1/* 2 * apply.c 3 * 4 * Copyright (C) Linus Torvalds, 2005 5 * 6 * This applies patches on top of some (arbitrary) version of the SCM. 7 * 8 */ 9#include "cache.h" 10#include "lockfile.h" 11#include "cache-tree.h" 12#include "quote.h" 13#include "blob.h" 14#include "delta.h" 15#include "builtin.h" 16#include "string-list.h" 17#include "dir.h" 18#include "diff.h" 19#include "parse-options.h" 20#include "xdiff-interface.h" 21#include "ll-merge.h" 22#include "rerere.h" 23 24/* 25 * --check turns on checking that the working tree matches the 26 * files that are being modified, but doesn't apply the patch 27 * --stat does just a diffstat, and doesn't actually apply 28 * --numstat does numeric diffstat, and doesn't actually apply 29 * --index-info shows the old and new index info for paths if available. 30 * --index updates the cache as well. 31 * --cached updates only the cache without ever touching the working tree. 32 */ 33static const char *prefix; 34static int prefix_length = -1; 35static int newfd = -1; 36 37static int unidiff_zero; 38static int state_p_value = 1; 39static int p_value_known; 40static int check_index; 41static int update_index; 42static int cached; 43static int diffstat; 44static int numstat; 45static int summary; 46static int check; 47static int apply = 1; 48static int apply_in_reverse; 49static int apply_with_reject; 50static int apply_verbosely; 51static int allow_overlap; 52static int no_add; 53static int threeway; 54static int unsafe_paths; 55static const char *fake_ancestor; 56static int line_termination = '\n'; 57static unsigned int p_context = UINT_MAX; 58static const char * const apply_usage[] = { 59 N_("git apply [<options>] [<patch>...]"), 60 NULL 61}; 62 63static enum ws_error_action { 64 nowarn_ws_error, 65 warn_on_ws_error, 66 die_on_ws_error, 67 correct_ws_error 68} ws_error_action = warn_on_ws_error; 69static int whitespace_error; 70static int squelch_whitespace_errors = 5; 71static int applied_after_fixing_ws; 72 73static enum ws_ignore { 74 ignore_ws_none, 75 ignore_ws_change 76} ws_ignore_action = ignore_ws_none; 77 78 79static const char *patch_input_file; 80static struct strbuf root = STRBUF_INIT; 81static int read_stdin = 1; 82 83static void parse_whitespace_option(const char *option) 84{ 85 if (!option) { 86 ws_error_action = warn_on_ws_error; 87 return; 88 } 89 if (!strcmp(option, "warn")) { 90 ws_error_action = warn_on_ws_error; 91 return; 92 } 93 if (!strcmp(option, "nowarn")) { 94 ws_error_action = nowarn_ws_error; 95 return; 96 } 97 if (!strcmp(option, "error")) { 98 ws_error_action = die_on_ws_error; 99 return; 100 } 101 if (!strcmp(option, "error-all")) { 102 ws_error_action = die_on_ws_error; 103 squelch_whitespace_errors = 0; 104 return; 105 } 106 if (!strcmp(option, "strip") || !strcmp(option, "fix")) { 107 ws_error_action = correct_ws_error; 108 return; 109 } 110 die(_("unrecognized whitespace option '%s'"), option); 111} 112 113static void parse_ignorewhitespace_option(const char *option) 114{ 115 if (!option || !strcmp(option, "no") || 116 !strcmp(option, "false") || !strcmp(option, "never") || 117 !strcmp(option, "none")) { 118 ws_ignore_action = ignore_ws_none; 119 return; 120 } 121 if (!strcmp(option, "change")) { 122 ws_ignore_action = ignore_ws_change; 123 return; 124 } 125 die(_("unrecognized whitespace ignore option '%s'"), option); 126} 127 128static void set_default_whitespace_mode(const char *whitespace_option) 129{ 130 if (!whitespace_option && !apply_default_whitespace) 131 ws_error_action = (apply ? warn_on_ws_error : nowarn_ws_error); 132} 133 134/* 135 * For "diff-stat" like behaviour, we keep track of the biggest change 136 * we've seen, and the longest filename. That allows us to do simple 137 * scaling. 138 */ 139static int max_change, max_len; 140 141/* 142 * Various "current state", notably line numbers and what 143 * file (and how) we're patching right now.. The "is_xxxx" 144 * things are flags, where -1 means "don't know yet". 145 */ 146static int state_linenr = 1; 147 148/* 149 * This represents one "hunk" from a patch, starting with 150 * "@@ -oldpos,oldlines +newpos,newlines @@" marker. The 151 * patch text is pointed at by patch, and its byte length 152 * is stored in size. leading and trailing are the number 153 * of context lines. 154 */ 155struct fragment { 156 unsigned long leading, trailing; 157 unsigned long oldpos, oldlines; 158 unsigned long newpos, newlines; 159 /* 160 * 'patch' is usually borrowed from buf in apply_patch(), 161 * but some codepaths store an allocated buffer. 162 */ 163 const char *patch; 164 unsigned free_patch:1, 165 rejected:1; 166 int size; 167 int linenr; 168 struct fragment *next; 169}; 170 171/* 172 * When dealing with a binary patch, we reuse "leading" field 173 * to store the type of the binary hunk, either deflated "delta" 174 * or deflated "literal". 175 */ 176#define binary_patch_method leading 177#define BINARY_DELTA_DEFLATED 1 178#define BINARY_LITERAL_DEFLATED 2 179 180/* 181 * This represents a "patch" to a file, both metainfo changes 182 * such as creation/deletion, filemode and content changes represented 183 * as a series of fragments. 184 */ 185struct patch { 186 char *new_name, *old_name, *def_name; 187 unsigned int old_mode, new_mode; 188 int is_new, is_delete; /* -1 = unknown, 0 = false, 1 = true */ 189 int rejected; 190 unsigned ws_rule; 191 int lines_added, lines_deleted; 192 int score; 193 unsigned int is_toplevel_relative:1; 194 unsigned int inaccurate_eof:1; 195 unsigned int is_binary:1; 196 unsigned int is_copy:1; 197 unsigned int is_rename:1; 198 unsigned int recount:1; 199 unsigned int conflicted_threeway:1; 200 unsigned int direct_to_threeway:1; 201 struct fragment *fragments; 202 char *result; 203 size_t resultsize; 204 char old_sha1_prefix[41]; 205 char new_sha1_prefix[41]; 206 struct patch *next; 207 208 /* three-way fallback result */ 209 struct object_id threeway_stage[3]; 210}; 211 212static void free_fragment_list(struct fragment *list) 213{ 214 while (list) { 215 struct fragment *next = list->next; 216 if (list->free_patch) 217 free((char *)list->patch); 218 free(list); 219 list = next; 220 } 221} 222 223static void free_patch(struct patch *patch) 224{ 225 free_fragment_list(patch->fragments); 226 free(patch->def_name); 227 free(patch->old_name); 228 free(patch->new_name); 229 free(patch->result); 230 free(patch); 231} 232 233static void free_patch_list(struct patch *list) 234{ 235 while (list) { 236 struct patch *next = list->next; 237 free_patch(list); 238 list = next; 239 } 240} 241 242/* 243 * A line in a file, len-bytes long (includes the terminating LF, 244 * except for an incomplete line at the end if the file ends with 245 * one), and its contents hashes to 'hash'. 246 */ 247struct line { 248 size_t len; 249 unsigned hash : 24; 250 unsigned flag : 8; 251#define LINE_COMMON 1 252#define LINE_PATCHED 2 253}; 254 255/* 256 * This represents a "file", which is an array of "lines". 257 */ 258struct image { 259 char *buf; 260 size_t len; 261 size_t nr; 262 size_t alloc; 263 struct line *line_allocated; 264 struct line *line; 265}; 266 267/* 268 * Records filenames that have been touched, in order to handle 269 * the case where more than one patches touch the same file. 270 */ 271 272static struct string_list fn_table; 273 274static uint32_t hash_line(const char *cp, size_t len) 275{ 276 size_t i; 277 uint32_t h; 278 for (i = 0, h = 0; i < len; i++) { 279 if (!isspace(cp[i])) { 280 h = h * 3 + (cp[i] & 0xff); 281 } 282 } 283 return h; 284} 285 286/* 287 * Compare lines s1 of length n1 and s2 of length n2, ignoring 288 * whitespace difference. Returns 1 if they match, 0 otherwise 289 */ 290static int fuzzy_matchlines(const char *s1, size_t n1, 291 const char *s2, size_t n2) 292{ 293 const char *last1 = s1 + n1 - 1; 294 const char *last2 = s2 + n2 - 1; 295 int result = 0; 296 297 /* ignore line endings */ 298 while ((*last1 == '\r') || (*last1 == '\n')) 299 last1--; 300 while ((*last2 == '\r') || (*last2 == '\n')) 301 last2--; 302 303 /* skip leading whitespaces, if both begin with whitespace */ 304 if (s1 <= last1 && s2 <= last2 && isspace(*s1) && isspace(*s2)) { 305 while (isspace(*s1) && (s1 <= last1)) 306 s1++; 307 while (isspace(*s2) && (s2 <= last2)) 308 s2++; 309 } 310 /* early return if both lines are empty */ 311 if ((s1 > last1) && (s2 > last2)) 312 return 1; 313 while (!result) { 314 result = *s1++ - *s2++; 315 /* 316 * Skip whitespace inside. We check for whitespace on 317 * both buffers because we don't want "a b" to match 318 * "ab" 319 */ 320 if (isspace(*s1) && isspace(*s2)) { 321 while (isspace(*s1) && s1 <= last1) 322 s1++; 323 while (isspace(*s2) && s2 <= last2) 324 s2++; 325 } 326 /* 327 * If we reached the end on one side only, 328 * lines don't match 329 */ 330 if ( 331 ((s2 > last2) && (s1 <= last1)) || 332 ((s1 > last1) && (s2 <= last2))) 333 return 0; 334 if ((s1 > last1) && (s2 > last2)) 335 break; 336 } 337 338 return !result; 339} 340 341static void add_line_info(struct image *img, const char *bol, size_t len, unsigned flag) 342{ 343 ALLOC_GROW(img->line_allocated, img->nr + 1, img->alloc); 344 img->line_allocated[img->nr].len = len; 345 img->line_allocated[img->nr].hash = hash_line(bol, len); 346 img->line_allocated[img->nr].flag = flag; 347 img->nr++; 348} 349 350/* 351 * "buf" has the file contents to be patched (read from various sources). 352 * attach it to "image" and add line-based index to it. 353 * "image" now owns the "buf". 354 */ 355static void prepare_image(struct image *image, char *buf, size_t len, 356 int prepare_linetable) 357{ 358 const char *cp, *ep; 359 360 memset(image, 0, sizeof(*image)); 361 image->buf = buf; 362 image->len = len; 363 364 if (!prepare_linetable) 365 return; 366 367 ep = image->buf + image->len; 368 cp = image->buf; 369 while (cp < ep) { 370 const char *next; 371 for (next = cp; next < ep && *next != '\n'; next++) 372 ; 373 if (next < ep) 374 next++; 375 add_line_info(image, cp, next - cp, 0); 376 cp = next; 377 } 378 image->line = image->line_allocated; 379} 380 381static void clear_image(struct image *image) 382{ 383 free(image->buf); 384 free(image->line_allocated); 385 memset(image, 0, sizeof(*image)); 386} 387 388/* fmt must contain _one_ %s and no other substitution */ 389static void say_patch_name(FILE *output, const char *fmt, struct patch *patch) 390{ 391 struct strbuf sb = STRBUF_INIT; 392 393 if (patch->old_name && patch->new_name && 394 strcmp(patch->old_name, patch->new_name)) { 395 quote_c_style(patch->old_name, &sb, NULL, 0); 396 strbuf_addstr(&sb, " => "); 397 quote_c_style(patch->new_name, &sb, NULL, 0); 398 } else { 399 const char *n = patch->new_name; 400 if (!n) 401 n = patch->old_name; 402 quote_c_style(n, &sb, NULL, 0); 403 } 404 fprintf(output, fmt, sb.buf); 405 fputc('\n', output); 406 strbuf_release(&sb); 407} 408 409#define SLOP (16) 410 411static void read_patch_file(struct strbuf *sb, int fd) 412{ 413 if (strbuf_read(sb, fd, 0) < 0) 414 die_errno("git apply: failed to read"); 415 416 /* 417 * Make sure that we have some slop in the buffer 418 * so that we can do speculative "memcmp" etc, and 419 * see to it that it is NUL-filled. 420 */ 421 strbuf_grow(sb, SLOP); 422 memset(sb->buf + sb->len, 0, SLOP); 423} 424 425static unsigned long linelen(const char *buffer, unsigned long size) 426{ 427 unsigned long len = 0; 428 while (size--) { 429 len++; 430 if (*buffer++ == '\n') 431 break; 432 } 433 return len; 434} 435 436static int is_dev_null(const char *str) 437{ 438 return skip_prefix(str, "/dev/null", &str) && isspace(*str); 439} 440 441#define TERM_SPACE 1 442#define TERM_TAB 2 443 444static int name_terminate(const char *name, int namelen, int c, int terminate) 445{ 446 if (c == ' ' && !(terminate & TERM_SPACE)) 447 return 0; 448 if (c == '\t' && !(terminate & TERM_TAB)) 449 return 0; 450 451 return 1; 452} 453 454/* remove double slashes to make --index work with such filenames */ 455static char *squash_slash(char *name) 456{ 457 int i = 0, j = 0; 458 459 if (!name) 460 return NULL; 461 462 while (name[i]) { 463 if ((name[j++] = name[i++]) == '/') 464 while (name[i] == '/') 465 i++; 466 } 467 name[j] = '\0'; 468 return name; 469} 470 471static char *find_name_gnu(const char *line, const char *def, int p_value) 472{ 473 struct strbuf name = STRBUF_INIT; 474 char *cp; 475 476 /* 477 * Proposed "new-style" GNU patch/diff format; see 478 * http://marc.info/?l=git&m=112927316408690&w=2 479 */ 480 if (unquote_c_style(&name, line, NULL)) { 481 strbuf_release(&name); 482 return NULL; 483 } 484 485 for (cp = name.buf; p_value; p_value--) { 486 cp = strchr(cp, '/'); 487 if (!cp) { 488 strbuf_release(&name); 489 return NULL; 490 } 491 cp++; 492 } 493 494 strbuf_remove(&name, 0, cp - name.buf); 495 if (root.len) 496 strbuf_insert(&name, 0, root.buf, root.len); 497 return squash_slash(strbuf_detach(&name, NULL)); 498} 499 500static size_t sane_tz_len(const char *line, size_t len) 501{ 502 const char *tz, *p; 503 504 if (len < strlen(" +0500") || line[len-strlen(" +0500")] != ' ') 505 return 0; 506 tz = line + len - strlen(" +0500"); 507 508 if (tz[1] != '+' && tz[1] != '-') 509 return 0; 510 511 for (p = tz + 2; p != line + len; p++) 512 if (!isdigit(*p)) 513 return 0; 514 515 return line + len - tz; 516} 517 518static size_t tz_with_colon_len(const char *line, size_t len) 519{ 520 const char *tz, *p; 521 522 if (len < strlen(" +08:00") || line[len - strlen(":00")] != ':') 523 return 0; 524 tz = line + len - strlen(" +08:00"); 525 526 if (tz[0] != ' ' || (tz[1] != '+' && tz[1] != '-')) 527 return 0; 528 p = tz + 2; 529 if (!isdigit(*p++) || !isdigit(*p++) || *p++ != ':' || 530 !isdigit(*p++) || !isdigit(*p++)) 531 return 0; 532 533 return line + len - tz; 534} 535 536static size_t date_len(const char *line, size_t len) 537{ 538 const char *date, *p; 539 540 if (len < strlen("72-02-05") || line[len-strlen("-05")] != '-') 541 return 0; 542 p = date = line + len - strlen("72-02-05"); 543 544 if (!isdigit(*p++) || !isdigit(*p++) || *p++ != '-' || 545 !isdigit(*p++) || !isdigit(*p++) || *p++ != '-' || 546 !isdigit(*p++) || !isdigit(*p++)) /* Not a date. */ 547 return 0; 548 549 if (date - line >= strlen("19") && 550 isdigit(date[-1]) && isdigit(date[-2])) /* 4-digit year */ 551 date -= strlen("19"); 552 553 return line + len - date; 554} 555 556static size_t short_time_len(const char *line, size_t len) 557{ 558 const char *time, *p; 559 560 if (len < strlen(" 07:01:32") || line[len-strlen(":32")] != ':') 561 return 0; 562 p = time = line + len - strlen(" 07:01:32"); 563 564 /* Permit 1-digit hours? */ 565 if (*p++ != ' ' || 566 !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' || 567 !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' || 568 !isdigit(*p++) || !isdigit(*p++)) /* Not a time. */ 569 return 0; 570 571 return line + len - time; 572} 573 574static size_t fractional_time_len(const char *line, size_t len) 575{ 576 const char *p; 577 size_t n; 578 579 /* Expected format: 19:41:17.620000023 */ 580 if (!len || !isdigit(line[len - 1])) 581 return 0; 582 p = line + len - 1; 583 584 /* Fractional seconds. */ 585 while (p > line && isdigit(*p)) 586 p--; 587 if (*p != '.') 588 return 0; 589 590 /* Hours, minutes, and whole seconds. */ 591 n = short_time_len(line, p - line); 592 if (!n) 593 return 0; 594 595 return line + len - p + n; 596} 597 598static size_t trailing_spaces_len(const char *line, size_t len) 599{ 600 const char *p; 601 602 /* Expected format: ' ' x (1 or more) */ 603 if (!len || line[len - 1] != ' ') 604 return 0; 605 606 p = line + len; 607 while (p != line) { 608 p--; 609 if (*p != ' ') 610 return line + len - (p + 1); 611 } 612 613 /* All spaces! */ 614 return len; 615} 616 617static size_t diff_timestamp_len(const char *line, size_t len) 618{ 619 const char *end = line + len; 620 size_t n; 621 622 /* 623 * Posix: 2010-07-05 19:41:17 624 * GNU: 2010-07-05 19:41:17.620000023 -0500 625 */ 626 627 if (!isdigit(end[-1])) 628 return 0; 629 630 n = sane_tz_len(line, end - line); 631 if (!n) 632 n = tz_with_colon_len(line, end - line); 633 end -= n; 634 635 n = short_time_len(line, end - line); 636 if (!n) 637 n = fractional_time_len(line, end - line); 638 end -= n; 639 640 n = date_len(line, end - line); 641 if (!n) /* No date. Too bad. */ 642 return 0; 643 end -= n; 644 645 if (end == line) /* No space before date. */ 646 return 0; 647 if (end[-1] == '\t') { /* Success! */ 648 end--; 649 return line + len - end; 650 } 651 if (end[-1] != ' ') /* No space before date. */ 652 return 0; 653 654 /* Whitespace damage. */ 655 end -= trailing_spaces_len(line, end - line); 656 return line + len - end; 657} 658 659static char *find_name_common(const char *line, const char *def, 660 int p_value, const char *end, int terminate) 661{ 662 int len; 663 const char *start = NULL; 664 665 if (p_value == 0) 666 start = line; 667 while (line != end) { 668 char c = *line; 669 670 if (!end && isspace(c)) { 671 if (c == '\n') 672 break; 673 if (name_terminate(start, line-start, c, terminate)) 674 break; 675 } 676 line++; 677 if (c == '/' && !--p_value) 678 start = line; 679 } 680 if (!start) 681 return squash_slash(xstrdup_or_null(def)); 682 len = line - start; 683 if (!len) 684 return squash_slash(xstrdup_or_null(def)); 685 686 /* 687 * Generally we prefer the shorter name, especially 688 * if the other one is just a variation of that with 689 * something else tacked on to the end (ie "file.orig" 690 * or "file~"). 691 */ 692 if (def) { 693 int deflen = strlen(def); 694 if (deflen < len && !strncmp(start, def, deflen)) 695 return squash_slash(xstrdup(def)); 696 } 697 698 if (root.len) { 699 char *ret = xstrfmt("%s%.*s", root.buf, len, start); 700 return squash_slash(ret); 701 } 702 703 return squash_slash(xmemdupz(start, len)); 704} 705 706static char *find_name(const char *line, char *def, int p_value, int terminate) 707{ 708 if (*line == '"') { 709 char *name = find_name_gnu(line, def, p_value); 710 if (name) 711 return name; 712 } 713 714 return find_name_common(line, def, p_value, NULL, terminate); 715} 716 717static char *find_name_traditional(const char *line, char *def, int p_value) 718{ 719 size_t len; 720 size_t date_len; 721 722 if (*line == '"') { 723 char *name = find_name_gnu(line, def, p_value); 724 if (name) 725 return name; 726 } 727 728 len = strchrnul(line, '\n') - line; 729 date_len = diff_timestamp_len(line, len); 730 if (!date_len) 731 return find_name_common(line, def, p_value, NULL, TERM_TAB); 732 len -= date_len; 733 734 return find_name_common(line, def, p_value, line + len, 0); 735} 736 737static int count_slashes(const char *cp) 738{ 739 int cnt = 0; 740 char ch; 741 742 while ((ch = *cp++)) 743 if (ch == '/') 744 cnt++; 745 return cnt; 746} 747 748/* 749 * Given the string after "--- " or "+++ ", guess the appropriate 750 * p_value for the given patch. 751 */ 752static int guess_p_value(const char *nameline) 753{ 754 char *name, *cp; 755 int val = -1; 756 757 if (is_dev_null(nameline)) 758 return -1; 759 name = find_name_traditional(nameline, NULL, 0); 760 if (!name) 761 return -1; 762 cp = strchr(name, '/'); 763 if (!cp) 764 val = 0; 765 else if (prefix) { 766 /* 767 * Does it begin with "a/$our-prefix" and such? Then this is 768 * very likely to apply to our directory. 769 */ 770 if (!strncmp(name, prefix, prefix_length)) 771 val = count_slashes(prefix); 772 else { 773 cp++; 774 if (!strncmp(cp, prefix, prefix_length)) 775 val = count_slashes(prefix) + 1; 776 } 777 } 778 free(name); 779 return val; 780} 781 782/* 783 * Does the ---/+++ line have the POSIX timestamp after the last HT? 784 * GNU diff puts epoch there to signal a creation/deletion event. Is 785 * this such a timestamp? 786 */ 787static int has_epoch_timestamp(const char *nameline) 788{ 789 /* 790 * We are only interested in epoch timestamp; any non-zero 791 * fraction cannot be one, hence "(\.0+)?" in the regexp below. 792 * For the same reason, the date must be either 1969-12-31 or 793 * 1970-01-01, and the seconds part must be "00". 794 */ 795 const char stamp_regexp[] = 796 "^(1969-12-31|1970-01-01)" 797 " " 798 "[0-2][0-9]:[0-5][0-9]:00(\\.0+)?" 799 " " 800 "([-+][0-2][0-9]:?[0-5][0-9])\n"; 801 const char *timestamp = NULL, *cp, *colon; 802 static regex_t *stamp; 803 regmatch_t m[10]; 804 int zoneoffset; 805 int hourminute; 806 int status; 807 808 for (cp = nameline; *cp != '\n'; cp++) { 809 if (*cp == '\t') 810 timestamp = cp + 1; 811 } 812 if (!timestamp) 813 return 0; 814 if (!stamp) { 815 stamp = xmalloc(sizeof(*stamp)); 816 if (regcomp(stamp, stamp_regexp, REG_EXTENDED)) { 817 warning(_("Cannot prepare timestamp regexp %s"), 818 stamp_regexp); 819 return 0; 820 } 821 } 822 823 status = regexec(stamp, timestamp, ARRAY_SIZE(m), m, 0); 824 if (status) { 825 if (status != REG_NOMATCH) 826 warning(_("regexec returned %d for input: %s"), 827 status, timestamp); 828 return 0; 829 } 830 831 zoneoffset = strtol(timestamp + m[3].rm_so + 1, (char **) &colon, 10); 832 if (*colon == ':') 833 zoneoffset = zoneoffset * 60 + strtol(colon + 1, NULL, 10); 834 else 835 zoneoffset = (zoneoffset / 100) * 60 + (zoneoffset % 100); 836 if (timestamp[m[3].rm_so] == '-') 837 zoneoffset = -zoneoffset; 838 839 /* 840 * YYYY-MM-DD hh:mm:ss must be from either 1969-12-31 841 * (west of GMT) or 1970-01-01 (east of GMT) 842 */ 843 if ((zoneoffset < 0 && memcmp(timestamp, "1969-12-31", 10)) || 844 (0 <= zoneoffset && memcmp(timestamp, "1970-01-01", 10))) 845 return 0; 846 847 hourminute = (strtol(timestamp + 11, NULL, 10) * 60 + 848 strtol(timestamp + 14, NULL, 10) - 849 zoneoffset); 850 851 return ((zoneoffset < 0 && hourminute == 1440) || 852 (0 <= zoneoffset && !hourminute)); 853} 854 855/* 856 * Get the name etc info from the ---/+++ lines of a traditional patch header 857 * 858 * FIXME! The end-of-filename heuristics are kind of screwy. For existing 859 * files, we can happily check the index for a match, but for creating a 860 * new file we should try to match whatever "patch" does. I have no idea. 861 */ 862static void parse_traditional_patch(const char *first, const char *second, struct patch *patch) 863{ 864 char *name; 865 866 first += 4; /* skip "--- " */ 867 second += 4; /* skip "+++ " */ 868 if (!p_value_known) { 869 int p, q; 870 p = guess_p_value(first); 871 q = guess_p_value(second); 872 if (p < 0) p = q; 873 if (0 <= p && p == q) { 874 state_p_value = p; 875 p_value_known = 1; 876 } 877 } 878 if (is_dev_null(first)) { 879 patch->is_new = 1; 880 patch->is_delete = 0; 881 name = find_name_traditional(second, NULL, state_p_value); 882 patch->new_name = name; 883 } else if (is_dev_null(second)) { 884 patch->is_new = 0; 885 patch->is_delete = 1; 886 name = find_name_traditional(first, NULL, state_p_value); 887 patch->old_name = name; 888 } else { 889 char *first_name; 890 first_name = find_name_traditional(first, NULL, state_p_value); 891 name = find_name_traditional(second, first_name, state_p_value); 892 free(first_name); 893 if (has_epoch_timestamp(first)) { 894 patch->is_new = 1; 895 patch->is_delete = 0; 896 patch->new_name = name; 897 } else if (has_epoch_timestamp(second)) { 898 patch->is_new = 0; 899 patch->is_delete = 1; 900 patch->old_name = name; 901 } else { 902 patch->old_name = name; 903 patch->new_name = xstrdup_or_null(name); 904 } 905 } 906 if (!name) 907 die(_("unable to find filename in patch at line %d"), state_linenr); 908} 909 910static int gitdiff_hdrend(const char *line, struct patch *patch) 911{ 912 return -1; 913} 914 915/* 916 * We're anal about diff header consistency, to make 917 * sure that we don't end up having strange ambiguous 918 * patches floating around. 919 * 920 * As a result, gitdiff_{old|new}name() will check 921 * their names against any previous information, just 922 * to make sure.. 923 */ 924#define DIFF_OLD_NAME 0 925#define DIFF_NEW_NAME 1 926 927static void gitdiff_verify_name(const char *line, int isnull, char **name, int side) 928{ 929 if (!*name && !isnull) { 930 *name = find_name(line, NULL, state_p_value, TERM_TAB); 931 return; 932 } 933 934 if (*name) { 935 int len = strlen(*name); 936 char *another; 937 if (isnull) 938 die(_("git apply: bad git-diff - expected /dev/null, got %s on line %d"), 939 *name, state_linenr); 940 another = find_name(line, NULL, state_p_value, TERM_TAB); 941 if (!another || memcmp(another, *name, len + 1)) 942 die((side == DIFF_NEW_NAME) ? 943 _("git apply: bad git-diff - inconsistent new filename on line %d") : 944 _("git apply: bad git-diff - inconsistent old filename on line %d"), state_linenr); 945 free(another); 946 } else { 947 /* expect "/dev/null" */ 948 if (memcmp("/dev/null", line, 9) || line[9] != '\n') 949 die(_("git apply: bad git-diff - expected /dev/null on line %d"), state_linenr); 950 } 951} 952 953static int gitdiff_oldname(const char *line, struct patch *patch) 954{ 955 gitdiff_verify_name(line, patch->is_new, &patch->old_name, 956 DIFF_OLD_NAME); 957 return 0; 958} 959 960static int gitdiff_newname(const char *line, struct patch *patch) 961{ 962 gitdiff_verify_name(line, patch->is_delete, &patch->new_name, 963 DIFF_NEW_NAME); 964 return 0; 965} 966 967static int gitdiff_oldmode(const char *line, struct patch *patch) 968{ 969 patch->old_mode = strtoul(line, NULL, 8); 970 return 0; 971} 972 973static int gitdiff_newmode(const char *line, struct patch *patch) 974{ 975 patch->new_mode = strtoul(line, NULL, 8); 976 return 0; 977} 978 979static int gitdiff_delete(const char *line, struct patch *patch) 980{ 981 patch->is_delete = 1; 982 free(patch->old_name); 983 patch->old_name = xstrdup_or_null(patch->def_name); 984 return gitdiff_oldmode(line, patch); 985} 986 987static int gitdiff_newfile(const char *line, struct patch *patch) 988{ 989 patch->is_new = 1; 990 free(patch->new_name); 991 patch->new_name = xstrdup_or_null(patch->def_name); 992 return gitdiff_newmode(line, patch); 993} 994 995static int gitdiff_copysrc(const char *line, struct patch *patch) 996{ 997 patch->is_copy = 1; 998 free(patch->old_name); 999 patch->old_name = find_name(line, NULL, state_p_value ? state_p_value - 1 : 0, 0);1000 return 0;1001}10021003static int gitdiff_copydst(const char *line, struct patch *patch)1004{1005 patch->is_copy = 1;1006 free(patch->new_name);1007 patch->new_name = find_name(line, NULL, state_p_value ? state_p_value - 1 : 0, 0);1008 return 0;1009}10101011static int gitdiff_renamesrc(const char *line, struct patch *patch)1012{1013 patch->is_rename = 1;1014 free(patch->old_name);1015 patch->old_name = find_name(line, NULL, state_p_value ? state_p_value - 1 : 0, 0);1016 return 0;1017}10181019static int gitdiff_renamedst(const char *line, struct patch *patch)1020{1021 patch->is_rename = 1;1022 free(patch->new_name);1023 patch->new_name = find_name(line, NULL, state_p_value ? state_p_value - 1 : 0, 0);1024 return 0;1025}10261027static int gitdiff_similarity(const char *line, struct patch *patch)1028{1029 unsigned long val = strtoul(line, NULL, 10);1030 if (val <= 100)1031 patch->score = val;1032 return 0;1033}10341035static int gitdiff_dissimilarity(const char *line, struct patch *patch)1036{1037 unsigned long val = strtoul(line, NULL, 10);1038 if (val <= 100)1039 patch->score = val;1040 return 0;1041}10421043static int gitdiff_index(const char *line, struct patch *patch)1044{1045 /*1046 * index line is N hexadecimal, "..", N hexadecimal,1047 * and optional space with octal mode.1048 */1049 const char *ptr, *eol;1050 int len;10511052 ptr = strchr(line, '.');1053 if (!ptr || ptr[1] != '.' || 40 < ptr - line)1054 return 0;1055 len = ptr - line;1056 memcpy(patch->old_sha1_prefix, line, len);1057 patch->old_sha1_prefix[len] = 0;10581059 line = ptr + 2;1060 ptr = strchr(line, ' ');1061 eol = strchrnul(line, '\n');10621063 if (!ptr || eol < ptr)1064 ptr = eol;1065 len = ptr - line;10661067 if (40 < len)1068 return 0;1069 memcpy(patch->new_sha1_prefix, line, len);1070 patch->new_sha1_prefix[len] = 0;1071 if (*ptr == ' ')1072 patch->old_mode = strtoul(ptr+1, NULL, 8);1073 return 0;1074}10751076/*1077 * This is normal for a diff that doesn't change anything: we'll fall through1078 * into the next diff. Tell the parser to break out.1079 */1080static int gitdiff_unrecognized(const char *line, struct patch *patch)1081{1082 return -1;1083}10841085/*1086 * Skip p_value leading components from "line"; as we do not accept1087 * absolute paths, return NULL in that case.1088 */1089static const char *skip_tree_prefix(const char *line, int llen)1090{1091 int nslash;1092 int i;10931094 if (!state_p_value)1095 return (llen && line[0] == '/') ? NULL : line;10961097 nslash = state_p_value;1098 for (i = 0; i < llen; i++) {1099 int ch = line[i];1100 if (ch == '/' && --nslash <= 0)1101 return (i == 0) ? NULL : &line[i + 1];1102 }1103 return NULL;1104}11051106/*1107 * This is to extract the same name that appears on "diff --git"1108 * line. We do not find and return anything if it is a rename1109 * patch, and it is OK because we will find the name elsewhere.1110 * We need to reliably find name only when it is mode-change only,1111 * creation or deletion of an empty file. In any of these cases,1112 * both sides are the same name under a/ and b/ respectively.1113 */1114static char *git_header_name(const char *line, int llen)1115{1116 const char *name;1117 const char *second = NULL;1118 size_t len, line_len;11191120 line += strlen("diff --git ");1121 llen -= strlen("diff --git ");11221123 if (*line == '"') {1124 const char *cp;1125 struct strbuf first = STRBUF_INIT;1126 struct strbuf sp = STRBUF_INIT;11271128 if (unquote_c_style(&first, line, &second))1129 goto free_and_fail1;11301131 /* strip the a/b prefix including trailing slash */1132 cp = skip_tree_prefix(first.buf, first.len);1133 if (!cp)1134 goto free_and_fail1;1135 strbuf_remove(&first, 0, cp - first.buf);11361137 /*1138 * second points at one past closing dq of name.1139 * find the second name.1140 */1141 while ((second < line + llen) && isspace(*second))1142 second++;11431144 if (line + llen <= second)1145 goto free_and_fail1;1146 if (*second == '"') {1147 if (unquote_c_style(&sp, second, NULL))1148 goto free_and_fail1;1149 cp = skip_tree_prefix(sp.buf, sp.len);1150 if (!cp)1151 goto free_and_fail1;1152 /* They must match, otherwise ignore */1153 if (strcmp(cp, first.buf))1154 goto free_and_fail1;1155 strbuf_release(&sp);1156 return strbuf_detach(&first, NULL);1157 }11581159 /* unquoted second */1160 cp = skip_tree_prefix(second, line + llen - second);1161 if (!cp)1162 goto free_and_fail1;1163 if (line + llen - cp != first.len ||1164 memcmp(first.buf, cp, first.len))1165 goto free_and_fail1;1166 return strbuf_detach(&first, NULL);11671168 free_and_fail1:1169 strbuf_release(&first);1170 strbuf_release(&sp);1171 return NULL;1172 }11731174 /* unquoted first name */1175 name = skip_tree_prefix(line, llen);1176 if (!name)1177 return NULL;11781179 /*1180 * since the first name is unquoted, a dq if exists must be1181 * the beginning of the second name.1182 */1183 for (second = name; second < line + llen; second++) {1184 if (*second == '"') {1185 struct strbuf sp = STRBUF_INIT;1186 const char *np;11871188 if (unquote_c_style(&sp, second, NULL))1189 goto free_and_fail2;11901191 np = skip_tree_prefix(sp.buf, sp.len);1192 if (!np)1193 goto free_and_fail2;11941195 len = sp.buf + sp.len - np;1196 if (len < second - name &&1197 !strncmp(np, name, len) &&1198 isspace(name[len])) {1199 /* Good */1200 strbuf_remove(&sp, 0, np - sp.buf);1201 return strbuf_detach(&sp, NULL);1202 }12031204 free_and_fail2:1205 strbuf_release(&sp);1206 return NULL;1207 }1208 }12091210 /*1211 * Accept a name only if it shows up twice, exactly the same1212 * form.1213 */1214 second = strchr(name, '\n');1215 if (!second)1216 return NULL;1217 line_len = second - name;1218 for (len = 0 ; ; len++) {1219 switch (name[len]) {1220 default:1221 continue;1222 case '\n':1223 return NULL;1224 case '\t': case ' ':1225 /*1226 * Is this the separator between the preimage1227 * and the postimage pathname? Again, we are1228 * only interested in the case where there is1229 * no rename, as this is only to set def_name1230 * and a rename patch has the names elsewhere1231 * in an unambiguous form.1232 */1233 if (!name[len + 1])1234 return NULL; /* no postimage name */1235 second = skip_tree_prefix(name + len + 1,1236 line_len - (len + 1));1237 if (!second)1238 return NULL;1239 /*1240 * Does len bytes starting at "name" and "second"1241 * (that are separated by one HT or SP we just1242 * found) exactly match?1243 */1244 if (second[len] == '\n' && !strncmp(name, second, len))1245 return xmemdupz(name, len);1246 }1247 }1248}12491250/* Verify that we recognize the lines following a git header */1251static int parse_git_header(const char *line, int len, unsigned int size, struct patch *patch)1252{1253 unsigned long offset;12541255 /* A git diff has explicit new/delete information, so we don't guess */1256 patch->is_new = 0;1257 patch->is_delete = 0;12581259 /*1260 * Some things may not have the old name in the1261 * rest of the headers anywhere (pure mode changes,1262 * or removing or adding empty files), so we get1263 * the default name from the header.1264 */1265 patch->def_name = git_header_name(line, len);1266 if (patch->def_name && root.len) {1267 char *s = xstrfmt("%s%s", root.buf, patch->def_name);1268 free(patch->def_name);1269 patch->def_name = s;1270 }12711272 line += len;1273 size -= len;1274 state_linenr++;1275 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, state_linenr++) {1276 static const struct opentry {1277 const char *str;1278 int (*fn)(const char *, struct patch *);1279 } optable[] = {1280 { "@@ -", gitdiff_hdrend },1281 { "--- ", gitdiff_oldname },1282 { "+++ ", gitdiff_newname },1283 { "old mode ", gitdiff_oldmode },1284 { "new mode ", gitdiff_newmode },1285 { "deleted file mode ", gitdiff_delete },1286 { "new file mode ", gitdiff_newfile },1287 { "copy from ", gitdiff_copysrc },1288 { "copy to ", gitdiff_copydst },1289 { "rename old ", gitdiff_renamesrc },1290 { "rename new ", gitdiff_renamedst },1291 { "rename from ", gitdiff_renamesrc },1292 { "rename to ", gitdiff_renamedst },1293 { "similarity index ", gitdiff_similarity },1294 { "dissimilarity index ", gitdiff_dissimilarity },1295 { "index ", gitdiff_index },1296 { "", gitdiff_unrecognized },1297 };1298 int i;12991300 len = linelen(line, size);1301 if (!len || line[len-1] != '\n')1302 break;1303 for (i = 0; i < ARRAY_SIZE(optable); i++) {1304 const struct opentry *p = optable + i;1305 int oplen = strlen(p->str);1306 if (len < oplen || memcmp(p->str, line, oplen))1307 continue;1308 if (p->fn(line + oplen, patch) < 0)1309 return offset;1310 break;1311 }1312 }13131314 return offset;1315}13161317static int parse_num(const char *line, unsigned long *p)1318{1319 char *ptr;13201321 if (!isdigit(*line))1322 return 0;1323 *p = strtoul(line, &ptr, 10);1324 return ptr - line;1325}13261327static int parse_range(const char *line, int len, int offset, const char *expect,1328 unsigned long *p1, unsigned long *p2)1329{1330 int digits, ex;13311332 if (offset < 0 || offset >= len)1333 return -1;1334 line += offset;1335 len -= offset;13361337 digits = parse_num(line, p1);1338 if (!digits)1339 return -1;13401341 offset += digits;1342 line += digits;1343 len -= digits;13441345 *p2 = 1;1346 if (*line == ',') {1347 digits = parse_num(line+1, p2);1348 if (!digits)1349 return -1;13501351 offset += digits+1;1352 line += digits+1;1353 len -= digits+1;1354 }13551356 ex = strlen(expect);1357 if (ex > len)1358 return -1;1359 if (memcmp(line, expect, ex))1360 return -1;13611362 return offset + ex;1363}13641365static void recount_diff(const char *line, int size, struct fragment *fragment)1366{1367 int oldlines = 0, newlines = 0, ret = 0;13681369 if (size < 1) {1370 warning("recount: ignore empty hunk");1371 return;1372 }13731374 for (;;) {1375 int len = linelen(line, size);1376 size -= len;1377 line += len;13781379 if (size < 1)1380 break;13811382 switch (*line) {1383 case ' ': case '\n':1384 newlines++;1385 /* fall through */1386 case '-':1387 oldlines++;1388 continue;1389 case '+':1390 newlines++;1391 continue;1392 case '\\':1393 continue;1394 case '@':1395 ret = size < 3 || !starts_with(line, "@@ ");1396 break;1397 case 'd':1398 ret = size < 5 || !starts_with(line, "diff ");1399 break;1400 default:1401 ret = -1;1402 break;1403 }1404 if (ret) {1405 warning(_("recount: unexpected line: %.*s"),1406 (int)linelen(line, size), line);1407 return;1408 }1409 break;1410 }1411 fragment->oldlines = oldlines;1412 fragment->newlines = newlines;1413}14141415/*1416 * Parse a unified diff fragment header of the1417 * form "@@ -a,b +c,d @@"1418 */1419static int parse_fragment_header(const char *line, int len, struct fragment *fragment)1420{1421 int offset;14221423 if (!len || line[len-1] != '\n')1424 return -1;14251426 /* Figure out the number of lines in a fragment */1427 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);1428 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);14291430 return offset;1431}14321433static int find_header(const char *line, unsigned long size, int *hdrsize, struct patch *patch)1434{1435 unsigned long offset, len;14361437 patch->is_toplevel_relative = 0;1438 patch->is_rename = patch->is_copy = 0;1439 patch->is_new = patch->is_delete = -1;1440 patch->old_mode = patch->new_mode = 0;1441 patch->old_name = patch->new_name = NULL;1442 for (offset = 0; size > 0; offset += len, size -= len, line += len, state_linenr++) {1443 unsigned long nextlen;14441445 len = linelen(line, size);1446 if (!len)1447 break;14481449 /* Testing this early allows us to take a few shortcuts.. */1450 if (len < 6)1451 continue;14521453 /*1454 * Make sure we don't find any unconnected patch fragments.1455 * That's a sign that we didn't find a header, and that a1456 * patch has become corrupted/broken up.1457 */1458 if (!memcmp("@@ -", line, 4)) {1459 struct fragment dummy;1460 if (parse_fragment_header(line, len, &dummy) < 0)1461 continue;1462 die(_("patch fragment without header at line %d: %.*s"),1463 state_linenr, (int)len-1, line);1464 }14651466 if (size < len + 6)1467 break;14681469 /*1470 * Git patch? It might not have a real patch, just a rename1471 * or mode change, so we handle that specially1472 */1473 if (!memcmp("diff --git ", line, 11)) {1474 int git_hdr_len = parse_git_header(line, len, size, patch);1475 if (git_hdr_len <= len)1476 continue;1477 if (!patch->old_name && !patch->new_name) {1478 if (!patch->def_name)1479 die(Q_("git diff header lacks filename information when removing "1480 "%d leading pathname component (line %d)",1481 "git diff header lacks filename information when removing "1482 "%d leading pathname components (line %d)",1483 state_p_value),1484 state_p_value, state_linenr);1485 patch->old_name = xstrdup(patch->def_name);1486 patch->new_name = xstrdup(patch->def_name);1487 }1488 if (!patch->is_delete && !patch->new_name)1489 die("git diff header lacks filename information "1490 "(line %d)", state_linenr);1491 patch->is_toplevel_relative = 1;1492 *hdrsize = git_hdr_len;1493 return offset;1494 }14951496 /* --- followed by +++ ? */1497 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))1498 continue;14991500 /*1501 * We only accept unified patches, so we want it to1502 * at least have "@@ -a,b +c,d @@\n", which is 14 chars1503 * minimum ("@@ -0,0 +1 @@\n" is the shortest).1504 */1505 nextlen = linelen(line + len, size - len);1506 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))1507 continue;15081509 /* Ok, we'll consider it a patch */1510 parse_traditional_patch(line, line+len, patch);1511 *hdrsize = len + nextlen;1512 state_linenr += 2;1513 return offset;1514 }1515 return -1;1516}15171518static void record_ws_error(unsigned result, const char *line, int len, int linenr)1519{1520 char *err;15211522 if (!result)1523 return;15241525 whitespace_error++;1526 if (squelch_whitespace_errors &&1527 squelch_whitespace_errors < whitespace_error)1528 return;15291530 err = whitespace_error_string(result);1531 fprintf(stderr, "%s:%d: %s.\n%.*s\n",1532 patch_input_file, linenr, err, len, line);1533 free(err);1534}15351536static void check_whitespace(const char *line, int len, unsigned ws_rule)1537{1538 unsigned result = ws_check(line + 1, len - 1, ws_rule);15391540 record_ws_error(result, line + 1, len - 2, state_linenr);1541}15421543/*1544 * Parse a unified diff. Note that this really needs to parse each1545 * fragment separately, since the only way to know the difference1546 * between a "---" that is part of a patch, and a "---" that starts1547 * the next patch is to look at the line counts..1548 */1549static int parse_fragment(const char *line, unsigned long size,1550 struct patch *patch, struct fragment *fragment)1551{1552 int added, deleted;1553 int len = linelen(line, size), offset;1554 unsigned long oldlines, newlines;1555 unsigned long leading, trailing;15561557 offset = parse_fragment_header(line, len, fragment);1558 if (offset < 0)1559 return -1;1560 if (offset > 0 && patch->recount)1561 recount_diff(line + offset, size - offset, fragment);1562 oldlines = fragment->oldlines;1563 newlines = fragment->newlines;1564 leading = 0;1565 trailing = 0;15661567 /* Parse the thing.. */1568 line += len;1569 size -= len;1570 state_linenr++;1571 added = deleted = 0;1572 for (offset = len;1573 0 < size;1574 offset += len, size -= len, line += len, state_linenr++) {1575 if (!oldlines && !newlines)1576 break;1577 len = linelen(line, size);1578 if (!len || line[len-1] != '\n')1579 return -1;1580 switch (*line) {1581 default:1582 return -1;1583 case '\n': /* newer GNU diff, an empty context line */1584 case ' ':1585 oldlines--;1586 newlines--;1587 if (!deleted && !added)1588 leading++;1589 trailing++;1590 if (!apply_in_reverse &&1591 ws_error_action == correct_ws_error)1592 check_whitespace(line, len, patch->ws_rule);1593 break;1594 case '-':1595 if (apply_in_reverse &&1596 ws_error_action != nowarn_ws_error)1597 check_whitespace(line, len, patch->ws_rule);1598 deleted++;1599 oldlines--;1600 trailing = 0;1601 break;1602 case '+':1603 if (!apply_in_reverse &&1604 ws_error_action != nowarn_ws_error)1605 check_whitespace(line, len, patch->ws_rule);1606 added++;1607 newlines--;1608 trailing = 0;1609 break;16101611 /*1612 * We allow "\ No newline at end of file". Depending1613 * on locale settings when the patch was produced we1614 * don't know what this line looks like. The only1615 * thing we do know is that it begins with "\ ".1616 * Checking for 12 is just for sanity check -- any1617 * l10n of "\ No newline..." is at least that long.1618 */1619 case '\\':1620 if (len < 12 || memcmp(line, "\\ ", 2))1621 return -1;1622 break;1623 }1624 }1625 if (oldlines || newlines)1626 return -1;1627 if (!deleted && !added)1628 return -1;16291630 fragment->leading = leading;1631 fragment->trailing = trailing;16321633 /*1634 * If a fragment ends with an incomplete line, we failed to include1635 * it in the above loop because we hit oldlines == newlines == 01636 * before seeing it.1637 */1638 if (12 < size && !memcmp(line, "\\ ", 2))1639 offset += linelen(line, size);16401641 patch->lines_added += added;1642 patch->lines_deleted += deleted;16431644 if (0 < patch->is_new && oldlines)1645 return error(_("new file depends on old contents"));1646 if (0 < patch->is_delete && newlines)1647 return error(_("deleted file still has contents"));1648 return offset;1649}16501651/*1652 * We have seen "diff --git a/... b/..." header (or a traditional patch1653 * header). Read hunks that belong to this patch into fragments and hang1654 * them to the given patch structure.1655 *1656 * The (fragment->patch, fragment->size) pair points into the memory given1657 * by the caller, not a copy, when we return.1658 */1659static int parse_single_patch(const char *line, unsigned long size, struct patch *patch)1660{1661 unsigned long offset = 0;1662 unsigned long oldlines = 0, newlines = 0, context = 0;1663 struct fragment **fragp = &patch->fragments;16641665 while (size > 4 && !memcmp(line, "@@ -", 4)) {1666 struct fragment *fragment;1667 int len;16681669 fragment = xcalloc(1, sizeof(*fragment));1670 fragment->linenr = state_linenr;1671 len = parse_fragment(line, size, patch, fragment);1672 if (len <= 0)1673 die(_("corrupt patch at line %d"), state_linenr);1674 fragment->patch = line;1675 fragment->size = len;1676 oldlines += fragment->oldlines;1677 newlines += fragment->newlines;1678 context += fragment->leading + fragment->trailing;16791680 *fragp = fragment;1681 fragp = &fragment->next;16821683 offset += len;1684 line += len;1685 size -= len;1686 }16871688 /*1689 * If something was removed (i.e. we have old-lines) it cannot1690 * be creation, and if something was added it cannot be1691 * deletion. However, the reverse is not true; --unified=01692 * patches that only add are not necessarily creation even1693 * though they do not have any old lines, and ones that only1694 * delete are not necessarily deletion.1695 *1696 * Unfortunately, a real creation/deletion patch do _not_ have1697 * any context line by definition, so we cannot safely tell it1698 * apart with --unified=0 insanity. At least if the patch has1699 * more than one hunk it is not creation or deletion.1700 */1701 if (patch->is_new < 0 &&1702 (oldlines || (patch->fragments && patch->fragments->next)))1703 patch->is_new = 0;1704 if (patch->is_delete < 0 &&1705 (newlines || (patch->fragments && patch->fragments->next)))1706 patch->is_delete = 0;17071708 if (0 < patch->is_new && oldlines)1709 die(_("new file %s depends on old contents"), patch->new_name);1710 if (0 < patch->is_delete && newlines)1711 die(_("deleted file %s still has contents"), patch->old_name);1712 if (!patch->is_delete && !newlines && context)1713 fprintf_ln(stderr,1714 _("** warning: "1715 "file %s becomes empty but is not deleted"),1716 patch->new_name);17171718 return offset;1719}17201721static inline int metadata_changes(struct patch *patch)1722{1723 return patch->is_rename > 0 ||1724 patch->is_copy > 0 ||1725 patch->is_new > 0 ||1726 patch->is_delete ||1727 (patch->old_mode && patch->new_mode &&1728 patch->old_mode != patch->new_mode);1729}17301731static char *inflate_it(const void *data, unsigned long size,1732 unsigned long inflated_size)1733{1734 git_zstream stream;1735 void *out;1736 int st;17371738 memset(&stream, 0, sizeof(stream));17391740 stream.next_in = (unsigned char *)data;1741 stream.avail_in = size;1742 stream.next_out = out = xmalloc(inflated_size);1743 stream.avail_out = inflated_size;1744 git_inflate_init(&stream);1745 st = git_inflate(&stream, Z_FINISH);1746 git_inflate_end(&stream);1747 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {1748 free(out);1749 return NULL;1750 }1751 return out;1752}17531754/*1755 * Read a binary hunk and return a new fragment; fragment->patch1756 * points at an allocated memory that the caller must free, so1757 * it is marked as "->free_patch = 1".1758 */1759static struct fragment *parse_binary_hunk(char **buf_p,1760 unsigned long *sz_p,1761 int *status_p,1762 int *used_p)1763{1764 /*1765 * Expect a line that begins with binary patch method ("literal"1766 * or "delta"), followed by the length of data before deflating.1767 * a sequence of 'length-byte' followed by base-85 encoded data1768 * should follow, terminated by a newline.1769 *1770 * Each 5-byte sequence of base-85 encodes up to 4 bytes,1771 * and we would limit the patch line to 66 characters,1772 * so one line can fit up to 13 groups that would decode1773 * to 52 bytes max. The length byte 'A'-'Z' corresponds1774 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.1775 */1776 int llen, used;1777 unsigned long size = *sz_p;1778 char *buffer = *buf_p;1779 int patch_method;1780 unsigned long origlen;1781 char *data = NULL;1782 int hunk_size = 0;1783 struct fragment *frag;17841785 llen = linelen(buffer, size);1786 used = llen;17871788 *status_p = 0;17891790 if (starts_with(buffer, "delta ")) {1791 patch_method = BINARY_DELTA_DEFLATED;1792 origlen = strtoul(buffer + 6, NULL, 10);1793 }1794 else if (starts_with(buffer, "literal ")) {1795 patch_method = BINARY_LITERAL_DEFLATED;1796 origlen = strtoul(buffer + 8, NULL, 10);1797 }1798 else1799 return NULL;18001801 state_linenr++;1802 buffer += llen;1803 while (1) {1804 int byte_length, max_byte_length, newsize;1805 llen = linelen(buffer, size);1806 used += llen;1807 state_linenr++;1808 if (llen == 1) {1809 /* consume the blank line */1810 buffer++;1811 size--;1812 break;1813 }1814 /*1815 * Minimum line is "A00000\n" which is 7-byte long,1816 * and the line length must be multiple of 5 plus 2.1817 */1818 if ((llen < 7) || (llen-2) % 5)1819 goto corrupt;1820 max_byte_length = (llen - 2) / 5 * 4;1821 byte_length = *buffer;1822 if ('A' <= byte_length && byte_length <= 'Z')1823 byte_length = byte_length - 'A' + 1;1824 else if ('a' <= byte_length && byte_length <= 'z')1825 byte_length = byte_length - 'a' + 27;1826 else1827 goto corrupt;1828 /* if the input length was not multiple of 4, we would1829 * have filler at the end but the filler should never1830 * exceed 3 bytes1831 */1832 if (max_byte_length < byte_length ||1833 byte_length <= max_byte_length - 4)1834 goto corrupt;1835 newsize = hunk_size + byte_length;1836 data = xrealloc(data, newsize);1837 if (decode_85(data + hunk_size, buffer + 1, byte_length))1838 goto corrupt;1839 hunk_size = newsize;1840 buffer += llen;1841 size -= llen;1842 }18431844 frag = xcalloc(1, sizeof(*frag));1845 frag->patch = inflate_it(data, hunk_size, origlen);1846 frag->free_patch = 1;1847 if (!frag->patch)1848 goto corrupt;1849 free(data);1850 frag->size = origlen;1851 *buf_p = buffer;1852 *sz_p = size;1853 *used_p = used;1854 frag->binary_patch_method = patch_method;1855 return frag;18561857 corrupt:1858 free(data);1859 *status_p = -1;1860 error(_("corrupt binary patch at line %d: %.*s"),1861 state_linenr-1, llen-1, buffer);1862 return NULL;1863}18641865/*1866 * Returns:1867 * -1 in case of error,1868 * the length of the parsed binary patch otherwise1869 */1870static int parse_binary(char *buffer, unsigned long size, struct patch *patch)1871{1872 /*1873 * We have read "GIT binary patch\n"; what follows is a line1874 * that says the patch method (currently, either "literal" or1875 * "delta") and the length of data before deflating; a1876 * sequence of 'length-byte' followed by base-85 encoded data1877 * follows.1878 *1879 * When a binary patch is reversible, there is another binary1880 * hunk in the same format, starting with patch method (either1881 * "literal" or "delta") with the length of data, and a sequence1882 * of length-byte + base-85 encoded data, terminated with another1883 * empty line. This data, when applied to the postimage, produces1884 * the preimage.1885 */1886 struct fragment *forward;1887 struct fragment *reverse;1888 int status;1889 int used, used_1;18901891 forward = parse_binary_hunk(&buffer, &size, &status, &used);1892 if (!forward && !status)1893 /* there has to be one hunk (forward hunk) */1894 return error(_("unrecognized binary patch at line %d"), state_linenr-1);1895 if (status)1896 /* otherwise we already gave an error message */1897 return status;18981899 reverse = parse_binary_hunk(&buffer, &size, &status, &used_1);1900 if (reverse)1901 used += used_1;1902 else if (status) {1903 /*1904 * Not having reverse hunk is not an error, but having1905 * a corrupt reverse hunk is.1906 */1907 free((void*) forward->patch);1908 free(forward);1909 return status;1910 }1911 forward->next = reverse;1912 patch->fragments = forward;1913 patch->is_binary = 1;1914 return used;1915}19161917static void prefix_one(char **name)1918{1919 char *old_name = *name;1920 if (!old_name)1921 return;1922 *name = xstrdup(prefix_filename(prefix, prefix_length, *name));1923 free(old_name);1924}19251926static void prefix_patch(struct patch *p)1927{1928 if (!prefix || p->is_toplevel_relative)1929 return;1930 prefix_one(&p->new_name);1931 prefix_one(&p->old_name);1932}19331934/*1935 * include/exclude1936 */19371938static struct string_list limit_by_name;1939static int has_include;1940static void add_name_limit(const char *name, int exclude)1941{1942 struct string_list_item *it;19431944 it = string_list_append(&limit_by_name, name);1945 it->util = exclude ? NULL : (void *) 1;1946}19471948static int use_patch(struct patch *p)1949{1950 const char *pathname = p->new_name ? p->new_name : p->old_name;1951 int i;19521953 /* Paths outside are not touched regardless of "--include" */1954 if (0 < prefix_length) {1955 int pathlen = strlen(pathname);1956 if (pathlen <= prefix_length ||1957 memcmp(prefix, pathname, prefix_length))1958 return 0;1959 }19601961 /* See if it matches any of exclude/include rule */1962 for (i = 0; i < limit_by_name.nr; i++) {1963 struct string_list_item *it = &limit_by_name.items[i];1964 if (!wildmatch(it->string, pathname, 0, NULL))1965 return (it->util != NULL);1966 }19671968 /*1969 * If we had any include, a path that does not match any rule is1970 * not used. Otherwise, we saw bunch of exclude rules (or none)1971 * and such a path is used.1972 */1973 return !has_include;1974}197519761977/*1978 * Read the patch text in "buffer" that extends for "size" bytes; stop1979 * reading after seeing a single patch (i.e. changes to a single file).1980 * Create fragments (i.e. patch hunks) and hang them to the given patch.1981 * Return the number of bytes consumed, so that the caller can call us1982 * again for the next patch.1983 */1984static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)1985{1986 int hdrsize, patchsize;1987 int offset = find_header(buffer, size, &hdrsize, patch);19881989 if (offset < 0)1990 return offset;19911992 prefix_patch(patch);19931994 if (!use_patch(patch))1995 patch->ws_rule = 0;1996 else1997 patch->ws_rule = whitespace_rule(patch->new_name1998 ? patch->new_name1999 : patch->old_name);20002001 patchsize = parse_single_patch(buffer + offset + hdrsize,2002 size - offset - hdrsize, patch);20032004 if (!patchsize) {2005 static const char git_binary[] = "GIT binary patch\n";2006 int hd = hdrsize + offset;2007 unsigned long llen = linelen(buffer + hd, size - hd);20082009 if (llen == sizeof(git_binary) - 1 &&2010 !memcmp(git_binary, buffer + hd, llen)) {2011 int used;2012 state_linenr++;2013 used = parse_binary(buffer + hd + llen,2014 size - hd - llen, patch);2015 if (used < 0)2016 return -1;2017 if (used)2018 patchsize = used + llen;2019 else2020 patchsize = 0;2021 }2022 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {2023 static const char *binhdr[] = {2024 "Binary files ",2025 "Files ",2026 NULL,2027 };2028 int i;2029 for (i = 0; binhdr[i]; i++) {2030 int len = strlen(binhdr[i]);2031 if (len < size - hd &&2032 !memcmp(binhdr[i], buffer + hd, len)) {2033 state_linenr++;2034 patch->is_binary = 1;2035 patchsize = llen;2036 break;2037 }2038 }2039 }20402041 /* Empty patch cannot be applied if it is a text patch2042 * without metadata change. A binary patch appears2043 * empty to us here.2044 */2045 if ((apply || check) &&2046 (!patch->is_binary && !metadata_changes(patch)))2047 die(_("patch with only garbage at line %d"), state_linenr);2048 }20492050 return offset + hdrsize + patchsize;2051}20522053#define swap(a,b) myswap((a),(b),sizeof(a))20542055#define myswap(a, b, size) do { \2056 unsigned char mytmp[size]; \2057 memcpy(mytmp, &a, size); \2058 memcpy(&a, &b, size); \2059 memcpy(&b, mytmp, size); \2060} while (0)20612062static void reverse_patches(struct patch *p)2063{2064 for (; p; p = p->next) {2065 struct fragment *frag = p->fragments;20662067 swap(p->new_name, p->old_name);2068 swap(p->new_mode, p->old_mode);2069 swap(p->is_new, p->is_delete);2070 swap(p->lines_added, p->lines_deleted);2071 swap(p->old_sha1_prefix, p->new_sha1_prefix);20722073 for (; frag; frag = frag->next) {2074 swap(frag->newpos, frag->oldpos);2075 swap(frag->newlines, frag->oldlines);2076 }2077 }2078}20792080static const char pluses[] =2081"++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";2082static const char minuses[]=2083"----------------------------------------------------------------------";20842085static void show_stats(struct patch *patch)2086{2087 struct strbuf qname = STRBUF_INIT;2088 char *cp = patch->new_name ? patch->new_name : patch->old_name;2089 int max, add, del;20902091 quote_c_style(cp, &qname, NULL, 0);20922093 /*2094 * "scale" the filename2095 */2096 max = max_len;2097 if (max > 50)2098 max = 50;20992100 if (qname.len > max) {2101 cp = strchr(qname.buf + qname.len + 3 - max, '/');2102 if (!cp)2103 cp = qname.buf + qname.len + 3 - max;2104 strbuf_splice(&qname, 0, cp - qname.buf, "...", 3);2105 }21062107 if (patch->is_binary) {2108 printf(" %-*s | Bin\n", max, qname.buf);2109 strbuf_release(&qname);2110 return;2111 }21122113 printf(" %-*s |", max, qname.buf);2114 strbuf_release(&qname);21152116 /*2117 * scale the add/delete2118 */2119 max = max + max_change > 70 ? 70 - max : max_change;2120 add = patch->lines_added;2121 del = patch->lines_deleted;21222123 if (max_change > 0) {2124 int total = ((add + del) * max + max_change / 2) / max_change;2125 add = (add * max + max_change / 2) / max_change;2126 del = total - add;2127 }2128 printf("%5d %.*s%.*s\n", patch->lines_added + patch->lines_deleted,2129 add, pluses, del, minuses);2130}21312132static int read_old_data(struct stat *st, const char *path, struct strbuf *buf)2133{2134 switch (st->st_mode & S_IFMT) {2135 case S_IFLNK:2136 if (strbuf_readlink(buf, path, st->st_size) < 0)2137 return error(_("unable to read symlink %s"), path);2138 return 0;2139 case S_IFREG:2140 if (strbuf_read_file(buf, path, st->st_size) != st->st_size)2141 return error(_("unable to open or read %s"), path);2142 convert_to_git(path, buf->buf, buf->len, buf, 0);2143 return 0;2144 default:2145 return -1;2146 }2147}21482149/*2150 * Update the preimage, and the common lines in postimage,2151 * from buffer buf of length len. If postlen is 0 the postimage2152 * is updated in place, otherwise it's updated on a new buffer2153 * of length postlen2154 */21552156static void update_pre_post_images(struct image *preimage,2157 struct image *postimage,2158 char *buf,2159 size_t len, size_t postlen)2160{2161 int i, ctx, reduced;2162 char *new, *old, *fixed;2163 struct image fixed_preimage;21642165 /*2166 * Update the preimage with whitespace fixes. Note that we2167 * are not losing preimage->buf -- apply_one_fragment() will2168 * free "oldlines".2169 */2170 prepare_image(&fixed_preimage, buf, len, 1);2171 assert(postlen2172 ? fixed_preimage.nr == preimage->nr2173 : fixed_preimage.nr <= preimage->nr);2174 for (i = 0; i < fixed_preimage.nr; i++)2175 fixed_preimage.line[i].flag = preimage->line[i].flag;2176 free(preimage->line_allocated);2177 *preimage = fixed_preimage;21782179 /*2180 * Adjust the common context lines in postimage. This can be2181 * done in-place when we are shrinking it with whitespace2182 * fixing, but needs a new buffer when ignoring whitespace or2183 * expanding leading tabs to spaces.2184 *2185 * We trust the caller to tell us if the update can be done2186 * in place (postlen==0) or not.2187 */2188 old = postimage->buf;2189 if (postlen)2190 new = postimage->buf = xmalloc(postlen);2191 else2192 new = old;2193 fixed = preimage->buf;21942195 for (i = reduced = ctx = 0; i < postimage->nr; i++) {2196 size_t l_len = postimage->line[i].len;2197 if (!(postimage->line[i].flag & LINE_COMMON)) {2198 /* an added line -- no counterparts in preimage */2199 memmove(new, old, l_len);2200 old += l_len;2201 new += l_len;2202 continue;2203 }22042205 /* a common context -- skip it in the original postimage */2206 old += l_len;22072208 /* and find the corresponding one in the fixed preimage */2209 while (ctx < preimage->nr &&2210 !(preimage->line[ctx].flag & LINE_COMMON)) {2211 fixed += preimage->line[ctx].len;2212 ctx++;2213 }22142215 /*2216 * preimage is expected to run out, if the caller2217 * fixed addition of trailing blank lines.2218 */2219 if (preimage->nr <= ctx) {2220 reduced++;2221 continue;2222 }22232224 /* and copy it in, while fixing the line length */2225 l_len = preimage->line[ctx].len;2226 memcpy(new, fixed, l_len);2227 new += l_len;2228 fixed += l_len;2229 postimage->line[i].len = l_len;2230 ctx++;2231 }22322233 if (postlen2234 ? postlen < new - postimage->buf2235 : postimage->len < new - postimage->buf)2236 die("BUG: caller miscounted postlen: asked %d, orig = %d, used = %d",2237 (int)postlen, (int) postimage->len, (int)(new - postimage->buf));22382239 /* Fix the length of the whole thing */2240 postimage->len = new - postimage->buf;2241 postimage->nr -= reduced;2242}22432244static int line_by_line_fuzzy_match(struct image *img,2245 struct image *preimage,2246 struct image *postimage,2247 unsigned long try,2248 int try_lno,2249 int preimage_limit)2250{2251 int i;2252 size_t imgoff = 0;2253 size_t preoff = 0;2254 size_t postlen = postimage->len;2255 size_t extra_chars;2256 char *buf;2257 char *preimage_eof;2258 char *preimage_end;2259 struct strbuf fixed;2260 char *fixed_buf;2261 size_t fixed_len;22622263 for (i = 0; i < preimage_limit; i++) {2264 size_t prelen = preimage->line[i].len;2265 size_t imglen = img->line[try_lno+i].len;22662267 if (!fuzzy_matchlines(img->buf + try + imgoff, imglen,2268 preimage->buf + preoff, prelen))2269 return 0;2270 if (preimage->line[i].flag & LINE_COMMON)2271 postlen += imglen - prelen;2272 imgoff += imglen;2273 preoff += prelen;2274 }22752276 /*2277 * Ok, the preimage matches with whitespace fuzz.2278 *2279 * imgoff now holds the true length of the target that2280 * matches the preimage before the end of the file.2281 *2282 * Count the number of characters in the preimage that fall2283 * beyond the end of the file and make sure that all of them2284 * are whitespace characters. (This can only happen if2285 * we are removing blank lines at the end of the file.)2286 */2287 buf = preimage_eof = preimage->buf + preoff;2288 for ( ; i < preimage->nr; i++)2289 preoff += preimage->line[i].len;2290 preimage_end = preimage->buf + preoff;2291 for ( ; buf < preimage_end; buf++)2292 if (!isspace(*buf))2293 return 0;22942295 /*2296 * Update the preimage and the common postimage context2297 * lines to use the same whitespace as the target.2298 * If whitespace is missing in the target (i.e.2299 * if the preimage extends beyond the end of the file),2300 * use the whitespace from the preimage.2301 */2302 extra_chars = preimage_end - preimage_eof;2303 strbuf_init(&fixed, imgoff + extra_chars);2304 strbuf_add(&fixed, img->buf + try, imgoff);2305 strbuf_add(&fixed, preimage_eof, extra_chars);2306 fixed_buf = strbuf_detach(&fixed, &fixed_len);2307 update_pre_post_images(preimage, postimage,2308 fixed_buf, fixed_len, postlen);2309 return 1;2310}23112312static int match_fragment(struct image *img,2313 struct image *preimage,2314 struct image *postimage,2315 unsigned long try,2316 int try_lno,2317 unsigned ws_rule,2318 int match_beginning, int match_end)2319{2320 int i;2321 char *fixed_buf, *buf, *orig, *target;2322 struct strbuf fixed;2323 size_t fixed_len, postlen;2324 int preimage_limit;23252326 if (preimage->nr + try_lno <= img->nr) {2327 /*2328 * The hunk falls within the boundaries of img.2329 */2330 preimage_limit = preimage->nr;2331 if (match_end && (preimage->nr + try_lno != img->nr))2332 return 0;2333 } else if (ws_error_action == correct_ws_error &&2334 (ws_rule & WS_BLANK_AT_EOF)) {2335 /*2336 * This hunk extends beyond the end of img, and we are2337 * removing blank lines at the end of the file. This2338 * many lines from the beginning of the preimage must2339 * match with img, and the remainder of the preimage2340 * must be blank.2341 */2342 preimage_limit = img->nr - try_lno;2343 } else {2344 /*2345 * The hunk extends beyond the end of the img and2346 * we are not removing blanks at the end, so we2347 * should reject the hunk at this position.2348 */2349 return 0;2350 }23512352 if (match_beginning && try_lno)2353 return 0;23542355 /* Quick hash check */2356 for (i = 0; i < preimage_limit; i++)2357 if ((img->line[try_lno + i].flag & LINE_PATCHED) ||2358 (preimage->line[i].hash != img->line[try_lno + i].hash))2359 return 0;23602361 if (preimage_limit == preimage->nr) {2362 /*2363 * Do we have an exact match? If we were told to match2364 * at the end, size must be exactly at try+fragsize,2365 * otherwise try+fragsize must be still within the preimage,2366 * and either case, the old piece should match the preimage2367 * exactly.2368 */2369 if ((match_end2370 ? (try + preimage->len == img->len)2371 : (try + preimage->len <= img->len)) &&2372 !memcmp(img->buf + try, preimage->buf, preimage->len))2373 return 1;2374 } else {2375 /*2376 * The preimage extends beyond the end of img, so2377 * there cannot be an exact match.2378 *2379 * There must be one non-blank context line that match2380 * a line before the end of img.2381 */2382 char *buf_end;23832384 buf = preimage->buf;2385 buf_end = buf;2386 for (i = 0; i < preimage_limit; i++)2387 buf_end += preimage->line[i].len;23882389 for ( ; buf < buf_end; buf++)2390 if (!isspace(*buf))2391 break;2392 if (buf == buf_end)2393 return 0;2394 }23952396 /*2397 * No exact match. If we are ignoring whitespace, run a line-by-line2398 * fuzzy matching. We collect all the line length information because2399 * we need it to adjust whitespace if we match.2400 */2401 if (ws_ignore_action == ignore_ws_change)2402 return line_by_line_fuzzy_match(img, preimage, postimage,2403 try, try_lno, preimage_limit);24042405 if (ws_error_action != correct_ws_error)2406 return 0;24072408 /*2409 * The hunk does not apply byte-by-byte, but the hash says2410 * it might with whitespace fuzz. We weren't asked to2411 * ignore whitespace, we were asked to correct whitespace2412 * errors, so let's try matching after whitespace correction.2413 *2414 * While checking the preimage against the target, whitespace2415 * errors in both fixed, we count how large the corresponding2416 * postimage needs to be. The postimage prepared by2417 * apply_one_fragment() has whitespace errors fixed on added2418 * lines already, but the common lines were propagated as-is,2419 * which may become longer when their whitespace errors are2420 * fixed.2421 */24222423 /* First count added lines in postimage */2424 postlen = 0;2425 for (i = 0; i < postimage->nr; i++) {2426 if (!(postimage->line[i].flag & LINE_COMMON))2427 postlen += postimage->line[i].len;2428 }24292430 /*2431 * The preimage may extend beyond the end of the file,2432 * but in this loop we will only handle the part of the2433 * preimage that falls within the file.2434 */2435 strbuf_init(&fixed, preimage->len + 1);2436 orig = preimage->buf;2437 target = img->buf + try;2438 for (i = 0; i < preimage_limit; i++) {2439 size_t oldlen = preimage->line[i].len;2440 size_t tgtlen = img->line[try_lno + i].len;2441 size_t fixstart = fixed.len;2442 struct strbuf tgtfix;2443 int match;24442445 /* Try fixing the line in the preimage */2446 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);24472448 /* Try fixing the line in the target */2449 strbuf_init(&tgtfix, tgtlen);2450 ws_fix_copy(&tgtfix, target, tgtlen, ws_rule, NULL);24512452 /*2453 * If they match, either the preimage was based on2454 * a version before our tree fixed whitespace breakage,2455 * or we are lacking a whitespace-fix patch the tree2456 * the preimage was based on already had (i.e. target2457 * has whitespace breakage, the preimage doesn't).2458 * In either case, we are fixing the whitespace breakages2459 * so we might as well take the fix together with their2460 * real change.2461 */2462 match = (tgtfix.len == fixed.len - fixstart &&2463 !memcmp(tgtfix.buf, fixed.buf + fixstart,2464 fixed.len - fixstart));24652466 /* Add the length if this is common with the postimage */2467 if (preimage->line[i].flag & LINE_COMMON)2468 postlen += tgtfix.len;24692470 strbuf_release(&tgtfix);2471 if (!match)2472 goto unmatch_exit;24732474 orig += oldlen;2475 target += tgtlen;2476 }247724782479 /*2480 * Now handle the lines in the preimage that falls beyond the2481 * end of the file (if any). They will only match if they are2482 * empty or only contain whitespace (if WS_BLANK_AT_EOL is2483 * false).2484 */2485 for ( ; i < preimage->nr; i++) {2486 size_t fixstart = fixed.len; /* start of the fixed preimage */2487 size_t oldlen = preimage->line[i].len;2488 int j;24892490 /* Try fixing the line in the preimage */2491 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);24922493 for (j = fixstart; j < fixed.len; j++)2494 if (!isspace(fixed.buf[j]))2495 goto unmatch_exit;24962497 orig += oldlen;2498 }24992500 /*2501 * Yes, the preimage is based on an older version that still2502 * has whitespace breakages unfixed, and fixing them makes the2503 * hunk match. Update the context lines in the postimage.2504 */2505 fixed_buf = strbuf_detach(&fixed, &fixed_len);2506 if (postlen < postimage->len)2507 postlen = 0;2508 update_pre_post_images(preimage, postimage,2509 fixed_buf, fixed_len, postlen);2510 return 1;25112512 unmatch_exit:2513 strbuf_release(&fixed);2514 return 0;2515}25162517static int find_pos(struct image *img,2518 struct image *preimage,2519 struct image *postimage,2520 int line,2521 unsigned ws_rule,2522 int match_beginning, int match_end)2523{2524 int i;2525 unsigned long backwards, forwards, try;2526 int backwards_lno, forwards_lno, try_lno;25272528 /*2529 * If match_beginning or match_end is specified, there is no2530 * point starting from a wrong line that will never match and2531 * wander around and wait for a match at the specified end.2532 */2533 if (match_beginning)2534 line = 0;2535 else if (match_end)2536 line = img->nr - preimage->nr;25372538 /*2539 * Because the comparison is unsigned, the following test2540 * will also take care of a negative line number that can2541 * result when match_end and preimage is larger than the target.2542 */2543 if ((size_t) line > img->nr)2544 line = img->nr;25452546 try = 0;2547 for (i = 0; i < line; i++)2548 try += img->line[i].len;25492550 /*2551 * There's probably some smart way to do this, but I'll leave2552 * that to the smart and beautiful people. I'm simple and stupid.2553 */2554 backwards = try;2555 backwards_lno = line;2556 forwards = try;2557 forwards_lno = line;2558 try_lno = line;25592560 for (i = 0; ; i++) {2561 if (match_fragment(img, preimage, postimage,2562 try, try_lno, ws_rule,2563 match_beginning, match_end))2564 return try_lno;25652566 again:2567 if (backwards_lno == 0 && forwards_lno == img->nr)2568 break;25692570 if (i & 1) {2571 if (backwards_lno == 0) {2572 i++;2573 goto again;2574 }2575 backwards_lno--;2576 backwards -= img->line[backwards_lno].len;2577 try = backwards;2578 try_lno = backwards_lno;2579 } else {2580 if (forwards_lno == img->nr) {2581 i++;2582 goto again;2583 }2584 forwards += img->line[forwards_lno].len;2585 forwards_lno++;2586 try = forwards;2587 try_lno = forwards_lno;2588 }25892590 }2591 return -1;2592}25932594static void remove_first_line(struct image *img)2595{2596 img->buf += img->line[0].len;2597 img->len -= img->line[0].len;2598 img->line++;2599 img->nr--;2600}26012602static void remove_last_line(struct image *img)2603{2604 img->len -= img->line[--img->nr].len;2605}26062607/*2608 * The change from "preimage" and "postimage" has been found to2609 * apply at applied_pos (counts in line numbers) in "img".2610 * Update "img" to remove "preimage" and replace it with "postimage".2611 */2612static void update_image(struct image *img,2613 int applied_pos,2614 struct image *preimage,2615 struct image *postimage)2616{2617 /*2618 * remove the copy of preimage at offset in img2619 * and replace it with postimage2620 */2621 int i, nr;2622 size_t remove_count, insert_count, applied_at = 0;2623 char *result;2624 int preimage_limit;26252626 /*2627 * If we are removing blank lines at the end of img,2628 * the preimage may extend beyond the end.2629 * If that is the case, we must be careful only to2630 * remove the part of the preimage that falls within2631 * the boundaries of img. Initialize preimage_limit2632 * to the number of lines in the preimage that falls2633 * within the boundaries.2634 */2635 preimage_limit = preimage->nr;2636 if (preimage_limit > img->nr - applied_pos)2637 preimage_limit = img->nr - applied_pos;26382639 for (i = 0; i < applied_pos; i++)2640 applied_at += img->line[i].len;26412642 remove_count = 0;2643 for (i = 0; i < preimage_limit; i++)2644 remove_count += img->line[applied_pos + i].len;2645 insert_count = postimage->len;26462647 /* Adjust the contents */2648 result = xmalloc(st_add3(st_sub(img->len, remove_count), insert_count, 1));2649 memcpy(result, img->buf, applied_at);2650 memcpy(result + applied_at, postimage->buf, postimage->len);2651 memcpy(result + applied_at + postimage->len,2652 img->buf + (applied_at + remove_count),2653 img->len - (applied_at + remove_count));2654 free(img->buf);2655 img->buf = result;2656 img->len += insert_count - remove_count;2657 result[img->len] = '\0';26582659 /* Adjust the line table */2660 nr = img->nr + postimage->nr - preimage_limit;2661 if (preimage_limit < postimage->nr) {2662 /*2663 * NOTE: this knows that we never call remove_first_line()2664 * on anything other than pre/post image.2665 */2666 REALLOC_ARRAY(img->line, nr);2667 img->line_allocated = img->line;2668 }2669 if (preimage_limit != postimage->nr)2670 memmove(img->line + applied_pos + postimage->nr,2671 img->line + applied_pos + preimage_limit,2672 (img->nr - (applied_pos + preimage_limit)) *2673 sizeof(*img->line));2674 memcpy(img->line + applied_pos,2675 postimage->line,2676 postimage->nr * sizeof(*img->line));2677 if (!allow_overlap)2678 for (i = 0; i < postimage->nr; i++)2679 img->line[applied_pos + i].flag |= LINE_PATCHED;2680 img->nr = nr;2681}26822683/*2684 * Use the patch-hunk text in "frag" to prepare two images (preimage and2685 * postimage) for the hunk. Find lines that match "preimage" in "img" and2686 * replace the part of "img" with "postimage" text.2687 */2688static int apply_one_fragment(struct image *img, struct fragment *frag,2689 int inaccurate_eof, unsigned ws_rule,2690 int nth_fragment)2691{2692 int match_beginning, match_end;2693 const char *patch = frag->patch;2694 int size = frag->size;2695 char *old, *oldlines;2696 struct strbuf newlines;2697 int new_blank_lines_at_end = 0;2698 int found_new_blank_lines_at_end = 0;2699 int hunk_linenr = frag->linenr;2700 unsigned long leading, trailing;2701 int pos, applied_pos;2702 struct image preimage;2703 struct image postimage;27042705 memset(&preimage, 0, sizeof(preimage));2706 memset(&postimage, 0, sizeof(postimage));2707 oldlines = xmalloc(size);2708 strbuf_init(&newlines, size);27092710 old = oldlines;2711 while (size > 0) {2712 char first;2713 int len = linelen(patch, size);2714 int plen;2715 int added_blank_line = 0;2716 int is_blank_context = 0;2717 size_t start;27182719 if (!len)2720 break;27212722 /*2723 * "plen" is how much of the line we should use for2724 * the actual patch data. Normally we just remove the2725 * first character on the line, but if the line is2726 * followed by "\ No newline", then we also remove the2727 * last one (which is the newline, of course).2728 */2729 plen = len - 1;2730 if (len < size && patch[len] == '\\')2731 plen--;2732 first = *patch;2733 if (apply_in_reverse) {2734 if (first == '-')2735 first = '+';2736 else if (first == '+')2737 first = '-';2738 }27392740 switch (first) {2741 case '\n':2742 /* Newer GNU diff, empty context line */2743 if (plen < 0)2744 /* ... followed by '\No newline'; nothing */2745 break;2746 *old++ = '\n';2747 strbuf_addch(&newlines, '\n');2748 add_line_info(&preimage, "\n", 1, LINE_COMMON);2749 add_line_info(&postimage, "\n", 1, LINE_COMMON);2750 is_blank_context = 1;2751 break;2752 case ' ':2753 if (plen && (ws_rule & WS_BLANK_AT_EOF) &&2754 ws_blank_line(patch + 1, plen, ws_rule))2755 is_blank_context = 1;2756 case '-':2757 memcpy(old, patch + 1, plen);2758 add_line_info(&preimage, old, plen,2759 (first == ' ' ? LINE_COMMON : 0));2760 old += plen;2761 if (first == '-')2762 break;2763 /* Fall-through for ' ' */2764 case '+':2765 /* --no-add does not add new lines */2766 if (first == '+' && no_add)2767 break;27682769 start = newlines.len;2770 if (first != '+' ||2771 !whitespace_error ||2772 ws_error_action != correct_ws_error) {2773 strbuf_add(&newlines, patch + 1, plen);2774 }2775 else {2776 ws_fix_copy(&newlines, patch + 1, plen, ws_rule, &applied_after_fixing_ws);2777 }2778 add_line_info(&postimage, newlines.buf + start, newlines.len - start,2779 (first == '+' ? 0 : LINE_COMMON));2780 if (first == '+' &&2781 (ws_rule & WS_BLANK_AT_EOF) &&2782 ws_blank_line(patch + 1, plen, ws_rule))2783 added_blank_line = 1;2784 break;2785 case '@': case '\\':2786 /* Ignore it, we already handled it */2787 break;2788 default:2789 if (apply_verbosely)2790 error(_("invalid start of line: '%c'"), first);2791 applied_pos = -1;2792 goto out;2793 }2794 if (added_blank_line) {2795 if (!new_blank_lines_at_end)2796 found_new_blank_lines_at_end = hunk_linenr;2797 new_blank_lines_at_end++;2798 }2799 else if (is_blank_context)2800 ;2801 else2802 new_blank_lines_at_end = 0;2803 patch += len;2804 size -= len;2805 hunk_linenr++;2806 }2807 if (inaccurate_eof &&2808 old > oldlines && old[-1] == '\n' &&2809 newlines.len > 0 && newlines.buf[newlines.len - 1] == '\n') {2810 old--;2811 strbuf_setlen(&newlines, newlines.len - 1);2812 }28132814 leading = frag->leading;2815 trailing = frag->trailing;28162817 /*2818 * A hunk to change lines at the beginning would begin with2819 * @@ -1,L +N,M @@2820 * but we need to be careful. -U0 that inserts before the second2821 * line also has this pattern.2822 *2823 * And a hunk to add to an empty file would begin with2824 * @@ -0,0 +N,M @@2825 *2826 * In other words, a hunk that is (frag->oldpos <= 1) with or2827 * without leading context must match at the beginning.2828 */2829 match_beginning = (!frag->oldpos ||2830 (frag->oldpos == 1 && !unidiff_zero));28312832 /*2833 * A hunk without trailing lines must match at the end.2834 * However, we simply cannot tell if a hunk must match end2835 * from the lack of trailing lines if the patch was generated2836 * with unidiff without any context.2837 */2838 match_end = !unidiff_zero && !trailing;28392840 pos = frag->newpos ? (frag->newpos - 1) : 0;2841 preimage.buf = oldlines;2842 preimage.len = old - oldlines;2843 postimage.buf = newlines.buf;2844 postimage.len = newlines.len;2845 preimage.line = preimage.line_allocated;2846 postimage.line = postimage.line_allocated;28472848 for (;;) {28492850 applied_pos = find_pos(img, &preimage, &postimage, pos,2851 ws_rule, match_beginning, match_end);28522853 if (applied_pos >= 0)2854 break;28552856 /* Am I at my context limits? */2857 if ((leading <= p_context) && (trailing <= p_context))2858 break;2859 if (match_beginning || match_end) {2860 match_beginning = match_end = 0;2861 continue;2862 }28632864 /*2865 * Reduce the number of context lines; reduce both2866 * leading and trailing if they are equal otherwise2867 * just reduce the larger context.2868 */2869 if (leading >= trailing) {2870 remove_first_line(&preimage);2871 remove_first_line(&postimage);2872 pos--;2873 leading--;2874 }2875 if (trailing > leading) {2876 remove_last_line(&preimage);2877 remove_last_line(&postimage);2878 trailing--;2879 }2880 }28812882 if (applied_pos >= 0) {2883 if (new_blank_lines_at_end &&2884 preimage.nr + applied_pos >= img->nr &&2885 (ws_rule & WS_BLANK_AT_EOF) &&2886 ws_error_action != nowarn_ws_error) {2887 record_ws_error(WS_BLANK_AT_EOF, "+", 1,2888 found_new_blank_lines_at_end);2889 if (ws_error_action == correct_ws_error) {2890 while (new_blank_lines_at_end--)2891 remove_last_line(&postimage);2892 }2893 /*2894 * We would want to prevent write_out_results()2895 * from taking place in apply_patch() that follows2896 * the callchain led us here, which is:2897 * apply_patch->check_patch_list->check_patch->2898 * apply_data->apply_fragments->apply_one_fragment2899 */2900 if (ws_error_action == die_on_ws_error)2901 apply = 0;2902 }29032904 if (apply_verbosely && applied_pos != pos) {2905 int offset = applied_pos - pos;2906 if (apply_in_reverse)2907 offset = 0 - offset;2908 fprintf_ln(stderr,2909 Q_("Hunk #%d succeeded at %d (offset %d line).",2910 "Hunk #%d succeeded at %d (offset %d lines).",2911 offset),2912 nth_fragment, applied_pos + 1, offset);2913 }29142915 /*2916 * Warn if it was necessary to reduce the number2917 * of context lines.2918 */2919 if ((leading != frag->leading) ||2920 (trailing != frag->trailing))2921 fprintf_ln(stderr, _("Context reduced to (%ld/%ld)"2922 " to apply fragment at %d"),2923 leading, trailing, applied_pos+1);2924 update_image(img, applied_pos, &preimage, &postimage);2925 } else {2926 if (apply_verbosely)2927 error(_("while searching for:\n%.*s"),2928 (int)(old - oldlines), oldlines);2929 }29302931out:2932 free(oldlines);2933 strbuf_release(&newlines);2934 free(preimage.line_allocated);2935 free(postimage.line_allocated);29362937 return (applied_pos < 0);2938}29392940static int apply_binary_fragment(struct image *img, struct patch *patch)2941{2942 struct fragment *fragment = patch->fragments;2943 unsigned long len;2944 void *dst;29452946 if (!fragment)2947 return error(_("missing binary patch data for '%s'"),2948 patch->new_name ?2949 patch->new_name :2950 patch->old_name);29512952 /* Binary patch is irreversible without the optional second hunk */2953 if (apply_in_reverse) {2954 if (!fragment->next)2955 return error("cannot reverse-apply a binary patch "2956 "without the reverse hunk to '%s'",2957 patch->new_name2958 ? patch->new_name : patch->old_name);2959 fragment = fragment->next;2960 }2961 switch (fragment->binary_patch_method) {2962 case BINARY_DELTA_DEFLATED:2963 dst = patch_delta(img->buf, img->len, fragment->patch,2964 fragment->size, &len);2965 if (!dst)2966 return -1;2967 clear_image(img);2968 img->buf = dst;2969 img->len = len;2970 return 0;2971 case BINARY_LITERAL_DEFLATED:2972 clear_image(img);2973 img->len = fragment->size;2974 img->buf = xmemdupz(fragment->patch, img->len);2975 return 0;2976 }2977 return -1;2978}29792980/*2981 * Replace "img" with the result of applying the binary patch.2982 * The binary patch data itself in patch->fragment is still kept2983 * but the preimage prepared by the caller in "img" is freed here2984 * or in the helper function apply_binary_fragment() this calls.2985 */2986static int apply_binary(struct image *img, struct patch *patch)2987{2988 const char *name = patch->old_name ? patch->old_name : patch->new_name;2989 unsigned char sha1[20];29902991 /*2992 * For safety, we require patch index line to contain2993 * full 40-byte textual SHA1 for old and new, at least for now.2994 */2995 if (strlen(patch->old_sha1_prefix) != 40 ||2996 strlen(patch->new_sha1_prefix) != 40 ||2997 get_sha1_hex(patch->old_sha1_prefix, sha1) ||2998 get_sha1_hex(patch->new_sha1_prefix, sha1))2999 return error("cannot apply binary patch to '%s' "3000 "without full index line", name);30013002 if (patch->old_name) {3003 /*3004 * See if the old one matches what the patch3005 * applies to.3006 */3007 hash_sha1_file(img->buf, img->len, blob_type, sha1);3008 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))3009 return error("the patch applies to '%s' (%s), "3010 "which does not match the "3011 "current contents.",3012 name, sha1_to_hex(sha1));3013 }3014 else {3015 /* Otherwise, the old one must be empty. */3016 if (img->len)3017 return error("the patch applies to an empty "3018 "'%s' but it is not empty", name);3019 }30203021 get_sha1_hex(patch->new_sha1_prefix, sha1);3022 if (is_null_sha1(sha1)) {3023 clear_image(img);3024 return 0; /* deletion patch */3025 }30263027 if (has_sha1_file(sha1)) {3028 /* We already have the postimage */3029 enum object_type type;3030 unsigned long size;3031 char *result;30323033 result = read_sha1_file(sha1, &type, &size);3034 if (!result)3035 return error("the necessary postimage %s for "3036 "'%s' cannot be read",3037 patch->new_sha1_prefix, name);3038 clear_image(img);3039 img->buf = result;3040 img->len = size;3041 } else {3042 /*3043 * We have verified buf matches the preimage;3044 * apply the patch data to it, which is stored3045 * in the patch->fragments->{patch,size}.3046 */3047 if (apply_binary_fragment(img, patch))3048 return error(_("binary patch does not apply to '%s'"),3049 name);30503051 /* verify that the result matches */3052 hash_sha1_file(img->buf, img->len, blob_type, sha1);3053 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))3054 return error(_("binary patch to '%s' creates incorrect result (expecting %s, got %s)"),3055 name, patch->new_sha1_prefix, sha1_to_hex(sha1));3056 }30573058 return 0;3059}30603061static int apply_fragments(struct image *img, struct patch *patch)3062{3063 struct fragment *frag = patch->fragments;3064 const char *name = patch->old_name ? patch->old_name : patch->new_name;3065 unsigned ws_rule = patch->ws_rule;3066 unsigned inaccurate_eof = patch->inaccurate_eof;3067 int nth = 0;30683069 if (patch->is_binary)3070 return apply_binary(img, patch);30713072 while (frag) {3073 nth++;3074 if (apply_one_fragment(img, frag, inaccurate_eof, ws_rule, nth)) {3075 error(_("patch failed: %s:%ld"), name, frag->oldpos);3076 if (!apply_with_reject)3077 return -1;3078 frag->rejected = 1;3079 }3080 frag = frag->next;3081 }3082 return 0;3083}30843085static int read_blob_object(struct strbuf *buf, const unsigned char *sha1, unsigned mode)3086{3087 if (S_ISGITLINK(mode)) {3088 strbuf_grow(buf, 100);3089 strbuf_addf(buf, "Subproject commit %s\n", sha1_to_hex(sha1));3090 } else {3091 enum object_type type;3092 unsigned long sz;3093 char *result;30943095 result = read_sha1_file(sha1, &type, &sz);3096 if (!result)3097 return -1;3098 /* XXX read_sha1_file NUL-terminates */3099 strbuf_attach(buf, result, sz, sz + 1);3100 }3101 return 0;3102}31033104static int read_file_or_gitlink(const struct cache_entry *ce, struct strbuf *buf)3105{3106 if (!ce)3107 return 0;3108 return read_blob_object(buf, ce->sha1, ce->ce_mode);3109}31103111static struct patch *in_fn_table(const char *name)3112{3113 struct string_list_item *item;31143115 if (name == NULL)3116 return NULL;31173118 item = string_list_lookup(&fn_table, name);3119 if (item != NULL)3120 return (struct patch *)item->util;31213122 return NULL;3123}31243125/*3126 * item->util in the filename table records the status of the path.3127 * Usually it points at a patch (whose result records the contents3128 * of it after applying it), but it could be PATH_WAS_DELETED for a3129 * path that a previously applied patch has already removed, or3130 * PATH_TO_BE_DELETED for a path that a later patch would remove.3131 *3132 * The latter is needed to deal with a case where two paths A and B3133 * are swapped by first renaming A to B and then renaming B to A;3134 * moving A to B should not be prevented due to presence of B as we3135 * will remove it in a later patch.3136 */3137#define PATH_TO_BE_DELETED ((struct patch *) -2)3138#define PATH_WAS_DELETED ((struct patch *) -1)31393140static int to_be_deleted(struct patch *patch)3141{3142 return patch == PATH_TO_BE_DELETED;3143}31443145static int was_deleted(struct patch *patch)3146{3147 return patch == PATH_WAS_DELETED;3148}31493150static void add_to_fn_table(struct patch *patch)3151{3152 struct string_list_item *item;31533154 /*3155 * Always add new_name unless patch is a deletion3156 * This should cover the cases for normal diffs,3157 * file creations and copies3158 */3159 if (patch->new_name != NULL) {3160 item = string_list_insert(&fn_table, patch->new_name);3161 item->util = patch;3162 }31633164 /*3165 * store a failure on rename/deletion cases because3166 * later chunks shouldn't patch old names3167 */3168 if ((patch->new_name == NULL) || (patch->is_rename)) {3169 item = string_list_insert(&fn_table, patch->old_name);3170 item->util = PATH_WAS_DELETED;3171 }3172}31733174static void prepare_fn_table(struct patch *patch)3175{3176 /*3177 * store information about incoming file deletion3178 */3179 while (patch) {3180 if ((patch->new_name == NULL) || (patch->is_rename)) {3181 struct string_list_item *item;3182 item = string_list_insert(&fn_table, patch->old_name);3183 item->util = PATH_TO_BE_DELETED;3184 }3185 patch = patch->next;3186 }3187}31883189static int checkout_target(struct index_state *istate,3190 struct cache_entry *ce, struct stat *st)3191{3192 struct checkout costate;31933194 memset(&costate, 0, sizeof(costate));3195 costate.base_dir = "";3196 costate.refresh_cache = 1;3197 costate.istate = istate;3198 if (checkout_entry(ce, &costate, NULL) || lstat(ce->name, st))3199 return error(_("cannot checkout %s"), ce->name);3200 return 0;3201}32023203static struct patch *previous_patch(struct patch *patch, int *gone)3204{3205 struct patch *previous;32063207 *gone = 0;3208 if (patch->is_copy || patch->is_rename)3209 return NULL; /* "git" patches do not depend on the order */32103211 previous = in_fn_table(patch->old_name);3212 if (!previous)3213 return NULL;32143215 if (to_be_deleted(previous))3216 return NULL; /* the deletion hasn't happened yet */32173218 if (was_deleted(previous))3219 *gone = 1;32203221 return previous;3222}32233224static int verify_index_match(const struct cache_entry *ce, struct stat *st)3225{3226 if (S_ISGITLINK(ce->ce_mode)) {3227 if (!S_ISDIR(st->st_mode))3228 return -1;3229 return 0;3230 }3231 return ce_match_stat(ce, st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE);3232}32333234#define SUBMODULE_PATCH_WITHOUT_INDEX 132353236static int load_patch_target(struct strbuf *buf,3237 const struct cache_entry *ce,3238 struct stat *st,3239 const char *name,3240 unsigned expected_mode)3241{3242 if (cached || check_index) {3243 if (read_file_or_gitlink(ce, buf))3244 return error(_("read of %s failed"), name);3245 } else if (name) {3246 if (S_ISGITLINK(expected_mode)) {3247 if (ce)3248 return read_file_or_gitlink(ce, buf);3249 else3250 return SUBMODULE_PATCH_WITHOUT_INDEX;3251 } else if (has_symlink_leading_path(name, strlen(name))) {3252 return error(_("reading from '%s' beyond a symbolic link"), name);3253 } else {3254 if (read_old_data(st, name, buf))3255 return error(_("read of %s failed"), name);3256 }3257 }3258 return 0;3259}32603261/*3262 * We are about to apply "patch"; populate the "image" with the3263 * current version we have, from the working tree or from the index,3264 * depending on the situation e.g. --cached/--index. If we are3265 * applying a non-git patch that incrementally updates the tree,3266 * we read from the result of a previous diff.3267 */3268static int load_preimage(struct image *image,3269 struct patch *patch, struct stat *st,3270 const struct cache_entry *ce)3271{3272 struct strbuf buf = STRBUF_INIT;3273 size_t len;3274 char *img;3275 struct patch *previous;3276 int status;32773278 previous = previous_patch(patch, &status);3279 if (status)3280 return error(_("path %s has been renamed/deleted"),3281 patch->old_name);3282 if (previous) {3283 /* We have a patched copy in memory; use that. */3284 strbuf_add(&buf, previous->result, previous->resultsize);3285 } else {3286 status = load_patch_target(&buf, ce, st,3287 patch->old_name, patch->old_mode);3288 if (status < 0)3289 return status;3290 else if (status == SUBMODULE_PATCH_WITHOUT_INDEX) {3291 /*3292 * There is no way to apply subproject3293 * patch without looking at the index.3294 * NEEDSWORK: shouldn't this be flagged3295 * as an error???3296 */3297 free_fragment_list(patch->fragments);3298 patch->fragments = NULL;3299 } else if (status) {3300 return error(_("read of %s failed"), patch->old_name);3301 }3302 }33033304 img = strbuf_detach(&buf, &len);3305 prepare_image(image, img, len, !patch->is_binary);3306 return 0;3307}33083309static int three_way_merge(struct image *image,3310 char *path,3311 const unsigned char *base,3312 const unsigned char *ours,3313 const unsigned char *theirs)3314{3315 mmfile_t base_file, our_file, their_file;3316 mmbuffer_t result = { NULL };3317 int status;33183319 read_mmblob(&base_file, base);3320 read_mmblob(&our_file, ours);3321 read_mmblob(&their_file, theirs);3322 status = ll_merge(&result, path,3323 &base_file, "base",3324 &our_file, "ours",3325 &their_file, "theirs", NULL);3326 free(base_file.ptr);3327 free(our_file.ptr);3328 free(their_file.ptr);3329 if (status < 0 || !result.ptr) {3330 free(result.ptr);3331 return -1;3332 }3333 clear_image(image);3334 image->buf = result.ptr;3335 image->len = result.size;33363337 return status;3338}33393340/*3341 * When directly falling back to add/add three-way merge, we read from3342 * the current contents of the new_name. In no cases other than that3343 * this function will be called.3344 */3345static int load_current(struct image *image, struct patch *patch)3346{3347 struct strbuf buf = STRBUF_INIT;3348 int status, pos;3349 size_t len;3350 char *img;3351 struct stat st;3352 struct cache_entry *ce;3353 char *name = patch->new_name;3354 unsigned mode = patch->new_mode;33553356 if (!patch->is_new)3357 die("BUG: patch to %s is not a creation", patch->old_name);33583359 pos = cache_name_pos(name, strlen(name));3360 if (pos < 0)3361 return error(_("%s: does not exist in index"), name);3362 ce = active_cache[pos];3363 if (lstat(name, &st)) {3364 if (errno != ENOENT)3365 return error(_("%s: %s"), name, strerror(errno));3366 if (checkout_target(&the_index, ce, &st))3367 return -1;3368 }3369 if (verify_index_match(ce, &st))3370 return error(_("%s: does not match index"), name);33713372 status = load_patch_target(&buf, ce, &st, name, mode);3373 if (status < 0)3374 return status;3375 else if (status)3376 return -1;3377 img = strbuf_detach(&buf, &len);3378 prepare_image(image, img, len, !patch->is_binary);3379 return 0;3380}33813382static int try_threeway(struct image *image, struct patch *patch,3383 struct stat *st, const struct cache_entry *ce)3384{3385 unsigned char pre_sha1[20], post_sha1[20], our_sha1[20];3386 struct strbuf buf = STRBUF_INIT;3387 size_t len;3388 int status;3389 char *img;3390 struct image tmp_image;33913392 /* No point falling back to 3-way merge in these cases */3393 if (patch->is_delete ||3394 S_ISGITLINK(patch->old_mode) || S_ISGITLINK(patch->new_mode))3395 return -1;33963397 /* Preimage the patch was prepared for */3398 if (patch->is_new)3399 write_sha1_file("", 0, blob_type, pre_sha1);3400 else if (get_sha1(patch->old_sha1_prefix, pre_sha1) ||3401 read_blob_object(&buf, pre_sha1, patch->old_mode))3402 return error("repository lacks the necessary blob to fall back on 3-way merge.");34033404 fprintf(stderr, "Falling back to three-way merge...\n");34053406 img = strbuf_detach(&buf, &len);3407 prepare_image(&tmp_image, img, len, 1);3408 /* Apply the patch to get the post image */3409 if (apply_fragments(&tmp_image, patch) < 0) {3410 clear_image(&tmp_image);3411 return -1;3412 }3413 /* post_sha1[] is theirs */3414 write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, post_sha1);3415 clear_image(&tmp_image);34163417 /* our_sha1[] is ours */3418 if (patch->is_new) {3419 if (load_current(&tmp_image, patch))3420 return error("cannot read the current contents of '%s'",3421 patch->new_name);3422 } else {3423 if (load_preimage(&tmp_image, patch, st, ce))3424 return error("cannot read the current contents of '%s'",3425 patch->old_name);3426 }3427 write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, our_sha1);3428 clear_image(&tmp_image);34293430 /* in-core three-way merge between post and our using pre as base */3431 status = three_way_merge(image, patch->new_name,3432 pre_sha1, our_sha1, post_sha1);3433 if (status < 0) {3434 fprintf(stderr, "Failed to fall back on three-way merge...\n");3435 return status;3436 }34373438 if (status) {3439 patch->conflicted_threeway = 1;3440 if (patch->is_new)3441 oidclr(&patch->threeway_stage[0]);3442 else3443 hashcpy(patch->threeway_stage[0].hash, pre_sha1);3444 hashcpy(patch->threeway_stage[1].hash, our_sha1);3445 hashcpy(patch->threeway_stage[2].hash, post_sha1);3446 fprintf(stderr, "Applied patch to '%s' with conflicts.\n", patch->new_name);3447 } else {3448 fprintf(stderr, "Applied patch to '%s' cleanly.\n", patch->new_name);3449 }3450 return 0;3451}34523453static int apply_data(struct patch *patch, struct stat *st, const struct cache_entry *ce)3454{3455 struct image image;34563457 if (load_preimage(&image, patch, st, ce) < 0)3458 return -1;34593460 if (patch->direct_to_threeway ||3461 apply_fragments(&image, patch) < 0) {3462 /* Note: with --reject, apply_fragments() returns 0 */3463 if (!threeway || try_threeway(&image, patch, st, ce) < 0)3464 return -1;3465 }3466 patch->result = image.buf;3467 patch->resultsize = image.len;3468 add_to_fn_table(patch);3469 free(image.line_allocated);34703471 if (0 < patch->is_delete && patch->resultsize)3472 return error(_("removal patch leaves file contents"));34733474 return 0;3475}34763477/*3478 * If "patch" that we are looking at modifies or deletes what we have,3479 * we would want it not to lose any local modification we have, either3480 * in the working tree or in the index.3481 *3482 * This also decides if a non-git patch is a creation patch or a3483 * modification to an existing empty file. We do not check the state3484 * of the current tree for a creation patch in this function; the caller3485 * check_patch() separately makes sure (and errors out otherwise) that3486 * the path the patch creates does not exist in the current tree.3487 */3488static int check_preimage(struct patch *patch, struct cache_entry **ce, struct stat *st)3489{3490 const char *old_name = patch->old_name;3491 struct patch *previous = NULL;3492 int stat_ret = 0, status;3493 unsigned st_mode = 0;34943495 if (!old_name)3496 return 0;34973498 assert(patch->is_new <= 0);3499 previous = previous_patch(patch, &status);35003501 if (status)3502 return error(_("path %s has been renamed/deleted"), old_name);3503 if (previous) {3504 st_mode = previous->new_mode;3505 } else if (!cached) {3506 stat_ret = lstat(old_name, st);3507 if (stat_ret && errno != ENOENT)3508 return error(_("%s: %s"), old_name, strerror(errno));3509 }35103511 if (check_index && !previous) {3512 int pos = cache_name_pos(old_name, strlen(old_name));3513 if (pos < 0) {3514 if (patch->is_new < 0)3515 goto is_new;3516 return error(_("%s: does not exist in index"), old_name);3517 }3518 *ce = active_cache[pos];3519 if (stat_ret < 0) {3520 if (checkout_target(&the_index, *ce, st))3521 return -1;3522 }3523 if (!cached && verify_index_match(*ce, st))3524 return error(_("%s: does not match index"), old_name);3525 if (cached)3526 st_mode = (*ce)->ce_mode;3527 } else if (stat_ret < 0) {3528 if (patch->is_new < 0)3529 goto is_new;3530 return error(_("%s: %s"), old_name, strerror(errno));3531 }35323533 if (!cached && !previous)3534 st_mode = ce_mode_from_stat(*ce, st->st_mode);35353536 if (patch->is_new < 0)3537 patch->is_new = 0;3538 if (!patch->old_mode)3539 patch->old_mode = st_mode;3540 if ((st_mode ^ patch->old_mode) & S_IFMT)3541 return error(_("%s: wrong type"), old_name);3542 if (st_mode != patch->old_mode)3543 warning(_("%s has type %o, expected %o"),3544 old_name, st_mode, patch->old_mode);3545 if (!patch->new_mode && !patch->is_delete)3546 patch->new_mode = st_mode;3547 return 0;35483549 is_new:3550 patch->is_new = 1;3551 patch->is_delete = 0;3552 free(patch->old_name);3553 patch->old_name = NULL;3554 return 0;3555}355635573558#define EXISTS_IN_INDEX 13559#define EXISTS_IN_WORKTREE 235603561static int check_to_create(const char *new_name, int ok_if_exists)3562{3563 struct stat nst;35643565 if (check_index &&3566 cache_name_pos(new_name, strlen(new_name)) >= 0 &&3567 !ok_if_exists)3568 return EXISTS_IN_INDEX;3569 if (cached)3570 return 0;35713572 if (!lstat(new_name, &nst)) {3573 if (S_ISDIR(nst.st_mode) || ok_if_exists)3574 return 0;3575 /*3576 * A leading component of new_name might be a symlink3577 * that is going to be removed with this patch, but3578 * still pointing at somewhere that has the path.3579 * In such a case, path "new_name" does not exist as3580 * far as git is concerned.3581 */3582 if (has_symlink_leading_path(new_name, strlen(new_name)))3583 return 0;35843585 return EXISTS_IN_WORKTREE;3586 } else if ((errno != ENOENT) && (errno != ENOTDIR)) {3587 return error("%s: %s", new_name, strerror(errno));3588 }3589 return 0;3590}35913592/*3593 * We need to keep track of how symlinks in the preimage are3594 * manipulated by the patches. A patch to add a/b/c where a/b3595 * is a symlink should not be allowed to affect the directory3596 * the symlink points at, but if the same patch removes a/b,3597 * it is perfectly fine, as the patch removes a/b to make room3598 * to create a directory a/b so that a/b/c can be created.3599 */3600static struct string_list symlink_changes;3601#define SYMLINK_GOES_AWAY 013602#define SYMLINK_IN_RESULT 0236033604static uintptr_t register_symlink_changes(const char *path, uintptr_t what)3605{3606 struct string_list_item *ent;36073608 ent = string_list_lookup(&symlink_changes, path);3609 if (!ent) {3610 ent = string_list_insert(&symlink_changes, path);3611 ent->util = (void *)0;3612 }3613 ent->util = (void *)(what | ((uintptr_t)ent->util));3614 return (uintptr_t)ent->util;3615}36163617static uintptr_t check_symlink_changes(const char *path)3618{3619 struct string_list_item *ent;36203621 ent = string_list_lookup(&symlink_changes, path);3622 if (!ent)3623 return 0;3624 return (uintptr_t)ent->util;3625}36263627static void prepare_symlink_changes(struct patch *patch)3628{3629 for ( ; patch; patch = patch->next) {3630 if ((patch->old_name && S_ISLNK(patch->old_mode)) &&3631 (patch->is_rename || patch->is_delete))3632 /* the symlink at patch->old_name is removed */3633 register_symlink_changes(patch->old_name, SYMLINK_GOES_AWAY);36343635 if (patch->new_name && S_ISLNK(patch->new_mode))3636 /* the symlink at patch->new_name is created or remains */3637 register_symlink_changes(patch->new_name, SYMLINK_IN_RESULT);3638 }3639}36403641static int path_is_beyond_symlink_1(struct strbuf *name)3642{3643 do {3644 unsigned int change;36453646 while (--name->len && name->buf[name->len] != '/')3647 ; /* scan backwards */3648 if (!name->len)3649 break;3650 name->buf[name->len] = '\0';3651 change = check_symlink_changes(name->buf);3652 if (change & SYMLINK_IN_RESULT)3653 return 1;3654 if (change & SYMLINK_GOES_AWAY)3655 /*3656 * This cannot be "return 0", because we may3657 * see a new one created at a higher level.3658 */3659 continue;36603661 /* otherwise, check the preimage */3662 if (check_index) {3663 struct cache_entry *ce;36643665 ce = cache_file_exists(name->buf, name->len, ignore_case);3666 if (ce && S_ISLNK(ce->ce_mode))3667 return 1;3668 } else {3669 struct stat st;3670 if (!lstat(name->buf, &st) && S_ISLNK(st.st_mode))3671 return 1;3672 }3673 } while (1);3674 return 0;3675}36763677static int path_is_beyond_symlink(const char *name_)3678{3679 int ret;3680 struct strbuf name = STRBUF_INIT;36813682 assert(*name_ != '\0');3683 strbuf_addstr(&name, name_);3684 ret = path_is_beyond_symlink_1(&name);3685 strbuf_release(&name);36863687 return ret;3688}36893690static void die_on_unsafe_path(struct patch *patch)3691{3692 const char *old_name = NULL;3693 const char *new_name = NULL;3694 if (patch->is_delete)3695 old_name = patch->old_name;3696 else if (!patch->is_new && !patch->is_copy)3697 old_name = patch->old_name;3698 if (!patch->is_delete)3699 new_name = patch->new_name;37003701 if (old_name && !verify_path(old_name))3702 die(_("invalid path '%s'"), old_name);3703 if (new_name && !verify_path(new_name))3704 die(_("invalid path '%s'"), new_name);3705}37063707/*3708 * Check and apply the patch in-core; leave the result in patch->result3709 * for the caller to write it out to the final destination.3710 */3711static int check_patch(struct patch *patch)3712{3713 struct stat st;3714 const char *old_name = patch->old_name;3715 const char *new_name = patch->new_name;3716 const char *name = old_name ? old_name : new_name;3717 struct cache_entry *ce = NULL;3718 struct patch *tpatch;3719 int ok_if_exists;3720 int status;37213722 patch->rejected = 1; /* we will drop this after we succeed */37233724 status = check_preimage(patch, &ce, &st);3725 if (status)3726 return status;3727 old_name = patch->old_name;37283729 /*3730 * A type-change diff is always split into a patch to delete3731 * old, immediately followed by a patch to create new (see3732 * diff.c::run_diff()); in such a case it is Ok that the entry3733 * to be deleted by the previous patch is still in the working3734 * tree and in the index.3735 *3736 * A patch to swap-rename between A and B would first rename A3737 * to B and then rename B to A. While applying the first one,3738 * the presence of B should not stop A from getting renamed to3739 * B; ask to_be_deleted() about the later rename. Removal of3740 * B and rename from A to B is handled the same way by asking3741 * was_deleted().3742 */3743 if ((tpatch = in_fn_table(new_name)) &&3744 (was_deleted(tpatch) || to_be_deleted(tpatch)))3745 ok_if_exists = 1;3746 else3747 ok_if_exists = 0;37483749 if (new_name &&3750 ((0 < patch->is_new) || patch->is_rename || patch->is_copy)) {3751 int err = check_to_create(new_name, ok_if_exists);37523753 if (err && threeway) {3754 patch->direct_to_threeway = 1;3755 } else switch (err) {3756 case 0:3757 break; /* happy */3758 case EXISTS_IN_INDEX:3759 return error(_("%s: already exists in index"), new_name);3760 break;3761 case EXISTS_IN_WORKTREE:3762 return error(_("%s: already exists in working directory"),3763 new_name);3764 default:3765 return err;3766 }37673768 if (!patch->new_mode) {3769 if (0 < patch->is_new)3770 patch->new_mode = S_IFREG | 0644;3771 else3772 patch->new_mode = patch->old_mode;3773 }3774 }37753776 if (new_name && old_name) {3777 int same = !strcmp(old_name, new_name);3778 if (!patch->new_mode)3779 patch->new_mode = patch->old_mode;3780 if ((patch->old_mode ^ patch->new_mode) & S_IFMT) {3781 if (same)3782 return error(_("new mode (%o) of %s does not "3783 "match old mode (%o)"),3784 patch->new_mode, new_name,3785 patch->old_mode);3786 else3787 return error(_("new mode (%o) of %s does not "3788 "match old mode (%o) of %s"),3789 patch->new_mode, new_name,3790 patch->old_mode, old_name);3791 }3792 }37933794 if (!unsafe_paths)3795 die_on_unsafe_path(patch);37963797 /*3798 * An attempt to read from or delete a path that is beyond a3799 * symbolic link will be prevented by load_patch_target() that3800 * is called at the beginning of apply_data() so we do not3801 * have to worry about a patch marked with "is_delete" bit3802 * here. We however need to make sure that the patch result3803 * is not deposited to a path that is beyond a symbolic link3804 * here.3805 */3806 if (!patch->is_delete && path_is_beyond_symlink(patch->new_name))3807 return error(_("affected file '%s' is beyond a symbolic link"),3808 patch->new_name);38093810 if (apply_data(patch, &st, ce) < 0)3811 return error(_("%s: patch does not apply"), name);3812 patch->rejected = 0;3813 return 0;3814}38153816static int check_patch_list(struct patch *patch)3817{3818 int err = 0;38193820 prepare_symlink_changes(patch);3821 prepare_fn_table(patch);3822 while (patch) {3823 if (apply_verbosely)3824 say_patch_name(stderr,3825 _("Checking patch %s..."), patch);3826 err |= check_patch(patch);3827 patch = patch->next;3828 }3829 return err;3830}38313832/* This function tries to read the sha1 from the current index */3833static int get_current_sha1(const char *path, unsigned char *sha1)3834{3835 int pos;38363837 if (read_cache() < 0)3838 return -1;3839 pos = cache_name_pos(path, strlen(path));3840 if (pos < 0)3841 return -1;3842 hashcpy(sha1, active_cache[pos]->sha1);3843 return 0;3844}38453846static int preimage_sha1_in_gitlink_patch(struct patch *p, unsigned char sha1[20])3847{3848 /*3849 * A usable gitlink patch has only one fragment (hunk) that looks like:3850 * @@ -1 +1 @@3851 * -Subproject commit <old sha1>3852 * +Subproject commit <new sha1>3853 * or3854 * @@ -1 +0,0 @@3855 * -Subproject commit <old sha1>3856 * for a removal patch.3857 */3858 struct fragment *hunk = p->fragments;3859 static const char heading[] = "-Subproject commit ";3860 char *preimage;38613862 if (/* does the patch have only one hunk? */3863 hunk && !hunk->next &&3864 /* is its preimage one line? */3865 hunk->oldpos == 1 && hunk->oldlines == 1 &&3866 /* does preimage begin with the heading? */3867 (preimage = memchr(hunk->patch, '\n', hunk->size)) != NULL &&3868 starts_with(++preimage, heading) &&3869 /* does it record full SHA-1? */3870 !get_sha1_hex(preimage + sizeof(heading) - 1, sha1) &&3871 preimage[sizeof(heading) + 40 - 1] == '\n' &&3872 /* does the abbreviated name on the index line agree with it? */3873 starts_with(preimage + sizeof(heading) - 1, p->old_sha1_prefix))3874 return 0; /* it all looks fine */38753876 /* we may have full object name on the index line */3877 return get_sha1_hex(p->old_sha1_prefix, sha1);3878}38793880/* Build an index that contains the just the files needed for a 3way merge */3881static void build_fake_ancestor(struct patch *list, const char *filename)3882{3883 struct patch *patch;3884 struct index_state result = { NULL };3885 static struct lock_file lock;38863887 /* Once we start supporting the reverse patch, it may be3888 * worth showing the new sha1 prefix, but until then...3889 */3890 for (patch = list; patch; patch = patch->next) {3891 unsigned char sha1[20];3892 struct cache_entry *ce;3893 const char *name;38943895 name = patch->old_name ? patch->old_name : patch->new_name;3896 if (0 < patch->is_new)3897 continue;38983899 if (S_ISGITLINK(patch->old_mode)) {3900 if (!preimage_sha1_in_gitlink_patch(patch, sha1))3901 ; /* ok, the textual part looks sane */3902 else3903 die("sha1 information is lacking or useless for submodule %s",3904 name);3905 } else if (!get_sha1_blob(patch->old_sha1_prefix, sha1)) {3906 ; /* ok */3907 } else if (!patch->lines_added && !patch->lines_deleted) {3908 /* mode-only change: update the current */3909 if (get_current_sha1(patch->old_name, sha1))3910 die("mode change for %s, which is not "3911 "in current HEAD", name);3912 } else3913 die("sha1 information is lacking or useless "3914 "(%s).", name);39153916 ce = make_cache_entry(patch->old_mode, sha1, name, 0, 0);3917 if (!ce)3918 die(_("make_cache_entry failed for path '%s'"), name);3919 if (add_index_entry(&result, ce, ADD_CACHE_OK_TO_ADD))3920 die ("Could not add %s to temporary index", name);3921 }39223923 hold_lock_file_for_update(&lock, filename, LOCK_DIE_ON_ERROR);3924 if (write_locked_index(&result, &lock, COMMIT_LOCK))3925 die ("Could not write temporary index to %s", filename);39263927 discard_index(&result);3928}39293930static void stat_patch_list(struct patch *patch)3931{3932 int files, adds, dels;39333934 for (files = adds = dels = 0 ; patch ; patch = patch->next) {3935 files++;3936 adds += patch->lines_added;3937 dels += patch->lines_deleted;3938 show_stats(patch);3939 }39403941 print_stat_summary(stdout, files, adds, dels);3942}39433944static void numstat_patch_list(struct patch *patch)3945{3946 for ( ; patch; patch = patch->next) {3947 const char *name;3948 name = patch->new_name ? patch->new_name : patch->old_name;3949 if (patch->is_binary)3950 printf("-\t-\t");3951 else3952 printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);3953 write_name_quoted(name, stdout, line_termination);3954 }3955}39563957static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)3958{3959 if (mode)3960 printf(" %s mode %06o %s\n", newdelete, mode, name);3961 else3962 printf(" %s %s\n", newdelete, name);3963}39643965static void show_mode_change(struct patch *p, int show_name)3966{3967 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {3968 if (show_name)3969 printf(" mode change %06o => %06o %s\n",3970 p->old_mode, p->new_mode, p->new_name);3971 else3972 printf(" mode change %06o => %06o\n",3973 p->old_mode, p->new_mode);3974 }3975}39763977static void show_rename_copy(struct patch *p)3978{3979 const char *renamecopy = p->is_rename ? "rename" : "copy";3980 const char *old, *new;39813982 /* Find common prefix */3983 old = p->old_name;3984 new = p->new_name;3985 while (1) {3986 const char *slash_old, *slash_new;3987 slash_old = strchr(old, '/');3988 slash_new = strchr(new, '/');3989 if (!slash_old ||3990 !slash_new ||3991 slash_old - old != slash_new - new ||3992 memcmp(old, new, slash_new - new))3993 break;3994 old = slash_old + 1;3995 new = slash_new + 1;3996 }3997 /* p->old_name thru old is the common prefix, and old and new3998 * through the end of names are renames3999 */4000 if (old != p->old_name)4001 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,4002 (int)(old - p->old_name), p->old_name,4003 old, new, p->score);4004 else4005 printf(" %s %s => %s (%d%%)\n", renamecopy,4006 p->old_name, p->new_name, p->score);4007 show_mode_change(p, 0);4008}40094010static void summary_patch_list(struct patch *patch)4011{4012 struct patch *p;40134014 for (p = patch; p; p = p->next) {4015 if (p->is_new)4016 show_file_mode_name("create", p->new_mode, p->new_name);4017 else if (p->is_delete)4018 show_file_mode_name("delete", p->old_mode, p->old_name);4019 else {4020 if (p->is_rename || p->is_copy)4021 show_rename_copy(p);4022 else {4023 if (p->score) {4024 printf(" rewrite %s (%d%%)\n",4025 p->new_name, p->score);4026 show_mode_change(p, 0);4027 }4028 else4029 show_mode_change(p, 1);4030 }4031 }4032 }4033}40344035static void patch_stats(struct patch *patch)4036{4037 int lines = patch->lines_added + patch->lines_deleted;40384039 if (lines > max_change)4040 max_change = lines;4041 if (patch->old_name) {4042 int len = quote_c_style(patch->old_name, NULL, NULL, 0);4043 if (!len)4044 len = strlen(patch->old_name);4045 if (len > max_len)4046 max_len = len;4047 }4048 if (patch->new_name) {4049 int len = quote_c_style(patch->new_name, NULL, NULL, 0);4050 if (!len)4051 len = strlen(patch->new_name);4052 if (len > max_len)4053 max_len = len;4054 }4055}40564057static void remove_file(struct patch *patch, int rmdir_empty)4058{4059 if (update_index) {4060 if (remove_file_from_cache(patch->old_name) < 0)4061 die(_("unable to remove %s from index"), patch->old_name);4062 }4063 if (!cached) {4064 if (!remove_or_warn(patch->old_mode, patch->old_name) && rmdir_empty) {4065 remove_path(patch->old_name);4066 }4067 }4068}40694070static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)4071{4072 struct stat st;4073 struct cache_entry *ce;4074 int namelen = strlen(path);4075 unsigned ce_size = cache_entry_size(namelen);40764077 if (!update_index)4078 return;40794080 ce = xcalloc(1, ce_size);4081 memcpy(ce->name, path, namelen);4082 ce->ce_mode = create_ce_mode(mode);4083 ce->ce_flags = create_ce_flags(0);4084 ce->ce_namelen = namelen;4085 if (S_ISGITLINK(mode)) {4086 const char *s;40874088 if (!skip_prefix(buf, "Subproject commit ", &s) ||4089 get_sha1_hex(s, ce->sha1))4090 die(_("corrupt patch for submodule %s"), path);4091 } else {4092 if (!cached) {4093 if (lstat(path, &st) < 0)4094 die_errno(_("unable to stat newly created file '%s'"),4095 path);4096 fill_stat_cache_info(ce, &st);4097 }4098 if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)4099 die(_("unable to create backing store for newly created file %s"), path);4100 }4101 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)4102 die(_("unable to add cache entry for %s"), path);4103}41044105static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)4106{4107 int fd;4108 struct strbuf nbuf = STRBUF_INIT;41094110 if (S_ISGITLINK(mode)) {4111 struct stat st;4112 if (!lstat(path, &st) && S_ISDIR(st.st_mode))4113 return 0;4114 return mkdir(path, 0777);4115 }41164117 if (has_symlinks && S_ISLNK(mode))4118 /* Although buf:size is counted string, it also is NUL4119 * terminated.4120 */4121 return symlink(buf, path);41224123 fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);4124 if (fd < 0)4125 return -1;41264127 if (convert_to_working_tree(path, buf, size, &nbuf)) {4128 size = nbuf.len;4129 buf = nbuf.buf;4130 }4131 write_or_die(fd, buf, size);4132 strbuf_release(&nbuf);41334134 if (close(fd) < 0)4135 die_errno(_("closing file '%s'"), path);4136 return 0;4137}41384139/*4140 * We optimistically assume that the directories exist,4141 * which is true 99% of the time anyway. If they don't,4142 * we create them and try again.4143 */4144static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)4145{4146 if (cached)4147 return;4148 if (!try_create_file(path, mode, buf, size))4149 return;41504151 if (errno == ENOENT) {4152 if (safe_create_leading_directories(path))4153 return;4154 if (!try_create_file(path, mode, buf, size))4155 return;4156 }41574158 if (errno == EEXIST || errno == EACCES) {4159 /* We may be trying to create a file where a directory4160 * used to be.4161 */4162 struct stat st;4163 if (!lstat(path, &st) && (!S_ISDIR(st.st_mode) || !rmdir(path)))4164 errno = EEXIST;4165 }41664167 if (errno == EEXIST) {4168 unsigned int nr = getpid();41694170 for (;;) {4171 char newpath[PATH_MAX];4172 mksnpath(newpath, sizeof(newpath), "%s~%u", path, nr);4173 if (!try_create_file(newpath, mode, buf, size)) {4174 if (!rename(newpath, path))4175 return;4176 unlink_or_warn(newpath);4177 break;4178 }4179 if (errno != EEXIST)4180 break;4181 ++nr;4182 }4183 }4184 die_errno(_("unable to write file '%s' mode %o"), path, mode);4185}41864187static void add_conflicted_stages_file(struct patch *patch)4188{4189 int stage, namelen;4190 unsigned ce_size, mode;4191 struct cache_entry *ce;41924193 if (!update_index)4194 return;4195 namelen = strlen(patch->new_name);4196 ce_size = cache_entry_size(namelen);4197 mode = patch->new_mode ? patch->new_mode : (S_IFREG | 0644);41984199 remove_file_from_cache(patch->new_name);4200 for (stage = 1; stage < 4; stage++) {4201 if (is_null_oid(&patch->threeway_stage[stage - 1]))4202 continue;4203 ce = xcalloc(1, ce_size);4204 memcpy(ce->name, patch->new_name, namelen);4205 ce->ce_mode = create_ce_mode(mode);4206 ce->ce_flags = create_ce_flags(stage);4207 ce->ce_namelen = namelen;4208 hashcpy(ce->sha1, patch->threeway_stage[stage - 1].hash);4209 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)4210 die(_("unable to add cache entry for %s"), patch->new_name);4211 }4212}42134214static void create_file(struct patch *patch)4215{4216 char *path = patch->new_name;4217 unsigned mode = patch->new_mode;4218 unsigned long size = patch->resultsize;4219 char *buf = patch->result;42204221 if (!mode)4222 mode = S_IFREG | 0644;4223 create_one_file(path, mode, buf, size);42244225 if (patch->conflicted_threeway)4226 add_conflicted_stages_file(patch);4227 else4228 add_index_file(path, mode, buf, size);4229}42304231/* phase zero is to remove, phase one is to create */4232static void write_out_one_result(struct patch *patch, int phase)4233{4234 if (patch->is_delete > 0) {4235 if (phase == 0)4236 remove_file(patch, 1);4237 return;4238 }4239 if (patch->is_new > 0 || patch->is_copy) {4240 if (phase == 1)4241 create_file(patch);4242 return;4243 }4244 /*4245 * Rename or modification boils down to the same4246 * thing: remove the old, write the new4247 */4248 if (phase == 0)4249 remove_file(patch, patch->is_rename);4250 if (phase == 1)4251 create_file(patch);4252}42534254static int write_out_one_reject(struct patch *patch)4255{4256 FILE *rej;4257 char namebuf[PATH_MAX];4258 struct fragment *frag;4259 int cnt = 0;4260 struct strbuf sb = STRBUF_INIT;42614262 for (cnt = 0, frag = patch->fragments; frag; frag = frag->next) {4263 if (!frag->rejected)4264 continue;4265 cnt++;4266 }42674268 if (!cnt) {4269 if (apply_verbosely)4270 say_patch_name(stderr,4271 _("Applied patch %s cleanly."), patch);4272 return 0;4273 }42744275 /* This should not happen, because a removal patch that leaves4276 * contents are marked "rejected" at the patch level.4277 */4278 if (!patch->new_name)4279 die(_("internal error"));42804281 /* Say this even without --verbose */4282 strbuf_addf(&sb, Q_("Applying patch %%s with %d reject...",4283 "Applying patch %%s with %d rejects...",4284 cnt),4285 cnt);4286 say_patch_name(stderr, sb.buf, patch);4287 strbuf_release(&sb);42884289 cnt = strlen(patch->new_name);4290 if (ARRAY_SIZE(namebuf) <= cnt + 5) {4291 cnt = ARRAY_SIZE(namebuf) - 5;4292 warning(_("truncating .rej filename to %.*s.rej"),4293 cnt - 1, patch->new_name);4294 }4295 memcpy(namebuf, patch->new_name, cnt);4296 memcpy(namebuf + cnt, ".rej", 5);42974298 rej = fopen(namebuf, "w");4299 if (!rej)4300 return error(_("cannot open %s: %s"), namebuf, strerror(errno));43014302 /* Normal git tools never deal with .rej, so do not pretend4303 * this is a git patch by saying --git or giving extended4304 * headers. While at it, maybe please "kompare" that wants4305 * the trailing TAB and some garbage at the end of line ;-).4306 */4307 fprintf(rej, "diff a/%s b/%s\t(rejected hunks)\n",4308 patch->new_name, patch->new_name);4309 for (cnt = 1, frag = patch->fragments;4310 frag;4311 cnt++, frag = frag->next) {4312 if (!frag->rejected) {4313 fprintf_ln(stderr, _("Hunk #%d applied cleanly."), cnt);4314 continue;4315 }4316 fprintf_ln(stderr, _("Rejected hunk #%d."), cnt);4317 fprintf(rej, "%.*s", frag->size, frag->patch);4318 if (frag->patch[frag->size-1] != '\n')4319 fputc('\n', rej);4320 }4321 fclose(rej);4322 return -1;4323}43244325static int write_out_results(struct patch *list)4326{4327 int phase;4328 int errs = 0;4329 struct patch *l;4330 struct string_list cpath = STRING_LIST_INIT_DUP;43314332 for (phase = 0; phase < 2; phase++) {4333 l = list;4334 while (l) {4335 if (l->rejected)4336 errs = 1;4337 else {4338 write_out_one_result(l, phase);4339 if (phase == 1) {4340 if (write_out_one_reject(l))4341 errs = 1;4342 if (l->conflicted_threeway) {4343 string_list_append(&cpath, l->new_name);4344 errs = 1;4345 }4346 }4347 }4348 l = l->next;4349 }4350 }43514352 if (cpath.nr) {4353 struct string_list_item *item;43544355 string_list_sort(&cpath);4356 for_each_string_list_item(item, &cpath)4357 fprintf(stderr, "U %s\n", item->string);4358 string_list_clear(&cpath, 0);43594360 rerere(0);4361 }43624363 return errs;4364}43654366static struct lock_file lock_file;43674368#define INACCURATE_EOF (1<<0)4369#define RECOUNT (1<<1)43704371static int apply_patch(int fd, const char *filename, int options)4372{4373 size_t offset;4374 struct strbuf buf = STRBUF_INIT; /* owns the patch text */4375 struct patch *list = NULL, **listp = &list;4376 int skipped_patch = 0;43774378 patch_input_file = filename;4379 read_patch_file(&buf, fd);4380 offset = 0;4381 while (offset < buf.len) {4382 struct patch *patch;4383 int nr;43844385 patch = xcalloc(1, sizeof(*patch));4386 patch->inaccurate_eof = !!(options & INACCURATE_EOF);4387 patch->recount = !!(options & RECOUNT);4388 nr = parse_chunk(buf.buf + offset, buf.len - offset, patch);4389 if (nr < 0) {4390 free_patch(patch);4391 break;4392 }4393 if (apply_in_reverse)4394 reverse_patches(patch);4395 if (use_patch(patch)) {4396 patch_stats(patch);4397 *listp = patch;4398 listp = &patch->next;4399 }4400 else {4401 if (apply_verbosely)4402 say_patch_name(stderr, _("Skipped patch '%s'."), patch);4403 free_patch(patch);4404 skipped_patch++;4405 }4406 offset += nr;4407 }44084409 if (!list && !skipped_patch)4410 die(_("unrecognized input"));44114412 if (whitespace_error && (ws_error_action == die_on_ws_error))4413 apply = 0;44144415 update_index = check_index && apply;4416 if (update_index && newfd < 0)4417 newfd = hold_locked_index(&lock_file, 1);44184419 if (check_index) {4420 if (read_cache() < 0)4421 die(_("unable to read index file"));4422 }44234424 if ((check || apply) &&4425 check_patch_list(list) < 0 &&4426 !apply_with_reject)4427 exit(1);44284429 if (apply && write_out_results(list)) {4430 if (apply_with_reject)4431 exit(1);4432 /* with --3way, we still need to write the index out */4433 return 1;4434 }44354436 if (fake_ancestor)4437 build_fake_ancestor(list, fake_ancestor);44384439 if (diffstat)4440 stat_patch_list(list);44414442 if (numstat)4443 numstat_patch_list(list);44444445 if (summary)4446 summary_patch_list(list);44474448 free_patch_list(list);4449 strbuf_release(&buf);4450 string_list_clear(&fn_table, 0);4451 return 0;4452}44534454static void git_apply_config(void)4455{4456 git_config_get_string_const("apply.whitespace", &apply_default_whitespace);4457 git_config_get_string_const("apply.ignorewhitespace", &apply_default_ignorewhitespace);4458 git_config(git_default_config, NULL);4459}44604461static int option_parse_exclude(const struct option *opt,4462 const char *arg, int unset)4463{4464 add_name_limit(arg, 1);4465 return 0;4466}44674468static int option_parse_include(const struct option *opt,4469 const char *arg, int unset)4470{4471 add_name_limit(arg, 0);4472 has_include = 1;4473 return 0;4474}44754476static int option_parse_p(const struct option *opt,4477 const char *arg, int unset)4478{4479 state_p_value = atoi(arg);4480 p_value_known = 1;4481 return 0;4482}44834484static int option_parse_space_change(const struct option *opt,4485 const char *arg, int unset)4486{4487 if (unset)4488 ws_ignore_action = ignore_ws_none;4489 else4490 ws_ignore_action = ignore_ws_change;4491 return 0;4492}44934494static int option_parse_whitespace(const struct option *opt,4495 const char *arg, int unset)4496{4497 const char **whitespace_option = opt->value;44984499 *whitespace_option = arg;4500 parse_whitespace_option(arg);4501 return 0;4502}45034504static int option_parse_directory(const struct option *opt,4505 const char *arg, int unset)4506{4507 strbuf_reset(&root);4508 strbuf_addstr(&root, arg);4509 strbuf_complete(&root, '/');4510 return 0;4511}45124513int cmd_apply(int argc, const char **argv, const char *prefix_)4514{4515 int i;4516 int errs = 0;4517 int is_not_gitdir = !startup_info->have_repository;4518 int force_apply = 0;4519 int options = 0;45204521 const char *whitespace_option = NULL;45224523 struct option builtin_apply_options[] = {4524 { OPTION_CALLBACK, 0, "exclude", NULL, N_("path"),4525 N_("don't apply changes matching the given path"),4526 0, option_parse_exclude },4527 { OPTION_CALLBACK, 0, "include", NULL, N_("path"),4528 N_("apply changes matching the given path"),4529 0, option_parse_include },4530 { OPTION_CALLBACK, 'p', NULL, NULL, N_("num"),4531 N_("remove <num> leading slashes from traditional diff paths"),4532 0, option_parse_p },4533 OPT_BOOL(0, "no-add", &no_add,4534 N_("ignore additions made by the patch")),4535 OPT_BOOL(0, "stat", &diffstat,4536 N_("instead of applying the patch, output diffstat for the input")),4537 OPT_NOOP_NOARG(0, "allow-binary-replacement"),4538 OPT_NOOP_NOARG(0, "binary"),4539 OPT_BOOL(0, "numstat", &numstat,4540 N_("show number of added and deleted lines in decimal notation")),4541 OPT_BOOL(0, "summary", &summary,4542 N_("instead of applying the patch, output a summary for the input")),4543 OPT_BOOL(0, "check", &check,4544 N_("instead of applying the patch, see if the patch is applicable")),4545 OPT_BOOL(0, "index", &check_index,4546 N_("make sure the patch is applicable to the current index")),4547 OPT_BOOL(0, "cached", &cached,4548 N_("apply a patch without touching the working tree")),4549 OPT_BOOL(0, "unsafe-paths", &unsafe_paths,4550 N_("accept a patch that touches outside the working area")),4551 OPT_BOOL(0, "apply", &force_apply,4552 N_("also apply the patch (use with --stat/--summary/--check)")),4553 OPT_BOOL('3', "3way", &threeway,4554 N_( "attempt three-way merge if a patch does not apply")),4555 OPT_FILENAME(0, "build-fake-ancestor", &fake_ancestor,4556 N_("build a temporary index based on embedded index information")),4557 /* Think twice before adding "--nul" synonym to this */4558 OPT_SET_INT('z', NULL, &line_termination,4559 N_("paths are separated with NUL character"), '\0'),4560 OPT_INTEGER('C', NULL, &p_context,4561 N_("ensure at least <n> lines of context match")),4562 { OPTION_CALLBACK, 0, "whitespace", &whitespace_option, N_("action"),4563 N_("detect new or modified lines that have whitespace errors"),4564 0, option_parse_whitespace },4565 { OPTION_CALLBACK, 0, "ignore-space-change", NULL, NULL,4566 N_("ignore changes in whitespace when finding context"),4567 PARSE_OPT_NOARG, option_parse_space_change },4568 { OPTION_CALLBACK, 0, "ignore-whitespace", NULL, NULL,4569 N_("ignore changes in whitespace when finding context"),4570 PARSE_OPT_NOARG, option_parse_space_change },4571 OPT_BOOL('R', "reverse", &apply_in_reverse,4572 N_("apply the patch in reverse")),4573 OPT_BOOL(0, "unidiff-zero", &unidiff_zero,4574 N_("don't expect at least one line of context")),4575 OPT_BOOL(0, "reject", &apply_with_reject,4576 N_("leave the rejected hunks in corresponding *.rej files")),4577 OPT_BOOL(0, "allow-overlap", &allow_overlap,4578 N_("allow overlapping hunks")),4579 OPT__VERBOSE(&apply_verbosely, N_("be verbose")),4580 OPT_BIT(0, "inaccurate-eof", &options,4581 N_("tolerate incorrectly detected missing new-line at the end of file"),4582 INACCURATE_EOF),4583 OPT_BIT(0, "recount", &options,4584 N_("do not trust the line counts in the hunk headers"),4585 RECOUNT),4586 { OPTION_CALLBACK, 0, "directory", NULL, N_("root"),4587 N_("prepend <root> to all filenames"),4588 0, option_parse_directory },4589 OPT_END()4590 };45914592 prefix = prefix_;4593 prefix_length = prefix ? strlen(prefix) : 0;4594 git_apply_config();4595 if (apply_default_whitespace)4596 parse_whitespace_option(apply_default_whitespace);4597 if (apply_default_ignorewhitespace)4598 parse_ignorewhitespace_option(apply_default_ignorewhitespace);45994600 argc = parse_options(argc, argv, prefix, builtin_apply_options,4601 apply_usage, 0);46024603 if (apply_with_reject && threeway)4604 die("--reject and --3way cannot be used together.");4605 if (cached && threeway)4606 die("--cached and --3way cannot be used together.");4607 if (threeway) {4608 if (is_not_gitdir)4609 die(_("--3way outside a repository"));4610 check_index = 1;4611 }4612 if (apply_with_reject)4613 apply = apply_verbosely = 1;4614 if (!force_apply && (diffstat || numstat || summary || check || fake_ancestor))4615 apply = 0;4616 if (check_index && is_not_gitdir)4617 die(_("--index outside a repository"));4618 if (cached) {4619 if (is_not_gitdir)4620 die(_("--cached outside a repository"));4621 check_index = 1;4622 }4623 if (check_index)4624 unsafe_paths = 0;46254626 for (i = 0; i < argc; i++) {4627 const char *arg = argv[i];4628 int fd;46294630 if (!strcmp(arg, "-")) {4631 errs |= apply_patch(0, "<stdin>", options);4632 read_stdin = 0;4633 continue;4634 } else if (0 < prefix_length)4635 arg = prefix_filename(prefix, prefix_length, arg);46364637 fd = open(arg, O_RDONLY);4638 if (fd < 0)4639 die_errno(_("can't open patch '%s'"), arg);4640 read_stdin = 0;4641 set_default_whitespace_mode(whitespace_option);4642 errs |= apply_patch(fd, arg, options);4643 close(fd);4644 }4645 set_default_whitespace_mode(whitespace_option);4646 if (read_stdin)4647 errs |= apply_patch(0, "<stdin>", options);4648 if (whitespace_error) {4649 if (squelch_whitespace_errors &&4650 squelch_whitespace_errors < whitespace_error) {4651 int squelched =4652 whitespace_error - squelch_whitespace_errors;4653 warning(Q_("squelched %d whitespace error",4654 "squelched %d whitespace errors",4655 squelched),4656 squelched);4657 }4658 if (ws_error_action == die_on_ws_error)4659 die(Q_("%d line adds whitespace errors.",4660 "%d lines add whitespace errors.",4661 whitespace_error),4662 whitespace_error);4663 if (applied_after_fixing_ws && apply)4664 warning("%d line%s applied after"4665 " fixing whitespace errors.",4666 applied_after_fixing_ws,4667 applied_after_fixing_ws == 1 ? "" : "s");4668 else if (whitespace_error)4669 warning(Q_("%d line adds whitespace errors.",4670 "%d lines add whitespace errors.",4671 whitespace_error),4672 whitespace_error);4673 }46744675 if (update_index) {4676 if (write_locked_index(&the_index, &lock_file, COMMIT_LOCK))4677 die(_("Unable to write new index file"));4678 }46794680 return !!errs;4681}