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 "cache-tree.h" 11#include "quote.h" 12#include "blob.h" 13#include "delta.h" 14#include "builtin.h" 15#include "string-list.h" 16#include "dir.h" 17#include "diff.h" 18#include "parse-options.h" 19 20/* 21 * --check turns on checking that the working tree matches the 22 * files that are being modified, but doesn't apply the patch 23 * --stat does just a diffstat, and doesn't actually apply 24 * --numstat does numeric diffstat, and doesn't actually apply 25 * --index-info shows the old and new index info for paths if available. 26 * --index updates the cache as well. 27 * --cached updates only the cache without ever touching the working tree. 28 */ 29static const char *prefix; 30static int prefix_length = -1; 31static int newfd = -1; 32 33static int unidiff_zero; 34static int p_value = 1; 35static int p_value_known; 36static int check_index; 37static int update_index; 38static int cached; 39static int diffstat; 40static int numstat; 41static int summary; 42static int check; 43static int apply = 1; 44static int apply_in_reverse; 45static int apply_with_reject; 46static int apply_verbosely; 47static int allow_overlap; 48static int no_add; 49static const char *fake_ancestor; 50static int line_termination = '\n'; 51static unsigned int p_context = UINT_MAX; 52static const char * const apply_usage[] = { 53 "git apply [options] [<patch>...]", 54 NULL 55}; 56 57static enum ws_error_action { 58 nowarn_ws_error, 59 warn_on_ws_error, 60 die_on_ws_error, 61 correct_ws_error 62} ws_error_action = warn_on_ws_error; 63static int whitespace_error; 64static int squelch_whitespace_errors = 5; 65static int applied_after_fixing_ws; 66 67static enum ws_ignore { 68 ignore_ws_none, 69 ignore_ws_change 70} ws_ignore_action = ignore_ws_none; 71 72 73static const char *patch_input_file; 74static const char *root; 75static int root_len; 76static int read_stdin = 1; 77static int options; 78 79static void parse_whitespace_option(const char *option) 80{ 81 if (!option) { 82 ws_error_action = warn_on_ws_error; 83 return; 84 } 85 if (!strcmp(option, "warn")) { 86 ws_error_action = warn_on_ws_error; 87 return; 88 } 89 if (!strcmp(option, "nowarn")) { 90 ws_error_action = nowarn_ws_error; 91 return; 92 } 93 if (!strcmp(option, "error")) { 94 ws_error_action = die_on_ws_error; 95 return; 96 } 97 if (!strcmp(option, "error-all")) { 98 ws_error_action = die_on_ws_error; 99 squelch_whitespace_errors = 0; 100 return; 101 } 102 if (!strcmp(option, "strip") || !strcmp(option, "fix")) { 103 ws_error_action = correct_ws_error; 104 return; 105 } 106 die(_("unrecognized whitespace option '%s'"), option); 107} 108 109static void parse_ignorewhitespace_option(const char *option) 110{ 111 if (!option || !strcmp(option, "no") || 112 !strcmp(option, "false") || !strcmp(option, "never") || 113 !strcmp(option, "none")) { 114 ws_ignore_action = ignore_ws_none; 115 return; 116 } 117 if (!strcmp(option, "change")) { 118 ws_ignore_action = ignore_ws_change; 119 return; 120 } 121 die(_("unrecognized whitespace ignore option '%s'"), option); 122} 123 124static void set_default_whitespace_mode(const char *whitespace_option) 125{ 126 if (!whitespace_option && !apply_default_whitespace) 127 ws_error_action = (apply ? warn_on_ws_error : nowarn_ws_error); 128} 129 130/* 131 * For "diff-stat" like behaviour, we keep track of the biggest change 132 * we've seen, and the longest filename. That allows us to do simple 133 * scaling. 134 */ 135static int max_change, max_len; 136 137/* 138 * Various "current state", notably line numbers and what 139 * file (and how) we're patching right now.. The "is_xxxx" 140 * things are flags, where -1 means "don't know yet". 141 */ 142static int linenr = 1; 143 144/* 145 * This represents one "hunk" from a patch, starting with 146 * "@@ -oldpos,oldlines +newpos,newlines @@" marker. The 147 * patch text is pointed at by patch, and its byte length 148 * is stored in size. leading and trailing are the number 149 * of context lines. 150 */ 151struct fragment { 152 unsigned long leading, trailing; 153 unsigned long oldpos, oldlines; 154 unsigned long newpos, newlines; 155 const char *patch; 156 int size; 157 int rejected; 158 int linenr; 159 struct fragment *next; 160}; 161 162/* 163 * When dealing with a binary patch, we reuse "leading" field 164 * to store the type of the binary hunk, either deflated "delta" 165 * or deflated "literal". 166 */ 167#define binary_patch_method leading 168#define BINARY_DELTA_DEFLATED 1 169#define BINARY_LITERAL_DEFLATED 2 170 171/* 172 * This represents a "patch" to a file, both metainfo changes 173 * such as creation/deletion, filemode and content changes represented 174 * as a series of fragments. 175 */ 176struct patch { 177 char *new_name, *old_name, *def_name; 178 unsigned int old_mode, new_mode; 179 int is_new, is_delete; /* -1 = unknown, 0 = false, 1 = true */ 180 int rejected; 181 unsigned ws_rule; 182 unsigned long deflate_origlen; 183 int lines_added, lines_deleted; 184 int score; 185 unsigned int is_toplevel_relative:1; 186 unsigned int inaccurate_eof:1; 187 unsigned int is_binary:1; 188 unsigned int is_copy:1; 189 unsigned int is_rename:1; 190 unsigned int recount:1; 191 struct fragment *fragments; 192 char *result; 193 size_t resultsize; 194 char old_sha1_prefix[41]; 195 char new_sha1_prefix[41]; 196 struct patch *next; 197}; 198 199/* 200 * A line in a file, len-bytes long (includes the terminating LF, 201 * except for an incomplete line at the end if the file ends with 202 * one), and its contents hashes to 'hash'. 203 */ 204struct line { 205 size_t len; 206 unsigned hash : 24; 207 unsigned flag : 8; 208#define LINE_COMMON 1 209#define LINE_PATCHED 2 210}; 211 212/* 213 * This represents a "file", which is an array of "lines". 214 */ 215struct image { 216 char *buf; 217 size_t len; 218 size_t nr; 219 size_t alloc; 220 struct line *line_allocated; 221 struct line *line; 222}; 223 224/* 225 * Records filenames that have been touched, in order to handle 226 * the case where more than one patches touch the same file. 227 */ 228 229static struct string_list fn_table; 230 231static uint32_t hash_line(const char *cp, size_t len) 232{ 233 size_t i; 234 uint32_t h; 235 for (i = 0, h = 0; i < len; i++) { 236 if (!isspace(cp[i])) { 237 h = h * 3 + (cp[i] & 0xff); 238 } 239 } 240 return h; 241} 242 243/* 244 * Compare lines s1 of length n1 and s2 of length n2, ignoring 245 * whitespace difference. Returns 1 if they match, 0 otherwise 246 */ 247static int fuzzy_matchlines(const char *s1, size_t n1, 248 const char *s2, size_t n2) 249{ 250 const char *last1 = s1 + n1 - 1; 251 const char *last2 = s2 + n2 - 1; 252 int result = 0; 253 254 /* ignore line endings */ 255 while ((*last1 == '\r') || (*last1 == '\n')) 256 last1--; 257 while ((*last2 == '\r') || (*last2 == '\n')) 258 last2--; 259 260 /* skip leading whitespace */ 261 while (isspace(*s1) && (s1 <= last1)) 262 s1++; 263 while (isspace(*s2) && (s2 <= last2)) 264 s2++; 265 /* early return if both lines are empty */ 266 if ((s1 > last1) && (s2 > last2)) 267 return 1; 268 while (!result) { 269 result = *s1++ - *s2++; 270 /* 271 * Skip whitespace inside. We check for whitespace on 272 * both buffers because we don't want "a b" to match 273 * "ab" 274 */ 275 if (isspace(*s1) && isspace(*s2)) { 276 while (isspace(*s1) && s1 <= last1) 277 s1++; 278 while (isspace(*s2) && s2 <= last2) 279 s2++; 280 } 281 /* 282 * If we reached the end on one side only, 283 * lines don't match 284 */ 285 if ( 286 ((s2 > last2) && (s1 <= last1)) || 287 ((s1 > last1) && (s2 <= last2))) 288 return 0; 289 if ((s1 > last1) && (s2 > last2)) 290 break; 291 } 292 293 return !result; 294} 295 296static void add_line_info(struct image *img, const char *bol, size_t len, unsigned flag) 297{ 298 ALLOC_GROW(img->line_allocated, img->nr + 1, img->alloc); 299 img->line_allocated[img->nr].len = len; 300 img->line_allocated[img->nr].hash = hash_line(bol, len); 301 img->line_allocated[img->nr].flag = flag; 302 img->nr++; 303} 304 305static void prepare_image(struct image *image, char *buf, size_t len, 306 int prepare_linetable) 307{ 308 const char *cp, *ep; 309 310 memset(image, 0, sizeof(*image)); 311 image->buf = buf; 312 image->len = len; 313 314 if (!prepare_linetable) 315 return; 316 317 ep = image->buf + image->len; 318 cp = image->buf; 319 while (cp < ep) { 320 const char *next; 321 for (next = cp; next < ep && *next != '\n'; next++) 322 ; 323 if (next < ep) 324 next++; 325 add_line_info(image, cp, next - cp, 0); 326 cp = next; 327 } 328 image->line = image->line_allocated; 329} 330 331static void clear_image(struct image *image) 332{ 333 free(image->buf); 334 image->buf = NULL; 335 image->len = 0; 336} 337 338static void say_patch_name(FILE *output, const char *pre, 339 struct patch *patch, const char *post) 340{ 341 fputs(pre, output); 342 if (patch->old_name && patch->new_name && 343 strcmp(patch->old_name, patch->new_name)) { 344 quote_c_style(patch->old_name, NULL, output, 0); 345 fputs(" => ", output); 346 quote_c_style(patch->new_name, NULL, output, 0); 347 } else { 348 const char *n = patch->new_name; 349 if (!n) 350 n = patch->old_name; 351 quote_c_style(n, NULL, output, 0); 352 } 353 fputs(post, output); 354} 355 356#define CHUNKSIZE (8192) 357#define SLOP (16) 358 359static void read_patch_file(struct strbuf *sb, int fd) 360{ 361 if (strbuf_read(sb, fd, 0) < 0) 362 die_errno("git apply: failed to read"); 363 364 /* 365 * Make sure that we have some slop in the buffer 366 * so that we can do speculative "memcmp" etc, and 367 * see to it that it is NUL-filled. 368 */ 369 strbuf_grow(sb, SLOP); 370 memset(sb->buf + sb->len, 0, SLOP); 371} 372 373static unsigned long linelen(const char *buffer, unsigned long size) 374{ 375 unsigned long len = 0; 376 while (size--) { 377 len++; 378 if (*buffer++ == '\n') 379 break; 380 } 381 return len; 382} 383 384static int is_dev_null(const char *str) 385{ 386 return !memcmp("/dev/null", str, 9) && isspace(str[9]); 387} 388 389#define TERM_SPACE 1 390#define TERM_TAB 2 391 392static int name_terminate(const char *name, int namelen, int c, int terminate) 393{ 394 if (c == ' ' && !(terminate & TERM_SPACE)) 395 return 0; 396 if (c == '\t' && !(terminate & TERM_TAB)) 397 return 0; 398 399 return 1; 400} 401 402/* remove double slashes to make --index work with such filenames */ 403static char *squash_slash(char *name) 404{ 405 int i = 0, j = 0; 406 407 if (!name) 408 return NULL; 409 410 while (name[i]) { 411 if ((name[j++] = name[i++]) == '/') 412 while (name[i] == '/') 413 i++; 414 } 415 name[j] = '\0'; 416 return name; 417} 418 419static char *find_name_gnu(const char *line, char *def, int p_value) 420{ 421 struct strbuf name = STRBUF_INIT; 422 char *cp; 423 424 /* 425 * Proposed "new-style" GNU patch/diff format; see 426 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2 427 */ 428 if (unquote_c_style(&name, line, NULL)) { 429 strbuf_release(&name); 430 return NULL; 431 } 432 433 for (cp = name.buf; p_value; p_value--) { 434 cp = strchr(cp, '/'); 435 if (!cp) { 436 strbuf_release(&name); 437 return NULL; 438 } 439 cp++; 440 } 441 442 /* name can later be freed, so we need 443 * to memmove, not just return cp 444 */ 445 strbuf_remove(&name, 0, cp - name.buf); 446 free(def); 447 if (root) 448 strbuf_insert(&name, 0, root, root_len); 449 return squash_slash(strbuf_detach(&name, NULL)); 450} 451 452static size_t sane_tz_len(const char *line, size_t len) 453{ 454 const char *tz, *p; 455 456 if (len < strlen(" +0500") || line[len-strlen(" +0500")] != ' ') 457 return 0; 458 tz = line + len - strlen(" +0500"); 459 460 if (tz[1] != '+' && tz[1] != '-') 461 return 0; 462 463 for (p = tz + 2; p != line + len; p++) 464 if (!isdigit(*p)) 465 return 0; 466 467 return line + len - tz; 468} 469 470static size_t tz_with_colon_len(const char *line, size_t len) 471{ 472 const char *tz, *p; 473 474 if (len < strlen(" +08:00") || line[len - strlen(":00")] != ':') 475 return 0; 476 tz = line + len - strlen(" +08:00"); 477 478 if (tz[0] != ' ' || (tz[1] != '+' && tz[1] != '-')) 479 return 0; 480 p = tz + 2; 481 if (!isdigit(*p++) || !isdigit(*p++) || *p++ != ':' || 482 !isdigit(*p++) || !isdigit(*p++)) 483 return 0; 484 485 return line + len - tz; 486} 487 488static size_t date_len(const char *line, size_t len) 489{ 490 const char *date, *p; 491 492 if (len < strlen("72-02-05") || line[len-strlen("-05")] != '-') 493 return 0; 494 p = date = line + len - strlen("72-02-05"); 495 496 if (!isdigit(*p++) || !isdigit(*p++) || *p++ != '-' || 497 !isdigit(*p++) || !isdigit(*p++) || *p++ != '-' || 498 !isdigit(*p++) || !isdigit(*p++)) /* Not a date. */ 499 return 0; 500 501 if (date - line >= strlen("19") && 502 isdigit(date[-1]) && isdigit(date[-2])) /* 4-digit year */ 503 date -= strlen("19"); 504 505 return line + len - date; 506} 507 508static size_t short_time_len(const char *line, size_t len) 509{ 510 const char *time, *p; 511 512 if (len < strlen(" 07:01:32") || line[len-strlen(":32")] != ':') 513 return 0; 514 p = time = line + len - strlen(" 07:01:32"); 515 516 /* Permit 1-digit hours? */ 517 if (*p++ != ' ' || 518 !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' || 519 !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' || 520 !isdigit(*p++) || !isdigit(*p++)) /* Not a time. */ 521 return 0; 522 523 return line + len - time; 524} 525 526static size_t fractional_time_len(const char *line, size_t len) 527{ 528 const char *p; 529 size_t n; 530 531 /* Expected format: 19:41:17.620000023 */ 532 if (!len || !isdigit(line[len - 1])) 533 return 0; 534 p = line + len - 1; 535 536 /* Fractional seconds. */ 537 while (p > line && isdigit(*p)) 538 p--; 539 if (*p != '.') 540 return 0; 541 542 /* Hours, minutes, and whole seconds. */ 543 n = short_time_len(line, p - line); 544 if (!n) 545 return 0; 546 547 return line + len - p + n; 548} 549 550static size_t trailing_spaces_len(const char *line, size_t len) 551{ 552 const char *p; 553 554 /* Expected format: ' ' x (1 or more) */ 555 if (!len || line[len - 1] != ' ') 556 return 0; 557 558 p = line + len; 559 while (p != line) { 560 p--; 561 if (*p != ' ') 562 return line + len - (p + 1); 563 } 564 565 /* All spaces! */ 566 return len; 567} 568 569static size_t diff_timestamp_len(const char *line, size_t len) 570{ 571 const char *end = line + len; 572 size_t n; 573 574 /* 575 * Posix: 2010-07-05 19:41:17 576 * GNU: 2010-07-05 19:41:17.620000023 -0500 577 */ 578 579 if (!isdigit(end[-1])) 580 return 0; 581 582 n = sane_tz_len(line, end - line); 583 if (!n) 584 n = tz_with_colon_len(line, end - line); 585 end -= n; 586 587 n = short_time_len(line, end - line); 588 if (!n) 589 n = fractional_time_len(line, end - line); 590 end -= n; 591 592 n = date_len(line, end - line); 593 if (!n) /* No date. Too bad. */ 594 return 0; 595 end -= n; 596 597 if (end == line) /* No space before date. */ 598 return 0; 599 if (end[-1] == '\t') { /* Success! */ 600 end--; 601 return line + len - end; 602 } 603 if (end[-1] != ' ') /* No space before date. */ 604 return 0; 605 606 /* Whitespace damage. */ 607 end -= trailing_spaces_len(line, end - line); 608 return line + len - end; 609} 610 611static char *find_name_common(const char *line, char *def, int p_value, 612 const char *end, int terminate) 613{ 614 int len; 615 const char *start = NULL; 616 617 if (p_value == 0) 618 start = line; 619 while (line != end) { 620 char c = *line; 621 622 if (!end && isspace(c)) { 623 if (c == '\n') 624 break; 625 if (name_terminate(start, line-start, c, terminate)) 626 break; 627 } 628 line++; 629 if (c == '/' && !--p_value) 630 start = line; 631 } 632 if (!start) 633 return squash_slash(def); 634 len = line - start; 635 if (!len) 636 return squash_slash(def); 637 638 /* 639 * Generally we prefer the shorter name, especially 640 * if the other one is just a variation of that with 641 * something else tacked on to the end (ie "file.orig" 642 * or "file~"). 643 */ 644 if (def) { 645 int deflen = strlen(def); 646 if (deflen < len && !strncmp(start, def, deflen)) 647 return squash_slash(def); 648 free(def); 649 } 650 651 if (root) { 652 char *ret = xmalloc(root_len + len + 1); 653 strcpy(ret, root); 654 memcpy(ret + root_len, start, len); 655 ret[root_len + len] = '\0'; 656 return squash_slash(ret); 657 } 658 659 return squash_slash(xmemdupz(start, len)); 660} 661 662static char *find_name(const char *line, char *def, int p_value, int terminate) 663{ 664 if (*line == '"') { 665 char *name = find_name_gnu(line, def, p_value); 666 if (name) 667 return name; 668 } 669 670 return find_name_common(line, def, p_value, NULL, terminate); 671} 672 673static char *find_name_traditional(const char *line, char *def, int p_value) 674{ 675 size_t len = strlen(line); 676 size_t date_len; 677 678 if (*line == '"') { 679 char *name = find_name_gnu(line, def, p_value); 680 if (name) 681 return name; 682 } 683 684 len = strchrnul(line, '\n') - line; 685 date_len = diff_timestamp_len(line, len); 686 if (!date_len) 687 return find_name_common(line, def, p_value, NULL, TERM_TAB); 688 len -= date_len; 689 690 return find_name_common(line, def, p_value, line + len, 0); 691} 692 693static int count_slashes(const char *cp) 694{ 695 int cnt = 0; 696 char ch; 697 698 while ((ch = *cp++)) 699 if (ch == '/') 700 cnt++; 701 return cnt; 702} 703 704/* 705 * Given the string after "--- " or "+++ ", guess the appropriate 706 * p_value for the given patch. 707 */ 708static int guess_p_value(const char *nameline) 709{ 710 char *name, *cp; 711 int val = -1; 712 713 if (is_dev_null(nameline)) 714 return -1; 715 name = find_name_traditional(nameline, NULL, 0); 716 if (!name) 717 return -1; 718 cp = strchr(name, '/'); 719 if (!cp) 720 val = 0; 721 else if (prefix) { 722 /* 723 * Does it begin with "a/$our-prefix" and such? Then this is 724 * very likely to apply to our directory. 725 */ 726 if (!strncmp(name, prefix, prefix_length)) 727 val = count_slashes(prefix); 728 else { 729 cp++; 730 if (!strncmp(cp, prefix, prefix_length)) 731 val = count_slashes(prefix) + 1; 732 } 733 } 734 free(name); 735 return val; 736} 737 738/* 739 * Does the ---/+++ line has the POSIX timestamp after the last HT? 740 * GNU diff puts epoch there to signal a creation/deletion event. Is 741 * this such a timestamp? 742 */ 743static int has_epoch_timestamp(const char *nameline) 744{ 745 /* 746 * We are only interested in epoch timestamp; any non-zero 747 * fraction cannot be one, hence "(\.0+)?" in the regexp below. 748 * For the same reason, the date must be either 1969-12-31 or 749 * 1970-01-01, and the seconds part must be "00". 750 */ 751 const char stamp_regexp[] = 752 "^(1969-12-31|1970-01-01)" 753 " " 754 "[0-2][0-9]:[0-5][0-9]:00(\\.0+)?" 755 " " 756 "([-+][0-2][0-9]:?[0-5][0-9])\n"; 757 const char *timestamp = NULL, *cp, *colon; 758 static regex_t *stamp; 759 regmatch_t m[10]; 760 int zoneoffset; 761 int hourminute; 762 int status; 763 764 for (cp = nameline; *cp != '\n'; cp++) { 765 if (*cp == '\t') 766 timestamp = cp + 1; 767 } 768 if (!timestamp) 769 return 0; 770 if (!stamp) { 771 stamp = xmalloc(sizeof(*stamp)); 772 if (regcomp(stamp, stamp_regexp, REG_EXTENDED)) { 773 warning(_("Cannot prepare timestamp regexp %s"), 774 stamp_regexp); 775 return 0; 776 } 777 } 778 779 status = regexec(stamp, timestamp, ARRAY_SIZE(m), m, 0); 780 if (status) { 781 if (status != REG_NOMATCH) 782 warning(_("regexec returned %d for input: %s"), 783 status, timestamp); 784 return 0; 785 } 786 787 zoneoffset = strtol(timestamp + m[3].rm_so + 1, (char **) &colon, 10); 788 if (*colon == ':') 789 zoneoffset = zoneoffset * 60 + strtol(colon + 1, NULL, 10); 790 else 791 zoneoffset = (zoneoffset / 100) * 60 + (zoneoffset % 100); 792 if (timestamp[m[3].rm_so] == '-') 793 zoneoffset = -zoneoffset; 794 795 /* 796 * YYYY-MM-DD hh:mm:ss must be from either 1969-12-31 797 * (west of GMT) or 1970-01-01 (east of GMT) 798 */ 799 if ((zoneoffset < 0 && memcmp(timestamp, "1969-12-31", 10)) || 800 (0 <= zoneoffset && memcmp(timestamp, "1970-01-01", 10))) 801 return 0; 802 803 hourminute = (strtol(timestamp + 11, NULL, 10) * 60 + 804 strtol(timestamp + 14, NULL, 10) - 805 zoneoffset); 806 807 return ((zoneoffset < 0 && hourminute == 1440) || 808 (0 <= zoneoffset && !hourminute)); 809} 810 811/* 812 * Get the name etc info from the ---/+++ lines of a traditional patch header 813 * 814 * FIXME! The end-of-filename heuristics are kind of screwy. For existing 815 * files, we can happily check the index for a match, but for creating a 816 * new file we should try to match whatever "patch" does. I have no idea. 817 */ 818static void parse_traditional_patch(const char *first, const char *second, struct patch *patch) 819{ 820 char *name; 821 822 first += 4; /* skip "--- " */ 823 second += 4; /* skip "+++ " */ 824 if (!p_value_known) { 825 int p, q; 826 p = guess_p_value(first); 827 q = guess_p_value(second); 828 if (p < 0) p = q; 829 if (0 <= p && p == q) { 830 p_value = p; 831 p_value_known = 1; 832 } 833 } 834 if (is_dev_null(first)) { 835 patch->is_new = 1; 836 patch->is_delete = 0; 837 name = find_name_traditional(second, NULL, p_value); 838 patch->new_name = name; 839 } else if (is_dev_null(second)) { 840 patch->is_new = 0; 841 patch->is_delete = 1; 842 name = find_name_traditional(first, NULL, p_value); 843 patch->old_name = name; 844 } else { 845 name = find_name_traditional(first, NULL, p_value); 846 name = find_name_traditional(second, name, p_value); 847 if (has_epoch_timestamp(first)) { 848 patch->is_new = 1; 849 patch->is_delete = 0; 850 patch->new_name = name; 851 } else if (has_epoch_timestamp(second)) { 852 patch->is_new = 0; 853 patch->is_delete = 1; 854 patch->old_name = name; 855 } else { 856 patch->old_name = patch->new_name = name; 857 } 858 } 859 if (!name) 860 die(_("unable to find filename in patch at line %d"), linenr); 861} 862 863static int gitdiff_hdrend(const char *line, struct patch *patch) 864{ 865 return -1; 866} 867 868/* 869 * We're anal about diff header consistency, to make 870 * sure that we don't end up having strange ambiguous 871 * patches floating around. 872 * 873 * As a result, gitdiff_{old|new}name() will check 874 * their names against any previous information, just 875 * to make sure.. 876 */ 877static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, const char *oldnew) 878{ 879 if (!orig_name && !isnull) 880 return find_name(line, NULL, p_value, TERM_TAB); 881 882 if (orig_name) { 883 int len; 884 const char *name; 885 char *another; 886 name = orig_name; 887 len = strlen(name); 888 if (isnull) 889 die(_("git apply: bad git-diff - expected /dev/null, got %s on line %d"), name, linenr); 890 another = find_name(line, NULL, p_value, TERM_TAB); 891 if (!another || memcmp(another, name, len + 1)) 892 die(_("git apply: bad git-diff - inconsistent %s filename on line %d"), oldnew, linenr); 893 free(another); 894 return orig_name; 895 } 896 else { 897 /* expect "/dev/null" */ 898 if (memcmp("/dev/null", line, 9) || line[9] != '\n') 899 die(_("git apply: bad git-diff - expected /dev/null on line %d"), linenr); 900 return NULL; 901 } 902} 903 904static int gitdiff_oldname(const char *line, struct patch *patch) 905{ 906 patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name, "old"); 907 return 0; 908} 909 910static int gitdiff_newname(const char *line, struct patch *patch) 911{ 912 patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name, "new"); 913 return 0; 914} 915 916static int gitdiff_oldmode(const char *line, struct patch *patch) 917{ 918 patch->old_mode = strtoul(line, NULL, 8); 919 return 0; 920} 921 922static int gitdiff_newmode(const char *line, struct patch *patch) 923{ 924 patch->new_mode = strtoul(line, NULL, 8); 925 return 0; 926} 927 928static int gitdiff_delete(const char *line, struct patch *patch) 929{ 930 patch->is_delete = 1; 931 patch->old_name = patch->def_name; 932 return gitdiff_oldmode(line, patch); 933} 934 935static int gitdiff_newfile(const char *line, struct patch *patch) 936{ 937 patch->is_new = 1; 938 patch->new_name = patch->def_name; 939 return gitdiff_newmode(line, patch); 940} 941 942static int gitdiff_copysrc(const char *line, struct patch *patch) 943{ 944 patch->is_copy = 1; 945 patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0); 946 return 0; 947} 948 949static int gitdiff_copydst(const char *line, struct patch *patch) 950{ 951 patch->is_copy = 1; 952 patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0); 953 return 0; 954} 955 956static int gitdiff_renamesrc(const char *line, struct patch *patch) 957{ 958 patch->is_rename = 1; 959 patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0); 960 return 0; 961} 962 963static int gitdiff_renamedst(const char *line, struct patch *patch) 964{ 965 patch->is_rename = 1; 966 patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0); 967 return 0; 968} 969 970static int gitdiff_similarity(const char *line, struct patch *patch) 971{ 972 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX) 973 patch->score = 0; 974 return 0; 975} 976 977static int gitdiff_dissimilarity(const char *line, struct patch *patch) 978{ 979 if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX) 980 patch->score = 0; 981 return 0; 982} 983 984static int gitdiff_index(const char *line, struct patch *patch) 985{ 986 /* 987 * index line is N hexadecimal, "..", N hexadecimal, 988 * and optional space with octal mode. 989 */ 990 const char *ptr, *eol; 991 int len; 992 993 ptr = strchr(line, '.'); 994 if (!ptr || ptr[1] != '.' || 40 < ptr - line) 995 return 0; 996 len = ptr - line; 997 memcpy(patch->old_sha1_prefix, line, len); 998 patch->old_sha1_prefix[len] = 0; 9991000 line = ptr + 2;1001 ptr = strchr(line, ' ');1002 eol = strchr(line, '\n');10031004 if (!ptr || eol < ptr)1005 ptr = eol;1006 len = ptr - line;10071008 if (40 < len)1009 return 0;1010 memcpy(patch->new_sha1_prefix, line, len);1011 patch->new_sha1_prefix[len] = 0;1012 if (*ptr == ' ')1013 patch->old_mode = strtoul(ptr+1, NULL, 8);1014 return 0;1015}10161017/*1018 * This is normal for a diff that doesn't change anything: we'll fall through1019 * into the next diff. Tell the parser to break out.1020 */1021static int gitdiff_unrecognized(const char *line, struct patch *patch)1022{1023 return -1;1024}10251026static const char *stop_at_slash(const char *line, int llen)1027{1028 int nslash = p_value;1029 int i;10301031 for (i = 0; i < llen; i++) {1032 int ch = line[i];1033 if (ch == '/' && --nslash <= 0)1034 return &line[i];1035 }1036 return NULL;1037}10381039/*1040 * This is to extract the same name that appears on "diff --git"1041 * line. We do not find and return anything if it is a rename1042 * patch, and it is OK because we will find the name elsewhere.1043 * We need to reliably find name only when it is mode-change only,1044 * creation or deletion of an empty file. In any of these cases,1045 * both sides are the same name under a/ and b/ respectively.1046 */1047static char *git_header_name(char *line, int llen)1048{1049 const char *name;1050 const char *second = NULL;1051 size_t len, line_len;10521053 line += strlen("diff --git ");1054 llen -= strlen("diff --git ");10551056 if (*line == '"') {1057 const char *cp;1058 struct strbuf first = STRBUF_INIT;1059 struct strbuf sp = STRBUF_INIT;10601061 if (unquote_c_style(&first, line, &second))1062 goto free_and_fail1;10631064 /* advance to the first slash */1065 cp = stop_at_slash(first.buf, first.len);1066 /* we do not accept absolute paths */1067 if (!cp || cp == first.buf)1068 goto free_and_fail1;1069 strbuf_remove(&first, 0, cp + 1 - first.buf);10701071 /*1072 * second points at one past closing dq of name.1073 * find the second name.1074 */1075 while ((second < line + llen) && isspace(*second))1076 second++;10771078 if (line + llen <= second)1079 goto free_and_fail1;1080 if (*second == '"') {1081 if (unquote_c_style(&sp, second, NULL))1082 goto free_and_fail1;1083 cp = stop_at_slash(sp.buf, sp.len);1084 if (!cp || cp == sp.buf)1085 goto free_and_fail1;1086 /* They must match, otherwise ignore */1087 if (strcmp(cp + 1, first.buf))1088 goto free_and_fail1;1089 strbuf_release(&sp);1090 return strbuf_detach(&first, NULL);1091 }10921093 /* unquoted second */1094 cp = stop_at_slash(second, line + llen - second);1095 if (!cp || cp == second)1096 goto free_and_fail1;1097 cp++;1098 if (line + llen - cp != first.len + 1 ||1099 memcmp(first.buf, cp, first.len))1100 goto free_and_fail1;1101 return strbuf_detach(&first, NULL);11021103 free_and_fail1:1104 strbuf_release(&first);1105 strbuf_release(&sp);1106 return NULL;1107 }11081109 /* unquoted first name */1110 name = stop_at_slash(line, llen);1111 if (!name || name == line)1112 return NULL;1113 name++;11141115 /*1116 * since the first name is unquoted, a dq if exists must be1117 * the beginning of the second name.1118 */1119 for (second = name; second < line + llen; second++) {1120 if (*second == '"') {1121 struct strbuf sp = STRBUF_INIT;1122 const char *np;11231124 if (unquote_c_style(&sp, second, NULL))1125 goto free_and_fail2;11261127 np = stop_at_slash(sp.buf, sp.len);1128 if (!np || np == sp.buf)1129 goto free_and_fail2;1130 np++;11311132 len = sp.buf + sp.len - np;1133 if (len < second - name &&1134 !strncmp(np, name, len) &&1135 isspace(name[len])) {1136 /* Good */1137 strbuf_remove(&sp, 0, np - sp.buf);1138 return strbuf_detach(&sp, NULL);1139 }11401141 free_and_fail2:1142 strbuf_release(&sp);1143 return NULL;1144 }1145 }11461147 /*1148 * Accept a name only if it shows up twice, exactly the same1149 * form.1150 */1151 second = strchr(name, '\n');1152 if (!second)1153 return NULL;1154 line_len = second - name;1155 for (len = 0 ; ; len++) {1156 switch (name[len]) {1157 default:1158 continue;1159 case '\n':1160 return NULL;1161 case '\t': case ' ':1162 second = stop_at_slash(name + len, line_len - len);1163 if (!second)1164 return NULL;1165 second++;1166 if (second[len] == '\n' && !strncmp(name, second, len)) {1167 return xmemdupz(name, len);1168 }1169 }1170 }1171}11721173/* Verify that we recognize the lines following a git header */1174static int parse_git_header(char *line, int len, unsigned int size, struct patch *patch)1175{1176 unsigned long offset;11771178 /* A git diff has explicit new/delete information, so we don't guess */1179 patch->is_new = 0;1180 patch->is_delete = 0;11811182 /*1183 * Some things may not have the old name in the1184 * rest of the headers anywhere (pure mode changes,1185 * or removing or adding empty files), so we get1186 * the default name from the header.1187 */1188 patch->def_name = git_header_name(line, len);1189 if (patch->def_name && root) {1190 char *s = xmalloc(root_len + strlen(patch->def_name) + 1);1191 strcpy(s, root);1192 strcpy(s + root_len, patch->def_name);1193 free(patch->def_name);1194 patch->def_name = s;1195 }11961197 line += len;1198 size -= len;1199 linenr++;1200 for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {1201 static const struct opentry {1202 const char *str;1203 int (*fn)(const char *, struct patch *);1204 } optable[] = {1205 { "@@ -", gitdiff_hdrend },1206 { "--- ", gitdiff_oldname },1207 { "+++ ", gitdiff_newname },1208 { "old mode ", gitdiff_oldmode },1209 { "new mode ", gitdiff_newmode },1210 { "deleted file mode ", gitdiff_delete },1211 { "new file mode ", gitdiff_newfile },1212 { "copy from ", gitdiff_copysrc },1213 { "copy to ", gitdiff_copydst },1214 { "rename old ", gitdiff_renamesrc },1215 { "rename new ", gitdiff_renamedst },1216 { "rename from ", gitdiff_renamesrc },1217 { "rename to ", gitdiff_renamedst },1218 { "similarity index ", gitdiff_similarity },1219 { "dissimilarity index ", gitdiff_dissimilarity },1220 { "index ", gitdiff_index },1221 { "", gitdiff_unrecognized },1222 };1223 int i;12241225 len = linelen(line, size);1226 if (!len || line[len-1] != '\n')1227 break;1228 for (i = 0; i < ARRAY_SIZE(optable); i++) {1229 const struct opentry *p = optable + i;1230 int oplen = strlen(p->str);1231 if (len < oplen || memcmp(p->str, line, oplen))1232 continue;1233 if (p->fn(line + oplen, patch) < 0)1234 return offset;1235 break;1236 }1237 }12381239 return offset;1240}12411242static int parse_num(const char *line, unsigned long *p)1243{1244 char *ptr;12451246 if (!isdigit(*line))1247 return 0;1248 *p = strtoul(line, &ptr, 10);1249 return ptr - line;1250}12511252static int parse_range(const char *line, int len, int offset, const char *expect,1253 unsigned long *p1, unsigned long *p2)1254{1255 int digits, ex;12561257 if (offset < 0 || offset >= len)1258 return -1;1259 line += offset;1260 len -= offset;12611262 digits = parse_num(line, p1);1263 if (!digits)1264 return -1;12651266 offset += digits;1267 line += digits;1268 len -= digits;12691270 *p2 = 1;1271 if (*line == ',') {1272 digits = parse_num(line+1, p2);1273 if (!digits)1274 return -1;12751276 offset += digits+1;1277 line += digits+1;1278 len -= digits+1;1279 }12801281 ex = strlen(expect);1282 if (ex > len)1283 return -1;1284 if (memcmp(line, expect, ex))1285 return -1;12861287 return offset + ex;1288}12891290static void recount_diff(char *line, int size, struct fragment *fragment)1291{1292 int oldlines = 0, newlines = 0, ret = 0;12931294 if (size < 1) {1295 warning("recount: ignore empty hunk");1296 return;1297 }12981299 for (;;) {1300 int len = linelen(line, size);1301 size -= len;1302 line += len;13031304 if (size < 1)1305 break;13061307 switch (*line) {1308 case ' ': case '\n':1309 newlines++;1310 /* fall through */1311 case '-':1312 oldlines++;1313 continue;1314 case '+':1315 newlines++;1316 continue;1317 case '\\':1318 continue;1319 case '@':1320 ret = size < 3 || prefixcmp(line, "@@ ");1321 break;1322 case 'd':1323 ret = size < 5 || prefixcmp(line, "diff ");1324 break;1325 default:1326 ret = -1;1327 break;1328 }1329 if (ret) {1330 warning(_("recount: unexpected line: %.*s"),1331 (int)linelen(line, size), line);1332 return;1333 }1334 break;1335 }1336 fragment->oldlines = oldlines;1337 fragment->newlines = newlines;1338}13391340/*1341 * Parse a unified diff fragment header of the1342 * form "@@ -a,b +c,d @@"1343 */1344static int parse_fragment_header(char *line, int len, struct fragment *fragment)1345{1346 int offset;13471348 if (!len || line[len-1] != '\n')1349 return -1;13501351 /* Figure out the number of lines in a fragment */1352 offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);1353 offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);13541355 return offset;1356}13571358static int find_header(char *line, unsigned long size, int *hdrsize, struct patch *patch)1359{1360 unsigned long offset, len;13611362 patch->is_toplevel_relative = 0;1363 patch->is_rename = patch->is_copy = 0;1364 patch->is_new = patch->is_delete = -1;1365 patch->old_mode = patch->new_mode = 0;1366 patch->old_name = patch->new_name = NULL;1367 for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {1368 unsigned long nextlen;13691370 len = linelen(line, size);1371 if (!len)1372 break;13731374 /* Testing this early allows us to take a few shortcuts.. */1375 if (len < 6)1376 continue;13771378 /*1379 * Make sure we don't find any unconnected patch fragments.1380 * That's a sign that we didn't find a header, and that a1381 * patch has become corrupted/broken up.1382 */1383 if (!memcmp("@@ -", line, 4)) {1384 struct fragment dummy;1385 if (parse_fragment_header(line, len, &dummy) < 0)1386 continue;1387 die(_("patch fragment without header at line %d: %.*s"),1388 linenr, (int)len-1, line);1389 }13901391 if (size < len + 6)1392 break;13931394 /*1395 * Git patch? It might not have a real patch, just a rename1396 * or mode change, so we handle that specially1397 */1398 if (!memcmp("diff --git ", line, 11)) {1399 int git_hdr_len = parse_git_header(line, len, size, patch);1400 if (git_hdr_len <= len)1401 continue;1402 if (!patch->old_name && !patch->new_name) {1403 if (!patch->def_name)1404 die(Q_("git diff header lacks filename information when removing "1405 "%d leading pathname component (line %d)",1406 "git diff header lacks filename information when removing "1407 "%d leading pathname components (line %d)",1408 p_value),1409 p_value, linenr);1410 patch->old_name = patch->new_name = patch->def_name;1411 }1412 if (!patch->is_delete && !patch->new_name)1413 die("git diff header lacks filename information "1414 "(line %d)", linenr);1415 patch->is_toplevel_relative = 1;1416 *hdrsize = git_hdr_len;1417 return offset;1418 }14191420 /* --- followed by +++ ? */1421 if (memcmp("--- ", line, 4) || memcmp("+++ ", line + len, 4))1422 continue;14231424 /*1425 * We only accept unified patches, so we want it to1426 * at least have "@@ -a,b +c,d @@\n", which is 14 chars1427 * minimum ("@@ -0,0 +1 @@\n" is the shortest).1428 */1429 nextlen = linelen(line + len, size - len);1430 if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))1431 continue;14321433 /* Ok, we'll consider it a patch */1434 parse_traditional_patch(line, line+len, patch);1435 *hdrsize = len + nextlen;1436 linenr += 2;1437 return offset;1438 }1439 return -1;1440}14411442static void record_ws_error(unsigned result, const char *line, int len, int linenr)1443{1444 char *err;14451446 if (!result)1447 return;14481449 whitespace_error++;1450 if (squelch_whitespace_errors &&1451 squelch_whitespace_errors < whitespace_error)1452 return;14531454 err = whitespace_error_string(result);1455 fprintf(stderr, "%s:%d: %s.\n%.*s\n",1456 patch_input_file, linenr, err, len, line);1457 free(err);1458}14591460static void check_whitespace(const char *line, int len, unsigned ws_rule)1461{1462 unsigned result = ws_check(line + 1, len - 1, ws_rule);14631464 record_ws_error(result, line + 1, len - 2, linenr);1465}14661467/*1468 * Parse a unified diff. Note that this really needs to parse each1469 * fragment separately, since the only way to know the difference1470 * between a "---" that is part of a patch, and a "---" that starts1471 * the next patch is to look at the line counts..1472 */1473static int parse_fragment(char *line, unsigned long size,1474 struct patch *patch, struct fragment *fragment)1475{1476 int added, deleted;1477 int len = linelen(line, size), offset;1478 unsigned long oldlines, newlines;1479 unsigned long leading, trailing;14801481 offset = parse_fragment_header(line, len, fragment);1482 if (offset < 0)1483 return -1;1484 if (offset > 0 && patch->recount)1485 recount_diff(line + offset, size - offset, fragment);1486 oldlines = fragment->oldlines;1487 newlines = fragment->newlines;1488 leading = 0;1489 trailing = 0;14901491 /* Parse the thing.. */1492 line += len;1493 size -= len;1494 linenr++;1495 added = deleted = 0;1496 for (offset = len;1497 0 < size;1498 offset += len, size -= len, line += len, linenr++) {1499 if (!oldlines && !newlines)1500 break;1501 len = linelen(line, size);1502 if (!len || line[len-1] != '\n')1503 return -1;1504 switch (*line) {1505 default:1506 return -1;1507 case '\n': /* newer GNU diff, an empty context line */1508 case ' ':1509 oldlines--;1510 newlines--;1511 if (!deleted && !added)1512 leading++;1513 trailing++;1514 break;1515 case '-':1516 if (apply_in_reverse &&1517 ws_error_action != nowarn_ws_error)1518 check_whitespace(line, len, patch->ws_rule);1519 deleted++;1520 oldlines--;1521 trailing = 0;1522 break;1523 case '+':1524 if (!apply_in_reverse &&1525 ws_error_action != nowarn_ws_error)1526 check_whitespace(line, len, patch->ws_rule);1527 added++;1528 newlines--;1529 trailing = 0;1530 break;15311532 /*1533 * We allow "\ No newline at end of file". Depending1534 * on locale settings when the patch was produced we1535 * don't know what this line looks like. The only1536 * thing we do know is that it begins with "\ ".1537 * Checking for 12 is just for sanity check -- any1538 * l10n of "\ No newline..." is at least that long.1539 */1540 case '\\':1541 if (len < 12 || memcmp(line, "\\ ", 2))1542 return -1;1543 break;1544 }1545 }1546 if (oldlines || newlines)1547 return -1;1548 fragment->leading = leading;1549 fragment->trailing = trailing;15501551 /*1552 * If a fragment ends with an incomplete line, we failed to include1553 * it in the above loop because we hit oldlines == newlines == 01554 * before seeing it.1555 */1556 if (12 < size && !memcmp(line, "\\ ", 2))1557 offset += linelen(line, size);15581559 patch->lines_added += added;1560 patch->lines_deleted += deleted;15611562 if (0 < patch->is_new && oldlines)1563 return error(_("new file depends on old contents"));1564 if (0 < patch->is_delete && newlines)1565 return error(_("deleted file still has contents"));1566 return offset;1567}15681569static int parse_single_patch(char *line, unsigned long size, struct patch *patch)1570{1571 unsigned long offset = 0;1572 unsigned long oldlines = 0, newlines = 0, context = 0;1573 struct fragment **fragp = &patch->fragments;15741575 while (size > 4 && !memcmp(line, "@@ -", 4)) {1576 struct fragment *fragment;1577 int len;15781579 fragment = xcalloc(1, sizeof(*fragment));1580 fragment->linenr = linenr;1581 len = parse_fragment(line, size, patch, fragment);1582 if (len <= 0)1583 die(_("corrupt patch at line %d"), linenr);1584 fragment->patch = line;1585 fragment->size = len;1586 oldlines += fragment->oldlines;1587 newlines += fragment->newlines;1588 context += fragment->leading + fragment->trailing;15891590 *fragp = fragment;1591 fragp = &fragment->next;15921593 offset += len;1594 line += len;1595 size -= len;1596 }15971598 /*1599 * If something was removed (i.e. we have old-lines) it cannot1600 * be creation, and if something was added it cannot be1601 * deletion. However, the reverse is not true; --unified=01602 * patches that only add are not necessarily creation even1603 * though they do not have any old lines, and ones that only1604 * delete are not necessarily deletion.1605 *1606 * Unfortunately, a real creation/deletion patch do _not_ have1607 * any context line by definition, so we cannot safely tell it1608 * apart with --unified=0 insanity. At least if the patch has1609 * more than one hunk it is not creation or deletion.1610 */1611 if (patch->is_new < 0 &&1612 (oldlines || (patch->fragments && patch->fragments->next)))1613 patch->is_new = 0;1614 if (patch->is_delete < 0 &&1615 (newlines || (patch->fragments && patch->fragments->next)))1616 patch->is_delete = 0;16171618 if (0 < patch->is_new && oldlines)1619 die(_("new file %s depends on old contents"), patch->new_name);1620 if (0 < patch->is_delete && newlines)1621 die(_("deleted file %s still has contents"), patch->old_name);1622 if (!patch->is_delete && !newlines && context)1623 fprintf_ln(stderr,1624 _("** warning: "1625 "file %s becomes empty but is not deleted"),1626 patch->new_name);16271628 return offset;1629}16301631static inline int metadata_changes(struct patch *patch)1632{1633 return patch->is_rename > 0 ||1634 patch->is_copy > 0 ||1635 patch->is_new > 0 ||1636 patch->is_delete ||1637 (patch->old_mode && patch->new_mode &&1638 patch->old_mode != patch->new_mode);1639}16401641static char *inflate_it(const void *data, unsigned long size,1642 unsigned long inflated_size)1643{1644 git_zstream stream;1645 void *out;1646 int st;16471648 memset(&stream, 0, sizeof(stream));16491650 stream.next_in = (unsigned char *)data;1651 stream.avail_in = size;1652 stream.next_out = out = xmalloc(inflated_size);1653 stream.avail_out = inflated_size;1654 git_inflate_init(&stream);1655 st = git_inflate(&stream, Z_FINISH);1656 git_inflate_end(&stream);1657 if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {1658 free(out);1659 return NULL;1660 }1661 return out;1662}16631664static struct fragment *parse_binary_hunk(char **buf_p,1665 unsigned long *sz_p,1666 int *status_p,1667 int *used_p)1668{1669 /*1670 * Expect a line that begins with binary patch method ("literal"1671 * or "delta"), followed by the length of data before deflating.1672 * a sequence of 'length-byte' followed by base-85 encoded data1673 * should follow, terminated by a newline.1674 *1675 * Each 5-byte sequence of base-85 encodes up to 4 bytes,1676 * and we would limit the patch line to 66 characters,1677 * so one line can fit up to 13 groups that would decode1678 * to 52 bytes max. The length byte 'A'-'Z' corresponds1679 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.1680 */1681 int llen, used;1682 unsigned long size = *sz_p;1683 char *buffer = *buf_p;1684 int patch_method;1685 unsigned long origlen;1686 char *data = NULL;1687 int hunk_size = 0;1688 struct fragment *frag;16891690 llen = linelen(buffer, size);1691 used = llen;16921693 *status_p = 0;16941695 if (!prefixcmp(buffer, "delta ")) {1696 patch_method = BINARY_DELTA_DEFLATED;1697 origlen = strtoul(buffer + 6, NULL, 10);1698 }1699 else if (!prefixcmp(buffer, "literal ")) {1700 patch_method = BINARY_LITERAL_DEFLATED;1701 origlen = strtoul(buffer + 8, NULL, 10);1702 }1703 else1704 return NULL;17051706 linenr++;1707 buffer += llen;1708 while (1) {1709 int byte_length, max_byte_length, newsize;1710 llen = linelen(buffer, size);1711 used += llen;1712 linenr++;1713 if (llen == 1) {1714 /* consume the blank line */1715 buffer++;1716 size--;1717 break;1718 }1719 /*1720 * Minimum line is "A00000\n" which is 7-byte long,1721 * and the line length must be multiple of 5 plus 2.1722 */1723 if ((llen < 7) || (llen-2) % 5)1724 goto corrupt;1725 max_byte_length = (llen - 2) / 5 * 4;1726 byte_length = *buffer;1727 if ('A' <= byte_length && byte_length <= 'Z')1728 byte_length = byte_length - 'A' + 1;1729 else if ('a' <= byte_length && byte_length <= 'z')1730 byte_length = byte_length - 'a' + 27;1731 else1732 goto corrupt;1733 /* if the input length was not multiple of 4, we would1734 * have filler at the end but the filler should never1735 * exceed 3 bytes1736 */1737 if (max_byte_length < byte_length ||1738 byte_length <= max_byte_length - 4)1739 goto corrupt;1740 newsize = hunk_size + byte_length;1741 data = xrealloc(data, newsize);1742 if (decode_85(data + hunk_size, buffer + 1, byte_length))1743 goto corrupt;1744 hunk_size = newsize;1745 buffer += llen;1746 size -= llen;1747 }17481749 frag = xcalloc(1, sizeof(*frag));1750 frag->patch = inflate_it(data, hunk_size, origlen);1751 if (!frag->patch)1752 goto corrupt;1753 free(data);1754 frag->size = origlen;1755 *buf_p = buffer;1756 *sz_p = size;1757 *used_p = used;1758 frag->binary_patch_method = patch_method;1759 return frag;17601761 corrupt:1762 free(data);1763 *status_p = -1;1764 error(_("corrupt binary patch at line %d: %.*s"),1765 linenr-1, llen-1, buffer);1766 return NULL;1767}17681769static int parse_binary(char *buffer, unsigned long size, struct patch *patch)1770{1771 /*1772 * We have read "GIT binary patch\n"; what follows is a line1773 * that says the patch method (currently, either "literal" or1774 * "delta") and the length of data before deflating; a1775 * sequence of 'length-byte' followed by base-85 encoded data1776 * follows.1777 *1778 * When a binary patch is reversible, there is another binary1779 * hunk in the same format, starting with patch method (either1780 * "literal" or "delta") with the length of data, and a sequence1781 * of length-byte + base-85 encoded data, terminated with another1782 * empty line. This data, when applied to the postimage, produces1783 * the preimage.1784 */1785 struct fragment *forward;1786 struct fragment *reverse;1787 int status;1788 int used, used_1;17891790 forward = parse_binary_hunk(&buffer, &size, &status, &used);1791 if (!forward && !status)1792 /* there has to be one hunk (forward hunk) */1793 return error(_("unrecognized binary patch at line %d"), linenr-1);1794 if (status)1795 /* otherwise we already gave an error message */1796 return status;17971798 reverse = parse_binary_hunk(&buffer, &size, &status, &used_1);1799 if (reverse)1800 used += used_1;1801 else if (status) {1802 /*1803 * Not having reverse hunk is not an error, but having1804 * a corrupt reverse hunk is.1805 */1806 free((void*) forward->patch);1807 free(forward);1808 return status;1809 }1810 forward->next = reverse;1811 patch->fragments = forward;1812 patch->is_binary = 1;1813 return used;1814}18151816static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)1817{1818 int hdrsize, patchsize;1819 int offset = find_header(buffer, size, &hdrsize, patch);18201821 if (offset < 0)1822 return offset;18231824 patch->ws_rule = whitespace_rule(patch->new_name1825 ? patch->new_name1826 : patch->old_name);18271828 patchsize = parse_single_patch(buffer + offset + hdrsize,1829 size - offset - hdrsize, patch);18301831 if (!patchsize) {1832 static const char *binhdr[] = {1833 "Binary files ",1834 "Files ",1835 NULL,1836 };1837 static const char git_binary[] = "GIT binary patch\n";1838 int i;1839 int hd = hdrsize + offset;1840 unsigned long llen = linelen(buffer + hd, size - hd);18411842 if (llen == sizeof(git_binary) - 1 &&1843 !memcmp(git_binary, buffer + hd, llen)) {1844 int used;1845 linenr++;1846 used = parse_binary(buffer + hd + llen,1847 size - hd - llen, patch);1848 if (used)1849 patchsize = used + llen;1850 else1851 patchsize = 0;1852 }1853 else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {1854 for (i = 0; binhdr[i]; i++) {1855 int len = strlen(binhdr[i]);1856 if (len < size - hd &&1857 !memcmp(binhdr[i], buffer + hd, len)) {1858 linenr++;1859 patch->is_binary = 1;1860 patchsize = llen;1861 break;1862 }1863 }1864 }18651866 /* Empty patch cannot be applied if it is a text patch1867 * without metadata change. A binary patch appears1868 * empty to us here.1869 */1870 if ((apply || check) &&1871 (!patch->is_binary && !metadata_changes(patch)))1872 die(_("patch with only garbage at line %d"), linenr);1873 }18741875 return offset + hdrsize + patchsize;1876}18771878#define swap(a,b) myswap((a),(b),sizeof(a))18791880#define myswap(a, b, size) do { \1881 unsigned char mytmp[size]; \1882 memcpy(mytmp, &a, size); \1883 memcpy(&a, &b, size); \1884 memcpy(&b, mytmp, size); \1885} while (0)18861887static void reverse_patches(struct patch *p)1888{1889 for (; p; p = p->next) {1890 struct fragment *frag = p->fragments;18911892 swap(p->new_name, p->old_name);1893 swap(p->new_mode, p->old_mode);1894 swap(p->is_new, p->is_delete);1895 swap(p->lines_added, p->lines_deleted);1896 swap(p->old_sha1_prefix, p->new_sha1_prefix);18971898 for (; frag; frag = frag->next) {1899 swap(frag->newpos, frag->oldpos);1900 swap(frag->newlines, frag->oldlines);1901 }1902 }1903}19041905static const char pluses[] =1906"++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";1907static const char minuses[]=1908"----------------------------------------------------------------------";19091910static void show_stats(struct patch *patch)1911{1912 struct strbuf qname = STRBUF_INIT;1913 char *cp = patch->new_name ? patch->new_name : patch->old_name;1914 int max, add, del;19151916 quote_c_style(cp, &qname, NULL, 0);19171918 /*1919 * "scale" the filename1920 */1921 max = max_len;1922 if (max > 50)1923 max = 50;19241925 if (qname.len > max) {1926 cp = strchr(qname.buf + qname.len + 3 - max, '/');1927 if (!cp)1928 cp = qname.buf + qname.len + 3 - max;1929 strbuf_splice(&qname, 0, cp - qname.buf, "...", 3);1930 }19311932 if (patch->is_binary) {1933 printf(" %-*s | Bin\n", max, qname.buf);1934 strbuf_release(&qname);1935 return;1936 }19371938 printf(" %-*s |", max, qname.buf);1939 strbuf_release(&qname);19401941 /*1942 * scale the add/delete1943 */1944 max = max + max_change > 70 ? 70 - max : max_change;1945 add = patch->lines_added;1946 del = patch->lines_deleted;19471948 if (max_change > 0) {1949 int total = ((add + del) * max + max_change / 2) / max_change;1950 add = (add * max + max_change / 2) / max_change;1951 del = total - add;1952 }1953 printf("%5d %.*s%.*s\n", patch->lines_added + patch->lines_deleted,1954 add, pluses, del, minuses);1955}19561957static int read_old_data(struct stat *st, const char *path, struct strbuf *buf)1958{1959 switch (st->st_mode & S_IFMT) {1960 case S_IFLNK:1961 if (strbuf_readlink(buf, path, st->st_size) < 0)1962 return error(_("unable to read symlink %s"), path);1963 return 0;1964 case S_IFREG:1965 if (strbuf_read_file(buf, path, st->st_size) != st->st_size)1966 return error(_("unable to open or read %s"), path);1967 convert_to_git(path, buf->buf, buf->len, buf, 0);1968 return 0;1969 default:1970 return -1;1971 }1972}19731974/*1975 * Update the preimage, and the common lines in postimage,1976 * from buffer buf of length len. If postlen is 0 the postimage1977 * is updated in place, otherwise it's updated on a new buffer1978 * of length postlen1979 */19801981static void update_pre_post_images(struct image *preimage,1982 struct image *postimage,1983 char *buf,1984 size_t len, size_t postlen)1985{1986 int i, ctx;1987 char *new, *old, *fixed;1988 struct image fixed_preimage;19891990 /*1991 * Update the preimage with whitespace fixes. Note that we1992 * are not losing preimage->buf -- apply_one_fragment() will1993 * free "oldlines".1994 */1995 prepare_image(&fixed_preimage, buf, len, 1);1996 assert(fixed_preimage.nr == preimage->nr);1997 for (i = 0; i < preimage->nr; i++)1998 fixed_preimage.line[i].flag = preimage->line[i].flag;1999 free(preimage->line_allocated);2000 *preimage = fixed_preimage;20012002 /*2003 * Adjust the common context lines in postimage. This can be2004 * done in-place when we are just doing whitespace fixing,2005 * which does not make the string grow, but needs a new buffer2006 * when ignoring whitespace causes the update, since in this case2007 * we could have e.g. tabs converted to multiple spaces.2008 * We trust the caller to tell us if the update can be done2009 * in place (postlen==0) or not.2010 */2011 old = postimage->buf;2012 if (postlen)2013 new = postimage->buf = xmalloc(postlen);2014 else2015 new = old;2016 fixed = preimage->buf;2017 for (i = ctx = 0; i < postimage->nr; i++) {2018 size_t len = postimage->line[i].len;2019 if (!(postimage->line[i].flag & LINE_COMMON)) {2020 /* an added line -- no counterparts in preimage */2021 memmove(new, old, len);2022 old += len;2023 new += len;2024 continue;2025 }20262027 /* a common context -- skip it in the original postimage */2028 old += len;20292030 /* and find the corresponding one in the fixed preimage */2031 while (ctx < preimage->nr &&2032 !(preimage->line[ctx].flag & LINE_COMMON)) {2033 fixed += preimage->line[ctx].len;2034 ctx++;2035 }2036 if (preimage->nr <= ctx)2037 die(_("oops"));20382039 /* and copy it in, while fixing the line length */2040 len = preimage->line[ctx].len;2041 memcpy(new, fixed, len);2042 new += len;2043 fixed += len;2044 postimage->line[i].len = len;2045 ctx++;2046 }20472048 /* Fix the length of the whole thing */2049 postimage->len = new - postimage->buf;2050}20512052static int match_fragment(struct image *img,2053 struct image *preimage,2054 struct image *postimage,2055 unsigned long try,2056 int try_lno,2057 unsigned ws_rule,2058 int match_beginning, int match_end)2059{2060 int i;2061 char *fixed_buf, *buf, *orig, *target;2062 struct strbuf fixed;2063 size_t fixed_len;2064 int preimage_limit;20652066 if (preimage->nr + try_lno <= img->nr) {2067 /*2068 * The hunk falls within the boundaries of img.2069 */2070 preimage_limit = preimage->nr;2071 if (match_end && (preimage->nr + try_lno != img->nr))2072 return 0;2073 } else if (ws_error_action == correct_ws_error &&2074 (ws_rule & WS_BLANK_AT_EOF)) {2075 /*2076 * This hunk extends beyond the end of img, and we are2077 * removing blank lines at the end of the file. This2078 * many lines from the beginning of the preimage must2079 * match with img, and the remainder of the preimage2080 * must be blank.2081 */2082 preimage_limit = img->nr - try_lno;2083 } else {2084 /*2085 * The hunk extends beyond the end of the img and2086 * we are not removing blanks at the end, so we2087 * should reject the hunk at this position.2088 */2089 return 0;2090 }20912092 if (match_beginning && try_lno)2093 return 0;20942095 /* Quick hash check */2096 for (i = 0; i < preimage_limit; i++)2097 if ((img->line[try_lno + i].flag & LINE_PATCHED) ||2098 (preimage->line[i].hash != img->line[try_lno + i].hash))2099 return 0;21002101 if (preimage_limit == preimage->nr) {2102 /*2103 * Do we have an exact match? If we were told to match2104 * at the end, size must be exactly at try+fragsize,2105 * otherwise try+fragsize must be still within the preimage,2106 * and either case, the old piece should match the preimage2107 * exactly.2108 */2109 if ((match_end2110 ? (try + preimage->len == img->len)2111 : (try + preimage->len <= img->len)) &&2112 !memcmp(img->buf + try, preimage->buf, preimage->len))2113 return 1;2114 } else {2115 /*2116 * The preimage extends beyond the end of img, so2117 * there cannot be an exact match.2118 *2119 * There must be one non-blank context line that match2120 * a line before the end of img.2121 */2122 char *buf_end;21232124 buf = preimage->buf;2125 buf_end = buf;2126 for (i = 0; i < preimage_limit; i++)2127 buf_end += preimage->line[i].len;21282129 for ( ; buf < buf_end; buf++)2130 if (!isspace(*buf))2131 break;2132 if (buf == buf_end)2133 return 0;2134 }21352136 /*2137 * No exact match. If we are ignoring whitespace, run a line-by-line2138 * fuzzy matching. We collect all the line length information because2139 * we need it to adjust whitespace if we match.2140 */2141 if (ws_ignore_action == ignore_ws_change) {2142 size_t imgoff = 0;2143 size_t preoff = 0;2144 size_t postlen = postimage->len;2145 size_t extra_chars;2146 char *preimage_eof;2147 char *preimage_end;2148 for (i = 0; i < preimage_limit; i++) {2149 size_t prelen = preimage->line[i].len;2150 size_t imglen = img->line[try_lno+i].len;21512152 if (!fuzzy_matchlines(img->buf + try + imgoff, imglen,2153 preimage->buf + preoff, prelen))2154 return 0;2155 if (preimage->line[i].flag & LINE_COMMON)2156 postlen += imglen - prelen;2157 imgoff += imglen;2158 preoff += prelen;2159 }21602161 /*2162 * Ok, the preimage matches with whitespace fuzz.2163 *2164 * imgoff now holds the true length of the target that2165 * matches the preimage before the end of the file.2166 *2167 * Count the number of characters in the preimage that fall2168 * beyond the end of the file and make sure that all of them2169 * are whitespace characters. (This can only happen if2170 * we are removing blank lines at the end of the file.)2171 */2172 buf = preimage_eof = preimage->buf + preoff;2173 for ( ; i < preimage->nr; i++)2174 preoff += preimage->line[i].len;2175 preimage_end = preimage->buf + preoff;2176 for ( ; buf < preimage_end; buf++)2177 if (!isspace(*buf))2178 return 0;21792180 /*2181 * Update the preimage and the common postimage context2182 * lines to use the same whitespace as the target.2183 * If whitespace is missing in the target (i.e.2184 * if the preimage extends beyond the end of the file),2185 * use the whitespace from the preimage.2186 */2187 extra_chars = preimage_end - preimage_eof;2188 strbuf_init(&fixed, imgoff + extra_chars);2189 strbuf_add(&fixed, img->buf + try, imgoff);2190 strbuf_add(&fixed, preimage_eof, extra_chars);2191 fixed_buf = strbuf_detach(&fixed, &fixed_len);2192 update_pre_post_images(preimage, postimage,2193 fixed_buf, fixed_len, postlen);2194 return 1;2195 }21962197 if (ws_error_action != correct_ws_error)2198 return 0;21992200 /*2201 * The hunk does not apply byte-by-byte, but the hash says2202 * it might with whitespace fuzz. We haven't been asked to2203 * ignore whitespace, we were asked to correct whitespace2204 * errors, so let's try matching after whitespace correction.2205 *2206 * The preimage may extend beyond the end of the file,2207 * but in this loop we will only handle the part of the2208 * preimage that falls within the file.2209 */2210 strbuf_init(&fixed, preimage->len + 1);2211 orig = preimage->buf;2212 target = img->buf + try;2213 for (i = 0; i < preimage_limit; i++) {2214 size_t oldlen = preimage->line[i].len;2215 size_t tgtlen = img->line[try_lno + i].len;2216 size_t fixstart = fixed.len;2217 struct strbuf tgtfix;2218 int match;22192220 /* Try fixing the line in the preimage */2221 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);22222223 /* Try fixing the line in the target */2224 strbuf_init(&tgtfix, tgtlen);2225 ws_fix_copy(&tgtfix, target, tgtlen, ws_rule, NULL);22262227 /*2228 * If they match, either the preimage was based on2229 * a version before our tree fixed whitespace breakage,2230 * or we are lacking a whitespace-fix patch the tree2231 * the preimage was based on already had (i.e. target2232 * has whitespace breakage, the preimage doesn't).2233 * In either case, we are fixing the whitespace breakages2234 * so we might as well take the fix together with their2235 * real change.2236 */2237 match = (tgtfix.len == fixed.len - fixstart &&2238 !memcmp(tgtfix.buf, fixed.buf + fixstart,2239 fixed.len - fixstart));22402241 strbuf_release(&tgtfix);2242 if (!match)2243 goto unmatch_exit;22442245 orig += oldlen;2246 target += tgtlen;2247 }224822492250 /*2251 * Now handle the lines in the preimage that falls beyond the2252 * end of the file (if any). They will only match if they are2253 * empty or only contain whitespace (if WS_BLANK_AT_EOL is2254 * false).2255 */2256 for ( ; i < preimage->nr; i++) {2257 size_t fixstart = fixed.len; /* start of the fixed preimage */2258 size_t oldlen = preimage->line[i].len;2259 int j;22602261 /* Try fixing the line in the preimage */2262 ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);22632264 for (j = fixstart; j < fixed.len; j++)2265 if (!isspace(fixed.buf[j]))2266 goto unmatch_exit;22672268 orig += oldlen;2269 }22702271 /*2272 * Yes, the preimage is based on an older version that still2273 * has whitespace breakages unfixed, and fixing them makes the2274 * hunk match. Update the context lines in the postimage.2275 */2276 fixed_buf = strbuf_detach(&fixed, &fixed_len);2277 update_pre_post_images(preimage, postimage,2278 fixed_buf, fixed_len, 0);2279 return 1;22802281 unmatch_exit:2282 strbuf_release(&fixed);2283 return 0;2284}22852286static int find_pos(struct image *img,2287 struct image *preimage,2288 struct image *postimage,2289 int line,2290 unsigned ws_rule,2291 int match_beginning, int match_end)2292{2293 int i;2294 unsigned long backwards, forwards, try;2295 int backwards_lno, forwards_lno, try_lno;22962297 /*2298 * If match_beginning or match_end is specified, there is no2299 * point starting from a wrong line that will never match and2300 * wander around and wait for a match at the specified end.2301 */2302 if (match_beginning)2303 line = 0;2304 else if (match_end)2305 line = img->nr - preimage->nr;23062307 /*2308 * Because the comparison is unsigned, the following test2309 * will also take care of a negative line number that can2310 * result when match_end and preimage is larger than the target.2311 */2312 if ((size_t) line > img->nr)2313 line = img->nr;23142315 try = 0;2316 for (i = 0; i < line; i++)2317 try += img->line[i].len;23182319 /*2320 * There's probably some smart way to do this, but I'll leave2321 * that to the smart and beautiful people. I'm simple and stupid.2322 */2323 backwards = try;2324 backwards_lno = line;2325 forwards = try;2326 forwards_lno = line;2327 try_lno = line;23282329 for (i = 0; ; i++) {2330 if (match_fragment(img, preimage, postimage,2331 try, try_lno, ws_rule,2332 match_beginning, match_end))2333 return try_lno;23342335 again:2336 if (backwards_lno == 0 && forwards_lno == img->nr)2337 break;23382339 if (i & 1) {2340 if (backwards_lno == 0) {2341 i++;2342 goto again;2343 }2344 backwards_lno--;2345 backwards -= img->line[backwards_lno].len;2346 try = backwards;2347 try_lno = backwards_lno;2348 } else {2349 if (forwards_lno == img->nr) {2350 i++;2351 goto again;2352 }2353 forwards += img->line[forwards_lno].len;2354 forwards_lno++;2355 try = forwards;2356 try_lno = forwards_lno;2357 }23582359 }2360 return -1;2361}23622363static void remove_first_line(struct image *img)2364{2365 img->buf += img->line[0].len;2366 img->len -= img->line[0].len;2367 img->line++;2368 img->nr--;2369}23702371static void remove_last_line(struct image *img)2372{2373 img->len -= img->line[--img->nr].len;2374}23752376static void update_image(struct image *img,2377 int applied_pos,2378 struct image *preimage,2379 struct image *postimage)2380{2381 /*2382 * remove the copy of preimage at offset in img2383 * and replace it with postimage2384 */2385 int i, nr;2386 size_t remove_count, insert_count, applied_at = 0;2387 char *result;2388 int preimage_limit;23892390 /*2391 * If we are removing blank lines at the end of img,2392 * the preimage may extend beyond the end.2393 * If that is the case, we must be careful only to2394 * remove the part of the preimage that falls within2395 * the boundaries of img. Initialize preimage_limit2396 * to the number of lines in the preimage that falls2397 * within the boundaries.2398 */2399 preimage_limit = preimage->nr;2400 if (preimage_limit > img->nr - applied_pos)2401 preimage_limit = img->nr - applied_pos;24022403 for (i = 0; i < applied_pos; i++)2404 applied_at += img->line[i].len;24052406 remove_count = 0;2407 for (i = 0; i < preimage_limit; i++)2408 remove_count += img->line[applied_pos + i].len;2409 insert_count = postimage->len;24102411 /* Adjust the contents */2412 result = xmalloc(img->len + insert_count - remove_count + 1);2413 memcpy(result, img->buf, applied_at);2414 memcpy(result + applied_at, postimage->buf, postimage->len);2415 memcpy(result + applied_at + postimage->len,2416 img->buf + (applied_at + remove_count),2417 img->len - (applied_at + remove_count));2418 free(img->buf);2419 img->buf = result;2420 img->len += insert_count - remove_count;2421 result[img->len] = '\0';24222423 /* Adjust the line table */2424 nr = img->nr + postimage->nr - preimage_limit;2425 if (preimage_limit < postimage->nr) {2426 /*2427 * NOTE: this knows that we never call remove_first_line()2428 * on anything other than pre/post image.2429 */2430 img->line = xrealloc(img->line, nr * sizeof(*img->line));2431 img->line_allocated = img->line;2432 }2433 if (preimage_limit != postimage->nr)2434 memmove(img->line + applied_pos + postimage->nr,2435 img->line + applied_pos + preimage_limit,2436 (img->nr - (applied_pos + preimage_limit)) *2437 sizeof(*img->line));2438 memcpy(img->line + applied_pos,2439 postimage->line,2440 postimage->nr * sizeof(*img->line));2441 if (!allow_overlap)2442 for (i = 0; i < postimage->nr; i++)2443 img->line[applied_pos + i].flag |= LINE_PATCHED;2444 img->nr = nr;2445}24462447static int apply_one_fragment(struct image *img, struct fragment *frag,2448 int inaccurate_eof, unsigned ws_rule,2449 int nth_fragment)2450{2451 int match_beginning, match_end;2452 const char *patch = frag->patch;2453 int size = frag->size;2454 char *old, *oldlines;2455 struct strbuf newlines;2456 int new_blank_lines_at_end = 0;2457 int found_new_blank_lines_at_end = 0;2458 int hunk_linenr = frag->linenr;2459 unsigned long leading, trailing;2460 int pos, applied_pos;2461 struct image preimage;2462 struct image postimage;24632464 memset(&preimage, 0, sizeof(preimage));2465 memset(&postimage, 0, sizeof(postimage));2466 oldlines = xmalloc(size);2467 strbuf_init(&newlines, size);24682469 old = oldlines;2470 while (size > 0) {2471 char first;2472 int len = linelen(patch, size);2473 int plen;2474 int added_blank_line = 0;2475 int is_blank_context = 0;2476 size_t start;24772478 if (!len)2479 break;24802481 /*2482 * "plen" is how much of the line we should use for2483 * the actual patch data. Normally we just remove the2484 * first character on the line, but if the line is2485 * followed by "\ No newline", then we also remove the2486 * last one (which is the newline, of course).2487 */2488 plen = len - 1;2489 if (len < size && patch[len] == '\\')2490 plen--;2491 first = *patch;2492 if (apply_in_reverse) {2493 if (first == '-')2494 first = '+';2495 else if (first == '+')2496 first = '-';2497 }24982499 switch (first) {2500 case '\n':2501 /* Newer GNU diff, empty context line */2502 if (plen < 0)2503 /* ... followed by '\No newline'; nothing */2504 break;2505 *old++ = '\n';2506 strbuf_addch(&newlines, '\n');2507 add_line_info(&preimage, "\n", 1, LINE_COMMON);2508 add_line_info(&postimage, "\n", 1, LINE_COMMON);2509 is_blank_context = 1;2510 break;2511 case ' ':2512 if (plen && (ws_rule & WS_BLANK_AT_EOF) &&2513 ws_blank_line(patch + 1, plen, ws_rule))2514 is_blank_context = 1;2515 case '-':2516 memcpy(old, patch + 1, plen);2517 add_line_info(&preimage, old, plen,2518 (first == ' ' ? LINE_COMMON : 0));2519 old += plen;2520 if (first == '-')2521 break;2522 /* Fall-through for ' ' */2523 case '+':2524 /* --no-add does not add new lines */2525 if (first == '+' && no_add)2526 break;25272528 start = newlines.len;2529 if (first != '+' ||2530 !whitespace_error ||2531 ws_error_action != correct_ws_error) {2532 strbuf_add(&newlines, patch + 1, plen);2533 }2534 else {2535 ws_fix_copy(&newlines, patch + 1, plen, ws_rule, &applied_after_fixing_ws);2536 }2537 add_line_info(&postimage, newlines.buf + start, newlines.len - start,2538 (first == '+' ? 0 : LINE_COMMON));2539 if (first == '+' &&2540 (ws_rule & WS_BLANK_AT_EOF) &&2541 ws_blank_line(patch + 1, plen, ws_rule))2542 added_blank_line = 1;2543 break;2544 case '@': case '\\':2545 /* Ignore it, we already handled it */2546 break;2547 default:2548 if (apply_verbosely)2549 error(_("invalid start of line: '%c'"), first);2550 return -1;2551 }2552 if (added_blank_line) {2553 if (!new_blank_lines_at_end)2554 found_new_blank_lines_at_end = hunk_linenr;2555 new_blank_lines_at_end++;2556 }2557 else if (is_blank_context)2558 ;2559 else2560 new_blank_lines_at_end = 0;2561 patch += len;2562 size -= len;2563 hunk_linenr++;2564 }2565 if (inaccurate_eof &&2566 old > oldlines && old[-1] == '\n' &&2567 newlines.len > 0 && newlines.buf[newlines.len - 1] == '\n') {2568 old--;2569 strbuf_setlen(&newlines, newlines.len - 1);2570 }25712572 leading = frag->leading;2573 trailing = frag->trailing;25742575 /*2576 * A hunk to change lines at the beginning would begin with2577 * @@ -1,L +N,M @@2578 * but we need to be careful. -U0 that inserts before the second2579 * line also has this pattern.2580 *2581 * And a hunk to add to an empty file would begin with2582 * @@ -0,0 +N,M @@2583 *2584 * In other words, a hunk that is (frag->oldpos <= 1) with or2585 * without leading context must match at the beginning.2586 */2587 match_beginning = (!frag->oldpos ||2588 (frag->oldpos == 1 && !unidiff_zero));25892590 /*2591 * A hunk without trailing lines must match at the end.2592 * However, we simply cannot tell if a hunk must match end2593 * from the lack of trailing lines if the patch was generated2594 * with unidiff without any context.2595 */2596 match_end = !unidiff_zero && !trailing;25972598 pos = frag->newpos ? (frag->newpos - 1) : 0;2599 preimage.buf = oldlines;2600 preimage.len = old - oldlines;2601 postimage.buf = newlines.buf;2602 postimage.len = newlines.len;2603 preimage.line = preimage.line_allocated;2604 postimage.line = postimage.line_allocated;26052606 for (;;) {26072608 applied_pos = find_pos(img, &preimage, &postimage, pos,2609 ws_rule, match_beginning, match_end);26102611 if (applied_pos >= 0)2612 break;26132614 /* Am I at my context limits? */2615 if ((leading <= p_context) && (trailing <= p_context))2616 break;2617 if (match_beginning || match_end) {2618 match_beginning = match_end = 0;2619 continue;2620 }26212622 /*2623 * Reduce the number of context lines; reduce both2624 * leading and trailing if they are equal otherwise2625 * just reduce the larger context.2626 */2627 if (leading >= trailing) {2628 remove_first_line(&preimage);2629 remove_first_line(&postimage);2630 pos--;2631 leading--;2632 }2633 if (trailing > leading) {2634 remove_last_line(&preimage);2635 remove_last_line(&postimage);2636 trailing--;2637 }2638 }26392640 if (applied_pos >= 0) {2641 if (new_blank_lines_at_end &&2642 preimage.nr + applied_pos >= img->nr &&2643 (ws_rule & WS_BLANK_AT_EOF) &&2644 ws_error_action != nowarn_ws_error) {2645 record_ws_error(WS_BLANK_AT_EOF, "+", 1,2646 found_new_blank_lines_at_end);2647 if (ws_error_action == correct_ws_error) {2648 while (new_blank_lines_at_end--)2649 remove_last_line(&postimage);2650 }2651 /*2652 * We would want to prevent write_out_results()2653 * from taking place in apply_patch() that follows2654 * the callchain led us here, which is:2655 * apply_patch->check_patch_list->check_patch->2656 * apply_data->apply_fragments->apply_one_fragment2657 */2658 if (ws_error_action == die_on_ws_error)2659 apply = 0;2660 }26612662 if (apply_verbosely && applied_pos != pos) {2663 int offset = applied_pos - pos;2664 if (apply_in_reverse)2665 offset = 0 - offset;2666 fprintf_ln(stderr,2667 Q_("Hunk #%d succeeded at %d (offset %d line).",2668 "Hunk #%d succeeded at %d (offset %d lines).",2669 offset),2670 nth_fragment, applied_pos + 1, offset);2671 }26722673 /*2674 * Warn if it was necessary to reduce the number2675 * of context lines.2676 */2677 if ((leading != frag->leading) ||2678 (trailing != frag->trailing))2679 fprintf_ln(stderr, _("Context reduced to (%ld/%ld)"2680 " to apply fragment at %d"),2681 leading, trailing, applied_pos+1);2682 update_image(img, applied_pos, &preimage, &postimage);2683 } else {2684 if (apply_verbosely)2685 error(_("while searching for:\n%.*s"),2686 (int)(old - oldlines), oldlines);2687 }26882689 free(oldlines);2690 strbuf_release(&newlines);2691 free(preimage.line_allocated);2692 free(postimage.line_allocated);26932694 return (applied_pos < 0);2695}26962697static int apply_binary_fragment(struct image *img, struct patch *patch)2698{2699 struct fragment *fragment = patch->fragments;2700 unsigned long len;2701 void *dst;27022703 if (!fragment)2704 return error(_("missing binary patch data for '%s'"),2705 patch->new_name ?2706 patch->new_name :2707 patch->old_name);27082709 /* Binary patch is irreversible without the optional second hunk */2710 if (apply_in_reverse) {2711 if (!fragment->next)2712 return error("cannot reverse-apply a binary patch "2713 "without the reverse hunk to '%s'",2714 patch->new_name2715 ? patch->new_name : patch->old_name);2716 fragment = fragment->next;2717 }2718 switch (fragment->binary_patch_method) {2719 case BINARY_DELTA_DEFLATED:2720 dst = patch_delta(img->buf, img->len, fragment->patch,2721 fragment->size, &len);2722 if (!dst)2723 return -1;2724 clear_image(img);2725 img->buf = dst;2726 img->len = len;2727 return 0;2728 case BINARY_LITERAL_DEFLATED:2729 clear_image(img);2730 img->len = fragment->size;2731 img->buf = xmalloc(img->len+1);2732 memcpy(img->buf, fragment->patch, img->len);2733 img->buf[img->len] = '\0';2734 return 0;2735 }2736 return -1;2737}27382739static int apply_binary(struct image *img, struct patch *patch)2740{2741 const char *name = patch->old_name ? patch->old_name : patch->new_name;2742 unsigned char sha1[20];27432744 /*2745 * For safety, we require patch index line to contain2746 * full 40-byte textual SHA1 for old and new, at least for now.2747 */2748 if (strlen(patch->old_sha1_prefix) != 40 ||2749 strlen(patch->new_sha1_prefix) != 40 ||2750 get_sha1_hex(patch->old_sha1_prefix, sha1) ||2751 get_sha1_hex(patch->new_sha1_prefix, sha1))2752 return error("cannot apply binary patch to '%s' "2753 "without full index line", name);27542755 if (patch->old_name) {2756 /*2757 * See if the old one matches what the patch2758 * applies to.2759 */2760 hash_sha1_file(img->buf, img->len, blob_type, sha1);2761 if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))2762 return error("the patch applies to '%s' (%s), "2763 "which does not match the "2764 "current contents.",2765 name, sha1_to_hex(sha1));2766 }2767 else {2768 /* Otherwise, the old one must be empty. */2769 if (img->len)2770 return error("the patch applies to an empty "2771 "'%s' but it is not empty", name);2772 }27732774 get_sha1_hex(patch->new_sha1_prefix, sha1);2775 if (is_null_sha1(sha1)) {2776 clear_image(img);2777 return 0; /* deletion patch */2778 }27792780 if (has_sha1_file(sha1)) {2781 /* We already have the postimage */2782 enum object_type type;2783 unsigned long size;2784 char *result;27852786 result = read_sha1_file(sha1, &type, &size);2787 if (!result)2788 return error("the necessary postimage %s for "2789 "'%s' cannot be read",2790 patch->new_sha1_prefix, name);2791 clear_image(img);2792 img->buf = result;2793 img->len = size;2794 } else {2795 /*2796 * We have verified buf matches the preimage;2797 * apply the patch data to it, which is stored2798 * in the patch->fragments->{patch,size}.2799 */2800 if (apply_binary_fragment(img, patch))2801 return error(_("binary patch does not apply to '%s'"),2802 name);28032804 /* verify that the result matches */2805 hash_sha1_file(img->buf, img->len, blob_type, sha1);2806 if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))2807 return error(_("binary patch to '%s' creates incorrect result (expecting %s, got %s)"),2808 name, patch->new_sha1_prefix, sha1_to_hex(sha1));2809 }28102811 return 0;2812}28132814static int apply_fragments(struct image *img, struct patch *patch)2815{2816 struct fragment *frag = patch->fragments;2817 const char *name = patch->old_name ? patch->old_name : patch->new_name;2818 unsigned ws_rule = patch->ws_rule;2819 unsigned inaccurate_eof = patch->inaccurate_eof;2820 int nth = 0;28212822 if (patch->is_binary)2823 return apply_binary(img, patch);28242825 while (frag) {2826 nth++;2827 if (apply_one_fragment(img, frag, inaccurate_eof, ws_rule, nth)) {2828 error(_("patch failed: %s:%ld"), name, frag->oldpos);2829 if (!apply_with_reject)2830 return -1;2831 frag->rejected = 1;2832 }2833 frag = frag->next;2834 }2835 return 0;2836}28372838static int read_file_or_gitlink(struct cache_entry *ce, struct strbuf *buf)2839{2840 if (!ce)2841 return 0;28422843 if (S_ISGITLINK(ce->ce_mode)) {2844 strbuf_grow(buf, 100);2845 strbuf_addf(buf, "Subproject commit %s\n", sha1_to_hex(ce->sha1));2846 } else {2847 enum object_type type;2848 unsigned long sz;2849 char *result;28502851 result = read_sha1_file(ce->sha1, &type, &sz);2852 if (!result)2853 return -1;2854 /* XXX read_sha1_file NUL-terminates */2855 strbuf_attach(buf, result, sz, sz + 1);2856 }2857 return 0;2858}28592860static struct patch *in_fn_table(const char *name)2861{2862 struct string_list_item *item;28632864 if (name == NULL)2865 return NULL;28662867 item = string_list_lookup(&fn_table, name);2868 if (item != NULL)2869 return (struct patch *)item->util;28702871 return NULL;2872}28732874/*2875 * item->util in the filename table records the status of the path.2876 * Usually it points at a patch (whose result records the contents2877 * of it after applying it), but it could be PATH_WAS_DELETED for a2878 * path that a previously applied patch has already removed.2879 */2880 #define PATH_TO_BE_DELETED ((struct patch *) -2)2881#define PATH_WAS_DELETED ((struct patch *) -1)28822883static int to_be_deleted(struct patch *patch)2884{2885 return patch == PATH_TO_BE_DELETED;2886}28872888static int was_deleted(struct patch *patch)2889{2890 return patch == PATH_WAS_DELETED;2891}28922893static void add_to_fn_table(struct patch *patch)2894{2895 struct string_list_item *item;28962897 /*2898 * Always add new_name unless patch is a deletion2899 * This should cover the cases for normal diffs,2900 * file creations and copies2901 */2902 if (patch->new_name != NULL) {2903 item = string_list_insert(&fn_table, patch->new_name);2904 item->util = patch;2905 }29062907 /*2908 * store a failure on rename/deletion cases because2909 * later chunks shouldn't patch old names2910 */2911 if ((patch->new_name == NULL) || (patch->is_rename)) {2912 item = string_list_insert(&fn_table, patch->old_name);2913 item->util = PATH_WAS_DELETED;2914 }2915}29162917static void prepare_fn_table(struct patch *patch)2918{2919 /*2920 * store information about incoming file deletion2921 */2922 while (patch) {2923 if ((patch->new_name == NULL) || (patch->is_rename)) {2924 struct string_list_item *item;2925 item = string_list_insert(&fn_table, patch->old_name);2926 item->util = PATH_TO_BE_DELETED;2927 }2928 patch = patch->next;2929 }2930}29312932static int apply_data(struct patch *patch, struct stat *st, struct cache_entry *ce)2933{2934 struct strbuf buf = STRBUF_INIT;2935 struct image image;2936 size_t len;2937 char *img;2938 struct patch *tpatch;29392940 if (!(patch->is_copy || patch->is_rename) &&2941 (tpatch = in_fn_table(patch->old_name)) != NULL && !to_be_deleted(tpatch)) {2942 if (was_deleted(tpatch)) {2943 return error(_("patch %s has been renamed/deleted"),2944 patch->old_name);2945 }2946 /* We have a patched copy in memory use that */2947 strbuf_add(&buf, tpatch->result, tpatch->resultsize);2948 } else if (cached) {2949 if (read_file_or_gitlink(ce, &buf))2950 return error(_("read of %s failed"), patch->old_name);2951 } else if (patch->old_name) {2952 if (S_ISGITLINK(patch->old_mode)) {2953 if (ce) {2954 read_file_or_gitlink(ce, &buf);2955 } else {2956 /*2957 * There is no way to apply subproject2958 * patch without looking at the index.2959 */2960 patch->fragments = NULL;2961 }2962 } else {2963 if (read_old_data(st, patch->old_name, &buf))2964 return error(_("read of %s failed"), patch->old_name);2965 }2966 }29672968 img = strbuf_detach(&buf, &len);2969 prepare_image(&image, img, len, !patch->is_binary);29702971 if (apply_fragments(&image, patch) < 0)2972 return -1; /* note with --reject this succeeds. */2973 patch->result = image.buf;2974 patch->resultsize = image.len;2975 add_to_fn_table(patch);2976 free(image.line_allocated);29772978 if (0 < patch->is_delete && patch->resultsize)2979 return error(_("removal patch leaves file contents"));29802981 return 0;2982}29832984static int check_to_create_blob(const char *new_name, int ok_if_exists)2985{2986 struct stat nst;2987 if (!lstat(new_name, &nst)) {2988 if (S_ISDIR(nst.st_mode) || ok_if_exists)2989 return 0;2990 /*2991 * A leading component of new_name might be a symlink2992 * that is going to be removed with this patch, but2993 * still pointing at somewhere that has the path.2994 * In such a case, path "new_name" does not exist as2995 * far as git is concerned.2996 */2997 if (has_symlink_leading_path(new_name, strlen(new_name)))2998 return 0;29993000 return error(_("%s: already exists in working directory"), new_name);3001 }3002 else if ((errno != ENOENT) && (errno != ENOTDIR))3003 return error("%s: %s", new_name, strerror(errno));3004 return 0;3005}30063007static int verify_index_match(struct cache_entry *ce, struct stat *st)3008{3009 if (S_ISGITLINK(ce->ce_mode)) {3010 if (!S_ISDIR(st->st_mode))3011 return -1;3012 return 0;3013 }3014 return ce_match_stat(ce, st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE);3015}30163017static int check_preimage(struct patch *patch, struct cache_entry **ce, struct stat *st)3018{3019 const char *old_name = patch->old_name;3020 struct patch *tpatch = NULL;3021 int stat_ret = 0;3022 unsigned st_mode = 0;30233024 /*3025 * Make sure that we do not have local modifications from the3026 * index when we are looking at the index. Also make sure3027 * we have the preimage file to be patched in the work tree,3028 * unless --cached, which tells git to apply only in the index.3029 */3030 if (!old_name)3031 return 0;30323033 assert(patch->is_new <= 0);30343035 if (!(patch->is_copy || patch->is_rename) &&3036 (tpatch = in_fn_table(old_name)) != NULL && !to_be_deleted(tpatch)) {3037 if (was_deleted(tpatch))3038 return error(_("%s: has been deleted/renamed"), old_name);3039 st_mode = tpatch->new_mode;3040 } else if (!cached) {3041 stat_ret = lstat(old_name, st);3042 if (stat_ret && errno != ENOENT)3043 return error(_("%s: %s"), old_name, strerror(errno));3044 }30453046 if (to_be_deleted(tpatch))3047 tpatch = NULL;30483049 if (check_index && !tpatch) {3050 int pos = cache_name_pos(old_name, strlen(old_name));3051 if (pos < 0) {3052 if (patch->is_new < 0)3053 goto is_new;3054 return error(_("%s: does not exist in index"), old_name);3055 }3056 *ce = active_cache[pos];3057 if (stat_ret < 0) {3058 struct checkout costate;3059 /* checkout */3060 memset(&costate, 0, sizeof(costate));3061 costate.base_dir = "";3062 costate.refresh_cache = 1;3063 if (checkout_entry(*ce, &costate, NULL) ||3064 lstat(old_name, st))3065 return -1;3066 }3067 if (!cached && verify_index_match(*ce, st))3068 return error(_("%s: does not match index"), old_name);3069 if (cached)3070 st_mode = (*ce)->ce_mode;3071 } else if (stat_ret < 0) {3072 if (patch->is_new < 0)3073 goto is_new;3074 return error(_("%s: %s"), old_name, strerror(errno));3075 }30763077 if (!cached && !tpatch)3078 st_mode = ce_mode_from_stat(*ce, st->st_mode);30793080 if (patch->is_new < 0)3081 patch->is_new = 0;3082 if (!patch->old_mode)3083 patch->old_mode = st_mode;3084 if ((st_mode ^ patch->old_mode) & S_IFMT)3085 return error(_("%s: wrong type"), old_name);3086 if (st_mode != patch->old_mode)3087 warning(_("%s has type %o, expected %o"),3088 old_name, st_mode, patch->old_mode);3089 if (!patch->new_mode && !patch->is_delete)3090 patch->new_mode = st_mode;3091 return 0;30923093 is_new:3094 patch->is_new = 1;3095 patch->is_delete = 0;3096 patch->old_name = NULL;3097 return 0;3098}30993100static int check_patch(struct patch *patch)3101{3102 struct stat st;3103 const char *old_name = patch->old_name;3104 const char *new_name = patch->new_name;3105 const char *name = old_name ? old_name : new_name;3106 struct cache_entry *ce = NULL;3107 struct patch *tpatch;3108 int ok_if_exists;3109 int status;31103111 patch->rejected = 1; /* we will drop this after we succeed */31123113 status = check_preimage(patch, &ce, &st);3114 if (status)3115 return status;3116 old_name = patch->old_name;31173118 if ((tpatch = in_fn_table(new_name)) &&3119 (was_deleted(tpatch) || to_be_deleted(tpatch)))3120 /*3121 * A type-change diff is always split into a patch to3122 * delete old, immediately followed by a patch to3123 * create new (see diff.c::run_diff()); in such a case3124 * it is Ok that the entry to be deleted by the3125 * previous patch is still in the working tree and in3126 * the index.3127 */3128 ok_if_exists = 1;3129 else3130 ok_if_exists = 0;31313132 if (new_name &&3133 ((0 < patch->is_new) | (0 < patch->is_rename) | patch->is_copy)) {3134 if (check_index &&3135 cache_name_pos(new_name, strlen(new_name)) >= 0 &&3136 !ok_if_exists)3137 return error(_("%s: already exists in index"), new_name);3138 if (!cached) {3139 int err = check_to_create_blob(new_name, ok_if_exists);3140 if (err)3141 return err;3142 }3143 if (!patch->new_mode) {3144 if (0 < patch->is_new)3145 patch->new_mode = S_IFREG | 0644;3146 else3147 patch->new_mode = patch->old_mode;3148 }3149 }31503151 if (new_name && old_name) {3152 int same = !strcmp(old_name, new_name);3153 if (!patch->new_mode)3154 patch->new_mode = patch->old_mode;3155 if ((patch->old_mode ^ patch->new_mode) & S_IFMT)3156 return error(_("new mode (%o) of %s does not match old mode (%o)%s%s"),3157 patch->new_mode, new_name, patch->old_mode,3158 same ? "" : " of ", same ? "" : old_name);3159 }31603161 if (apply_data(patch, &st, ce) < 0)3162 return error(_("%s: patch does not apply"), name);3163 patch->rejected = 0;3164 return 0;3165}31663167static int check_patch_list(struct patch *patch)3168{3169 int err = 0;31703171 prepare_fn_table(patch);3172 while (patch) {3173 if (apply_verbosely)3174 say_patch_name(stderr,3175 "Checking patch ", patch, "...\n");3176 err |= check_patch(patch);3177 patch = patch->next;3178 }3179 return err;3180}31813182/* This function tries to read the sha1 from the current index */3183static int get_current_sha1(const char *path, unsigned char *sha1)3184{3185 int pos;31863187 if (read_cache() < 0)3188 return -1;3189 pos = cache_name_pos(path, strlen(path));3190 if (pos < 0)3191 return -1;3192 hashcpy(sha1, active_cache[pos]->sha1);3193 return 0;3194}31953196/* Build an index that contains the just the files needed for a 3way merge */3197static void build_fake_ancestor(struct patch *list, const char *filename)3198{3199 struct patch *patch;3200 struct index_state result = { NULL };3201 int fd;32023203 /* Once we start supporting the reverse patch, it may be3204 * worth showing the new sha1 prefix, but until then...3205 */3206 for (patch = list; patch; patch = patch->next) {3207 const unsigned char *sha1_ptr;3208 unsigned char sha1[20];3209 struct cache_entry *ce;3210 const char *name;32113212 name = patch->old_name ? patch->old_name : patch->new_name;3213 if (0 < patch->is_new)3214 continue;3215 else if (get_sha1(patch->old_sha1_prefix, sha1))3216 /* git diff has no index line for mode/type changes */3217 if (!patch->lines_added && !patch->lines_deleted) {3218 if (get_current_sha1(patch->old_name, sha1))3219 die("mode change for %s, which is not "3220 "in current HEAD", name);3221 sha1_ptr = sha1;3222 } else3223 die("sha1 information is lacking or useless "3224 "(%s).", name);3225 else3226 sha1_ptr = sha1;32273228 ce = make_cache_entry(patch->old_mode, sha1_ptr, name, 0, 0);3229 if (!ce)3230 die(_("make_cache_entry failed for path '%s'"), name);3231 if (add_index_entry(&result, ce, ADD_CACHE_OK_TO_ADD))3232 die ("Could not add %s to temporary index", name);3233 }32343235 fd = open(filename, O_WRONLY | O_CREAT, 0666);3236 if (fd < 0 || write_index(&result, fd) || close(fd))3237 die ("Could not write temporary index to %s", filename);32383239 discard_index(&result);3240}32413242static void stat_patch_list(struct patch *patch)3243{3244 int files, adds, dels;32453246 for (files = adds = dels = 0 ; patch ; patch = patch->next) {3247 files++;3248 adds += patch->lines_added;3249 dels += patch->lines_deleted;3250 show_stats(patch);3251 }32523253 print_stat_summary(stdout, files, adds, dels);3254}32553256static void numstat_patch_list(struct patch *patch)3257{3258 for ( ; patch; patch = patch->next) {3259 const char *name;3260 name = patch->new_name ? patch->new_name : patch->old_name;3261 if (patch->is_binary)3262 printf("-\t-\t");3263 else3264 printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);3265 write_name_quoted(name, stdout, line_termination);3266 }3267}32683269static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)3270{3271 if (mode)3272 printf(" %s mode %06o %s\n", newdelete, mode, name);3273 else3274 printf(" %s %s\n", newdelete, name);3275}32763277static void show_mode_change(struct patch *p, int show_name)3278{3279 if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {3280 if (show_name)3281 printf(" mode change %06o => %06o %s\n",3282 p->old_mode, p->new_mode, p->new_name);3283 else3284 printf(" mode change %06o => %06o\n",3285 p->old_mode, p->new_mode);3286 }3287}32883289static void show_rename_copy(struct patch *p)3290{3291 const char *renamecopy = p->is_rename ? "rename" : "copy";3292 const char *old, *new;32933294 /* Find common prefix */3295 old = p->old_name;3296 new = p->new_name;3297 while (1) {3298 const char *slash_old, *slash_new;3299 slash_old = strchr(old, '/');3300 slash_new = strchr(new, '/');3301 if (!slash_old ||3302 !slash_new ||3303 slash_old - old != slash_new - new ||3304 memcmp(old, new, slash_new - new))3305 break;3306 old = slash_old + 1;3307 new = slash_new + 1;3308 }3309 /* p->old_name thru old is the common prefix, and old and new3310 * through the end of names are renames3311 */3312 if (old != p->old_name)3313 printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,3314 (int)(old - p->old_name), p->old_name,3315 old, new, p->score);3316 else3317 printf(" %s %s => %s (%d%%)\n", renamecopy,3318 p->old_name, p->new_name, p->score);3319 show_mode_change(p, 0);3320}33213322static void summary_patch_list(struct patch *patch)3323{3324 struct patch *p;33253326 for (p = patch; p; p = p->next) {3327 if (p->is_new)3328 show_file_mode_name("create", p->new_mode, p->new_name);3329 else if (p->is_delete)3330 show_file_mode_name("delete", p->old_mode, p->old_name);3331 else {3332 if (p->is_rename || p->is_copy)3333 show_rename_copy(p);3334 else {3335 if (p->score) {3336 printf(" rewrite %s (%d%%)\n",3337 p->new_name, p->score);3338 show_mode_change(p, 0);3339 }3340 else3341 show_mode_change(p, 1);3342 }3343 }3344 }3345}33463347static void patch_stats(struct patch *patch)3348{3349 int lines = patch->lines_added + patch->lines_deleted;33503351 if (lines > max_change)3352 max_change = lines;3353 if (patch->old_name) {3354 int len = quote_c_style(patch->old_name, NULL, NULL, 0);3355 if (!len)3356 len = strlen(patch->old_name);3357 if (len > max_len)3358 max_len = len;3359 }3360 if (patch->new_name) {3361 int len = quote_c_style(patch->new_name, NULL, NULL, 0);3362 if (!len)3363 len = strlen(patch->new_name);3364 if (len > max_len)3365 max_len = len;3366 }3367}33683369static void remove_file(struct patch *patch, int rmdir_empty)3370{3371 if (update_index) {3372 if (remove_file_from_cache(patch->old_name) < 0)3373 die(_("unable to remove %s from index"), patch->old_name);3374 }3375 if (!cached) {3376 if (!remove_or_warn(patch->old_mode, patch->old_name) && rmdir_empty) {3377 remove_path(patch->old_name);3378 }3379 }3380}33813382static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)3383{3384 struct stat st;3385 struct cache_entry *ce;3386 int namelen = strlen(path);3387 unsigned ce_size = cache_entry_size(namelen);33883389 if (!update_index)3390 return;33913392 ce = xcalloc(1, ce_size);3393 memcpy(ce->name, path, namelen);3394 ce->ce_mode = create_ce_mode(mode);3395 ce->ce_flags = namelen;3396 if (S_ISGITLINK(mode)) {3397 const char *s = buf;33983399 if (get_sha1_hex(s + strlen("Subproject commit "), ce->sha1))3400 die(_("corrupt patch for subproject %s"), path);3401 } else {3402 if (!cached) {3403 if (lstat(path, &st) < 0)3404 die_errno(_("unable to stat newly created file '%s'"),3405 path);3406 fill_stat_cache_info(ce, &st);3407 }3408 if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)3409 die(_("unable to create backing store for newly created file %s"), path);3410 }3411 if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)3412 die(_("unable to add cache entry for %s"), path);3413}34143415static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)3416{3417 int fd;3418 struct strbuf nbuf = STRBUF_INIT;34193420 if (S_ISGITLINK(mode)) {3421 struct stat st;3422 if (!lstat(path, &st) && S_ISDIR(st.st_mode))3423 return 0;3424 return mkdir(path, 0777);3425 }34263427 if (has_symlinks && S_ISLNK(mode))3428 /* Although buf:size is counted string, it also is NUL3429 * terminated.3430 */3431 return symlink(buf, path);34323433 fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);3434 if (fd < 0)3435 return -1;34363437 if (convert_to_working_tree(path, buf, size, &nbuf)) {3438 size = nbuf.len;3439 buf = nbuf.buf;3440 }3441 write_or_die(fd, buf, size);3442 strbuf_release(&nbuf);34433444 if (close(fd) < 0)3445 die_errno(_("closing file '%s'"), path);3446 return 0;3447}34483449/*3450 * We optimistically assume that the directories exist,3451 * which is true 99% of the time anyway. If they don't,3452 * we create them and try again.3453 */3454static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)3455{3456 if (cached)3457 return;3458 if (!try_create_file(path, mode, buf, size))3459 return;34603461 if (errno == ENOENT) {3462 if (safe_create_leading_directories(path))3463 return;3464 if (!try_create_file(path, mode, buf, size))3465 return;3466 }34673468 if (errno == EEXIST || errno == EACCES) {3469 /* We may be trying to create a file where a directory3470 * used to be.3471 */3472 struct stat st;3473 if (!lstat(path, &st) && (!S_ISDIR(st.st_mode) || !rmdir(path)))3474 errno = EEXIST;3475 }34763477 if (errno == EEXIST) {3478 unsigned int nr = getpid();34793480 for (;;) {3481 char newpath[PATH_MAX];3482 mksnpath(newpath, sizeof(newpath), "%s~%u", path, nr);3483 if (!try_create_file(newpath, mode, buf, size)) {3484 if (!rename(newpath, path))3485 return;3486 unlink_or_warn(newpath);3487 break;3488 }3489 if (errno != EEXIST)3490 break;3491 ++nr;3492 }3493 }3494 die_errno(_("unable to write file '%s' mode %o"), path, mode);3495}34963497static void create_file(struct patch *patch)3498{3499 char *path = patch->new_name;3500 unsigned mode = patch->new_mode;3501 unsigned long size = patch->resultsize;3502 char *buf = patch->result;35033504 if (!mode)3505 mode = S_IFREG | 0644;3506 create_one_file(path, mode, buf, size);3507 add_index_file(path, mode, buf, size);3508}35093510/* phase zero is to remove, phase one is to create */3511static void write_out_one_result(struct patch *patch, int phase)3512{3513 if (patch->is_delete > 0) {3514 if (phase == 0)3515 remove_file(patch, 1);3516 return;3517 }3518 if (patch->is_new > 0 || patch->is_copy) {3519 if (phase == 1)3520 create_file(patch);3521 return;3522 }3523 /*3524 * Rename or modification boils down to the same3525 * thing: remove the old, write the new3526 */3527 if (phase == 0)3528 remove_file(patch, patch->is_rename);3529 if (phase == 1)3530 create_file(patch);3531}35323533static int write_out_one_reject(struct patch *patch)3534{3535 FILE *rej;3536 char namebuf[PATH_MAX];3537 struct fragment *frag;3538 int cnt = 0;35393540 for (cnt = 0, frag = patch->fragments; frag; frag = frag->next) {3541 if (!frag->rejected)3542 continue;3543 cnt++;3544 }35453546 if (!cnt) {3547 if (apply_verbosely)3548 say_patch_name(stderr,3549 "Applied patch ", patch, " cleanly.\n");3550 return 0;3551 }35523553 /* This should not happen, because a removal patch that leaves3554 * contents are marked "rejected" at the patch level.3555 */3556 if (!patch->new_name)3557 die(_("internal error"));35583559 /* Say this even without --verbose */3560 say_patch_name(stderr, "Applying patch ", patch, " with");3561 fprintf(stderr, " %d rejects...\n", cnt);35623563 cnt = strlen(patch->new_name);3564 if (ARRAY_SIZE(namebuf) <= cnt + 5) {3565 cnt = ARRAY_SIZE(namebuf) - 5;3566 warning(_("truncating .rej filename to %.*s.rej"),3567 cnt - 1, patch->new_name);3568 }3569 memcpy(namebuf, patch->new_name, cnt);3570 memcpy(namebuf + cnt, ".rej", 5);35713572 rej = fopen(namebuf, "w");3573 if (!rej)3574 return error(_("cannot open %s: %s"), namebuf, strerror(errno));35753576 /* Normal git tools never deal with .rej, so do not pretend3577 * this is a git patch by saying --git nor give extended3578 * headers. While at it, maybe please "kompare" that wants3579 * the trailing TAB and some garbage at the end of line ;-).3580 */3581 fprintf(rej, "diff a/%s b/%s\t(rejected hunks)\n",3582 patch->new_name, patch->new_name);3583 for (cnt = 1, frag = patch->fragments;3584 frag;3585 cnt++, frag = frag->next) {3586 if (!frag->rejected) {3587 fprintf_ln(stderr, _("Hunk #%d applied cleanly."), cnt);3588 continue;3589 }3590 fprintf_ln(stderr, _("Rejected hunk #%d."), cnt);3591 fprintf(rej, "%.*s", frag->size, frag->patch);3592 if (frag->patch[frag->size-1] != '\n')3593 fputc('\n', rej);3594 }3595 fclose(rej);3596 return -1;3597}35983599static int write_out_results(struct patch *list)3600{3601 int phase;3602 int errs = 0;3603 struct patch *l;36043605 for (phase = 0; phase < 2; phase++) {3606 l = list;3607 while (l) {3608 if (l->rejected)3609 errs = 1;3610 else {3611 write_out_one_result(l, phase);3612 if (phase == 1 && write_out_one_reject(l))3613 errs = 1;3614 }3615 l = l->next;3616 }3617 }3618 return errs;3619}36203621static struct lock_file lock_file;36223623static struct string_list limit_by_name;3624static int has_include;3625static void add_name_limit(const char *name, int exclude)3626{3627 struct string_list_item *it;36283629 it = string_list_append(&limit_by_name, name);3630 it->util = exclude ? NULL : (void *) 1;3631}36323633static int use_patch(struct patch *p)3634{3635 const char *pathname = p->new_name ? p->new_name : p->old_name;3636 int i;36373638 /* Paths outside are not touched regardless of "--include" */3639 if (0 < prefix_length) {3640 int pathlen = strlen(pathname);3641 if (pathlen <= prefix_length ||3642 memcmp(prefix, pathname, prefix_length))3643 return 0;3644 }36453646 /* See if it matches any of exclude/include rule */3647 for (i = 0; i < limit_by_name.nr; i++) {3648 struct string_list_item *it = &limit_by_name.items[i];3649 if (!fnmatch(it->string, pathname, 0))3650 return (it->util != NULL);3651 }36523653 /*3654 * If we had any include, a path that does not match any rule is3655 * not used. Otherwise, we saw bunch of exclude rules (or none)3656 * and such a path is used.3657 */3658 return !has_include;3659}366036613662static void prefix_one(char **name)3663{3664 char *old_name = *name;3665 if (!old_name)3666 return;3667 *name = xstrdup(prefix_filename(prefix, prefix_length, *name));3668 free(old_name);3669}36703671static void prefix_patches(struct patch *p)3672{3673 if (!prefix || p->is_toplevel_relative)3674 return;3675 for ( ; p; p = p->next) {3676 if (p->new_name == p->old_name) {3677 char *prefixed = p->new_name;3678 prefix_one(&prefixed);3679 p->new_name = p->old_name = prefixed;3680 }3681 else {3682 prefix_one(&p->new_name);3683 prefix_one(&p->old_name);3684 }3685 }3686}36873688#define INACCURATE_EOF (1<<0)3689#define RECOUNT (1<<1)36903691static int apply_patch(int fd, const char *filename, int options)3692{3693 size_t offset;3694 struct strbuf buf = STRBUF_INIT;3695 struct patch *list = NULL, **listp = &list;3696 int skipped_patch = 0;36973698 /* FIXME - memory leak when using multiple patch files as inputs */3699 memset(&fn_table, 0, sizeof(struct string_list));3700 patch_input_file = filename;3701 read_patch_file(&buf, fd);3702 offset = 0;3703 while (offset < buf.len) {3704 struct patch *patch;3705 int nr;37063707 patch = xcalloc(1, sizeof(*patch));3708 patch->inaccurate_eof = !!(options & INACCURATE_EOF);3709 patch->recount = !!(options & RECOUNT);3710 nr = parse_chunk(buf.buf + offset, buf.len - offset, patch);3711 if (nr < 0)3712 break;3713 if (apply_in_reverse)3714 reverse_patches(patch);3715 if (prefix)3716 prefix_patches(patch);3717 if (use_patch(patch)) {3718 patch_stats(patch);3719 *listp = patch;3720 listp = &patch->next;3721 }3722 else {3723 /* perhaps free it a bit better? */3724 free(patch);3725 skipped_patch++;3726 }3727 offset += nr;3728 }37293730 if (!list && !skipped_patch)3731 die(_("unrecognized input"));37323733 if (whitespace_error && (ws_error_action == die_on_ws_error))3734 apply = 0;37353736 update_index = check_index && apply;3737 if (update_index && newfd < 0)3738 newfd = hold_locked_index(&lock_file, 1);37393740 if (check_index) {3741 if (read_cache() < 0)3742 die(_("unable to read index file"));3743 }37443745 if ((check || apply) &&3746 check_patch_list(list) < 0 &&3747 !apply_with_reject)3748 exit(1);37493750 if (apply && write_out_results(list))3751 exit(1);37523753 if (fake_ancestor)3754 build_fake_ancestor(list, fake_ancestor);37553756 if (diffstat)3757 stat_patch_list(list);37583759 if (numstat)3760 numstat_patch_list(list);37613762 if (summary)3763 summary_patch_list(list);37643765 strbuf_release(&buf);3766 return 0;3767}37683769static int git_apply_config(const char *var, const char *value, void *cb)3770{3771 if (!strcmp(var, "apply.whitespace"))3772 return git_config_string(&apply_default_whitespace, var, value);3773 else if (!strcmp(var, "apply.ignorewhitespace"))3774 return git_config_string(&apply_default_ignorewhitespace, var, value);3775 return git_default_config(var, value, cb);3776}37773778static int option_parse_exclude(const struct option *opt,3779 const char *arg, int unset)3780{3781 add_name_limit(arg, 1);3782 return 0;3783}37843785static int option_parse_include(const struct option *opt,3786 const char *arg, int unset)3787{3788 add_name_limit(arg, 0);3789 has_include = 1;3790 return 0;3791}37923793static int option_parse_p(const struct option *opt,3794 const char *arg, int unset)3795{3796 p_value = atoi(arg);3797 p_value_known = 1;3798 return 0;3799}38003801static int option_parse_z(const struct option *opt,3802 const char *arg, int unset)3803{3804 if (unset)3805 line_termination = '\n';3806 else3807 line_termination = 0;3808 return 0;3809}38103811static int option_parse_space_change(const struct option *opt,3812 const char *arg, int unset)3813{3814 if (unset)3815 ws_ignore_action = ignore_ws_none;3816 else3817 ws_ignore_action = ignore_ws_change;3818 return 0;3819}38203821static int option_parse_whitespace(const struct option *opt,3822 const char *arg, int unset)3823{3824 const char **whitespace_option = opt->value;38253826 *whitespace_option = arg;3827 parse_whitespace_option(arg);3828 return 0;3829}38303831static int option_parse_directory(const struct option *opt,3832 const char *arg, int unset)3833{3834 root_len = strlen(arg);3835 if (root_len && arg[root_len - 1] != '/') {3836 char *new_root;3837 root = new_root = xmalloc(root_len + 2);3838 strcpy(new_root, arg);3839 strcpy(new_root + root_len++, "/");3840 } else3841 root = arg;3842 return 0;3843}38443845int cmd_apply(int argc, const char **argv, const char *prefix_)3846{3847 int i;3848 int errs = 0;3849 int is_not_gitdir = !startup_info->have_repository;3850 int force_apply = 0;38513852 const char *whitespace_option = NULL;38533854 struct option builtin_apply_options[] = {3855 { OPTION_CALLBACK, 0, "exclude", NULL, "path",3856 "don't apply changes matching the given path",3857 0, option_parse_exclude },3858 { OPTION_CALLBACK, 0, "include", NULL, "path",3859 "apply changes matching the given path",3860 0, option_parse_include },3861 { OPTION_CALLBACK, 'p', NULL, NULL, "num",3862 "remove <num> leading slashes from traditional diff paths",3863 0, option_parse_p },3864 OPT_BOOLEAN(0, "no-add", &no_add,3865 "ignore additions made by the patch"),3866 OPT_BOOLEAN(0, "stat", &diffstat,3867 "instead of applying the patch, output diffstat for the input"),3868 OPT_NOOP_NOARG(0, "allow-binary-replacement"),3869 OPT_NOOP_NOARG(0, "binary"),3870 OPT_BOOLEAN(0, "numstat", &numstat,3871 "shows number of added and deleted lines in decimal notation"),3872 OPT_BOOLEAN(0, "summary", &summary,3873 "instead of applying the patch, output a summary for the input"),3874 OPT_BOOLEAN(0, "check", &check,3875 "instead of applying the patch, see if the patch is applicable"),3876 OPT_BOOLEAN(0, "index", &check_index,3877 "make sure the patch is applicable to the current index"),3878 OPT_BOOLEAN(0, "cached", &cached,3879 "apply a patch without touching the working tree"),3880 OPT_BOOLEAN(0, "apply", &force_apply,3881 "also apply the patch (use with --stat/--summary/--check)"),3882 OPT_FILENAME(0, "build-fake-ancestor", &fake_ancestor,3883 "build a temporary index based on embedded index information"),3884 { OPTION_CALLBACK, 'z', NULL, NULL, NULL,3885 "paths are separated with NUL character",3886 PARSE_OPT_NOARG, option_parse_z },3887 OPT_INTEGER('C', NULL, &p_context,3888 "ensure at least <n> lines of context match"),3889 { OPTION_CALLBACK, 0, "whitespace", &whitespace_option, "action",3890 "detect new or modified lines that have whitespace errors",3891 0, option_parse_whitespace },3892 { OPTION_CALLBACK, 0, "ignore-space-change", NULL, NULL,3893 "ignore changes in whitespace when finding context",3894 PARSE_OPT_NOARG, option_parse_space_change },3895 { OPTION_CALLBACK, 0, "ignore-whitespace", NULL, NULL,3896 "ignore changes in whitespace when finding context",3897 PARSE_OPT_NOARG, option_parse_space_change },3898 OPT_BOOLEAN('R', "reverse", &apply_in_reverse,3899 "apply the patch in reverse"),3900 OPT_BOOLEAN(0, "unidiff-zero", &unidiff_zero,3901 "don't expect at least one line of context"),3902 OPT_BOOLEAN(0, "reject", &apply_with_reject,3903 "leave the rejected hunks in corresponding *.rej files"),3904 OPT_BOOLEAN(0, "allow-overlap", &allow_overlap,3905 "allow overlapping hunks"),3906 OPT__VERBOSE(&apply_verbosely, "be verbose"),3907 OPT_BIT(0, "inaccurate-eof", &options,3908 "tolerate incorrectly detected missing new-line at the end of file",3909 INACCURATE_EOF),3910 OPT_BIT(0, "recount", &options,3911 "do not trust the line counts in the hunk headers",3912 RECOUNT),3913 { OPTION_CALLBACK, 0, "directory", NULL, "root",3914 "prepend <root> to all filenames",3915 0, option_parse_directory },3916 OPT_END()3917 };39183919 prefix = prefix_;3920 prefix_length = prefix ? strlen(prefix) : 0;3921 git_config(git_apply_config, NULL);3922 if (apply_default_whitespace)3923 parse_whitespace_option(apply_default_whitespace);3924 if (apply_default_ignorewhitespace)3925 parse_ignorewhitespace_option(apply_default_ignorewhitespace);39263927 argc = parse_options(argc, argv, prefix, builtin_apply_options,3928 apply_usage, 0);39293930 if (apply_with_reject)3931 apply = apply_verbosely = 1;3932 if (!force_apply && (diffstat || numstat || summary || check || fake_ancestor))3933 apply = 0;3934 if (check_index && is_not_gitdir)3935 die(_("--index outside a repository"));3936 if (cached) {3937 if (is_not_gitdir)3938 die(_("--cached outside a repository"));3939 check_index = 1;3940 }3941 for (i = 0; i < argc; i++) {3942 const char *arg = argv[i];3943 int fd;39443945 if (!strcmp(arg, "-")) {3946 errs |= apply_patch(0, "<stdin>", options);3947 read_stdin = 0;3948 continue;3949 } else if (0 < prefix_length)3950 arg = prefix_filename(prefix, prefix_length, arg);39513952 fd = open(arg, O_RDONLY);3953 if (fd < 0)3954 die_errno(_("can't open patch '%s'"), arg);3955 read_stdin = 0;3956 set_default_whitespace_mode(whitespace_option);3957 errs |= apply_patch(fd, arg, options);3958 close(fd);3959 }3960 set_default_whitespace_mode(whitespace_option);3961 if (read_stdin)3962 errs |= apply_patch(0, "<stdin>", options);3963 if (whitespace_error) {3964 if (squelch_whitespace_errors &&3965 squelch_whitespace_errors < whitespace_error) {3966 int squelched =3967 whitespace_error - squelch_whitespace_errors;3968 warning(Q_("squelched %d whitespace error",3969 "squelched %d whitespace errors",3970 squelched),3971 squelched);3972 }3973 if (ws_error_action == die_on_ws_error)3974 die(Q_("%d line adds whitespace errors.",3975 "%d lines add whitespace errors.",3976 whitespace_error),3977 whitespace_error);3978 if (applied_after_fixing_ws && apply)3979 warning("%d line%s applied after"3980 " fixing whitespace errors.",3981 applied_after_fixing_ws,3982 applied_after_fixing_ws == 1 ? "" : "s");3983 else if (whitespace_error)3984 warning(Q_("%d line adds whitespace errors.",3985 "%d lines add whitespace errors.",3986 whitespace_error),3987 whitespace_error);3988 }39893990 if (update_index) {3991 if (write_cache(newfd, active_cache, active_nr) ||3992 commit_locked_index(&lock_file))3993 die(_("Unable to write new index file"));3994 }39953996 return !!errs;3997}