704 * however, deleting or adding an entry may have cascading effect.2705 *2706 * Say the "foo/bar/file" has become untracked, we need to tell the2707 * untracked_cache_dir of "foo" that "bar/" is not an untracked2708 * directory any more (because "bar" is managed by foo as an untracked2709 * "file").2710 *2711 * Similarly, if "foo/bar/file" moves from untracked to tracked and it2712 * was the last untracked entry in the entire "foo", we should show2713 * "foo/" instead. Which means we have to invalidate past "bar" up to2714 * "foo".2715 *2716 * This function traverses all directories from root to leaf. If there2717 * is a chance of one of the above cases happening, we invalidate back2718 * to root. Otherwise we just invalidate the leaf. There may be a more2719 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to2720 * detect these cases and avoid unnecessary invalidation, for example,2721 * checking for the untracked entry named "bar/" in "foo", but for now2722 * stick to something safe and simple.2723 */2724static intinvalidate_one_component(struct untracked_cache *uc,2725struct untracked_cache_dir *dir,2726const char*path,int len)2727{2728const char*rest =strchr(path,'/');27292730if(rest) {2731int component_len = rest - path;2732struct untracked_cache_dir *d =2733lookup_untracked(uc, dir, path, component_len);2734int ret =2735invalidate_one_component(uc, d, rest +1,2736 len - (component_len +1));2737if(ret)2738invalidate_one_directory(uc, dir);2739return ret;2740}27412742invalidate_one_directory(uc, dir);2743return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;2744}27452746voiduntracked_cache_invalidate_path(struct index_state *istate,2747const char*path)2748{2749if(!istate->untracked || !istate->untracked->root)2750return;2751invalidate_one_component(istate->untracked, istate->untracked->root,2752 path,strlen(path));2753}27542755voiduntracked_cache_remove_from_index(struct index_state *istate,2756const char*path)2757{2758untracked_cache_invalidate_path(istate, path);2759}27602761voiduntracked_cache_add_to_index(struct index_state *istate,2762const char*path)2763{2764untracked_cache_invalidate_path(istate, path);2765}27662767/* Update gitfile and core.worktree setting to connect work tree and git dir */2768voidconnect_work_tree_and_git_dir(const char*work_tree_,const char*git_dir_)2769{2770struct strbuf gitfile_sb = STRBUF_INIT;2771struct strbuf cfg_sb = STRBUF_INIT;2772struct strbuf rel_path = STRBUF_INIT;2773char*git_dir, *work_tree;27742775/* Prepare .git file */2776strbuf_addf(&gitfile_sb,"%s/.git", work_tree_);2777if(safe_create_leading_directories_const(gitfile_sb.buf))2778die(_("could not create directories for%s"), gitfile_sb.buf);27792780/* Prepare config file */2781strbuf_addf(&cfg_sb,"%s/config", git_dir_);2782if(safe_create_leading_directories_const(cfg_sb.buf))2783die(_("could not create directories for%s"), cfg_sb.buf);27842785 git_dir =real_pathdup(git_dir_,1);2786 work_tree =real_pathdup(work_tree_,1);27872788/* Write .git file */2789write_file(gitfile_sb.buf,"gitdir:%s",2790relative_path(git_dir, work_tree, &rel_path));2791/* Update core.worktree setting */2792git_config_set_in_file(cfg_sb.buf,"core.worktree",2793relative_path(work_tree, git_dir, &rel_path));27942795strbuf_release(&gitfile_sb);2796strbuf_release(&cfg_sb);2797strbuf_release(&rel_path);2798free(work_tree);2799free(git_dir);2800}28012802/*2803 * Migrate the git directory of the given path from old_git_dir to new_git_dir.2804 */2805voidrelocate_gitdir(const char*path,const char*old_git_dir,const char*new_git_dir)2806{2807if(rename(old_git_dir, new_git_dir) <0)2808die_errno(_("could not migrate git directory from '%s' to '%s'"),2809 old_git_dir, new_git_dir);28102811connect_work_tree_and_git_dir(path, new_git_dir);2812}