apply: move libified code from builtin/apply.c to apply.{c,h}

As most of the apply code in builtin/apply.c has been libified by a number of
previous commits, it can now be moved to apply.{c,h}, so that more code can
use it.

Helped-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Helped-by: Ramsay Jones <ramsay@ramsayjones.plus.com>
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

apply.c

@@ -1,5 +1,23 @@
+/*
+ * apply.c
+ *
+ * Copyright (C) Linus Torvalds, 2005
+ *
+ * This applies patches on top of some (arbitrary) version of the SCM.
+ *
+ */
+
 #include "cache.h"
+#include "blob.h"
+#include "delta.h"
+#include "diff.h"
+#include "dir.h"
+#include "xdiff-interface.h"
+#include "ll-merge.h"
 #include "lockfile.h"
+#include "parse-options.h"
+#include "quote.h"
+#include "rerere.h"
 #include "apply.h"
 
 static void git_apply_config(void)
@@ -125,3 +143,4716 @@ int check_apply_state(struct apply_state *state, int force_apply)
 
 	return 0;
 }
+
+static void set_default_whitespace_mode(struct apply_state *state)
+{
+	if (!state->whitespace_option && !apply_default_whitespace)
+		state->ws_error_action = (state->apply ? warn_on_ws_error : nowarn_ws_error);
+}
+
+/*
+ * 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.
+ */
+struct fragment {
+	unsigned long leading, trailing;
+	unsigned long oldpos, oldlines;
+	unsigned long newpos, newlines;
+	/*
+	 * 'patch' is usually borrowed from buf in apply_patch(),
+	 * but some codepaths store an allocated buffer.
+	 */
+	const char *patch;
+	unsigned free_patch:1,
+		rejected:1;
+	int size;
+	int linenr;
+	struct fragment *next;
+};
+
+/*
+ * 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
+
+/*
+ * This represents a "patch" to a file, both metainfo changes
+ * such as creation/deletion, filemode and content changes represented
+ * as a series of fragments.
+ */
+struct patch {
+	char *new_name, *old_name, *def_name;
+	unsigned int old_mode, new_mode;
+	int is_new, is_delete;	/* -1 = unknown, 0 = false, 1 = true */
+	int rejected;
+	unsigned ws_rule;
+	int lines_added, lines_deleted;
+	int score;
+	unsigned int is_toplevel_relative:1;
+	unsigned int inaccurate_eof:1;
+	unsigned int is_binary:1;
+	unsigned int is_copy:1;
+	unsigned int is_rename:1;
+	unsigned int recount:1;
+	unsigned int conflicted_threeway:1;
+	unsigned int direct_to_threeway:1;
+	struct fragment *fragments;
+	char *result;
+	size_t resultsize;
+	char old_sha1_prefix[41];
+	char new_sha1_prefix[41];
+	struct patch *next;
+
+	/* three-way fallback result */
+	struct object_id threeway_stage[3];
+};
+
+static void free_fragment_list(struct fragment *list)
+{
+	while (list) {
+		struct fragment *next = list->next;
+		if (list->free_patch)
+			free((char *)list->patch);
+		free(list);
+		list = next;
+	}
+}
+
+static void free_patch(struct patch *patch)
+{
+	free_fragment_list(patch->fragments);
+	free(patch->def_name);
+	free(patch->old_name);
+	free(patch->new_name);
+	free(patch->result);
+	free(patch);
+}
+
+static void free_patch_list(struct patch *list)
+{
+	while (list) {
+		struct patch *next = list->next;
+		free_patch(list);
+		list = next;
+	}
+}
+
+/*
+ * 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;
+#define LINE_COMMON     1
+#define LINE_PATCHED	2
+};
+
+/*
+ * This represents a "file", which is an array of "lines".
+ */
+struct image {
+	char *buf;
+	size_t len;
+	size_t nr;
+	size_t alloc;
+	struct line *line_allocated;
+	struct line *line;
+};
+
+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;
+}
+
+/*
+ * 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 whitespaces, if both begin with whitespace */
+	if (s1 <= last1 && s2 <= last2 && isspace(*s1) && isspace(*s2)) {
+		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;
+}
+
+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++;
+}
+
+/*
+ * "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".
+ */
+static void prepare_image(struct image *image, char *buf, size_t len,
+			  int prepare_linetable)
+{
+	const char *cp, *ep;
+
+	memset(image, 0, sizeof(*image));
+	image->buf = buf;
+	image->len = len;
+
+	if (!prepare_linetable)
+		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++;
+		add_line_info(image, cp, next - cp, 0);
+		cp = next;
+	}
+	image->line = image->line_allocated;
+}
+
+static void clear_image(struct image *image)
+{
+	free(image->buf);
+	free(image->line_allocated);
+	memset(image, 0, sizeof(*image));
+}
+
+/* fmt must contain _one_ %s and no other substitution */
+static void say_patch_name(FILE *output, const char *fmt, struct patch *patch)
+{
+	struct strbuf sb = STRBUF_INIT;
+
+	if (patch->old_name && patch->new_name &&
+	    strcmp(patch->old_name, patch->new_name)) {
+		quote_c_style(patch->old_name, &sb, NULL, 0);
+		strbuf_addstr(&sb, " => ");
+		quote_c_style(patch->new_name, &sb, NULL, 0);
+	} else {
+		const char *n = patch->new_name;
+		if (!n)
+			n = patch->old_name;
+		quote_c_style(n, &sb, NULL, 0);
+	}
+	fprintf(output, fmt, sb.buf);
+	fputc('\n', output);
+	strbuf_release(&sb);
+}
+
+#define SLOP (16)
+
+static int read_patch_file(struct strbuf *sb, int fd)
+{
+	if (strbuf_read(sb, fd, 0) < 0)
+		return error_errno("git apply: failed to read");
+
+	/*
+	 * 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.
+	 */
+	strbuf_grow(sb, SLOP);
+	memset(sb->buf + sb->len, 0, SLOP);
+	return 0;
+}
+
+static unsigned long linelen(const char *buffer, unsigned long size)
+{
+	unsigned long len = 0;
+	while (size--) {
+		len++;
+		if (*buffer++ == '\n')
+			break;
+	}
+	return len;
+}
+
+static int is_dev_null(const char *str)
+{
+	return skip_prefix(str, "/dev/null", &str) && isspace(*str);
+}
+
+#define TERM_SPACE	1
+#define TERM_TAB	2
+
+static int name_terminate(int c, int terminate)
+{
+	if (c == ' ' && !(terminate & TERM_SPACE))
+		return 0;
+	if (c == '\t' && !(terminate & TERM_TAB))
+		return 0;
+
+	return 1;
+}
+
+/* remove double slashes to make --index work with such filenames */
+static char *squash_slash(char *name)
+{
+	int i = 0, j = 0;
+
+	if (!name)
+		return NULL;
+
+	while (name[i]) {
+		if ((name[j++] = name[i++]) == '/')
+			while (name[i] == '/')
+				i++;
+	}
+	name[j] = '\0';
+	return name;
+}
+
+static char *find_name_gnu(struct apply_state *state,
+			   const char *line,
+			   const char *def,
+			   int p_value)
+{
+	struct strbuf name = STRBUF_INIT;
+	char *cp;
+
+	/*
+	 * Proposed "new-style" GNU patch/diff format; see
+	 * http://marc.info/?l=git&m=112927316408690&w=2
+	 */
+	if (unquote_c_style(&name, line, NULL)) {
+		strbuf_release(&name);
+		return NULL;
+	}
+
+	for (cp = name.buf; p_value; p_value--) {
+		cp = strchr(cp, '/');
+		if (!cp) {
+			strbuf_release(&name);
+			return NULL;
+		}
+		cp++;
+	}
+
+	strbuf_remove(&name, 0, cp - name.buf);
+	if (state->root.len)
+		strbuf_insert(&name, 0, state->root.buf, state->root.len);
+	return squash_slash(strbuf_detach(&name, NULL));
+}
+
+static size_t sane_tz_len(const char *line, size_t len)
+{
+	const char *tz, *p;
+
+	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;
+}
+
+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;
+}
+
+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);
+	}
+
+	/* 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;
+
+	n = sane_tz_len(line, end - line);
+	if (!n)
+		n = tz_with_colon_len(line, end - line);
+	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;
+}
+
+static char *find_name_common(struct apply_state *state,
+			      const char *line,
+			      const char *def,
+			      int p_value,
+			      const char *end,
+			      int terminate)
+{
+	int len;
+	const char *start = NULL;
+
+	if (p_value == 0)
+		start = line;
+	while (line != end) {
+		char c = *line;
+
+		if (!end && isspace(c)) {
+			if (c == '\n')
+				break;
+			if (name_terminate(c, terminate))
+				break;
+		}
+		line++;
+		if (c == '/' && !--p_value)
+			start = line;
+	}
+	if (!start)
+		return squash_slash(xstrdup_or_null(def));
+	len = line - start;
+	if (!len)
+		return squash_slash(xstrdup_or_null(def));
+
+	/*
+	 * 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))
+			return squash_slash(xstrdup(def));
+	}
+
+	if (state->root.len) {
+		char *ret = xstrfmt("%s%.*s", state->root.buf, len, start);
+		return squash_slash(ret);
+	}
+
+	return squash_slash(xmemdupz(start, len));
+}
+
+static char *find_name(struct apply_state *state,
+		       const char *line,
+		       char *def,
+		       int p_value,
+		       int terminate)
+{
+	if (*line == '"') {
+		char *name = find_name_gnu(state, line, def, p_value);
+		if (name)
+			return name;
+	}
+
+	return find_name_common(state, line, def, p_value, NULL, terminate);
+}
+
+static char *find_name_traditional(struct apply_state *state,
+				   const char *line,
+				   char *def,
+				   int p_value)
+{
+	size_t len;
+	size_t date_len;
+
+	if (*line == '"') {
+		char *name = find_name_gnu(state, 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(state, line, def, p_value, NULL, TERM_TAB);
+	len -= date_len;
+
+	return find_name_common(state, line, def, p_value, line + len, 0);
+}
+
+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(struct apply_state *state, const char *nameline)
+{
+	char *name, *cp;
+	int val = -1;
+
+	if (is_dev_null(nameline))
+		return -1;
+	name = find_name_traditional(state, nameline, NULL, 0);
+	if (!name)
+		return -1;
+	cp = strchr(name, '/');
+	if (!cp)
+		val = 0;
+	else if (state->prefix) {
+		/*
+		 * Does it begin with "a/$our-prefix" and such?  Then this is
+		 * very likely to apply to our directory.
+		 */
+		if (!strncmp(name, state->prefix, state->prefix_length))
+			val = count_slashes(state->prefix);
+		else {
+			cp++;
+			if (!strncmp(cp, state->prefix, state->prefix_length))
+				val = count_slashes(state->prefix) + 1;
+		}
+	}
+	free(name);
+	return val;
+}
+
+/*
+ * Does the ---/+++ line have 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+)?"
+		" "
+		"([-+][0-2][0-9]:?[0-5][0-9])\n";
+	const char *timestamp = NULL, *cp, *colon;
+	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)) {
+			warning(_("Cannot prepare timestamp regexp %s"),
+				stamp_regexp);
+			return 0;
+		}
+	}
+
+	status = regexec(stamp, timestamp, ARRAY_SIZE(m), m, 0);
+	if (status) {
+		if (status != REG_NOMATCH)
+			warning(_("regexec returned %d for input: %s"),
+				status, timestamp);
+		return 0;
+	}
+
+	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);
+	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));
+}
+
+/*
+ * Get the name etc info from the ---/+++ lines of a traditional patch header
+ *
+ * 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.
+ */
+static int parse_traditional_patch(struct apply_state *state,
+				   const char *first,
+				   const char *second,
+				   struct patch *patch)
+{
+	char *name;
+
+	first += 4;	/* skip "--- " */
+	second += 4;	/* skip "+++ " */
+	if (!state->p_value_known) {
+		int p, q;
+		p = guess_p_value(state, first);
+		q = guess_p_value(state, second);
+		if (p < 0) p = q;
+		if (0 <= p && p == q) {
+			state->p_value = p;
+			state->p_value_known = 1;
+		}
+	}
+	if (is_dev_null(first)) {
+		patch->is_new = 1;
+		patch->is_delete = 0;
+		name = find_name_traditional(state, second, NULL, state->p_value);
+		patch->new_name = name;
+	} else if (is_dev_null(second)) {
+		patch->is_new = 0;
+		patch->is_delete = 1;
+		name = find_name_traditional(state, first, NULL, state->p_value);
+		patch->old_name = name;
+	} else {
+		char *first_name;
+		first_name = find_name_traditional(state, first, NULL, state->p_value);
+		name = find_name_traditional(state, second, first_name, state->p_value);
+		free(first_name);
+		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 {
+			patch->old_name = name;
+			patch->new_name = xstrdup_or_null(name);
+		}
+	}
+	if (!name)
+		return error(_("unable to find filename in patch at line %d"), state->linenr);
+
+	return 0;
+}
+
+static int gitdiff_hdrend(struct apply_state *state,
+			  const char *line,
+			  struct patch *patch)
+{
+	return 1;
+}
+
+/*
+ * 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..
+ */
+#define DIFF_OLD_NAME 0
+#define DIFF_NEW_NAME 1
+
+static int gitdiff_verify_name(struct apply_state *state,
+			       const char *line,
+			       int isnull,
+			       char **name,
+			       int side)
+{
+	if (!*name && !isnull) {
+		*name = find_name(state, line, NULL, state->p_value, TERM_TAB);
+		return 0;
+	}
+
+	if (*name) {
+		int len = strlen(*name);
+		char *another;
+		if (isnull)
+			return error(_("git apply: bad git-diff - expected /dev/null, got %s on line %d"),
+				     *name, state->linenr);
+		another = find_name(state, line, NULL, state->p_value, TERM_TAB);
+		if (!another || memcmp(another, *name, len + 1)) {
+			free(another);
+			return error((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"), state->linenr);
+		}
+		free(another);
+	} else {
+		/* expect "/dev/null" */
+		if (memcmp("/dev/null", line, 9) || line[9] != '\n')
+			return error(_("git apply: bad git-diff - expected /dev/null on line %d"), state->linenr);
+	}
+
+	return 0;
+}
+
+static int gitdiff_oldname(struct apply_state *state,
+			   const char *line,
+			   struct patch *patch)
+{
+	return gitdiff_verify_name(state, line,
+				   patch->is_new, &patch->old_name,
+				   DIFF_OLD_NAME);
+}
+
+static int gitdiff_newname(struct apply_state *state,
+			   const char *line,
+			   struct patch *patch)
+{
+	return gitdiff_verify_name(state, line,
+				   patch->is_delete, &patch->new_name,
+				   DIFF_NEW_NAME);
+}
+
+static int gitdiff_oldmode(struct apply_state *state,
+			   const char *line,
+			   struct patch *patch)
+{
+	patch->old_mode = strtoul(line, NULL, 8);
+	return 0;
+}
+
+static int gitdiff_newmode(struct apply_state *state,
+			   const char *line,
+			   struct patch *patch)
+{
+	patch->new_mode = strtoul(line, NULL, 8);
+	return 0;
+}
+
+static int gitdiff_delete(struct apply_state *state,
+			  const char *line,
+			  struct patch *patch)
+{
+	patch->is_delete = 1;
+	free(patch->old_name);
+	patch->old_name = xstrdup_or_null(patch->def_name);
+	return gitdiff_oldmode(state, line, patch);
+}
+
+static int gitdiff_newfile(struct apply_state *state,
+			   const char *line,
+			   struct patch *patch)
+{
+	patch->is_new = 1;
+	free(patch->new_name);
+	patch->new_name = xstrdup_or_null(patch->def_name);
+	return gitdiff_newmode(state, line, patch);
+}
+
+static int gitdiff_copysrc(struct apply_state *state,
+			   const char *line,
+			   struct patch *patch)
+{
+	patch->is_copy = 1;
+	free(patch->old_name);
+	patch->old_name = find_name(state, line, NULL, state->p_value ? state->p_value - 1 : 0, 0);
+	return 0;
+}
+
+static int gitdiff_copydst(struct apply_state *state,
+			   const char *line,
+			   struct patch *patch)
+{
+	patch->is_copy = 1;
+	free(patch->new_name);
+	patch->new_name = find_name(state, line, NULL, state->p_value ? state->p_value - 1 : 0, 0);
+	return 0;
+}
+
+static int gitdiff_renamesrc(struct apply_state *state,
+			     const char *line,
+			     struct patch *patch)
+{
+	patch->is_rename = 1;
+	free(patch->old_name);
+	patch->old_name = find_name(state, line, NULL, state->p_value ? state->p_value - 1 : 0, 0);
+	return 0;
+}
+
+static int gitdiff_renamedst(struct apply_state *state,
+			     const char *line,
+			     struct patch *patch)
+{
+	patch->is_rename = 1;
+	free(patch->new_name);
+	patch->new_name = find_name(state, line, NULL, state->p_value ? state->p_value - 1 : 0, 0);
+	return 0;
+}
+
+static int gitdiff_similarity(struct apply_state *state,
+			      const char *line,
+			      struct patch *patch)
+{
+	unsigned long val = strtoul(line, NULL, 10);
+	if (val <= 100)
+		patch->score = val;
+	return 0;
+}
+
+static int gitdiff_dissimilarity(struct apply_state *state,
+				 const char *line,
+				 struct patch *patch)
+{
+	unsigned long val = strtoul(line, NULL, 10);
+	if (val <= 100)
+		patch->score = val;
+	return 0;
+}
+
+static int gitdiff_index(struct apply_state *state,
+			 const char *line,
+			 struct patch *patch)
+{
+	/*
+	 * index line is N hexadecimal, "..", N hexadecimal,
+	 * and optional space with octal mode.
+	 */
+	const char *ptr, *eol;
+	int len;
+
+	ptr = strchr(line, '.');
+	if (!ptr || ptr[1] != '.' || 40 < ptr - line)
+		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 = strchrnul(line, '\n');
+
+	if (!ptr || eol < ptr)
+		ptr = eol;
+	len = ptr - line;
+
+	if (40 < len)
+		return 0;
+	memcpy(patch->new_sha1_prefix, line, len);
+	patch->new_sha1_prefix[len] = 0;
+	if (*ptr == ' ')
+		patch->old_mode = strtoul(ptr+1, NULL, 8);
+	return 0;
+}
+
+/*
+ * 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.
+ */
+static int gitdiff_unrecognized(struct apply_state *state,
+				const char *line,
+				struct patch *patch)
+{
+	return 1;
+}
+
+/*
+ * 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(struct apply_state *state,
+				    const char *line,
+				    int llen)
+{
+	int nslash;
+	int i;
+
+	if (!state->p_value)
+		return (llen && line[0] == '/') ? NULL : line;
+
+	nslash = state->p_value;
+	for (i = 0; i < llen; i++) {
+		int ch = line[i];
+		if (ch == '/' && --nslash <= 0)
+			return (i == 0) ? NULL : &line[i + 1];
+	}
+	return NULL;
+}
+
+/*
+ * This is to extract the same name that appears on "diff --git"
+ * 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.
+ */
+static char *git_header_name(struct apply_state *state,
+			     const char *line,
+			     int llen)
+{
+	const char *name;
+	const char *second = NULL;
+	size_t len, line_len;
+
+	line += strlen("diff --git ");
+	llen -= strlen("diff --git ");
+
+	if (*line == '"') {
+		const char *cp;
+		struct strbuf first = STRBUF_INIT;
+		struct strbuf sp = STRBUF_INIT;
+
+		if (unquote_c_style(&first, line, &second))
+			goto free_and_fail1;
+
+		/* strip the a/b prefix including trailing slash */
+		cp = skip_tree_prefix(state, first.buf, first.len);
+		if (!cp)
+			goto free_and_fail1;
+		strbuf_remove(&first, 0, cp - first.buf);
+
+		/*
+		 * second points at one past closing dq of name.
+		 * find the second name.
+		 */
+		while ((second < line + llen) && isspace(*second))
+			second++;
+
+		if (line + llen <= second)
+			goto free_and_fail1;
+		if (*second == '"') {
+			if (unquote_c_style(&sp, second, NULL))
+				goto free_and_fail1;
+			cp = skip_tree_prefix(state, sp.buf, sp.len);
+			if (!cp)
+				goto free_and_fail1;
+			/* They must match, otherwise ignore */
+			if (strcmp(cp, first.buf))
+				goto free_and_fail1;
+			strbuf_release(&sp);
+			return strbuf_detach(&first, NULL);
+		}
+
+		/* unquoted second */
+		cp = skip_tree_prefix(state, second, line + llen - second);
+		if (!cp)
+			goto free_and_fail1;
+		if (line + llen - cp != first.len ||
+		    memcmp(first.buf, cp, first.len))
+			goto free_and_fail1;
+		return strbuf_detach(&first, NULL);
+
+	free_and_fail1:
+		strbuf_release(&first);
+		strbuf_release(&sp);
+		return NULL;
+	}
+
+	/* unquoted first name */
+	name = skip_tree_prefix(state, line, llen);
+	if (!name)
+		return NULL;
+
+	/*
+	 * since the first name is unquoted, a dq if exists must be
+	 * the beginning of the second name.
+	 */
+	for (second = name; second < line + llen; second++) {
+		if (*second == '"') {
+			struct strbuf sp = STRBUF_INIT;
+			const char *np;
+
+			if (unquote_c_style(&sp, second, NULL))
+				goto free_and_fail2;
+
+			np = skip_tree_prefix(state, sp.buf, sp.len);
+			if (!np)
+				goto free_and_fail2;
+
+			len = sp.buf + sp.len - np;
+			if (len < second - name &&
+			    !strncmp(np, name, len) &&
+			    isspace(name[len])) {
+				/* Good */
+				strbuf_remove(&sp, 0, np - sp.buf);
+				return strbuf_detach(&sp, NULL);
+			}
+
+		free_and_fail2:
+			strbuf_release(&sp);
+			return NULL;
+		}
+	}
+
+	/*
+	 * Accept a name only if it shows up twice, exactly the same
+	 * form.
+	 */
+	second = strchr(name, '\n');
+	if (!second)
+		return NULL;
+	line_len = second - name;
+	for (len = 0 ; ; len++) {
+		switch (name[len]) {
+		default:
+			continue;
+		case '\n':
+			return NULL;
+		case '\t': case ' ':
+			/*
+			 * 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(state, name + len + 1,
+						  line_len - (len + 1));
+			if (!second)
+				return NULL;
+			/*
+			 * 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))
+				return xmemdupz(name, len);
+		}
+	}
+}
+
+/* Verify that we recognize the lines following a git header */
+static int parse_git_header(struct apply_state *state,
+			    const char *line,
+			    int len,
+			    unsigned int size,
+			    struct patch *patch)
+{
+	unsigned long offset;
+
+	/* A git diff has explicit new/delete information, so we don't guess */
+	patch->is_new = 0;
+	patch->is_delete = 0;
+
+	/*
+	 * 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.
+	 */
+	patch->def_name = git_header_name(state, line, len);
+	if (patch->def_name && state->root.len) {
+		char *s = xstrfmt("%s%s", state->root.buf, patch->def_name);
+		free(patch->def_name);
+		patch->def_name = s;
+	}
+
+	line += len;
+	size -= len;
+	state->linenr++;
+	for (offset = len ; size > 0 ; offset += len, size -= len, line += len, state->linenr++) {
+		static const struct opentry {
+			const char *str;
+			int (*fn)(struct apply_state *, const char *, struct patch *);
+		} 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 },
+			{ "rename old ", gitdiff_renamesrc },
+			{ "rename new ", gitdiff_renamedst },
+			{ "rename from ", gitdiff_renamesrc },
+			{ "rename to ", gitdiff_renamedst },
+			{ "similarity index ", gitdiff_similarity },
+			{ "dissimilarity index ", gitdiff_dissimilarity },
+			{ "index ", gitdiff_index },
+			{ "", gitdiff_unrecognized },
+		};
+		int i;
+
+		len = linelen(line, size);
+		if (!len || line[len-1] != '\n')
+			break;
+		for (i = 0; i < ARRAY_SIZE(optable); i++) {
+			const struct opentry *p = optable + i;
+			int oplen = strlen(p->str);
+			int res;
+			if (len < oplen || memcmp(p->str, line, oplen))
+				continue;
+			res = p->fn(state, line + oplen, patch);
+			if (res < 0)
+				return -1;
+			if (res > 0)
+				return offset;
+			break;
+		}
+	}
+
+	return offset;
+}
+
+static int parse_num(const char *line, unsigned long *p)
+{
+	char *ptr;
+
+	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,
+		       unsigned long *p1, unsigned long *p2)
+{
+	int digits, ex;
+
+	if (offset < 0 || offset >= len)
+		return -1;
+	line += offset;
+	len -= offset;
+
+	digits = parse_num(line, p1);
+	if (!digits)
+		return -1;
+
+	offset += digits;
+	line += digits;
+	len -= digits;
+
+	*p2 = 1;
+	if (*line == ',') {
+		digits = parse_num(line+1, p2);
+		if (!digits)
+			return -1;
+
+		offset += digits+1;
+		line += digits+1;
+		len -= digits+1;
+	}
+
+	ex = strlen(expect);
+	if (ex > len)
+		return -1;
+	if (memcmp(line, expect, ex))
+		return -1;
+
+	return offset + ex;
+}
+
+static void recount_diff(const char *line, int size, struct fragment *fragment)
+{
+	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 '\\':
+			continue;
+		case '@':
+			ret = size < 3 || !starts_with(line, "@@ ");
+			break;
+		case 'd':
+			ret = size < 5 || !starts_with(line, "diff ");
+			break;
+		default:
+			ret = -1;
+			break;
+		}
+		if (ret) {
+			warning(_("recount: unexpected line: %.*s"),
+				(int)linelen(line, size), line);
+			return;
+		}
+		break;
+	}
+	fragment->oldlines = oldlines;
+	fragment->newlines = newlines;
+}
+
+/*
+ * Parse a unified diff fragment header of the
+ * form "@@ -a,b +c,d @@"
+ */
+static int parse_fragment_header(const char *line, int len, struct fragment *fragment)
+{
+	int offset;
+
+	if (!len || line[len-1] != '\n')
+		return -1;
+
+	/* Figure out the number of lines in a fragment */
+	offset = parse_range(line, len, 4, " +", &fragment->oldpos, &fragment->oldlines);
+	offset = parse_range(line, len, offset, " @@", &fragment->newpos, &fragment->newlines);
+
+	return offset;
+}
+
+/*
+ * Find file diff header
+ *
+ * Returns:
+ *  -1 if no header was found
+ *  -128 in case of error
+ *   the size of the header in bytes (called "offset") otherwise
+ */
+static int find_header(struct apply_state *state,
+		       const char *line,
+		       unsigned long size,
+		       int *hdrsize,
+		       struct patch *patch)
+{
+	unsigned long offset, len;
+
+	patch->is_toplevel_relative = 0;
+	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;
+	for (offset = 0; size > 0; offset += len, size -= len, line += len, state->linenr++) {
+		unsigned long nextlen;
+
+		len = linelen(line, size);
+		if (!len)
+			break;
+
+		/* Testing this early allows us to take a few shortcuts.. */
+		if (len < 6)
+			continue;
+
+		/*
+		 * Make sure we don't find any unconnected patch fragments.
+		 * That's a sign that we didn't find a header, and that a
+		 * patch has become corrupted/broken up.
+		 */
+		if (!memcmp("@@ -", line, 4)) {
+			struct fragment dummy;
+			if (parse_fragment_header(line, len, &dummy) < 0)
+				continue;
+			error(_("patch fragment without header at line %d: %.*s"),
+				     state->linenr, (int)len-1, line);
+			return -128;
+		}
+
+		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)) {
+			int git_hdr_len = parse_git_header(state, line, len, size, patch);
+			if (git_hdr_len < 0)
+				return -128;
+			if (git_hdr_len <= len)
+				continue;
+			if (!patch->old_name && !patch->new_name) {
+				if (!patch->def_name) {
+					error(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)",
+							state->p_value),
+						     state->p_value, state->linenr);
+					return -128;
+				}
+				patch->old_name = xstrdup(patch->def_name);
+				patch->new_name = xstrdup(patch->def_name);
+			}
+			if (!patch->is_delete && !patch->new_name) {
+				error("git diff header lacks filename information "
+					     "(line %d)", state->linenr);
+				return -128;
+			}
+			patch->is_toplevel_relative = 1;
+			*hdrsize = git_hdr_len;
+			return offset;
+		}
+
+		/* --- followed by +++ ? */
+		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
+		 * minimum ("@@ -0,0 +1 @@\n" is the shortest).
+		 */
+		nextlen = linelen(line + len, size - len);
+		if (size < nextlen + 14 || memcmp("@@ -", line + len + nextlen, 4))
+			continue;
+
+		/* Ok, we'll consider it a patch */
+		if (parse_traditional_patch(state, line, line+len, patch))
+			return -128;
+		*hdrsize = len + nextlen;
+		state->linenr += 2;
+		return offset;
+	}
+	return -1;
+}
+
+static void record_ws_error(struct apply_state *state,
+			    unsigned result,
+			    const char *line,
+			    int len,
+			    int linenr)
+{
+	char *err;
+
+	if (!result)
+		return;
+
+	state->whitespace_error++;
+	if (state->squelch_whitespace_errors &&
+	    state->squelch_whitespace_errors < state->whitespace_error)
+		return;
+