git/apply.c

/*
 * apply.c
 *
 * Copyright (C) Linus Torvalds, 2005
 *
 * This applies patches on top of some (arbitrary) version of the SCM.
 *
 */

#include "cache.h"
#include "abspath.h"
#include "alloc.h"
#include "config.h"
#include "object-store.h"
#include "blob.h"
#include "delta.h"
#include "diff.h"
#include "dir.h"
#include "environment.h"
#include "gettext.h"
#include "hex.h"
#include "xdiff-interface.h"
#include "ll-merge.h"
#include "lockfile.h"
#include "object-name.h"
#include "parse-options.h"
#include "quote.h"
#include "rerere.h"
#include "apply.h"
#include "entry.h"
#include "setup.h"
#include "wrapper.h"

struct gitdiff_data {
	struct strbuf *root;
	int linenr;
	int p_value;
};

static void git_apply_config(void)
{
	git_config_get_string("apply.whitespace", &apply_default_whitespace);
	git_config_get_string("apply.ignorewhitespace", &apply_default_ignorewhitespace);
	git_config(git_xmerge_config, NULL);
}

static int parse_whitespace_option(struct apply_state *state, const char *option)
{
	if (!option) {
		state->ws_error_action = warn_on_ws_error;
		return 0;
	}
	if (!strcmp(option, "warn")) {
		state->ws_error_action = warn_on_ws_error;
		return 0;
	}
	if (!strcmp(option, "nowarn")) {
		state->ws_error_action = nowarn_ws_error;
		return 0;
	}
	if (!strcmp(option, "error")) {
		state->ws_error_action = die_on_ws_error;
		return 0;
	}
	if (!strcmp(option, "error-all")) {
		state->ws_error_action = die_on_ws_error;
		state->squelch_whitespace_errors = 0;
		return 0;
	}
	if (!strcmp(option, "strip") || !strcmp(option, "fix")) {
		state->ws_error_action = correct_ws_error;
		return 0;
	}
	/*
	 * Please update $__git_whitespacelist in git-completion.bash
	 * when you add new options.
	 */
	return error(_("unrecognized whitespace option '%s'"), option);
}

static int parse_ignorewhitespace_option(struct apply_state *state,
						 const char *option)
{
	if (!option || !strcmp(option, "no") ||
	    !strcmp(option, "false") || !strcmp(option, "never") ||
	    !strcmp(option, "none")) {
		state->ws_ignore_action = ignore_ws_none;
		return 0;
	}
	if (!strcmp(option, "change")) {
		state->ws_ignore_action = ignore_ws_change;
		return 0;
	}
	return error(_("unrecognized whitespace ignore option '%s'"), option);
}

int init_apply_state(struct apply_state *state,
		     struct repository *repo,
		     const char *prefix)
{
	memset(state, 0, sizeof(*state));
	state->prefix = prefix;
	state->repo = repo;
	state->apply = 1;
	state->line_termination = '\n';
	state->p_value = 1;
	state->p_context = UINT_MAX;
	state->squelch_whitespace_errors = 5;
	state->ws_error_action = warn_on_ws_error;
	state->ws_ignore_action = ignore_ws_none;
	state->linenr = 1;
	string_list_init_nodup(&state->fn_table);
	string_list_init_nodup(&state->limit_by_name);
	strset_init(&state->removed_symlinks);
	strset_init(&state->kept_symlinks);
	strbuf_init(&state->root, 0);

	git_apply_config();
	if (apply_default_whitespace && parse_whitespace_option(state, apply_default_whitespace))
		return -1;
	if (apply_default_ignorewhitespace && parse_ignorewhitespace_option(state, apply_default_ignorewhitespace))
		return -1;
	return 0;
}

void clear_apply_state(struct apply_state *state)
{
	string_list_clear(&state->limit_by_name, 0);
	strset_clear(&state->removed_symlinks);
	strset_clear(&state->kept_symlinks);
	strbuf_release(&state->root);

	/* &state->fn_table is cleared at the end of apply_patch() */
}

static void mute_routine(const char *msg UNUSED, va_list params UNUSED)
{
	/* do nothing */
}

int check_apply_state(struct apply_state *state, int force_apply)
{
	int is_not_gitdir = !startup_info->have_repository;

	if (state->apply_with_reject && state->threeway)
		return error(_("options '%s' and '%s' cannot be used together"), "--reject", "--3way");
	if (state->threeway) {
		if (is_not_gitdir)
			return error(_("'%s' outside a repository"), "--3way");
		state->check_index = 1;
	}
	if (state->apply_with_reject) {
		state->apply = 1;
		if (state->apply_verbosity == verbosity_normal)
			state->apply_verbosity = verbosity_verbose;
	}
	if (!force_apply && (state->diffstat || state->numstat || state->summary || state->check || state->fake_ancestor))
		state->apply = 0;
	if (state->check_index && is_not_gitdir)
		return error(_("'%s' outside a repository"), "--index");
	if (state->cached) {
		if (is_not_gitdir)
			return error(_("'%s' outside a repository"), "--cached");
		state->check_index = 1;
	}
	if (state->ita_only && (state->check_index || is_not_gitdir))
		state->ita_only = 0;
	if (state->check_index)
		state->unsafe_paths = 0;

	if (state->apply_verbosity <= verbosity_silent) {
		state->saved_error_routine = get_error_routine();
		state->saved_warn_routine = get_warn_routine();
		set_error_routine(mute_routine);
		set_warn_routine(mute_routine);
	}

	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

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;
	}
}

void release_patch(struct patch *patch)
{
	free_fragment_list(patch->fragments);
	free(patch->def_name);
	free(patch->old_name);
	free(patch->new_name);
	free(patch->result);
}

static void free_patch(struct patch *patch)
{
	release_patch(patch);
	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 *end1 = s1 + n1;
	const char *end2 = s2 + n2;

	/* ignore line endings */
	while (s1 < end1 && (end1[-1] == '\r' || end1[-1] == '\n'))
		end1--;
	while (s2 < end2 && (end2[-1] == '\r' || end2[-1] == '\n'))
		end2--;

	while (s1 < end1 && s2 < end2) {
		if (isspace(*s1)) {
			/*
			 * Skip whitespace. We check on both buffers
			 * because we don't want "a b" to match "ab".
			 */
			if (!isspace(*s2))
				return 0;
			while (s1 < end1 && isspace(*s1))
				s1++;
			while (s2 < end2 && isspace(*s2))
				s2++;
		} else if (*s1++ != *s2++)
			return 0;
	}

	/* If we reached the end on one side only, lines don't match. */
	return s1 == end1 && s2 == end2;
}

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)

/*
 * apply.c isn't equipped to handle arbitrarily large patches, because
 * it intermingles `unsigned long` with `int` for the type used to store
 * buffer lengths.
 *
 * Only process patches that are just shy of 1 GiB large in order to
 * avoid any truncation or overflow issues.
 */
#define MAX_APPLY_SIZE (1024UL * 1024 * 1023)

static int read_patch_file(struct strbuf *sb, int fd)
{
	if (strbuf_read(sb, fd, 0) < 0 || sb->len >= MAX_APPLY_SIZE)
		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 strbuf *root,
			   const char *line,
			   int p_value)
{
	struct strbuf name = STRBUF_INIT;
	char *cp;

	/*
	 * Proposed "new-style" GNU patch/diff format; see
	 * https://lore.kernel.org/git/7vll0wvb2a.fsf@assigned-by-dhcp.cox.net/
	 */
	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 (root->len)
		strbuf_insert(&name, 0, root->buf, 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 strbuf *root,
			      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 (root->len) {
		char *ret = xstrfmt("%s%.*s", root->buf, len, start);
		return squash_slash(ret);
	}

	return squash_slash(xmemdupz(start, len));
}

static char *find_name(struct strbuf *root,
		       const char *line,
		       char *def,
		       int p_value,
		       int terminate)
{
	if (*line == '"') {
		char *name = find_name_gnu(root, line, p_value);
		if (name)
			return name;
	}

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

static char *find_name_traditional(struct strbuf *root,
				   const char *line,
				   char *def,
				   int p_value)
{
	size_t len;
	size_t date_len;

	if (*line == '"') {
		char *name = find_name_gnu(root, line, 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(root, line, def, p_value, NULL, TERM_TAB);
	len -= date_len;

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

/*
 * 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->root, 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 (starts_with(name, state->prefix))
			val = count_slashes(state->prefix);
		else {
			cp++;
			if (starts_with(cp, state->prefix))
				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[] =
		"^[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, epoch_hour, hour, minute;
	int status;

	for (cp = nameline; *cp != '\n'; cp++) {
		if (*cp == '\t')
			timestamp = cp + 1;
	}
	if (!timestamp)
		return 0;

	/*
	 * YYYY-MM-DD hh:mm:ss must be from either 1969-12-31
	 * (west of GMT) or 1970-01-01 (east of GMT)
	 */
	if (skip_prefix(timestamp, "1969-12-31 ", &timestamp))
		epoch_hour = 24;
	else if (skip_prefix(timestamp, "1970-01-01 ", &timestamp))
		epoch_hour = 0;
	else
		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;
	}

	hour = strtol(timestamp, NULL, 10);
	minute = strtol(timestamp + m[1].rm_so, NULL, 10);

	zoneoffset = strtol(timestamp + m[3].rm_so + 1, (char **) &colon, 10);
	if (*colon == ':')