apply.c 112.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * apply.c
 *
 * Copyright (C) Linus Torvalds, 2005
 *
 * This applies patches on top of some (arbitrary) version of the SCM.
 *
 */
#include "cache.h"
10
#include "cache-tree.h"
11
#include "quote.h"
12
#include "blob.h"
J
Junio C Hamano 已提交
13
#include "delta.h"
P
Peter Eriksen 已提交
14
#include "builtin.h"
15
#include "string-list.h"
16
#include "dir.h"
17
#include "diff.h"
M
Miklos Vajna 已提交
18
#include "parse-options.h"
19 20
#include "xdiff-interface.h"
#include "ll-merge.h"
21
#include "rerere.h"
L
Linus Torvalds 已提交
22

23 24 25 26 27 28 29 30 31
/*
 *  --check turns on checking that the working tree matches the
 *    files that are being modified, but doesn't apply the patch
 *  --stat does just a diffstat, and doesn't actually apply
 *  --numstat does numeric diffstat, and doesn't actually apply
 *  --index-info shows the old and new index info for paths if available.
 *  --index updates the cache as well.
 *  --cached updates only the cache without ever touching the working tree.
 */
32 33
static const char *prefix;
static int prefix_length = -1;
34
static int newfd = -1;
35

36
static int unidiff_zero;
37
static int p_value = 1;
38
static int p_value_known;
39
static int check_index;
40
static int update_index;
41 42 43 44 45
static int cached;
static int diffstat;
static int numstat;
static int summary;
static int check;
46
static int apply = 1;
47
static int apply_in_reverse;
J
Junio C Hamano 已提交
48
static int apply_with_reject;
J
Junio C Hamano 已提交
49
static int apply_verbosely;
50
static int allow_overlap;
51
static int no_add;
52
static int threeway;
53
static const char *fake_ancestor;
54
static int line_termination = '\n';
M
Miklos Vajna 已提交
55 56
static unsigned int p_context = UINT_MAX;
static const char * const apply_usage[] = {
57
	N_("git apply [options] [<patch>...]"),
M
Miklos Vajna 已提交
58 59
	NULL
};
L
Linus Torvalds 已提交
60

61 62 63 64
static enum ws_error_action {
	nowarn_ws_error,
	warn_on_ws_error,
	die_on_ws_error,
65
	correct_ws_error
66
} ws_error_action = warn_on_ws_error;
67
static int whitespace_error;
68
static int squelch_whitespace_errors = 5;
69
static int applied_after_fixing_ws;
70 71 72

static enum ws_ignore {
	ignore_ws_none,
73
	ignore_ws_change
74 75 76
} ws_ignore_action = ignore_ws_none;


77
static const char *patch_input_file;
78 79
static const char *root;
static int root_len;
M
Miklos Vajna 已提交
80 81
static int read_stdin = 1;
static int options;
L
Linus Torvalds 已提交
82

83 84 85
static void parse_whitespace_option(const char *option)
{
	if (!option) {
86
		ws_error_action = warn_on_ws_error;
87 88 89
		return;
	}
	if (!strcmp(option, "warn")) {
90
		ws_error_action = warn_on_ws_error;
91 92
		return;
	}
J
Junio C Hamano 已提交
93
	if (!strcmp(option, "nowarn")) {
94
		ws_error_action = nowarn_ws_error;
J
Junio C Hamano 已提交
95 96
		return;
	}
97
	if (!strcmp(option, "error")) {
98
		ws_error_action = die_on_ws_error;
99 100 101
		return;
	}
	if (!strcmp(option, "error-all")) {
102
		ws_error_action = die_on_ws_error;
103 104 105
		squelch_whitespace_errors = 0;
		return;
	}
106 107
	if (!strcmp(option, "strip") || !strcmp(option, "fix")) {
		ws_error_action = correct_ws_error;
108 109
		return;
	}
110
	die(_("unrecognized whitespace option '%s'"), option);
111 112
}

113 114 115 116 117 118 119 120 121 122 123 124
static void parse_ignorewhitespace_option(const char *option)
{
	if (!option || !strcmp(option, "no") ||
	    !strcmp(option, "false") || !strcmp(option, "never") ||
	    !strcmp(option, "none")) {
		ws_ignore_action = ignore_ws_none;
		return;
	}
	if (!strcmp(option, "change")) {
		ws_ignore_action = ignore_ws_change;
		return;
	}
125
	die(_("unrecognized whitespace ignore option '%s'"), option);
126 127
}

128 129
static void set_default_whitespace_mode(const char *whitespace_option)
{
130 131
	if (!whitespace_option && !apply_default_whitespace)
		ws_error_action = (apply ? warn_on_ws_error : nowarn_ws_error);
132 133
}

134 135 136 137 138 139 140
/*
 * For "diff-stat" like behaviour, we keep track of the biggest change
 * we've seen, and the longest filename. That allows us to do simple
 * scaling.
 */
static int max_change, max_len;

141 142 143 144 145
/*
 * Various "current state", notably line numbers and what
 * file (and how) we're patching right now.. The "is_xxxx"
 * things are flags, where -1 means "don't know yet".
 */
146
static int linenr = 1;
147

148 149 150 151 152 153 154
/*
 * This represents one "hunk" from a patch, starting with
 * "@@ -oldpos,oldlines +newpos,newlines @@" marker.  The
 * patch text is pointed at by patch, and its byte length
 * is stored in size.  leading and trailing are the number
 * of context lines.
 */
155
struct fragment {
156
	unsigned long leading, trailing;
157 158
	unsigned long oldpos, oldlines;
	unsigned long newpos, newlines;
159 160 161 162
	/*
	 * 'patch' is usually borrowed from buf in apply_patch(),
	 * but some codepaths store an allocated buffer.
	 */
163
	const char *patch;
164 165
	unsigned free_patch:1,
		rejected:1;
166
	int size;
167
	int linenr;
168 169 170
	struct fragment *next;
};

171 172 173 174 175 176 177 178 179
/*
 * When dealing with a binary patch, we reuse "leading" field
 * to store the type of the binary hunk, either deflated "delta"
 * or deflated "literal".
 */
#define binary_patch_method leading
#define BINARY_DELTA_DEFLATED	1
#define BINARY_LITERAL_DEFLATED 2

180 181 182 183 184
/*
 * This represents a "patch" to a file, both metainfo changes
 * such as creation/deletion, filemode and content changes represented
 * as a series of fragments.
 */
185
struct patch {
186
	char *new_name, *old_name, *def_name;
187
	unsigned int old_mode, new_mode;
188
	int is_new, is_delete;	/* -1 = unknown, 0 = false, 1 = true */
J
Junio C Hamano 已提交
189
	int rejected;
190
	unsigned ws_rule;
191
	int lines_added, lines_deleted;
192
	int score;
193
	unsigned int is_toplevel_relative:1;
194 195 196 197
	unsigned int inaccurate_eof:1;
	unsigned int is_binary:1;
	unsigned int is_copy:1;
	unsigned int is_rename:1;
198
	unsigned int recount:1;
199
	unsigned int conflicted_threeway:1;
200
	unsigned int direct_to_threeway:1;
201
	struct fragment *fragments;
202
	char *result;
203
	size_t resultsize;
204 205
	char old_sha1_prefix[41];
	char new_sha1_prefix[41];
206
	struct patch *next;
207 208 209

	/* three-way fallback result */
	unsigned char threeway_stage[3][20];
210
};
211

212
static void free_fragment_list(struct fragment *list)
213
{
214 215 216 217 218 219
	while (list) {
		struct fragment *next = list->next;
		if (list->free_patch)
			free((char *)list->patch);
		free(list);
		list = next;
220
	}
221 222 223 224 225
}

static void free_patch(struct patch *patch)
{
	free_fragment_list(patch->fragments);
226 227 228
	free(patch->def_name);
	free(patch->old_name);
	free(patch->new_name);
J
Junio C Hamano 已提交
229
	free(patch->result);
230 231 232 233 234 235 236 237 238
	free(patch);
}

static void free_patch_list(struct patch *list)
{
	while (list) {
		struct patch *next = list->next;
		free_patch(list);
		list = next;
239 240 241
	}
}

242 243 244 245 246 247 248 249 250
/*
 * A line in a file, len-bytes long (includes the terminating LF,
 * except for an incomplete line at the end if the file ends with
 * one), and its contents hashes to 'hash'.
 */
struct line {
	size_t len;
	unsigned hash : 24;
	unsigned flag : 8;
251
#define LINE_COMMON     1
252
#define LINE_PATCHED	2
253 254 255 256 257 258 259 260 261
};

/*
 * This represents a "file", which is an array of "lines".
 */
struct image {
	char *buf;
	size_t len;
	size_t nr;
262
	size_t alloc;
263 264 265 266
	struct line *line_allocated;
	struct line *line;
};

267 268 269 270 271
/*
 * Records filenames that have been touched, in order to handle
 * the case where more than one patches touch the same file.
 */

272
static struct string_list fn_table;
273

274 275 276 277 278 279 280 281 282 283 284 285
static uint32_t hash_line(const char *cp, size_t len)
{
	size_t i;
	uint32_t h;
	for (i = 0, h = 0; i < len; i++) {
		if (!isspace(cp[i])) {
			h = h * 3 + (cp[i] & 0xff);
		}
	}
	return h;
}

286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338
/*
 * Compare lines s1 of length n1 and s2 of length n2, ignoring
 * whitespace difference. Returns 1 if they match, 0 otherwise
 */
static int fuzzy_matchlines(const char *s1, size_t n1,
			    const char *s2, size_t n2)
{
	const char *last1 = s1 + n1 - 1;
	const char *last2 = s2 + n2 - 1;
	int result = 0;

	/* ignore line endings */
	while ((*last1 == '\r') || (*last1 == '\n'))
		last1--;
	while ((*last2 == '\r') || (*last2 == '\n'))
		last2--;

	/* skip leading whitespace */
	while (isspace(*s1) && (s1 <= last1))
		s1++;
	while (isspace(*s2) && (s2 <= last2))
		s2++;
	/* early return if both lines are empty */
	if ((s1 > last1) && (s2 > last2))
		return 1;
	while (!result) {
		result = *s1++ - *s2++;
		/*
		 * Skip whitespace inside. We check for whitespace on
		 * both buffers because we don't want "a b" to match
		 * "ab"
		 */
		if (isspace(*s1) && isspace(*s2)) {
			while (isspace(*s1) && s1 <= last1)
				s1++;
			while (isspace(*s2) && s2 <= last2)
				s2++;
		}
		/*
		 * If we reached the end on one side only,
		 * lines don't match
		 */
		if (
		    ((s2 > last2) && (s1 <= last1)) ||
		    ((s1 > last1) && (s2 <= last2)))
			return 0;
		if ((s1 > last1) && (s2 > last2))
			break;
	}

	return !result;
}

339 340 341 342 343 344 345 346 347
static void add_line_info(struct image *img, const char *bol, size_t len, unsigned flag)
{
	ALLOC_GROW(img->line_allocated, img->nr + 1, img->alloc);
	img->line_allocated[img->nr].len = len;
	img->line_allocated[img->nr].hash = hash_line(bol, len);
	img->line_allocated[img->nr].flag = flag;
	img->nr++;
}

348 349 350 351 352
/*
 * "buf" has the file contents to be patched (read from various sources).
 * attach it to "image" and add line-based index to it.
 * "image" now owns the "buf".
 */
353 354 355 356 357
static void prepare_image(struct image *image, char *buf, size_t len,
			  int prepare_linetable)
{
	const char *cp, *ep;

358
	memset(image, 0, sizeof(*image));
359 360 361
	image->buf = buf;
	image->len = len;

362
	if (!prepare_linetable)
363 364 365 366 367 368 369 370 371 372
		return;

	ep = image->buf + image->len;
	cp = image->buf;
	while (cp < ep) {
		const char *next;
		for (next = cp; next < ep && *next != '\n'; next++)
			;
		if (next < ep)
			next++;
373
		add_line_info(image, cp, next - cp, 0);
374 375
		cp = next;
	}
376
	image->line = image->line_allocated;
377 378 379 380 381
}

static void clear_image(struct image *image)
{
	free(image->buf);
382 383
	free(image->line_allocated);
	memset(image, 0, sizeof(*image));
384 385
}

386 387
/* fmt must contain _one_ %s and no other substitution */
static void say_patch_name(FILE *output, const char *fmt, struct patch *patch)
J
Junio C Hamano 已提交
388
{
389 390
	struct strbuf sb = STRBUF_INIT;

J
Junio C Hamano 已提交
391 392
	if (patch->old_name && patch->new_name &&
	    strcmp(patch->old_name, patch->new_name)) {
393 394 395
		quote_c_style(patch->old_name, &sb, NULL, 0);
		strbuf_addstr(&sb, " => ");
		quote_c_style(patch->new_name, &sb, NULL, 0);
396
	} else {
J
Junio C Hamano 已提交
397 398 399
		const char *n = patch->new_name;
		if (!n)
			n = patch->old_name;
400
		quote_c_style(n, &sb, NULL, 0);
J
Junio C Hamano 已提交
401
	}
402 403 404
	fprintf(output, fmt, sb.buf);
	fputc('\n', output);
	strbuf_release(&sb);
J
Junio C Hamano 已提交
405 406
}

407
#define SLOP (16)
L
Linus Torvalds 已提交
408

409
static void read_patch_file(struct strbuf *sb, int fd)
L
Linus Torvalds 已提交
410
{
411
	if (strbuf_read(sb, fd, 0) < 0)
412
		die_errno("git apply: failed to read");
413 414 415 416 417 418

	/*
	 * Make sure that we have some slop in the buffer
	 * so that we can do speculative "memcmp" etc, and
	 * see to it that it is NUL-filled.
	 */
419 420
	strbuf_grow(sb, SLOP);
	memset(sb->buf + sb->len, 0, SLOP);
L
Linus Torvalds 已提交
421 422
}

423
static unsigned long linelen(const char *buffer, unsigned long size)
L
Linus Torvalds 已提交
424 425 426 427 428 429 430 431 432 433
{
	unsigned long len = 0;
	while (size--) {
		len++;
		if (*buffer++ == '\n')
			break;
	}
	return len;
}

434 435 436 437 438
static int is_dev_null(const char *str)
{
	return !memcmp("/dev/null", str, 9) && isspace(str[9]);
}

439 440
#define TERM_SPACE	1
#define TERM_TAB	2
441 442 443 444 445 446 447 448 449 450 451

static int name_terminate(const char *name, int namelen, int c, int terminate)
{
	if (c == ' ' && !(terminate & TERM_SPACE))
		return 0;
	if (c == '\t' && !(terminate & TERM_TAB))
		return 0;

	return 1;
}

452 453 454 455 456
/* remove double slashes to make --index work with such filenames */
static char *squash_slash(char *name)
{
	int i = 0, j = 0;

457 458 459
	if (!name)
		return NULL;

460 461 462 463 464 465 466 467 468
	while (name[i]) {
		if ((name[j++] = name[i++]) == '/')
			while (name[i] == '/')
				i++;
	}
	name[j] = '\0';
	return name;
}

469
static char *find_name_gnu(const char *line, const char *def, int p_value)
L
Linus Torvalds 已提交
470
{
471 472
	struct strbuf name = STRBUF_INIT;
	char *cp;
473

474 475 476 477 478 479 480 481
	/*
	 * Proposed "new-style" GNU patch/diff format; see
	 * http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
	 */
	if (unquote_c_style(&name, line, NULL)) {
		strbuf_release(&name);
		return NULL;
	}
482

483 484 485 486 487
	for (cp = name.buf; p_value; p_value--) {
		cp = strchr(cp, '/');
		if (!cp) {
			strbuf_release(&name);
			return NULL;
488
		}
489 490 491 492 493 494 495 496 497
		cp++;
	}

	strbuf_remove(&name, 0, cp - name.buf);
	if (root)
		strbuf_insert(&name, 0, root, root_len);
	return squash_slash(strbuf_detach(&name, NULL));
}

498
static size_t sane_tz_len(const char *line, size_t len)
L
Linus Torvalds 已提交
499
{
500
	const char *tz, *p;
501

502 503 504 505 506 507 508 509 510 511 512 513 514 515
	if (len < strlen(" +0500") || line[len-strlen(" +0500")] != ' ')
		return 0;
	tz = line + len - strlen(" +0500");

	if (tz[1] != '+' && tz[1] != '-')
		return 0;

	for (p = tz + 2; p != line + len; p++)
		if (!isdigit(*p))
			return 0;

	return line + len - tz;
}

516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533
static size_t tz_with_colon_len(const char *line, size_t len)
{
	const char *tz, *p;

	if (len < strlen(" +08:00") || line[len - strlen(":00")] != ':')
		return 0;
	tz = line + len - strlen(" +08:00");

	if (tz[0] != ' ' || (tz[1] != '+' && tz[1] != '-'))
		return 0;
	p = tz + 2;
	if (!isdigit(*p++) || !isdigit(*p++) || *p++ != ':' ||
	    !isdigit(*p++) || !isdigit(*p++))
		return 0;

	return line + len - tz;
}

534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
static size_t date_len(const char *line, size_t len)
{
	const char *date, *p;

	if (len < strlen("72-02-05") || line[len-strlen("-05")] != '-')
		return 0;
	p = date = line + len - strlen("72-02-05");

	if (!isdigit(*p++) || !isdigit(*p++) || *p++ != '-' ||
	    !isdigit(*p++) || !isdigit(*p++) || *p++ != '-' ||
	    !isdigit(*p++) || !isdigit(*p++))	/* Not a date. */
		return 0;

	if (date - line >= strlen("19") &&
	    isdigit(date[-1]) && isdigit(date[-2]))	/* 4-digit year */
		date -= strlen("19");

	return line + len - date;
}

static size_t short_time_len(const char *line, size_t len)
{
	const char *time, *p;

	if (len < strlen(" 07:01:32") || line[len-strlen(":32")] != ':')
		return 0;
	p = time = line + len - strlen(" 07:01:32");

	/* Permit 1-digit hours? */
	if (*p++ != ' ' ||
	    !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' ||
	    !isdigit(*p++) || !isdigit(*p++) || *p++ != ':' ||
	    !isdigit(*p++) || !isdigit(*p++))	/* Not a time. */
		return 0;

	return line + len - time;
}

static size_t fractional_time_len(const char *line, size_t len)
{
	const char *p;
	size_t n;

	/* Expected format: 19:41:17.620000023 */
	if (!len || !isdigit(line[len - 1]))
		return 0;
	p = line + len - 1;

	/* Fractional seconds. */
	while (p > line && isdigit(*p))
		p--;
	if (*p != '.')
		return 0;

	/* Hours, minutes, and whole seconds. */
	n = short_time_len(line, p - line);
	if (!n)
		return 0;

	return line + len - p + n;
}

static size_t trailing_spaces_len(const char *line, size_t len)
{
	const char *p;

	/* Expected format: ' ' x (1 or more)  */
	if (!len || line[len - 1] != ' ')
		return 0;

	p = line + len;
	while (p != line) {
		p--;
		if (*p != ' ')
			return line + len - (p + 1);
609 610
	}

611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627
	/* All spaces! */
	return len;
}

static size_t diff_timestamp_len(const char *line, size_t len)
{
	const char *end = line + len;
	size_t n;

	/*
	 * Posix: 2010-07-05 19:41:17
	 * GNU: 2010-07-05 19:41:17.620000023 -0500
	 */

	if (!isdigit(end[-1]))
		return 0;

628 629 630
	n = sane_tz_len(line, end - line);
	if (!n)
		n = tz_with_colon_len(line, end - line);
631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656
	end -= n;

	n = short_time_len(line, end - line);
	if (!n)
		n = fractional_time_len(line, end - line);
	end -= n;

	n = date_len(line, end - line);
	if (!n)	/* No date.  Too bad. */
		return 0;
	end -= n;

	if (end == line)	/* No space before date. */
		return 0;
	if (end[-1] == '\t') {	/* Success! */
		end--;
		return line + len - end;
	}
	if (end[-1] != ' ')	/* No space before date. */
		return 0;

	/* Whitespace damage. */
	end -= trailing_spaces_len(line, end - line);
	return line + len - end;
}

657 658 659 660 661 662 663
static char *null_strdup(const char *s)
{
	return s ? xstrdup(s) : NULL;
}

static char *find_name_common(const char *line, const char *def,
			      int p_value, const char *end, int terminate)
664 665 666 667
{
	int len;
	const char *start = NULL;

668 669
	if (p_value == 0)
		start = line;
670
	while (line != end) {
671
		char c = *line;
672

673
		if (!end && isspace(c)) {
674 675 676 677 678
			if (c == '\n')
				break;
			if (name_terminate(start, line-start, c, terminate))
				break;
		}
679 680 681 682 683
		line++;
		if (c == '/' && !--p_value)
			start = line;
	}
	if (!start)
684
		return squash_slash(null_strdup(def));
685 686
	len = line - start;
	if (!len)
687
		return squash_slash(null_strdup(def));
688 689 690 691 692 693 694 695 696 697

	/*
	 * Generally we prefer the shorter name, especially
	 * if the other one is just a variation of that with
	 * something else tacked on to the end (ie "file.orig"
	 * or "file~").
	 */
	if (def) {
		int deflen = strlen(def);
		if (deflen < len && !strncmp(start, def, deflen))
698
			return squash_slash(xstrdup(def));
L
Linus Torvalds 已提交
699
	}
700

701 702 703 704 705
	if (root) {
		char *ret = xmalloc(root_len + len + 1);
		strcpy(ret, root);
		memcpy(ret + root_len, start, len);
		ret[root_len + len] = '\0';
706
		return squash_slash(ret);
707 708
	}

709
	return squash_slash(xmemdupz(start, len));
710 711
}

712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742
static char *find_name(const char *line, char *def, int p_value, int terminate)
{
	if (*line == '"') {
		char *name = find_name_gnu(line, def, p_value);
		if (name)
			return name;
	}

	return find_name_common(line, def, p_value, NULL, terminate);
}

static char *find_name_traditional(const char *line, char *def, int p_value)
{
	size_t len = strlen(line);
	size_t date_len;

	if (*line == '"') {
		char *name = find_name_gnu(line, def, p_value);
		if (name)
			return name;
	}

	len = strchrnul(line, '\n') - line;
	date_len = diff_timestamp_len(line, len);
	if (!date_len)
		return find_name_common(line, def, p_value, NULL, TERM_TAB);
	len -= date_len;

	return find_name_common(line, def, p_value, line + len, 0);
}

743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764
static int count_slashes(const char *cp)
{
	int cnt = 0;
	char ch;

	while ((ch = *cp++))
		if (ch == '/')
			cnt++;
	return cnt;
}

/*
 * Given the string after "--- " or "+++ ", guess the appropriate
 * p_value for the given patch.
 */
static int guess_p_value(const char *nameline)
{
	char *name, *cp;
	int val = -1;

	if (is_dev_null(nameline))
		return -1;
765
	name = find_name_traditional(nameline, NULL, 0);
766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787
	if (!name)
		return -1;
	cp = strchr(name, '/');
	if (!cp)
		val = 0;
	else if (prefix) {
		/*
		 * Does it begin with "a/$our-prefix" and such?  Then this is
		 * very likely to apply to our directory.
		 */
		if (!strncmp(name, prefix, prefix_length))
			val = count_slashes(prefix);
		else {
			cp++;
			if (!strncmp(cp, prefix, prefix_length))
				val = count_slashes(prefix) + 1;
		}
	}
	free(name);
	return val;
}

788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805
/*
 * Does the ---/+++ line has the POSIX timestamp after the last HT?
 * GNU diff puts epoch there to signal a creation/deletion event.  Is
 * this such a timestamp?
 */
static int has_epoch_timestamp(const char *nameline)
{
	/*
	 * We are only interested in epoch timestamp; any non-zero
	 * fraction cannot be one, hence "(\.0+)?" in the regexp below.
	 * For the same reason, the date must be either 1969-12-31 or
	 * 1970-01-01, and the seconds part must be "00".
	 */
	const char stamp_regexp[] =
		"^(1969-12-31|1970-01-01)"
		" "
		"[0-2][0-9]:[0-5][0-9]:00(\\.0+)?"
		" "
806 807
		"([-+][0-2][0-9]:?[0-5][0-9])\n";
	const char *timestamp = NULL, *cp, *colon;
808 809 810 811 812 813 814 815 816 817 818 819 820 821 822
	static regex_t *stamp;
	regmatch_t m[10];
	int zoneoffset;
	int hourminute;
	int status;

	for (cp = nameline; *cp != '\n'; cp++) {
		if (*cp == '\t')
			timestamp = cp + 1;
	}
	if (!timestamp)
		return 0;
	if (!stamp) {
		stamp = xmalloc(sizeof(*stamp));
		if (regcomp(stamp, stamp_regexp, REG_EXTENDED)) {
823
			warning(_("Cannot prepare timestamp regexp %s"),
824 825 826 827 828 829 830 831
				stamp_regexp);
			return 0;
		}
	}

	status = regexec(stamp, timestamp, ARRAY_SIZE(m), m, 0);
	if (status) {
		if (status != REG_NOMATCH)
832
			warning(_("regexec returned %d for input: %s"),
833 834 835 836
				status, timestamp);
		return 0;
	}

837 838 839 840 841
	zoneoffset = strtol(timestamp + m[3].rm_so + 1, (char **) &colon, 10);
	if (*colon == ':')
		zoneoffset = zoneoffset * 60 + strtol(colon + 1, NULL, 10);
	else
		zoneoffset = (zoneoffset / 100) * 60 + (zoneoffset % 100);
842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860
	if (timestamp[m[3].rm_so] == '-')
		zoneoffset = -zoneoffset;

	/*
	 * YYYY-MM-DD hh:mm:ss must be from either 1969-12-31
	 * (west of GMT) or 1970-01-01 (east of GMT)
	 */
	if ((zoneoffset < 0 && memcmp(timestamp, "1969-12-31", 10)) ||
	    (0 <= zoneoffset && memcmp(timestamp, "1970-01-01", 10)))
		return 0;

	hourminute = (strtol(timestamp + 11, NULL, 10) * 60 +
		      strtol(timestamp + 14, NULL, 10) -
		      zoneoffset);

	return ((zoneoffset < 0 && hourminute == 1440) ||
		(0 <= zoneoffset && !hourminute));
}

861
/*
J
Junio C Hamano 已提交
862
 * Get the name etc info from the ---/+++ lines of a traditional patch header
863
 *
864 865 866
 * FIXME! The end-of-filename heuristics are kind of screwy. For existing
 * files, we can happily check the index for a match, but for creating a
 * new file we should try to match whatever "patch" does. I have no idea.
867
 */
868
static void parse_traditional_patch(const char *first, const char *second, struct patch *patch)
869 870 871
{
	char *name;

872 873
	first += 4;	/* skip "--- " */
	second += 4;	/* skip "+++ " */
874 875 876 877 878 879 880 881 882 883
	if (!p_value_known) {
		int p, q;
		p = guess_p_value(first);
		q = guess_p_value(second);
		if (p < 0) p = q;
		if (0 <= p && p == q) {
			p_value = p;
			p_value_known = 1;
		}
	}
884
	if (is_dev_null(first)) {
885 886
		patch->is_new = 1;
		patch->is_delete = 0;
887
		name = find_name_traditional(second, NULL, p_value);
888
		patch->new_name = name;
889
	} else if (is_dev_null(second)) {
890 891
		patch->is_new = 0;
		patch->is_delete = 1;
892
		name = find_name_traditional(first, NULL, p_value);
893
		patch->old_name = name;
894
	} else {
895 896 897 898
		char *first_name;
		first_name = find_name_traditional(first, NULL, p_value);
		name = find_name_traditional(second, first_name, p_value);
		free(first_name);
899 900 901 902 903 904 905 906 907
		if (has_epoch_timestamp(first)) {
			patch->is_new = 1;
			patch->is_delete = 0;
			patch->new_name = name;
		} else if (has_epoch_timestamp(second)) {
			patch->is_new = 0;
			patch->is_delete = 1;
			patch->old_name = name;
		} else {
908 909
			patch->old_name = name;
			patch->new_name = xstrdup(name);
910
		}
911 912
	}
	if (!name)
913
		die(_("unable to find filename in patch at line %d"), linenr);
914 915
}

916
static int gitdiff_hdrend(const char *line, struct patch *patch)
917 918 919 920
{
	return -1;
}

921 922 923 924 925 926 927 928 929
/*
 * We're anal about diff header consistency, to make
 * sure that we don't end up having strange ambiguous
 * patches floating around.
 *
 * As a result, gitdiff_{old|new}name() will check
 * their names against any previous information, just
 * to make sure..
 */
930 931 932 933
#define DIFF_OLD_NAME 0
#define DIFF_NEW_NAME 1

static char *gitdiff_verify_name(const char *line, int isnull, char *orig_name, int side)
934 935
{
	if (!orig_name && !isnull)
936
		return find_name(line, NULL, p_value, TERM_TAB);
937 938

	if (orig_name) {
939 940 941
		int len;
		const char *name;
		char *another;
942 943 944
		name = orig_name;
		len = strlen(name);
		if (isnull)
945
			die(_("git apply: bad git-diff - expected /dev/null, got %s on line %d"), name, linenr);
946
		another = find_name(line, NULL, p_value, TERM_TAB);
947
		if (!another || memcmp(another, name, len + 1))
948 949 950
			die((side == DIFF_NEW_NAME) ?
			    _("git apply: bad git-diff - inconsistent new filename on line %d") :
			    _("git apply: bad git-diff - inconsistent old filename on line %d"), linenr);
951
		free(another);
952 953
		return orig_name;
	}
954 955 956
	else {
		/* expect "/dev/null" */
		if (memcmp("/dev/null", line, 9) || line[9] != '\n')
957
			die(_("git apply: bad git-diff - expected /dev/null on line %d"), linenr);
958 959
		return NULL;
	}
960 961
}

962
static int gitdiff_oldname(const char *line, struct patch *patch)
963
{
964
	char *orig = patch->old_name;
965 966
	patch->old_name = gitdiff_verify_name(line, patch->is_new, patch->old_name,
					      DIFF_OLD_NAME);
967 968
	if (orig != patch->old_name)
		free(orig);
969 970 971
	return 0;
}

972
static int gitdiff_newname(const char *line, struct patch *patch)
973
{
974
	char *orig = patch->new_name;
975 976
	patch->new_name = gitdiff_verify_name(line, patch->is_delete, patch->new_name,
					      DIFF_NEW_NAME);
977 978
	if (orig != patch->new_name)
		free(orig);
979 980 981
	return 0;
}

982
static int gitdiff_oldmode(const char *line, struct patch *patch)
983
{
984
	patch->old_mode = strtoul(line, NULL, 8);
985 986 987
	return 0;
}

988
static int gitdiff_newmode(const char *line, struct patch *patch)
989
{
990
	patch->new_mode = strtoul(line, NULL, 8);
991 992 993
	return 0;
}

994
static int gitdiff_delete(const char *line, struct patch *patch)
995
{
996
	patch->is_delete = 1;
997 998
	free(patch->old_name);
	patch->old_name = null_strdup(patch->def_name);
999
	return gitdiff_oldmode(line, patch);
1000 1001
}

1002
static int gitdiff_newfile(const char *line, struct patch *patch)
1003
{
1004
	patch->is_new = 1;
1005 1006
	free(patch->new_name);
	patch->new_name = null_strdup(patch->def_name);
1007
	return gitdiff_newmode(line, patch);
1008 1009
}

1010
static int gitdiff_copysrc(const char *line, struct patch *patch)
1011
{
1012
	patch->is_copy = 1;
1013
	free(patch->old_name);
1014
	patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
1015 1016 1017
	return 0;
}

1018
static int gitdiff_copydst(const char *line, struct patch *patch)
1019
{
1020
	patch->is_copy = 1;
1021
	free(patch->new_name);
1022
	patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
1023 1024 1025
	return 0;
}

1026
static int gitdiff_renamesrc(const char *line, struct patch *patch)
1027
{
1028
	patch->is_rename = 1;
1029
	free(patch->old_name);
1030
	patch->old_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
1031 1032 1033
	return 0;
}

1034
static int gitdiff_renamedst(const char *line, struct patch *patch)
1035
{
1036
	patch->is_rename = 1;
1037
	free(patch->new_name);
1038
	patch->new_name = find_name(line, NULL, p_value ? p_value - 1 : 0, 0);
1039 1040 1041
	return 0;
}

1042
static int gitdiff_similarity(const char *line, struct patch *patch)
1043
{
1044 1045
	if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
		patch->score = 0;
1046
	return 0;
L
Linus Torvalds 已提交
1047 1048
}

1049 1050
static int gitdiff_dissimilarity(const char *line, struct patch *patch)
{
1051 1052
	if ((patch->score = strtoul(line, NULL, 10)) == ULONG_MAX)
		patch->score = 0;
1053 1054 1055
	return 0;
}

1056 1057
static int gitdiff_index(const char *line, struct patch *patch)
{
1058 1059
	/*
	 * index line is N hexadecimal, "..", N hexadecimal,
1060 1061 1062 1063 1064 1065
	 * and optional space with octal mode.
	 */
	const char *ptr, *eol;
	int len;

	ptr = strchr(line, '.');
1066
	if (!ptr || ptr[1] != '.' || 40 < ptr - line)
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079
		return 0;
	len = ptr - line;
	memcpy(patch->old_sha1_prefix, line, len);
	patch->old_sha1_prefix[len] = 0;

	line = ptr + 2;
	ptr = strchr(line, ' ');
	eol = strchr(line, '\n');

	if (!ptr || eol < ptr)
		ptr = eol;
	len = ptr - line;

1080
	if (40 < len)
1081 1082 1083 1084
		return 0;
	memcpy(patch->new_sha1_prefix, line, len);
	patch->new_sha1_prefix[len] = 0;
	if (*ptr == ' ')
1085
		patch->old_mode = strtoul(ptr+1, NULL, 8);
1086 1087 1088
	return 0;
}

1089 1090 1091 1092
/*
 * This is normal for a diff that doesn't change anything: we'll fall through
 * into the next diff. Tell the parser to break out.
 */
1093
static int gitdiff_unrecognized(const char *line, struct patch *patch)
1094 1095 1096 1097
{
	return -1;
}

1098 1099 1100 1101 1102
/*
 * Skip p_value leading components from "line"; as we do not accept
 * absolute paths, return NULL in that case.
 */
static const char *skip_tree_prefix(const char *line, int llen)
1103
{
1104
	int nslash;
1105 1106
	int i;

1107 1108 1109 1110
	if (!p_value)
		return (llen && line[0] == '/') ? NULL : line;

	nslash = p_value;
1111 1112
	for (i = 0; i < llen; i++) {
		int ch = line[i];
1113
		if (ch == '/' && --nslash <= 0)
1114
			return (i == 0) ? NULL : &line[i + 1];
1115 1116 1117 1118
	}
	return NULL;
}

1119 1120
/*
 * This is to extract the same name that appears on "diff --git"
1121 1122 1123 1124 1125 1126
 * line.  We do not find and return anything if it is a rename
 * patch, and it is OK because we will find the name elsewhere.
 * We need to reliably find name only when it is mode-change only,
 * creation or deletion of an empty file.  In any of these cases,
 * both sides are the same name under a/ and b/ respectively.
 */
1127
static char *git_header_name(const char *line, int llen)
1128
{
1129 1130
	const char *name;
	const char *second = NULL;
1131
	size_t len, line_len;
1132

1133 1134 1135 1136 1137
	line += strlen("diff --git ");
	llen -= strlen("diff --git ");

	if (*line == '"') {
		const char *cp;
1138 1139
		struct strbuf first = STRBUF_INIT;
		struct strbuf sp = STRBUF_INIT;
1140 1141 1142

		if (unquote_c_style(&first, line, &second))
			goto free_and_fail1;
1143

1144 1145 1146
		/* strip the a/b prefix including trailing slash */
		cp = skip_tree_prefix(first.buf, first.len);
		if (!cp)
1147
			goto free_and_fail1;
1148
		strbuf_remove(&first, 0, cp - first.buf);
1149

1150 1151
		/*
		 * second points at one past closing dq of name.
1152 1153 1154 1155 1156 1157
		 * find the second name.
		 */
		while ((second < line + llen) && isspace(*second))
			second++;

		if (line + llen <= second)
1158
			goto free_and_fail1;
1159
		if (*second == '"') {
1160 1161
			if (unquote_c_style(&sp, second, NULL))
				goto free_and_fail1;
1162 1163
			cp = skip_tree_prefix(sp.buf, sp.len);
			if (!cp)
1164
				goto free_and_fail1;
1165
			/* They must match, otherwise ignore */
1166
			if (strcmp(cp, first.buf))
1167 1168
				goto free_and_fail1;
			strbuf_release(&sp);
1169
			return strbuf_detach(&first, NULL);
1170 1171 1172
		}

		/* unquoted second */
1173 1174
		cp = skip_tree_prefix(second, line + llen - second);
		if (!cp)
1175
			goto free_and_fail1;
1176
		if (line + llen - cp != first.len ||
1177 1178
		    memcmp(first.buf, cp, first.len))
			goto free_and_fail1;
1179
		return strbuf_detach(&first, NULL);
1180 1181 1182 1183 1184

	free_and_fail1:
		strbuf_release(&first);
		strbuf_release(&sp);
		return NULL;
1185 1186
	}

1187
	/* unquoted first name */
1188 1189
	name = skip_tree_prefix(line, llen);
	if (!name)
1190
		return NULL;
1191

1192 1193
	/*
	 * since the first name is unquoted, a dq if exists must be
1194 1195 1196 1197
	 * the beginning of the second name.
	 */
	for (second = name; second < line + llen; second++) {
		if (*second == '"') {
1198
			struct strbuf sp = STRBUF_INIT;
1199
			const char *np;
1200 1201 1202 1203

			if (unquote_c_style(&sp, second, NULL))
				goto free_and_fail2;

1204 1205
			np = skip_tree_prefix(sp.buf, sp.len);
			if (!np)
1206 1207 1208 1209
				goto free_and_fail2;

			len = sp.buf + sp.len - np;
			if (len < second - name &&
1210 1211 1212
			    !strncmp(np, name, len) &&
			    isspace(name[len])) {
				/* Good */
1213
				strbuf_remove(&sp, 0, np - sp.buf);
1214
				return strbuf_detach(&sp, NULL);
1215
			}
1216 1217 1218 1219

		free_and_fail2:
			strbuf_release(&sp);
			return NULL;
1220 1221 1222
		}
	}

1223 1224 1225 1226
	/*
	 * Accept a name only if it shows up twice, exactly the same
	 * form.
	 */
1227 1228 1229 1230
	second = strchr(name, '\n');
	if (!second)
		return NULL;
	line_len = second - name;
1231
	for (len = 0 ; ; len++) {
1232
		switch (name[len]) {
1233 1234 1235
		default:
			continue;
		case '\n':
1236
			return NULL;
1237
		case '\t': case ' ':
1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
			/*
			 * Is this the separator between the preimage
			 * and the postimage pathname?  Again, we are
			 * only interested in the case where there is
			 * no rename, as this is only to set def_name
			 * and a rename patch has the names elsewhere
			 * in an unambiguous form.
			 */
			if (!name[len + 1])
				return NULL; /* no postimage name */
			second = skip_tree_prefix(name + len + 1,
						  line_len - (len + 1));
1250 1251
			if (!second)
				return NULL;
1252 1253 1254 1255 1256 1257
			/*
			 * Does len bytes starting at "name" and "second"
			 * (that are separated by one HT or SP we just
			 * found) exactly match?
			 */
			if (second[len] == '\n' && !strncmp(name, second, len))
P
Pierre Habouzit 已提交
1258
				return xmemdupz(name, len);
1259 1260 1261 1262
		}
	}
}

L
Linus Torvalds 已提交
1263
/* Verify that we recognize the lines following a git header */
1264
static int parse_git_header(const char *line, int len, unsigned int size, struct patch *patch)
L
Linus Torvalds 已提交
1265
{
1266 1267 1268
	unsigned long offset;

	/* A git diff has explicit new/delete information, so we don't guess */
1269 1270
	patch->is_new = 0;
	patch->is_delete = 0;
1271

1272 1273 1274 1275 1276 1277
	/*
	 * Some things may not have the old name in the
	 * rest of the headers anywhere (pure mode changes,
	 * or removing or adding empty files), so we get
	 * the default name from the header.
	 */
1278
	patch->def_name = git_header_name(line, len);
1279 1280 1281 1282 1283 1284 1285
	if (patch->def_name && root) {
		char *s = xmalloc(root_len + strlen(patch->def_name) + 1);
		strcpy(s, root);
		strcpy(s + root_len, patch->def_name);
		free(patch->def_name);
		patch->def_name = s;
	}
1286

1287 1288 1289 1290 1291 1292
	line += len;
	size -= len;
	linenr++;
	for (offset = len ; size > 0 ; offset += len, size -= len, line += len, linenr++) {
		static const struct opentry {
			const char *str;
1293
			int (*fn)(const char *, struct patch *);
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
		} optable[] = {
			{ "@@ -", gitdiff_hdrend },
			{ "--- ", gitdiff_oldname },
			{ "+++ ", gitdiff_newname },
			{ "old mode ", gitdiff_oldmode },
			{ "new mode ", gitdiff_newmode },
			{ "deleted file mode ", gitdiff_delete },
			{ "new file mode ", gitdiff_newfile },
			{ "copy from ", gitdiff_copysrc },
			{ "copy to ", gitdiff_copydst },
1304 1305
			{ "rename old ", gitdiff_renamesrc },
			{ "rename new ", gitdiff_renamedst },
L
Linus Torvalds 已提交
1306 1307
			{ "rename from ", gitdiff_renamesrc },
			{ "rename to ", gitdiff_renamedst },
1308
			{ "similarity index ", gitdiff_similarity },
1309
			{ "dissimilarity index ", gitdiff_dissimilarity },
1310
			{ "index ", gitdiff_index },
1311
			{ "", gitdiff_unrecognized },
1312 1313
		};
		int i;
L
Linus Torvalds 已提交
1314 1315

		len = linelen(line, size);
1316
		if (!len || line[len-1] != '\n')
L
Linus Torvalds 已提交
1317
			break;
1318
		for (i = 0; i < ARRAY_SIZE(optable); i++) {
1319 1320 1321 1322
			const struct opentry *p = optable + i;
			int oplen = strlen(p->str);
			if (len < oplen || memcmp(p->str, line, oplen))
				continue;
1323
			if (p->fn(line + oplen, patch) < 0)
1324
				return offset;
1325
			break;
1326
		}
L
Linus Torvalds 已提交
1327 1328
	}

1329
	return offset;
L
Linus Torvalds 已提交
1330 1331
}

1332
static int parse_num(const char *line, unsigned long *p)
1333 1334
{
	char *ptr;
1335 1336 1337 1338 1339 1340 1341 1342

	if (!isdigit(*line))
		return 0;
	*p = strtoul(line, &ptr, 10);
	return ptr - line;
}

static int parse_range(const char *line, int len, int offset, const char *expect,
1343
		       unsigned long *p1, unsigned long *p2)
1344
{
1345 1346 1347 1348 1349 1350 1351
	int digits, ex;

	if (offset < 0 || offset >= len)
		return -1;
	line += offset;
	len -= offset;

1352 1353
	digits = parse_num(line, p1);
	if (!digits)
1354 1355 1356 1357 1358 1359
		return -1;

	offset += digits;
	line += digits;
	len -= digits;

L
Linus Torvalds 已提交
1360
	*p2 = 1;
1361 1362 1363 1364 1365 1366 1367 1368 1369 1370
	if (*line == ',') {
		digits = parse_num(line+1, p2);
		if (!digits)
			return -1;

		offset += digits+1;
		line += digits+1;
		len -= digits+1;
	}

1371 1372 1373 1374 1375 1376 1377 1378 1379
	ex = strlen(expect);
	if (ex > len)
		return -1;
	if (memcmp(line, expect, ex))
		return -1;

	return offset + ex;
}

1380
static void recount_diff(const char *line, int size, struct fragment *fragment)
1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407
{
	int oldlines = 0, newlines = 0, ret = 0;

	if (size < 1) {
		warning("recount: ignore empty hunk");
		return;
	}

	for (;;) {
		int len = linelen(line, size);
		size -= len;
		line += len;

		if (size < 1)
			break;

		switch (*line) {
		case ' ': case '\n':
			newlines++;
			/* fall through */
		case '-':
			oldlines++;
			continue;
		case '+':
			newlines++;
			continue;
		case '\\':
1408
			continue;
1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
		case '@':
			ret = size < 3 || prefixcmp(line, "@@ ");
			break;
		case 'd':
			ret = size < 5 || prefixcmp(line, "diff ");
			break;
		default:
			ret = -1;
			break;
		}
		if (ret) {
1420
			warning(_("recount: unexpected line: %.*s"),
1421 1422 1423 1424 1425 1426 1427 1428 1429
				(int)linelen(line, size), line);
			return;
		}
		break;
	}
	fragment->oldlines = oldlines;
	fragment->newlines = newlines;
}

1430 1431 1432 1433
/*
 * Parse a unified diff fragment header of the
 * form "@@ -a,b +c,d @@"
 */
1434
static int parse_fragment_header(const char *line, int len, struct fragment *fragment)
1435 1436 1437 1438 1439 1440 1441
{
	int offset;

	if (!len || line[len-1] != '\n')
		return -1;

	/* Figure out the number of lines in a fragment */
1442 1443
	offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
	offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
1444 1445 1446 1447

	return offset;
}

1448
static int find_header(const char *line, unsigned long size, int *hdrsize, struct patch *patch)
L
Linus Torvalds 已提交
1449 1450 1451
{
	unsigned long offset, len;

1452
	patch->is_toplevel_relative = 0;
1453 1454 1455 1456
	patch->is_rename = patch->is_copy = 0;
	patch->is_new = patch->is_delete = -1;
	patch->old_mode = patch->new_mode = 0;
	patch->old_name = patch->new_name = NULL;
1457
	for (offset = 0; size > 0; offset += len, size -= len, line += len, linenr++) {
L
Linus Torvalds 已提交
1458 1459 1460 1461 1462 1463 1464 1465 1466
		unsigned long nextlen;

		len = linelen(line, size);
		if (!len)
			break;

		/* Testing this early allows us to take a few shortcuts.. */
		if (len < 6)
			continue;
1467 1468

		/*
1469
		 * Make sure we don't find any unconnected patch fragments.
1470 1471 1472 1473
		 * That's a sign that we didn't find a header, and that a
		 * patch has become corrupted/broken up.
		 */
		if (!memcmp("@@ -", line, 4)) {
1474 1475
			struct fragment dummy;
			if (parse_fragment_header(line, len, &dummy) < 0)
1476
				continue;
1477
			die(_("patch fragment without header at line %d: %.*s"),
1478
			    linenr, (int)len-1, line);
1479 1480
		}

L
Linus Torvalds 已提交
1481 1482 1483 1484 1485 1486 1487 1488
		if (size < len + 6)
			break;

		/*
		 * Git patch? It might not have a real patch, just a rename
		 * or mode change, so we handle that specially
		 */
		if (!memcmp("diff --git ", line, 11)) {
1489
			int git_hdr_len = parse_git_header(line, len, size, patch);
1490
			if (git_hdr_len <= len)
L
Linus Torvalds 已提交
1491
				continue;
1492 1493
			if (!patch->old_name && !patch->new_name) {
				if (!patch->def_name)
1494 1495 1496 1497 1498 1499
					die(Q_("git diff header lacks filename information when removing "
					       "%d leading pathname component (line %d)",
					       "git diff header lacks filename information when removing "
					       "%d leading pathname components (line %d)",
					       p_value),
					    p_value, linenr);
1500 1501
				patch->old_name = xstrdup(patch->def_name);
				patch->new_name = xstrdup(patch->def_name);
1502
			}
1503 1504 1505
			if (!patch->is_delete && !patch->new_name)
				die("git diff header lacks filename information "
				    "(line %d)", linenr);
1506
			patch->is_toplevel_relative = 1;
1507
			*hdrsize = git_hdr_len;
L
Linus Torvalds 已提交
1508 1509 1510
			return offset;
		}

1511
		/* --- followed by +++ ? */
L
Linus Torvalds 已提交
1512 1513 1514 1515 1516 1517
		if (memcmp("--- ", line,  4) || memcmp("+++ ", line + len, 4))
			continue;

		/*
		 * We only accept unified patches, so we want it to
		 * at least have "@@ -a,b +c,d @@\n", which is 14 chars
1518
		 * minimum ("@@ -0,0 +1 @@\n" is the shortest).
L
Linus Torvalds 已提交
1519 1520 1521 1522 1523 1524
		 */
		nextlen = linelen(line + len, size - len);
		if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
			continue;

		/* Ok, we'll consider it a patch */
1525
		parse_traditional_patch(line, line+len, patch);
L
Linus Torvalds 已提交
1526
		*hdrsize = len + nextlen;
1527
		linenr += 2;
L
Linus Torvalds 已提交
1528 1529 1530 1531 1532
		return offset;
	}
	return -1;
}

1533
static void record_ws_error(unsigned result, const char *line, int len, int linenr)
1534
{
W
Wincent Colaiuta 已提交
1535
	char *err;
1536

W
Wincent Colaiuta 已提交
1537 1538
	if (!result)
		return;
1539 1540 1541 1542

	whitespace_error++;
	if (squelch_whitespace_errors &&
	    squelch_whitespace_errors < whitespace_error)
1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555
		return;

	err = whitespace_error_string(result);
	fprintf(stderr, "%s:%d: %s.\n%.*s\n",
		patch_input_file, linenr, err, len, line);
	free(err);
}

static void check_whitespace(const char *line, int len, unsigned ws_rule)
{
	unsigned result = ws_check(line + 1, len - 1, ws_rule);

	record_ws_error(result, line + 1, len - 2, linenr);
1556 1557
}

L
Linus Torvalds 已提交
1558
/*
1559 1560 1561 1562
 * Parse a unified diff. Note that this really needs to parse each
 * fragment separately, since the only way to know the difference
 * between a "---" that is part of a patch, and a "---" that starts
 * the next patch is to look at the line counts..
L
Linus Torvalds 已提交
1563
 */
1564
static int parse_fragment(const char *line, unsigned long size,
1565
			  struct patch *patch, struct fragment *fragment)
L
Linus Torvalds 已提交
1566
{
1567
	int added, deleted;
L
Linus Torvalds 已提交
1568
	int len = linelen(line, size), offset;
1569
	unsigned long oldlines, newlines;
1570
	unsigned long leading, trailing;
L
Linus Torvalds 已提交
1571

1572
	offset = parse_fragment_header(line, len, fragment);
L
Linus Torvalds 已提交
1573 1574
	if (offset < 0)
		return -1;
1575 1576
	if (offset > 0 && patch->recount)
		recount_diff(line + offset, size - offset, fragment);
1577 1578
	oldlines = fragment->oldlines;
	newlines = fragment->newlines;
1579 1580
	leading = 0;
	trailing = 0;
L
Linus Torvalds 已提交
1581 1582 1583 1584

	/* Parse the thing.. */
	line += len;
	size -= len;
1585
	linenr++;
1586
	added = deleted = 0;
1587 1588 1589
	for (offset = len;
	     0 < size;
	     offset += len, size -= len, line += len, linenr++) {
L
Linus Torvalds 已提交
1590 1591 1592 1593 1594 1595 1596 1597
		if (!oldlines && !newlines)
			break;
		len = linelen(line, size);
		if (!len || line[len-1] != '\n')
			return -1;
		switch (*line) {
		default:
			return -1;
1598
		case '\n': /* newer GNU diff, an empty context line */
L
Linus Torvalds 已提交
1599 1600 1601
		case ' ':
			oldlines--;
			newlines--;
1602 1603 1604
			if (!deleted && !added)
				leading++;
			trailing++;
L
Linus Torvalds 已提交
1605 1606
			break;
		case '-':
1607
			if (apply_in_reverse &&
1608
			    ws_error_action != nowarn_ws_error)
1609
				check_whitespace(line, len, patch->ws_rule);
1610
			deleted++;
L
Linus Torvalds 已提交
1611
			oldlines--;
1612
			trailing = 0;
L
Linus Torvalds 已提交
1613 1614
			break;
		case '+':
1615
			if (!apply_in_reverse &&
1616
			    ws_error_action != nowarn_ws_error)
1617
				check_whitespace(line, len, patch->ws_rule);
1618
			added++;
L
Linus Torvalds 已提交
1619
			newlines--;
1620
			trailing = 0;
L
Linus Torvalds 已提交
1621
			break;
1622

1623 1624
		/*
		 * We allow "\ No newline at end of file". Depending
1625 1626
                 * on locale settings when the patch was produced we
                 * don't know what this line looks like. The only
1627 1628 1629 1630
                 * thing we do know is that it begins with "\ ".
		 * Checking for 12 is just for sanity check -- any
		 * l10n of "\ No newline..." is at least that long.
		 */
1631
		case '\\':
1632
			if (len < 12 || memcmp(line, "\\ ", 2))
1633
				return -1;
1634
			break;
L
Linus Torvalds 已提交
1635 1636
		}
	}
L
Linus Torvalds 已提交
1637 1638
	if (oldlines || newlines)
		return -1;
1639 1640 1641
	fragment->leading = leading;
	fragment->trailing = trailing;

1642 1643
	/*
	 * If a fragment ends with an incomplete line, we failed to include
1644 1645 1646
	 * it in the above loop because we hit oldlines == newlines == 0
	 * before seeing it.
	 */
1647
	if (12 < size && !memcmp(line, "\\ ", 2))
1648 1649
		offset += linelen(line, size);

1650 1651
	patch->lines_added += added;
	patch->lines_deleted += deleted;
1652 1653

	if (0 < patch->is_new && oldlines)
1654
		return error(_("new file depends on old contents"));
1655
	if (0 < patch->is_delete && newlines)
1656
		return error(_("deleted file still has contents"));
L
Linus Torvalds 已提交
1657 1658 1659
	return offset;
}

1660 1661 1662 1663 1664 1665 1666 1667
/*
 * We have seen "diff --git a/... b/..." header (or a traditional patch
 * header).  Read hunks that belong to this patch into fragments and hang
 * them to the given patch structure.
 *
 * The (fragment->patch, fragment->size) pair points into the memory given
 * by the caller, not a copy, when we return.
 */
1668
static int parse_single_patch(const char *line, unsigned long size, struct patch *patch)
L
Linus Torvalds 已提交
1669 1670
{
	unsigned long offset = 0;
1671
	unsigned long oldlines = 0, newlines = 0, context = 0;
1672
	struct fragment **fragp = &patch->fragments;
L
Linus Torvalds 已提交
1673 1674

	while (size > 4 && !memcmp(line, "@@ -", 4)) {
1675 1676 1677
		struct fragment *fragment;
		int len;

1678
		fragment = xcalloc(1, sizeof(*fragment));
1679
		fragment->linenr = linenr;
1680
		len = parse_fragment(line, size, patch, fragment);
L
Linus Torvalds 已提交
1681
		if (len <= 0)
1682
			die(_("corrupt patch at line %d"), linenr);
1683 1684
		fragment->patch = line;
		fragment->size = len;
1685 1686 1687
		oldlines += fragment->oldlines;
		newlines += fragment->newlines;
		context += fragment->leading + fragment->trailing;
1688 1689 1690

		*fragp = fragment;
		fragp = &fragment->next;
L
Linus Torvalds 已提交
1691 1692 1693 1694 1695

		offset += len;
		line += len;
		size -= len;
	}
1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717

	/*
	 * If something was removed (i.e. we have old-lines) it cannot
	 * be creation, and if something was added it cannot be
	 * deletion.  However, the reverse is not true; --unified=0
	 * patches that only add are not necessarily creation even
	 * though they do not have any old lines, and ones that only
	 * delete are not necessarily deletion.
	 *
	 * Unfortunately, a real creation/deletion patch do _not_ have
	 * any context line by definition, so we cannot safely tell it
	 * apart with --unified=0 insanity.  At least if the patch has
	 * more than one hunk it is not creation or deletion.
	 */
	if (patch->is_new < 0 &&
	    (oldlines || (patch->fragments && patch->fragments->next)))
		patch->is_new = 0;
	if (patch->is_delete < 0 &&
	    (newlines || (patch->fragments && patch->fragments->next)))
		patch->is_delete = 0;

	if (0 < patch->is_new && oldlines)
1718
		die(_("new file %s depends on old contents"), patch->new_name);
1719
	if (0 < patch->is_delete && newlines)
1720
		die(_("deleted file %s still has contents"), patch->old_name);
1721
	if (!patch->is_delete && !newlines && context)
1722 1723 1724 1725
		fprintf_ln(stderr,
			   _("** warning: "
			     "file %s becomes empty but is not deleted"),
			   patch->new_name);
1726

L
Linus Torvalds 已提交
1727 1728 1729
	return offset;
}

1730 1731 1732 1733 1734 1735 1736 1737 1738 1739
static inline int metadata_changes(struct patch *patch)
{
	return	patch->is_rename > 0 ||
		patch->is_copy > 0 ||
		patch->is_new > 0 ||
		patch->is_delete ||
		(patch->old_mode && patch->new_mode &&
		 patch->old_mode != patch->new_mode);
}

1740 1741
static char *inflate_it(const void *data, unsigned long size,
			unsigned long inflated_size)
J
Junio C Hamano 已提交
1742
{
1743
	git_zstream stream;
1744 1745 1746 1747 1748 1749 1750 1751 1752
	void *out;
	int st;

	memset(&stream, 0, sizeof(stream));

	stream.next_in = (unsigned char *)data;
	stream.avail_in = size;
	stream.next_out = out = xmalloc(inflated_size);
	stream.avail_out = inflated_size;
1753 1754 1755
	git_inflate_init(&stream);
	st = git_inflate(&stream, Z_FINISH);
	git_inflate_end(&stream);
1756 1757 1758 1759 1760 1761 1762
	if ((st != Z_STREAM_END) || stream.total_out != inflated_size) {
		free(out);
		return NULL;
	}
	return out;
}

1763 1764 1765 1766 1767
/*
 * Read a binary hunk and return a new fragment; fragment->patch
 * points at an allocated memory that the caller must free, so
 * it is marked as "->free_patch = 1".
 */
1768 1769 1770 1771 1772
static struct fragment *parse_binary_hunk(char **buf_p,
					  unsigned long *sz_p,
					  int *status_p,
					  int *used_p)
{
1773 1774
	/*
	 * Expect a line that begins with binary patch method ("literal"
1775 1776 1777
	 * or "delta"), followed by the length of data before deflating.
	 * a sequence of 'length-byte' followed by base-85 encoded data
	 * should follow, terminated by a newline.
J
Junio C Hamano 已提交
1778 1779 1780 1781 1782 1783 1784 1785
	 *
	 * Each 5-byte sequence of base-85 encodes up to 4 bytes,
	 * and we would limit the patch line to 66 characters,
	 * so one line can fit up to 13 groups that would decode
	 * to 52 bytes max.  The length byte 'A'-'Z' corresponds
	 * to 1-26 bytes, and 'a'-'z' corresponds to 27-52 bytes.
	 */
	int llen, used;
1786 1787 1788 1789
	unsigned long size = *sz_p;
	char *buffer = *buf_p;
	int patch_method;
	unsigned long origlen;
J
Junio C Hamano 已提交
1790
	char *data = NULL;
1791 1792
	int hunk_size = 0;
	struct fragment *frag;
J
Junio C Hamano 已提交
1793

J
Junio C Hamano 已提交
1794 1795
	llen = linelen(buffer, size);
	used = llen;
1796 1797

	*status_p = 0;
J
Junio C Hamano 已提交
1798

1799
	if (!prefixcmp(buffer, "delta ")) {
1800 1801
		patch_method = BINARY_DELTA_DEFLATED;
		origlen = strtoul(buffer + 6, NULL, 10);
J
Junio C Hamano 已提交
1802
	}
1803
	else if (!prefixcmp(buffer, "literal ")) {
1804 1805
		patch_method = BINARY_LITERAL_DEFLATED;
		origlen = strtoul(buffer + 8, NULL, 10);
J
Junio C Hamano 已提交
1806 1807
	}
	else
1808 1809 1810
		return NULL;

	linenr++;
J
Junio C Hamano 已提交
1811
	buffer += llen;
J
Junio C Hamano 已提交
1812 1813 1814 1815 1816
	while (1) {
		int byte_length, max_byte_length, newsize;
		llen = linelen(buffer, size);
		used += llen;
		linenr++;
1817 1818 1819 1820
		if (llen == 1) {
			/* consume the blank line */
			buffer++;
			size--;
J
Junio C Hamano 已提交
1821
			break;
1822
		}
1823 1824
		/*
		 * Minimum line is "A00000\n" which is 7-byte long,
J
Junio C Hamano 已提交
1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843
		 * and the line length must be multiple of 5 plus 2.
		 */
		if ((llen < 7) || (llen-2) % 5)
			goto corrupt;
		max_byte_length = (llen - 2) / 5 * 4;
		byte_length = *buffer;
		if ('A' <= byte_length && byte_length <= 'Z')
			byte_length = byte_length - 'A' + 1;
		else if ('a' <= byte_length && byte_length <= 'z')
			byte_length = byte_length - 'a' + 27;
		else
			goto corrupt;
		/* if the input length was not multiple of 4, we would
		 * have filler at the end but the filler should never
		 * exceed 3 bytes
		 */
		if (max_byte_length < byte_length ||
		    byte_length <= max_byte_length - 4)
			goto corrupt;
1844
		newsize = hunk_size + byte_length;
J
Junio C Hamano 已提交
1845
		data = xrealloc(data, newsize);
1846
		if (decode_85(data + hunk_size, buffer + 1, byte_length))
J
Junio C Hamano 已提交
1847
			goto corrupt;
1848
		hunk_size = newsize;
J
Junio C Hamano 已提交
1849 1850 1851
		buffer += llen;
		size -= llen;
	}
1852 1853 1854

	frag = xcalloc(1, sizeof(*frag));
	frag->patch = inflate_it(data, hunk_size, origlen);
1855
	frag->free_patch = 1;
1856 1857 1858 1859 1860 1861 1862 1863 1864 1865
	if (!frag->patch)
		goto corrupt;
	free(data);
	frag->size = origlen;
	*buf_p = buffer;
	*sz_p = size;
	*used_p = used;
	frag->binary_patch_method = patch_method;
	return frag;

J
Junio C Hamano 已提交
1866
 corrupt:
J
Junio C Hamano 已提交
1867
	free(data);
1868
	*status_p = -1;
1869
	error(_("corrupt binary patch at line %d: %.*s"),
1870 1871 1872 1873 1874 1875
	      linenr-1, llen-1, buffer);
	return NULL;
}

static int parse_binary(char *buffer, unsigned long size, struct patch *patch)
{
1876 1877
	/*
	 * We have read "GIT binary patch\n"; what follows is a line
1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897
	 * that says the patch method (currently, either "literal" or
	 * "delta") and the length of data before deflating; a
	 * sequence of 'length-byte' followed by base-85 encoded data
	 * follows.
	 *
	 * When a binary patch is reversible, there is another binary
	 * hunk in the same format, starting with patch method (either
	 * "literal" or "delta") with the length of data, and a sequence
	 * of length-byte + base-85 encoded data, terminated with another
	 * empty line.  This data, when applied to the postimage, produces
	 * the preimage.
	 */
	struct fragment *forward;
	struct fragment *reverse;
	int status;
	int used, used_1;

	forward = parse_binary_hunk(&buffer, &size, &status, &used);
	if (!forward && !status)
		/* there has to be one hunk (forward hunk) */
1898
		return error(_("unrecognized binary patch at line %d"), linenr-1);
1899 1900 1901 1902 1903 1904 1905 1906
	if (status)
		/* otherwise we already gave an error message */
		return status;

	reverse = parse_binary_hunk(&buffer, &size, &status, &used_1);
	if (reverse)
		used += used_1;
	else if (status) {
1907 1908
		/*
		 * Not having reverse hunk is not an error, but having
1909 1910 1911 1912 1913 1914 1915 1916 1917 1918
		 * a corrupt reverse hunk is.
		 */
		free((void*) forward->patch);
		free(forward);
		return status;
	}
	forward->next = reverse;
	patch->fragments = forward;
	patch->is_binary = 1;
	return used;
J
Junio C Hamano 已提交
1919 1920
}

1921 1922 1923 1924 1925 1926 1927
/*
 * Read the patch text in "buffer" taht extends for "size" bytes; stop
 * reading after seeing a single patch (i.e. changes to a single file).
 * Create fragments (i.e. patch hunks) and hang them to the given patch.
 * Return the number of bytes consumed, so that the caller can call us
 * again for the next patch.
 */
1928
static int parse_chunk(char *buffer, unsigned long size, struct patch *patch)
L
Linus Torvalds 已提交
1929 1930
{
	int hdrsize, patchsize;
1931
	int offset = find_header(buffer, size, &hdrsize, patch);
L
Linus Torvalds 已提交
1932 1933 1934 1935

	if (offset < 0)
		return offset;

1936 1937 1938 1939
	patch->ws_rule = whitespace_rule(patch->new_name
					 ? patch->new_name
					 : patch->old_name);

1940 1941
	patchsize = parse_single_patch(buffer + offset + hdrsize,
				       size - offset - hdrsize, patch);
L
Linus Torvalds 已提交
1942

1943
	if (!patchsize) {
1944 1945 1946 1947 1948
		static const char *binhdr[] = {
			"Binary files ",
			"Files ",
			NULL,
		};
J
Junio C Hamano 已提交
1949
		static const char git_binary[] = "GIT binary patch\n";
1950 1951 1952 1953
		int i;
		int hd = hdrsize + offset;
		unsigned long llen = linelen(buffer + hd, size - hd);

J
Junio C Hamano 已提交
1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965
		if (llen == sizeof(git_binary) - 1 &&
		    !memcmp(git_binary, buffer + hd, llen)) {
			int used;
			linenr++;
			used = parse_binary(buffer + hd + llen,
					    size - hd - llen, patch);
			if (used)
				patchsize = used + llen;
			else
				patchsize = 0;
		}
		else if (!memcmp(" differ\n", buffer + hd + llen - 8, 8)) {
1966 1967 1968 1969
			for (i = 0; binhdr[i]; i++) {
				int len = strlen(binhdr[i]);
				if (len < size - hd &&
				    !memcmp(binhdr[i], buffer + hd, len)) {
J
Junio C Hamano 已提交
1970
					linenr++;
1971
					patch->is_binary = 1;
J
Junio C Hamano 已提交
1972
					patchsize = llen;
1973 1974 1975
					break;
				}
			}
J
Junio C Hamano 已提交
1976
		}
1977

J
Junio C Hamano 已提交
1978 1979 1980
		/* Empty patch cannot be applied if it is a text patch
		 * without metadata change.  A binary patch appears
		 * empty to us here.
1981 1982
		 */
		if ((apply || check) &&
J
Junio C Hamano 已提交
1983
		    (!patch->is_binary && !metadata_changes(patch)))
1984
			die(_("patch with only garbage at line %d"), linenr);
1985
	}
1986

L
Linus Torvalds 已提交
1987 1988 1989
	return offset + hdrsize + patchsize;
}

J
Johannes Schindelin 已提交
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
#define swap(a,b) myswap((a),(b),sizeof(a))

#define myswap(a, b, size) do {		\
	unsigned char mytmp[size];	\
	memcpy(mytmp, &a, size);		\
	memcpy(&a, &b, size);		\
	memcpy(&b, mytmp, size);		\
} while (0)

static void reverse_patches(struct patch *p)
{
	for (; p; p = p->next) {
		struct fragment *frag = p->fragments;

		swap(p->new_name, p->old_name);
		swap(p->new_mode, p->old_mode);
		swap(p->is_new, p->is_delete);
		swap(p->lines_added, p->lines_deleted);
		swap(p->old_sha1_prefix, p->new_sha1_prefix);

		for (; frag; frag = frag->next) {
			swap(frag->newpos, frag->oldpos);
			swap(frag->newlines, frag->oldlines);
		}
	}
}

2017 2018 2019 2020
static const char pluses[] =
"++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++";
static const char minuses[]=
"----------------------------------------------------------------------";
2021 2022 2023

static void show_stats(struct patch *patch)
{
2024
	struct strbuf qname = STRBUF_INIT;
2025 2026
	char *cp = patch->new_name ? patch->new_name : patch->old_name;
	int max, add, del;
2027

2028
	quote_c_style(cp, &qname, NULL, 0);
2029

2030 2031 2032 2033 2034 2035
	/*
	 * "scale" the filename
	 */
	max = max_len;
	if (max > 50)
		max = 50;
2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047

	if (qname.len > max) {
		cp = strchr(qname.buf + qname.len + 3 - max, '/');
		if (!cp)
			cp = qname.buf + qname.len + 3 - max;
		strbuf_splice(&qname, 0, cp - qname.buf, "...", 3);
	}

	if (patch->is_binary) {
		printf(" %-*s |  Bin\n", max, qname.buf);
		strbuf_release(&qname);
		return;
2048
	}
2049 2050 2051

	printf(" %-*s |", max, qname.buf);
	strbuf_release(&qname);
2052 2053 2054 2055

	/*
	 * scale the add/delete
	 */
2056
	max = max + max_change > 70 ? 70 - max : max_change;
2057 2058 2059
	add = patch->lines_added;
	del = patch->lines_deleted;

2060
	if (max_change > 0) {
2061
		int total = ((add + del) * max + max_change / 2) / max_change;
2062 2063 2064
		add = (add * max + max_change / 2) / max_change;
		del = total - add;
	}
2065 2066
	printf("%5d %.*s%.*s\n", patch->lines_added + patch->lines_deleted,
		add, pluses, del, minuses);
2067 2068
}

2069
static int read_old_data(struct stat *st, const char *path, struct strbuf *buf)
2070 2071 2072
{
	switch (st->st_mode & S_IFMT) {
	case S_IFLNK:
2073
		if (strbuf_readlink(buf, path, st->st_size) < 0)
2074
			return error(_("unable to read symlink %s"), path);
2075
		return 0;
2076
	case S_IFREG:
2077
		if (strbuf_read_file(buf, path, st->st_size) != st->st_size)
2078
			return error(_("unable to open or read %s"), path);
2079
		convert_to_git(path, buf->buf, buf->len, buf, 0);
2080
		return 0;
2081 2082 2083 2084 2085
	default:
		return -1;
	}
}

2086 2087 2088 2089 2090 2091 2092
/*
 * Update the preimage, and the common lines in postimage,
 * from buffer buf of length len. If postlen is 0 the postimage
 * is updated in place, otherwise it's updated on a new buffer
 * of length postlen
 */

2093 2094 2095
static void update_pre_post_images(struct image *preimage,
				   struct image *postimage,
				   char *buf,
2096
				   size_t len, size_t postlen)
2097
{
2098 2099 2100
	int i, ctx;
	char *new, *old, *fixed;
	struct image fixed_preimage;
2101

2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112
	/*
	 * Update the preimage with whitespace fixes.  Note that we
	 * are not losing preimage->buf -- apply_one_fragment() will
	 * free "oldlines".
	 */
	prepare_image(&fixed_preimage, buf, len, 1);
	assert(fixed_preimage.nr == preimage->nr);
	for (i = 0; i < preimage->nr; i++)
		fixed_preimage.line[i].flag = preimage->line[i].flag;
	free(preimage->line_allocated);
	*preimage = fixed_preimage;
2113

2114
	/*
2115 2116 2117 2118 2119 2120 2121
	 * Adjust the common context lines in postimage. This can be
	 * done in-place when we are just doing whitespace fixing,
	 * which does not make the string grow, but needs a new buffer
	 * when ignoring whitespace causes the update, since in this case
	 * we could have e.g. tabs converted to multiple spaces.
	 * We trust the caller to tell us if the update can be done
	 * in place (postlen==0) or not.
2122
	 */
2123 2124 2125 2126 2127
	old = postimage->buf;
	if (postlen)
		new = postimage->buf = xmalloc(postlen);
	else
		new = old;
2128 2129 2130 2131 2132 2133 2134 2135 2136
	fixed = preimage->buf;
	for (i = ctx = 0; i < postimage->nr; i++) {
		size_t len = postimage->line[i].len;
		if (!(postimage->line[i].flag & LINE_COMMON)) {
			/* an added line -- no counterparts in preimage */
			memmove(new, old, len);
			old += len;
			new += len;
			continue;
2137
		}
2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148

		/* a common context -- skip it in the original postimage */
		old += len;

		/* and find the corresponding one in the fixed preimage */
		while (ctx < preimage->nr &&
		       !(preimage->line[ctx].flag & LINE_COMMON)) {
			fixed += preimage->line[ctx].len;
			ctx++;
		}
		if (preimage->nr <= ctx)
2149
			die(_("oops"));
2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163

		/* and copy it in, while fixing the line length */
		len = preimage->line[ctx].len;
		memcpy(new, fixed, len);
		new += len;
		fixed += len;
		postimage->line[i].len = len;
		ctx++;
	}

	/* Fix the length of the whole thing */
	postimage->len = new - postimage->buf;
}

2164 2165 2166
static int match_fragment(struct image *img,
			  struct image *preimage,
			  struct image *postimage,
2167
			  unsigned long try,
2168
			  int try_lno,
2169
			  unsigned ws_rule,
2170
			  int match_beginning, int match_end)
2171
{
2172
	int i;
2173
	char *fixed_buf, *buf, *orig, *target;
2174 2175
	struct strbuf fixed;
	size_t fixed_len;
2176
	int preimage_limit;
2177

2178 2179 2180 2181 2182 2183 2184 2185
	if (preimage->nr + try_lno <= img->nr) {
		/*
		 * The hunk falls within the boundaries of img.
		 */
		preimage_limit = preimage->nr;
		if (match_end && (preimage->nr + try_lno != img->nr))
			return 0;
	} else if (ws_error_action == correct_ws_error &&
2186
		   (ws_rule & WS_BLANK_AT_EOF)) {
2187
		/*
2188 2189 2190 2191 2192
		 * This hunk extends beyond the end of img, and we are
		 * removing blank lines at the end of the file.  This
		 * many lines from the beginning of the preimage must
		 * match with img, and the remainder of the preimage
		 * must be blank.
2193 2194 2195 2196 2197 2198 2199 2200
		 */
		preimage_limit = img->nr - try_lno;
	} else {
		/*
		 * The hunk extends beyond the end of the img and
		 * we are not removing blanks at the end, so we
		 * should reject the hunk at this position.
		 */
2201
		return 0;
2202
	}
2203 2204

	if (match_beginning && try_lno)
2205
		return 0;
2206

2207
	/* Quick hash check */
2208
	for (i = 0; i < preimage_limit; i++)
2209 2210
		if ((img->line[try_lno + i].flag & LINE_PATCHED) ||
		    (preimage->line[i].hash != img->line[try_lno + i].hash))
2211 2212
			return 0;

2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246
	if (preimage_limit == preimage->nr) {
		/*
		 * Do we have an exact match?  If we were told to match
		 * at the end, size must be exactly at try+fragsize,
		 * otherwise try+fragsize must be still within the preimage,
		 * and either case, the old piece should match the preimage
		 * exactly.
		 */
		if ((match_end
		     ? (try + preimage->len == img->len)
		     : (try + preimage->len <= img->len)) &&
		    !memcmp(img->buf + try, preimage->buf, preimage->len))
			return 1;
	} else {
		/*
		 * The preimage extends beyond the end of img, so
		 * there cannot be an exact match.
		 *
		 * There must be one non-blank context line that match
		 * a line before the end of img.
		 */
		char *buf_end;

		buf = preimage->buf;
		buf_end = buf;
		for (i = 0; i < preimage_limit; i++)
			buf_end += preimage->line[i].len;

		for ( ; buf < buf_end; buf++)
			if (!isspace(*buf))
				break;
		if (buf == buf_end)
			return 0;
	}
2247

2248 2249 2250 2251 2252 2253 2254 2255 2256
	/*
	 * No exact match. If we are ignoring whitespace, run a line-by-line
	 * fuzzy matching. We collect all the line length information because
	 * we need it to adjust whitespace if we match.
	 */
	if (ws_ignore_action == ignore_ws_change) {
		size_t imgoff = 0;
		size_t preoff = 0;
		size_t postlen = postimage->len;
2257 2258 2259 2260
		size_t extra_chars;
		char *preimage_eof;
		char *preimage_end;
		for (i = 0; i < preimage_limit; i++) {
2261
			size_t prelen = preimage->line[i].len;
2262
			size_t imglen = img->line[try_lno+i].len;
2263

2264 2265
			if (!fuzzy_matchlines(img->buf + try + imgoff, imglen,
					      preimage->buf + preoff, prelen))
2266 2267
				return 0;
			if (preimage->line[i].flag & LINE_COMMON)
2268 2269
				postlen += imglen - prelen;
			imgoff += imglen;
2270 2271 2272 2273
			preoff += prelen;
		}

		/*
2274 2275 2276
		 * Ok, the preimage matches with whitespace fuzz.
		 *
		 * imgoff now holds the true length of the target that
2277 2278 2279 2280 2281 2282
		 * matches the preimage before the end of the file.
		 *
		 * Count the number of characters in the preimage that fall
		 * beyond the end of the file and make sure that all of them
		 * are whitespace characters. (This can only happen if
		 * we are removing blank lines at the end of the file.)
2283
		 */
2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299
		buf = preimage_eof = preimage->buf + preoff;
		for ( ; i < preimage->nr; i++)
			preoff += preimage->line[i].len;
		preimage_end = preimage->buf + preoff;
		for ( ; buf < preimage_end; buf++)
			if (!isspace(*buf))
				return 0;

		/*
		 * Update the preimage and the common postimage context
		 * lines to use the same whitespace as the target.
		 * If whitespace is missing in the target (i.e.
		 * if the preimage extends beyond the end of the file),
		 * use the whitespace from the preimage.
		 */
		extra_chars = preimage_end - preimage_eof;
2300 2301 2302 2303
		strbuf_init(&fixed, imgoff + extra_chars);
		strbuf_add(&fixed, img->buf + try, imgoff);
		strbuf_add(&fixed, preimage_eof, extra_chars);
		fixed_buf = strbuf_detach(&fixed, &fixed_len);
2304
		update_pre_post_images(preimage, postimage,
2305
				fixed_buf, fixed_len, postlen);
2306 2307 2308
		return 1;
	}

2309 2310 2311
	if (ws_error_action != correct_ws_error)
		return 0;

2312
	/*
2313
	 * The hunk does not apply byte-by-byte, but the hash says
2314 2315 2316
	 * it might with whitespace fuzz. We haven't been asked to
	 * ignore whitespace, we were asked to correct whitespace
	 * errors, so let's try matching after whitespace correction.
2317 2318 2319 2320
	 *
	 * The preimage may extend beyond the end of the file,
	 * but in this loop we will only handle the part of the
	 * preimage that falls within the file.
2321
	 */
2322
	strbuf_init(&fixed, preimage->len + 1);
2323 2324
	orig = preimage->buf;
	target = img->buf + try;
2325
	for (i = 0; i < preimage_limit; i++) {
2326 2327
		size_t oldlen = preimage->line[i].len;
		size_t tgtlen = img->line[try_lno + i].len;
2328 2329
		size_t fixstart = fixed.len;
		struct strbuf tgtfix;
2330 2331 2332
		int match;

		/* Try fixing the line in the preimage */
2333
		ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);
2334 2335

		/* Try fixing the line in the target */
2336 2337
		strbuf_init(&tgtfix, tgtlen);
		ws_fix_copy(&tgtfix, target, tgtlen, ws_rule, NULL);
2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348

		/*
		 * If they match, either the preimage was based on
		 * a version before our tree fixed whitespace breakage,
		 * or we are lacking a whitespace-fix patch the tree
		 * the preimage was based on already had (i.e. target
		 * has whitespace breakage, the preimage doesn't).
		 * In either case, we are fixing the whitespace breakages
		 * so we might as well take the fix together with their
		 * real change.
		 */
2349 2350 2351
		match = (tgtfix.len == fixed.len - fixstart &&
			 !memcmp(tgtfix.buf, fixed.buf + fixstart,
					     fixed.len - fixstart));
2352

2353
		strbuf_release(&tgtfix);
2354 2355 2356 2357 2358
		if (!match)
			goto unmatch_exit;

		orig += oldlen;
		target += tgtlen;
2359 2360
	}

2361 2362 2363 2364 2365 2366 2367 2368

	/*
	 * Now handle the lines in the preimage that falls beyond the
	 * end of the file (if any). They will only match if they are
	 * empty or only contain whitespace (if WS_BLANK_AT_EOL is
	 * false).
	 */
	for ( ; i < preimage->nr; i++) {
2369
		size_t fixstart = fixed.len; /* start of the fixed preimage */
2370 2371 2372 2373
		size_t oldlen = preimage->line[i].len;
		int j;

		/* Try fixing the line in the preimage */
2374
		ws_fix_copy(&fixed, orig, oldlen, ws_rule, NULL);
2375

2376 2377
		for (j = fixstart; j < fixed.len; j++)
			if (!isspace(fixed.buf[j]))
2378 2379 2380 2381 2382
				goto unmatch_exit;

		orig += oldlen;
	}

2383 2384 2385 2386 2387
	/*
	 * Yes, the preimage is based on an older version that still
	 * has whitespace breakages unfixed, and fixing them makes the
	 * hunk match.  Update the context lines in the postimage.
	 */
2388
	fixed_buf = strbuf_detach(&fixed, &fixed_len);
2389
	update_pre_post_images(preimage, postimage,
2390
			       fixed_buf, fixed_len, 0);
2391 2392 2393
	return 1;

 unmatch_exit:
2394
	strbuf_release(&fixed);
2395
	return 0;
2396 2397
}

2398 2399 2400 2401
static int find_pos(struct image *img,
		    struct image *preimage,
		    struct image *postimage,
		    int line,
2402
		    unsigned ws_rule,
2403
		    int match_beginning, int match_end)
2404
{
2405 2406 2407
	int i;
	unsigned long backwards, forwards, try;
	int backwards_lno, forwards_lno, try_lno;
2408

2409
	/*
2410
	 * If match_beginning or match_end is specified, there is no
2411 2412 2413 2414 2415 2416 2417 2418
	 * point starting from a wrong line that will never match and
	 * wander around and wait for a match at the specified end.
	 */
	if (match_beginning)
		line = 0;
	else if (match_end)
		line = img->nr - preimage->nr;

2419 2420 2421 2422 2423 2424
	/*
	 * Because the comparison is unsigned, the following test
	 * will also take care of a negative line number that can
	 * result when match_end and preimage is larger than the target.
	 */
	if ((size_t) line > img->nr)
2425 2426
		line = img->nr;

2427 2428 2429
	try = 0;
	for (i = 0; i < line; i++)
		try += img->line[i].len;
2430

2431 2432 2433 2434
	/*
	 * There's probably some smart way to do this, but I'll leave
	 * that to the smart and beautiful people. I'm simple and stupid.
	 */
2435 2436 2437 2438 2439
	backwards = try;
	backwards_lno = line;
	forwards = try;
	forwards_lno = line;
	try_lno = line;
2440

2441
	for (i = 0; ; i++) {
2442
		if (match_fragment(img, preimage, postimage,
2443
				   try, try_lno, ws_rule,
2444 2445
				   match_beginning, match_end))
			return try_lno;
2446 2447

	again:
2448
		if (backwards_lno == 0 && forwards_lno == img->nr)
2449
			break;
2450 2451

		if (i & 1) {
2452
			if (backwards_lno == 0) {
2453 2454
				i++;
				goto again;
2455
			}
2456 2457
			backwards_lno--;
			backwards -= img->line[backwards_lno].len;
2458
			try = backwards;
2459
			try_lno = backwards_lno;
2460
		} else {
2461
			if (forwards_lno == img->nr) {
2462 2463
				i++;
				goto again;
2464
			}
2465 2466
			forwards += img->line[forwards_lno].len;
			forwards_lno++;
2467
			try = forwards;
2468
			try_lno = forwards_lno;
2469 2470 2471
		}

	}
2472 2473 2474
	return -1;
}

2475
static void remove_first_line(struct image *img)
2476
{
2477 2478 2479 2480
	img->buf += img->line[0].len;
	img->len -= img->line[0].len;
	img->line++;
	img->nr--;
2481 2482
}

2483
static void remove_last_line(struct image *img)
2484
{
2485
	img->len -= img->line[--img->nr].len;
2486 2487
}

2488 2489 2490 2491 2492
/*
 * The change from "preimage" and "postimage" has been found to
 * apply at applied_pos (counts in line numbers) in "img".
 * Update "img" to remove "preimage" and replace it with "postimage".
 */
2493 2494 2495 2496
static void update_image(struct image *img,
			 int applied_pos,
			 struct image *preimage,
			 struct image *postimage)
2497
{
2498
	/*
2499 2500
	 * remove the copy of preimage at offset in img
	 * and replace it with postimage
2501
	 */
2502 2503 2504
	int i, nr;
	size_t remove_count, insert_count, applied_at = 0;
	char *result;
2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518
	int preimage_limit;

	/*
	 * If we are removing blank lines at the end of img,
	 * the preimage may extend beyond the end.
	 * If that is the case, we must be careful only to
	 * remove the part of the preimage that falls within
	 * the boundaries of img. Initialize preimage_limit
	 * to the number of lines in the preimage that falls
	 * within the boundaries.
	 */
	preimage_limit = preimage->nr;
	if (preimage_limit > img->nr - applied_pos)
		preimage_limit = img->nr - applied_pos;
2519

2520 2521 2522 2523
	for (i = 0; i < applied_pos; i++)
		applied_at += img->line[i].len;

	remove_count = 0;
2524
	for (i = 0; i < preimage_limit; i++)
2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540
		remove_count += img->line[applied_pos + i].len;
	insert_count = postimage->len;

	/* Adjust the contents */
	result = xmalloc(img->len + insert_count - remove_count + 1);
	memcpy(result, img->buf, applied_at);
	memcpy(result + applied_at, postimage->buf, postimage->len);
	memcpy(result + applied_at + postimage->len,
	       img->buf + (applied_at + remove_count),
	       img->len - (applied_at + remove_count));
	free(img->buf);
	img->buf = result;
	img->len += insert_count - remove_count;
	result[img->len] = '\0';

	/* Adjust the line table */
2541 2542
	nr = img->nr + postimage->nr - preimage_limit;
	if (preimage_limit < postimage->nr) {
2543
		/*
2544 2545
		 * NOTE: this knows that we never call remove_first_line()
		 * on anything other than pre/post image.
2546
		 */
2547 2548
		img->line = xrealloc(img->line, nr * sizeof(*img->line));
		img->line_allocated = img->line;
2549
	}
2550
	if (preimage_limit != postimage->nr)
2551
		memmove(img->line + applied_pos + postimage->nr,
2552 2553
			img->line + applied_pos + preimage_limit,
			(img->nr - (applied_pos + preimage_limit)) *
2554 2555 2556 2557
			sizeof(*img->line));
	memcpy(img->line + applied_pos,
	       postimage->line,
	       postimage->nr * sizeof(*img->line));
2558 2559 2560
	if (!allow_overlap)
		for (i = 0; i < postimage->nr; i++)
			img->line[applied_pos + i].flag |= LINE_PATCHED;
2561
	img->nr = nr;
2562 2563
}

2564 2565 2566 2567 2568
/*
 * Use the patch-hunk text in "frag" to prepare two images (preimage and
 * postimage) for the hunk.  Find lines that match "preimage" in "img" and
 * replace the part of "img" with "postimage" text.
 */
2569
static int apply_one_fragment(struct image *img, struct fragment *frag,
2570 2571
			      int inaccurate_eof, unsigned ws_rule,
			      int nth_fragment)
2572
{
2573
	int match_beginning, match_end;
2574
	const char *patch = frag->patch;
2575
	int size = frag->size;
2576 2577
	char *old, *oldlines;
	struct strbuf newlines;
2578
	int new_blank_lines_at_end = 0;
2579 2580
	int found_new_blank_lines_at_end = 0;
	int hunk_linenr = frag->linenr;
2581
	unsigned long leading, trailing;
2582 2583 2584
	int pos, applied_pos;
	struct image preimage;
	struct image postimage;
2585

2586 2587
	memset(&preimage, 0, sizeof(preimage));
	memset(&postimage, 0, sizeof(postimage));
2588
	oldlines = xmalloc(size);
2589
	strbuf_init(&newlines, size);
2590

2591
	old = oldlines;
2592
	while (size > 0) {
J
Johannes Schindelin 已提交
2593
		char first;
2594
		int len = linelen(patch, size);
2595
		int plen;
2596
		int added_blank_line = 0;
2597
		int is_blank_context = 0;
2598
		size_t start;
2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609

		if (!len)
			break;

		/*
		 * "plen" is how much of the line we should use for
		 * the actual patch data. Normally we just remove the
		 * first character on the line, but if the line is
		 * followed by "\ No newline", then we also remove the
		 * last one (which is the newline, of course).
		 */
2610
		plen = len - 1;
2611
		if (len < size && patch[len] == '\\')
2612
			plen--;
J
Johannes Schindelin 已提交
2613
		first = *patch;
2614
		if (apply_in_reverse) {
J
Johannes Schindelin 已提交
2615 2616 2617 2618 2619
			if (first == '-')
				first = '+';
			else if (first == '+')
				first = '-';
		}
2620

J
Johannes Schindelin 已提交
2621
		switch (first) {
2622 2623 2624 2625 2626
		case '\n':
			/* Newer GNU diff, empty context line */
			if (plen < 0)
				/* ... followed by '\No newline'; nothing */
				break;
2627
			*old++ = '\n';
2628
			strbuf_addch(&newlines, '\n');
2629 2630
			add_line_info(&preimage, "\n", 1, LINE_COMMON);
			add_line_info(&postimage, "\n", 1, LINE_COMMON);
2631
			is_blank_context = 1;
2632
			break;
2633
		case ' ':
2634 2635
			if (plen && (ws_rule & WS_BLANK_AT_EOF) &&
			    ws_blank_line(patch + 1, plen, ws_rule))
2636
				is_blank_context = 1;
2637
		case '-':
2638 2639
			memcpy(old, patch + 1, plen);
			add_line_info(&preimage, old, plen,
2640
				      (first == ' ' ? LINE_COMMON : 0));
2641
			old += plen;
J
Johannes Schindelin 已提交
2642
			if (first == '-')
2643 2644 2645
				break;
		/* Fall-through for ' ' */
		case '+':
2646 2647 2648 2649
			/* --no-add does not add new lines */
			if (first == '+' && no_add)
				break;

2650
			start = newlines.len;
2651 2652 2653
			if (first != '+' ||
			    !whitespace_error ||
			    ws_error_action != correct_ws_error) {
2654
				strbuf_add(&newlines, patch + 1, plen);
2655 2656
			}
			else {
2657
				ws_fix_copy(&newlines, patch + 1, plen, ws_rule, &applied_after_fixing_ws);
2658
			}
2659
			add_line_info(&postimage, newlines.buf + start, newlines.len - start,
2660 2661
				      (first == '+' ? 0 : LINE_COMMON));
			if (first == '+' &&
2662 2663
			    (ws_rule & WS_BLANK_AT_EOF) &&
			    ws_blank_line(patch + 1, plen, ws_rule))
2664
				added_blank_line = 1;
2665 2666 2667 2668 2669
			break;
		case '@': case '\\':
			/* Ignore it, we already handled it */
			break;
		default:
2670
			if (apply_verbosely)
2671
				error(_("invalid start of line: '%c'"), first);
2672 2673
			return -1;
		}
2674 2675 2676
		if (added_blank_line) {
			if (!new_blank_lines_at_end)
				found_new_blank_lines_at_end = hunk_linenr;
2677
			new_blank_lines_at_end++;
2678
		}
2679 2680
		else if (is_blank_context)
			;
2681 2682
		else
			new_blank_lines_at_end = 0;
2683 2684
		patch += len;
		size -= len;
2685
		hunk_linenr++;
2686
	}
2687
	if (inaccurate_eof &&
2688
	    old > oldlines && old[-1] == '\n' &&
2689
	    newlines.len > 0 && newlines.buf[newlines.len - 1] == '\n') {
2690
		old--;
2691
		strbuf_setlen(&newlines, newlines.len - 1);
2692
	}
2693 2694 2695

	leading = frag->leading;
	trailing = frag->trailing;
2696 2697

	/*
2698 2699
	 * A hunk to change lines at the beginning would begin with
	 * @@ -1,L +N,M @@
2700 2701
	 * but we need to be careful.  -U0 that inserts before the second
	 * line also has this pattern.
2702
	 *
2703 2704 2705 2706 2707
	 * And a hunk to add to an empty file would begin with
	 * @@ -0,0 +N,M @@
	 *
	 * In other words, a hunk that is (frag->oldpos <= 1) with or
	 * without leading context must match at the beginning.
2708
	 */
2709 2710
	match_beginning = (!frag->oldpos ||
			   (frag->oldpos == 1 && !unidiff_zero));
2711 2712 2713 2714 2715 2716 2717 2718

	/*
	 * A hunk without trailing lines must match at the end.
	 * However, we simply cannot tell if a hunk must match end
	 * from the lack of trailing lines if the patch was generated
	 * with unidiff without any context.
	 */
	match_end = !unidiff_zero && !trailing;
2719

2720
	pos = frag->newpos ? (frag->newpos - 1) : 0;
2721 2722
	preimage.buf = oldlines;
	preimage.len = old - oldlines;
2723 2724
	postimage.buf = newlines.buf;
	postimage.len = newlines.len;
2725 2726 2727
	preimage.line = preimage.line_allocated;
	postimage.line = postimage.line_allocated;

2728
	for (;;) {
2729

2730 2731
		applied_pos = find_pos(img, &preimage, &postimage, pos,
				       ws_rule, match_beginning, match_end);
2732 2733

		if (applied_pos >= 0)
2734 2735 2736 2737 2738
			break;

		/* Am I at my context limits? */
		if ((leading <= p_context) && (trailing <= p_context))
			break;
2739 2740
		if (match_beginning || match_end) {
			match_beginning = match_end = 0;
2741 2742
			continue;
		}
2743

2744 2745 2746 2747
		/*
		 * Reduce the number of context lines; reduce both
		 * leading and trailing if they are equal otherwise
		 * just reduce the larger context.
2748 2749
		 */
		if (leading >= trailing) {
2750 2751
			remove_first_line(&preimage);
			remove_first_line(&postimage);
2752 2753 2754 2755
			pos--;
			leading--;
		}
		if (trailing > leading) {
2756 2757
			remove_last_line(&preimage);
			remove_last_line(&postimage);
2758
			trailing--;
2759
		}
2760 2761
	}

2762
	if (applied_pos >= 0) {
2763
		if (new_blank_lines_at_end &&
2764
		    preimage.nr + applied_pos >= img->nr &&
2765 2766
		    (ws_rule & WS_BLANK_AT_EOF) &&
		    ws_error_action != nowarn_ws_error) {
2767 2768
			record_ws_error(WS_BLANK_AT_EOF, "+", 1,
					found_new_blank_lines_at_end);
2769 2770 2771 2772
			if (ws_error_action == correct_ws_error) {
				while (new_blank_lines_at_end--)
					remove_last_line(&postimage);
			}
2773
			/*
2774 2775 2776 2777 2778
			 * We would want to prevent write_out_results()
			 * from taking place in apply_patch() that follows
			 * the callchain led us here, which is:
			 * apply_patch->check_patch_list->check_patch->
			 * apply_data->apply_fragments->apply_one_fragment
2779
			 */
2780 2781
			if (ws_error_action == die_on_ws_error)
				apply = 0;
2782
		}
2783

2784 2785 2786 2787
		if (apply_verbosely && applied_pos != pos) {
			int offset = applied_pos - pos;
			if (apply_in_reverse)
				offset = 0 - offset;
2788 2789 2790 2791 2792
			fprintf_ln(stderr,
				   Q_("Hunk #%d succeeded at %d (offset %d line).",
				      "Hunk #%d succeeded at %d (offset %d lines).",
				      offset),
				   nth_fragment, applied_pos + 1, offset);
2793 2794
		}

2795 2796 2797 2798 2799 2800
		/*
		 * Warn if it was necessary to reduce the number
		 * of context lines.
		 */
		if ((leading != frag->leading) ||
		    (trailing != frag->trailing))
2801 2802 2803
			fprintf_ln(stderr, _("Context reduced to (%ld/%ld)"
					     " to apply fragment at %d"),
				   leading, trailing, applied_pos+1);
2804 2805 2806
		update_image(img, applied_pos, &preimage, &postimage);
	} else {
		if (apply_verbosely)
2807
			error(_("while searching for:\n%.*s"),
2808
			      (int)(old - oldlines), oldlines);
2809
	}
2810

2811
	free(oldlines);
2812
	strbuf_release(&newlines);
2813 2814 2815 2816
	free(preimage.line_allocated);
	free(postimage.line_allocated);

	return (applied_pos < 0);
2817 2818
}

2819
static int apply_binary_fragment(struct image *img, struct patch *patch)
J
Junio C Hamano 已提交
2820 2821
{
	struct fragment *fragment = patch->fragments;
2822 2823
	unsigned long len;
	void *dst;
J
Junio C Hamano 已提交
2824

2825
	if (!fragment)
2826
		return error(_("missing binary patch data for '%s'"),
2827 2828 2829 2830
			     patch->new_name ?
			     patch->new_name :
			     patch->old_name);

2831 2832 2833 2834 2835 2836 2837
	/* Binary patch is irreversible without the optional second hunk */
	if (apply_in_reverse) {
		if (!fragment->next)
			return error("cannot reverse-apply a binary patch "
				     "without the reverse hunk to '%s'",
				     patch->new_name
				     ? patch->new_name : patch->old_name);
2838
		fragment = fragment->next;
2839 2840
	}
	switch (fragment->binary_patch_method) {
J
Junio C Hamano 已提交
2841
	case BINARY_DELTA_DEFLATED:
2842
		dst = patch_delta(img->buf, img->len, fragment->patch,
2843 2844 2845
				  fragment->size, &len);
		if (!dst)
			return -1;
2846 2847 2848
		clear_image(img);
		img->buf = dst;
		img->len = len;
2849
		return 0;
J
Junio C Hamano 已提交
2850
	case BINARY_LITERAL_DEFLATED:
2851 2852 2853 2854 2855
		clear_image(img);
		img->len = fragment->size;
		img->buf = xmalloc(img->len+1);
		memcpy(img->buf, fragment->patch, img->len);
		img->buf[img->len] = '\0';
2856
		return 0;
J
Junio C Hamano 已提交
2857
	}
2858
	return -1;
J
Junio C Hamano 已提交
2859 2860
}

2861 2862 2863 2864 2865 2866
/*
 * Replace "img" with the result of applying the binary patch.
 * The binary patch data itself in patch->fragment is still kept
 * but the preimage prepared by the caller in "img" is freed here
 * or in the helper function apply_binary_fragment() this calls.
 */
2867
static int apply_binary(struct image *img, struct patch *patch)
2868
{
J
Junio C Hamano 已提交
2869
	const char *name = patch->old_name ? patch->old_name : patch->new_name;
J
Junio C Hamano 已提交
2870
	unsigned char sha1[20];
J
Junio C Hamano 已提交
2871

2872 2873
	/*
	 * For safety, we require patch index line to contain
J
Junio C Hamano 已提交
2874 2875 2876 2877 2878 2879 2880 2881
	 * full 40-byte textual SHA1 for old and new, at least for now.
	 */
	if (strlen(patch->old_sha1_prefix) != 40 ||
	    strlen(patch->new_sha1_prefix) != 40 ||
	    get_sha1_hex(patch->old_sha1_prefix, sha1) ||
	    get_sha1_hex(patch->new_sha1_prefix, sha1))
		return error("cannot apply binary patch to '%s' "
			     "without full index line", name);
J
Junio C Hamano 已提交
2882

J
Junio C Hamano 已提交
2883
	if (patch->old_name) {
2884 2885
		/*
		 * See if the old one matches what the patch
J
Junio C Hamano 已提交
2886
		 * applies to.
J
Junio C Hamano 已提交
2887
		 */
2888
		hash_sha1_file(img->buf, img->len, blob_type, sha1);
J
Junio C Hamano 已提交
2889 2890 2891 2892 2893 2894 2895 2896
		if (strcmp(sha1_to_hex(sha1), patch->old_sha1_prefix))
			return error("the patch applies to '%s' (%s), "
				     "which does not match the "
				     "current contents.",
				     name, sha1_to_hex(sha1));
	}
	else {
		/* Otherwise, the old one must be empty. */
2897
		if (img->len)
J
Junio C Hamano 已提交
2898 2899 2900
			return error("the patch applies to an empty "
				     "'%s' but it is not empty", name);
	}
J
Junio C Hamano 已提交
2901

J
Junio C Hamano 已提交
2902
	get_sha1_hex(patch->new_sha1_prefix, sha1);
D
David Rientjes 已提交
2903
	if (is_null_sha1(sha1)) {
2904
		clear_image(img);
J
Junio C Hamano 已提交
2905
		return 0; /* deletion patch */
J
Junio C Hamano 已提交
2906
	}
J
Junio C Hamano 已提交
2907

J
Junio C Hamano 已提交
2908
	if (has_sha1_file(sha1)) {
J
Junio C Hamano 已提交
2909
		/* We already have the postimage */
2910
		enum object_type type;
J
Junio C Hamano 已提交
2911
		unsigned long size;
2912
		char *result;
J
Junio C Hamano 已提交
2913

2914 2915
		result = read_sha1_file(sha1, &type, &size);
		if (!result)
J
Junio C Hamano 已提交
2916 2917 2918
			return error("the necessary postimage %s for "
				     "'%s' cannot be read",
				     patch->new_sha1_prefix, name);
2919 2920 2921
		clear_image(img);
		img->buf = result;
		img->len = size;
2922
	} else {
2923 2924
		/*
		 * We have verified buf matches the preimage;
J
Junio C Hamano 已提交
2925 2926
		 * apply the patch data to it, which is stored
		 * in the patch->fragments->{patch,size}.
J
Junio C Hamano 已提交
2927
		 */
2928
		if (apply_binary_fragment(img, patch))
2929
			return error(_("binary patch does not apply to '%s'"),
J
Junio C Hamano 已提交
2930
				     name);
J
Junio C Hamano 已提交
2931

J
Junio C Hamano 已提交
2932
		/* verify that the result matches */
2933
		hash_sha1_file(img->buf, img->len, blob_type, sha1);
J
Junio C Hamano 已提交
2934
		if (strcmp(sha1_to_hex(sha1), patch->new_sha1_prefix))
2935
			return error(_("binary patch to '%s' creates incorrect result (expecting %s, got %s)"),
2936
				name, patch->new_sha1_prefix, sha1_to_hex(sha1));
J
Junio C Hamano 已提交
2937
	}
2938

J
Junio C Hamano 已提交
2939 2940 2941
	return 0;
}

2942
static int apply_fragments(struct image *img, struct patch *patch)
J
Junio C Hamano 已提交
2943 2944 2945
{
	struct fragment *frag = patch->fragments;
	const char *name = patch->old_name ? patch->old_name : patch->new_name;
2946 2947
	unsigned ws_rule = patch->ws_rule;
	unsigned inaccurate_eof = patch->inaccurate_eof;
2948
	int nth = 0;
J
Junio C Hamano 已提交
2949 2950

	if (patch->is_binary)
2951
		return apply_binary(img, patch);
J
Junio C Hamano 已提交
2952

2953
	while (frag) {
2954 2955
		nth++;
		if (apply_one_fragment(img, frag, inaccurate_eof, ws_rule, nth)) {
2956
			error(_("patch failed: %s:%ld"), name, frag->oldpos);
J
Junio C Hamano 已提交
2957 2958 2959 2960
			if (!apply_with_reject)
				return -1;
			frag->rejected = 1;
		}
2961 2962
		frag = frag->next;
	}
2963
	return 0;
2964 2965
}

2966
static int read_blob_object(struct strbuf *buf, const unsigned char *sha1, unsigned mode)
2967
{
2968
	if (S_ISGITLINK(mode)) {
2969
		strbuf_grow(buf, 100);
2970
		strbuf_addf(buf, "Subproject commit %s\n", sha1_to_hex(sha1));
2971 2972
	} else {
		enum object_type type;
2973 2974 2975
		unsigned long sz;
		char *result;

2976
		result = read_sha1_file(sha1, &type, &sz);
2977
		if (!result)
2978
			return -1;
2979 2980
		/* XXX read_sha1_file NUL-terminates */
		strbuf_attach(buf, result, sz, sz + 1);
2981 2982 2983 2984
	}
	return 0;
}

2985 2986 2987 2988 2989 2990 2991
static int read_file_or_gitlink(struct cache_entry *ce, struct strbuf *buf)
{
	if (!ce)
		return 0;
	return read_blob_object(buf, ce->sha1, ce->ce_mode);
}

2992 2993
static struct patch *in_fn_table(const char *name)
{
2994
	struct string_list_item *item;
2995 2996 2997 2998

	if (name == NULL)
		return NULL;

2999
	item = string_list_lookup(&fn_table, name);
3000 3001 3002 3003 3004 3005
	if (item != NULL)
		return (struct patch *)item->util;

	return NULL;
}

3006 3007 3008 3009
/*
 * item->util in the filename table records the status of the path.
 * Usually it points at a patch (whose result records the contents
 * of it after applying it), but it could be PATH_WAS_DELETED for a
3010 3011 3012 3013 3014 3015 3016
 * path that a previously applied patch has already removed, or
 * PATH_TO_BE_DELETED for a path that a later patch would remove.
 *
 * The latter is needed to deal with a case where two paths A and B
 * are swapped by first renaming A to B and then renaming B to A;
 * moving A to B should not be prevented due to presense of B as we
 * will remove it in a later patch.
3017
 */
3018
#define PATH_TO_BE_DELETED ((struct patch *) -2)
3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030
#define PATH_WAS_DELETED ((struct patch *) -1)

static int to_be_deleted(struct patch *patch)
{
	return patch == PATH_TO_BE_DELETED;
}

static int was_deleted(struct patch *patch)
{
	return patch == PATH_WAS_DELETED;
}

3031 3032
static void add_to_fn_table(struct patch *patch)
{
3033
	struct string_list_item *item;
3034 3035 3036 3037 3038 3039 3040

	/*
	 * Always add new_name unless patch is a deletion
	 * This should cover the cases for normal diffs,
	 * file creations and copies
	 */
	if (patch->new_name != NULL) {
3041
		item = string_list_insert(&fn_table, patch->new_name);
3042 3043 3044 3045 3046 3047 3048 3049
		item->util = patch;
	}

	/*
	 * store a failure on rename/deletion cases because
	 * later chunks shouldn't patch old names
	 */
	if ((patch->new_name == NULL) || (patch->is_rename)) {
3050
		item = string_list_insert(&fn_table, patch->old_name);
3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062
		item->util = PATH_WAS_DELETED;
	}
}

static void prepare_fn_table(struct patch *patch)
{
	/*
	 * store information about incoming file deletion
	 */
	while (patch) {
		if ((patch->new_name == NULL) || (patch->is_rename)) {
			struct string_list_item *item;
3063
			item = string_list_insert(&fn_table, patch->old_name);
3064 3065 3066
			item->util = PATH_TO_BE_DELETED;
		}
		patch = patch->next;
3067 3068 3069
	}
}

3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081
static int checkout_target(struct cache_entry *ce, struct stat *st)
{
	struct checkout costate;

	memset(&costate, 0, sizeof(costate));
	costate.base_dir = "";
	costate.refresh_cache = 1;
	if (checkout_entry(ce, &costate, NULL) || lstat(ce->name, st))
		return error(_("cannot checkout %s"), ce->name);
	return 0;
}

3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102
static struct patch *previous_patch(struct patch *patch, int *gone)
{
	struct patch *previous;

	*gone = 0;
	if (patch->is_copy || patch->is_rename)
		return NULL; /* "git" patches do not depend on the order */

	previous = in_fn_table(patch->old_name);
	if (!previous)
		return NULL;

	if (to_be_deleted(previous))
		return NULL; /* the deletion hasn't happened yet */

	if (was_deleted(previous))
		*gone = 1;

	return previous;
}

3103 3104 3105 3106 3107 3108 3109 3110 3111 3112
static int verify_index_match(struct cache_entry *ce, struct stat *st)
{
	if (S_ISGITLINK(ce->ce_mode)) {
		if (!S_ISDIR(st->st_mode))
			return -1;
		return 0;
	}
	return ce_match_stat(ce, st, CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE);
}

3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137
#define SUBMODULE_PATCH_WITHOUT_INDEX 1

static int load_patch_target(struct strbuf *buf,
			     struct cache_entry *ce,
			     struct stat *st,
			     const char *name,
			     unsigned expected_mode)
{
	if (cached) {
		if (read_file_or_gitlink(ce, buf))
			return error(_("read of %s failed"), name);
	} else if (name) {
		if (S_ISGITLINK(expected_mode)) {
			if (ce)
				return read_file_or_gitlink(ce, buf);
			else
				return SUBMODULE_PATCH_WITHOUT_INDEX;
		} else {
			if (read_old_data(st, name, buf))
				return error(_("read of %s failed"), name);
		}
	}
	return 0;
}

3138 3139 3140 3141 3142 3143 3144
/*
 * We are about to apply "patch"; populate the "image" with the
 * current version we have, from the working tree or from the index,
 * depending on the situation e.g. --cached/--index.  If we are
 * applying a non-git patch that incrementally updates the tree,
 * we read from the result of a previous diff.
 */
3145 3146
static int load_preimage(struct image *image,
			 struct patch *patch, struct stat *st, struct cache_entry *ce)
3147
{
3148
	struct strbuf buf = STRBUF_INIT;
3149 3150
	size_t len;
	char *img;
3151 3152
	struct patch *previous;
	int status;
3153

3154 3155 3156 3157 3158
	previous = previous_patch(patch, &status);
	if (status)
		return error(_("path %s has been renamed/deleted"),
			     patch->old_name);
	if (previous) {
3159
		/* We have a patched copy in memory; use that. */
3160
		strbuf_add(&buf, previous->result, previous->resultsize);
3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175
	} else {
		status = load_patch_target(&buf, ce, st,
					   patch->old_name, patch->old_mode);
		if (status < 0)
			return status;
		else if (status == SUBMODULE_PATCH_WITHOUT_INDEX) {
			/*
			 * There is no way to apply subproject
			 * patch without looking at the index.
			 * NEEDSWORK: shouldn't this be flagged
			 * as an error???
			 */
			free_fragment_list(patch->fragments);
			patch->fragments = NULL;
		} else if (status) {
3176
			return error(_("read of %s failed"), patch->old_name);
3177 3178
		}
	}
3179

3180
	img = strbuf_detach(&buf, &len);
3181 3182 3183
	prepare_image(image, img, len, !patch->is_binary);
	return 0;
}
3184

3185 3186 3187 3188 3189 3190
static int three_way_merge(struct image *image,
			   char *path,
			   const unsigned char *base,
			   const unsigned char *ours,
			   const unsigned char *theirs)
{
3191 3192 3193
	mmfile_t base_file, our_file, their_file;
	mmbuffer_t result = { NULL };
	int status;
3194

3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211
	read_mmblob(&base_file, base);
	read_mmblob(&our_file, ours);
	read_mmblob(&their_file, theirs);
	status = ll_merge(&result, path,
			  &base_file, "base",
			  &our_file, "ours",
			  &their_file, "theirs", NULL);
	free(base_file.ptr);
	free(our_file.ptr);
	free(their_file.ptr);
	if (status < 0 || !result.ptr) {
		free(result.ptr);
		return -1;
	}
	clear_image(image);
	image->buf = result.ptr;
	image->len = result.size;
3212

3213
	return status;
3214 3215
}

3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254
/*
 * When directly falling back to add/add three-way merge, we read from
 * the current contents of the new_name.  In no cases other than that
 * this function will be called.
 */
static int load_current(struct image *image, struct patch *patch)
{
	struct strbuf buf = STRBUF_INIT;
	int status, pos;
	size_t len;
	char *img;
	struct stat st;
	struct cache_entry *ce;
	char *name = patch->new_name;
	unsigned mode = patch->new_mode;

	if (!patch->is_new)
		die("BUG: patch to %s is not a creation", patch->old_name);

	pos = cache_name_pos(name, strlen(name));
	if (pos < 0)
		return error(_("%s: does not exist in index"), name);
	ce = active_cache[pos];
	if (lstat(name, &st)) {
		if (errno != ENOENT)
			return error(_("%s: %s"), name, strerror(errno));
		if (checkout_target(ce, &st))
			return -1;
	}
	if (verify_index_match(ce, &st))
		return error(_("%s: does not match index"), name);

	status = load_patch_target(&buf, ce, &st, name, mode);
	if (status < 0)
		return status;
	else if (status)
		return -1;
	img = strbuf_detach(&buf, &len);
	prepare_image(image, img, len, !patch->is_binary);
3255 3256 3257
	return 0;
}

3258 3259
static int try_threeway(struct image *image, struct patch *patch,
			struct stat *st, struct cache_entry *ce)
3260
{
3261 3262 3263
	unsigned char pre_sha1[20], post_sha1[20], our_sha1[20];
	struct strbuf buf = STRBUF_INIT;
	size_t len;
3264
	int status;
3265 3266 3267 3268
	char *img;
	struct image tmp_image;

	/* No point falling back to 3-way merge in these cases */
3269
	if (patch->is_delete ||
3270 3271
	    S_ISGITLINK(patch->old_mode) || S_ISGITLINK(patch->new_mode))
		return -1;
3272

3273
	/* Preimage the patch was prepared for */
3274 3275 3276 3277
	if (patch->is_new)
		write_sha1_file("", 0, blob_type, pre_sha1);
	else if (get_sha1(patch->old_sha1_prefix, pre_sha1) ||
		 read_blob_object(&buf, pre_sha1, patch->old_mode))
3278
		return error("repository lacks the necessary blob to fall back on 3-way merge.");
3279 3280 3281

	fprintf(stderr, "Falling back to three-way merge...\n");

3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293
	img = strbuf_detach(&buf, &len);
	prepare_image(&tmp_image, img, len, 1);
	/* Apply the patch to get the post image */
	if (apply_fragments(&tmp_image, patch) < 0) {
		clear_image(&tmp_image);
		return -1;
	}
	/* post_sha1[] is theirs */
	write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, post_sha1);
	clear_image(&tmp_image);

	/* our_sha1[] is ours */
3294 3295 3296 3297 3298 3299 3300 3301 3302
	if (patch->is_new) {
		if (load_current(&tmp_image, patch))
			return error("cannot read the current contents of '%s'",
				     patch->new_name);
	} else {
		if (load_preimage(&tmp_image, patch, st, ce))
			return error("cannot read the current contents of '%s'",
				     patch->old_name);
	}
3303 3304 3305 3306
	write_sha1_file(tmp_image.buf, tmp_image.len, blob_type, our_sha1);
	clear_image(&tmp_image);

	/* in-core three-way merge between post and our using pre as base */
3307 3308 3309 3310 3311 3312 3313 3314 3315
	status = three_way_merge(image, patch->new_name,
				 pre_sha1, our_sha1, post_sha1);
	if (status < 0) {
		fprintf(stderr, "Failed to fall back on three-way merge...\n");
		return status;
	}

	if (status) {
		patch->conflicted_threeway = 1;
3316 3317 3318 3319
		if (patch->is_new)
			hashclr(patch->threeway_stage[0]);
		else
			hashcpy(patch->threeway_stage[0], pre_sha1);
3320 3321 3322 3323 3324
		hashcpy(patch->threeway_stage[1], our_sha1);
		hashcpy(patch->threeway_stage[2], post_sha1);
		fprintf(stderr, "Applied patch to '%s' with conflicts.\n", patch->new_name);
	} else {
		fprintf(stderr, "Applied patch to '%s' cleanly.\n", patch->new_name);
3325 3326 3327 3328
	}
	return 0;
}

3329
static int apply_data(struct patch *patch, struct stat *st, struct cache_entry *ce)
3330
{
3331 3332 3333 3334
	struct image image;

	if (load_preimage(&image, patch, st, ce) < 0)
		return -1;
3335

3336 3337
	if (patch->direct_to_threeway ||
	    apply_fragments(&image, patch) < 0) {
3338 3339
		/* Note: with --reject, apply_fragments() returns 0 */
		if (!threeway || try_threeway(&image, patch, st, ce) < 0)
3340 3341
			return -1;
	}
3342 3343
	patch->result = image.buf;
	patch->resultsize = image.len;
3344
	add_to_fn_table(patch);
3345
	free(image.line_allocated);
3346

3347
	if (0 < patch->is_delete && patch->resultsize)
3348
		return error(_("removal patch leaves file contents"));
3349

3350
	return 0;
3351 3352
}

3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363
/*
 * If "patch" that we are looking at modifies or deletes what we have,
 * we would want it not to lose any local modification we have, either
 * in the working tree or in the index.
 *
 * This also decides if a non-git patch is a creation patch or a
 * modification to an existing empty file.  We do not check the state
 * of the current tree for a creation patch in this function; the caller
 * check_patch() separately makes sure (and errors out otherwise) that
 * the path the patch creates does not exist in the current tree.
 */
3364
static int check_preimage(struct patch *patch, struct cache_entry **ce, struct stat *st)
3365 3366
{
	const char *old_name = patch->old_name;
3367 3368
	struct patch *previous = NULL;
	int stat_ret = 0, status;
3369
	unsigned st_mode = 0;
3370

3371 3372
	if (!old_name)
		return 0;
3373

3374
	assert(patch->is_new <= 0);
3375
	previous = previous_patch(patch, &status);
J
Junio C Hamano 已提交
3376

3377 3378 3379 3380
	if (status)
		return error(_("path %s has been renamed/deleted"), old_name);
	if (previous) {
		st_mode = previous->new_mode;
3381
	} else if (!cached) {
3382 3383
		stat_ret = lstat(old_name, st);
		if (stat_ret && errno != ENOENT)
3384
			return error(_("%s: %s"), old_name, strerror(errno));
3385
	}
J
Junio C Hamano 已提交
3386

3387
	if (check_index && !previous) {
3388 3389 3390 3391
		int pos = cache_name_pos(old_name, strlen(old_name));
		if (pos < 0) {
			if (patch->is_new < 0)
				goto is_new;
3392
			return error(_("%s: does not exist in index"), old_name);
3393 3394 3395
		}
		*ce = active_cache[pos];
		if (stat_ret < 0) {
3396
			if (checkout_target(*ce, st))
3397 3398 3399
				return -1;
		}
		if (!cached && verify_index_match(*ce, st))
3400
			return error(_("%s: does not match index"), old_name);
3401 3402 3403
		if (cached)
			st_mode = (*ce)->ce_mode;
	} else if (stat_ret < 0) {
3404
		if (patch->is_new < 0)
3405
			goto is_new;
3406
		return error(_("%s: %s"), old_name, strerror(errno));
3407
	}
3408

3409
	if (!cached && !previous)
3410 3411 3412 3413 3414 3415 3416
		st_mode = ce_mode_from_stat(*ce, st->st_mode);

	if (patch->is_new < 0)
		patch->is_new = 0;
	if (!patch->old_mode)
		patch->old_mode = st_mode;
	if ((st_mode ^ patch->old_mode) & S_IFMT)
3417
		return error(_("%s: wrong type"), old_name);
3418
	if (st_mode != patch->old_mode)
3419
		warning(_("%s has type %o, expected %o"),
3420
			old_name, st_mode, patch->old_mode);
3421
	if (!patch->new_mode && !patch->is_delete)
3422
		patch->new_mode = st_mode;
3423 3424 3425 3426 3427
	return 0;

 is_new:
	patch->is_new = 1;
	patch->is_delete = 0;
3428
	free(patch->old_name);
3429 3430 3431 3432
	patch->old_name = NULL;
	return 0;
}

3433 3434 3435 3436 3437

#define EXISTS_IN_INDEX 1
#define EXISTS_IN_WORKTREE 2

static int check_to_create(const char *new_name, int ok_if_exists)
3438 3439
{
	struct stat nst;
3440 3441 3442 3443 3444 3445 3446 3447

	if (check_index &&
	    cache_name_pos(new_name, strlen(new_name)) >= 0 &&
	    !ok_if_exists)
		return EXISTS_IN_INDEX;
	if (cached)
		return 0;

3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460
	if (!lstat(new_name, &nst)) {
		if (S_ISDIR(nst.st_mode) || ok_if_exists)
			return 0;
		/*
		 * A leading component of new_name might be a symlink
		 * that is going to be removed with this patch, but
		 * still pointing at somewhere that has the path.
		 * In such a case, path "new_name" does not exist as
		 * far as git is concerned.
		 */
		if (has_symlink_leading_path(new_name, strlen(new_name)))
			return 0;

3461 3462
		return EXISTS_IN_WORKTREE;
	} else if ((errno != ENOENT) && (errno != ENOTDIR)) {
3463
		return error("%s: %s", new_name, strerror(errno));
3464
	}
3465 3466 3467
	return 0;
}

3468 3469 3470 3471
/*
 * Check and apply the patch in-core; leave the result in patch->result
 * for the caller to write it out to the final destination.
 */
3472
static int check_patch(struct patch *patch)
3473 3474 3475 3476 3477 3478
{
	struct stat st;
	const char *old_name = patch->old_name;
	const char *new_name = patch->new_name;
	const char *name = old_name ? old_name : new_name;
	struct cache_entry *ce = NULL;
3479
	struct patch *tpatch;
3480 3481 3482 3483 3484 3485 3486 3487 3488 3489
	int ok_if_exists;
	int status;

	patch->rejected = 1; /* we will drop this after we succeed */

	status = check_preimage(patch, &ce, &st);
	if (status)
		return status;
	old_name = patch->old_name;

3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503
	/*
	 * A type-change diff is always split into a patch to delete
	 * old, immediately followed by a patch to create new (see
	 * diff.c::run_diff()); in such a case it is Ok that the entry
	 * to be deleted by the previous patch is still in the working
	 * tree and in the index.
	 *
	 * A patch to swap-rename between A and B would first rename A
	 * to B and then rename B to A.  While applying the first one,
	 * the presense of B should not stop A from getting renamed to
	 * B; ask to_be_deleted() about the later rename.  Removal of
	 * B and rename from A to B is handled the same way by asking
	 * was_deleted().
	 */
3504
	if ((tpatch = in_fn_table(new_name)) &&
3505
	    (was_deleted(tpatch) || to_be_deleted(tpatch)))
3506 3507 3508 3509
		ok_if_exists = 1;
	else
		ok_if_exists = 0;

3510 3511
	if (new_name &&
	    ((0 < patch->is_new) | (0 < patch->is_rename) | patch->is_copy)) {
3512 3513
		int err = check_to_create(new_name, ok_if_exists);

3514 3515 3516
		if (err && threeway) {
			patch->direct_to_threeway = 1;
		} else switch (err) {
3517 3518 3519
		case 0:
			break; /* happy */
		case EXISTS_IN_INDEX:
3520
			return error(_("%s: already exists in index"), new_name);
3521 3522 3523 3524 3525 3526
			break;
		case EXISTS_IN_WORKTREE:
			return error(_("%s: already exists in working directory"),
				     new_name);
		default:
			return err;
3527
		}
3528

3529
		if (!patch->new_mode) {
3530
			if (0 < patch->is_new)
3531 3532 3533 3534
				patch->new_mode = S_IFREG | 0644;
			else
				patch->new_mode = patch->old_mode;
		}
3535
	}
3536 3537 3538 3539 3540

	if (new_name && old_name) {
		int same = !strcmp(old_name, new_name);
		if (!patch->new_mode)
			patch->new_mode = patch->old_mode;
3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552
		if ((patch->old_mode ^ patch->new_mode) & S_IFMT) {
			if (same)
				return error(_("new mode (%o) of %s does not "
					       "match old mode (%o)"),
					patch->new_mode, new_name,
					patch->old_mode);
			else
				return error(_("new mode (%o) of %s does not "
					       "match old mode (%o) of %s"),
					patch->new_mode, new_name,
					patch->old_mode, old_name);
		}
3553
	}
3554

3555
	if (apply_data(patch, &st, ce) < 0)
3556
		return error(_("%s: patch does not apply"), name);
J
Junio C Hamano 已提交
3557
	patch->rejected = 0;
3558
	return 0;
3559 3560
}

3561
static int check_patch_list(struct patch *patch)
3562
{
3563
	int err = 0;
3564

3565
	prepare_fn_table(patch);
3566
	while (patch) {
J
Junio C Hamano 已提交
3567 3568
		if (apply_verbosely)
			say_patch_name(stderr,
3569
				       _("Checking patch %s..."), patch);
3570 3571
		err |= check_patch(patch);
		patch = patch->next;
3572
	}
3573
	return err;
3574 3575
}

3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589
/* This function tries to read the sha1 from the current index */
static int get_current_sha1(const char *path, unsigned char *sha1)
{
	int pos;

	if (read_cache() < 0)
		return -1;
	pos = cache_name_pos(path, strlen(path));
	if (pos < 0)
		return -1;
	hashcpy(sha1, active_cache[pos]->sha1);
	return 0;
}

3590 3591
/* Build an index that contains the just the files needed for a 3way merge */
static void build_fake_ancestor(struct patch *list, const char *filename)
3592 3593
{
	struct patch *patch;
3594
	struct index_state result = { NULL };
3595
	int fd;
3596 3597 3598 3599 3600 3601 3602

	/* Once we start supporting the reverse patch, it may be
	 * worth showing the new sha1 prefix, but until then...
	 */
	for (patch = list; patch; patch = patch->next) {
		const unsigned char *sha1_ptr;
		unsigned char sha1[20];
3603
		struct cache_entry *ce;
3604 3605 3606
		const char *name;

		name = patch->old_name ? patch->old_name : patch->new_name;
3607
		if (0 < patch->is_new)
3608
			continue;
3609
		else if (get_sha1_blob(patch->old_sha1_prefix, sha1))
3610 3611
			/* git diff has no index line for mode/type changes */
			if (!patch->lines_added && !patch->lines_deleted) {
3612
				if (get_current_sha1(patch->old_name, sha1))
3613 3614 3615 3616 3617 3618
					die("mode change for %s, which is not "
						"in current HEAD", name);
				sha1_ptr = sha1;
			} else
				die("sha1 information is lacking or useless "
					"(%s).", name);
3619 3620
		else
			sha1_ptr = sha1;
3621

3622
		ce = make_cache_entry(patch->old_mode, sha1_ptr, name, 0, 0);
3623
		if (!ce)
3624
			die(_("make_cache_entry failed for path '%s'"), name);
3625 3626
		if (add_index_entry(&result, ce, ADD_CACHE_OK_TO_ADD))
			die ("Could not add %s to temporary index", name);
3627
	}
3628 3629 3630 3631 3632 3633

	fd = open(filename, O_WRONLY | O_CREAT, 0666);
	if (fd < 0 || write_index(&result, fd) || close(fd))
		die ("Could not write temporary index to %s", filename);

	discard_index(&result);
3634 3635
}

3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646
static void stat_patch_list(struct patch *patch)
{
	int files, adds, dels;

	for (files = adds = dels = 0 ; patch ; patch = patch->next) {
		files++;
		adds += patch->lines_added;
		dels += patch->lines_deleted;
		show_stats(patch);
	}

3647
	print_stat_summary(stdout, files, adds, dels);
3648 3649
}

J
Junio C Hamano 已提交
3650 3651 3652 3653
static void numstat_patch_list(struct patch *patch)
{
	for ( ; patch; patch = patch->next) {
		const char *name;
3654
		name = patch->new_name ? patch->new_name : patch->old_name;
3655 3656 3657
		if (patch->is_binary)
			printf("-\t-\t");
		else
3658 3659
			printf("%d\t%d\t", patch->lines_added, patch->lines_deleted);
		write_name_quoted(name, stdout, line_termination);
J
Junio C Hamano 已提交
3660 3661 3662
	}
}

3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707
static void show_file_mode_name(const char *newdelete, unsigned int mode, const char *name)
{
	if (mode)
		printf(" %s mode %06o %s\n", newdelete, mode, name);
	else
		printf(" %s %s\n", newdelete, name);
}

static void show_mode_change(struct patch *p, int show_name)
{
	if (p->old_mode && p->new_mode && p->old_mode != p->new_mode) {
		if (show_name)
			printf(" mode change %06o => %06o %s\n",
			       p->old_mode, p->new_mode, p->new_name);
		else
			printf(" mode change %06o => %06o\n",
			       p->old_mode, p->new_mode);
	}
}

static void show_rename_copy(struct patch *p)
{
	const char *renamecopy = p->is_rename ? "rename" : "copy";
	const char *old, *new;

	/* Find common prefix */
	old = p->old_name;
	new = p->new_name;
	while (1) {
		const char *slash_old, *slash_new;
		slash_old = strchr(old, '/');
		slash_new = strchr(new, '/');
		if (!slash_old ||
		    !slash_new ||
		    slash_old - old != slash_new - new ||
		    memcmp(old, new, slash_new - new))
			break;
		old = slash_old + 1;
		new = slash_new + 1;
	}
	/* p->old_name thru old is the common prefix, and old and new
	 * through the end of names are renames
	 */
	if (old != p->old_name)
		printf(" %s %.*s{%s => %s} (%d%%)\n", renamecopy,
3708
		       (int)(old - p->old_name), p->old_name,
3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740
		       old, new, p->score);
	else
		printf(" %s %s => %s (%d%%)\n", renamecopy,
		       p->old_name, p->new_name, p->score);
	show_mode_change(p, 0);
}

static void summary_patch_list(struct patch *patch)
{
	struct patch *p;

	for (p = patch; p; p = p->next) {
		if (p->is_new)
			show_file_mode_name("create", p->new_mode, p->new_name);
		else if (p->is_delete)
			show_file_mode_name("delete", p->old_mode, p->old_name);
		else {
			if (p->is_rename || p->is_copy)
				show_rename_copy(p);
			else {
				if (p->score) {
					printf(" rewrite %s (%d%%)\n",
					       p->new_name, p->score);
					show_mode_change(p, 0);
				}
				else
					show_mode_change(p, 1);
			}
		}
	}
}

3741 3742 3743 3744 3745 3746 3747
static void patch_stats(struct patch *patch)
{
	int lines = patch->lines_added + patch->lines_deleted;

	if (lines > max_change)
		max_change = lines;
	if (patch->old_name) {
3748 3749 3750
		int len = quote_c_style(patch->old_name, NULL, NULL, 0);
		if (!len)
			len = strlen(patch->old_name);
3751 3752 3753 3754
		if (len > max_len)
			max_len = len;
	}
	if (patch->new_name) {
3755 3756 3757
		int len = quote_c_style(patch->new_name, NULL, NULL, 0);
		if (!len)
			len = strlen(patch->new_name);
3758 3759 3760
		if (len > max_len)
			max_len = len;
	}
3761 3762
}

3763
static void remove_file(struct patch *patch, int rmdir_empty)
3764
{
3765
	if (update_index) {
3766
		if (remove_file_from_cache(patch->old_name) < 0)
3767
			die(_("unable to remove %s from index"), patch->old_name);
3768
	}
3769
	if (!cached) {
3770
		if (!remove_or_warn(patch->old_mode, patch->old_name) && rmdir_empty) {
3771
			remove_path(patch->old_name);
3772 3773
		}
	}
3774 3775 3776 3777 3778 3779 3780 3781 3782
}

static void add_index_file(const char *path, unsigned mode, void *buf, unsigned long size)
{
	struct stat st;
	struct cache_entry *ce;
	int namelen = strlen(path);
	unsigned ce_size = cache_entry_size(namelen);

3783
	if (!update_index)
3784 3785
		return;

3786
	ce = xcalloc(1, ce_size);
3787 3788
	memcpy(ce->name, path, namelen);
	ce->ce_mode = create_ce_mode(mode);
3789 3790
	ce->ce_flags = create_ce_flags(0);
	ce->ce_namelen = namelen;
3791 3792 3793 3794
	if (S_ISGITLINK(mode)) {
		const char *s = buf;

		if (get_sha1_hex(s + strlen("Subproject commit "), ce->sha1))
3795
			die(_("corrupt patch for subproject %s"), path);
3796 3797 3798
	} else {
		if (!cached) {
			if (lstat(path, &st) < 0)
3799
				die_errno(_("unable to stat newly created file '%s'"),
3800
					  path);
3801 3802 3803
			fill_stat_cache_info(ce, &st);
		}
		if (write_sha1_file(buf, size, blob_type, ce->sha1) < 0)
3804
			die(_("unable to create backing store for newly created file %s"), path);
3805
	}
3806
	if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
3807
		die(_("unable to add cache entry for %s"), path);
3808 3809
}

3810 3811
static int try_create_file(const char *path, unsigned int mode, const char *buf, unsigned long size)
{
3812
	int fd;
3813
	struct strbuf nbuf = STRBUF_INIT;
3814

3815 3816 3817 3818 3819 3820 3821
	if (S_ISGITLINK(mode)) {
		struct stat st;
		if (!lstat(path, &st) && S_ISDIR(st.st_mode))
			return 0;
		return mkdir(path, 0777);
	}

3822
	if (has_symlinks && S_ISLNK(mode))
3823 3824 3825
		/* Although buf:size is counted string, it also is NUL
		 * terminated.
		 */
3826
		return symlink(buf, path);
3827 3828 3829 3830 3831

	fd = open(path, O_CREAT | O_EXCL | O_WRONLY, (mode & 0100) ? 0777 : 0666);
	if (fd < 0)
		return -1;

3832 3833 3834
	if (convert_to_working_tree(path, buf, size, &nbuf)) {
		size = nbuf.len;
		buf  = nbuf.buf;
3835
	}
3836 3837
	write_or_die(fd, buf, size);
	strbuf_release(&nbuf);
3838

3839
	if (close(fd) < 0)
3840
		die_errno(_("closing file '%s'"), path);
3841 3842 3843
	return 0;
}

3844 3845 3846 3847 3848
/*
 * We optimistically assume that the directories exist,
 * which is true 99% of the time anyway. If they don't,
 * we create them and try again.
 */
3849
static void create_one_file(char *path, unsigned mode, const char *buf, unsigned long size)
3850
{
3851 3852
	if (cached)
		return;
3853 3854
	if (!try_create_file(path, mode, buf, size))
		return;
3855

3856
	if (errno == ENOENT) {
3857 3858
		if (safe_create_leading_directories(path))
			return;
3859 3860
		if (!try_create_file(path, mode, buf, size))
			return;
3861 3862
	}

J
Johannes Schindelin 已提交
3863
	if (errno == EEXIST || errno == EACCES) {
3864 3865 3866 3867
		/* We may be trying to create a file where a directory
		 * used to be.
		 */
		struct stat st;
3868
		if (!lstat(path, &st) && (!S_ISDIR(st.st_mode) || !rmdir(path)))
3869 3870 3871
			errno = EEXIST;
	}

3872 3873
	if (errno == EEXIST) {
		unsigned int nr = getpid();
3874

3875
		for (;;) {
3876 3877
			char newpath[PATH_MAX];
			mksnpath(newpath, sizeof(newpath), "%s~%u", path, nr);
3878 3879 3880
			if (!try_create_file(newpath, mode, buf, size)) {
				if (!rename(newpath, path))
					return;
3881
				unlink_or_warn(newpath);
3882 3883 3884 3885
				break;
			}
			if (errno != EEXIST)
				break;
3886 3887
			++nr;
		}
3888
	}
3889
	die_errno(_("unable to write file '%s' mode %o"), path, mode);
3890 3891
}

3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910
static void add_conflicted_stages_file(struct patch *patch)
{
	int stage, namelen;
	unsigned ce_size, mode;
	struct cache_entry *ce;

	if (!update_index)
		return;
	namelen = strlen(patch->new_name);
	ce_size = cache_entry_size(namelen);
	mode = patch->new_mode ? patch->new_mode : (S_IFREG | 0644);

	remove_file_from_cache(patch->new_name);
	for (stage = 1; stage < 4; stage++) {
		if (is_null_sha1(patch->threeway_stage[stage - 1]))
			continue;
		ce = xcalloc(1, ce_size);
		memcpy(ce->name, patch->new_name, namelen);
		ce->ce_mode = create_ce_mode(mode);
3911 3912
		ce->ce_flags = create_ce_flags(stage);
		ce->ce_namelen = namelen;
3913 3914 3915 3916 3917 3918
		hashcpy(ce->sha1, patch->threeway_stage[stage - 1]);
		if (add_cache_entry(ce, ADD_CACHE_OK_TO_ADD) < 0)
			die(_("unable to add cache entry for %s"), patch->new_name);
	}
}

3919 3920
static void create_file(struct patch *patch)
{
3921
	char *path = patch->new_name;
3922 3923 3924 3925 3926 3927
	unsigned mode = patch->new_mode;
	unsigned long size = patch->resultsize;
	char *buf = patch->result;

	if (!mode)
		mode = S_IFREG | 0644;
3928
	create_one_file(path, mode, buf, size);
3929 3930 3931 3932 3933

	if (patch->conflicted_threeway)
		add_conflicted_stages_file(patch);
	else
		add_index_file(path, mode, buf, size);
3934 3935
}

3936 3937
/* phase zero is to remove, phase one is to create */
static void write_out_one_result(struct patch *patch, int phase)
3938 3939
{
	if (patch->is_delete > 0) {
3940
		if (phase == 0)
3941
			remove_file(patch, 1);
3942 3943 3944
		return;
	}
	if (patch->is_new > 0 || patch->is_copy) {
3945 3946
		if (phase == 1)
			create_file(patch);
3947 3948 3949 3950 3951 3952
		return;
	}
	/*
	 * Rename or modification boils down to the same
	 * thing: remove the old, write the new
	 */
3953
	if (phase == 0)
3954
		remove_file(patch, patch->is_rename);
3955
	if (phase == 1)
J
Junio C Hamano 已提交
3956
		create_file(patch);
3957 3958
}

J
Junio C Hamano 已提交
3959 3960
static int write_out_one_reject(struct patch *patch)
{
3961 3962
	FILE *rej;
	char namebuf[PATH_MAX];
J
Junio C Hamano 已提交
3963
	struct fragment *frag;
3964
	int cnt = 0;
3965
	struct strbuf sb = STRBUF_INIT;
J
Junio C Hamano 已提交
3966

3967
	for (cnt = 0, frag = patch->fragments; frag; frag = frag->next) {
J
Junio C Hamano 已提交
3968 3969
		if (!frag->rejected)
			continue;
3970 3971 3972
		cnt++;
	}

J
Junio C Hamano 已提交
3973 3974 3975
	if (!cnt) {
		if (apply_verbosely)
			say_patch_name(stderr,
3976
				       _("Applied patch %s cleanly."), patch);
3977
		return 0;
J
Junio C Hamano 已提交
3978
	}
3979 3980 3981 3982 3983

	/* This should not happen, because a removal patch that leaves
	 * contents are marked "rejected" at the patch level.
	 */
	if (!patch->new_name)
3984
		die(_("internal error"));
3985

J
Junio C Hamano 已提交
3986
	/* Say this even without --verbose */
3987 3988 3989 3990 3991 3992
	strbuf_addf(&sb, Q_("Applying patch %%s with %d reject...",
			    "Applying patch %%s with %d rejects...",
			    cnt),
		    cnt);
	say_patch_name(stderr, sb.buf, patch);
	strbuf_release(&sb);
J
Junio C Hamano 已提交
3993

3994 3995 3996
	cnt = strlen(patch->new_name);
	if (ARRAY_SIZE(namebuf) <= cnt + 5) {
		cnt = ARRAY_SIZE(namebuf) - 5;
3997
		warning(_("truncating .rej filename to %.*s.rej"),
3998 3999 4000 4001 4002 4003 4004
			cnt - 1, patch->new_name);
	}
	memcpy(namebuf, patch->new_name, cnt);
	memcpy(namebuf + cnt, ".rej", 5);

	rej = fopen(namebuf, "w");
	if (!rej)
4005
		return error(_("cannot open %s: %s"), namebuf, strerror(errno));
4006 4007 4008 4009 4010 4011 4012 4013

	/* Normal git tools never deal with .rej, so do not pretend
	 * this is a git patch by saying --git nor give extended
	 * headers.  While at it, maybe please "kompare" that wants
	 * the trailing TAB and some garbage at the end of line ;-).
	 */
	fprintf(rej, "diff a/%s b/%s\t(rejected hunks)\n",
		patch->new_name, patch->new_name);
4014
	for (cnt = 1, frag = patch->fragments;
4015 4016 4017
	     frag;
	     cnt++, frag = frag->next) {
		if (!frag->rejected) {
4018
			fprintf_ln(stderr, _("Hunk #%d applied cleanly."), cnt);
4019
			continue;
J
Junio C Hamano 已提交
4020
		}
4021
		fprintf_ln(stderr, _("Rejected hunk #%d."), cnt);
4022
		fprintf(rej, "%.*s", frag->size, frag->patch);
J
Junio C Hamano 已提交
4023
		if (frag->patch[frag->size-1] != '\n')
4024
			fputc('\n', rej);
J
Junio C Hamano 已提交
4025
	}
4026 4027
	fclose(rej);
	return -1;
4028 4029
}

4030
static int write_out_results(struct patch *list)
4031
{
4032
	int phase;
J
Junio C Hamano 已提交
4033 4034
	int errs = 0;
	struct patch *l;
4035
	struct string_list cpath = STRING_LIST_INIT_DUP;
4036 4037

	for (phase = 0; phase < 2; phase++) {
J
Junio C Hamano 已提交
4038
		l = list;
4039
		while (l) {
J
Junio C Hamano 已提交
4040 4041
			if (l->rejected)
				errs = 1;
4042
			else {
J
Junio C Hamano 已提交
4043
				write_out_one_result(l, phase);
4044 4045 4046 4047 4048 4049 4050 4051
				if (phase == 1) {
					if (write_out_one_reject(l))
						errs = 1;
					if (l->conflicted_threeway) {
						string_list_append(&cpath, l->new_name);
						errs = 1;
					}
				}
J
Junio C Hamano 已提交
4052
			}
4053 4054
			l = l->next;
		}
4055
	}
4056 4057 4058 4059 4060 4061 4062 4063

	if (cpath.nr) {
		struct string_list_item *item;

		sort_string_list(&cpath);
		for_each_string_list_item(item, &cpath)
			fprintf(stderr, "U %s\n", item->string);
		string_list_clear(&cpath, 0);
4064 4065

		rerere(0);
4066 4067
	}

J
Junio C Hamano 已提交
4068
	return errs;
4069 4070
}

4071
static struct lock_file lock_file;
4072

J
Junio C Hamano 已提交
4073 4074 4075 4076 4077 4078
static struct string_list limit_by_name;
static int has_include;
static void add_name_limit(const char *name, int exclude)
{
	struct string_list_item *it;

4079
	it = string_list_append(&limit_by_name, name);
J
Junio C Hamano 已提交
4080 4081
	it->util = exclude ? NULL : (void *) 1;
}
4082 4083 4084

static int use_patch(struct patch *p)
{
J
Jason Riedy 已提交
4085
	const char *pathname = p->new_name ? p->new_name : p->old_name;
J
Junio C Hamano 已提交
4086 4087 4088
	int i;

	/* Paths outside are not touched regardless of "--include" */
4089 4090 4091 4092 4093 4094
	if (0 < prefix_length) {
		int pathlen = strlen(pathname);
		if (pathlen <= prefix_length ||
		    memcmp(prefix, pathname, prefix_length))
			return 0;
	}
J
Junio C Hamano 已提交
4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108

	/* See if it matches any of exclude/include rule */
	for (i = 0; i < limit_by_name.nr; i++) {
		struct string_list_item *it = &limit_by_name.items[i];
		if (!fnmatch(it->string, pathname, 0))
			return (it->util != NULL);
	}

	/*
	 * If we had any include, a path that does not match any rule is
	 * not used.  Otherwise, we saw bunch of exclude rules (or none)
	 * and such a path is used.
	 */
	return !has_include;
4109 4110
}

J
Junio C Hamano 已提交
4111

4112
static void prefix_one(char **name)
4113
{
4114 4115 4116 4117 4118
	char *old_name = *name;
	if (!old_name)
		return;
	*name = xstrdup(prefix_filename(prefix, prefix_length, *name));
	free(old_name);
4119 4120 4121 4122
}

static void prefix_patches(struct patch *p)
{
4123
	if (!prefix || p->is_toplevel_relative)
4124 4125
		return;
	for ( ; p; p = p->next) {
4126 4127
		prefix_one(&p->new_name);
		prefix_one(&p->old_name);
4128 4129 4130
	}
}

4131 4132 4133 4134
#define INACCURATE_EOF	(1<<0)
#define RECOUNT		(1<<1)

static int apply_patch(int fd, const char *filename, int options)
L
Linus Torvalds 已提交
4135
{
4136
	size_t offset;
4137
	struct strbuf buf = STRBUF_INIT; /* owns the patch text */
4138
	struct patch *list = NULL, **listp = &list;
4139
	int skipped_patch = 0;
L
Linus Torvalds 已提交
4140

4141
	patch_input_file = filename;
4142
	read_patch_file(&buf, fd);
L
Linus Torvalds 已提交
4143
	offset = 0;
4144
	while (offset < buf.len) {
4145 4146 4147
		struct patch *patch;
		int nr;

4148
		patch = xcalloc(1, sizeof(*patch));
4149 4150
		patch->inaccurate_eof = !!(options & INACCURATE_EOF);
		patch->recount =  !!(options & RECOUNT);
4151
		nr = parse_chunk(buf.buf + offset, buf.len - offset, patch);
L
Linus Torvalds 已提交
4152 4153
		if (nr < 0)
			break;
4154
		if (apply_in_reverse)
J
Johannes Schindelin 已提交
4155
			reverse_patches(patch);
4156 4157
		if (prefix)
			prefix_patches(patch);
4158 4159 4160 4161
		if (use_patch(patch)) {
			patch_stats(patch);
			*listp = patch;
			listp = &patch->next;
4162 4163
		}
		else {
4164
			free_patch(patch);
4165 4166
			skipped_patch++;
		}
L
Linus Torvalds 已提交
4167 4168
		offset += nr;
	}
4169

4170
	if (!list && !skipped_patch)
4171
		die(_("unrecognized input"));
4172

4173
	if (whitespace_error && (ws_error_action == die_on_ws_error))
4174 4175
		apply = 0;

4176 4177
	update_index = check_index && apply;
	if (update_index && newfd < 0)
4178 4179
		newfd = hold_locked_index(&lock_file, 1);

4180 4181
	if (check_index) {
		if (read_cache() < 0)
4182
			die(_("unable to read index file"));
4183 4184
	}

J
Junio C Hamano 已提交
4185 4186 4187
	if ((check || apply) &&
	    check_patch_list(list) < 0 &&
	    !apply_with_reject)
4188 4189
		exit(1);

4190 4191 4192 4193 4194 4195
	if (apply && write_out_results(list)) {
		if (apply_with_reject)
			exit(1);
		/* with --3way, we still need to write the index out */
		return 1;
	}
4196

4197 4198
	if (fake_ancestor)
		build_fake_ancestor(list, fake_ancestor);
4199

4200 4201
	if (diffstat)
		stat_patch_list(list);
4202

J
Junio C Hamano 已提交
4203 4204 4205
	if (numstat)
		numstat_patch_list(list);

4206 4207 4208
	if (summary)
		summary_patch_list(list);

4209
	free_patch_list(list);
4210
	strbuf_release(&buf);
4211
	string_list_clear(&fn_table, 0);
L
Linus Torvalds 已提交
4212 4213 4214
	return 0;
}

4215
static int git_apply_config(const char *var, const char *value, void *cb)
4216
{
4217 4218
	if (!strcmp(var, "apply.whitespace"))
		return git_config_string(&apply_default_whitespace, var, value);
4219 4220
	else if (!strcmp(var, "apply.ignorewhitespace"))
		return git_config_string(&apply_default_ignorewhitespace, var, value);
4221
	return git_default_config(var, value, cb);
4222 4223
}

M
Miklos Vajna 已提交
4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256
static int option_parse_exclude(const struct option *opt,
				const char *arg, int unset)
{
	add_name_limit(arg, 1);
	return 0;
}

static int option_parse_include(const struct option *opt,
				const char *arg, int unset)
{
	add_name_limit(arg, 0);
	has_include = 1;
	return 0;
}

static int option_parse_p(const struct option *opt,
			  const char *arg, int unset)
{
	p_value = atoi(arg);
	p_value_known = 1;
	return 0;
}

static int option_parse_z(const struct option *opt,
			  const char *arg, int unset)
{
	if (unset)
		line_termination = '\n';
	else
		line_termination = 0;
	return 0;
}

4257 4258 4259 4260 4261 4262 4263 4264 4265 4266
static int option_parse_space_change(const struct option *opt,
			  const char *arg, int unset)
{
	if (unset)
		ws_ignore_action = ignore_ws_none;
	else
		ws_ignore_action = ignore_ws_change;
	return 0;
}

M
Miklos Vajna 已提交
4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289
static int option_parse_whitespace(const struct option *opt,
				   const char *arg, int unset)
{
	const char **whitespace_option = opt->value;

	*whitespace_option = arg;
	parse_whitespace_option(arg);
	return 0;
}

static int option_parse_directory(const struct option *opt,
				  const char *arg, int unset)
{
	root_len = strlen(arg);
	if (root_len && arg[root_len - 1] != '/') {
		char *new_root;
		root = new_root = xmalloc(root_len + 2);
		strcpy(new_root, arg);
		strcpy(new_root + root_len++, "/");
	} else
		root = arg;
	return 0;
}
4290

4291
int cmd_apply(int argc, const char **argv, const char *prefix_)
L
Linus Torvalds 已提交
4292 4293
{
	int i;
J
Junio C Hamano 已提交
4294
	int errs = 0;
4295
	int is_not_gitdir = !startup_info->have_repository;
M
Miklos Vajna 已提交
4296
	int force_apply = 0;
4297

4298
	const char *whitespace_option = NULL;
L
Linus Torvalds 已提交
4299

M
Miklos Vajna 已提交
4300
	struct option builtin_apply_options[] = {
4301 4302
		{ OPTION_CALLBACK, 0, "exclude", NULL, N_("path"),
			N_("don't apply changes matching the given path"),
M
Miklos Vajna 已提交
4303
			0, option_parse_exclude },
4304 4305
		{ OPTION_CALLBACK, 0, "include", NULL, N_("path"),
			N_("apply changes matching the given path"),
M
Miklos Vajna 已提交
4306
			0, option_parse_include },
4307 4308
		{ OPTION_CALLBACK, 'p', NULL, NULL, N_("num"),
			N_("remove <num> leading slashes from traditional diff paths"),
M
Miklos Vajna 已提交
4309 4310
			0, option_parse_p },
		OPT_BOOLEAN(0, "no-add", &no_add,
4311
			N_("ignore additions made by the patch")),
M
Miklos Vajna 已提交
4312
		OPT_BOOLEAN(0, "stat", &diffstat,
4313
			N_("instead of applying the patch, output diffstat for the input")),
R
René Scharfe 已提交
4314 4315
		OPT_NOOP_NOARG(0, "allow-binary-replacement"),
		OPT_NOOP_NOARG(0, "binary"),
M
Miklos Vajna 已提交
4316
		OPT_BOOLEAN(0, "numstat", &numstat,
4317
			N_("show number of added and deleted lines in decimal notation")),
M
Miklos Vajna 已提交
4318
		OPT_BOOLEAN(0, "summary", &summary,
4319
			N_("instead of applying the patch, output a summary for the input")),
M
Miklos Vajna 已提交
4320
		OPT_BOOLEAN(0, "check", &check,
4321
			N_("instead of applying the patch, see if the patch is applicable")),
M
Miklos Vajna 已提交
4322
		OPT_BOOLEAN(0, "index", &check_index,
4323
			N_("make sure the patch is applicable to the current index")),
M
Miklos Vajna 已提交
4324
		OPT_BOOLEAN(0, "cached", &cached,
4325
			N_("apply a patch without touching the working tree")),
M
Miklos Vajna 已提交
4326
		OPT_BOOLEAN(0, "apply", &force_apply,
4327
			N_("also apply the patch (use with --stat/--summary/--check)")),
4328
		OPT_BOOL('3', "3way", &threeway,
J
Junio C Hamano 已提交
4329
			 N_( "attempt three-way merge if a patch does not apply")),
4330
		OPT_FILENAME(0, "build-fake-ancestor", &fake_ancestor,
4331
			N_("build a temporary index based on embedded index information")),
M
Miklos Vajna 已提交
4332
		{ OPTION_CALLBACK, 'z', NULL, NULL, NULL,
4333
			N_("paths are separated with NUL character"),
M
Miklos Vajna 已提交
4334 4335
			PARSE_OPT_NOARG, option_parse_z },
		OPT_INTEGER('C', NULL, &p_context,
4336 4337 4338
				N_("ensure at least <n> lines of context match")),
		{ OPTION_CALLBACK, 0, "whitespace", &whitespace_option, N_("action"),
			N_("detect new or modified lines that have whitespace errors"),
M
Miklos Vajna 已提交
4339
			0, option_parse_whitespace },
4340
		{ OPTION_CALLBACK, 0, "ignore-space-change", NULL, NULL,
4341
			N_("ignore changes in whitespace when finding context"),
4342 4343
			PARSE_OPT_NOARG, option_parse_space_change },
		{ OPTION_CALLBACK, 0, "ignore-whitespace", NULL, NULL,
4344
			N_("ignore changes in whitespace when finding context"),
4345
			PARSE_OPT_NOARG, option_parse_space_change },
M
Miklos Vajna 已提交
4346
		OPT_BOOLEAN('R', "reverse", &apply_in_reverse,
4347
			N_("apply the patch in reverse")),
M
Miklos Vajna 已提交
4348
		OPT_BOOLEAN(0, "unidiff-zero", &unidiff_zero,
4349
			N_("don't expect at least one line of context")),
M
Miklos Vajna 已提交
4350
		OPT_BOOLEAN(0, "reject", &apply_with_reject,
4351
			N_("leave the rejected hunks in corresponding *.rej files")),
4352
		OPT_BOOLEAN(0, "allow-overlap", &allow_overlap,
4353 4354
			N_("allow overlapping hunks")),
		OPT__VERBOSE(&apply_verbosely, N_("be verbose")),
M
Miklos Vajna 已提交
4355
		OPT_BIT(0, "inaccurate-eof", &options,
4356
			N_("tolerate incorrectly detected missing new-line at the end of file"),
M
Miklos Vajna 已提交
4357 4358
			INACCURATE_EOF),
		OPT_BIT(0, "recount", &options,
4359
			N_("do not trust the line counts in the hunk headers"),
M
Miklos Vajna 已提交
4360
			RECOUNT),
4361 4362
		{ OPTION_CALLBACK, 0, "directory", NULL, N_("root"),
			N_("prepend <root> to all filenames"),
M
Miklos Vajna 已提交
4363 4364 4365 4366
			0, option_parse_directory },
		OPT_END()
	};

4367
	prefix = prefix_;
4368
	prefix_length = prefix ? strlen(prefix) : 0;
4369
	git_config(git_apply_config, NULL);
4370 4371
	if (apply_default_whitespace)
		parse_whitespace_option(apply_default_whitespace);
4372 4373
	if (apply_default_ignorewhitespace)
		parse_ignorewhitespace_option(apply_default_ignorewhitespace);
4374

4375
	argc = parse_options(argc, argv, prefix, builtin_apply_options,
M
Miklos Vajna 已提交
4376
			apply_usage, 0);
4377

4378 4379 4380 4381 4382 4383 4384 4385 4386
	if (apply_with_reject && threeway)
		die("--reject and --3way cannot be used together.");
	if (cached && threeway)
		die("--cached and --3way cannot be used together.");
	if (threeway) {
		if (is_not_gitdir)
			die(_("--3way outside a repository"));
		check_index = 1;
	}
M
Miklos Vajna 已提交
4387 4388 4389 4390 4391
	if (apply_with_reject)
		apply = apply_verbosely = 1;
	if (!force_apply && (diffstat || numstat || summary || check || fake_ancestor))
		apply = 0;
	if (check_index && is_not_gitdir)
4392
		die(_("--index outside a repository"));
M
Miklos Vajna 已提交
4393 4394
	if (cached) {
		if (is_not_gitdir)
4395
			die(_("--cached outside a repository"));
M
Miklos Vajna 已提交
4396 4397 4398
		check_index = 1;
	}
	for (i = 0; i < argc; i++) {
L
Linus Torvalds 已提交
4399 4400 4401 4402
		const char *arg = argv[i];
		int fd;

		if (!strcmp(arg, "-")) {
4403
			errs |= apply_patch(0, "<stdin>", options);
4404
			read_stdin = 0;
L
Linus Torvalds 已提交
4405
			continue;
4406
		} else if (0 < prefix_length)
4407 4408
			arg = prefix_filename(prefix, prefix_length, arg);

L
Linus Torvalds 已提交
4409 4410
		fd = open(arg, O_RDONLY);
		if (fd < 0)
4411
			die_errno(_("can't open patch '%s'"), arg);
4412
		read_stdin = 0;
4413
		set_default_whitespace_mode(whitespace_option);
4414
		errs |= apply_patch(fd, arg, options);
L
Linus Torvalds 已提交
4415 4416
		close(fd);
	}
4417
	set_default_whitespace_mode(whitespace_option);
4418
	if (read_stdin)
4419
		errs |= apply_patch(0, "<stdin>", options);
4420 4421 4422 4423 4424
	if (whitespace_error) {
		if (squelch_whitespace_errors &&
		    squelch_whitespace_errors < whitespace_error) {
			int squelched =
				whitespace_error - squelch_whitespace_errors;
4425 4426 4427 4428
			warning(Q_("squelched %d whitespace error",
				   "squelched %d whitespace errors",
				   squelched),
				squelched);
4429
		}
4430
		if (ws_error_action == die_on_ws_error)
4431 4432 4433 4434
			die(Q_("%d line adds whitespace errors.",
			       "%d lines add whitespace errors.",
			       whitespace_error),
			    whitespace_error);
4435
		if (applied_after_fixing_ws && apply)
4436 4437
			warning("%d line%s applied after"
				" fixing whitespace errors.",
4438 4439
				applied_after_fixing_ws,
				applied_after_fixing_ws == 1 ? "" : "s");
4440
		else if (whitespace_error)
4441 4442 4443 4444
			warning(Q_("%d line adds whitespace errors.",
				   "%d lines add whitespace errors.",
				   whitespace_error),
				whitespace_error);
4445
	}
4446

4447
	if (update_index) {
4448
		if (write_cache(newfd, active_cache, active_nr) ||
B
Brandon Casey 已提交
4449
		    commit_locked_index(&lock_file))
4450
			die(_("Unable to write new index file"));
4451 4452
	}

J
Junio C Hamano 已提交
4453
	return !!errs;
L
Linus Torvalds 已提交
4454
}